Conformation of Amine-Modified DNA: 2-Aminofluorene-and 2

Monophosphates with All Possible Nearest Neighbors. A Comparison of Search and Optimization Methods. Robert Shapiro,* *-* Dany Sidawi,* Yong-Su Miao,*...
0 downloads 0 Views 3MB Size
Chem. Res. Toxicol. 1994, 7,239-253

239

Conformation of Amine-Modified DNA 2-Aminofluoreneand 2- (Acety1amino)fluorene-Modified Deoxydinucleoside Monophosphates with All Possible Nearest Neighbors. A Comparison of Search and Optimization Methods Robert Shapiro,’f+Dany Sidawi,? Yong-Su Miao,? Brian E. Hingerty,i Kevin E. Schmidt,# Jules Moskowitz,+and Suse Broydell Chemistry and Biology Departments, New York University, New York, New York 10003,Health Sciences Research Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831,and Physics and Astronomy Department, Arizona State University, Tempe, Arizona 85287 Received October 18,1993

Although a significant part of the replication fork exists as single-stranded DNA, little is known about the effect of carcinogens and mutagens on single-strand conformation. Largescale conformational searches with potential energy minimization, using the torsion angle space molecular mechanics program DUPLEX, were employed to explore the conformation of all 16 deoxydinucleoside monophosphates bearing 2-aminofluorene (AF) or 2-(acetylamino)fluorene (AAF) modification on guanine. We have thus examined the effect of 3’ versus 5‘ modification, the presence or absence of the acetyl group, and the effect of four different neighbors in each case. The principal effect of the acetyl group appeared to be the destabilization of anti (and, to a lesser degree, borderline anti) conformations for modified guanine. This mattered little in the 5’-substituted dimers, where one conformational type predominated in the low-energy structures for the adducts of both AAF and AF: It was right-handed, withsyn-guanine, imperfect base-base stacking, and fluorene to 3’-sugar contacts. Greater divergence was seen in the 3’substituted series. The AAF-substituted 3’-adducts primarily displayed good base-fluorene stacking, with syn-guanine in contact with the 5’-sugar. The AF-substituted 3’-adducts displayed a variety of structures which included base-base and carcinogen-base stacked forms. Two novel forms were encountered [global minima for d(ApG-AF) and d(GpGAF)l, whose unusual structures suggest mutagenic capability. In order to address the multiple minimum problem, we conducted our searches of conformation space using two alternative optimization methods that also employ differing search strategies. We used the Powell algorithm, BOTM, with starting conformations that are selected combinations of rotamers, and the method of simulated annealing (SA), with random or arbitrary starting conformations. While both approaches were effective in defining the most important structures, SA was more successful than BOTM in locating the structures of lowest energy.

Introduction The genetic effects of the carcinogens 2-aminofluorene (AF)’ and its N-acetyl derivative, 2-(acety1amino)fluorene (AAF),’ have been so frequently investigated that these compounds have been termed “superb tools for the exploration of the mechanism of carcinogenesis” (1). A recent review (2)that celebrated 50 years of AAF research called it “one of the most versatile compounds in experimental cancer research”. Activated forms of both AF and AAF bind to DNA, and the most prominent adducts are those that result from modification of position 8 of guanine (Figure 1). The mutagenic specificity of these compounds has been explored by a number of groups. Those conducted in a site-specific manner have been of greatest interest to us, as they reveal the dramatic effect of DNA sequence upon mutagenesis. Mutagenesis by AF and AAF. Fuchs and co-workers (3-6)have reported that AAF produces frameshift mut Chemistry Department, New York 1 Oak Ridge National Laboratory.

University.

Arizona State University. New York University. * Abstract published in Advance ACS Abstracts, March 1, 1994. 1 Abbreviations: AF, 2-aminofluorene;AAF, 2-facety1amino)fluorene; SA, simulated annealing. 11 Biology Department,

tations in a bacterial system. Runs of guanines and the sequence GzCGCz (third guanine modified) were particular hotspots. Other workers also found that oligo-G sequences are a hotspot for -1 deletions by AAF in bacterial systems (7,8).A mechanism for AAF-induced deletions has been proposed in which the AAF-modified guanine stabilizes the formation of a bulge in DNA (6,9,10). In vitro studies of the mechanism have shown that the sequence surrounding the adduct is an important factor (9). Some bacterial studies have reported a greater diversity of evenb, including frameshifts at runs of A’s and C’s, and substitutions (8). A somewhat different picture of AAF mutagenesis emerged from studies in mammalian cells. In certain cases, base substitutions predominated, particularly transversions (11,12).In another study, mutagenesis by the AAF adduct was associated with deletions and other complex gene rearrangements (13,14). AF, while related to AAF in structure, differs in part in its mutagenic spectrum. In bacterial studies using a modified plasmid, mainly point substitutions, particularly GC -TA transversions, were reported (15).Other workers in bacterial systems have encountered additional types of mutations, however, including other substitutions, single

0S93-22~x/94/2101-0239$04.50/0 0 1994 American Chemical Society

240 Chem. Res. Toxicol., Vol. 7, No. 2, 1994

Shapiro et al.

/C’\ C!

ll

/c. c9

\

C6 I I

c” CS

I

Figure 1. Structure, numbering scheme, and variable dihedral angle designations for d(CpG-AAF). In the AF adduct the acetyl group, C(O)C, (C,is themethylgroup), is replacedby ahydrogen. The dihedral angles A-B-C-D are defined as follows: x’,x (pyr): Ol’-Cl’-Nl-C2; x’, x (pur): Ol’-Cl’-N944; a: C5’-05’-P03’; 8: C4‘-C5’-05’-P; y: C3’44’-C5’-05’; t: P-O3’-C3’-C4’; 05’-P-O3‘-C3’; CYT(thymine methyl): C6-C5-C-H; a’: N9C8-N-C2; 8’: C&N-C2-C1; 7’: C8-N-C-Cm; 6’: N-C-C,-H. The angle A-B-C-D is measured by a clockwise rotation of D with respect to A, looking down the B-C bond. Sugar pucker is defined by the pseudorotation parameter P (40). PI is the 3’linked sugar and Pz is the 5’4inked sugar.

base deletions, and a variety of frameshifts (8, 16-18). Two groups of workers using AF in mammalian cells have TA found that point mutations, particularly GC transversions, predominate (12, 14). In the case of AF, then, the results from bacteria and mammalian cells appear similar. The above studies suggest that the structure of DNA in the vicinity of the adducts is an important factor in mutagenesis and that the sequence around the adduct has an important role in determining this structure. Very few experiment-based high-resolution views have been provided, however, of the structure of AF- or AAF-modified DNA. Studies on the Conformation of AF- and AAFModified Oligomers. Pate1 and co-workers have used high-resolution NMR, together with our potential energy calculations, to define a double-helical undecamer containing a guanine-AF to adenine mismatch (19). In this structure the modified guanine is syn and the AF resides in the B-DNA minor groove. NMR studies on an AFmodified 15-mer have revealed a mixture of conformers, a major one with AF in the major groove and a second one which involved AF-base stacking (20). Two conformers of this type have also been deduced from NMR data in an AF-modified c-H-rad protooncogene sequence in the laboratory of Krugh (21). Interestingly, these two kinds of conformers had been predicted for AF-modified duplexes in our early work (22). Krugh and co-workers have combined NMR with our minimized potential energy calculations and molecular dynamics simulations to describe an AAF-modified double-stranded nonamer, with cytosine opposite the modified G (23). The modified guanine is syn, displaced from its normal intrahelical position, the Go-C pair at the modification site is disrupted,

-

and the AAF is largely inserted into the normal position of the guanine, as had been predicted by the base displacement (24) or insertion-denaturation models (25). Earlier studies on AAF-modified DNA were less definitive and are summarized in ref 23. The above investigations embedded the adduct within a length of double-stranded DNA. Such structures are relevant to mutagenesis in that they are possible substrates for error-prone repair. When mutation takes place during replication, however, the adduct may appear in singlestranded DNA, in the single-stranded region of the replication fork, or in a bulge produced by a slippage mechanism. For this reason, knowledge of the effect of AF and AAF on single-strand conformation is also important to an understanding of mutagenesis. Nuclear magnetic resonance and circular dichroism studies have been performed on a number of short singlestranded ribo- and deoxyribonucleotides of defined sequence, bearing a single AF or AAF bound to C-8of guanine (24,26-32). The most important result to emerge at the dimer level was the demonstration of significant stacking between the fluorene ring and an adjacent 5’-base. In the same sequence, stacking in the AAF series always exceeded that in the AF series. Detailed conformations were not defined by this work, however. In the past, we carried out minimized potential energy calculationson the conformation of AF- and AAF-modified single- and double-stranded oligonucleotides, using a limited number of trials in an alternating CG sequence context (22,33-35). We have now initiated a much more comprehensive study of single-stranded amine-modified oligonucleotides in a wide variety of sequence contexts. We have sampled in depth the conformation space of all 8 AF- and all 8 AAF-modified (at the guanine 8-position) deoxydinucleoside monophosphates (dimers), to investigate the interaction of the modified residue with the various 5’- or 3‘-neighbors, running 4897 trials in each case. Further, this effort allowed us to compare the effects of AF and AAF on conformation with one another in all these contexts. In order for us to extend this type of broad search, which makes no assumptions about the adduct conformation, to larger fragments of DNA, however, more effective ways will be needed to locate important minima. For four modified dimers, we tested the value of amuchwider search (31104 additional trials for each were run). In addition, we evaluated a different approach to the location of minima in all of the 16AF- and AAF-modified dimers: the method of simulated annealing (SA)‘ as an optimization technique vis-a-vis our usual Powell algorithm (361,BOTM. SA has the virtue of being able to make uphill moves on the potential energy surface which BOTM cannot do.

