Nucleic Acids Research

Vectors

We previously used a site-specific mutagenesis approach to investigate the mutational properties of oxidized adenine in mammalian cells (25). We selected mutated plasmid molecules that were resistant to a restriction enzyme. In that study, we observed that most of the resistant plasmids contained random deletions (~10% of the total population). Thus, it is necessary to minimize this error-prone replication in mammalian cells (31,32) to select the mutant plasmids that are resistant to a restriction enzyme. Moreover, whether a DNA lesion induces deletions around itself during replication is another important question to be resolved. Therefore, we constructed new vectors in which the targeted site is located between two essential sites, the SV40 origin and the ampicillin resistance gene, to minimize the random deletions (Fig. 1A).

The (+)-strand of pSVKAM2 [we define the (+)-strand as the same strand of ss pSVK3 (the parental vector of pSVKAM2) that is produced by the infection of helper phage] is replicated during lagging strand synthesis around the region where the linker DNA with a photolesion was inserted (Fig. 1B and C). Accordingly, the (-)-strand of pSVK3 around the inserted site is replicated by the leading strand apparatus. Ds vectors with a photolesion were constructed by the insertion of ds oligonucleotides by the use of the ApaI and EcoRI sites (Fig. 1B).

We used the following nomenclature: (+)-T(6-4)T and (+)-T=T as the vectors containing T(6-4)T and T=T, respectively, in the (+)-strand. The (+)-control vector has the same sequence with an unmodified TpT instead of a photolesion. The (-)-T(6-4)T, (-)-T=T, and (-)-control are named similarly, although the strand of interest is the (-)-strand (Fig. 1).

Cytotoxicity of T(6-4)T and T=T in COS-7 cells

We first evaluated the cytotoxicities of T(6-4)T and T=T in COS-7 cells to determine the degree of the replication block. As preliminary experiments, we transfected various amounts of the pSVKAM2 vector into COS-7 cells and the plasmids recovered from the cells were transfected into E.coli DH5[alpha] cells. We observed that the number of E.coli colonies and the amount of DNA transfected into COS-7 cells exhibited a linear correlation between 0 and 5 ng under our conditions (data not shown). This means that the number of bacterial colonies reflects the relative amount of plasmid DNA replicated in COS-7 cells. Thus, the number of E.coli colonies was a good indicator for the evaluation of the photolesion cytotoxicity in COS cells.

Four ng of the modified and unmodified vectors were transfected into COS-7 cells and were allowed to replicate in the cells. The plasmid DNAs recovered from the cells were transfected into E.coli DH5[alpha] cells and the colonies formed were counted. The number of E.coli colonies derived from (+)-T(6-4)T was 93% of the (+)-control (Table 1). On the other hand, the number of E.coli colonies derived from (+)-T=T was 79% of the control (Table 1). Since the photolesions in the (+)-vector were located on the template strand of lagging strand synthesis (Fig. 1C), these results mean that the two photodimers weakly blocked the lagging strand synthesis. Similarly, the relative replication efficiencies of (-)-T(6-4)T and (-)-T=T were estimated to be 71 and 58%, respectively (Table 1). These results suggest that T=T is more cytotoxic than T(6-4)T during either leading or lagging strand synthesis. Moreover, the degree of the replication block by either photolesion is higher during leading strand synthesis than during lagging strand synthesis.

T(6-4)T and T=T are mutagenic in mammalian cells

We introduced the photolesions into a unique restriction enzyme site (5[prime]-CTTAAG, an AflII site), and thus we could select mutants as bacterial cells with a plasmid resistant to the restriction enzyme. The AflII treatment was carried out for the plasmids recovered from the transfected COS-7 cells, and the digested DNAs were introduced into E.coli. The hybridization experiment was employed to discriminate mutated and normal sequences, and to remove `mutants' that contained a large (>4 bp) deletion. By these techniques, we selected plasmids mutated within the AflII site.

Table 1. Cytotoxicities of photolesions in COS-7 cells^a

Vector	Double strand (+) lagging	Double strand (-) leading
unmodified	100b	100b
T(6-4)T	93	71
T=T	79	58

^aPercentage of colonies resulting from transformation of E.coli DH5[alpha] cells with the plasmid DNA recovered from COS-7 cells. The values represent the average of six separate experiments.
^bThis value is defined as 100 for each experiment. The actual number of colonies ranged between 1000 and 3000.

Table 2. Mutation frequencies of photolesions in COS-7 cells^a

Vector	Double strand (+) lagging	Double strand (-) leading
unmodified	0.11	0.01
T(6-4)T	4.70	2.30
T=T	0.68	0.20

^aPercentage of colonies containing a plasmid mutated within the 5[prime]-CTTA sequence. Mutants containing a large (>4 bp) deletion are excluded in the calculation as described in Materials and Methods. The values represent the average of three separate experiments.

