In search of authentic inhibitors of HIV-1 integration

Transcription

1 Antiviral Chemistry & Chemotherapy 13:1 15 Review In search of authentic inhibitors of HIV-1 integration Zeger Debyser*, Peter Cherepanov, Bénédicte Van Maele, Erik De Clercq and Myriam Witvrouw Rega Institute for Medical Research, KULeuven, Leuven, Flanders, Belgium *Corresponding author: Tel: ; Fax: ; Current strategies for the treatment of HIV infection are based on cocktails of drugs that target the viral reverse transcriptase or protease enzymes. At present, the clinical benefit of this combination therapy for HIV-infected patients is considerable, although it is not clear how long this effect will last taking into account the emergence of multiple drug-resistant viral strains. Addition of new anti- HIV drugs targeting additional steps of the viral replication cycle may increase the potency of inhibition and prevent resistance development. During HIV replication, integration of the viral genome into the cellular chromosome is an essential step catalysed by the viral integrase. Although HIV integrase is an attractive target for antiviral therapy, so far all research efforts have led to the identification of only one series of compounds that selectively inhibit the integration step during HIV replication, namely the diketo acids. In this review we try to address the question why it has proven so difficult to find potent and selective integrase inhibitors. We point to potential pitfalls in defining an inhibitor as an authentic integrase inhibitor, and propose new strategies and technologies for the discovery of authentic HIV integration inhibitors. Keywords: AIDS, HIV, integrase, antiviral Introduction Considerable progress has been made in the treatment of patients infected with HIV. Combination regimens that include potent reverse transcriptase (RT) and protease inhibitors (PIs) are in clinical practice (De Clercq, 1995; 2001). Highly active antiretroviral therapy (HAART) is of great benefit for most patients, although long-term therapeutic success may be jeopardized by the emergence of drug-resistant virus strains. Therefore, it remains essential to develop drugs targeted at alternative steps of the viral replication cycle. The integration of retro-transcribed viral DNA into the host chromosome is an essential step in the replication cycle of retroviruses. After integration the proviral DNA is replicated and genetically transmitted as part of the cellular genome. Thus, integration defines a point of no return in the life cycle of HIV. The integrase of HIV is an attractive target for selective antiviral therapy since there is no known functional homologue in human cells (LaFemina et al., 1992). However, a decade of HIV integrase research has not resulted in the development of an integrase inhibitor approved or under consideration for clinical use. In our description of the biochemistry of the integration process we will accentuate the unknown and indicate new antiviral targets for drug development. We describe the pitfalls of currently used integrase assays, and the solutions and technologies to circumvent misidentification of integrase inhibitors. Biochemistry of HIV-1 integration HIV-1 integration encompasses a series of molecular events that follows the completion of reverse transcription in the cytoplasm of the infected cell and that ends with the initiation of transcription from the inserted proviral DNA. An overview of these events during retroviral DNA integration is shown in Figure 1 (for a review on the molecular biology of integration see Brown, 1997). The DNA product formed after reverse transcription is a double-stranded linear DNA with Long Terminal Repeat (LTR) sequences at each end. This molecule is the substrate for HIV-1 integrase that will catalyze and facilitate several steps that finally result in the insertion of the viral DNA into the chromosome of the host cell. It is not known at which point during the reverse transcription process the integrase associates with the retrotranscribed DNA. Since integrase binds non-specifically to DNA, it is possible that the enzyme interacts with the first stretch of double-stranded DNA formed, namely the 3 LTR. HIV-1 integrase has been found to oligomerize in the test tube upon DNA 2002 International Medical Press /02/$

2 Z Debyser et al. Figure 1. utline of the integration reaction in vivo +pgt TG AC CA-H GTCA Host chromosomal DNA AC TG AC Linear viral DNA CAGT-3 GTCA-5 p 5 -ACTG 3 -TGAC p ACTG H-AC Integrated HIV-1 3 -processing Strand transfer CA GT +pgt Repair of remaining DNA CA GT The integration reaction is concerted; both viral DNA ends are inserted into to the host chromosomal DNA at the same time. In the case of HIV-1, the distance between the integration sites of both ends is always 5 bp apart. Repair of the remaining gaps in the chromosomal DNA results in a 5 bp duplication of the host cell genome. The viral ends are probably bridged by cellular host-factors, like BAF or HMGI(Y). binding (Vercammen et al., unpublished observations). It is therefore plausible that an active oligomer forms at the 3 LTR; upon completion of the 5 LTR a second oligomer may form. How this oligomer is positioned at the 5 end and its exact stoichiometry are unknown, although the enzyme is believed to function at least as a tetramer. There is evidence for a physical and a functional interaction between HIV-1 reverse transcriptase and integrase (Wu et al., 1999; Tasara et al., 2001). This may help correct positioning of integrase. The integrase-mediated reactions take place in the cytoplasm in a subviral particle first referred to as RT complex, later as pre-integration complex (PIC). It is clear that there is a continuity between these complexes. The first catalytic step takes place in the cytoplasm of the infected cell, and removes the terminal GT dinucleotides from each LTR 3 end. This site-specific endonucleolytic activity is referred to CA as the 3 -processing reaction. After the PIC is transported through the nuclear pore, the viral DNA is joined to the cellular DNA. This strand-transfer step includes a staggered cleavage of host DNA, and a ligation of the processed 3 -H ends of the viral DNA. Chemically this involves a nucleophilic attack by the 3 -H group generated after processing and a concomitant transesterification to the 5 -phosphate of the cleaved cellular DNA. The sites of integration in the two target DNA strands are separated by five nucleotides. The 3 -ends of the target DNA remain unjoined after