| Nucleic Acids Research | Pages |
DNA minor groove recognition by a tetrahydropyrimidinium analogue of Hoechst 33258: NMR and molecular dynamics studies of the complex with d(GGTAATTACC)2
Introduction
Materials And Methods
DNA synthesis
Sample preparation
NMR analysis
Structure calculations
Results And Discussion
Complex stability
Structural analysis by NMR: DNA resonance assignments
Assignment of drug resonances in the complex
Intermolecular NOEs
Identification of a single A-tract binding site
Conformation of the bound ligand
Comparison with other solution structures
Acknowledgements
References
DNA minor groove recognition by a tetrahydropyrimidinium analogue of Hoechst 33258: NMR and molecular dynamics studies of the complex with d(GGTAATTACC)2
ABSTRACT
INTRODUCTION
Hoechst 33258 (H33258) is the best known of the bis-benzimidazole family of minor groove binders. Since the early 1970s, H33258 has been used widely as a fluorescent DNA stain (1-3). Although a useful biochemical tool, H33258 has limited biological activity, finding use as an anthelmintic (4,5) and possessing some activity against L1210 and P388 leukaemias in mice (6). Several analogues have shown potent activity against a number of microorganisms including those that lead to AIDS-related opportunistic infections (7-9). This family of drugs therefore provides a starting point for rational drug design as well as a good model system with which to investigate the molecular basis for DNA sequence recognition and binding (10). Hoechst 33258 has been shown by footprinting studies, fluorescence and [125I]DNA cleavage patterns to have a marked affinity for AT rich sequences (11-13). The presence of the guanine NH2 sterically impedes the drug from binding across GC pairs, although footprinting studies indicate that a GC base pair is generally present at the end of the binding site.
DNA complexes of H33258 and various analogues have been widely studied by both X-ray crystallography (14-23) and NMR spectroscopy (24-28), but these studies have raised questions concerning the degree of GC tolerance of these compounds which, in all cases, involves the bulky piperazine ring, with the planar benzimidazole and phenyl rings staying firmly in the narrower A-T tract. The binding sites reported to date are presented in Table 1, and show that even when complexed with the same oligonucleotide some structural heterogeneity is observed, in part reflecting the conditions under which the crystals were grown or the temperature at which the data were collected. Further confusion arises from authors' definitions of binding sites. For example, the final two entries in Table 1, with the drug bound to the same AAATTT tract, are described alternatively as AATTT and ATTTG sites even though the intermolecular hydrogen bonding pattern involves the same DNA bases. Those bases involved in hydrogen bonding to the benzimidazole NHs of the drug are underlined in Table 1, providing a common point of reference. The solution NMR structural models reported to date (24-27) have only been studied at low resolution, permitting rather limited comparison with the X-ray data. However, one possible conclusion to be drawn from this body of crystallographic data is that any structural heterogeneity that is apparent, may arise from the interpretation of a more dynamic model of binding in terms of a number of different snapshots. The possibility of such a dynamic model has not yet been elucidated by NMR, although one study has already alluded to a degree of mobility within the groove (27).
Table 1.
| Oligonucleotide | Binding Site | Method of study | Reference |
| CGCGAATTCGCG | AATT | X-ray | 18 |
| CGCGAATTCGCG | AATTa | X-ray | 15 |
| CGCGAATTCGCG | AATT | NMR | 26 |
| CGCGAATTCGCG | ATTC | X-ray | 14 |
| CGCGAATTCGCG | AATTa | X-ray | 16 |
| CGCGATATCGCG | TATC | X-ray | 17 |
| GTGGAATTCCAC | AATT | NMR | 27,39 |
| GGTAATTACC | AATT | NMR | 24 |
| CTTTTGCAAAAG | TTTTb | NMR | 25 |
| CGCAAATTTGCG | AATTT | X-ray | 19 |
| CGCAAATTTGCG | ATTTG | X-ray | 20 |
While a number of studies have turned attention to structural modifications aimed at enhancing the degree of G-C recognition, notably through a meta-hydroxy phenyl analogue (21,28), we have chosen to investigate the relative importance of minor groove complementarity and specific hydrogen bond mediated recognition in a number of analogues in which the N-methylpiperazine ring has been replaced by less bulky substituents with the capacity for directed hydrogen bonding to the floor of the groove. Here we describe NMR studies of the complex of Hoechst 43254 (H43254) in which the N-methylpiperazine ring of H33258 has been replaced by a 2,3,4,5-tetrahydropyrimidin-1-ium group (Fig. 1), which we have shown leads to a greater thermal stability of the complex with d(GGTAATTACC)2 (29). At physiological pH the pyrimidinium ring also carries a positive charge but has the potential for A-T specific minor groove-directed hydrogen bonding with adenine N3/thymine O2 through an NH. The incompatibility of this hydrogen bond donor with the 2-amino group of guanine, which protrudes into the groove, suggests a much lower degree of GC tolerance. The question that arises is whether the difference in stability between the two complexes is due to hydrogen bond formation (better electrostatic interactions) or due to the enhanced groove complementarity of the less bulky half chair conformation of the pyrimidinium ring (30). The structure of a complex of a 4,5-dihydro-3H-imidazol-1-ium homologue bound to d(CGCGAATTCGCG)2 has recently been determined by X-ray crystallography (10,23) and has shown that the more planar imidazolium ring can be accommodated within the narrow A-T minor groove but does not possess the correct geometry to hydrogen bond to the groove floor. Despite this, the melting temperature of the complex is raised above that of the complex with H33258 suggesting that better complementarity with the A-T minor groove rather than hydrogen bonding imparts the extra stability.
