Introduction

Top Summary Introduction Procedures Results Discussion References

DNA topoisomerase II is a ubiquitous enzyme that regulates DNA topological features; during its catalytic cycle, it cleaves a DNA duplex, allows the passage of another DNA segment through the cut, and finally reseals the strand breaks (Berger et al., 1996). Eukaryotic type II topoisomerases are the targets of highly effective antitumor drugs such as m-AMSA, bisantrene, anthracyclines, and etoposide. The drugs interfere with the DNA breakage-reunion reaction by freezing a transient DNA-enzyme complex in which the DNA strands are cut and covalently linked to the protein (Liu, 1989). The pharmacological activity of these antitumor compounds is thus related to DNA cleavage stimulation in living tumor cells (Pommier, 1993).

Several lines of evidence indicate that topoisomerase II poisons bind to the enzyme/DNA interface at the site of DNA cleavage (Capranico et al., 1997). These drugs are commonly composed of a planar ring system, with DNA-intercalation or -intercalation-like properties, and one or two protruding side chains, which are believed to recognize the enzyme side of the cleavable complex (Osheroff et al., 1994). However, these drugs do not share a unique pharmacophore (Capranico et al., 1997) and can only define a "loose" one, as earlier suggested (Leteurtre et al., 1992). The lack of structural restrictions for the pharmacophore of topoisomerase II poisons can mainly be ascribed to receptor heterogeneity, which is determined by the variability of the local nucleic acid sequence at the enzyme/DNA interface. This is reflected in marked differences of cleavage patterns observed in sequencing gels and local base preferences among the different classes of chemical poisons (Capranico et al., 1990, 1993; Pommier et al., 1991).

In previous attempts to define the structural determinants of this important family of anticancer agents, we demonstrated that, although many chemically unrelated poisons can stimulate DNA cleavage in the presence of topoisomerase II, only compounds sharing common steric and electronic features can trap the enzyme at the same nucleic acid sites (Capranico et al., 1994). This is the case for bisantrene and m-AMSA, which have identical A+1 base requirements for topoisomerase II-mediated DNA cleavage stimulation. In fact, a remarkable similarity in space occupancy and charge distribution was demonstrated for the two compounds, which might account for the equivalent base requirements (Capranico et al., 1994). Stimulated by these findings, we recently investigated a number of bisantrene congeners, including a 10-aza-bioisoster, bearing the same 4,5-dihydro-1H-imidazol-2-yl-hydrazone side chain at positions 1, 4, or 9 of anthracene (Capranico et al., 1998). Moving the bisantrene side chain along the planar ring system (from C9 and C5 to C1 and C4) dramatically affected the base preference and intensity patterns of poison-stimulated DNA cleavage. In contrast, switching the planar aromatic systems of bisantrene and m-AMSA did not substantially alter the sequence specificity of drug action. Base preferences at the cleavage site are reported in Table 1. Indeed, a common pharmacophore was shared by bisantrene, m-AMSA, and the 9-substituted analogues, whereas the 1-substituted regioisomer showed radically changed pharmacophoric properties.

Because topoisomerase poisons, including m-AMSA, likely interact at the protein/DNA interface (Freudenreich and Kreuzer, 1994; Capranico et al., 1997), their DNA-binding activity can be affected by the structural and chemical changes induced by protein/nucleic acid contacts. In the case of anthracyclines, the base preferences exhibited by the drug for naked DNA and for the topoisomerase II-DNA complex are different, indicating that alterations of the drug interaction with the double helix occur in the ternary complex. Amsacrine, on the other hand, shows the same (5'-TA) dinucleotide preference for binding to protein-free DNA and for stabilizing the enzyme-DNA complex (Chen et al., 1988; Pommier et al., 1991). Hence, a thorough understanding of the structural and functional aspects of the effects of the protein on poison-DNA interactions is still missing. In this connection, we were interested in investigating the DNA-binding affinity and specificity of the series of bisantrene/m-AMSA congeners shown in Fig. 1 and comparing them with the sequence preferences found in the topoisomerase II-poisoning experiments (Capranico et al., 1998).

