Development of new HIV drug targeting a cofactor of the cell
- Ledgins - Pfizer - pdf attached
Download the PDF here |
17 May 2010
We report in Nature Chemical Biology on the development of a new class of anti-HIV agents: the LEDGINs. Because HIV becomes resistant against current drugs, it remains important to develop new drugs preferentially directed against novel targets. In 2003 our virology group discovered how the virus uses LEDGF/p75, a protein of the cell, to integrate in the genome of the host cell. LEDGF/p75 acts as a molecular tether bridging the viral integration complex and the chromatin. In the context of the FP6 TRIoH project, we collaborated with many European partners to validate LEDGF/p75 as a genuine antiviral target. Today Dr. Debyser, who presently heads the FP7 integrated project THINC, announces the development of highly potent anti-HIV agents with clinical potential. This achievement required the input from a multidisciplinary team of medicinal chemists, virologists, structural biologists and modelers.
Nature Biological Chemistry Christ et al.This breakthrough proves that it is possible to block HIV replication by targeting the interaction between a viral protein and a cellular cofactor. Researchers warn that years of clinical research still lie ahead.
Research was financed in part by the FP7 integrated project THINC.
Nature Chemical Biology
Rational design of small-molecule inhibitors of the LEDGF/p75-integrase interaction and HIV replication - pdf attached
"We have now designed and synthesized a new class of antiviral agents that effectively target the LEDGF/p75-IN interaction, not only demonstrating the feasibility of exploiting PPIs as targets for drug discovery but also validating a new paradigm in anti-HIV research......Because of their limited genome size, viruses in general, and HIV in particular, must rely on host proteins, referred to as cofactors, to achieve the multiple steps in HIV replication. Such protein-protein interactions (PPIs) between viral proteins and cofactors constitute a pool of potential new antiviral targets against which small-molecule protein-protein interaction inhibitors (SMPPIIs) can be designed. Genome and proteome analyses highlight the key role of PPIs in biological processes; thus, PPIs represent a large and important class of therapeutic targets36, 37. Nevertheless, these interactions were left aside for some time because drug development against targets such as enzymes and receptors has so far been more cost effective. However the expense of developing these classical drugs is increasing, and new target proteins and/or PPIs have come within reach......we determined its activity in primary isolates from human donors (primary peripheral blood mononuclear cells and macrophages) (Supplementary Table 3 and Supplementary Methods). Compound 6 retained antiviral activity in these primary cells as do raltegravir and azidothymidine (AZT). Moreover, the compound proved to be active against viral strains using either the CXCR4 (IIIB) or the CCR5 (BAL and YU2) chemokine co-receptor......We did not detect any significant cross-resistance of 6 with NRTI, NNRTI or entry inhibitor-resistant virus strains. Notably, 6 retained full activity against all five raltegravir-resistant strains tested, underlining its divergent mode of action. Furthermore, the IBD-resistant strain (first row in Table 2) with the A128T E170G double mutation was fully resistant to 6 but retained sensitivity to all other antiviral agents tested."
Frauke Christ1,7, arnout Voet2,7, arnaud Marchand5,7, stefan nicolet3,6, Belete a desimmie1, damien Marchand5, dorothee Bardiot5, nam Joo Van der Veken1, Barbara Van remoortel1, sergei V strelkov3, Marc de Maeyer2, patrick Chaltin4,5 & Zeger debyser1*
1laboratory for Molecular virology and Gene therapy, division of Molecular Medicine, Katholieke universiteit leuven (Kuleuven), leuven, belgium. 2laboratory for biomolecular Modelling, department of chemistry, division of biochemistry, Molecular and Structural biology, Kuleuven, Heverlee, belgium. 3laboratory for biocrystallography, department of pharmaceutical Sciences, Kuleuven, leuven, belgium. 4centre for drug design and development, Kuleuven R&d, Kuleuven, leuven, belgium. 5centre for innovation and Stimulation of Medicines development, leuven, belgium. 6M.e. Muller institute for Structural biology, biozentrum, university of basel, cH-4056 basel, Switzerland. 7these authors contributed equally to this work. *e-mail: email@example.com
16 May 2010
Lens epithelium-derived growth factor (LEDGF/p75) is a cellular cofactor of HIV-1 integrase that promotes viral integration by tethering the preintegration complex to the chromatin. By virtue of its crucial role in the early steps of HIV replication, the interaction between LEDGF/p75 and integrase represents an attractive target for antiviral therapy. We have rationally designed a series of 2-(quinolin-3-yl)acetic acid derivatives (LEDGINs) that act as potent inhibitors of the LEDGF/p75-integrase interaction and HIV-1 replication at submicromolar concentration by blocking the integration step. A 1.84-Å resolution crystal structure corroborates the binding of the inhibitor in the LEDGF/p75-binding pocket of integrase. Together with the lack of cross-resistance with two clinical integrase inhibitors, these findings define the 2-(quinolin-3-yl)acetic acid derivatives as the first genuine allosteric HIV-1 integrase inhibitors. Our work demonstrates the feasibility of rational design of small molecules inhibiting the protein-protein interaction between a viral protein and a cellular host factor.
