Conformational Analysis of the Major DNA Adduct Derived from the

The heterocyclic amine 2-amino-3-methylimidazo[4,5-f]quinoline (IQ) is one of a number of carcinogens found in barbecued meat and fish. It is mutageni...
0 downloads 0 Views 276KB Size
Chem. Res. Toxicol. 1999, 12, 895-905

895

Conformational Analysis of the Major DNA Adduct Derived from the Food Mutagen 2-Amino-3-methylimidazo[4,5-f]quinoline Xiangyang Wu,† Robert Shapiro,‡ and Suse Broyde*,§ Departments of Physics, Chemistry, and Biology, New York University, New York, New York 10003 Received June 16, 1999

The heterocyclic amine 2-amino-3-methylimidazo[4,5-f]quinoline (IQ) is one of a number of carcinogens found in barbecued meat and fish. It is mutagenic in bacterial and mammalian assays and induces tumors in mammals. IQ is biochemically activated to a derivative which reacts with DNA to form a major covalent adduct at carbon 8 of guanine. This adduct may deform the DNA and consequently cause a mutation, which may be responsible for initiating IQ’s carcinogenicity. Atomic resolution structures of the IQ-damaged DNA are not yet available experimentally. We have carried out an extensive molecular mechanics energy minimization search to locate feasible structures for the major IQ-DNA adduct in the representative sequence d(5′-G1-G2-C3-G4-C5-C6-A7-3′)‚d(5′-T8-G9-G10-C11-G12-C13-C14-3′) with IQ modification at G4; this contains the GGCGCC mutational hotspot sequence known as NarI. The molecular mechanics program AMBER 5.0 with the force field of Cornell et al. [(1995) J. Am. Chem. Soc. 117, 5179-5197] was employed, including explicit Na+ counterions and an implicit treatment for solvation. However, key parameters, the partial charges, bond lengths, bond angles, and dihedral parameters of the modified residue, are not available in the AMBER database. We carefully parametrized the force field, created 800 starting conformations which uniformly sampled at 18° intervals each of the three flexible torsion angles that govern the IQ-DNA orientation, and minimized their energy. A conformational mix of structural types, including major groove, minor groove, and base-displaced intercalated carcinogen positions, was generated. This mixture may be related to the diversity of mutational outcomes induced by IQ.

Introduction 2-Amino-3-methylimidazo[4,5-f]quinoline (IQ)1 is a prominent member of a group of potent mutagens and carcinogens, the heterocyclic amines, that are formed during the cooking of protein-rich foods, such as meats and fish (1-3), and it has also been detected as a component of cigarette smoke (4). It exhibits mutagenic activity in bacterial (5) and mammalian (6) assays, and is a potent carcinogen, inducing tumors in numerous organs of rodents and in the liver of monkeys (7, 8). Although IQ is less prevalent than many of the other members of the group, it compensates to some extent by its greater biological activity (9, 10). Because of the widespread exposure of humans to these heterocyclic amines through food and other sources, they are considered to be likely causative agents for human cancer (11, 12). Although mutagenesis and especially carcinogenesis are complex, multistep processes, it is widely accepted that polycyclic aromatic chemicals, after metabolic activation, initiate them by reaction with DNA (13, 14). IQ, for example, forms a major product by substitution at * To whom correspondence should be addressed. Phone: (212) 9988231. Fax: (212) 995-4015. E-mail: [email protected]. † Department of Physics. ‡ Department of Chemistry. § Department of Biology. 1 Abbreviations: IQ, 2-amino-3-methylimidazo[4,5-f]quinoline; AF, 2-aminofluorene; AAF, 2-acetylaminofluorene.

′ ′

Figure 1. Structure and numbering scheme of the IQ-modified dG major adduct. Torsion angle definitions are as follows: χ, O4′-C1′-N9-C4; R′, N9-C8-N(IQ)-C2(IQ); and β′, C8N(IQ)-C2(IQ)-N3(IQ).

C8 of guanine (Figure 1) as well as a minor one by reaction at N2 of guanine (15-21). These adducts can cause mutations during DNA replication. If the mutation occurs in a protein coding region, it may cause the synthesis of aberrant proteins, which can lead to cancer if the protein is one that is involved in cell cycle control, such as an oncogene or tumor suppressor gene product (13). The structure and conformation of a carcinogen-DNA adduct is believed to play an important role in determining whether a given DNA adduct will cause mutations during replication. In addition, repair enzymes may

10.1021/tx990108w CCC: $18.00 © 1999 American Chemical Society Published on Web 09/30/1999

