Discovery of Novel Small-Molecule HIV-1 Replication Inhibitors That Stabilize Capsid Complexes

Abstract

The identification of novel antiretroviral agents is required to provide alternative treatment options for HIV-1-infected patients. The screening of a phenotypic cell-based viral replication assay led to the identification of a novel class of 4,5-dihydro-1H-pyrrolo[3,4-c]pyrazol-6-one (pyrrolopyrazolone) HIV-1 inhibitors, exemplified by two compounds: BI-1 and BI-2. These compounds inhibited early postentry stages of viral replication at a step(s) following reverse transcription but prior to 2 long terminal repeat (2-LTR) circle formation, suggesting that they may block nuclear targeting of the preintegration complex. Selection of viruses resistant to BI-2 revealed that substitutions at residues A105 and T107 within the capsid (CA) amino-terminal domain (CA_NTD) conferred high-level resistance to both compounds, implicating CA as the antiviral target. Direct binding of BI-1 and/or BI-2 to CA_NTD was demonstrated using isothermal titration calorimetry and nuclear magnetic resonance (NMR) chemical shift titration analyses. A high-resolution crystal structure of the BI-1:CA_NTD complex revealed that the inhibitor bound within a recently identified inhibitor binding pocket (CA_NTD site 2) between CA helices 4, 5, and 7, on the surface of the CA_NTD, that also corresponds to the binding site for the host factor CPSF-6. The functional consequences of BI-1 and BI-2 binding differ from previously characterized inhibitors that bind the same site since the BI compounds did not inhibit reverse transcription but stabilized preassembled CA complexes. Hence, this new class of antiviral compounds binds CA and may inhibit viral replication by stabilizing the viral capsid.

INTRODUCTION

The advent of highly active antiretroviral therapy has led to significant reductions in morbidity and mortality associated with HIV/AIDS. There are currently 26 FDA-approved drugs for the treatment of HIV-1 (1). These drugs fall into six distinct classes that target different sites on 4 of the 15 viral proteins, in addition to one host protein. Although these drugs are generally effective, poor adherence, toxicity associated with long-term treatment, and multidrug resistance can ultimately limit their efficacy. The identification of novel inhibitors of HIV-1 replication that exhibit novel mechanisms of action and favorable resistance and safety profiles will expand potential treatment options.

The viral Gag polyprotein mediates the assembly and budding of immature virions (2–4). As the virus buds, Gag is cleaved by the viral protease to create a series of smaller proteins (MA, CA, and NC) and peptides (SP1, SP2, and p6). The newly processed proteins then rearrange in a process called maturation. Mature virions contain a conical core particle that has an outer shell (the “capsid”) composed of CA subunits. The capsid surrounds a ribonucleoprotein complex comprising the viral RNA genome, the NC protein, and the viral enzymes reverse transcriptase (RT) and integrase (IN) (2, 3). The conical capsid lattice follows the geometry of a fullerene cone, with ∼200 CA hexamers comprising the body of the cone and the required declination provided by 12 CA pentamers: 7 at the wide end and 5 at the narrow end (5, 6). The amino-terminal domain of CA (CA_NTD, amino acid residues 1 to 146) forms the hexameric (or pentameric) rings, whereas the carboxyl-terminal domain of CA (CA_CTD, amino acid residues 151 to 231) forms a “belt” around the rings and makes dimeric interactions that connect adjacent rings (7–9).

Amino acid substitutions within HIV-1 CA can impair either the late-stage event of virion assembly or early postentry events such as reverse transcription, capsid uncoating, and/or nuclear entry (2, 10–12). Two observations of particular relevance to the current study are that (i) CA amino acid substitutions such as E128A/R132A that appear to stabilize the viral capsid also reduce the efficiency of reverse transcription (12), and (ii) other detrimental CA amino acid substitutions, such as Q63A/Q67A, can increase the levels of CA associated with the preintegration complex (PIC), suggesting that they may impair capsid uncoating (13).

There is growing interest in HIV-1 CA as a target of antiviral inhibitors, and several peptides and small molecules that bind CA and inhibit viral replication have been identified (reviewed in reference 14). A phage display approach led to the identification of a peptide that binds the CA_CTD and inhibits the in vitro assembly of both immature and mature particles (15, 16). A small molecule, CAP-1, was shown to target a pocket (site 1) at the base of the CA_NTD formed by helices 1 to 4 (17, 18), and more potent inhibitors that bind this pocket have subsequently been reported (19–21). All of these compounds inhibit CA assembly in vitro but can have distinct effects in inhibiting either virion production or capsid assembly (20). A distinct family of small molecules was recently reported to bind to a separate site on CA_NTD, site 2, formed by helices 3, 4, 5, and 7 (22). These compounds perturb viral capsid assembly and appear to both enhance the rate of CA multimerization in vitro and accelerate capsid dissociation in cells (22, 23).

Here we describe a new family of 4,5-dihydro-1H-pyrrolo[3,4-c]pyrazol-6-one (pyrrolopyrazolone) small molecules that bind within CA_NTD site 2 and inhibit HIV-1 replication. These compounds differ from previously reported site 2 inhibitors (22, 23) since they stabilize HIV-1 CA assemblies and prevent uncoating of viral capsids in vitro. The pyrrolopyrazolones also reduce the accumulation of 2 long terminal repeat (2-LTR) circles without affecting reverse transcription in infected cells. Thus, these compounds may exert their primary effect at the level of nuclear import of the PIC and therefore represent a new class of anti-HIV inhibitors.

MATERIALS AND METHODS

Plasmids.

