Molecular determinants of site-specific inhibition of human DNA topoisomerase I by fagaronine and ethoxidine. Relation to DNA binding.

DNA topoisomerase (top) I inhibition activity of the natural alkaloid fagaronine (NSC157995) and its new synthetic derivative ethoxidine (12-ethoxy-benzo[c]phenanthridine) has been correlated with their molecular interactions and sequence specificity within the DNA complexes. Flow linear dichroism shows that ethoxidine exhibits the same inhibition of DNA relaxation as fagaronine at the 10-fold lower concentration. The patterns of DNA cleavage by top I show linear enhancement of CPT-dependent sites at the 0.016-50 microM concentrations of fagaronine, whereas ethoxidine suppress both top I-specific and CPT-dependent sites. Suppression of top I-mediated cleavage by ethoxidine is found to be specific for the sites, including strand cut between A and T. Fagaronine and ethoxidine are DNA major groove intercalators. Ethoxidine intercalates DNA in A-T sequences and its 12-ethoxy-moiety (absent in fagaronine) extends into the DNA minor groove. These findings may explain specificity of suppression by ethoxidine of the strong top I cleavage sites with the A(+1), T(-1) immediately adjacent to the strand cut. Fagaronine does not show any sequence specificity of DNA intercalation, but its highly electronegative oxygen of hydroxy group (absent in ethoxidine) is shown to be an acceptor of the hydrogen bond with the NH(2) group of G base of DNA. Ability of fagaronine to stabilize top I-mediated ternary complex is proposed to be determined by interaction of its hydroxy group with the guanine at position (+1) of the DNA cleavage site and of quaternary nitrogen interaction with top I. The model proposed provides a guidance for screening new top I-targeted drugs in terms of identification of molecular determinants responsible for their top I inhibition effects.

The benzo[c]phenanthridine alkaloid fagaronine (Fig. 1), isolated from the roots of Fagara zanthoxyloides Lam. (Rutaceae) (1), exhibits antitumor activity against P388 and L1210 murine leukemias in vivo and toward colon 26 (1,2). It has been shown to induce differentiation in murine erythroid Friend cells, human K562 erythroleukemia and promyelocytic HL60 cells (3)(4)(5). Fagaronine was proved to be a DNA intercalator (6), it inhibits DNA and RNA polymerase activities and protein synthesis (7,8). Fagaronine also inhibits reverse transcriptases from different sources (5, 9 -10) and was proposed to act through at least two different mechanisms: inhibition of nucleic acid synthesis due to interaction with DNA and inhibition of the elongation step of protein synthesis (7). Further studies revealed that fagaronine is able to stabilize top I 1 ternary cleavable complexes at low concentrations and to inhibit both top I and top II ternary complexes at higher concentrations (12,13). The most potent fagaronine derivative nitidine (Fig. 1), isolated from extract of a climbing shrub Zanthoxyulum nitidum (14), was observed to trap both, top I-and II-cleavable complexes. Nearly 100 naturally occurring alkaloids in this class have been isolated from plants, and many more have been synthesized, but they are generally not markedly better than nitidine and fagaronine (15) and do not exhibit any significant activity against solid tumors. It is worth noting that the only benzo[c]phenanthridine alkaloids found thus far to stabilize the top I-cleavable complexes are those that have previously been shown to have antitumor activity in experimental animal models (12). So, new structural analogues of benzophenanthridines, top I inhibitors with an enlarged spectrum of activity, are highly desirable.
In terms of structure-activity relationship, two main points can be emphasized: (i) all the compounds synthesized and studied so far carry the iminium charge on the benzo[c]phenanthridine ring (Fig. 1), which seems to be necessary for their biological action, and (ii) the reactivity of the iminium toward nucleophilic attack has been put forward to explain the antileukemia activity of these series (15). Recently, the iminium bond electrophilicity within the benzo[c]phenanthridines was shown to be a factor which requires consideration in ternary complex formation with reverse transcriptase (16). The other molecular determinants playing the key role in the benzo[c] phenanthridines anticancer or enzymes-inhibition activity are not known yet and need to be identified.
