Two 1 : 1 binding modes for distamycin in the minor groove of d(GGCCAATTGG) Koen Uytterhoeven 1 , Jiri Sponer 2 and Luc Van Meervelt 1 1 Biomolecular Architecture, Department of Chemistry, Katholieke Universiteit Leuven, Belgium; 2 Institute of Biophysics, Academy of Sciences of the Czech Republic, and National Center for Biomolecular Research, Brno, Czech Republic Single-crystal X-ray structure determinations of the complex between the minor-groove binder distamycin and d(GGCCAATTGG) reveal two 1 : 1 binding modes which differ in the orientation of the drug molecule in the minor groove. The two crystals were grown from different cry- stallization conditions and found to diffract to 2.38 and 1.85 A ˚ , respectively. The structures were refined to comple- tion using SHELXL-93, resulting in a residual R factor of 20.30% for the 2.38-A ˚ resolution structure (including 46 water molecules) and 19.74% for the 1.85-A ˚ resolution structure (including 74 water molecules). In both orienta- tions, bifurcated hydrogen bonds are formed between the amide nitrogen atoms of the drug and AT base pairs. With a binding site of at least five base pairs, close contacts between the terminal distamycin atoms and guanine amino groups are inevitable. The detailed nature of several of these inter- actions was further investigated by ab initio quantum chemical methods. Keywords: distamycin; drug–DNA complex; minor groove binder; quantum chemical calculations; X-ray structure. Distamycin A (Fig. 1) is a member of a family of naturally occuring oligopeptides showing antiviral and antibiotic properties. Like other minor-groove binder drugs, distamy- cin binds noncovalently in the minor groove of DNA with a binding preference for stretches of AT-rich sequences [1], thereby preventing DNA and RNA synthesis by inhibition of the corresponding polymerase reaction. The crystal structure determination of a 1 : 1 distamycin–d(CGCAAA TTTGCG) complex (12-dista) at 2.2 A ˚ resolution shows that the drug covers five of the six AT base pairs [2]. The amide nitrogen atoms of the drug form hydrogen bonds to N3(A) and/or O2(T) atoms in the minor groove. The complex is further stabilized by van der Waals’ and electrostatic interactions. The selectivity for AT-rich sequences of minor-groove binders was first thought to have sterical reasons: the bulky NH 2 group at the floor of the minor groove of CG- containing regions can prevent binding of these drugs [3]. More recently, factors such as minor-groove width influen- cing the extent of van der Waals’ interactions [4] and electrostatic interactions between the positively charged drug and the more negatively charged minor groove in the case of AT sequences [5] were added. Solution NMR studies have also discovered side-by- side binding of two distamycin molecules in the minor groove of d(CGCAATTGCG) [6]. More structural information about this 2 : 1 binding mode was first provided by the crystal structure of d(ICICICIC)–dista- mycin [7] and later by side-by-side complexes of dista- mycin with natural targets d(ICITACIC), d(ICATATIC) and d(GTATATAC) [8,9]. Owing to the overlap of about 75%, the two staggered antiparallel distamycin molecules span almost eight base pairs and are kept together by dipole–dipole interactions between stacking pyrrole rings and amide bonds. Each drug hydrogen-bonds with the bases of only one DNA strand and stacks with the sugar rings. We have previously reported the structure determin- ation at 1.9-A ˚ resolution of the complex of the shor- ter minor-groove binder 4¢,6-diamidino-2-phenylindole (DAPI) with d(GGCCAATTGG) (10-DAPI), revealing a novel off-centered binding with a hydrogen bond between the drug and a CG base pair [10]. In an attempt to use similar crystal engineering techniques to improve the resolution of 1 : 1 distamycin–DNA complexes (currently 2.2 A ˚ for 12-dista and 2.0 A ˚ for the dista- mycin–d(CGCGAATTC + GCG) complex where C + ¼ 5-methylcytidine (NDB entry code GDLB41), we have cocrystallized distamycin with the decamer d(GGCCAA TTGG). Intensity measurements for two crystals obtained from different crystallization conditions were carried out to 2.38 and 1.85 A ˚ resolution. Whereas for one crystal the distamycin orientation and binding site is the same as in 12-dista, the orientation of the drug is inverted in the other crystal. Both orientations show interactions between thedrugandguanineNH 2 groups. For the inverted orientation, the DNA–distamycin interaction is also characterized by ab initio methods. Correspondence to L. Van Meervelt, Biomolecular Architecture, Department of Chemistry, Katholieke Universiteit Leuven, Celestijnenlaan 200F, B-3001 Leuven (Heverlee), Belgium. Fax: + 32 16 327990, Tel.: + 32 16 327609, E-mail: Luc.VanMeervelt@chem.kuleuven.ac.be Abbreviations: 12-dista, crystal structure of the d(CGCAAATTTGCG) complex (2); 10-DAPI, crystal structure of the d(GGCCAATTGG)–DAPI complex (10); MPD, 2-methyl-2, 4-pentanediol; DAPI, 4¢,6-diamidino-2-phenylindole; HF, Hartree- Fock; MP2, Moeller–Plesset perturbational theory. Note: a web page is available at http://www.chem.kuleuven.ac.be/research/bma/ (Received 20 December 2001, revised 10 April 2002, accepted 23 April 2002) Eur. J. Biochem. 