The Relative Roles in Vivo of Saccharomyces cerevisiae Pol η, Pol ζ, Rev1 Protein and Pol32 in the Bypass and Mutation Induction of an Abasic Site, T-T (6-4) Photoadduct and T-T cis-syn Cyclobutane Dimer
Peter E. M. Gibbs, John McDonald, Roger Woodgate, Christopher W. Lawrence


We have investigated the relative roles in vivo of Saccharomyces cerevisiae DNA polymerase η, DNA polymerase ζ, Rev1 protein, and the DNA polymerase δ subunit, Pol32, in the bypass of an abasic site, T-T (6-4) photoadduct and T-T cis-syn cyclobutane dimer, by transforming strains deleted for RAD30, REV3, REV1, or POL32 with duplex plasmids carrying one of these DNA lesions located within a 28-nucleotide single-stranded region. DNA polymerase η was found to be involved only rarely in the bypass of the T-T (6-4) photoadduct or the abasic sites in the sequence context used, although, as expected, it was solely responsible for the bypass of the T-T dimer. We argue that DNA polymerase ζ, rather than DNA polymerase δ as previously suggested, is responsible for insertion in bypass events other than those in which polymerase η performs this function. However, DNA polymerase δ is involved indirectly in mutagenesis, since the strain lacking its Pol32 subunit, known to be deficient in mutagenesis, shows as little bypass of the T-T (6-4) photoadduct or the abasic sites as those deficient in Pol ζ or Rev1. In contrast, bypass of the T-T dimer in the pol32Δ strain occurs at the wild-type frequency.

AS well as a variety of processes that repair damage to its genome, the yeast Saccharomyces cerevisiae possesses mechanisms that lead to the tolerance of DNA damage by promoting continued elongation at replication forks that have stalled when encountering unrepaired lesions. Translesion replication, one of the mechanisms for such continued elongation, employs DNA polymerases Pol η (Johnson et al. 1999), encoded by RAD30 (McDonald et al. 1997), and Pol ζ (Nelson et al. 1996a), encoded by REV3 (Lemontt 1971) and REV7 (Lawrence et al. 1985a), which are specialized enzymes required for bypassing lesions that cause the replicase to stall. In addition to these enzymes, translesion replication often requires Rev1 protein, which appears to be essential for Pol ζ activity (Lawrence and Hinkle 1996; Lawrence 2002) and also possesses a deoxycytidyl transferase activity (Nelson et al. 1996b), and Pol32p, a subunit of DNA polymerase δ that is required for induced mutagenesis (Huang et al. 1997, 2000). The various functions of these proteins in translesion replication are still unclear. On the basis of in vitro studies of the enzyme, it has been proposed that Pol ζ is an accurate DNA polymerase incapable of nucleotide insertion opposite lesions, and that its role in template damage bypass is restricted to extension from nucleotides inserted by other polymerases (Johnson et al. 2000). Pol η, in contrast, inserts nucleotides relatively accurately opposite lesions capable of base pairing, such as T-T cyclobutane dimers, and efficiently extends past them (Johnson et al. 1999; Washington et al. 2000). It also inserts nucleotides opposite other lesions, but cannot extend from these termini. A number of these are misinsertions, making Pol η responsible for a high proportion of mutations.

Several observations, however, appear to cast doubt on this model. All enzymatic studies of Pol ζ have been carried out in the absence of a functional association with Rev1p, required for Pol ζ function in vivo (Lawrence and Hinkle 1996; Lawrence 2002), and also other factors occurring in the cell that may enhance its activity, such as proliferating cell nuclear antigen (PCNA), replication factor C (RFC), and replication protein A (RPA), which have been shown to enhance the activity of Pol η (Haracska et al. 2001b). Further, extension by Pol ζ from a Pol η misinsertion opposite the 3′ nucleotide of a bipyrimidine lesion, as suggested for a T-T (6-4) UV photoadduct (Johnson et al. 2001), necessarily requires insertion opposite the 5′-thymine of the lesion by Pol ζ, indicating that this enzyme can, in fact, insert nucleotides opposite a lesion. In addition, cells lacking Pol ζ are almost entirely incapable of DNA damage-induced mutagenesis, whereas in those lacking Pol η the reduction is much smaller (Lawrence 2002). Although we (Lawrence and Hinkle 1996; Lawrence et al. 2000) and others (Johnson et al. 2000) have shown that Pol ζ has a marked facility for extension from terminal mismatches, consistent with its role in the proposed model, these observations do not preclude a role for this enzyme in nucleotide insertion opposite a lesion.

