A Novel Phenotype Links HIV-1 Capsid Stability to cGAS-Mediated DNA Sensing

The HIV-1 capsid, which is made from individual viral capsid proteins (CA), is a target for a number of antiviral compounds, including the small-molecule inhibitor PF74. In the present study, we utilized PF74 to identify a transmitted/founder (T/F) strain that shows increased capsid stability. Interestingly, PF74-resistant variants prevented cGAS-dependent innate immune activation under a condition where the other T/F strains induced type I interferon. These observations thus reveal a new CA-specific phenotype that couples capsid stability to viral DNA recognition by cytosolic DNA sensors.

ABSTRACT

The HIV-1 capsid executes essential functions that are regulated by capsid stability and host factors. In contrast to increasing knowledge on functional roles of capsid-interacting host proteins during postentry steps, less is known about capsid stability and its impact on intracellular events. Here, using the antiviral compound PF-3450074 (PF74) as a probe for capsid function, we uncovered a novel phenotype of capsid stability that has a profound effect on innate sensing of viral DNA by the DNA sensor cGAS. A single mutation, R143A, in the capsid protein conferred resistance to high concentrations of PF74, without affecting capsid binding to PF74. A cell-free assay showed that the R143A mutant partially counteracted the capsid-destabilizing activity of PF74, pointing to capsid stabilization as a resistance mechanism for the R143A mutant. In monocytic THP-1 cells, the R143A virus, but not the wild-type virus, suppressed cGAS-dependent innate immune activation. These results suggest that capsid stabilization improves the shielding of viral DNA from innate sensing. We found that a naturally occurring transmitted founder (T/F) variant shares the same properties as the R143A mutant with respect to PF74 resistance and DNA sensing. Imaging assays revealed delayed uncoating kinetics of this T/F variant and the R143A mutant. All these phenotypes of this T/F variant were controlled by a genetic polymorphism located at the trimeric interface between capsid hexamers, thus linking these capsid-dependent properties. Overall, this work functionally connects capsid stability to innate sensing of viral DNA and reveals naturally occurring phenotypic variation in HIV-1 capsid stability.

IMPORTANCE The HIV-1 capsid, which is made from individual viral capsid proteins (CA), is a target for a number of antiviral compounds, including the small-molecule inhibitor PF74. In the present study, we utilized PF74 to identify a transmitted/founder (T/F) strain that shows increased capsid stability. Interestingly, PF74-resistant variants prevented cGAS-dependent innate immune activation under a condition where the other T/F strains induced type I interferon. These observations thus reveal a new CA-specific phenotype that couples capsid stability to viral DNA recognition by cytosolic DNA sensors.

INTRODUCTION

The capsid of HIV-1, a cone-shaped shell, is made of the capsid protein (CA) subunit and encases essential components, including two copies of viral RNA and the enzymes reverse transcriptase (RT) and integrase (IN) (1, 2). Upon delivery into the cytoplasm, the viral capsid engages the host cytoskeleton network to migrate toward the nucleus, where the reverse-transcribed viral DNA is inserted into the host chromatin (3–5). During this early phase of the viral replication cycle, the HIV-1 core transforms itself into physically distinct subviral structures, called reverse transcription complexes (RTCs) and preintegration complexes (PICs), through a process called uncoating or capsid disassembly (6, 7). Uncoating, a temporally and spatially regulated process, results in the shedding of CA subunits from the core.

The molecular and mechanistic details of capsid disassembly remain incompletely understood, but a consensus view is that uncoating is exquisitely controlled by both intrinsic capsid stability and extrinsic elements, including capsid-binding cellular factors. Intrinsic capsid stability is governed by interactions between CA subunits that generate CA oligomers (hexamers and pentamers), the basic units of the viral capsid that interact with each other to form a capsid lattice in a fullerene cone (8–10). Both the N-terminal domain (NTD) and the C-terminal domain (CTD) of CA are linked with a short stretch of a flexible linker and involved in the assembly of viral capsids. Interactions between individual CA subunits, which are mediated by NTD-NTD, NTD-CTD, and CTD-CTD interfaces (9), are critical for capsid stability (11–14). To what extent each of these interfaces contributes to capsid stability is incompletely understood. As described above, capsid stability and uncoating are also regulated by cellular factors. At least a dozen host proteins are known to directly bind to the viral capsid (reviewed in reference 15), although how each of the molecules either stabilizes or destabilizes the viral capsid is controversial. Nonetheless, a significant role played by capsid-interacting cellular proteins is well illustrated by the rhesus monkey variant of TRIM5α (16), a potent antiviral molecule that interacts with the viral capsid and accelerates uncoating to prevent reverse transcription and, as a result, severely limits viral infection. In addition, a small molecule called inositol hexakisphosphate, which has recently been shown to be a cofactor for HIV-1 assembly (17), displayed the ability to regulate capsid stability (18).