269, 2868–2877 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.02952.x EXPERIMENTAL PROCEDURES Crystallization and data collection The DNA decamer d(GGCCAATTGG) was purchased from Oswel DNA service (University of Southampton, UK), distamycin from Serva Biochemica (Heidelberg, Germany). Crystals were grown at 16 °C using the sitting drop method from two different conditions containing 54.4/33.25 m M sodium cacodylate buffer (pH 6.0), 35.0/105.0 m M MgCl 2 , 70 m M NaCl, 8.8% 2-methyl-2,4-pentanediol (MPD), 10.5 m M spermine, 0.25/0.42 m M ssDNA and 0.125/ 0.21 m M distamycin against a 50/35% MPD stock solution. From a bar-shaped crystal of dimensions 0.4 · 0.1 · 0.05 mm from condition 1, intensity data were collected at 100 K on a MAR345 imaging plate detector at beamline X11inanEMBLHamburg(k ¼ 0.9116 A ˚ ) over a 105 ° u range with increments of 1.5 ° using cryocooling techniques with a crystal-to-detector distance of 350 mm. A well-diffracting crystal of dimensions 0.2 · 0.1 · 0.05 mm from condition 2 was mounted for data collection at 100 K using a similar protocol at beamline BW7b in an EMBL Hamburg (150 ° u range, crystal-to-detector dis- tance 250 mm, k ¼ 0.8423 A ˚ ). Data were processed using the DENZO /scalepack [11] suite of programs. Data collection statistics for both crystals are given in Table 1. The final resolution limit of the diffraction pattern was 2.38 A ˚ for crystal 1 and 1.85 A ˚ for crystal 2. Structure solution and refinement Unit cell parameters and space group indicated isomorph- ism with the d(GGCCAATTGG)–DAPI structure, which was used as a starting model (NDB entry code DD0002, except DAPI and solvent molecules) for further refinement on F 2 using SHELXL-93 [12]. The nucleotides of strand 1 are labeled G1–G10 in the 5¢fi3¢ direction and G11–G20 on strand 2, and the drug is labeled D. After determination of the weighting factor and positional adjustment of parts of the DNA structure, water molecules were added, but not in the minor-groove region. At this stage of the refinement, the distamycin molecule was located in the (F o ) F c ) Fourier difference map. In subsequent refinement cycles, more water molecules were gradually added. During the conjugate- gradient refinement, no torsion angle or hydrogen-bond restraints were applied. The 1,2 and 1,3 distances used as dictionary values for distamycin were based on netropsin [13], except for the formamide end which was based on fragments retrieved from the Cambridge Structural Database [14]. Fig. 1. Structure and numbering scheme of distamycin. Hydrogen atoms attached to pyrrole and alkyl groups are not shown. Table 1. Data collection and refinement statistics of the d(GGCCAATTGG)–distamycin complexes. NA, Not available. Crystal 1 Crystal 2 Data collection statistics Space group P2 1 2 1 2 1 P2 1 2 1 2 1 Unit cell (A ˚ )a¼ 26.011, b ¼ 40.861, c ¼ 53.164 a ¼ 25.289, b ¼ 36.439, c ¼ 53.047 Total no. of measured reflections 27524 43223 No. of independent reflections 2538 4223 Resolution (A ˚ ) 2.38 1.85 Multiplicity 5.5 4.3 v 2 1.035 1.142 R symm (%) 4.1 (100.0–2.38 A ˚ ) 4.9 (100.0–1.85 A ˚ ) 25.1 (2.42–2.38 A ˚ ) 19.5 (1.92–1.85 A ˚ ) Completeness (%) 92.7 (100.0–2.38 A ˚ ) 93.0 (100.0–1.85 A ˚ ) 93.4 (2.42–2.38 A ˚ ) 96.9 (1.92–1.85 A ˚ ) Mean I/r(I) 22.6 21.1 Reflections with I >3 r(I) (%) 80.3 (100.0–2.38 A ˚ ) 80.3 (100.0–1.85 A ˚ ) 46.3 (2.42–2.38 A ˚ ) 55.3 (1.92–1.85 A ˚ ) Refinement statistics Resolution range (A ˚ ) 100.0–2.38 100.0–1.85 R value/R free value (%) 20.30/NA 19.74/27.80 No. of nonsolvent atoms 445 445 No. of water molecules 46 74 Average B values of DNA (A ˚ 2 ) 66.6 28.2 Average B values of distamycin (A ˚ 2 ) 79.1 30.8 Average B values of water molecules (A ˚ 2 ) 73.0 43.0 Rmsd of bond lengths (A ˚ ) 0.018 0.019 Rmsd of bond angles (°) 3.20 2.70 Ó FEBS 2002 Two 1 : 1 binding modes for distamycin (Eur. J. Biochem. 