The function of the Saccharomyces cerevisiae REV1 gene is required for translesion replication and mutagenesis induced by a wide variety of DNA-damaging agents. We showed previously that Rev1p possesses a deoxycytidyl transferase activity, which incorporates dCMP opposite abasic sites in the DNA template, and that dCMP insertion is the major event during bypass of an abasic site in vivo. However, we now find that Rev1p function is needed for the bypass of a T–T (6–4) UV photoproduct, a process in which dCMP incorporation occurs only very rarely, indicating that Rev1p possesses a second function. In addition, we find that Rev1p function is, as expected, required for bypass of an abasic site. However, replication past this lesion was also much reduced in the G-193R rev1-1 mutant, which we find retains substantial levels of deoxycytidyl transferase activity. This mutant is, therefore, presumably deficient principally in the second, at present poorly defined, function. The bypass of an abasic site and T–T (6–4) lesion also depended on REV3 function, but neither it nor REV1 was required for replication past the T–T dimer; bypass of this lesion presumably depends on another enzyme.
REV1 is one of at least three genes whose functions are required for translesion replication (also termed lesion bypass) and DNA damage-induced mutagenesis in budding yeast, Saccharomyces cerevisiae (Lawrence and Hinkle, 1996; Lawrence and Woodgate, 1999). Of the other two, REV3 encodes the catalytic subunit of DNA polymerase ζ, and REV7 encodes a subunit of this enzyme of presently unknown function (Nelson et al., 1996a). The 112 kDa Rev1 gene product shows 24% sequence identity over a region of 150 residues with the 47 kDa Escherichia coli UmuC protein (Larimer et al., 1989), which is also required for translesion replication, and both are members of a superfamily of enzymes that are found widely in prokaryotes and eukaryotes (McDonald et al., 1997; Roush et al., 1998). We have shown previously that Rev1p possesses a deoxycytidyl transferase activity (Nelson et al., 1996b), which adds dCMP opposite an abasic site in the template and, subsequently, other members of the superfamily were also shown to be DNA polymerases (reviewed by Friedberg and Gerlach, 1999). The existence of this activity, which all members possess, has diverted attention from the possibility that the transferase activity may not be the main function of the Rev1p subfamily. The Rev1p transferase activity could clearly be of importance in the bypass of abasic sites in yeast, and cytosine is in fact preferentially inserted opposite these lesions in vivo (Gibbs and Lawrence, 1995). However, REV1 function is also required for the induction of basepair substitutions by UV, ionizing radiation and a variety of chemical mutagens, events that often do not involve the insertion of dCMP, suggesting that Rev1p may possess another function in addition to its transferase activity. More specifically, dCMP insertion occurs only very rarely during the bypass of a T–T cis–syn cyclobutane dimer (Gibbs et al., 1993) or a T–T pyrimidine (6–4) pyrimidinone adduct (Gibbs et al., 1995) in yeast cells. It has not yet been established whether replication past these lesions does in fact depend on REV1 but, if this proves to be the case, a second Rev1p function must presumably be employed. To obtain more direct evidence for the existence of this postulated second function, we have therefore determined the bypass frequencies of these lesions, and also of an abasic site, in yeast strains deleted for REV1 or REV3 or carrying the rev1-1 mutation. The deleted strains, together with the wild type, were transformed with duplex vectors that contained one of these lesions located centrally within a 28 nucleotide single-stranded region. The proportion of replicated plasmids resulting from translesion replication and the identity of the nucleotide insertion or deletion events that had occurred during bypass were then determined by DNA sequence analysis. The existence of a second Rev1p function was evident in the data for the T–T (6–4) adduct and abasic site. With the latter lesion, differentiation between the two Rev1p activities was achieved with the aid of the rev1-1 mutant; we show that Rev1-1p possesses a substantial level of transferase activity, but is markedly deficient in the second function. Replication past the T–T cyclobutane dimer required neither the two Rev1p functions nor that of Rev3p. Bypass of this lesion is presumably carried out by a different DNA polymerase.
