The Ability of a Variety of Polymerases to Synthesize Past Site-specific cis-syn, trans-syn-II, (6–4), and Dewar Photoproducts of Thymidylyl-(3′→5′)-thymidine*

The role of photoproduct structure, 3′ → 5′ exonuclease activity, and processivity on polynucleotide synthesis past photoproducts of thymidylyl-(3′ → 5′)-thymidine was investigated. Both Moloney murine leukemia virus reverse transcriptase and 3′ → 5′ exonuclease-deficient (exo−) Vent polymerase were blocked by all photoproducts, whereas Taq polymerase could slowly bypass the cis-syn dimer. T7 RNA polymerase was able to bypass all the photoproducts in the order cis-syn> Dewar > (6-4) > trans-syn-II. Klenow fragment could not bypass any of the photoproducts, but an exo−mutant could bypass the cis-syn dimer to a greater extent than the others. Likewise T7 DNA polymerase, composed of the T7 gene 5 protein and Escherichia coli thioredoxin, was blocked by all the photoproducts, but the exo− mutant Sequenase 2.0 was able to bypass them all in the order cis-syn > Dewar > trans-syn-II > (6-4). No bypass occurred with an exo− gene 5 protein in the absence of the thioredoxin processivity factor. Bypass of the cis-syn andtrans-syn-II products by Sequenase 2.0 was essentially non-mutagenic, whereas about 20% dTMP was inserted opposite the 5′-T of the Dewar photoproduct. A mechanism involving a transient abasic site is proposed to account for the preferential incorporation of dAMP opposite the 3′-T of the photoproducts.

Dipyrimidine sites are the major sites of UV-induced photoproducts and mutations (1)(2)(3)(4)(5)(6)(7). The four main classes of photoproducts formed by ultraviolet light at dipyrimidine sites (shown in Fig. 1 for a TpT 1 site) are the cis-syn and trans-syn (trans-syn-I (8) and trans-syn-II (9)) cyclobutane dimers and the (6 -4) pyrimidine-pyrimidone photoproducts and their Dewar valence isomers (10 -13). All of these photoproducts have been found to lead to mutations in Escherichia coli under SOS conditions by use of site-specific photoproduct-containing bacteriophage vectors, but the (6 -4) and Dewar photoproducts are far more mutagenic than either the cis-syn and trans-syn isomers (14 -18). The extent to which a particular photoproduct contributes to UV-induced mutations at a particular site not only depends on its mutagenicity, but also depends on its rate of induction, repair, and DNA synthesis bypass (7,12). At the moment, the relative contribution of individual DNA photoproducts to UV-induced mutations is not known, nor are the detailed mechanisms by which DNA photoproducts are repaired or bypassed. Recently, we have prepared homogeneous 49-mer oligonucleotides containing the four major photoproduct classes of TpT (19) 2 for use as substrates for the necessary in vitro and in vivo mechanistic studies. Herein, we report the use of these 49-mers and 72-mers containing a T7 promoter to study the role of photoproduct structure and the 3Ј 3 5Ј exonuclease activity and processivity of polymerases on DNA and RNA synthesis past these photoproducts.

EXPERIMENTAL PROCEDURES
Enzymes, Reagents, and Equipment-The preparation of the photoproduct-containing 49-mers has been reported elsewhere (19). 2 Other oligonucleotides were purchased at a local facility and purified by ion exchange high performance liquid chromatography. Oligonucleotide concentrations were measured by absorbance at 260 nm using estimated extinction coefficients (20). T4 polynucleotide kinase and exo Ϫ Vent (21) were purchased from New England Biolabs. Taq DNA polymerase, wild-type T7 DNA polymerase, Sequenase Version 1.0 (22), Sequenase 2.0 (␦28 K118-R145; Ref. 23), Klenow fragment (KF), exo Ϫ KF (D355A/E357A; Ref. 24), T7 RNA polymerase, and deoxynucleotide triphosphates were purchased from U. S. Biochemical Corp. Concentrations of commercial enzymes were calculated from data obtained from the supplier. The D5A/E7A T7 gene 5 protein (25) and E. coli thioredoxin components of T7 DNA polymerase were a generous gift of Isaac Wong and Kenneth Johnson (University of Pennsylvania). Moloney murine leukemia virus reverse transcriptase (MMLV RT) was purchased from Life Technologies, Inc. dNTPs were from Fisher, and NTPs were from Sigma. [␥-32 P]ATP (2 M, 10 Ci/l) was purchased from Amersham Pharmacia Biotech. Dideoxy sequencing mixes were prepared in Sequenase buffer (40 mM Tris⅐HCl, 10 MgCl 2 , and 5 mM DTT) with each nucleotide triphosphate at 300 M, with the eponymous nucleotide triphosphate in a ratio of 1:3 ddNTP to dNTP. Unless otherwise stated, all electrophoresis was conducted on 0.4-mm-thick, 37.5cm-long, 7 M urea, 1:19 cross-linked, 15% acrylamide gel at 1800 V. DNA fragments were visualized by autoradiography with Kodak XAR-5 film. Densitometry was performed on a Joyce-Loel Chromoscan 3 or a Molecular Dynamics computing densitometer model 300A. The percentage of a primer-elongated product is computed as the percentage of the total amount of extended products.