the strand transfer. The integration reaction is completed by the removal of the two unpaired nucleotides at the 5 -end of the viral DNA and the repair of the single stranded gaps created between the viral and target DNA. This step is probably carried out by cellular DNA repair enzymes (Yoder & Bushman, 2000), although a possible role of RT and IN (Chow et al., 1992) has been proposed. Staggered strand transfer and gap repair result in the duplication of host cell sequences immediately flanking the inserted proviral DNA. Structure of HIV-1 integrase The HIV-1 integrase is a 32-kDa protein, composed of 288 amino acids that can be separated in three domains (Figure 2). The amino-terminal domain, the catalytic core and the carboxy-terminal domain fold independently. Highly conserved among the integrases of retroviruses, retrotransposons, and even many prokaryotic transposable elements, are the zinc finger motif in the N-terminal domain and the DD(35)E motif in the catalytic core (Kulkosky et al., 1992). Because of the relative insolubility of recombinant full-length integrase, the structures of the individual domains have been determined separately. The N-terminal domain (residues 1 50) is characterized by the HHCC motif containing two histidines and two cysteines. In vitro binding studies with the N-terminal part of IN failed to show affinity for LTR-specific or non-specific DNA (Mumm & Grandgenett, 1991); Bushman et al., 1993); Vink et al., 1993). The N-terminal part was shown to bind zinc (Burke et al., 1992), and, although the zinc finger is not responsible for DNA binding, it is required for the 3 -processing and strand transfer activity (Engelman & Craigie, 1992; van Gent et al., 1992; Vink et al., 1993), but not for the disintegration reaction (Chow et al., 1992). Elucidation of the structure by NMR revealed the putative zinc finger of IN to be composed of three α-helices (Cai et al., 1997). This structure is stabilized by the zinc-binding pocket through Zn 2+ -metal coordination. The binding of zinc ions to the N-terminal domain apparently promotes tetramerisation of full-length integrase in vitro and may also stimulate catalytic activity (Zheng et al., 1996; Lee et al., 1997) International Medical Press

3 Inhibitors of HIV-1 integration Figure 2. Schematic representation of the individual domains of HIV-1 integrase N-terminus Central core C-terminus HH CC D(64) D(116) E(152) Binds Zn 2+ Retroviruses Retrotransposons Polynucleotidyl transfer Retroviruses Retrotransposons Transposons Binds DNA The main properties of each domain are given below. The homology with other polynucleotidyl transfer enzymes is also presented. The catalytic core (residues ) of integrase is found in the protease-resistant central domain. It contains the highly conserved catalytic triad DD(35)E. These residues are required for catalysis by coordination of divalent metal ions (Mg 2+ or Mn 2+ ). The conserved acidic residues are essential for both processing and strand transfer activity, supporting a model in which the central region of integrase encodes a single catalytic core serving both reactions (Khan et al., 1991; Kulkosky et al., 1992). The location of the catalytic site in the central region is further demonstrated by the disintegration activity of the central core (Engelman & Craigie, 1992; Kulkosky et al., 1992; Leavitt et al., 1993). The crystal structure of a soluble mutant of the isolated core domain of HIV-1 IN has been resolved (Dyda et al., 1994). The monomeric structure contains five β-sheets, surrounded by six α-helices, typical for all members of the polynucleotidyl transfer enzymes, like HIV-1 RNaseH (Davies et al., 1991) or bacteriophage MuA transposase (Rice & Mizuuchi, 1995). D64 and D116 of the DD(35)E motif are in close proximity, but E152 is located on a disordered loop. The two active sites in the dimer are too far apart to permit five base pair staggered cleavage of the target DNA, suggesting that IN functions as at least a tetramer (Rice et al., 1996; Cai et al., 1997). A diketo derivative (5-CITEP) was co-crystallized with the HIV-1 integrase core domain. The inhibitor bound to a site spanning both the DDE motif and the mononucleotide-binding site (Goldgur et al., 1999). The C-terminal domain (residues ) of integrase is the domain that shows the least sequence conservation among integrases (Lutzke et al., 1994; Cannon et al., 1996). It is essential both for processing and for strand transfer, but not for disintegration. The C-terminal domain (residues ) binds non-specifically to DNA and forms a dimer of parallel monomers, as shown by NMR (Eijkelenboom et al., 1995; Lodi et al., 1995). The structure of each monomer consists of five anti-parallel β strands that fold into a β-barrel, strikingly similar to the SH3 domain (Src homology 3). It contains a saddle-shaped groove that might accommodate double-stranded DNA. K264; an important DNA-binding residue is located within this groove (Lutzke et al., 1994; Lodi et al., 1995; Eijkelenboom et al.,1995). Although structural information about each individual domain is now available, the spatial arrangement of the three domains within one IN monomer, and the individual positioning into a multimeric structure remain to be resolved. Chen et al. (2000) were able to obtain crystallization of a two-domain integrase protein. The structure of the core, linked to the C-terminus was resolved at 2.8 Å. The structure of the core in this Y-shaped dimer was identical to the structure resolved by Dyda et al. (1994). Recently, the structure of a two-domain fragment composed of an N-terminus and central core has been elucidated (Wang et al., 2001). The elucidation of the complete structure of HIV-1 IN would be very useful for rational drug design based on molecular modelling. The main obstacle so far is the poor solubility of the enzyme. Based on the two resolved two-domain structures a plausible model for a full-length integrase tetramer has been put forward (Wang et al., 2001). Diketo acids, the only authentic integrase inhibitors identified so far, only bind to integrase after the enzyme has formed a complex with target DNA (Espeseth et al., 2000). It follows that a cocrystal of IN with DNA would be required to reveal the binding site of the diketo acids. Cofactors of integration as potential antiviral targets Cofactors The retro-transcribed viral DNA is associated with a dense nucleoprotein complex derived from the virion core. This nucleoprotein complex is called the pre-integration complex (PIC) and is competent to insert endogenous cdna into a target DNA supplied in vitro (Ellison et al., 1990; Farnet & Haseltine, 1990). Functional HIV-1 PICs were found to contain integrase, reverse transcriptase, matrix protein p17 and small amounts of capsid protein p24 (Bukrinsky et al., 1993; Miller et al., 1995). Viral DNA is compacted to fit in the PIC. Since the sites of HIV cdna integration in each target DNA are always separated by a defined 5 bp distance, it seems likely that the cdna ends are held together by proteins. Although some cofactors have been identified, their respective roles and mechanism are less well understood. The yeast two-hybrid system was used to isolate a cellular binding partner for HIV-1 IN (Kalpana & Goff, 1993; Kalpana et al., 1994). Integrase Antiviral Chemistry & Chemotherapy 13:1 3

