Introduction

Top Summary Introduction Procedures Results & Discussion References

Several classes of inhibitors of HIV-1 IN have been reported (for recent reviews, see Refs. 1 and 2). Among these inhibitors, hydroxylated aromatics, which are present in a variety of natural products, have consistently exhibited remarkable potency in in vitro assays specific for IN. Although they are from a variety of unrelated families, the majority of these inhibitors share a common structural feature of two aryl units separated by a central linker (examples are shown in Fig. 1). The recurring requirement of a catechol moiety is a common theme among hydroxylated aromatic inhibitors of IN. Catechols, however, do not have desirable properties in the cell-based assay against HIV-1 infection; many of them retain high cytotoxicity. Oxidation of catechol units in situ to quinones may contribute to their cytotoxicity. Recent reports by Robinson et al. (3, 4) indicate that the caffeoylquinic acids and chicoric acid, both of which contain two catechol moieties, exhibit remarkable antiviral activity with high potency against IN.

View larger version (23K): [in this window] [in a new window]   Fig. 1.   Structures of several known hydroxylated aromatic inhibitors of HIV-1 integrase. CAPE, caffeic acid phenethyl ester.

Several other studies on the antiviral activity and cytotoxicity of chicoric acids have appeared since the isolation and synthesis of the parent compound were first reported by Scarpati and Oriente (5). For example, chicoric acid and caffeic acid reduced infectivity of vesicular stomatitis virus in mouse L-929 cells only after infection with vesicular stomatitis virus (6). These compounds provide no prophylactic properties; no antiviral effects were observed when L-929 cells were treated before vesicular stomatitis virus infection (6). A variety of polyhydroxylated carboxylic acids have been reported to inhibit HIV-1 replication, perhaps through interaction with gp120 and prevention of virus binding to the CD₄ receptor. For example, Mahmood et al. (7) reported that 4,5-di-O-caffeoylquinic acid interacts irreversibly with gp120. Moreover, galloylquinic acids were reported to possess biological activities that included inhibition of HIV-1 reverse transcriptase and human DNA polymerases , , and  (8).

Other biological properties of plant polyphenols include anticarcinogenic, anti-inflammatory, antibacterial, immune-stimulating, antiallergic, and ostrogenic effects and inhibition of cyclo-oxygenase, lipoxygenase, and phospholipase A (9). Plant polyphenols are in general multifunctional and can act as reducing agents, hydrogen-donating antioxidants, metal chelators, and singlet-oxygen quenchers (9).

In our search for novel IN inhibitors, we have identified many hydroxylated aromatic inhibitors of IN (1, 2). Although they are in general cytotoxic or inactive against HIV-1 replication in cell-based assays, their use as lead compounds for the design of noncatechol-containing inhibitors has been indispensable, and their contribution to our structure-activity studies has helped to elucidate the structural requirement for cytotoxicity and activity. We previously identified a series of novel inhibitors of IN based on three-point pharmacophores that were used in searches of the National Cancer Institute 3D database to identify numerous IN inhibitors (10-12). Evidence was obtained that suggested that some of these three-point pharmacophores were in fact part of a four-point pharmacophore; we have now extended those studies to characterize such a pharmacophore. This pharmacophore has been built from the low energy conformations of chicoric acid and caffeoylquinic acids, and its use in a search of the National Cancer Institute 3D database yielded several novel inhibitors; among them, a series of tetracyclines were identified as potent inhibitors and studied further to characterize their molecular interaction with IN.

	Experimental Procedures

Top Summary Introduction Procedures Results & Discussion References

Chemicals. Tetracycline, oxytetracycline, doxycycline, methacycline, and rolitetracycline were obtained from Sigma Chemical (St. Louis, MO). All of the other compounds used in this study were obtained from the National Cancer Institute chemical repository through the Drug Synthesis and Chemistry Branch (Bethesda, MD). All compounds were dissolved in DMSO, and the stock solutions were stored at 20°.

