A novel norindenoisoquinoline structure reveals a common interfacial inhibitor paradigm for ternary trapping of the topoisomerase I-DNA covalent complex

DNA topoisomerase I, an essential enzyme in metazoans (1, 2), has been identified as the target of camptothecin and its derivatives (3, 4). Topoisomerase I is ubiquitous and highly conserved in eukaryotes (5–7). The enzymatic activity of topoisomerase I, like that of other topoisomerases, breaks the DNA backbone reversibly by replacing a DNA phosphodiester bond with a bond between a phosphate at one end of the DNA break and a tyrosine residue of the topoisomerase I protein. In the case of topoisomerase I, the tyrosine links to the phosphate at the 3′-hydroxyl end of the DNA break (Fig. 1A-C). DNA relaxation takes place by rotation of the free end of the broken DNA around the intact DNA strand (Fig. 1B; ref. 8).

Camptothecin was discovered as the active principle of extracts from the Chinese tree Camptotheca acuminata by Wall and Wani (9). Camptothecin is a planar heterocycle consisting of five rings, including an important α-hydroxylactone E-ring (Fig. 1F). Camptothecin has several remarkable features. First, the natural alkaloid is the 20S-camptothecin enantiomer; this is the form that is active against topoisomerase I and experimental cancers, whereas the synthetic 20R enantiomer lacks both activities (9). Second, camptothecin traps the topoisomerase I-DNA cleavage complex, forming a potentially lethal DNA lesion. The anticancer activity thus is not due to inhibition of the topoisomerase I enzyme activity per se. This distinction was shown in topoisomerase I–deficient yeast strains, which completely lack the camptothecin sensitivity of wild-type strains (10–12). Third, camptothecin does not bind significantly to either DNA or purified topoisomerase I: both must be present together in the form of cleavage complexes (9). These characteristics led to the hypothesis that camptothecin forms a ternary complex with topoisomerase I and its DNA substrate by binding to a stereospecific site in the topoisomerase I-DNA cleavage complex (Fig. 1C; ref. 13). The DNA sequence at the cleavage sites has a strong bias for a guanine residue at the 5′ side of the DNA break (14). In view of the aromatic planar structure of camptothecin resembling a fused DNA base pair, it was hypothesized that camptothecin forms a ternary complex with topoisomerase I and DNA, stacking against a guanine-containing base pair on the 5′ side of the break while at the same time binding to sites on the topoisomerase I protein (14). Camptothecin thus is stabilized at a topoisomerase I-DNA interface in a structure similar to a normal intermediate in the topoisomerase reaction (Fig. 1C). By convention, the base pairs flanking the break are called −1 and +1. Position −1 corresponds to the nucleotide covalently linked to topoisomerase I and +1 to the nucleotide at the free 5′-hydroxyl terminus (Fig. 1A-C). The atomic structure of a ternary topoisomerase I cleavage complex was revealed by X-ray crystallography with the clinical camptothecin derivative topotecan (see structure in Fig. 1F). This structure confirmed the drug-stacking hypothesis (15). More recently, we found the binding of the natural camptothecin alkaloid to the topoisomerase I-DNA complex to be remarkably similar to the topotecan structure (16).

Because of the anticancer activity of camptothecins, many derivatives have been synthesized for clinical development (17). Two of these have been approved by the U.S. Food and Drug Administration: topotecan (Hycamtin) for ovarian and lung cancers and irinotecan (CPT11; Campto) for colon carcinomas. Several other derivatives are in preclinical development or clinical trial (Exatecan, 9-nitrocamptothecin, BAY 38-3441; refs. 18, 19). Camptothecins, however, have limitations. The α-hydroxylactone in the E-ring (see Fig. 1F) is rapidly converted in the circulation to a carboxylate whose tight binding to serum albumin limits the available active drug (20). In addition, camptothecins are actively exported from the cell by drug efflux membrane “pumps” (21).