Recently, a new fagaronine derivative ethoxidine (Fig. 1), has been synthesized by one of us (17), and its activity against human immunodeficiency virus, type 1 reverse transcriptase has been described (16). Our recent Raman, surface-enhanced Raman scattering (SERS), and flow linear dichroism (FLD) comparative studies of ethoxidine and fagaronine DNA complexes showed that the new derivative is an intercalator with its DNA binding mode different from that of fagaronine (18). Our preliminary results demonstrate also 10-fold higher top I inhibition activity of ethoxidine, as well as its much lower IC 50 values in the human K562 and A549 cancer cell lines, as compared with fagaronine (19).
In this paper we present the results of the study with a wide range of biochemical and biophysical techniques aiming to: (i) compare mechanisms of top I inhibition by fagaronine and ethoxidine, (ii) correlate sequence specificity and molecular interactions of these compounds within the DNA complexes with their mechanisms of top I inhibition, and (iii) identify molecular determinants of the drug chromophores responsible for their DNA binding and top I poisoning or catalytic inhibition.

EXPERIMENTAL PROCEDURES
Materials-CT DNA and the double-stranded poly(dA-dT)⅐poly(dA-dT) and poly(dG-dC)⅐poly(dG-dC) polymers were purchased from Sigma. Their concentrations in the DNA base pairs were determined by using molar extinction coefficients of 13,200, 13,900, and 13,200 M Ϫ1 cm Ϫ1 , respectively (20). CPT was purchased from Sigma and fagaronine was supplied by National Cancer Institute (Bethesda, MD). The synthesis of ethoxidine was described previously (17). Fagaronine and ethoxidine were prepared as 1 mM stock solutions in methanol and were diluted by buffer to desired concentration. DNA and polymers were dissolved in PBS to 5 mg/ml stock solution. Drug-DNA complexes were prepared by mixing the drug stock solutions with the DNA solution in PBS.
Plasmid pGEM7Z(fϩ) and restriction endonucleases were purchased from Promega and Escherichia coli strain "Sure" was purchased from Stratagene. Bovine pancreatic DNase I and Klenow fragment of E. coli DNA polymerase I were purchased from Sigma and Roche Molecular Biochemicals, respectively.
Recombinant 68-kDa human DNA top I was purified to homogeneity from insect cells using a two-step procedure as described (21,22). Specific activity of top I used in our assays was found to be 1.8 ϫ 10 6 units/mg, where one unit of activity is an amount of enzyme yielding 100% of relaxation of 300 ng of supercoiled pGEM7Z(fϩ) plasmid DNA in 30 min at 37°C.
DNA Plasmid Constructs-Preparation of the top I DNA substrates in the form of plasmid constructs containing top I-specific and CPT-dependent cleavage sites was described (23). The constructs were purified from the cells and analyzed by DNA sequencing method of Sanger (24).
For 3Ј-end labeling, plasmid DNA constructs were cleaved with Hin-dIII and ApaI and labeled with [␣-33 P]dATP in the presence of the Klenow fragment of DNA polymerase I according to (24). The 3Ј-labeled DNA fragments were purified by electrophoresis on a nondenaturing 5% (w/v) polyacrylamide gel and isolated by electroelution followed by ethanol precipitation.
Topoisomerase Cleavage Assays-Cleavage was carried out by incubating 50 units of top I with a 5 l of the solution of the radioactively labeled DNA fragment (3,000 -10,000 cpm) in 10 mM Tris-HCl (pH 7.8), 5% glycerol, 0.5 mM EDTA, 0.3 mM 2-mercaptoethanol (final volume 20 l). For the analysis of DNA cleavage by top I in the presence of the drugs, reaction mixtures were incubated at 25°C for 20 min, then SDS and proteinase K were adjusted to 0.5% (w/v) and 1 mg/ml, respectively. After incubation for a further 45 min at 37°C, DNA was purified by phenol extraction, precipitated with ethanol, washed with 70% ethanol, and dried.
Gel Electrophoresis-The samples of DNA were dissolved in 1.5 l of the formamide-dye mixture (90% formamide containing 15 mM EDTA (pH 8)), heated 1 min at 90°C, and applied to 8% denaturing polyacrylamide gel. Electrophoresis was proceeded for 65 min at 65 watts (2,500 V). The gels were fixed with 10% acetic acid and dried on glass pretreated with Bind-silane (Amersham Pharmacia Biotech). Cleavage products were identified by comparison with "A ϩ G" Maxam-Gilbert sequencing ladder.