Our aim in this work was to investigate whether the functions of the yeast REV1 and REV3 genes were required for replication past an abasic site, T–T cis–syn cyclobutane dimer or T–T pyrimidine (6–4) pyrimidinone adduct and, if so, to determine the nature of the nucleotide insertions (or other genetic events) that occurred, as a means of examining whether the Rev1p dCMP transferase activity might be involved in the bypass. To this end, we measured the frequency of replication past these three lesions with the aid of duplex vectors that possessed a 28 nucleotide single-stranded region within which one of these lesions was located at a designated site, together with lesion-free but otherwise identical control constructs (Gibbs and Lawrence, 1995). Unlike single-stranded constructs (Gibbs et al., 1993), these vectors contain a duplex origin of replication and are therefore replicated efficiently. Placing the lesion in a single-stranded region makes it immune to repair and ensures that fully replicated molecules are genetically homogeneous. Strain CL1265-7C, its isogenic mutant derivatives carrying rev1 or rev3 deletions, or CL1212–7D carrying the rev1-1 mutation, were transformed with a lesion-containing vector, and the bypass frequency was estimated from the number of resulting colonies normalized to the number obtained by transforming with an equal amount of the control (lesion-free) vector. The nucleotides inserted opposite the lesion were determined by sequence analysis in a sample of the replicated vector products. Sequence analysis also identified transformants that arose without lesion bypass having taken place; most arose from the small fraction of vector molecules that remained uncut or retained the 28-mer needed for construction of the vector, but normally removed by denaturation. The latter were detected by the presence of a C–C mismatch opposite the lesion. The proportions of transformants resulting from such incorrectly constructed vector molecules were estimated from the sample subjected to sequence analysis and excluded from the calculations of bypass frequencies. Because the proportion of incorrectly constructed molecules was small, corrections to the control frequencies were negligible. In the mutants, however, in which bypass frequencies are very low, these corrections were substantial.
We found that REV1 function was required for the bypass of an abasic site, as might be expected if this depended on the deoxycytidyl transferase activity, but also that it was required for replication past a T–T pyrimidine (6–4) pyrimidinone, in which the transferase activity has not been implicated (Table 1). Interestingly, replication past a T–T cyclobutane dimer requires the function of neither of the REV genes; another polymerase presumably performs the bypass. Both the abasic site and the (6–4) adduct appreciably inhibit translesion replication in the wild-type strain, and bypass occurs in only 25% and 19% of vector molecules that carry these lesions respectively. However, this proportion drops to only 1–3% in rev1Δ or rev3Δ mutants. In some of the transformants in these strains, a deletion of one or more nucleotides occurred at the site of the lesion, possibly resulting from a misalignment mechanism without nucleotide insertion opposite the lesion, rather than translesion replication in the strict sense. If these events are excluded, the bypass frequency (given in parentheses in Table 1 and headed no Δ nuc) drops to 1% or less in the rev1Δ and rev3Δ mutants, emphasizing the important role of the Rev1 and Rev3 proteins in replication past these strongly blocking lesions.
Table 1. Bypass frequencies of an abasic site, T-T cis-syn cyclobutane dimer, and T-T pyrimidine (6–4) pyrimidinone adduct in REV + , rev1Δ, rev1–1, and rev3Δ strains
Bypass frequencies are the unweighted average of from two to six replicate experiments. Total number of colonies counted resulting from transformation with the control (lesion-free) construct for the abasic site experiments were: REV+, 514, 489, 322, 878, 1346, 630; rev1Δ, 433, 191, 45, 144, 1763, 773; rev1-1, 101, 295; rev3Δ, 428, 199, 660, 1019. Control counts for the T–T dimer experiments were: REV+, 290, 240; rev1Δ, 232, 99; rev1-1, 79, 164; rev3Δ, 297, 93. Control counts for the T–T (6–4) experiments were: REV+, 780, 7744; rev1Δ, 1438, 5348; rev1-1, 16736, 326; rev3Δ, 2422, 278.
. Bypass frequencies excluding deletion events.
. Number of bypass events analysed with insertions.
. Number of bypass events analysed with deletions.