Since the first description of AIDS in 1981, more than 25 million people worldwide have fallen victim to HIV infections. Despite the enormous efforts in developing new effective antiviral agents and the introduction of highly active antiretroviral therapy (HAART), the incidence of HIV infections continues to rise. Although HIV replication can be chronically suppressed with proper HAART, no cure is in sight. Therefore, the quest for new antiviral agents to complement existing treatment strategies remains one of the main goals in HIV drug discovery. Classical drugs target the viral enzymes reverse transcriptase, protease and integrase. Raltegravir (MK-518) interferes with the strand-transfer reaction of viral integrase (IN) and has recently been approved for clinical use1. Inhibition of integration by raltegravir is accompanied by an extremely strong reduction in viral load2. However, in contrast to previous predictions based on in vitro experimentation, raltegravir resistance evolves readily in the clinic3, necessitating the efforts to develop second-generation integrase inhibitors. Potent second-generation strand-transfer inhibitors with beneficial resistance profiles have not been described so far, to our knowledge. As integration is a highly organized multistep process that relies on cellular factors for completion4, we have taken an alternative approach to block the HIV integration step. Instead of targeting the catalytic activity of integrase, we intended to interfere with the integrase-LEDGF/p75 interaction5 and, consequently, block integration allosterically.
LEDGF/p75, a transcriptional coactivator that mediates stress response6, 7, 8, was initially identified as an integrase-interacting partner by coimmunoprecipitation from cells overexpressing HIV-1 integrase9. The crucial role of LEDGF/p75 in HIV replication was evidenced by mutagenesis, RNA interference (RNAi), transdominant overexpression of the integrase-binding domain (IBD) of LEDGF/p75 and knockout studies9, 10, 11, 12, 13, 14, 15, 16, 17. A chromatin-binding domain (PWWP) is located at the N terminus of LEDGF/p75, and the C terminus contains the IBD. LEDGF/p75 depletion impairs binding of integrase to chromatin18, 19. Together with the overall structure of LEDGF/p75, this finding supports the theory that this cofactor tethers the viral preintegration complex (PIC) to cellular chromatin. Analysis of integration sites in human cells depleted of LEDGF/p75 by RNAi or in embryonic fibroblasts derived from p75 knockout mice, corroborated this role of LEDGF/p75 as tethering and targeting factor of HIV14, 20.
As a proof-of-concept to exploit the LEDGF/p75-IN interaction as antiviral target, we overexpressed the C-terminal domain of LEDGF/p75 in human cells10. This LEDGF/p75 fragment, which lacked the chromatin-binding domain, could efficiently compete with the endogenous cofactor and inhibited HIV replication and integration to nearly undetectable levels. Moreover, by repeated passaging of HIV in cells overexpressing this LEDGF/p75 fragment, we selected a virus strain that was resistant to this phenotype12. Notably, two mutations in integrase rendered integrase resistant: A128T and E170G. As seen in the cocrystal structure of the IBD in complex with the integrase catalytic core domain21 both amino acids are crucial for the LEDGF/p75-IN interaction (Fig. 1). On the basis of this evidence, we embarked on a rational drug design program to discover LEDGINs, small molecules that target the LEDGF/p75-IN interaction, and to thereby inhibit HIV replication.