Methods DUPLEX/BOTM. The potential energy minimizations were carried out with DUPLEX (37),a molecular mechanics program for nucleic acids and carcinogen-modified nucleic acids. DUPLEX performs minimizations in the reduced variable domain of torsion angle space. In torsion space small movements of bond lengths, bond angles,and dihedral angles that vary little in nature (such as ones governing deviation from planarity in an aromatic ring) are not permitted; only the large movements of torsions and puckers of the deoxyribose rings (indicated in Figure 1)are allowed. The resulting vast diminution in the number of variables that must be simultaneously optimized to minimize the potential

Conformation of Amine-Modified DNA energy, compared to Cartesian space minimizations with 3 N 6 (N = number of atoms) variables, permits larger movements from a given starting conformation;it also assurea realistic internal geometry. These advantages are at essentially no cost to our objective of defining overall conformational features of carcinogen-modified polynucleotides (of course, all degrees of freedom must be allowed to flex when pathways between forms are under investigation). DUPLEX employs a force field for nucleic acids that is similar to one devised in the laboratory of Olson (38,39). Minimizations are carried out with the Powell algorithm, BOTM (36).BOTM is computationally simple in that it does not require derivatives of the energy function, but it does not permit uphill moves on the potential energy surface. In overview BOTM works as follows: a given variable (torsion angle or sugar pseudorotation parameter) (40) is set to a starting value and two selected surrounding values, the energy is computed for each of these, a parabola is interpolated from these data, and ita minimum is designated as the current variable value; the procedure is repeated for all variables,and this constitutes a single iteration. Numerous iterations are performed near the local energy minimum, all variables are flexed simultaneously, and the minimum of the multidimensionalparaboloid is interpolated. This continues until a preset condition is met in which further small conformational changes produce very little energy change. At this point we have a local minimum energy conformation. We can then choose another starting conformation. Although the sugar conformation in DUPLEX is governed by the pseudorotation parameter as an independent variable, the puckering amplitude and other geometric parameters of the sugar are not constant. Rather, previously optimized values corresponding to each pseudorotation parameter value are employed. These are detailed in ref 34. Full details about the geometry of the fluorene moiety and linkage site, as well as all force field parameters relating to the AF or AAF and linkage site, including partial charges and barriers to rotation, are given in refs 22,33, and 34. Starting conformations for the poten$ial energy minimizations with BOTM alone are given in the supplementary material (Tables S-1 and 5-2). A total of 4897 trials (Table S-l), selected from our knowledge of preferred domains in nucleic acids (41, 42) were carried out for each AF- and AAF-modified DNA dimer. An additional 31104 trials for d(CpG-AF), d(CpG-AAF), d(GAF-pC), and d(G-AAFpC) were also carried out. Rather than a targeted search employing our best knowledge of where lowenergy conformations of dihedral angles may be found, represented by the 4897 trials, this much broader (and costlier) search simply employs all combinations of favored domains (41) for each dihedral angle. Domains represented by the starting conformations employed are centered at the following: angle XI

x'

P 6,

5; a,8, Y

deg (Domain) 195,235 (anti) 300,320,340 (high anti) 60 (syn) 18 (CSj-endo) 162 (C2j-endo) 60 (gauche+) 180 (trans) 300 (gauche-)

DUPLEX/SA. In the present workDUPLEX was interfaced with the SA optimization method. DUPLEXJSA uses SA to search for the lowest potential energy well from high-energy starting positions; this is followed by the Powell algorithm which finds the well minimum. Simulated annealing is an optimization technique that was proposed separately by Kirkpatrick and Cerny in the 1980s (43, 44). It is founded on the Metropolis algorithm, which uses a Monte Carlo method to minimize a cost function such as the potential energy (45). After defining an initial state of the system such as the variable torsion angles that define the structure, SA

Chem. Res. Toxicol., Vol. 7, No. 2, 1994 241 makes a small, random change, LiD, in one randomly selected variable (46-51). Next, the energy associated with the new conformation, E,,,, is calculated. If the new conformation is a t a lower energy than the old one (E, < E,M,the change in energy AE = E,-E,ld is negative),then adownhill move on the potential energysurface has beenmade, and the random change is accepted. If the new conformation is a t a higher energy than the old one (E, > Eold,the change in energy AE is positive), then an uphill move on the potential energy surface was attempted, and the Metropolis criterion is employed to decide whether the new structure should be retained. According to this criterion, the of accepting the perturbed state is probability P = compared to a random number; k is the Boltzmann constant in J/K, and T is the temperature, an assigned control parameter used to minimize the potential energy,which does not correspond to a physical temperature. When P 2 X,where X is the random number in the interval (O,l), the new structure is accepted and used for the next random alteration. If P < X,the perturbed state is rejected and the old conformation is reused as the starting point for the subsequent random change. After many such random movements at a given T,the Metropolis algorithm produces a probability distribution of states that approaches Boltzmann's (52-54). I t can be seen from the probabilityrelation that a higher T translates into a higher e-(a/kn, increasing the likelihood of e-(*lkn being greater than the random number. Consequently, the chances of the perturbed state being accepted are greater at a high T. As T decreases, the likelihood of e-(alk"J being greater than the random number is diminished and consequently the probability that the perturbed state is accepted is smaller. In other words, at low T,the old structure is more frequently retained and subjected to further dihedral rotations. SA is attractive because it is mathematically simple in that it does not need to use derivatives of the energy function, and unlike other minimizers, it can accept moves that lead to incrementa in the energy (55-57). It is these uphill walks on the potentialenergy surface which give the SA technique its unique power, for they enable the molecule to climb out of higher energy local minima; this improves the chances of reaching the lower energy wells (49, 58, 59). The ability of SA to locate the lowest energy wells depends critically on the selected rate of change of T,known as the cooling schedule (60,61). Our optimizationof the cooling schedule for the present study is described in Sidawia2Details of the search strategy and simulated annealing protocol are given in the Appendix. Computations were carried out on a Cray Y-MP supercomputer at the NSF San Diego Supercomputer Center and on a Cray Y-MP at the DOE National Energy Research Supercomputer Center a t Livermore, CA. The CPU time for the 26 SA runs that constitute one SA trial for a deoxydinucleoside monophosphate modified by either AF or AAF did not exceed 30 min. A single trial with BOTM alone takes less than 30 8. The number of BOTM and SA runs attempted for the various carcinogenmodified dimers are listed in Table A-3 (Appendix).

Results M i n i m a were calculated for the eight AF-modified dimers d(G-AF-pN) and d(NpG-AF) as well as the eight AAF-modifieddimers d(G-AAF-pN) and d(NpG-AAF) ( N = A, C, G, and T ) . Two alternative methods of locating potential energy m i n i m a were employed: the Powell algorithm BOTM (36)and SA. T h e s e two methods also employed different s t a r t i n g conformations. W i t h SA, arbitrary or completely r a n d o m s t a r t i n g conformations were used. W i t h t h e Powell m e t h o d t h e s t a r t i n g conformations were selected combinations of preferred rotamer domains i n nucleic acids for t h e flexible torsion angles and sugar puckers. The method generally involved 4897

* Ph.D. Thesis, in preparation.

242 Chem. Res. Toxicol., Vol. 7,No. 2,1994

Shapiro et al.

Table 1. Global and Other Significant Minima Min. #

Fig.

AI{

Mthd

1

2(e)

00

SA

fl 74 74

---

x’

Pf 32 30

y 301 301

38 39

_--

7

a‘

fl’

6

287 286

104 105

202 203

$’

7’

290 289

a

fl

292 295

184 175

y

65 69

P2 144 159

x

6

52 231

187 186

AAF d(G-AAF-DA\

Min.#

AR

Fig.

,112000 I

Min. # d(TpG-AAF) d(TpG-AAF)

,~

1 2

I I

Fig. 6(a) 6(b)

An 00 00

1 I

:”

SA

Mthd

__, ,

Mthd

BOTMlSA SA

2,

$ y 74) 301

$

:I’ 1851

I

y

PI 34

I

Pf

x’

a‘

40 1285

x’

a

I

106 I 3 5 7

T

s

6’

C

911 2401 3121 2771 3151 1881 2971 1701 2391 1801 1811 101

621

trials (short search),but in the four structures abovewhere N = C, a long search of 31104 additional trials was used as well. The SA method employed no more than 1000 trials per molecule. The angle combinations employed in the short and long searches with BOTM are listed in the supplementary material; the SA search protocol is given in the Appendix. The minima found for each structure using the two searches with BOTM and the SA search have been combined and ranked in order of increasing energy to 3.0 kcal/mol in Tables S-3/S-18 (supplementary material). To limit the length of the tables, higher energy minima which differed from a more stable one only in one or more of the 5’-terminal angles @ and y, the 3’-terminal angle e, and the thymine (CYT)and acetyl methyl (6’) rotations were omitted. These variables have relatively little effect on the overall shape of the molecules. In addition, minima which differed by less than 20’ in all nonterminal angles from the values of a more stable minimum were considered to be equivalent and omitted. In general, we found that rotation of p’ (the variable defining the rotation about the bond linking the amine N to the AF ring) by 180’ afforded another stable minimum. Such minima pairs could differ in energy by 1.0 kcal/mol or more, but generally had similar shapes. Both forms are listed. Similarly, in the AAF series, the value of y’ (the variable defining rotation about the amide bond in the acetyl group) could be near 0’ or 180’. The energies of

6

I 309

a

5

a

I203 I289 I294

$

y

P2

$

y

I 172 I

71

x

a‘

&

aT

6

1 163 I 2 4 0 I 184 I 2 8 9

$’

y’

6’

6

2581 671 1551 1741 501 991 2891 3531 1891 185 1941 1871 571 1551 1441 1221 2681 1721 1831 186,

the two forms differed, but the shape of the remainder of the molecule was unaffected. Both minima were listed when they were found. In addition, Table 1lists global and other most significant minima for all the modified d’imers. The Structures: d(G-AF-PC), d(G-AF-pT), d(GAAF-pC), and d(G-AAF-pT) (Tables 5-3-5-6). These structures are considered together because of the great similarity in their global minima and in several others. The computed globalminima are virtually identical (apart from the termini and methyls described above) with a small difference in the glycosyl torsion angle x (231’ for C, 240’ for T) distinguishing the C and T series. The presence or absence of the acetyl group in the AAF series does not affect the remainder of the structure. The lowest energy conformations are illustrated for d(GAF-pC) (Figure 2a) and d(G-AAF-pT) (Figure 2b). They feature syn-guanine,with a’near 285O, and a right-handed backbone. In this arrangement, the plane of the bonds to the amine nitrogen is almost perpendicular to that of guanine, and the aminofluorene ring is oriented in the direction of the 3’-neighbor. The AF ring is also almost perpendicular to the amine nitrogen bonds (p’ is close to 105’). This type of diarylamine conformation has been termed “helical”(62). The two rings relate to one another roughly as do the pages of a partly opened book (in outline, a V-like structure), and the inside surface of this structure

Cottformation of Amine-Modified DNA

Figure 2. Stereoviewsof (a) d(G-AF-pC)#1,(b)d(GAAF-pT) #l; (c) d(G-AF-pC)#2, (d) d(GtAAF-pC)#4, and (e) d(G-AF-PA) #l. These stereoviewsare prepared for use with a stereoviewer. To view with crossed eyes, the left and right images must be interchanged.

will be called the binding pocket. The 3‘-nucleotide is nested within this pocket. In this conformer type, the contacts of the AF ring are primarily with the underside of the 3’-sugar. They vary slightly from structure to structure but may include a subset of the H-1’, H-4’, and H-2” protons and oxygens 3’ and 4’. In addition, the AB’ring contacts the 2-carbonyl group of the pyrimidine. The pyrimidine contacts both sides of the G-AFV-like structve and has imperfect basebase stacking with guanine (thymine turns slightly more toward guanine than cytosine). In the AAF structures, the acetyl group is unhindered, with the carbonyl group oriented toward AAF, and the methyl toward G. Relatively few additional minima occur in this series below 3.0 kcal/mol. With one exception, those below 2.0 kcal/mol differ from the global minima only in that the values of @’ or y’ (acetyl torsion) have been changed by 180’. The reversal of the orientation of the amine ring or the acetyl group has little effect on the remainder of the overall structure. The exception is minimum #2 (0.6 kcal/ mol) for d(G-AF-pC), pictured in Figure 2c. This conformer features a 2-11backbone (41) and appears to have better base-base stacking than the global minimum, but the aminofluorene residue is, for the most part, exposed. This conformer does not appear below 3.0 kcal/mol for the other conformers of this group. Another conformer that is worthy of attention is d(GAAF-pC) #4, 2.2 kcal/mol (Figure 2d). The principal difference between this conformer and the global minimum is a change in 8 to 196d from 176*. This has repositioned the cytosine ring in the V made by G and AAF so that it now stacks more fully with AAF. The distal ring of AAF still maintains contact with the underside of the 3’-sugar. Thus only a modest angle charige serves to convert d