In the case of the abasic site, the role of Rev1p in bypass appears to depend, at least in part, on its dCMP transferase activity (Table 1); 16/18 (89%) of the bypass events analysed in the REV+ strain resulted from the insertion of dCMP opposite the abasic site, with dAMP insertion occurring in the remaining 2/18 (11%) of events. These results are in good agreement with those observed previously (Gibbs and Lawrence, 1995), in which the respective values were 83% and 13%. Further, no dCMP insertion was detected in the rev1Δ mutant; among the few bypass events occurring in this strain, all those analysed (8/8) resulted from dAMP insertion. Interestingly, the same was true in the rev3Δ mutant, even though this strain is REV1+ and therefore has wild-type levels of the transferase; presumably, DNA polymerase ζ, of which Rev3p is the catalytic subunit, is required for extension of the terminus created by the addition of dCMP opposite the abasic site in the template. Pol ζ may also be responsible for the bypass events in which dAMP is inserted opposite the abasic site, as about threefold more of these occur in the REV+ strain than in the rev3Δ mutant (REV+, 23% bypass × 11% dA = 2.5%; rev3Δ, 0.9% bypass × 100% dA = 0.9%; 2.5/0.9 ≅ 3), although a larger data set will be needed firmly to establish this conclusion. In contrast to its role in the bypass of an abasic site, however, the Rev1p transferase function does not appear to play a significant part in replication past the T–T (6–4) photoadduct, as dCMP insertion opposite the photoproduct occurs only rarely. In 50 of the 72 transformants analysed (69%), dAMP was correctly inserted opposite both of the thymines, with dGMP and dTMP being inserted opposite the 3′ thymine in 18 (25%) and three (4%) transformants respectively. Only a single instance of dCMP insertion was detected, in this case opposite the 5′ thymine. Again, such results are similar to those obtained previously (Gibbs et al., 1995).
These data point strongly to the existence of a second Rev1p function and raise the question of whether it too plays a role in the bypass of the abasic site, together with the transferase activity. Evidence from the rev1-1 mutant (G-193R) shows that this is indeed the case (Table 1). Even though the bypass frequency of the abasic site in the rev1-1 strain was not as low as in the rev1Δ strain, it was nevertheless much reduced, indicating that some REV1 function is impaired. However, the rev1-1 mutant appears to retain some deoxycytidyl transferase activity. In contrast to the absence of dCMP incorporation seen in the rev1Δ strain, three out of four of the bypass events analysed in the rev1-1 mutant result from the insertion of dCMP. Primer extension assays of the transferase activity of Gst-Rev1p and Gst-Rev1-1p confirmed that this is indeed the case (Fig. 1). The substrate for the reactions was formed by annealing a 5′32P-labelled 20-mer to a 71-mer template containing an abasic site at the first position for incorporation. Reactions using 0, 6, 30 and 150 ng of the fusion proteins and 100 µM dCTP were incubated for 2.5 min, and the amount of primer and its +1 nucleotide extension product were quantified by phosphorimaging. These results indicate that Rev1-1p retains about 60% of the transferase activity. The G-193R mutation lies within a postulated BRCT domain, which may be concerned with protein–protein interactions (Gibbs et al., 2000).
We find that Rev1p possesses a second, as yet poorly defined, function in addition to its deoxycytidyl transferase activity. Whereas the transferase activity seems to be mainly, perhaps exclusively, concerned with the insertion of dCMP opposite abasic sites in the DNA template, the second function probably plays a role in the bypass of a broad range of lesions that have the potential to block replication. As shown by the data in Table 1, it is clearly used in the bypass of an abasic site and T–T (6–4) lesion. Moreover, rev1 mutants are deficient with respect to mutagenesis induced by an array of mutagens that includes ionizing radiation and chemical mutagens in which dCMP insertion is minor (Prakash, 1976; McKee and Lawrence, 1979a, b). The enzymatic basis for the second Rev1p function is not yet known. One possibility is that Rev1p interacts transiently with pol ζ and stimulates this enzyme to insert nucleotides opposite replication-inhibiting lesions, or enhances primer extension. No evidence for the formation of a stable association of Rev1p with pol ζ could be found by purifying pol ζ from cells simultaneously