Trans -4-hydroxy-2-nonenal (4-HNE), a major electrophilic by-product of lipid peroxidation, is able to interact with DNA to form exocyclic guanine adducts. 4-HNE is a mutagen and a significant amount of 4-HNE–guanine adduct has been detected in normal cells. Recently, it has been reported that exposure of the wild-type p53 human lymphoblastoid cell line to 4-HNE causes a high frequency of G to T transversion mutations at the third base of codon 249 (-AGG*-) in the p53 gene, a mutational hotspot in human cancers, particularly hepatocellular carcinoma. These findings raise a possibility that 4-HNE could be an important etiological agent for human cancers that have a mutation at codon 249 of the p53 gene. However, to date, the sequence specificity of 4-HNE–DNA binding remains unclear due to the lack of methodology. To address this question, we have developed a method, using UvrABC nuclease, a nucleotide excision repair enzyme complex isolated from Escherichia coli , to map the distribution of 4-HNE–DNA adducts in human p53 gene at the nucleotide sequence level. We found that 4-HNE–DNA adducts are preferentially formed at the third base of codon 249 in the p53 gene. The preferential binding of 4-HNE was also observed at codon 174, which has the same sequence and the same nearest neighbor sequences (-G AGG* C-) as codon 249. These results suggest that 4-HNE may be an important etiological agent for human cancers that have a mutation at codon 249 of the p53 gene.
Trans -4-hydroxy-2-nonenal (4-HNE) (Figure 1 ) is a major electrophilic product of lipid peroxidation caused by oxidative stress, which is formed by radical-initiated degradation of ω-6-polyunsaturated fatty acids such as linoleic and arachidonic acids, two relatively abundant fatty acids in human cells ( 1 ). Although it has been proposed that 4-HNE can be further metabolized to an epoxide form that can interact with DNA to form exocyclic etheno-guanine, -adenine and -cytosine adducts ( 2 ), a significant amount of 6-(1-hydroxyhexanyl)-8-hydroxy 1, N2 -propano-2′-deoxyguansine (4-HNE-dG) adduct (Figure 1 ), a bulky exocyclic DNA adduct, has been found in various tissues of human and rat ( 3 – 7 ). It also has been found that the amount of 4-HNE-dG adducts significantly increases in the liver of rats treated with carbon tetrachloride, a condition that can induce lipid peroxidation and liver carcinogenesis in rat ( 6 , 8 ). These results emphasize the potentially important role that 4-HNE and 4-HNE-dG may play in tumor initiation and promotion.
Recently, it has been reported that exposure of the wild-type p53 human lymphoblastoid cell line to 4-HNE causes a high frequency of G to T transversion mutations at the third base of codon 249 (-AGG-) of the p53 gene, a mutational hotspot in many human cancers particularly in hepatocellular carcinoma ( 9 ). Human p53 gene is the most frequently mutated gene in human cancers; >50% of human cancers have mutations in the p53 gene, which are distributed along >200 codons. It has been found that in human cancers most of the mutations in the p53 gene are located in the DNA binding domain of the p53 protein ( 10 – 13 ). Interestingly, >30% of the p53 gene mutations are at methylated CpG sites. Most of the mutational hotspots in the p53 gene in human cancers are at methylated CpG sites (such as codons 157,158, 175, 245, 248, 273 and 282), although there are few exceptions, such as codon 249 in liver and lung cancers, and codons 280 and 285 in bladder cancers ( 10 – 13 ). Previously, we have found that various bulky chemical carcinogens, including the epoxide forms of polycyclic aromatic hydrocarbons (PAHs) [such as benzo[ a ]pyrene diol epoxide (BPDE), benzo[ g ]- chrysene diol epoxide (BCDE)], N -acetoxy-2-acetylaminofluorene (NAAAF), aflatoxin B1 8,9-diol epoxide (AFB1-DE), 4-aminobiphenyl, preferentially form DNA adducts at the aforementioned CpG sites corresponding to the mutational hotspots in human cancers, especially in cigarette smoking-related lung cancer ( 14 – 18 ). Furthermore, adducts formed at these CpG sites are poorly repaired ( 19 ). Our results strongly suggest that targeted DNA damage in addition to growth selection may contribute greatly to the mutational spectrum in the p53 gene in human cancers, particularly for cigarette smoking-related lung cancer. However, we have found that none of the epoxide forms of PAHs in cigarette smoke, AFB1-DE, NAAAF or 4-aminobiphenyl binds strongly at codon 249 of the p53 gene ( 14 – 18 ). Combined together, these findings raise an important question of whether 4-HNE is an important etiological agent for human cancers, which have a mutation at codon 249 of the p53 gene. In light of the recent finding that, in liver cells, the p53 gene containing a mutation at codon 249 has a dominant-negative effect ( 20 – 22 ), it is possible that both targeted DNA damage at codon 249 and the strong dominant-negative effect of this 4-HNE-induced mutation at codon 249 may contribute to shape the mutational spectrum in human cancers, particularly in hepatocellular carcinoma, in which codon 249 is a mutational hotspot. However, to date, the sequence specificity of 4-HNE–DNA binding remains unclear, thus, a method for mapping the distribution of 4-HNE–DNA adducts in the p53 gene at the nucleotide sequence level is needed to address this question.