UV-visible and Circular Dichroism Spectroscopy-Uv-visible spectra were recorded with a JASCO V-530 UV-visible scanning spectrophotometer. CD spectra were recorded in the region 200 -500 nm with a Jobin Yvon Mark III dichrograph. CD and UV-visible measurements were performed using quartz cells of 1 and 0.5 cm, respectively.
Flow Linear Dichroism Spectroscopy-FLD spectra were recorded in the region 220 -450 nm with a Jobin Yvon Mark III dichrograph equipped with a self-made achromatic /4 device. The self-made flow cell described in Ref. 21 Linear dichroism (⌬A) is the difference between the absorbance for light polarized parallel (A ʈ ) and perpendicular (A Ќ ) to the flow. The reduced linear dichroism (LD r ) is defined by LD r ϭ ⌬A/A ϭ (A ʈ Ϫ A Ќ )/A, where A is the isotropic absorbance of the sample.
Measurements of the linear dichroism in the region of absorption of the drug was used to determine the drugs chromophore orientation relative to the DNA axis. The angle ␤ between the transition moment of the dye chromophore and the orientation axis of the DNA was calculated from the measured ratios of LD r for the DNA bases and for the drugs, where ␣ ϭ 86°is the angle between transition moment of the bases and the orientation axis of the DNA molecule (25). Surface-enhanced Raman Scattering Spectroscopy-SERS spectra were recorded with a spectrometer Coderg, model PHO, with double monochromator in the frequency range 300 -1800 cm Ϫ1 . Ar ϩ -ion laser (Coherent Radiation, model Innova 2020) operating at 457.9 nm (for fagaronine) or at 488 nm (for ethoxidine) wavelengths was used for spectra excitation. SERS spectra were recorded for 1 scan with a 1 s time constant. Silver hydrosol was prepared according to the protocol published before (26). ( Fig. 2) was used as a top I substrate. To compare the abilities of fagaronine and ethoxidine to modulate the top I-mediated DNA cleavage, we have analyzed the DNA cleavage pattern of the ␣-33 P-labeled DNA substrate by enzyme with and without drugs. The cleavable complex formation was estimated by analysis of distribution of the single strand DNA breaks after the SDS and proteinase K treatment. The specific DNA cleavages by top I without any drugs and with 10 M CPT were used as a control.

Modulation of Topoisomerase I-mediated DNA Cleavage by
In the absence of drugs, top I-specific cleavage sites (designated as A, B, C, D, E, G, H) with different cleavage intensities have been revealed (Fig. 2). Addition of 10 M CPT in the reaction mixture induces typical modulation of the pattern of DNA cleavage by top I. The intensities of cleavage in the sites C and H were found to be unchanged, the sites A, B, D, E, and G were enhanced, and a new site (F) was induced. As expected, all of the enhanced sites have a T base at the cleavage position (Ϫ1), and most of them have G at the cleavage position (ϩ1) in accordance with the data published previously (27).
Fagaronine induces the same dose-dependent modulation of the top I-mediated DNA cleavage as camptothecin with canonical G(ϩ1) and T(Ϫ1) immediately adjacent to the strand cut (Fig. 2). These results confirm the data published previously (12, 13) and were used as a reference to comparative study of the pattern of DNA cleavage modulated by ethoxidine. The pattern of top I-mediated DNA cleavage in the presence of ethoxidine was found to be completely different from that of fagaronine. The presence of ethoxidine in the reaction mixture at the same concentrations as fagaronine induces strong suppression of top I-specific and CPT-dependent cleavage sites (Fig. 2). All these sites were strongly reduced by 2 M ethoxidine, and some of these sites disappeared at 10 M, while others at 50 M concentration of the drug.
