Bloom syndrome DNA helicase mitigates mismatch repair-dependent apoptosis.

Generation of O6-methylguanine (O6-meG) by DNA-alkylating agents such as N-methyl N-nitrosourea (MNU) activates the multiprotein mismatch repair (MMR) complex and the checkpoint response involving ATR/CHK1 and ATM/CHK2 kinases, which may in turn trigger cell cycle arrest and apoptosis. The Bloom syndrome DNA helicase BLM interacts with the MMR complex, suggesting functional relevance to repair and checkpoint responses. We observed a strong interaction of BLM with MMR proteins in HeLa cells upon treatment with MNU as evidenced by co-immunoprecipitation as well as colocalization in the nucleus as revealed by dual immunofluorescence staining. Knockout of BLM sensitized HeLa MR cells to MNU-induced cell cycle disruption and enhanced expression of the apoptosis markers cleaved caspase-9 and PARP1. MNU-treated BLM-deficient cells also exhibited a greater number of 53BP1 foci and greater phosphorylation levels of H2AX at S139 and RPA32 at S8, indicating the accumulation of DNA double-strand breaks. These findings suggest that BLM prevents double-strand DNA breaks during the MMR-dependent DNA damage response and mitigates O6-meG-induced apoptosis.

FANCD2 counteracts O6-methylguanine-induced mismatch repair-dependent apoptosis.

BACKGROUND: Sn1-type alkylating agents methylate the oxygen atom on guanine bases thereby producing O6-methylguanine. This modified base could pair with thymine and cytosine, resulting in the formation of O6-methylguanine/thymine mismatch during DNA replication, recognized by the mismatch repair (MMR) complex, which then initiates the DNA damage response and subsequent apoptotic processes. In our investigation of the molecular mechanisms underlying MMR-dependent apoptosis, we observed FANCD2 modification upon the activity of alkylating agent N-methyl-N-nitrosourea (MNU). This observation led us to hypothesize a relevant role for FANCD2 in the apoptosis induction process. METHODS AND RESULTS: We generated FANCD2 knockout cells using the CRISPR/Cas9 method in the human cervical cancer cell line HeLa MR. FANCD2-deficient cells exhibited MNU hypersensitivity. Upon MNU exposure, FANCD2 colocalized with the MMR complex. MNU-treated FANCD2 knockout cells displayed severe S phase delay followed by increased G2/M arrest and MMR-dependent apoptotic cell death. Moreover, FANCD2 knockout cells exhibited impaired CtIP and RAD51 recruitment to the damaged chromatin and DNA double-strand break accumulation, indicated by simultaneously observed increased γH2AX signal and 53BP1 foci. CONCLUSIONS: Our data suggest that FANCD2 is crucial for recruiting homologous recombination factors to the sites of the MMR-dependent replication stress to resolve the arrested replication fork and counteract O6-methylguanine-triggered MMR-dependent apoptosis.

SMARCAD1-mediated recruitment of the DNA mismatch repair protein MutLα to MutSα on damaged chromatin induces apoptosis in human cells.

The mismatch repair (MMR) complex is composed of MutSα (MSH2-MSH6) and MutLα (MLH1-PMS2) and specifically recognizes mismatched bases during DNA replication. O6-Methylguanine is produced by treatment with alkylating agents, such as N-methyl-N-nitrosourea (MNU), and during DNA replication forms a DNA mismatch (i.e. an O6-methylguanine/thymine pair) and induces a G/C to A/T transition mutation. To prevent this outcome, cells carrying this DNA mismatch are eliminated by MMR-dependent apoptosis, but the underlying molecular mechanism is unclear. In this study, we provide evidence that the chromatin-regulatory and ATP-dependent nucleosome-remodeling protein SMARCAD1 is involved in the induction of MMR-dependent apoptosis in human cells. Unlike control cells, SMARCAD1-knockout cells (ΔSMARCAD1) were MNU-resistant, and the appearance of a sub-G1 population and caspase-9 activation were significantly suppressed in the ΔSMARCAD1 cells. Furthermore, the MNU-induced mutation frequencies were increased in these cells. Immunoprecipitation analyses revealed that the recruitment of MutLα to chromatin-bound MutSα, observed in SMARCAD1-proficient cells, is suppressed in ΔSMARCAD1 cells. Of note, the effect of SMARCAD1 on the recruitment of MutLα exclusively depended on the ATPase activity of the protein. On the basis of these findings, we propose that SMARCAD1 induces apoptosis via its chromatin-remodeling activity, which helps recruit MutLα to MutSα on damaged chromatin.