It has been well established that UvrABC nuclease, the nucleotide excision repair enzyme complex isolated from Escherichia coli , is able to specifically and quantitatively incise a variety of bulky carcinogen-induced DNA damage. This UvrABC nuclease incision method has been successfully applied to determine the sequence specificity of bulky chemical carcinogen-induced DNA damage and their repair ( 14 , 23 – 29 ). In this report, we have characterized the mode of UvrABC nuclease incision toward DNA modified with 4-HNE. We found that UvrABC nuclease is able to incise 4-HNE–DNA adducts quantitatively and specifically. Using this method, we have mapped the 4-HNE–DNA binding spectrum in exons 5, 7 and 8 of human p53 gene, and have found that codon 249, a mutational hotspot in human cancers, is a preferential site for 4-HNE binding. We have also investigated the effect of cytosine methylation at CpG sites on 4-HNE–DNA binding spectrum in the p53 gene.
Materials and methods
Preparation of plasmid DNA and DNA fragment
Plasmid pGEM inserted with the adenine phosphoribosyltransferase ( APRT ) gene from Chinese hamster ovary cells (pGEM- APRT ) ( 30 ) was purified via cesium chloride density centrifugation and dialyzed extensively against TE buffer [10 mM Tris and 1 mM EDTA (pH 8.0)]. The 32 P-end-labeled fragments of human p53 gene exons 5, 7 and 8 were prepared as described previously ( 31 ). Briefly, the 247 bp 5′-end-labeled exon 5 DNA fragment, the 187 bp 5′-end-labeled exon 7 DNA fragment and the 210 bp 5′-end-labeled exon 8 DNA fragment were amplified from the p53 gene-containing plasmid (pAT153p53p; obtained from L.Crawford and S.P.Tuck, Imperial Cancer Fund Laboratories, London, UK). The following primers (Midland Certified Reagent Company, Midland, TX) were used for amplification of each exon: (a) for exon 5, 5′-TGCCCTGACTTTCAACTCTGTCTCC-3′ and 5′-TCTCTCCA- GCCCCAGCTGCTCAC-3′, with the former being 5′ 32 P-end-labeled with [γ- 32 P]ATP (NEN, Boston, MA); (b) for exon 7, 5′-GCACTGGCC- TCATCTTGGGCCTG-3′, and 5′-CACAGCAGGCCAGTGTGCAGGGT-3′, with the former labeled at the 5′ end; and (c) for exon 8, 5′-ACT- GCCTCTTGCTTCTCTTTTCCTATCC-3′, 5′-CTTGGTCTCCTCCACCGC- TTCTTG-3′, with the former labeled at the 5′ end. The 5′- 32 P-single-end-labeled DNA fragments were purified through 8% non-denaturing polyacrylamide gels.
The 141 bp 3′- 32 P-end-labeled fragment of the p53 gene exon 7 was isolated from the above 187 bp amplified exon 7 DNA fragment with unlabeled primers. The 187 bp fragment was digested with Hif I restriction enzyme (Promega, Madison, WI), and the 141 bp product was isolated from a 2% agarose gel. The fragment was then 3′-end-labeled with [α- 32 P]dATP (NEN) ( 31 ). The resulting 141 bp singly 3′-end-labeled fragment was purified through an 8% non-denaturing polyacrylamide gel.
