Chem. Res. Toxicol. 1994, 7, 209-218
209
Mutation and Repair Induced by the Carcinogen 2-(Hydroxyamino)- 1-methyl-6-phenylimidazo[ 4,5- blpyridine (N-OH-PhIP)in the Dihydrofolate Reductase Gene of Chinese Hamster Ovary Cells and Conformational Modeling of the dG-CS-PhIP Adduct in DNA Adelaide M. Carothers,*vtWei Yuan,+Brian E. Hingerty,i Suse Broyde,s Dezider Grunberger,?and Elizabeth G. Snyderwinell Institute of Cancer Research, Columbia University, New York, New York 10032, Health Sciences Research Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, Department of Biology, New York University, New York, New York 10003, and Laboratory of Experimental Carcinogenesis, Division of Cancer Etiology, National Cancer Institute, National Institutes of Health, 37-3C28, Bethesda, Maryland 20892 Received November 1 5 , 1 9 9 3
Three experiments using 20 pM 2-(hydroxyamino)-l-methyl-6-phenylimidaz0[4,5-b1 pyridine (N-OH-PhIP)were performed to induce mutations in the dihydrofolate reductase (DHFR)gene of a hemizygous Chinese hamster ovary (CHO) cell line (UA21). Metabolized forms of this chemical primarily bind a t the C-8 position of guanine in DNA. In total, 21 independent induced mutants were isolated and 20 were characterized. DNA sequencing showed that the preferred mutation type found in 75 % of the induced DHFR- clones was G C T-A single and tandem double transversions. In addition to base substitutions, one mutant carried a -1 frameshift and another one had lost the entire locus by deletion. The induced changes affected purine targets on the nontranscribed strand of the gene in nearly all of the mutants sequenced (18/19). At the time that the first two experiments were performed, the initial adduct levels were quantitated in treated cells a t the mutagenic dose by 32P-postlabeling. While the induced frequency of mutation was relatively low (-5 X 109,the adduct levels after a 1-h exposure of UA21 cells to 20 pM N-OH-PhIP were relatively high (13 adducts X 108 nucleotides). This latter method was then employed to learn if the induced mutation frequency correlated with rapid overall genome repair of PhIP-DNA adducts. Total adduct levels, determined using DNA samples from treated cells collected after intervals of time, were reduced by about 50% after 6 h, and about 70% after 24 h. Since overall genome repair in CHO cells is relatively slow compared with preferential gene repair, the removal of dG-C8-PhIP adducts was apparently efficient. In order to better understand the mutational and repair results, we performed computational modeling to determine the lowest energy structure for the major dG-C8-PhIP adduct in a repetitively mutated duplex sequence opposite dA. Results of this analysis indicate that the PhIP-modified base resembles previous structural determinations of (deoxyguanosin-8-y1)aminofluorene; the carcinogen is in the B-DNA minor groove and it adopts a syn conformation mispaired with an anti A. The implications of this conformational distortion in DNA structure for damage recognition by cellular repair enzymes are discussed.
-
Introduction
cigarette smoke (6). PhIP is metabolically activated by the hepatic monooxygenase system to the proximal 2-(hydroxyamino)form which, in turn, is converted to the ultimate N-sulfate and N-acetoxy derivatives by sulfotransferase and acetyltransferase, respectively (7-1 1). These latter species bind covalently to DNA. HPLC analysis of enzymaticallyhydrolyzed PhIP-modified DNA, as well as 32P-postlabelinganalysis, has shown that the
A diverse group of heterocyclic amines form upon cooking muscle tissue from meat or fish ( I ) . These compounds are generated from the condensation of amino acids with creatinine during pyrolysis. In the human diet, they are present at parts per billion levels and may pose a cancer risk (2). Indeed, the Nz-hydroxylated metabolite of one such amine, 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP),l is mutagenic in model systems 1 Abbreviations: PhIP, 2-amino-l-methyl-&phenylimidazo[4,5-blpyr(2)and carcinogenic in rodents (3-5). In addition to being idine;N-OH-PhIP,2-(hydroxyamino)-l-methyl-&phenylimidam[4,6-b]found in cooked foods, PhIP is also present in mainstream pyridine; dG, 2’-deoxyguanosine;dG-C8-PhIP,N-(deoxyguanosin-8-y1)pyridine;AAAF,N-acetoxy-N2-amino-l-methyI-6-phenylimidazo[4,5-b]
* Correspondence should be addressed to this author at Columbia
University, 701 W. 168th St., Room 532, New York, NY 10032. Phone: 212-305-6923;FAX: 212-306-5328. t Columbia University. 8 Oak Ridge National Laboratory. 1 New York University. II National Institutes of Health. * Abstract published in Advance ACS Abstracts, March 1, 1994.