As shown Table 2, the two photolesions induced mutations in mammalian cells. The T(6-4)T lesion in the (+)-DNA induced mutations with an efficiency of 4.70%. The maximum MF was expected to be 50%, because the complementary strand is also replicated. Thus, at least 9.40% of the photoproducts were estimated to induce the misincorporation of nucleotides during lagging strand synthesis. On the other hand, the MF of T=T was 0.68% during lagging strand synthesis (Table 2).

The T(6-4)T and T=T lesions in the ds (-)-vectors induced mutations with efficiencies of 2.30 and 0.20%, respectively (Table 2). The fact that the MFs of these photolesions in different strands were not identical indicates that the ratios of misincorporation were affected by the leading or lagging strand synthesis during replication.

Mutation spectra of photolesions in COS-7 cells

We analyzed the sequences of the mutants obtained with the (+)-T(6-4)T. The mutation detected most frequently was a 3[prime]-T->C transition (17 colonies, Table 3). This result indicates that dGMP was incorporated opposite the 3[prime]-T of T(6-4)T. Transversions of the 3[prime]-T were also elicited during lagging strand synthesis (eight colonies). A similar mutation spectrum was observed with T(6-4)T in the (-)-strand, during leading strand synthesis (Table 3). However, the ratio of colonies with the 3[prime]-T->C transition was lower than that during lagging strand synthesis, and no 3[prime]-T->G transversions appeared to be elicited during leading strand synthesis. An interesting feature was that C->T transitions occurred at the 5[prime]-flanking position of T(6-4)T during leading strand synthesis. This phenomenon was not observed during lagging strand synthesis.

Table 3. Mutations induced by T(6-4)T in COS-7 cells^a

Sequenceb	Vector
	double strand (+) lagging	double strand (-) leading
5[prime]-TC-3[prime]	17 (65)	10 (50)
5[prime]-TG-3[prime]	3 (12)	0 (0)
5[prime]-TA-3[prime]	5 (19)	4 (20)
5[prime]-CT-3[prime]	1 (4)	1 (5)
5[prime]-GT-3[prime]	0 (0)	1 (5)
5[prime]-AT-3[prime]	0 (0)	0 (0)
Others	0 (0)	4 (20)c

^aNumber of mutants observed in three separate experiments. Percentage of each targeted mutation is represented in parentheses.
^bMutated sites are indicated by underlines.
^cThe C residue at the 5[prime]-flanking position of TT changed to T (5[prime]-CTTA to TTTA).

Table 4. Mutations induced by T=T in COS-7 cells^a

Sequenceb	Vector
	double strand (+) lagging	double strand (-) leading
5[prime]-TC-3[prime]	6 (19)	4 (17)
5[prime]-TG-3[prime]	7 (23)	2 (8)
5[prime]-TA-3[prime]	1 (3)	1 (4)
5[prime]-CT-3[prime]	4 (13)	2 (8)
5[prime]-GT-3[prime]	2 (6)	2 (8)
5[prime]-AT-3[prime]	2 (6)	3 (13)
5[prime]-AC-3[prime]	1 (3)	3(13)
[Delta]T	2 (6)c	2(8)
[Delta]TT	0 (0)	1(4)e
+A	0 (0)	2(8)f
Others	6 (19)d	2 (8)g

^aNumber of mutants observed in three separate experiments. Percentage of each targeted mutation is represented in parentheses.
^bMutated sites are indicated by underlines.
^cOne clone has a sequence deleted at C in addition to T (5[prime]-CTTA to TA).
^dOne clone has a 5[prime]-GTTA sequence and four clones have a 5[prime]-CTTT sequence. One clone has a sequence deleted at A (5[prime]-CTTA to CTT).
^eThis clone has a sequence deleted at C in addition to TT (5[prime]-CTTA to A).
^fThese clones have a 5[prime]-CTATA sequence.
^gOne clone has a 5[prime]-GTTA sequence and one clone has a 5[prime]-CTTT sequence.

Large deletions induced by T(6-4)T and T=T

We compared the percentage of large (>4 bp) deletions induced by the two photolesions. Large deletions occurred in 1.7-1.8% of the cases of vectors containing T(6-4)T (Table 5). The T=T lesion elicited deletions with a frequency of 2.0 and 1.3% during lagging and leading strand syntheses, respectively. On the other hand, deletions occurred in only 0.4% of the unmodified vectors. Thus, it appeared that the two photodimers induced deletion mutations. Moreover, the frequencies of the deletions were higher in the vector containing T=T in the (+)-strand than that containing T=T in the (-)-strand. This result suggests that the deletion mutations occurred more frequently during lagging strand synthesis than during leading strand synthesis with T=T. However, the frequencies of the induced deletions were similar in the case of T(6-4)T.