Chem. Res. Toxicol., Vol. 7, No. 2, 1994 243

stfucture that could be called base-base stacked into one that has base-amine stacking. The Structures: d(G-AF-PA), d(G-AF-pG), d(GAAF-PA), and d(G-AAF-pG) (Tables S-7-5-10). The overall shapes of the principal minima in this group resembled those in the above neighboring pyrimidine series. They were generally right-handed, wiul modified guanine invariably syn and a’(except for a rare left-handed exception which is discussed below) in the range 257321O. They featured contacts between the AF ring and the underside of the 3‘-sugar, with imperfect base-base stacking. The group as a whole was more diverse, and the numbel) of minima to 3.0 kcal/mol was greater, ranging up to 42 in the case of d(G-AAF-PA). One great source of this variability was the glycosyl angle of the neighboring purine which could assume a variety of values in the anti, high anti, and syn range. Greater variability was also observed in 8’ and in the phosphodiester backbone. The global minimum for d(G-AF-PA) (Figure 2e) and the structure for d(G-AAF-PA) #2 are close in their backbone angles to the global minima in the d(G-AF-pPy) and d(G-AAF-pPy)series, with, however adenine syn.The global minimum for d(G-AAF-PA) [as well as d(G-AFPA) #91 is also quite similar, except that 8’ is 279O (a rotation of 180O). Neither variant occurs to 3.0 kcal/mol for d(G-AF-pG)and d(G-AAF-pG). Again, this structural type features contacts of the AF ring with the underside of the 3’-sugar (H-1‘, H-4’, and the 3’-OH) and imperfect base-base stacking (the dihedral angle between the base planes is about 28O). The AF ring also contacts H-8 of adenine. In the 8’ = 279’ conformer, the CH2 group of AF is near the H-8of adenine. In contrast to the case in which the 3‘-neighbor is a pyrimidine, the neighboring purine ring stacks only with guanine, and not in part with AF. The type of diarylamine conformation in these structures is closer to the type called ‘perpendicular” (62). In the global minimum for d(G-AF-pG) [Figure 3a; this type of structure occurs also in d(G-AF-PA) #31, we again have an imperfect base-base stack (H-2’ of the 5’-sugar is inserted between the bases), with contacts of AF to the underside of the 3’-sugar (H-4’, 0-4’, and the 3’-OHgroup). In this structure, the neighboring purine has x of 297O, on the border of the high anti range, and 8’ of 341O. This latter value brings the AF ring &nost to coplanarity with the amine bond plane and at a sharp angle to the plane of guanine [a conformation termed “perpendicular” (62)l. No partial AF to neighboring base stack exists, as in the case of the global minima with a 3’-neighboring pyrimidine. The global minimum for d(G-AAF-pG) (Figure 3b) provides another variant on this theme. The unmodified guanine is anti and the phosphodiester angle a has a trans value. The guanines stack with a dihedral angle of about 35” between planes, and the overlap of the unmodified G is confined to its 6-membered ring. The distal ring of AAF contacts the bottom-to-side of the 3’-sugu (H-1’, H-4’, 0-4’, and one H-5’ proton). Another low-energy variant of the imperfect base-basestacked, AF-3‘-sugar contacts group is displayed by d(G* AF-PA) #2,0.4 kcal/mol, and d(G-AAF-pG) #3,0.2 kcal/ mol (Figure 3c). [This occurs at 1.9 kcal/mol for d(GAF-pG) and 1.5 kcal/mol for d(G-AAF-PA) as well. This was not observed for a 3‘-pyrimidine neighbor.] The neighboring purine is close to high anti (309O), and c has a less common value of 259O. In this structure, AF overlaps both sugar rings; the center and distal rings contact the

244

Chem. Res. Toxicol., Vol. 7, No. 2, 1994

Figure 3. Stereoviewsof (a) d(G-AF-pG)#1,(b)d(G-&V‘-pG) #1, (c) dfG-AAF-pG)#3, (d) d(G-AAF-pA)#9, and (e) d(CpGAF) #l. See the caption to Figure 2. side (H-4’ and one H-5’, 0-4’) of the 3’-sugar while the proximal ring contacts H-1’ and H-2” of the 5’-sugar. The base-base stack is partial with a dihedral angle of about 26’ between the base planes and is limited to the amino group region of adenine (or the carbonyl area of guanine). The remainder of the adenine ring contacts the 5’-sugar. A number of other low-energy conformers exist with base-base stacking and AF-3’-sugar contacts. In some cases, they combined these features with very unusual backbones; for example, d(G-AF-PA)#6has {, a,0, y values of 13’, 182’, 71°, and 151’. The most striking variant is d(G-AAF-PA) #9 (Figure 3d, 1.2 kcal/mol). This conformer, together with an almost identical minimum at 2.4 kcal/mol for d(G-AF-PA) and two close relatives at higher energy in the d(G-AAF-PA) table, are the only 5’-AF- or AAF-modified conformers to 3.0 kcal/mol with a’ (65’) outside the range 270-321’. The modified guanine i s syn, and adenine anti, yet the backbone angles e, f, a of 282’, 91’, and 55’ are close to those for the CpG step of Z-DNA. The bases are parallel, overlap in part, and extend their hydrogen-bonding edges in normal Watson-Crick alignment, as in Z-DNA. The pseudorotations of the 5’- and 3’-sugars are 163’ and l o , respectively. The AAF residue is somewhat exposed, as if it were located in a major groove. Its proximal ring contacts the side of the 3‘-sugar: one 5‘-proton, H-3’, and the 3’-OH, as well as H-8 of adenine. The Structures: d(CpG-AF) and d(TpG-AF) (Tables S-ll and S-12). A much greater diversity of lowenergy conformers was present in the 3’-AF-modified dimers, when compared to the 5’-modified isomers. The global minima for neighboring C and T were quite different, for example, and those for neighboring A and G (seebelow) illustrated two additional possibilities. Of the 16 different dimers considered in this report, the subgroup d(NpGAF), then, shows the most profound sensitivityof structure to sequence.

Shapiro et al.

Figure 4. Stereoviews of (a) d(TpG-AF)#1,(b)d(CpG-AF)#2, (c)d(TpG-AF)#Z,(d) d(ApG-AF)#1, and (e) d(ApG-AF)#2. See the caption to Figure 2. The most reliable indicator of structural class proved to be the glycosyl angle of the modified guanine, which generally assumed a value in the vicinity of one of the following: 230’ (anti), 347’ (an unusual very low value of syn not far from high anti), 60’ (syn), and 150’ (an uncommon part of anti, toward high syn). The resulting structures will be described using specific examples. The d(CpG-AF) glqbal minimum ( x for G = 223’) is illustrated in Figure 3e. In this right-handed structure, the combination of anti-G md a’of 270’ serves to orient the AF ring toward the 5’-neighbor, while in d(G-AF-pC), an a‘ near 270’ and syn-G had pointed AF toward a 3‘neighbor. In this structure, the bases stack imperfectly (the dihedral angle between their planes is about 25’1, while AF is positioned above the 5’-sugar, its proximal ring interacting with H-5” of that sugar and the unsubstituted phosphate oxygens. Similar structures exist in the d(TpG-AF) (#6,2.2 kcal/mol) and d(ApG-AF) (#9,1.5 kcal/mol) series, but a t higher energy, relative to the global minima. The d(TpG-AF) global minimum (Figure 4a) is quite different. A left-handed backbone ({and a haveg+values characteristic of the CpG step of Z-DNA) is combined with an intermediate value of 219’ for a’and a x value a t 347’ for modified G to produce a structure with parallel T to AF stacking. Thymine contacts the proximal and center AF rings from a position on the outer, convex face of the G-AF segment, rather than in the space (binding pocket) enclosed by G and AF. The thymine methyl group is over the proximal ring and close to the 2’-CH2 group of the modified residue. Thymine stacks using the face that would normally interact with a 3’-neighbor in B-DNA. The two sugars have an antiparallel relation, as in Z-DNA. Guanine is relatively unstacked, centered above ita own 5’-0 and the phosphate group, A similar structure is found in the d(CpG-AF) (#3,1.5 kcal/mol), d(ApG-AF) (#4,0.8 kcal/mol), and G (#3,0.6kcal/mol) series. In the cases of the purine neighbors, both heterocyclic rings interact with

Conformation of Amine-Modified DNA

the three rings of AF. A related structure with a low syn value for G, and with AF to base stacking, also occurs a t relatively low energy in this series: d(CpG-AF) (#2, 1.0 kcal/mol, Figure 4b), d(TpG-AF) (#3, 1.8 kcal/mol), d(ApG-AF) (#17,2.2 kcal/ mol), and d(GpG-AF) (#25,2.3kcal/mol). This conformer type featured a tg+combination for phosphodiester angles rand a and a value of 164’ for y. In Figure 4b, cytosine and AF are close to parallel, and the cytosine is centered over the proximal and center rings of AF, but is now located within the binding pocket (the concave G-AF surface). In contrast with the structure of Figure 4a, the pyrimidine stacks with AF using the face that would interact with a 5’-neighbor in B-DNA. Further, the orientation of the pyrimidine ring relative to AF is reversed. Thus in d(TpGAF) #3, the methyl group of T is over the distal AF ring. The two sugars are close to one another, but with their axes askew. Guanine stacks on one phosphate oxygen, and on H-4’ and 0-3’ of the 5’-sugar. AF to guanine stacking can also be achieved in other ways in this group. Minimum #2 for d(TpG-AF) [related structures occur at higher energy, relative to the global minima, for the other d(NpG-AF)I achieves this relation with a glycosyl angle of 150’ (anti, but shifted toward high syn, with N-3 of G close to the 2’-CH2 group of its sugar) for modified G and values of 16’ and 189O for phosphodiester angles rand a (Figure 4c). In this conformer, the thymine ring has only a partial stack with the proximal AF ring, with its methyl group oriented away from AF and close to the 0-6 and N-7 of G, in a small pocket. Guanine contacts H-6 of T, the top of the 5’-sugar, H-2’, one of the 5’-CH2 protons, and the 5’-OH. The structure appears compact except that the distal ring of AF is exposed to the medium. A final group that features close AF to neighboring base stacking occurs with modified G syn (x near 60’). This type closely resembles the global minima in the d(NpGAAF) series, but occurs at higher energies here: d(CpGAF) #5 (2.0 kcal/mol), d(TpG-AF) #11 (2.6 kcal/mol), d(ApG-AF) #14 (1.7 kcal/mol), and d(GpG-AF) #lo (1.3 kcal/mol). The neighboring pyrimidines in this group are anti and the purines syn. This type of structure is illustrated in the d(NpG-AAF) section. The Structures: d(ApG-AF)and d(GpG-AF) (Tables 5-13 and 5-14). These series have already been discussed to some extent in the above section, as the same conformers are displayed with a neighboring purine as with an adjacent pyrimidine. d(ApG-AF) and d(GpGAF) display additional conformers, however, including their global minima, in which the neighboring base adopts an unusual anti glycosyl angle in the range 149-178’. (This range is not seen, to 3.0 kcal/mol, for the pyrimidine neighbors.) The conformers all display base-base stacking with modified guanine anti, in the normal 220-230’ range, or near the less common 167”. Some of these structures contain other novel features of interest, however. In the global minimum for d(ApG-AF) (Figure 4d) both A and G have a novel anti value for the glycosyl angle of about 165’. The bases are almost parallel and stack with their six-membered rings. However, the faces that interact with one another are the opposite ones that would normally be used in an ApG sequence in DNA; in effect, they have exchanged positions. The two aligned glycosyl bonds are at an angle of about 55’ to one another, creating a novel H-bonding situation in which 0-6 of G is positioned above

