DNA sequence recognition by under-methylated analogues of triostin A

Preferred binding sites for the bifunctional intercalator TANDEM determined using DNA fragments that contain every symmetrical hexanucleotide sequence

We have prepared novel DNA footprinting substrates that contain all 64 symmetrical hexanucleotide sequences. These were contained in two restriction fragments that were cloned into the pUC19 polylinker site; each fragment was also obtained in both orientations. These fragments were used to assess the sequence binding preferences of the synthetic quinoxaline antibiotic TANDEM. We found that, although the ligand binds to most TpA steps, the affinity is affected by the flanking sequences. The best binding sites contain the tetranucleotide sequence ATAT, although YATATR is a better site than RATATY. TTAA always is a poor binding site, especially TTTAAA. The binding to GTAC is strongly dependent on the flanking bases, with good binding to GGTACC but none at all to CGTACG.

Bisintercalator natural products with potential therapeutic applications: isolation, structure determination, synthetic and biological studies

Echinomycin is the prototypical bisintercalator, a molecule that binds to DNA by inserting two planar chromophores between the base-pairs of duplex DNA, placing its cyclic depsipeptide backbone in the minor groove. As such, it has been the focus of an extensive number of investigations into its biological activity, nucleic acid binding and, to some extent, its structure-activity relationships. However, echinomycin is also the parent member of an extended family of natural products that interact with DNA by a similar mechanism of bisintercalation. The structural variety in these compounds leads to changes in sequence selectivity and and biological activity, particularly as anti-tumour and anti-viral agents. One of the more recently identified marine natural products that is moving close to clinical development is thiocoraline, and it therefore seems timely to review the various bisintercalator natural products.

Solution-phase synthesis of nucleobase-substituted analogues of triostin A

A synthesis of novel analogues of triostin A presenting two identical or different nucleobases instead of the original quinoxaline substituents has been developed. The DNA bisintercalator triostin A (1) with its rigid backbone provides an optimal scaffold for a parallel preorganization of the intercalating moieties. The bicyclic octadepsipeptide is built up stepwise in solution and modified with various nucleobase-substituted acetic acids at a late stage. The choice of orthogonal protecting groups allows for the synthesis of triostin analogues bearing two different substituents.

Preferred binding sites for [N-MeCYs(3), N-MeCys(7)]TANDEM determined using a universal footprinting substrate

We have prepared a novel footprinting substrate which contains all 136 tetranucleotide sequences and have used this to determine the preferred binding sites for the synthetic quinoxaline antibiotic [N-MeCys(3),N-MeCys(7)]TANDEM. We find that, although the ligand binds to all TpA steps, it binds best to the tetranucleotide sequence ATAT and shows only weak interaction with TTAA and GTAC. The best binding sites contain the sequences ATAX and XTAT.

DNA recognition by quinoline antibiotics: use of base-modified DNA molecules to investigate determinants of sequence-specific binding of luzopeptin

The luzopeptin antibiotics contain a cyclic decadepsipeptide to which are attached two quinoline chromophores that bisintercalate into DNA. Although they bind DNA less tightly than the structurally related quinoxaline antibiotics echinomycin and triostin A, the molecular basis of their interaction remains unclear. We have used the PCR in conjunction with novel nucleotides to create specifically modified DNA for footprinting experiments. In order to study the influence that removal, addition or relocation of the guanine 2-amino group, which normally identifies G.C base pairs from the minor groove, has on the interaction of luzopeptin antibiotics with DNA. The presence of a purine 2-amino group is not strictly required for binding of luzopeptin to DNA, but the exact location of this group can alter the position of preferred drug binding sites. It is, however, not the sole determinant of nucleotide sequence recognition in luzopeptin-DNA interaction. Nor can the selectivity of luzopeptin be attributed to the quinoline chromophores, suggesting that an analogue mode of DNA recognition may be operative. This is in contrast to the digital readout that seems to predominate with the quinoxaline antibiotics.