dG-CsAF, acetylaminofluorene;N-OH-AF,N-hydroxy-2-aminofluorene; N-(deoxyguanosin-8-yl)-2-aminofluorene; dG-C8-AAF, N-(deoxyguanosin-&yl)-N-ace~l-2-~ofluorene; 1-NOP, 1-nitroeopyrene;BcPHDE, ( ~ ) - 3 a , 4 ~ - d i h y d r o ~ - l c r , 2 a - e p o ~ - l , 2 , 3 , ~ t e ~ ~ y d r o ~ ~ o ~ c l p h e n a n threne; BPDE, ( * : ) - 7 ~ , 8 a - ~ d r o x y - 9 a , l ~ ~ ~ x y - 7 , 8 , 9 , l ~ t e t r ~ y & o ~ n zo[alpyrene;PAH,polycyclic aromatichydrocarbon;DHFR, dihydrofolate reductase;CHO, Chinesehamsterovary;DMSO,dimethyl sulfoxide;SDS, sodium dodecyl sulfate; EDTA, ethylenediaminetetraacetic acid; kpb, kilobase pair; PCR, polymerase chain reaction.
0 ~ 9 3 - 2 2 a ~ 1 9 4 1 2 ~ 0 ~ - 0 2 0 9 ~ o 4 . ~00 11994 o American Chemical Society
Carothers et al.
210 Chem. Res. Toricol., Vol. 7, No. 2, 1994 0 II
0”” 4
5
Figure 1. The PhIP adduct at the C-8 position of guanosine.
major adduct formed in vivo by reaction of DNA with N-OH-PHP is N2-(2’-deoxyguanosin-8-yl)-PhIP(dG-CSPhIP) (12-16). In order to examine the mutagenic specificity of N-OHPhIP, we have induced a collection of Chinese hamster ovary (CHO) cell mutants that are deficient in dihydrofolate reductase (DHFR) activity. DHFR- mutants are readily selectable (17)using a hemizygous cell line (UA21) that bears a single copy of the wild-type gene (18).Adduct levels at the initial time after carcinogen treatment in the parental cells were quantitated by the 32P-postlabeling method (19). Also using this method, repair of PhIPDNA adducts from the genome overall of treated cells was determined. DNA sequence analysis of mutants showed that, like the other aromatic amines we have previously (AAAF) analyzed, N-acetoxy-N-acetyl-2-aminofluorene (20)and N-hydroxy-2-aminofluorene(N-OH-AF) (211, N-OH-PhIP primarily induced single-base G C T-A transversions. These arylamine carcinogens also bind to the C8 position of 2’-deoxyguanosine (dG) (22).To better understand the mechanisms of certain carcinogen-induced mutations and cytotoxicity, as well as how different DNA lesions are recognized by repair enzymes, we have been interested in analyzing the nature of structural alterations in DNA by physical and computational means. Thus, the conformation of the N-(deoxyguanosin-8-yl)-2-aminofluorene adduct (dG-C8-AF) opposite dA (23)and of the N- (deoxyguanosin-8-yl)-N-acetyl-2-aminofluorene adduct (dG-Ca-AAF) opposite dC (24)was determined in DNA by two-dimensional NMR and potential energy computations. To learn if there are structural similarities between these arylamines and PhIP adducts, computational modeling studies were performed for the major dGC8-PhIP adduct in duplex DNA (Figure 1). Among independent PhJP-induced DHFR- mutants, transversions arose at a specific target sequence at different sites. Conformational searches with potential energy minimization were made using one of these target sequences [d(Cl-C2-A3-[PhIP]G4-G5-T&A7.d(T&A9-C10-&-T12G13-G14)], because this site yielded the same mutation twice independently. The partner strand of this sequence had the Watson-Crick complement at all positions except the lesion, which was mispaired with A, since G.C T.A base changes were the predominant observed mutation. Given the results of previous studies with arylamines (20, 21, 23, 24) and the well-known mutagencity of dG-C8 adducts in targeted systems that use site-specifically modified copstructa (see ref 25 for review), we think it is reasonable to assume this mispair represents a prevalent PhIP-induced mutagenic lesion.
-
-
Materials and Methods Materials. Caution: P h I P and the derivative N-OH-PhIP are carcinogenic to rodents and should be handled carefully.
