Effects of 6-Thioguanine and S6-Methylthioguanine on Transcription in Vitro and in Human Cells*

Background: Thiopurine prodrugs can result in the formation of S6-methylthioguanine (S6mG) and 6-thioguanine (SG) in DNA. Results: We examined how SG and S6mG affect DNA transcription in vitro and in human cells. Conclusion: S6mG, but not SG, causes a strong mutagenic and inhibitory effects on transcription. Significance: This work provides new knowledge that S6mG-mediated transcriptional alternations might contribute to thiopurine-induced cytoxicity and potential therapy-related cancers. Thiopurine drugs are extensively used as chemotherapeutic agents in clinical practice, even though there is concern about the risk of therapy-related cancers. It has been previously suggested that the cytotoxicity of thiopurine drugs involves their metabolic activation, the resultant generation of 6-thioguanine (SG) and S6-methylthioguanine (S6mG) in DNA, and the futile mismatch repair triggered by replication-induced SG:T and S6mG:T mispairs. Disruption of transcription is known to be one of the major consequences of DNA damage induced by many antiviral and antitumor agents; however, it remains undefined how SG and S6mG compromise the efficiency and fidelity of transcription. Using our recently developed competitive transcription and adduct bypass assay, herein we examined the impact of SG and S6mG on transcription in vitro and in human cells. Our results revealed that, when situated on the transcribed strand, S6mG exhibited both inhibitory and mutagenic effects during transcription mediated by single-subunit T7 RNA polymerase or multisubunit human RNA polymerase II in vitro and in human cells. Moreover, we found that the impact of S6mG on transcriptional efficiency and fidelity is modulated by the transcription-coupled nucleotide excision repair capacity. In contrast, SG did not considerably compromise the efficiency or fidelity of transcription, and it was a poor substrate for NER. We propose that S6mG might contribute, at least in part, to thiopurine-mediated cytotoxicity through inhibition of transcription and to potential therapy-related carcinogenesis via transcriptional mutagenesis.

most effective chemotherapeutic drugs in clinical practice. They are widely used as anti-inflammatory, anticancer, and immunosuppressive agents in the treatment of a variety of human diseases including childhood acute leukemia and inflammatory disorders, even though there is concern about therapy-induced malignancies resulting from long term thiopurine use (1)(2)(3)(4). Despite their successful clinical applications for over half a century, the exact molecular mechanisms by which thiopurines exert their cytotoxic effects remain incompletely understood.
Thiopurines are prodrugs, and to achieve their efficacy and cytotoxicity, they need to be converted to active metabolites, i.e., S G nucleotides, and subsequently incorporated into DNA (2,3,5). The incorporation of S G nucleotides into DNA has been detected in several different types of mammalian cells and tissues after treatment with thiopurines, although the levels of DNA S G vary widely among different biological samples (6 -10). S G in DNA has been implicated in the mitochondrial pathway of apoptosis by suppressing the activation of Rac1 GTPase through its binding to S GTP instead of GTP (11). In addition, DNA S G is known to perturb some other cellular pathways that involve global cytosine demethylation or the formation of oxidative damage products such as DNA interstrand cross-links and DNA-protein cross-links (6,7,(12)(13)(14)(15).
Postreplicative mismatch repair (MMR) system is also thought to play an important role in thiopurine-mediated cytotoxicity (3). MMR-deficient cells in culture are more resistant to thiopurine treatment than MMR-proficient cells (16 -19). Some earlier studies have suggested that the toxicity of thiopurines involves methylation of the S 6 position of S G in DNA by S-adenosyl-L-methionine to form S 6 -methylthioguanine (S 6 mG) (Fig. 1A), misincorporation of thymidine opposite the S 6 mG during DNA replication, and recognition of the resultant mispairs by the MMR system (8,20). Our recent studies further revealed that S 6 mG is capable of inducing G 3 A mutations at frequencies of 94 and 39% in Escherichia coli and mammalian cells, respectively (8,(21)(22)(23). Additionally, replicative bypass of S G is also mildly mutagenic in these cells, and the resulting S G:T base pairs can be efficiently recognized by MMR proteins; thus, S G may also trigger the postreplicative MMR system and contribute to thiopurine toxicity (21,22,24,25).