Suppression of DNA cleavage sites by ethoxidine was found to be sequence-specific. To establish dependence of ethoxidineinduced suppression of DNA cleavage by top I on the local base sequence immediately adjacent to the cleavage site, we have The FLD technique enables to determine the relative orientation of the plane of the drug chromophore to the plane of DNA bases: the linear dichroism of intercalators is known to be negative, whereas the minor groove binders induce the positive signal (28). FLD signals from fagaronine and ethoxidine bound to DNA at the saturation ratios were found to be negative in the regions of all electronic transitions (Fig. 4B). The angles between the short axis electronic transition of the fagaronine and ethoxidine chromophores and the axis of the DNA molecule calculated with Equation 1 are ϳ73°and 79°, respectively. So, Deprotonation of fagaronine OH group results in disappearance of bands corresponding to 1L b electronic transition of the chromophore and an increase with a bathochromic shift of the bands of the 1L a transition (Fig. 4A). The pH dependence of the fagaronine's UV-visible spectrum is determined by its OH group with pK ϭ 8.0 (Fig. 4, inset). Therefore, at the physiological pH the solution contains both protonated and deprotonated forms of fagaronine.
UV-visible spectra of ethoxidine in Me 2 SO, ethanol, methanol, and PBS are found to be practically identical (spectra not shown). On the other hand, the profile and relative intensities of the fagaronine spectra are modified upon Me 2 SO-ethanolmethanol-PBS transitions (Fig. 4C). The molecules of polar solvents are presumed to form the hydrogen bonds with the oxygen of OH group of fagaronine, and this effect leads to a decrease of the influence of the strong negative charge of the oxygen on the conjugated chromophore system. So, the distribution of electronic density in fagaronine chromophore in polar solutions becomes more similar to this in ethoxidine, and the UV-visible spectra of fagaronine and ethoxidine in PBS are found to be closer than their spectra in Me 2 SO. The trace 6 in Fig. 5A shows the differential spectra of fagaronine in PBS minus Me 2 SO. The profile of this difference spectrum may be used as a reference for the effects induced in the case of formation of hydrogen bond between the oxygen of fagaronine's OH group and a less electronegative moiety.
UV-visible Spectra of Drugs in the Complexes with DNA and with Alternating Double-stranded Polynucleotides-Addition of CT DNA to fagaronine or ethoxidine solution results in a hypsochromic shift in their absorption spectra and an increase in the band at ϳ400 nm accompanied by relative changes in the bands in the 270 -340 nm region (Fig. 5). Characteristic difference spectra of fagaronine and ethoxidine within the DNA complexes are shown in Fig. 5, trace 3. These pronounced spectral modifications induced by DNA binding were used to evaluate the binding constants. For that purpose, DNA was titrated by the drugs (not all the spectral curves are presented in the Fig. 5 for clarity), and the drug/DNA binding constants were determined (Table I).
We have analyzed the spectra and determined the binding constants of the drugs within poly(dA-dT)⅐poly(dA-dT) and poly(dG-dC)⅐poly(dG-dC) complexes. In terms of the values of its binding constants, fagaronine does not show base preference of intercalation (Table I). The binding constants on the level of 10 6 M Ϫ1 were found for its interaction with all used sequences as well as with CT DNA. Contrary, ethoxidine shows strong preference of intercalation within the AT sequences. Specificity of ethoxidine binding with the poly(dG-dC)⅐poly(dG-dC) is 25fold lower than with poly(dA-dT)⅐poly(dA-dT) ( Table I).
The differences of molecular interactions of fagaronine and ethoxidine within the specific DNA sequences may be revealed from analysis of the profiles of the difference UV-visible spectra of drugs within the complexes (Fig. 5). The spectral profiles of the difference spectra of fagaronine complexes with poly(dG-dC)⅐poly(dG-dC) and with CT DNA are very similar, and they are drastically different from those of fagaronine-poly(dA-dT)⅐poly(dA-dT) complexes (compare trace 4 with the traces 3 and 5, Fig. 5A). Otherwise, we did not observe any differences in the spectral profiles of difference spectra of ethoxidine complexed with CT DNA and with a number of alternating polynucleotides (Fig. 5B, traces 3 -5).