expressing Rev1p, Gst-Rev3p and Rev7p but, as each is derived from cloned genes, there is no guarantee that all subunits have been identified; indeed, indirect evidence suggests that other subunits may exist. Rev1p may therefore be incorporated into a larger complex with pol ζ by interaction with a subunit that remains to be discovered. Even though other members of the UmuC, DinB, Rev1 superfamily are polymerases that use all four nucleotide triphosphates (Gerlach et al., 1999; Johnson et al., 1999; Masutani et al., 1999; Tang et al., 1999; Wagner et al., 1999), there is no evidence to suggest that the second Rev1p function constitutes the addition of dAMP, dGMP or dTMP. In contrast to other members of the superfamily, Rev1p was essentially incapable of primer extension with these nucleotides on any template tested so far, and also appeared to be unable to insert any nucleotide opposite a T–T dimer (Nelson et al., 1996a). Although it is possible that nucleotides other than dCMP are specifically incorporated opposite some specific lesions, this seems unlikely. Whatever the nature of the second function, it seems to be concerned more with nucleotide incorporation than with elongation of misaligned primers. REV1 function was required for UV-induced base substitution mutation at 10 out of 12 genetic sites tested, but only three out of eight sites tested for UV-induced frameshift mutation (Lawrence and Christensen, 1978; Lawrence et al., 1984). In contrast, REV3 function was required for both kinds of events at all sites tested (Lawrence and Christensen, 1979; Lawrence et al., 1984). If, as appears likely, the second function of Rev1p is not in fact a polymerase activity, it will be of interest to investigate other members of the superfamily for the presence of this function, both those within the REV1 subgroup and those in the umuC and dinB subgroups, with which the REV1 genes share significant identity. Finally, neither of the Rev1p functions, nor that of Rev3p, appear to be used in the bypass of a T–T dimer, and thus another enzyme must be used for this purpose. A likely candidate is the recently discovered pol η, encoded by RAD30, which has been shown to replicate efficiently past a T–T dimer in vitro (Johnson et al., 1999). As is evident from Table 1, yeast pol η clearly does not bypass either abasic sites or T–T (6–4) lesions in vivo. A comparison between the phenotypes of rev3 and rad30 mutants suggests that the substrate range for pol η may be quite small. A specialized DNA polymerase devoted to replication past pyrimidine dimers might be biologically advantageous because UV is a ubiquitous environmental mutagen, and dimers are a common photoproduct. Moreover, dimerized pyrimidines are capable of correct basepairing, offering the possibility of accurate bypass.
Construction of the site-specifically modified vectors
Duplex vectors with a 28 nucleotide single-stranded region, within which a lesion was located at a specific site, were constructed by ligating lesion-containing 36-mers into plasmid pYDV1 (Gibbs and Lawrence, 1995) that had been linearized by digestion with EcoRI (which has a 5′ overhang) and PstI (which has a 3′ overhang). pYDV1 was constructed by inserting a 2 kb fragment comprising the 0.9 kb HincII–XbaI 2 µ plasmid minimal ori and the 1.1 kb HindIII URA3 fragment into the FspI site of M13mp7L1 (Banerjee et al., 1988), and replacing the polylinker region with a 24 nucleotide sequence that preserved one of the EcoRI sites and introduced a single PstI sequence. Efficient insertion of the 36-mer into the pYDV1 vector was achieved using a two-step ligation (Fig. 2). In the first step, a fivefold molar excess of kinased 36-mer annealed to an unkinased oligomer complementary to its central 28 nucleotides was ligated at a total DNA concentration of ≈ 800 ng µl−1 to EcoRI-linearized, dephosphorylated pYDV1 plasmid DNA. Restriction analysis of a sample of the product of this reaction showed that, under these conditions, a single copy of the 28-mer/36-mer insert was ligated to > 95% of each of the two ends of the linearized vector. The reaction product from this first stage was then cut with PstI, and recircularization was achieved by a second ligation reaction at a total DNA concentration of 2.5 ng µl−1; at this low concentration, ligations involving two vector molecules are unlikely. About 50–80% of the vector molecules were recircularized. Ligation efficiency was enhanced by carrying it out in the presence of MunI at the first stage, and PstI at the second. Because