Molecular Basis of Transcriptional Mutagenesis at 8-Oxoguanine*

Structure-function analysis has revealed the mechanism of yeast RNA polymerase II transcription at 8-oxoguanine (8-oxoG), the major DNA lesion resulting from oxidative stress. When polymerase II encounters 8-oxoG in the DNA template strand, it can misincorporate adenine, which forms a Hoogsteen bp with 8-oxoG at the active center. This requires rotation of the 8-oxoG base from the standard anti- to an uncommon syn-conformation, which likely occurs during 8-oxoG loading into the active site. The misincorporated adenine escapes intrinsic proofreading, resulting in transcriptional mutagenesis that is observed directly by mass spectrometric RNA analysis.

Here, we investigated the molecular basis of 8-oxoG transcription by pol II. We show that pol II from the yeast Saccharomyces cerevisiae can misincorporate adenine at 8-oxoG and that the misincorporated adenine escapes intrinsic polymerase proofreading and remains in the transcript upon elongation, resulting in transcriptional mutagenesis. The mutant transcript was observed directly by mass spectrometry. We then used x-ray analysis to show that the misincorporated adenine forms a Hoogsteen bp with 8-oxoG at the pol II active center. In contrast, 8-oxoG forms a standard Watson-Crick bp with a correctly incorporated cytosine. Finally, we suggest a mechanism for mutagenic transcription and its suppression by TFIIS.

EXPERIMENTAL PROCEDURES
Sample Preparation-S. cerevisiae pol II containing a hexahistidine-tagged Rpb3 subunit (strain kindly provided by the M. Kashlev laboratory) was purified as described (8). Briefly, cells were lysed by bead beating. The lysate was cleared by centrifugation and ultracentrifugation. The cleared lysate was precipitated by the addition of saturated ammonium sulfate. The pellet was subjected to nickel-nitrilotriacetic acid affinity chromatography. The eluted protein was subjected to anion exchange chromatography (Mono Q, GE Healthcare). The last elution peak was collected and concentrated. The concentrated pol II was precipitated by the addition of ammonium sulfate, and the pellets were stored at Ϫ80°C. Recombinant TFIIS was prepared as described (9). DNA strands containing 8-oxoG were synthesized by BioSpring.
Elongation Complex Assembly and RNA Extension and Cleavage Assays-For RNA extension assays, synthetic oligonucleotides were annealed for scaffold assembly as described (10). Elongation complexes (ECs) were obtained by incubating pol II with 2 M eq of nucleic acid scaffold in transcription buffer (20 mM HEPES (pH 7.6), 60 mM (NH 4 ) 2 SO 4 , 8 mM MgSO 4 , 10 M ZnCl 2 , 10% (v/v) glycerol, and 10 mM dithiothreitol) at 20°C for 30 min. For RNA extension assays, different amounts of NTPs were added, and the mixture was incubated at 28°C as described (10). ECs containing complete complementary scaffolds were assembled essentially as described (11). Briefly, the DNA non-template was 5Ј-end-labeled with biotin using a TTTTT linker. The RNA was 5Ј-end-labeled with 6-carboxyfluorescein and a UGCAU linker. For EC assembly, pol II was incubated with a hybrid of the DNA template strand annealed to the RNA (2-fold excess) for 15 min at 20°C, subsequently with the biotinylated DNA non-template strand (4-fold excess) for 10 min at 25°C, and finally with recombinant Rpb4/7 (5-fold excess) for 10 min at 25°C. Bead-based RNA extension assays were carried out as described (12) with minor changes. In summary, beads (Dynabeads MyOne TM streptavidin T1, Invitrogen) were added and incubated for 30 min at 25°C. Beads were subsequently washed with transcription buffer containing 0.1% Triton-X, transcription buffer containing 0.2 M (NH 4 ) 2 SO 4 , and transcription buffer. Beads were resuspended in transcription buffer and incubated with NTPs at 28°C. The reaction was stopped by the addition of 100 mM EDTA. The beads were transferred into urea loading buffer, and samples were loaded onto a 20% polyacrylamide gel containing 7 M urea. The RNA products 5Ј-end-labeled with 6-carboxyfluorescein were visualized with a Typhoon 9400 scanner (GE Healthcare). Gel bands were quantified using ImageQuant (GE Healthcare assuming Michaelis-Menten kinetics using SigmaPlot. For MALDI time-of-flight analysis, reactions were incubated with the indicated NTPs, stopped, and analyzed as described (10). Because the amounts of product required for MALDI analysis could not be obtained with the use of complementary scaffolds, we used minimal scaffolds. Crystal Structure Analysis-The 8-oxoG-containing scaffolds were co-crystallized, and the structures were determined FIGURE 1. 8-oxoG directs slow adenine misincorporation. a, base pairing properties of 8-oxoG. b, nucleic acid scaffolds (color coding used throughout). RNA was 5Ј-end-labeled with 6-carboxyfluorescein and a UGCAU linker. Filled circles denote nucleotides present in the structures (see Fig. 3). c, RNA extension with EC A. d, time courses of incorporation with CTP or misincorporation with ATP. The indicated complementary scaffold was used. The DNA non-template was 5Ј-end-labeled with biotin using a TTTTT linker. The RNA was 5Ј-end-labeled with 6-carboxyfluorescein and a UGCAU linker. essentially as described (13) with minor changes. The crystallization solution lacked magnesium ions (200 mM ammonium acetate, 300 mM sodium acetate, 50 mM HEPES (pH 7.0), 4 -6% (w/v) polyethylene glycol 6000, and 5 mM tris(2-carboxyethyl)phosphine). Diffraction data were collected at a wavelength of 0.919 Å at the protein crystallography beamline X06SA of the Swiss Light Source using a PILATUS 6M pixel detector (see Table 1) (14). Raw data were processed with XDS (15). Structures were solved by molecular replacement with the program Phaser (16) using the structure of the complete 12-subunit pol II without nucleic acids (Protein Data Bank code 1Y1W) (9). The molecular replacement solution was subjected to rigid body refinement with CNS Version 1.2 (17). Model building was done with Coot (18) and Moloc (Gerber Molecular Design). The nucleic acids were built stepwise into unbiased F o Ϫ F c electron density. The register of the nucleic acids was unambiguously defined by bromine labeling as described (10). Refinement was monitored with the free R-factor, calculated from the same set of excluded reflections as in the refinement of the complete pol II complex (19) and the complete pol II EC (9, 10, 13).