Preparation of oligonucleotide substrates. The high performance liquid chromatography-purified oligonucleotides AE117, 5-ACTGCTAGAGATTTTCCACAC-3; AE118, 5-GTGTGGAAAATCTCTAGCAGT-3; AE157, 5-GAAAGCGACCGCGCC-3; AE146, 5-GGACGCCATAGCCCCGGCGCGGTCGCTTTC-3; AE156, 5-GTGTGGAAAATCTCTAGCAGGGGCTATGGCGTCC-3; RM22M, 5-TACTGCTAGAGATTTTCCACAC-3; and RMAB2, 5-GTGTGGAAAATCTCTAGCUGT-3 were purchased from Midland Certified Reagent Company (Midland, TX). Purified recombinant IN deletion mutant IN^50-212, expression system for the wild-type IN and the IN^50-212 (F185K), were generous gifts of Drs. T. Jenkins and R. Craigie (Laboratory of Molecular Biology, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD). To analyze the extents of 3-processing and strand transfer using 5-end-labeled substrates, AE118 was 5-end labeled using T₄ polynucleotide kinase (GIBCO BRL, Gaithersburg, MD) and [-³²P]ATP (DuPont-New England Nuclear, Boston, MA). To determine the extent of 30-mer target strand generation during disintegration, AE157 was 5-end labeled and annealed to AE156, AE146, and AE117. The kinase was heat inactivated, and AE117 was added to the same final concentration. The mixture was heated at 95°, allowed to cool slowly to room temperature, and run through a G-25 Sephadex quick-spin column (Boehringer-Mannheim Biochemicals, Indianapolis, IN) to separate annealed double-stranded oligonucleotide from unincorporated label.

IN assays. To determine the extent of 3-processing and strand transfer, IN was preincubated at a final concentration of 200 nM with the inhibitor in reaction buffer [containing 50 mM NaCl, 1 mM HEPES, pH 7.5, 50 µM EDTA, 50 µM dithiothreitol, 10% glycerol (w/v), 7.5 mM MnCl₂, 0.1 mg/ml bovine serum albumin, 10 mM 2-mercaptoethanol, 10% DMSO, and 25 mM 3-(N-morpholino)propanesulfonic acid, pH 7.2] at 30° for 30 min (14). Then, 20 nM of the 5-end ³²P-labeled linear oligonucleotide substrate was added, and incubation was continued for an additional hour. Reactions were quenched by the addition of an equal volume (16 µl) of loading dye (98% deionized formamide, 10 mM EDTA, 0.025% xylene cyanol, 0.025% bromphenol blue). An aliquot (5 µl) was electrophoresed on a denaturing 20% polyacrylamide gel (0.09 M Tris-borate, pH 8.3, 2 mM EDTA, 20% acrylamide, 8 M urea).

DNA binding assay using Schiff base formation. This assay was described in detail recently (13). Briefly, IN (200 nM) was preincubated with the inhibitor (at the indicated concentration) for 30 min at 30°. Subsequently, an oligonucleotide that contained an abasic site (13) (see Fig. 9A) in reaction buffer (described above) was added for 2 min at room temperature. A freshly prepared solution of sodium borohydride (0.1 M final concentration) was added, and reaction was continued for an additional 2 min. An equal volume (16 µl) of 2× SDS-polyacrylamide gel electrophoresis buffer (100 mM Tris, pH 6.8, 4% 2-mercaptoethanol, 4% SDS, 0.2% bromphenol blue, 20% glycerol) was added to each reaction, and the reaction was heated at 95° for 3 min before loading a 20-µl aliquot onto a 12% SDS-polyacrylamide gel. The gel was run at 120 V for 1.5 hr, dried, and exposed in a PhosphorImager cassette. Gels were analyzed using a Molecular Dynamics PhosphorImager.