PhIP was purchased from the Nard Institute (Osaka, Japan). N-OH-PhIP was prepared by reduction of the nitro derivative of PhIP as described (14);it was confirmed to be >95% pure by HPLC. Cell culture media used in this work and materials for DNA sequence and Southern blot analyses of mutants were the same as previously indicated (18,26). All other reagents were purchased from Sigma (St. Louis, MO). Cell Culture and Carcinogen Treatment. The parental cell line from which all mutants were generated was UA21, a CHO derivative that is hemizygous at the DHFR locus (18).The methods used for the growth of these cells and the selection of DHFR-deficient mutants were detailed elsewhere (17). To minimize the possiblity of selecting a preexisting mutant in the induced populations, UA21 cells were not continuously cultured. Rather, separate carcinogen treatments were performed using (2-3) X 106 cells expanded once from a single cell clone. For every 10 dishes of treated cells subjected to the selection procedure, a dish of untreated (solvent only) cells was included as a negative control. Three separate mutagenesis experiments were performed. For these three carcinogen treatments, the medium was removed from each dish, and the cells were washed twice withwarm (37 “C)phosphate-buffered saline 12.7 mM KC1, 1.2 mM KH2PO4, 138 mM NaC1, 8.1 mM Na2HP04.7H20 (pH 7.0)]. Each experimental dish then received 10 mL of Ham’s F12 medium without serum. N-OH-PhIP was added using a stock solution (20 mg/mL dissolved in DMSO) such that the final concentration was 20 pM. Treated and untreated dishes were next incubated for 1 h in a C02 incubator at 37 OC. Subsequently, the medium was aspirated, and the dishes were all washed once with warm F12 medium containing 10% bovine fetal calf serum and washed twice with warm PBS. The cells were then trypsinized, resuspended in 10 mL of complete F12 medium, and replated. Selection for mutants followed a 6-day expression period (17). All colonies that arose after selection were isolated and subcloned. To confirm the phenotype of putative mutants, cell extracts from each were tested for failure to bind the substrate analog [3H]methotrexate (27). Mutants were deemed independent even if they were obtained from the same dish after DNA sequencing verified that their genotypes were different. For repair studies, cells were prelabeled with [3Hlthymidine and subcultured into complete F12 medium 1day prior toN-OHPhIP treatment as described previously (28).After the carcinogen exposure, cells used for the determination of initial adduct levels were lysed immediately with lysis buffer [lo mM Tris-HC1, 1 mM EDTA (pH 8.0), 0.5% SDS, and 1mg/mL proteinase Kl. Treated cells used for repair analysis were incubated in medium supplemented with bromodeoxyuridineand fluorodeoxyuridine at 106 and 10-8 M final concentrations, respectively. After allowing intervals of incubation time for repair, DNA samples were prepared from cells that were twice washed with PBS and lysed. The DNA was subsequently purified by CsCl gradient fractionation (28). a2P-Postlabeling Analysis. This procedure resolves P2P1ATP-labeled bisphosphonucleotide adducts as fingerprints on autoradiograms after poly(ethy1ene imine)-cellulose thin-layer chromatography (19). Purified DNA from N-OH-PhIP-treated parentalUA21 cells was analyzed by the S2P-postlabelingmethod using intensification conditions (13, 19). Adduct levels were determined from the Cerenkov assay of excised adduct spots after autoradiography (29) and by the previously described procedure (13). The arbitrary numbering of the PhIP adduct spots already reported (13, 16) was maintained here. Mutation Analyses. DNA from independent PhIP-induced DHFR- clones was prepared by a crude method (30) for amplification by the polymerase chain reaction (PCR) or for Southern analysis as described (26). Mutational changes were determined by a direct sequencing method that used doublestranded PCR amplificationproducts (30)and the primer system previously detailed (20). Southern blot analysis of one PhIPinduced mutant and parental UA21 DNA was performed with
PhIP-Induced Mutations, Repair, and Conformation hybridizationto a nick-translatedmixed probe of 10DHFR cloned DNA fragmenta as described earlier (26'). Computations. Minimized potential energy calculationswere carried out with DUPLEX, a molecular mechanics program for nucleic acids that performs potential energy minimizations in the reduced variable domain of torsion angle space (31). A hydrogen bond penalty function (31,32)is employed in all firststage minimizations to aid the location of any type of designated hydrogen-bondedstructure or denatured site when the function is not implemented at a particular base pair. This penalty function makes no contribution to the energy when a selected hydrogen-bondingscheme has been achieved, but penalizes the energywhen the selectedpattern is not found. Thus,the function guides the minimization algorithm toward structures with the chosen hydrogen bonds. It is subsequently released in the terminal minimization so that final computed structures are unrestrained minimum-energy conformations. In the present work, the adjustable weight in this penalty function was assigned a value of 30 kcal/ (mol.A2). The geometry of the carcinogen was generated by building the major C8 adduct of PhIP to guanine with MacroModel V3.5X (33) and minimizing it using the MacroModel implementation of the AMBER force field (AMBER*)and the default parameters. The solvent and extended nonbonded cutoff option were employed. The resulting bond lengths and bond angles shown in Figure 1A and 2A (Appendix)are all within normal ranges, and the aromatic portions of PhIP are planar. These values,together with fixed dihedral angles of Oo or 180° in the aromatic rings, were used in the DUPLEX coordinate generator. The flexible torsion angles governing the dG-PhIP adduct orientation and flexibilities within the PhIP moiety are defined in Table 1A (Appendix). Partial charges employed for PhIP and the linkage site in DUPLEX were computed for the dG-PhIP adduct with the CNDO module of CHEM-X (1987Version, Chemical Design Limited,Mahwah,NJ). These areshowninTable 2A (Appendix) and are compatible with the partial charges employed by DUPLEX (34). Torsional potentials and rotation barriers for these flexible bonds were assigned as follows: a' and p' were the same as those employed for other dG-C8adductsat the carcinogen-base linkage site (35),while the linkage site to thephenylring,y', was assigned a value identical to the linkage to phenyl in 4-aminobiphenyl (36). The methyl rotation on PhIP was modeled on the methyl of thymine (31). These quantities are summarized in Table 1A (Appendix).