5-C cytosine methylation at CpG sites
The 5′- 32 P-end-labeled DNA fragments were subjected to Sss I methylase (New England Biolabs, Beverly, MA) treatment in the presence of S -adenosylmethionine (SAM), to methylate all cytosines at CpG sites, according to manufacturer's instructions.
4-HNE modifications of supercoiled plasmid DNA and DNA fragments
4-HNE was synthesized according to the method described previously ( 32 ). Stock solution was prepared by dissolving 4-HNE in methanol at a concentration of 100 mg/ml. DNA fragments and purified supercoiled pGEM- APRT plasmid DNA were dissolved in 70 ml TE buffer (pH 7.0), mixed with 30 ml 4-HNE solution of various concentrations, and incubated at 37°C for 20 h. The unreacted 4-HNE was removed by repeated phenol and diethyl ether extractions, and the DNA was then ethanol precipitated, washed with 75% ethanol and dried under vacuum.
Detection and quantification of 4-HNE-dG adducts in DNA fragments by 32P-post-labeling/HPLC method
Method for 4-HNE-dG adduct analysis was similar as described previously with a few modifications ( 33 ). 4-HNE-dG 3′-monophosphate and 3′,5′-biphosphate were synthesized according to the method described previously ( 6 ). The DNA fragment of the p53 gene exon 7 (50–150 mg) modified with or without 4-HNE was digested with micrococcal nuclease (Sigma, Louis, MO) and spleen phosphodiesterase (Worthington, Lakewood, NJ), applied onto a C18 solid-phase extraction cartridge and the unmodified nucleotides were removed by washing the cartridge with 50 mM ammonium formate (pH 7.0). The 4-HNE-dG 3′-monophosphate was eluted with 50% methanol and labeled with [γ- 32 P]ATP. The labeled mixture was further purified by the aforementioned procedure and reconstituted with 100 μl H 2 O. The synthetic 4-HNE-dG 3′,5′-bisphosphates were added as UV standards. The 32 P-labeled 4-HNE-dG 3′,5′-bisphosphates were purified on a C18 reversed-phase HPLC with a solvent system of 50 mM NaH 2 PO 4 (pH 5.8) as buffer A, 50% CH 3 OH in H 2 O as buffer B, and a solvent gradient from 0 to 75% buffer B in 75 min at a flow rate of 1 ml/min. The quantification and identity confirmation was accomplished by the ring-opening/reduction reaction, characteristic property of enal-derived cyclic propano-dG adducts ( 6 ). Briefly, the 4-HNE-dG adduct fractions collected from the HPLC were treated with 30 μl of 10 N NaOH and an excess of NaBH 4 for 10 min at room temperature to convert 4-HNE-dG 3′,5′-bisphosphates adduct to corresponding ring-opened 4-HNE-dG adduct. The reaction mixture was then neutralized by H 3 PO 4 and analyzed by the same reversed-phase HPLC system with a dual detection of UV and β-Ram radioflow detector (IN/US System, Tampa, FL).
Purification of UvrA, UvrB and UvrC proteins
The UvrA, UvrB and UvrC proteins were isolated from the E.coli K12 strain CH296 carrying plasmids pUNC45 ( uvrA ), pUNC21( uvrB ) or pDR3274( uvrC ) ( 34 ). These plasmids and the CH296 strain were kindly provided by Dr A.Sancar, University of North Carolina, Chapel Hill, NC. The purification procedures were the same as described previously ( 29 , 35 ).
UvrABC incision assay of 4-HNE–DNA adducts
Standard UvrABC incision assays were carried out in 25 μl solution containing 100 mM KCl, 1 mM ATP, 10 mM MgCl 2 , 10 mM Tris (pH 7.5) and 1 mM EDTA. UvrABC nuclease was added to 4-HNE-modified DNA at a molar ratio of 6:1. The reactions were carried out at 37°C for various times. For pGEM- APRT supercoiled plasmid DNA, the reactions were stopped after 60 min by adding 0.1% sodium dodecyl sulfate and heating at 65°C for 5 min, and the plasmid DNA was then electrophoresed in an 1% agarose gel in TAE buffer [40mM Tris–acetate (pH8.0), 1mM EDTA] at 1V/cm overnight, stained with ethidium bromide (0.5 μg/ml) and visualized with UV light. For 32 P-labeled DNA fragments, the reactions were stopped by phenol and ether extractions followed by ethanol precipitation of the DNA.