Disruption of transcription and related processes is considered as one of the major factors contributing to the cytotoxicity of potent antiviral and anticancer agents currently used in the clinic or in clinical trials (26 -30). Many of these drugs interfere with transcription through chemical modifications of cellular DNA (26,31). In this context, some drug-induced DNA damage may inhibit the initiation of RNA synthesis by altering the binding of some transcription factors to DNA, whereas others may act as a physical impediment to transcription elongation by RNA polymerase (RNAP) (26,27,31). The latter type of DNA lesions usually trigger transcription-coupled nucleotide excision repair (TC-NER) that preferentially removes lesions on the transcribed strand of DNA (32). On the other hand, transcriptional bypass of some DNA lesions may lead to the generation of mutant transcripts in a process called transcriptional mutagenesis (33,34). A recent study revealed that S G in DNA is only marginally inhibitory to transcription in vitro (10). However, it remains undefined whether S G compromises transcription fidelity and how S 6 mG affects DNA transcription.
To address these questions, here we have examined quantitatively the impact of S G and S 6 mG on the transcriptional efficiency and accuracy in vitro and in vivo by using our recently developed competitive transcription and adduct bypass assay (35). We also investigated whether TC-NER is involved in the removal of these lesions in human cells.

EXPERIMENTAL PROCEDURES
Materials and Cell Culture Conditions-Unmodified oligodeoxyribonucleotides (ODNs), [␥-32 P]ATP, and shrimp alkaline phosphatase were purchased from Integrated DNA Technologies, PerkinElmer Life Technologies, and USB Corporation, respectively. Chemicals and enzymes unless otherwise specified were obtained from Sigma-Aldrich and New England Biolabs, respectively. The 293T human embryonic kidney epithelial cells were purchased from ATCC. The human skin fibroblast cell lines GM00637 and GM04429 were kindly provided by Prof. Gerd P. Pfeifer (City of Hope). The cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Invitrogen), 100 units/ml penicillin, and 100 g/ml streptomycin (ATCC) and incubated at 37°C in 5% CO 2 atmosphere.
Transcription Template Preparation-To construct lesionfree control vector for S G and S 6 mG, a 52-mer ODN with the sequence of 5Ј-CTAGCGCTACCGGACTCAGATCTCGAG-CTCTAGCTTTGCGCAAGCGACTCCG-3Ј was annealed with the complementary strand and ligated to an NheI-EcoRI restriction fragment from the nonreplicating pTGFP-T7-Hha10 plasmids (35). We next treated the control plasmids with the nicking enzyme Nt.BstNBI and subsequently produced a gapped vector by removing a 33-mer single-stranded ODN. The gapped plasmid was subsequently annealed with a 5Ј-phosphorylated S G-or S 6 mG-bearing ODN and ligated with T4 DNA ligase. The ligation mixture was incubated with ethidium bromide, and the resulting supercoiled lesion-bearing plasmid was isolated by agarose gel electrophoresis, following recently published procedures (21,36,37). The competitor vector used for this study was previously constructed for studying the impact of N 2 -(1-carboxyethyl)-2Ј-deoxyguanosine on transcrip-tion (35). Purified control or lesion-bearing genome was mixed with a competitor genome and used as transcription templates, with a molar ratio of competitor vector to control or lesion-bearing genome being 1:6 for all experiments in this study.
In Vitro Transcription Assay-Multiple rounds of transcription reactions using T7 RNAP or HeLa nuclear extract (Promega) were performed as described elsewhere (35). Briefly, T7 RNAP-mediated reaction was incubated at 37°C for 1 h in a 20-l mixture containing 20 units of T7 RNAP, 10 units of RNase inhibitor, 0.5 mM each ATP, CTP, GTP, and UTP, and 50 ng of NotI-linearized DNA template. The hRNAPII-mediated reaction was incubated at 30°C for 1 h in a 25-l mixture containing 8 units of HeLa nuclear extract, 10 units of RNase inhibitor, 0.4 mM each of all four ribonucleotides, and 50 ng of NotIlinearized DNA template.