896

Chem. Res. Toxicol., Vol. 12, No. 10, 1999

recognize and repair the damage or fail to do so, depending on the adduct conformation. The conformation adopted by an adduct is influenced by a number of factors, including the DNA sequence context. The same carcinogen-DNA adduct may be much more mutagenic in a hotspot sequence context than a coldspot context (22-29), and differing conformations may be responsible for this phenomenon (24). Atomic resolution structures of DNA adducts in various sequence contexts are needed to elucidate the structural reasons for their propensity to cause replication errors, and to gain an understanding of the likelihood for repair. Unfortunately, atomic resolution DNA adduct structures are very hard to obtain experimentally, whether by high-resolution NMR in solution or by X-ray diffraction analysis of crystals. Pure, site-specific adducts are difficult to synthesize, especially in the amounts (several milligrams) needed for these techniques. IQ is one of the carcinogens whose DNA adducts have not yet been prepared in sufficient quantities for these purposes. In such cases, computer simulation can play an important role in delineating feasible conformations. Since quantum mechanical calculations are not possible for systems of this size, the classical approach of molecular mechanics and molecular dynamics must be used to simulate their structures. This approach requires a well-parametrized force field for the system being investigated, and a reasonable technique for addressing the multiple-minimum problem with adequate, unbiased, uniform sampling of the potential energy surface. The work presented here is focused on the structure of the major IQ-DNA adduct in the DNA sequence d(5′G1-G2-C3-G4-C5-C6-A7-3′)‚d(5′-T8-G9-G10-C11-G12-C13C14-3′) with IQ modification at G4. Recent bacterial mutagenicity studies (30) have revealed that this sequence, known as NarI, is a mutational hotspot for the deletion of a G‚C base pair upon modification with an activated form of IQ. The sequence had earlier been shown to be a hotspot for the same deletion by the carcinogen 2-acetylaminofluorene (AAF) (31, 32). Although the overall mutagenic efficiency of the two adducts was similar, IQ mutagenesis was independent of the bacterial SOS repair system, while AAF mutagenesis was SOS-dependent. In this respect, IQ mutagenesis resembled mutagenesis by 2-aminofluorene (AF), which however was much less potent at producing -2 deletions in this sequence (33). It was suggested (30) that the SOS response is required by adducts, such as those of AAF, which strongly interfere with translesion synthesis by the polymerase. In the case of mutagenesis by adducts such as AF, and possibly IQ, replication is not blocked as severely (30). Although a number of high-resolution NMR solution studies have been carried out with AF-modified DNA in various sequence contexts, reviewed by Patel et al. (34), little is known about IQ-modified DNA. We have started by considering a double helix containing the NarI sequence. We employ the molecular mechanics program AMBER 5.0 (35) with the force field of Cornell et al. (36) which has been demonstrated to reproduce well the structures of DNA [reviewed by Darden (37)]. To investigate the IQ-DNA adduct, we carefully parametrized the force field to obtain partial charge, bond length, bond angle, and torsional parameters of the modified residue that are consistent with the rest of the force field. Then, we created 800 different starting conformations, which

Wu et al. Table 1. Conformations Used in Partial Charge Computation χ (deg)

R′ (deg)

β′ (deg)

χ (deg)

R′ (deg)

β′ (deg)

173 243 227

218 127 189

80 207 195

60 67 67

205 193 55

187 226 186

uniformly sampled in combination at 18° intervals the flexible torsion angles R′ and β′, in two domains of the glycosidic torsion angle χ, syn and anti (see Figure 1). χ, R′, and β′ govern the orientation of the carcinogen with respect to the DNA. The energy of each structure was minimized, and the resultant structures were sorted into different families according to their structural similarities in the R′, β′, and χ domains. Interestingly, a conformational mix of structural types is suggested from the results which may be related to the diversity of mutational outcomes that IQ has been observed to induce.

Methods Force Field. We employed AMBER 5.0 (35) with the force field of Cornell et al. (36) using the PARM96.DAT parameters for the DNA. AMBER carries out energy minimizations in Cartesian space. The x, y, and z coordinates of all atoms are the flexible parameters, and hence, all bond lengths, bond angles, and dihedral angles are mobile; thus, no special treatment is needed to permit the sugar puckers to vary. Parametrization. (1) Partial Charges. We obtained partial charge sets from a number of different conformations to avoid as much as possible bias in the charge set that is conformationdependent, using six different conformations which had been obtained in preliminary energy minimization studies by Y. Li (38) in our laboratory (Table 1). These included syn and anti orientations for the glycosidic torsion angle χ. Hartree-Fock calculations with basis set 6-31G* in GAUSSIAN 94 (39) were used to calculate the electrostatic potential at a large number of grid points (∼2500 points) around the modified nucleoside of Figure 1. The least-squares charge fitting algorithm, RESP (40), provided with AMBER 5.0 was then used to fit the charge to each atomic center in the molecule. Partial charges were then averaged and normalized to avoid charge imbalance (see the Appendix). (2) Torsional Parameters. In the AMBER force field, the dihedral energy term is

Edihedrals )