On the basis of a set of 25 drug-DNA NOEs, that define with some precision the binding site of H43254 in the minor groove of d(GGTAATTACC)2, we have employed a restrained molecular dynamics protocol using three different starting structures (each with a canonical B-DNA structure but with the drug located at different sites along the groove) to determine the preferred binding site and gain insight into the possible dynamic nature of the interaction. The accurate prediction of DNA structure requires careful simulations with a high quality forcefield. We have used the AMBER 4.1 suite of programs (31), treated solvation explicitly under periodic boundary conditions, and used the particle mesh Ewald (PME) method to treat the important long range electrostatic interactions (31-33). We arrive at three closely similar structures which we consider to represent a single converged binding site in the centre of the AATT tract. The structural analysis and the quality of the fit to the experimental NMR restraints suggests that the complex is well defined by a single binding site model.
Figure 1. Chemical structure of H43254 showing atom number, short hand notation for subunits and torsion angles ([alpha]1, [alpha]2 and [alpha]3) between subunits. The sequence of the oligonucleotide used in this study is also shown, with nucleotides on one strand distinguished from those of the complementary strand by an asterisk.
MATERIALS AND METHODS
DNA synthesis
The DNA decamer d(GGTAATTACC) was synthesised using standard solid-phase phosphoramidite chemistry, and purified trityl-on by reverse-phase HPLC using TEAA buffer, pH 7.0 and an acetonitrile gradient. The trityl group was cleaved by treatment with 50% aqueous acetic acid for 1 h at 35°C. Acetic acid was extracted into ether and the oligonucleotide finally dialysed to remove residual TEAA and acetic acid and to introduce Na+ as the counter ion. The decamer was shown to be >95% pure by 1H NMR spectroscopy.
Sample preparation
The sample of H43254 was kindly provided by Hoechst, Frankfurt, Germany and was used without further purification after checking by 1H NMR. The drug was insoluble in water at pH 7.0, so the drug-DNA complex was formed by adding a small excess of the solid material to the oligonucleotide in 100 mM NaCl, 10 mM NaD2PO4 and the mixture stirred overnight at 4°C. The sample was centrifuged to remove uncomplexed drug, leaving the 1:1 complex in solution at a concentration of 2 mM, as determined for the oligonucleotide by UV absorption measurements at 260 nm. The sodium salt of trimethylsilylpropionate was added as an internal reference compound, 0.1% sodium azide as an anti-bacterial agent and 0.1% EDTA to complex any heavy metal ions.
NMR analysis
NMR data were collected at 500 MHz on a Bruker DRX500 spectrometer and processed on an R4600PC Silicon Graphics Indy workstation using XWINNMR software. Standard phase-sensitive 2D NMR pulse sequences were used throughout, including NOESY, DQF-COSY, TOCSY and jump-and-return NOESY for solvent suppression in 90% H2O solutions. NOESY spectra were acquired at mixing times between 50 and 300 ms and typically 1024 complex data points were collected for each of 512 t1 increments with 64 transients for each. A HMQC experiment was acquired with 160 transients for each of 512 t1 increments. Spectra were zero filled to 2k × 1k prior to Fourier transformation.