Non-camptothecin topoisomerase I inhibitors are therefore of great pharmaceutical interest to overcome the limitations of camptothecins. Moreover, because drugs from different chemical families that share the same molecular target generally have different spectra of clinical activity, non-camptothecin topoisomerase I inhibitors may be effective against different tumors. For instance, this is the case for the tubulin and the topoisomerase II inhibitors (17). Indenoisoquinoline and indolocarbazole derivatives (Fig. 1F) are being pursued for therapeutic development. The first indenoisoquinoline topoisomerase I inhibitor was discovered by searching the National Cancer Institute cell screen database for compounds resembling camptothecin in their cytotoxicity profile against 60 human cancer cell lines (22). In the absence of definitive structural information of the drug target site, derivatives were initially designed with amino groups that would bind electrostatically to the polyanionic phosphodiester backbone. Over 300 indenoisoquinoline derivatives have been synthesized to date, and several potent and selective topoisomerase I inhibitors have been identified as potential anticancer agents (17, 23). We recently reported the first crystal structure of an indenoisoquinoline (MJ-238; Fig. 1F) bound to a topoisomerase I cleavage complex (16). In the same study, we also reported the structure of the natural camptothecin alkaloid and of an indolocarbazole (SA315F; Fig. 1F) bound to the topoisomerase I-DNA complex (16).

This study describes the crystal structure of a potent norindenoisoquinoline (24), topoisomerase I inhibitor (AI-III-52), in a complex with DNA and topoisomerase I. Comparison with the corresponding ternary complexes of the MJ-238 indenoisoquinoline (16) and with the structures of camptothecin (16), topotecan (15), and an indolocarbazole (16) reveals a common molecular mechanism of drug action. These topoisomerase I inhibitors all bind at the topoisomerase I-DNA interface by intercalating and stacking between the base pairs flanking the DNA cleavage site and by forming critical hydrogen bonds with topoisomerase I amino acid residues. These same topoisomerase I residues have been implicated previously in enzyme catalysis and/or resistance to camptothecins (25). We discuss topoisomerase I inhibitors as paradigms for drugs that trap macromolecular interfaces and the generality and implications of the interfacial inhibitor concept (26, 27).

A 70-kDa construct of human topoisomerase I (topoisomerase 70), residues Lys¹⁷⁵ to Phe⁷⁶⁵, was expressed and purified from baculovirus insect cells (Sf9) as described previously (28). Crystallization trials were conducted with wild-type topoisomerase 70 and duplex DNA oligomers containing a 5′ bridging phosphorothiolate linkage at the site of DNA cleavage (15, 16). Crystals were grown by sitting drop vapor diffusion by preparing drops containing 2 mL precipitant, 1.5 mL of 50 mmol/L duplex DNA, 0.5 mL compound, and 1.5 mL of 4.2 mg/mL protein. Precipitant was 10% to 12% polyethylene glycol 8000, 100 mmol/L MES (pH 6.4), and 200 mmol/L lithium sulfate. The compound solutions used were 1 to 5 mmol/L ligand dissolved in DMSO (0–10%, v/v) in water. Crystals were cryoprotected for data collection by passing them through precipitant plus 30% (v/v) polyethylene glycol 400. Data were collected at 100 K at Beamline 5.0.3 (Advanced Light Source, Lawrence Berkeley National Laboratory, Berkeley, CA). Structures were solved by molecular replacement using the program AMORE (29) with protein coordinates from the topoisomerase 70 ternary complex structure with topotecan (15). Refinements were conducted using CNX (30) and REFMAC (31) and iterative model adjustments with XtalView (see Table 1; ref. 32). DNAs were placed into the ∣Fo∣ − ∣Fc∣ electron density and refined. Drug molecules were then placed into the to ∣Fo∣ − ∣Fc∣ electron density using QUANTA module X-LIGAND (33). The diffraction resolution of the crystals obtained was 3.1 Å and this did not allow the placement of water molecules.