(Vn/2)[1 + cos(nφ - γ)]

dihedrals

where φ is the variable torsion angle. The rotation barrier height Vn, the phase angle γ, and the periodicity n define the torsional potential energy function. Each bonded series of atoms (I-J-K-L) must have at least one set of these dihedral parameters in the force field. For our IQ adduct, the important dihedral parameters for R′ and β′ at the linkage between IQ and guanine, i.e., the dihedral parameters for the linkage with AMBER atom types X-N2-CK-X (36) which include the rotation barrier, the dihedral angle, and the periodicity, are missing. The adduct of Figure 1 with methyl replacing sugar was used to compute these quantities. We generated 24 structures in which the dihedral angle β′ was uniformly distributed in its 360° conformational space at 15° intervals. We chose to investigate β′ because it is further removed from the methyl group used to model the sugar. χ and R′ were maintained at 173° and 218°, respectively (Table 1), an energy minimum obtained in a preliminary study in our laboratory (38). We then carried out point ab initio quantum mechanical calculations (MP2/6-31G*) for each structure (without geometry optimization) using GAUSSIAN 94 (39). Hartree-Fock calculations with basis set 6-31G*, followed by a Moller-Plesset energy correction truncated at second-order (MP2 method), were employed. An

IQ-DNA Adduct Conformation Table 2. Starting Structures for Energy Minimization torsion angle

angle value (deg)

χ R′ β′

60 (syn), 251 (anti) 0, 18, 36, 54, 72, ..., 342 (0-360° at 18° intervals) 0, 18, 36, 54, 72, ..., 342 (0-360° at 18° intervals)

energy profile was derived. The lowest energy that was computed was assigned a value of zero, and other energies are given relative to it, since we are interested only in the conformational component of the total energy. Then a similar series of computations was carried out using AMBER 5.0 with the dihedral term set to zero, which produced a second curve. We subtract this curve from the quantum curve. The difference curve is cosinelike, and the amplitude, phase, and periodicity supply the needed force field terms. The philosophy underlying this approach is that the dihedral force field term corrects the other nonbonded terms so that the sum is in agreement with the quantum mechanical relative energies. In this work, we followed the approach recommended by the AMBER development team (www.amber.ucsf.edu/amber/newparams.html). (3) Bond Length and Bond Angle Equilibrium Values and Force Constants. To obtain equilibrium bond lengths and angles, we carried out geometry optimization for the IQ-modified nucleoside (Figure 1) with GAUSSIAN 94 using the HartreeFock method with basis set STO-3G. The force constants for the carcinogen bond lengths and bond angles were assigned by employing values in the force field database for chemically similar bonds and angles. Counterions and Solvation. Twelve explicit sodium ions for modeling counterion condensation were initially placed 5.0 Å from the phosphorus atom along the pendant phosphate oxygen bisector. Hydration of these counterions was modeled by an increase in the van der Waals radius to 5 Å and a van der Waals energy of 0.1 kcal/mol in the Lennard-Jones potential. In addition, we implemented and employed the Hingerty sigmoidal distance-dependent dielectric function (41) into AMBER 5.0. This function mimics implicitly the solvent environment in the Coulombic term of the force field. Starting Conformations. The starting structure of the DNA was an energy-minimized ideal B-form structure (42). To survey the potential energy surface, we created 800 different starting structures, which uniformly sampled at 18° intervals in combination the possible rotations about the two flexible torsion angles R′ and β′ in two domains of χ, syn and anti. A torsion driver (a program which generates Cartesian coordinates for the IQ-DNA adduct from selected χ, R′, and β′ torsion angle values in the frame of reference of the stationary B-DNA) was employed for this purpose. The starting structures are given in Table 2. Energy Minimization. We employed 100 steps using the steepest-descent algorithm to quickly remove collisions, followed by conjugate-gradient minimization. A structure was deemed converged if the energy difference in successive steps was less than 0.005 kcal/mol. If convergence was not achieved in 14 000 steps, the trial was ended. Modeling and visualization were carried out with INSIGHTII, from Molecular Simulations, Inc. Helix parameters were computed with the program developed by M. Babcock, E. P. D. Pednault, and W. K. Olson (43). Computations were carried out on Cray Supercomputers at the Department of Energy’s National Energy Research Supercomputer Center, at the National Science Foundation’s San Diego Supercomputer Center, and on our own SGI workstations.