View larger version (17K): [in this window] [in a new window]   Fig. 1.   Chemical structures of the test drugs.

We show that not only the affinity for the nucleic acid is affected; the DS is also dramatically modified upon movement of the side-chain groups along the anthracene ring system. Most interestingly, the changes in specificity exhibited by structural isomers in the binary drug-DNA system precisely correspond to the specificity changes observed for enzyme-mediated DNA cleavage. This information could facilitate further elucidation of the details of the mechanism of action of topoisomerase poisons and will be valuable in the rational design of novel poisons with a predefined site selectivity for DNA cleavage.

	Experimental Procedures

Top Summary Introduction Procedures Results Discussion References

Materials. m-AMSA and bisantrene were obtained from the Drug Synthesis and Chemistry Branch, National Cancer Institute (Bethesda, MD), and from Lederle, respectively. The imidazolyl-hydrazone compounds were synthesized as described elsewhere (Zagotto et al., 1998). All compounds used were stored at 20° in dimethylsulfoxide or deionized water and were diluted in deionized water before use.

1

1

Stability of the test drugs in aqueous media. All compounds were stable in aqueous solution at the working concentrations and did not undergo self-aggregation phenomena up to at least 100 µM. A notable exception was 1,4-IHA, whose molar extinction coefficient decreased with increasing concentration. This phenomenon has been described for molecules containing hydrophobic planar systems that could stack on each other, including bisantrene and mitoxantrone. The dependence of the extinction coefficient on the drug concentration was investigated at different ionic strengths. The experimental data obtained by spectroscopic titrations were consistent with a model of indefinite aggregation (Schwartz et al., 1970), which takes into account aggregation processes beyond a simple monomer-dimer equilibrium. As could be anticipated, considering the presence of protonated side chains, with increasing ionic strength the drug charge becomes shielded, which reduces the repulsive effect and stabilizes the aggregates. An association constant value of approximately 10⁴ M¹ was found.

DNA binding studies. Measurements were carried out at 25° in ETN buffer (1 mM EDTA, 10 mM Tris, pH 7.0, with NaCl to obtain the desired ionic strength). Binding was monitored spectrophotometrically or fluorometrically, in the ligand absorption or emission region, respectively, after addition of scalar amounts of DNA to a freshly prepared drug solution. To avoid large systematic inaccuracies resulting from experimental errors in extinction coefficients or fluorescence quantum yield, the range of bound drug fractions was 0.15-0.85. Data were evaluated according to the equation of McGhee and Von Hippel (1974) for noncooperative ligand-lattice interactions, where r is the molar ratio of bound ligand to DNA, m is the free ligand concentration, K_i is the intrinsic binding constant, and n is the exclusion parameter. Spectroscopic measurements were made with a Perkin-Elmer Lambda 5 apparatus and a MPF66 fluorometer, both equipped with a Haake F3-C thermostat.

T_m studies. Melting curves were measured using a Uvikon 943 spectrophotometer coupled to a Neslab RTE111 cryostat. For each series of measurements, 12 samples were placed in a thermostatically controlled cell-holder (path length, 10 mm), and the quartz cuvettes were heated by circulating water. The measurements were made in BPE buffer (pH 7.1). The temperature inside the cuvette was monitored by using a thermocouple in contact with the solution. The absorbance at 260 nm was measured over the range of 20-100°, with a heating rate of 1°/min. The T_m was taken as the midpoint of the hyperchromic transition.

DNA-unwinding assays. Plasmid DNA (pBR322) was incubated for 30 min with increasing concentrations (0-100 µM) of 9-IHA, aza-9-IHA, and 1-IHA. Aliquots containing 400 ng of DNA (10 µl) were then loaded onto 1% agarose gels, after addition of 2 µl of loading buffer (0.01% bromphenol blue, 0.01% xylene cyanol, 40% sucrose). Gels were run in TBE buffer (89 mM Tris-borate, pH 8.3, 1 mM EDTA) at 10 V/cm. The gels were then stained with ethidium bromide (0.5 mg/liter) and scanned using a BioRad Gel Doc 1000 system.

