Capsid-CPSF6 Interaction Is Dispensable for HIV-1 Replication in Primary Cells but Is Selected during Virus Passage In Vivo

ABSTRACT

Cleavage and polyadenylation specificity factor subunit 6 (CPSF6), a host factor that interacts with the HIV-1 capsid (CA) protein, is implicated in diverse functions during the early part of the HIV-1 life cycle, including uncoating, nuclear entry, and integration targeting. Preservation of CA binding to CPSF6 in vivo suggests that this interaction is fine-tuned for efficient HIV-1 replication in physiologically relevant settings. Nevertheless, this possibility has not been formally examined. To assess the requirement for optimal CPSF6-CA binding during infection of primary cells and in vivo, we utilized a novel CA mutation, A77V, that significantly reduced CA binding to CPSF6. The A77V mutation rendered HIV-1 largely independent from TNPO3, NUP358, and NUP153 for infection and altered the integration site preference of HIV-1 without any discernible effects during the late steps of the virus life cycle. Surprisingly, the A77V mutant virus maintained the ability to replicate in monocyte-derived macrophages, primary CD4⁺ T cells, and humanized mice at a level comparable to that for the wild-type (WT) virus. Nonetheless, revertant viruses that restored the WT CA sequence and hence CA binding to CPSF6 emerged in three out of four A77V-infected animals. These results suggest that the optimal interaction of CA with CPSF6, though not absolutely essential for HIV-1 replication in physiologically relevant settings, confers a significant fitness advantage to the virus and thus is strictly conserved among naturally circulating HIV-1 strains.

IMPORTANCE CPSF6 interacts with the HIV-1 capsid (CA) protein and has been implicated in nuclear entry and integration targeting. Preservation of CPSF6-CA binding across various HIV-1 strains suggested that the optimal interaction between CA and CPSF6 is critical during HIV-1 replication in vivo. Here, we identified a novel HIV-1 capsid mutant that reduces binding to CPSF6, is largely independent from the known cofactors for nuclear entry, and alters integration site preference. Despite these changes, virus carrying this mutation replicated in humanized mice at levels indistinguishable from those of the wild-type virus. However, in the majority of the animals, the mutant virus reverted back to the wild-type sequence, hence restoring the wild-type level of CA-CPSF6 interactions. These results suggest that optimal binding of CA to CPSF6 is not absolutely essential for HIV-1 replication in vivo but provides a fitness advantage that leads to the widespread usage of CPSF6 by HIV-1 in vivo.

INTRODUCTION

The HIV-1 capsid (CA) protein forms the protein core that envelopes viral enzymes and genetic material following fusion with the plasma membrane (1–3). Uncoating of the viral core is a fine-tuned process that regulates downstream events, including reverse transcription, nuclear entry, and integration targeting (4–9). It is likely that HIV-1 exploits multiple host factors that can directly interact with the viral CA protein for proper uncoating in virus-infected cells. These host factors likely include cyclophilin A (CypA) and cleavage and polyadenylation specificity factor subunit 6 (CPSF6), which can bind to CA and affect multiple steps in the early phase of the HIV-1 life cycle (10–16).

One suggested role for CypA and CPSF6 is to regulate HIV-1 nuclear entry (11–13, 17). Other host proteins that physically interact with CA, such as the beta-karyopherin transportin 3 (TNPO3) (also known as TRN-SR2) and nucleoporins (NUPs) NUP358, NUP153, and NUP98 (11, 12, 17–31), also seem to aid HIV-1 replication primarily by acting at the nuclear entry step. The N74D mutation, which abolishes CA binding to CPSF6, relieves the dependence of HIV-1 on TNPO3, NUP358, and NUP153 (11–13, 20–23, 25–27, 30, 31). Roles of CypA and CPSF6 in shaping the unique pattern of HIV-1 integration targeting (32) have been suggested based on the observation that mutations that block CA interactions with CPSF6 or CypA significantly alter the target site selection (12, 33). Moreover, recent studies using CPSF6 knockout cells have provided unambiguous evidence for a critical role for CPSF6 in targeting HIV-1 integration to transcriptionally active chromatin (34, 35). Finally, Rasaiyaah et al. provided data supporting a model in which recruitment of CPSF6 and CypA enables HIV-1 to evade detection by the cytosolic DNA sensor cyclic GMP-AMP (cGAS) in macrophages (14). CypA may play a similar yet distinct role in dendritic cells to fine-tune the uncoating process (36). These observations highlight the diverse functions played by CPSF6, together with CypA, during the early steps of the HIV-1 replication cycle.

The binding sites on HIV-1 CA differ between CypA and CPSF6. HIV-1 utilizes a flexible linker between alpha helices 4 and 5, called the CypA-binding loop, to interact with CypA (10, 37, 38), whereas a hydrophobic pocket formed by helices 3, 4, and 5 within the N-terminal domain of CA contains key amino acids that mediate direct binding to CPSF6 (13). Additionally, the C-terminal domain of CA also provides binding determinants for CPSF6 (39, 40). Amino acid sequences of the CypA-binding loop vary across different primate immunodeficiency viruses (41–43) as well as between various strains of HIV-1 (44). This variation may be an adaptive response to evade recognition by host restriction factors, such as TRIM5alpha (45–48) and Mx2 (49). In contrast, amino acids that constitute the CPSF6-binding pocket are highly conserved among different lineages of primate lentiviruses (13, 40). Consistently, HIV-1 CA mutations that block CA interactions with CPSF6 significantly impair viral replicative capacity in primary cells (12, 14, 50, 51). Furthermore, our previous study revealed strong selective pressure to maintain CA-CPSF6 interactions during viral replication in vivo (52).

These observations suggest a strict requirement for CA-CPSF6 interactions during HIV-1 replication. However, it remains unclear to what extent HIV-1 can tolerate reduced binding of CA to CPSF6 and accompanying alterations in the nuclear entry pathway and integration targeting during in vivo viral replication. This question is significant as there is growing interest in antiviral compounds that target a highly conserved pocket in HIV-1 CA bound by CPSF6 (39, 40, 53–56). In the present study, we utilized a unique capsid mutant virus that markedly reduced CPSF6 binding in a humanized mouse model to elucidate the in vivo requirement for optimal CA-CPSF6 interactions during viral replication. Our data show that a significant decrease in CA-CPSF6 interactions can be accommodated both in primary cells and in humanized mice without overtly affecting HIV-1 replication. However, in three out of the four infected animals, reversion events that restored the utilization of CPSF6 occurred. In addition, the wild-type (WT) virus almost completely dominated in in vivo competition experiments with the mutant virus. These results argue that although optimal CA-CPSF6 interactions are not absolutely required for HIV-1 replication in macrophages, primary cells, and humanized mice, they nevertheless offer significant advantages to the replicative capacity of HIV-1. We would conclude that this in turn underlies the sequence conservation of the CPSF6-binding pocket in CA across pandemic HIV-1 strains.

MATERIALS AND METHODS

Plasmid DNAs.

Molecular infectious clones based on the LAI strain of HIV-1 (pBru3ori), carrying either the intact or defective env gene, were used in the present study. We also used a previously described HIV-1 Gag-Pol expression vector, pCRV1-Gag-Pol (LAI) (52). Various CA mutations were introduced into these clones using standard cloning procedures as described previously (52). Most of the in vitro experiments with molecular clones based on the plasmid pBru3ori were performed by using clones that carry the green fluorescent protein (GFP) gene in place of the nef gene (57), except for in vitro competition experiments in which virus was generated by using a full-length molecule clone carrying E2-Crimson in place of the nef gene. Plasmid DNA encoding the vesicular stomatitis virus G (VSV-G) glycoprotein (pHCMV-G) was described previously (58).

Cell culture.