4 Z Debyser et al. interactor 1 (Ini1) has amino acid similarity with the yeast transcriptional activator SNF5, a component of the multiprotein SWI/SNF complex (Wang et al.,1996). This complex activates transcription by remodelling the chromatin. Complexes similar to yeast SWI/SNF have been isolated from mammalian cells. They were demonstrated to contain a similar set of protein components as the yeast complex. Ini1 is part of the mammalian SWI/SNF complex. Because Ini1 directly interacts with HIV IN and is involved in chromatin remodelling, it was put forward that Ini1 may target the retroviral integration machinery to open chromatin regions (Kalpana et al., 1994). The minimal integrase interacting binding domain has been characterized, as well as the minimal domain responsible for the binding to DNA (Morozov et al., 1998). Ini1, however, was not detected in the pre-integration complex. Recently, it was claimed that Ini1 is specifically incorporated into HIV-1 virions and that expression of a cytoplasmically located fragment of Ini1 potently inhibited HIV-1 particle production and replication in a transdominant manner, suggesting a role during late events of the viral life-cycle (Yung et al.,2001). In contrast, incoming retroviral PICs were shown to trigger the exportin-mediated cytoplasmic export of Ini1, implying a role during early steps of HIV replication (Turelli et al., 2001). These paradoxical data await full explanation. While developing methods to purify PICs from freshly infected cells, Farnet and Bushman (1997) identified HMGI(Y) as a factor important for integration in vitro. This is a non-histone chromosomal protein important for transcriptional control and chromosomal architecture. Addition of HMGI(Y) (or Ini1) stimulates in vitro integration reactions with recombinant integrase (Kalpana et al., 1994; Farnet & Bushman, 1997). However, the function of each host factor during in vivo integration remains to be determined. In PIC in vitro integration assays, there is a strong preference for intermolecular integration, instead of intramolecular integration. Lee and Craigie (1998) were able to identify the barrier to auto-integration factor (BAF), a polypeptide of 89 amino acids. BAF was able to restore the integration activity of salt-stripped PICs at nanomolar concentrations, whereas for HMGI(Y) micromolar concentrations were required (Chen & Engelman, 1998). BAF is a highly evolutionarily conserved DNA binding protein, and its structure has been resolved by NMR (Cai et al., 1998). BAF is largely helical, and each subunit is composed of five helices. Recently, the crystal structure of homodimeric BAF was resolved at a 1.9 Å resolution, and it was determined that the fold of the BAF monomer resembles that of the second domain of RuvA of Escherichia coli, a protein involved in processing Holliday junctions formed during recombination and DNA repair (Umland et al., 2000). This comparison revealed the presence of the helix-hairpin-helix non-specific DNA binding motif within BAF. The presence of this DNA binding motif explains how BAF could bridge dsdna both intra- as well as intermolecularly. Nuclear import Lentiviral integration requires transport of the pre-integration complex through the nuclear membrane pores via an active, energy-dependent process (Bukrinsky et al., 1992). The role of various viral proteins in this process is not well understood. At least three components of the HIV-1 preintegration complex seem to contribute to nuclear transport: Vpr (Heinzinger et al., 1994; Fouchier et al., 1998; Popov et al., 1998), matrix protein (MA) (Bukrinsky et al., 1993; Gallay et al., 1995) and integrase (Gallay et al., 1997). These different proteins interact with each other and with the cellular import machinery. There is, however, no consensus on the minimum requirements for import of the PIC into the nucleus and the results seem to depend on the experimental conditions used. Whereas the karyophilic properties of MA have been questioned recently (Depienne et al., 2000), those of integrase have been confirmed. We constructed an N-terminal fusion of IN with GFP (Pluymers et al., 1999). A reasonable amount of GFP-IN was expressed and it localized in the nucleus. A similar study was reported by Depienne et al. (2000). Although it was originally proposed that the interaction between IN and karyopherin-α, that is required for nuclear import, is mediated by the recognition of a bipartite nuclear localization signal in the integrase protein (Gallay et al., 1997), it was found later, using digitonin-permeabilized cells, that nuclear accumulation of IN did not involve karyopherinmediated pathways (Depienne et al., 2001). A non-canonical NLS within the catalytic core domain of HIV-1 IN has been described recently (Bouyac-Bertoia et al., 2001). The authors claimed this NLS to be essential for productive infection by HIV. Interestingly, virus defective in Vpr and in the NLS of MA can still infect γ-irradiated P4 cells (Gallay et al., 1997), pointing as well to an important role of integrase in nuclear import of HIV. However, IN is not the only factor involved in nuclear import of the proviral DNA (Petit et al., 2000) many proteins probably act synergistically to ensure nuclear import through the nucleopore. Finally, not only viral proteins play a role. The central DNA flap seems important as well for the import of viral DNA into the nucleus (Zennou et al., 2000). During reverse transcription a central strand displacement event, consecutive to central initiation and termination of plus strand synthesis, creates a plus strand overlap, the central flap. Mutant viral DNA, which lacks this DNA flap, accumulates in infected cells as unintegrated linear DNA, at the vicinity of the nuclear membrane (Zennou et al., 2000). Whatever the mechanism of nuclear transport of HIV-1 DNA might be, it would be interesting to investigate in International Medical Press