DNA recognition by quinoxaline antibiotics: use of base-modified DNA molecules to investigate determinants of sequence-specific binding of triostin A and TANDEM

The methodology of DNAase I footprinting has been adapted to investigate the sequence-specific binding of two quinoxaline drugs to DNA fragments containing natural and modified bases. In order to help comprehend the molecular origin of selectivity in the bis-intercalation of triostin A and TANDEM at CpG and TpA sites respectively, we have specifically examined the effect of the 2-amino group of guanine on their sequence specificity by using DNA in which that group has been either removed from guanine, added to adenine or both. Previous studies suggested that the recognition of particular nucleotide sequences by these drugs might be dependent upon the placement of the purine 2-amino group, serving as a positive or a negative effector for triostin A and TANDEM respectively. However, the footprinting data reported here indicate that this is not entirely correct, since they show that the 2-amino group of guanine is important for the binding of triostin A to DNA but has absolutely no influence on the interaction of TANDEM with TpA steps. Apparently the binding of triostin A to CpG sites is primarily due to hydrogen bonding interaction between the cyclic peptide of the antibiotic and the 2-amino group of guanine residues, whereas the selective binding of TANDEM to TpA sites is not hydrogen-bond driven and probably originates mainly from steric and/or hydrophobic interactions, perhaps involving indirect recognition of a suitable minor groove structure.

Dissociation of the AT-specific bifunctional intercalator [N-MeCys3,N-MeCys7]TANDEM from TpA sites in DNA

We have examined the dissociation of [N-MeCys3,N-MeCys7]TANDEM, an AT-selective bifunctional intercalator, from TpA sites in mixed-sequence DNAs by a modification of the footprinting technique. Dissociation of complexes between the ligand and radiolabelled DNA fragments was initiated by adding a vast excess of unlabelled calf thymus DNA. Portions of this mixture were subjected to DNAse I footprinting at various times after adding the competitor DNA. Dissociation of the ligand from each site was seen by the time-dependent disappearance of the footprinting pattern. Within a natural DNA fragment (tyrT) the ligand dissociates from TTAT faster than from ATAT. We found that the stability of complexes with isolated TpA steps decreases in the order ATAT > TTAA > TATA. Dissociation from each of these sites is much faster than from longer regions of (AT)n. These results confirm the requirement for A and T base-pairs surrounding the TpA step and suggest that the interaction is strongest with regions of alternating AT, possibly as a result of its unusual structure. The ligand dissociates more slowly from the centre of (AT)n tracts than from the edges, suggesting that variations in dissociation rate arise from sequence-dependent variations in local DNA structure.

DNA recognition by intercalators and hybrid molecules

Experiments are described which probe the role of the 2-amino group of guanine as a critical determinant of the recognition of nucleotide sequences in DNA by specific ligands. Homologous samples of tyrT DNA substituted with inosine or 2,6-diaminopurine residues in place of guanosine or adenine respectively yield characteristically modified footprinting patterns when challenged with sequence-selective antibiotics such as echinomycin, actinomycin or netropsin. The capacity of small molecules to recognise particular DNA sequences is exploited in the 'combilexin' strategy to target small molecules to defined sites in DNA. A composite molecule containing a distamycin moiety linked to an intercalating ellipticine derivative has been synthesised and shown to bind tightly to DNA but without much sequence-selectivity. Refinement of this molecule based on predictions from molecular modelling has led to the synthesis of a second generation derivative bearing an additional positive charge: this new hybrid molecule is strongly selective for binding to AT-rich tracts in DNA.

Sequence-specific binding of [N-MeCys3,N-MeCys7]TANDEM to TpA

The sequence selective binding of [N-MeCys3,N-MeCys7]TANDEM to DNA has been studied by footprinting experiments on DNA fragments containing the self-complementary sequences CGCGATATCGCG, CGCGTATACGCG, CGCGTTAACGCG and CGCGAATTCGCG. DNAase I and micrococcal nuclease reveal drug-induced footprints with the central sequences ATAT, TATA and TTAA, but not AATT, suggesting that the ligand binds to the dinucleotide TpA. The ligand renders certain adenines hyper-reactive to diethyl pyrocarbonate. These are observed with ATAT, TATA and TTAA, but not AATT, and are located both within, and distal to, the TpA-binding sites.