Primer Extension by Taq Polymerase-Primer extensions were conducted by incubating 7 nM 15-mer primer annealed to 70 nM 49-mer template with 100 or 200 M dNTPs and 0.5 units of enzyme in a total volume of 10 l (67 mM Tris⅐HCl, pH 8.8 at 25°C, 17 mM (NH 4 ) 2 SO 4 , 6.7 mM MgCl 2 , 1 mM DTT, and 20 g/ml BSA) for 30 min at 60°C. The polymerase was added last to the prewarmed solution. The reactions were quenched by the addition of 15 l of 95% formamide.
Primer Extension by KF and exo Ϫ KF-Primer extensions were conducted by incubating 7 nM 15-mer primer annealed to 70 nM 49-mer template with 100 nM enzyme (0.8 units of KF, 1.9 units of exo Ϫ KF) and 100 M dNTPs in a total volume of 10 l (50 mM Tris⅐HCl, pH 7.5, 10 mM MgCl 2 , 1 mM DTT, and 100 g/ml BSA) for 30 min at 37°C. The reactions were quenched by addition of 15 l of 95% formamide.
Primer Extension by Wild Type T7, Sequenase 2.0, and D5A/E7A Gene 5 Protein with and without Thioredoxin-Primer extension were conducted by incubating 7 nM 15-mer primer annealed to 70 nM 49-mer template with 100 M dNTPs and 100 nM enzyme (0.1 units of wild-type T7, 1.2 units of Sequenase 2.0), preincubated with or without 2 M thioredoxin in a total volume of 10 l (40 mM Tris⅐HCl, pH 7.5, 50 mM NaCl, 20 mM MgCl 2 , and 10 mM DTT) for 30 min at 37°C. The reactions were quenched by the addition of 15 l of 95% formamide.
Sequencing the Sequenase 2.0 Bypass Products-Thirty pmol of 5Јlabeled primer was annealed to 20 pmol of template. To each primed template was added 13 units of Sequenase 2.0 with buffer and dNTPs for a total volume of 5 l (200 M dNTPs, 40 mM Tris⅐HCl, pH 7.6, 10 mM MgCl 2 , and 5 mM DTT. The reaction was incubated at 37°C for 90 min, quenched with 20 l of 95% formamide, and electrophoresed. The fulllength product was excised, eluted, and dialyzed. An estimated 50 fmol or less of each bypass product was annealed to 25 fmol of 5Ј-end labeled 16-mer primer, d(AGCTACCATGCCTGCA). Sequenase version 1.0 (2.6 units) was added, for a final volume of 8 l (40 mM Tris⅐HCl, pH 7.6, 10 mM MgCl 2 , and 5 mM DTT). To each of four tubes containing 5 l of dideoxy mix was added 2 l of the annealed primed template. The reactions were incubated for 5 min at 37°C, and quenched by the addition of 9 l of 95% formamide. The reaction mixtures were heatdenatured, electrophoresed, and visualized by autoradiography

Substrates-
The photoproduct-containing 49-mers ( Fig. 1) were designed to be suitable for a variety of repair and replication studies, and their preparation and characterization have been reported previously (19). 2 The substrate for transcription was constructed by primer extension of a hybrid between a 66-mer containing a T7 RNA promoter (26) and the photoproduct-containing 49-mers. DNA and RNA synthesis reactions opposite the photoproducts with well characterized polymerases were principally undertaken to determine the enzymatic properties and conditions that facilitate bypass of DNA photoproducts. For the experiments described herein, the important design feature of the 49-mers is that the photoproducts are centrally located in a deoxyoligonucleotide long enough to serve as a template for primer extension by polymerases. With the exception of the trans-syn-II containing 49-mer, these photoproduct containing 49-mers have also been incorporated into M13 vectors and used to obtain photoproduct mutation spectra in E. coli under SOS conditions (18).
