|Home | About | Journals | Submit | Contact Us | Français|
PAHs (polycyclic aromatic hydrocarbons) are suspect lung cancer carcinogens that must be metabolically converted into DNA-reactive metabolites. P4501A1/P4501B1 plus epoxide hydrolase activate PAH to (±)-anti-benzo[a]pyrene diol epoxide ((±)-anti-BPDE), which causes bulky DNA adducts. Alternatively, aldo-keto reductases (AKRs) convert intermediate PAH trans-dihydrodiols to o-quinones, which cause DNA damage by generating reactive oxygen species (ROS). In lung cancer, the types or pattern of mutations in p53 are predominantly G to T transversions. The locations of these mutations form a distinct spectrum characterized by single point mutations in a number of hotspots located in the DNA binding domain. One route to the G to T transversions is via oxidative DNA damage. An RP-HPLC-ECD assay was used to detect the formation of 8-oxo-dGuo in p53 cDNA exposed to representative quinones, BP-7,8-dione, BA-3,4-dione, and DMBA-3,4-dione under redox cycling conditions. Concurrently, a yeast reporter system was used to detect mutations in the same cDNA samples. Nanomolar concentrations of PAH o-quinones generated 8-oxo-dGuo (detected by HPLC-ECD) in a concentration dependent manner that correlated in a linear fashion with mutagenic frequency. By contrast, micromolar concentrations of (±)-anti-BPDE generated (+)-trans-anti-BPDE-N2-dGuo adducts (detected by stable-isotope dilution LC/MS methodology) in p53 cDNA that correlated in a linear fashion with mutagenic frequency, but no 8-oxo-dGuo was detected. Previous studies found that mutations observed with PAH o-quinones were predominately G to T transversions and those observed with (±)-anti-BPDE were predominately G to C transversions. However, mutations at guanine bases observed with either PAH-treatment occurred randomly throughout the DNA-binding domain of p53. Here, we find that when the mutants were screened for dominance, the dominant mutations clustered at or near hotspots primarily at the protein–DNA interface, whereas the recessive mutations are scattered throughout the DNA binding domain without resembling the spectra observed in cancer. These observations, if extended to mammalian cells, suggest that mutagenesis can drive the pattern of mutations but that biological selection for dominant mutations drives the spectrum of mutations observed in p53 in lung cancer.
The p53 tumor suppressor gene is mutated in a large portion of human cancer, including lung cancer. Studies of databases compiling p53 mutations from over 20 000 tumor samples have identified three properties of the p53 mutations in lung cancer (1-4). The first feature is a predominance of G to T transversions in p53. Other types of cancers show different mutational patterns, generally dominated by G to A transitions, suggesting that the carcinogens responsible for the mutations are different. This is the most unambiguous signature of lung cancer. The second property, called a strand bias, is that guanine bases are preferentially mutated in the nontranscribed strand suggesting that transcription-coupled repair occurs when the G lesion occurs on the transcribed strand. This causes a preponderance of G to T transversions relative to the complementary change of C to A. The third feature is mutation of a number of hotspot codons, which account for about 50% of all the reported mutations. The most common hotspots are codons 248, 273, 249, 245, 158, and 157 but, with the exception of 157 and 158, most hotspot codons are also mutated in nonlung cancers. This suggests that, for the most part, the location of mutations is mutagen independent. Any molecular mechanism for lung-cancer initiation must account for these observations.
It is estimated that 85–90% of all lung cancer is observed in individuals that smoke (5). Tobacco smoke can cause oxidative stress and also contains a number of chemical carcinogens, including two major classes, the tobacco-specific nitrosamine 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK) and polycyclic aromatic hydrocarbons (PAH) (6). PAH are metabolically activated to either reactive diol-epoxides (7-10), (e.g., (±)-anti-BPDE)1 by the combined action of P4501A1/1B1 and epoxide hydrolase or the intermediate trans-dihydrodiols are converted to reactive and redox-active PAH o-quinones by the action of aldo-keto reductases (AKR (1A1 and 1C1–1C4)) (11-15). In previous studies, we examined the mutagenicity of (±)-anti-BPDE and PAH o-quinones using a yeast reporter gene assay that scored transcriptional competency of p53 (16, 17). We reported that PAH o-quinones were highly mutagenic provided they were allowed to redox-cycle, and that the mutation pattern that dominated was G to T transversions. These mutations were abolished by reactive oxygen species (ROS) scavengers and, under the redox-cycling conditions employed, 8-oxo-dGuo could be detected by RP-HPLC-ECD in salmon testis DNA (18, 19). In the same assay, (±)-anti-BPDE was 80-fold less mutagenic than PAH o-quinones, and the mutation pattern observed was predominated by G to C transversions (17). Under these same reaction conditions, (+)-trans-anti-BPDE-N2-dGuo ((+)-anti-BPDE-N2-dGuo) bulky adducts could be detected in calf-thymus DNA using a stable-isotope dilution LC/MS assay (20). The detection of 8-oxo-dGuo adducts and (+)-anti-BPDE-N2-dGuo adducts in bulk DNA did not prove that these lesions could account for the mutations observed in p53. In our prior studies, adduct measurements in p53 were not performed. Moreover, whereas PAH o-quinones yielded G to T transversions in p53, the mutations observed occurred randomly throughout the DNA-binding domain and did not recapitulate the mutational spectrum observed in lung cancer (16, 17, 21).