269) 2869 For crystal 1, the R value converged to 20.30% after addition of 46 water molecules (R free was not used to avoid further reduction of the number of data per parameter at this resolution). For crystal 2, the first maps already indicated an inverted orientation (hereafter called orienta- tion B) of the drug molecule with respect to crystal 1 (orientation A). Therefore refinement for crystal 2 was monitored using R free calculated for a reference set of 10% of the reflections. Addition of 29 water molecules and distamycin led to the following R values: R ¼ 23.88%, R free ¼ 32.31% for orientation A, and R ¼ 22.79%, R free ¼ 30.75% for orientation B. Final R values for crystal 2wereR ¼ 19.74%, R free ¼ 28.01% for orientation B. An independent refinement with orientation A resulted in R ¼ 20.30%, R free ¼ 29.84%. As a consequence and in agreement with the electron density maps, orientation B was retained for crystal 2. Figure 2 shows the final (2F o ) F c ) electron-density maps in the minor-groove region for both crystals. Refinement statistics are presented in Table 1. The helical parameters in accordance with the Tsukuba Workshop guidelines [15], and torsion angles were calcula- tedwiththe3 DNA program [16]. Quantum chemical calculations Ab initio quantum chemical calculations were used to investigate intrinsic molecular interactions of a number of close intergroup contacts observed in the crystal. Special attention was given to contacts involving the guanine amino groups. Appropriate fragments of the drug and several proximal bases have been taken from the crystal structure. Intermolecular positions of the interacting species were frozen based on the crystal data using a set of constraints involving three (in some cases more) appropriate non- hydrogen atoms on each monomer. The rest of the structure including all the hydrogen positions was relaxed using gradient optimization. This procedure has been extensively used in the past to investigate local contacts seen in DNA crystal structures and allows full relaxation of the electronic structure and hydrogen positions while keeping the systems Fig. 2. Final (F o ) F c ) electron-density maps in the minor groove of the crystal structure of the d(GGCCAATTGG)–distamycin complex in which the drug has been omitted from the final refined model. (A) Crystal 1 in orientation A; (B) crystal 2 in orientation B. The refined distamycin position is superimposed on the difference density for reference, contouring at 1 r (yellow) and 2 r level (green). The figure was prepared with BOBSCRIPT [34] and RASTER 3 D [35]. 2870 K. Uytterhoeven et al.(Eur. J. Biochem. 269) Ó FEBS 2002 studied in the experimental geometry [10,17–20]. The optimizations were carried out within the Hartree–Fock (HF) approximation with the standard polarized 6–31G* basis set. Although this level of calculations underestimates the flexibility of amino groups, it nevertheless is sufficient to reveal when the amino group is activated by molecular interactions towards a partial sp 3 hybridization [21–23]. Interaction energy calculations between drug fragments and nucleobases were carried out assuming the quantum chemical-optimized geometries with inclusion of electron correlation effects using the second-order Moeller-Plesset perturbational theory (MP2) with the 6–31G basis set augmented by diffuse d-polarization functions to all sec- ond-row elements [exponents of 0.25, designated as 6–31G*(0.25)] to properly account for the dispersion attraction [22,23]. The calculations were corrected for the basis set superimposition error using the standard counter- poise procedure. The quantum chemical procedure used in this study allows reliable semiquantitative characterization of the nature and intrinsic Ôin vacuoÕ strength of the observed intermolecular contacts. All quantum chemical calculations were carried out using the GAUSSIAN 94 program suite [24]. RESULTS AND DISCUSSION Both complexes (Fig. 3) have a conformation that closely resembles the native decamer structure [25] and its DAPI complex [10]: a central octamer d(CCAATTGG) consisting of normal Watson–Crick base pairs is at both ends flanked by two overhanging guanines forming triplets in the crystal packing. All three drug–decamer complexes differ from the native 1.15-A ˚ resolution structure in the phosphate confor- Fig. 3. Stereoscopic representation of the distamycin–d(GGCCAATTGG) complexes. (A) Crystal 1 in orientation A; (B) crystal 2 in orientation B. The DNA is drawn with open bonds, and the distamycin with solid bonds. Nitrogen, oxygen and phosphor atoms are grey, and carbon atoms are white. The figure was prepared using B OBSCRIPT [34]. Ó FEBS 2002 Two 1 : 1 binding modes for distamycin (Eur. J. Biochem. 