Results Mutant Characterizations. Three separate mutagenesis experiments were performed identically using UA21 cells (18)treated with 20 pM N-OH-PhIP. This carcinogen concentration in each experiment yielded about a 22 % survival of the treated population by colony-forming assay. From the first experiment, 15independent DHFRmutants were isolated from 11of 20 treated dishes (a total of 3 X lo6 challenged cells), giving an induced frequency of 5 X 10-6. Subsequent PhIP treatments produced only an additional 6 independent mutants. By fluctuation analysis, the spontaneous rate of mutation a t this locus is 1.3 X lo-' (37). Given that mutation frequency is approximately 10-fold higher than mutation rate, the PhIP-induced frequency at the DHFR locus was relatively low. We cannot exclude the possibility that 1-3 of the mutants in our combined collection were spontaneous in origin since a total of 9 X 106 cells were challenged by selection in the three experiments. However, in the course of this work, selection of 9 X 106 untreated cells failed to yield any spontaneous colonies. Moreover, in similar negative control experiments recently performed using other chemical carcinogens [N-OH-AF, syn-benzo[gl-
Chem. Res. Toxicol., Vol. 7,No. 2, 1994 211 chrysene diol epoxide, and (-)-anti-3a,4@-dihydroxyla,2~-epoxy-1,2,3,4-tetrahydrobenzo[clphenanthrene (BcPHDE)] no spontaneous mutants were obtained. The characteristics of the 20 N-OH-PhIP-induced mutants are summarized in Table 1. The predominant change detected in 16 (80%) of the mutants was a single base substitution. The favored mutation type was G-C T.A transversion which arose in 13/16 point mutants and twice among tandem double base changes. The mutation in DPh7 which affected an adenine does not appear to be targeted to the premutagenic lesion since N-OH-PhIP forms adducts exclusively at guanines. Thus, this mutation may have occurred as a spontaneous event. Carcinogeninduced mutations in the DHFR gene are frequently observed to be strand-biased, and base substitutions affect purines on the nontranscribed strand of the gene (39).In this PhIP-induced mutant collection, again all except one of the mutations arose at guanines located on the nontranscribed strand of the gene (DPhl, Table 1). The different heterocyclic aromatic amine and PAH carcinogens we have used formerly to induce DHFRmutants have produced similar types of base substituents T transversions), but have displayed (e.g., purine distinct target site specificities within the 25 kilobase pair (kbp)gene. Among the 17distinct PhIP-induced mutation sites, 11(65%) are novel targets not previously altered in approximately 150 DHFR- point mutants induced by treatments with various agents. The frequency and distribution of mutation sites in the DHFR gene recently were summarized (40).Mutations at new target sites in the gene occurred at the following cDNA positions: 10, 74,159,160,161,184,282,304,353,423,424,and 484, and at the +1position of the splice donor site of exon 3 (see Table 1). The mutation in one clone (DPh21) from the first mutagenesis experiment was not found although these cells were subcloned twice and rechecked for phenotype, and the sequencing of genomic PCR fragments containing each of the 6 coding exons was repeated. Sequencing of DHFR cDNA from these cells resulting from reverse transcription of total RNA and subsequent PCR amplification was not performed. Thus, the mutation in this clone may affect splicing via a remote base change in one of the introns. Mutant DPh20 failed to yield any product following PCR amplification of the exon-bearing fragments. Southern analysis using a probe that scans a continuous 34 kbp comprising the DHFR locus (26)confirmed that the entire gene was deleted in DPh2O cells (data not shown). Adduct Characterization, Quantitation, and Repair. The level of PhIP-DNA adducts at the mutagenic dose (20 pM) was evaluated in UA21 cells by 32Ppostlabeling, in part to learn if low levels of modification by this chemical could explain the induced mutation frequency. Three PhIP adducts profiled in Figure 2 were found in DNAfromN-OH-PhIP-treatedUA21 cells. These adducts were identical to those seen in tissues of animals fed PhIP (13,16) and in DNA reacted in vitro with N-acetoxy-PhIP (16). The major adduct spot (indicated by the number "1")was identified as the dG-C8 adduct by comigration with synthetic standard (16).Although they are both modifications of guanine, the structural identities of the minor adducts 2 and 3 are not yet known with certainty. Recently, it was shown that inclusion of an additional P1 nuclease hydrolysis step following the labeling reaction in the 32P-postlabelingassay eliminated
-
-
Carothers et at.