Yeast pol II Slowly Misincorporates Adenine at 8-oxoG-We
reconstituted an S. cerevisiae pol II EC containing 8-oxoG at position ϩ1 of the template strand, opposite the NTP-binding site (Fig. 1b, Scaffold A) (13). Incubation of the resulting EC A with NTPs led to RNA extension and accumulation of the runoff product (Fig. 1c, lanes 1-3; "Experimental Procedures"; and supplemental Fig. 1). Incubation with individual NTPs showed that both cytosine and adenine are incorporated opposite the lesion (Fig. 1c, lanes 4 -11). Uridine was also incorporated, but this was also observed for undamaged DNA (supplemental Fig. 1) and is caused by a lesion-independent mechanism called template misalignment (20). Incubation with GTP also resulted in some misincorporation (Fig. 1c, lanes 4 -11), and for undamaged DNA, some misincorporation of ATP was also observed (supplemental Fig. 1), but these misincorporations are at a level expected from yeast pol II in the absence of TFIIS (8). To further investigate adenine misincorporation and cytosine incorporation opposite the lesion, we performed a time course analysis with an EC that contained a fully complementary complete scaffold. This revealed that adenine misincorporation was ϳ34-fold slower than cytosine incorporation (Fig. 1d and "Experimental Procedures"), suggesting a misincorporation frequency of a few percent, consistent with data for mammalian pol II (7).
Adenine Misincorporation Results in Transcriptional Mutagenesis-To test whether misincorporation occurs when CTP is present and whether the misincorporated adenine remains in the RNA, we analyzed RNA products by MALDI mass spectrometry. We could distinguish RNAs containing cytosine or adenine, which differ in mass by only 24 Da (Fig. 2). RNA with the misincorporated adenine was detected when an excess of ATP over CTP was used (Fig. 2a). At an equimolar ATP:CTP ratio, misincorporation could not be detected, apparently because the mass of the misincorporation product was indistinguishable from that of the cytosine incorporation product associated with a sodium ion (Fig. 2a). When EC A was incubated with ATP before NTP addition, runoff RNA contained the misincorporated adenine (Fig. 2b). In summary, 8-oxoG can cause adenine misincorporation in the presence of CTP, and the inserted adenine escapes polymerase intrinsic proofreading and remains in the RNA, leading to transcriptional mutagenesis.
TFIIS Removes a Misincorporated Adenine-Because TFIIS enhances intrinsic pol II RNA cleavage (21,22), we investigated whether TFIIS can stimulate removal of a misincorporated adenine opposite the 8-oxoG lesion. We added recombinant TFIIS to ECs B and C, which contained cytosine or adenine, opposite the lesion (Fig. 1b), and incubated the mixture for 5 min. This led to cleavage of an RNA 3Ј-dinucleotide, which was not observed for undamaged DNA containing a G-C match (Fig. 2c, Sc. B undamaged ). Dinucleotide cleavage generally occurs in the presence of TFIIS (23) and was also observed for undamaged DNA containing a G-A mismatch (Fig. 2c, Sc. C undamaged ). Upon subsequent incubation with NTPs for 5 min, RNA was elongated (Fig. 2c). MALDI analysis of runoff RNA produced by EC C, incubated with both TFIIS and NTPs for 5 min, indicated replacement of the RNA 3Ј-adenine by cytosine (data not shown). This showed that TFIIS had stimulated removal of the misincorporated adenine from the RNA prior to elongation. These data suggest that TFIIS suppresses transcriptional mutagenesis by stimulating proofreading, but they do not demonstrate proofreading under rapid elongation conditions. During rapid elongation, proofreading apparently does not occur to a large extent, at least with mammalian pol II (7).