Effect of Polymerase and dNTP Concentration on Photoproduct Bypass-The results of the primer extension reactions as a function of polymerase and dNTP concentration are displayed in Figs. 2 and 3. All polymerases were able to fully extend the primers on the undamaged templates even at the low dNTP concentration, except for MMLV RT, which did not fully extend at 1 M dNTPs. Not surprisingly, then, MMLV RT was almost completely blocked by all the photoproducts even at 100 M dNTPs, stopping primarily one nucleotide prior to the 3Ј-T of the photoproducts, and terminating primarily three nucleotides prior to the photoproducts at low dNTPs ( Fig. 2A). The exo Ϫ KF primarily stopped one nucleotide prior to and opposite the 3Ј-T of all the photoproducts at low dNTPs, and primarily opposite the 3Ј-T at high dNTPs (Fig. 2B). In this experiment, exo Ϫ Klenow was also able to bypass the cis-syn dimer in 2, 8, and 19% yields at 1, 10, and 100 M dNTPs, respectively. The exo Ϫ Vent could not bypass any of the lesions, and stopped primarily one prior to all the photoproducts at low dNTPs, but stopped primarily opposite the 3Ј-T at high dNTPs (Fig. 2C). Interestingly, Vent was able to advance the primer opposite the 5Ј-T for the (6 -4) and Dewar products, but not the cis-syn and trans-syn-II products. Previous unpublished work in this laboratory found that Taq DNA polymerase was able to bypass the cis-syn dimer, but not the trans-syn-I dimer at 60°C (27). In this study, Taq polymerase was also found to bypass the cis-syn dimer in about 10% yield at 60°C in the presence of either 100 or 200 M dNTPs, but could not bypass the trans-syn-II, (6 -4), or Dewar products, even at 200 M dNTPs (data not shown). In all cases, synthesis stopped primarily opposite the 3Ј-T of all of the photoproducts (Ͼ53%) with significant amounts of termination one prior and opposite the 5Ј-T of the photoproducts (7-15%). Sequenase 2.0 could bypass all the lesions at high dNTP concentrations, and stopped primarily opposite to the 3Ј-T of the photoproducts at 10 M dNTPs (Fig. 3). Extension opposite the undamaged template was too fast to measure, but the pseudo first-order rate constant was at least 15 min Ϫ1 , and should be greater than 500 min Ϫ1 based on published kinetic data (25). Of all the photoproducts, the cis-syn dimer was bypassed the fastest by Sequenase 2.0, with a pseudo first order rate constant of 0.49 min Ϫ1 . The Dewar product was bypassed the second fastest at 0.071 min Ϫ1 and the trans-syn-II dimer slightly more slowly at 0.040 min Ϫ1 . The (6 -4) product was bypassed the slowest of all with a rate constant of 0.0063 min Ϫ1 .