Plasmids used to produce vesicular stomatitis virus envelope glycoprotein (VSV-G) pseudotyped replication-incompetent HIV-1 (HIV-1 helper plasmid, pTV-Luc transfer plasmid, and pIADL-VSV-G envelope plasmid) and to express HIV-1_NL4-3 CA-NC were described previously (20). BI-2 drug resistance mutations were introduced into the HIV-1 helper plasmid and CA-NC expression plasmids using the QuikChange II site-directed mutagenesis kit (Agilent Technologies), according to the manufacturer's instructions.

Cells.

C8166, SupT1, and 293FT cells were obtained from John Sullivan (University of Massachusetts Medical Center), ATCC (CRL-1942), and Life Technologies (R700-07), respectively, and maintained at 37°C, 5% CO₂. The establishment of the C8166-LTR-Luc cells was described previously (20). C8166, C8166-LTR-Luc, and SupT1 cells were maintained in RPMI 1640 medium (Wisent Inc.) supplemented with 10% fetal bovine serum (FBS; HyClone) and 1% penicillin-streptomycin (Life Technologies). C8166-LTR-Luc cells were cultured in the presence of 5 μg/ml blasticidin S-HCl (Life Technologies), but blasticidin S-HCl was omitted from all antiviral assays. 293FT cells were maintained in Dulbecco's modified Eagle medium (DMEM; Wisent Inc.) supplemented with 10% FBS and 1% penicillin-streptomycin.

Preparation of test compounds for antiviral assays.

Ten-point serial dilutions of test compounds were prepared in RPMI 1640 medium supplemented with 10% FBS and 1% penicillin-streptomycin. All samples, including the negative and positive controls, contained the same concentration of dimethyl sulfoxide (DMSO) in assay medium (≤0.5%). Diluted compound (30 or 100 μl) was added to quadruplicate or triplicate wells of 384- or 96-well assay plates, prior to single-cycle (384-well) or multicycle (96-well) viral replication assays, respectively.

Single-cycle viral replication assays.

The activity of the BI-1 and BI-2 inhibitors in the postentry phase of the HIV-1 replication cycle was evaluated by infecting SupT1 cells with VSV-G pseudotyped viral vectors in the presence of the compounds. VSV-G pseudotyped HIV-1 was prepared as described previously (20). Cells and virus were mixed and added to 384-well assay plates (30 μl of medium containing 15,000 SupT1 cells and virus equivalent to 1 ng of CA). Uninfected cells were included as negative controls. Seventy-two hours postinfection, firefly luciferase activity was evaluated by adding 10 μl/well of Steady-Glo (Promega). Luminescence was measured using a TopCount-HTS plate reader (PerkinElmer Life Sciences). The activity of the BI-1 and BI-2 inhibitors in the postintegration (or viral assembly) phase of the HIV-1 replication cycle was evaluated by adding test compounds to 293FT cells during the production of VSV-G pseudotyped HIV-1, exactly as described previously (20). The 50% effective concentration (EC₅₀) is the concentration of compound that produced half-maximal inhibition of HIV-1 viral replication in both assays. EC₅₀ values were calculated by nonlinear regression analysis using SAS software (SAS Institute).

Multiple-cycle viral replication assays.

Antiviral activity in a 72-h multicycle viral replication assay was determined using C8166-LTR-Luc cells and HIV-1 2.12, exactly as described in reference 20.

MTT cytotoxicity assays.

The antiviral and cytotoxicity activities of compounds were measured in parallel. Fifty thousand SupT1 or C8166-LTR-Luc cells in 100 μl of media were added to 96-well plates that contained 100 μl of diluted compound. After 72 h, mitochondrial function and cell viability were assessed by evaluating the reduction of the tetrazolium salt MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl-tetrazolium bromide; Sigma-Aldrich) with the addition of 40 μl of a 5-mg/ml MTT solution in RPMI 1640 medium lacking FBS. Assay plates were incubated 3 to 5 h, 150 μl of medium was removed, and the formazan product was solubilized following the addition of 50 μl 10% Triton X-100 in 0.01N HCl. Formazan concentrations were quantified by measuring the optical density at 570 nm using a SpectraMax plate reader (Molecular Devices). The 50% cytotoxic concentration (CC₅₀) represents the concentration of test compound that reduced cell viability by 50%. CC₅₀ values were calculated by nonlinear regression analysis using SAS software (SAS Institute).

Quantitative PCR assays.

Twenty-four hours prior to infection, 40,000 C8166 cells/well were seeded in 96-well plates (Corning) precoated with 25 μg/ml poly-d-lysine (Sigma-Aldrich). Immediately prior to infection, VSV-G pseudotyped HIV-1 virus was treated with DNase I (Life Technologies) for 60 min at 37°C. Cells were infected for 2 h at 37°C with virions (equivalent to 25 ng of CA per well) in the presence of DMSO, 1.1 μM nevirapine, 70 μM BI-1, or 20 nM of the integrase strand transfer inhibitor L-870,810 (24) (in a volume of 200 μl). Cells were then washed once with phosphate-buffered saline (PBS) to remove unbound virus, and fresh inhibitor was applied. At 10 or 24 h postinfection, the cells were washed once with PBS and total DNA was extracted using the DNA blood minikit (Qiagen) according to the manufacturer's instructions. DNA was eluted in 200 μl AE buffer (supplied with the kit), and 5-μl samples were used in quantitative PCRs. Primers and TaqMan probes specific for late reverse transcription (RT) products or 2-LTR circles were multiplexed with primers and TaqMan probes specific for mitochondrial DNA, as described previously (25). Absolute copy numbers were quantified from standard curves prepared using cloned PCR products. Late RT products and 2-LTR levels were normalized against mitochondrial levels in each sample.

Selection of HIV-1 variants resistant to BI-2.

C8166 cells were infected with HIV-1 NL4.3, and resistance to BI-2 was selected starting at an inhibitor concentration of approximately twice its EC₅₀ (5 μM), in accordance with the protocol described previously (20). Resistance mutations were identified by clonal sequencing.