A “targeted mutagenesis” model has been proposed to account for the spectrum of mutations in lung cancer based on sequence specific DNA-adducts formed by (±)-anti-BPDE. In both treated cells and purified DNA, (±)-anti-BPDE will form adducts preferentially at many of the hotspots in p53 including codons 157, 248, and 249 (22) in one study and codons 157, 248, and 273 in another study (23). Sequence specificity is enhanced by 5′-methylation of 5′-CpG-3′ islands (24, 25). (±)-Anti-BPDE predominantly causes G to T transversions in most mutagenesis studies (21, 22, 26). Taken together, this suggests that (±)-anti-BPDE is an ultimate carcinogen by forming adducts at specific sites in p53 to cause G to T transversions. However, the targeted mutagenesis model has been challenged by Rodin and Rodin who examined the p53 database and failed to find significant differences in the spectrum of mutations between smokers and nonsmokers although they confirmed the predominance of G to T transversions in lung cancers (27, 28). They proposed that the lung cancer spectra of p53 mutations resulted from biological selection and that smoke exposure enhanced the effects of an endogenous mutagen. Rodin and Rodin further speculated that reactive oxygen species (ROS), which have long been suspected as an ultimate carcinogen, may play the predominant role in lung carcinogenesis (27-29). In their model, ROS would cause the formation of 8-oxo-2′-deoxyguanosine (8-oxo-dGuo), leading to G to T transversions, whereas genetic selection for the most advantageous mutations would determine the spectrum. Tests of either hypothesis have been inconclusive because, although adducts can sometimes be observed in hotspot codons, mutagenesis experiments have not reproduced both the pattern of mutations (type of base changes) and spectrum of mutations (location of mutations by codon) observed in lung cancer.
To address these issues, we measured p53 mutation with PAH o-quinones and (±)-anti-BPDE in the yeast reporter gene assay and used a portion of the same DNA for adduct analysis (8-oxo-dGuo and (+)-anti-BPDE-N2-dGuo). We found that in both instances a linear correlation existed between DNA-adducts and mutation frequency. To address the role of selection, we sorted PAH derived mutations into dominant and recessive mutants. With this filter in place, we showed that dominant mutations now cluster to hotspots observed in p53 in lung cancer. Because G to T transversions are preferentially formed with PAH o-quinones under redox-cycling conditions, we speculate that the mutation pattern observed in lung cancer may be attributed to 8-oxo-dGuo formation but that the mutational spectrum requires selection for dominance.
All PAHs are potentially hazardous and should be handled in accordance with the NIH Guidelines for the Laboratory Use of Chemical Carcinogens.
BP-7,8-dione, BA-3,4-dione, and 7,12-DMBA-3,4-dione were synthesized according to published methods (30). (±)-Anti-BPDE was obtained from the National Cancer Institute, Chemical Carcinogen Standard Reference Repository (Midwest-Research Institute, Kansas City, Missouri). All compounds were analyzed for purity and identity by LC/MS before use. YEASTMAKER Yeast Transformation and Plasmid Isolation Kit and all yeast culture media were purchased from Clontech (Palo Alto, CA). Adenine, l-leucine, l-tryptophan, DNase I (type II), alkaline phosphatase (type III from Escherichia coli), deferoxamine mesylate (desferal), and cupric chloride were purchased from Sigma-Aldrich Chemical Co. (St. Louis, MO). Phosphodiesterase I (PDE I, type II) from Crotalus adamanteus venom was acquired from Worthington Biochemical Corp. (Lakewood, NJ). Shrimp alkaline phosphatase (SAP) was acquired from Roche Diagnostics (Indianapolis, IN). 2′-Deoxyguanosine and [15N5]-dGuo were obtained from ICN Biomedical Inc. (Irvine, CA) and Spectra Stable Isotope (Columbia, MD), respectively. All other chemicals and enzymes were of the highest grade available, and all solvents for HPLC-ECD and LC-MS analysis were HPLC grade.