5 Inhibitors of HIV-1 integration Figure 3. Analysis of the integration reaction using radiolabelled oligonucleotides CA Separate on a denaturating polyacrylamide gel Strand transfer products CAG CA +GT Substrate Cleaved Compound X Compound Y A schematic representation (left) and an autoradiogram (right) of the integrase-mediated reactions. A 20-bp oligonucleotide substrate consisting of the terminal U5 LTR sequence is used to evaluate the integrase-mediated reactions. In this experiment, inhibition of the overall integration by a dilution series of compound X and Y was evaluated. The 3 -processing reaction cleaves off the terminal GT dinucleotide (formation of the 2 band, black arrow) and this cleaved substrate will be covalently inserted into another DNA oligonucleotide in the strand transfer reaction (higher bands, grey arrow). more detail the minimal necessities for the import. Most probably, the various protein and DNA signals can mediate import of the PIC through the nuclear membrane. In vivo these import signals may act synergistically. ligonucleotide-based integrase assays used to identify integrase inhibitors The use of synthetic oligonucleotides, that represent the terminal sequences of the viral DNA, and recombinant HIV-1 integrase, expressed in E. coli, has led to an in vitro test system for integration (Bushman et al., 1990; Craigie et al., 1990; Katz et al., 1990; Sherman & Fyfe, 1990). The assay proved to be useful to investigate in detail the reactions catalyzed by the enzyme (for a schematic representation of the in vitro integration assay see Figure 3). In vitro, an apparent reversal of the strand transfer joining reaction, called disintegration, can be catalyzed by the core domain of integrase. Disintegration has not been shown to occur in vivo. The in vitro assays have been reviewed recently (Debyser et al., 2001; Marchand et al.,2001). Integrase inhibitors with mistaken identity A vast series of compounds have been reported to inhibit the integrase activity in oligonucleotide assays (Pommier & Neamati, 1999; Neamati et al., 2000; Pommier et al., 2000) (Figure 4), but so far only for the diketo acids (Hazuda et al., 2000) was it unambiguously shown that inhibition of the viral replication was due to interference with the integration step. Most integrase inhibitors do not show antiviral activity or are too toxic in cell culture. Moreover, for two classes of integrase inhibitors with antiviral activity, our group has demonstrated that viral entry, and not integration, is the predominant antiviral target in cell culture (Esté et al., 1998, Pluymers et al., 2000). G-quartets G-quartets are oligonucleotides that can form a highly stable intramolecular four-stranded DNA structure containing two stacked guanosine-quartets (G-quartets) (jwang et al., 1995). The prototypical AR177 (Zintevir), a 17-mer composed of deoxyguanosine and deoxythymidine residues Antiviral Chemistry & Chemotherapy 13:1 5

6 Z Debyser et al. Figure 4. Structures of HIV-1 integrase inhibitors H2C H3C CH C H 3 C H 3 C CH CH 3 S N S H 3 C CH 3 H H C C H CH 3 H Integric acid Naphtothiazepine Granulatine H HC CH H H H H H NH HN L-chicoric acid Salicylhydrazide Cl H H N N N N NH H CAPE 5-C ltep F N CH L-731,988 Integric acid (Hazuda et al., 1999), naphtothiazepine (Neamati et al., 1999), granulatine (Neamati et al., 1997), caffeic acid phenylethyl ester (CAPE) (Fesen et al., 1994; Neamati et al., 1997), salicylhydrazide (Zhao et al., 1997; Neamati et al., 1998), L-chicoric acid (Robinson et al., 1996), 5-CITEP, 1-(5-chloroindol-3-yl)-3-hydroxy-3-(2H-tetrazol-5-yl)-propenone (Goldgur et al., 1999) and the diketo acid L-731,988 (Hazuda et al., 2000). nly diketo acids are authentic inhibitors of HIV-1 integration. (Rando et al., 1995), is a potent inhibitor of HIV replication in cell culture (jwang et al., 1995). Although G- quartets inhibit HIV IN activity in the nanomolar range (Mazumder et al., 1996; Cherepanov et al., 1997), we have clearly shown by selecting and sequencing drug-resistant HIV-1 strains that the antiviral activity in cell culture is due to the inhibition of viral entry rather than integration (Cherepanov et al., 1997; Esté et al., 1998). L-chicoric acid The dicaffeoyl quinic acid and L-chicoric acid (L-CA) derivatives inhibit HIV-1 IN in vitro and HIV International Medical Press

7 Inhibitors of HIV-1 integration replication in cell culture (Robinson et al., 1996a,b). An HIV-1 strain resistant to L-chicoric acid was selected, and the resistance mutation was mapped in the IN gene (G140S) (King & Robinson, 1998). We have selected HIV-1 strains that are resistant to L-CA. ur strains acquired mutations in the viral envelope glycoprotein gp120 but not in the integrase gene. Recombinant virus strains that carried the env gene of the resistant strain were also resistant to L-CA. Moreover, recombinant integrase carrying the G140S mutation was inhibited to the same extent by L-CA as the wild-type enzyme (Pluymers et al., 2000). Authentic integrase inhibitors Diketo acid derivatives inhibit HIV-1 replication at micromolar concentrations through a specific inhibition of the DNA strand transfer step (Hazuda et al., 2000). Drugresistant HIV strains were selected and shown to carry mutations in the integrase gene. In recombinant integrase, these mutations conferred partial resistance. The diketo derivative 5-CITEP was cocrystallized with the HIV-1 integrase core domain (Goldgur et al., 1999). Some diketo acids have been shown to only bind to integrase after the enzyme has bound target DNA (Espeseth et al., 2000). Therefore the relevance of the cocrystal model should be questioned because no target DNA was present at the time of crystallization. The diketo acids were confirmed to be authentic integrase inhibitors, owing their antiviral effect to an inhibition of the integration process during viral replication (W Pluymers et al., unpublished observations). More potent congeners were synthesized more recently, capable of inhibiting HIV-1 replication in the nanomolar range (Wai et al., 2000). Why is it so difficult to find selective integrase inhibitors? Many characteristics of the biochemistry and the molecular biology of integration render it a process difficult to target effectively with classical drugs. In sharp