In addition to Pol η and Pol ζ, translesion replication also often requires the function of REV1 and POL32, but the molecular basis for their requirement is not known. To further investigate some of the above issues, we have transformed yeast strains deleted for RAD30, REV3, REV1, or POL32 with duplex vectors that carry an abasic site, T-T (6-4) photoadduct, or T-T cis-syn cyclobutane dimer specifically located within a 28-nucleotide single-stranded region, and determined the lesion bypass frequency and spectrum of nucleotide insertions opposite the lesions. We find that, as expected, Pol η is entirely responsible for the bypass of the T-T dimer and does so accurately. However, Pol η plays at best only a small part in replication past the other lesions. Finally, we find that Pol32p is required for the bypass of the T-T (6-4) photoadduct and abasic sites, but not for the bypass of a T-T cyclobutane dimer.


Yeast strains and plasmids:

All yeast strains used in these experiments were derived from the strain CL1265-7C (MATα arg4-17 his3Δ-1 leu2-3,112 trp1 ura3-52). Deletions of the rev3, rev1, and rad1 genes were introduced as described (Alani et al. 1987), using suitable plasmid constructs. The rad30Δ::HIS3 mutation was introduced using the plasmid pJM82 (McDonald et al. 1997), and the pol32Δ::KanMX mutation was created using a plasmid described (Huang et al. 1997) for the yeast open reading frame YJR043c. Sequence analysis of the POL32 gene in tetrads segregating for the rev5-1 mutation (Lawrence et al. 1985b) has recently shown that this is an allele of POL32 (M. Villasmil and P. E. M. Gibbs, unpublished data). The plasmid pYPOG1, used in most experiments, is derived from pUC19 with modification of the polylinker sequence and insertion of the URA3-2μ ori cassette from pYDV1 at the unique AatII site. A few of the initial experiments used the M13mp7RF-based plasmid pYDV1 (Gibbs and Lawrence 1995). The two plasmids gave essentially identical results.

Construction of gapped-circular plasmids:

Gapped-circular plasmids containing a uniquely placed lesion within a 28-nucleotide single-stranded region of an otherwise duplex plasmid were constructed essentially as described (Gibbs and Lawrence 1995). Briefly, 36-mers containing a specifically located T-T cyclobutane dimer, T-T (6-4) photoadduct, or dUMP nucleotide were constructed by ligating together a 13-mer, 11-mers carrying these entities, and a 12-mer, using a 50-mer scaffold, followed by purification of the full-length species by gel electrophoresis. The 11-mers containing the UV photoproducts were produced and purified as described (Banerjee et al. 1988; Leclerc et al. 1991) and were >99.5% pure. These 36-mers were then annealed to an unphosphorylated complementary 28-mer and ligated into EcoRI-PstI-digested pYPOG1, using a two-step process that promoted ligation efficiency (Gibbs and Lawrence 1995). Following denaturation of the 28-mer in the presence of a molar excess of its complementary 28-mer, gapped-circular material was digested with EcoRI to linearize the small fraction of molecules that retain the 28-mer and then purified by electrophoresis through a 0.8% agarose gel (SeaKem GTG; Cambrex). The desired band, which comigrates with the nicked-circular form of the vector, was excised from the gel and purified using Qiaquick spin columns (QIAGEN, Valencia, CA). The concentration of the recovered plasmid DNA was determined by picogreen fluorometry (Molecular Probes, Eugene, OR) using a Turner Quantech fluorometer. A few preliminary experiments were carried out with unpurified construct. These gave similar experimental results except for a higher background of transformants arising from incorrectly assembled plasmid constructs, which were identified by sequence analysis and excluded from the reported results. Abasic sites were introduced into the purified dUMP-containing gapped circular molecules by treatment with uracil N-glycosylase (New England Biolabs, Beverly, MA) immediately prior to transformation of the yeast strains.