269) 2871 mation of residue C13, which correlates with the difference in resolution as described previously [10]. As expected for B-DNA, the sugar puckering modes for the central octamer duplex are situated in the normal C3¢-exo to O4¢-endo range, the southern part (C2¢-endo) of the pseudorotation cycle. However, for the overhanging guanines, C3 and C13, the puckering modes are closer to those for A-DNA, which is a consequence of triplet formation. DNA–distamycin interactions Distamycin binds in the minor groove of the central octamer. The expected binding site of at least five base pairs makes interactions with GC base pairs inevitable. Both structures not only differ in resolution, but also in the orientation of the distamycin molecule (Fig. 4). In the lower resolution structure of crystal 1, the orientation (orienta- tion A) is similar to that described for 12-dista [2]. The higher-resolution structure of crystal 2 shows an inverted orientation of distamycin (orientation B). Position of distamycin in the lower-resolution structure (orientation A, Fig. 4A) Distamycin binds to five base pairs covering the sequence d(AATTG). The positively charged amidinium end group is orientated toward the A5.T18 base pair with atom N9(D) lying deep in the minor groove interacting with N3(A5) (2.70 A ˚ ) and the partial negative O4¢ atoms O4¢(A5) (3.29 A ˚ )andO4¢(A6) (2.80 A ˚ ). A close contact with the amino group of G19 is avoided (N9(D)…N2(G19) 3.86 A ˚ ). The nearest neighbor of N8(D) is O4¢(3.41 A ˚ ). Bifurcated hydrogen bonds are formed between the amide nitrogen atoms simultaneously to both N3(A6) and O2(T18) for N7(D), O2(T7) and O2(T17) for N5(D), and N3(A15) and N2(G9) for N1(D). N3(D) only has one close contact to N3(A16) and is 3.59 A ˚ away from O2(T8). At the other end of distamycin, the formamide O1(D) is oriented away from the groove. The nearest neighbor of O1(D) is a sugar O4¢(A15) atom at 3.50 A ˚ . The three pyrrole rings are rotated to each other to follow closely the curvature of the groove. Rings A and B make an angle of 22.9 °, rings B and C 28.4 ° and ring C makes an angle of 35.9 ° with the terminal amidinium group. The drug–DNA complex is further stabilized by van der Waals’ interactions mainly between carbon atoms of the three pyrrole rings and the sugar–phosphate backbone. No water molecules are in direct contact with the drug molecule. Position of distamycin in the higher-resolution structure (orientation B, Fig. 4B) It was clear from the density maps of the 1.85-A ˚ resolution structure that the drug has to be rotated over 180 ° with respect to orientation A. As a consequence, the amidinium group is now orientated toward the G9–C12 base pair and distamycin now interacts with six base pairs. The N9(D) atom of the amidinium group is in close contact with amino group N2 of guanine G9 (3.16 A ˚ )and hydrogen-bonded with N3(A15) (2.85 A ˚ ). This N9(D) atom is also in contact with O2(C14) by two intermediate water molecules. The O1(D) atom of the formamide group now points toward the C4–G19 base pair with a very close contact with N2(G19)ofonly2.57A ˚ . Atom O1(D) is further in contact with O2(C4) (3.39 A ˚ ), O4¢(A5) (2.94 A ˚ )andtwowater molecules (2.77 and 2.95 A ˚ ). In the same formamide group, atom N1(D) interacts with N3(A5) (3.12 A ˚ )andO4¢(A6) (3.03 A ˚ ). Fig. 4. Schematic view of the distamycin interactions in the minor groove of d(GGCCAATTGG)–distamycin structures. (A) Crystal 1; (B) crystal 2. 2872 K. Uytterhoeven et al.(Eur. J. Biochem. 269) Ó FEBS 2002 The nitrogen atoms of the amide linkers between the pyrrole rings form bifurcated hydrogen bonds with the thymine or adenine bases. The central N5(D) atom of the drug is hydrogen-bonded to the two O2 atoms of the thymine bases of the central AT steps. Nitrogen atoms N3(D) and N7(D) bridge N3(adenine) and O2(thymine) atoms of opposite strands. Furthermore, several stabilizing van der Waals’ contacts between the O4¢ atoms of the DNA sugars and distamycin atoms are observed. Rings A and B make an angle of 8.6 °,ringsBandC 21.1 °, and ring C makes an angle of 15.7 ° with the terminal amidinium group. Quantum chemical analysis of the base–distamycin contacts For both orientations, the crystal structure reveals a number of close intermolecular contacts including interactions between the guanine amino groups and the terminal distamycin atoms. Such amino-group interactions are often assumed to be repulsive steric clashes. To clarify the nature of the selected interactions, we used quantum chemical calculations carried out at the MP2/6–31*(0.25)//HF/ 6–31G* approximation (see Experimental procedures). As the interactions were most surprising for the 1.85-A ˚ resolution structure, the calculations were only performed for orientation B. Further, such calculations are usually only performed for systems of resolution 2 A ˚ and better, as outcomes of such calculations may be spoiled by inaccurate positioning of the interacting groups. The contact area of the distamycin amidinium end was investigated by evaluating the hydrogen positions and interaction energy terms for the interactions between the terminal drug segment and the bases of base pair A15–T8 and base G9 (Table 2, Fig. 5). The drug binding to A15 is very strong, with substantial electrostatic and dispersion components indicating a strong hydrogen bond between N9(D) and N3(A15). The intrinsic strength of this interac- tion is halfway between CG