Molecular visualization was done using PyMOL. The PDB codes for the different topoisomerase I crystal structures were 1LT8 for norindenoisoquinoline (24), 1SC7 for MJ-238 (16), 1SEU for indolocarbazole (16), 1K4T for topotecan (15), 1T81 for camptothecin (16), and 1A31 for topoisomerase I in the absence of inhibitor (34).

The overall structure the AI-III-52 norindenoisoquinoline in a topoisomerase I cleavage complex is presented in Fig. 1D and E. The drug is bound deep inside topoisomerase I (Fig. 1D), intercalated between the base pairs flanking the DNA cleavage site (positions −1 and +1; Fig. 1E). The topoisomerase I protein encircles the DNA in a closed clamp conformation (34–36). The polypeptide hinge is behind the DNA, and the tight narrow cleft through which the DNA has presumably entered the enzyme is represented in front of the DNA. The topoisomerase I enzyme used in these experiments comprises the core, linker, and catalytic domain. It does not include the topoisomerase I NH₂-terminal domain that encompasses residues 1 to 174 and that is not well conserved among eukaryotic topoisomerase I enzymes. The NH₂ terminus has been shown both in vitro and in vivo to be dispensable with respect to the biochemical and cytotoxic effects of topoisomerase I poisons, including camptothecin and topotecan (28, 37). Technical aspects of structure determination are summarized in Table 1.

Figure 2A shows two expanded views of AI-III-52 bound to the DNA in the topoisomerase I cleavage complex. Fig. 2A (left) is in the same view orientation as Fig. 1D and E, with the DNA cleavage site seen from the minor groove. Fig. 2A (right) is a 90° rotation showing AI-III-52 under the +1 nucleotide pair. The +1 bp (shown as capped sticks) covers and stacks against the entire drug (shown in space filling model). Extensive stacking is also the case for the −1 bp, which is completely covered by the drug and therefore not shown in Fig. 2A (right). Figure 3B shows the relative positions of the base pairs flanking the intercalated norindenoisoquinoline molecule. The +1 bp is oriented as in Fig. 2A (right), and the drug molecule is not shown. The longitudinal axes of the +1 and −1 bp (shown as thin dashed lines) are almost parallel. Thus, the twist angle between the base pairs flanking the intercalated drug is reduced to 10°.

DNA unwinding is a characteristic of DNA intercalators, caused by the increased distance between flanking base pairs required to accommodate the intercalated molecule. For instance, intercalation of ethidium reduces the twist angle between the flanking bases pairs to 7° compared with the normal 33°. The reduction in the twist angle between flanking base pairs for the norindenoisoquinoline AI-III-52 intercalated in the topoisomerase I-DNA cleavage complex (Fig. 3B) is comparable with that produced by intercalated ethidium. The twist angle reduction observed with AI-III-52 is not due to topoisomerase I cleavage, because the twist angle between the base pair flanking the topoisomerase I cleavage site in the absence of drug remains near normal (Fig. 3A; ref. 34). The twist angle of 58° between +1 and +2 bp in the norindenoisoquinoline structure indicates slight overwinding of the DNA immediately adjacent to the drug binding (Fig. 3C). This overwinding probably compensates for the unwinding produced by the norindenoisoquinoline.

Figure 2B shows that the indenoisoquinoline MJ-238 (see structure in Fig. 1F), like the norindenoisoquinoline AI-III-52, stacks and intercalates between bp −1 and +1 on each side of the cleavage site generated by topoisomerase I. For both drugs, their aromatic portion is parallel to the −1 and +1 bp long axes. However, the two indenoisoquinolines are flipped 180° (Figs. 2 and 4, compare A and B). Numbering of the corresponding atoms makes this difference apparent. For AI-III-52, the sequence of atoms in a clockwise orientation is 1, 10, and 6, whereas for MJ-238 the sequence is 1, 6, and 10. In other words, AI-III-52 is intercalated with the nitrogen N6 in the minor groove. This nitrogen N6 can hydrogen bond with the topoisomerase I R364 residue. By contrast, MJ-238 has the nitrogen N6 in the major groove of the DNA and the C11 carbonyl hydrogen bonded to R364. The side chain attached to the nitrogen N6 protrudes in the DNA major groove (Fig. 2B).