Yeast transformation and sequence analysis:

Yeast cultures grown in 1% yeast extract, 2% peptone, 2% dextrose liquid medium to early log phase were made competent with lithium acetate and transformed with 5-ng aliquots of gapped-circle plasmids. Samples of these cells were plated on synthetic medium lacking uracil and grown for 3 days at 30°. Lesion bypass efficiency was estimated by normalizing the number of colonies resulting from transformation with lesion-containing plasmids to the number obtained with an equal quantity of lesion-free construct, assembled by an identical procedure. Bypass frequencies were corrected for the very low frequency of transformations resulting from plasmids lacking a lesion, either because of errors of construction (detected by sequence analysis) or from a small fraction of lesion-free contaminants in the 11-mer samples (estimated by transformation of uninduced cells of the Escherichia coli strains SMH10 or SMH99); on average, the latter fractions were 0.3% in oligonucleotides carrying T-T (6-4) photoadducts, 0.4% in those carrying abasic sites, and 1.1% in those carrying T-T dimer. Experimental data were collected using at least three independently assembled, purified, and quantitated samples of construct, to diminish any bias in the bypass estimates resulting from minor differences of inherent transformability of the plasmids, presumably caused by small amounts of DNA damage incurred during construction.

Sequence analysis was carried out on plasmid DNA extracted as described (Gibbs and Lawrence 1995) from cells derived from single-transformant colonies grown in patches on selective media for 3–4 days. DNA in the aqueous phase of the initial extract was concentrated approximately sixfold by precipitation with an equal volume of 2-propanol, centrifugation, and solubilization of the pellet in tris-EDTA buffer pH 8. An aliquot of this DNA was amplified by PCR, using the primers 5′ CGCAACTG TTGGGAAGGG 3′ and 5′ GCGCCCAATACGCAAACC 3′; the samples were incubated at 94° for 2 min, followed by 40 cycles of 94° for 20 sec, 54° for 10 sec and 72° for 1 min. The PCR products were then screened by hybridization with a 15-mer probe specific for the TT sequence as described (Gibbs and Lawrence 1995) to detect the nonmutant sequence at the lesion site and by direct sequencing of samples that failed to hybridize. PCR products were purified prior to sequencing using Qiaquick columns, either on single columns by centrifugation or in a 96-well format on a vacuum manifold. The sequencing reactions were carried out with the primer 5′ CTTCGCTATTACGCCAGCTG 3′ and the Big-Dye sequencing reagent from Applied Biosystems (Foster City, CA). The sequencing reagent was used at one-quarter of the recommended concentration, with appropriate buffer supplementation with a dilution buffer as recommended by the manufacturer. For the few experiments using pYDV1, an aliquot of the aqueous phase was used to transfect the E. coli strain JM101, which had been made competent for DNA uptake using CaCl2. Single plaques from each transfection were then analyzed by hybridization followed by sequencing of those plaques that failed to hybridize.


We have investigated the relative roles in vivo of Pol ζ, Pol η, Rev1p, and Pol32p, a subunit of DNA polymerase δ, in replication past an abasic site, a T-T (6-4) photoadduct, and a T-T cis-syn cyclobutane dimer by transforming an isogenic series of strains with a duplex vector that carried one of these lesions within a 28-nucleotide single-stranded region. Each lesion was placed in the same sequence context, with abasic residues located within the same T-T site as the UV photoproducts, replacing either the 5′- or 3′-thymine. We obtained two kinds of information from these experiments: the frequency of lesion bypass and the types of nucleotide insertion or other event that occurred during translesion replication. The lesion bypass frequency was estimated from the number of transformants normalized to the number of transformants obtained using an equal amount of lesion-free plasmid; transformation cannot occur unless the 28-nucleotide gap is filled, which requires lesion bypass. To improve the precision of these estimates, we not only employed fluorimetry to accurately quantify the plasmid DNA, but also used at least three independently constructed batches of plasmid, to minimize any bias that might arise from random differences in the inherent transformability of the plasmid DNA resulting from the small amounts of spontaneous DNA damage that occur during construction. We also corrected lesion bypass frequencies for the very small fraction attributable to contaminating lesion-free plasmids. Because the bypass frequencies of the different lesions vary considerably in the wild-type strain, ranging from 3.3% for the T-T (6-4) photoadduct to 60% for the T-T dimer, we express the frequencies in the mutant strains as a percentage of the wild-type frequency; this procedure provides a clearer picture across lesions of the relative contributions of the different gene products examined. The types of nucleotide insertions accompanying lesion bypass were determined by sequence analysis of replicated plasmids.