Virtual screening of inhibitors of LEDGF/p75-IN interaction
As a starting point, we used a limited set of 200,000 commercially available compounds. We filtered this set of compounds in several rounds, each time discarding the molecules that did not meet our criteria. In a first selection step, the filters were based on chemoinformatical rules defining chemical properties of small-molecule inhibitors of protein-protein interactions. We incorporated these filters into the MOE QSAR module (Fig. 1b). We then screened the remaining compounds (160,000) against our pharmacophore model, which we used as a query to select compounds having features compatible with the LEDGF/p75-binding pocket. We created this pharmacophore model by means of a thorough analysis of the different crystal structures of HIV-1 integrase and, more particularly, the complex of the IBD with the integrase core domain21 (Fig. 1a), which demonstrates a well-defined binding pocket for LEDGF/p75 at the interface of the integrase core dimer. The structure of tetraphenylarsonium bound in the LEDGF/p75-binding pocket22 and the crystal packing of IN core structures23 gave further valuable insight in the overall architecture of the binding pocket and the potential interactions to include in rational design (Fig. 1a).
On the basis of all this structural information, we created a consensus pharmacophore query for pharmacophore-based filtering (Supplementary Results). In a following step, we docked molecules that passed this second filter (2,000) into the binding pocket (reconstructed from the 1HYV crystal structure)2 using the docking program GOLD (3.0; default settings)24. We ranked the resulting solutions based on GoldScore, DrugScore25 and ChemScore26 scoring functions. The molecules scoring the best for all three scoring functions were visually inspected, and we retained those that matched the pharmacophore model after docking. Finally, we selected the 25 most promising compounds for biological evaluation.
Optimization of 2-(quinolin-3-yl)acetic acids
We acquired the 25 molecules commercially and evaluated their inhibitory activity against the LEDGF/p75-IN interaction using the AlphaScreen assay. Out of these 25 compounds, we identified a hit molecule (1) that showed 36% inhibition of the LEDGF/p75-IN interaction at 100 µM. By a similarity search using the MACCS daylight fingerprints with a similarity of higher than 75%, we selected several analogs of this hit compound. Once again, we tested the commercially acquired analogs in the AlphaScreen assay and determined an IC50 value for the most potent molecule (2, IC50 = 27.72 µM). At that stage, the chemical properties of this hit molecule were not fully satisfactory because of the presence of a ketimine moiety; therefore, we selected and evaluated new analogs that lack this chemically unstable functional group. Aside from the removal of the ketimine moiety, we were also interested to see whether the tetrazole moiety could be replaced by a known bioisostere, so we also selected some analogs bearing a carboxylic acid moiety. This process led to the identification of 2-(6-chloro-2-oxo-4-phenyl-1,2-dihydroquinolin-3-yl)acetic acid (3, IC50 = 12.2 ± 3.4 µM), which we then evaluated in a standard anti-HIV activity assay in cell culture. Compound 3 had a moderate antiviral activity (EC50 = 41.9 ± 1.1 µM) and did not affect cell viability (CC50 > 150 µM) (Table 1). In addition, we were able to soak 3 in a crystal of the IN core domain (Supplementary Fig. 2) and obtain information for further hit-to-lead optimization.
On the basis of the available information, we synthesized several close analogs of 3 and discovered three new compounds (4, 5 and 6) with increased activity in vitro and in cell culture, thereby achieving an approximately tenfold increase of activity in AlphaScreen (IC50 = 1.37 ± 0.36 µM) and a nearly 20-fold increase in antiviral activity (EC50 = 2.35 ± 0.28 µM) for the most active compound, 6. The interaction of LEDGF/p75 with neither JPO2 nor PogZ (cellular binding partners of LEDGF/p7527, 28, 29) was inhibited, suggesting binding of the compound to integrase and not to LEDGF/p75. Compound 6 did not affect integrase-DNA binding and only weakly inhibited the catalytic activities of integrase (Table 1 and Supplementary Table 1). These findings point to a different mechanism of action than that observed for the strand-transfer inhibitors raltegravir and elvitegravir1, 30. We detected no inhibitory activity of 6 against HIV-1 reverse transcriptase (data not shown).