Watson–Crick and AT Watson–Crick base pairs [22,23]. Furthermore the drug also recognizes T8 by a long-range electrostatic interaction between N7(D) and O2(T8). The long-range nature of this contact can be deduced considering both the geometry and the negligible correlation component to the interaction energy. The interaction with G9 is an energetically neutral van der Waals’ contact with an attractive correlation (dispersion) component and thus a repulsive HF compo- nent. This means no active recognition. The guanine amino group is certainly influenced by this contact. It has a substantial pyramidal character: both hydrogen atoms deviate 0.39 and 0.49 A ˚ from the best plane through the base atoms. The amino-group nonplanarity is modestly enhanced over isolated guanine optimized by the same method (hydrogen deviations 0.10 and 0.41 A ˚ , respectively). Note also that, because of increased propeller twisting of the GC base pair, the cytosine base (its O2 atom) not involved in the calculation helps to lead the guanine amino-group hydrogen atoms away from the drug. A search in the Cambridge Structural Database [14] shows that most amino groups in guanine fragments are almost planar. Note, however, that in vacuo the amino group is substantially nonplanar [21–23], and this electronic structure feature may reoccur in DNA when intermolecular interactions profit from the nonplanarity [10,17–19,21–23]. This has been observed previously for the DAPI–DNA complex [10] and is seen to a lesser extent here also. The interaction between the neutral formamide part of the drug at the other end and the proximal base pairs A5–T18 and C4–G19 is less favourable than the amidi- nium–DNA interactions because of the neutral character of this end (Table 2, Fig. 5). The composition of the interac- tion energy shows that the drug interacts with the AT pair more in a van der Waals’ than in a hydrogen-bonding manner. (Hydrogen-bonded systems are always dominated by a strong HF component, which could be obtained as the difference between the total interaction energy and the correlation component in Table 2.) The contact with the CG base pair is weakly repulsive, but not enough to prevent binding. The contact with C4 is indeed not favorable because of the proximity of the two oxygen atoms O1(D) and O2(C4) (3.39 A ˚ ). Despite the very short distance between O1(D) and N2(G19), the energy components show that this contact is not a strong hydrogen bond at all. This is also clear from the amino-group hydrogen positions, pointing away from O1(D). To obtain an upper limit of the possible attraction that can be achieved by this part of the drug and a CG base pair, we alleviated all constraints during the optimization. A completely unconstrained gas phase optimization of the formamide part and base pair C4–G19 illustrates that not much of the hydrogen-bonding potential is used. In the most optimal binding, the interaction energy between G19 and the formamide fragment would be )35.1 kJÆmol )1 (E cor ¼ )4.2 kJÆmol )1 ). However, this optimal geometry is not possible because of the interactions of the rest of the distamycin with the oligonucleotide. We also carried out some additional optimizations with differ- ent constraints (not shown). These confirmed that in the refined crystal geometry the drug is rather too tightly packed against the G19 base. The interaction can be improved by locally relaxing the drug geometry, but no hydrogen bond can be obtained close to the experimental geometry. It should also be noted that the CG–formamide end interaction might be supported by two crystallographic water molecules (O116 and O126). Although several Table 2. Total interaction energy and correlation component for the interaction between terminal distamycin fragments and bases in close contact with distamycin evaluated at the MP2/6–31G*(0.25)//HF/ 6–31G* approximation and utilizing constraints taken from the crystal. The electron correlation component of the interaction energy includes the dispersion attraction and corrections to electrostatic and polar- ization terms. Base Interaction energy (kJÆmol )1 ) Correlation component (kJÆmol )1 ) Amidinium end A15 )81.1 )23.4 T8 )48.5 )1.3 G9 2.5 )19.6 Formamide end A5 )19.6 )18.8 T18 )12.5 )1.3 C4 10.9 )5.4 G19 0.0 )14.6 Ó FEBS 2002 Two 1 : 1 binding modes for distamycin (Eur. J. Biochem. 