Chem. Res. Toricol., Vol. 7, No. 2, 1994 246

Figure 5. Stereoviews of (a) d(GpG-AF) #1,(b)d(GpG-AF) #4, (c) d(CpG-AAF) #1, (d) d(CpG-AAF) #2, and (e) d(CpG-AAF) #4. See the caption to Figure 2.

N-7 of A, the N-1 to H-1 bond of G is above the C-6 to amino bond of A, and the C-2 to amino bond of G is above the lone pair on N-1 of A. The principal change from a B-DNA backbone in this conformer lies in its value for phosphodiester bond a,which is 215’, in the trans domain rather than the more usual gauche-. AF contacts the side of the 5’-sugar (H-4’, one 5’-proton, 3’-0, and the proton of the 5’-OH). The AF (proximal ring) contacts the 5’CH2 group of the 3’-sugar and (middle ring) a phosphate oxygen and 3’-0 of the 5‘-sugar. A normal d(ApG-AF) structure with both bases anti would have an a‘ value near 270°, to orient the AF toward its 5’-neighbor residue. The a’value of 119’ in this case reflects the inverted orientation of the residues. The proximal and center rings of AF contact the back of the 5’-sugar, (H-1’, H-4’, 0-4’, and the proton of the 5’-OH), while the distal ring contacts H-4’. This novel structure also appears at low energy in the d(GpG-AF) series: #2, 0.5 kcal/mol above the global minimum. The coglobal minimum (#2) for d(ApG-AF) (Figure 4e) has a more normal anti value for G, but x’ of A is at 178’. In the backbone, both a and P are trans. The bases are aligned in normal Watson-Crick orientation, but stacking looks poor. Overlap is limited to the six-membered ring of A and the five-membered ring of modified G, with the dihedral angle between their planes about 40’. AF is positioned above the 5’-sugar, in contact with H-3’, one H-5’ proton, the proton of the 5’-OH, and one P I 0 oxygen. A similar structure exists in the d(GpG-AF)series: #8 (0.8 kcal/mol). The d(GpG-AF) global minimum (Figure 5a) provides another novelty. The modified G has an anti x value of 232’, in the normal range, while the value of the 5’-G is 153’. The only uncommon backbone angle is y, at 170’. In this conformer, the six-membered rings of the two

246 Chem. Res. Toxicol., Vol. 7, No. 2, 1994

guanines stack perfectly, but the modified one has been turned so that the face it presents to its 5’-neighbor is the one that would normally be presented to a 3‘aeighbor. As a result, the amino group of one G lies approximately under the N-1 H of the other, and vice versa. Each amino nitrogen lies above N-7 of the other. The hydrogen-bonding face of one base is presented in reverse Watson-Crick orientation, compared to the other. The aminofluorene (proximal ring only) interacts with the 5’-0 of its own sugar, one phosphate oxygen, and H-2” of the neighboring sugar. A similar conformation can be seen in d(ApG-AF) #3 (0.7 kcal/mol). In this structure, guanine C-6, N-1, and H-1 are above adenine N-1, C-6, and N-6, respectively. Reverse Watson-Crick base orientation is also achieved in d(GpG-AF) #6 (0.7 kcal/mol) but through a very different set of torsion angles. In these minima, CY’is 90°, y is 311°,and the unmodified and modified glycosyl angles are 297’ and 173’, respectively. The modified G again stacks with the unmodified one using the face that would be turned to its 3’-neighbor in B-DNA, and the amino group of one G is roughly over the carbonyl of the other. AF stacks on the phosphate group. A similar conformation is seen in d(ApG-AF) #7. Finally, d(ApG-AF) #5 (0.9 kcal/mol) and d(GpG-AF) #4(0.7 kcal/mol, Figure 5b) combine aconventional righthanded backbone with normal anti (228’) modified G and x’ of 173’ for the 5’-neighbor. This more conventional structure features an imperfect base-base stack (Partial overlap, base planesat about a 30’ angle). The AF residue has close contacts with the top of the 5’-sugar(one5’proton, H-2’, and H-3’ and partial overlap with phosphate). The Structures: d(NpG-AAF) (Tables S-15-S-18). In contrast to the series with 5‘-residue modification, the presence of the acetyl group on the aromatic amine has a pronounced effect here. Structures with x values for modified G of 230’ (anti) and 347’ (a low value of syn), which were important for d(NpG-AF),do not appear below 3.0 kcal/mol. The tables are filled for the most part with conformers that have x values for modified G of 60’ (syn) with a few (never the lowest energy conformation) that have x of 150° (an uncommon, low region of anti, toward high syn). The value of CY’ remains near 90’ (with a single exception), which in these structures orients the AAF in the direction of the 5’-neighbor. A large number of conformers exist below 3.0 kcal/mol, 47 for d(CpG-AAF), 17 for d(TpG-AAF), 31 for d(ApGAAF), and 13 for d(GpG-AAF). All with energies of 1.0 kcal/mol or less (and many above) were examined in detail. They displayed [with the exception of d(CpG-AAF) #4 and #5] good base to AAF stacking, with the modified guanine contacting the top or side of the 5’-sugar. This type of structure was compatible witha number of different arrangements of the backbone, none of them conventional B-DNA. The global minimum #1for d(CpG-AAF) (Figure 5c) (d(TpG-AAF) #4 is very similar) and d(CpG-AAF) #2 (Figure 5d) both have close to B-DNA geometry with, however, e near 270’. They differ from one another primarily in the pseudorotation of the 5’-residue,but this modest change shifts the stack of cytosine from the proximal AAF ring in #lto the distal ring in #2 and the guanine contacts from the top of the 5’-sugar (in #1)to the side of the sugar (in #2). Also pictured (Figure 5e) is d(CpG-AAF)#4 (#5 is the same, except for the acetylgroup, which is reversed). In this conformer, neither G nor AAF

Shapiro et al.

Figure 6. Stereoviews of (a) d(TpG-kAF) #1,(b)d(TpG-AAF) #2, (c) d(GpG-AAF) #A, (d) d(GpG-AAF) #2, and (e) d(ApGAAF) #2. See the caption to Figure 2.

stacks in a parallel mode with cytosine. Both contact the 5’-sugar, while cytosine interacts with the edge of the AAF ring system. d(CpG-AAF) #7 has syn-cytosine, but in ita features otherwise resembles #2. The two coglobalminima for d(TpG-AAF) are illustrated in Figure 6, panels a and b. Although the backbone angles e, I;, a,p, y differ considerably (g,g,g,g+, t for #1,Figure 6a and t, g+,t, t,g+ for #2, Figure 6b), both achieve a stack between cytosine and the AAF proximal ring, and guanine to 5’-sugar contacts [in #1,guanine contacts the side (H2’, H-2”, H-3’) and the proton of the 5’-OH group; in #2 guanine sits above the 5’-sugar, contacting H-2’, one 5’ proton, and the 5’-hydroxyl proton]. Minimum #2 is fairly close to d(CpG-AAF) #5 in its torsion angles, but a change of 24’ in p and lesser changes in the backbone anglessuffice to shift the AAF stack from the 5’-sugar in the latter to cytosine in the former. The global minima for d(GpG-AAF) and d(ApG-AAF) were quite similar and are illustrated for the latter (Figure 6c). The 5’-purine is syn, and the backbone angles e, S; cy, p, y are g-, g-,g,t, g+. The near-parallel purine to AAF stack places the five-membered rings of eachsystem above one another, while the six-membered purine ring is above the proximal AAF ring. Guanine lies above the 5’-sugar, contacting H-3’, one 5‘-proton, and the H of the 5’-hydroxyl group. The #2 minima for d(ApG-AAF) and d(GpG-AAF) were also about the same (illustrated in Figure 6, panels d and e, respectively). The backbone angles in this case were of a type (g, g+, g+, t, t ) associated with the CpG step of Z-DNA. The parallel purine to AAF stack now had its orientation reversed from that in the global minima. The five-membered purine ring was above the AAF proximal ring and the six-membered purine ring above the AAF center ring. In this case, the modified guanine contacted H-1’ of the neighboring sugar and H-8 of the 5’-purine.

Chem. Res. Toxicol., Vol. 7, No. 2, 1994 247

Conformation of Amine-Modified DNA

Table 2. Comparison of SA and BOTM in Locating Important Minima total no. of min

dimer

Table

d(G-AF-pC) d(G-AF-pT) d(G-AAF-pC) d(G-AAF-pT) d(G-AF-pA) d(G-AF-pG) d(G-AAF-PA) d(G-AAF-pG) d(CpG-AF d(TDG-AF) d(ApG-AF)

5-3 5-4 5-5 S-6 s-7 S-9 S-10 s-11 s-12 S-13

13 6 6 5 35 18 42 30 21 17 37

d(GpG-AF) d(CpG-AAF) d(TDG-AAF)

S-14 S-15 S-16

d(ApG-AAF) d(GpG-AAF) total

5-17 S-18

5-8

no. of min SAn 6 3 4 4 22 4 29 8

no. of min BOTMb 7 2 2 1

no. of min BOTM/SAC 0 1 0 0 0 0 0 0

15 28

13 14 13 22 7 0 6

54 47 17

46 33 16

4 13 0

4

31 13 392

25 10 264

6 2 112

0 1 16

11

3 2 3 1 1

origin of global min SAf BOTM SA = BOTM SAf + BOTM SAf + BOTM SA only SAf + BOTM SA only SAf BOTM SA = BOTM SA only #1SA only %2SAf BOTM SAf + BOTM SA only it1 SA BOTM #2 SA only SAf + BOTM SA = BOTM

min, SA on19

min, BOTM only

+

+

+

#3 #3 #1-3 #1 #I,7 #2,3,6 #I,3 , 4 4 #2-4

#4 #44 #5

#7

#2,3 #2

0 Minima found by SA uniquely, or equivalent forms were produced by BOTM, but the one found by SA waa of lower energy. Minima found by BOTM uniquely, or equivalent forms were produced by SA, but the one found by BOTM was of lower energy. Minima found by both SA and BOTM at identical energy. Minima 1.0kcal/mol or less above the global found by SA uniquely. BOTM did not fiid these minima below 3.0 kcal/mol. e Minima 1.0 kcal/mol or less above the global located by BOTM uniquely. SA did not locate these minima below 3.0 kcal/mol. f Equivalent forms were produced by BOTM and SA, but the one found by SA waa of lower energy.

