Diketo acid inhibitor mechanism and HIV-1 integrase: Implications for metal binding in the active site of phosphotransferase enzymes

Abstract

The process of integrating the reverse-transcribed HIV-1 DNA into the host chromosomal DNA is catalyzed by the virally encoded enzyme integrase (IN). Integration requires two metal-dependent reactions, 3′ end processing and strand transfer. Compounds that contain a diketo acid moiety have been shown to selectively inhibit the strand transfer reaction of IN in vitro and in infected cells and are effective as inhibitors of HIV-1 replication. To characterize the molecular basis of inhibition, we used functional assays and binding assays to evaluate a series of structurally related analogs. These studies focused on investigating the role of the conserved carboxylate and metal binding. We demonstrate that an acidic moiety such as a carboxylate or isosteric heterocycle is not required for binding to the enzyme complex but is essential for inhibition and confers distinct metal-dependent properties on the inhibitor. Binding requires divalent metal and resistance is metal dependent with active site mutants displaying resistance only when the enzymes are evaluated in the context of Mg²⁺. The mechanism of action of these inhibitors is therefore likely a consequence of the interaction between the acid moiety and metal ion(s) in the IN active site, resulting in a functional sequestration of the critical metal cofactor(s). These studies thus have implications for modeling active site inhibitors of IN, designing and evaluating analogs with improved efficacy, and identifying inhibitors of other metal-dependent phosphotransferases.

An essential step in HIV replication is the integration of the reverse-transcribed viral genome into host chromosomal DNA by the virally encoded integrase (IN) protein (1–3). Integration is required for efficient long terminal repeat-driven transcription of the provirus for the production of viral proteins and RNA progeny. IN represents an important chemotherapeutic target, as its inactivation, either by mutagenesis or inhibition, blocks productive infection by HIV-1 (4–7).

Integration is carried out in the cell in a series of distinct steps (8–10). First, IN cleaves the two terminal nucleotides from each 3′ end of the viral DNA. The 3′ processing reaction is carried out concurrently with or soon after reverse transcription in the cytoplasm. In the second step, strand transfer, IN catalyzes staggered nicking of the target chromosomal DNA and joining of each 3′ end of the viral DNA to the 5′ ends of the host DNA. Strand transfer is temporally and spatially separated from 3′ processing and occurs after transport of the preintegration complex from the cytoplasm into the nucleus.

Divalent metals such as Mg²⁺ or Mn²⁺ are required for both 3′ processing and strand transfer and for the assembly of IN onto specific viral donor DNA to form a complex competent to carry out either function (11–13). Mg²⁺ is the likely metal cofactor in vivo; however, Mn²⁺ frequently is used in biochemical studies because IN activity is generally more robust in its presence. Binding of the metal cofactors in the active site of IN is mediated by a triad of acidic residues called the DDE motif (D64, D116, and E152 in HIV-1 IN). The DDE motif is highly conserved among a superfamily of nucleases and polynucleotidyltransferases that include the retroviral INs and bacterial transposases. For IN and related enzymes that catalyze phosphoryl transfer reactions, the number of metals present and required in the active site remains controversial. Structural studies of IN reveal a single binding site for Mg²⁺ or Mn²⁺ (14–16), although a second metal has been observed with Zn²⁺ or Cd²⁺ (17).

L-731,988 is the first in a novel class of diketo acids (DKAs) identified as potent inhibitors of IN-catalyzed strand transfer (7). The activity of L-731,988 is remarkable in that it discriminates between the two catalytic activities of IN-inhibiting strand transfer with an IC₅₀ of 100 nM and 3′ processing with an IC₅₀ of 5 μM. Unlike many previously identified inhibitors of IN that interfere with assembly (18), L-731,988 and related DKAs inhibit integration and viral replication in cell culture (7). Mutations that confer resistance to DKAs have been identified and map to IN residues adjacent to D64 and E152. The inhibitor 5CITEP [1-(5-chloroindol-3-yl)-3-hydroxy-3-(2H-tetrazol-5-yl)-propenone] also binds within the IN active site (19). In the crystal structure of the core domain of IN (amino acids 50–212) with 5CITEP, the inhibitor contacts IN residues that are identical or close to those predicted to be important for the DKAs on the basis of resistance.