269) 2873 optimization attempts (not shown) did not result in a converged structure, the intermediate structures suggest that the water molecules are actively involved. Influence of the binding of distamycin on the DNA conformation One of the most important effects of drug binding on the DNA conformation is the widening of the minor groove as illustrated in Fig. 6. The width of the groove is measured by taking the shortest H5¢–H4¢ distances between the two DNA chains. In the native structure, the minor-groove width is symmetric. Both distamycin and DAPI have a similar asymmetric opening effect: the widening is more pronounced at the 3¢ end of the first strand. Where DAPI opens the groove over a distance of three to four base pairs, distamycin has an effect on at least five base pairs. The opening of the groove by distamycin (by  4A ˚ )ismore pronounced than with DAPI (1.5 A ˚ ). In the 2 : 1 side-by- side complexes, the minor groove expands to  7.7 A ˚ in order to accommodate two distamycin molecules. Distamycin makes more close contacts with the atoms in the minor groove in orientation A than in orientation B (Table 3). For both orientations, the number of contacts with both strands is almost equal, which was less empha- sized for 12-dista. The complexation of distamycin has no major effect on interbase parameters (buckle, propellor twist, and opening) Fig. 6. Comparison of the minor-groove width based on H4¢–H5¢ dis- tances for the native decamer (blue) and its complex with distamycin (black for the 2.38 A ˚ and red for the 1.85 A ˚ resolution structure) and DAPI (green). Fig. 5. Optimized geometry based on HF/6–31G* calculations of the interaction between (A) the amidinium end and bases A15, T8 and G9, and (B) the formamide end and bases C4 and G19. Intermolecular geometry frozen according to the crystal data. Drug fragment atoms are yellow. 2874 K. Uytterhoeven et al.(Eur. J. Biochem. 269) Ó FEBS 2002 and cartesian neighboring base parameters (tilt, roll, twist, shift slide, and rise). However, some small adaptions such as the increased propeller twist of G14–G9 in orientation B are observed, and are necessary to optimize the complexation. Also the base stacking patterns are very similar to those observed for the native decamer. CONCLUSIONS Both current crystal structures describe the interaction of distamycin in the minor groove of the central CAATTG sequence in a 1 : 1 binding mode. The present tight crystal packing due to the triplet formation is not compatible with a 2 : 1 drug binding mode, which requires a much broader minor groove. The opposite drug orientations in the minor groove are despite the pseudo-twofold symmetry of the palindromic sequence distinguishable because of the different triplet formation at both ends of the oligonucleotide. The length of the minor-groove binder makes contacts with the G-NH 2 group at both ends of the drug inevitable. For the novel orientation B, analyis of the absolute interaction energies obtained by quantum chemical methods shows that the presence of both G-NH 2 does not destabilize the distamycin binding to an extent that it prevents complexa- tion. The amidinium end of the drug does not recognize G9 actively, but this region optimizes its conformation with respect to the available space. The amidinium fragment ÔsitsÕ on the G base; as a consequence, the G-NH 2 group becomes pyramidal and the propeller twist of base pair G9–C14 increases by 5 ° compared with the native decamer, helping the pyramidalization. Thus the DNA structure adapts to host the drug molecule, including a modest amino-group pyramidalization. The contacts of the atoms O1(D) and N1(D) at the other side of the drug are more complicated. The interaction energy with A5 of )19.6 kJÆmol )1 is close to those of a water dimer. Combined with the N1(D)…N3(A5) distance of 3.12 A ˚ , this could possibly lead to the conclusion that a good hydrogen bond is formed. However, the composition of the interaction energy (the dominating dispersion component) and the N1(D)-H…N3(A5) angle (137 °)do not support this view. The short contact between atoms O1(D) and N2(G19) (2.57 A ˚ ) again cannot be classified as a strong hydrogen bond. Optimal hydrogen bonding in this region as located by an unconstrained gas phase optimi- zation is not possible here because of the other DNA– distamycin interactions. It is helpful to check the electron density again in this region: both O1(D) and C1(D) are not in reasonable (2F o – F c ) electron density, whereas the rest of the drug molecule fits these maps nicely (Fig. 2B). Also the temperature factors of these two atoms are much higher than those of the other distamycin atoms (Fig. 7). Most probably, this end of the drug has more than one conformation, the average of which is observed. This illustrates the use of quantum chemical calculations in further analysis of crystallographic results, a combination Table 3. Close contacts of atoms in the minor groove of d(GGCCAATTGG) or d(CGCAAATTTGCG) and distamycin atoms. Distances less than 3.6 A ˚ are considered close contacts. d(GGCCAATTGG) + distamycin (orientation A) C4 G19 C4¢,O4¢,C5¢ 3 A5 C1¢,O4¢, N3, C4 T18 C4¢,C5¢, O2, O4 8 A6 C1¢,C4¢,O4¢,C5¢, N3 T17 C1¢,C4¢,O4¢,O2 9 T7O3¢,C4¢,O4¢,C5¢, O2 A16 C4¢,O4¢,N3 8 T8O3¢,C4¢,C5¢, O2 A15 C1¢,O4¢, C2, N3, C4 