DNA sequencing, gel electrophoresis and autoradiography
Singly 5′- or 3′- 32 P-end-labeled DNA fragments were sequenced by the Maxam–Gilbert chemical cleavage method ( 36 ). The 32 P-labeled DNA fragments, with or without UvrABC enzyme treatment, were suspended in sequencing tracking dye (80% v/v deionized formamide, 0.1% xylene cyanol and 0.1% bromophenol blue), heated at 90°C for 5 min, and separated by electrophoresis in 8% denaturing polyacrylamide gels. The gels were dried, and initially exposed to a phosphor screen and then to Kodak films. The intensity of bands was determined by a PhosphorImager (Cyclone, Packard, Meriden, CT).
4-HNE-dG is the major adduct formed in DNA modified with 4-HNE
It has been well established that modification of DNA with 4-HNE in phosphate buffer produces four major 4-HNE-dG adducts: 1, 2, 3 and 4 (Figure 1 ); these four isomers can be separated by HPLC. Since the reaction of 4-HNE with DNA is a slow process, a significant amount of depurination is expected to occur during long reaction time (>24 h) under this condition. We, therefore, modified DNA fragment of the p53 gene exon 7 with 4-HNE in TE buffer at neutral pH for 20 h to reduce DNA depurination and degradation. The DNA adducts formed under this condition were then analyzed by a 32 P post-labeling/HPLC method. The results in Figure 2 show that the four isomeric 4-HNE-dG adducts are formed in the DNA fragment modified with 4-HNE, and indistinguishable from the synthetic 4-HNE-dG standards.
Supercoiled plasmid DNA modified with 4-HNE is sensitive to UvrABC nuclease incision
It has been reported that the nucleotide excision repair pathway in mammalian cells is involved in the repair of 1, N2 -propano-deoxyguanosine adducts ( 37 ). Thus, it is reasonable to expect that UvrABC nuclease, a nucleotide excision repair enzyme complex isolated from E.coli , may be able to incise 4-HNE–DNA adducts. To test this possibility, we modified pGEM- APRT supercoiled plasmid DNA with different concentrations of 4-HNE and subsequently treated the modified DNA with UvrABC nuclease. Results in Figure 3 show that while UvrABC does not cut unmodified plasmid DNA, it does cut 4-HNE-modified supercoiled plamsid DNA and turns supercoiled plasmid DNA from covalently closed circle (CCC) to open circle (OC). Because of the lack of radioactively labeled 4-HNE, we were unable to calculate the amount of 4-HNE–DNA adducts cut by UvrABC nuclease. Nonetheless, we have found that the extent of cutting is proportional to the concentrations of 4-HNE used for modification, which indicates that UvrABC nuclease is able to incise 4-HNE–DNA adducts quantitatively.
UvrABC nuclease incises 7 nucleotides 5′ to and 3–4 nucleotides 3′ to a 4-HNE–DNA adduct
The 4-HNE-modification-induced UvrABC nuclease incision of supercoiled plasmid DNA is most likely due to UvrABC nuclease recognizing and incising of 4-HNE–DNA adducts. However, it could also be due to the UvrABC nuclease incising a secondary structure induced by 4-HNE-dG adducts in supercoiled plasmid DNA, which may not be located in the immediate vicinity of the modified bases. To distinguish between these two possibilities, we determined the UvrABC incision sites on a defined DNA fragment modified with 4-HNE. The hallmark of damaged DNA recognition and incision by UvrABC nuclease is its dual incisions; the enzyme complex incises six to eight bases 5′ and three to five bases 3′ to the damaged base(s) ( 23 – 27 ). The DNA fragment of the p53 gene exon 7 was labeled with 32 P at a single 5′ or 3′ end, and treated with 4-HNE at concentration of 30 mg/ml, which renders the maximal UvrABC incision. These DNA fragments were then reacted with UvrABC nuclease. Figure 4 shows the electrophoretic separation of the resultant DNAs in a denaturing polyacrylamide gel. In the 5′- 32 P-end-labeled DNA fragment of the p53 gene exon 7, 4-HNE modification induced one strong and many weak UvrABC nuclease incision bands in the exon region; all of these bands were generated by UvrABC incision 7 nucleotides 5′ to a 4-HNE–DNA adduct (Figure 4A ). UvrABC incision also generated one strong and many weak bands in 4-HNE-modified 3′-end-labeled DNA fragment of the p53 gene exon 7. Again, all these incision bands can be attributed to UvrABC incision 3–4 nucleotides 3′ to a 4-HNE–DNA adduct (Figure 4B ). These results suggest that the UvrABC nuclease makes dual incisions 7 bases 5′ and 3–4 bases 3′ to 4-HNE–DNA adducts. These results are consistent with published results showing that purified UvrABC typically makes dual incisions 6–8 bases 5′ and 3–5 bases 3′ to a damaged base ( 23 – 27 ). Results in Figure 4 also show that 4-HNE–DNA adduct formation in exon 7 of the p53 gene appears to have significant sequence selectivity: 4-HNE–DNA adducts are preferentially formed at codon 249 (-G AGG* C-) (the codon sequence is underlined and the adducted guanine base is marked with a * symbol), and a strong 4-HNE binding site labeled with X(-CAGG*A-) is also observed in the intron 7 region (Figure 4A ).