Trace 6 of Fig. 5A shows the difference spectrum of fagaronine solutions in PBS and in Me 2 SO. At the same time, the spectra of ethoxidine in Me 2 SO and PBS solutions were found to be identical (no difference spectrum). Moreover, the profile of the fagaronine difference (PBS minus Me 2 SO) spectrum in the 1L b spectral region corresponds exactly to effect of fagaronine binding with poly(dG-dC)⅐poly(dG-dC) and with CT DNA (compare trace 6 with traces 3 and 5 in Fig. 5A). So, participation of the oxygen of the OH group of fagaronine in the hydrogen bond with the molecules of the polar solvent induces the same spectral effect as its molecular interactions upon DNA and poly(dG-dC)⅐poly(dG-dC) binding. It is reasonable to suggest that the oxygen of the fagaronine OH group is an acceptor of proton coming from the guanine NH 2 group (see "Discussion").
Induced Circular Dichroism Spectra of Drug-DNA Complexes-Since the fagaronine and ethoxidine are the planar and achiral chromophores, only those molecules complexed to the asymmetric DNA matrices are able to display induced CD (Fig.  6). These induced CD signals, which are indicative of interactions between the drug and host DNA duplex, can be used to detect and to monitor any CD-active DNA binding mode(s). The CD spectra of fagaronine and ethoxidine contain all bands corresponding to 1L b (400 -420 nm) and 1L a (270 -340 nm) electronic transitions. As is known, the 1L a (oriented along chromophore's long axis) and 1L b electronic transitions lie in the plane of the chromophore's aromatic system and are normal to each other (29,30). The fact that all the bands of the both electronic transitions become optically active indicate that both electric dipole moments of the transition participate in DNA interaction which may be possible only in the case of drug-DNA intercalation.
Titration of the DNA with the varied amounts of ligands may give a hint about the character of ligand-DNA interactions, the number of the binding centers, and relative orientation of the plane of the DNA bases with respect to the drug's chromophore. An increase of the fagaronine and ethoxidine content does not induce modification of the profile of their CD spectra clearly indicating the only one binding center for each molecule (Fig. 6).
Both ethoxidine and fagaronine, when complexed with the DNA, reveal similar induced CD effects: an increase of the 1L b group of bands with a concomitant decrease of the bands corresponding to the 1L a electronic transition (Fig. 6). Nevertheless, one important difference is obvious: CD signal for the all 1L a bands of ethoxidine is positive, whereas the 304 nm band of 1L a electronic transition of fagaronine is negative. This fact implies the difference in geometry of fagaronine and ethoxidine DNA complexes.
Comparison of the symmetry groups of numerous intercalators as well as the results of molecular modeling (16, 29 -31) show that the fagaronine long axis is located between the 2-OH-and 8-methoxy group with some shift in the side of OH-substituent (Fig. 1). This is the direction of 1L a electronic transition. 1L b electronic is perpendicular to 1L a and is directed along the short axis of the chromophore. Orientation of the ethoxidine's long axis should be different due to equivalence of its 2-and 3-methoxy-substituents, which are less electronegative than the 2-hydroxy group of fagaronine. The theory of nondegenerate and degenerate coupled oscillator CD and its practical applications for intercalators (31) shows that the sign of cosinus of angle between the orientation of the chromophore long axis and direction of the base pair electronic transition is determined by polarization of electronic transition of interca-  lator and its orientation within the "pocket" between the DNA bp. So, if the long axis of the polarized electronic transition of intercalator is parallel to the long axis of the electronic transition of the bp (the angle is close to "0"), the signal of induced CD should be negative. If the long axis is issuing from the DNA groove and the angle is not equal to 0, the positive CD signal will be induced.
The fact that the induced CD signals for all 1L a bands of ethoxidine-DNA complex are positive indicates that the part of the long axis of ethoxidine chromophore penetrates the DNA double helix and points out from the DNA groove. The chromophore plane is not totally parallel to the plane of the DNA base pairs. This fact determines an additional possibility of excitonic interaction with the DNA bases so that in the induced CD spectra of ethoxidine, a split band with the positive Cotton effect at 289 nm and the negative at 263 nm appears.
Finally, geometry of ethoxidine intercalation presumes protrusion of its long axis within the DNA minor groove and slight rotation of the plane of chromophore relative to the plane of the DNA bases due to the steric effect of its spacious 12-ethoxy group disposed within the DNA minor groove.
Induced CD spectra of fagaronine-DNA complexes are typical for the classical DNA major groove intercalator (Fig. 6A). The band at 304 nm of its 1L a electronic transition is negative, clearly indicating typical DNA intercalation with the parallel orientation of the chromophore relative to the plane of the DNA bases.