Strand orientation of [alpha]-oligodeoxynucleotides in triple helix structures: dependence on nucleotide sequence

The aims of the present theoretical study of the conformations of [alpha]-oligodeoxynucleotides forming triple helices with DNA duplexes are to understand the structural and energetic factors involved in [alpha]-triple helix formation by means of energy minimization, and to explain the experimentally observed dependence of strand orientation on the nucleotide sequence. It is found that the energetically preferred orientation of the [alpha]-oligonucleotide with respect to the homopurine strand depends on the sequence of the homopurine.homopyrimidine tracts. This is a consequence of the structural heteromorphism of base triplets in the intrinsically more stable reverse Hoogsteen hydrogen bonding configuration. Practical rules are proposed for determining the orientation of the nuclease-resistant [alpha]-oligodeoxynucleotide strand which will form the most stable triple helix.

Proton NMR studies of [N-MeCys3,N-MeCys7]TANDEM binding to DNA oligonucleotides: sequence-specific binding at the TpA site

[N-MeCys3,N-MeCys7]TANDEM, an undermethylated analogue of Triostin A, contains two N-methyl groups on the cysteine residues only. Footprinting results showed that [N-MeCys3,N-MeCys7]TANDEM binds strongly to DNA rich in A.T residues [Low, C. M. L., Fox, K. R., Olsen, R. K., & Waring, M. J. (1986) Nucleic Acids Res. 14, 2015-2033]. However, it was not known whether specific binding of [N-MeCys3,N-MeCys7]TANDEM requires a TpA step or an ApT step. In 1:1 saturated complexes with the octamers [d(GGATATCC)]2 and [d(GGTTAACC)]2, [N-MeCys3,N-MeCys7]TANDEM binds to each octamer as a bis-intercalator bracketing the TpA step. The octadepsipeptide ring binds in the minor groove of the DNA. Analysis of sugar coupling constants from the phase-sensitive COSY data indicates that the sugar of the thymine in the TpA binding site adopts predominantly an N-type sugar conformation, while the remaining sugars on the DNA adopt an S-type conformation, as has been observed in other Triostin A and echinomycin complexes. The drug does not bind to the octamer [d(GGAATTCC)]2 as a bis-intercalator. Only weak nonintercalative binding is observed to this DNA octamer. These results show unambiguously that [N-MeCys3,N-MeCys7]TANDEM binds sequence specifically at TpA sites in DNA. The factors underlying the sequence specificity of [N-MeCys3,N-MeCys7]TANDEM binding to DNA are discussed.

Echinomycin binding to alternating AT

We have studied the binding of echinomycin to DNA fragments containing GC-rich regions flanked by blocks of alternating AT by DNase I footprinting and diethylpyrocarbonate modification. Regions of alternating AT flanking the sequences CCCG, CCGC, CGGC and GG show a four base pair DNase I cleavage pattern and reaction of alternate adenines with diethylpyrocarbonate. This pattern is strongest when the AT-block is immediately adjacent to the CpG ligand binding site. We explain these phenomena by suggesting that echinomycin binds to the dinucleotide step ApT in a cooperative fashion. The cooperative effects can be transmitted through the dinucleotide step GC but not CC or AA. No such repetitive patterns are seen with surrounding regions of (ATT).(AAT). Evidence is presented for secondary drug binding sites at CpC and TpG with weaker interaction at the CpG site within the hexanucleotide TTCGAA.