Previous studies have shown that the binding of DKA inhibitors requires that IN be assembled into a nucleoprotein complex competent to catalyze strand transfer and that binding of the inhibitor and the target DNA substrate are mutually exclusive (20). These studies suggest a biochemical basis for the strand transfer selectivity of these inhibitors. Here, we have used a series of derivatives of the DKAs and 5CITEP in enzymatic assays and binding assays to explore the molecular basis of this inhibition. We demonstrate that (i) binding to IN is in part mediated by the interaction of the acid functionality with metals in the active site, (ii) binding to IN is not itself sufficient to inhibit strand transfer, and (iii) mutations that confer resistance likely work by affecting metal coordination. The data presented suggest sequestration of the active site metal is required for inhibition and are consistent with a two-metal model for the active site of IN. As compounds that contain a DKA moiety have also been shown to inhibit the Cap-dependent endonuclease of influenza virus and the 5B polymerase of hepatitis C (21, 22), these studies have implications for a variety of metal-dependent phosphotransferase enzymes of therapeutic interest.

Materials and Methods

Oligonucleotides.

Oligonucleotides (preprocessed U5 top strand, 5′-biotin-ACCCTTTTAGTCAGTGTGGAAAATCTCTAGCA; U5 bottom strand, 5′-ACTGCTAGAGATTTTCCACACTGACTAAAAG) were synthesized and gel-purified by Midland Certified Reagent, Midland, TX.

Enzymes.

Full-length HXB2 IN was expressed and purified as described (23). Deletions and point mutations were engineered by using standard techniques, and mutants were purified as for the wild-type protein.

³H-DKAs.

³H-l-731,988 (4-[1-(4-fluorobenzyl)pyrrole-2-yl-2,4-diketobutanoic acid) (³H-I) was prepared by catalytic tritiation of the iodopyrrole (24).§ The synthesis of compounds I and II [³H-II, 4-(3-benzylphenyl)-2,4-dioxo-butyric acid] was as described (25). Specific activity of the purified radiolabeled compounds was 20–30 Ci/mmol.

IN Activity Assays.

IN activity was determined with a microtiter plate assay performed as described (23) with the following modifications. Donor DNA biotinylated on the 5′ end of the strand processed by IN was immobilized onto streptavidin plates (Black Reactibind, Pierce) by using 1.5 pmol/well. IN was assembled in reaction buffer [20 mM Hepes, pH 7.6/40 mM NaCl/5 mM β-mercaptoethanol (βME)/50 μg/ml BSA] containing 25 mM MnCl₂. Excess IN was removed and the complexes were washed before the addition of 3′ FITC-labeled target DNA substrate. Strand transfer reactions were performed in 2.5 mM MgCl₂ or MnCl₂ by using 0.5, 5.0, or 25.0 nM target DNA. Products were detected by using an anti-FITC antibody conjugated with alkaline phosphatase (Roche Molecular Biochemicals) and a chemiluminescence substrate (Sapphire II, Tropix, Bedford, MA). To assess effects on assembly and/or 3′ end processing, compounds were added during assembly and removed before the addition of target DNA. For strand transfer, compounds were added after assembly, before the addition of target DNA.

Scintillation Proximity Assay (SPA) for Binding of Tritiated Inhibitor.

Streptavidin-coated polyvinyltoluene SPA beads (≈100 pmol biotin binding sites/mg beads) were dissolved to 10 mg/ml in buffer A (27.8 mM Hepes, pH 7.8/27.8 mM MnCl₂/111.1 μg/ml BSA/5.56 mM βME). Beads were incubated with 360 nM biotinylated donor DNA, washed, and then resuspended at 5 mg/ml in buffer A. IN was added to a final concentration of 500 nM and incubated for 1 h at room temperature. For binding assays, 10 μl of ³H-(I or II) diluted in DMSO was mixed with 70 μl buffer A containing 57.1 mM NaCl and 20 μl of the bead suspension to achieve the following assay conditions: 25 mM Hepes, 25 mM MnCl₂, 100 μg/ml BSA, 5 mM β-ME, 5 nM DNA, 100 nM IN, and 50 mM NaCl. Samples were equilibrated overnight, and bound ligand was determined by scintillation counting. Metal-dependence studies were conducted by assembling the complexes in the absence of inhibitor, washing the beads, and adding radiolabeled inhibitor with metal as noted.