Capsid assembly assays.

Purification of CA-NC (wild type and A42D) and the immobilized CA-NC assembly assay were performed as described previously (20).

EM analyses of CA-NC tubes.

CA-NC tubes were assembled in solution by incubating 9 μM CA-NC and 1 μM oligonucleotide [d(TG)₅₀] in 100 μl of assay buffer containing 50 mM Tris-Cl (pH 8.0), 400 mM NaCl, and 5 mM β-mercaptoethanol, in the presence of either 20 or 200 μM BI-2 or DMSO alone (vehicle control), for 16 h at 4°C. Five-microliter suspensions of assembled CA-NC tubes were processed and analyzed by electron microscopy (EM) as described previously (26).

Capsid stabilization assay.

Reacti-Bind NeutrAvidin-coated black 384-well microplates (Pierce) were washed once with 80 μl/well of buffer B (50 mM Tris-Cl [pH 8.0], 350 mM NaCl, 10 μM ZnSO₄, 0.0025% 3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfate [CHAPS] (wt/vol), 50 μg/ml bovine serum albumin [BSA], 1 mM dithiothreitol [DTT]). To each well was added a 25 nM solution (50 μl/well) of 5′-biotin-labeled d(TG)₂₅ oligonucleotide (Integrated DNA Technologies) in buffer C (same composition as buffer B, except containing 5 mg/ml BSA). Following an overnight incubation (∼12 to 16 h), the wells were washed twice with buffer B (80 μl/well). CA-NC complexes were assembled by incubation of 2 μM CA-NC protein in buffer B supplemented with 100 nM 5′-fluorescein-labeled d(TG)₂₅ oligonucleotide (Integrated DNA Technologies) in a total volume of 60 μl/well for 2 h at room temperature. Unassembled material was removed by washing once with 80 μl/well of buffer B. Test compounds (60 μl/well), serially diluted in buffer D (same composition as buffer B, except containing 250 mM NaCl) to a final DMSO concentration of 1%, were added to these preassembled CA-NC complexes. Following a 2-h incubation at room temperature, disassembled material was removed by two washes with buffer B (80 μl/well). During this disassembly incubation, complexes spontaneously disassembled at a given rate, and the amount of fluorescein-labeled oligonucleotide remaining in each well, which is proportional to the stability of the complexes, was determined following the addition of 80 μl/well of buffer B supplemented with 0.1% SDS (wt/vol) and quantification of fluorescence on a Victor² plate reader (PerkinElmer Life Sciences). The negative-control wells contained no compound but were incubated for 2 h to assess the extent of disassembly during that period. The positive-control wells had no compound, and the disassembly incubation was omitted. The fluorescence associated with these wells was determined following the addition of buffer B supplemented with 0.1% SDS (wt/vol) and represented 100% of the assembled CA-NC complexes present prior to the disassembly incubation. The capacity of a test compound to stabilize capsid complexes (i.e., inhibit the dissociation of the preassembled CA-NC complexes) was considered to be proportional to the captured fluorescence remaining following the disassembly incubation. For each well, the percent stabilization was calculated using the following equation: % stabilization = {1 − [(F_{positive control} − F_{test well})/(F_{positive control} − F_{negative control})]} ∗ 100, where F_{positive control} and F_{negative control} are the averages of each fluorescence (F) value from a series of wells that were incubated under the relevant conditions, as described above. In the absence of compound, the percent stabilization is 0%, and in the absence of the disassembly incubation, the percent stabilization is 100%. The 50% inhibitory concentration (IC₅₀) is the concentration of compound that produced half-maximal stabilization of CA complexes. IC₅₀ values were calculated by nonlinear regression analysis using SAS software (SAS Institute).

ITC.

Isothermal titration calorimetry (ITC) was performed using 200 μM CA_NTD and CA_CTD-NC in a buffer containing 50 mM Tris-Cl (pH 8.0), 350 mM NaCl, and 1% DMSO using a VP-ITC microcalorimeter (GE Healthcare Life Sciences). Titrations were performed at 25°C, as described previously (20).

NMR spectroscopy.

To prepare ¹⁵N- and ¹³C/¹⁵N-labeled CA_NTD, the proteins were expressed, as described previously (20), in Spectra 9 media (¹⁵N, 98% and ¹³C, 98%; ¹⁵N, 98%) (Cambridge Isotope Laboratories Inc.). At the last step of purification, the proteins (at 720 μM) were exchanged into 25 mM sodium phosphate buffer (pH 5.5) and 2 mM deuterated-DTT using a NAP-5 column (GE Healthcare Life Sciences). For ¹H-¹⁵N chemical shift perturbation studies, 12.5 μl of ¹⁵N CA_NTD was added to 527 μl of 25 mM sodium phosphate (pH 5.5) in H₂O, followed by 60 μl of D₂O. To this sample, increasing amounts of a 10 mM stock solution of BI-1 in DMSO-d6 were added. The CA_NTD–BI-1 final concentrations were as follows: 15 μM CA_NTD and 0 μM BI-1; 15 μM CA_NTD and 15 μM BI-1; 15 μM CA_NTD and 30 μM BI-1; 15 μM CA_NTD and 60 μM BI-1; and 15 μM CA_NTD and 120 μM BI-1.¹H-¹⁵N nuclear magnetic resonance (NMR) experiments were acquired at 35°C using a 600-MHz Avance II spectrometer (Bruker BioSpin) equipped with a 5-mm z-gradient triple-resonance cryoprobe. Two-dimensional ¹H-¹⁵N heteronuclear single quantum coherence (HSQC) and ¹H-¹⁵N transverse relaxation optimized spectroscopy (TROSY) spectra were acquired using standard pulse sequences and parameters. Backbone resonances were assigned using a 1 mM ¹³C/¹⁵N-labeled HIV-1 CA_NTD (residues 1 to 146) sample in 25 mM sodium phosphate (pH 5.5) via standard triple-resonance assignment strategies as previously reported (27). All NMR data were processed and analyzed using Topspin (Bruker BioSpin). Binding isotherms from ¹H-¹⁵N NMR titration experiments were calculated using Prism 4 software (GraphPad Software, Inc.).

