Roles of Capsid-Interacting Host Factors in Multimodal Inhibition of HIV-1 by PF74

ABSTRACT

The viral capsid of HIV-1 interacts with a number of host factors to orchestrate uncoating and regulate downstream events, such as reverse transcription, nuclear entry, and integration site targeting. PF-3450074 (PF74), an HIV-1 capsid-targeting low-molecular-weight antiviral compound, directly binds to the capsid (CA) protein at a site also utilized by host cell proteins CPSF6 and NUP153. Here, we found that the dose-response curve of PF74 is triphasic, consisting of a plateau and two inhibitory phases of different slope values, consistent with a bimodal mechanism of drug action. High PF74 concentrations yielded a steep curve with the highest slope value among different classes of known antiretrovirals, suggesting a dose-dependent, cooperative mechanism of action. CA interactions with both CPSF6 and cyclophilin A (CypA) were essential for the unique dose-response curve. A shift of the steep curve at lower drug concentrations upon blocking the CA-CypA interaction suggests a protective role for CypA against high concentrations of PF74. These findings, highlighting the unique characteristics of PF74, provide a model in which its multimodal mechanism of action of both noncooperative and cooperative inhibition by PF74 is regulated by interactions of cellular proteins with incoming viral capsids.

IMPORTANCE PF74, a novel capsid-targeting antiviral against HIV-1, shares its binding site in the viral capsid protein (CA) with the host factors CPSF6 and NUP153. This work reveals that the dose-response curve of PF74 consists of two distinct inhibitory phases that are differentially regulated by CA-interacting host proteins. PF74's potency depended on these CA-binding factors at low doses. In contrast, the antiviral activity of high PF74 concentrations was attenuated by cyclophilin A. These observations provide novel insights into both the mechanism of action of PF74 and the roles of host factors during the early steps of HIV-1 infection.

INTRODUCTION

The emergence of HIV-1 variants resistant to currently approved antiretrovirals necessitates the development of novel classes of inhibitors that possess high levels of genetic barriers to resistance (1). Among viral proteins that are not exploited as an antiviral target, the viral capsid (CA) protein is an attractive target for antiviral interventions (2, 3). CA, a genetically fragile protein (4), exhibits limited tolerance to genetic changes and, hence, would predictably temper the evolution of drug resistance (5). The mutational intolerance of CA is caused by structural and functional constraints (6–8). CA, which is the major virion core structural protein, generated by protease-mediated cleavage of the precursor Gag Pr55 protein, plays essential roles during both particle assembly and disassembly (9, 10). Perhaps it is this genetic fragility that makes CA highly vulnerable to host immune responses, such as CD8-specific adaptive immunity (11) and TRIM5α-mediated intrinsic immunity (12), both of which target highly conserved portions of CA.

Recent work discovered several novel small-molecule compounds that target CA (13–15). Among them, PF-3450074 (hereinafter abbreviated as PF74) has been extensively studied owing to its unique properties (5, 13, 16–26). PF74 (13) was a derivative of a small compound identified in a high-throughput screen for antivirals capable of inhibiting HIV-1 replication (27). PF74 binds directly to the N-terminal domain (NTD) of CA at a preformed pocket made by helices 3, 4, 5, and 7 (13, 28). More recent studies revealed that PF74 binds an intermolecular interface between the NTD of one subunit and the C-terminal domain (CTD) of the neighboring subunit within the same CA hexamer (21, 23, 29). Importantly, the PF74-binding site in CA is shared by the host proteins cleavage and polyadenylation specificity factor 6 (CPSF6) and nucleoporin 153 (NUP153) (20, 21, 23, 28).

HIV-1 inhibition by PF74 appears to involve multiple mechanisms (21, 22, 24, 26). High concentrations of PF74 (more than 5 μM) reduce the level of newly synthesized viral DNA, but lower concentrations (∼1 μM) do not (13, 16, 21, 22, 24, 26). The reverse transcription step is intricately linked with the process of core disassembly (30, 31). It is this uncoating process that high concentrations of PF74 (∼5 to 10 μM) appear to act upon: namely, PF74 at high concentrations irreversibly accelerates the uncoating step to block reverse transcription (16, 18, 22, 23), although recent work using a different experimental approach did not find a change in capsid integrity (26). It appears that PF74-mediated inhibition of HIV-1 can also target a step(s) following the completion of reverse transcription. The magnitudes of the inhibitory effects of PF74 at high concentrations on viral infectivity are not accounted for by the levels of decrease in viral DNA (13, 16, 21, 22, 24, 26). Furthermore, lower concentrations of PF74 (∼2 μM) did block HIV-1 infection but barely affected reverse transcription (21, 22, 24, 26), while reducing the number of two-long-terminal-repeat (2-LTR)-containing circles, a molecular signature of virus nuclear import (22, 24). These observations suggested that reverse transcription and nuclear entry are targeted by PF74. This is consistent with the observation that cell cycle-dependent CA mutant viruses, which can bypass nuclear translocation through nuclear pores, became less sensitive to PF74 (17).

It remains to be determined precisely how PF74 blocks HIV-1 nuclear entry. As described above, one relevant observation is that PF74 binds to the same pocket of CA as NUP153 and CPSF6 (20, 21, 23, 28), host molecules implicated in HIV-1 nuclear entry (8, 32, 33). Matreyek et al. observed that PF74 competed with the C-terminal domain of NUP153 for binding to HIV-1 CA and that its dose-response curve was significantly altered either by NUP153 depletion or ectopic expression of the artificial TRIM-NUP153C restriction factor (20). Furthermore, it was shown that PF74-resistant mutants are less dependent on NUP153 (5). These observations suggested a model in which PF74 antagonizes CA engagement of NUP153 to prevent HIV-1 nuclear entry (20).

In contrast to NUP153, the role of CPSF6 in the antiviral mechanisms of PF74 has yet to be defined. CPSF6 may play a critical role in guiding intracellular viral complexes to utilize a known set of nuclear entry cofactors, such as transportin 3 (TNPO3), NUP358, and NUP153 (28, 32–35). CPSF6 interacts with HIV-1 CA at the same binding site as PF74 (21, 23, 28, 29). Accordingly, the interaction of CPSF6 with HIV-1 CA can be competitively blocked by PF74 both in vitro (20–22, 28) and in vivo (24). However, it remains unclear whether and how the competitive inhibition of CA-CPSF6 interactions by PF74 affects antiviral activity.

The present study was initiated to elucidate the role of CPSF6 in antiviral mechanisms of PF74, hoping that it will also provide novel insights into elusive functions of CPSF6 during HIV-1 replication. We first observed that PF74 exhibited a unique dose-response curve, which was triphasic and consisted of two inhibitory phases encompassing a plateau phase. The slopes of the two inhibitory phases differed significantly from each other in their steepness. The curve at higher PF74 concentrations (5 to 10 μM) displayed the steepest slope value among different antiretrovirals. Preventing CA-CPSF6 interactions by small interfering RNA (siRNA), gene knockout, or CA mutations drastically altered the curve such that it no longer displayed three phases. The potency of PF74 at lower concentrations was also reduced, but the slope value at higher concentrations was not affected. Blocking interactions between CA and cyclophilin A (CypA) also resulted in the loss of the three phases, but CypA differed from CPSF6 in that its depletion shifted the steep curve toward a lower range of drug concentrations, suggesting a protective role of CypA against PF74 at high drug concentrations. Altogether, we conclude that CPSF6 and CypA modulate the viral capsid to significantly affect the antiviral mechanism of PF74 in distinct ways.