Results Force Field Parameters. (1) Partial Charges. RESP partial charges for the anti family are listed in Table S1 (Supporting Information) and for the syn family in Table S2 (Supporting Information). Average RESP partial charges for the anti and syn families are listed

Chem. Res. Toxicol., Vol. 12, No. 10, 1999 897 Table 3. RESP Partial Charges after Modification To Account for the Residuala 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56

atom

AMBER atom type

RESP charge

P O1P O2P O5′ C5′ H5′1 H5′2 C4′ H4′ O1′ C1′ H1′ N9 C8 N (amine) H10 C2 (IQ) N10 (IQ) C10A (IQ) C9A (IQ) C9 (IQ) H23 (IQ) C8 (IQ) H22 (IQ) C7 (IQ) H21 (IQ) N6 (IQ) C5A (IQ) C5 (IQ) H18 (IQ) C4 (IQ) H19 (IQ) C3A (IQ) N3 (IQ) C (methyl) H24A H24B H24C N7 C5 C6 O6 N1 H1 C2 N2 HN1 HN2 N3 C4 C3′ H3′ C2′ H2′1 H2′2 O3′

P O2 O2 OH CT H1 H1 CT H1 OS CT H2 N* CK N2 H CK NB CB CA CA HA CA HA CA HA NC CA CA HA CA HA CB N* CT H1 H1 H1 NB CB C O NA H CA N2 H H NC CB CT H1 CT HC HC OH

1.1662 -0.7760 -0.7760 -0.6747 0.1107 0.0520 0.0711 0.1109 0.1492 -0.3096 0.0343 0.1298 0.0140 0.3380 -0.3767 0.2816 0.3933 -0.4837 0.1286 -0.0323 -0.0207 0.1468 -0.3121 0.1521 0.2507 0.0981 -0.6087 0.4933 -0.2858 0.1758 -0.2501 0.1889 -0.0105 -0.0009 -0.1474 0.0927 0.0749 0.1100 -0.4978 0.1327 0.4988 -0.5487 -0.4283 0.3429 0.6750 -0.9183 0.4417 0.4434 -0.5175 0.1169 0.2488 0.0284 -0.0503 0.0290 0.0537 -0.7282

a Partial charges for P, O1P, and O2P were taken from Cornell et al. (36).

in Table S3. The difference in the average RESP charges between the syn and anti conformations is relatively small, although individual conformer differences can be much higher at certain atoms (Tables S1 and S2 of the Supporting Information). The sum of the partial charges in the nucleoside of Figure 1 (without hydrogens at the 3′- and 5′-hydroxyls) must equal that of the unmodified nucleoside in the AMBER database to avoid charge imbalance. To achieve this, the difference, called the residual, must be distributed among the nucleoside atoms. The modification should affect the partial charge of each atom as little as possible. To do this, we distribute the residual proportionally to the charges of each of the

898

Chem. Res. Toxicol., Vol. 12, No. 10, 1999

Wu et al.

Figure 2. Torsion angle β′ parametrization. The points are the differences between the AMBER force field calculation without the dihedral term for β′ and the quantum mechanical calculation. The dashed curve is the cosine fitting curve. Table 4. Parameters Added to the Force Field Bond Parameters AMBER bond type

force constant Kr (kcal mol-1Å-2)

equilibrium length req (Å)

CK-N2

484.5

1.338

Angle Parameters AMBER angle type

force constant Kθ (kcal mol-1 rad-2)

equilibrium angle θeq (deg)

N*-CK-N2 CK-N2-H CK-N2-CK N2-CK-NB CA-CA-NC CA-NC-CA HA-CA-NC CA-CB-N*

70 35 50 70 70 70 50 70

115.38 117.05 123.18 128.68 122.03 119.88 116.73 133.02

Torsional Parameters AMBER torsion type

torsion barrier Vn/2 (kcal/mol)

phase angle γ (deg)

periodicity n

X-CK-N2-X

2.25

173

2

atoms according to a formula devised in our laboratory (38) (see the Appendix). The partial charges after modification are listed in Table 3. (2) Torsional Parameters. Quantum mechanical calculations described in the Methods produced an energy profile as a function of β′. Structures with β′ values of 0 ( 60° were extremely high-energy with steric clashes, and were eliminated from the data set. A similar set of calculations with AMBER 5.0 without the dihedral term yielded a second curve. The conformational energy difference between these as a function of the dihedral angle β′ and the fitting curve is shown in Figure 2. From this figure, we obtained the parameters of interest: the rotation barrier for β′, Vn, equals 4.5 kcal/mol, the periodicity n equals 2, and the phase angle γ equals 173°. This result is given in Table 4. (3) Bond and Angle Parameters. Table 4 lists computed equilibrium bond length and angle values for the IQ as well as the respective force constants, which were assigned by analogy to chemically similar atom types in the AMBER database. AMBER atom type assignments are given in Table 3.

Figure 3. Number of conformers in 1 kcal/mol energy bins: (A) anti domain and (B) syn domain.