X-ray crystallography.

Cocrystallization with BI-1 was carried out in a similar manner as described previously (20, 28, 29). A small molecule, BI-3, which is structurally distinct from BI-1 and BI-2 and binds to a different site on CA_NTD, was used to facilitate crystallization of the CA_NTD (29). A protein solution comprised of 15 mg/ml CA_NTD (residues 1 to 146), 15 mM morpholineethanesulfonic acid (MES) (pH 6.5), 1.5 mM BI-1, 1.5 mM BI-3 (PDB ligand 0OE), and 2% DMSO was mixed in a 1:1 ratio with a solution containing 22.5% PEG8000, 0.1 M TAPS [N-Tris(hydroxymethyl)methyl-3-aminopropanesulfonic acid] (pH 9.0), and 0.1 M sodium acetate, microseeded, and suspended over a reservoir containing the same solution. Large, block-like crystals appeared within a day and continued to grow for 1 to 2 weeks. Immediately prior to flash-freezing in liquid nitrogen, individual crystals were briefly transferred to a solution containing 22.5% PEG8000, 0.1 M TAPS (pH 9.0), 0.1 M sodium acetate, 1.5 mM BI-1, and 12% DMSO for cryoprotection. Data were collected at 100 K on an FR-E X-ray generator equipped with Osmic HiRes² optics and an MA345dtb image plate detector. Data reduction and scaling were performed using HKL2000 (30). Preliminary models were obtained via rigid body refinement in Phenix using a model ultimately derived from PDB ID 1GWP (31). Further iterations of refinement using Phenix and manual model building using Coot (32) yielded the final model. The data processing and final model refinement statistics are provided in Table S1 in the supplemental material. The stereochemical quality of the model was assessed with the program Molprobity (33); there were no Ramachandran outliers.

Protein structure accession number.

The coordinates for the CA_NTD-BI-1 structure have been deposited in the PDB under accession code 4J93.

RESULTS

Identification of a small-molecule inhibitor of HIV-1 replication.

A novel class of HIV-1 inhibitors was identified using a cell-based screen that employed a single-cycle HIV-1 reporter vector pseudotyped with the vesicular stomatitis viral envelope glycoprotein (VSV-G). Compounds were added at the time of infection and could therefore target only the early stages of replication (postentry events). A preselected subset of ∼60,000 compounds that reflected the diversity and complexity of our entire corporate compound collection were screened. One of the hits identified in this screen, BI-1, a representative of the pyrrolopyrazolones, was active in both single and multicycle viral replication assays with EC₅₀ values of 8.2 and 7.5 μM, respectively (Table 1). A more potent analogue, BI-2, was identified from our full compound collection using similarity searches. BI-2 displayed EC₅₀s of 1.8 and 1.4 μM, in single and multicycle viral replication assays, respectively (Table 1). Neither BI-1 nor BI-2 altered the production of infectious virions when added to producer cells (Table 1), implying that these compounds were not active during the late phase of the HIV-1 replication cycle (postintegration to virus maturation). Thus, the pyrrolopyrazolone family of compounds targets the early stage of the HIV-1 replication cycle.

Table 1.

Antiviral profiles of BI compounds 1 and 2^a

^aValues represent the averages of at least three independent experiments; standard deviation of the mean is indicated in parentheses.

Mechanistic studies.

Given that the pyrrolopyrazolones are active only during the early phase of the replication cycle, BI-1 and BI-2 were tested for activity against viruses with known nonnucleoside reverse transcriptase inhibitor (NNRTI) resistance mutations. Both compounds maintained activity against all mutant viruses tested (see Table S2 in the supplemental material). To begin to define the mode of action of the pyrrolopyrazolone compounds, quantitative PCR (qPCR) assays were performed to monitor the accumulation of late reverse transcription products and 2-LTR circles within cells infected with VSV-G pseudotyped HIV-1 and treated with BI-1. Positive controls for this experiment included treatment with the NNRTI nevirapine (NVP) and the integrase strand transfer inhibitor L-870,810 (24). Late reverse transcript levels and 2-LTR circles were normalized against mitochondrial DNA levels (mtDNA) present in each sample and are reported relative to a DMSO control sample (Fig. 1). Both positive controls behaved as expected: NVP inhibited reverse transcription (Fig. 1A) and consequently, also reduced the accumulation of 2-LTR circles (Fig. 1B), whereas L-870,810 had no effect on reverse transcription (Fig. 1A) but enhanced the levels of 2-LTR circles (Fig. 1B). As shown in Fig. 1A, BI-1 did not affect the production of late reverse transcripts but did reduce the levels of 2-LTR circles by nearly 1 order of magnitude at both 10 and 24 h postinfection (Fig. 1B). BI-1 did not inhibit either RT or IN enzyme activity in a series of biochemical assays or PIC activity in assays performed using lysates from infected cells treated with inhibitors (data not shown). Instead, the reduction in 2-LTR circles and the lack of an effect on either reverse transcription or integration suggested that BI-1 might inhibit nuclear import of the PIC.

Fig 1.