Effect of Exonuclease Activity and Processivity on Bypass-The effect of the 3Ј 3 5Ј exonuclease activity was determined by comparing the action of wild-type KF and T7 polymerase with corresponding exonuclease-deficient mutants on the photoproduct containing templates (Figs. 4 and 5). There was a dramatic increase in the ability of both polymerases to extend opposite and past the lesions in the absence of exonucleolytic proofreading ability. Although wild-type KF led to Ͻ3% bypass of any of the photoproducts at 100 M dNTPs, exo Ϫ KF led to 47% bypass of the cis-syn dimer, and smaller, but significant amount (2-4%) of bypass of the other photoproducts under identical conditions (Fig. 4). Likewise, wild-type T7 DNA polymerase was unable to bypass any of the photoproducts at 100 M dNTPs, but Sequenase 2.0 led to Ͼ23% bypass of all the photoproducts (Fig. 5). The effect of processivity on bypass was determined by comparing the action of the exonuclease-deficient T7 polymerase with and without its processivity cofactor (Fig. 6). T7 polymerase is composed of a 1:1 complex of the T7 gene 5 protein and E. coli thioredoxin (28). Without thioredoxin, a D5A/E7A exonuclease-deficient T7 gene 5 protein was unable to bypass any of the photoproducts and terminated primarily prior to the 3Ј-T of the photoproducts. Upon the addition of thioredoxin, however, a significant fraction of the cis-syn and Dewar products were bypassed (Ͼ23%) and termination occurred primarily opposite the 3Ј-T of the photoproducts (Fig. 6). Unlike what was observed for the ⌬28 exonuclease-deficient Sequenase 2.0, only about 3% of the trans-syn-II and (6 -4) products were bypassed by the D5A/E7A exonucle-ase-deficient mutant in the presence of excess thioredoxin.
Sequence Determination of the Sequenase 2.0 Bypass Products-With the exception of the (6 -4) product, the bypass products of the photoproduct-containing 49-mers were obtained in sufficient quantity to be sequenced several times by the dideoxy method (Fig. 7). The bypass products of the undamaged, cis-syn dimer, and trans-syn-II dimer templates did not appear to have any mutations above a background level of about 5% for all sites that are produced during sequencing. Densitometric analysis of the dideoxy sequencing bands of the bypass product of the Dewar photoproduct in comparison to that of the undamaged template indicated that about 20% of thymidine had been introduced opposite the 5Ј-T of the photoproduct in place of deoxyadenosine.
To confirm that T could indeed be incorporated by the polymerase opposite the 5Ј-T of the Dewar product, and to investigate the selectivity of nucleotide incorporation opposite the (6 -4) product, the rates of dTMP and dAMP incorporation opposite both products were determined (data not shown). Extension of a 15-nucleotide primer terminating in A opposite the 3Ј-T of the (6 -4) product with 100 M dATP occurred 25 faster than with 100 M dTTP, compared with 7.4 times faster for the Dewar product. Extension of a 16-nucleotide primer terminating in T opposite the 5Ј-T of the (6 -4) product was 26 times faster than that terminating in A, compared with 40 times faster for the Dewar product, demonstrating that the mutagenic products could be further elongated. Because the polymerase is exonuclease-deficient, and both mutagenic and nonmutagenic products can be readily elongated, the selectivity of nucleotide incorporation opposite the 5Ј-T in the bypass product is solely governed by the selectivity of nucleotide incorporation in the elongation step opposite the 5Ј-T. Thus, one might expect that dTMP is also incorporated opposite the 5Ј-T of the (6 -4) product in the bypass product, but possibly at a lower frequency than for the Dewar product.
Transcription Past the Photoproducts-RNA synthesis opposite the dimers was also briefly investigated with T7 RNA polymerase and the 72-mer duplex containing the T7 RNA promoter. At 800 M NTPs, all the photoproducts could be bypassed with almost the same relative order as observed for the exo Ϫ T7 DNA polymerase, Sequenase 2.0, except that the trans-syn-II isomer was bypassed the slowest (Fig. 8). In 1 h, the ratio of bypass products to termination products was 4.9, 0.4, 0.72, and 2 for the cis-syn, trans-syn-II, (6 -4), and Dewar photoproducts, respectively. Arrest was found to occur at multiple sites surrounding the photoproduct site, but could not be accurately assigned, though it does appear that the T7 RNA polymerase could advance one nucleotide further opposite the cis-syn photoproduct than any of the other photoproducts. In addition to the predominant full-length products, small amounts of ϩ1 and Ϫ1 full-length product were observed, which have also been observed by others (29), as well as some nonspecific longer products that may be self-encoded run-offs (30). DISCUSSION One of our primary goals was to determine how differences in polymerase and lesion structure and properties would affect extension opposite and past DNA photoproducts. A second was to find a DNA polymerase that could bypass all the lesions with high enough efficiency to allow the effects of 3Ј 3 5Ј exonuclease activity and processivity on the bypass reactions, and to allow isolation and sequencing of the bypass products and eventually study the bypass mechanisms in detail. Polymerases deficient in 3Ј 3 5Ј exonuclease activity were selected for study first, because 3Ј 3 5Ј exonucleolytic cleavage is known to compete with elongation opposite and past DNA damage and slow down or completely inhibit DNA damage bypass (31)(32)(33). Primer extension opposite the photoproducts was examined as a function of dNTP concentration, as the rate of bypass was expected to increase with increasing dNTP concentration based on early studies on heterogeneous templates (31) and later studies with site-specific cis-syn dimers (34). Of the six 3Ј 3 5Ј exonuclease-deficient polymerases studied, MMLV RT, Taq, exo Ϫ Vent, exo Ϫ KF, Sequenase 2.0, and T7 RNA polymerase, only four of them were able to bypass at least one of the photoproducts, and only two were able to bypass all four photoproducts. Both T7 RNA polymerase and the exo Ϫ T7 DNA polymerase Sequenase 2.0 were able to bypass all the lesions (Figs. 3 and 8), whereas exo Ϫ KF and Taq were only able to significantly bypass the cis-syn dimer. When slightly different conditions were used, including a longer reaction time and the addition of 100 g/ml BSA, exo Ϫ KF was able to synthesize past a small fraction of the trans-syn-II, (6 -4), and Dewar products (Fig. 4).