Energy Minimization Results. Five hundred fortytwo local energy minima were achieved in our energy minimization searches. Of 400 starting structures in the anti domain of χ, 337 structures converged according to our criterion, while 63 structures did not converge after 14 000 steps; of 400 starting structures in the syn domain, 205 structures converged, while 195 structures did not converge after 14 000 steps. The lowest-energy structure is located in the syn domain, and its associated energy is -363.0 kcal/mol. There are 22 structures with energies within 3.0 kcal/mol of that of the lowest-energy structure (19 in the syn domain and three in the anti domain), 56 with energies within 5.0 kcal/mol of that of the lowest-energy structure (40 in the syn domain and 16 in the anti domain), and 211 structures with energies within 10.0 kcal/mol of that of the lowest-energy structure (59 in the syn domain and 152 in the anti domain). The number of conformers versus energy in 1 kcal/mol energy bins is shown in Figure 3. The full data set that was obtained is given in the Supporting Information, Tables S4 (anti starting domain of glycosidic torsion angle χ) and S5 (syn starting domain of glycosidic torsion angle χ). The structures were sorted into different families according to the key torsion angles χ, R′, and β′ which govern the orientation of the carcinogen with respect to the DNA. Structures whose χ, R′, and β′ values are all within 20° of those of the lowest-energy variant are grouped into the same structural family. The lowestenergy variants of each family in both anti and syn

IQ-DNA Adduct Conformation

Chem. Res. Toxicol., Vol. 12, No. 10, 1999 899

Table 5. Structural Families Computed for the IQ-dG Adduct no.

IQ positiona

1 2b

minor groove, 3′ major groove, 5′ distorted minor groove, 5′ major groove major groove, 5′ distorted major groove minor groove, 5′ major groove, 3′ minor groove, 3′ major groove, 5′ distorted major groove, 3′ major groove, 3′ major groove, 3′ major groove, 5′ distorted major groove, 3′ base-displaced intercalated

3 4 5c 6 7d 8 9e 10c 11f 12f 13f 14c 15 16

∆E (kcal/mol)

no. in family

χ, R′, β′ (deg)

0.0 1.6

32 1

65 (syn), 207, 173 165 (anti), 194, 171

1.7 2.5 3.9

42 46 1

64 (syn), 48, 187 235 (anti), 192, 192 120 (anti), 195, 236

4.4 4.4 4.6 5.3 5.4

32 2 57 10 16

228 (anti), 159, 299 36 (syn), 351, 204 247 (anti), 109, 183 68 (syn), 189, 206 172 (anti), 245, 68

6.7 7.5 8.3 9.1

3 1 2 4

244 (anti), 111, 162 267 (anti), 110, 197 246 (anti), 112, 138 157 (anti), 201, 231

9.4 10.5

60 3

247 (anti), 119, 70 71 (syn), 60, 149

a 3′ and 5′ designate the IQ long axis orientation along the modified strand. Anti domain glycosidic torsion angle χ is in the range of 90-270°; syn domain glycosidic torsion angle χ is in the range of -90° (270°) to 90°. When not designated, the IQ long axis is approximately perpendicular to helix axis. b Minor groove of G1‚C14 and G2‚C13 close to IQ due to distortion at C3‚G12. c Variant of 2. d Variant of 3. e Variant of 1. f Variant of 8.

domains up to a ∆E of about 10 kcal/mol are described in Table 5. We consider conformers with energies of up to about 10 kcal/mol to be of interest in relation to mutagenesis because a mutagenic structure may be very rare and yet be biologically highly significant. For this reason, we have not considered additional energy corrections involving non-Coulombic (hydrophobic) solvation effects and conformational entropies which can amount to about 2 kcal/mol (44-46). The number of conformers in a given family may be a rough indicator of relative conformational entropy (Table 5). We also bear in mind that specific interactions with cellular proteins are likely to influence the structures and their energetics. Representative structural types are illustrated in Figures 4-6. The structures described in Table 5 fall into specific IQ orientational families, which are named for the location of the IQ residue with respect to the double helix: minor groove, major groove, major groove distorted, and base-displaced intercalated. The minor groove structures all contain syn IQ-modified guanine; the guanine remains stacked within the helix. The IQ is nicely sandwiched into the minor groove. These structures all fit the curve of the N1 to C9 edge of IQ (Figure 1) snugly to the curve of the DNA minor groove. The IQ-modified G4‚C11 base pair is stabilized by one non-Watson-Crick hydrogen bond: O6 of G4 with H4N4 of C11 (Table 5, structures 1 and 7) or N7 of G4 with H4N4 of C11 (Table 5, structures 3 and 9). In the case of structure 3 (Table 5), there is also a hydrogen bond between N1 of IQ and H2N2 of G12. The essential stabilization of these structures clearly stems from the favorable interactions between the faces of the IQ with the walls of the minor groove, burying much of the hydrophobic aromatic domain of IQ within the groove. The methyl group faces away from the minor groove interior, protruding somewhat, but achieves maximal interaction with DNA by interacting with the groove.