Effects of BI-1 on late reverse transcription and 2-LTR circle formation. C8166 cells were infected in the presence of DMSO, 70 μM BI-1, 1.1 μM nevirapine (NVP), or 20 nM L-870,810 for 2 h at 37°C. Cells were washed with PBS, and fresh inhibitor or DMSO was added to the cells. The cells were harvested 10 or 24 h postinfection, and total DNA was isolated and used in quantitative PCRs specific for late RT (A) or 2-LTR circle (B) products. DNA levels were normalized to mitochondrial DNA levels and then to HIV-1 DNA levels in control, DMSO-treated cells (defined as 100%). Shown are data from a representative experiment (of two independent experiments) with the averages and standard deviations of the means for duplicate samples.

CA substitution mutations confer BI-2 resistance.

Resistance selection studies were performed to identify the target of the BI compounds. To select for resistance, HIV-1 NL4-3 was cultured in C8166 cells in the presence of increasing concentrations of the more potent BI-2 inhibitor. As the cytopathic effect escalated (implying viral replication), the drug concentration was increased 2-fold. Proviral DNA was isolated, amplified, and sequenced at passages 4, 6, 8, 10, and 12. These analyses revealed the accumulation of substitution mutations at positions A105 and T107 within the N-terminal domain of CA. The initial mutation to emerge was an A105T substitution. This mutation was still maintained during passages 8 to 12 but was found at lower frequency as additional mutations emerged at position 107. Viruses with T107 substitutions, either T107A or T107N, emerged during these later passages, and ultimately these became the most prevalent mutations in the population.

To confirm that the A105T, T107A, and T107N substitutions were responsible for conferring resistance to the BI compounds, each mutation was introduced separately into an otherwise wild-type HIV-1 vector. None of the individual resistance mutations affected viral replication significantly in the absence of inhibitor treatment (data not shown). However, both BI-1 and BI-2 were less effective against all three mutants than against the wild-type virus, thereby confirming that these substitutions were responsible for conferring inhibitor resistance (Table 2). Specifically, the EC₅₀ for BI-1 increased 6-fold against the T107A mutant virus and more than 8-fold against the A105T and T107N mutants, remaining inactive at 70 μM, the highest concentration of BI-1 tested. Similarly, the EC₅₀ for BI-2 increased 5- to 10-fold against all three mutant viruses (Table 2). In contrast, each of the mutant viruses maintained susceptibility to NVP (Table 2), demonstrating that resistance was specific for the BI compounds. The identification of drug resistance mutations within the CA_NTD implies that CA is the functional target of the BI compounds.

Table 2.

Susceptibility of viruses containing CA resistance mutations to antiviral inhibition by BI-1 and BI-2

Virus inhibitor	Avg EC50 fold change (SD)a
	CA A105T	CA T107A	CA T107N
BI-1	>8	>6	>8
BI-2	5 (1)	7 (0)	10 (4)
Nevirapine	0.8 (0.2)	1 (0.5)	0.8 (0.4)

^aThe average EC₅₀ fold change values represent the ratio of the EC₅₀ against the mutant virus to the EC₅₀ against the wild-type virus. Values represent the averages of at least 3 independent experiments, except for T107A, for which the average of 2 independent experiments is shown.

BI compounds bind directly to CA_NTD.

Isothermal titration calorimetry and NMR chemical shift experiments were used to test whether the BI compounds bound directly to pure recombinant CA_NTD. As summarized in Table 3, ITC measurements demonstrated that BI-1 and BI-2 bound to CA_NTD in a 1:1 stoichiometry, with dissociation constants (K_d) of 20 and 3.0 μM, respectively. Control experiments with a CA_CTD-NC fusion protein showed no compound binding (not shown), indicating that the interactions observed for CA_NTD were specific.

Table 3.

Inhibitor binding to CA_NTD measured by ITC^a

Inhibitor	Kd(μM)	ΔG(kcal/mol)	ΔH(kcal/mol)	−TΔS(kcal/mol)
BI-1	20	−6.3	−15.3	9.0
BI-2	3.0	−7.5	−14.9	7.4

^aValues represent the averages of two and three independent experiments for BI-1 and BI-2, respectively. ΔG, Gibbs free energy change; ΔH, enthalpy change; T, temperature; ΔS, entropy change.

Binding of BI-1 to CA_NTD was also evident in NMR chemical shift perturbation studies (Fig. 2). Figure 2A shows overlays of two-dimensional (2D) ¹H-¹⁵N HSQC TROSY spectra of ¹⁵N-labeled CA_NTD in the presence of increasing concentrations of BI-1. Although most resonances were unaffected by the addition of BI-1, a significant subset were shifted or broadened beyond detection, indicating site-specific inhibitor binding. Residues that were maximally affected by inhibitor binding are labeled in Fig. 2A, and their positions are mapped onto the structure of the free CA_NTD protein in Fig. 2B. The shifted residues cluster on one face of CA_NTD defined by helices 3, 4, 5, and 7. ¹H and ¹⁵N chemical shift changes (Δδ) for residues D103, G106, T108, and L138 were plotted as a function of inhibitor concentration and fit to a 1:1 binding model, yielding an average K_d of 20 ± 2.2 μM (Fig. 2C). Thus, our NMR and ITC measurements of BI-1 compound binding to CA_NTD are in excellent agreement and together reveal that both BI-1 and BI-2 bind directly and specifically to the amino-terminal domain of HIV-1 CA.

Fig 2.

Identification of the inhibitor binding site within CA_NTD. (A) Overlay of 2D ¹H-¹⁵N HSQC TROSY spectra obtained upon titration of BI-1 into CA_NTD (15 μM ¹⁵N CA_NTD, residues 1 to 146). Color coding and CA_NTD to BI-1 ratios were as follows: black, 1:0; red, 1:1; cyan, 1:4; blue, 1:8. Backbone NH resonances that shift upon inhibitor addition are labeled with the corresponding residue number. (B) Ribbon diagram showing the CA_NTD structure and highlighting the residues most affected by BI-1 binding. Significantly shifted resonances were F32, A47, Q50, L52, T54, L56, N57, T58, G60, G61, H62, Q63, K70, E71, T72, N74, A77, D81, D103, A105, G106, T107, T108, I134, and L138. (C) ¹H and ¹⁵N NMR chemical shifts for CA residues D103, G106, T108, and L138. Fits show 1:1 binding models corresponding to the designated dissociation constants (average K_d, 20 ± 2.2 μM).