Drug-DNA interproton distances were derived by integration of NOEs at 100 and 200 ms mixing times and calibrated to a number of fixed reference distances: cytosine H5/H6, Bz1 H6/H7 and Bz2 H6/H7. The values were extrapolated back to zero mixing time according to the method of Baleja et al. (34). An error bound of 15% was added to distances below 3.5 Å and 25% added to longer distances. Distances calculated from the H2O spectrum at 100 ms are intrinsically less accurate due to the use of only one mixing time and the sinusoidal wave of intensity caused by jump-and-return water suppression and so error bounds of 35% were added.
Structure calculations
Molecular modelling was carried out on R5000SC and R10000SC Silicon Graphics work stations using AMBER 94 software (31) for energy minimisation and restrained dynamics calculations. Partial charges for the drug atoms were determined by a semi-empirical approach within Spartan 3.1 (35) using the AM1 method (see supplementary material in NAR Online). These values differ somewhat from those used by Fede et al. (27) which were calculated using the MNDO method. MNDO has been shown to calculate point charges closer to those calculated by ab initio methods than AM1 (36), but AM1 point charges can produce a more accurate molecular electrostatic potential, although this is somewhat molecule-dependent (37). Atom types were assigned using standard AMBER 94 rules and missing force-field parameters estimated from known values for similar types of molecular fragments. Improper torsion angle restraints were centred around both ends of bonds between the individual aromatic moieties to keep them close to the favoured planar orientation.
The drug was docked with canonical B-DNA using the Leap module of AMBER. Counterions were added to neutralise charges (17 sodium ions) and the system was solvated to a minimum distance of 5 Å around the solute using boxes of 216 TIP3P waters. The drug was docked in three different positions to ensure the binding site found was not biased by the starting structure; position (b) is the site judged by eye to be a close fit to the NOEs and positions (a) and (c) shift the drug by 1 bp in either direction along the groove and also out of the groove by [sim]3 Å to ensure that no insurmountable energy barriers were encountered during relaxation to the optimal interaction geometry. Prior to the main molecular dynamics simulations, the systems were subjected to a careful conditioning protocol. First, a 5000 step conjugate gradient minimisation was applied to the water and then the whole system. Then, with the coordinates of the DNA, counterions and drug frozen, the water was subjected to 10 ps molecular dynamics at 100 K. In successive 10 ps steps, with the PME method activated, the counterions were released, then the drug, before finally the system was heated to 300 K over 5 ps and held there for a further 5 ps.
A set of 25 key NOE restraints to the groove floor was then applied to the three conditioned structures while the DNA was restrained with a force of 100 kcal mol-1 Å-2. The drug-DNA restraint force constants were ramped up over 10 ps at 300 K; after a further 40 ps of constant temperature dynamics the system was cooled to 1 K over 10 ps. With no restraints applied, the whole system was then energy minimised. RMS deviations between starting and final energy minimised structures were calculated using MOLMOL (38).
RESULTS AND DISCUSSION
Complex stability
We have previously shown using UV melting data for the free DNA and the complexes with both H33258 and H43254 (29) that both drugs appreciably stabilise the duplex to thermal denaturation. The effect is most pronounced for the tetrahydropyrimidinium analogue with a Tm [sim]7°C higher than for the N-methylpiperazine relative (52 versus 45°C, respectively), equating to a difference in ligand binding energies of [sim]5 kJ mol-1 (10). The NMR experiments show that both drugs lift the C2v symmetry of the DNA duplex such that the two strands are non-equivalent in the complex. This is clearly evident from the observation of six thymine methyl resonances in the spectrum of the complex indicating that the drug is located principally at a single high affinity site. A number of chemical exchange cross-peaks are identified between corresponding protons on the non-equivalent strands of the decamer, indicating that the drug is dissociating slowly from the complex at 298 K and is able to re-bind in a 180° related orientation. Such exchange cross-peaks in NOESY spectra have proved useful in confirming resonance assignments in a number of cases. Quantitative examination of exchange cross-peak intensities suggests that the rate of exchange is broadly similar to rates previously reported in other complexes (400-800 ms at 298 K) (25,39).