View larger version (91K): [in this window] [in a new window]   Fig. 9.   IN-DNA cross-linking by formation of a Schiff base between IN and a duplex oligonucleotide that contains an abasic site. A, A 21-mer blunt-end oligonucleotide that corresponds to the U5 end of HIV proviral DNA that contains a uracil (underlined) in place of an adenine. B, Effect of tetracycline 25 and rolitetracycline 26 on the extent of DNA binding of HIV-1 integrase in the presence of 2 mM manganese chloride (A), 2 mM magnesium chloride (B), or 2 mM EDTA (C). PhosphorImager image that shows the inhibition of the 39-kDa product that corresponds to the IN-DNA covalent complex (lane 2) in the presence of 1000, 250, 100, 50, 25, 10, and 1 µM concentration of drugs (lanes 3-10).

Topoisomerase reactions. Reactions were performed in 10 µl of reaction buffer (0.01 M Tris·HCl, pH 7.5, 150 mM KCl, 5 mM MgCl₂, 0.1 mM EDTA, 15 mg/ml bovine serum albumin) with the following duplex oligonucleotide substrate (17) labeled with -³²P-cordycepin at the 3-end of the upper strand (*): 5-GATCTAAAAGACTT  GGAAAAATTTTTAAAAAA*ATTTTCTGAA-CCTTTTTAAAAATTTTTTCTAG-5. This oligonucleotide contains a single topoisomerase I cleavage site (caret in upper strand). Approximately 50 fmol of oligonucleotide/reaction was incubated with 10 units of calf thymus DNA topoisomerase I (GIBCO BRL). Reactions were stopped by the addition of SDS (0.5% final concentration). Proteolysis was halted by the addition of 36 µl of 2.5× loading buffer (98% formamide, 0.01 M EDTA, 1 mg/ml xylene cyanol, 1 mg/ml bromphenol blue). An aliquot (5 µl) was electrophoresed on a denaturing 20% polyacrylamide gel (0.09 M Tris-borate, pH 8.3, 2 mM EDTA, 20% acrylamide, 8 M urea).

National Cancer Institute 3D database and search software. The National Cancer Institute 3D database and the Chem-X program used in both 3D database build and search processes have been described previously (18-20). The current version of the National Cancer Institute 3D database consists of 206,876 open and 201,036 discrete (proprietary) structures, for a total of 407,912 structures. All searches reported in this article were conducted only within the open part of the database. The conformational flexible search algorithm implemented in Chem-X (July 1994 version; Chemical Design, Oxford, UK) running on a Silicon Graphics IRIS Indigo workstation (Mountain View, CA) was used. For flexible compounds, multiple conformations are generated and analyzed during both building and searching of the database. A pharmacophore in this study refers to the 3D arrangement of those atoms in a compound that is responsible for the biological activity of the compound. Three-dimensional database pharmacophore searching facilitates identification of molecules that meet the requirements specified in the pharmacophore query. This approach has gained attention recently for its use in the discovery of new leads in drug development programs, and we have built a searchable 3D database (18) with a total of 407,000 structures from the two-dimensional structures of the National Cancer Institute Drug Information System database (19) using the program Chem-X. Use of 3D database searching has enabled us to identify a number of novel protein kinase C agonists (21), as well as HIV-1 protease (22) and IN inhibitors (10-12).

Molecular modeling. All molecular modeling studies were performed with the QUANTA 4.0/CHARMm 22 (Molecular Simulations, Burlington, MA) molecular modeling package running on a Silicon Graphics IRIS Indigo workstation. Energy minimization was typically computed with 5000 iterations or until convergence (defined as an energy gradient of 0.001 Kcal/mol/Å), using an Adjusted Basis Newton-Raphson algorithm as implemented in CHARMm. The structures of compounds were built using the ChemNote module with QUANTA and were energy-minimized using CHARMm. Conformational searches were conducted using the Monte Carlo random search algorithm implemented in QUANTA.