Adherent cell lines (HeLa, HOS, 293T, and TZM-bl) were cultured in Dulbecco's modified Eagle's medium (Cellgro) supplemented with 10% fetal bovine serum (FBS) (Cellgro) and 1× penicillin-streptomycin (P/S) (Cellgro). Immortalized suspension cells (MT4, MT4-CCR5, and C8166-R5) were cultured in RPMI (Cellgro) supplemented with 10% FBS (Sigma), 1× P/S, and 2 mM 2-glutamine (Cellgro). HeLa cells overexpressing CPSF6-358 (a gift from V. KewalRamani), MT4-CCR5, and C8166-CCR5 (a gift from J. Robinson) were previously described (11, 59). Expression of CPSF6-358 was confirmed by Western blotting with antibody against hemagglutinin (HA) (HA.11; Covance) and fluorescent secondary antibody (IRDye 800CW; Li-Cor Biosciences). Peripheral blood mononuclear cells (PBMC) were isolated from whole blood obtained from anonymous blood donors (New York Blood Center) using a standard Ficoll (Cellgro) procedure. Primary CD4⁺ T cells isolated using the human CD4⁺ T cell enrichment kit per the manufacturer's instruction (Stemcell) were activated using Dynabeads Human T-Activator CD3/CD28 (Life Technologies) and cultured in suspension cell medium supplemented with 30 units per ml of interleukin-2 (IL-2) (PeproTech). Monocyte-derived macrophages (MDM) were prepared from PBMC as follows. PBMC were allowed to attach to a 10-cm dish for 2 h and then vigorously washed with phosphate-buffered saline (PBS). Attached cells were subjected to differentiation in RPMI supplemented with 10% FBS, 1× P/S, 2 mM 2-glutamine, and 100 ng per ml of granulocyte-macrophage colony-stimulating factor (GM-CSF) (PeproTech) for 4 days. MDM were also derived by using 40 ng per ml of M-CSF (PeproTech) with the same conditions except with 10% human serum (Sigma) instead of 10% FBS.

siRNA knockdowns.

Transient small interfering RNA (siRNA) knockdowns for TNPO3 and NUP153 were done as described previously (52) and were shown to reproducibly deplete these proteins (52). Briefly, HeLa cells were plated at 5 × 10⁵ cells per well of a 6-well plate and transfected with 30 pmol of siRNA using Lipofectamine RNAiMAX (Invitrogen) or with the transfection reagent alone (control). Specific siRNAs were anti-TNPO3 (L-019949-01-0010; Thermo Scientific) and anti-NUP153 (GGACUUGUUAGAUCUAGUUUU). The following day, transfection was repeated with the same procedure. At 4 h posttransfection, cells were seeded at 5 × 10⁵ cells per 24-well or 96-well plate for infection. NUP358 was stably depleted from HeLa cells by using a puromycin-encoding lentiviral vector (pLKO.1; Addgene), which encodes the identical short hairpin RNA (shRNA) as a similar lentiviral vector used in our previous work (52). HeLa cell subclones (F1-5F and F1-8C) generated by two rounds of single-cell cloning, which were used in this study, were shown to exhibit a decreased level of NUP358 by a Western blot assay (56). In addition, silencing of each gene was validated by using quantitative reverse transcription-PCR (qRT-PCR) as described previously (51). Briefly, mRNA extracted from cells using the RNeasy minikit (Qiagen) at 3 days after siRNA transfection was used to synthesize cDNA using high-capacity cDNA reverse transcription kits (Life Technologies) and to quantify expression of each gene with a SYBR green-based quantitative PCR assay. Expression of the GAPDH (glyceraldehyde-3-phosphate dehydrogenase) gene was quantified in parallel as housekeeping gene control to determine the changes of mRNA levels using the 2^−ΔΔCT method (60).

Infection.

All viruses were generated by transfecting 293T cells using polyethylenimine (PEI) (PolySciences). Viruses generated by constructs that lack the intact env gene (such as the full-length molecule clone pBru3oriGFP3ΔEnv and the Gag-Pol expression vector pCRV1-GagPol) were pseudotyped with the VSV-G envelope glycoprotein. The Gag-Pol expression vector was also cotransfected with a packageable GFP-encoding HIV-1 vector (pHRsin-CSGW). All infections using HeLa cells were performed at 5 × 10⁵ cells per plate. Immortalized T cell lines and primary CD4⁺ T cells were plated at 2.5 × 10⁵ cells per ml and 1 × 10⁶ cells per ml, respectively. MDM were plated at 1 × 10⁶ cells per ml. Virus infectivity was examined by measuring GFP-positive or Crimson-positive cells using an LSRII flow cytometer (BD Biosciences) or Guava easyCyte (Millipore).

CA-binding assay.

Protein binding to HIV-1 CA tubes was performed as described previously (52). Briefly, Escherichia coli-derived CA assemblies (5 μM) were incubated with HeLa cell extracts for 1 h with gentle mixing in 100-μl reaction mixtures in binding buffer (50 mM Tris-HCl [pH 8], 150 mM NaCl, 5 mM MgCl₂, and 0.5 mM EDTA). Reaction mixtures were then centrifuged for 5 min at 5,000 × g, and the pellets were washed with 100 μl of binding buffer before pelleting again. The pellets were fractionated by SDS-PAGE and immunoblotted with antibodies against CA (183-H12-5C) and CPSF6 (Novus Biologicals). Signals originating from fluorescent secondary antibodies were quantitated with a Li-Cor Odyssey scanner (Li-Cor Biosciences).

ITC.

The CA N-terminal domain (NTD) (residues 1 to 151) was prepared as previously described (61). The CA-binding peptide of CPSF6 (residues 313 to 327, PVLFPGQPFGQPPLG) was synthesized by GeneScript. Aliquots of CPSF6 peptide (6.2 or 6.7 mM) were added to the CA NTD (600 or 800 μM) in a buffer containing 50 mM sodium phosphate (pH 7.4) and 150 mM NaCl at 25°C. Isothermal titration calorimetry (ITC) experiments were carried out by using a MicroCal PEAQ-ITC (Malvern Instruments Ltd.).

Integration site analysis.

Genomic DNA extracted from MDM or HOS cells infected with GFP-encoding viruses were used for integration site analysis. Briefly, MDM were prepared from CD14⁺ cells purified with the EasySep human monocyte enrichment kit (Stemcells). Cells were allowed to differentiate in RPMI supplemented with 10% FBS, 1× P/S, 2 mM 2-glutamine, and 100 ng per ml of GM-CSF (PeproTech) for 4 days. VSV-G-pseudotyped viruses generated from three plasmid DNAs (packageable GFP-encoding HIV-1 vector, HIV-1 Gag-Pol expression vector, and VSV-G-encoding vector) were used to infect both MDM and HOS cells. Replication-competent R5-tropic viruses that carried part of the env gene from the macrophage-tropic Bal strain were also used to infect MDM. All virus stocks were treated with Turbo DNase (Life Technologies) at 37°C for 60 min prior to infection. Cells infected with VSV-G-pseudotyped viruses and with replication-competent viruses were harvested for DNA extraction at 5 days and 12 days after infection, respectively. Genomic DNA was extracted from infected cells with the DNeasy blood and tissue kit (Qiagen) according to the manufacturer's instructions.

HIV-1 integration sites were determined by Illumina sequencing of DNA libraries generated from ligation-mediated PCR (LM-PCR) amplification of digested genomic DNA essentially as previously described (62, 63). In brief, DNA was digested overnight with 100 U each of MseI and BglII and purified using the QIAquick PCR purification kit (Qiagen). Unique double-stranded linkers were prepared for each DNA sample (see references 62 and 63 for an example of linker oligonucleotide design). Linker DNA was ligated to digested cellular DNA (1 μg) overnight at 16°C in four parallel reactions, and ligation reaction products were purified using the QIAquick PCR purification kit. Seminested PCR employed linker DNA and HIV-1 U5-specific primers multiplexed into eight separate samples per PCR stage. Linker primers, though unique for each DNA sample, were the same for both PCR rounds, while first- and second-round U5 primers were nested. Linker primers and second-round U5 primers encoded Illumina adapter sequences, 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. PCR mixtures in each round were incubated at 94°C for 2 min, followed by 30 cycles of 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. Pooled PCR products were purified using the QIAquick PCR purification kit and, after the second round, were sequenced 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 previously described (62, 63).

Time-of-addition experiment.