MLH1-mediated recruitment of FAN1 to chromatin for the induction of apoptosis triggered by O6 -methylguanine.

O6 -Methylguanines (O6 -meG), which are produced in DNA by the action of alkylating agents, are mutagenic and cytotoxic, and induce apoptosis in a mismatch repair (MMR) protein-dependent manner. To understand the molecular mechanism of O6 -meG-induced apoptosis, we performed functional analyses of FANCD2 and FANCI-associated nuclease 1 (FAN1), which was identified as an interacting partner of MLH1. Immunoprecipitation analyses showed that FAN1 interacted with both MLH1 and MSH2 after treatment with N-methyl-N-nitrosourea (MNU), indicating the formation of a FAN1-MMR complex. In comparison with control cells, FAN1-knockdown cells were more resistant to MNU, and the appearances of a sub-G1 population and caspase-9 activation were suppressed. FAN1 formed nuclear foci in an MLH1-dependent manner after MNU treatment, and some were colocalized with both MLH1 foci and single-stranded DNA (ssDNA) created at damaged sites. Under the same condition, FANCD2 also formed nuclear foci, although it was dispensable for the formation of FAN1 foci and ssDNA. MNU-induced formation of ssDNA was dramatically suppressed in FAN1-knockdown cells. We therefore propose that FAN1 is loaded on chromatin through the interaction with MLH1 and produces ssDNA by its exonuclease activity, which contributes to the activation of the DNA damage response followed by the induction of apoptosis triggered by O6 -meG.

ROS control in human iPS cells reveals early events in spontaneous carcinogenesis.

Reactive oxygen species (ROS) generated during cellular respiration oxidize various cellular constituents, which cause carcinogenesis. Because most studies on the role of ROS in carcinogenesis have mainly been performed using tumor-derived cell lines, which harbor various types of mutation, it has been difficult to determine the molecular details that lead to cancer formation. To overcome this difficulty, we established human-induced pluripotent stem cell lines in which the intracellular ROS levels are controlled at various differentiation stages by manipulating the ROS-yielding mitochondria. By introducing a specific amino acid substitution (I69E) into the succinate dehydrogenase complex, subunit C protein, a component of mitochondrial respiratory chain complex II, the ROS level increased considerably. When ROS-overproducing cells at the early stage of endoderm differentiation were subcutaneously inoculated into the backs of nude mice, we observed tumor formation. These tumor-initiating cells were subjected to a comprehensive analysis by RNA sequencing. It was revealed that tumor-initiating cells showed 27 upregulated transcripts compared with control cells. The newly identified genes include those coding for PAX8 and FOSB (transcription factors) as well as FGF22, whose expressions are known to increase in developing embryos. These results suggest that these genes may play a pivotal role in cancer formation at the very early stages of cell differentiation.

Differential genomic destabilisation in human cells with pathogenic MSH2 mutations introduced by genome editing.

Repeat destabilisation is variously associated with human disease. In neoplastic diseases, microsatellite instability (MSI) has been regarded as simply reflecting DNA mismatch repair (MMR) deficiency. However, several discrepancies have been pointed out. Firstly, the MSI+ phenotype is not uniform in human neoplasms. Established classification utilises the frequency of microsatellite changes, i.e. MSI-H (high) and -L (low), the former regarded as an authentic MMR-defective phenotype. In addition, we have observed the qualitatively distinct modes of MSI, i.e. Type A and Type B. One discrepancy we previously pointed out is that tumours occurring in MMR gene knockout mice exhibited not drastic microsatellite changes typical in MSI-H tumours (i.e. Type B mode) but minor and more subtle alterations (i.e. Type A mode). In the present study, MSH2 mutations reported in Lynch syndrome (LS) kindred have been introduced into HeLa cells using the CRISPR/Cas9 system. The established mutant clones clearly exhibited MMR-defective phenotypes with alkylating agent-tolerance and elevated mutation frequencies. Nevertheless, microsatellites were not markedly destabilised as in MSI-H tumours occurring in LS patients, and all the observed alterations were uniformly Type A, which confirms the results in mice. Our findings suggest added complexities to the molecular mechanisms underlying repeat destabilisation in human genome.