Crystal structure of the BI-1:CA_NTD complex.

Cocrystals of CA_NTD and BI-1 produced high-quality diffraction data to 1.74-Å resolution, and the structure was refined to 19.7% R_cryst and 23.3% R_free (see Table S1 in the supplemental material for full details). Density for BI-1 was clear and revealed that the inhibitor bound in a relatively shallow binding pocket formed by residues from helices 3, 4, and 5 (Fig. 3), in good agreement with the ¹H-¹⁵N HSQC TROSY data (Fig. 2). The pocket appears to be preformed because BI-1 binding did not induce significant protein backbone movements. The pyrrolopyrazolone core stacks against the amide side chain of CA residue N53, parallel to the surface of the protein, and buttressed on both sides by residues N57 and T107. The N57 side chain becomes more ordered upon inhibitor binding, forming one direct and one water-mediated hydrogen bond with the N2 and N1 nitrogens of the inhibitor, respectively. The T107 side chain rotates to form a hydrogen bond between the T107 Oγ and the inhibitor N5 atom. A water-mediated hydrogen bond is also formed between the carbonyl oxygens of BI-1 and the G106 backbone. The C3 phenyl substituent of BI-1 is rotated 12° relative to the pyrrolopyrazolone core and sits in a hydrophobic environment created by the side chains of L56, M66, L69, K70, and I73. It is clear that only the S enantiomer at position C5 of BI-1 can bind CA_NTD because inversion of the chiral center would produce a highly unfavorable steric clash between the pyridine ring and the protein. The C5 pyridine is oriented perpendicular to the plane of the inhibitor, and the nitrogen likely faces the solvent, although carbon and nitrogen cannot be discriminated at this resolution. The pyridine binding pocket is more open than the C3 phenyl binding pocket and is formed primarily by residues K70, I73, N74, and Y130.

Fig 3.

Interactions of BI-1 within the CA_NTD site 2 pocket. Semitransparent surface representation of the BI-1 binding pocket, shaded from dark to light by increasing solvent exposure. BI-1 (thick sticks) and residues forming the pocket (thin sticks) are colored by atom type. Water molecules are indicated by red spheres and hydrogen bonds by cyan lines.

BI-1 promotes HIV-1 CA-NC tube assembly.

Pure recombinant CA and CA-NC proteins can assemble in vitro, forming helical tubes that are composed of hexagonal arrays of CA rings that mimic the surface lattice of the viral capsid (6, 34). The influence of BI-1 binding on CA-NC assembly was analyzed using a fluorescence assay that measured the degree of CA-NC assembly on the surface of NeutrAvidin-coated microplates (20). CA-NC assembly was nucleated by immobilized, biotin-conjugated single-stranded d(TG)₂₅ oligonucleotides, and the extent of assembly was quantified by measuring the levels of soluble, fluorescein-labeled d(TG)₂₅ oligonucleotides that remained incorporated following a wash step. Remarkably, BI-1 stimulated CA-NC assembly in a concentration-dependent manner in this assay, resulting in a 2-fold increase in the fluorescence signal at a BI-1 concentration of 25 μM and an NaCl concentration of 350 mM (Fig. 4A). CA-NC assembly is salt dependent, and BI-1 also stimulated CA-NC assembly at NaCl concentrations of 250 mM and 200 mM NaCl, with maximal increases of 9- and 5-fold, respectively (Fig. 4B).

Fig 4.

BI-1 and BI-2 enhance assembly and stability of CA-NC tubes. (A) Assembly of wild-type and A42D CA-NC was assayed in the presence of increasing concentrations of BI-1 (2-fold dilutions, 1.6 to 25 μM) in a buffer containing 350 mM NaCl. Assembly assays were performed in NeutrAvidin-coated microplates, precoupled with d(TG)₂₅ 5′ biotin-labeled oligonucleotide, by incubating 2 μM CA-NC proteins and 100 nM d(TG)₂₅ 5′ fluorescein-labeled oligonucleotide for 2 h at 23°C. Assembled CA-NC tubes were quantified by measuring the fluorescein signal remaining following a wash step. (B) Wild-type CA-NC assembly levels at different NaCl concentrations (150, 200, and 250 mM) in the presence of increasing BI-1 concentrations. (C) Assembly of CA-NC tubes in solution [9 μM CA-NC, 1 μM d(TG)₅₀, 400 mM NaCl] during 16 h of incubation at 4°C in the presence of DMSO (control) or 20 or 200 μM BI-2. Suspended CA-NC tubes were visualized by negative-staining electron microscopy. (D) Effects of BI-1 and BI-2 on CA-NC tube stability. CA-NC tubes were preassembled in microplates and washed, the buffer salt concentration was reduced from 350 to 250 mM in serial dilutions of BI-1 or BI-2, and the CA-NC tubes were allowed to disassemble for 2 h at 23°C. The remaining fluorescein signal was measured following a wash step and used to quantify the extent of CA stabilization (0% represents no stabilization). Shown are data from a representative experiment (of at least two independent experiments) with the averages and standard deviations of the means for duplicate samples.