Compound 6 inhibits HIV replication
The selectivity of 6 in cell culture (Table 1) allowed us to perform a thorough virological characterization. To evaluate the potential of this new antiviral and to proceed from bench to bedside, we determined its activity in primary isolates from human donors (primary peripheral blood mononuclear cells and macrophages) (Supplementary Table 3 and Supplementary Methods). Compound 6 retained antiviral activity in these primary cells as do raltegravir and azidothymidine (AZT). Moreover, the compound proved to be active against viral strains using either the CXCR4 (IIIB) or the CCR5 (BAL and YU2) chemokine co-receptor.
To unambiguously pinpoint a new antiviral target in the HIV replication cycle, a thorough analysis of the mode of action in cell culture is required31, 32. The strongest evidence can be obtained from the cross-resistance profile of a new compound in comparison with established antiviral agents and resistance selection against the new drug candidate. Time of addition (TOA) and quantitative PCR (Q-PCR) experiments give further valuable input. Table 2 shows the cross-resistance profile of 6. We analyzed 6 in parallel with drugs targeting other early steps of HIV replication (entry, reverse transcription and integration). Next to viral strains selected to be resistant to the integrase inhibitors raltegravir and elvitegravir, we analyzed strains resistant to the nucleoside reverse transcriptase inhibitor (NRTI) AZT, or to the non-nucleoside reverse transcriptase inhibitor (NNRTI) efavirenz, or to AMD3100, a CXCR4 chemokine receptor antagonist. Moreover, we also included a strain selected to be resistant to transdominant inhibition by the IBD of LEDGF/p7512. This strain carries two mutations (A128T and E170G) in the LEDGF/p75-binding pocket of IN. As expected, all strains were resistant to the compounds they had been selected with (Table 2, resistant strains indicated in bold type). In the case of the IN strand-transfer inhibitors (raltegravir and elvitegravir), we observed a more than tenfold resistance for strains containing multiple mutations in the integrase gene. We did not detect any significant cross-resistance of 6 with NRTI, NNRTI or entry inhibitor-resistant virus strains. Notably, 6 retained full activity against all five raltegravir-resistant strains tested, underlining its divergent mode of action. Furthermore, the IBD-resistant strain (first row in Table 2) with the A128T E170G double mutation was fully resistant to 6 but retained sensitivity to all other antiviral agents tested.
As HIV evolves rapidly, any given antiviral drug will select resistant strains, thus limiting the drug's efficacy33. In vitro selection of resistance against a new class of antiviral agents ultimately corroborates the antiviral target. To select a 6-resistant virus strain, we subjected HIV-1 NL4.3 to selective pressure in cell culture (Supplementary Methods). After six passages in cell culture, the single point mutation A128T was selected in the integrase gene in 13% of the virus population (Fig. 2a and 2b). As discussed above, A128T was one out of two mutations previously selected in the cell lines overexpressing the IBD12 and is located at the entrance of the LEDGF/p75-binding pocket. We did not detect any mutations in the envelope and reverse transcriptase genes nor in the long terminal repeats (LTRs). After 39 passages and at a compound concentration of 55 µM, which corresponds to a 25-fold excess of the 50% efficient concentration (EC50 = 2.35µM), we observed full conversion (Fig. 2a and 2b). The selected mutant strain and a molecular clone of HIV-1 NL4.3 with the A128T mutation showed no reduction in susceptibility to inhibition by raltegravir or AZT. Likewise, both raltegravir and AZT retained full activity against 6-resistant strains, proving once more the different mode of action of both integrase inhibitor classes (Fig. 2c). Probing the efficacy of 6 against different HIV-1 strains (NL4.3, HXB2D, IIIB), HIV-2 (EHO, ROD) and SIV (MAC251) revealed that, as for NNRTIs (efavirenz), the 2-(quinolin-3-yl)acetic acid derivatives are specific for HIV-1 strains (Supplementary Table 2). As rational design was based on the structural information of the LEDGF/p75-HIV-1 IN interaction, this did not come as a surprise. A128 is an important contact point at the interface, so we included it in the design of the pharmacophore model. In contrast to HIV-1 IN, both the HIV-2 (Fig. 2b) and SIV INs carry a methionine instead of an alanine at this position, preventing binding of the 2-(quinolin-3-yl)acetic acid derivatives. As expected, 6 did not inhibit the catalytic activity of either HIV-2 or A128T HIV-1 IN (Supplementary Table 1). This demonstrates that the modest inhibition of integrase activity requires direct binding to the LEDGF/p75-binding pocket and supports the idea that these compounds function as allosteric inhibitors of HIV-1 integrase activity.