In Vivo Transcription Assay-Human skin fibroblast cells (GM00637 and GM04429) were grown to ϳ70% confluence in 24-well plates and co-transfected with 50 ng of mixed genome and 450 ng of carrier plasmid (self-ligated pGEM-T; Promega) with Lipofectamine 2000 (Invitrogen), according to the manufacturer's protocol. To prepare the carrier plasmid, we first treated the linearized pGEM-T vector (Promega) containing 3Ј-T overhangs with T4 DNA polymerase, followed by selfligating the resultant plasmid DNA, which contains blunt ends. Highly efficient depletion of cockayne syndrome group B (CSB) and xeroderma pigmentosum group C (XPC) genes was achieved by siRNA treatments as previously described (35), and all siRNAs were obtained from Dharmacon: CSB SMARTpool (L-004888), XPC SMARTpool (L-016040), and siGENOME Non-Targeting siRNA (D-001210). Briefly, for siRNA experiments, 293T cells were grown to 40 -60% confluence in 24-well plates and transfected with either or both XPC and CSB siRNAs (25 pmol each). After a 48-h incubation, 50 ng of mixed genome, 450 ng of carrier DNA, and another aliquot of siRNAs were co-transfected into the cells with Lipofectamine 2000 (Invitrogen). For all in vivo transcription assays, the cells were harvested for RNA extraction 24 h after transfection with the mixed genome.
RNA Extraction and RT-PCR-RNA transcripts arising from in vitro or in vivo transcription were isolated using the Total RNA Kit I (Omega) and subjected to two rounds of DNase I treatment with the DNA-free kit (Ambion) to eliminate the DNA contamination. cDNA was generated by using M-MLV reverse transcriptase (Promega) and a mixture of an oligo(dT) 16 primer and a gene-specific primer (5Ј-TCGGTGTTGCTGT-GAT-3Ј). RT-PCR amplification for PAGE and LC-MS/MS analyses were then performed by using Phusion high fidelity DNA polymerase and a pair of primers spanning the lesion site as described in a recent study (35). Real time quantitative RT-PCR for evaluating the extent of siRNA knockdown was performed by using the iQ SYBR Green Supermix kit (Bio-Rad) and gene-specific primers for CSB, XPC, or the control gene GAPDH as described elsewhere (35).
PAGE Analysis-PAGE analysis of transcription products of S G and S 6 mG was performed as described elsewhere (35). Briefly, a portion of the above RT-PCR fragments was incubated in a 10-l reaction containing 1ϫ NEB buffer 4, 10 units of SacI, and 1 unit of shrimp alkaline phosphatase at 37°C for 1 h and then at 80°C for 20 min. The resulting dephosphory-lated DNA was incubated in a 15-l solution containing 1ϫ NEB buffer 4, ATP (50 pmol of cold, premixed with 1.66 pmol of [␥-32 P]ATP), and 10 units of T4 polynucleotide kinase. The mixture was then incubated at 37°C for 30 min and at 65°C for 20 min, after which 10 units of FspI was added and incubated at 37°C for 1 h. The resulting 32 P-labeled restriction fragments were resolved by using 30% polyacrylamide gel (acrylamide:bisacrylamide ϭ 19:1) with 8 M urea and quantified by phosphorimaging analysis (21,22). The relative mutation frequency (RMF) was determined by the ratio of the amount of detectable mutant restriction product to the total amounts of restriction fragments arising from the lesion-bearing genome. The relative bypass efficiency (RBE) was calculated using the following formula, RBE ϭ (lesion signal/competitor signal)/(nonlesion control signal/competitor signal) (22,35,38).
LC-MS/MS Analysis-To identify the transcription products of S G and S 6 mG using LC-MS/MS, RT-PCR products were treated with 50 units of SacI, 50 units of FspI, and 20 units of shrimp alkaline phosphatase in 250 l of NEB buffer 4 at 37°C for 4 h, followed by heating at 80°C for 20 min. The resulting solution was extracted with phenol/chloroform/isoamyl alcohol (25:24:1, v/v/v), and the aqueous portion was dried with a SpeedVac, desalted with HPLC, and dissolved in water. The resultant ODN mixture was subjected to LC-MS/MS analysis following previously described conditions (21,22,35). Briefly, a 0.5 ϫ 150-mm Zorbax SB-C18 column (Agilent Technologies) was used. The flow rate was 8.0 l/min, and a 5-min linear gradient of 5-35% methanol followed by a 15 min of 35-95% methanol in 400 mM 1,1,1,3,3,3-hexafluoro-2-propanol buffer (pH was adjusted to 7.0 by the addition of triethylamine) was employed for the separation. The LTQ linear ion trap mass spectrometer was set up for monitoring the fragmentation of the [M-3H] 3Ϫ ions of the 14-mer (d(GCAAAMCTAGAGCT), where M designates A, T, C, or G) ODNs.