Proper disassembly of the capsid is critical for nearly every step during the early phase of HIV-1 replication, up until viral integration into the host chromosomes. For instance, defects of CA mutations in viral DNA synthesis suggest a role of uncoating in the progression of reverse transcription (11, 19–21). These functional interplays are not unidirectional; postentry steps also affect the progression of capsid disassembly. A notable example is the effect of reverse transcription on the extent and kinetics of uncoating (22–25). The functional roles of CA and its shedding from intracellular viral complexes are extended to the nucleus (26–30). CA is the major determinant for HIV-1 nuclear entry (31–37) and critically contributes to the unique preference of HIV-1 integration targeting (32, 38–41). Thus, viral uncoating can be viewed as a complex series of events in which the viral capsid and its matured complexes shed CA subunits in both the cytoplasm and the nucleus.

Another function of the viral capsid that has been proposed in the literature is to shield nascent viral DNA from innate immune sensors (42–44). HIV-1 reverse transcription takes place in the cytoplasm, and thus, viral DNA can be recognized by cytosolic DNA sensors to trigger an innate immune response (45). Cytoplasmic sensing of HIV-1 DNA appears to be context dependent: for example, HIV-1 infection of monocyte-derived macrophages (MDMs) does not appear to induce a robust innate immune response (46, 47). Nonetheless, macrophages are equipped with functional cytosolic DNA-sensing machinery, which consists of cyclic GMP-AMP synthase (cGAS) and the adaptor protein stimulator of interferon genes (STING), as well as other regulators and downstream effectors, such as interferon gamma-inducible factor 16 (IFI16), PQBP1, TANK-binding kinase 1 (TBK-1), and interferon regulatory factor 3 (IRF3) (45, 48, 49). The lack of a robust type I interferon (IFN) response upon HIV-1 infection of macrophages can be explained by the presence of the cytosolic exonuclease TREX1 (47), as well as the negative regulation of host factors by viral accessory proteins (reviewed in reference 50). Additionally, it has been proposed that the capsid cloaks viral DNA from being sensed by cytosolic DNA sensors (51, 52). Perturbation of CA interactions with the host cell factors cyclophilin A (CypA) and cleavage and polyadenylation specificity factor subunit 6 (CPSF6) was shown to trigger innate immune responses and IFN production in macrophages (52). The involvement of CypA in innate sensing of HIV-1 DNA was also observed in monocyte-derived dendritic cells (51).

Capsid stability is the major intrinsic property that regulates uncoating (6, 7). HIV-1 capsid stability, which is dictated by fine-tuned interactions between individual CA subunits, must be optimal to ensure proper uncoating, which is critical for viral replication (11). Our current understanding of HIV-1 capsid stability is largely built upon the biochemical characterization of viral cores prepared in vitro (11). Core yield, the quantity of CA that is associated with isolated cores, is a standard measure for capsid stability (53). A similar core isolation technique can be combined with microscopy-based observations of the physical associations of CA with core particles (54). There is general agreement between biochemical and microscopic techniques on the behavior of representative CA mutants with altered capsid stability, although some discrepancies have been noted (22, 54, 55). An inevitable drawback of these powerful techniques is their laborious experimental procedures that preclude large-scale studies of various mutants or diverse naturally occurring variants. A complementary approach, such as a recently described assay exploiting the exposure of a virion-associated mRNA reporter (56), is needed to further deepen our understanding of capsid stability.

In the present work, we exploited PF-3450074 (PF74), a capsid-binding small-molecule compound (57), as a tool to study capsid functions. PF74 was shown to destabilize the viral capsid in certain assay systems (55, 58–61), although the compound did not affect capsid stability and even stabilized cores in imaging-based assays (27, 54, 62). We used the capsid-targeting activity of PF74, together with cell-free and cell-based assays, to reveal a novel naturally occurring phenotype of capsid stability that drastically alters cGAS-dependent sensing of HIV-1 DNA and highlights an underappreciated capacity of HIV-1 to accommodate phenotypic variation in the viral capsid.