MATERIALS AND METHODS

Cloning.

The molecular infectious clones used in the present study are based on the LAI (pBru3) (36, 37) or NL4-3 (pNLX.Luc.R−.ΔAvrII) (38) strains of HIV-1, carrying defective env genes and either the gene for firefly luciferase or the gene for green fluorescent protein (GFP) in place of the nef gene. Standard cloning procedures were used to introduce various CA mutations into these clones, as described in our previous work (39, 40).

Cell culture.

Dulbecco modified Eagle medium (DMEM; Cellgro) supplemented with 10% fetal bovine serum (FBS; Cellgro) and 1× penicillin-streptomycin (Pen-Str; Cellgro) was used to culture adherent cell lines. CPSF6 knockout HEK293T cells that carry an empty expression vector or express the 551-amino-acid isoform of CPSF6 (CPSF6[551]) were described previously (41). RPMI medium (Cellgro) supplemented with 10% FBS (Sigma), 1× Pen-Str and 2 mM 2-glutamine (Cellgro) was used to culture an immortalized T cell line (MT4).

Depletion of cellular genes by RNA interference.

HeLa cells plated at 5 × 10⁵ cells per well of a 6-well plate were transfected with 30 pmol siRNA using Lipofectamine RNAiMAX (Invitrogen). The specific siRNAs used for knockdowns were as follows: CPSF6 siRNA (CGUCAUAAAUCCCGUAGUA), CypA siRNA (GAUGAACUUCAUCCAGACUUU), and NUP153 siRNA(GGACUUGUUAGAUCUAGUUUU). Transfection was repeated on the following day with the same procedure. Four hours after the second transfection, cells were seeded at 5 × 10⁵ cells per 96-well plate for infection. These knockdown procedures were performed as described in our previous study (40) and were shown to reproducibly deplete target proteins (40). NUP358 was stably depleted from HeLa cells by using a puromycin-encoding lentiviral vector (pLKO.1; Addgene) encoding a small hairpin RNA (shRNA) identical to that used in a similar lentiviral vector in our previous work (40). Single-cell clones of HeLa cells stably transduced with the puromycin-encoding lentiviral vector were established by a standard procedure. One of the clones (F1), which had a reduced amount of NUP358 as determined by Western blotting, was used to obtain subclones. Two subclones of the parental F1 clone, F1-5F and F1-8C, were used in this study. The effects of knockdown were determined by examining protein levels as described previously (40), using anti-CPSF6 (15489-1-AP; Proteintech), anti-CypA, (PA1-025; ThermoFisher), anti-NUP358 (ab2938; Abcam), anti-NUP153 (ab24700; Abcam), or anti-tubulin (T6074; Sigma-Aldrich) antibody. Anti-rabbit (65-6120; Invitrogen) and anti-mouse (sc-2005; Santa Cruz) horseradish peroxidase-conjugated secondary antibodies were used in Western blotting for CypA, NUP153, and tubulin. The secondary antibodies used for detection of CPSF6 and NUP358 were IRDye 800CW goat anti-rabbit secondary antibody (LI-COR Biosciences) and IRDye 800CW goat anti-mouse secondary antibody (LI-COR Biosciences).

Generation of PPIA knockout cells.

HeLa cells were transduced with the packaged lentiviral construct lentiCRISPRv2 (utilizing clustered regularly interspaced short palindromic repeats [CRISPR] and CRISPR-associated gene 9 [CRISPR-Cas9] technology) (42), into which the phosphorylated double-stranded oligodeoxyribonucleotide formed by annealing of 5′-CACCGTTCTTCGACATTGCCGTCGA-3′ and 5′-AAACTCGACGGCAATGTCGAAGAA-3′ was inserted, resulting in the expression of the guide RNA 5′-UUCUUCGACAUUGCCGUCGA-3′. Transduced cells were selected in puromycin and cloned by limiting dilution. A control cell line transduced with the empty lentiCRISPRv2 vector was also generated. Clones were expanded, and cytoplasmic extracts were analyzed for CypA expression by immunoblotting using rabbit anti-CypA antibody from Millipore (catalog no. 07-313). Clones lacking detectable CypA were expanded and used for experiments.

For reexpression of CypA in the knockout HeLa lines, we engineered silent mutations in the PPIA sequence targeted by the guide RNA and inserted the corresponding cDNA into the EcoRV site in EF.IRES.blasti, a lentiviral vector containing the eEF1A1 promoter upstream from the cloning site and an internal ribosome entry site (IRES)-blasticidin resistance gene cassette. This vector was engineered by replacement of the cytomegalovirus-red fluorescent protein (CMV-RFP) cassette in EF.CMV.RFP (Addgene) (43) with the IRES-blasti region from pMX-IRES-blasti (Cell Biolabs, Inc.) using the unique EcoRV and XbaI restriction sites. The silent mutations altered CypA codons 8 to 12 to UUU GAU AUA GCA GUG. The vector, which was packaged by cotransfection with psPAX2 and pHCMV-G (43), was transduced into the HeLa knockout clones. Control cell lines were generated by transduction with the empty EF.IRES.blasti vector. Stable populations, which were selected with blasticidin, were analyzed for CypA levels by immunoblotting.

Infection.

Polyethylenimine (PEI; PolySciences) was used to generate virus stocks by transient transfection of HEK293T cells. Plasmid pHCMV-G was cotransfected with env-deficient reporter virus constructs. All infections using adherent cells (HeLa and HEK293T) were performed at 5 × 10⁵ cells per 96-well plate, whereas MT4 cells were seeded at 3 × 10⁶ cells per 96-well plate. Infection of HeLa cells was enhanced by spinoculation (1,200 × g for 30 min) in the presence of 20 μg per ml of DEAE-dextran. For infection experiments using luciferase-encoding viruses, cells were lysed at 2 days after infection with luciferase cell culture lysis 5× reagent (Promega) and used to measure luciferase activity with a luciferase assay kit (Promega) on a luminometer. For infection experiments using GFP-encoding viruses, viral infectivity was determined by using Guava easyCyte (Millipore) 2 days after infection. PF74 was purchased from Glixx Laboratories and Sigma.

Quantitative PCR.

HeLa cells were plated 1 day before infection (1.5 × 10⁶ cells per 6-well plate). Viruses were treated with 80 units per ml (final concentration) of Turbo DNase (Thermo Fisher Scientific) at 37°C for 1 h prior to infection. Infection was done as described above using cells treated with 1.25 or 10 μM PF74, 5 μM nevirapine, or 1.25 μM raltegravir. DNA was extracted from virus-infected cells 24 h after infection using the DNeasy blood and tissue kit (Qiagen) or NucleoSpin 8 tissue kit (Macherey-Nagel). The DNA extracted was used for quantitative PCR to measure the late reverse transcription product using TaqMan universal master mix II (Applied Biosystems). The primers and probe used here were described previously (44). MH531 (5′-TGTGTGCCCGTCTGTTGTGT-3′) and MH532 (5′-GAGTCCTGCGTCGAGAGAGC-3′) were used as the forward and reverse primers, respectively. LRTP (5′-FAM-CAGTGGCGCCCGAACAGGGA-TAMRA-3′ [FAM, 6-carboxyfluorescein; TAMRA, 6-carboxytetramethylrhodamine]) was used as the probe. A standard curve was generated using quantities of plasmids that ranged from 12 to 1.2 × 10⁷ copies per reaction mixture volume. The following program was run on the 7500 Fast real-time PCR system (Thermo Scientific): 50°C for 2 min and 95°C for 10 min, and then 40 amplification cycles of 15 s at 95°C, followed by 1 min at 60°C.