Table 5. Deletion frequencies of photolesions in COS-7 cells^a

Vector	Double strand (+) lagging	Double strand (-) leading
unmodified	0.39	0.41
T(6-4)T	1.74	1.81
T=T	1.96	1.28

^aPercentage of colonies containing a large (>4 bp) deletion to total E.coli colonies. The values represent the average of three separate experiments.

DISCUSSION

We have used a site-specific mutagenesis approach to investigate the mutational properties of the two photolesions, T(6-4)T and T=T, in mammalian cells. The two major aims were to investigate the mutational properties of T(6-4)T and T=T in ds DNA and to compare the effects of the photolesions on replication by leading and lagging strand replication proteins in mammalian cells.

We found that the replication efficiencies of the vectors containing T=T were lower than those of the vectors with T(6-4)T (Table 1). These results indicate that T(6-4)T blocks DNA replication more weakly than T=T. It has been reported that the repair of (6-4)photoproducts is more rapid than that of cyclobutane dimers (12,13,33-35). Moreover, it was shown that human cell extracts repair T(6-4)T more rapidly than T=T, using ds DNA containing each of the photoproducts (36). The weaker replication block by T(6-4)T may be due to its more rapid repair, at least in part. In accordance with these findings, Thomas and Kunkel reported that cyclobutane pyrimidine dimers are responsible for the inhibition of replication conducted by human cell extracts with UV-irradiated DNA (37). The same conclusion was achieved by Carty et al. by the use of human and monkey cell extracts (38).

The mutation frequencies of T(6-4)T were 4.70 and 2.30% with (+)- and (-)-T(6-4)T, respectively (Table 2). On the other hand, (+)- and (-)-T=T induced mutations at frequencies of 0.68 and 0.20%, respectively (Table 2). Obviously, T(6-4)T was more mutagenic than T=T in mammalian cells. The MF of small lesions, such as 8-hydroxyguanine and 2-hydroxyadenine, in ds DNA is between 0.1 and 1% (25,39,40). Thus, T(6-4)T is a highly mutagenic lesion.

The mutation found most frequently was a 3[prime]-T->C transition with T(6-4)T (Table 3). This type of mutation is found as the major mutation induced by T(6-4)T in yeast and SOS-induced bacteria (41-43). Thus, it is shown for the first time that the mutation spectra of T(6-4)T in the cells of three different organisms (E.coli, yeast and mammals) are coincident. It was reported that a duplex containing a Gua base opposite the 3[prime]-T of T(6-4)T is thermodynamically more stable than oligonucleotides containing other pairs involving the 3[prime]-T of T(6-4)T (44). The 3[prime]-T->C transition observed in our present study can be explained by this base pair formation. Transversion mutations were also detected at the 3[prime]-T site, although 3[prime]-T->G mutations were not detected with (-)-T(6-4)T (Table 3). Thus, the 3[prime]-T was the primary target site of the mutations induced by T(6-4)T. In contrast, fewer mutations were found at the 5[prime]-T of T(6-4)T (Table 3). Interestingly, C->T transitions at the 5[prime]-flanking position of the TT site were found in the case of (-)-T(6-4)T (four colonies). Mutations found at the 5[prime]-flanking site of a modified base have been reported in mammalian cells (39,40,45,46). Although the precise mechanism of this type of mutation has not been resolved, it may be caused by misincorporation of nucleotides by DNA polymerase(s), due to the local structure perturbation.

The mutation spectrum of T=T in COS-7 cells was more complex (Table 4). The mutations found most frequently were 3[prime]-T->C and 3[prime]-T->G. The 3[prime]-T->A mutation was not induced frequently. The ratios of the substitution mutations at the 3[prime]-T position to those at the 5[prime]-T site were 14:8 and 7:7 during lagging and leading strand syntheses, respectively. Thus, the mutation spectrum was affected by the strand within which the T=T was located. Interestingly, tandem mutations (5[prime]-TT->AC) were induced by T=T in COS-7 cells. The deletion of T or TT also occurred in the presence of T=T. The tandem and deletion mutations were observed more frequently with the (-)-T=T vector than the (+)-T=T vector.

These T=T mutation spectra differ from previous findings that T=T mainly induces 3[prime]-T->A and 3[prime]-T->C mutations in E.coli (47) and 3[prime]-T->A transversions in NIH3T3 cells (15). A purine-rich oligonucleotide (5[prime]-GCAAGT=TGGAG) was employed in the E.coli study (47), and the T=T lesion was located at the junction of a pyrimidine-rich sequence (5[prime]-TTCTTCT=TGGCC) in the NIH3T3 study (15). The mutation spectra of T=T may be sensitive to the sequence contexts.