Results

SPA to Detect Binding of Tritiated Ligands to HIV-1 IN.

IN binds with high affinity to nonspecific DNA and specific sequences corresponding to the HIV-1 DNA ends (11, 12, 26, 27). To evaluate the binding of ligands to IN, we used a SPA that exploits this interaction (20). Biotinylated HIV-1 long terminal repeat sequence oligonucleotides (or unrelated DNAs) bound to streptavidin-coated polyvinyltoluene beads are used to capture IN. Immobilization of the donor substrate before assembly limits the potential for strand transfer with a second DNA molecule functioning as target. The assay involves three steps: (i) immobilization of the DNA, (ii) assembly of IN onto the immobilized DNA, and (iii) addition of radiolabeled ligand and detection. The binding experiments described here use two DKA inhibitors of strand transfer, ³H-I (7) and ³H-II (24, 28).§ Compound I or L-731,988 has an IC₅₀ of 100 nM in strand transfer and inhibits HIV-1 replication with an IC₅₀ of 2 μM. Compound II is 10- to 20-fold more potent.

DKA Binding Correlates with Assembly of an IN Complex Competent to Catalyze Strand Transfer.

In the SPA, a signal is generated only when radioligand is enriched on the surface of the beads; therefore, DKA binding depended on the amount of immobilized donor DNA and not on the concentration of IN. Binding of each compound was saturable, not cooperative (Hill coefficent ≈1), and required both IN and DNA. As reported (20), binding correlated with activity of the strand transfer complex: (i) binding was detected with full-length IN and not with truncated enzymes competent to catalyze disintegration but not strand transfer, and (ii) binding was detected when IN was assembled onto viral DNA ends and not nonspecific sequence or single-stranded DNAs (data not shown). In saturation binding studies using the U5 long terminal repeat DNA sequence, dissociation constants of 15 and 5 nM were determined for compounds I and II, respectively (Fig. 1).

Figure 1.

Binding of radiolabeled DKAs to IN assembled on donor DNA ends. IN assembled onto U5 viral end DNA that was immobilized on SPA beads was equilibrated with ³H-I (●) or ³H-II (○).

Competition Assay to Detect Binding of Compounds to IN.

The format of the SPA can be used to carry out either direct or competitive binding studies. Given the limitation in the number of ligands that can be radiolabeled, we used ³H-II in a competitive binding mode to investigate structurally distinct inhibitors and novel analogs. The latter (Table 1) were synthesized to address similarities between the DKAs and 5CITEP.

Table 1.

Inhibition of integrase activities by structurally diverse inhibitors

Assays performed in MnCl₂. 

Although a decrease in radioligand binding can result from competition for a mutual binding site, indirect effects are also possible. To distinguish between these alternatives, we performed binding studies using ³H-II at both 5 nM and 50 nM (Fig. 2). For a direct competitor the IC₅₀ should vary as a function of the concentration of labeled ligand and the IC₅₀ determined by using the radioligand at its K_d (5 nM for ³H-II) should reflect the K_d of the competitor. With 5 nM ³H-II, the IC₅₀s for unlabeled I and II in the competitive binding assay (4.7 nM and 7.1 nM) were consistent with the K_ds determined in direct binding assays (5 nM and 15 nM, respectively). As expected, the IC₅₀s of I and II were higher when the labeled ligand was increased to 50 nM (Fig. 2A).

Figure 2.