Incision of 4-HNE–DNA adduct by UvrABC nuclease is sequence independent
It is possible that the distinct difference of 4-HNE–DNA adduct incision by UvrABC at codon 249 could be due to UvrABC nuclease preferentially incising 4-HNE–DNA adduct formed at codon 249 compared with adducts formed at other sequences. If this is the case, then the rate constant of UvrABC incision at codon 249 should be higher than that at other sequences; i.e. if the UvrABC nuclease incision has particular sequence preferences, then the rate constant of incision at different sequences should be different. On the other hand, if UvrABC incision has no sequence preference, then the rate constant of incision at different sequences should be similar and, therefore, the extent of incision at different sequences should reflect the amount of adduct formation at these sequences. Results in Figure 5 show that there are no significant differences in the rate constant of UvrABC incision at codon 249 in comparison with other codons in exons 7 and 8 of the p53 gene. Since under our reaction conditions the molar ratio of UvrABC/DNA is 6, and UvrABC nuclease remains active at the end of incubation (data not shown); furthermore, the UvrABC incision is an irreversible reaction, these results strongly suggest that DNA sequence does not play a significant role in determining the efficiency of UvrABC incision and that the degree of UvrABC incision is proportional to the extent of 4-HNE–DNA adduct formation. The results indicate that under our standard reaction conditions the intensity of UvrABC incision at different sequences should represent the sequence preference of 4-HNE–DNA adduct formation. Therefore, the UvrABC incision method can be used to determine the sequence selectivity of 4-HNE–DNA binding. Results in Figure 6 show that the relative intensity of UvrABC cutting at the 5′-side is comparable with the intensity of the 3′-side cutting, which further reflects the dual incisions of UvrABC nuclease on 4-HNE–DNA adducts. Therefore, either 5′- or 3′-end-labeled fragments can be used for determining the 4-HNE binding spectrum.
Binding spectrum of 4-HNE in exons 5, 7 and 8 of human p53 gene
We determined 4-HNE binding in exons 5, 7 and 8 of the p53 gene, because >80% of p53 mutations and most of the p53 mutational hotspots in human cancer are located in this region ( 10 – 13 ). DNA fragments of human p53 gene exons 5 and 8 amplified from the p53- containing plasmid were 5′-end-labeled, modified with 4-HNE, reacted with UvrABC nuclease, and the resultant DNA products were gel separated (Figure 7A ). Figures 6 and 7B show the relative intensity of 4-HNE–DNA adduct formation in exons 5, 7 and 8 of the p53 gene. Results demonstrate that 4-HNE–DNA adduct formation in the p53 gene has significant sequence selectivity. The 4-HNE–DNA adducts are preferentially formed at codon 249 (-G AGG* C-) of exon 7, and at site X(-CAGG*A-) in the intron 7 region (Figures 4 and 6 ). Codon 174 in exon 5 is also a preferential 4-HNE binding site. Both codon 174 and codon 249 have the same sequence and the same nearest neighbor sequences (-G AGG* C-). Although 4-HNE binding distribution in exon 8 is quite even, binding at codon 286, which contains a similar sequence (-GAG G*AA -) as codon 249, is the strongest.