DNA purification and labeling. Plasmid pBS (Stratagene) was isolated from Escherichia coli by a standard sodium dodecyl sulfate-sodium hydroxide lysis procedure and was purified by banding in CsCl-ethidium bromide gradients. Ethidium was removed by several isopropanol extractions followed by exhaustive dialysis against Tris-EDTA buffer. The purified plasmid was then precipitated and resuspended in appropriate buffer before digestion with the restriction enzymes. The 265-base pair DNA fragment was prepared by 5'-³²P-end-labeling of the EcoRI/alkaline phosphatase-treated plasmid using [-³²P]ATP (6000 Ci/mmol) and T4 polynucleotide kinase, followed by treatment with PvuII. The digestion products were separated on a 6% polyacrylamide gel under native conditions in TBE buffer. After autoradiography, the band of DNA was excised, crushed, and soaked in elution buffer (500 mM ammonium acetate, 10 mM magnesium acetate) overnight at 37°. This suspension was filtered through a Millipore 0.22-µm filter, and the DNA was precipitated with ethanol. After washing with 70% ethanol and vacuum drying of the precipitate, the labeled DNA was resuspended in 10 mM Tris (adjusted to pH 7.0) containing 10 mM NaCl.

DNase I footprinting experiments. Experiments were performed essentially as previously described (Bailly and Waring, 1995). Briefly, reactions were conducted in a total volume of 10 µl. Samples (3 µl) of the labeled DNA fragment were incubated with 5 µl of the buffer solution containing the ligand at the appropriate concentration. After 30 min of incubation at 37° to ensure equilibration of the binding reaction, the digestion was initiated by the addition of 2 µl of a DNase I solution, the concentration of which was adjusted to yield a final enzyme concentration of approximately 0.01 unit/ml in the reaction mixture. After 3 min, the reaction was stopped by freeze-drying. Samples were lyophilized and resuspended in 5 µl of an 80% formamide solution containing tracking dyes. The DNA samples were then heated at 90° for 4 min and chilled on ice for 4 min before electrophoresis.

Electrophoresis and quantitation by storage phosphor imaging. DNA cleavage products were resolved by polyacrylamide gel electrophoresis under denaturing conditions (0.3-mm thick gels, 8% acrylamide containing 8 M urea). After electrophoresis (approximately 2.5 hr at 60 W and 1600 V in TBE buffer; BRL sequencer model S2), gels were soaked in 10% acetic acid for 10 min, transferred to Whatman 3MM paper, and dried under vacuum at 80°. A Molecular Dynamics 425E PhosphorImager was used to collect data from the storage screens exposed to dried gels overnight at room temperature. Base-line-corrected scans were analyzed by integrating all densities between two selected boundaries using ImageQuant version 3.3 software. Each resolved band was assigned to a particular bond within the DNA fragment by comparison of its position with those of sequencing standards generated by treatment of DNA with dimethylsulfate (guanine) and/or formic acid (guanine plus adenine), followed by piperidine-induced cleavage at the modified bases in the DNA.

Molecular modeling. This study involved the use of the consensus dinucleotide intercalation geometries d(ApT) and d(GpC), which were initially derived from the crystal structures of adriamycin (Frederick et al., 1990) and proflavine (Neidle et al., 1988) intercalation complexes, respectively. The d(ApT) and d(GpC) intercalation sites were located at the center of a decanucleotide duplex having the sequence d(5'-ATATA-3')₂ and d(5'-GCGCG-3')₂. Decamers in the B-form were built using the nucleic acid builder function of the HyperChem software program (release 4.5; Hypercube Inc., Gainesville, FL). Both decanucleotides were minimized using the Amber all-atom force field (Weiner et al., 1986) of the Macromodel 5.0 molecular modeling package (Mohamadi et al., 1990), until the root mean square of the conjugate gradient was <0.05 kcal/mol/Å. The dielectric constant was assumed to be distance independent, with a magnitude of 4.