	Results and Discussion

Top Summary Introduction Procedures Results & Discussion References

Identification of a unique four-point pharmacophore. Recently, we reported the identification of two distinct three-point pharmacophores among a series of lichen acids (10). Searches of the National Cancer Institute 3D database with these pharmacophores yielded 800 compounds that contained either of these pharmacophores; the dicaffeoylquinic acids were identified in this way as potent inhibitors of IN. Robinson et al. (3, 4) reported anti-HIV-1 activity for the caffeoylquinic class of compounds. To determine the biologically active conformation of 1-MO-3,5-DCQA, 3,5-DCQA, and L-chicoric acid, 5000 conformations were generated in each case with the use of a random sample generator in QUANTA. Each of the conformations was minimized with the Adjusted Basis Newton-Raphson algorithm in CHARMm (23) using 5000 iterations or until convergence defined as an energy gradient of 0.001 Kcal/mol/Å. The unique four-point pharmacophore was identified in each case from analysis of the low energy (E < 4.7 Kcal/mol) conformations of these three compounds. The pharmacophore query shown in Fig. 2 was built from the 3D structures of 1-MO-3,5-DCQA, 3,5-DCQA, and L-chicoric acid, whose structures are shown in Fig. 1. These compounds possess the three-point pharmacophores defined previously (10-12); it was envisaged that four-point pharmacophores should be both more precise and more specific. Compounds that contained a four-point pharmacophore should bind more effectively than those with only a three-point pharmacophore. The pattern of six distances in a four-point pharmacophore is more stringent than that of the three distances in a three-point pharmacophore, and fewer hits should be expected in a search. Molecular modeling studies of 1-MO-3,5-DCQA, 3,5-DCQA, and L-chicoric acid enabled us to identify the novel four-point pharmacophore shown in Fig. 2. These three IN inhibitors fit this four-point pharmacophore in low energy conformations (E = 1.25, 4.37, and 4.67 Kcal/mol, respectively), and the superimposition on one another of the four-point pharmacophores in the best-fit conformations of these inhibitors, a view of which is shown in Fig. 3, gave RMS values of 0.334, 0.209, and 0.055Å. From this four-point pharmacophore, we can generate four independent three-point pharmacophores with distance patterns: 9.5, 8.6, and 2.6 Å; 11.5, 9.5, and 2.6 Å; 10.5, 8.6, and 2.6 Å; and 11.5, 10.5, and 2.6 Å.

View larger version (24K): [in this window] [in a new window]   Fig. 2.   Dimensions of four-point pharmacophore queries. In searching the National Cancer Institute 3D database, a basic nitrogen atom was allowed in place of oxygen.

View larger version (52K): [in this window] [in a new window]   Fig. 3.   Four-point pharmacophore superimposition (RMS = 0.334, 0.209, and 0.055 Å) of chicoric acid (dark gray lines), 3,4-DCQA acid (light gray lines), and 3,5-DCQA (thick lines).

3D database search. The four-point pharmacophore was used in a search of the National Cancer Institute 3D database. The distance tolerances in the pharmacophore query were set to 0.7 Å, as shown in Fig. 2. Every atom in the pharmacophore was allowed to be either oxygen or an amino nitrogen. The 3D database search of 206,876 structures yielded a total of 179 compounds that contain this four-point pharmacophore distance pattern in one or more of their conformations. Thirty-nine compounds (Fig. 4) were manually selected for bioassay based on considerations of structural diversity and sample availability, with the most probable four-point pharmacophore centers identified. It is important to note that several of the compounds selected by the four-point pharmacophore presented in this study would also fit the previously described three-point pharmacophore. However, the three-point pharmacophore in general yielded a higher number of hits. On the other hand, the four-point pharmacophore yielded a greater number of active inhibitors (see below).

View larger version (22K): [in this window] [in a new window]   View larger version (24K): [in this window] [in a new window]   View larger version (23K): [in this window] [in a new window]   Fig. 4.   Structures of compounds that contain the four-point pharmacophore.