Using this procedure, we find that DNA polymerase η rarely contributes to the bypass of the T-T (6-4) photoadduct or abasic sites but, as expected, is solely responsible for the bypass of the T-T dimer. As shown in Table 1, the majority of events bypassing either a T-T (6-4) photoadduct or an abasic site depend on REV3 and REV1, but not on RAD30. With regard to the T-T (6-4) photoadduct, the frequency of bypass in the rad30Δ strain was reduced by only 7.5% from the wild-type frequency, compared to reductions of 97 and 99% in rev3Δ and rev1Δ strains, respectively. Somewhat greater reductions were observed for bypass of the abasic lesions in the rad30Δ strain, although they were still small. In the rad30Δ strain, the bypass frequency of the 5′-abasic residue (O-T) was reduced by 13% and that of the 3′ lesion (T-O) by 19%. In the rev3Δ strain, however, these frequencies were reduced by 96 and 95% respectively, and in the rev1Δ strain the reductions were 97 and 96%, respectively.

View this table:

Bypass frequencies, expressed as percentage of the wild-type frequency, for abasic sites, T-T (6-4) photoadducts, and T-T cyclobutane dimers in rad30Δ, rev3Δ, rev1Δ, and pol32Δ strains

The spectra of nucleotide insertions opposite these lesions (Table 2) also support the conclusion that Pol η is rarely involved in the bypass the T-T (6-4) photoadduct and abasic sites in vivo. The spectra of insertions opposite the 3′-thymine of the T-T (6-4) lesion were similar in the wild-type and rad30Δ strains in showing a preponderance of dAMP insertions and only a small fraction of dGMP insertions. Frequencies of dGMP insertions were lower in the rad30Δ than in the wild-type strain (4 vs. 10%), but not eliminated, indicating that they can also be generated by an enzyme other than Pol η, and the proportion of dAMP insertions is increased (92 vs. 85%). Overall, the two insertion spectra are significantly different (P < 0.001), although the differences are small and indicate that Pol η plays only a minor role in the bypass of the T-T (6-4) photoadduct in this sequence context. Spectra in the rev3Δ and rev1Δ strains are similar to the rad30Δ spectrum, although the frequencies of dAMP insertions are even higher (rev3Δ, 98%; rev1Δ, 97%; rad30Δ, 92%). It is possible, however, that some of the dAMP insertions seen in the rev3Δ and rev1Δ strains may arise from a low level of contamination of the T-T (6-4) insert with the undamaged sequence; at the very low bypass frequencies found in these strains a low level of such contamination may have an appreciable influence.

View this table:

Nucleotide insertions opposite the 3′-site of a T-T (6-4) photoadduct, abasic sites, and the 3′-site of a T-T cyclobutane dimer

The nucleotide insertion spectra for abasic sites also indicate that Pol ζ together with Rev1p is responsible for most bypass events. Nucleotide insertion frequencies for the 5′ abasic site are virtually identical in the wild-type and rad30Δ strain, but very different in the rev3Δ and rev1Δ strains. In the wild-type and rad30Δ strains by far the largest fraction of bypass events results in the incorporation of dCMP opposite the lesion (wild type, 91%; rad30Δ, 92%) whereas dAMP insertion occurs only infrequently (wild type, 7%; rad30Δ, 7%), as we observed previously (Gibbs and Lawrence 1995; Nelson et al. 2000). In marked contrast, no bypass events resulting in dCMP incorporation were observed in the rev3Δ and rev1Δ strains, almost all entailing dAMP incorporation (rev3Δ dAMP, 100%; rev1Δ dAMP, 96%), reflecting the absence or ineffectiveness of the Rev1p deoxycytidyl transferase activity in these strains. At the 3′-abasic site, insertion of dCMP was also the most frequent event in wild-type and rad30Δ strains (61 and 75%, respectively), with smaller frequencies of dAMP insertion (32 and 21%, respectively. The differences between the two strains, even though small, are nevertheless significant (P = 0.02), suggesting that Pol η competes with Rev1p in ∼10–15% of bypass events.