Minimum #3 in the case of d(ApG-AAF) [not encountered for d(GpG-AAF)] displayed the same stacking orientation as #2, but with yet another backbone, and with guanine above and in contact with the 5’-sugar. In this case, the glycosyl angle for modified G was 148’. Conformers were also encountered, with slightly higher energy, in which the neighboring purine was anti, for example, d(ApG-AAF) #6 (1.5 kcal/mol). In this case, the adenine to AAF stack differed in position, with the center and distal rings of AAF overlapping the five- and six-membered rings of adenine, respectively. Effectiveness of Search Strategies. A search of conformational space was conducted initially using 4897 trials (”short search”) for each of the 16 modified deoxydinucleoside phosphates with the Powell minimization algorithm, BOTM. The input torsion angles and sugar pseudorotation parameters used were based upon our knowledge of preferred combinations and are given in the Methods section. For the four cytosine-containingdimers, an extended search involving 31104 additional trials (detailed in the methods section) was carried out (“long search”). An independent search was also carried out for each dimer using the simulated annealing optimization method and arbitrary or random starting conformations. This was carried out for each case until a global minimum as stable as that located by BOTM had been found. The conformers were combined to produce a single table for each dimer (Tables S-3-S-181, eliminating redundancy as described in the Methods section. Short Search vs Long Search. In the four cases examined, the long search produced no new global minima. The following minima (15 of the 392 in Tables S-3-S-18) were contributed by the long search: d(G-AF-pC) (Table S-3) #4,9, and 11; d(G-AAF-pC) (Table S-5) #5; d(CpGAF) (Table S-11)#12-14and 19;d(CpG-AAF) (Table S-15) #7,11,14,16,19,26,33. Of this list, all are more than 1.5 kcal/mol above the global minima except #7 and #11for d(CpG-AAF). The most stable, #7 (0.9 kcal/mol above the global), differs from #2 in the same table only in that it has syn-cytosine, and the position of guanine in its contacts with the Y-sugar has shifted modestly. (Several

other minima found by the long search also feature syncytosine. Some minima of this type turned up in simulated annealing runs, but none in the short search, as syncytosinewas not included among the input conformations.) Minimum #11for d(CpG-AAF) was almost identical to #7, except that cytosine was anti. On the whole, the results from the long search were sparse, given the large investment of computational time it required. Simulated Annealing (SA)vs Powell Minimization (BOTM)withLongand Short Searches. Acompariaon of the two methods is made in Table 2. The first row in this table, for example, awards 6 of the 13 minima found for d(G-AF-pC)to thesimulated annealingmethod. Either these minima were found by SA uniquely or equivalent forms were produced by both, but the one found by SA was of lower energy, and the one found by the alternative method was deleted from the table. Using this criterion for all dimers, the number of minima found by SA (264) exceeded that found by BOTM (112). The column titled “origin of global min” indicates the origin of the 18 global minima in the tables (in two cases there were coglobal minima). Fourteen of them were awarded to SA, and in four cases, the two methods produced the global a t identical energy. It is significant that all of the globals awarded to BOTM were also found by SA, in a higher-energy variant in some cases, but six of the global minima located by SA did not appear to 3.0 kcal/mol in the searches with BOTM for that dimer. The last two columns list minima 1.0 kcal/mol or less above the global that were located by one method only. (The other method did not find it below 3.0 kcal/mol.) Most important structural types were located by BOTM, but a few were missed, notably the one in which neighboring purines stack, but invert their positions (relative to those normal in B-DNA). This structure was seen in d(ApG-AF) #1and d(GpG-AF) #2. BOTM also missed a “no stack structure” observed in d(CpG-AAF) #4 and #5. In the series d(G-AF-pPu) and d(G-AAF-pPu), BOTM missed all structures [including d(G-AAF-PA) #1-3 and d(G-AF-PA) #31 in which the purine neighbor had a conventional syn (near 5 5 O ) orientation.

Shapiro et al.

248 Chem. Res. Toxicol., Vol. 7, No. 2, 1994

The omissions of the SA method were less significant: for example, some conformations of intermediate energy with syn-cytosine, and d(G-AAF-PA) #4, a right-handed, adenine anti (261') base-base stacked conformer, which however was found by SA in the d(G-AAF-pG) series.

Discussion Our survey has allowed us to explore the effect of three different variables on the conformation of aminofluorenemodified DNA dimers: 5'-modificationvs 3'-modification, the presence or absence of an acetyl group, and the identity of the neighboring base (C, T, A, or G). In addition, we have compared different search and optimization methods as to their effectiveness in locating significant minima. The data have been compiled in 16tables and is illustrated selectively in 5 figures. The topics mentioned are discussed in separate sections below, but at this point we can make some overall observations. The feature that the most stable conformers share is that they are compact and minimize the surface exposed to the outside medium. The aminofluorene and bases present rigid flat extended surfaces, so it is advantageous to place them over one another, or on a suitable surface of sugar or phosphate. Either aminofluorene or the modified guanine will usually stack with the neighboring base, while the one that is not so occupied stacks with the neighboring sugar. The 5'-modified series is dominated by a conformer class that features base-base interactions (though often at a sharp angle), while aminofluorene has close sugar contacts. The 3'-modified series offers greater diversity. In the case of AAF modification, the most stable conformers feature a near-parallel amine to neighboring base stack, while modified guanine contacts the neighboring sugar. Some amine-base stacked structures also appear in the 3'-AF-modified series, though they differ in detail from their AAF-modified counterparts. The 3'AF-modified group, however, also displays a variety of different base-base, amine-sugar stacked structures, some of which are novel and suggest mutagenic mechanisms. A Comparison of Search Strategies and Optimization Methods. Molecular mechanics methods have the potential ability to determine the most stable conformation for a fragment of DNA, but a number of unresolved issues must be faced before this can be assured. One major difficulty is the multiple minimum problem, which is encountered when minimization methods are used. The location of the global energy minimum and of all important local minima cannot be guaranteed. The deoxydinucleoside monophosphates considered here represent the smallest DNA unit above the nucleotide level, yet even with bond lengths, bond angles, and most dihedral angles held constant, they present 14 torsional and sugar pseudorotation variables for AF modification and 16 for AAF modification (if thymine is present, the methyl group introduces an additional variable). By limiting the input angles to domains that are preferred in nucleic acids and sampling those for which the data base is small, for example, the aromatic amine torsions, at 45' intervals, we have been able to conduct a reasonably thorough search of conformation space with 4897 trials using DUPLEX/ BOTM. We were interested, however, in testing the effectivenessof our strategy in locatingthe global minimum and important classes of low-energy minima. In one approach which was tested on the four cytosine-containing dimers, an additional 31104 trials were run, allowingother

angle combinations to be explored. This expansion, however, produced only 15 of the 392 minima that appeared in the final four tables of results. No new global minimum was found, and the most stable new one that occurred was 0.9 kcal/mol above the global one. It differed from a more stable one found in the short search only in that cytosine had rotated from anti to syn. It was clearly not worth the extra investment of computer time to perform the additional trials. An alternative optimization and search method, simulated annealing with arbitrary or random starting conformations, proved fruitful. This completely independent approach located six new global or coglobal minima not found by our previous method, and in the remaining cases identified the existing global minima (in some cases the energy was slightly higher due to changes in the terminal angles which did not affect the important features of the conformer). The SA method also identified several lowenergy types of conformers not located by the earlier method, though it missed a smaller number identified by the other method. The results are summarized in Table 2.

While it cannot be said that a SA trial takes less CPU time than a BOTM trial (since a SA trial, composed of 25 runs, takes less than 30 min while a trial with BOTM, comprised of 1 run, takes less than 30 s), the SA method found more low-energy forms in fewer runs using fewer CPU hours (see Table A-3 in the Appendix). With computers that are massively parallel and thus orders of magnitude faster than the Y-MP now available, SA optimization could prove even more worthwhile because of its power. It will be of interest to compare optimization by BOTM and by SA in larger single strands to ascertain how the two algorithms perform for systems with more variables. Recently, Brower employed simulated annealing to examine conformations of small polypeptides; their results indicated an exponential increase in time to find a global minimum as a function of length (63). By comparison, approximate timings for BOTM on a Cray Y-MP are as follows for one trial in unmodified single strands: dimer, 15s;trimer, 60s; tetramer, 150s. However, many thousands of trials would be needed to assure a reasonable survey of the potential energy surface. In a previous study in which dimer minima were combined to provide trimer starting conformations, 3753 trimer trials were made (37). Comprehensive searches for tetramers have not been attempted; however, even the shortcut approach of combining dimer minima to yield tetramers would necessitate tens of thousands of trials. Conformation of Dimers: d(G-AF-pN) and d(GAAF-pN) (N = A, C, G, T). The global minima and most of the low-energy structures for all eight of the 5'-modified dimers studied fall into a single conformational group. In almost all of the minima, guanine is syn, the guanine-toamine N angle, a', is near 290°, and the backbone is righthanded. This arrangement produces conformers in which the modified guanine stacks at an angle with ita 3'neighboring base, while the arylamine ring has close contacts with the bottom or side of the neighboring sugar. In these structures, the guanine plane is close to perpendicular with the plane of the amine bonds, and the fluorene is also at a sharp angle with those bonds. This circumstance creates a wedge-shaped binding pocket. The 3'nucleotide binds within the concave surface of the pocket, and the amine NH or acetyl protrudes from the less