Influence of conformational energy penalty and geometric fitness. All 39 compounds were initially tested in the IN assay at 100 µg/ml (Table 1). Compounds with inhibitory potency >50% were further assayed at different concentrations to determine their IC₅₀ values. Twenty (51%) exhibited IC₅₀ values of <20 µM and 24 (62%) exhibited IC₅₀ values at <100 µM (Table 1), which suggests that the pharmacophore query used for searching is effective in identifying active compounds. However, 15 of the compounds were inactive (IC₅₀ > 100 µM), so the four-point pharmacophore query defined in Fig. 2 seems to be necessary but not sufficient to determine effective binding of the ligand to IN. Other factors such as hydrophobicity and molecular shape effects could be responsible for the biological inactivity of compounds that contain the pharmacophore. Two factors that contribute to the structure-activity relationships are the conformational energy penalty and the geometric fitness of the pharmacophore in the ligand binding conformation. The larger the conformational energy penalty, the more difficult it will be for the inhibitor to adopt the desired conformation and less precise its binding will be to the enzyme. The better the enzyme-ligand fit, the more potent is the inhibitor.

View this table: [in this window] [in a new window]   TABLE 1 Correlation between anti-HIV-1 integrase activities of a series of four-point pharmacophore containing drugs with their corresponding RMS and E1 values Purified recombinant HIV-1 integrase was treated with different concentrations of drugs as described in Experimental Procedures to obtain IC50 values. RMS values refer to root-mean-square of superimposition of indicated compound onto the four-point pharmacophore. E1 values were obtained from energy differences between lowest energy conformer and superposition conformers of indicated compounds onto the four-point pharmacophore.

View larger version (12K): [in this window] [in a new window]   Fig. 5.   Correlation between HIV-1 integrase inhibitory potency and F values according to the equation F = (E1  E2)(RMS). The natural logarithm of IC50 values for 3-processing (A) and strand transfer (B) is plotted against the natural logarithm of F.

Selection of compounds that contain the four-point pharmacophore. Structures of the 39 four-point pharmacophore-containing compounds assayed in this study are shown in Fig. 4. The pharmacophore query was built on the basis of the hydrogen bond donating capabilities of the inhibitors because in many cases, the hydrogen bond interaction provides a major contribution to the overall free energy of binding. Because a basic nitrogen may be capable of behaving as either a hydrogen bond donor or acceptor, a basic nitrogen atom was allowed in place of oxygen in the search, and several of the retrieved compounds (e.g., 6, 8, 12, 13, 15, 19, 21-24, and 25-39) contain at least one basic nitrogen atom in the pharmacophore assembly.

In vitro inhibition of HIV-1 IN. Catechol-containing compounds in general are potent inhibitors of IN in vitro (1). In many instances, this inhibition is associated with nonspecific binding to other cellular targets. Thus, collateral toxicity associated with this class of inhibitors is of major concern in anti-HIV drug development. We have used 3D database searching of the National Cancer Institute drug screening program to identify novel inhibitors of IN and identified novel non-catechol-containing inhibitors (1), some of which are presented in this study. Several of the inhibitors reported are analogs of compounds reported previously; for example, we identified a series of flavones as inhibitors of IN (25). Novel flavones, such as the dihydrochalcone derivative 1 and the phylloflavan 3, inhibited IN <2 µM. In common with the chicoric and caffeoylquinic acids, compounds 1-3 contain two catechol moieties. All of the catechol-containing compounds exhibited similar potency as those for the chicoric and caffeoylquinic acids, apparently regardless of the linker length and linker geometry. Compound 11, with two seven-membered rings, also exhibited remarkable potency, with IC₅₀ values of 1.4 and 2.3 µM for 3-processing and strand transfer, respectively. However, when the OH groups are protected, as in compounds 9, 10, and 20, total loss of activity was observed (Table 1). These results are in accord with our recent investigation of a series of monohydroxylated arylamide inhibitors of IN (26). Derivatization of a hydroxyl group in a series of catechol-containing inhibitors abolished anti-IN activities (26), and the implication is that a hydrogen bond donor (---OH) is necessary at these positions: a hydrogen bond acceptor (---O---) is inadequate. Anthracycline antibiotics 22 and 23 exhibited IC₅₀ values of 1.4 ± 0.4 and 2.3 ± 0.8 µM for 3-processing and 1.5 ± 0.7 and 4.1 ± 2.0 µM for strand transfer, respectively. In accord with our earlier observations (27), several other anthracyclines also showed similar potency range (data not shown). N-Caffeoylglycine-L-phenylalanine (6), with IC₅₀ values of 104.4 ± 21.3 µM for 3-processing and 32.0 ± 17.7 µM for strand transfer, was moderately active. The disulfides 7 and 8 exhibited moderate activity, whereas the sulfonylamides 14 and 15 were less active.