The ability of Sequenase 2.0 to bypass all four products is not unexpected, as T7 DNA polymerases have been reported to bypass a wide variety of bulky adducts and intrastrand crosslinked species. The exo Ϫ T7 polymerases have been reported to bypass the bulky 2-aminofluorene (AF) adduct of guanine (35,36), and the bulky 7-bromomethylbena[a]anthracene (37) and styrene oxide (38) adducts of adduct of dA. The exo Ϫ -deficient T7 polymerase also bypasses photochemically cross-linked TA sites (39), and cis-diamminedichloroplatinum(II) cross-linked purines in GG, AG, and GCG (40) though no bypass was observed in different sequence context for the GG and AG sites (41). On the other hand, exo Ϫ T7 polymerase has not been found to bypass acetylaminofluorene (AAF) (32,36) or benzopyrenediolexpoxide (BPDE) (42) adducts. KF has also been found to bypass a variety of bulky lesions, including the AF adduct of guanine (35,43), and the bulkier AAF adduct of guanine (43) and C4Ј-modified bases (44). It can also bypass model estrogen DNA adducts (45), styrene oxide adducts (38) and certain stereoisomers of BPDE adducts of G (46). In contrast to what we observe for the dipyrimidine photoproducts, the TA* photoproduct is more easily bypassed by exo Ϫ KF than by Sequenase 2.0 (39). A similar trend was observed for a cis-platinum adduct of a GG site in one sequence context (41) but not in another (40).
The ability of the thermostable Taq polymerase to bypass some forms of DNA damage is not unprecedented, as it has been recently reported to bypass the cis-syn thymine dimer and a (6 -4) product to a small extent (47), as well as 7, 8-dihydro-8-oxoadenine, a lesion that causes little distortion to the DNA duplex (48). Because exo Ϫ Vent did not bypass any of the photoproducts, it may be a better choice for quantifying these products in genes by methods based on quantifying full-length polymerase chain reaction products (49,50) or polymerase chain reaction termination products (47,51,52) Our finding that all the photoproducts studied, which have often been classified as bulky adducts, can be bypassed by T7 RNA polymerase, contrasts with results observed for prokaryotic and eukaryotic RNA polymerases. Cyclobutane dimers have been shown to halt Escherichia coli RNA polymerase both in vitro (53) as well as in vivo (54). In transcription-coupled repair, RNA polymerase arrest initiates a series of events that involve excision of a small section of DNA containing the damage followed by new gap-filling DNA synthesis (Ref. 55; reviewed in Ref. 56). Similarly, eukaryotic RNA polymerase II is also fully inhibited by UV-induced adducts (57), and proceeds to initiate what is thought to be a similar series of events as its prokaryotic counterpart. On the other hand, T7 RNA polymerase is often found to be able to transcribe past many types of DNA damage. For example, modified bases such as 8-oxoguanine and an abasic site analog do not block transcription, whereas AF and AAF adducts show an increasing ability to block transcription (58). Cytosine arabinoside (29), as well as single nucleotide gaps (59 -61) also prove unable to arrest T7 RNA polymerase, though they may result in miscoding by the polymerase. The bulky BPDE DNA adducts inhibit transcription by T7 RNA polymerase to varying degrees which depend on the stereochemistry of the adduct (62).