Several lines of evidence indicate that the BI compounds specifically stimulated the formation of authentic CA-NC tubes in this assay. First, the fluorescent signal did not increase when BI-1 was added to a control CA-NC protein that carried the CA A42D amino acid substitution, which has previously been shown to inhibit assembly of the hexagonal CA lattice (Fig. 4A) (10, 11). Second, BI-2 also stimulated the formation of CA-NC tubes in a homogeneous solution assay (Fig. 4C; see Fig. S1 in the supplemental material). In this assay, 9 μM CA-NC and 1 μM d(TG)₅₀ were incubated together in a buffer that contained 300, 400, or 500 mM NaCl and different concentrations of BI-2. CA-NC tube formation was analyzed both by quantifying the amount of pelletable CA-NC tubes (see Fig. S1 in the supplemental material) and by electron microscopy with negative staining (Fig. 4C). Consistent with the results from the heterogeneous plate-based fluorescence assembly assay, the number of CA-NC tubes increased with increasing BI-2 concentrations, and the CA-NC tubes had normal sizes and morphologies (Fig. 4C and data not shown). We therefore conclude that the BI compounds bind HIV-1 CA in a conformation that stimulates assembly of a hexagonal lattice.

BI compounds stabilize HIV-1 CA-NC complexes.

Given that the BI compounds stimulated CA-NC assembly, we hypothesized that they might also stabilize assembled CA-NC tubes. To test this hypothesis, we established an in vitro CA stabilization assay. Briefly, CA-NC tubes were preassembled on NeutrAvidin-coated microplates in 350 mM NaCl and then allowed to disassemble in the presence or absence of compound, in a buffer that contained only 250 mM NaCl. The degree of CA-NC tube stabilization was evaluated by measuring the fluorescence remaining following a final wash step (see Materials and Methods for details). In the absence of compound, ∼70% of the CA-NC tubes disassembled during the low-salt incubation (defined as 0% stabilization). As shown in Fig. 4D, both BI-1 and BI-2 stabilized CA-NC tubes in a concentration-dependent manner. The IC₅₀ values for BI-1 and BI-2 were 19 and 3.3 μM, respectively, in good agreement with their binding affinities for free CA_NTD (Table 3 and Fig. 2C). The compounds also modestly stabilized purified HIV-1 cores (data not shown) against dissociation in vitro (35), consistent with the CA assembly experiments. Taken together, our data demonstrate that BI compound binding must increase the thermodynamic stability of the hexagonal CA lattice because the compounds enhance assembly and inhibit disassembly of CA complexes.

Resistance mutations prevent BI-2 stabilization of CA-NC tubes in vitro.

To test whether the activity of BI-2 in the CA stabilization assay correlated with CA_NTD pocket binding and antiviral activity, we examined the effects of the A105T and T107A resistance mutations on CA-NC stability. The intrinsic activities of the mutant CA-NC proteins were similar to that of the wild-type protein in both the CA assembly and stabilization assays (data not shown). However, BI-2 failed to stabilize either of the mutant CA-NC proteins in the CA stabilization assay (IC₅₀, >30 μM compared to 3.3 μM for WT CA-NC; fold change [FC], >10), consistent with the loss of susceptibility of each of these mutants to BI-2 in viral replication assays (Table 2).

CA residue Asn-57 contributes to BI compound binding by forming a hydrogen bond with the pyrazole ring (Fig. 3). To test whether this residue also contributes to BI-2 activity, we analyzed the CA-NC N57S mutant protein in CA stabilization and viral replication assays. This amino acid substitution reduced CA-NC assembly but did not reduce the intrinsic rate of CA-NC dissociation in the CA stabilization assay (data not shown). Importantly, BI-2 failed to stabilize the CA-NC N57S mutant in the CA stabilization assay (IC₅₀ > 30 μM; FC > 10). Similarly, a CA N57S mutant virus was resistant to BI-2 (EC₅₀ > 65 μM, FC > 36). Thus, it was possible to use the BI-1:CA_NTD structure to introduce an amino acid substitution that conferred resistance to BI-2 stabilization and antiviral activity. The excellent correspondence between the requirements for BI inhibitor binding, activity in the CA assembly and stabilization assays, and antiviral activity implies that inhibitor binding within CA_NTD pocket 2 is responsible for antiviral activity.

DISCUSSION

Cell-based antiviral phenotypic screens can identify novel inhibitors of viral replication that target either viral or cellular targets. Such screens have led to the discovery of anti-HIV-1 inhibitors with novel mechanisms of action (22, 36–40) that probably could not have been identified in more specifically targeted assays. Although it can be challenging to identify their functional targets and mechanisms of action, such compounds can also provide valuable tools for identifying and analyzing key steps in viral replication.

In the present case, we screened for inhibitors of postentry steps in HIV-1 infection. Our screen identified a novel series of pyrrolopyrazolone small-molecule inhibitors that are mechanistically distinct from other classes of HIV-1 antiviral compounds. Quantitative PCR analysis of reverse transcription products and 2-LTR circles indicated that the BI compounds target a step following reverse transcription, but prior to (or concomitant with) nuclear import of the preintegration complex (PIC). Amino acid substitutions that confer inhibitor resistance map to a series of residues that surround a conserved pocket within the N-terminal domain of CA (site 2), and BI-1 was shown to bind in this pocket by ITC, NMR, and X-ray crystallographic analyses. Thus, CA_NTD site 2 is the functional target of the BI inhibitors.

To replicate efficiently, HIV-1 must form capsids of the appropriate stability (3, 10–12), and we speculate that the BI compounds may function, at least in part, by stabilizing the viral capsid. We have shown that the BI compounds accelerate CA-NC tube assembly and stabilize preassembled CA-NC tubes against dissociation. However, the mechanism by which the BI pyrrolopyrazolones stabilize capsid complexes is uncertain because the inhibitor binding site does not reside directly within any inter- or intrahexamer CA interfaces that stabilize the capsid lattice (see Fig. S2 in the supplemental material). It is possible, however, that the compounds act by reducing CA_NTD flexibility and/or stabilizing CA interfaces allosterically. The latter possibility is consistent with the observation that BI-1 binding alters the backbone amide chemical shift environment of several residues that lie outside the binding pocket. Distal shifted residues include Phe32 and His62, which reside within the flexible loops between α1/α2 and α3/α4, respectively, both of which are positioned near intrahexamer CA interfaces. In other contexts, amino acid substitutions that stabilize the HIV-1 capsid can inhibit viral replication at a step between reverse transcription and integration, as was recently shown for the E45A mutant (10–12, 41). Similarly, the Q63A/Q67A mutant exhibits delayed uncoating in target cells and is impaired for PIC nuclear entry and integration (13). Thus, it seems possible that the BI compounds act in a similar manner to prevent nuclear localization of the PIC.