Although Pol η is engaged in bypass of the T-T (6-4) photoadduct or of the abasic sites only rarely, it is, as expected (Johnson et al. 1999), entirely responsible for the bypass of the T-T dimer. Interestingly, however, it is not essential for such bypass. As seen in Table 1, ∼15% of the lesions were bypassed in the absence of Pol η (rad30Δ strain) and, from the absence of bypass in the rad30Δ rev3Δ (0%) and near absence in the rad30Δ rev1Δ strains (1.0%), it is clear that most of this bypass requires both Pol ζ and Rev1. Such a bypass can be inaccurate (see Table 2), with misinsertions of both dGMP and dTMP, giving 3′ T → C and 3′ T → A mutations, the same types of mutations as generated by PolV in E. coli (Banerjee et al. 1988). Among 557 transformants analyzed, there were 30 of the former and 20 of the latter, for a mutation frequency of ∼9%. Nevertheless, even though Pol ζ and Rev1p are concerned with replication past a T-T dimer in the absence of Pol η, they are unlikely to be so in its presence. A single 3′ T → C, but no 3′ T → A, mutation was observed in a total of 1918 transformants analyzed in a RAD30+ strain (Gibbs et al. 1993; P. E. M. Gibbs, unpublished data). Given that such substitutions occur in 9% of bypass events employing Pol ζ and Rev1p, these data suggest that, in the presence of Pol η, no more than ∼0.6% of bypass events utilize the former enzymes.

In addition to the relative roles of Pol η, Pol ζ, and Rev1 in the bypass of the lesions, we also examined the function of Pol32, a subunit of Pol δ that has been shown to be involved in lesion bypass and mutagenesis (Huang et al. 2000). As can be seen in Table 1, the frequency of bypass of the abasic sites in the pol32Δ strain (3.6 and 4.7%) was reduced to much the same level seen in the rev3Δ (5.5 and 4.3%) and rev1Δ strains (2.9 and 3.8%), indicating the essential role of the subunit in these events. Bypass frequencies for the T-T (6-4) photoadduct in these strains were also comparable (pol32Δ, 0.0%; rev3Δ, 3.2%; rev1Δ, 1.0%). Pol ζ activity therefore appears to require, either directly or indirectly, the Pol32 subunit of Pol δ. In turn, this suggests that stalling by Pol δ at template lesions triggers bypass replication on the plasmid, as it does on the chromosome. Interestingly, however, Pol32 is not required for the bypass of the T-T dimer by Pol η; in the pol32Δ strain, the bypass frequency for this lesion (112.1%) is no different from that in the wild-type strain (Table 1). Similarly, deletion of POL32 has no effect on the highly accurate bypass of the dimer (Table 2).


It has been proposed that Pol ζ and Pol η have distinctly different and nonoverlapping roles in lesion bypass (Johnson et al. 2000). According to this model, Pol ζ is an accurate enzyme that incorporates dNTPs opposite DNA lesions only very inefficiently and therefore is incapable of replicating past them. Indeed, from a steady-state kinetic analysis of nucleotide misincorporation on undamaged templates by Pol ζ, it was suggested that this enzyme is more accurate than Pol α and has a fidelity approaching that of Pol δ, despite having no associated 3′ → 5′ proofreading activity (Johnson et al. 2000). At the same time, incorporation of NTPs opposite a T-T dimer, T-T (6-4) photoproduct, or abasic site was found to be very inefficient. However, inefficiency of insertion opposite some lesions, such as the 3′-nucleotide of a T-T (6-4) photoadduct, does not necessarily exclude Pol ζ from this role in vivo, because bypass of this lesion is itself very inefficient. In contrast to its insertion inefficiency, Pol ζ has a marked facility for extension from terminally mismatched primers, as we and others have shown (Lawrence and Hinkle 1996; Lawrence et al. 2000; Johnson et al. 2000), suggesting a general capability for elongation from abnormal ends. From such results, it has been concluded that Pol ζ extends only from insertions opposite lesions performed by other enzymes, particularly, in yeast, by Pol η. Pol η, on the other hand, can not only efficiently bypass lesions capable of base pairing, such as cyclobutane pyrimidine dimers, and accurately insert correct nucleotides opposite them (Johnson et al. 1999; Washington et al. 2000), but also misinsert nucleotides opposite lesions that it is incapable of bypassing (Johnson et al. 2001). Completion of bypass in this circumstance therefore requires extension by Pol ζ, explaining its essential and widespread role in mutagenesis.