The effect of C5 cytosine methylation at CpG sites on 4-HNE–DNA binding
Most of the mutational hotspots in the p53 gene in human cancers are at CpG sites (such as codons 157, 158, 175, 245, 248, 273 and 282) although there are few exceptions, such as codon 249 in liver and lung cancers, and codons 280 and 285 in bladder cancer ( 10 – 13 ). We have found that methylation of cytosine greatly enhances carcinogen–DNA binding at these CpG sites in the p53 gene, indicating that cytosine methylation is the major cause for preferential carcinogen–adduct formation at CpG sites in human p53 gene ( 14 – 18 , 31 ). We have also found that C5 cytosine methylation at CpG sites may affect adduct formation at surrounding sequences and the effect is dependent on the nature of the carcinogens and the sequences surrounding the CpG sites ( 17 , 38 ). It has been shown that all cytosine residues at CpG sites in the coding region of human p53 gene are highly methylated in a variety of tissues ( 39 ). These findings strongly suggest that cytosine methylation may greatly affect the chemical– DNA binding spectrum. Since the surrounding sequences of two 4-HNE strong binding sites in the coding sequence of the p53 gene—codon 249 (-CGG AGG -) and codon 174 (- AGG CGC-)—contain a CpG site, it is possible that C5 cytosine methylation may affect 4-HNE binding at codons 249 and 174 in the p53 gene. We therefore determine the effect of C5 cytosine methylation on 4-HNE–DNA binding in the p53 gene. The 5′- 32 P-end-labeled DNA fragments were subjected to Sss I methylase treatment in the presence of SAM to methylate all cytosines at CpG sites. DNA fragments with or without methylation treatment were then modified with 4-HNE and the adduct distribution at various sequence positions were mapped by the UvrABC incision method. The extent of cytosine methylation was determined by the Maxam–Gilbert pyrimidine reaction ( 36 ). Hydrazine is unable to modify 5-C-methylated cytosines; consequently, no cytosine ladders are observed at methylated cytosines ( 40 ). Figure 8A shows that under our methylation conditions, all of the CpG sites in the DNA fragments of the p53 gene exons 5, 7 and 8 are methylated.
Results in Figure 8A and 8B show that cytosine methylation greatly enhances 4-HNE binding at CpG sites in codons 152, 154, 156, 157, 282 and 290; however, only codon 282 becomes a 4-HNE strong binding site. In contrast, 4-HNE binding at CpG sites in codons 175, 170, 245 and 273 was not affected by C5 cytosine methylation. Furthermore, 4-HNE binding at codons 249 and 174 was not affected by C5 cytosine methylation.
Increasing evidence is emerging to indicate that DNA damage induced endogenously may be involved in disease processes, including carcinogenesis ( 41 – 44 ). A great effort has recently been exerted to determine the role that aldehydes, generated during lipid peroxidation, play in carcinogenesis. Lipid peroxidation is a cellular process, which commonly takes place, even under normal physiological conditions. This process becomes significant when cells are under oxidative stress, exposed to xenobiotics, and subjected to bacterial and viral infection ( 1 ). Aldehydes are also generated during autoxidation of fats and oils, particularly those rich in polyunsaturated fatty acids. Among the many species of aldehydes found in mammalian cells, 4-HNE is one of the most abundant and cytotoxic ( 1 ) and it is known that 4-HNE induces a wide variety of cellular responses ( 45 , 46 ). However, to date, the interaction of 4-HNE with DNA remain elusive, which is probably due to the difficulties in detecting 4-HNE–DNA adducts. In this study we have attempted to develop a method for detecting 4-HNE–DNA adduct formation at the nucleotide sequence level in the hope that it will be a useful tool for addressing the genotoxicity, mutagenicity and carcinogenicity of 4-HNE–DNA adduct.
Our results show that the E.coli nucleotide excision repair enzyme UvrABC nuclease incises 7 bases 5′ and 3–4 bases 3′ to 4-HNE–DNA adducts. The kinetics of incision at different sequences in a DNA fragment are the same, and the extent of UvrABC incision is proportional to the 4-HNE concentration used for modification. These results demonstrate that the UvrABC nuclease is able to incise 4-HNE–DNA adducts specifically and quantitatively. This finding allowed us to determine the 4-HNE binding spectrum in human p53 gene.