8-oxoG and the Misincorporated Adenine Form a Hoogsteen
Pair-To investigate the molecular basis for 8-oxoG transcription, we solved crystal structures of pol II ECs containing 8-oxoG (see "Experimental Procedures"). An interpretable electron density map was obtained with an 8-oxoG-C bp at the penultimate position in the hybrid (Fig. 1b, Scaffold D). The structure was solved at 3.7 Å resolution ( Fig. 3 and Table 1). Bromine labeling of the DNA revealed that the EC was not post-translocated but adopted a state with a frayed RNA 3Ј-end at register ϩ1. This RNA fraying does not influence our analysis, as it occurred also in a control structure with undamaged DNA (supplemental Fig. 2). We recently published a detailed structural analysis of RNA fraying and showed that it occurs with different types of terminal RNA nucleotides and that frayed nucleotides can be stabilized by a neighboring mismatch pair (8).

DISCUSSION
We have shown that yeast pol II can misincorporate adenine at an 8-oxoG lesion in the DNA template strand, that the misincorporated adenine forms a Hoogsteen bp with the lesion at the polymerase active center, and that the misincorporated adenine escapes polymerase intrinsic proofreading and remains in the RNA as a product of transcriptional mutagenesis. How may misincorporation and proofreading occur? In incoming DNA, 8-oxoG likely adopts a standard anti-conformation within a Watson-Crick bp, but its base must rotate by 180°t o the syn-conformation in the 8-oxoG-A DNA-RNA Hoogsteen bp at the active center. Modeling shows that base rotation in the templating position at register ϩ1 would result in a clash with the bridge helix and the DNA base at register Ϫ1. We therefore propose that base rotation occurs during 8-oxoG translocation from the downstream position ϩ2 to the templating position ϩ1 in the recently characterized pre-templating position (24), where the 8-oxoG base would be unpaired and free to rotate from anti to syn before being loaded into the templating position. In the templating site, 8-oxoG likely forms a Hoogsteen bp with an incoming ATP, leading to adenine misincorporation. TFIIS induces backtracking of 8-oxoG from the active center to downstream DNA. When repeated translocation preserves the anti-conformation, correct cytosine incorporation occurs, and RNA is elongated error-free.
Transcriptional mutagenesis as demonstrated here in vitro may be a threat in vivo because a yeast strain lacking the gene encoding TFIIS is sensitive to oxidative stress (25,26). This in vivo effect may result at least in part from the accumulation of mutant RNAs that give rise to mutant proteins with impaired function. Transcriptional mutagenesis does not occur at dinucleotide lesions in the template, although these also cause misincorporation (10,13). At a cyclobutane pyrimidine dimer, misincorporation triggers pol II stalling (10), whereas misin-FIGURE 2. Transcriptional mutagenesis. a, adenine misincorporation in EC A in the presence of CTP at different ratios (1 mM ATP, 5-min incubation). RNA products containing adenine (5900 Da) or cytosine (5924 Da) were identified by mass spectrometry (see "Experimental Procedures"). b, misincorporated adenine remains in the runoff (RO) RNA. EC A was incubated with NTPs (1 mM, 60 min) or with ATP (1 mM, 5 min), followed by incubation with NTPs (1 mM, 60 min) as indicated. Mass spectrometry identified runoff RNAs containing cytosine (10,393 Da) or adenine (10,417 Da). c, effect of the cleavage stimulatory factor TFIIS on RNA elongation with ECs B and C. Incubation with TFIIS for 5 min resulted in cleavage products (lanes 3, 10, and 11). The presence of NTPs (1 mM) increased runoff formation (lanes 6 and 14) independent of the order of addition (lanes 7, 8, 15, and 16). * and #, RNA impurities and a cleavage product after backtracking, respectively. Controls included incubations of fluorescent RNA with pol II (lanes  4 and 12) and of lesion-free ECs B undamaged and C undamaged with TFIIS (lanes 2 and 10).
corporation is a result of stalling at a cisplatin cross-link lesion (13).