A Structural Basis for a Phosphoramide Mustard-Induced DNA Interstrand Cross-Link at 5'-D(GAC) QING DONG*, DANIEL Barskyt, MICHAEL E

Proc. Natl. Acad. Sci. USA Vol. 92, pp. 12170-12174, December 1995 Biochemistry

A structural basis for a phosphoramide mustard-induced DNA interstrand cross-link at 5'-d(GAC) QING DONG*, DANIEL BARSKYt, MICHAEL E. COLVINt, CARL F. MELIUSt, SUSAN M. LUDEMANt, JONATHAN F. MORAVEK§, 0. MICHAEL COLVINO, DARELL D. BIGNER*, PAUL MODRICHII, AND HENRY S. FRIEDMAN*§** Departments of *Pathology and §Pediatrics, IDuke Comprehensive Cancer Center, ItHoward Hughes Medical Institute and Department of Biochemistry, Duke University Medical Center, Durham, NC 27710; tM.S. 9214, Sandia National Laboratories, Livermore, CA 94550-0969; and tOncology Center, Johns Hopkins University, Baltimore, MD 21287 Contributed by Paul Modrich, September 14, 1995

ABSTRACT Phosphoramide mustard-induced DNA in- and quantum chemistry were applied to study the structural terstrand cross-links were studied both in vitro and by com- basis for the linkage induced by nitrogen mustards. puter simulation. The local determinants for the formation of phosphoramide mustard-induced DNA interstrand cross- MATERIALS AND METHODS links were defined by using different pairs of synthetic oligonucleotide duplexes, each of which contained a single poten- In Vitro Studies for PM-Induced DNA Cross-link Forma- tially cross-linkable site. Phosphoramide mustard was found tion. T4 polynucleotide kinase was purchased from New to cross-link dG to dG at a 5'-d(GAC)-3'. The structural basis England Biolabs. [,y-32P]ATP (6000 Ci/mmol; 1 Ci = 37 GBq) for the formation of this 1,3 cross-link was studied by molec- was purchased from Amersham. Oligonucleotides were pur- ular dynamics and quantum chemistry. Molecular dynamics chased from Oligo, etc. (Wilsonville, OR). PM (cyclohexylam- indicated that the geometrical proximity of the binding sites monium salt) was a gift from the Drug Synthesis and Chem- also favored a 1,3 dG-to-dG linkage over a 1,2 dG-to-dG istry Branch, National Cancer Institute. Dechlorophosphor- linkage in a 5'-d(GCC)-3' sequence. While the enthalpies of amide mustard [HOP(O)(NH2)N(CH2CH2Cl)(CH2CH3)] was 1,2 and 1,3 mustard cross-linked DNA were found to be very made by a multistep organic synthesis procedure. (The syn- close, a 1,3 structure was more flexible and may therefore be thesis protocol will be made available by S.M.L.) PM or in a considerably higher entropic state. dechlorophosphoramide mustard was dissolved in 50 mM sodium phosphate, pH 7.0, as 500 mM stock immediately Cyclophosphamide is an important alkylating agent that has before drug treatment. been widely used in treating various malignant tumors (1). Oligonucleotides were purified by electrophoresis in a de- DNA interstrand cross-links appear to be the primary cyto- naturing 20% polyacrylamide gel (ref. 13, pp. 11.23-11.28) and toxic adduct and the base of antitumor activity of phosphor- eluted by soaking the crushed gel pieces containing the correct amide compounds (2, 3). Many studies have suggested that size in distilled water at 14°C. The oligonucleotide eluate was DNA interstrand cross-links induced by alkylating agents passed through a Sephadex G-25 column to remove urea and prevent the separation of the strands of the DNA helix and thus salts, and the oligonucleotides were concentrated in a Speed- inhibit DNA replication (4-6). Further clarification of the Vac (Savant), divided into aliquots, and stored at -70°C. One molecular interactions between DNA and cross-linking agents of the oligonucleotides in each pair was phosphorylated at the is necessary to understand the cellular responses to DNA 5' end by T4 polynucleotide kinase (ref. 13, pp. 5.68-5.71). The interstrand cross-links and