212 Chem. Res. Toxicol., Vol. 7, No.2,1994
Table 1. Characteristics of D E . Mutants Induced. by 2-Amino-l-methyl-G-phenylimida~[4~~]pyridine Designation
DPhl’
DPh2’ DPh3’ DPh4’ DPh5’ DPh6’ DPh7’ DPha3 DPh92
DPhlO’ DPhll’
DPh123 DPhl3’ DPhl4’ DPhl5’ DPhl6‘
Sequence Context Exon Positionb
CCC CCT TCA TCC CCT C M CTA
E E E E
ACC
E E E E E
ATT TTC GAG CCA
CCA CCA
TCC’ CCC
1
74
*A
2
-1
c-4
CCC
3 3 3 3 4 4 4 4 5 6 6 6 6
161 187 +1 +4 244 304 350 353 484
*T C+T *T A+T *T *T C+T *T
TM CTC MC MC AGG CAC ACT
CTA
TTT
E E E
DPhl7’
TCT
E CAC
DPh1a3
TAT
a GCC
DPhl9’
CCA
MT
GAG
SINGLE BASE SUBSTITUTIONS 1 10 *T
C M
ACA
E
Change
CTA M A M C
-1 -1
517 541
C+T C+T
*T *T C+T
Phenotypic Consequence
missense Pro3+Tht nonsense (opal) s p l i c e acceptor defect miss ense C1ys3+Va1 nonsense (amber) s p l i c e donor defect s p l i c e donor defect nonsense (ochre) nonsense (ochre) missense Clyl16+Val missense Clyl17+Val nonsense (opal) splice acceptor defect s p l i c e acceptor defect nonsense (ochre) nonsense (ochre)
S I N G L E BASE FRAMESHIFT
4
282
-C
COMPLEX MUTATION CC+TT 3 159,160 5
423,424
CC+TT
nonsense a t 293
double missense MetS2 -Clys3+I1e - Cys nonsense (ochre)
GENE DISRUPTION DPh203 Superscripted numbers to the right of the mutant designation indicate which mutants came from separate mutagenesis experiments (1-3). Unless otherwise indicated, the target sequence represented is the nontranscribed (coding) strand in a 5’ to 3’ orientation. Mutated bases are underlined. Position numbers refer to the protein-coding region of the cDNA where the first base is the adenine of the ATG translation initiation codon (38). For mutations that occurred in introns, the position relativeto the nearest exon is given, with “+* indicating downstream from the exon and “-* meaning upstream from the exon. Denotes that the sequence shown is the transcribed (template) strand in a 5’ to 3’ orientation assuming adduct formation took place at the guanine.
*
Q
2
.or
Figure 2. a2P-Postlabeling profile of PhIP adduds in DNA of UA21 cells treated with 20 p M N-OH-PhIP. The assay was performed using intensification conditions (13,16). Adduct 1 corresponds to the C-8guanine adduct of PhIP.Autoradiography was at -70 “C for 30 min.
the two minor adducts, suggesting that one or both of these spots may result from partial DNA hydrolysis (41). Initial adduct levels (e.g., after the l-h exposure to the carcinogen) evaluated contemporaneously with the first and second mutagenesis experiments resulted in modifications of 12.94 and 13.05 total PhIP adducts/106 nucleotides, respectively. Thus, a 20 p M N-OH-PhIP concentrationproduced approximately0.65 lesion/25kbp, which is the size of the DHFR gene. This level of initial damage is high relative to the initial adduct levels resulting from treatment of mammalian cells with polycyclic aromatic hydrocarbon (PAH) carcinogens (discussedbelow). Another explanation for the low induced mutation frequency is that the removal of PhIP-DNA adducts is rapid. To test this idea, we quantitated adduct levels in treated cells that were collected after intervals of time
(1-24 h) to allow for repair. Gradient-fractionated DNA samples were used in these 32P-postlabeling experiments so that the decline in adduct levels over time represents only repair and not de nouo DNA replication. The results of independent repair assays are presented in Figure 3. The combined data show that more than half of totalPhIP adducts are removed from DNA in these CHO cells at 6 h after treatment. Energy Minimization Modeling. In view of the mutagenesisand repair similaritiesupon treatment of cells with N-OH-PhIP andN-OH-AF (see Discussion),we next performed potential energy minimization searchesto learn if there are also structural similarities between the major dG-C8-PhIP adduct and that of dG-C8-AF (23). Two unequivocally independent PhIP-induced mutations occurred at a single site in the splice acceptor sequence of exon 6 (DPhl3 and DPhl4, Table 1). The sequence affected in these mutants (5’-CA G G-3’; the underlined base was mutated) was also mutated at two additional sites among mutants obtained from the first N-OH-PhIP experiment (DPh3, DPhl9, Table 1). Assuming that this sequence was the preferred mutagenic context, a conformational search with energy minimization was performed using a duplex 7-mer with the sequence d(Cl-C2-A3[PhIP] G4-G5-T6-A7)*d( T8-A9-C10-~-T12-G13-G14). The GOAmismatch was studied because this pairing represents the in vivo mutagenic specificity of PhIP adducts.