Competitive binding assays at two concentrations of tritiated ligand. (A) Compounds I and II were assayed in competitive binding assays using either 5 nM or 50 nM ³H-II: compound I vs. 5 nM ³H-II (○); compound I vs. 50 nM ³H-II (▿); compound II vs. 5 nM ³H-II (●); compound II vs. 50 nM ³H-II (▾). (B) 5CITEP titrated vs. 5 nM (●) or 50 nM (○) ³H-II. (C) l-chicoric acid titrated vs. 5 nM (●)or 50 nM (○) ³H-II.

Competitive binding was next used to investigate structurally diverse inhibitors: 5CITEP (19) and l-chicoric acid (18). 5CITEP is somewhat similar to the DKAs consisting of a diketone separating tetrazole and chloroindole moieties (19). We also observed that 5CITEP selectively inhibits strand transfer and that, like the DKAs, inhibition depends on the concentration of target substrate (Table 1). In contrast, l-chicoric acid did not inhibit strand transfer and was effective only if added during assembly.

In competitive binding assays using ³H-II, both 5CITEP and l-chicoric acid effected a dose-dependent reduction in binding (Fig. 2 B and C). As expected for a compound with similar mechanism, competition by 5CITEP was sensitive to the concentration of labeled DKA, suggesting a common binding site. The observation that l-chicoric acid was not sensitive to the concentration of ³H-II suggests a discrete binding site and is consistent with its distinct mode of action.

The Acid Moiety Is Required for Inhibition but Not Binding of DKA Derivatives.

Binding and mechanistic studies suggest the DKAs and 5CITEP are structural homologs. To explore role of the common functionalities in these compounds, a tetrazole derivative of I and a carboxylate derivative of 5CITEP were synthesized. The chimeric compounds, as well as I and 5CITEP, were evaluated for binding and inhibition in Mg²⁺ and Mn²⁺. As shown in Table 2, both hybrid molecules were active; however, in the context of either template, the carboxylate (I and IV, respectively) was more potent. The choice of metal did not affect the affinity or potency of either carboxylate analog. In contrast the tetrazoles (III and 5CITEP) exhibited reduced binding and inhibition in Mg²⁺ relative to Mn²⁺. The tetrazole therefore can functionally replace the carboxylate but confers a distinct metal dependence on activity. The weak activity of the tetrazole analog of I in Mg²⁺ may explain the poor antiviral activity (IC₅₀ >50 μM, data not shown) versus compound I (IC₅₀ = 5 μM) despite comparable activity in Mn²⁺.

Table 2.

Binding and inhibition by hybrid integrase inhibitors

Additional derivatives were synthesized to evaluate the requirement for the acid and keto substituents (Table 3) (24).§ Compounds devoid of both the carboxylate and α and γ-ketone (IX and X) did not either bind or inhibit. Compounds lacking an acid functionality but retaining the diketo substituents (compounds V–VIII) bound with submicromolar to micromolar affinity. Surprisingly, despite binding, these analogs were inactive as inhibitors (in both Mn²⁺ and Mg²⁺ >100 μM). This finding is in contrast to the carboxylate or tetrazole analogs in which binding and inhibition were observed with comparable potency (Table 3). The acidic tetrazole/carboxylate moiety therefore is not required for binding but is essential for inhibition.

Table 3.

Requirement of the DKA motif in binding and inhibition

Binding of the DKAs to IN Is Metal Dependent.

The observation that the acid substituent confers metal dependence on these inhibitors suggests divalent metal may be required for binding. To address whether binding is metal dependent, complexes were assembled and then washed extensively before performing binding studies in the presence and absence of Mg²⁺ or Mn²⁺. As shown in Fig. 3, compound II bound in a strictly metal-dependent manner and approximately 20-fold more Mg²⁺ was required. Similar results also were obtained for compound I (data not shown).

Figure 3.

DKA binding requires divalent metal. IN was assembled onto the beads in the presence of 27 mM MnCl₂. The beads were washed in buffer lacking divalent metal, and the complex was incubated with 5 nM of ³H-II in MnCl₂ (●) or MgCl₂ (□) as indicated. cpm recorded were normalized to maximal binding calculated for each curve (%B_max) The average of at least three separate determinations is shown.