Like the indenoisoquinolines, the two camptothecins intercalate between −1 and +1 bp at the cleavage site of the topoisomerase I-DNA complex (Fig. 2C and D; refs. 15, 16). Both camptothecins stack in the same orientation. Their five-membered planar heterocyclic structure coincides with the long axis of the base pairs, and their shapes closely match that of a base pair. The hydroxyl substituents at the 20-position are in close proximity with the sugar of the −1 nucleotide and the DNA phosphodiester backbone and contribute to the optimum stacking of the natural 20S-camptothecin enantiomer. This explains why the 20R-camptothecin enantiomer cannot trap the topoisomerase I catalytic complex (13). At the other end of the camptothecin, the A ring is not immediately in contact with the DNA phosphodiester backbone opposite from the cleavage site. This space can accommodate the 10-hydroxy substituent of topotecan. Filling this space probably accounts for the greater potency of 10-hydroxy-substituted camptothecins and for the even greater activity of 10,11-dimethoxy-camptothecin derivatives (13, 38).

The indolocarbazole SA315F is also bound with maximum stacking interactions. The indolocarbazole aromatic portion is aligned with the axis of the −1 and +1 bp (Fig. 2E; ref. 16). The deoxyribose is positioned in the major groove.

For the indenoisoquinoline MJ-238, topotecan, and the indolocarbazole SA315F, the drug aromatic polycyclic ring is parallel to the +1 bp axis and the bulky substituents encroach the major groove. This common orientation would suggest that drug substitutions that impinge on the minor groove may be disfavored. Consistent with this possibility, substitutions on the camptothecins at positions 1 and 12 (see Fig. 1F) preclude topoisomerase I inhibition (13). In addition, minor groove substitutions with benzo[a]pyrene diol epoxide adducts on deoxyguanine N2 prevent topoisomerase I from cleaving DNA when they occupy the topoisomerase I site in the DNA minor groove (39). In contrast, substitutions toward the minor groove on indenoisoquinolines at position 11 (see Fig. 1F) result in more potent compounds for cytotoxicity and topoisomerase I inhibition (40–42). This suggests that these substitutions directly interact with topoisomerase I residues and strengthen the hydrogen bond network for these particular compounds.

A schematic representation of the hydrogen bond network for the two indenoisoquinolines and the two camptothecins is shown in Fig. 4. Two of the topoisomerase I catalytic residues are shown. Y723 forms the covalent bond with DNA. K532 serves a dual function. K532 forms a hydrogen bond on the minor groove side with the −1 base on the scissile strand (8, 15) and probably also functions as a general acid during cleavage to protonate the leaving 5′ oxygen (43). Three other topoisomerase I residues form hydrogen bonds with the intercalated drugs: N722, R364, and D533. None of these residues are essential for topoisomerase I catalytic activity. However, each of them has been implicated in resistance to camptothecins (4), indolocarbazoles (44), and indenoisoquinolines (23).

The norindenoisoquinoline AI-III-52 forms two hydrogen bonds, one with R364 and one with N722 (Fig. 4A). Only one hydrogen bond was detected for MJ-238 (16): with R364 (Fig. 4B). The weaker topoisomerase I inhibitory activity of MJ-238 compared with AI-III-52 (16, 24) can therefore be attributed to its reduced hydrogen bonding with topoisomerase I and to reduced stacking because of the smaller surface of its aromatic portion (Figs. 2 and 4, compare A and B).