contrast to reverse transcriptase that is catalyzing a polymerization reaction, 3 processing and DNA strand transfer reactions are carried out only once at each LTR-end during the normal replication cycle. Integrase inhibition is thus an all-ornone event. However, the identification of the diketo acids as authentic inhibitors of DNA strand transfer has provided a proof-of-principle that true inhibitors of integration, acting in cell culture, can be found. The catalytic activities leading to integration take place in a DNA-protein complex, the pre-integration complex. Steric hindrance may limit accessibility of drugs to the catalytic sites. In this complex, integrase is tightly associated with the processed DNA substrate. There might be a short window, early after infection, where formation of the integrase-dna association could be prevented. Unfortunately, the negatively charged inhibitors that interfere with IN-DNA complex formation, such as G-quartets, do not enter cells. Cell culture based assays have been developed with cell lines and laboratory strains that facilitate HIV replication. It is not clear, whether inhibition of integration is assayed at best under these artificial culture conditions. In the oligonucleotide-based enzymatic tests not all steps of the integration process are assayed for, and the inhibitors discovered may target biochemical interactions (for example, Mg 2+ chelation) shared by many other enzymes, which would make it difficult, if not impossible, to dissociate activity and toxicity of the compounds. Therefore new assays should be developed that allow identification of authentic integrase inhibitors and search for the additional subtargets during the integration process. The lack of three-dimensional structures of full-length integrase impedes the structure-based design of integrase inhibitors. Moreover, because of the known conformational changes in integrase induced upon DNA binding (Espeseth et al., 2000), cocrystallization of integrase with a model DNA substrate would be required. Novel targets for drug intervention HIV-1 integration is a complex process that starts with a partially synthesized cdna in the cytoplasm and ends with proviral DNA inserted in a transcriptionally active region of the chromosome. Although integrase by itself can carry out the basic biochemical reactions in vitro, other viral and cellular proteins play important roles in the integration process. Although at this time largely unexplored, all steps represent potential targets for therapeutic intervention (Table 1). New assays have to be developed to detect inhibition of these cellular factors. Although inhibition of host factors could be associated with cellular toxicity, this can be overcome if integration of HIV can be prevented at a concentration that is not toxic for the cells. The following targets could be envisaged. (1) The final stages of reverse transcription and the first steps of integration are concurrent, and a physical and functional interaction between RT and IN have been described (Engelman et al., 1995; Wu et al., 1999; Tasara et al., 2001). (2) A specific integrase-dna complex is formed, possibly emerging from the reverse transcription process. Many integrase inhibitors interfere with integrase-dna binding in vitro. It is not clear if inhibition of this step within the PIC is feasible in the cell. (3) Multimerization of IN, which is required for activity, may be targeted by transdominant negative integrase mutants provided by gene therapy. Antiviral Chemistry & Chemotherapy 13:1 7

8 Z Debyser et al. Table 1. Potential drug targets of the integration process Cellular Replication step localisation Antiviral target Antiviral approach Inhibitors identified References 1. Interaction between Cytoplasm RT, IN Specific antibodies, peptides, Wu et al., 1999; RT and IN compounds that can disrupt the Tasara et al., 2001 interaction between IN and RT 2. Integrase Virion, IN Specific antibodies to epitopes Fab 35 Barsov et al., 1996; multimerization cytoplasm (?) important for multimerization Lutzke & Plasterk, 1998; Peptides that mimic the MAb17 Kalpana et al., 1999; multimerization domains Yi et al., 2000 Compounds that interfere with multimerization 3. Integrase-DNA Cytoplasm IN Compounds that interfere with G-quartets, L-chicoric Lutzke et al., 1994; Lee-Huang complex formation DNA binding acid, MAP30, GAP31* et al., 1995; Cherepanov et al., Antibodies directed to the minimal DNA 1997; Wang et al., 1996; binding region Neamati et al., 2000; Pommier et al., 2000; Pluymers et al., processing Cytoplasm IN Low molecular weight inhibitors Specific inhibitors of 3 -processing have not been described 5. DNA juxtaposition Cytoplasm IN, BAF, HMGI(Y) Compounds, antibodies, peptides that Farnet & Bushman, 1997; (LTR bridging) interfere with the activity of the co-factors Chen & Engelman, 1998; Selective AT-rich DNA binders Neamati et al., 1998; (distamycin, lexitropins) Umland et al., Nuclear import of Nuclear pore Vpr, MA(?), IN, cppt Peptides mimicking the nuclear CNI-H0294** Gallay et al., 1995; the pre-integration localization signal Bcvir Pluymers et al., 1999; complex Inhibitors of nuclear transport Petit et al., 2000; antisense inhibition of DNA flap Zennou et al., Selection of Nucleus IN, chromosomal Prevent integration into Kalpana et al., 1994; integration site architecture, INI1(?) actively transcribed regions Goulaouic et al., 1996; Interfere with chromosome remodelling Katz et al., Strand transfer Nucleus IN, INI1(?) Low molecular weight inhibitors Diketo acids Hazuda et al., 2000, Wai et al., 2001, Pluymers et al., Circle formation Nucleus IN, BAF Promotion of autointegration Hazuda et al., DNA repair of Nucleus DNA PK (?), Inhibition of DNA repair Inhibitors of DNA Daniel et al., 1999; Baekelandt the remaining gaps ATM kinase (?), Inhibition of the interaction between IN repair (specificity?) et al., 2000; Brin et al., 2000; PARP-1 (?) and DNA repair proteins Yoder & Bushman, 2000; Ha et al., 2001 * The majority of compounds identified as inhibitors of HIV-1 integrase-mediated reactions are inhibitors of IN-DNA binding. They are not selective and inhibit other DNA modifying enzymes as well. ** CNI-H0294 was proffered as a compound that inhibits nuclear targeting of HIV-1 derived preintegration complexes by inactivating the NLS of the HIV-1 matrix antigen (Popov et al., 1996). BCvir is a peptidomimetic that functionally mimics the NLS of HIV-1 matrix antigen (Friedler et al., 1998). Conflicting results have been published regarding the role of DNA-PK, ATM and PARP during retroviral integration (Daniel et al., 1999; Baekelandt et al., 2000; Ha et al., 2001) International Medical Press