	Results

Top Summary Introduction Procedures Results Discussion References

Thermodynamics of DNA Binding

Spectroscopic studies. Upon addition of DNA (polynucleotide) solutions, all test compounds exhibited dramatic changes in their spectroscopic properties. In particular, the fluorescence of 9-IHA and aza-9-IHA was progressively quenched (Fig. 2B), and the absorption spectrum of 1-IHA underwent a bathochromic and hypochromic shift, generating an isosbestic point (Fig. 2A). The case of 1,4-IHA deserves additional comment. The drug tends to aggregate to a remarkable extent (see Experimental Procedures) and was very effective in inducing DNA condensation and precipitation even at the lowest binding ratios examined. Therefore, a quantitative evaluation of the DNA binding constants for this compound was not possible, but it can be safely stated that this drug exhibits a very high affinity for the nucleic acid under the experimental conditions used.

View larger version (26K): [in this window] [in a new window]   Fig. 2.   Examples of spectrophotometric (A) and fluorometric (B) titration experiments for the binding of test drugs to double-stranded DNA from calf thymus, in ETN buffer (0.1 M, pH 7.0). Curve 1, free drug; curve 2, totally bound drug. A, 1-IHA (15.3 µM); B, aza-9-IHA (1.6 µM).

1

View larger version (14K): [in this window] [in a new window]   Fig. 3.   Dependence of the binding constant Ki for the test congeners on ionic strength (IS).

Mode of interaction. Compounds 9-IHA, aza-9-IHA, and 1-IHA were examined for their abilities to affect the electrophoretic mobility of supercoiled plasmid DNA. As expected from the close structural similarities with the parent drugs, which are well known intercalators, all of them caused a progressive decrease of plasmid mobility in the gel, followed by a leveling off and finally an increase with increasing drug concentration (data not shown). This behavior is clearly indicative of an intercalating mechanism of binding for the test drugs.

Binding to single-stranded DNA. To evaluate the effect of DNA structure on the binding efficiency of the test drugs, we also determined the affinity constants with single-stranded nucleic acid. The quantitative data are reported in Table 3. Interestingly, bisantrene congeners retained a remarkable capacity for binding to the single-stranded structure. In comparison with the K_i values found for double-stranded DNA, 9-IHA showed a 2-fold increase, 1-IHA maintained the same values, and aza-9-IHA exhibited a modest reduction. On the other hand, bisantrene, although undergoing a 6-fold reduction in K_i, still exhibited an affinity constant of the order of 10⁶ M¹.

Melting experiments. Effective intercalating drugs are well known for their ability to shift the T_m of double-stranded DNA to higher values (Foye et al., 1986). Although increases in T_m cannot be immediately correlated with binding stability, high T_m values are indicative of tight interactions. The results of melting experiments performed with the test drugs using a number of natural and synthetic DNAs are reported in Table 4. No useful data could be obtained with poly(dG-dC) because of the very high T_m (97°) exhibited by the polynucleotide. Values of T_m always increased with increasing drug concentration and were as high as 44° in the case of 1,4-IHA, whereas they never exceeded 21° for the monosubstituted congeners. Considering that T_m values for bisantrene did not exceed 42°, this confirms the outstanding DNA-binding efficiency of the 1,4-disubstituted analogue, in agreement with the results presented above.

DNA Sequence Selectivity

Spectroscopic measurements. Spectrophotometric and fluorometric methods were used to evaluate the affinities of the test drugs for DNAs with different base compositions. The results are reported in Table 5. 9-IHA and aza-9-IHA demonstrated a clear preference for A-T steps. In contrast, 1-IHA was able to bind to AT- or GC-rich DNAs almost equally well, losing the selectivity shown by the 9-substituted congeners. As expected from previous reports on intercalating ligands, binding of the test compounds to poly(purine)·poly(pyrimidine) sequences was substantially disfavored (data not shown).