Stabilization of MAPO1 by specific binding with folliculin and AMP-activated protein kinase in O6-methylguanine-induced apoptosis.

When DNA is damaged by alkylating agents, apoptosis is induced to exclude cells carrying DNA lesions in order to prevent mutations and cancer. MAPO1, identified as a component involved in the induction of apoptosis, interacts with AMP-activated protein kinase (AMPK) and folliculin (FLCN). We herein report that MAPO1 is stabilized during the course of apoptosis, triggered by alkylation-induced O(6)-methylguanine in DNA. An immunoblotting analysis revealed that the amount of MAPO1 increased gradually after treatment with N-methyl-N-nitrosourea (MNU), although the level of mRNA for MAPO1 was unchanged. When cells were exposed to a proteasome inhibitor, MG132, the MAPO1 level significantly increased. On the other hand, application of a protein synthesis inhibitor, cycloheximide, caused a decrease in the MAPO1 content, implying that proteasome-mediated degradation is involved. In FLCN-knockdown cells, the MAPO1 level decreased, and no increases occurred even after MNU treatment. In contrast, stabilization of MAPO1 occurred in AMPKα-knockdown cells even without MNU treatment. While MAPO1 retains its ability to stably bind to FLCN, it dissociates gradually from AMPK after exposure to MNU. It seems that the proapoptotic function of MAPO1 may be regulated by AMPK and FLCN through stabilization of MAPO1 itself.

Endogenous ROS production in early differentiation state suppresses endoderm differentiation via transient FOXC1 expression.

Oxidative stress plays a pivotal role in the differentiation and proliferation of cells and programmed cell death. However, studies on the role of oxidative stress in differentiation have mainly employed the detection of reactive oxygen species (ROS) during differentiation or generated by ROS inducers. Therefore, it is difficult to clarify the significance of endogenous ROS production in the differentiation of human cells. We developed a system to control the intracellular level of ROS in the initial stage of differentiation in human iPS cells. By introducing a specific substitution (I69E) into the SDHC protein, a component of the mitochondrial respiratory chain complex, the endogenous ROS level increased. This caused impaired endoderm differentiation of iPS cells, and this impairment was reversed by overproduction of mitochondrial-targeted catalase, an anti-oxidant enzyme. Expression of tumor-related FOXC1 transcription factor increased transiently as early as 4 h after ROS-overproduction in the initial stage of differentiation. Knockdown of FOXC1 markedly improved impaired endoderm differentiation, suggesting that endogenous ROS production in the early differentiation state suppresses endoderm differentiation via transient FOXC1 expression.

Novel plasmids for the fluorescence-based evaluation of DNA mismatch repair in human cells.

Mismatch repair (MMR) is a highly conserved DNA repair pathway that corrects mismatched bases during DNA replication. The biological significance of MMR in human cells is underscored by the fact that dysfunction of the MMR pathway results in Lynch syndrome, which is associated with a genetic predisposition to different cancer types. We have previously established a reporter mismatch plasmid to evaluate MMR using fluorescent proteins in living cells. However, the preparation of these plasmids requires significant amounts of time and money, which reduces their broad applicability. To overcome the abovementioned limitations, we produced in this study a novel reporter plasmid, pBSII NLS-MC-EGFP-tdTomato (pBET2), that can be used in the oligo swapping method. In this method, a nicking endonuclease produces a single-stranded DNA gap on a double-stranded DNA plasmid that can be replaced by ligation with synthetic oligonucleotides. It is significantly easier and more user-friendly than previous assays, which require in vitro DNA synthesis with single-stranded plasmid DNA and purification using ultracentrifugation in cesium chloride-ethidium bromide gradients. The plasmid also contains a nicking site that allows the MMR repair machinery to efficiently distinguish the newly synthesized strand as a target for repair. In addition, a nuclear localization signal facilitates green fluorescent protein expression in the nucleus, which helps to verify the effectiveness of MMR using fluorescence microscopy. Similar to the previous reporter plasmid, this construct facilitates the assessment of MMR proficiency in human living cells via the expression of fluorescent proteins while overcoming many of the negative aspects of the previous protocol.

DNA polymerase delta Exo domain stabilizes mononucleotide microsatellites in human cells.