Determination of PF74's affinity for recombinant WT and K182R CA hexamers.

Disulfide bond-stabilized wild-type (WT) HIV-1 and K182R mutant (containing a change of K to R at position 182) CA hexamers were produced by assembly of purified recombinant CA containing four amino acid substitutions, as previously reported (45). Equilibrium dialysis was performed with the rapid equilibrium dialysis plate (Thermo Scientific) by adding 500 μl of various concentrations (0.10 to 1.0 μM) of [³H]PF74, which was produced from triiodo-PF74 as described previously (25), into the buffer chamber and 300 μl of 1.0 μM CA hexameric protein into the sample chamber. Each [³H]PF74 concentration was tested in duplicate. The plates were incubated at 37°C for 24 h with rotation at 100 rpm. Samples were removed from each side of the chamber, and ³H in each sample was quantified by liquid scintillation counting in a TopCount (PerkinElmer) scintillation counter. The [³H]PF74 concentrations were determined by comparison to a reference sample consisting a known quantity of [³H]PF74. K_d (dissociation constant) values were calculated by fitting the data to a one-site binding model with GraphPad Prism software (GraphPad software).

Integration site analysis.

Genomic DNA was extracted from HEK293T cells infected with HIV-1_NLX.Luc.R− in the presence of 1.39 μM PF74 or dimethyl sulfoxide (DMSO) 5 days after infection using the DNeasy blood and tissue kit (Qiagen) for integration site analysis. Illumina sequencing of DNA libraries generated from ligation-mediated (LM)-PCR amplification of genomic DNA was used to determine HIV-1 integration sites essentially as described previously (46, 47). Briefly, DNA digestion by MseI and BglII was performed overnight, followed by purification by using the QIAquick PCR purification kit (Qiagen). Unique double-stranded linkers for each DNA sample were prepared as described previously (47). Overnight incubation allowed linker DNA to be ligated to 1 μg of digested DNA. Linker DNA and HIV-1 U5-specific primers were used in a seminested PCR. Linker primers were unique for each DNA sample but were the same for both PCR rounds, while first- and second-round U5 primers were nested. Illumina adapter sequences were encoded by linker primers and second-round U5 primers, and second-round U5 primers additionally encoded 6-bp indices for multiplexing during sequencing. The sequences of all oligonucleotides used in this study will be made available upon request. PCRs were incubated at 94°C for 2 min, followed by 30 cycles at 94°C for 15 s, 55°C for 30 s, and 68°C for 45 s, which was followed by a final extension for 10 min at 68°C. The QIAquick PCR purification kit was used to purify pooled PCR products. Next-generation sequencing was performed on the Illumina MiSeq platform at the Dana-Farber Cancer Institute Molecular Biology Core Facilities. Sequences were mapped to the hg19 version of the human genome using BLAT, ensuring that the genomic match started immediately after GGAAAATCTCTAGCA, which corresponds to the integrase-processed HIV-1 U5 end. Bioinformatics analysis of HIV-1 integration sites was performed as described previously (46, 47).

Calculation of slope values.

We calculated slope (m) values according to previous work by the Siliciano laboratory (48). The dose-response curves of PF74 under some conditions were triphasic and included two inhibitory phases, as described in Results. Thus, slope values that corresponded to the two inhibitory phases were determined separately. For instance, in order to obtain m values for PF74 concentrations of between 0.313 and 1.25 μM, we set the viral infectivity at 0.313 μM PF74 as 100% and obtained the 50% inhibitory concentrations (IC₅₀s) from the relative infectivities at 0.313, 0.625, and 1.25 μM to derive the m value. Similarly, we calculated m values for higher PF74 concentrations (5 to 10 μM) by normalizing the viral infectivity at 5 μM PF74 to 100% and by just utilizing the infectivity at 10 μM PF74, whereas the IC₅₀s were derived by using the values of 5, 7.5, and 10 μM.

Statistical analysis.

Differences in infectivities or the magnitudes of changes in viral infectivity between various conditions (e.g., between control and knockdown or between WT and mutants) were examined by using the unpaired Student t test. In integration site analysis, P values were calculated by Fisher's exact test or the Wilcoxon rank sum test. P values of less than 0.05 were considered statistically significant.

RESULTS

The dose-response curve of PF74 is triphasic.

We first examined the antiviral activity of PF74 against wild-type (WT) HIV-1_LAI in HeLa cells. PF74 potently inhibited infection by the WT virus, with an IC₅₀ (mean ± standard deviation) of 0.70 ± 0.16 μM and an IC₉₀ of 5.5 ± 5.3 μM in a single-cycle replication assay (n = 14 experiments). A distinctive feature of PF74, which was also shown by previous work (17, 21), was its unique dose-response curve: the curve, which was triphasic, consisted of two inhibitory phases flanking a plateau phase (Fig. 1A, left). Infectivity decreased steadily upon the addition of drug up to a concentration of 1.25 μM PF74, remained unchanged or even increased slightly between 1.25 and 5 μM, and then was further reduced at higher drug concentrations of 7.5 μM and 10 μM (Fig. 1A, left). In contrast, as shown previously (20), the N57A CA mutant virus was completely resistant to PF74 antiviral activity (Fig. 1A, right). Similar dose-response curves were observed using GFP reporter-labeled virus and MT4 T cells (Fig. 1B).

FIG 1.

Triphasic dose-response curve of PF74 displays a steep slope at high doses. (A) Dose-response curves of PF74 for the WT and N57A mutant viruses were generated by infecting HeLa cells with luciferase-encoding reporter viruses in the presence of increasing amounts of PF74. The results are representative of at least five independent experiments. (B) Dose-response curves of PF74 for GFP-encoding reporter viruses were measured using MT4 cells. Mean values from two independent experiments are shown, with error bars denoting the standard errors of the means. (C) The magnitudes of PF74-mediated inhibition of HIV-1 were quantified for two different ranges of drug concentrations by compiling the results from at least 10 independent experiments. Each dot represents the value for one experiment. ***, P < 0.0001. (D) The slope value (m) for each range of drug concentrations was calculated as described in Materials and Methods. The data were generated by compiling the results from at least 10 independent experiments. Each dot represents the value for one experiment, and the bars denote the means of the groups. ***, P < 0.0001.

We found that the steepness of the slopes differed between low (0.313 to 1.25 μM) and high (5 to 10 μM) drug concentrations (Fig. 1A). The degree of PF74-mediated inhibition of HIV-1 was proportional to the drug concentration at low concentrations; a 2-fold increase in drug concentration from 0.313 to 0.625 μM (Fig. 1C) or from 0.625 μM to 1.25 μM (data not shown) resulted in an approximately twofold decrease in viral infectivity. In contrast, the potency of higher PF74 concentrations (more than 5 μM PF74) increased disproportionately; the same twofold increase of the drug dose at a higher range of concentrations (5 to 10 μM) caused an eightfold decrease in viral infectivity (Fig. 1C). The steepness at higher drug concentrations was not due to an artifact, such as cell toxicity, because the PF74-resistant mutant virus N57A was completely unaffected by PF74 treatment both at 5 and 10 μM (Fig. 1A, right, and B).