Primary CD4⁺ T cells, which were activated by Dynabeads Human T-Activator CD3/CD28, were infected with replication-competent viruses by spinoculation at 1,200 × g for 30 min. Infected cells were washed twice and resuspended with RPMI supplemented with 10% FBS, 1× P/S, 2 mM 2-glutamine, and 30 units per ml of IL-2. Raltegravir (1 μM at the final concentration) was added to the culture at different time points. Viral infectivity was measure by counting GFP-positive cells at 48 h after infection.

Quantification of virion production.

Reverse transcriptase (RT) activity in the supernatants of 293T cells transfected with pBru3-based constructs carrying the luciferase gene in place of the nef gene in the absence of the VSV-G envelope to prevent infection of cells with newly produced virions was determined using a SYBR green-based quantitative PCR assay (64). Transfection efficiency was assessed by measuring the luciferase activity of cells lysed at 48 h after transfection with the Luciferase Cell Culture Lysis 5× reagent and luciferase assay system (Promega).

For comparison of virion production between WT and A77V viruses by Western blotting, 293T cells were seeded in 12-well plates at a concentration of 3.3 × 10⁵ cells per well and were transfected the following day using PEI with different amounts of WT or A77V proviral plasmids. At 48 h posttransfection, the culture supernatants were filtered through a 0.45-μm polyvinylidene fluoride membrane. The physical particle yield was determined by layering 900 μl of the virion-containing supernatant onto 500 μl of 20% sucrose in PBS, followed by centrifugation at 20,000 × g for 2 h at 4°C. Pelleted virions and cell lysates were resuspended in 1× lithium dodecyl sulfate buffer containing 0.1% β-mercaptoethanol. Proteins were resolved on NuPAGE Novex 4 to 12% Bis-Tris minigels (Invitrogen) and transferred onto nitrocellulose membranes. The blots were then blocked with 3% FBS in PBS and probed with mouse anti-p24 HIV-1 capsid (183-H12-5C) antibody. The bound primary antibodies were detected using IRDye 800CW goat anti-mouse secondary antibody. Fluorescent signals were quantitated with a Li-Cor Odyssey scanner (Li-Cor Biosciences).

IFN bioassay.

Interferon (IFN) secretion from MDM infected with replication-competent virus carrying the Bal env gene was examined by using a previously described reporter cell line (HEK293-ISRE-Luc) (65) which expresses luciferase under the control of an IFN-inducible promoter carrying the IFN-stimulated response element (ISRE). For this assay, MDM plated on 24-well plates at 2 × 10⁵ cells per well were infected in triplicate with GFP-encoding Bal-Env virus carrying either the WT, N74D, or A77V CA at a multiplicity of infection (MOI) of 0.2 or 1. The MOI was determined by using the WT virus titer on MT4-CCR5 cells and an equal amount of virus input as normalized by RT activity for the mutants. In one experimental setting (two donors), the amount that is three times more than the WT virus was used for the N74D virus. At 1 day postinfection, the medium was replaced with fresh medium, and culture supernatant was harvested at 3 and 6 days postinfection. HEK293-ISRE-Luc cells were plated on 96-well plates at 3 × 10⁵ cells per well 1 day before adding 50 μl of culture supernatants. Cells were lysed, and luciferase activity was measured as described above. Serial dilutions of recombinant interferon alpha A/D (Sigma) were added to the cells in each experiment. The limit of detection of this assay was 1 unit per ml. Viral replication was examined by monitoring RT activity in the culture supernatant as described above.

Ethics statement.

All animal experiments were carried out in strict accordance with the Policy on Humane Care and Use of Laboratory Animals of the U.S. Public Health Service. The protocol was approved by the Institutional Animal Care and Use Committee (IACUC) at The Rockefeller University (Assurance no. A3081-01). CO₂ was used for euthanasia, and all efforts were made to minimize suffering. Human fetal liver samples were obtained via a nonprofit partner (Advanced Bioscience Resources, Alameda, CA) without any information that would identify the subjects from whom they were derived and did not require Institutional Review Board (IRB) approval for their use, as previously described (66).

Generation of humanized mice.

NOD.Cg-Prkdc^scid IL2rg^tmWjl/Sz (NSG) mice were purchased from the Jackson Laboratories and maintained under specific-pathogen-free conditions in the animal facilities of The Rockefeller University. To generate human immune system mice, NSG mice were first transduced with human genes encoding HLA-DR1 and human cytokines, including GM-CSF, M-CSF, IL-3, IL-4, IL-7, and IL-15, by using adeno-associated virus serotype 9 (AAV9) vectors. Certain human cytokines, such as GM-CSF, IL-3, and IL-15, have been shown to be important for the differentiation and development of human lymphocytes from human hematopoietic stem cells (HSC). In fact, these human cytokines are shown to facilitate the reconstitution of matured human lymphocyte lineage (67, 68). The details of generation of AAV vectors were described previously (66). Two weeks later, mice were irradiated with 150 rads. The next day, mice were given human CD34⁺ hematopoietic cells from an HLA-DR1-positive donor. AAV9-DR1/cytokine-transduced NSG mice were monitored at 6, 10, and 14 weeks after HSC engraftment by determining the percentage of human CD45⁺ cells in the peripheral blood using flow cytometric analyses as described previously (66, 69, 70).

Infection of humanized mice.

After confirming reconstitution of human cells, mice were infected with 2 × 10⁵ infectious units (IU) of CCR5-tropic WT or A77V mutant viruses or both together in coinfection experiments via caudal vein. The IU of the viruses were titrated on an indicator cell line based on MT4, which has an integrated long terminal repeat (LTR)-GFP and stably expresses CCR5 (T. M. Zang and P. D. Bieniasz, unpublished data).

Monitoring human CD4⁺ T cells in the peripheral blood.

Peripheral human CD4⁺ T cells were monitored as described previously (66). Briefly, PBMC were purified from peripheral blood collected from the mice. After blocking with normal mouse serum supplemented with anti-CD16/CD32 (clone 93; BioLegend), cells were stained for 30 min at room temperature in the dark with the following antibodies: Pacific Blue anti-human CD45 (clone HI30; BioLegend), peridinin chlorophyll protein (PerCP)/Cy5.5–anti-mouse CD45 (clone 30-F11; BioLegend), phycoerythrin (PE)/Cy7–anti-human CD3 (clone UCHT1; BioLegend), allophycocyanin (APC)/Cy7-anti-human CD4 (clone RPA-T4; BioLegend), Alexa Fluor 700–anti-human CD8 (clone HIT8a; BioLegend), Alexa Fluor 647–anti-human CD161 (clone HP-3G10; BioLegend), PE–anti-human CD19 (clone HIB19; BioLegend), and APC–anti-human CD3 (clone HIT3a; BioLegend). After staining, the cells were washed twice with PBS containing 2% FBS, fixed with 1% paraformaldehyde, and analyzed using an LSR II instrument (BD Biosciences).

Plasma viral loads.

The quantification of plasma viral loads was done as described previously (71) with slight modification. Briefly, total RNA was collected from mouse plasma using a High Pure viral RNA kit (Roche Diagnostics). Copy numbers of viral RNA were quantified with a quantitative real-time PCR system using the TaqMan RNA-to-Ct 1-Step kit (Life Technologies). The primers and probe used in this study were as follows: forward primer HIVgag683 (+) (5′-CTCTCGACGCAGGACTCGGCTTGCT-3′), reverse primer: HIVgag803 (−) (5′-GCTCTCGCACCCATCTCTCTCCTTCTAGCC-3′), and probe HIVgag TaqMan 720R748 (6-carboxyfluorescein [FAM]-GCAAGAGGCGAGGGGAGGCGACTGGTGAG-6-carboxytetramethylrhodamine [TAMRA]). Quantification and data analysis were performed using the 7500 Fast real-time PCR system (Applied Biosystems). Results were normalized to the volume of plasma extracted and expressed as HIV-1 RNA copies per ml of plasma. The detection limit of this assay was ∼1,000 copies per ml of plasma.

Sequence analysis of CA in plasma viral RNA.