The ade reporter yeast strain, yIG397, and gap-repair expression vector pss16 were kindly provided by Dr. Richard Iggo (Swiss Institute for Experimental Cancer Research, 1066 Epalinges, Basel, Switzerland) (31, 32). Plasmid pTS76, which expresses wild-type p53 using the TRP1 selectable marker, was a generous gift from Dr. Gilberto Fronza (33). Basic methods for yeast manipulations were carried out as described (34). Liquid media contained 0.67% yeast nitrogen base, 2% dextrose, 1% casamino acids, and 20 μg/mL of adenine. Solid media for prototrophic selection of appropriate plasmids contained 0.67% yeast nitrogen base, 2% dextrose, and 2% agar with complete additions minus the relevant amino acids and nucleosides to select for auxotrophic markers.
Mutagenesis was performed using the p53 cDNA fragment and pSS16 Gap-Repair vector as described previously (16). For the p53 gap repair assay, 0.5 μg of the mutagen-treated p53 cDNA with carrier DNA was dissolved in 50 μL of TE buffer (pH 8.0). Next, 15 μL of the DNA sample (150 ng p53 cDNA plus 100 μg carrier DNA) was then mixed with 100 ng of the pSS16 gapped-vector and cotransformed into the yeast host strain yIG397 (grown to a OD600 0.6–0.9) using the lithium acetate procedure according to the YEASTMAKER™ Yeast transformation System Kit (CLONTECH). Yeast colonies expressing wild-type p53 are white and yeast colonies expressing mutant p53 are red. Red colonies were clearly identifiable after three days at 30 °C but the color is more intense after an additional 2 days at 4 °C. The spontaneous or background rate of mutation frequency in the assay was 0.4%. The mutation frequency was expressed as: [(number of red colonies − number of spontaneous red colonies)/total number of colonies] × 100 as previously described (16). The plasmids tested for dominance were described previously (17). To test dominance, we coexpressed wild-type p53 from pTS76, a TRP1-derived plasmid (33), along with individual mutants (originally isolated as red colonies using a LEU2-based plasmid) and tested colonies on leu/trp plates. Recessive mutants become white, whereas dominant mutants become pink, although in most cases the colony was not as red as the original, and a range of color intensity was observed; all pink colonies were scored as dominant.
The precipitated p53 cDNA was recovered by centrifugation at 13 000 rpm for 30 min at 4 °C and rinsed with 70% ethanol. The pellet was dissolved in Chelex-treated 10 mM potassium phosphate buffer (pH 6.5), and the concentration of the hydrated DNA was measured using a Beckman DU640 UV/vis spectrophotometer. The DNA was divided into 4 and 24 μg aliquots, respectively. A 4 μg aliquot of the DNA was exposed to 125–700 nM of three PAH o-quinones (8% DMSO v/v), 180 μM NADPH, and 100 μM CuCl2, and a 24 μg DNA aliquot was exposed to 0–20 μM (±)-anti-BPDE (8% DMSO v/v) without NADPH and CuCl2. The samples were incubated for 3 h at 37 °C. After incubation, 0.5 μg of DNA was taken from the sample and immediately mixed with 300 μg of carrier DNA (herring testis) for the p53 gap repair assay. Also, either 3.5 μg of DNA was additionally saved from the PAH exposed samples for the analysis of 8-oxo-dGuo by RP-HPLC-ECD, or 20 μg of DNA was saved for the analysis of (+)-anti-BPDE-N2-dGuo by stable isotope dilution LC/MS.
To detect 8-oxo-dGuo in p53 cDNA fragments, the quantitative digestion of p53 cDNA was conducted as previously described, with minor modifications using precautions to prevent the adventitious oxidation of guanine (18, 19). The level of 8-oxo-dGuo was quantified by HPLC-ECD and expressed as 8-oxo-dGuo per 105 dGuo as previously described (18, 19).
To detect (+)-anti-BPDE-N2-dGuo adducts in the p53 cDNA fragment, quantitative digestion of p53 cDNA was conducted according to our previous protocol (16) with the following modification. The 20 μg of (±)-anti-BPDE-treated p53 cDNA pellet was dissolved in 300 μL of 10 mM Tris-HCl (pH 7.4) containing 100 mM MgCl2. Ten microliters of DNase 1 (2 mg/3 mL 10 mM MOPS/100 mM MgCl2 buffer) (Amersham Biosciences, Piscataway, NJ) was added and incubated for 1.5 h at 37 °C. Next, 150 μL of 0.2 M glycine buffer (pH 10) was added along with 1 unit of PDE I, and the incubation was continued for 2 h at 37 °C. At the end of the incubation, 150 μL of 50 mM Tris-HCl (pH 7.4) was added with 150 μL of SAP 10X buffer and 30 units of SAP. The sample was incubated further for 2 h at 37 °C. An internal standard (12.5 μg) of (+)-trans-anti-BPDE-[15N5]-N2-dGuo was added to the digest. The p53 cDNA digest was evaporated to dryness using a vacuum concentrator. The DNA was dissolved in 200 μL of MeOH:H2O, 1:1 (v/v) and filtered through Costar Spin-X nylon centrifuge filter tube (Corning Incorporated, Corning, NY) before LC-MS and base analysis.