We have found that the strongest 4-HNE binding sites in exons 5 and 7 are at the second guanine residue in codons 174 and 249; both codons contain the same sequence and have the same nearest neighbor sequence (-G AGG* C-). Although 4-HNE binding at exon 8 is more evenly distributed, binding at codon 286, which contains the -GAG G*AA - sequence, is the strongest. The -CAGG*A- sequence in the intron 7 region is also a strong 4-HNE binding site. In contrast, 4-HNE does not bind at the -GAGGT- sequence, such as codon 172 in exon 5. Together, these findings suggest that the -AGG* (C/A)- sequence is a preferential site for 4-HNE binding. It is probable that this sequence aligns well with the long alkyl chain of 4-HNE to form a stable complex, which allows efficient Michael addition and cyclization between the exocyclic amino group of guanine and CC double bond of the aldehyde of 4-HNE to occur ( 1 ). Future molecular modeling may shed more insight into the interaction between 4-HNE and DNA.
We have shown previously that C5 cytosine methylation at CpG sites greatly enhances the binding of many carcinogens at this sequence ( 17 , 31 ). Our current results also show that C5 cytosine methylation at CpG sites does enhance 4-HNE binding at CpG sites; however, only the CpG site at codon 282 becomes a prominent 4-HNE binding site after methylation. The nature of this sequence-dependent C5 cytosine methylation effect on 4-HNE–DNA binding remains unclear; nonetheless, these results do further demonstrate the sequence specificity of 4-HNE–DNA binding.
Our results show that codon 249 and codon 174 are preferential sites for 4-HNE binding. Ample evidence supports that in animal models severe lipid peroxidation in liver is induced by conditions such as choline-deficient diet or carbon tetrachloride treatment ( 8 , 47 ). Since 4-HNE is a major product of lipid peroxidation in human liver, if lipid oxidation can cause hepatocellular carcinoma in humans, then an intriguing question is raised: why only codon 249 is a p53 mutational hotspot in human hepatocellular carcinoma while both codons 174 and 249 are preferential sites for 4-HNE binding? It is possible that in vivo only codon 249 is a preferential site for 4-HNE binding because of chromatin structure. However, it has been found that only the codon 249-mutated p53 gene product has a strong dominant-negative effect in liver cells ( 20 – 22 ). Liver cells with a mutation in codon 249 appear to have stronger p53 -minus phenotypes, such as increased resistance to apoptosis and higher tendencies to be carcinogenic, than liver cells with a mutation at other codons ( 20 , 22 ). In combination with our current results, these findings suggest that the reasons why codon 249 of the p53 gene is a mutational hotspot in hepatocellular carcinoma could be that this site is a preferential binding site for DNA damaging agents endogenously generated in liver and that liver cells with a mutation at this site have a growth advantage.
It should also be noted that epidemiological studies have shown that the geographic regions with a high incidence of hepatocellular carcinoma, such as Northern Africa and Eastern China, have a high incidence of AFB1 contamination in their food source. Importantly, >50% of hepatocellular carcinoma cases studied in these areas have a G to T transversion at the second guanine of codon 249 of the p53 gene ( 48 ). AFB1 has been suspected to be the etiologic agent for hepatocellular carcinoma in the regions that have a mutation at codon 249 of the p53 gene. Consistent with this notion is the recent finding from Essigmann's laboratory that AFB1 formamidopyrimidine adduct, an imidazole ring open product derived from the AFB1-dG, induces a high frequency of G to T transversion in E.coli cells ( 49 ). However, we found that in cultured hepatocytes AFB1-DE binds strongly at codons 248, 273, 157 and 158 of the p53 gene but not at codon 249 of this gene (data not shown), which is consistent with the results from Pfeifer's laboratory ( 50 ). The causative effect of AFB1 adduct formation and mutation at codon 249 of the p53 gene in human hepatocellular carcinoma remains to be determined. Interestingly, hepatitis B virus infection is common in regions with a high incidence of hepatocellular carcinoma, and epidemiological evidence indicates that there is a synergistic interaction between AFB1 exposure and hepatitis B virus infection on the induction of hepatocellular carcinoma, and the mutation at codon 249 of the p53 gene is associated with AFB1 exposure and hepatitis B virus infection ( 48 , 51 ). It is possible that hepatitis B virus infection and AFB1 exposure could also greatly elevate lipid peroxidation in liver, and the aldehydes, particularly the 4-HNE, generated by hepatitis B virus infection and AFB1 exposure, may damage codon 249 of the p53 gene preferentially and contribute to the mutation at this site and subsequently initiate hepatocarcinogenesis.
We thank Drs O.Bhanot and Yen-Yee Tang for their critical review of the manuscript. This work was supported by NIH grants ES03124, ES08389 and ES10344 (M.-s.T.).