Polymerases That Were Incapable of Bypassing the Photoproducts-MMLV RT led to very little if any bypass of any of the photoproducts, stopping almost exclusively one nucleotide prior to all four photoproducts at 100 M dNTP concentrations, which may make it useful for mapping the location of photoproducts in irradiated DNA. Although we are not aware of any reports of MMLV RT used in studies of damage bypass, the reverse transcriptase from avian myeloblastosis virus has been found to terminate at DNA photoproducts, but unlike MMLV RT, termination appeared to occur opposite the 3Ј-T of the dimer (63). Avian myeloblastosis virus RT has also been found to bypass cis-thymine glycol lesions (64) and abasic sites (65). Human immunodeficiency virus RT has been shown to be blocked by all but one of the six stereoisomeric BPDE adducts of G (66). The exo Ϫ Vent was also unable to bypass any of the lesions, but stopped at different positions depending on the lesion and dNTP concentration (Fig. 2C). At 100 M dNTPs, exo Ϫ Vent stalled opposite the 3Ј-base (first base) of the cis-syn and transsyn-II dimers, but partly extended opposite the 5Ј-base (second base) of the (6 -4) and Dewar products. It is difficult to understand why incorporation opposite the 5Ј-base of the (6 -4) and Dewar products is easier than for the cis-syn and trans-syn-II dimers, and we are unaware of any other reports using exo Ϫ Vent on damaged templates with which to compare our results.
Effect of Photoproduct Structure on Bypass-Of all the photoproducts, the cis-syn dimer was bypassed most easily by any given polymerase. Sequenase 2.0 bypassed the cis-syn dimer about 7 times faster than the Dewar product, about 11 times faster than the trans-syn-II dimer, and about 77 times faster than the (6 -4) product. We interpret this as a consequence of relatively close resemblance of the cis-syn dimer structure to a undamaged dithymine site (Fig. 1), and to the relatively little distortion that it causes to normal DNA duplex (67)(68)(69). The correspondingly slow bypass of the (6 -4) product of TT can likewise be attributed to the fact that it has been found to greatly distort DNA structure and not to base pair to an opposed A based on an NMR structure (69) or only weakly so based on an unrestrained molecular dynamics calculation (70). Though no structure for the trans-syn-II dimer in a duplex exists, its lower rate of bypass can be attributed to its 3Ј-T, which is locked into a syn orientation and places the methyl group in the base pairing region, thereby sterically blocking the addition of nucleotides to the primer terminus (Fig. 1). The finding that the Dewar product is more rapidly bypassed than the (6 -4) product was predicted previously on the basis of molecular modeling studies of the dinucleotide products indicating that the Dewar product could be fit to a B DNA structure better than could the (6 -4) product (71). Photoisomerization of the pyrimidone ring of the (6 -4) product converts it from an extended flat planar ring to a more compact tentlike structure.
Effect of Exonuclease Activity and Processivity on Bypass-Two properties that have been suggested as important in the ability of a polymerase to bypass a lesion are the presence and activity of a proofreading (3Ј 3 5Ј) exonuclease (31,64) and the processivity of the polymerase (72). KF and T7 DNA polymerases were chosen to study the effects of exonuclease activity on bypass, because both enzymes could be obtained in exonuclease-proficient and -deficient versions, and have been the subject of a number of detailed kinetic studies (25,73,74). The exo Ϫ Klenow was able to extend one nucleotide further than wild-type Klenow on all the lesions before stalling, and was able bypass 47% the cis-syn dimer compared with 3% for the wild-type Klenow under otherwise identical conditions (Fig. 4). Likewise, no bypass was seen with wild-type T7 polymerase, but Sequenase 2.0 gave substantial amounts of bypass of all lesions (Fig. 5). These results are in accord with the observation that synthesis opposite irradiated templates was increased when exo Ϫ KF and T7 DNA polymerase were used in place of the wild-type enzymes (75). Recently, elimination of the 3Ј 3 5Ј exonuclease activity of KF, has been shown to greatly accelerate bypass of an abasic site analog (33).