Conformational heterogeneity of quinoxaline peptides in solution

The conformational heterogeneity of several quinoxaline antibiotics, a class of naturally occurring quinoxaline peptides with antitumor properties, and their synthetic analogues was investigated in polar and nonpolar solvents by high performance liquid chromatography (HPLC) with uv photodiode array detection, uv-absorbance, low-temperature phosphorescence, and nmr techniques. Multiple peak formation and interconversion in the HPLC and 1H-nmr analysis of triostin A, its under-N-methylated synthetic analogues (des-N-tetramethyltriostin A [TANDEM] and [N-MeCys3, N-MeCys7]-TANDEM [MCTAN-DEM]), and echinomycin were examined as a function of temperature, solvent polarity, and residence time in solution prior to analysis. Slow interconversion between HPLC peaks, ascribed to the presence of multiple solution conformers, was exhibited by these peptides although at very different interconversion rates. Among the triostins, the rate of interconversion appeared to vary with the degree of N-methylation of the residues in the cyclic depsipeptide chain. Interconversion of the n and p conformers of triostin A in chloroform occurred on a chromatographic timescale (a few minutes with kn----p calculated to be 0.02 s-1 at 25 degrees C) while the solution conformers of TANDEM in methanol equilibrated very slowly to one preferred conformer over a period of several weeks at ambient temperature. MCTANDEM, a synthetic analogue of triostin A with an intermediate degree of N-methylation of the residues in the peptide ring, consisted of an equilibrium mixture of n and p conformers in methanol that interconverted on a chromatographic time scale. Two additional conformers of MCTANDEM developed within a few weeks' residence time in methanol at ambient temperature. Echinomycin was found to exist in methanol as an interconverting mixture of at least four minor conformers in addition to the major isoform (95% by peak area) of the peptide. The solution conformers of the quinoxaline peptides investigated in this report are most likely a consequence of hindered rotation about the N-methylated peptide bonds in the depsipeptide ring and/or intramolecular hydrogen bonding.

Sequence-selective binding of an ellipticine derivative to DNA

The DNA sequence specificity of an ellipticine derivative bearing an aminoalkyl side chain has been determined by a variety of footprinting methods. The drug exhibits sequence selective binding and discriminates against runs of adenines or thymines. Binding is shown to occur at various sequences with a preference for GC rich regions of DNA. A large enhancement of DNAase I and of hydroxyl radical cleavage in regions rich in A's or T's is observed together with hyperreactivity of adenines towards diethylpyrocarbonate in the presence of drug. This indicates the occurrence of drug-induced changes in critical conformational features of DNA. The total absence of hyperreactivity of guanine residues towards diethylpyrocarbonate appears to be related to the sequence selectivity of drug binding. No alteration of the dimethyl sulphate and methylene blue-induced cleavage of DNA is observed. Irradiation of ellipticine derivative-DNA complexes with UV light followed by alkali treatment leads to selective photocleavage at guanine residues, consistent with the deduced degree of selectivity of the binding reaction.

Liquid chromatographic investigation of quinoxaline antibiotics and their analogues by means of ultraviolet diode-array detection

Numerical formats for evaluation of spectral purity and for spectral comparison of ultraviolet diode-array detector data, together with library search routines, were applied to the liquid chromatographic analysis of echinomycin, triostin A and their synthetic and biosynthetic analogues. Samples of monoquinoline and bisquinoline analogues of echinomycin were found to contain echinomycin and the other respective analogue. Triostin A and its undermethylated synthetic analogues, des-N-tetramethyltriostin A (TANDEM) and [MeCys3,MeCys7]-TANDEM, were each composed of two or more components. Triostin A primarily consisted of a major chromatographic component and a minor component with very similar ultraviolet spectral features. TANDEM exhibited three chromatographic components with nearly identical ultraviolet spectral characteristics. Apparent conformational interconversion of at least two forms of the [MeCys3,MeCys7]-TANDEM analogue was observed by reversed-phase liquid chromatography. An activation energy of 15 kcal/mol was estimated for the interconversion based upon an Arrhenius plot of the data.

Two-dimensional 1H and 31P NMR spectra of a decamer oligodeoxyribonucleotide duplex and a quinoxaline ((MeCys3, MeCys7)(TANDEM) drug duplex complex

Assignment of the 1H and 31P NMR spectra of a decamer oligodeoxyribonucleotide duplex, d(CCCGATCGGG), and its quinoxaline ((MeCys3, MeCys7]TANDEM) drug duplex complex has been made by two-dimensional 1H-1H and heteronuclear 31P-1H correlated spectroscopy. The 31P chemical shifts of this 10 base pair oligonucleotide follow the general observation that the more internal the phosphate is located within the oligonucleotide sequence, the more upfield the 31P resonance occurs. While the 31P chemical shifts show sequence-specific variations, they also do not generally follow the Calladine "rules" previously demonstrated. 31P NMR also provides a convenient monitor of the phosphate ester backbone conformational changes upon binding of the drug to the duplex. Although the quinoxaline drug, [MeCys3, MeCys7]TANDEM, is generally expected to bind to duplex DNA by bis-intercalation, only small 31P chemical shift changes are observed upon binding the drug to duplex d(CCCGATCGGG). Additionally, only small perturbations in the 1H NMR and UV spectra are observed upon binding the drug to the decamer, although association of the drug stabilizes the duplex form relative to the other states. These results are consistent with a non-intercalative mode of association of the drug. Modeling and molecular mechanics energy minimization demonstrate that a novel structure in which the two quinoxaline rings of the drug binds in the minor groove of the duplex is possible.