PhZP-Znduced Mutations, Repair, and Conformation
cn W
T I
60
vh
0 I-
50
-I
o
40
r.
30
E!
3 0 r
z I-
20
2 0
10
o a
0 0
5
10
15
20
25
TIME (Hours) Figure 3. Overall genome repair of PhIP-DNA adducts in treatedUA21 cells. Density-labeled DNA from UA21 cellstreated with 20 gM N-OH-PhIP was prepared after a l-h exposure (the initial damage) to the carcinogen, as well as after 2-, 6-, and 24-h incubationsposttreatment. DNA samples, preparedas described in Materials and Methods, were analyzed by S2P-postlabeling to quantitate total adducts remaining over time. Each value is the mean standard error of 5 (the initial damage time point) or 3 separate experiments. The adducts are represented by the following symbols: 1 (circles),2 (squares), and 3 (triangles). AD€NINEGUANINE
HYDROGEN BONDING
R
R
Figure 4. The four adenine-guanine pairing schemes. The search strategy employed was as follows: The hydrogen bond penalty function (31) was used to locate structures with four different possible hydrogen-bonding schemes at the G-A mismatch. Three of these bonding schemes were previouslyobserved in crystals of unmodified DNA with G mispaired with A (42-47), while the fourth was found in the solution structure of a duplex sequence containing dG-C8-AF opposite dA (23). These pairing schemes are illustrated in Figure 4 and are summarized in Table 3A (Appendix). The hydrogen bond penalty function was also employed to locate standard WatsonCrick base pairs at the other sites. Torsion angles for the DNA starting conformation were those of the B-DNA fiber diffraction model (481, except that syn bases at the lesion site were oriented with glycosidic torsion angles 04'-C1'-
Chem. Res. Toxicol., Vol. 7, No. 2, 1994 213
N W 4 of 60'. Sixteen orientations of the carcinogenbase linkage torsion angles were employed as starting conformations for the energy minimizations of each hydrogen-bonding type: a' = ,'O 90°, 180°, 270' in combination with 8' = Oo, 90°, 180°, 270'. The torsion angles y' and 6' were always started at 60' and 45', respectively. Thus, a total of 64 starting conformations were employed. Following these first-stagetrials, asecond minimization was performed for each resulting structure without the hydrogen bond penalty function. Finally, those structures that employed protonated bases were deprotonated and minimized again as in earlier work (32). This latter step is necessary because energiesof protonated and unprotonated forms cannot be compared since they contain different numbers of atoms. The lowest energy structure that was computed from the 64 starting conformations places the carcinogen in the B-DNA minor groove (Figures 5 and 6). The carcinogen edges are exposed, and the aromatic moiety is mostly shielded from solvent. The modified guanosine is syn ( x = 61°), it is stacked within the helix, and the partner adenosine is anti. This lowest energy structure is derived from the hydrogen bond target scheme 06(G).-(Hl+Nl) (A) (Figure 4, bottom right), except that the final structure has no hydrogen bonds between G and A (see Materials and Methods). However, the N1 of A is poised to pair with 0 6 of G if the N1 were protonated. The PhIP adduct is oriented so that its long axis is directed toward the 3' end of the modified strand. The torsion angles a' and 8' that defiie this orientation are 214' and 140°,respectively. The imidazopyridine moiety lies over the sugar of G5, and the phenyl ring lies over the sugar of T6. The DNA torsion angles given in Table 4A (Appendix)are all within normal B-DNA ranges (49) except for the pucker at the sugar of the modified G. The pseudorotation parameter (SO) for this sugar is 111' and corresponds to an 04'-endo-Cl-exo pucker, whereas the standard B-DNA pucker is in the C2'-endo domain with the pseudorotation parameter centered at 180'. The phenyl ring is twisted 21' out of the plane of the attached pyridine moiety, and it protrudes out from the groove with one surface exposed. Twists between phenyl rings in biphenyl analogs were observed in solution (51) and were predicted (36)for the dG-C8 adduct of 4-aminobiphenyl. The PhIP methyl group faces outward so that the other edge of the carcinogen is able to fit closely against the DNA. Other syn guanine structures were more than 5 kcal/ mol higher in energy and anti guanines were disfavored by over 12 kcal/mol. Moreover, a syn conformation of the modified dG is consistent with the proton NMR results of Turesky et al. (52) for the related heterocyclic amine 2-amino-3-methylimidazo[4,5-flquinoIine adduct at dGC8.