9 G9 C4¢,O4¢,C5¢, O1P, N2, N3 C14 6 d(GGCCAATTGG) + distamycin (orientation B) C4 O2 G19 C4¢,O4¢,C5¢,N2 5 A5 C1¢,O4¢, N3 T18 C4¢,O4¢,C5¢,O2 7 A6 C4¢,O4¢, N3 T17 C4¢,O4¢,O2 6 T7C4¢,O4¢,C5¢, O2 A16 C1¢,C4¢,O4¢, C2, N3 9 T8C4¢,O4¢,C5¢, O2 A15 C2, N3 6 G9 O4¢, N2, N3 C14 3 d(CGCAAATTTGCG) + distamycin (12-dista) A4 C2, N3 T21 C4¢,O4¢,C5¢,O2 6 A5 C2, N3 T20 C4¢,O4¢,C5¢,O2 6 A6 C1¢,C4¢,O4¢,C5¢, C2, N3, C4 T19 7 T7C4¢ A18 N3 2 T8 O2 A17 1 T9C4¢,O4¢,C5¢ A16 3 Fig. 7. Temperature factors for distamycin in orientation B (crystal 2). Red are high (B  55 A ˚ 2 ) and white are low (B  20A ˚ 2 ) temperature factors. The figure was prepared using BOBSCRIPT [34]. Ó FEBS 2002 Two 1 : 1 binding modes for distamycin (Eur. J. Biochem. 269) 2875 that is so far unique in the field of biological macromol- ecules. We plan to investigate this binding mode by an extensive explicit-solvent molecular dynamics simulation in the near future. Competition experiments demonstrate that distamycin is capable of replacing netropsin in its 1 : 1 and 2 : 1 complexes with DNA [26]. Compared with netropsin, both distamycin orientations indeed bind better in the minor groove [27,28]. Different 1 : 1 binding modes and orientations have been reported for several minor-groove binders such as netropsin [27,28] and Hoechst 33258 [29–32]. In the case of two orientations fitting the electron density equally well, one can conclude that both orientations in the minor groove are energetically very similar, or that the resolution of the crystal structure determination is not high enough. We have shown that, for DAPI and distamycin, crystal engineering techniques may overcome the problem of interpreting electron-density maps. In our case in which both orienta- tions occur in different crystals, one can also conclude that the two distamycin orientations in the d(GGCCAATTGG) minor groove are energetically equivalent. However, as the 1 : 1 association of distamycin to AT-rich sequences is extremely fast [33], why is there no disorder in our structures? Here the charged character of dystamycin may play an important role. The three orthogonal twofold screw axes present in the space group P2 1 2 1 2 1 in general prevent similar parts of a molecule being in each others neighbor- hood. Adaption of this packing scheme by the drug without orientational disorder means that the distance between the positive distamycin ends is always maximal, a situation that is electrostatically more favourable. ACKNOWLEDGEMENTS This work was partly supported by the European Community–Access to Research Infrastructure Action of the Improving Human Potential Programme to the EMBL Hamburg Outstation, contract number: HPRI-CT-1999-00017, and by the Fund for Scientific Research (Flanders). We thank the staff of the EMBL Hamburg Outstation for assistance. J. S. was supported by grant LN00A016 (National Center for Biomolecular Research), MSMT CR. All quantum chemical calculations were carried out at the Supercomputer Center, Brno. REFERENCES 1. Zimmer, C. & Wahnert, U. (1986) Nonintercalating DNA-bind- ing ligands: specificity of the interaction and their use as tools in biophysical, biochemical and biological investigations of the genetic material. Prog. Biophys. Mol. Biol. 47, 31–112. 2. Coll, M., Frederick, C.A., Wang, A.H J. & Rich, A. (1987) A bifurcated hydrogen-bonded conformation in the d(AT) base pair of the DNA dodecamer d(CGCAAATTTGCG) and its complex with distamycin. Proc. Natl Acad. Sci. USA 84, 8385–8389. 3. Wartell, R.M., Larson, J.E. & Wells, R.D. (1974) Netropsin- specific probe for A-T regions of duplex deoxyribonucleic acid. J. Biol. Chem. 249, 6719–6731. 4. Neidle, S. (1992) Minor-groove width and accessibility in B-DNA drug and protein complexes. FEBS Lett. 298, 97–99. 5. Pullman, B. (1983) Electrostatics of polymorphic DNA. J. Biomol. Struct. Dyn. 1, 773–794. 6.Pelton,J.G.&Wemmer,D.E.(1989)Bindingmodesof dis- tamycin A with d(CGCAAATTTGCG) 2 determined by two- dimensional NMR. J. Am. Chem. Soc. 112, 1393–1399. 7. Chen, X., Ramakrishnam, B., Rao, S.T. & Sundaralingam, M. (1994) Binding of two distamycin A molecules in the minor groove of an alternating B-DNA duplex. Nat. Struct. Biol. 1, 169–175. 8. Chen, X., Ramakrishnan, B. & Sundaralingam, M. (1997) Crystal structures of the side-by-side binding of distamycin to AT-con- taining DNA octamers d(ICITACIC) and d(ICATATIC). J. Mol. Biol. 267, 1157–1170. 9. Mitra, S.N., Wahl, M.C. & Sundaralingam, M. (1999) Structure of the side-by-side binding of distamycin to d(GTATATAC) 2 . Acta Crystallogr. D55, 602–609. 10. Vlieghe, D., Sponer, J. & Van Meervelt, L. (1999) Crystal structure of d(GGCCAATTGG) complexed with DAPI reveals novel binding mode. Biochemistry 38, 16443–16451. 11. Otwinowski, Z. & Minor, W. (1997) Processing of X-ray diffrac- tion data collected in oscillation mode. Methods Enzymol. 