Conformation of Amine-Modified DNA hindered convex side. Two subclasses can be defined, depending on whether the neighboring base is a pyrimidine or purine. In the former case, illustrated in Figures 2a,b, the pyrimidine forms a triangle with the sides of the pocket. It stacks with guanine, but also overlaps the fluorene to a lesser extent. Only a small bond rotation is needed, however, to shift to a closely related but energetically less favorable minimum in which the pyrimidine stacks with the aminofluorene ring. In the most stable minima with a neighboring purine (Figures 2e and 3a,b), the neighboring base stacked only with the modified guanine, while aminofluorene stacked with the 5‘-sugar (in one conformer type, it contacted portions of both sugars; see Figure 3c). These features were displayed in a variety of minima that were reasonably close in energy. The diversity was produced by a variation of the glycosyl angle of the neighboring purine, which could assume a number of values in the syn, anti, and high anti range, and (at a modest energy penalty) by the use of unconventional backbones. The tables, of course, do not provide a reason for the dominance of the above structural type. Alternatives can be found at energies above the cutoff point. One minimum for d(G-AF-pC) at 3.3 kcal/mol, for example (data not shown), featured anti-guanine, with a’ near 90’. This conformer featured a reasonable base-base stack, but AF was positioned above the 5’-sugar and was relatively exposed to the medium (64). This last feature may contribute to the lesser stability of this alternative conformer. Conformation of Dimers: d(NpG-AF)(N = A, G, C, T). The dimer class in which the 3’-residue is modified by aminofluorene shows the most diverse behavior. It is the only group in which the base sequence has a profound effect on the nature of the most stable structure(s). Very different structures, in fact, are displayed in the global minima from the four different neighboring bases. A rough division of the minima into groups could be made on the basis of the glycosyl angle of the modified guanine. (a) x near 230’ (anti), with base-base contacts: this group itself contained some very different structures. The d(CpG-AF) global minimum, for example (Figure 3e), features an imperfect base-base stack, while AF contacts the phosphate and the top of the 5’-sugar. The combination of anti-G,a’near 270°, a n d g (right-handed) values for f and a serve to orient the AF toward the 5‘aeighbor. The d(ApG-AF) coglobal minimum (#2, Figure 4e) has roughly the same shape, but changes in a’ (to 229’) and in the backbone have altered the contacts. Those of AF remain roughly the same, but base-base overlap is partial and at a sharp angle. The d(GpG-AF) global minimum (Figure 5a) has arighthanded backbone, with a’near 270°, and modified G anti. It displays AF-phosphate contacts and a parallel basebase stack which is quite novel, however, in terms of its base orientation. They stack in parallel fashion, but the modified G uses the face that would be turned toward its 3’-neighbor in B-DNA. As a result, the hydrogen-bonding edge of one base presents its functional groups in reverse order to that of the other. If this arrangement could be incorporated in a B-DNA context, mispairing could be a likely consequence. (b) x for both modified G and the neighbor base close to 165’ (an unusually low variant of anti), with base-base contacts, as in t h e d(ApG-AF) global minimum

Chem. Res. Toxicol., Vol. 7, No. 2, 1994 249 (Figure 4d): In this right-handed structure, both bases stack using the “wrong” face. If this structure could be incorporated into a normal B-DNA stack, the two bases would appear as if they had changed positions. This might provide a possible mechanism for a two-base inversion. The a’value of 119O reflects this inversion of position and allows the aminofluorene to contact the 5’-sugar. (c) x for modified G at 345’ (an unusually low value for syn near high anti), with parallel T to AF (proximal and center rings) stacking, with guanine relatively unstacked: This is illustrated in the global minimum for d(TpG-AF) (Figure 4a). This structure has a left-handed backbone (f and a are g+). A variant on this theme [see Figure 4b, d(CpG-AF) #21 has a t , g+ combination for f and a,with the neighboring C using a different face in the stack to AF than that used by d(TpG-AF) #l. (d) x for modified G in the vicinity of 60’ (eyn) with AF-base stacking and a right-handed backbone: This structure dominates the d(NpG-AAF) series, but occurs at 1.3 kcal/mol or higher for d(NpG-AF). (e) A final type of conformer with base-AF stacking, a very different structural type: This structure has a x for modified G a t 150°, with f and a of g+, t and guanine to 5’-sugar contacts [see d(TpG-AF) #2, Figure 4c). Conformation of Dimers: d(NpG-AAF)(N = A, G, C, T). One set of features dominates the numerous lowenergy conformers in this series: they exhibit near-parallel fluorene to base stacking, with guanine contacting the side (and top in some cases) of the 5’-sugar. The modified guanine is syn (x near 60’) in most of the conformers [see, for example, Figures 5c and 6c, d(CpG-AAF) #1 and d(GpG-AAF) #13, with a few in which modified G has x near 150’ [d(TpG-AAF)#2, Figure 6bl. (One low-energy exception to this general pattern exists in d(CpG-AAF) #4 and 5, in which neither cytosine to AAF nor cytosine to G stacking exists.) The various conformers differ in the glycosyl torsion of the neighboring base, in the rings of the AAF system used in the stack, and in the backbone. No structures appear, to 3.0 kcal/mol, in which the modified guanine has x near 347’ or 230’. Conformers of this type were important in the d(NpG-AF) series, but presumably are destabilized here because of steric hindrance caused by the presence of the acetyl group. Summary of Dimer Studies. (1) AminofluoreneBase Stackingvs Base-Base Stacking. In some earlier literature, these categories were often considered as exclusive alternatives. In some cases, they are closely related, interconvertible by a modest torsion angle shift. In any event, in the low-energy minima, both guanine and aminofluorene generallyfiid stacking partners: one selects the neighboring base, the other the neighboring sugar or phosphate. In conformers with aminofluorene-base stacking, the amine and base planes are commonly near-parallel, while base-base stacking will sometimes occur at dihedral angles of 20’ or greater. In such cases, greater stabilization may result from optimizing the fluorene-sugar contacts than from achieving a good base-base interaction. (2) The Effect of the Acetyl Group. In the deoxydinucleoside monophosphates, the principal effect of the acetyl group appears to be the destabilization of certain conformations for guanine. The glycosyl torsion for modified guanine, x’ for 5‘-modification and x for 3‘modification, remained within the regions 15-66’ and 145155’ in all of the minima to 3.0 kcal/mol. Much of the syn

Shapiro et al.

250 Chem. Res. Toxicol., Vol. 7, No. 2, 1994

Chart A-1. SA Protocol for AF- and AAF-Modified Arbitrary starting Positions ~ l iln i t i a l angles equal to one " m e r of one of the f o l l o w m g sets: 0' 100' 90' 210'

45' 225' 315*

135'

20' 65'

Random Startlng P 0 5 1 t l O n 5 ~ 1 initial 1 angles have random values 116 trials1

110' ZOO' 290' 155' 2 4 5 ' 395'

I vlth the l o w e s t energy and

TaYe the f i n a l

with the lowest energy and all conformations with e n e m i e s UD to 3.0 kcal/mol above It and go to step D .

I

.

I !5ss

vary the carcinogen angles randomlv 18 trials1

I If final conformations have energies > than those o f step C

If final conformations have energies < than those o f s t e p C 1

t h e conformation V L c h the lowest energy and all

conformations with energies up to 3.0 kcal/mol above it

EsQ

Submit conformations of

-

1 m termini angles

vary 14 trlalsi

randamry

(1 One trial is for one starting conformation and is submitted 25 times to DUPLEX/SA.

Table A-1. Simulated Annealing Cooling Schedule beta (moUkcal) 0.2 nblock 20 180.0 bfact 1.04 delta (deg) nstep 25 nloop 40

range and almost all of the anti range were excluded. This limitation should be carried over to larger DNA structures which would be even more hindered. This restriction on AAF conformation did not produce any important difference between AF and AAF modification for the 5'substituted dimers, where the most stable AF-modified conformations involved syn-guanine, and quite similar ones were favored in the AAF series. In the 3'-substituted series, however, the most favored conformations for AF involved glycosyl torsions excluded for AAF. The most stable AAF-modified conformers then involved aminebase stacked structures which occurred at higher energy in the AF series. (3) 5'-Substitution vs 3/-Substitution. One conformational type predominated in the low-energy structures for the 5'mbstituted adducts of both AF and AAF: They were right-handed, with syn-guanine,imperfect base-base stacking, and aminofluorene to 3'-sugar contacts. The AAF-substituted 3'-adducts primarily displayed good base-fluorene stacking, with syn-guanine in contact with the 5'-sugar. The AF-substituted 3'-adducts were more diverse. They included a variety of low-energy structures with guanine anti, good to poor base-base stacking, and fluorene to 5'-sugar and phosphate contacts, as well as several different aminofluorene to base stacked conformer types. (4) The Effect of the Neighboring Base. This was important only in the 3'-AF-substituted series, where the 4 different neighbors exhibited 4 very different global minima. The result suggests that the identity of the 5'neighbor would have the most significant effect on

conformation in larger AF-modified structures. Comparison to Experiment. The earliest structural studies on AAF-modified dinucleoside monophosphates were carried out with ribodimers. In a comparison of r(GAAF-PA) and r(ApG-AAF), a "spectacular" difference in the circular dichroism spectra of the two was noted. In particular, a manyfold enhancement (in comparison to the sum of monomer spectra) in the magnitude of a peak near 260 nm was noted for the latter, but not for the former (24). This differencevanished when the spectra were taken in methanol rather than water (26). This effect was attributed to a strong stacking interaction between adenine and AAF in r(ApG-AAF). Upfield shifts of aromatic protons in r(ApG-AAF) [but also in r(G-AAF-PA)] further supported an adenine-AAF stacking interaction. These results might not necessarily hold for modified DNA dimers, but our calculations certainly suggest that strong adenine-AAF stacking does take place in d(ApG-AAF). The circular dichroism spectra of d(G-AAF-pC) and d(CpG-AAF) have been published (29) and show significant differences in shape and magnitude (the latter has greater intensity). In another study, d(ApG-AF) and d(ApG-AAF) were compared by NMR and CD (27,28). The authors concluded that fluorene-base stacking was more significant in the latter than the former. A detailed NMR study of d(CpG-AAF) has been carried out by Evans and Levine (31),who found strong cytosineAAF stacking, with guanine syn, and the guanine C(8) to amine N bond (a')near 90°. The pseudorotation of the 5/-sugar was 2'-endo, and several protons had shifted upfield in position, attributed to a stacking interaction. The 3'-sugar was 3'-endo in pseudorotation (another conformer was also present), with a trans value of p and gauche+ value for y (another conformer was also present). Our calculations for d(CpG-AAF) (Table 53-15) have revealed four conformers with energies up to 0.3 kcal/mol, and several others up to 1.0 kcal/mol. It is noteworthy that the global minimum and #2 (which differs from #1 only in the 5'-sugar pseudorotation, by 22O) both display all of the above features. They differ from Evans' structure only in one additional detail: his assignment of y' (the amide torsion of the acetyl group) is 180°, while it is reversed, near Oo, in our lowest energy structures. The acetyl-reversed variant of the global minimum occurs at 2.1 kcallmol in Table S-15. Rslsvance to Mutagenesis. AF and AAFmodification of deoxydinucleoside monophosphates has been shown to furnish a number of unexpected and provocative structures. In some cases they represent global minima. If comparable structures were to occur in single-stranded DNA near the replication fork, or served tostabilize bulges, loops, or hairpins in double-stranded DNA, then they could cause mutations by a number of mechanisms. It will be necessary first to incorporate these structures in larger DNA fragments, and to see whether they persist. The present work represents an important prerequisite in defining the most favorable structures and their sequence contexts.