Plasma viral RNA obtained from infected mice was reverse transcribed into cDNA using high-capacity cDNA reverse transcription kits (Life Technologies) in the presence of RiboLock RNase inhibitor (Life Technologies). A CA fragment containing amino acid position 77 was amplified from cDNA with the Phusion high-fidelity PCR kit (NEB) using forward primer 5′-GAGCCACCCCACAAGATTTAAACAC-3′ and reverse primer 5′-CATGCACTGGATGCACTCTATC-3′. PCR products were purified with Wizard SV gel or PCR clean-up system (Promega) and subjected to sequencing analysis (Macrogen).

Statistical analysis.

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

RESULTS

The A77V mutant virus retains WT levels of infectivity in immortalized cell lines.

The requirement for CA-CPSF6 interactions during HIV-1 replication in macrophages or in vivo has been difficult to study due to the lack of viral mutants defective for CPSF6 binding that are capable of completing reverse transcription in primary cells (50). We became interested in the alanine-to-valine substitution at position 76 (A76V) of SIVmac239 CA since this mutant maintains replicative capacity in primary cells (72) despite the genetic change within the CPSF6-binding pocket. To characterize the effects of this mutation in the context of HIV-1, the corresponding position in HIV-1 CA, A77, was mutated to V. The HIV-1 CA A77V mutant was characterized together with the prototypic CPSF6-binding-deficient HIV-1 CA N74D mutant (11). We first determined the relative infectivities of these two CA mutants compared to that of the WT virus. An equal amount of each VSV-G-pseudotyped virus, as normalized by RT activity, was used to infect two adherent and two suspension cell lines in a single-cycle replication assay. The A77V mutant virus exhibited infectivity comparable to that of the WT virus (Fig. 1) in all the cell types tested. In contrast, infectivity of the N74D mutant virus was slightly impaired in MT4 cells, although it was as infectious as the WT virus in the other three cell lines (Fig. 1), a finding consistent with previous work (11, 50).

FIG 1.

The A77V mutant virus retains WT-level infectivity in cell lines. Single-cycle infection assays were carried out to determine the infectivity of the A77V mutant relative to those of the WT virus and N74D mutant in four different cell lines. Equal amounts of virus particles normalized by RT activity were used for each virus. Results for HeLa and HOS are means of titers obtained by using three different inocula in two different experiments, while those for MT4 and C8166-CCR5 are means from three independent experiments. Viral infectivity by the WT was significantly lower in HeLa and HOS cells than those by N74D (P < 0.05 in HeLa and P < 0.01 in HOS) and A77V (P < 0.001 in both cell types). WT virus showed significantly higher infectivity in MT4 and C8166-CCR5 cells than N74D (P < 0.001) and in MT4 cells than A77V (P < 0.01).

The A77V mutation reduces CA interactions with CPSF6.

A77 is located within the pocket that mediates CA binding to NUP153 and CPSF6 (13, 39, 40, 73). Thus, we asked if the A77V mutation affects CA binding to CPSF6. To address this question comprehensively, four different experimental approaches were utilized. First, we used a genetic assay based on the ability of a C-terminally truncated form of CPSF6, CPSF6-358, to restrict HIV-1 infection. CPSF6-358 localizes to the cell cytoplasm and binds to incoming viral capsids to potently block HIV-1 infection (11). As expected, the WT virus was effectively inhibited in cells expressing CPSF6-358 compared to control cells, whereas the infectivities of N74D and A105T, carrying another CA mutation located in the CPSF6-binding pocket (52), in CPSF6-358-expressing cells were indistinguishable from that in control cells (Fig. 2A). A77V infection was also completely resistant to CPSF6-358 (Fig. 2A), suggesting that the A77V mutation significantly decreases binding of CA to CPSF6.

FIG 2.

The A77V mutation reduces CA binding to CPSF6. (A) Expression levels of a truncated form of CPSF6 (CPSF6-358). Western blots of cell lysates extracted from either control or CPSF6-358-expressing stable cell clones were probed with anti-HA antibody (upper panel) or anti-tubulin antibody (lower panel). HeLa cells stably transduced with the control LPCX vector or one encoding CPSF6-358 were infected with VSV-G-pseudotyped GFP reporter viruses. The means from two independent experiments are shown, with error bars denoting standard error of mean (SEM). Viral infectivity by the WT was significantly lower in CPSF6-358-expressing cells than in control cells (**, P < 0.01). (B) HeLa cells were infected with the WT virus at a multiplicity of infection (MOI) of 0.01. Equivalent virus particles normalized by RT activity were used for infection of all viruses. The results are the means from three independent experiments, with error bars denoting SEM. Addition of either N74D or A77V to RKLM significantly increased viral infectivity (**, P < 0.01). (C) HIV-1 CA-NC tubular complexes were incubated with cell lysates from HeLa cells. A small amount of the mixture was saved before centrifugation and used as input. Representative data from four to six experiments are shown. (D) Summary of binding experiments using HIV-1 CA-NC tubes for three mutants. Shown is binding relative to the WT CA-NC, which was calculated by using values in the pellet (CPSF6) and normalizing to the pellet values of CA-NC. The means from four (A105T) or six (others) independent experiments are shown, with SEM as error bars. Reduction of CA-CPSF6 binding by CA mutations was statistically significant (*, P < 0.05; **, P < 0.01). (E) ITC traces for CPSF6 peptide binding to the CA NTD (residues 1 to 151) of the WT, N74D, and A77V viruses. The data were best fitted to a single-site binding isotherm to determine the binding affinity. The results are representative of two independent experiments.

The genetic assay described above relies on overexpression of the artificial CPSF6-358 molecule to probe potential CA-CPSF6 binding. In contrast, we previously discovered that endogenous CPSF6 can bind and inhibit certain CA mutants (52). For instance, the R132K/L136M (RKLM) mutant virus is blocked by endogenous CPSF6. This block is rescued by transient depletion of CPSF6 or by mutations in the CPSF6-binding domain that reduce the binding to CPSF6, such as N74D. We took advantage of this inhibitory activity of endogenous CPSF6 against the RKLM mutant to examine the effects of A77V on CA interactions with endogenous CPSF6. Consistent with our previous work (52), addition of the N74D mutation to the RKLM mutant virus increased viral infectivity to a level similar to that of the WT virus (Fig. 2B). The A77V mutation phenocopied the N74D mutation and rescued RKLM infectivity (Fig. 2B).

Next, we extended these observations using an in vitro protein binding assay, which examines the ability of assembled CA-nucleocapsid (NC) complexes to pull down endogenous CPSF6 in cell lysates (52). CA-NC tubes made by the WT CA efficiently pulled down endogenous CPSF6 (Fig. 2C), whereas introduction of the N74D mutation severely reduced the amount of endogenous CPSF6 in the pellet (Fig. 2C). The A77V mutation also potently decreased the amount of CA in the pellet fraction (Fig. 2C). Among the three mutants tested, the N74D mutation blocked CA-CPSF6 interactions most strongly, whereas the A77V and A105T mutations reduced CA binding to CPSF6 by ∼70% and ∼20%, respectively (Fig. 2D). The reductions in CA-CPSF6 binding by the three different mutants were statistically significant (Fig. 2D). Finally, we utilized ITC to determine the binding affinity (K_d) of recombinant HIV-1 CA N-terminal domain (CA^N) to a peptide that corresponds to CPSF6 residues 313 to 327 (CPSF6_313–327). The peptide was shown to bind directly to CA^N (13). As expected from work by the James lab, we found that WT CA^N bound CPSF6_313–327 with an affinity of 730 μM, whereas the N74D mutation almost completely abolished binding (Fig. 2E). The A77V mutation reduced the binding of CA^N to CPSF6_313–327 almost to the same extent as N74D. Collectively, these data demonstrate that the A77V mutation, like N74D, significantly reduces CA binding to CPSF6.

The A77V CA mutation relieves the requirement for TNPO3, NUP358, and NUP153 during HIV-1 infection of HeLa cells.