the 28-mer/36-mer insert was designed to have a MunI cohesive end, which is compatible with an EcoRI end but has a different flanking basepair, this procedure helped to maintain the concentration of insert monomer by cutting any dimers that formed. Ligation junctions with the vector were immune to restriction. The presence of PstI at the second stage prevented religation of the released fragment and intermolecular ligation. The desired ligation junction of insert and the PstI end of the vector was again designed to be immune to the restriction enzyme. The sequence of the 36-mer was 5′-AATTGATTCAGTGGCAAGTTGGAGAATTCACTTGCA-3′ in all cases, with lesions placed at the underlined T–T site. The sequence of the 28-mer was complementary to the central sequence of the 36-mer, except that it possessed a C–C mismatch opposite the T–T site. To produce abasic site-containing oligonucleotide, the 36-mer was synthesized with uracil replacing the 5′ thymine at the lesion site, and the construct was treated with E. coli uracil N-glycosylase after both ligation steps had been completed. More than 99% of the vector molecules contained the abasic site, as judged by its inability to form plaques when transfected into non-SOS-induced E. coli, under conditions in which lesion-free vector produced many plaques. The 36-mers containing UV photoproducts were constructed by ligating a flanking 13-mer to the 5′ end of an 11-mer carrying the photoproduct, ligating a flanking 12-mer to the 3′ end and purifying the fully ligated species by gel electrophoresis. Oligonucleotides containing a T–T dimer or T–T (6–4) adduct were > 99.5% and > 98% pure, respectively, and were produced as described previously (Banerjee et al., 1988; Lawence et al., 1990; LeClerc et al., 1991). Further details of the vector construction method can be found in Gibbs and Lawrence (1995).
Strains, transformation and sequence analysis
Cells of strain CL1265-7C (MATα, REV+arg4-17 his3Δleu2-3,12 trp ura3-52), its isogenic rev1Δ or rev3Δ derivatives or of strain CL1212-7D (MATαrev1-1 arg4-17 his5-2 lys1-1 ura3-52), were made competent with LiAc and transformed with 25 ng of lesion-containing construct or its lesion-free control counterpart, together with carrier DNA (Schiestl and Gietz, 1989). Construct DNA samples were quantified with Hoechst 33258 and fluorimetry. Before transformation, construct DNA was heated at 85°C in the presence of 1000-fold molar excess of the complement of the 28-mer to denature the 28-mer from the gap and prevent it from reannealing. The 28-mer had a C–C mismatch opposite the T–T site (or, in the case of the abasic site X, the XT site), so that failure to denature or prevent reannealing could be detected by the presence of a C–C sequence in the template strand of the replicated vector products. This genetic marker is very reliable, as tandem double mutations are extremely rare, and no difference was observed with an A–A mismatch (Gibbs and Lawrence, 1995), which is less desirable, however, because it disrupts annealing to a greater extent. Templates for sequence analysis were obtained by transfecting JM101 with DNA extracted from transformants and sequence analysed by hybridization and the dideoxy method. Further details of the methods used can be found in Gibbs et al. (1993) and Gibbs and Lawrence (1995).
Assay of Rev1-1p transferase activity
The G-193R mutation found in rev1-1 was transferred to the Gst–REV1 fusion gene by plasmid gap repair, the presence of the mutation verified by sequencing, and Gst-Rev1p and Gst-Rev1-1p fusion proteins purified as described previously (Nelson et al., 1996a). Protein was quantified by Coomassie staining, and 0, 6, 30 or 150 ng of either Rev1p or Rev1-1p was added to 10 µl reactions containing 25 mM potassium phosphate buffer (pH 7.4), 6 mM MgCl2, 1 mM dithiothreitol, 0.1 mg ml−1 acetylated BSA, 10% glycerol, 20 nM 5′32P-labelled 20-mer primer annealed to a 71-mer template. The primer was 5′-CGACGGCCAGTGAATTCTCC-3′, and the template was 5′-ACAGGAAACAGCTATGACCATGATTCAGTGGCAAGTOGGAGAATTCACTGGCCGTCGTTTTACAACGTCGT-3′, in which O is the abasic site. Reactions were incubated at 30°C for 2.5 min, terminated with 20 ml of 20 mM EDTA, 95% formamide, and products were analysed on 12% polyacrylamide gels. Phosphorimaging was used to quantify the primer and +1 nucleotide product bands.
We thank Susan Wallace (University of Vermont, Burlington, USA) for her generous gift of purified E. coli uracil glycosylase. This work was supported by grant no. GM21858 from the National Institutes of Health.