the manner by which such lesions reaction was terminated by adding EDTA to a final concen- are recognized and repaired in mammalian cells. tration of 10 mM, and the solution was passed through a G-25 Phosphoramide mustard [PM, HOP(O)(NH2)N(CH2CH2Cl)2], column to remove ATP. The labeled oligonucleotide was the pharmacologically active metabolite of cyclophosphamide, purified again as described above. primarily alkylates the N7 atom of guanines in DNA (7, 8). It Two complementary oligonucleotides (1 ,uM each) were has been postulated that the DNA sequence 5'-CG-3' imparts mixed in 50 mM sodium phosphate, pH 7.0, heated at 68°C for a topologic and geometric condition favoring formation of 3 min, and allowed to cool to room temperature over several DNA interstrand cross-links by nitrogen mustard compounds hours just before the drug treatment. The annealed DNA (4). An early computer modeling study suggested that PM oligonucleotide duplexes or single-strand oligonucleotides (at favors a 5'-GC-3' DNA sequence (resulting in a 1,2 cross-link) 1 j,M) were treated with various concentrations of PM in 50 (9); however, several experimental studies with nitrogen mus- mM sodium phosphate for 4 hr at room temperature. The tard [CH3N(CH2CH2Cl)2] have indicated that 5'-GNC-3' (a reaction mixture was stored at -20°C or electrophoresed 1,3 cross-link) is the sequence for the nitrogen mustard- immediately. The gel was run at 1600 V in 90 mM Tris borate/2 induced DNA cross-link (10-12). Since DNA cross-links in- mM EDTA buffer, pH 8.0, for at least 2 hr before loading. An duced by nitrogen mustard and PM are structurally similar, it end-labeled oligonucleotide ladder was loaded on the same gel is conceivable that PM can form a 1,3 instead of a 1,2 linkage. (Pharmacia). After electrophoresis at 1800 V for 2 hr at 50°C, However, no experimental evidence has been published to DNA species were visualized by autoradiography. Cross-linked support either model. oligonucleotide duplexes visualized in this manner were ex- We have designed several pairs of synthetic oligonucleotide cised from the gel and purified as described above. Oligonu- duplexes to define the local determinants of PM-induced cleotides were sequenced by the Maxam-Gilbert method, cross-link formation in vitro. In addition, molecular dynamics using the G, G+A, C, and T+C reactions (14, 15).

The publication costs of this article were defrayed in part by page charge Abbreviations: PM, phosphoramide mustard; NNM, nornitrogen mus- payment. This article must therefore be hereby marked "advertisement" in tard. accordance with 18 U.S.C. §1734 solely to indicate this fact. **To whom reprint requests should be addressed. 12170 Downloaded by guest on September 24, 2021 Biochemistry: Dong et al. Proc. Natl. Acad. Sci. USA 92 (1995) 12171 Structural Studies of Nornitrogen Mustard (NNM) DNA The internal energies of the bare NNM linkages were deter- Cross-link The structures of both 1,2 and 1,3 linkages were mined by ab initio quantum chemistry calculations. To this end, simulated by using molecular dynamics and were analyzed by the NNM cross-link was isolated from the DNA at regular quantum chemistry. Three 12-bp duplexes-i.e., pairs P1 and P3 intervals along the molecular dynamics trajectory, and the gua- (see Table 1) and a control pair, d(CATGTAG3G2CTAA)-d- nine N7s were replaced with hydrogens whose positions were (TTAG'CCTACATG)-were built in the canonical B-DNA adjusted to the appropriate C-H bond distance. Each 16-atom form by using the program QUANTA4.o/CHARMm (16). Because structure was then optimized with the GAUSSLAN 94 program (30) the primary interest was the effect of the physical linkage on at the Hartree-Fock level oftheory, using a 6-31G* basis set while the DNA and because molecular dynamics parameters