Figure 2. Portion of the 200 ms NOESY spectrum of the H43254-d(GGTAATTACC)2 complex recorded in D2O at 298 K illustrating the d(1[prime];6,8) sequential connectivity pathway used in the assignment process. Connectivities within each strand are illustrated separately in the two panels which otherwise show identical regions of the NOESY spectrum. Intranucleotide d(1[prime];6,8) NOEs for each nucleotide are labelled. In the left hand panel we are unable to unambiguously identify the A8*-C9* correlation. A number of other drug-DNA NOEs are also highlighted (a-e), and are assigned as follows: (a) ph H2/6-A4 H1[prime]; (b) ph H2/6-A5 H1[prime]; (c) bz1 H4-T6* H1[prime]; (d) bz1 H4-T6 H1[prime]; and (e) ph H3/5-T7* H1[prime].
Structural analysis by NMR: DNA resonance assignments
Deoxyribose spin systems have been identified using TOCSY data in combination with patterns of cross-peaks identified from NOESY data collected at mixing times of 100 and 200 ms. Cytosine H5/H6 and thymine 5-CH3/H6 are unambiguously identified in TOCSY spectra, while sequential assignments for deoxyribose spin systems and base H6 or H8 were determined using established procedures based on patterns of internucleotide NOEs via the d(1[prime];6,8) (Fig. 2) and d(2[prime],2[prime][prime];6,8) independent connectivity pathways (40). While the latter pathway yields a complete assignment for both strands, we were unable to identify the H1[prime] of A8* using the d(1[prime];6,8) route. NOE connectivities observed in the aromatic region of the spectrum between sequentially neighbouring base H6 or H8 confirmed the base proton assignments established using the other two routes, and also the integrity of base stacking interactions in the complex. A number of sequential adenine H2-H2 NOEs are also apparent. Most imino and amino proton resonances have been identified from H2O NOESY data sets collected using the 1-1 jump-return solvent suppression scheme, and are illustrated in part in Figure 3. All sequential imino-imino NOEs could be identified in the 12-15 p.p.m. region of the spectrum, and TNH to AH2 NOEs confirmed the latter assignments. A full list of chemical shift assignments is available as supplementary material via NAR Online.
Figure 3. Portion of the 300 ms 1-1 NOESY spectrum of the H43254-d(GGTAATTACC)2 complex recorded in 90% H2O/10 D2O at 298 K illustrating drug-DNA NOEs (in boxes). These are highlighted as (a-m), and are assigned as follows: (a) bz1 H4-T6 NH; (b) bz1 H4-T6* NH; (c) ph H2/6-T7* NH; (d) ph H3/5-T7* NH; (e) bz2 H4-T7* NH; (f) ph H3/5-T3 NH and bz2 NH; (g) bz1 NH-A4* H2; (h) bz1 NH-A5* H2; (i) tp NH-A4* H2; (j) tp NH-A5* H2; (k) bz1 H4-A4* H2; (l) ph H3/5-A4 H2; (m) ph H2/6-A4 H2. In addition, several intradrug NOEs (n-q) are highlighted that are important for assignment purposes: (n) bz1 NH-bz1 H4/bz2 H4; (o) tp NH-bz1 H4; (p) tp NH-bz1 H7; and (q) tp NH-bz1 H6. The corresponding 1D spectrum is illustrated along the top axis and the positions of the drug bz1 NH, bz2 NH and tp NH resonances are indicated. T3 NH and bz2 NH have very similar chemical shifts ([sim]13.4 p.p.m.) at this temperature.