In support of this two-polymerase model, in vitro studies showed that Pol η inserted dGMP opposite the 3′ T of a T-T (6-4) adduct, but was unable to extend beyond this lesion; extension could, however, be accomplished by Pol ζ (Johnson et al. 2001). Insertion of dGMP at this position of the T-T (6-4) lesion generates a 3′ T → C mutation, commonly observed in vivo (Gibbs et al. 1995; Bresson and Fuchs 2002). In a context where this mutation was the only kind found, Bresson and Fuchs (2002) observed that their induction was indeed entirely dependent on Pol η. However, our results for this lesion (Tables 1 and 2), located in a different sequence context, are unlike those of Bresson and Fuchs in several respects; in the wild-type strain, only 10% rather than 60% of the bypass events entailed insertion of dGMP opposite the 3′ T, and they were accompanied by a further 5% of mutations of other kinds. Moreover, although dGMP insertion is significantly less frequent in the rad30Δ mutant, decreasing from 10 to 4%, it is not abolished in this Pol η-deficient strain. Clearly, some other relatively inaccurate enzyme is also capable of generating these insertions. Further, in keeping with these sequence data indicating that Pol η rarely inserts nucleotides opposite the 3′ T of the T-T (6-4) adduct, replication past this lesion was decreased by only 7.5% in the rad30Δ mutant (Table 1). The data of Bresson and Fuchs (2002) and of this report are more similar, however, in showing a high frequency of dAMP insertion opposite the 3′ T of the T-T (6-4) adduct in the rad30Δ mutant, which was 100% in the work of Bresson and Fuchs and 92% in these results (Table 2).

Which DNA polymerase is responsible for this insertion, as well as the misinsertions of nucleotides other than dGMP, opposite the 3′ T of the T-T (6-4) adduct in the rad30Δ mutant? If, as suggested, Pol ζ is viewed as being incapable of insertion opposite lesions (Johnson et al. 2000), this function in yeast is presumably performed by Pol δ (or possibly Pol ε). It has been further proposed that Pol δ inserts opposite a thymine glycol lesion, an incorporation event that Pol ζ appears to carry out very efficiently, because it is the first polymerase to encounter the lesion (Johnson et al. 2003). There are several reasons, however, for believing that Pol δ is unlikely to perform the Pol η-independent insertions opposite the 3′-nucleotide of a T-T (6-4) photoadduct and insertions opposite other lesions, such as a thymine glycol. As a replicase, Pol δ is highly sensitive to template defects and misinserts nucleotides only extremely rarely (Goodman et al. 1993). Moreover, replicases possess 3′ → 5′ proofreading activity, suggesting that insertions opposite a lesion could be removed. Proofreading can be a significant impediment to stable nucleotide insertion. For example, in the absence of PolV and all other DNA polymerases, proofreading-deficient (mutD5) DNA polymerase III of E. coli can replicate both efficiently and accurately past a T-T cyclobutane dimer in vivo, but its proofreading-competent counterpart cannot (Vandewiele et al. 1998). While not from a eukaryote, such an example suggests that proofreading in Pol δ might also offer an impediment to insertion opposite a lesion. In contrast to a replicase, Pol ζ is much better suited to insertion opposite sites of template damage, as attested by its demonstrated ability to extend past lesions and insert opposite a thymine glycol (Johnson et al. 2003) or the 5′-nucleotide of the T-T (6-4) photoadduct. Moreover, in addition to its lack of a 3′ → 5′ proofreading activity, Pol ζ does not appear to be an accurate enzyme, contrary to the conclusion of Johnson et al. (2000). Johnson and co-workers' suggestion that Pol ζ is more accurate than Pol α and has nearly the same fidelity as Pol δ was based on a comparison between their data from a steady-state kinetic analysis of the nucleotide insertion fidelity of Pol ζ on undamaged templates (Johnson et al. 2000) and those for mutation rates in the LacZα gene in M13mp2 replicated in vitro either by purified calf thymus Pol δ or by Hela cell Pol α (Thomas et al. 1991). A comparison between these two sets of data is perhaps open to question, as they were collected by substantially different procedures. However, of greater importance, the comparison does not appear to support the conclusion. Rather, they appear to show that Pol ζ is much less accurate than Pol δ and even slightly less accurate than Pol α. The accuracy of Pol α was greater than the accuracy of Pol ζ by 2-fold or more in 7 of the 12 template base/incoming mispaired nucleotide combinations investigated, and about the same in the remaining 5. Overall, Pol δ is >10-fold more accurate than Pol ζ, and ranges from 2-fold to >700-fold more accurate in the same set of template base/incoming mispaired nucleotide combinations. Such data suggest that Pol ζ is probably the least accurate enzyme within the B-family, as might be expected from its lack of a proofreading function and from its role in translesion replication.