CPSF6 is implicated in HIV-1 nuclear entry. The loss of CPSF6-CA interaction caused by the N74D mutation correlates with the infection of cells independent from cofactors involved in HIV-1 nuclear entry, such as TNPO3, NUP358, and NUP153 (11). Therefore, we tested whether the A77V mutation also altered the cofactor utilization for nuclear entry by using RNA interference. Transient gene knockdown by siRNA, which was successfully used in our previous study (52, 56), was validated by qRT-PCR of total cellular RNA. In two independent experiments, expression of TNPO3 and NUP153 was downregulated on average by 4.8-fold and 9.5-fold, respectively. NUP358 was downregulated by 2.3-fold for subclone F1-5F and by 2.5-fold for subclone F1-8C in shRNA-expressing cells versus control cells. Transient or stable depletion of each of three nuclear entry host cofactors reduced infection of the WT virus by 3- to 5-fold but did not affect or only minimally affected infection of the N74D or A105T mutant viruses (Fig. 3). The phenotype of A77V was indistinguishable from that of N74D, except for NUP358-depleted cells, where viral infectivity was reduced to a slightly larger extent than for N74D (Fig. 3B). These data indicate that nuclear entry of A77V is largely independent of the pathway utilized by the WT virus, which includes TNPO3, NUP358, and NUP153.

FIG 3.

The A77V mutation in the CPSF6-binding pocket relieves the dependence on TNPO3, NUP358, and NUP153 for HIV-1 infection. (A) VSV-G-pseudotyped GFP reporter viruses were used to infect HeLa cells after siRNA knockdown of TNPO3. (B) Two HeLa subclones stably expressing shRNA targeting NUP358 were infected with WT and CA mutant viruses. (C) VSV-G-pseudotyped GFP reporter viruses were used to infect HeLa cells after siRNA knockdown of NUP153. The results are the means from at least four independent experiments, with error bars denoting SEM. Depletion of each gene product significantly decreased infectivity of WT HIV-1 (*, P < 0.05; **, P < 0.01).

A77V differs from the WT but resembles N74D in integration site preferences.

HIV-1 integrates its genome into host chromosomes semirandomly. Genomic features favored for HIV-1 integration include transcription units, active genes, and gene-dense regions (32). HIV-1 integration site selection is, in part, determined by the viral nuclear import pathway (12, 33, 74). For instance, N74D, the prototypic mutant that replicates in cell lines independently from the TNPO3/NUP358/NUP153-dependent nuclear entry pathway, displayed altered integration targeting, with no preference to integrate into gene-rich regions of chromosomes (12, 33). Furthermore, recent work demonstrated that CPSF6 plays a critical role in targeting HIV-1 integration to transcriptionally active chromatin (34, 35). Because A77V can infect cells that are depleted of the known nuclear entry pathway cofactors (Fig. 3) at a level comparable to the N74D mutant virus and significantly reduces CA binding to CPSF6 (Fig. 2), we asked if the A77V mutation altered HIV-1 integration preferences. We initially generated integration site libraries by infecting HOS cells with the A77V mutant as well as control viruses (WT and N74D), and extracted genomic DNA at 5 days postinfection. Integration site analysis was done as described previously (33, 62, 63, 75). Briefly, LM-PCR was utilized to generate libraries of DNA fragments, which were subsequently sequenced on the Illumina platform. Parsing the data for unique integration sites yielded between 2,826 and 4,169 sites for the different infections. These data were compared among the three viruses and to 50,000 random sites that were generated in silico (matched random control [MRC]) for genomic features such as genes, CpG islands, transcriptional start sites (TSSs), and gene density (Table 1).

TABLE 1.

WT and capsid mutant viral integration sites in HOS cells

Genomic feature	Value per genomic annotationa
	WT	N74D	A77V	MRCc
Unique integration sites	2,826	3,969	4,169	50,000
In RefSeq genesb	2,273 (80.4)	2,407 (60.6)	2,525 (60.6)	22,328 (44.7)
CpG islands (±2.5 kb)	146 (5.2)	52 (1.3)	54 (1.3)	2,100 (4.2)
TSSs (±2.5 kb)	120 (4.3)	67 (1.7)	78 (1.9)	2,010 (4.0)
Genes per Mb	18.8	7.1	6.9	8.7

^aNumbers in parentheses are percentage values.

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

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

The A77V mutation altered HIV-1 integration site preferences for all of the genomic features analyzed. Specifically, the A77V mutation ablated the ability of HIV-1 to integrate into gene-rich regions, yielding a mean value of 6.9 genes per Mb, which was not only lower than the WT value of 18.8 genes per Mb but also less than the random MRC value of 8.7 genes per Mb (Table 1 and Fig. 4A). Similarly, statistically significant reductions in the number of integration events near CpG islands, TSSs, and within genes were observed for the A77V mutant compared to the WT virus (Table 1 and Fig. 4B). Strikingly, the changes caused by the A77V mutation in HIV-1 integration site patterns were statistically indistinguishable from those caused by the N74D mutation (Table 1; Fig. 4).

FIG 4.

The A77V mutation eliminates the preference for HIV-1 to integrate within gene-rich regions of chromosomes in HOS cells. (A) Percent integration sites (y axis) plotted against number of genes per Mb (x axis) for infections conducted using the WT virus and mutants (N74D and A77V) as well as for the matched random control (MRC). (B) P values are shown for comparison of WT HIV-1 integration site distribution in HOS cells to that of capsid mutant viruses N74D and A77V, as well as the MRC. Numbers of integration sites within RefSeq genes and nearby CpG islands and TSSs, as well as regional gene density profiles, are listed in Table 1. P values above 0.05 are highlighted in bold and italic. a, P values calculated by Fisher's exact test; b, P values calculated by Wilcoxon rank sum test.

Effects of A77V on viral replication in primary CD4⁺ T cells.

The above findings showed that the A77V mutant virus shares multiple biological phenotypes with the N74D mutant virus. However, other phenotypes of these two mutants may differ in primary cells. First, a chimeric HIV-1 carrying SIVmac CA with the A76V mutation (counterpart of A77V) retained replicative capability in PBMC (72). Second, the A77V mutant displayed viral infectivity that was equivalent to that of the WT virus in immortalized cell lines, in contrast to the N74D virus, which was modestly impaired in T cell lines (Fig. 1). Thus, we next asked whether the A77V mutation preserves the ability of HIV-1 to replicate in primary CD4⁺ T cells. Primary CD4⁺ T cells were infected with GFP-encoding replication-competent viruses expressing the LAI env gene at a multiplicity of infection of 0.005 (0.5% of cells would become GFP positive at 2 days after infection). We found that the A77V mutant replicated in CD4⁺ T cells with an efficiency comparable to that of the WT virus, as opposed to the N74D mutant, which was significantly impaired for replication in primary CD4⁺ T cells (Fig. 5A). We do note a modest but noticeable decrease in the peak viremia of A77V compared to the WT (Fig. 5A). The difference was small yet statistically significant (P = 0.0006) when the number of virus-infected cells at the peak time points was normalized by the number of virus-infected cells at 2 days postinfection.

FIG 5.

Effects of the A77V mutation on viral replication in primary CD4⁺ T cells. (A) Spreading infection of replication-competent GFP reporter viruses in primary CD4⁺ T cells. Primary CD4⁺ T cells were infected with viruses carrying the intact env gene at an MOI of 0.005 (0.5% of cells become GFP-positive at 2 days after infection). As the susceptibility of these cells to HIV-1 infection varied according to blood donor, different amounts of virus inputs were used for each individual donor. Each replication experiment was performed in triplicate, and mean values are plotted, with error bars denoting standard deviation. (B) Viral infectivity of VSV-G-pseudotyped viruses in primary CD4⁺ T cells from three donors. Each symbol represents infectivity in primary CD4⁺ T cells using one of the three donors. At least three independent experiments were performed for each donor. The means are shown as horizontal lines. Viral infectivity by WT was significantly different from that by N74D (**, P < 0.01) whereas the difference between WT and A77V was not significant (P = 0.275). (C) Inhibition of the A77V mutant virus by raltegravir is comparable to that of the WT virus. Primary activated CD4⁺ T cells were infected with GFP reporter viruses, and raltegravir (1 μM) was added to infected cultures at the indicated time points. The means from five independent experiments are shown with error bars that represent SEM.