for the constraining the terminal hydrogens and the ,3 carbons to their phosphoramide moiety have not been developed, all modeling molecular dynamics-derived positions (31). was performed on NNM [HN(CH2CH2Cl)2] instead of PM- i.e., HN in place of HOP(O)(NH2)N (1). This substitution could affect steric and electrostatic interactions. NNM, like RESULTS other nitrogen mustard diadducts, nitrogen mustard (17), and Design of Duplex Oligonucleotides Containing Only a Single melphalan, is expected to have a neutral net charge at pH 7. Cross-linkable Site. Different pairs of oligonucleotides were However, chlorambucil (18) and PM have been shown to be synthesized that contained a single potentially cross-linkable site. anions at pH 7 (19). Sterically, the much larger phosphoramide As shown in Table 1, pairs P1, P2, and P4 were expected to be moiety may reduce the range of available conformations of the link, but this is not expected to alter the computational results (see discussion below). The interstrand NNM cross-link was added to pairs P1 and P3 by replacing the terminal chloride atoms with the appropriate N7 atoms, whereas the control pair was left unmodified. For visualization purposes a PM cross- link was built into pair P1 as shown in Fig. 1. In our simulations the NNM link atoms were only weakly charged, and all three systems had a net charge of -22 owing to the 22 phosphate groups along the backbones. Counter-ions were not included because they may require a long equilibration time in the presence of explicit water (more than 50 ps) (20), and a comparative study found that inclusion of counter-ions in the absence of explicit water did not significantly improve the resulting DNA structural properties (21). The linked structures were minimized slightly to remove the unusually large strains introduced by inserting the link. Using the program X-PLOR (22), we added a sphere of radius 48 A ofTIP3P (23) water molecules within 2.6 A of the solute (DNA and NNM) molecules. Each total system consisted of approximately 6150 atoms. The original water density was maintained by applying a 49.2-A repulsive spherical shell having a harmonic force constant of 20 kcal/mol per A2 (1 kcal = 4.18 kJ). The SETLE algorithm (24) was used to maintain the internal equilibrium of each water molecule for improved energy conservation. After the initial construction and minimization, a recom- mended preparation protocol was followed wherein the DNA was held fixed while the water was heated and cooled twice from 20 to 300 K over equal intervals totaling 4 ps (25). With no atoms held fixed, the systems were then subjected to 400 steps of conjugate gradient minimization and then heated from 0 K to 300 K by periodically rescaling the velocities over a period of 1 ps. Similar results were achieved when a heating period of 10 ps with a correspondingly shorter equilibration period thereafter was used. All molecular dynamics and energy minimizations were carried out by an in-house program CCEMD (26, 27) which employed the parameters and force fields found in the CHARMm22 program and parameter ("RTF") files (28). A 12-A electrostatic cutoff for all atoms was employed in the method devised by Levitt, which avoids some of the traditional problems such as local heating where the cutoff occurs (29). The dynamics are accomplished by Verlet integration with time steps of 1.0 fs. All the results from dynamics studies presented here are from 200-ps trajectories carried out in the microcanonical ensemble (constant number, volume, and energy) where the first 20 ps of each simulation has been regarded as an equilibration period and is not included in the analyses. As has been noted by a referee, although a B-DNA form is the most reasonable starting structure, it creates some bias in