RESULTS

Resistance to effects of high doses of PF74.

In the present study, we utilized PF74 to study capsid functions. A unique dose-response curve of PF74 (Fig. 1A) corresponds to two distinct mechanisms of action, in which low doses block a step following reverse transcription, whereas high doses block reverse transcription (26, 27, 57–59, 61, 63, 64). Among a panel of CA mutants examined for their sensitivity to PF74 (unpublished results), the R143A mutant was distinct because its difference from the wild-type (WT) virus of the LAI strain was more pronounced at high drug concentrations (Fig. 1A). Namely, the antiviral activity by PF74 at 2.5, 5, and 10 μM against the R143A mutant was significantly less than that against the WT virus (Fig. 1A and B). However, when PF74 dissociation constants (K_d) were measured using equilibrium dialysis, the R143A substitution did not markedly alter the affinity of CA hexamers to PF74 (Fig. 1C). This finding suggests that PF74 can bind to the capsid of the R143A virus, with the caveat that stabilized CA hexamers, and not native cores, were used in the binding assay.

FIG 1.

A CA mutation confers resistance to antiviral activity at high PF74 concentrations while maintaining PF74 binding. (A) GFP-encoding reporter viruses carrying either the WT or R143A capsid were used to infect HeLa cells transduced with the empty LPCX vector with increasing amounts of PF74. Relative infectivity was calculated by setting the number of GFP-positive cells without PF74 treatment as 100%. The mean values were obtained from five independent experiments. Error bars denote the standard error of the mean (SEM). *, P < 0.05 (calculated with the unpaired Student's t test). (B) The degree of PF74-mediated inhibition at high drug doses was quantified by use of the results shown in panel A. Individual dots correspond to data points from each experiment. The results were analyzed with the unpaired Student's t test. (C) The affinity of PF74 for WT and R143A hexamers was determined using equilibrium dialysis. One representative result from two independent experiments with similar results is shown here. The K_d value for the R143A mutant in the other experiment was 0.244 μM.

The R143A mutant specifically resisted antiviral activity at high PF74 concentrations, even though it did not exhibit a substantially altered CA hexamer affinity for PF74 (Fig. 1). As high drug concentrations were shown to destabilize the viral capsid in certain assays (55, 58–61), one possible mechanism of the observations is that the R143A mutant neutralizes the core-destabilizing activity by PF74. To test this hypothesis, we examined the effects of PF74 on purified HIV-1 cores using a biochemical approach (11). In control experiments performed to validate our experimental system, we observed that addition of a detergent, Triton X-100, in the core isolation procedure led to the appearance of a broad peak of reverse transcriptase (RT) in dense fractions (data not shown). The density of one of the fractions (1.25 g/ml) was within the range of densities known to contain retroviral cores in previous studies (11, 65–67). Furthermore, two representative CA mutant viruses that have either hyperstable (the E45A mutant) or unstable (the Q219A mutant) cores (11) behaved as expected in our experimental system (data not shown). Consistent with previous observations (58), pretreatment of the virus with PF74 reduced the level of p24 CA in fractions that contain cores (Fig. 2A, fractions 6 and 7). Furthermore, the quantity of RT molecules associated with similar high-molecular-weight complexes was substantially decreased upon PF74 treatment (Fig. 2A, bottom) in a dose-dependent manner (Fig. 2B). In contrast, the amount of RT in the same dense fractions was barely changed for the PF74-resistant mutant 5mut, which encodes five amino acid substitutions (Q67H, K70R, H87P, T107N, and L111I) (57, 58) (Fig. 2C). Thus, consistent with the previous work (58), PF74 appears to destabilize the viral capsid in this particular assay.

FIG 2.