Both 3′- and 5′-orientations of the IQ long axis with respect to the modified strand are found, with the 3′-orientation providing the lowest-energy structure, which is shown in a space filling view in Figure 6. The 5′- or 3′-orientation is governed by the torsion angle R′ which is in the range of 180-207° for the 3′-orientation and the range of 351° (-9°) to 48° for the 5′-orientation. The torsion angle β′ largely governs the flexibility relating to rotation about the IQ long axis and is in the range of 173-204°, which places the 9-10 edge into the groove and the 4-5-6 edge away. The normal major groove structures all contain Watson-Crick base pairing and undistorted B-DNA structures, with the IQ-modified guanine glycosidic torsion angle in the usual B-DNA anti domain (∼230-250°) (47). The large major groove can accommodate flexibility in the IQ orientation. Consequently, the torsion angle β′ adopts values ranging from 70° to 299°, corresponding to rotation about the IQ long axis. The torsion angle R′ largely governs the IQ long axis orientation in relation to the helix axis. Two conformational domains are found, one with the IQ long axis nearly parallel to the helix axis and a second where it is nearly perpendicular. In the parallel structures, the IQ’s long axis is 3′-directed along the modified strand, and R′ is in the range of 109-119°. In the long axis perpendicular structures, the R′ range is 159-192°. In all of these major groove structures, the IQ is exposed to solvent, which is the price paid for retaining the undistorted double helix. In the distorted major groove structures, by contrast, the IQ achieves significant burial of one aromatic face, but at the expense of an abnormal anti region [120-172° compared to normal values of ∼230-260° (47)] for the IQ-modified guanine glycosidic torsion so that the G4‚ C11 base pair is severely distorted (81° buckle, -54° propeller twist, 39° tilt, and 96° roll). In addition, the C3‚ G12 base pair is also highly distorted in tilt and roll (0° buckle, -3° propeller twist, -73° tilt, and 41° roll), and the normal Watson-Crick hydrogen bond between O6 of G12 and H4N4 of C3 does not form. Instead, there is a weak cross-strand hydrogen bond between N4 of C3 and H4N4 of C11 [N4(C3)-N4(C11) distance of 3.34 Å and N4(C3)-H4(C11)-N4(C11) angle of 150.1°], which might be relevant to a possible role in mutagenicity for this conformer (see the Discussion). These distortions permit the IQ to orient 5′ along the modified strand, into the minor groove region of the G1‚C14 and G2‚C13 base pairs, contacting one minor groove wall of residues C13 and C14 on the partner strand. As in the normal minor groove structures, the fit of the N1-C9 edge of the IQ into the curve of the groove is a stabilizing feature. The external methyl achieves some stabilizing interaction with one groove wall. The last structural family, base-displaced intercalated, places the IQ partially into an intercalated position within the double helix, with its distal ring protruding into the minor groove. There is some stacking between IQ and the neighboring C3‚G12 base pair as well as with C5 of the C5‚G10 pair. The IQ-modified guanine is displaced into the major groove, to make room for the IQ. The modified guanine glycosidic torsion angle is syn. There are no hydrogen bonds between IQ-modified G4 and its partner C11, but the amino group hydrogens of G4 bond with pendant oxygens of the phosphate groups that separate residues 2 and 3, and 3 and 4.

900

Chem. Res. Toxicol., Vol. 12, No. 10, 1999

Wu et al.

Figure 4. Representative structural types in stereoview. Anti domain. A (structure 2 in Table 5), B (structure 4 in Table 5), C (structure 6 in Table 5), and D (structure 8 in Table 5). Hydrogens have been deleted for clarity.

The DNA duplexes remained within observed B-DNA ranges (47) in torsion angles and sugar pseudorotation parameters (48). Coordinates of all structures are available electronically in the Supporting Information.

Discussion Comparison of IQ- and AF-Modified Double Helices. The carcinogen IQ is roughly comparable in its bulk to 2-aminofluorene, with both containing a fused three-ring aromatic system, with two six-membered rings and one five-membered ring. IQ differs from AF in that it contains three nitrogens within its ring system and an exocyclic methyl group. AF-modified DNA duplexes have been the subject of considerably more study, both experimental and theoretical. Experimental high-resolution NMR studies have been reviewed by Patel et al. (34). Most relevant is a combined experimental and theoretical study of a dodecamer duplex which contained within it the sequence used in the study presented here. That modified duplex exhibited a conformational mix which exchanged slowly on the NMR time scale. The mixture contained about equal amounts of a base-displaced intercalated structure (guanine syn) and one with gua-

nine in the major groove (guanine anti) (49, 50). Conformational mixes of these two states were also encountered in NMR studies with the adduct in other base sequences (49-54). The proportion of the two forms was exquisitely dependent on the neighboring base sequence context. In a recent theoretical study (with no experimental input) with two other sequences, we found minor and major groove structures approximately competitive in energy, with base-displaced intercalated structures being 3-4 kcal/mol less stable (55). A minor groove structure was also encountered in an NMR study of a duplex containing a mismatched A or I opposite the modified G (56, 57). One interesting difference between the AF and the IQ structures concerns the observed presence of rotamers, conformers rapidly interconverting via a ∼180° rotation about the long axis (torsion angle β′) in AF major groove and base-displaced intercalated conformers (34). The substituent nonplanar methyl group on IQ might make such rotated structures less likely in the base-displaced intercalated conformation of IQ than in the flat AF. Rotated variants are several kilocalories per mole higher in energy than base-displaced intercalated structure 16 of Table 5 (Table S4 of the Supporting Information).