The previously reported PF-3450074 inhibitor also binds in the CA_NTD site 2 pocket (22), and PF-3450074 and BI-1 make similar interactions with many of the same residues (Fig. 5A). Nevertheless, the activities of these two chemically distinct inhibitors appear to differ significantly. Key differences include the following: (i) PF-3450074 is active during both the early and late phases of the replication cycle (22, 23), whereas the BI pyrrolopyrazolones are active only during the early phase; (ii) reverse transcription is inhibited in infected cells treated with PF-3450074 (22) but not in infected cells treated with the BI compounds; and (iii) PF-3450074 stimulates capsid dissociation (23), whereas the BI compounds stabilize CA assemblies. Thus, the two different chemical classes of inhibitors appear to exert distinct effects on early steps of viral replication, despite binding in the same pocket on CA_NTD. Although the structural basis for their differing activities remains to be determined, we note that the indole moiety of PF-3450074 extends beyond the region of site 2 that is occupied by the BI pyrrolopyrazolones (Fig. 5A), resulting in additional contacts with CA residues Met66, Gln67, Lys70, and Gln63 (22). We speculate that these additional contacts beyond the core of site 2 may result in capsid destabilization.

Fig 5.

BI-1, PF-3450074, and CPSF6_313–327 all bind in CA_NTD pocket 2. (A) Superposition of the BI-1:CA_NTD (PDB 4J93) and PF-3450074:CA_NTD (PDB 2XDE) structures. The CA_NTD surface is shown with atomic color coding. The bound BI-1 and PF-3450074 inhibitors (sticks) are shown in green and silver, respectively. To create this figure, the CA_NTD complexes were superimposed, and the figure shows the superimposed inhibitors, mapped onto the CA_NTD structure from the BI-1:CA_NTD complex. Note that the PF-3450074 inhibitor appears to clash sterically with Lys-70 because this side chain occupies a different position in the PF-3450074:CA_NTD structure. (B) Superposition of BI-1:CA_NTD (PDB 4J93) and CPSF6_313–327:CA_NTD (PDB 4B4N) structures. The orientations and color coding of the CA_NTD and BI-1 inhibitors are the same as in Fig. 5A, and the CPSF6_313–327 backbone is shown in tan, with three key binding side chains shown explicitly and with residue labels.

HIV-1 CA binds several host factors that influence HIV-1 activity, including cyclophilin A, TRIM5α, and TNPO3 (reviewed in references 42–46). A role for CA in the nuclear import of the PIC is well established, in part through the interactions of CA with nucleoporins (47–52). We speculate that BI pyrrolopyrazolones binding to CA site 2 may disrupt the interaction of some host factors with capsid cores. TNPO3 binds HIV-1 CA proteins (53, 54) and CA-NC tubes (54) and stimulates core dissociation in vitro (55). TNPO3 depletion reduces viral infectivity (49, 53–62), and CA amino acid substitutions, both within CA_NTD site 2 (T54A, N57A, K70R, N74D, A105T, and T107N) and elsewhere, can render the virus resistant to this effect (49, 54, 60–62). The interaction between TNPO3 and CA has not yet been characterized in molecular detail, however, and it is not yet clear how the BI inhibitors would affect TNPO3 binding.

A C-terminally truncated mutant of CPSF6 (CSPF6_1–358) restricts the infectivity of wild-type HIV-1 (47). Like the BI pyrrolopyrazolones, CSPF6_1–358 has no effect on reverse transcription but prevents subsequent accumulation of 2-LTR circles (47, 61, 63). CA_NTD amino acid substitutions within site 2, like N74D, that inhibit CPSF6_1–358 binding, abrogate viral restriction (47, 63, 64), and this amino acid substitution renders HIV-1 noninfectious in macrophages, suggesting that host factor binding to site 2 may be important in some contexts (47). The region of CSPF6_1–358 required for HIV-1 restriction maps to residues 301 to 358 (63), and a peptide encompassing CPSF6 residues 313 to 327 has been cocrystallized in complex with CA_NTD (64). Moreover, CA complexes are stabilized in infected cells expressing CSPF6_1–358 (61, 62). As illustrated in Fig. 5B, CPSF6_313–327 binds in CA_NTD pocket 2, overlapping the BI-1 and PF-3450074 binding sites. PF-3450074 competitively inhibits CPSF6_313–327 binding (64), and it is clear that BI compound binding is also likely to sterically occlude CPSF6_313–327 binding. We therefore suggest that the BI compounds may mimic the action of CSPF6_1–358 (47, 61, 62, 64) and/or TNPO3 (61, 62) and anticipate that the BI compounds will be useful tools for helping to define precisely how CPSF6 functions in viral replication.

In summary, we used a cell-based antiviral screen to identify novel inhibitors of HIV-1 replication that bind and stabilize HIV-1 capsids. Our discovery of a new class of antiviral compounds that bind CA_NTD site 2 and inhibit distinct steps in HIV replication indicates that this pocket warrants continued investigation as an antiretroviral target. Further optimization of BI-1 and BI-2 or the discovery of a new chemical series with increased potency may enable the identification of compounds suitable for the treatment of HIV-1.