PF74 reduces CA and RT molecules in core-containing fractions. (A) HIV-1 cores were separated on a discontinuous sucrose gradient after incubation in the presence or absence of 10 μM PF74 at 37°C for 1 h. The amount of CA (top) and RT (bottom) in each fraction was measured by a p24 CA ELISA and the SG-PERT assay, respectively. A representative profile from two independent experiments is shown on the left. Additional experiments were performed to measure CA and RT for core-containing fractions 6 and 7. Each dot represents the amount measured in a single experiment. Dots that represent results originating from the same experiment are connected with lines. Differences between the two conditions were studied with the unpaired t test. (B) Virus particles were treated with increasing amounts of PF74 and subjected to the core isolation procedure. The amount of RT molecules in the indicated fractions was measured by the SG-PERT assay. The mean from three independent experiments is shown, with error bars (shown in gray) denoting the SEM. Best-fit curves were generated by nonlinear regression using log-transformed data. (C) A drug-resistant mutant (5mut) was compared to the WT virus as described above. The data shown here are representative of those from two independent experiments.

We next asked how the R143A mutation affects the sensitivity to the capsid-destabilizing activity of PF74 (Fig. 3). Having shown that PF74 decreases core-associated CA and RT molecules (Fig. 2), we utilized the PCR-based RT assay in the following experiments, as it is more sensitive than the CA p24 enzyme-linked immunosorbent assay (ELISA). Consistent with the cell-based assay, in which R143A was resistant to antiviral activity by high PF74 doses (Fig. 1A and B), the R143A mutation partially neutralized the core-destabilizing activity of PF74 (Fig. 3). Specifically, the magnitude of the PF74-mediated reduction of RT molecules in core-containing fractions (fractions 6 and 7) was smaller for the R143A mutant than for the WT virus (Fig. 3A). The difference was statistically significant across five independent experiments, although the magnitude of the decrease varied between these experiments (Fig. 3B). The amount of RT in fraction 6 in the absence of the drug, which appeared to correlate with capsid stability, did not differ between the two viruses (Fig. 3C). These observations suggest that the R143A substitution confers resistance to the antiviral activity of PF74 by counteracting its capsid-destabilizing activity.

FIG 3.

The R143A mutant virus partially resists the core-destabilizing activity of PF74. (A) The core yield of the R143A mutant was compared to that of the WT virus in the presence or absence of 10 μM PF74 by the method described in the legend to Fig. 2. A representative result from five independent experiments is shown. (B) The core-destabilizing activity of PF74 against two viruses is depicted as the amount of RT in fraction 6 treated with 10 μM PF74 relative to that with no drug treatment. The data shown here are the results from five independent experiments and were examined using the two-tailed paired t test. (C) Core yields are shown as the RT amount in fraction 6 normalized to the amount of input used in each experiment. Data were compiled from six experiments and analyzed by the two-tailed unpaired t test. NS, not significant.

A naturally occurring genetic polymorphism at the trimeric hexamer interface confers resistance to PF74-induced capsid destabilization.

The results of our studies of the lab-generated mutant R143A uncovered a novel CA-specific phenotype. We extended our study to primary strains for understanding the natural phenotypic variation of the HIV-1 capsid. Ten transmitted founder (T/F) viruses (68), which were assessed for their sensitivity to PF74, displayed considerable variations in dose-response curves against PF74 (data not shown). Among them, one strain (CH040) was distinct because its resistance to antiviral activity was more pronounced at high drug concentrations, a phenotype similar to that of the aforementioned R143A mutant (Fig. 1). Transfer of the DNA sequence encoding the entire CA domain of the Gag protein from CH040 to the LAI backbone conferred this phenotype, in which PF74 lost its potency at high, but not low, drug concentrations (Fig. 4A). Because the LAI-based chimeric virus carrying CA from CH040 was very similar to the R143A mutant virus in their PF74 resistance profiles, we reasoned that the capsid of CH040 resists PF74-mediated inhibition by neutralizing its capsid destabilization activity. To test this idea, the PF74 effects on the core yield were examined by using the same spin-through procedure (Fig. 4B). Similar to the R143A virus, the chimeric virus carrying CA from CH040 partially neutralized the core-destabilizing activity by PF74, as it yielded a higher level of RT molecules in core-containing fractions than the WT LAI strain (Fig. 4B) with statistical significance (Fig. 4C). Isolated CH040 cores had a 2-fold larger quantity of RT than the LAI capsid in the absence of PF74 (Fig. 4D).

FIG 4.