9 Inhibitors of HIV-1 integration (4) The endonucleolytic 3 -processing reaction is followed by (5) the juxtaposition of both LTR ends within the PIC to ensure concerted integration of both LTR ends. This step probably involves a cellular cofactor, such as BAF or HMGI(Y). (6) The cleaved viral DNA within the PIC is actively transported through the nucleopore into the nucleus. Many protein and DNA determinants have been described that act synergistically. Because of this redundancy, it is not clear whether inhibition of one determinant will be sufficient to block the import. (7) After nuclear import the complex will be directed to the chromosomes where the integration site is selected. The mechanism of integration site selection ought to be better understood before inhibitors can be identified. Ini1 may target the proviral DNA into actively transcribed regions. Alternatively, transcriptional activation may be induced by integration. (8) The DNA strand transfer reaction is a valid target, as shown by the inhibition of viral replication by diketo acids. (9) In an unproductive pathway of integration, DNA circles are formed. Promotion of DNA circles would abort HIV infection. Inhibition of DNA strand transfer by diketo acids is accompanied by an increase in 2-LTR circles. (10) Integration is completed by DNA repair, probably carried out by cellular DNA repair enzymes. It will be difficult to obtain selectivity by inhibition of DNA repair. Novel technologies to corroborate integrase as antiviral target Novel PIC integration assays PICs can be partially purified from cells infected with HIV-1 and used as a source of integrase activity in vitro (Brown et al., 1987; Ellison et al., 1990; Farnet & Haseltine, 1990). PICs can direct joining of both ends of the viral cdna in a co-ordinated fashion yielding a product resembling the gapped integration intermediate. Many of the inhibitors that were found using the oligonucleotide assay and recombinant integrase proved inactive in the PIC assays (Farnet & Bushman, 1996; Farnet et al., 1996). In the past PIC assays have not been widely used due to technical difficulties in handling large amounts of infectious HIV. Recently, however, novel PIC integration assays have been put forward that may facilitate their use. ne PIC integration assay used DNA-coated microtitre plates to speed up in vitro assaying of PIC integration (Hansen et al., 1999), while Brooun et al. (2001) have developed a highly sensitive PCR-based PIC integration assay with long target DNA molecules. PCR-based quantification Quantitative, real-time PCR techniques have been developed that allow quantification of the various HIV DNA species during HIV infection by selecting specific primer/probe combinations (Butler et al., 2001). HIV-1 derived vectors are used to ensure single-round infection, although differences may exist in kinetics of HIV infection and HIV vector transduction (Figure 5) (B Van Maele et al., unpublished observations). A first Q-PCR quantifies total HIV DNA. During the first hours of the replication cycle, the amount of total HIV DNA is the result of the ongoing reverse transcription. After multiple passaging, total HIV DNA will reflect integration, although DNA circles may confound this analysis. The 2-LTR circles can be quantified directly by a specific primer pair detecting the LTR linkage. Integrated proviral DNA is quantified by Alu-PCR. Alu-repeats are defined as short interspersed nuclear elements (SINE) of approximately 300 bp and occupy approximately 5% of the human genome. n average, an Alu-repeat is present every 5000 bp. Using one Alurepeat and one HIV-1 specific primer, one is able to PCRamplify a fraction of integrated proviral DNA. A standard is obtained by measuring integrated DNA present in transduced and passaged cells with the Q-PCR for total DNA. The signal obtained by Q-Alu PCR on this DNA standard can then be attributed to a defined amount of integrated DNA. In the presence of an RT inhibitor, a relative decrease in total HIV DNA synthesis will be observed, as well as a drop in 2-LTR circles and integrated DNA. A strand transfer inhibitor, in contrast, will not affect HIV DNA synthesis but lead to an increase in the amount of 2-LTR circles. The Alu-PCR signal will be significantly lower. This technology, theoretically, allows identification of the antiviral target in cell culture. However, in its current version, the assay is labour-intensive, time-consuming and expensive. The sensitivity of the Alu-PCR is poor, since integrants too far away from an Alu-repeat will not be amplified. Therefore, technological improvements of this methodology will be required. An alternative PCR assay to quantify HIV integration in cell culture has been described (Vandegraaff et al., 2001). Integrated HIV DNA is detected by a nested Alu-PCR; extrachromosomal HIV DNA is amplified in parallel. In this assay PCR products are analysed by Southern hybridization. The LTR circles, as such, are not detected. Diketo acids were shown to block HIV integration using this method. However, quantification by classical PCR is less accurate than when using realtime PCR. Chimeric virus technology Selection of drug-resistant HIV strains in cell culture followed by sequencing of viral genes is an important approach to identify the antiviral target of a particular drug. In the case of G-quartets and L-chicoric acid, absence of mutations in the integrase gene but presence of mutations in gp120 of virus strains selected until resistant in cell culture, convinced us that integration was not the antiviral Antiviral Chemistry & Chemotherapy 13:1 9