To examine the effects of processivity on bypass, we again made use of the T7 DNA polymerase system, by taking advantage of the 1000-fold increase in processivity conferred on T7 gene 5 polymerase subunit by thioredoxin, (28). Thioredoxin enhances the processivity of the gene 5 protein by increasing the lifetime of the polymerase-DNA complex (76). The difference in bypass ability of an exo Ϫ T7 gene 5 protein (25) in the presence or absence of thioredoxin (Fig. 6), was similar to the difference seen in the absence or presence of the exonuclease (Fig. 5). In the presence of thioredoxin, the exo Ϫ gene 5 protein was able to bypass all the photoproducts, stalling opposite the first and second bases of the photoproducts. In the absence of thioredoxin, however, the gene 5 protein was not able to bypass any of the photoproducts, and stalled one base prior to every photoproduct site. These results are similar to those that we observed to occur in the bypass of cis-syn and trans-syn-I dimers by calf thymus polymerase ␦ (pol ␦) in the presence and absence of its processivity factor, proliferating cell nuclear antigen (PCNA) (77). Addition of PCNA has also been shown to greatly increase the bypass of abasic sites by pol ␦ (78).
The A-rule Revisited: Origin for the Incorporation of A Opposite the 3Ј-T of Photoproducts by Sequenase 2.0 -One general hypothesis for the preferential incorporation of A opposite DNA damage is the A-rule (79,80), which proposes that preferential dATP binding by the polymerase governs nucleotide incorporation when it encounters a non-instructional lesion, typified by an abasic site. Because the mutation spectra of UV-irradiated DNA could be explained by incorporation of A, dipyrimidine photoproducts were originally classified as non-instructional. Recently this classification has been called into question, and hence the use of the A-rule to explain mutations caused by DNA photoproducts. Lawrence and co-workers have argued that the cis-syn dimer must be an instructive lesion by virtue of its high coding specificity in E. coli relative to abasic sites (81), and its ability to engage in near normal hydrogen bonding to adenine (67)(68)(69). Furthermore, G is incorporated opposite the C in a cis-syn dimer of TC in E. coli (82) and opposite the 3Ј-base in the (6 -4) products of both TT and TC (83). In contrast, Sequenase 2.0 puts A opposite the 3Ј-T of all the photoproducts, irrespective of structure. Why then would one polymerase add G opposite the 3Ј-T of the (6 -4) product and another polymerase add A? The argument that DNA photoproducts are instructional and not subject to the A-rule was based on observations in E. coli under SOS conditions and may in fact not apply to all polymerases. A possible expla-nation for the incorporation of A opposite the 3Ј-T of all the photoproducts by Sequenase 2.0 comes from examination of the recent crystal structure of a complex between an exo Ϫ T7 polymerase and a template primer in the presence of a ddNTP. In this structure, the template is forced to take a sharp 90°turn at the polymerase active site following the nucleotide opposite which the dNTP is incorporated (84). Because all dipyrimidine photoproducts covalently link two nucleotides together, the 3Ј-pyrimidine of a photoproduct cannot be made to occupy the site opposite which the dNTP resides, and is instead forced out of the active site with the rest of the template (Fig. 9). This would create an empty site, much like an abasic site, which would therefore be non-instructional, and lead to the preferential incorporation of A. Bending the template at the active site may be an important and general mechanism for preventing or greatly attenuating translesion synthesis at the sites of intrastrand cross-linked nucleotides such as dipyrimidine photoproducts, cis-platinum adducts, and other intrastrand cross-linked nucleotides, and has been observed in human pol ␤ (85) and Bacillus polymerase I (86). Once the A is incorporated, the template can move by one nucleotide, and the entire photoproduct can now be bound in the active site. Incorporation of a nucleotide opposite the 5Ј-pyrimidine will then be mediated by the instructional properties of the 5Ј-pyrimidine and the fit of the photoproduct in the active site.