The capsid from a primary strain confers resistance to antiviral activity and the core destabilization effects of PF74. (A) Dose-response curves of LAI and LAI encoding CH040 CA were generated on TZM-bl (left; n = 2) and MT4 (right; n = 7 for LAI, n = 4 for CH040) cells using an env/nef-defective GFP reporter virus. Luciferase activity was used as the readout for TZM-bl cells, whereas the number of GFP-positive cells was used for MT4 cells. Relative infectivity was calculated by setting the data points without PF74 treatment as 100%. The average values are shown, with error bars denoting SEM. **, P < 0.01 (calculated with the unpaired Student's t test). (B) The core yields of the virus carrying the capsid from CH040 were compared to those of the parental LAI virus in the presence or absence of 10 μM PF74 by the method described in the legend to Fig. 2. A representative result from eight independent experiments is shown. (C) The effects of PF74 on capsid stability were assessed by examining the amount of RT in fraction 6 of PF74-treated samples compared to that in samples without PF74 treatment. Results from seven independent experiments are shown and were analyzed using the two-tailed, paired t test. Values obtained from the same experiment are connected by lines. (D) RT amounts in a core-containing fraction (fraction 6) were used to determine the core yield. Results are shown as relative values of core yields that were normalized to those by the LAI virus. The data were compiled from 11 independent experiments. The P value was derived using the Wilcoxon matched-pairs signed-rank test.

The CA-coding sequence of CH040 differs from that of LAI (Fig. 5A). Among nine amino acid differences between these two sequences, a threonine-to-isoleucine change (T216I; from LAI to CH040) was noticeable, as this change is located at the trimeric interface for CA hexamer-hexamer interactions (12, 14, 69) and the same T216I substitution emerged as a second-site suppressor mutation to rescue the P38A CA mutant, the capsid of which is unstable (70). To determine whether this genetic polymorphism dictates the difference in core-related phenotypes between LAI and CH040 viruses, the T216I substitution was introduced into the LAI strain, whereas the reciprocal isoleucine-to-threonine (I216T) substitution was introduced into the clone encoding the CH040 CA protein. These two new viruses exhibited opposing phenotypes; the T216I change conferred the LAI strain with PF74 resistance, specifically at high PF74 concentrations, whereas the I216T substitution rendered the chimeric virus carrying CH040 CA more susceptible to antiviral activity at high doses of PF74 (Fig. 5B). We also compared the effects of PF74 on the core yields of LAI carrying the T216I mutation to those on the WT LAI strain (Fig. 6). The LAI mutant carrying the T216I substitution differed from the parental LAI virus (Fig. 6A) but resembled the LAI chimera with CH040 CA (Fig. 4B) in the abundance of RT molecules associated with core-containing fractions (Fig. 6B) in the presence of PF74. It was difficult to judge the effects of T216I on capsid stability in the absence of PF74 treatment due to the presence of an outlier (Fig. 6C). Without the outlier, the core yield of the T216I mutant in the absence of PF74 was lower than that of the parental WT virus by 3-fold (P = 0.0048). These results indicate that a naturally occurring genetic polymorphism at the trimeric interface between CA hexamers allowed HIV-1 to gain PF74 resistance.

FIG 5.

Amino acids at position 216 in CA govern the sensitivity to antiviral activity of high PF74 doses. (A) Alignment of HIV-1 CA amino acid sequences of LAI and CH040. Amino acids of CH040 that are identical to those of LAI are shown as dots. The amino acids at position 216 are highlighted in red. (B) Dose-response curves of PF74 for these viruses were generated by infecting MT4 cells with GFP-encoding reporter viruses in the presence of increasing amounts of PF74. Results are displayed as relative infectivity normalized to the infectivity in control cells without PF74. The mean values from at least three independent experiments are shown, with error bars denoting SEM (n = 7 for LAI, n = 3 for LAI + T216I, n = 3 for CH040, and n = 5 for CH040 + I216T). **, P < 0.01 (two-tailed, unpaired t test).

FIG 6.

The T216I substitution confers partial resistance to PF74-induced core destabilization. (A) PF74 effects on HIV-1 cores. Virus particles of two LAI-based viruses were subjected to the spin-through core isolation procedure after incubation with or without 10 μM PF74. (B) PF74 effects on the core yield are depicted as the normalized values of the RT amounts in fraction 6 relative to those obtained without drug treatment. Log₁₀-transformed values are shown and were used for analysis by the two-tailed paired t test. (C) Core yields were derived by the amount of RT in fraction 6. The normalized amounts of core yield to the WT LAI virus from four independent experiments are shown. Data were analyzed by the two-tailed unpaired t test.