In addition to exhibiting low fidelity on undamaged templates, Pol ζ also has poor fidelity when inserting opposite lesions. An illustration of this is provided by insertions opposite the 5′ nucleotide of the T-T (6-4) adduct, a site at which Pol ζ performs the insertions according to all models for translesion replication. In the wild-type strain, 2.2% (10/466) of the insertions opposite the 5′-nucleotide of the T-T (6-4) adduct (labeled “other” in Table 2) resulted in mutations, while in the rad30Δ mutant the frequency was 2.6% (12/454), together with 0.7% (3/454) single-nucleotide deletions (Δ nuc), which could be either a 3′ or 5′ event, but which was not seen in the wild type. Another example indicating that Pol ζ possesses low fidelity during lesion bypass is probably provided by the high mutation frequency (9%, see Table 2) observed at the site of the 3′ T of the T-T dimer in the rad30Δ strain, where, as argued above, Pol ζ is likely to be responsible for insertion. These results indicate that Pol ζ can insert opposite lesions and is far from accurate during lesion bypass in vivo, that is, in circumstances in which, unlike those for enzymatic studies in vitro, a functional association with Rev1p and any other factors, such as PCNA and RPA, is present. The absence of these factors in the enzymatic studies may well explain the apparent contrast between the bypass frequencies observed in vitro and in vivo. Results from one group of investigators (Johnson et al. 2000, 2001) indicated no bypass of a T-T (6-4) photoadduct by Pol ζ, whereas those from another group (Guo et al. 2001) indicated very inefficient bypass of this lesion despite the use of a substantial molar excess of enzyme over primer/template, a condition that favors bypass.

In addition to investigating the relative roles of Pol ζ and Pol η in lesion bypass, we have also examined the function in this process of the Pol δ subunit, Pol32. This subunit is required for UV-induced mutagenesis (Huang et al. 2000) and was found in a two-hybrid analysis (Huang et al. 2000) to associate with the Srs2 helicase (Rong et al. 1991). This helicase disrupts the formation of Rad51 nucleoprotein filaments, perhaps preventing blocked replication forks from generating substrates for homologous recombination repair (Krejci et al. 2003; Veaute et al. 2003). Srs2 may be unable to function in the absence of Pol32, leading to the diversion of the blocked forks into this recombination pathway or possibly generating lethal intermediates, which in either case removes all substrates for translesion replication and hence for mutagenesis. In keeping with this idea, bypass frequencies for the abasic sites and T-T (6-4) photoadduct are decreased in the pol32Δ strain by much the same extent as in the rev3Δ and rev1Δ mutants (Table 1). Interestingly, and in contrast to the other lesions, bypass of the T-T dimer is unchanged in the absence of the Pol32 subunit, suggesting that Pol η replicates past the lesion before fork collapse can occur, thereby avoiding any requirement for Srs2.