The steepness of slopes within a dose-response curve can be described by the slope parameter (m). The m value is a critical measure to assess the potency of different classes of antiretrovirals (48, 49), with higher m values showing higher inhibitory potentials (48). We found that the slope values differed significantly between lower and higher concentrations of PF74 (Fig. 1D). The m value at lower PF74 concentrations was an average of 1.7, whereas the m value at higher PF74 concentrations was 4.5, which is among the highest m values exhibited by different classes of antiretrovirals (Fig. 1D) (48). Thus, PF74 is unusual among all the antiretroviral drugs in that its dose-response curve consists of three different phases with different slope values.

Cooperative block at reverse transcription and multistep inhibition contribute to the steep slope at high PF74 concentrations.

The slope (m) values are analogous to the Hill coefficient, which can be used to quantitatively describe the magnitude of cooperativity in ligand binding to an allosteric protein. For drug action, slope values greater than 1 are interpreted as cooperativity (48) and are caused by cooperative drug binding to targets or by multistep inhibition (48, 50). PF74 can block multiple steps, including reverse transcription and nuclear entry (13, 16, 21, 22, 24, 26). To determine the underlying mechanism of the steep slope found for PF74, we quantitatively assessed the effects of PF74 on reverse transcription. We first examined the process of reverse transcription at both low (1.25 μM) and high (10 μM) concentrations (Fig. 2A). Consistent with previous observations (13, 16, 21, 24, 26), high but not low concentrations of PF74 significantly reduced the amount of de novo-generated viral cDNA (Fig. 2A). To further examine the effects of high concentrations of PF74, we measured the copy numbers of viral DNA in cells treated with 5, 7.5, or 10 μM PF74 (Fig. 2B and C). Increasing the PF74 concentration from 5 μM to 10 μM resulted in ninefold and sevenfold decreases in viral infectivity and DNA copy number, respectively. A similar trend was observed in another experiment (4.5-fold and 5.2-fold decreases in infectivity and DNA copy number, respectively; data not shown).

FIG 2.

Cooperative inhibition of reverse transcription by high PF74 concentrations. The progression of reverse transcription was examined by quantitating the amount of late products of reverse transcription in HeLa cells at 24 h after infection. (A) Effects of both low (1.25 μM) and high (10 μM) PF74 concentrations on viral DNA synthesis were examined. Nevirapine (Nev) and raltegravir (Ral) were used as controls. (B) Viral infectivity was quantified by measuring the numbers of GFP-positive cells using flow cytometry in the presence of a range of high PF74 concentrations. (C) DNA extracted 24 h after infection was used for PCR to quantify late products of reverse transcription. Results of quantitative PCR (A and C) are shown as copy numbers normalized to DNA concentrations. Mean values from two independent experiments are shown, with error bars denoting the standard errors of the means.

The defect during the process of reverse transcription can largely account for the steep slope at high PF74 concentrations. However, high PF74 concentrations seemingly affect a step other than reverse transcription: the 15-fold reverse transcription defect at 10 μM PF74 cannot fully account for the 30-fold decrease in viral infectivity. Previous work indicated decreased copy numbers of 2-LTR circles at high PF74 concentrations (22, 24). To obtain an independent piece of evidence for a PF74-mediated block to steps other than reverse transcription, a time-of-addition experiment was performed using PF74 and the nonnucleoside reverse transcriptase inhibitor nevirapine at concentrations that block infection by 2 orders of magnitude. Nevirapine retained over 90% of its inhibitory activity when added up to 4 h after infection and lost activity thereafter, with a midpoint around 8 h after infection (Fig. 3). In contrast, PF74 remained potent, with ∼75% of maximum activity, even when added at 8 h after infection, suggesting that PF74 inhibits a step(s) after reverse transcription. These results thus accord with a model in which the steep slope of the dose response curve at high PF74 concentrations is explained by a cooperative block of reverse transcription, as well as inhibition of step(s) that occur after reverse transcription.

FIG 3.

HIV-1 remains sensitive to inhibition by PF74 longer than to nevirapine. Time-of-addition experiments were performed by infecting HeLa cells with HIV-1 reporter virus encoding luciferase. Two different drugs (1 μM nevirapine or 10 μM PF74) were added at indicated time points. Viral infectivity values, measured by determining luciferase activity of cell lysates prepared 2 days after infection, are shown as relative infectivity normalized to infectivity in the absence of drugs. Mean values from four independent experiments are shown, with error bars denoting the standard errors of the means. **, P < 0.01; *, P < 0.05.

Preventing CA-CPSF6 interactions eliminates the triphasic dose-response curve of PF74.

The observations described above help explain why high concentrations of PF74 display a high slope value. As shown in the current study (Fig. 2) and previous work (21, 24, 26), low concentrations of PF74 do not block reverse transcription, and hence, they must inhibit a subsequent step, such as nuclear entry. This implicates roles of host cofactors for HIV-1 nuclear entry in the antiviral functions of PF74. Among them, NUP153 and CPSF6 seemed particularly relevant, as they bind at a site within CA very similar to that of PF74 (20–22, 28). A role for NUP153 but not CPSF6 in the antiviral activity of PF74 has been examined previously (20). To determine how CPSF6 affects antiviral mechanisms of PF74, we first examined the antiviral activity of PF74 against HIV-1 CA mutants with mutations that block CA interactions with CPSF6 (Fig. 4). These mutations (N74D and T107N) within the CPSF6-binding pocket of CA (28) reduced CA-CPSF6 interactions (25, 28, 32). Our results indicated that the dose-response curves of PF74 for N74D and T107N mutants differed from that for the WT virus and, importantly, did not display three phases (Fig. 4). These changes in the dose-response curves persisted when viral infectivity was normalized (data not shown).

FIG 4.

CA mutations alter the PF74 dose-response curve. HeLa cells were infected with three different luciferase-encoding reporter viruses in the presence or absence of PF74. Viral infectivity was measured by determining luciferase activity of cell lysates prepared 2 days after infection. Results are displayed as relative infectivity normalized to the infectivity in control cells without drug treatment. Mean values from two independent experiments are shown, with error bars denoting the standard errors of the means.

It was difficult to draw a strong conclusion from the data using N74D and T107N mutant viruses because the observed changes in dose-response curves could be caused by reduced CA binding to PF74. For instance, the T107N mutation was shown to reduce CA binding to PF74 (25), although this was not seen in another study (23). To separate PF74 binding to CA from CPSF6 binding to CA, we utilized the K182R mutation, which is located away from the sites where PF74 interacts with CA but nevertheless reduced CA-CPSF6 binding (21, 23). We validated this observation in our hands by showing that the K182R mutant virus was highly resistant to the block in infection exerted by the truncated version of cytoplasmic CPSF6, called CPSF6-358 (Fig. 5A) (32). We next investigated whether the K182R mutation altered PF74 binding by performing equilibrium dialysis using recombinant CA hexamers as described before (25). Importantly, PF74 binding to CA carrying the K182R mutation displayed a K_d of 0.284 μM, which was almost indistinguishable from the binding affinity of the WT CA hexamer (K_d, 0.243 μM) (Fig. 5B). Despite the negligible effects of the K182R mutation on PF74's binding to CA, the same mutation drastically altered the dose-response curve of PF74, which significantly differed from that for the WT virus and did not seem to have three distinct phases (Fig. 5C): the K182R mutant virus had higher infectivity than the WT in the presence of lower concentrations of PF74, while it displayed a comparable level of infectivity with high drug concentrations (Fig. 5C). The altered dose-response curve of PF74 for the K182R mutant was also observed when the slightly attenuated infectivity of the K182R mutant was normalized to that of the WT virus (data not shown).

FIG 5.