Single-virus imaging studies in vitro and in cells reveal that PF74-resistant variants have stable capsids.

Our observations suggested that CH040 differs from LAI in capsid stability and that the T216I substitution is responsible for the difference. To test this possibility, we utilized two complementary imaging assays established in previous work (54, 55). In the first approach, we colabeled HIV-1 pseudoviruses by incorporating integrase fused to mNeonGreen (INmNeonGreen [INmNG]) and cyclophilin A (CypA) tagged with DsRed (CypA-DsRed). We have previously demonstrated the utility of the CypA-DsRed fusion protein as a noninvasive marker for the viral capsid that allows monitoring the loss of HIV-1 CA from single cores in vitro and in living cells (54, 71). To assess the core stability in vitro, double-labeled pseudoviruses were adhered to coverslips and permeabilized with saponin, and the loss of CypA-DsRed from single virions was monitored by time-lapse imaging. These experiments revealed that the kinetics of CypA-DsRed loss from PF74-resistant viruses were delayed compared to those of the LAI virus (Fig. 7A and B). Specifically, a significantly higher fraction of single INmNG-labeled LAI cores than PF74-resistant virus (R143A, T216I, and CH040 mutant) cores lost CypA-DsRed (Fig. 7A and C). In agreement with the previously published results, the number of INmNG spots remained relatively constant within 30 min after permeabilization. These observations imply that the capsids of PF74-resistant viruses are more stable than those of the WT LAI virus (Fig. 7B and C).

FIG 7.

Microscopic analysis of HIV-1 core stability in vitro. (A) Images of single CypA-DsRed and INmNeonGreen spots at time zero and 30 min after permeabilization with saponin (the same field of view is shown for both time points). Bars, 2 μm. (B) Kinetics of single-virus uncoating in vitro, as measured by the loss of CypA-DsRed spots. An arrow indicates the time of CsA addition. Error bars are SEM. A representative result of four experiments for LAI WT, LAI + T216I, and CH040 and two experiments for R143A is shown. (C) The bar chart shows the average stable fraction of CypA-DsRed and INmNeonGreen spots at 30 min after permeabilization in multiple experiments. ***, P < 0.0001 (unpaired t test).

We next examined single-HIV-1 uncoating in the cytoplasm using the same INmNG- and CypA-DsRed-labeled viruses pseudotyped with vesicular stomatitis virus G (VSV-G) glycoprotein (Fig. 8A and B). In this cell-based assay, VSV-G-mediated viral fusion releases the HIV-1 cores into the cytoplasm, where different outcomes are observed: (i) highly unstable cores exhibiting abrupt uncoating followed by proteasomal degradation of viral complexes and (ii) stable cores that survive for several hours in the cytoplasm and that usually exhibit a very slow/gradual loss of CA/CypA-DsRed (54). A minor fraction of the relatively stable postfusion HIV-1 cores is readily identified by treating cells with cyclosporine (CsA), which displaces CypA-DsRed selectively from cores in the cytoplasm but not from intact viruses trapped in endosomes (Fig. 8A). Thus, the fraction of long-lived cores that does not complete uncoating within the first 90 min of infection correlates with the intrinsic capsid stability in living cells (54). We observed a significantly higher fraction of stable postfusion INmNG-labeled cores (i.e., cores that lost CypA-DsRed after CsA treatment) (54) for the PF74-resistant variants than for the LAI WT at 90 min postinfection (Fig. 8B). This result further supports the notion that the PF74 resistance is associated with increased capsid stability.

FIG 8.

The CH040 capsid and T216I substitution both stabilize cytoplasmic HIV-1 cores. Single-virus core stability was studied with two different cell-based imaging assays. The core stability of HIV-1 LAI and CH040 WT viruses or LAI-based CA mutants R143A and T216I was measured in TZM-bl cells (A and B) or in HeLa cells (C). (A) Postfusion cores that did not uncoat during the first 90 min postinfection were readily identified by the loss of CypA-DsRed (arrows) in response to CsA treatment, whereas endosomal viral complexes retained CypA-DsRed under these conditions (yellow dashed circles). Images are maximum-intensity projections of the same cell. Bar, 5 μm. (B) Bar chart for the fraction of stable postfusion HIV-1 cores for different capsid variants. The results represent the averages from three experiments. Statistical analysis was performed using Student's t test. *, P < 0.05; **, P <0.01. (C) HeLa cells were infected with three HIV-1 variants containing fluorescently tagged integrase and stained for EU (viral RNA) at different time points. Data were compiled from the results of three independent experiments and normalized with the data set for the first time point (15 min postinfection). Mean values for the normalized data are shown, with error bars denoting SEM. Capsid permeabilization kinetics were analyzed by multiple linear regression modeling in the SAS statistical package to test if the slopes were equivalent. Using the F statistic, the three slopes were not equivalent, with P being equal to 0.025. Using the F statistic with the Bonferroni correction for pairwise comparisons, P values of less than 0.0167 were considered statistically significant. The P value for CH040 versus LAI + T216I was 0.59. The comparison of LAI WT to LAI + T216I or to CH040 yielded P values of 0.011 and 0.038, respectively.