IQ-DNA Adduct Conformation

Chem. Res. Toxicol., Vol. 12, No. 10, 1999 901

Figure 5. Representative structural types in stereoview. Syn domain. A (structure 1 in Table 5), B (structure 3 in Table 5), and C (structure 16 in Table 5). Hydrogens have been deleted for clarity.

Figure 6. Lowest-energy conformation as a space filling model (structure 1 in Table 5). IQ is in B-DNA minor groove. All DNA atoms are white except phosphorus which is magenta. The IQ atoms are colored as follows: carbon, green; hydrogen, white; and nitrogen, blue.

From the study presented here, we infer that IQ shares with AF the ability to provide a conformational mix. In the IQ case, the most favored structure in the sequence being investigated contained syn guanine with the IQ in the minor groove and directed 3′ along the modified strand. Minor groove NMR solution structures for AF modification also contained a 3′-oriented carcinogen (56, 57). A preference for syn guanine was demonstrated earlier at the nucleoside level for IQ modification by NMR

(16). Energetically feasible structures with anti guanine were also computed. These consisted of structures with a normal B-DNA double helix, with the carcinogen in the major groove, as in the AF structures, or in a distorted duplex where the IQ partly contacts the minor groove. The most stable base-displaced intercalated structure found in the work presented here lies more than 10 kcal/ mol above the minor groove structure which furnished the lowest-energy structure. Perhaps this structural type is less favored for IQ than for AF in this sequence because of IQ’s exocyclic methyl group which may impede insertion into the helix, although it is also possible that the lowest-energy variant of this structural type was not located due to the multiple-minimum problem. Structures of this type are of interest in relation to mutagenesis and repair, and the complex context of the cell might stabilize it more (see below). Relevance to IQ Mutagenesis. Different studies have shown a variety of mutagenic effects for IQ, although none of these involved site-specific modification with a defined adduct. The most spectacular mutagenicity reported for IQ occurs in Salmonella typhimurium strains TA98 and 1535 (Ames assay). These systems have been designed to assess the deletion of a G‚C base pair from a sequence containing a number of consecutive G‚C pairs. The mutagenic potency of IQ in this assay was about 7800 times that of 4-aminobiphenyl and 600 times that of AAF (58). In studies using a variety of plasmids in Escherichia coli, IQ was also found to be effective in inducing the loss of a G from a run of G’s, and of a G‚C pair from a (G‚C)n sequence, but its efficiency was found to be comparable to rather than superior to that of AAF (30). Similarly, Watanabe and Ohta (59) found that 99.5% of mutations in another E. coli gene were frameshift,

902

Chem. Res. Toxicol., Vol. 12, No. 10, 1999

especially involving the loss of a C‚G pair. The mutagenic potency of IQ was reduced by a factor of 100 (compared to the results with strains TA98 and 1538) when it was assayed in S. typhimurium strains TA100 and 1535. The latter strains are employed to detect point mutations. The changes induced by IQ were predominantly (84%) G‚C to T‚A transversions (10). The same transversions also predominated in a study with human fibroblasts (20). Other studies have given contrasting results, however, with G‚C to A‚T transitions predominating in a yeast gene within E. coli (60) and a mixture of deletions and various point mutations in mammalian systems (21, 61). The analyses of mutations in tumor-related genes from rats treated with IQ have revealed G‚C to T‚A and G‚C to C‚G transversions, as well as the loss of G from a run of G’s (62, 63). The above differences can be ascribed to a number of factors: (1) the sequence context of the target gene, (2) differences in metabolic activation between cell types, (3) level of modification, (4) nature of the adduct, a guanine C8 major adduct (presumably the most important contributor) or a guanine N2 minor adduct, (5) the different DNA damage-processing enzymes used by prokaryotic and eukaryotic cells, and (6) alternative adduct conformations in different sequence contexts. Such factors have been considered to influence mutagenic outcome in this and other polycyclic aromatic carcinogen-DNA adducts (23, 24, 26-30, 32, 58-63). In addition, the polymerase varies from one system to another, and it has been shown that differing polymerases may induce differing mutagenic outcomes (64-67). A fuller study of the mutagenic effect of IQ requires understanding the conformations of the modified DNA in a complex with polymerase at a replication fork. This study would include the effect of the modification in stabilizing a bulged-out DNA intermediate, if the mechanism of frameshift mutagenesis at repetitive sequences is presumed to proceed through a slippage mechanism (66, 68-72). Nonetheless, our results with doublestranded DNA can offer some insights. They have located via an unbiased, systematic search a number of structural types, including those with the carcinogen in the major groove, in the minor groove, and in a basedisplaced intercalated position, suggesting the possibility of a conformational equilibrium. Conformational heterogeneity has been proposed as an origin for the different kinds of mutations engendered by the same stereochemically defined benzo[a]pyrene diol epoxide adduct by Loechler and co-workers (23, 24, 7376), and a similar phenomenon may well play a role in the variety of mutagenic outcomes engendered by IQ, AF, and other polycyclic aromatic carcinogens. Minor groove conformers with abnormal syn-modified guanine are plausible candidates as mutagenic structures. Conformers of this type are stable in specific sequence contexts in the AF case, namely, in the case of mismatches at the lesion site (56, 57). A syn guanine minor groove-positioned structure has also been previously computed for the guanine C8 adduct of the related food mutagen 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) in a duplex containing a mismatched adenine opposite the modified guanine (77). Both basedisplaced intercalated and minor groove structures contain the syn-modified guanines when the modification is at C8, and interestingly, this pair of conformers appears