In prokaryotes and yeasts, DNA polymerase proofreading (PPR) and DNA mismatch repair (MMR) cooperatively counteracts replication errors leading to repeat sequence destabilization (i.e. insertions/deletions of repeat units). However, PPR has not thus far been regarded as a mechanism stabilizing repeat sequences in higher eukaryotic cells. In a human cancer cell line, DLD-1, which carries mutations in both MSH6 and the Exo domain of POLD1, we previously observed that mononucleotide microsatellites were markedly destabilized whereas being stable in the simple MMR-defective backgrounds. In this study, we introduced the Exo domain mutation found in DLD-1 cells into MSH2-null HeLa cell clones, using CRISPR/Cas9 system. In the established Exo-/MMR-mutated HeLa clones, mononucleotide repeat sequences were remarkably destabilized as in DLD-1 cells. In contrast, dinucleotide microsatellites were readily destabilized in the parental MMR-deficient backgrounds, and the instability was not notably increased in the genome-edited HeLa clones. Here, we show an involvement of the Exo domain functions of DNA polymerase delta in mononucleotide repeat stabilization in human cells, which also suggests a possible role division between DNA polymerase and MMR in repeat maintenance in the human genome.

The WD40-repeat protein Pwp1p associates in vivo with 25S ribosomal chromatin in a histone H4 tail-dependent manner.

The tails of core histones (H2A, H2B, H3 and H4) are critical for the regulation of chromatin dynamics. Each core histone tail is specifically recognized by various tail binding proteins. Here we screened for budding yeast histone H4-tail binding proteins in a protein differential display approach by two-dimensional gel electrophoresis (2DGE). To obtain highly enriched chromatin proteins, we used a Mg2+-dependent chromatin oligomerization technique. The Mg2+-dependent oligomerized chromatin from H4-tail deleted cells was compared with that from wild-type cells. We used mass spectrometry to identify 22 candidate proteins whose amounts were reduced in the oligomerized chromatin from the H4-tail deleted cells. A Saccharomyces Genome Database search revealed 10 protein complexes, each of which contained more than two candidate proteins. Interestingly, 7 out of the 10 complexes have the potential to associate with the H4-tail. We obtained in vivo evidence, by a chromatin immunoprecipitation assay, that one of the candidate proteins, Pwp1p, associates with the 25S ribosomal DNA (rDNA) chromatin in an H4-tail-dependent manner. We propose that the complex containing Pwp1p regulates the transcription of rDNA. Our results demonstrate that the protein differential display approach by 2DGE, using a histone-tail mutant, is a powerful method to identify histone-tail binding proteins.

Modes of actions of two types of anti-neoplastic drugs, dacarbazine and ACNU, to induce apoptosis.

O(6)-Methylguanine and O(6)-chloroethylguanine, which are the primary cytotoxic DNA lesions produced by 5-(3,3-dimethyl-1-triazeno)imidazole-4-carboxamide (dacarbazine) and 1-(4-amino-2-methyl-5-pyrimidinyl)methyl-3-(2-chloroethyl)-3-nitrosourea (ACNU), respectively, can be repaired by O(6)-methylguanine-DNA methyltransferase (MGMT), coded by the MGMT gene. However, the two types of drugs exhibit different effects on cells defective in both MGMT and MLH1 functions, the latter being related to the cellular activity to recognize mismatched bases of DNA for inducing apoptosis. Human cells deficient in both MGMT and MLH1 are resistant to the killing effect of dacarbazine and exhibit an increased mutant frequency after treatment with dacarbazine. On the other hand, these doubly deficient cells are sensitive to the killing action of ACNU and there is no significant increase in ACNU-induced mutant frequency. A mismatch recognition complex, composed of MSH2, MSH6, MLH1, PMS2 and PCNA, is formed after exposing MGMT-deficient cells to dacarbazine, but not in cells treated with ACNU. In contrast, the phosphorylation of Chk1 efficiently occurs in cells treated with dacarbazine as well as with ACNU, the former being in MLH1-dependent manner, whereas the latter in MLH1-independent manner. Therefore, the signals delivered from different sources would merge at the step of Chk1 activation or at an earlier step, and the subsequent process leading to apoptosis appears to be common.