Discussion Treatment of UA21 cells with 20pMN-OH-PhIP yielded 21 independent DHFR- mutants in three separate experiments. The induced change in 75% of the 20 characterized mutants was G C T-A single and tandem double transversions. The results using this heterocyclic amine were similar to those previously obtained using the aromatic amine N-OH-AF (21) with respect to induced mutation frequency, specificity, and repair. Both chemicals form covalent adducts a t the C8 position of guanine. The induced frequency of mutation was also low using N-OH-AF, and we maintained that the collection of DHFR-
-
214 Chem. Res. Toxicol., Vol. 7, No. 2,1994
Carothers et al.
Figure 5. Stereo space-filling representations of d(C1-C2-A3-[PhIP]G4-G5-T6-A7~d(T8-A9-C10-A11-T12-G13-G14) lowest energy conformation. The PhIP is in the B-DNAminor groove oriented in the 3’ direction of the modified strand. The modified G is yellow, the PhIP is red, the adenine opposite the modified G is green, and the rest of the DNA is cyan (blue-green). (Upper) View emphasizing the wedging of PhIP into the minor groove. (Lower) View into the major groove, emphasizing the syn orientation of the modified guanine which nevertheless remains stacked with adjacent bases.
Figure 6. Stereo ball-and-stick representations of the PhIP-modified oligomer. The sequence and structure are the same as those depicted in Figure 5, upper and lower views, respectively.
mutants was indeed induced because a third of the changes in independent clones occurred at a single hotspot in separate experiments (21) that was also repetitively mutated using AAAF (20). H-OH-AF also induced mostly G C T*Atransversions, and these were found in 61% (11/15) of mutants. In this latter collection,only guanines
-
were mutated and the altered base was located on the nontranscribed strand in 88% of the DHFR- mutants containing single base changes. In part because of these similarities with N-OH-AFinduced mutants, we think the mutations presented in Table 1that are predominantly strand-biased G-C T*A
-
PhZP-Znduced Mutations, Repair, and Conformation transversions were induced by N-OH-PhIP. Different carcinogensin this system have tended to induce mutations at distinct target sites in this mammalian gene, and the locations of mutations induced by PhIP accordingly were different from those identified previously. Thus, novel sites in the gene were mutated in 55% (11/20)of the PhIPinduced collection. In the present study, only three mutations occurred at sites also mutated using N-OH-AF (21)and AAAF (20). These mutations arose at DHFR cDNA positions 244 (DPhB), 350 (DPhlO), and 541 (DPhl6). A preferred PhIP-induced mutational target sequence of 5'-CA G G-3' may have been demonstrated in this work since transversions were detected in 4 independent mutants at 3 different sites in the DHFR gene (the -1 splice acceptor site of both exons 2 and 6 and cDNAposition 423). Altogether in the 561base pair DHFR cDNA and consensus splice site sequences, 5-CA G G-3' occurs at 8positions on the nontranscribed (coding)strand. We did not observed this specific change, however, in independent mutagenesis experiments,which is a criterion of a hotspot sequence. Using the 32P-postlabelingmethod (13, 16), we showed that the initial level of PhIP adducts in DNA of treated UA21 cells was relatively high (about 13 adducts/106 nucleotides or 1adduct in 38.5 kbp). Both N-OH-AF and N-OH-PhIP are unstable in aqueous solutions; nonetheless, in the latter case the low frequency of induced mutation, especiallyin the second mutagenesis experiment, could not be explained by a trivial failure of the chemical to damage cellularDNA. This initial adduct level contrasts with the roughly 21.7-fold lower level of 6 adducts per lo7 nucleotides extrapolated to form in CHO cellstreated with the racemic PAH BcPHDE from 32P-postlabelingquantitations after a 1-h treatment (53)at the mutagenic dose (0.1 pM) (54). Treatment of UA21 cells with 0.1 pM BcPHDE resulted in less lethality than occurred after a 20 FM PhIP treatment (78% killing with PhIP compared with 54% killing with BcPHDE) (54). Interestingly, the induced mutation frequency for BcPHDE (1.7 X 10"') at the DHFR locus was at least 34-fold higher than observed here for PhIP (54). Repair of BcPHDE adducts from these cells, in contrast to PhIP adducts, was considerably slower and reduced the PAH damage by 2-fold only from the transcribed strand of the DHFR gene after 24 h (53).Taken together, these results suggest that the higher initial PhIP adduct levels contributed to cytotoxicity but were poorly mutagenic. Repair of the arylamine adducts, dG-AF and dG-AAF, from the genome overall has been shown to be rapid in both CHO (55)and human fibroblast cells (56). Consistent with results from these studies, we observed a relatively rapid overall genome repair rate for PhIP-DNA adducts in UA21 cells. In one representative PhIP repair experiment, the adduct level remaining at 24 h was 23.9% of the initial damage and reduced the total adduct level to about half the highest value by 6 h after treatment (Figure 3). Furthermore, the similarity in the decline of the three adduct spots observed by 32P-postlabelingsuggests that rather than representing different PhIP modifications, they may either reflect partial hydrolysis products of a single dG-C8-PhIP adduct or oxidation products formed in vitro during the course of sample isolation. Oxidation products derived from dG-AF adducts are formed upon alkali treatment of DNA samples, but have not been proved to exist in vivo (57-59).