Crystal structures of 3 and 6 in LEDGF/p75-binding pocket
Although the full-length HIV-1 IN structure has not been resolved, several groups have described the crystallization of the HIV-1 IN core domain22, 23, 35. To validate our pharmacophore model, we soaked IN core crystals with 3 and 6 and carried out X-ray structure determination (Supplementary Table 5 and Supplementary Methods). Both crystal structures reveal a well defined electron density for all atoms of the compound (Fig. 4a and Supplementary Fig. 3). The compounds are bound in a cleft between the two monomers of the IN core dimer. Notably, their binding modes in the crystal are almost identical to those predicted by molecular modeling. For instance, the root mean square (r.m.s.) deviation between the predicted and observed atom positions for 6 is 1.2 Å. Upon comparing the 3 and 6 structures, it becomes apparent that the extra propyl group in 6 is readily accommodated in an appropriate void deep inside the cleft that is, behind the bound compound (Fig. 4a), just as designed.
Furthermore, overlaying the compound-bound structures with the IBD-IN core cocrystal structure21 confirms that the compound occupies the same space in the LEDGF/p75-binding pocket; thus, the binding is mutually exclusive (Fig. 4b). In particular, the main chain nitrogens of residues Glu170 and His171 of integrase form hydrogen bonds with the carboxyl moiety of 6. In the IBD-IN core complex, equivalent hydrogen bonds are formed with the side chain of Asp366 located in the LEDGF/p75 IBD loop (Fig. 4c). Furthermore, the apolar side chain of Ala128 of integrase packs directly against the 6 compound, so that the Cβ atom of Ala128 is located 3.6 Å away from the sole chlorine atom of 6 and at an equal distance of 3.8 Å from the nearest atom of either the double ring or the phenyl moiety (Fig. 4a). Finally, when comparing the new structures with the IBD-IN core complex, no noticeable expansion or contraction of the binding pocket is found, as the r.m.s. deviation of the protein backbone forming the pocket is only 0.25 Å. Moreover, the binding of the compounds does not lead to any significant conformation changes of the side chains in the pocket. Minor side chain changes can be observed for the Gln95 and the Gln168, but these are also present in the different crystal structures21, 22, 23 of the IN core and do not interact directly with the bound compounds.
As discussed above, the HIV-1 mutant A128T and the HIV-2 IN, which carries a methionine residue in this position, are resistant to 6. The crystal structure hints at an explanation for this observation. Indeed, mutating this residue in silico to incorporate a threonine or methionine side chain in the most abundant rotamer results in a substantial steric clash, as in either case the Cγ atom is located only 2.5 Å from the chlorine atom. These insights can be exploited to create a new pharmacophoric model in such a way that new compounds can be designed to avoid steric hindrance by A128T or to even have a beneficial influence by this mutation. Adding volumetric restrictions to the A128 area in the pharmacophore query will take this into account and will probably select for new compounds that overcome the HIV-2 resistance and restraining resistance selection in cell culture.
These findings ultimately indicate that 6 is a well-optimized protein-protein interaction inhibitor of LEDGF/p75 binding acting at an allosteric site of HIV-1 integrase.
Design of more potent 2-(quinolin-3-yl)acetic acids
While the virological profiling of 6 was in progress, we also carried out chemical synthesis of more congeners to broaden the structure-activity relationship (SAR) and to increase the potency of the 2-(quinolin-3-yl)acetic acids. One such compound that we designed and synthesized was 7 (Supplementary Table 4). Compound 7 inhibited the LEDGF/p75-integrase interaction in the submicromolar range (IC50 = 0.58 ± 0.30 µM). Concomitantly, the anti-HIV activity increased to the same extent (EC50 = 0.76 ± 0.08 µM), resulting in a selectivity index of nearly 100.
Because of their limited genome size, viruses in general, and HIV in particular, must rely on host proteins, referred to as cofactors, to achieve the multiple steps in HIV replication. Such protein-protein interactions (PPIs) between viral proteins and cofactors constitute a pool of potential new antiviral targets against which small-molecule protein-protein interaction inhibitors (SMPPIIs) can be designed. We have now designed and synthesized a new class of antiviral agents that effectively target the LEDGF/p75-IN interaction, not only demonstrating the feasibility of exploiting PPIs as targets for drug discovery but also validating a new paradigm in anti-HIV research.