DNase I footprinting experiments. Footprinting studies were performed to gain further insight into the sequence preferences exhibited by the tested anthracene derivatives. A representative example is shown in Fig. 4. The gel and the corresponding differential cleavage plots reveal that the nuclease activity was only slightly affected in the presence of 1-IHA and 1,4-IHA (20 µM each). Bisantrene exhibited a more pronounced effect (it was used at 5 µM), whereas both 9-IHA and aza-9-IHA showed absolutely no effects even at concentrations as high as 100 µM (gel not shown). Quantitative analysis of the footprinting gels (the one shown in Fig. 4 and others with various DNA fragments) provided two interesting pieces of information. First, in the presence of the 1-substituted derivatives, but not with the 9-substituted congeners, the DNase I cleavage was enhanced at sequences containing contiguous A-T pairs. Both bisantrene and 1,4-IHA frequently promoted DNase I cleavage at oligopurine·oligopyrimidine tracts; conversely, they slightly reduced the cleavage activity at alternating purine-pyrimidine sequences. The preference for mixed purine-pyrimidine sequences is also apparent from the T_m measurements. Indeed, much higher T_m values were obtained with poly (dA-dC)·poly(dG-dT) than with poly(dA-dG)·poly(dC-dT) in the presence of 1,4-IHA (Table 4). Second, the sequences where cleavage by the nuclease was slightly reduced were neither GC- nor AT-rich. In most cases, they contained at least one 5'-GT-AC or 5'-TG-CA dinucleotide step. The footprinting data reinforce the conclusion, drawn from the spectroscopic measurements, that 1-IHA can bind to both AT- and GC-containing sequences. The lack of effective interference of the 9-substituted drugs with the cleavage of DNA by the enzyme was anticipated. It might be attributable to the relatively low affinity of the drugs and/or to the fast on and off kinetics. It is well known that amsacrine and related 9-aminoacridine analogues do not interfere with the DNase I cleavage reaction (Wakelin et al., 1990; Bailly et al., 1992; Crenshaw et al., 1995). Although DNase footprinting and topoisomerase stimulation experiments were performed with different DNA fragments, a comparison between them is significant because it is based on a statistical analysis.

View larger version (79K): [in this window] [in a new window]   Fig. 4.   DNase I footprinting with the 5'-end-labeled 265-mer PvuII-EcoRI restriction fragment cut from the plasmid pBS, in the presence of 5 µM bisantrene (Bst) or its derivatives (20 µM each). Control tracks (Cont) contained no drug. Guanine-specific sequence markers (G) obtained by treatment of the DNA with dimethylsulfate, followed by piperidine cleavage, were also run. Numbers at the top of the gels refer to the numbering scheme of the fragment. The differential cleavage plots compare the susceptibility of the fragment to DNase I cutting in the presence of bisantrene, 1-IHA, and 1,4-IHA. Negative values correspond to a ligand-protected site and positive values represent enhanced cleavage. The vertical scale is in units of ln (fa)  ln (fc), where fa is the fractional cleavage at any bond in the presence of the drug and fc is the fractional cleavage of the same bond in the control. Only the region of the restriction fragment analyzed by densitometry is shown. , Sites of drug-induced enhanced reactivity toward DNase I.

Molecular Modeling Studies

Docking experiments were performed to calculate the energy changes and optimal geometries corresponding to the intercalative binding of 9-IHA and 1-IHA to A-T and G-C steps in a double-helical structure. In agreement with the binding experiments reported above, only alternating purine-pyrimidine steps were considered. In all cases, intercalation from the minor groove was preferred by 6-8 kcal/mol. The energy balance for the most favorable binding of 9-IHA and 1-IHA to A-T and G-C base pairs is presented in Table 6.

View this table: [in this window] [in a new window]   TABLE 6 Calculated and experimental binding energy values for the intercalation process (L + oligomer  L-oligomer)

Interestingly, an effective sequence discrimination appears to operate for 9-IHA; an energy difference of 22 kcal/mol was observed between intercalation into the more favored A-T step and that into the less preferred G-C step. In contrast, 1-IHA showed substantially reduced selectivity, with G-C steps being slightly favored (<3 kcal/mol) over A-T steps. It should be emphasized that the calculated interaction energy values were approximated; they do not take into account entropy changes and solvation effects, which might provide relevant contributions to the binding process. It is, however, encouraging that the rank of calculated intercalation energies corresponds satisfactorily to the rank of experimentally determined free energies of binding (Fig. 5).