Construction of Plasmids Containing a Site-specifically
Inserted S G or S 6 mG-To investigate how S G or S 6 mG perturbs the efficiency and fidelity of transcription, we first constructed nonreplicating double-stranded plasmids with a single S G or S 6 mG at a defined site on the transcribed strand (Fig. 1B). The lesions were positioned 63 and 44 nucleotides downstream of the transcription start sites of the cytomegalovirus and T7 promoters, respectively. In this context, we chose T7 RNA polymerase (T7 RNAP) as a model for assessing the potential impact of these lesions on transcription of mitochondrial genome, because there is a high degree of homology between T7 RNAP and eukaryotic mitochondrial RNAPs (39).
Effects of S G and S 6 mG on Transcription in Vitro-To examine the impact of S G or S 6 mG on transcription in vitro, the lesion-bearing or nonlesion control plasmids were individually premixed with a competitor genome and used as DNA templates for multiple rounds of transcription with T7 RNAP or HeLa cell nuclear extract. The latter served as a source of human RNA polymerase II (hRNAPII) transcription machinery. Relative to the control plasmids containing an unmodified guanine in lieu of a lesion, the competitor genome has three additional nucleotides near the relevant site (Fig. 1B). After in vitro transcription, the RNA products of interest were identified and quantified by restriction digestion of RT-PCR products, and PAGE and LC-MS/MS analyses of appropriate restriction fragments (Figs. 1B and 2A). The effects of S G and S 6 mG on transcriptional efficiency and fidelity were determined, as previously described (35), by the RBE and RMF, respectively (see "Experimental Procedures").
PAGE analysis showed that the 10-mer DNA fragments (d(CTAGNTTTGC), where N designates A, T, C, or G) with a single-nucleotide difference opposite the lesion site can be resolved from each other (Fig. 2B, lanes 7-10). Except for the 10-mer 32 P-labeled product with the wild-type sequence (i.e., d(CTAGCTTTGC)), no mutated products were detectable for S G-bearing plasmid by T7 RNAP and hRNAPII in vitro (Fig. 2B,  lanes 1 and 4). By using LC-MS and MS/MS analyses, we monitored the fragmentation of the [M-3H] 3Ϫ ions of the complementary 14-mer fragments (d(GCAAAMCTAGAGCT), where M designates A, T, C, or G). Again, we found that only the wild-type sequence (d(GCAAAGCTAGAGCT)) could be detected in the restriction mixtures arising from the in vitro transcription of S G-containing substrates (data not shown). In addition, a single S G on the transcribed strand did not considerably affect the transcriptional bypass efficiencies of T7 RNAP or hRNAPII in vitro (Fig. 2C).  NOVEMBER 30, 2012 • VOLUME 287 • NUMBER 49 PAGE analysis also showed that, unlike S G, transcriptional bypass of S 6 mG in vitro with both T7 RNAP and hRNAPII generated at least one type of mutant transcript (Fig. 2B, lanes 2  and 5), which contains a uridine misincorporation opposite the S 6 mG (Fig. 3A). We further confirmed the identity of this mutation by LC-MS and MS/MS analyses. Again, we monitored the fragmentation of the [M-3H] 3Ϫ ions of the complementary 14-mer fragments (d(GCAAAMCTAGAGCT), where M des-ignates A, T, C, or G) and found that only the wild-type 14-mer sequence (i.e., d(GCAAAGCTAGAGCT)) and another 14-mer fragment with the mutated sequence (i.e., d(GCAAAAC-TAGAGCT)) were detectable (Fig. 3, B and C). The quantification data from PAGE analysis revealed that S 6 mG was highly mutagenic in in vitro transcription by T7 RNAP and hRNAPII, with RMF values of 49 and 85%, respectively (Fig. 2D). In addition, we found that S 6 mG acted as a modest inhibitor of tran-

FIGURE 2. Transcriptional alternations induced by S G and S 6 mG in in vitro transcription systems using T7 RNAP or HeLa nuclear extract (hRNAPII).