Induced Circular Dichroism Spectra of Drugs in the Complexes with Alternating Polynucleotides-The UV-visible data on specific molecular interactions of fagaronine within the (GC) sequences are clearly confirmed by CD spectroscopy. The in-duced CD spectral profiles of fagaronine complexes with the poly(dA-dT)⅐poly(dA-dT) or with the poly(dG-dC)⅐poly(dG-dC) polymers were found to be completely different (data not shown). The spectral profiles of induced CD spectra imply the difference of molecular interactions of fagaronine's chromophore within these complexes. Moreover, the spectral profile of fagaronine complex with the poly(dG-dC)⅐poly(dG-dC) appears very similar as in case of CT DNA, whereas its complexation with the poly(dA-dT)⅐poly(dA-dT) induces completely different spectral features. On the other hand, the profiles of CD spectra induced by ethoxidine were measured to be the same for its complexes with the DNA, poly(dA-dT)⅐poly(dA-dT) or poly(dG-dC)⅐poly(dG-dC). These effects were found to be very similar to those for UV-visible analysis described above.
SERS Spectra of Fagaronine and Ethoxidine DNA Complexes-The SERS spectra of fagaronine and ethoxidine in solution have been analyzed by us, the assignments of the spectral bands have been made, and the bands sensitive to environment of quaternary nitrogen, O-CH 3 , or OH group of fagaronine have been identified (18). Fig. 7 demonstrates the most informative regions of fagaronine and ethoxidine SERS spectra. The spectra of free drugs were compared with those of their DNA complexes at a 1/200 bp ratio, where all the drug molecules are ensured to be DNAbound, and no contribution of the free drugs is present.
As was shown in Ref. 18, the bands in the region 1360 -1400 cm Ϫ1 are sensitive to modification of the N ϩ environment. So, the spectral modifications of ethoxidine spectrum within this region upon DNA binding indicate possible involvement of its quaternary nitrogen in the drug-DNA interaction. Moreover, the band at 1113 cm Ϫ1 (assigned to the (C-O) vibration) indicates that the exterior O-CH 3 moiety is involved in interaction with the DNA.
The SERS spectral changes observed for fagaronine-DNA complexes (Fig. 7) are more significant than in case of ethoxidine. If modifications of the quaternary nitrogen are less pronounced than that in ethoxidine, the bands attributed to vibrations involving OH group motions will be strongly modified. So, the band at 1273 cm Ϫ1 corresponds to (C-O) vibration, and it was shown to be sensitive to the formation of hydrogen bond with participation of OH groups of various chromophores (18). Addition of DNA induces splitting of this band in two peaks ( Fig. 7). Appearance of two peaks in expected position C-O vibration in the spectrum of drug-DNA complex may be explained in terms of the two, protonated and deprotonated forms of fagaronine at physiological pH (Fig. 4, inset). Participation of these both forms of fagaronine's OH group in the hydrogen bonding with the less electronegative atom (e.g. NH 2 group of the guanine) should certainly induce the detected splitting of this vibration. Therefore, it is reasonable to suggest that the oxygen in the deprotonated and protonated OH group is an acceptor of proton coming from the guanine NH 2 group, and this conclusion supports UV-visible data of analysis of fagaronine-DNA complexes. DISCUSSION DNA-binding agents can inhibit top I activity by at least two different mechanisms: (i) stabilizing the top I-DNA-cleavable complex and (ii) suppressing DNA cleavage. Consequently, there are two main groups of top I inhibitors with different mechanisms of effect on the top I specific activity, top I suppressors and top I poisons (7). Both groups of drugs inhibit plasmid DNA relaxation by top I. The suppressors mechanism of activity is determined by the possibility to prevent top I-DNA recognition and binding of the enzyme to its DNA substrate, whereas the poisons stabilize DNA-top I ternary cleavable complex.