The anti-BPDE-N2-dGuo adducts in the DNA digest were measured by an Agilent 1100 HPLC system (Agilent Technology, Paolo Alto, CA) equipped with a CTC autosampler (Leap Technology, Carrboro, NC) coupled to an MDS-Sciex API-4000 triple-quadrupole mass spectrometer (Applied Biosystems, Foster City, CA). The analysis of stable anti-BPDE-N2-dGuo adducts was conducted on a YMC J'sphere M80 Column (4 μm; 150 mm × 2.0 mm, 80 Å) using 5 mM NH4OAc with 0.02% formic acid as mobile phase A and methanol as mobile phase B at a flow of 200 μL/min. The linear gradient started at 49% B and increased to 51% B in the first 20 min, followed by 51 to 65% B in 10 min, and continued to 100% B in 5 min. The mass spectrometer parameters were as follows: CID gas, 10 units; curtain gas, 30 units; ion source gas-1, 30 units; ion source gas-2, 10 units; ion spray voltage, 5.0 kV; ionization temperature, 500 °C; decluster potential, 50 V; entrance potential, 8 V; collision energy, 40 eV; collision cell exit potential, 22 V. MRM analyses were conducted in positive ESI mode using the following mass transitions: m/z 570.2(MH+, anti-BPDE-N2-dGuo) → m/z 257.1 (MH+-dGuo-H2O-CO); m/z 575.2 (MH+, anti-BPDE-[15N5]-N2-dGuo) → m/z 257.1 (MH+-[15N5]-dGuo-H2O-CO). The injection volume was 25 μL for the samples. The anti-BPDE-N2-dGuo: anti-BPDE-[15N5]-N2-dGuo ratio was used to calculate adduct concentration based on interpolation of the calibration curve.
DNA base analysis was conducted on a Hitachi Elite Chrom HPLC system (Hitachi High Technologies, San Jose, CA) equipped with a UV detector. The separation employed an XTerra MS5 C18 column (5 μm; 250 mm × 4.6 mm, 125 Å). As a mobile phase, Solvent A (5 mM NH4OAc, 0.02% formic acid) and solvent B (100% methanol) was developed an XTerra MS5 C18 column (5 μm; 250 mm × 4.6 mm, 125 Å). The flow rate was set to 1 mL/min, and the nucleosides were eluted with 15% B. After 15 min, solvent B was increased to 100% for 5 min to wash off the column. This was followed by equilibration at initial conditions for another 5 min. The sample injection volume was 40 μL, and DNA base levels were calculated by interpolation from the calibration curve. Overall adduct formation is given as a ratio of number of (+)-anti-BPDE-N2-dGuo adducts per 105 DNA bases.
Data were analyzed by the Student's t test. Differences between treatment groups were considered significant at p < 0.05. All experiments were repeated at least 3 times. The data are presented as the mean ( SE values.
Submicromolar concentrations of PAH o-quinones caused G to T transversions in p53 cDNA but only when the o-quinones were allowed to redox cycle in the presence of NADPH and CuCl2. In the absence of redox-cycling, little to no mutagenesis was observed, and no mutations were observed when NADPH and CuCl2 were tested alone (16, 17). The requirement for ROS and the predominance of G to T transversions suggest that the responsible lesion is 8-oxo-dGuo, a highly mutagenic modification that is often observed in smokers. Moreover, 8-oxo-dGuo is produced in salmon testis and calf thymus DNA by redox cycling o-quinones (18, 19). Because other oxidation products of guanine can occur, studies were also performed using the aldehydic reactive probe to detect aldehydic sites in bulk DNA following treatment with PAH o-quinones under redox cycling conditions. It was found that there was a linear correlation of aldehydic sites detected in the presence of 8-oxo-guanine glycosylase with 8-oxo-dGuo detected by HPLCECD. Furthermore, it was found that the rank order of oxidative lesions observed were 8-oxo-dGuo >> oxidized pyrimidines = abasic sites (19). To determine if 8-oxo-dGuo was responsible for o-quinone-induced mutations in p53, the p53 gap repair assay was conducted using p53 cDNA treated with o-quinones under redox-cycling conditions. Samples were divided and the treated DNA was analyzed both for 8-oxo-dGuo by RP-HPLC-ECD, and p53 mutagenic frequency by the yeast p53 gap repair assay. As a control, p53 cDNA was treated with (±)-anti-BPDE and the sample analyzed for anti-BPDE-N2-dGuo adducts by stable isotope dilution LC/MS, 8-oxo-dGuo by RP-HPLC-ECD, and p53 mutagenic frequency by the yeast p53 gap repair assay.