A, sample processing for PAGE analysis (p* indicates 32 P-labeled phosphate group). The RT-PCR products were first treated with SacI, and the 5Ј-phosphate groups in the resulting products were removed with shrimp alkaline phosphatase. The dephosphorylated DNA was radiolabeled on the 5Ј end with [␥-32 P]ATP and T4 polynucleotide kinase (T4 PNK), after which the second enzyme (FspI) was added to produce the 32 P-labeled restriction fragments for PAGE analysis. Only the RT-PCR product arising from the lesion-containing genome is shown. B, representative gel images showing the restriction fragments released from the RT-PCR products arising from transcription templates containing S G (lanes 1 and 4), S 6 mG (lanes 2 and 5), or normal guanine (G, lanes 3 and 6). 10mer-TA, 10mer-TG, 10mer-TT, and 10mer-TC represent standard ODNs d(CTAGNTTTGC), where N is A, G, T, and C, respectively (lanes 7-10, respectively). 13mer-Comp represents standard ODN d(CATCAAGCTTTGC) (lane 11), which corresponds to the restriction fragment arising from the competitor genome. C and D, the RBE values of S G and S 6 mG (C) and RMF values of S 6 mG (D) in in vitro transcription systems using T7 RNAP or HeLa nuclear extract (hRNAPII). The data represent the means and standard error of results from three independent experiments. scription by T7 RNAP while posing a major impediment to transcription by hRNAPII, with the RBE values being 69 and 24%, respectively (Fig. 2C).
Effects of S G and S 6 mG on Transcription in Human Cells-We then performed the competitive transcription and adduct bypass assay to examine how S G and S 6 mG affect DNA transcription in vivo. Similar to the aforementioned in vitro experiments, we premixed either lesion-bearing or control plasmids with the competitor genome and co-transfected the mixed DNA substrates into NER-deficient (GM04429, lacking XPA) and repair-proficient (GM00637) human skin fibroblasts for in vivo transcription studies. After a 24-h incubation, the RNA products were isolated and processed as described above, and the resulting restriction fragments were subjected to PAGE and LC-MS/MS analysis (Fig. 1B).
PAGE analysis showed that transcriptional bypass of S G did not produce any detectable mutant transcripts in human skin fibroblast cells (Fig. 4A, lanes 6 and 9). We also confirmed these results with LC-MS and MS/MS analyses (data not shown). Furthermore, S G only modestly compromised transcriptional efficiency in both GM00637 and GM04429 cells at 24 h after transfection, with similar RBE values of ϳ 65% (Fig. 4B). On the  NOVEMBER 30, 2012 • VOLUME 287 • NUMBER 49 other hand, ϳ20% of the transcripts contained a uridine misincorporation opposite the S 6 mG in GM00637 cells, and a higher degree of this mutation (ϳ 75%) was observed in XPA-deficient GM04429 cells (Fig. 4, A and C). In addition, the RBE value for S 6 mG was significantly (p Ͻ 0.01) lower in GM04429 cells (ϳ29%) than in the repair-proficient GM00637 cells (ϳ43%) (Fig. 4B). Thus, these results indicated that XPA, a core component of NER machinery (32), contributes to the removal of S 6 mG, but not S G, in human cells.