Assignment of drug resonances in the complex
Several sets of resonances are readily attributed to the bound drug from an analysis of DQF-COSY and TOCSY data. The aromatic region of the spectrum is devoid of DNA scalar coupling cross-peaks below 7.0 p.p.m. Thus, the three pairs of doublets can be assigned as bz1 H6/H7, bz2 H6/H7 and ph H2,6/H3,5. Coupling between methylene protons of the tetrahydropyrimidinium (tp) ring are readily identified in the aliphatic region of the spectrum. We observe only a single resonance for the four protons of the tp 3-CH2 and 5-CH2 at 3.80 p.p.m., which are coupled to a single resonance for the 4-CH2 group at 2.40 p.p.m. Equivalence between tp 3-CH2 and 5-CH2 signals suggests rapid conformational averaging between the two sides of the symmetrical pyrimidinium ring, consistent with 180° flips around the C1-C5 (bz1) bond. This dynamic process is also suggested by the tp NH resonance which is readily identified as a two-proton intensity signal at 9.40 p.p.m. Its assignment is established by a strong NOE to the tp 3-CH2/5-CH2 signal at 3.80 p.p.m. The assignment of the tp NH signal at 9.40 p.p.m. enables us to identify the bz1 H4 and bz1 H6 signals on the basis of strong NOEs (as highlighted in the H2O NOESY spectrum of the complex shown in Fig. 3). The coupled protons bz1 H6/H7 are therefore readily distinguished from bz2 H6/H7, while the ph H2,6/H3,5 are identified on account of their double signal intensity, again reflecting dynamic averaging of the environment on either side of the phenyl ring. A strong NOE from the p-OCH3 group (4.17 p.p.m.) of the phenyl ring readily distinguishes between ph H2,6 and ph H3,5 resonances. Identification of the remaining unassigned signal from bz2 H4 is facilitated by the use of 1H-13C correlations identified in HMQC spectra. Although the 1H chemical shifts of the drug H4 and adenine H2 signals are very similar, the 13C shifts of the attached carbons are quite different, falling into a region (145-152 p.p.m.), some 20 p.p.m. upfield of the 13C shifts from the C2 resonances of the adenine bases. It is evident that the H4 of bz1 and bz2 have closely similar 1H shifts in the complex but are resolved in the HMQC spectrum. Of the six AH2 signals, only four 1H-13C correlations are resolved in the HMQC spectrum. The remaining two are too close to the noise level, but have been identified through NOE connectivities to the imino protons. In the H2O NOESY spectrum (Fig. 3) the bz1NH resonance is identified on the basis of strong NOEs to both bz1 H4 and bz2 H4, and the bz2 NH on the basis of NOEs to bz2 H4 and ph H3/5.
Intermolecular NOEs
On the basis of the assignments described above for drug and DNA resonances we have been able to identify a large number of intermolecular NOEs from both exchangeable and non-exchangeable protons that define the position and orientation of H43254 in the minor groove, spanning the TAATTA tract. Many of these NOEs are highlighted in the various portions of the NOESY spectra already illustrated. The majority of interactions between drug and DNA involve protons located on the concave edge of the drug and the floor (AH2 and TNH) and walls (H1[prime]) of the minor groove. However, we also identify a number of strong NOEs from benzimidazole H6 and H7, located on the convex (solvent exposed) edge of the drug, to deoxyribose H4[prime] and H5[prime]/5[prime][prime]. These interactions are indicative of a deep narrow minor groove with steep walls to maximise van der Waals contacts with the bound ligand. Such interactions have not previously been reported in NMR studies of similar complexes perhaps because of poor spectral resolution in the H4[prime] and H5[prime]/H5[prime][prime] region of the spectrum. In the present complex bz1 H7 is a particularly well-resolved doublet in the 1D spectrum of the complex and is the furthest downfield resonance observed. As a consequence NOEs are readily identified with H4[prime] and H5[prime]/5[prime][prime] of both T7 and T7* which are also well resolved, while bz1 H6 is observed to give NOEs to A8 H5[prime]/H5[prime][prime]. The short range contacts implied by these NOEs are entirely consistent with distances measured from a number of X-ray structures of H33258 (15,18).
NOEs are observed from all four of the structural units of the drug defining intermolecular interactions along the full length of the A-T tract. In light of the fact that the tetrahydropyrimidinium ring appears to enhance binding to d(GGTAATTACC)2 compared with the corresponding N-methylpiperazine analogue, we looked closely for intermolecular NOEs from the tp NH and methylene protons (tp 3-CH2/5-CH2) that might provide some insight as to the origin of the difference. Two weak NOEs are observed from tp NH to A5*H2 and A4*H2, but a large number of medium to strong NOEs are observed to the methylene protons suggesting that a number of van der Waals contacts may be present with the floor of the minor groove. As noted above, the interaction appears to be a dynamic one since the environments on either side of the ring appear to be averaged by rapid 180° ring flips. We have measured a temperature-dependent shift for the tp NH of [sim]3.0 p.p.b./°C (compared with 1.0 p.p.b./°C for bz1NH) suggesting that the temperature coefficient is reflecting averaging between solvent accessible and solvent excluded environments while bz1 NH is buried at the drug-DNA interface.