We first measured the formation of 8-oxo-dGuo. Under redox-cycling conditions, a concentration-dependent formation of 8-oxo-dGuo in p53 was observed with each of the PAH o-quinones tested. At 125–750 nM PAH o-quinone, the amount of 8-oxo-dGuo detected increased linearly to 150 ~ 600 adducts per 105 dGuo. The resultant rank order for 8-oxo-dGuo generation in p53 cDNA was BA-3,4-dione > 7,12-DMBA-3,4-dione > BP-7,8-dione, in agreement with the previous data using salmon testis DNA (19), suggesting that the purified p53 DNA was damaged by reactive oxygen. In contrast, no 8-oxo-dGuo was detected in p53 when (±)-anti-BPDE was tested as the mutagen (part D of Figure 1). Stable isotope dilution LC-MS analysis revealed that stable (+)-anti-BPDE-N2-dGuo adducts were generated in the p53 cDNA treated with (±)-anti-BPDE (Figure 2). Between 2.5 μM and 20 μM, the amounts of (+)-anti-BPDE-N2-dGuo adducts linearly increased to 150 ~ 940 adducts per 105 dGuo. We note that it required >20-fold more (±)-anti-BPDE than PAH o-quinone to achieve the same level of DNA adducts, consistent with previous mutagenesis data showing that >20-fold more (±)-anti-BPDE was required for mutagenesis (16, 17).
We next compared the mutagenic frequency with the levels of DNA adducts (Figure 3). All three quinones showed a linear corelationship between 8-oxo-dGuo and mutagenic frequencies (R2 values for 7,12-DMBA-3,4-dione, BA-3,4-dione, and BP-7,8-dione were 0.896, 0.7896, and 0.825, respectively), suggesting that 8-oxo-dGuo was the mutagenic lesion responsible for the G to T transversions observed in PAH o-quinone-treated p53. On the other hand, the mutagenic effects of (±)-anti-BPDE on p53 were not related to 8-oxo-dGuo formation but correlated with the formation of (+)-anti-BPDE-N2-dGuo adducts (Figure 4, R2) 0.8411). This suggests that the mutagenic lesion caused by PAH o-quinones is 8-oxo-dGuo, whereas the mutagenic lesion caused by (±)-anti-BPDE is a stable bulky adduct. The slopes of each of these plots are quite comparable, suggesting that the mutagenic potential of 8-oxo-dGuo and (+)-anti-BPDE-N2-dGuo were similar.
Because many tumorderived mutations, especially hotspot codons, are dominant negatives in transcriptional assays (33, 35, 36), we hypothesized that the spectrum of mutants from a randomly generated collection would more closely match the spectrum in tumors if only the dominant mutants were screened. To determine the spectrum of dominant mutations, we sorted p53 mutations generated by o-quinones and (±)-anti-BPDE, previously reported and sequenced, into dominant and recessive mutants, and then plotted the spectra. The mutations generated by BA-3,4-dione and DMBA-3,4-dione were combined because they produce a common mutagen ROS. Not all of the mutations observed with BP-7,8-dione could be analyzed because of technical difficulties, but these showed the same trends observed with the other o-quinones. We also tested a set generated by (±)-anti-BPDE with methylated p53 cDNA, because adducts have been reported to form more specifically at known hotspots in the methylated DNA (25). To test dominance, we coexpressed wild-type p53 from a TRP1-selectable plasmid, along with individual mutants (originally isolated from a LEU2 selectable plasmid), and tested colonies on leu/trp plates. Recessive mutants become white, whereas dominant mutants become pink. In nearly all cases, the dominant mutants were not as red as the original colony, and a range of color intensity was observed as listed in Table 1. We found that 27 out of 129 plasmids with single point mutations were dominant (Table 1). The pattern of mutations in the dominant mutants was about the same as that when all mutants were scored and were dominated by G > T and C > A transversions in PAH o-quinone treated samples and G > C transversions in (±)-anti-BPDE treated samples. The most commonly mutated amino acid in the dominant mutants was arginine with 12 mutants (44%), which is also the most commonly mutated amino acid in the database (31%). (To reduce confusion over the mutations causing dominance, we plotted only those plasmids with single point mutations, although plasmids with multiple mutations are included in Table 1).