Relevance to antiviral activity. We based our study on the caffeoylquinic and chicoric acids that had been reported to exhibit antiviral activity with high potency against IN (3, 4). Interestingly, none of the caffeoylquinic acids tested in the National Cancer Institute Antiviral Drug Screening Program exhibited detectable activity against HIV-1-infected CEM cells (10). In addition, none of the tested compounds presented in Table 1 exhibited significant antiviral activity. Compounds 13-17, 21, 34, and 39 showed remarkable cytotoxicity. Interestingly, several of the tetracyclines were noncytotoxic, which suggests that they are devoid of nonspecific cellular effects. Further rational drug design requires synthesis of derivatives that would be more stable by modifying the R group (Fig. 4) to prevent intracellular cleavage, leading to the parent tetracycline, which has only marginal potency against IN.

Mechanism of IN inhibition by tetracyclines. On infection of cells, effective retroviral replication requires the integration of a double-stranded DNA copy of the viral single-stranded RNA genome into a host chromosome. HIV-1 IN is known to catalyze two consecutive reactions. Initially, it processes linear viral DNA by removing the nucleotides 3 from the conserved CA, leaving the recessed 3-OH termini. Subsequently, the processed 3-ends of the viral DNA are then covalently joined to host DNA (1, 31, 32). These two steps, known respectively as 3-processing and DNA strand transfer, can be readily measured in an in vitro assay using purified recombinant IN, a 21-mer duplex oligonucleotide that corresponds to the U5 end of the HIV long terminal repeat sequence (Fig. 6A) and divalent metal ion (Mn²⁺ or Mg²⁺). Fig. 6B shows a representative gel that illustrates inhibition of both 3-processing and strand transfer reactions by tetracyclines 29, 28, 26, and 25. The substituted tetracyclines 26, 28, and 29 were 5-100-fold more potent than the parent compound 25. When we examined the inhibition of 3-processing and strand transfer in the presence of Mg²⁺, no significant differences in IC₅₀ values were observed from those in the presence of Mn²⁺ (not shown). The IC₅₀ values for 3-processing and strand transfer, respectively, were 2.7 and 2.4 µM for 29, 3.2 and 3.1 µM for 28, 49.8 and 43.2 µM for 26, and 341.0 and 333.0 µM for 25.

View larger version (66K): [in this window] [in a new window]   Fig. 6.   HIV-1 integrase catalytic assays. A, A 21-mer blunt-end oligonucleotide that corresponds to the U5 end of the HIV-1 proviral DNA, 5-end-labeled with 32P, is reacted with purified HIV-1 integrase. The initial step involves nucleolytic cleavage of two bases from the 3-end, which results in a 19-mer oligonucleotide. Subsequently, 3-ends are covalently joined to another identical oligonucleotide, which serves as the target DNA. B, Concentration-dependent inhibition of HIV-1 IN by tetracyclines 29, 28, 26, and 25. Above each lane, drug concentrations in micromolar.

View larger version (66K): [in this window] [in a new window]   Fig. 7.   Choice of nucleophile for the 3-processing reaction by HIV-1 integrase. A, Nucleophilic substitution of water, glycerol, or the 3-hydroxyl group of the viral DNA terminus yielding a linear trinucleotide with a 5-phosphate (L), a linear trinucleotide with a glycerol esterified to the 5-phosphate (G), and a cyclic trinucleotide (C). Global inhibition of nucleophilic attack in the 3-processing reaction by tetracyclines 29, 28, 26, and 25 (B). Above each lane, drug concentrations in micromolar.