To further characterize the A77V virus in primary CD4⁺ T cells, single-cycle infectivity assays were performed using env-deficient reporter viruses pseudotyped with the VSV-G protein. In contrast to the case for the N74D virus, only marginal differences between the A77V and WT viruses were observed (Fig. 5B). These differences did not reach statistical significance across multiple experiments (P = 0.275) and therefore suggest that the early steps of viral replication are completed with comparable efficiencies by these viruses. We further defined the kinetics of the early stages of viral infection in primary CD4⁺ T cells by a time-of-addition experiment using the integrase inhibitor raltegravir. There was no change in the timing when intracellular viral complexes became resistant to the antiviral activity of the inhibitor, suggesting that the cumulative kinetics of entry, reverse transcription, and integration by the A77V mutant virus are similar to those of the WT virus in primary CD4⁺ T cells (Fig. 5C).

Finally, we examined the effects of the A77V mutation on the production of virion particles. As we were unable to achieve efficient DNA transfection of primary CD4⁺ T cells, the level and rate of virus particle assembly were determined by measuring virion-associated RT activity in culture supernatants of transfected 293T cells. The luciferase reporter gene present in these constructs served as an internal control for transfection efficiency, which was comparable across the conditions tested (Fig. 6A). The A77V mutation did not impact the magnitude or rate of RT accumulation in cell supernatants over two time points (Fig. 6B). Furthermore, Western blot analysis showed that the A77V mutation did not affect the steady-state levels of intracellular Pr55 Gag expression or the accumulation of CA p24 in the culture supernatant of transfected cells (Fig. 6C). Overall, our studies on the early and late events of HIV-1 replication demonstrated that the A77V mutation did not tangibly affect virus titers or replication kinetics even though this mutation altered the proviral distribution in the host genome rather dramatically.

FIG 6.

The A77V mutation does not affect virion production in 293T cells. (A) 293T cells were transfected with increasing amounts of plasmids carrying luciferase reporter viruses. Cells were lysed at 48 h after transfection and used for measuring luciferase activity. RLU, relative light units. (B) Culture supernatant was collected at 24 h and 48 h after transfection, and exogenous RT activity was measured with a SYBR green-based quantitative PCR assay. (C) Western blot analysis of lysates and virions harvested from 293T cells following transfection with increasing amounts of WT or A77V plasmids. The samples were probed with an anti-CA antibody. HIV-1 protein products are labeled; *, cellular proteins cross-reactive with the anti-CA antibody. Representative data from two independent experiments are shown.

The A77V mutation alters integration site preferences but not the replicative capacity in macrophages.

We next examined the ability of the A77V virus to replicate in macrophages, as one important property of the N74D mutant was its impaired infectivity in this cell type. MDM were infected with GFP-encoding replication-competent viruses carrying the env gene from the CCR5-tropic Bal strain. Virus input (as normalized by RT activity) was selected so that the WT virus would infect 0.5% to 1.0% of the cells at 2 days after infection. WT virus successfully replicated (Fig. 7A), yielding ∼10 to 20% GFP-positive cells at peak time points. As expected (12, 14, 50), the N74D mutant failed to replicate (Fig. 7A). Similarly, the A105T CA mutant exhibited impaired replication in MDM (Fig. 7A). In sharp contrast, the A77V mutant virus replicated as efficiently as WT virus in MDM (Fig. 7A). Similar observations were made with MDM from at least three different donors as well as MDM prepared by using M-CSF instead of GM-CSF (data not shown). Infection of MDM with equal amounts of single-cycle VSV-G-pseudotyped WT and CA mutant viruses revealed similar results, in that the A77V virus infected MDM with an efficiency similar to that of the WT but the N74D and A105T mutants were impaired (Fig. 7B).

FIG 7.

A77V mutant virus efficiently replicates in macrophages without eliciting a detectable IFN response. (A) Spreading infection in MDM. Replication-competent GFP reporter viruses carrying the CCR5-tropic env gene from the Bal strain were used to infect MDM. An MOI of 0.005 was used for initiating infection for the WT virus, such that ∼0.5% of the cells become infected at 2 days after infection. Equivalent amounts of infectious virus particles based on TZM-bl cell titers were used for the CA mutants. Data from two different donors are shown. (B) Infection of MDM with VSV-G-pseudotyped GFP reporter viruses. MDM were infected with the same infectious units of WT and CA mutant viruses (MOI of 0.1). Due to donor variability, actual infection by the WT virus ranged from 4.3% to 18.3% of GFP-positive cells at 2 or 3 days after infection. Infectious units of all the viruses were determined on MT4 cells. Results are shown as the means from three independent experiments. Error bars are SEM. Viral infectivity by the WT was significantly higher than those by N74D and A105T (**, P < 0.01). (C to E) HEK293-ISRE-Luc cells were treated with either IFN (C), poly(I·C) (D), or culture supernatant harvested from MDM infected with two different MOIs of replication-competent GFP reporter viruses (E). One day after incubation, the cells were lysed and analyzed for luciferase activity. (F) The replication kinetics of HIV-1 in MDM from the blood donors used for panel E were assessed by determining RT activity in culture supernatants.

Rasaiyaah et al. reported that the HIV-1 CA N74D mutant induced host innate immune responses upon infection of MDM, leading to a role for CPSF6 in shielding viral nucleic acids from innate immune sensors (14). To address whether the A77V mutant, which similarly reduced CA binding to CPSF6 (Fig. 2), activates innate immune responses, we utilized an IFN bioassay based on a cell line that carries the luciferase reporter gene under the control of the ISRE (65) (Fig. 7C). Transfection of poly(I·C), a known ligand for Toll-like receptor 3 (76), led to the production of type I interferon, as demonstrated by an elevated luciferase activity in the culture supernatant (Fig. 7D). In contrast, infection of MDM with the WT or mutant viruses failed to produce any detectable IFN activity (Fig. 7E). Moreover, IFN activity was not detected when MDM were infected with an increased dose of the N74D virus (three times more RT than for the WT) (data not shown). In our hands, this assay was able to routinely detect a minimum of 1 unit of IFN per ml (Fig. 7C). Therefore, in contrast to the study by Rasaiyaah et al., we did not find that the N74D mutation induced IFN production in MDM. Nevertheless, N74D replication in MDM was significantly attenuated compared to that of either the WT or A77V virus, which both replicated to equivalent levels (Fig. 7F).

The ability of the A77V mutant to infect and replicate in primary MDM allowed us to determine if the changes in the integration site selection observed using an immortalized cell line (Table 1; Fig. 4) were reproducible in a more physiologically relevant cell target. To this end, MDM were infected with VSV-G-pseudotyped HIV-1 carrying either the WT or A77V CA sequence. Integration targeting in MDM was studied as described above, with 1,106 and 904 integration sites obtained for the WT and A77V viruses, respectively. We found that the A77V mutation changed integration target site preferences in MDM just as it did in HOS cells. Specifically, integration into genes and gene-rich regions, as well as nearby CpG islands and TSSs, was favored by the WT virus; however, these preferences were altered by the A77V mutation (Table 2; Fig. 8). As was observed in HOS cells, not only did the A77V mutation ablate the preferences for gene-rich regions, CpG islands, and TSSs, but the mutant virus actively avoided these regions compared to the MRC (Table 2; Fig. 8). The stark differences in integration site preferences between the WT and A77V viruses were validated in a spreading-infection setting where viruses carrying the CCR5-tropic env gene from the Bal strain were allowed to replicate in macrophages for 12 days. The results of integration site analysis for MDM infected with replication-competent viruses were virtually identical to those for MDM infected with VSV-pseudotyped viruses; A77V completely eliminated the integration bias near CpG islands and TSSs and within gene-rich regions of chromosomes (Table 2; Fig. 8). Thus, the drastic alteration in viral integration site selection caused by the A77V mutation is reproduced in macrophages.

TABLE 2.

Distribution of WT and CA mutant A77V integration in MDM

Genomic feature	Value per genomic annotationa
	WT (VSV.G)	A77V (VSV.G)	Bal-WT	Bal-A77V	MRCc
Unique integration sites	1,106	904	20,051	24,231	50,000
In RefSeq genesb	836 (75.6)	579 (64.0)	16,434 (82.0)	16,896 (69.7)	22,328 (44.7)
CpG islands (±2.5 kb)	87 (7.9)	18 (2.0)	1,247 (6.2)	543 (2.2)	2,100 (4.2)
TSSs (±2.5 kb)	76 (6.9)	17 (1.9)	963 (4.8)	714 (2.9)	2,010 (4.0)
Genes per Mb	19.7	8.3	22.6	7.2	8.7

^aNumbers in parentheses indicate percentage values.