10 Z Debyser et al. Figure 5. Analysis of HIV integration by quantitative PCR (a) (b) (c) Copies/cell Copies/cell Copies/cell NAC (untransduced cells) control +AZT (750 nm) +L708,906 (25 µm) Time after transduction (h) Time after transduction (h) Time after transduction (h) 293T cells were transduced with HIV-1 vectors at multiplicity of infection=10 in the absence of inhibitor ( ) or in the presence of 750 nm AZT ( ) or 25 µm L708, 906 ( ). At different time points after infection, DNA extracts were prepared and analysed by realtime PCR (Van Maele et al., submitted for publication). Late reverse transcripts were quantified (a), as well as 2-LTR circles (b) and integrated proviral DNA (c). A no amplification control was run in parallel ( ). target (Esté et al., 1998, Pluymers et al.,2000). Full proof is provided by recovery of resistance after recombination of the gene carrying the mutations in a wild type background. We have designed Chimeric Virus Technology that allows the generation of viable virus by homologous recombination of a PCR-amplified selected strain-derived IN gene into a IN-deleted laboratory HIV-1 strain (Witvrouw et al., unpublished observations). We have selected drugresistant strains in the presence of the diketo acid L-708, 906 (Witvrouw et al., unpublished observations). Introduction of the IN gene of this selected strain into the wild-type background resulted in fully resistant HIV virus. Although resistance selection and recombination, provides full proof of the antiviral target, the approach is not always feasible. To select resistance, a gradual increase in drug concentration during cell culture is required. For drugs with poor solubility and/or low selectivity indices (SI), as is often the case with initial lead compounds, selection of drug resistance will not always be feasible. Lead optimization by chemical modifications may be required to develop a compound with a selectivity index that is sufficiently high for resistance selection. Cellular integration assay A cell-based integration assay would circumvent the problems associated with the cell-free enzymatic screening of integrase inhibitors. Moreover, such a system would enable the discovery of new cellular targets for anti-hiv agents. ur laboratory has been working on the development of a cellular integration system. A cell-based integration assay has to meet following requirements: (1) an efficient eukaryotic expression system for HIV-1 integrase has to be designed; (2) the integrase expressed has to be enzymatically active (3) a test system for detecting integration (and integration inhibitors) in the cell has to be developed. At this stage we have obtained high level expression of HIV-1 integrase in mammalian cells. This technological breakthrough was achieved by designing a synthetic integrase gene that circumvents the intrinsic instability of the mrna of the wild-type gene. The enzyme proved active as demonstrated by trans-complementation of integration-defective HIV-1-derived vector particles (Cherepanov et al., 2000). Stable cell lines (293T-IN S ) that constitutively express high levels of HIV-1 integrase were selected. We have shown that integrase remains stably associated to the chromosomes during mitosis. The final goal of our work is the development of a cell-based integration assay that could be used to detect selective integrase inhibitors. In a more simplified format, the test could serve as a second-line specificity check for inhibitors of the in vitro integrase assays, to prevent investment in inhibitors with mistaken International Medical Press

11 Inhibitors of HIV-1 integration identity. We are currently evaluating different strategies using reporter genes. ne of the problems encountered, is the fairly high-level of non-homologous recombination in mammalian cells that is capable of inserting DNA substrates in the host chromosome in a non-integrase-mediated way. A cell based integration system will enable the study of the cell biology of HIV-1 integration and more particularly the role of host factors associated therewith. This may possibly lead to the identification of new cellular targets for anti-hiv agents. Algorithm to corroborate authenticity of integrase inhibition As listed above, a vast array of integrase inhibitors has been discovered in vitro. For an IN inhibitor to be included in a clinically useful antiviral combination regimen, it is essential to corroborate its selectivity towards the integration step of the viral replication cycle. Until now only the class of diketo acids have been recognized as inhibitors that selectively target integrase (Hazuda et al.,2000). When a new integrase inhibitor is discovered that also displays antiviral activity in cell culture, it does not follow automatically that this inhibitor targets the integration step of HIV replication. Further investigations are required to pinpoint the target of action of this compound. In the absence of a cell-based specific integration assay, the following experiments should be carried out before the authenticity of integrase inhibition can be claimed. Time-of-addition experiment By adding the inhibitor at various time points after infection it is often possible to define the replication step targeted in cell culture (Figure 6). This so-called time-ofaddition experiment (TA) should be carried out in parallel with known inhibitors of various replication steps. Theoretically, integrase inhibitors act after reverse transcription and before retroviral gene expression is initiated. Using this assay it was shown that L-chicoric acid interferes with viral entry (0 1 h after infection) and not integration (Pluymers et al., 2000), and that the diketo acid L-708, 906 acts at a time-point (6 8 h) that co-incides with the DNA strand transfer step of integration (W Pluymers et al., unpublished observations). Although this assay can provide valuable information, there are some caveats: (1) the real moment of interaction of a particular compound may be later if the compound has to be metabolized prior to being active; (2) not the molecular target but the moment of interaction is given by a TA assay; and (3) the formation of the IN-DNA complex and the 3 -processing reaction probably occur before reverse transcription is completed. Diketo acids inhibit DNA strand transfer preferentially. It remains to be seen whether authentic Figure 6. Time-of-addition experiment log p24 (pg/ml) Time (h) Entry Reverse Strand Maturation transcription transfer (protease) MT-4 cells were infected with HIV-1(III B ) at a multiplicity of infection of 0.5, and the test compounds were added at different times post infection (Pluymers et al., 2002). Viral p24 Ag production was determined at 31 h post infection and is expressed as the log 10 of the p24 Ag content in pg/ml. ( ), control; ( ), dextran sulphate (20 µm); ( ), AZT (1.9 µm); ( ), ritonavir (2.8 µm); ( ), the diketo acid L-708,906 (173 µm). inhibitors of DNA binding and/or 3 -processing can be fully discriminated from RT inhibitors in this assay. Exclusion of other targets In order to exclude activity of the compounds against other targets, it is possible to test the compounds in several in vitro assays (for example, reverse transcriptase assay). Such an approach is labour-intensive and only gives indirect evidence of selectivity. Moreover, for many potential targets no in vitro assays have been developed. If the TA experiment suggests inhibition of the integration step, it may be helpful to assay for inhibition of reverse transcription and gene expression as well. PCR based quantification of HIV DNA species during replication Quantification of total HIV DNA, 2-LTR DNA circles and integrated proviral DNA (through Alu-PCR) during HIV infection, provides a direct way to discriminate between RT and IN inhibitors (Butler et al.,2001). In the presence of an RT inhibitor (zidovudine), total HIV DNA synthesis, 2-LTR circle formation and proviral DNA integration are inhibited. Inhibition of DNA strand transfer by authentic IN inhibitors, such as the diketo Antiviral Chemistry & Chemotherapy 13:1 11