We defined hotspots as the codons that together account for 50% of the mutations in lung cancer based on the IARC database release R11 (this list differs from release R10 by the addition of codon 234 and is listed by the order of frequency in the legend to Table 2). Because the hotspots account for 24 of the 213 amino acids sequenced, a random distribution is 11.2%. Under all mutagenesis paradigms, the dominant mutations were enriched in hotspots. As shown in columns 4 and 2 of Table 2, 33% (9 of 27, P = 0.0002) of the dominant mutants had single point mutations at hotspots mutated in lung cancer. If instead we compared the dominant mutants against nonlung cancers, the incidence of hotspots was 44% (12 of 27) because hotspots (e.g., codon 213 is a nonlung cancer hotspot) vary slightly in the different tumors (Table 2). When analyzed by the PAH-treatment paradigm, 6 of the 16 dominant mutants (38%, P = 0.001) from the PAH o-quinone treated samples were at hotspots. A similar trend was seen with (±)-anti-BPDE treated samples with 3 of 11 of the dominant mutants (27%) were at hotspots, but the numbers of mutants were too low to be significant (P = 0.062). In our previous screens, we found that lung cancer hotspots are mutated about 20% of the time with o-quinones and 10% of the time with (±)-anti-BPDE, demonstrating that selection for dominance used in this study increases the percentage of mutations in hotspots (16, 17). As we define hotspots as the mutations that account for >50% of the mutations in tumors, our incidence of occurrence in the dominant mutants approaches that seen in the database. We believe that our data are underestimates because two plasmids with mutations in hotspot codon 248 were not included due to a second mutation. Additionally, the dominant mutations that were not located at hotspots were usually one or two residues away from a major hotspot, such as 244 and 279, which are near 245 and 280, respectively, suggesting they may interfere with DNA binding. Mutational spectra show that the majority of the dominant mutations, 88% (14 of the 16 isolated after PAH-o-quinone treatment), clustered in domains IV and V (Figure 5). The PAH o-quinone derived dominant mutations cluster near the protein-DNA interface and bear a striking resemblance to the cluster formed by the top 10 hotspots in lung cancer (parts A and B of Figure 6). The (±)-anti-BPDE derived dominant mutants did not cluster as strongly at the protein-DNA interface but did cluster in other hotspot regions, domain III, and codons 156–158, which lie between domains II and III. Thus, by combining mutagenesis with dominance selection, we found that the mutations clustered in all of the major hotspot regions.
In contrast to the clustering seen in the dominant mutations, only 15% of the recessive mutants (15 of 102) mapped to hotspots and were scattered throughout the DNA binding domain (Figure 5). Many mapped to the interdomain region between domains IV and V, which is rarely mutated in cancers and where no dominant mutations were found. There was a noticeable absence of recessive mutations in the DNA–protein interface (Figure 6). Several mutants were isolated multiple times, in which one plasmid was dominant while the other was recessive. We note that the plasmids were obtained through a random mutagenesis and recombination paradigm, so they may contain additional mutations that could influence the dominance as well as the stability of the expressed p53. However, data suggest that while mutagenesis mechanisms determine the pattern of mutations, the spectrum of mutations is predominantly driven by biological selection.
This study provides additional support for p53-mutagenesis by PAH o-quinones mediated by reactive oxygen species and that aldo-keto reductases involved in their formation may contribute to lung-cancer initiation. We find a linear correlation between the generation of 8-oxo-2′-deoxyguanosine and p53-mutagenesis. Whereas ROS can oxidize bases other than guanine, and other oxidation products of guanine are sometimes observed, 8-oxo-2′-deoxyguanosine has been studied most extensively because it is highly mutagenic and most prone to cause the G to T transversions characteristic of lung cancer (37, 38). 8-Oxo-2′-deoxyguanosine is also the most common lesion in PAH o-quinone treated DNA (19). As expected, we also observed a linear correlation between (+)-anti-BPDE-N2-dGuo and mutagenesis in (±)-anti-BPDE treated DNA but were unable to detect any 8-oxo-2′-deoxyguanosine. This correlation strongly implicates 8-oxo-2′-deoxyguanosine as the DNA adduct responsible for quinone induced p53 mutagenesis. Additionally, we show for the first time in a mutagenesis assay, a close correlation of the tumor-derived spectrum by dominant but not by recessive or unselected mutants, suggesting that selection drives the tumor-derived spectrum.
Our p53 mutagenesis assay allowed us to compare the potency of the different adducts by correlating levels of adducts with mutagenic frequency. We found that 8-oxo-dGuo and (+)-anti-BPDE-N2-dGuo adducts gave a similar mutagenic frequency based on adduct number. Thus, in our assays these two lesions are equipotent in producing mutations. However, the level of 8-oxo-dGuo adducts were achieved with >20-fold less PAH metabolite than required to produce the (+)-anti-BPDE-N2-dGuo lesions. We are unaware of a direct comparison of the mutagenic frequencies of these lesions previously.