K182R mutation changes the PF74 dose-response curve without altering drug binding. (A) Vesicular stomatitis virus G glycoprotein (VSV-G)-pseudotyped GFP reporter viruses were used to infect HeLa cells stably transduced with the control vector or vector ectopically expressing CPSF6-358. Mean values from three independent experiments are shown, with error bars denoting the standard errors of the means. (B) Equilibrium dialysis was used to measure the affinity of PF74 for WT and K182R CA hexamers. One representative data set from two independent experiments is shown. The other experiment showed very similar K_d values (0.278 μM for WT and 0.308 μM for K182R). (C) Dose-response curves of PF74 against the WT or K182R mutant virus were generated by plotting relative viral infectivity values as described in the legend to Fig. 1. Mean values from four independent experiments are shown, with error bars denoting the standard errors of the means.

To directly elucidate a role for CPSF6 in the antiviral activity of PF74, we performed siRNA-mediated depletion of CPSF6. As previously reported (40, 41, 51), CPSF6 depletion by itself did not largely affect HIV-1 infection, with an ∼20% increase in WT virus infectivity being observed. However, we found that CPSF6 depletion counteracted the triphasic dose-response curve of the drug against the WT virus (Fig. 6A and B). Specifically, CPSF6 depletion, which was confirmed by Western blotting (Fig. 6C), decreased the antiviral activity of PF74 at lower concentrations: at concentrations ranging from 0.313 to 2.5 μM, virus infectivity was significantly increased upon CPSF6 depletion (Fig. 6B). CPSF6 depletion also resulted in the loss of the plateau phase at concentrations between 0.625 and 5 μM (Fig. 6A). The effects of CPSF6 depletion on PF74 activity were independent of the multiplicity of infection, as CPSF6 depletion rescued the infectivity of the WT virus over a range of virus inputs (data not shown), and were specific for CA, because CPSF6 depletion did not affect the antiviral activity of PF74 against the N74D (data not shown) or K182R (Fig. 6A and B) mutant viruses. Finally, a similar change in the dose-response curve of PF74 was observed upon CPSF6 knockout in HEK293T cells (Fig. 6D). Importantly, restoration of CPSF6 expression by stable transduction of a vector expressing the 551-amino-acid isoform of CPSF6 (CPSF6[551]) enhanced the antiviral activity of PF74 at low drug concentrations (Fig. 6D).

FIG 6.

CPSF6 depletion eliminates the triphasic nature of the PF74 dose-response curve. (A) HeLa cells transfected with either nontargeting (Ctrl) or CPSF6-targeting (CPSF6 k/d [knock down]) siRNA were used for a single-cycle replication assay as described in the legend to Fig. 1. Mean values from three independent experiments are shown, with error bars denoting the standard errors of the means. (B) The effects of CPSF6 depletion on the antiviral activity of PF74 were examined to assess statistical significance by compiling data from four independent experiments. **, P < 0.01; *, P < 0.05. (C) Western blot analysis of HeLa cells transfected with nontargeting siRNA or siRNA targeting CPSF6. (D) The effects of CPSF6 deletion on the PF74 dose-response curve were examined using HEK293T cells carrying the empty expression vector (CTRL-EV), HEK293T CPSF6 knockout cells (B8-CKO-EV), and HEK293T CPSF6 knockout cells ectopically expressing CPSF6[551] (B8-CKO-551) infected with GFP-encoding reporter viruses. Mean values from three independent experiments are shown, with error bars denoting the standard errors of the means.

We next determined the effects of CA-CPSF6 interactions on the steepness of the slopes within the PF74 dose-response curves, which differed between low and high drug concentrations (Fig. 1A, C, and D). We first determined the magnitude of PF74-mediated inhibition by quantifying the decrease of viral infectivity at small ranges of drug concentrations. CPSF6 depletion significantly reduced the potency of PF74 at low concentrations (0.313 to 1.25 μM) (Fig. 7A) but not when the decrease in viral infectivity was examined at high concentrations of 5 to 10 μM (Fig. 7B). A similar pattern was observed when the slope (m) values were examined (Fig. 7C and D). The m values derived at lower concentrations (0.313 to 0.625 μM) were significantly lower in CPSF6-depleted cells than in control cells (Fig. 7C). In contrast, there was no significant change caused by CPSF6 depletion in the slope values at higher concentrations (5 to 10 μM) (Fig. 7D). The absence of decreased infectivity between 1.25 and 5 μM made it impossible to calculate m values (Fig. 6A). Thus, we examined the levels of reduction in viral infectivity from 1.25 to 5 μM PF74. As evident from a rather drastic change in the dose-response curve of PF74 upon CPSF6 depletion, which eliminated the plateau phase (Fig. 6A), the levels of decrease in viral infectivity at these PF74 concentrations (1.25 to 5 μM) were significantly greater upon CPSF6 depletion (Fig. 7E). Taken together, these findings reveal a key role of CPSF6 in contributing to the unique dose-response curve of PF74.

FIG 7.

Two opposing roles of CPSF6 in PF74's antiviral activity. (A, B, and E) The degree of inhibition by PF74 in control cells (Ctrl) was compared to that in CPSF6-depleted cells (CypA k/d) by plotting the decreases in viral infectivity between given ranges of drug concentrations. (C and D) The slope value (m) for each range of drug concentrations was calculated as described in Materials and Methods. Each dot represents the value for one experiment, and the bars denote the means of the groups. **, P < 0.01; *, P < 0.05.

A similar yet distinct role for cyclophilin A in determining the PF74 dose-response curve.

CypA shares properties with CPSF6: both factors bind CA directly, and CA mutant viruses that carry changes that alter host factor binding display altered integration site preferences (38, 52, 53). As CypA also modulates the antiviral activity of PF74 (16), we wished to understand its effect on the unique dose-response curve of PF74. CypA depletion by siRNA-mediated gene knockdown (Fig. 8A), which altered the dose-response curve (Fig. 8B), reduced the potency at lower drug concentrations (Fig. 8B) but acted oppositely at high drug concentrations by increasing the antiviral activity (Fig. 8B). Importantly, the change in the dose-response curve occurred even when CypA depletion barely affected viral infectivity in the absence of PF74 (data not shown). Assessment of the steepness of the slopes at different ranges of drug concentrations indicated that the slopes were significantly changed in all of the ranges we tested (Fig. 8C to E) except for the highest range of PF74 concentrations (Fig. 8F). Importantly, CypA differed from CPSF6 in the midrange of drug concentrations, where CypA depletion significantly increased the magnitude of inhibition by PF74 (Fig. 8D). This was also observed when slope values for the same drug range were examined: CypA-depleted cells displayed an m value of 3.83 ± 1.38, which was higher than the m value (1.3 ± 0.95) derived on control cells. Thus, it appeared that CypA depletion shifted the steep slope toward lower drug concentrations. Substantially similar observations were made when CypA-CA interactions were blocked with cyclosporine (CsA) and by G89V and P90A mutations within CA (Fig. 9A and B). Namely, both CsA treatment (Fig. 9A) and the G89V and P90A mutations (Fig. 9B) removed the triphasic dose-response curve and increased the magnitude of inhibition at medium concentrations. These changes in the curves were also observed when inocula of CA mutants were normalized by infectivity to that of the WT virus. Importantly, this finding was observed using cell clones in which the PPIA gene was knocked out by using CRISPR-Cas9 technology (Fig. 9C to E). CypA knockout eliminated the unique triphasic dose-response curve (Fig. 9D) and increased the potency of high PF74 concentrations, which was observed in an experimental setting using higher virus input (Fig. 9E). Importantly, reverse complementation of CypA expression restored the antiviral activity at lower concentrations (Fig. 9D) but reduced the antiviral effects at high concentrations (Fig. 9E), although the reverse complementation did not fully restore the antiviral profile seen in the original cell clone transduced with the empty expression vector. Similar observations were made with two other CypA knockout clones.