Table 2.
| a | b | c | |
| a | 4.27 | 12.6 | |
| b | 0.95 | 8.97 | |
| c | 1.34 | 0.80 |
Identification of a single A-tract binding site
On the basis of a set of 25 drug-DNA NOEs, that define intermolecular interactions along the full length of the A-T tract, we have used a restrained dynamics protocol (implemented using AMBER 4.1) to generate an unbiased family of structures and so assess the extent to which the experimental data define the position and orientation of the drug within the minor groove. Three different starting positions of the drug were considered with respect to a canonical B-DNA model of the sequence with pairwise RMS deviations between drug molecules in the range 4.3-12.6 Å. Each system was explicitly solvated, including counterions to neutralise phosphate charges. After equilibration and conditioning (see Materials and Methods), NOE restraints between drug and DNA only were gradually introduced over 10 ps and each structure subjected to constant temperature dynamics (300 K), followed by cooling to 1 K over a period of 10 ps and finally unrestrained minimisation. Each of the three starting conformations converged to a similar structure (Fig. 4A) with a pairwise RMSD between drug molecules over all heavy atoms of 0.80-1.34 Å (Table 2), and pairwise RMSDs over all heavy atoms of the DNA of 0.75-0.93 Å. Two distance restraints are consistently violated by >0.1 Å, but these are both long distances, one of them involving a labile proton which is intrinsically less accurately defined than other distances. In one structure these are the only two distances to be violated, and in the other two, they account for the greater part of the distance penalty. Average deviations from the input restraints over the final 5 ps of dynamics are for each structure (a) 0.119 Å ± 0.007, (b) 0.035 Å ± 0.006 and (c) 0.077 Å ± 0.008, although these values rise to 0.364, 0.194 and 0.234 Å respectively after cooling and minimisation.
Figure 4. (A) Stereoview of the three final structures: (a) blue, (b) red and (c) green, after molecular dynamics refinement. The pairwise RMSD between drug molecules is 0.8-1.34 Å, and between DNA structures is 0.75-0.93 Å. (B) Stereoview of H43254 bound at the AATT site in complex (b). Bases only are shown and intermolecular hydrogen bonds are identified as dotted lines. Only hydrogen atoms involved in hydrogen bonds are included for clarity. Hydrogen bonding in the complexes was analysed in the low temperature minimised structures using both heavy atom-proton distances and heavy atom-heavy atom distances. A significant degree of propeller twisting of the central AT base pairs is evident which appears to be instrumental in accommodating the intermolecular hydrogen bonds with the bound ligand (Fig. 4A). A generally similar pattern of interactions is evident between the benzimidazole NHs in all three structures, as represented in Figure 5; distances are summarised in Table 3. However, there are subtle differences. While structures (a) and (b) both suggest that bz1 NH is involved in bifurcated hydrogen bonds, in (c) one of the hydrogen bonds is shorter, but the other is longer than is consistent with a bifurcated interaction. Considering bz2 NH, analogous conclusions are reached: structures (b) and (c) are consistent with bifurcated bonds with adenine N3 and thymine O2, but in (a) only a single hydrogen bond is present to bz2 NH. Only structure (b) has both bz1 NH and bz2 NH compatible with both bifurcated interactions (Fig. 4B). No individual structural parameter appears to account for these differences other than small relative displacements along the groove. We do not see any justification for describing the three conformers as different binding modes but rather small fluctuations within an otherwise converged binding site at the centre of the AATT tract. The quality of the fit to the experimental NMR restraints suggests that the complex is well defined by a single binding site model. The proposed pattern of hydrogen bonding interactions compares favourably with several of the X-ray structures in which H33258 is complexed with the AATT tract of d(CGCGAATTCGCG)2 (15,18). In addition, the distance of the pyrimidinium NH to A5 N3 is, in two out of three structures, close to optimal for hydrogen bond formation. In structure (c) the ring has a slightly different orientation corresponding to rotation around the C1-C5(bz1) bond that lengthens this interaction, but suggests a weak bifurcated interaction with both an N3 and thymine O2, a feature seen in all structures prior to cooling and minimisation. This result suggests that the pyrimidinium portion of the structure is, perhaps not surprisingly, less well constrained within the groove, being on the end of a fairly rigid extended aromatic ring system. The calculations show that a number of conformations are possible that satisfy the constraints but all lie entirely within the confines of the same binding site. This apparent orientational flexibility of the pyrimidinium ring, and the dynamic nature of the hydrogen bond that must be synonymous with this, are consistent with the NMR data that suggest that the ring flips rapidly in the bound state averaging the chemical shift environments on either side of the pyrimidinium ring. The persistance of a short range hydrogen bond between the pyrimidinium ring and the floor of the groove in two of the three structures suggests that this interaction may impart some degree of enhanced AT sequence specifity and higher binding affinity over H33258. Figure 5. Schematic representation of intermolecular hydrogen bonding interactions involving adenine N3 and thymine O2 in the DNA minor groove with bz1 NH, bz2 NH and tp NH of the bound drug. Hydrogen bond distances measured from the three structures in Figure 4A are presented in Table 3. 