these sim- FIG. 1. An oligonucleotide duplex cross-linked by PM as demon- ulations which cannot be expected to vanish in the 200-ps strated by computer modeling. Oligonucleotide duplex (P1) was simulation time. Further details and parameters are available constructed in the program MACROMODEL V4.o. A PM linkage (arrow) from the authors upon request (Internet: [email protected]). was built between the N7 positions of guanines 1 and 3. Downloaded by guest on September 24, 2021 12172 Biochemistry: Dong et al. Proc. Natl. Acad. Sci. USA 92 (1995) Table 1. Oligonucleotides used as substrates for DNA interstrand tard, which forms only monoadducts. As shown in Fig. 2, slow- cross-link formation migrating bands were generated with pair P1 upon treatment with Pair DNA sequence Strand PM, with a mobility similar to that of the 32-mer. The upper band was the cross-linked duplex as confirmed by DNA chemical Pi AATTCATGTAGACTAA Sv1 sequencing. The lower band existed across all experimental GTACATCTGATTTCGA SV2 conditions in the overexposed film (with or without PM treat- PiC AATTCATGTATACTAA Sv1c ment). In the control pair P1C, which differed from pair P1 only GTACATATGATTTCGA SV2C by a single base pair, no PM-induced DNA cross-link was formed P2 AATTAATGTTGACTAA SV3 under identical conditions. At a 10-mM dose, about 0.1% of the TTACAACTGATTTCGA SV4 duplex was cross-linked after 4-hr drug treatment as judged by P2C AATTAATGTTTACTAA SV3C direct measurement of radioactivity. The amount of cross-linked TTACAAATGATTTCGA SV4C DNA was further increased (up to 1%) upon postincubation P3 AATTCATGTAGCATAA SV5 without PM at room temperature for another 24 hr. In addition, GTACATCGTATTTCGA SV6 a significant reduction in mobility of both single- and double- P4 AATTAATGTAGACTAA SV7 stranded oligonucleotides was observed when they were treated TTACATCTGATTTCGA SV8 with PM. This effect, however, was not observed with dechloro- P5 AATTCATGTGAACTAA SV9 phosphoramide mustard at a concentration up to 100 mM (data GTACACTTGATTTCGA Sv1o not shown). As depicted in Fig. 3, PM-induced DNA cross-links Oligonucleotides were hybridized to their complementary strands were formed only in duplexes that contain a 1,3 potentially and used as the substrates for DNA cross-link formation. The bold- cross-linkable site (P1, P2, and P4) and not in duplexes that had faced sequences are restriction sites for Acc I in pair P1 and pair P4 a 1,2 potentially cross-linkable site (P3). Pair P5, which contains and HincIl in pair P2. The underlined bases indicate a single nucleotide a 1,4 potentially cross-linkable site, was not cross-linked by PM change in the control pairs. under the identical conditions (data not shown). substrates for 1,3 DNA cross-link formation, while pairs P3 and As noted in Fig. 3, a 5'-TGAC-3' site (in P2) was also P5 were expected to support 1,2 and 1,4 DNA cross-link forma- cross-linked by PM. Its cross-linking efficiency was the same as tion, respectively. The potentially cross-linkable site resided in the compared with the P1 duplex despite the presence of T in front recognition sequence forAcc I restriction enzyme in pairs P1 and of G. This result does not concur with an earlier computer P4 and in a HincII recognition sequence in pair P2. The A12 and modeling study, which predicted that thymine methyl groups C13 in pair P1 were swapped to form pair P3 in order to provide adjacent to guanine will inhibit interstrand cross-link forma- a site for 1,2 DNA cross-link formation. Pair P4 had the same tion at 5'-GC-3' (9). sequence as pair P1, except the fifth G-C pair was changed to A-T. DNA Chemical Sequencing of the Cross-linked Product. The In addition, two control pairs corresponding to pairs P1 and P2 volatile secondary amine piperidine has been shown to create with only a