S 6 -Methylthioguanine Impairs Transcription
Because XPC and CSB are key players in global-genome repair and TC-NER subpathways, respectively (32), we used siRNAs to diminish the expression of these two genes, individually or in combination, in 293T cells and assessed their roles in the transcriptional alternations induced by S G and S 6 mG. Real time PCR results revealed that the siRNA knockdown was highly efficient for both CSB and XPC genes (Fig. 5). The results from PAGE analysis showed that depletion of CSB alone caused a considerable increase in transcriptional mutagenesis induced by S 6 mG, with the RMF values significantly (p Ͻ 0.01) increas-ing from 10% in control siRNA-treated cells to 28% in CSBdepleted cells (Fig. 6, A, lanes 7 and 10, and B). In contrast, knockdown of XPC alone did not change appreciably the mutagenic properties of S 6 mG (Fig. 6, A, lanes 7 and 13, and B), and simultaneous knockdown of CSB and XPC gave a similar RMF value for S 6 mG as knockdown of CSB alone (Fig. 6, A, lanes 10  and 16, and B). Likewise, although depletion of XPC had no effect (Fig. 6, A, lanes 7 and 13, and C), siRNA knockdown of CSB conferred a significant (p Ͻ 0.01) reduction in RBE value for S 6 mG compared with control siRNA treatment (Fig. 6, A,  lanes 7 and 10, and C). In addition, we found that S G modestly decreases transcriptional efficiency without introducing a mutation in 293T cells, and neither CSB nor XPC was required for the removal of S G in human cells (Fig. 6, A and C).

DISCUSSION
It has been previously shown that a single S G on the transcribed strand weakly inhibits transcription by yeast RNAPII in vitro (10). In agreement with this finding, our results demon-  8 and 11). 10mer-TA, 10mer-TG, 10mer-TT, and 10mer-TC represent standard ODNs d(CTAGNTTTGC), where N is A, G, T, and C, respectively (lanes 1-4, respectively). 13mer-Comp represents standard ODN d(CATCAAGCTTTGC) (lane 5), which corresponds to the restriction fragment arising from the competitor genome. The band between 13mer-Comp and 10mer-TT is most likely a nonspecific digestion product and is not lesion-dependent, because it is also observed with lesion-free control templates. B and C, the RBE values of S G and S 6 mG (B) and RMF values of S 6 mG (C) based on in vivo transcription experiments using GM04429 or GM00637 cells. The data represent the means and standard error of results from three independent experiments. **, p Ͻ 0.01; ***, p Ͻ 0.001. The p values were calculated by using unpaired two-tailed Student's t test.
strated that S G in DNA does not considerably compromise transcription elongation mediated by single-subunit T7 RNAP or multisubunit hRNAPII in vitro and in human cells. In addition, we found that depletion of NER factors, including XPA, CSB, and XPC, does not lead to a change in the transcription bypass efficiency for S G. Again, these results are consistent with previously published data indicating that DNA S G is a poor substrate for NER in human cells (10,21). Moreover, to our knowledge, we showed for the first time that S G is not mutagenic during transcription in vitro or in vivo. In this vein, it was reported recently that S G does not substantially block DNA replication, although it can induce G 3 A mutations at frequencies of 8 -10% in E. coli and mammalian cells (21,22). Unlike S G, a single S 6 mG in the transcribed strand was found to appreciably impede transcription by hRNAPII and induce transcriptional mutagenesis in vitro and in human cells. A similar but less pronounced effect of S 6 mG on transcription was also observed in T7 RNAP-mediated in vitro transcription system. More importantly, we found that the impact of S 6 mG on transcriptional efficiency and fidelity is modulated by XPA and CSB, but not XPC. These findings indicate that, when located on the template strand of an actively transcribed gene, S 6 mG is primarily repaired by TC-NER in human cells. Similar to our results, it was reported that O 6 -methylguanine, a structural analog of S 6 mG, contributes to mutagenesis and blockage of   6, 9, 12, and 15), S 6 mG (lanes 7, 10, 13, and 16), or normal guanine (G, lanes 8, 11, 14, and 17). 10mer-TA, 10mer-TG, 10mer-TT, and 10mer-TC represent standard ODNs d(CTAGNTTTGC), where N is A, C, G, and T, respectively (lanes 1-4, respectively). 13mer-Comp represents standard ODN d(CATCAAGCTTTGC) (lane 5), which corresponds to the restriction fragment arising from the competitor genome. The band observed between 13mer-Comp and 10mer-TT is most likely a nonspecific digestion product and is not lesion-dependent, because it is also obtained with lesion-free control templates. B and C, the RMF values of S 6 mG (B) and the RBE values of S G and S 6 mG (C) in 293T cells treated with either or both CSB and XPC siRNAs. The data represent the means and standard error of results from three independent experiments. *, p Ͻ 0.05; **, p Ͻ 0.01. The p values were calculated by using unpaired two-tailed Student's t test.