Conformation of the bound ligand
The torsion angles [alpha]1, [alpha]2 and [alpha]3 (Fig. 1) define the conformation of the ligand in the minor groove. From the dynamics trajectories of structures (a), (b) and (c), we have determined average values for [alpha]1, [alpha]2 and [alpha]3 by considering 10 structures sampled over the last 5 ps of the dynamics run at 300 K. The data are presented in Table 4. It is evident that the conformation of the bound ligand in each structure is determined with some precision with relatively small fluctuations in conformation about the mean values. Moreover, the three structures are very similar to each other, identifying a common converged bound conformation derived from the three independent starting structures. In all cases the rings are reasonably close to co-planarity with a maximum deviation for [alpha]3 of only [sim]20°, while values for [alpha]2, which defines the relative orientation of the two benzimidazole rings, fall in the narrow range [sim]6-14°. We have also analysed the torsion angles in the cooled (1 K), minimised structures and find that the deviations from coplanarity are even smaller (Table 4), consistent with the conclusions from low temperature X-ray structures (18). The drug-DNA interactions appear to be sufficiently accommodating to allow the drug to adopt its lowest energy conformation at low temperature. Noticeably, in our low temperature structures the NOE restraint violation energy increases very slightly, indicating that some deviation from coplanarity is necessary to allow the drug to optimally match the curvature of the minor groove.
Table 3.
| Structure | bz2 NH-A5 N3 | bz2 NH-T7* O2 | bz1 NH-T6 O2 | bz1 NH-T6* O2 | tp NH-T7O2tp | tp NH-A5* N3 |
| a | 2.09 (3.00) | 3.84 (4.24) | 2.49 (3.24) | 2.61 (3.09) | 2.04 (3.02) | 3.15 (3.43) |
| b | 2.33 (3.14) | 2.66 (3.09) | 2.42 (3.09) | 2.14 (2.81) | 3.39 (3.85) | 2.14 (3.02) |
| c | 2.50 (3.36) | 2.67 (3.12) | 3.20 (3.76) | 1.90 (2.79) | 3.17 (3.60) | 2.04 (2.97) |
Table 4.
| a1 (300 K) |
b1 (300 K) |
c1 (300 K) |
a2 (1 K) |
b2 (1 K) |
c2 (1 K) |
NMR3 | X-ray4 I |
X-ray5 II |
|
| [alpha]1 | 0.8 ± 4.9 | 2.3 ± 6.7 | -2.1 ± 4.9 | 0.5 | 3.1 | -0.1 | (I) 23 to 70 | (i) 8 (8) | |
| (II) 39 to 84 | (ii) 10 | 8 | |||||||
| (iii) 30 | |||||||||
| [alpha]2 | 11.9 ± 4.0 | 3.7 ± 2.3 | 6.0 ± 4.0 | 2.6 | 3.6 | -0.4 | (I) 11 to 27 | (i) 20 (15) | |
| (II) -23 to -33 | (ii) 12 | 32 | |||||||
| (iii) 2 | |||||||||
| [alpha]3 | 22.2 ± 5.3 | 19.4 ± 6.2 | 17.3 ± 6.5 | 9.3 | 20.3 | 3.9 | (I) -60 to -106 | (i) 9 (3) | |
| 56 to 116 | (ii) 8 | 14 | |||||||
| (II) -52 to -64 | (iii) 6 |
2Values from one single low temperature structure after energy minimisation at 1 K.
3Data from the NMR structures (I and II) of Fede et al. (27).
4Data from the X-ray structure of Quintana et al. (18) at (i) 273 K piperazine up, (in parentheses) piperazine down (ii) 248 K (iii) 173 K.
5Data from the X-ray structure of Teng et al. (15) at 298 K.