Gentil et al. reported the mutational properties of T(6-4)T and T=T on ss DNA in COS-7 cells (16). They found that the MF of T(6-4)T was ~60% and that of T=T was 2.4%. These higher MFs were probably caused by the very low repair efficiencies of the two photolesions on ss DNA, and by the absence of the complementary strand. Surprisingly, they reported that >80% of the total mutations induced by T(6-4)T were G->T transversions at the 5[prime]-flanking site (16). This is in contrast to our present observation that T(6-4)T elicited targeted mutations (Table 3). Although we observed mutations at the 5[prime]-flanking site, they were minor and were found only during leading strand synthesis (Table 3). The differences in the mutational properties between our experiments and those of Gentil et al. could be attributable to differences in the replication of ss and ds vectors or to the sequence context of the adduct in the vector.

Carty et al. reported that ds plasmids containing either T(6-4)T or T=T at a predetermined site are poorly mutagenic (MFs <0.3 and 0.2 %, respectively) in in vitro DNA replication using human HeLa cell extracts (21). Moreover, they analyzed the replication efficiencies of plasmids containing T(6-4)T and T=T, and found that T(6-4)T is more inhibitory to the replication than T=T. Their results are in contrast to our present results that T(6-4)T and T=T were mutagenic, and that the plasmids containing T(6-4)T were replicated more efficiently than those containing T=T (Table 1). The reason for these discrepancies between our present study and that of Carty et al. is not clear. One possibility could be the differences in the sequence context. Alternatively, some factors necessary for bypass of the photolesions may be present in COS-7 cells, but in relatively low amounts in HeLa extracts, because COS-7 cells are green monkey cells, a natural host of the SV40 virus, and express a high level of the SV40 large tumor antigen.

Both T(6-4)T and T=T induced mutations during lagging strand synthesis more frequently than during leading strand synthesis in COS-7 cells, at least in the sequence context used in our study (Table 2). The T(6-4)T and T=T lesions induced mutations by the lagging replication apparatus 2.0 and 3.4 times, respectively, as frequently as by the leading replication apparatus. We found similar results when DNA containing oxidized adenine (2-hydroxyadenine) was replicated in COS-7 cells (25). Higher error rates during lagging strand replication in E.coli were reported by the use of a plasmid containing a single modified base at unique site (17,19), although a similar MF was reported in the case of 8-hydroxyguanine (18). Thus, it is likely that the fidelity in lagging strand synthesis may be lower than that in leading strand synthesis. The T(6-4)T and T=T lesions were less cytotoxic during lagging strand synthesis than during leading strand synthesis (Table 1). Taken together, these results indicate that the lagging strand apparatus can bypass the photolesion more easily than the leading strand apparatus, and that misincorporation of deoxynucleotides occurs during this translesional synthesis. Moreover, the mutation spectra of the two photodimers were affected by the orientation of DNA synthesis (Tables and ). Thus, the fidelities during the leading and lagging strand syntheses appear to be different, as reported in previous papers (17,19,25). Further studies with ds vectors will address the molecular mechanisms of these phenomena.

The bulky DNA lesions, T(6-4)T and T=T, induced deletions (>4 bp) in mammalian cells (Table 5). It is known that DNA polymerases are blocked in in vitro DNA synthesis with UV-irradiated DNA or an oligonucleotide containing T=T (48,49), and that the replication block appeared to cause cytotoxicity (Table 1). When T=T or T(6-4)T is present in the lagging template strand, Okazaki fragment formation around the photolesion may be inhibited. Indeed, gap formation during lagging strand synthesis was observed during in vitro SV40-origin dependent DNA synthesis with DNA containing T=T (50). When the cellular components join the two terminals of the gap, deletion mutations may occur. The enhanced frequency of deletion mutations by T=T and T(6-4)T may be derived from this mechanism, in the case of lagging strand synthesis.

It was reported that DNA synthesis is terminated by the presence of T=T in the leading template strand during in vitro SV40-origin dependent DNA synthesis (50,51). This blockage caused uncoupling of the replication fork, to generate ss regions (50,51), and the cellular components could bypass the T=T lesion after the transient replication block (51). Some of the deletions induced by T=T, and likewise T(6-4)T, in the leading template strand may be due to a `jump' of the DNA replication machinery on the template strand after the arrest.

In this study, we have shown the mutational properties of the two major photoproducts of TpT in mammalian cells. Further investigations of the properties of other photoproducts in mammalian cells will help us to understand the mechanisms of UV light-induced mutations and skin cancer.

ACKNOWLEDGEMENTS

This work was supported in part by a Grant from the University of Occupational and Environmental Health and Grants-in-Aid from the Ministry of Education, Science, Sports and Culture of Japan.