Origin of the Preference for Incorporation of A Opposite the 5Ј-T of the Photoproducts-In considering the rate and selectivity of nucleotide incorporation opposite the 5Ј-pyrimidine, it is useful to examine the results of the primer extension reactions by Sequenase 2.0 opposite the 5Ј-T of the photoproducts in light of the mechanism by which T7 DNA polymerase maintains high fidelity on undamaged templates (74). The rate-determining step during processive synthesis is a conformation change after dNTP binding, and before bond formation. In this induced-fit model, it is speculated that the conformational change selects the correct dNTP by recognizing its correct Watson-Crick geometry. Likewise, 3Ј-end mismatches drastically slow the conformational change necessary to incorporate the next nucleotide. By analogy, dNTP incorporation opposite and past lesions will be governed by the how closely the nascent base pair resembles the Watson-Crick geometry of normal B DNA. This model argues against the importance of hydrogen bonding, and argues for the importance of resemblance to Watson-Crick geometry as providing the instruction to the polymerase. Indeed, the selection of correct ge-ometry has been suggested by several groups as the primary means by which polymerases maintain high fidelity (74,(87)(88)(89)(90)(91). This is also borne out by recent crystal structures of a Bacillus DNA polymerase I complexed to template-primers and T7 DNA polymerase and pol␤ complexed to template-primers and ddNTPs (84,85). For the case of the dinucleotide photoproducts studied, all have a 5Ј-T that retains the H-bonding properties of T, though the conformation of the 5Ј-T depends on the particular photoproduct. Thus, the preferential incorporation of A opposite the 5Ј-T of all these products is likely to be the result of the ability to form a Watson Crick-like base pair. The differing rates of bypass of the photoproducts is probably the result of deviations of the geometry from an ideal Watson-Crick base pair caused by distortions induced by the photoproducts that slow phosphodiester bond formation during the rate-determining step in bypass. For all the photoproducts, the major termination band corresponds to termination opposite the 3Ј-T of the dimer suggesting that the slowest or rate-determining step in bypass involves the extension step opposite the 5Ј-T (step E2 of Fig. 9).
Biological Implications-Of the four major dipyrimidine photoproducts of TT, the cis-syn dimer was found to be the most easily bypassed by the polymerases studied, which correlates with the fact that it is the least disruptive of DNA structure (69). The cis-syn thymine dimers have also been found to be more easily bypassed than trans-syn-I dimers by pol ␦/PCNA (77), and a vector containing a cis-syn thymine dimer was found to be more efficiently replicated by lagging strand synthesis than a (6 -4) product in cell free HeLa extracts (92). The greater efficiency of cis-syn dimer bypass would support the notion that cis-syn dimers have the highest mutagenic potential of the four major photoproducts (12), as they are also the most slowly repaired by excision repair systems (93)(94)(95). Although there is no direct evidence at this point, it may be that cis-syn dimers may also be the most easily bypassed by transcription systems, and therefore also the least readily repaired by transcription-coupled repair. It is known, however, that cis-syn dimers of TT sites are not very mutagenic when bypassed by KF in vitro (34,96) or in E. coli under SOS conditions or in yeast (14,97). Likewise, bypass of cis-syn thymine dimer in an SV40 vector by HeLa cell free extracts appears to be non-mutagenic (92). On the other hand, it has been shown that the deamination products of C-containing cis-syn dimers are highly mutagenic, almost exclusively causing C 3 T mutations, FIG. 9. Proposed model for elongation opposite dipyrimidine photoproducts by T7 DNA polymerase. A, model for elongation opposite undamaged DNA, which is based on the crystal structure of a primer-template, and ddNTP complex with T7 DNA polymerase, which reveals that the template is forced to make a right angle turn after the catalytic site for primer extension. B, a model for elongation opposite a photoproduct between the two Ts (denoted by an equal sign). Each successive step in elongation opposite a photoproduct of TT is indicated by En. Because the bases in dipyrimidine photoproducts are covalently linked together, the 3Ј-nucleotide of these photoproducts cannot be accommodated in the active site during primer elongation opposite this site. the major mutation induced by UV light (15, 98 -100). Thus, cis-syn dimers have four features (7) that make them prime candidates as the principal products involved in mutagenesis at dipyrimidine sites. 1) They are the major photoproducts induced by UV light, 2) they are the least rapidly repaired of the dipyrimidine photoproducts, 3) they are the most easily bypassed, and 4) they can be highly mutagenic. The observation that the Dewar isomer of the (6 -4) product is more easily bypassed than the (6 -4) product by Sequenase 2.0, as previously predicted based on its structure (71), would also confer a higher mutagenic potential on this product than its (6 -4) isomer.