Acknowledgment. This work was supported by NIH CA28038 (S.B. and R.S.), NIH RR 06458 (S.B. and R.S.), U.S.DOE FG0290ER 60931 (S.B. and R.S.), US. DOE Contract DE-AC05-840R21400 with Martin-Marietta Energy Systems (B.E.H.), U S . DOE OHER Field Work Proposal ERKP931 (B.E.H.), and NSF CHE-9015337 (K.E.S. and J.M.). Computations were carried out at the

Conformation of Amine-Modified DNA

initial final a Energy

Chem. Res. Tonicol., Vol. 7, No. 2, 1994 251

Table A-2. Random Starting Conformations for d(ApG-AF) and Energy-Minimized Final Conformations‘ energy B Y PI x’ t t a B Y P2 x d d 73.5 115 287 19 152 131 269 102 283 287 220 259 172 64 -33.5 65 178 193 164 181 264 215 182 62 145 167 119 43

c 88

187

in kcal/mol, torsion angles and pseudorotation parameters in deg.

Table A-3. Comparison of Number of Runs and CPU Hours for DUPLEX/BOTM and DUPLEX/SA*.b AF AAF DUPLEX/BOTM DUPLEX/SA DUPLEX/BOTM DUPLEX/SA no. of runs CPU (h) no. of runs CPU (h) no. of runs CPU (h) no. of rune CPU (h) 16 4897 41 625 12 4897 41 800 8 4897 + 31104 300 700 14 4897 + 31104 300 400 41 400 8 41 lo00 20 4897 4897 10 4897 41 650 13 4897 41 600 8 4897 41 400 8 4897 41 400 300 300 6 300 100 2 4897 + 31104 4897 + 31104 2 4897 41 100 2 4897 41 100 4 4897 41 400 8 4897 41 200

*

a One run of DUPLEX/BOTM constitutes one BOTM trial; 25 runs of DUPLEX/SA constitute one SA trial. Computed at 30 e/run for DUPLEX/BOTM and 30 min/25 runs for DUPLEX/SA.

National Science Foundation’s San Diego Supercomputer Center and the Department of Energy’s National Energy Research Supercomputer Center at Livermore, CA. We thank the Minicomputer Center a t the Teaneck Campus of Fairleigh Dickinson University for access privileges graciously afforded to D.S.

Appendix Search Strategies for SA. The search strategy was developed using the present DUPLEX/BOTM study as a benchmark. The stages were designed to allow the SA trials for the AF- and AAF-modified dimers to reach a final energy equal to or lower than the final energy found by DUPLEX/BOTM. Chart A-1 summarizes the stages. Initial conformations for the AF- and AAF-modified dimers were selected from arbitrary or random starting positions (Chart A-1, steps A and A’), and DUPLEX/SA was employed to yield the energy associated with each initial set of torsion angles. Following the SA runs employing these arbitrary or random (see Table A-2) starting geometries (steps A and A’), two alternate paths are taken. For AF or AAF dimers modified at the 5’-end, the structure with the lowest energy (step B) has its terminal torsion angles (8,y,t) varied randomly and then it is again subjected to DUPLEX/SA (step G). If the carcinogen modification is at the 3’-end, all structures with energies up to 3.0 kcal/mol yielded from steps A and A’ are identified (step C), and the carcinogen torsion angles (a’, 8’ for AF; CY’,8’’ y’, 6’for AAF) are randomly changed before each structure is submitted to DUPLEX/SA (step D). If the geometries generated at this point have energies lower than those of step C, the termini angles of all conformations with energies up to 3.0 kcal/mol generated from step D are randomly altered before applying DUPLEX/SA (steps F and G, respectively). If the energies of the structures resulting from step D are higher than those of step C, then the conformations of step C are again submitted to DUPLEX/SA after having their termini torsions randomly changed (steps E and G, respectively). Simulated Annealing Protocol. Each of the initial conformations was annealed using a starting annealing T of approximately 2517 K with beta, the variable which is the inverse of kT, set at 0.2 mol/kcal. Step I. A randomly chosen angle from the initial structure is rotated by a random amount varying from

-delta/2 to +delta/Z, with delta set a t 180.0°;this translates into a range of -90° to +goo. The energy associated with this new conformation is calculated. Step 11. If the new structure is at a lower energy than the old one, the former is accepted; if the new geometry is at a higher energy, the Metropolis criterion (45) is employed to decide whether the new conformation should be retained. The random selection and rotation of a dihedral angle are repeated 24 more times a t T = 2517 K, since m t e p , the variable which controls the total number of dihedral rotations made before the acceptance is calculated, is assigned the value of 25 in this protocol. Step 111. At the termination of the 25th dihedral rotation, the number of acceptances of dihedral moves is calculated. Steps 1-111 are executed 19 more times a t T = 2517 K since nblock, the parameter which determines the total number of dihedral changes to be attempted at a given T , is equal to 20 in our protocol. Thus, the total number of dihedral variations made at T = 2517 K is 500 ( m t e p X nblock = 25 X 20). Step IV. After the first 500 torsional rotations, the acceptance rate, which is the number of acceptances of dihedral moves per total number of dihedral rotations, is determined. If the acceptance rate is greater than or equal to 25%, beta is multiplied by bfact, whose value in the present protocol is 1.04, to generate a new beta. This decreases T to 2420 K, and steps 1-111 are performed a t this new, lower T for another 500 dihedral perturbations. If the acceptance rate is less than 25 % ,delta, which governs the permitted range of random rotation of the randomly chosen angle, is divided in half (-45O to +45O). Consequently, for the subsequent 500 dihedral rotations, steps 1-111 are performed at the same T as the previous 500 dihedral perturbations, with the range of random rotation of the dihedral angles cut in half. This is necessary to maintain an acceptance rate of at least 25 75, Steps I-IV are repeated a total of 40 times, a t which point the SA run ends, since nloop, the parameter which determines the number of T’s in one SA run equals 40 in our protocol. The variables beta, delta, mtep, nblock, bfact,and nloop constitute the cooling schedule of SA. The current schedule is an optimized one, derived from trials that varied these parameters. The cooling schedule optimization is

252 Chem. Res. Toxicol., Vol. 7, No. 2, 1994

described in Sidawi,2 and the current protocol is summarized in Table A-1. Supplementary Material Available: Tables giving starting conformations (deg) for the potential energy minimizations with BOTM alone, for 4897 trials (Table S-1) and 31104 trials (Table S-2))and minimum-energy conformations found for each structure using the two searches with BOTM and the SA search, of d(G-AF-pC) (Table S-3)) d(G-AF-pT) (Table S-4)) d(G-AAFpC) (Table S-5), d(G-AAF-pT) (Table S-6),d(G-AF-PA) (Table S-7), d(G-AF-pG) (Table S-8)) d(G-AAF-PA) (Table S-9), d(GAAF-pG) (Table S-10)) d(CpG-AF) (Table S-111, d(TpG-AF) (Table S-12))d(ApG-AF)(Tables-13))d(GpG-AF) (Table S-14), d(CpG-AAF) (Table S-15), d(TpG-AAF) (Table S-16)) d(ApGAAF) (Table S-17), and d(GpG-AAF) (Table 5-18) (24 pages). Ordering information is given on any current masthead page.

References Weissburger,J. H. (1988)Past, present and future role of carcinogen and mutagenic N-substituted aryl compounds in human cancer causation. In Carcinogenicand Mutagenic Responses to Aromatic Amines and Nitroarenes (King,C. M., Romano,L. J., and Schueltze, D., Eds.) pp 3-19, Elsevier, New York. Kriek, E. (1992)Fifty years of research onN-acetyl-2-aminofluorene, one of the most versatile compoundsin experimentalcancerresearch. J. Cancer Res. Clin. Oncol. 118,481-489, Burnouf, D., Koehl, P., and Fuchs, R. P. P. (1989) Single adduct mutagenesis: Strong effect of the position of a single acetulaminofluorene adduct &hin a mutation hot spot. Pro;. NatLAcad. Sci. U.S.A. 86,4147-4151. Veaute, X., and Fuchs, R. P. P. (1991) Polymorphism in N-2acetylaminofluoreneinduced DNA structure as revealed by DNase I footprinting. Nucleic Acids Res. 19, 5603-5606. Belguise-Valladier,P., and Fuchs, R. P. P. (1991)Strong sequencedependent polymorphism in adduct-induced DNA structure: analysis of singleN-2-acetylaminofluoreneresidues bound with the NarI mutation hot spot. Biochemistry 30, 10091-10100. Lambert, I. B., Napolitano, I. B., and Fuchs, R. P. P. (1992) Carcinogen induced frameshift mutagenesisin repetitive sequences. Proc. Natl. Acad. Sci. U.S.A. 89, 1310-1314. Shibutani, S., Moriya, M., and Johnson, F. (1989) Chemistry and mutagenic properties of acetylaminofluorene-derivedDNA adducta. Proc. Am. Assoc. Cancer Res. 30,128. Gupta, P. K., Pandrangi, R. G., Lee, M.-S., and King, C. M. (1991) Induction of mutations by N-acetoxy-N-acetyl-2-aminofluorene modified M13 viral DNA. Carcinogenesis 12,819-824. Shibutani, S., and Grollman, A. P. (1993) On the mechanism of frameshift (deletion)mutagenesisinvitro. J.Biol. Chem.268,1170311710. Garcia, A., Lambert, I. B., and Fuchs, R. P. P. (1993) DNA adductinduced stabilization of slipped frameshift intermediates within repetitive sequences: implications for mutagenesis. Proc. Natl. Acad. Sci. U.S.A. 90, 5989-5993. Moriya, M., Takeshita, M., Johnson, F., Peden, K., Will, S., and Grollman, A. P. (1988) Targeted mutations induced by a single acetylaminofluoreneDNA adduct in mammalian cells and bacteria. Roc. Natl. Acad. Sei. U.S.A 85, 1586-1589. Mah, M. C.-M., Boldt, J., Culp, S. J., Maher, V. M., and McCormick, J. J. (1991) Replication of acetylaminofluorene-adductedplasmids in human cells: Spectrum of base substitutions and evidence of excision repair. Proc. Natl. Acad. Sci. U.S.A. 88,10193-10197. Carothers, A. M., Urlaub, G., Steigerwalt, R. W., Chasin, L. A., and Grunberner. D. (1986) Characterization of mutations induced bv 24N-acebxy-N-acetyl) aminofluorenein thedihydrofolate redu& gene of cultured Chinese hamster ovarycells. Proc. Natl. Acad. Sci. U.S.A. 83,9556-9560. (14) Carothers, A. M., Steigerwalt, R. W., Urlaub, G., Chasin, L. A., and Grunberger, D. (1989) DNA base changes and RNA levels in N-acetoxy-2-aminofluorene-induceddihydrofolate reductase mutanta of Chinese hamster ovary cells. J. Mol. Biol. 208, 417-428. (15) Bichara, M., and Fuchs, R. P. P. (1985)DNA binding and mutation spectra of the carcinogenN-2-aminofluorenein E. coli: a correlation between the conformation of the premutagenic lesion and the mutation specificity. J. Mol. Biol. 183, 341-351. (16) Gupta, P. K., Lee, M. S., and King, C. M. (1988) Comparison of mutagenesis induced in single-and double-stranded M13viral DNA by treatment with N-hydroxy-2-aminofluorene.Carcinogenesis 9, 1337-1345. (17) Gupta, P. K., Johnson, D. L., Reid, T. M., Lee, M.-S., Romano, L. J., and King, C. M. (1989) Mutagenesis by single site-specific arylamine-DNA adducts. J. Biol. Chem. 264, 20120-20130.