IN Mutations Confer Resistance in Mg²⁺ but Not Mn²⁺.

Substitutions in several amino acids near the metal coordinating residues D64 and E152 of IN confer resistance to DKA inhibitors (7). Specifically, combinations of T66I and either S153Y or M154I engender a 2–6× loss of sensitivity to I in antiviral assays. To examine the effects of these mutations in vitro, purified IN containing T66I/M154I was tested for the ability to bind ³H-II in Mg²⁺ and Mn²⁺ (Fig. 4). The mutations had no effect on binding in Mn²⁺ but dramatically compromised binding in Mg²⁺. The ability of the mutant to bind the DKAs correlated with the effect on strand transfer activity (data not shown). When assayed in Mn²⁺, the mutant enzyme was inhibited with little change in potency. In contrast, in Mg²⁺ the IC₅₀s of II was higher for the mutant. Interestingly, IN mutations that conferred DKA resistance also diminished the activity of 5CITEP, but the effect was also apparent only in Mg²⁺ (data not shown).

Figure 4.

Binding of DKA to wild-type and DKA-resistant IN. After assembly onto biotinylated U5 donor DNA, binding of ³H-II to IN was assessed as follows: wild type-IN, 25 mM MnCl₂ (●); T66I,M154I-IN, 25 mM MnCl₂ (○); wild type-IN, 25 mM MgCl₂ (▾); T66I,M154I-IN, 25 mM MgCl₂ (▿).

Model for the Interaction of DKAs with IN.

We developed a model for the binding of the DKA analogs to IN that is consistent with the requirement for the acid functionality and observed metal dependence of these inhibitors. In this model, the inhibitor coordinates two metals bound at the active site by the conserved DDE motif of IN (Fig. 5). The bond lengths and angles for the acid moiety used in developing this model are based on the crystal structure of 5CITEP (19). As observed in the 5CITEP structure, the DKA portion of the molecule is planar. Assuming 2.0 Å as the bond length between oxygen atoms of the inhibitor and metals and that each metal is equidistant from either the ketone or carboxylate oxygen atoms, the distance between the metals is calculated to be 3.61 Å. The distance between the two metals in the active site of avian sarcoma virus IN is 3.62 Å and thus in agreement. This model is consistent with the observation that both binding and inhibition of IN involves an interaction with the metal cofactor.

Figure 5.

Model for the binding of two divalent metals by DKA inhibitors.

Discussion

In this article, we explored similarities between 5CITEP and the DKAs and examined the requirements for DKA binding and inhibition of IN. We present evidence that suggests a direct interaction with divalent metal in the IN active site. Specifically, we have demonstrated: (i) Consistent with data for L-731,988 previously reported (20), these inhibitors selectively inhibit strand transfer and require assembly of a full-length IN onto donor DNA. (ii) Binding of the DKAs requires divalent metal and binding and activity of analogs with isosteric acid replacements can exhibit distinct metal specificity. (iii) Binding is not sufficient for inhibition of strand transfer and the acidic function is required. (iv) The effects of IN mutations that confer resistance are manifest in the presence of Mg²⁺ and not Mn²⁺. Based on these observations, we suggest sequestration of the divalent cofactor by the DKA is responsible for inhibition and have developed a model in which binding and activity involves direct coordination of two divalent metals in the IN active site. This model is consistent with structural and biochemical studies of INs and transposases and has implications for exploiting this mechanism of action to identify inhibitors of related, metal-dependent enzymes.

Although the number of metals is controversial, both biochemical and structural studies suggest that a model in which the IN active site binds two metals is plausible. Cysteine mutagenesis studies of Tn10 transposase (29) demonstrated that all three DDE residues function in divalent metal binding and that the glutamate residue has a role in target capture. All IN DDE mutants display reduced 3′ end processing, strand transfer, and disintegration, and mutation of any of these residues in related transposases and polynucleotidyltransferases diminishes their respective activities.