Actinomycin D induced DNase I cleavage enhancement caused by sequence specific propagation of an altered DNA structure

Two DNA hexadecamers containing one central 5'-GC-3' base step have been examined by footprinting methodology in the presence and absence of actinomycin D. The results of these studies, coupled with imino proton NMR measurements indicate that the antitumor drug causes a change in DNA conformation at a distance from the actinomycin intercalation site in a molecule of sequence d[ATATATAGCTATATAT] that does not occur in d[AAAAAAAGCTTTTTTT]. The experiments demonstrate that DNase I rate enhancements associated with actinomycin D binding are caused by ligand alteration of equilibrium DNA structure.

Diethyl pyrocarbonate can detect a modified DNA structure induced by the binding of quinoxaline antibiotics

The reactivity of the 160 bp tyrT DNA fragment towards diethyl pyrocarbonate (DEPC) has been investigated in the presence of bis-intercalating quinoxaline antibiotics and the synthetic depsipeptide TANDEM. At moderate concentrations of each ligand, specific purine residues (mainly adenosines) exhibit enhanced reactivity towards the probe, and several sites of enhancement appear to be related to the sequence selectivity of drug binding. Further experiments were performed with echinomycin at pH 5.5 and 4.6 to facilitate the protonation of cytosine required for formation of Hoogsteen GC base pairs. No significant increase in reactivity was observed under these conditions. Additionally, no protection of deoxyguanosine residues from methylation by dimethyl sulphate was observed in the presence of echinomycin. We conclude that the structural anomaly giving rise to drug-dependent enhanced DEPC reaction is not simply the formation of Hoogsteen base pairs adjacent to antibiotic binding sites. Nor is it due to a general unwinding of the double helix, since we show that conditions which are supposed to unwind the helix lead to a uniform increase in purine reactivity, regardless of the surrounding nucleotide sequence.

Sequence-specific binding of luzopeptin to DNA

We have examined the binding of luzopeptin, an antitumor antibiotic, to five DNA fragments of varying base composition. The drug forms a tight, possibly covalent, complex with the DNA causing a reduction in mobility on nondenaturing polyacrylamide gels and some smearing of the bands consistent with intramolecular cross-linking of DNA duplexes. DNAase I and micrococcal nuclease footprinting experiments suggest that the drug binds best to regions containing alternating A and T residues, although no consensus di- or trinucleotide sequence emerges. Binding to other sites is not excluded and at moderate ligand concentrations the DNA is almost totally protected from enzyme attack. Ligand-induced enhancement of DNAase I cleavage is observed at both AT and GC-rich regions. The sequence selectivity and characteristics of luzopeptin binding are quite different from those of echinomycin, a bifunctional intercalator of related structure.

The use of micrococcal nuclease as a probe for drug-binding sites on DNA

The cutting pattern produced by micrococcal nuclease on three DNA fragments has been determined in the absence and presence of various DNA-binding drugs. The enzyme itself cuts almost exclusively at pA and pT bonds, showing a greater activity at (A-T)n than in homopolymeric runs of A and T. Each drug produces distinct changes in the cleavage pattern. The protected regions can not be pinpointed with sufficient precision to assess the exact drug-binding sites on account of the sequence selectivity of the enzyme, although where a direct comparison is possible these include most of those seen as DNAase I footprints. The enzyme is most useful for assessing the selectivity of drugs which bind to AT-rich regions. Several drugs protect the DNA from micrococcal nuclease attack in regions which do not contain their acknowledged best binding sites. It appears that micrococcal nuclease is sensitive to the existence of secondary drug-binding sites which are not evident with other footprinting techniques.