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

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

FIG 8.

The A77V mutation reduces HIV-1 integration preference for gene-dense regions in macrophages. (A) Percent integration sites (y axis) plotted against number of genes per Mb (x axis) for infections conducted using the WT virus and A77V mutant virus as well as for the MRC. Integration sites derived from cells infected with replication-competent viruses are indicated by dashed lines. (B) P values are shown for comparison of the VSV-G-pseudotyped (VSV.G) or replication-competent (Bal) WT HIV-1 integration site distribution in MDM to those of VSV.G and Bal A77V mutant viruses, as well as the MRC. Numbers of integration sites within RefSeq genes and nearby CpG islands and TSSs, as well as regional gene density profiles, are listed in Table 2. P values above 0.05 are highlighted in bold and italic. a, P values calculated by Fisher's exact test. b, P values calculated by Wilcoxon rank sum test.

A77V mutant viral replication in humanized mice.

The utilization of CPSF6 is under selective pressure during in vivo viral replication (13, 40, 52). Thus, we further examined the replicative capacity of the A77V mutant virus in a humanized mouse model of HIV-1 infection. While the humanized mice we used in this study do not possess intact immune systems, they offer multiple advantages over in vitro viral replication assays, including their ability to support HIV-1 replication in the absence of exogenous stimuli. Infection of three humanized mice with CCR5-tropic HIV-1 expressing the WT capsid caused high levels of viremia, with ∼10⁶ copies per ml of plasma viral RNA at peak viremia and a high viral load set point (∼10⁵ copies per ml) in two out of the three animals (Fig. 9A, left panel). The A77V mutant virus also displayed peak plasma viremia at ∼10⁶ RNA copies per ml in all four infected animals (Fig. 9A, right panel). At the first time point viremia was measured (2 weeks postinfection [wpi]), two A77V-infected animals (animals 751 and 753) had ∼2 × 10⁵ viral RNA copies per ml, a level indistinguishable from that in WT-infected animals. We noted a delay in the other two A77V-infected animals (animals 399 and 752), in which plasma viral RNA copy numbers were 10-fold lower than those in WT-infected animals at 2 wpi; however, both animals eventually exhibited peak viral loads similar to those in WT-infected animals (more than 10⁶ copies per ml of viral RNA in plasma at 4 wpi for animal 399 and at 6 wpi for animal 752). We noted that the level of the decline in viral RNA after peak viremia differed between the two viruses. While there was a 20- to 100-fold decrease of viral RNA copies in the WT-infected animals from 6 wpi to 8 wpi, the decline in the two A77V-infected animals, which were the only animals available after 6 wpi (animals 751 and 753), was more moderate, with less than a 4-fold decrease in viral loads from 6 wpi to 8 wpi (Fig. 9A).

FIG 9.

Replication and evolution of the A77V virus in humanized mice. (A) Plasma viremia in HIV-1-infected mice. Copy numbers of viral RNA in plasma were analyzed at different time points. (B) Frequency of human CD4⁺ T cells in blood of HIV-1-infected mice. The frequency of peripheral CD4⁺ T cells (human CD45⁺ CD3⁺ CD4⁺ cells) was analyzed at 0, 2, 3, 6, 10, and 14 wpi. (C) Evolution of RNA sequences encoding the amino acid at position 77 in the HIV-1 CA sequence. RNA sequences of viral genomes in four A77V-infected mice were periodically analyzed.

All of the mice infected with the WT virus suffered a decline in CD4⁺ T cells over the time course of the experiment (Fig. 9B). This decline in CD4⁺ T cells is specific for HIV-1 infection because uninfected humanized animals retain a steady level of CD4⁺ T cells for up to 6 months after their generation (unpublished observation). Although we were able to test only three animals for the WT virus, the differences in the percentage of CD4-positive T cells between 0 and 6 wpi were statistically significant (P = 0.01) (Fig. 9B). By comparison, A77V-infected animals exhibited a smaller CD4⁺ T cell decline when values at 0 wpi were compared to those at 6 wpi (Fig. 9B), and this difference was statistically significant (P = 0.04). Accordingly, the level of human CD4⁺ T cells in the WT-infected animals seemed to be somewhat lower at 6 wpi than that in the A77V-infected animals, although the difference was not statistically significant (P = 0.052).

Due to the nature of the genetic code, only a single nucleotide substitution would be required to change a valine codon back to alanine. It therefore seemed possible that the high levels of replication in A77V-infected animals could be due to reversion of the CA mutation back to the WT sequence. However, sequence analysis of bulk populations of PCR-amplified DNA that corresponded to the nucleotide sequence coding for CA position 77 did not support this. Namely, while there was clear evidence for reversion of A77V to WT at later time points (8 wpi), the major genotype during the acute viremia (∼6 wpi) maintained A77V; sequences identified at the first three time points contained exclusively valine at 77, except for the sequence of animal 399 at 6 wpi, in which there was a second small peak for cytosine at the position that would revert the valine back to alanine (Fig. 9C). Since the vast majority of viruses maintained the A77V allele during early time points (∼6 wpi), we conclude that the A77V variant has the capacity to replicate at WT levels in vivo at least during the acute phase of infection.

A fitness advantage conferred by the WT CA for HIV-1 replication in humanized mice.

The A77V virus was almost indistinguishable from the WT virus during the acute phase of viral replication in infected animals. However, as described above, we noted the appearance of the WT sequence, i.e., a valine-to-alanine conversion at position 77 in CA, in three out of the four A77V-infected animals (Fig. 9C) at later time points. This observation suggested that the WT virus might have a growth advantage over the A77V mutant virus evident at later stages of replication. Of the four A77V-infected animals two, animals 399 and 753, died very early, at around 6 wpi. At that point (6 wpi), evidence of reversion to the WT sequence was detected in animal 399 but not animal 753 (Fig. 9C). In contrast, animals 751 and 752 survived longer, and therefore we determined the frequency of each variant (WT and A77V) by sequencing the bulk population of PCR products encompassing the portion of CA containing position 77 at later time points postinfection. We found that alanine (encoded by GCT) at position 77 gradually dominated in both animals, although the valine at position 77 in the variant (encoded by GTT) remained present (Fig. 9C, right panel). PCR fragments containing the entire CA regions of viruses obtained from both animals (animals 751 and 752) at 14 wpi were analyzed, which revealed no changes other than the valine-to-alanine reversion at position 77 (data not shown).

To directly compare the fitness of the A77V variant to that of the WT virus, we performed an in vivo competition experiment in which humanized mice were simultaneously coinfected with the same dose of both WT and A77V viruses, as normalized on an indicator cell line. Infection of three humanized mice with the two viruses together resulted in robust viral replication (Fig. 10A) as well as a steady decline in CD4⁺ T cells (Fig. 10B). We were able to monitor these animals only until 8 wpi because they died for unknown reasons. Sequence analysis of partial gag sequences encompassing the CA domain isolated from plasma viral RNA revealed that the WT virus was the dominant species in the viral population in virtually all animals (Fig. 10C). In fact, the A77V variant was very rare; its existence was detected only at the first time point examined in one animal. Specifically, the viral sequence derived from animal 145 at 1 wpi consisted of a mixture of both the WT (GCT) and A77V mutant (GTT) alleles (Fig. 10C). We further explored this difference by carrying out an ex vivo competition experiment in which the WT virus and the A77V mutant virus, which carried two different fluorescent genes, were used to coinfect primary human CD4⁺ T cells. Consistent with the observation in humanized mice, the WT virus dominated in the competitive replication assay, regardless of the starting amounts of both viruses (Fig. 11). These results suggest that there is a subtle but significant growth advantage of the virus carrying the WT capsid that is particularly important in primary cells in vitro and in vivo.

FIG 10.