PCNA-MutSalpha-mediated binding of MutLalpha to replicative DNA with mismatched bases to induce apoptosis in human cells.

Modified bases, such as O6-methylguanines, are produced in cells exposed to alkylating agents and cause apoptosis. In human cells treated with N-methyl-N-nitrosourea, we detected a protein complex composed of MutSalpha, MutLalpha and PCNA on damaged DNA by immunoprecipitation method using chromatin extracts, in which protein-protein interactions were stabilized by chemical crosslinking. Time course experiments revealed that MutSalpha, consisting of MSH2 and MSH6 proteins, and PCNA bind to DNA to form an initial complex, and MutLalpha, composed of MLH1 and PMS2, binds to the complex when the DNA is damaged. This sequential mode of binding was further confirmed by the findings that the association of PCNA-MutSalpha complex on chromatin was observed even in the cells that lack MLH1, whereas in the absence of MSH2 no association of MutLalpha with the chromatin was achieved. Moreover, reduction in the PCNA content by small-interfering RNA or inhibition of DNA replication by aphidicolin, an inhibitor of DNA polymerase, significantly reduced the levels of the PCNA-MutSalpha-MutLalpha complex and also suppressed an increase in the caspase-3 activity, a hallmark for the induction of apoptosis. These observations imply that the induction of apoptosis is coupled with the progression of DNA replication through the action of PCNA.

Nutrient-induced FNIP degradation by SCFß-TRCP regulates FLCN complex localization and promotes renal cancer progression.

Folliculin-interacting protein 1 and 2 (FNIP1 and FNIP2) play critical roles in preventing renal malignancy through their association with the tumor suppressor FLCN. Mutations in FLCN are associated with Birt-Hogg-Dubé (BHD) syndrome, a rare disorder with increased risk of renal cancer. Recent studies indicated that FNIP1/FNIP2 double knockout mice display enlarged polycystic kidneys and renal carcinoma, which phenocopies FLCN knockout mice, suggesting that these two proteins function together to suppress renal cancer. However, the molecular mechanism functionally linking FNIP1/FNIP2 and FLCN remains largely elusive. Here, we demonstrated that FNIP2 protein is unstable and subjected to proteasome-dependent degradation via ß-TRCP and Casein Kinase 1 (CK1)-directed ubiquitination in a nutrition-dependent manner. Degradation of FNIP2 leads to lysosomal dissociation of FLCN and subsequent lysosomal association of mTOR, which in turn promotes the proliferation of renal cancer cells. These results indicate that SCFß-TRCP negatively regulates the FLCN complex by promoting FNIP degradation and provide molecular insight into the pathogenesis of BHD-associated renal cancer.

The recognition and modification sites for the bacterial type I restriction systems KpnAI, StySEAI, StySENI and StySGI.

Using an in vivo plasmid transformation method, we have determined the DNA sequences recognized by the KpnAI, StySEAI, StySENI and StySGI R-M systems from Klebsiella oxytoca strain M5a1, Salmonella eastbourne, Salmonella enteritidis and Salmonella gelsenkirchen, respectively. These type I restriction-modification systems were originally identified using traditional phage assay, and described here is the plasmid transformation test and computer program used to determine their DNA recognition sequences. For this test, we constructed two sets of plasmids, pL and pE, that contain phage lambda and Escherichia coli K-12 chromosomal DNA fragments, respectively. Further, using the methylation sensitivities of various known type II restriction enzymes, we identified the target adenines for methylation (listed in bold italics below as A or T in case of the complementary strand). The recognition sequence and methylation sites are GAA(6N)TGCC (KpnAI), ACA(6N)TYCA (StySEAI), CGA(6N)TACC (StySENI) and TAAC(7N)RTCG (StySGI). These DNA recognition sequences all have a typical type I bipartite pattern and represent three novel specificities and one isoschizomer (StySENI). For confirmation, oligonucleotides containing each of the predicted sequences were synthesized, cloned into plasmid pMECA and transformed into each strain, resulting in a large reduction in efficiency of transformation (EOT).

Function of high-mobility group A proteins in the DNA damage signaling for the induction of apoptosis.