The pattern of mutations observed in most studies with (±)-anti-BPDE are usually dominated by G > T transversions, which occur 70–80% of the time with most of the remainder being A > C transversions (26). One study found that (±)-anti-BPDE also caused G > T transversions in the same yeast assay that we use (21). However, other studies found a preference for G > C transversions in a yeast mutagenesis assay with (±)-anti-BPDE (39, 40). The preference for G to C over G to T transversions in yeast may be governed by the trans-lesional bypass polymerase Polζ, which has been shown to preferentially incorporate G opposite an (+)-anti-BPDE-N2-dGuo adduct in yeast strains proficient in mutagenesis (39, 40). Thus, while it was somewhat unexpected to observe fewer G > T mutations by (±)-anti-BPDE, we speculate that the repertoire of translesion bypass polymerases that predominate in yeast differed from the studies of Yoon et al. Our observation that about 90% of the mutations seen with (±)-anti-BPDE were at GC base pairs is consistent with the formation of the (+)-anti-BPDEN2-dGuo when we measured adduct formation.
The spectrum of p53 mutations in lung and other cancers are characterized by a number of hotspots within the DNA-binding domain. Because several smoke-derived mutagens including (±)-anti-BPDE and acrolein (41) form adducts preferentially at guanine bases located in these hotspots, it has been argued that they preferentially target these sequences (23). Many of these sequences contain 5′CpG-3′ islands and methylation of the 5′-cytosine may enhance hydrophobic interaction and reactivity (24, 25). However, studies on targeting these sequences in p53 did not measure adduct formation directly but instead located sites of adduction by ligation-mediated PCR.
In ligation-mediated PCR, the UvrABC endonuclease of the E. coli nucleotide excision repair pathway is used to cleave DNA on either side of a bulky DNA-adduct. After cleavage, a ligation-assisted PCR assay is used to amplify the DNA, and the products are run on sequencing gels to determine the site of adduction. If cleavage rates of adducted sequences differ based on sequence context, the assay may show codon bias. Alternatively, nonlinear amplification by PCR may amplify small differences in adduct formation at different codons. New direct methods to quantify (+)-anti-BPDE-N2-dGuo adducts at specific codons in p53 have been developed using ds-oligonucleotides encoding exons 5, 7, and 8 site-specifically labeled with [15N5]-dGuo coupled with MS. Although site-specific modification was observed at codons 156, 157, and 158 treated with (±)-anti-BPDE, the (+)-anti-BPDE-N2-dGuo adduct was predominately seen at codon 156, which is not a hot spot mutated in lung cancer (42, 43). Whereas these studies also found preference for adducts at codons 245, 248, and 273, more recent studies found that ROS could preferentially target codons 245 and 248, depending on the oxidant (44).
To date, (±)-anti-BPDE (23), acrolein (41), ROS (44), and even Chromium (45) have been shown to react preferentially with the same codons. Because the same codons react with so many mutagens, studies showing that a reactive intermediate forms adducts at specific codons do not conclusively incriminate that mutagen. We also find that when the mutational spectrum of the gene is considered rather than the adduct spectrum, little, if any, specificity is observed in the spectra in the absence of biological selection.
As an alternative to targeted mutagenesis, Rodin and Rodin have argued that genetic selection determines the p53-mutational spectrum in lung cancer (27, 28, 46). Several yeast mutational studies have generated patterns of mutations in p53 dominated by lesions at guanines, but the mutations selected in yeast are scattered throughout the DNA binding region, showing no resemblance to the spectrum observed in tumors (16, 17, 21). Yet, when sorted into dominant and recessive mutations, we find a strong correlation between the tumor-derived spectrum and the dominant mutant spectrum. The similarity is perhaps most striking when considered by domain. We find that domains IV and V in particular stand out with clusters of dominant mutants in these two hotspot regions, whereas the interdomain region is a cold spot. These domains correspond to the surfaces of p53 that contact DNA where most of the mutations in tumors are found (Figure 6). Our data strongly suggests that tumor-derived mutational spectra are predominantly driven by biological selection for the strongest, for example, dominant mutants. We note some exceptions to this conclusion.
There are several observations that do not fit the simple conclusion that genetic dominance completely drives the spectrum. We, and others, have found that codon 279 is a strongly dominant mutant, and with its presence in four independent isolates, it was one of our most commonly isolated dominant mutations (33, 35). Yet in tumors, codon 279 is a cold spot and is rarely mutated. However, codon 279 is located in the DNA binding helix of domain V between the DNA contact residues 278 (a hotspot in all cancers but not lung cancer) and 280 (which is also a hotspot). We speculate that codon 279 mutations would provide a strong growth advantage to cells.
Of the dominant mutations isolated in this study, 19 have been previously reported as dominant, 4 have been reported as recessive (codons 213, 256, 178, 196), while no data are available on three (codons 156, 251, and 257) (3). Of the previously reported recessive mutants, we found that 256, 178, and 196 were weakly dominant, which may explain why they escaped detection. Codon 213, which was mutated in 3 plasmids that were strongly dominant, falls just outside of our definition of a lung cancer hotspot, ranking 25th, with 19 entries in the database. However, codon 213 is a hotspot in nonlung cancers and ranked 9th, with 320 entries. We found that half of the mutations in codon 213, even dominant plasmids, contained nonsense mutations that would truncate the protein. The dominance of some mutations in codon 213 may explain its relatively high frequency in the database.