FIG 8.

CypA depletion alters the dose-response curve of PF74 in a unique manner. Control (nontargeting-siRNA-treated [Ctrl]) or CypA-depleted cells (CypA k/d) were used for a single-cycle infection assay using luciferase-encoding reporter virus. (A) Western blot detection of HeLa cells transfected with nontargeting siRNA or siRNA targeting CypA. (B) Infectivity values relative to the result with no drug are plotted to display PF74's antiviral activity. Mean values from four independent experiments are shown, with error bars denoting the standard errors of the means. (C to F) The magnitudes of PF74-mediated inhibition of HIV-1 in control cells were compared to the levels of inhibition in CypA- or CPSF6-depleted cells by plotting the decreases in viral infectivity between various ranges of drug concentrations. Each dot represents the value for one experiment, and the bars denote the means of the groups. **, P < 0.01; *, P < 0.05.

FIG 9.

The CypA-CA interaction protects HIV-1 from the antiviral activity of high PF74 concentrations. (A) The WT virus encoding luciferase was used to infect HeLa cells in the presence and absence of CsA. Mean values from three independent experiments are shown, with error bars denoting the standard errors of the means. (B) The sensitivity to PF74 was determined for different CA mutants (G89V and P90A) that fail to bind to CypA. Viral infectivity was determined as described in the legend to Fig. 1. Results are plotted as relative infectivity, normalized to infectivity without PF74. Mean values from four independent experiments are shown, with error bars denoting the standard errors of the means. (C) Western blot analysis of HeLa cell lines transduced with CRISPR vectors and control or CypA expression vectors. (D) PF74 dose-response curves were examined on the cell clone transfected with nontargeting CRISPR vector and transduced with the empty expression vector (EV), G1-C4 knockout cells transduced with EV, and G1-C4 cells transduced with the CypA expression vector. Data are the average results of two independent experiments, with error bars showing the standard errors of the means. (E) The effects of CypA on the antiviral activities of high PF74 doses (5 μM and 10 μM) were studied by using a high inoculum. Data are the average results of two independent experiments, with error bars showing the standard errors of the means. (F) The potency of 5 μM PF74 was compared between different conditions by compiling results from different experiments. Each dot represents the value for one experiment, and the bars denote the means of the groups. Values shown are infectivity (%) in the presence of 5 μM PF74 relative to no drug. **, P < 0.01. *, P < 0.05.

The shift of the steep slope toward lower drug concentrations caused by preventing CA-CypA interactions by genetic or pharmacological means (Fig. 8D and 9A and B) suggested that while CypA sensitizes the virus for PF74 antiviral activity at lower concentrations (Fig. 8 and 9) (16), it has a protective role against high PF74 concentrations. In fact, we observed greater magnitudes of reduction in viral infectivity at 5 μM (Fig. 9F) or higher concentrations (7.5 and 10 μM, data not shown) under conditions where CypA was prevented from interacting with CA than were seen in the respective control values. A similar trend was observed for CPSF6-depleted cells, although the difference between the results for CPSF6-depleted and control cells was modest compared to the effects of CypA (Fig. 9F). These results demonstrated that CypA-CA interactions, similar to CPSF6-CA interactions, are required for the triphasic dose-response curve for PF74 but that CypA uniquely possesses the ability to neutralize the antiviral activity of high PF74 doses.

NUP153 and NUP358 are essential for the full antiviral activity of PF74.

HIV-1 depends on nuclear pore complexes for optimal transport of preintegration complexes (PICs) into the nucleus by utilizing constituents of nuclear pore complexes, such as NUP358 and NUP153. NUP153 is particularly relevant to the molecular mechanisms of PF74-mediated HIV-1 inhibition because it binds to CA at the same site as PF74 (20, 21, 23). Furthermore, NUP153 competes with PF74 for binding to CA, and its depletion drastically changed the drug's 90% effective concentration (EC₉₀) value (20). Having shown that PF74 exhibited a triphasic dose-response curve with different slope values (Fig. 1), we wished to reassess the role of NUP153 in the unique antiviral profile of PF74. NUP153 depletion by siRNA (Fig. 10A), which caused a 2-fold decrease in viral infectivity in the absence of PF74, predictably altered the dose-response curve of PF74 (Fig. 10B) and removed the three distinct phases observed for the curve under the control condition. These changes were reproducibly observed regardless of the multiplicity of infection used (data not shown). The reduction of PF74 potency by NUP153 depletion was evident from the overall upward shift of the curve with a wide range of drug concentrations (Fig. 10B). However, NUP153 depletion affected the steepness only at lower drug concentrations (Fig. 10C) and not at intermediate (Fig. 10D) or higher (Fig. 10E) concentrations. NUP153 affected PF74 activity similarly to CPSF6 up to a concentration of 1.25 μM (Fig. 10B and C) but differently at higher concentrations (Fig. 10B and D). However, if we focused our attention on the steepness of the curve as measured by the extent of inhibition observed between 5 and 10 μM, the slope seemed to remain unchanged between CPSF6 and NUP153 depletion (Fig. 10E). Thus, it appears that NUP153 was distinct from CPSF6 in that its depletion maintained the plateau phase (Fig. 10B and D).

FIG 10.

NUP153 depletion reduces the potency of PF74 in a manner distinct from the manner of CPSF6 depletion. (A) Western blot analysis of HeLa cells transfected with nontargeting siRNA (Control) or siRNA targeting NUP153 (NUP153 k/d). (B) A single-cell infectivity assay using GFP-encoding virus was performed in the presence of various amounts of PF74 by using cells treated with nontargeting siRNA (Ctrl), siRNA targeting NUP153 (NUP153 k/d), or siRNA targeting CPSF6 (CPSF6 k/d). Mean values from three independent experiments are shown, with error bars denoting the standard errors of the means. (C to E) The magnitudes of PF74-mediated inhibition of HIV-1 were assessed for three different ranges of drug concentrations. **, P < 0.01; *, P < 0.05. Each dot represents the value for an independent infection sample, and the bars denote the means of the groups.

NUP358 can also participate in HIV-1 nuclear entry (53–56). Thus, we also examined the role of NUP358 in the dose-response of PF74. We found that NUP358 depletion recapitulated NUP153 depletion. The dose-response curve of PF74 on cells from a subclone (F1-5F) depleted of NUP358 (Fig. 11A and B) was very similar to the one obtained on cells depleted of NUP153 (Fig. 10B). An almost identical dose-response curve was observed when the other subclone (F1-8C) was used for infection (data not shown). NUP358 depletion reduced the potency of most of the drug concentrations we tested (Fig. 11B). However, when the steepness of the slopes was examined by measuring the extent of inhibition by PF74, the only differences between control and NUP358-depleted cells were observed at lower drug concentrations (Fig. 11C to E), a trend that was the same as the effects of NUP153 depletion on PF74 activity. NUP358 depletion caused a 10-fold decrease in the infectivity of the WT virus. Thus, we also increased virus input for infection of NUP358-depleted cells so that a comparable level of infectivity was achieved between control and NUP358-depleted cells in the absence of PF74 treatment. However, this did not significantly alter our observations, as the dose-response curve of PF74 in NUP358-deleted cells was significantly different from that in control cells. All these findings highlight the effects of complex host-virus interactions on PF74 activity but reveal that the steep slope at high drug concentrations was largely independent from the effects of host factors.