transcription and thus might invoke TC-NER (40,41). Previous replication studies have also shown that S 6 mG is capable of inducing high frequencies of G 3 A transition in E. coli and mammalian cells (21,22). However, S 6 mG is not a strong block to DNA replication, and deficiency in NER factor XPA does not considerably alter the effects of S 6 mG on the efficiency and fidelity of DNA replication (21,22).
Transcriptional inhibition has been suggested to be a potential strategy for cancer therapy (29 -31). It has been shown that the survival of tumor cells requires the expression of antiapoptotic factors, and tumor cells appear to be more sensitive to inhibition of transcription than normal cells (42). In this context, the half-lives of mRNAs encoding antiapoptosis proteins are generally shorter than those of apoptosis-promoting factors. In addition, many of these antiapoptosis genes are on average larger in size than the apoptosis-promoting genes. Thus, transcription inhibitors are thought to selectively impede the expression of these antiapoptosis factors and induce the apoptosis of cancer cells (29,43,44). Many transcriptional inhibitors have been clinically tested as anticancer agents, including cisplatin and other platinum-based drugs (31,(45)(46)(47). Given that S 6 mG, but not S G, exhibits a strong inhibitory effect on transcription in our experimental systems, it is therefore possible that S G in human genomic DNA has little or no direct role, at the level of transcription, in the cytotoxic effects of thiopurine drugs; however, S 6 mG, a methylation damage derived from S G in DNA, may contribute, at least in part, to the cytotoxicity of thiopurine drugs through serving as a transcriptional inhibitor. On the other hand, it has been previously suggested that both S 6 mG and S G can induce replication errors and subsequently trigger futile cycles of MMR, which is widely regarded as a major contributor to thiopurine toxicity (8, 20 -22, 24, 25).
The mechanisms through which S 6 mG interferes with transcription may be similar to those of platinum drug-induced DNA damage (31). It has been shown that DNA damage induced by platinum-based agents may block transcription at both the initiation and elongation stages (31,45,46). Thus, aside from its effect on transcription elongation observed in this study, it would be interesting to investigate, in the future, whether S 6 mG also affects transcription initiation by altering the binding of transcription factors to DNA.
Long term use of thiopurines in patients is associated with an elevated incidence of certain iatrogenic cancers (48 -51). Previous studies have provided some clues that the mutations induced by S 6 mG and S G during replication might have a role in the carcinogenicity of thiopurine drugs (21,22). In this vein, transcriptional mutagenesis has also been proposed as one of the principal inducers of cancer and other human diseases, despite the lack of direct evidence linking transcriptional mutations to cancer (33,40,52,53). Notably, a previous study has provided important implications for the role of transcriptional mutagenesis in tumorigenesis in that mutagenic transcriptional bypass of 8-oxoguanine could lead to activation of an oncogenic pathway (53). Despite the very low frequency of formation after thiopurine treatment, S 6 mG in cellular DNA appears to be at a similar level as O 6 -methylguanine and abundant oxidative DNA lesions such as 8-oxoguanine (7,8,54). The observation that S 6 mG, but not S G, can lead to the generation of mutant transcripts at high frequency, suggests that S 6 mG-induced transcriptional mutagenesis may also contribute to the development of thiopurine therapy-related cancers and other complications.
Taken together, the present study demonstrates that, when located on the transcribed DNA strand, S 6 mG, but not S G, can cause strong mutagenic and inhibitory effects on transcription and act as an efficient substrate for TC-NER. These findings suggest that, at the level of transcription, S G has little or no direct impact on the activities of thiopurine drugs, but its methylation product (i.e., S 6 mG) might have two opposite effects: one contributes to thiopurine-induced cytotoxicity through inhibition of transcription, and the other leads to thiopurine therapy-related cancers by inducing mutant transcripts. Therefore, the work reported here improves our understanding of the mechanisms by which thiopurine drugs exert their cytotoxic and potential carcinogenic effects.