View larger version (38K): [in this window] [in a new window]   Fig. 8.   HIV-1 integrase disintegration assay using the IN core domain. A, The substrate oligonucleotide mimics a strand transfer step product (i.e., a Y oligonucleotide that contains a 15-mer oligonucleotide 5-end-labeled with 32P). HIV-1 integrase mediates the disintegration generating a 30-mer oligonucleotide. B, Concentration-dependent inhibition of IN50-212 (F185K)-catalyzed disintegration step by compounds 26, 25 and 28. The electrophoresis in a 20% denaturing acrylamide gels showing the original 15-mer and the 30-mer disintegration products. Above each lane, drug concentrations in micromolar. C, Graph that shows the quantification of results in B. 25, ; 26, ; 28, . Percent inhibition was calculated after PhosphorImager quantification.

View larger version (16K): [in this window] [in a new window]   Fig. 10.   Concentration- and time-dependent inhibition of IN-DNA binding by rolitetracycline 26. Preincubation indicates that drug was incubated with the enzyme for 30 min before the addition of DNA. Postincubation indicates that drug was added 5 min after IN and DNA were mixed. Reaction was performed in the presence of 2 mM manganese chloride.

Tetracyclines do not affect eukaryotic DNA topoisomerase I. Eukaryotic topoisomerase I assays can be used to probe the selectivity of compounds toward other DNA binding proteins. To determine the extent of topoisomerase I cleavable complex formation and inhibition of cleavable complex trapped by camptothecin, a 33-mer oligonucleotide bearing a unique topoisomerase I cleavage site in its center was used (17). and formation of the cleavable complex was monitored. None of the four tetracyclines (29, 28, 26, and 25) tested induced any detectable cleavable complex (data not shown) or inhibited the ability of topoisomerase I to generate camptothecin-mediated cleavable complex at concentrations that effectively inhibited IN (Fig. 11). These results indicate that the tested compounds exhibit some selectivity for IN and suggest that their inhibitory potency is not due to nonspecific binding to the IN or the DNA. Compounds 29, 28, and 26 exhibited no significant inhibition at the highest concentration tested, and tetracycline 25 showed an IC₅₀ value of 400 µM against topoisomerase I (Fig. 11).

View larger version (74K): [in this window] [in a new window]   Fig. 11.   Inhibition of camptothecin-induced cleavable complex formation by compounds 29, 28, 26, and 25. Lane 1, DNA alone. Lane 2, DNA plus topoisomerase I. Lanes 3-21, DNA and topoisomerase I plus camptothecin (10 µM); numbers above lanes, drug concentration in micromolar.

Tetracycline as lead compounds. A unique class of compounds identified in this study are the tetracyclines 25-39, which inhibited IN at low micromolar concentrations. Tetracyclines reported in this study are water soluble and potential candidates for cocrystallization with wild-type IN. In addition, tetracyclines are strong chelating agents and therefore of interest in elucidating the role of metal ions in the mechanism of IN function. It is puzzling why the prototype tetracycline with no substituent on the carboxamide group at the 2 position is considerably less potent than the carboxamido protected derivatives. This disparity is perhaps partially due to the nature of chelating properties of the tetracyclines. Alternatively, substituted tetracyclines could undergo intramolecular rearrangement on the side group (R group in Fig. 4) to release reactive electrophiles at the IN catalytic site. Although all the structures presented in this study contain the four-point pharmacophores, which can reasonably be superimposed on one another, the requirement for potency against IN is stringent, and subtle changes in structure can greatly affect their potency. Therefore, a more precise pharmacophore determination may have to await X-ray or NMR structural information. We are currently engaged in determining the cocrystal structure of several of the water-soluble inhibitors with HIV-1 IN.