Genome and proteome analyses highlight the key role of PPIs in biological processes; thus, PPIs represent a large and important class of therapeutic targets36, 37. Nevertheless, these interactions were left aside for some time because drug development against targets such as enzymes and receptors has so far been more cost effective. However the expense of developing these classical drugs is increasing, and new target proteins and/or PPIs have come within reach. X-ray structures have shown that much of the surface area of a typical protein-protein interface is buried, with the atoms packed closely together, implying that only a few cavities are available for small-molecule binding. As shown in Figures 1 and 4, the LEDGF/p75-IN interaction is an example of a well defined cavity in one interacting partner (integrase) penetrated by a loop of the other partner (LEDGF/p75). This constellation encouraged us to search for LEDGINs, small molecules competing for the binding of LEDGF/p75 to HIV-1 integrase.
Structural biology helps greatly in our understanding of how a small molecule can compete directly with the natural protein inhibitor. A classical example of a novel drug target in cancer treatment is the human protein double minute 2 (HDM2)38. HDM2 binds to the tumor suppressor protein p53 and increases its degradation. Inhibition of this interaction can stimulate p53 activity in wild-type cancer cells, driving them toward apoptosis. X-ray crystallography of the p53-MDM2 complex reveals that this PPI is primarily mediated by a few key amino acids of p53 and a small but deep hydrophobic cleft in MDM2 (refs. 39,40). Nutlins are a series of tetrasubstituted imidazoles that potently disrupt the MDM2-p53 interaction. It was observed that nutlins mimic the binding of the helical region of p53 by interacting with the hydrophobic cleft of MDM2. Furthermore, these compounds are active against tumor xenografts in vivo41. Analogously to nutlins, the specific features of the LEDGF/p75-IN interaction with the interhelical loop protruding in the integrase core pocket may explain our success in developing LEDGINs with antiviral activity.
Because the specific interaction between a cellular cofactor and a viral protein is targeted, inhibition ought not to be associated with cellular toxicity. In fact, we have not observed prominent cellular toxicity with any of the analyzed compounds.
An earlier report described the discovery of a small molecule (D77) capable of interrupting the LEDGF/p75-IN interaction and potentially inhibiting HIV replication in cell culture42. Furthermore, we have reported a new class of compounds that inhibit the LEDGF/p75-IN interaction in vitro43. In both reports, no evidence was provided that the described molecules act on the LEDGF/p75-IN interaction in cell culture. In our hands, D77 furthermore rather proved to be toxic in cell culture (F.C. and B.V.R., unpublished data).
Here we present a new class of experimental antiviral agents that target the LEDGF/p75-IN interaction. Multiple lines of evidence pinpoint the molecular mechanism of inhibition of the 2-(quinolin-3-yl)acetic acid derivatives to a block in integration via an inhibition of the interaction of IN with LEDGF/p75. LEDGF/p75 by itself is known to act as an allosteric activator of IN activity44. Therefore, it does not come as a surprise that the 2-(quinolin-3-yl)acetic acid derivatives Ñ aside from their inhibition of the LEDGF/p75-IN interaction Ñ also moderately inhibit the enzymatic activity of HIV-1 IN. Although all compounds tested so far were more active on the cofactor binding than on the enzymatic activity, this allosteric function might add to the potent antiviral profile of this compound class.
Looking at the high-resolution cocrystal structure and the chemical nature of 6, both pharmacological properties and antiviral potency of the present compounds can still be optimized. In fact, more potent congeners with submicromolar antiviral activity have already been identified by us (see the activity of 7 in Supplementary Table 4). After our initial discovery, Boehringer Ingelheim has reported the anti-HIV activity of some 2-(quinolin-3-yl)acetic acid derivatives in a patent application45. It is likely that these compounds share a similar mechanism of action as that disclosed here. Our work reinforces a general approach to dissecting the interplay of HIV with its intracellular cofactors in the search of new targets for antiviral therapy. It stresses again the feasibility of developing small-molecule PPI inhibitors. Along with the clinical potential of an allosteric integrase inhibitor to cope with raltegravir resistance, this new class of inhibitors will also aid the elucidation of the HIV-1 replication cycle in further detail.