The WT virus dominates in coinfected animals. (A) Plasma viremia in HIV-1-infected mice. Copy numbers of viral RNA in plasma were analyzed at indicated time points. (B) Frequency of human CD4⁺ T cells in blood of HIV-1-infected mice. The frequency of peripheral CD4⁺ T cells (human CD45⁺ CD3⁺ CD4⁺ cells) was analyzed at different time points. (C) Sequence analysis of plasma viral RNA from coinfected animals. DNA sequences encoding the amino acid at position 77 in the HIV-1 CA sequence are shown. Viral genomes in the three coinfected mice were periodically analyzed.

FIG 11.

The WT virus dominates A77V in primary CD4⁺ T cells. (A) Competitive replication of GFP-encoding viruses (WT or A77V) with Crimson-encoding WT virus in primary CD4⁺ T cells. Primary CD4⁺ T cells were infected with viruses at an MOI of 0.005 (0.5% of cells become GFP or Crimson positive at 2 days after infection). Results are shown as means, with error bars denoting standard deviation, for triplicate samples. (B) Frequency of GFP-positive cells relative to Crimson-positive cells. The viral infectivity of replication-competent viruses in primary CD4⁺ T cells from three blood donors was measured at 2, 4, 6, and 8 days after infection. Each data point represents the relative frequency of GFP-positive cells in one donor. The values were calculated as 100 × % GFP-positive cells/(% GFP-positive cells + % Crimson-positive cells). The results from the donor used for panel A are also included in this figure. Results are shown as means, with error bars denoting standard deviation, for triplicate samples. The relative GFP positivity by A77V at 8 days after infection was significantly lower than that at 2 days after infection (P < 0.01).

DISCUSSION

In the present work, we found that a unique capsid mutation that significantly reduces CA binding to CPSF6 largely did not affect HIV-1 replication in primary cells and acute infection in humanized mice. However, prolonged in vivo replication revealed an impetus to select viruses that restored CA binding to CPSF6. These results suggest that although optimal CA-CPSF6 interactions are not absolutely necessary for viral replication, they provide a significant growth advantage. It is possible that the selection for the WT CA sequence was driven by a function unrelated to CPSF6, but we think that this is unlikely, as there was no phenotypic change associated with the A77V mutation in the efficiency of capsid assembly (Fig. 6). These observations accord with a strong selective pressure to preserve CA-CPSF6 interactions of HIV-1 in HLA-B27-positive subjects (52).

What drives selection for optimal CA interactions with CPSF6 during HIV-1 replication in vivo? A prominent role suggested for CPSF6 during HIV-1 replication involves the utilization of host factors active in nuclear entry (TNPO3, NUP358, and NUP153) (11, 31). The A77V mutant infected an immortalized cell line largely independent from these cofactors. Thus, the route of nuclear entry mediated by TNPO3, NUP358, and NUP153 may be one major driving force for selection against the A77V mutation in vivo. One limitation of our work is the inability to determine the cofactor usage in primary cells. Thus, it is theoretically possible, although we think it is unlikely, that the utilization of nuclear entry cofactors may differ between immortalized cell lines and primary cells and by extension that the A77V virus may utilize TNPO3, NUP358, and NUP153 in primary cells. Nonetheless, our observation accords with the fact that utilization of these nuclear entry molecules is a well-conserved property among primate immunodeficiency viruses, thus arguing for an evolutionary advantage in maintaining interactions with these molecules during infection by HIV-1 and related viruses (12, 23, 25, 27, 52, 77).

Another potential advantage in selecting for optimal CA-CPSF6 interactions during in vivo viral replication is HIV-1 integration site selection. Recent work demonstrated that CPSF6 plays a prominent role in HIV-1 integration targeting (34, 35). Consistently, the A77V virus, which significantly reduced CA binding to CPSF6, drastically altered the integration pattern, including the loss of the tendency to integrate near CpG islands or TSSs, and reduced the preference for integration within genes and gene-dense regions (Fig. 4 and 8; Tables 1 and 2). HIV-1 integration can occur in close proximity to nuclear pores (78), which may provide chromatin organization that facilitates efficient viral replication (79, 80). In fact, HIV-1 integration sites can affect the basal transcription of the LTR promoter (81). Therefore, it is reasonable to assume that the specific pattern of HIV-1 integration site preference promotes viral replication (79, 80, 82, 83). This idea appears to be consistent with our observation that in vivo viral evolution selected for virus that restored the codon for alanine at position 77 in CA and hence the WT pattern of integration (Fig. 9). The subtle relationship between integration site selection and viral replication in vitro accords with a recent study in which HIV-1 integrase polymorphisms that altered integration site selection and that were associated with rapid disease progression did not grossly alter viral gene transcription, viral replication, or virus-mediated cytotoxicity in cell culture (84).

However, it is prudent to exercise caution when drawing direct links between HIV-1 integration site targeting and viral replication. Our experiments uncovered robust viral replication in both primary cell and in vivo settings by the A77V mutant virus (Fig. 5, 7, and 9) under conditions where HIV-1 integration sites differed dramatically from those for the WT virus. Thus, any effects of the altered integration site preference caused by the A77V mutation on viral replication should be modest at best and may be miniscule. This interpretation is consistent with observations that redirecting integration to novel sites through the use of artificial LEDGF/p75 fusion proteins did not exert strong effects on integration efficiency or on the expression of the resultant proviruses, suggesting that the location of integrated DNA has only marginal effects on viral transcription in cell line models (85, 86). Importantly, while the A77V mutant virus completely lost the preference to integrate into gene-rich regions, it still favored integration within genes (∼20% reduction compared to WT) (Fig. 4 and 8; Tables 1 and 2). Thus, one possibility is that integration within genes may be more relevant for promoting viral replication.

Another proposed function of CPSF6 is to promote viral replication in macrophages (12, 14, 50, 51). Consistent with previous observations (12, 14, 50, 51), we found that the N74D mutant virus was significantly impaired in macrophages, as was another CPSF6-binding-defective CA mutant, A105T (Fig. 7). These observations are in line with a report by Zhou et al., who observed that mutations that reduced CA binding to CPSF6 also impaired viral replication in macrophages (51). In contrast, the A77V mutant, in which the CA-CPSF6 interaction also was significantly altered (Fig. 2), retained the ability to replicate in macrophages as efficiently as WT virus (Fig. 7). Although the A77V mutation did not reduce CA-CPSF6 binding to the same extent as the N74D mutation in an in vitro CA-NC copelleting assay (Fig. 2C), the A77V change reduced the extent of the protein-protein interaction as efficiently as N74D in the ITC experiment (Fig. 2E). Additionally, A77V, like N74D, was completely resistant to inhibition by CPSF6-358 (Fig. 2A). The A105T mutation, which had the smallest effect on CA-CPSF6 binding among the three mutants tested, reduced HIV-1 infection of macrophages to a degree comparable to that of N74D, which abolished the CA-CPSF6 interaction biochemically (Fig. 2 and 7). Therefore, although the replication of certain CPSF6-binding-defective mutants is significantly attenuated in macrophages, the A77V mutant robustly grows under this condition.

One of the suggested mechanisms for the attenuated replication of CPSF6-binding-defective mutants in macrophages is their ability to activate innate immune responses. To address this potential mechanism, we employed a bioassay (65). None of the viruses examined, including the N74D mutant virus, reproducibly induced any detectable IFN in this bioassay (Fig. 7), which has a detection limit of 1 unit of IFN per ml. Currently, it is unclear why we were unable to replicate the observation by Rasaiyaah et al. (14). As their observation was made with primary macrophages, donor variation and/or subtle experimental differences (e.g., procedures for macrophage and virus preparation, titration techniques, and IFN detection assays) might significantly affect outcomes. Nevertheless, our data suggest that the impairment of N74D virus replication in macrophages is not due to induction of an IFN response and support alternative explanations such a block to N74D virus infection that occurs prior to reverse transcription, as described by Ambrose et al. (50).

Overall, our data demonstrate that the deficiency in CPSF6 interaction alone cannot account for attenuated virus replication in macrophages. They also support the findings that CPSF6 interactions are critical for target site selection during integration and that optimal CA-CPSF6 interactions, although apparently dispensable in a number of assays, confer a significant fitness advantage during HIV-1 replication in vivo.