FIG 11.

NUP358 depletion neutralizes antiviral activity of PF74. (A) Western blot analysis of cell lysates prepared from control HeLa cells and two subclones (F1-5F and F1-8C) carrying shRNA targeting NUP358. Cell lysates were subjected to SDS-PAGE, transferred onto membrane, and probed with an anti-NUP358 antibody or anti-tubulin antibody as a loading control. (B) GFP-encoding reporter virus was titrated on increasing amounts of PF74 in a single-cycle replication assay using control HeLa cells or HeLa cells in which NUP358 was depleted by stable expression of shRNA targeting NUP358 (F1-5F). Results are plotted as relative infectivity, normalized to infectivity without PF74. Mean values from four independent experiments are shown, with error bars denoting the standard errors of the means. (C to E) The magnitudes of PF74-mediated inhibition of HIV-1 were assessed for three different ranges of drug concentrations. **, P < 0.01; *, P < 0.05. Each dot represents the value for an independent infection sample, and the bars denote the means of the groups.

PF74 retargets HIV-1 integration site distribution.

Our findings provide strong support for key roles of CA-binding host factors in the antiviral mechanism of PF74. Various CA-interacting host proteins influence integration site distribution (38, 41, 53, 57). PF74 targets CA at a site shared by CPSF6 and NUP153, and mutations within the site that block CA binding to these host molecules affect integration targeting (38, 53). Moreover, recent studies using CPSF6 knockout cells revealed a critical role for CPSF6 in targeting HIV-1's integration to transcriptionally active chromatin (41, 58). Thus, the effect of PF74 on the integration site preferences of HIV-1 was examined to provide independent evidence for the participation of these host factors in the antiviral activity of PF74. Integration site libraries were generated by infecting HEK293T cells with WT virus and extracting genomic DNA 5 days after infection. Integration sites were amplified by ligation-mediated (LM)-PCR and sequenced using the Illumina platform, and unique integration sites were mapped with respect to various genomic annotations (46, 47, 59). The pattern with respect to integration into genes or gene-dense regions or near transcription start sites (TSSs) or CpG islands in PF74-treated cells was compared to that of DMSO-treated control cells or 50,000 random sites generated in silico (matched random control [MRC]) (Table 1). In DMSO-treated cells, HIV-1 preferred to integrate into genes and gene-dense regions; however, little preference was observed for integration near CpG islands or TSSs (Table 1) (60). PF74 treatment significantly decreased HIV-1 integration within each of the genomic features (Table 1; Fig. 12). PF74 decreased integration into genes by ∼8% and near promoters by ∼1.3% (Table 1; Fig. 12). Furthermore, integration into gene-dense regions was prominently affected (Table 1; Fig. 12). The presence of PF74 lowered integration into gene-dense regions to an average of 12.2 genes Mb⁻¹, a value that was significantly lower than the 21.5 genes Mb⁻¹ observed in the control (Table 1; Fig. 12).

TABLE 1.

Effect of PF74 on HIV-1 integration sites in 293T cells

Genomic feature	Valuea (%) per genomic feature
	DMSO	PF74	MRCc
Unique integration sites	13,037	2,678	50,000
RefSeq genesb	10,514 (80.6)	1,951 (72.9)	22,328 (44.7)
CpG islands (±2.5 kb)	750 (5.8)	85 (3.2)	2100 (4.2)
TSSs (±2.5 kb)	543 (4.2)	77 (2.9)	2010 (4.0)
Gene-dense regions	21.5	12.2	8.7

^aValues are the number of sites per genomic feature except those for gene-dense regions, which are the number of genes per Mb. Percentages are relative to the number of unique integration sites.

^bAnnotations for RefSeq genes, CpG islands, and TSSs were obtained from human genome build hg19.

^cMRC, matched random control; contains coordinates for 50,000 computer-generated integration sites within the vicinity of MseI and BglII restriction sites in hg19.

FIG 12.

Statistical analyses of HIV-1 integration site distributions. P values are shown for comparison of HIV-1 integration site distribution in control cells to that in cells treated with 1.39 μM PF74, as well as the matched random control (MRC). Table 1 contains information on the numbers of integration sites within RefSeq genes and nearby CpG islands and TSSs, as well as average gene-density profiles at the site of integration. ^a, P values calculated by Fisher's exact test; ^b, P values calculated by Wilcoxon rank sum test.

DISCUSSION

PF74 has generated broad interest not only because it potentially represents a new scaffold from which to develop clinically relevant inhibitors but also because the drug can be utilized as a probe to understand poorly elucidated steps mediated by the viral CA protein, such as uncoating, nuclear entry, and integration site selection (6, 8). A detailed understanding of the antiviral mechanisms of PF74 may provide insights into roles of CA-binding host factors during HIV-1 replication, as the PF74 binding site within CA is also bound by CPSF6 and NUP153. This study provides evidence that CA engagement of a number of cellular cofactors is essential for the unique triphasic dose-response curve, lending support for the model in which CA-binding host factors regulate PF74 antiviral activity both positively and negatively.

Cooperative inhibition of HIV-1 by high-dose PF74.

One distinctive feature of PF74, which was also observed in previous work (17, 21), was its triphasic dose-response curve, consisting of two inhibitory phases flanking a plateau phase (Fig. 1A). We found that the slope values of these two inhibitory phases differed from each other (Fig. 1C and D), with the slope values of high PF74 concentrations (m of ∼4.5) being the highest among different classes of antiretrovirals (48). It has been argued that high slope values largely account for the favorable potency of protease inhibitors in clinical settings (48). PF74, whose IC₉₀ is in the lower micromolar range, is not potent enough for direct clinical application; however, its unique mechanism of action along with the steep slope may provide a renewed hope to utilize PF74 as a starting material for further drug development. A directly relevant property of PF74 is that it targets a highly conserved portion of CA (21, 23, 28, 29, 40), which by itself is a surprisingly fragile viral protein (4), and hence, studies of antiviral drugs targeting this particular pocket of CA may temper the development of highly fit drug-resistant variants (5, 25). In fact, the fitness of CA mutants that are partially resistant to PF74 is compromised, especially in primary cells (25).

Slope parameters, equivalent to the Hill coefficient, are used as a measure of cooperativity. Slope values greater than 1, suggestive of a cooperative mechanism in drug action, can be explained by intra- or intermolecular cooperativity, as well as multistep inhibition (48, 50). PF74 binding to individual CA molecules is monovalent (13), but its binding to CA hexamers or pentamers could proceed as polyvalent interactions. The binding profile of PF74 to CA hexamers does not appear cooperative (25), and thus, the cooperative mechanism of action of PF74 may be caused by a step that follows its binding to CA. CA-CA interactions within or between hexamers are tightly regulated to form the conical capsid of HIV-1 with stability that facilitates virion morphogenesis and the uncoating process (9, 61). Tipping the balance of these CA interactions alters the stability of HIV-1 cores and drastically impairs viral infectivity (62). The binding of PF74 to CA at the interfaces between neighboring CA subunits within a CA hexamer (21, 23, 29) causes changes at both intra- and interhexamer interfaces (29) and may affect the stability of CA cores. This could be the underlying molecular basis for the ability of high PF74 concentrations to destabilize the viral capsid (13, 16, 23). These observations, together with our novel findings, suggest a model in which the destabilizing activity of PF74 is disproportionally enhanced as an increasing number of PF74 molecules interact with the viral capsid, which is the molecular basis for positive cooperativity that leads to the steep slopes observed with high PF74 concentrations. Supporting this model is the steep slope in the dose-response curve for viral DNA copy numbers (Fig. 2C), which largely accounts for the high slope parameter for inhibition of viral infectivity.