In addition to interacting with Srs2p, Pol32p is believed to interact with several other proteins involved in DNA repair and cell-cycle checkpoints, suggesting an additional model. The interacting proteins include Rad9p, the protein kinase that activates a checkpoint cascade, and three subunits of an alternate clamp loader, Ctf18p (Tong et al. 2004), Elg1p (Bellaoui et al. 2003; Tong et al. 2004), and Rad24p (Tong et al. 2004). The latter interaction is particularly intriguing since Pol32 is also thought to interact with Rad17 and Ddc1 (Tong et al. 2004), two members of the Rad17/Mec3/Ddc1 checkpoint clamp (Majka and Burgers 2003). Although the biochemical function of the checkpoint clamp is currently unknown, it is clear that it plays a significant role in UV-induced mutagenesis, as cells with mutations in Rad17, Rad24, or Mec3 are poorly mutable by UV-light (Paulovich et al. 1998). Pol32 may therefore participate in the Rad17/Mec3/Ddc1-dependent loading of pol ζ at sites of DNA damage. Both of these models imply that dependence of induced mutagenesis on Pol32p does not necessarily indicate that Pol δ is responsible for insertion opposite the lesion.

In addition to the issues above, we have also examined the question of whether mechanisms of lesion bypass on the plasmid are similar to those in the genome. Although it has been asserted that plasmids are different in this respect (Haracska et al. 2001a), a variety of data suggest otherwise. The similarities between lesion bypass on the plasmid and in the genome are underscored by the identical roles performed on them by both Pol ζ and Pol η, and by the involvement of Pol32, and hence the use of Pol δ to replicate the plasmid. As described above, Pol η is alone responsible for the bypass of the T-T cyclobutane dimer on the plasmid and, in at least some events, also inserts dGMP opposite the 3′-thymine of the T-T (6-4) photoadduct as predicted from in vitro studies (Johnson et al. 2001), but requires Pol ζ to extend from this incorporation event because it is unable to accomplish this itself. The exact parallel between the roles of these gene products on plasmid and genome strongly supports the validity of the plasmid model system.

However, the issue prompting the contrary conclusion concerned the role of the deoxycytidyl transferase activity of Rev1p in the bypass of abasic sites in genomic DNA (Haracska et al. 2001a). In contrast to plasmid constructs, in which dCMP is clearly inserted opposite an abasic site by Rev1p (Gibbs and Lawrence 1995; Nelson et al. 2000; Table 2), it was inferred that dAMP was principally incorporated opposite this lesion by Pol δ in the yeast genome (Haracska et al. 2001a). Support for this conclusion was derived from sequence analysis of forward mutations to canavanine resistance in an apn1Δ apn2Δ strain treated with methyl methanesulfonate. Unlike the use of plasmids carrying an abasic site, however, such a procedure is incapable of providing unbiased estimates of nucleotide incorporation opposite the lesion. In particular, it cannot detect the insertion of dCMP opposite this lesion when it results from the loss of a guanine base because, rather than generating a canR mutation, it restores the wild-type sequence. Since 79% of the canR mutations sequenced involved alterations at G · C base pairs (Haracska et al. 2001a), it is likely that a major category of insertion events involving dCMP insertion was not detected. Moreover, unlike results from the plasmid constructs, there is no direct evidence that the genomic mutations analyzed were necessarily induced by abasic sites. Finally, direct evidence for dCMP incorporation opposite abasic sites in yeast genomic DNA by Rev1p deoxycytidyl transferase activity has been obtained by transforming yeast strains with oligonucleotides carrying this lesion, together with sequence analysis of the genomic site in transformants (Otsuka et al. 2002; K. Negishi, personal communication). In REV1+ strains, dCMP was incorporated opposite the abasic site in 63% of the transformants. Insertion of dCMP was much reduced, not only in rev1Δ strains, but also in rev1 mutants that encoded D467A, E468A substitutions abolishing the transferase activity. The plasmid model system therefore appears to accurately reflect genomic events in each of the respects examined.


We thank Candace Brayfield, Michelle Coleman, Michelle Villasmil, Erin Zahradnik, and Kimberley Colern for assistance with these experiments and George Kampo and Laura Ascroft for electrophoretic analysis of dye-terminated sequencing products. This work was supported in part by the National Institutes of Health (NIH) intramural program and in part by grant GM60652 to C.W.L. from the NIH.


  • Communicating editor: S. T. Lovett

  • Received August 9, 2004.
  • Accepted October 25, 2004.


View Abstract