Structures have been solved for HIV-1 and avian sarcoma virus (ASV) IN in the absence of DNA (14–17, 30) and for Tn5 transposase with a bound donor end (30). Crystals of the HIV-1 and ASV IN core domains with Mg²⁺ or Mn²⁺ contain one metal coordinated by D64 and D116 (site I; positions in HIV-1 IN); however, an additional metal coordinated by D64 and E152 (site II) was observed in ASV IN by using either Zn²⁺ or Cd²⁺ (17). The structure of Tn5 transposase bound to donor DNA also reveals a single metal (31). However, in the Tn5 transposase structure the metal is coordinated at site II. The HIV-1 RNase H active site contains two metals coordinated concurrently by the DDE motif (32). The results of these and other studies on a variety of phosphotransferases suggest that there are two discrete metal binding sites, that both sites can be occupied concurrently, and that occupancy at the individual sites may be affected by DNA (33, 34). IN appears to bind a metal ion at site I but not at site II in the absence of donor DNA. If binding of the DKA requires a metal at site II and metal binding at this site is contingent upon formation of strand transfer complex, the model would explain why inhibitor binding is observed only in the context of donor DNA.

When the DKAs are bound to the site II metal ion by the ketone moiety, it is likely that binding of metal at site I is required to neutralize the negative charge of the carboxylate or coordinate the electron pair provided by the heterocycle nitrogen. Esters, which lack an acidic functionality that requires neutralization, would not interact with the metal at site I and would thus be inactive as inhibitors. A direct interaction between the inhibitor carboxylate or heterocycle with a metal at site I can also account for the differences in binding and inhibition observed in Mg²⁺ and Mn²⁺. Carboxylates bind both Mn²⁺ and Mg²⁺ effectively whereas nitrogen-containing heterocycles exhibit higher affinity for Mn²⁺. An alternative possibility is that the active site is structurally different in Mn²⁺ and Mg²⁺. However, this seems unlikely based on the equivalent binding of the carboxylates in either metal.

IN mutations that confer virologic resistance in vitro reduce the affinity and activity of DKAs in Mg²⁺ but not Mn²⁺. These results illustrate the importance of assessing IN activity in the appropriate metal and provide evidence supporting Mg²⁺ as the relevant cofactor for integration in vivo. These observations are also consistent with the proposed model in which the DKAs coordinate metals in the IN active site and suggest a nonobvious mechanism in which resistance mutations act indirectly by affecting metal binding. The amino acids associated with resistance are generally not directed into the active site and thus are not likely to interact directly with the inhibitor. The mutations may shift the active site residues altering the affinity or position of the metals. An effect on metal binding would also account for the decreased catalytic activity of the enzyme and the decreased replication capacity of viruses harboring these mutations (7). Mn²⁺ is compatible with a wider range of reactive group geometries than Mg²⁺ (35) and thus Mn²⁺-catalyzed IN reactions and inhibition may be less sensitive to changes in reactive group geometries.

The DKAs are the first biologically validated small molecule inhibitors of IN, and 5CITEP is the only active site inhibitor of HIV-1 IN for which structural information is available. Each therefore represents an important prototype in efforts to develop novel therapeutics against this enzyme. Insights into the mechanism of these agents and the role of metal coordination thus have important implications for modeling active site inhibitors and generating analogs with improved potency. A model in which these inhibitors work by sequestering metal ions in the active site explains the observation that structurally discrete DKAs inhibit the cap-dependent endonuclease of influenza virus (21), the 5B polymerase of hepatitis (22), and the RAG recombination enzyme (36) and suggests metal sequestration may be a mechanism that can be exploited to develop inhibitors of many biologically interesting polynucleotidyltransferase enzymes.

Abbreviations

IN

integrase

DKA

diketo acid

β-ME

β-mercaptoethanol

SPA

scintillation proximity assay

5CITEP

1-(5-chloroindol-3-yl)-3-hydroxy-3-(2H-tetrazol-5-yl)-propenone

³H-I

4-[1-(4-fluorobenzyl)pyrrole-2-yl-2,4-diketobutanoic acid

³H-II

³H-4-(3-benzylphenyl)-2,4-dioxo-butyric acid