An alternative model for the steep slope involves multistep inhibition. Previous work demonstrated that the inhibition at multiple steps by protease inhibitors (50) contributes to a cooperative mechanism of action. As described above, high PF74 concentrations target not only reverse transcription but also subsequent steps, likely including nuclear entry (Fig. 2). This was supported by the results of time-of-addition experiments, in which virus remained sensitive to PF74 for a longer period of time than to nevirapine (Fig. 3). Therefore, it is also possible that inhibition of multiple steps by PF74 contributes to the cooperative mechanism of action by high PF74 concentrations.

We also found that the steep slope is largely independent from the CA interactions of host cellular factors, perhaps with the exception of CypA. Namely, depletion of every protein tested did reduce the antiviral activity of PF74 but did not change the slope values at high concentrations (Fig. 7B and D, 8F, 10E, and 11E). One potential exception was the appearance of a rather high slope value (m of ∼3.8) for the dose-response curve on CypA-depleted cells between PF74 concentrations of 0.625 μM and 2.5 μM (Fig. 8D). Similar observations were made with CsA (Fig. 9A) and CA G89V and P90A mutants (Fig. 9B). Consistent with this notion, the extent of inhibition by 5 μM PF74 was significantly enhanced by blocking of interactions between CA and CypA or CPSF6 but not with NUP153 and NUP358 (Fig. 9F, 10B, and 11B). These observations suggest that CypA has a protective role against the antiviral function of high PF74 doses. The roles of CypA during HIV-1 replication have been elusive (63), but CypA has been implicated in modulating the uncoating step and has both stabilizing and destabilizing activity depending on the cell types or assay systems used (17, 64–66). As previous data indicate that high PF74 concentrations destabilize the viral core (16, 18, 22, 23) and CypA seems to counteract this function, one mechanism of the CypA function, which is consistent with some previous reports (17, 66), may be to stabilize the incoming viral capsid. Importantly, depletion of CPSF6 and CypA had similar yet distinct effects on the antiviral potency of PF74 (Fig. 6 to 9), suggesting a functional difference between these CA-binding molecules during the early steps of HIV-1 infection.

Antiviral mechanisms at low concentrations of PF74.

Lower concentrations of PF74 (∼1.25 μM) do not inhibit HIV-1 infection by blocking reverse transcription (Fig. 2) (21, 24, 26). As described above, this observation supports the idea that the antiviral functions of PF74 include a block that occurs after the completion of reverse transcription (21, 24, 26), consistent with the data from our time-of-addition experiment (Fig. 3). We found that PF74 treatment significantly decreased integration targeting into transcriptional units, gene-dense regions, CpG islands, and TSSs (Fig. 12; Table 1). However, the magnitude of the reduction by PF74 was not nearly that of the N74D mutant virus (38, 53), whose infectivity is similar to that of the WT virus in these cell types (32, 67). Thus, the contribution of the altered integration site distribution to drug potency is likely minor. These observations, together with data showing that the formation of 2-LTR circles was reduced in the presence of intermediate concentrations of PF74 (22, 24), suggest nuclear entry as a step blocked by low PF74 concentrations. Our data therefore are in line with models that propose the coupling of integration to PIC nuclear import (33, 46, 68).

Our results revealed that the full activity of PF74 at low concentrations depends on CPSF6 (Fig. 6). This finding does not support a model in which competitive inhibition of CA binding to CPSF6 by PF74 is responsible for the PF74-mediated inhibition of HIV-1 (24). If this were the case, the PF74 potency would be enhanced. However, this was not what we observed (Fig. 6). Our observation was not completely unexpected, because preventing CA-CPSF6 interactions, such as by CA mutations or gene depletion, does not overtly affect infectivity by HIV-1 (32, 40, 41). Thus, while CA-CPSF6 interactions may be prevented by PF74, as shown by biochemical and imaging methods (20, 22, 24, 28), preventing CA-CPSF6 interactions is not likely a mechanism of the PF74-mediated inhibition of HIV-1.

How does PF74 block HIV-1 infection at low concentrations? CPSF6-CA interactions, though insufficient, seem to correlate with the utilization of known cofactors involved in HIV-1 nuclear entry, such as NUP358 and NUP153 (20, 28, 32, 53). In particular, NUP153 may hold a key role in the antiviral activity of PF74 because NUP153-CA interactions can be competitively blocked by PF74 and depletion of NUP153 neutralizes the antiviral activity of PF74 at low concentrations (Fig. 10) (20). Interestingly, PF74-resistant variants became less dependent on NUP153 utilization (5). In this scenario, PF74 blocks the nuclear translocation of PICs by interfering with viral utilization of NUP153. Similar to mutations that cause CPSF6-binding deficiency (e.g., N74D), CPSF6 depletion may alter the nuclear entry pathway such that PICs become independent of NUP153. This scenario is consistent with the observation that depletion of CypA also rendered PF74 less potent at lower concentrations (Fig. 8) (16), as CypA has also been implicated for viral utilization of NUP153 (69). NUP358 phenocopied NUP153 with regard to their effects on PF74 activity upon depletion (Fig. 10 and 11). The utilization of these two nucleoporins by HIV-1 is correlated (32), and thus, the engagement of PICs with NUP358 may be a prerequisite for NUP153 utilization.

A model for two opposing effects of PF74 on the viral capsid.

Our findings, together with those from previous reports (8, 16, 18–24), allow us to propose a model in which PF74 exerts two opposing effects on the viral capsid. In this model, low concentrations of PF74 bind and stabilize the capsid and sterically inhibit the engagement of NUP153 or other NUPs that contain FG repeats capable of interacting with the viral capsid. CPSF6 and CypA may both contribute to the antiviral activity in this phase by additionally stabilizing the capsid, thus cooperating with PF74 to delay uncoating and prevent nuclear entry. This model can account for the effects of depletion of these host factors on antiviral activity at low PF74 concentrations, since the virus may utilize alternative nuclear entry pathways (32). Further experiments will be needed to test whether low PF74 doses, as well as CPSF6, retain stabilizing effects on the capsid. In contrast, high concentrations of PF74 result in high-occupancy binding to the capsid, thus irreversibly disintegrating the capsid structure and resulting in abortive reverse transcription (16, 23). This would occur prior to the capsid's engagement with NUPs. In this way, the two opposing effects of PF74 (capsid stabilization at low concentrations and irreversible capsid destruction at high concentrations) result in the inflection in the dose-response curve at intermediate PF74 concentrations, accounting for the presence of the plateau phase.

CA association with functional PICs.

Time-of-addition experiments showed that intracellular viral complexes were sensitive to PF74 for a longer period of time than to nevirapine (Fig. 3). This finding conflicts with an earlier observation that PF-3450071, a derivative of PF74, lost its activity 2 or 3 h after infection (13) but is largely consistent with recent work (21, 26). The fate of CA molecules associated with intracellular viral complexes and mechanisms of uncoating has been a subject of intense debate (6–8, 34, 70). In this regard, a more broadly relevant implication is that this observation supports the claim that a subset of CA remains associated with functional PICs following the completion of reverse transcription (26).

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.