Recently synthesized (17) new fagaronine derivative ethoxidine ( Fig. 1) is found to be ϳ10-fold more potent inhibitor of top I DNA relaxation activity as fagaronine (Fig. 3A). Our preliminary data show also 3-8-fold lower IC 50 values of ethoxidine in the human K562 and A549 cancer cell lines, compared with fagaronine (19). In this work we applied a variety of biochemical and biophysical techniques to the comparative study of mechanisms of top I inhibition by fagaronine and ethoxidine and identification of molecular determinants of the drugs responsible for their DNA binding and top I poisoning or catalytic inhibition.
It is known that fagaronine belongs to the group of top I poisons (12,13). In the top I cleavage assays, we have employed the model DNA substrate constructs, containing randomly distributed top I-specific and CPT-dependent cleavage sites (Fig.  2). Fagaronine showed the typical CPT pattern of modulation of top I-mediated DNA cleavage, all CPT-dependent sites of the DNA construct (involving G(ϩ1) and T(Ϫ1) bases immediately adjacent the strand cut) were equivalently enhanced, and no detectable differences of cleavage intensities between the individual sites were found. The mechanism of modulation of top I-mediated DNA cleavage by ethoxidine was found to be completely different from that of fagaronine. The presence of ethoxidine at the same concentrations as fagaronine induced strong suppression of both top I-specific and CPT-dependent cleavage sites in a concentration-dependent manner (Fig. 2). Site-by-site densitometric analysis of intensities of DNA cleavage as a function of ethoxidine concentration demonstrate the sequence specificity of the effect: the sites, including A(ϩ1) and T(Ϫ1) bases immediately adjacent to the strand cut (within the top I-specific sites), were found to be suppressed by ethoxidine much more effectively than the others. The results of these assays show that the 12-ethoxy substitution of fagaronine chromophore changes completely the mechanism of top I inhibition by the drug from top I poisoning by fagaronine to suppression of top I-DNA recognition by ethoxidine.
Finally, the biochemical data emphasize that ethoxidine is not able to trap top I-cleavage complex, but suppresses the DNA cleavage in the sequence-specific manner. Comparative structural analysis of fagaronine and ethoxidine DNA complexes was further employed to identify molecular determinants of DNA binding and top I poisoning by the drugs.
Two related structural factors are thought to account for the expected biological effects of intercalated drugs. One is geometry of the intercalation complex, allowing an orientation of the drug propitious to interactions with proteins such as DNA topoisomerases, polymerases, or transcriptional factors. The second factor is the presumed sequence-dependent conformational perturbations in DNA induced by the drug, which may lead to disturbance of the DNA-protein recognition (32) or, inversely, to stimulate the recognition of specific topologies by such an enzymes as DNA topoisomerase I (33).
Fagaronine and ethoxidine are shown to be the DNA major groove intercalators (15). FLD analysis of drug-DNA complexes (Fig. 3B) shows that fagaronine and ethoxidine intercalate into DNA with the stoichiometries 1/2 and 1/4 DNA bp, respectively. Lower stoichiometry of ethoxidine DNA intercalation may be explained by the sterical limits due to its spacious 12-alkoxy substituent excluding close approach of the neighboring DNA intercalated chromophores. Induced CD spectra of ethoxidine DNA complexes show that its long axis penetrates within the DNA minor groove. The plane of ethoxidine chromophore is slightly rotated relative to the plane of the DNA bases due to the steric effect of its spacious 12-ethoxy substituent being disposed within the DNA minor groove.
The CT DNA binding constants for ethoxidine and fagaronine are very similar (Table I), whereas they reveal different sequence specificity of DNA binding. Ethoxidine shows a 25fold higher binding constant with the poly(dA-dT)⅐poly(dA-dT) than with poly(dG-dC)⅐poly(dG-dC), whereas fagaronine binds both polymers with nearly the same binding constants (Table  I). AT specificity of ethoxidine intercalation may be explained by the lower rigidity of AT than GC base pares, so, its spacious 12-ethoxy substituent may penetrate more easily within d(AT) versus d(GC) DNA duplexes. UV-visible difference spectral analysis as well as induced CD spectra of drug-DNA complexes show specific molecular interactions of fagaronine within the (GC) sequences, whereas no any sequence specificity of ethoxidine interactions was detected. UV-visible, CD, and SERS spectroscopy show participation of the oxygen of fagaronine hydroxy group in hydrogen bonding with the less electronegative group (e.g. amino group of the guanine) upon DNA intercalation. SERS spectroscopy shows also involvement of ethoxidine quaternary nitrogen in the DNA interaction, whereas this effect is not so clear for fagaronine DNA complexes.