276, 307–326. 12. Sheldrick, G.M. & Schneider, T.R. (1997) SHELXL: high- resolution refinement. Methods Enzymol. 277, 319–343. 13.Berman,H.,Neidle,S.,Zimmer,C.&Thrum,H.(1979) Netropsin, a DNA-binding oligopeptide structural and binding studies. Biochim. Biophys. Acta 561, 124–131. 14. Allen, F.H. & Kennard, O. (1993) 3D search and research using the Cambridge Structural Database. Chem. Design Automation News 8, 1 & 31–37. 15. Olson, W.K., Bansal, M., Burley, S.K., Dickerson, R.E., Gerstein, M., Harvey, S.C., Heinemann, U., Lu, X J., Neidle, S., Shakked, Z., Sklenar, H., Suzuki, M., Tung, C S., Westhof, E., Wolberger, C. & Berman, H.M. (2001) A standard reference frame for the description of nucleic acid base-pair geometry. J. Mol. Biol. 313, 229–237. 16. Lu, X J., Shakked, Z. & Olson, W.K. (2000) A-DNA con- formational motifs in ligand-bound double helices. J. Mol. Biol. 300, 819–840. 17. Sponer, J. & Hobza, P. (1994) Bifurcated hydrogen bonds in DNA crystal structures. An ab initio quantum chemical study. J. Am. Chem. Soc. 116, 709–714. 18. Sponer, J., Florian, J., Leszczynski, J. & Hobza, P. (1996) Non- planar DNA base pairs. J. Biomol. Struct. Dyn. 13, 827–833. 19. Luisi, B., Orozco, M., Sponer, J., Luque, F.J. & Shakked, Z. (1998) On the potential role of the amino nitrogen atom as a hydrogen acceptor in macromolecules. J. Mol. Biol. 279, 1123– 1136. 20. Sponer, J., Florian, J., Ng, H K., Sponer, J.E. & Spackova, N. (2000) Local conformational variations observed in B-DNA crystals do not improve base stacking. Computational analysis of base stacking in d(CATGGGCCCATG) 2 B«Aintermediate crystal structure. Nucleic Acids Res. 24, 4893–4902. 21. Sponer, J. & Hobza, P. (1994) Nonplanar geometries of DNA bases. Second order Moller–Plesset study. J. Phys. Chem. 98, 3161–3164. 22. Hobza, P. & Sponer, J. (1999) Structure, energetics, and dynamics of the nucleic acid base pairs: nonempirical ab initio calculations. Chem. Rev. 99, 3247–3276. 23. Sponer, J., Leszczynski, J. & Hobza, P. (2001) Electronic prop- erties, hydrogen bonding, stacking and cation-binding of DNA and RNA bases. Biopolymers 61, 3–21. 24. Frisch, M.J., Trucks, G.W., Schlegel, H.B., Gill, P.M.W., Johnson, B.G., Robb, M.A., Cheeseman, J.R., Keith, T., Petersson, G.A., Montgomery, J.A. et al. (1995) GAUSSIAN. Gaussian, Inc, Pittsburgh, PA. 25. Vlieghe, D., Van Meervelt, L., Dautant, A., Gallois, B., Precigoux, G. & Kennard, O. (1996) Parallel and antiparallel (G.GC) 2 triple helix fragments in a crystal structure. Science 273, 1702–1705. 26. Lah, J. & Vesnaver, G. (2000) Binding of distamycin A and netropsin to the 12mer DNA duplexes containing mixed AT.GC 2876 K. Uytterhoeven et al.(Eur. J. Biochem. 269) Ó FEBS 2002 sequences with at most five or three successive AT base pairs. Biochemistry 39, 9317–9326. 27. Coll, M., Aymami, J., Van der Marel, G.A., van Boom, J.H., Rich, A. & Wang, A.H J. (1989) Molecular structure of the netropsin–d(CGCGATATCGCG) complex: DNA conformation in an alternating AT segment. Biochemistry 28, 310–320. 28. Goodsell, D.S., Kopka, M.L. & Dickerson, R.E. (1995) Refine- ment of netropsin bound to DNA: bias and feedback in electron density map interpretation. Biochemistry 34, 4983–4993. 29. Quintana, J.R., Lipanov, A.A. & Dickerson, R.E. (1991) Low- temperature crystallographic analysis of the binding of Hoechst 33258 to the double-helical DNA dodecamer C-G-C-G-A-A-T-T- C-G-C-G. Biochemistry 30, 10294–10306. 30. Pjura, P., Grzeskowiak, K. & Dickerson, R.E. (1987) Binding of Hoechst 33258 to the minor groove of B-DNA. J. Mol. Biol. 197, 257–271. 31. Teng, M.K., Frederick, C.A., Usman, N. & Wang, A.H J. (1988) The molecular structure of the complex of Hoechst 33258 and the DNA dodecamer d(CGCGAATTCGCG). Nucleic Acids Res. 16, 2671–2690. 32. Carrondo, M., Coll, M., Aymami, J., Wang, A.H J., Van der Marel, G.A., van Boom, J.H. & Rich, A. (1989) Binding of Hoechst dye to d(CGCGATATCGCG) and its influence on the conformation of the DNA fragment. Biochemistry 28, 7849–7859. 33. Baliga, R. & Crothers, D.M. (2000) On the kinetics of distamycin binding to its target sites on duplex DNA. Proc. Natl Acad. Sci. USA 97, 7814–7818. 34. Esnouf, R.M. (1999) Further additions to Molscript version 1.4 including reading and contouring electron-density maps. Acta Cryst. D55, 938–940. 35. Merritt, E.A. & Murphy, M.E.P. (1994) RASTER3D version 2. Acta Crystallogr. D50, 869–873. Ó FEBS 2002 Two 1 : 1 binding modes for distamycin (Eur. J. Biochem. 269) 2877 . Two 1 : 1 binding modes for distamycin in the minor groove of d(GGCCAATTGG) Koen Uytterhoeven 1 , Jiri Sponer 2 and Luc Van Meervelt 1 1 Biomolecular. Republic Single-crystal X-ray structure determinations of the complex between the minor- groove binder distamycin and d(GGCCAATTGG) reveal two 1 : 1 binding modes