single base pair changed at the 1,3 cross-linkable site strand breaks at sites of 7-alkylguanines (16, 32). To reveal the were also synthesized. guanines that were alkylated by PM in the duplex oligonucleo- Cross-link Forma- tides, DNA chemical sequencing and piperidine hydrolysis of the Evidence of PM-Induced DNA Interstrand cross-linked products were employed. The cross-linked product tion. One ofthe oligonucleotides in each pair was end-labeled and was purified from the denaturing PAGE, then heated at 90°C hybridized with the complementary strand. Oligonucleotides in were treated with various concentrations of PM for different either with or without piperidine. As shown Fig. 4, without lengths of time, and formation of cross-linked products was piperidine treatment, the majority of the cross-linked material evaluated by denaturing PAGE. For control purposes, oligonu- (P1) remained intact. In contrast, the cross-linked product (SV1- cleotides were also treated with dechlorophosphoramide mus- AGC GAC GAC TGAC GAC TAC 0 510 0 510 0 510 0 510 ss ds ss ds 0 5 10 0 5 10 0 5 10 0 5 10 ,* -XL ...... 1*, L - 32 mer

- XL - 32 mer d ,- 1 e 40

- 16mer ,,_ 16mer

- 8 mer

FIG. 3. Formation of PM-induced DNA cross-link in different FIG. 2. Formation of PM-induced DNA cross-link in a duplex of pairs of oligonucleotides. Odd-numbered oligonucleotides of each pair oligonucleotides. Oligonucleotides (pairs P1 and P1C) were hybridized were end-labeled and hybridized to the corresponding complementary with their complementary strands (ds) and incubated with 0, 5, or 10 strand. Different duplexes were then incubated with 0, 5, and 10 mM mM PM for 4 hr at room temperature as indicated. SV1 and SV1C PM at room temperature for 4 hr. AGC (P3), GAC (P1, left), GAC (Table 1) from each pair (ss) were treated in the same manner as the (P4, right), and TGAC (P2) indicate the potentially cross-linkable sites control. XL indicates the cross-linked oligonucleotide duplex; 8-, 16-, of each pair. On the far right is the oligonucleotide ladder, ranging and 32-mer indicate the positions of radiolabeled oligonucleotide from 8-mer to 32-mer with increments of 2. XL indicates the cross- ladder electrophoresed on the same gel. linked oligonucleotide duplex. Downloaded by guest on September 24, 2021 Biochemistry: Dong et al. Proc. Natl. Acad. Sci. USA 92 (1995) 12173

T G G T intervals along the molecular dynamics trajectories, were used + + + + as constraints for quantum chemistry studies of the linkage. A 1 CC G 2 3 4 G A CC The quantum chemical energies fluctuated broadly over a

....I .v 5-kcal/mol range, favoring neither the 1,2 nor the 1,3 linkage. gm -AL These results suggest that reaction enthalpy alone does not favor the 1,3 linkage. If the 1,3 linkage is favored by nature, some other effect must be determining this outcome. If the cross-linking reactions are in equilibrium, then a favorable entropy change must dominate the change in the free energy (AAG); i.e., the 1,3 cross-link reaction yields a more entropic product than the 1,2 3' cross-link reaction. Alternatively, if the distribution of products is

C- kinetically determined, then the 1,3 linkage must have a lower A- reaction barrier AGt or be statistically more likely to achieve a G- 1^T conformation favoring the cross-linking reaction (greater Arrhe- S' nius A factor). Since the reaction enthalpies for the 1,2 and 1,3 -G cross-link reactions are comparable, the Hammond postulate (34) .._.... implies that the reactions have comparable transition enthalpies - ... 