12 Z Debyser et al. acid L-708, 906, results in an increase in 2-LTR circles, an inhibition of proviral DNA formation but no effect on DNA synthesis by reverse transcription (Figure 4). Selection and sequencing of drug-resistant strains By passaging the virus in the presence of the drug, resistant strains can be selected. The identification of drug-induced mutations by sequencing the integrase gene can provide direct evidence for the antiviral target. HIV strains that are resistant to other inhibitors (such as diketo acids) can be used to evaluate cross-resistance. However, whenever a resistant HIV strain is generated and mutations in the integrase gene identified, it remains necessary to re-introduce the observed mutations in the wild-type integrase gene, preferentially in the context of a recombinant virus. When the compound, in the presence of which the strain was selected, is no longer able to inhibit recombinant integrase or virus that is carrying the specific mutation(s) the compound may be surmised to target integrase. Conclusions After more than a decade of an intensive search for inhibitors of HIV integrase, only one class of authentic integrase inhibitors has been identified, the diketo acids (Hazuda et al., 2000). At present it is not clear whether the compounds will make it to the clinic; still they provide the proof-of-principle that integrase inhibitors may block viral replication. Lentiviral DNA integration can be considered as a bona fide antiviral target. The elucidation of a co-crystal of full length IN complexed with DNA, the development of an efficient and selective system for the evaluation of integrase inhibitors in cell culture, as well as the clarification of the exact role of host co-factors during integration, will facilitate the development of new integrase inhibitors. For the moment, we have to rely on a combination of experimental approaches to prove the authenticity of integrase inhibition in cell culture. Authenticity of integrase inhibition and avoidance of mistaken identity is a prerequisite for delineating the appropriate structure-activity relationship, which in turn, is a prerequisite for identifying the appropriate lead compound for clinical development. Acknowledgements Zeger Debyser has a postdoctoral fellowship from the Flemish Fund for Scientific Research (FW). Bénédicte Van Maele is funded by a grant from the Flemish Institute supporting Scientific-Technological Research in Industry (IWT). References Baekelandt V, Claeys A, Cherepanov P, De Clercq E, De Strooper B, Nuttin B & Debyser Z 2000 DNA-dependent protein kinase is not required for efficient lentivirus integration. Journal of Virology 74: Barsov EV, Huber WE, Marcotrigiano J, Clark PK, Clark AD, Arnold E & Hughes SH (1996) Inhibition of human immunodeficiency virus type 1 integrase by the Fab fragment of a specific monoclonal antibody suggests that different multimerization states are required for different enzymatic functions. Journal of Virology 70: Bouyac-Bertoia M, Dvorin JD, Fouchier RA, Jenkins Y, Meyer BE, Wu LI, Emerman M & Malim MH (2001) HIV-1 infection requires a functional integrase NLS. Molecular Cell 7: Brin E, Yi J, Skalka AM & Leis J (2000) Modeling the late steps in HIV-1 retroviral integrase-catalyzed DNA integration. Journal of Biological Chemistry 275: Brooun A, Richman DD & Kornbluth RS (2001) HIV-1 preintegration complexes preferentially integrate into longer target DNA molecules in solution as detected by a sensitive, polymerase chain reaction-based integration assay. Journal of Biological Chemistry 276: Brown P, Bowerman B, Varmus HE & Bishop JM (1987) Correct integration of retroviral DNA in vitro. Cell 49: Brown P (1997) Integration of retroviral DNA. Current Topics in Microbiological Immunology 157: Bukrinsky MI, Sharova N, Dempsey MP, Stanwick TL, Bukrinskaya AG, Haggerty S & Stevenson M (1992) Active nuclear import of human immunodeficiency virus type 1 preintegration complexes. Proceedings of the National Academy of Sciences USA 89: Bukrinsky MI, Sharova N, McDonald TL, Pushkarskaya T, Tarpley WG & Stevenson M (1993) Association of integrase, matrix, and reverse transcriptase antigens of human immunodeficiency virus type 1 with viral nucleic acids following acute infection. Proceedings of the National Academy of Sciences USA 90: Bukrinsky MI, Haggerty S, Dempsey MP, Sharova N, Adzhubel A, Spitz L, Lewis P, Goldfarb D, Emerman M & Stevenson M (1993) A nuclear localization signal within HIV-1 matrix protein that governs infection of non-dividing cells. Nature 365: Burke CJ, Sanyal G, Bruner MW, Ryan JA, LaFemina RL, Robbins HL, Zeft AS, Middaugh CR & Cordingley MG (1992) Structural implications of spectroscopic characterization of a putative zinc finger peptide from HIV-1 integrase. Journal of Biological Chemistry 267: Bushman FD, Fujiwara T & Craigie R (1990) Retroviral DNA integration directed by HIV integration protein in vitro. Science 249: Bushman FD, Engelman A, Palmer I, Wingfield P & Craigie R (1993) Domains of the integrase protein of human immunodeficiency virus type 1 responsible for polynucleotidyl transfer and zinc binding. Proceedings of the National Academy of Sciences USA 90: Butler SL, Hansen MS & Bushman FD (2001) A quantitative assay for HIV DNA integration in vivo. Nature Medicine 7: Cai M, Zheng R, Caffrey M, Craigie R, Clore GM & Gronenborn AM (1997) Solution structure of the N-terminal zinc binding domain of HIV-1 integrase. Nature Structural Biology 4: Cai M, Huang Y, Zheng R, Wei SQ, Ghirlando R, Lee MS, Craigie R, Gronenborn AM & Clore GM (1998) Solution structure of the cellular factor BAF responsible for protecting International Medical Press