Wu et al.

to be easily interconvertible since torsion angles χ, R′, and β′ can be very close (compare conformers 3 and 16 in Table 5). Base-displaced intercalated conformers have been hypothesized to play a role in mutagenesis (74-76, 78). Structures of this type found in our current study had energies of ∼10 kcal/mol, but they could possibly be more prominent in another duplex sequence, or at a model replication fork. The stability of such conformers is strongly governed by sequence context in the case of AF adducts to guanine C8 (34) and trans-anti-benzo[a]pyrene diol epoxide adducts to guanine N2 (25), and this could also be the case for IQ adducts, where they appear to be less favored in the duplex. In particular, model replication fork structures, with carcinogen modification at the first unpartnered base in the single-stranded region following the single strand-double strand junction, adopt base-displaced intercalated conformations in both the AF (79) and (+)-trans-anti-benzo[a]pyrene diol epoxide adducts (80) in solution as deduced from high-resolution NMR studies. Our earlier computations have also suggested that a conformation of this type stabilizes a bulged, slipped mutagenic intermediate in the case of AF and AAF modification (81). The base-displaced intercalated conformation may be more stable in these replication-relevant contexts compared to a duplex, since Watson-Crick hydrogen bonding is disrupted. The distorted major groove structure, unique so far to modification by IQ, must also be considered in connection with mutagenesis. In particular, the presence of a crossstrand hydrogen bond between N4 of C3 and H4N4 of C11, together with the distorted G4‚C11 pairing, suggests the possibility of slippage at a replication fork, to provide a -1 deletion by the following mechanism involving a C‚C mismatch:

Similar types of cross-strand hydrogen bonds, involving interactions between mismatched misaligned bases with attendant distortion of the normal pairing, have been observed in a crystal structure of the unmodified NarI sequence, and it is speculated that there is a relationship with the mutation hotspot properties of this sequence (82). Recently, Timsit (83) proposed slippage mechanisms akin to the one described above that are founded on crossstrand hydrogen bonds observed in crystal structures. Site-specific IQ modification studies involving mutagenesis and repair, as well as experimental structural investigations, will be helpful in further investigating these hypotheses.

Conclusion Our unbiased, systematic conformational search has located preferred structures with the modified guanine syn and the IQ carcinogen in the minor groove, as well as structures with the IQ in the major groove and guanine anti in a DNA duplex containing the NarI mutational hotspot sequence. In addition, a higher-

IQ-DNA Adduct Conformation

Chem. Res. Toxicol., Vol. 12, No. 10, 1999 903

energy syn guanine base-displaced intercalated structure was found. The results suggest conformational heterogeneity whose population balance could be governed by sequence and other environmental contexts, both in solution and in the cell. This situation may relate to the diverse mutagenic outcomes induced by IQ. The syn guanine and the distorted major groove structures are plausible candidates in relation to mutagenic replication.

Appendix The sum of the partial charges for the fragment being replaced, the nucleoside, is -0.5942 in the force field of Cornell et al. (36). Our combination of charges does not equal -0.5942. The difference between our summed charges and -0.5942 must be distributed about the adducted nucleoside so a charge difference at the modified nucleoside can be avoided. The modification should affect the partial charge of each atom as little as possible. To do this, we keep the average of the magnitude of the sum of positive charges and the sum of negative charges while modifying their difference so that it equals -0.5942, and then distribute the change proportionally to the charge of each atom. Steps to accomplish the goal are as follows. Determine R, the sum of the partial charges, qi, on the unmodified fragment.

R)

∑i qi

Determine the average in magnitude of the sum of positive charges, S+, and the sum of negative charges, S-, on the modified fragment:

S+before ) S-before )

∑ qi

qi>0

∑ abs(qi)

qi