5' I (AHt), leaving the transition entropy and kinetic feasibility as possible discriminants. A reasonable diagnostic of the relative reaction and transition entropies of the cross-linked DNA is the flexibility of the cross-link, which is directly related to the a dG-to-dG N7-to-N7 separation distance. Table 2 presents the ON mean N7-to-N7 distances and their standard deviations for both the 1,2 and the 1,3 linkages, calculated for the 175 ps of 0. . the simulation. The standard deviations are about twice as large for the 1,3 linkage as for the 1,2 linkage, indicating that the former is much more flexible, hence, is in a more entropic FIG. 4. Hydrolysis of oligonucleotide duplexes cross-linked by PM. state. Chemical sequencing ladders of SV1 (Table 1) are displayed on the left The kinetic feasibility of the cross-linking reaction can be (C, T+C, G+A, and G lanes). Lanes labeled 1 and 2 contained the estimated from the molecular dynamics trajectories; the cross- PM-cross-linked duplexes (SV1-labeled) with or without piperidine hydrolysis, respectively. Lanes 3 and 4 show the cross-linked product whose 5' end of SV2 was labeled, without and with piperidine 12.0 1,2 unlinked hydrolysis, respectively. The right-hand side shows the sequencing 1,2 unlinked ladder of SV2 (G, G+A, T+C, and C lanes). Side labels indicate the .... 1,2 linked bases of the cross-linkable site. 10.0 end-labeled) was specifically hydrolyzed by piperidine at the

position of Gi1. The cross-linked product (SV2-end-labeled) was a1) hydrolyzed only at G8 in the oligonucleotide. We thus conclude 0c; 8.0 that these two guanines make up the site for the formation of the ._L x DNA cross-link induced by PM. bRI !tIll 'arT Structure Study of PM-Induced DNA Interstrand Cross- 6.0 link The calculated energies of the two linked systems (1,2 and 1,3 cross-linked DNA) were not directly comparable because ' > '.*tg[ :f et.; I,j .sile, they contained different numbers of solvent molecules. For 4.0 this reason, at 1-ps intervals along the molecular dynamics 0.0 50.0 100.0 150.0 200.0 trajectories, the total solute energies of the bare DNA were time calculated. To reduce the errors that would occur for gas-phase [ps] energies calculated by molecular mechanics, such as a loss of 12.0 screening between charged groups, these energies were calculated by using an algorithm based on the solvent-accessible surface area (33) in the absence of explicit solvent molecules. The continuum-solvated energies of the cross-linked DNA 10.0 structures were compared over the course of two 200-ps dynamics studies. After a 25-ps equilibration period, the a) energies became comparable, but they fluctuated broadly over C. 8.0 a 100-kcal/mol range, indicating that neither structure was lower in energy. To remove the large, randomly fluctuating contributions to these energies by the 10 non-cross-linked base 6.0 pairs, the coordinates of the link atoms, collected at regular

Table 2. dG-to-dG N7 distances averaged over the last 175 ps of the simulation 4.0 Distance, A 0.0 50.0 100.0 150.0 200.0 time [ps] Pair Linked Unlinked 1,3 6.40 0.31 9.00 0.70 FIG. 5. The dG-to-dG (N7-to-N7) distances over the time course 1,2 5.48 0.18 7.90 0.40 of a 200-ps simulation for the control structure. The dG-to-dG distances for a 1,2 linkage (Upper) and a 1,3 linkage (Lower) are as Results are mean SD. indicated above. In our analysis, the first 25 ps was ignored. Downloaded by guest on September 24, 2021 12174 Biochemistry: Dong et aL Proc. Natl. Acad. Sci. USA 92 (1995) linking can occur only if the monoadducted DNA can adopt a 1. Colvin, M. & Chabner, B. A. (1990) in Cancer Chemotherapy favorable conformation. To determine if this occurs, the Principles and Practice, eds. Chabner, B. A. & Colins, J. M. minima of the guanine N7-to-N7 distance from the "unlinked" (Lippincott, Philadelphia), pp. 276-313. pair of curves were compared to the maxima of the respective 2. Colvin, M. & Hilton, J. (1981) Cancer Treat. Rep. 65, Suppl. 3, "linked" counterparts. Fig. 5 reveals that the 1,3 N7-to-N7 89-95. 