The two camptothecins have a more similar H-bond network with the topoisomerase I-DNA complex than the indenoisoquinolines (Fig. 4). For camptothecin, three hydrogen bonds have been reported (Fig. 4C; ref. 16). Two are direct. One is between the camptothecin 20-hydroxy and D533, and the other between the camptothecin N1 and R364. The third H-bond has been proposed to be water mediated between the 17-carbonyl on the camptothecin D-ring and N722 of topoisomerase I (16). For topotecan, only two of the three H-bonds (with D533 and N722) were initially reported (15) and are shown in Fig. 4D. Reexamination of the topotecan ternary structure reveals that the R364 side chain is also in close proximity with N1 of topotecan and is positioned for hydrogen bonding (15).

To our knowledge, camptothecins were the first inhibitors hypothesized (14, 45) and shown (15) to bind at the “interface of two macromolecules” (topoisomerase I and its DNA substrate), as these macromolecules undergo a catalytic conformational change (the “cleavage complex”; see Fig. 1A-C). This mechanism can be generalized as it also applies to three non-camptothecin inhibitors: two indenoisoquinolines and an indolocarbazole (see Fig. 2). The implications of these structures are twofold. First, they will aid for the rational synthesis of novel topoisomerase I inhibitors. However, the example of the two indenoisoquinolines that bind in flipped orientation (see Fig. 2A and B) shows the potential limitations of simple extrapolations from crystal structures, the need to correlate structural biology predictions with biological testing and ultimately additional atomic resolutions, and, hence, the need of an iterative process from chemistry to biology to structural biology, and again.

A second implication of the interfacial binding of topoisomerase I inhibitors is its apparent generality. The “interfacial inhibition” paradigm can be extended to a wide range of drugs and macromolecular targets (Table 2; refs. 26, 27). We therefore define “interfacial inhibitors” as drugs that bind at the interface of two (or more) macromolecules as the multimeric complex undergoes a structural transition. These multimeric complexes can present a drug binding site at their protein-protein interface as in the case of brefeldin A and tubulin inhibitors [colchicine, vinblastine, and paclitaxel (Taxol)] or a drug-binding site at nucleic acid-protein interfaces as in the case of topoisomerase I inhibitors (present study), the RNA polymerase II inhibitor α-amanitin, and the antibiotic ribosome inhibitors [Table 2 (26, 27) and Fig. 3 (in a recent review; ref. 27)].

Interfacial inhibition is reversible and uncompetitive (26, 27). Interfacial inhibitors bind to and trap a catalytic intermediate of the macromolecular complex in a specific conformation (Fig. 5A). Interfacial inhibition represents a paradigm for “uncompetitive inhibition” (Fig. 5B and C: Lineweaver-Burke plot equations). The pharmacologic activity of interfacial inhibitors is due at least in part to the conversion of the trapped macromolecular intermediate into lethal lesions by other macromolecules that interact with the normally transient ternary drug-macromolecular complex. For instance, in the case of topoisomerase I inhibitors, the cleavage complexes are only lethal when they are converted into replication double-strand breaks by the collision of a DNA polymerase with a topoisomerase I cleavage complex reversibly trapped by the drug (46). This explains why the antiproliferative activity of topoisomerase I inhibitors increases with the levels of topoisomerase I. Conversely, a common resistance mechanism to topoisomerase I inhibitors is down-regulation of topoisomerase I (4). Hence, interfacial inhibitors differ from competitive inhibitors. Increasing drug target increases resistance to competitive inhibitors, whereas it sensitizes to interfacial inhibitors.

The interfacial inhibitor paradigm has implications for drug discovery as interfacial inhibitors stabilize rather than inhibit the formation of macromolecular complexes. Assays should therefore be designed to look for drugs that stabilize rather than inhibit the formation of macromolecular complexes. Interfacial inhibitor screening using various methods to measure the binding of macromolecules has the potential to discover highly selective inhibitors for one particular receptor site and/or enzyme substrate.