Shapiro et al. (18) Reid, T. M., Lee, M.-S., and King, C. M. (1990) Mutagenesis by site-specific arylamine adducts in plasmid DNA Enhancing replication of the adducted strands alters mutation frequency. Biochemistry 29,61534161. (19) Norman, D., Abuaf, P., Hingerty, B. E., Live, D., Grunberger, D., Broyde, S., and Patel, D. J. (1989) NMR and computational characterizationof the N-(deoxyguancmin-&yl)aminofluorene adduct [(AF)Gl opposite adenosine in DNA (AF)G[synl.A[antil pair formation and ita pH dependence. Biochemistry 28, 7462-7476. (20) Cho,B.P.,Beland,F. A.,andMarques,M.M. (1994)NMRstructural studies of a 15-mer DNA duplex from a rw protooncogene modified with the carcinogen 2-aminofluorene: conformationalheterogeneity. Biochemistry 33 (in press). (21) Eckel, L. M., and Krugh, T. R. (1994) 2-Aminofluorene modified DNA duplex exista in two interchangeable conformations. Nature: Structural Biology 1, 89-94. (22) Broyde, S.,and Hingerty, B. E. (1983) Conformation of 2-aminofluorene-modified DNA. Biopolymers 22,2423-2441. (23) OHandley, S. F., Sanford, D. G., Xu, R., Lester, C., Hingerty, B. E., Broyde, S., and Krugh, T. R. (1993)Structural characterization of an N-acetyl-2-aminofluorene(AAF)modified DNA oligomer by NMR, energyminimizationand moleculardynamics. Biochemistry 32, 2481-2497. (24) Grunberzer. D.. Nelson. J. H.. Cantor. C. R.. and Weinatein. I. B. (1970) Cking and conformational properties of oligonucleotides modifiedwiththe carcinogenN-2-acetylaminofluorene. Roc. Natl. Acad. Sci. U.S.A. 66, 488-494. Fucha, R. P. P., and Daune, M. (1971) Changes of etability and conformationof DNA followingthe covalentbindingof a carcinogen. FEBS Lett. 14, 206-208. Nelson, J. H., Grunberger, D., Cantor, C. R., and Weinatein, I. B. (1971)Modificationof ribonucleic acid by chemicalcarcinogensIV. Circular dichroism and proton magnetic resonance studies of oligonucleotidesmodified with N-2-acetylaminofluorene. J. Mol. Biol. 62, 331-346. Leng, M., Ptak, M., and Rio, P. (1980) Conformation of acetylaminofluorene and aminofluorene modified guanosine and guanosine derivatives. Biochem. Biophys. Res. Commun. 96, 1095. Santella, R. M., Kriek, E., and Grunberger, D. (1980) Circular dichroism and proton magnetic resonance studies of d(ApdG) modified with 2-aminofluorene and 2-acetylaminofluorene. Carcinogenesis l, 897-902. Rio, P., Malfoy, B., Sage, E., and Leng, M. (1983) Conformation of oligonucleotidesand nucleic acids modified with 2-aminofluorene or 2-acetylaminofluorene. Enuiron. Health Perspect. 49,117-123. Alderfer, J. L., Lilga, K. T., French, J. B., and Box, H. C. (1984) W-NMR studies of the effect of the carcinogenacetylaminofluorene on the conformation of dinucleoside monophosphate. Chem.-Biol. Interact. 48, 69-80. Evans, F. E., and Levine, R. A. (1988) NMR study of stacking interactions and conformational adjustments in the dinucleotidecarcinogenadduct 2’-deoxycytidyl-(3~5)-2’-deoxy-8-(N-fluoren-2y1acetamido)guanosine. Biochemistry 27,3046-3055. van Houte, L. P. A., Westra, J. G., Retel, J., and van Grondelle, R. (1991) A circular dichroism study on the conformation of d(CGT) modified with N-acetyl-2-aminofluoreneor 2-aminofluorene. J. Biomol. Struct. Dyn. 9, 45-59. Hingerty, B. E., and Broyde, S. (1982) Conformation of the deoxydinuclecmidemonophosphate d(CpdG) modified at carbon 8 of guanine with 2-(acetylaminofluorene).Biochemistry 21, 32433252. Hingerty, B. E., and Broyde, S. (1982) Base displacement in AAF and AF modified dCpdG syn and anti guanine. Int. J. Quantum Chem., Quantum Biol. Symp. 9, 125-136. Shapiro, R., Hingerty, B. E., and Broyde, S. (1989) Minor-groove binding models for acetylaminofluorenemodifiedDNA. J.Biomol. Struct. Dun. 7, 493-513. Powell, M. (1964) An efficient method for finding the minimum of a function of several variables without calculating derivatives. Comput. J. 7, 155-162. Hingerty, B. E., Figueroa, S.,Hayden, T., and Broyde, S. (1989) Prediction of DNA structure from sequence: a build-up technique. Biopolymers 28, 1195-1222. Srinivasan, A. R., and Oleon, W. K. (1980) Polynucleotide conformations in real solutions-a preliminary theoretical estimate. Fed. Proc., Fed. Am. Soc. Exp. Biol. 39, 2199. Taylor, E. R., and Olson, W. K. (1983)Theoretical studies of nucleic acid interactions. I. Estimates of conformational mobility in intercalated chains. Biopolymers 22,2667-2702. Altona, C., and Sundaralingam, M. (1972) Conformatienal analysis of the sugar ring in nucleosides and nucleotides. A new description 94,8205using the concept of pseudorotation. J. Am. Chem. SOC. 8212. Saenger, W. (1984) Principles of Nucleic Acid Structure, Springer Verlag, New York.

Conformation of Amine-Modified DNA (42) Broyde, S., Wartell, R., Stellman, S., and Hingerty B. (1978) Minimumenergyconformatiom of DNA dimericsubunits potential energy calculations for d(GpdC), d(ApdA), d(CpdC), d(GpdG) and d(TpdT). Biopolymers 17,1485-1506. (43) Kirkpatrick, S., Gelatt, C. D., and Vecchi, M. P. (1983)Optimization by simulated annealing. Science 220, 671-680. (44) Cerny, V. (1985) Thermodynamical approach to the traveling salesmanproblem: An efficientsimulationalgorithm. J.Opt. Theory Appl. 46,41-51. (45) Metropolis, N., Rosenbluth, A. W., Rosenbluth, M. N., and Teller, A. H. (1953) Equation of state calculations by fast computing machines. J. Chem. Phys. 21, 1087-1092. (46) Lukashin, A. V., Engelbrecht, J., and Brunak, S. (1992) Multiple alignment using simulated annealing: branch point definition in human mRNA splicing. Nucleic Acids Res. 20, 2511-2516. (47) Moskowitz, J. W., Schmidt, K. E., Wilson, S. R., and Cui, W. (1988) The application of simulated annealing to problems of molecular mechanics. Znt. J. Quantum Chem. 22,611417. (48) Donnelly, R. A. (1987) Geometry optimization by simulated annealing. Chem. Phys. Lett. 136, 274-278. (49) Bounds, D. G. (1987) New optimization methods from physics and biology. Nature 329, 215-219. (50) Aarta,E., and Korst, J. (1989)Simulated Annealingand Boltzmann Machines; Wiley, New York. (51) Van Laarhoven, P. J. M. (1988) Theoretical and Computational Aspects of Simulated Annealing, Centre for Mathematics and Computer Science, Amsterdam. (52) Chandler, D. (1987)Introduction to Modern Statistical Mechanics, Oxford University Press, New York. (53) Van Laarhoven, P. J. M., and A&, E. H. L. (1987) Simulated Annealing: Theory and Applications, D. Reidel, Boston. (64) Wilson,S.R., Cui, W., Moskowitz, J. W., and Schmidt, K. E. (1988) Conformationalanalysisof flexiblemolecules: Locationof the global minimum energyconformation by the simulated annealing method. Tetrahedron Lett. 29,4373-4376. (55) Rajasekaran, S. (1992) Nested annealing: a provable improvement to simulated annealing. Theor. Comput. Sci. 99, 157-176.

Chem. Res. Toxicol., Vol. 7, No. 2, 1994 263 (56) Tovey, C. A. (1988) Simulated annealing. In Simulated Annealing (SA)and Optimization: Modern Algorithms with VLS-I,Optimal Design and Missile Defense Applications (Johnson,M. W., Ed.) pp 389-407, American Sciences Press, Syracuse, NY. (57) Brooks, D. G., and Verdini, W. A. (1988) Computational experience with generalizedsimulated annealing over continuouavariables. In Simulated Annealing (SA)and Optimization: Modern Algorithms with VLS-Z, Optimal Design and Missile Defense Applications (Johnson, M. W., Ed.) pp 425-449, American Sciences Press, Syracuse, NY. (58) Howard, A. E., and Kollman, P. E. (1988) An analysis of current methodologiesfor conformational searching of complex molecules. J. Med. Chem. 31,1669-1675. (59) Morales, L. B., Garduno-Juarez, R., and Romero, D. J. (1991) Applicationsof Simulatedannealingto the multiple-minimaproblem in small peptides. J. Biomol. Struct. Dyn. 8, 721-735. (60) Brunger, A. T., and Krukomki, A. (1990)Slow-coolingprotocolsfor crystallographic refinement by simulated annealing. Acta Crystallogr. A46, 585-593. (61) Basu, A., and Fraser, N. (1990) Rapid determination of the critical temperature in simulated annealing inversion. Science 249,14091412. (62) Evane,F.E.,andLevine,R.A. (1986)Conformationandconfiiation at the centralamine nitrogen of a nucleotide adduct of the carcinogen &(acetylaminofluorene)as studied by 1SC and 1BN NMR spectroscopy. J. Biomol. Struct. Dyn. 3, 923-934. (63) Brower, R. C., Vasmatzis, G., Silverman, G., and Delisi, C. (1993) Exhaustive conformational searches and eipulated annealing for models of lattice peptides. Biopolymers 33, 329-334. (64) Miao, Y.-S. (1992)Calculationsand Build-up Techniquesto Predict DNA Subunits Modified by the Carcinogens2-Aminofluoreneand 2-Acetylaminofluorene. Ph.D. Thesis, Department of Chemistry, New York University. (65) Wang, A., Quigley, G., Kolpak, F., Van der Marel, G.,van Boom, J., and Rich, A. (1981) Left handed double helical DNA variations in the backbone conformation. Science 211, 171-176.