3. Sladek, N. E. (1988) Pharmacol. Ther. 37, 301-355. separation for the unlinked DNA came within 0.03 A of the 1,3 4. Brooks, P. & Lawley, P. D. (1961) Biochem. J. 80, 496-503. N7-to-N7 separation for the linked DNA, whereas the 1,2 5. Lawley, P. D. & Brooks, P. (1963) Biochem. J. 89, 127-138. N7-to-N7 separation for the unlinked DNA never came closer 6. Kohn, K. W., Spears, C. L. & Doty, P. (1966) J. Mol. Biol. 19, than 0.41 Ato the 1,2 N7-to-N7 separation for the linked DNA. 266-288. 7. Colvin, M., Brundrett, R. B., Kan, M. N., Jardine, I. & Fenselau, C. (1976) Cancer Res. 36, 1121-1126. DISCUSSION 8. Hemminki, K. & Ludlum, D. (1984) J. Natl. Cancer Inst. 73, Several DNA cross-linking agents have been demonstrated to 1021-1028. form 1,3 instead of 1,2 interstrand cross-links, such as nitrogen 9. Hausheer, F. H., Singh, U. C., Saxe, J. D. & Colvin, 0. M. (1989) mustard and hepsulfam (10-12, 35). We have shown here that AntiCancer Drug Des. 4, 281-294. PM produces DNA interstrand cross-links by alkylating gua- 10. Ojwang, J. O., Grueneberg, D. A. & Loechler, E. L. (1989) nines of the opposite strands in a 1,3 fashion. These Cancer Res. 49, 6529-6537. findings 11. Millard, J. T., Raucher, S. & Hopkins, P. B. (1990) J. Am. Chem. have shown that a 5'-dGAC (or GNC) sequence is required for Soc. 112, 2459-2460. the formation of DNA interstrand cross-links induced by these 12. Rink, S. M., Solomon, M. S., Taylor, M. J., Rajur, S. B., cross-linking agents. Elucidation of the local determinants for McLaughlin, L. W. & Hopkins, P. B. (1993) J. Am. Chem. Soc. cross-linking will allow us to study the repair mechanism of 115, 2551-2557. PM-induced DNA cross-links by nuclear extracts in vitro. This 13. Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989) Molecular finding also provides a stepping stone towards understanding Cloning: A Laboratory Manual (Cold Spring Harbor Lab. Press, how a 1,3 cross-link relates to the cytotoxicity and antitumor Plainview, NY), 2nd Ed. activity produced by PM. 14. Maxam, A. M. & Gilbert, W. (1977) Proc. Natl. Acad. Sci. USA Our results of the molecular dynamics studies must be 74, 560-564. interpreted in terms of enthalpy, entropy, and kinetics. Com- 15. Maxam, A. M. & Gilbert, W. (1980) Methods Enzymol. 65, parisons of the internal energies yield the relative enthalpy; 499-560. 16. Molecular Simulations Inc. (1994) QUANTA 4.0/CHARMm (Molec- measures of the relative flexibility are suggestive of the relative ular Simulations, Burlington, MA), QUANTA version 4.0. entropy of the reaction; and the accessibility of conformations 17. Rink, S. M. & Hopkins, P. B. (1995) Biochemistry 34, 1439-1445. favorable for the reaction to take place is a measure of the 18. Remias, M. G., Lee, C.-S. & Haworth, I. S. (1995) J. Biomol. kinetics of the cross-linking reaction. Struct. Dyn. 12, 911-936. 19. Gamscsik, M. P., Ludeman, S. M., Shulman-Roskes, E. M., The i1, linkage is more flexible, as observed in the measured McLennan, I. J., Colvin, M. E. & Colvin, 0. M. (1993) J. Am. interstrand dG-to-dG N7 distances. Although the 1,3 N7-to-N7 Chem. Soc. 36, 3636-3645. distance is on average 0.9 A larger than the 1,2 distances, it spans 20. Fritsch, V., Ravishanker, G., Beveridge, D. L. & Westhof, E. nearly the entire range of the 1,2 distance during this simulation (1993) Biopolymers 33, 1537-1552. while the 21. Singh, U. C., Weiner, S. J. & Kollman, P. A. (1985) Proc. Natl. (cf Fig. 5). Thus, 1,2 and 1,3 linkages maybe very similar Acad. Sci. USA 82, 755-759. enthalpically, the 1,3 linkage is favored entropically. 22. Brunger, A. T. (1992) X-PLOR, A System for X-ray Crystallogra- The 1,3 cross-link is also kinetically favored. The distribu- phy and NMR (Howard Hughes Medical Inst. and Dept. of tions of 1,3 dG-to-dG N7-to-N7 distances for the linked and Molecular Biophysics and Biochemistry, Yale Univ., New Haven, unlinked DNA are nearly overlapping (distributions not CT), Version 3.1. 