O(6)-Methylguanine produced in DNA can pair with thymine during DNA replication, thus leading to a G-to-A transition mutation. To prevent such outcomes, cells harboring O(6)-methylguanine-containing mispair undergo apoptosis that requires the function of mismatch repair (MMR) protein complex. To identify the genes involved in the induction of apoptosis, we performed gene-trap mutagenesis and isolated a clone of mouse cells exhibiting an increased resistance to the killing effect of an alkylating agent, N-methyl-N-nitrosourea (MNU). The mutant carries an insertion in the Hmga2 gene, which belongs to a gene family encoding the high-mobility group A non-histone chromatin proteins. To elucidate the function of HMGA proteins in the apoptosis pathway, we introduced siRNAs for HMGA1 and/or HMGA2 into human HeLa MR cells defective in O(6)-methylguanine-DNA methyltransferase. HMGA1- and HMGA2-single knockdown cells showed an increased resistance to MNU, and HMGA1/HMGA2-double knockdown cells exhibited further increased tolerance compared to the control. The phosphorylation of ATR and CHK1, the appearance of a sub-G1 population, and caspase-9 activation were suppressed in the knockdown cells, although the formation of mismatch recognition complex was unaffected. These results suggest that HMGA family proteins function at the step following the damage recognition in the process of apoptosis triggered by O(6)-methylguanine.

The identification of a novel gene, MAPO2, that is involved in the induction of apoptosis triggered by O6-methylguanine.

O6-Methylguanine, one of alkylated DNA bases, is especially mutagenic. Cells containing this lesion are eliminated by induction of apoptosis, associated with the function of mismatch repair (MMR) proteins. A retrovirus-mediated gene-trap mutagenesis was used to isolate new genes related to the induction of apoptosis, triggered by the treatment with an alkylating agent, N-methyl-N-nitrosourea (MNU). This report describes the identification of a novel gene, MAPO2 (O6-methylguanine-induced apoptosis 2), which is originally annotated as C1orf201. The MAPO2 gene is conserved among a wide variety of multicellular organisms and encodes a protein containing characteristic PxPxxY repeats. To elucidate the function of the gene product in the apoptosis pathway, a human cell line derived from HeLa MR cells, in which the MAPO2 gene was stably knocked down by expressing specific miRNA, was constructed. The knockdown cells grew at the same rate as HeLa MR, thus indicating that MAPO2 played no role in the cellular growth. After exposure to MNU, HeLa MR cells and the knockdown cells underwent cell cycle arrest at G2/M phase, however, the production of the sub-G1 population in the knockdown cells was significantly suppressed in comparison to that in HeLa MR cells. Moreover, the activation of BAK and caspase-3, and depolarization of mitochondrial membrane, hallmarks for the induction of apoptosis, were also suppressed in the knockdown cells. These results suggest that the MAPO2 gene product might positively contribute to the induction of apoptosis triggered by O6-methylguanine.

Activation of AMP-activated protein kinase by MAPO1 and FLCN induces apoptosis triggered by alkylated base mismatch in DNA.

O6-methylguanine produced in DNA by the action of simple alkylating agents, such as N-methyl-N-nitrosourea (MNU), causes base-mispairing during DNA replication, thus leading to mutations and cancer. To prevent such outcomes, the cells carrying O6-methylguanine undergo apoptosis in a mismatch repair protein-dependent manner. We previously identified MAPO1 as one of the components required for the induction of apoptosis triggered by O6-methylguanine. MAPO1, also known as FNIP2 and FNIPL, forms a complex with AMP-activated protein kinase (AMPK) and folliculin (FLCN), which is encoded by the BHD tumor suppressor gene. We describe here the involvement of the AMPK-MAPO1-FLCN complex in the signaling pathway of apoptosis induced by O6-methylguanine. By the introduction of siRNAs specific for these genes, the transition of cells to a population with sub-G1 DNA content following MNU treatment was significantly suppressed. After MNU exposure, phosphorylation of AMPKα occurred in an MLH1-dependent manner, and this activation of AMPK was not observed in cells in which the expression of either the Mapo1 or the Flcn gene was downregulated. When cells were treated with AICA-ribose (AICAR), a specific activator of AMPK, activation of AMPK was also observed in a MAPO1- and FLCN-dependent manner, thus leading to cell death which was accompanied by the depolarization of the mitochondrial membrane, a hallmark of the apoptosis induction. It is therefore likely that MAPO1, in its association with FLCN, may regulate the activation of AMPK to control the induction of apoptosis triggered by O6-methylguanine.