Comparison with other solution structures
Only one NMR investigation to date has attempted to determine in detail the structure and dynamics of the interaction between H33258 and the DNA minor groove from studies of the complex with d(GGTGAATTCCAC)2 (27,39). Using a combination of random docking and subsequent structural refinement based upon a simulated annealing protocol using 23 intermolecular NOE restraints, these authors have identified two families of conformers. These differ primarily in lying in one of two well-defined narrow ranges of the [alpha]2 torsion angle, which defines the relative orientation between the two benzimidazole rings. By comparison a much broader range of values are observed for [alpha]1 and [alpha]3, consistent with a greater degree of mobility for the N-methylpiperazine and phenol rings (Table 4). In all of the NMR studies to date, these 2-fold symmetric substituents (including the pyrimidinium ring in this study) are undergoing rapid rotation within the lifetime of the bound ligand. The conclusion drawn from these studies is that the benzimidazole rings are much less mobile and are well-defined in terms of the these two families of conformers. In family I of the solution conformers described by Fede et al. (27), [alpha]2 deviates from planarity in the range 11-27°. This family of conformers compares favourably with our own family where [alpha]2 lies in a similar, partially overlapping, range 2-16° (taken from 30 structures; see above). These ranges are in good agreement with those from the X-ray structures (2-22°) determined by Quintana et al. (18) at a number of temperatures, but are slightly smaller than the value of 32° measured by Teng et al. (15) in their structure with the drug also located at the centre of an AATT tract. In contrast, family II of the solution conformers determined by Fede et al. (27) have negative [alpha]2 values in the range ca -23 to -33° for which no counterparts are observed in the X-ray structures or in our NMR structures.
We have shown, using intermolecular NOE restraints and a restrained dynamics protocol based on three independent starting structures, that H43254, a tetrahydropyrimidinium analogue of H33258, binds at a well-defined site positioned centrally in the AATT tract. In all three structures the pyrimidinium ring deviates relatively little from co-planarity with the adjoining benzimidazole ring ([alpha]1 = -7 to +12°) fitting snugly between the walls of the narrow minor groove. In two of the three structures the pyrimidinium ring forms short range hydrogen bonds with N3 of A5, while in the third structure a weak bifurcated interaction is observed with the same N3 but also T7 O2. These favourable electrostatic interactions may account for some part of the higher binding affinity ([sim]5 kJ mol-1) of this drug compared to H33258. The highly complementary `fit' of the tetrahydropyrimidinium ring between the walls of the minor groove suggests that van der Waals interactions, and even a smaller cost in free energy in perturbing the DNA conformation, are all important factors in accounting for the effects of replacing the N-methylpiperazine ring with a tetrahydropyrimidinium analogue.
ACKNOWLEDGEMENTS
We thank Hoechst, Frankfurt, Germany for a sample of Hoechst 43254 and John Keyte for oligonucleotide synthesis. The coordinates of all calculated structures are available from the authors on request. We thank the EPSRC for a research studentship for C.E.B.-S. M.S.S. is grateful to The Royal Society, Nuffield Foundation and the Department of Chemistry (Nottingham) for financial support. C.A.L. gratefully acknowledges the University of Nottingham for a New Lecturers Award.
See supplementary material available in NAR Online.
REFERENCES
This page is run by Oxford University Press, Great Clarendon Street, Oxford OX2 6DP, as part of the OUP Journals
Comments and feedback: www-admin{at}oup.co.uk
Last modification: 24 Mar 1998
Copyright© Oxford University Press, 1998.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  C. Bailly, G. Chessari, C. Carrasco, A. Joubert, J. Mann, W. D. Wilson, and S. Neidle Sequence-specific minor groove binding by bis-benzimidazoles: water molecules in ligand recognition Nucleic Acids Res., March 1, 2003; 31(5): 1514 - 1524. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. E. Bostock-Smith, S. A. Harris, C. A. Laughton, and M. S. Searle Induced fit DNA recognition by a minor groove binding analogue of Hoechst 33258: fluctuations in DNA A tract structure investigated by NMR and molecular dynamics simulations Nucleic Acids Res., February 1, 2001; 29(3): 693 - 702. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Gavathiotis, G. J. Sharman, and M. S. Searle Sequence-dependent variation in DNA minor groove width dictates orientational preference of Hoechst 33258 in A-tract recognition: solution NMR structure of the 2:1 complex with d(CTTTTGCAAAAG)2 Nucleic Acids Res., February 1, 2000; 28(3): 728 - 735. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||