23. Jorgensen, W., Chandrasekar, J., Madura, J., Impey, R. & Klein, shown, but four very close approaches can be seen for the 1,3 M. (1983) J. Chem. Phys. 79, 925-935. linkage in Fig. 5 Lower), whereas the corresponding distribu- 24. Miyamoto, S. & Kollman, P.A. (1992) J. Comp. Chem. 13, tions for the 1,2 case are more than 0.4 A apart. Remias et al. 952-962. (18) monitored a monoadduct nitrogen mustard in a similar 25. Beveridge, L., McConnell, K. J., Nirmala, R., Young, M. A., molecular dynamics simulation. Vijayakumar, S. & Ravishanker, G. (1994) in Structure and Much the same behavior was Reactivity in Aqueous Solutions, eds. Cramer, C. J. & Truhlar, observed in watching for the unbound active group to come D. G. (Am. Chem. Soc., Washington, DC), pp. 381-394. within a reasonable distance of the relevant guanine N7: the 26. Judson, R., McGarrah, D. B., Melius, C. F., Mori, E., Barsky, D., 1,3 linkage became accessible to nucleophilic attack by the N7 Tan, Y. T., Windemuth, A., Tresurywala, A. M., Jaeger, E. P., in an 80-ps simulation, while the 1,2 linkage did not. Note that Meza, J. C. & Plantenga, T. (1994) Tech. Rep. SAND95-8258 (Sandia National Labs., Livermore, CA). simulations with the bulkier PM moiety instead of NNM 27. Windemuth, A. & Schulten, K. (1991) Mol. Simul. 5, 353-361. should further favor the less-hindered 1,3 linkage. 28. Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States, D. J., In summary, the main computational findings are that the Swaminathan, S. & Karplus, M. (1983) J. Comp. Chem. 4, enthalpies of 1,2 and 1,3 NNM cross-linked DNA are very 187-217. but the structure is more 29. Levitt, M., Hirschberg, M., Sharon, R. & Daggert, V. (1995) similar, 1,3 considerably flexible and Comput. Phys. Commun., in press. thus is expected to be in a considerably higher entropic state. 30. Frisch, M. J., Gill, P. M. W., Johnson, B. G., Wong, M. W., The dynamics of the unlinked DNA also kinetically favors the Foresman, J. B., Robb, M. A., Head-Gordon, M., Replogle, E. S., formation of the 1,3 linkage simply because the 1,3 guanine Gomperts, R., Andres, J. L., Raghavachari, K., Binkley, J. S., N7-to-N7 distance is closer to that distance in the cross-linked Gonzalez, C., Martin, R. L., Fox, D. J., Defrees, D. J., Baker, J., Stewart, J. J. P. & Pople, J. A. (1993) GAUSSIAN 92/DFT (Gaussian, product. These findings represent at least a partial explanation Inc., Pittsburgh), Revision G.4. for the experimental observations of the production of a 1,3 31. Hirst, D. M. (1990) A Computational Approach to Chemistry, cross-link in DNA by bifunctional nitrogen mustards. (Blackwell, Oxford), pp. 394-436. 32. Mattes, W. B., Hartley, J. A. & Kohn, K. W. (1986) Biochim. The authors thank Drs. Guomin Li, Yinchin Hsieh, Woei-horng Fang, Biophys. Acta 868, 71-76. 33. Mohamadi, F., Richards, N. G. J., Guida, W. G., Liskamp, R., and Richard Judson and Mr. Thomas C. Bishop for helpful discussions Lipton, M., Caufield, C., Chang, G., Hendrickson, T. & Still, and technical assistance. This work was supported by National Institutes W. C. (1990) J. Comp. Chem. 11, 440-467. of Health Grants NS 20023 and NS 30245 and American Cancer Society 34. Lowry, T. H. & Richardson, K. S. (1987) Mechanism and Theory Grant DHP 67E to H.S.F.; Department of Energy Contract DE-AC04- in Organic Chemistry (Harper & Row, New York), 3rd Ed. 94AL85000 to M.E.C.; National Cancer Institute Grant CA 16783 to 35. Streeper, R. T., Cotter, R. J., Colvin, M. E., Hilton, J. & Colvin, O.M.C.; and National Institutes of Health Grant GM 23719 to P.M. 0. M. (1995) Cancer Res. 55, 1491-1498. Downloaded by guest on September 24, 2021