View larger version (11K): [in this window] [in a new window]   Fig. 5.   Relationship between the calculated energy of intercalation (E) and the experimental free energy of binding (G) corresponding to the interaction of 9-IHA (circles) and 1-IHA (squares) with A-T (closed symbols) and G-C (open symbols) polynucleotides.

The structures corresponding to the most stable intercalation geometries of 9-IHA and 1-IHA are presented in Figs. 6 and 7. In general, the side-chain groups protrude from the central part of the minor groove, producing effective electrostatic contacts with the negative charge density of the DNA. To optimize side-chain locations, the planar portions of the two drugs are considerably displaced in the intercalation pocket, with respect to each other. More efficient stacking interactions with A-T base pairs appear to be primarily responsible for the base preference exhibited by 9-IHA, compared with its 1-substituted isomer.

View larger version (28K): [in this window] [in a new window]   Fig. 6.   Best intercalation geometries for intercalation of 9-IHA and 1-IHA into A-T base pairs, as obtained from the docking experiments. Left, view from the minor groove; right, view along the DNA axis. Hydrogen atoms are not shown. Atoms of the drug structure are in black.

View larger version (29K): [in this window] [in a new window]   Fig. 7.   Best intercalation geometries for intercalation of 9-IHA and 1-IHA into G-C base pairs, as obtained from the docking experiments. Left, view from the minor groove; right, view along the DNA axis. Hydrogen atoms are not shown. Atoms of the drug structure are in black.

	Discussion

Top Summary Introduction Procedures Results Discussion References

The results obtained with the bisantrene congeners and presented here show unequivocally that the relative locations of the side chain and the planar ring system dramatically affect the affinity and specificity of these drugs for DNA. Loss of one of the two side chains of bisantrene clearly reduces binding to a major extent, which demonstrates the effective participation of both charged side-chain groups in the binding process. This is confirmed by the calculated value of 2 for the m' parameter (Table 2). The contribution of the dihydroimidazolyl lateral groups is not only electrostatic but must include noncharged contributions, as demonstrated by the G° values extrapolated from the data measured at different ionic strengths. In fact, the G° value for bisantrene is approximately 1 kcal/mol more negative than the G° values for the 9-substituted analogues. The effects of bioisosteric substitution (C10 to N) essentially do not affect either K_i or G°. This indicates very similar arrangements of the planar moieties of 9-IHA and aza-9-IHA when the drugs are bound to DNA. On the other hand, a shift of the side chain from position 9 to position 1 reduces the interaction free energy (in absolute values) by an additional 1 kcal/mol (G°  5 kcal/mol). Hence, less favorable contacts occur between the drug and DNA when the dihydroimidazolyl group is located on the lateral rings of anthracene. The melting experiment results are in agreement with these findings. Unfortunately, because of the self-aggregation and DNA-precipitating ability of 1,4-IHA, it was not possible to make a sound comparison between this compound and the isomer bisantrene.

Not only the binding thermodynamics are affected by a shift of the side chain to a different position; the DS is also dramatically changed with movement of the lateral group from position 9 to position 1. As shown in Table 5, the strong preference of the 9-substituted compounds for A-T base pairs vanishes with the 1-substituted drug. This is confirmed by the docking experiment results, which indicate a more stable intercalation into A-T pairs for 9-IHA and an essential loss of discrimination for A-T versus G-C steps by 1-IHA. Further elucidation of sequence specificity, which was attempted with footprinting studies, was only marginally successful because of the poor footprinting ability of the monosubstituted derivatives. Nevertheless, these experiments confirm a wider sequence acceptance exhibited by 1-IHA.

With elucidation of the effects of the chain shift, in terms of DS, it is interesting to observe that DS and TS coincide for the test drugs. Indeed, whereas the 9-substituted compounds confirm the A+1 preference of m-AMSA and bisantrene, the 1-substituted congener does not show any specificity in stimulating DNA cleavage (Capranico et al., 1998) (Table 1).