By plotting the pooled recessive mutations, we also found a category of mutation, which we refer to as frequently isolated recessive mutations. We isolated multiple plasmids with mutations at codons 156, 196, 283, and 213 (213 was frequently found in both dominant and recessive plasmids; see above), qualifying them as recessive hotspots. All were also scored at least once as a dominant mutant. Interestingly, all four codons encode Arg residues and are located within CpG islands, which may be preferred sites of adduct addition. We speculate that these mutations cause strong loss-of-function mutants in p53 and are readily isolated in yeast screens.
While generally regarded as a tumor suppressor, there are numerous properties of p53 that suggest that dominance plays a central role in tumorigenesis. The spectrum of p53 stands out from the spectra of other tumor suppressors, such as NF1, NF2, and PTEN. The spectra of mutations in these tumor suppressors are characterized by clear loss-of-function mutations, whereas the p53 spectrum is characterized by missense mutations. Mutant p53 proteins are usually long-lived stable proteins, unlike wild-type p53, suggesting that the mutant proteins contribute to cell transformation. Some mutants even transform cells in cell transfection experiments. The case for dominance is also supported by numerous functional studies on mutants, showing that they are dominant in transcription assays, acting by forming nonfunctional tetramers with the wild-type p53. If dominance provides a growth advantage to the emerging tumor cell that is not provided by a recessive mutation, then the mutational spectrum would be enriched by dominant mutations. In fact, in some tumors dominant p53 mutations correlate with early disease onset, but the relationship is not well established (47). Whereas we have focused our studies on selection for dominance, we note that there are other properties attributed to mutant p53 proteins that may contribute to tumorigenesis, including cell transformation, NFkappaB activation, and stimulation of genetic instability (48, 49). These may account for some of the differences between our spectra of dominant mutations and the p53 spectra seen in tumors.
Smokers have reduced levels of antioxidants, and excrete 8-oxo-2′-deoxyguanosine in there urine, suggesting that they are under oxidative stress. However, the measurements of 8-oxo-dGuo are often suspect due to problems of adventitious oxidation of guanine bases (50). Moreover, the detection of 8-oxo-dGuo in the urine suggests that this is repaired by nucleotide excision repair when base-excision repair normally repairs this lesion. Tobacco smoke can generate ROS through a number of compounds, including semiquinones (non-AKR derived), NO, and ozone (6, 29, 51). More recently, AKRs have emerged as an alternative pathway of tobacco-generated ROS. The ability of AKRs to form PAH o-quinones with the concomitant production of ROS is likely to occur in tobacco-related cancer because AKRs are consistently overexpressed in human lung adenocarcinoma (A549) cells, NSCLC, and SCLC, bronchial epithelial cells derived from NSCLC, or oral cancer cells (52-58). In addition, one allele of hOGG1 (8-oxo-guanine glycosylase, the base-excision repair enzyme responsible for the removal of 8-oxo-guanine) is absent in 50% of NSCLC patients suggesting that the reduced rates of repair of oxidatively damaged DNA may increase susceptibility to oxidative DNA damage (59, 60).
In conclusion, we provide additional evidence for the AKR pathway of PAH o-quinone mutagenesis. We have implicated 8-oxo-dGuo as the most likely base lesion that causes the G to T transversions that predominate in p53 mutational patterns. The spectrum of mutations is random under low-stringency selection, suggesting that adducts formed preferentially at specific codons are diluted out by nonbiased adducts when they proceed from a DNA lesion to a mutation. Instead, genetic selection, which we approximate by testing for dominance, is the predominant determinant of the spectrum.
We are left with one additional characteristic of the p53 mutations observed in lung cancer that requires explanation, that is the strand bias for mutations on the nontranscribed strand. This could be explained by the loss of hOGG1 in lung cancer, which preferentially repairs the nontranscribed strand (61, 62). The strand bias can be examined in future studies involving the deletion of the yeast homologue of hOGG1 in our p53 yeast reporter gene assay.
We thank Dr. Gilberto Fronza for the gift of plasmid pTS76. This work was supported by grants R01 CA39504 and P01 CA92537 awarded to T.M.P. and by grants R01 GM48241 and R01 ES015662, and pilot project support from 1P30 ES013508–01 to J.F. The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the NIEHS, NIH. We thank the NCI Chemical Carcinogen Standard Reference Repository for (±)-anti-BPDE.
1Abbreviations: (±)-anti-BPDE, anti-benzo[a]pyrene diol epoxide; BPQ, benzo[a]pyrene-7,8-dione; BAQ, benz[a]anthracene-3,4-dione; DMBAQ, dimethylbenz[a]anthracene-3,4-dione; PAH, polycyclic aromatic hydrocarbons.