Computer-aided drug design.
Chemoinformatical analysis and filtering of compounds as well as pharmacophore modeling was done using the program MOE (Chemical Computing Group). Using Omega2 (OpenEye Scientific Software), we generated three-dimensional conformations of small molecules. Pharmacophore analysis of protein-ligand complexes was done using LigandScout 1.0 (Inte:Ligand). Small molecules were docked into the protein receptor using Gold3.0 (The Cambridge Crystallographic Data Centre). All computations were performed on an Apple G5 Xserve cluster and Intel Dualcore2 3.40GHz workstations.
Compounds 1, 2 and 3 were purchased from Interbioscreen. Compounds 4, 5, 6 and 7 were synthesized by CISTIM Leuven Vzw under an agreement with CD3. Compound 6 (2-(6-chloro-2-methyl-4-phenylquinolin-3-yl)pentanoic acid) was obtained according to the following procedure (see Supplementary Figure 2 for details): (2-amino-5-chlorophenyl)(phenyl)methanone and levulinic acid were condensed in a mixture of sulfuric acid and acetic acid. After protection of the carboxylic acid as a methyl ester, the n-propyl side chain was introduced by nucleophilic substitution of allylbromide and hydrogenation of the terminal double bond in the presence of a palladium catalyst (Pd/C). The deprotection of the methyl ester, performed in basic conditions (10 M NaOH in DMSO), afforded the title compound as a white solid.
The AlphaScreen assay was performed according to the manufacturer's protocol (Perkin Elmer, Benelux). Reactions were performed in a 25 µl final volume in 384-well Optiwell microtiter plates (Perkin Elmer). The reaction buffer contained 25 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM MgCl2, 0.01% (v/v) Tween-20 and 0.1% (w/v) BSA. His6-tagged integrase (300 nM final concentration) was incubated with the compounds for 30 min at 4 °C. The compounds were added at varying concentrations spanning a wide range from 0.1 µM to 100 µM. Afterwards, 100 nM Flag-LEDGF/p75 was added and incubation was prolonged for an additional 1 h at 4 °C. Subsequently, 5 µl of nickel chelate-coated acceptor beads and 5 µl anti-Flag donor beads were added to a final concentration of 20 µg ml-1 of both beads. Proteins and beads were incubated for 1 h at 30 °C to allow association to occur. Exposure of the reaction to direct light was omitted as much as possible and the emission of light from the acceptor beads was measured in the EnVision plate reader (Perkin Elmer, Benelux) and analyzed using the EnVision manager software.
X-ray structure determination.
Integrase core crystals were grown for 3 d at 4 °C to a maximal size of 200 µm using the hanging drop technique with reservoir solution containing 10% (w/v) PEG8000, 0.1 M sodium cacodylate, pH 6.5, 0.1 M (NH4)2SO4, 5 mM DTT (conditions modified from ref. 22). Drops were composed of 1 µl of recombinant HIV-1 IN core domain protein at 4.4 mg ml-1 and 1 µl of the reservoir. Obtained crystals were soaked for 12 h at 4 °C in the crystallization solution supplemented with 8 mM inhibitor that had been solubilized in 0.1 M DMSO. Crystals were then transferred in a 0.1 M sodium cacodylate, pH 6.5, 20% (w/v) PEG8000, 25% (w/v) PEG200, 0.2 M (NH4)2SO4 and 5 mM DTT solution and flash-frozen in liquid nitrogen.
Diffraction data were collected at 100K using 1.0-Å radiation at the X06DA beamline of the Swiss Light Source, Paul Scherrer Institut, Villigen, Switzerland, using a Mar225 CCD detector, indexed with iMosflm46 and scaled with Scala46. The structure PDB 1HYV22 (without tetraphenylarsonium compound) was used as a starting point for the structural refinement in Refmac 5.5 (ref. 46). Although the integrase core domain was crystallized essentially as described in ref. 22, we observed small differences in the integrase residues that could be located in the electron density maps. Our structure includes residues 55-145, 153-188 and 193-209, whereas the structure from ref. 22 includes residues 57-140, 148-189 and 193-210. The r.m.s. deviation between both structures is 0.315 Å. The inhibitor structure was built into the difference (Fo - Fc) map using Coot46.