Chem. Res. Toxicol., Vol. 7, No. 2, 1994 215 Detection of adducts in the overall genome by cellular enzymes may be different for adducts at the C8 position of dG relative to those affecting the N2 position of this base. A differential repair rate was demonstrated in human cells using (f)-7/3,8a-dihydroxy9a,lOa-epoxy7,8,9,lO-tetrahydrobenzo[a]pyrene(BPDE) and 1-nitrosopyrene (1-NOP) (60). The PAH, BPDE, binds to dGN2,whereas 1-NOP binds to dG-C8. In this latter study, BPDE adducts were repaired more slowly from the overall genome relative to the nearly8096 repair of 1-NOPadducts by 24 h. Consistent with results presented here, 1-NOP was also less effective than BPDE in inducing mutations using a modified shuttle vector (60). That the modificationposition on the base (consequently the adduct conformation) and not the nature of the adducted moiety per se influences the repair rate is indicated by the study of adduct removal after treatment of primary cultures of rat hepatocytes with N-hydroxyN-acetyl-2-aminofluorene (61). In this work, the dG-N2AAF adduct, which represented 6% of the initial bound adducts, persisted at its initial level for 24 h, unlike the dG-C&AAF and -AF adducts that were repaired. We conclude from the present study that the low mutation frequency at the DHFR locus using N-OH-PhIP is in part due to rapid adduct removal from the DNA of treated cells. An alternative explanation, though, that PhIP-DNA adducb are chemically unstable and result in depurination, has not been excluded by the data presented here. A possible explanation for the efficient removal of dGC8 adducts in general is that damage recognition is influenced by the lesion geometry and that rotation of a base about the glycosidicbond from anti to s y n may favor this recognition. Therefore, computational modeling was performed to learn if d G P h I P adducts produced a distortion in DNA structure that is similar to dG-AAF or dG-AF. The results of this analysis showed that the lowest energy conformation of a G P h I P adduct in a mutagenic context causes the modified base that is aligned with an anti A to adopt a s y n conformation. The rotated G is still stacked with adjacent bases. Furthermore, the carcinogen plane is sandwiched between the DNA strands in the minor groove. .Hydrophobic interactionsstabilizethis orientation of the carcinogen moiety without hydrogen bonding. This structure is analogous to the conformation of dG-C8-AF previously described (231, but distinct from the basedisplacement conformation of dG-C8-AAF (24). Supporting the idea that DNA conformation may influence damage recognition, the rate of mismatch correction, in addition to that of nucleotide excision repair, is a180 more rapid for a dG in a syn conformation. Thus, the efficiency of G.G mismatch correction in mammalian cells was 92 % (621,and the crystal structure of a duplex oligonucleotide carrying a G-G mismatch has antissyn glycosidic angles (63).
Acknowledgment. We thank Dr. Herman A. J. Schut, Department of Pathology, Medical College of Ohio, for 32P-postlabelinganalysis. We also thank ProfessorsRobert Shapiro and James Canary, Chemistry Department, New York University: we are grateful to Dr. Shapiro for his insightful perspective, helpful discussions, and assistance in preparation of some of the figures, and to Dr. Canary for assistance with MacroModel. We thank Dr. Bin Li for preparation of the color illustrations. This investigation was supported by National Institutes of Health Grants CA28038, RR06458, CA21111, and CA39547; by DOE
Carothers et al.
216 Chem. Res. Toxicol., Vol. 7, No. 2, 1994
Grant DE FG02 90ER60931,US.DOE Contract DE AC05840R21400 with Martin Marietta Energy Systems, and US.DOE, Office of Health and Environmental Research, Field Work Proposal ERKP931; and by the Lucille P. Markey Charitable Trust. Computations were carried out on Cray Supercomputers at the Department of Energy’s National Energy Research Supercomputer Center and the National Science Foundation’s San Diego Supercomputer Center. Table IA. PhIP Torsion Angles and Barriers. torsion angles
atoms sign N9-C&N-C2 C&N-C2-N1 C ~ - N ~ - C M ~ - H -t C7-C6-Cl’-C2’
a’
8’ 7’
6’
m 2 2 3 2
Vo/2(kcal/mol) 3.54 3.54 -1.34 3.00
Figure 1A. Bond lengths of PhIP employed in the modeling study. Methyl hydrogen bond lengths are 1.090 A.
C 8