An attractive explanation for all these results is to propose the following structural model of drug-DNA complex (Fig. 8). While the ethoxidine chromophore interacts with the DNA, its spacious 12-ethoxy chain determines specificity of intercalation within the less stable AT sequences. It is more favorable for the spacious groups to unwind locally the DNA in AT sequences and to penetrate within the double helix than within the GC sequences. The long axis of ethoxidine's chromophore protrudes within the DNA outside the double helix so that the 12-ethoxy chain is disposed within the DNA minor groove and suppresses recognition of AT sequences by top I. The studies reported here provide experimental evidence that the nature of the side chain in the 12-position is of primary importance in eliciting the observed poisoning of top I. There is no doubt that the substituent at position 12 strongly affects the catalytic activity of the enzyme.
We found it reasonable to suggest three factors determining top I poisoning activity of fagaronine: (i) specific orientation of chromophore within the DNA complex, (ii) molecular interactions of fagaronine hydroxy group with the 2-amino group of DNA G base (probably in a (ϩ1) position adjacent to the strand cut), and (iii) the electrophilic iminium bond of benzo[c] phenantridine can be a subject of a nucleophilic attack from the top I and responsible for formation of a labile covalent bond between the top I and benzo[c]phenantridine.
The structure and geometry of the fagaronine-DNA complex may determine top I poisoning activity of the drug. Orientation of fagaronine chromophore deduced from the FLD and CD spectroscopic data presumes projection of its OH group from the surface of the minor groove in such a fashion as to promote interaction with the 2-amino group of guanine (the only hydrogen bond donor group exposed in the minor groove). The exact molecular mechanism of the fagaronine's hydroxy group interaction with G base for the top I poisoning effect is not clear, whereas its role in this process is evident. It is well known that steric and electronic features rather than the chemical nature of different substituents may be critical in determining the positional sequence specificity of the poisons-enhanced DNA cleavage by the topoisomerases (34). The most potent among the benzo[c]phenantridines top I poisons, fagaronine and nitidine ( Fig. 1), contain strong electronegative charges in the position 2 (OH group of fagaronine) or 2 and 3 (nitidine), whereas their analogues (O-methylfagaronine, fagaridine, isofagaridine, and chelerythrine) do not and fail to stabilize top I-DNA complex. It is possible that the interactions of fagaronine and nitidine with the DNA G base may induce local alterations of DNA structure at the specific nucleic acid sequence to which the enzyme is bound. This hypothesis will be tested at a later date.
Orientation of fagaronine chromophore within the DNA complex deduced from the spectroscopic data presumes that the quaternary nitrogen is partially exposed in the DNA minor groove and is acceptable to top I attack (Fig. 8). For all benzo-[c]phenanthridines so far synthesized and studied, the iminium charge on the ring (Fig. 1) seems to be necessary for their biological action. Moreover, the reactivity of the iminium toward nucleophilic attack was put forward to explain the antileukemia activity of these series (15) and to play a role for in vivo hydration and alkylation (35). This iminium bond may be attacked by carbon, nitrogen, and sulfur nucleophiles (35) or by the enzyme's amide bond or by cysteine, serine, threonine, or tyrosine amino acid residue (36). Reverse transcriptase inhibitory activity of fagaronine and O-methylfagaronine was shown to be associated with the ternary complex formation between the enzyme, alkaloid, and DNA (11). This finding was used in the model of reverse transcriptase inhibition by benzo[c-]phenanthridines proposed in Ref. 16.
We propose that exposed DNA minor groove position of fagaronine iminium bond (Fig. 8) enables the nucleophilic attack by some group of top I, formation of labile covalent bond between the enzyme and the quaternary nitrogen of the drug, and stabilization of the top I-fagaronine-DNA ternary cleavable complex. It is not clear if fagaronine is able to interact directly with the top I through its iminium bond. These experiments are in progress now.
The proposed model (Fig. 8) is based on extended experimental biochemical and biophysical data and may be used for screening of new top I-targeted benzo[c]phenanthridine structural analogues, top I inhibitors with an enlarged spectrum of activity.