Recently, we have been interested in identification of the structural drug determinants for TS (Capranico et al., 1994; Palumbo et al., 1994). A key step in the enzyme-mediated mechanism of action is the formation of a ternary topoisomerase II-DNA-drug complex that establishes proper contacts between the drug pharmacophores and the enzyme and nucleic acid counterparts (Capranico et al., 1997). Although these contacts occur in a particular structure in which the original conformations of the macromolecular species are mutually distorted, some of the recognition features involved in binary drug-enzyme and drug-nucleic acid interactions should still be operative in the ternary system. In particular, for intercalating agents, the characteristics that allow preferential stacking onto given base pairs should also influence the locations at which drug-mediated cleavage stimulation occurs.

In the case presented here, it appears that, notwithstanding the changes occurring in the cleavable complex, recognition of DNA by the drug is preserved as it was in the absence of the enzyme. This fact is confirmed by the parallel changes in DS and TS that occur when the drug structure is isomerized.

The finding that TS corresponds to DS for the test drugs is even more surprising when we consider that the +1 specificity involves the base directly linked to the catalytic tyrosine (Capranico et al., 1997). Because of the special amino acid-nucleotide phosphodiester bond formed and the distortion caused by the new protein-linked arrangement, the base at +1 should be more profoundly affected (both electronically and structurally) than the base at position 1, with the latter remaining essentially paired in a double-helical conformation, as indicated by base-mismatch experiments (Bigioni et al., 1996). Indeed, considering that the central portion of the cleavage region (positions +1 to +4) must be melted to allow DNA strand-passing and that it is the religation step that is mainly inhibited by drugs such as m-AMSA (Robinson and Osheroff, 1991), it is possible that compounds having +1 specificity stabilize a catalytic intermediate in which the target base is unpaired. This is in line with the aforementioned base-mismatch experiments (Bigioni et al., 1996). In fact, whereas mismatches introduced at position 1, relative to the enzyme cleavage site, were shown to abolish DNA cleavage, those at positions +1 and +2 increased the level of DNA breakage instead. In the single-stranded region of DNA in the cleavable complex, the creation of efficient  interactions between the planar systems of the preferred base and of the drug should be crucial to avoid dissociation from the cleavable complex and loss of cleavage stimulation. Hence, efficient recognition of the DNA by the drug system through stacking interactions at position +1 in the cleavable complex should correlate with the base preferences in the enzyme-free system. This was, indeed, our observation. Moreover, the remarkable affinities shown by bisantrene and its congeners for stacking interactions with single-stranded DNA (Table 3) are consistent with the proposed recognition mechanism. In agreement with this concept, only purine bases, which exhibit an extended stacking surface, were previously found to be preferred for cleavage stimulation at position +1, whereas any base can be specifically recognized by different drugs at position 1 (Capranico et al., 1997). Our model might also be effective for topoisomerase I poisons, as suggested by very recent work on the crystal structure of human topoisomerase I in covalent and noncovalent complexes with DNA (Redimbo et al., 1998). In fact, based on chemical, structural, and biochemical evidence, the binding mode proposed for the anticancer drug camptothecin (exhibiting G+1 specificity) at the topoisomerase I-DNA complex suggests an effective stacking interaction of the guanine residue with the planar ring system of the drug. This is confirmed by the fact that TS corresponds to DS for indolocarbazole derivatives having +1 specificity (Bailly et al., 1997, 1998).

From the aforementioned results, some useful insight can be gained into the molecular basis of the TS. It appears that, for +1-specific agents, the drug-DNA contacts are primarily responsible for base selectivity; the interaction between the protein and the poison does not play a major role in stabilizing a specific ternary complex intermediate. As a result, it should be possible to predict the likely location of +1 cleavage stimulation along the DNA chain from binary complex studies. As a corollary to the model presented here, a +1 cleavage-stimulating agent should be characterized by high affinity for single-stranded DNA. These findings have useful implications for the design of novel topoisomerase poisons. Studies on the DNA-binding properties of other topoisomerase II poisons exhibiting +1 specificity, such as saintopin (Leteurtre et al., 1994) and NSC665517 [2-methyl-3-[2-(D-xylopyranosylamino)phenyl]-6,8-dibromo-4(³H)-quinazolinone] (Gupta et al., 1995) (both guanine specific), are warranted to confirm the general applicability of our model and further elucidate the mode of base selection in drug-stimulated DNA cleavage.