To further examine whether the stable cores were intact or partially opened, we utilized an alternative imaging method, which takes advantage of the dependence of viral RNA staining on the opening of the viral capsid (55, 72, 73). Previous work showed that the 5-ethynyl uridine (EU) incorporated into HIV-1 RNA during virus production can be stained in cells after virus infection, with a decrease in RNA detection over time occurring concurrently with uncoating and reverse transcription (55). The rate of RNA staining is affected by capsid stability, such that viruses with hyperstable capsids have slower RNA staining kinetics and a loss of RNA signal than WT HIV-1 capsid and such that viruses with unstable capsids show faster RNA staining. In this study, RNA staining of HIV-1 with the CH040 capsid was distinct from that of WT HIV-1 (Fig. 8C). WT LAI virus had a steady decrease in RNA staining after the initial time point, similar to previous results (55, 72, 73). In contrast, higher numbers of RNA-positive particles were observed at 60 and 90 min postinfection for the CH040 virus than for the LAI virus, consistent with delayed core permeabilization. The T216I substitution in the LAI capsid phenocopied the CH040 capsid in this core permeabilization assay (Fig. 8C). These results also suggest that the capsid of CH040 is more stable than that of LAI and that this difference is regulated by the T216I substitution.

PF74-resistant variants prevent innate sensing of HIV-1 DNA by cGAS.

One potential consequence of stabilized capsids and delayed uncoating in target cells is the prevention of viral DNA exposure to cytosolic innate sensors. To test this idea, we used a previously described experimental system based on the monocytic cell line THP-1, in which HIV-1 infection activates cGAS-mediated IFN signaling (45). As reported previously (45), THP-1 infection with green fluorescent protein (GFP) reporter HIV-1 induced type I IFN as well as IP-10, one of the interferon-stimulated genes (ISGs) (Fig. 9A). This innate immune activation was blocked by nevirapine, a reverse transcriptase inhibitor (Fig. 9A), but not by raltegravir or dolutegravir, integrase inhibitors (data not shown). Depletion of cGAS significantly reduced the ability of HIV-1 to induce type I IFN upon viral infection (Fig. 9B).

FIG 9.

The R143A CA mutation suppresses HIV-induced innate responses in THP-1 cells. (A) THP-1 cells were infected with the HIV-1 GFP reporter virus at an MOI of ∼1 in the absence or presence of an RT inhibitor, nevirapine (Nev), at 5 μM. Virus infection was assessed by counting the GFP-positive cells at 2 days after infection (left). A fraction of the culture supernatant was used to determine the level of type I IFN using a reporter cell line that expresses luciferase under the control of the ISRE promoter (middle). The induction of IP-10 was measured by using qRT-PCR (right). The value for mock-infected cells was set as the basal level of IP-10 (right). The averages of the results from four independent experiments are shown, with error bars denoting SEM. (B) Stable gene knockdown by shRNA was validated using qRT-PCR. Expression of the GAPDH gene was quantified in parallel as a control (Ctrl) to determine the changes of mRNA levels using the 2^−ΔΔCT method. Cells stably expressing shRNA directed against cGAS were harvested on three different occasions for mRNA extraction and qRT-PCR (left). The effects of cGAS depletion on IFN production by HIV-1 were determined as described above. In these experiments, increasing amounts of virus input were used to assess the relationship between viral infectivity and the ability to stimulate type I IFN production. A representative result from two experiments is shown (middle). The results shown in the right panel, where IFN production was normalized to the number of GFP-positive cells (i.e., virus-infected cells), were compiled from data generated in four independent experiments. Differences between the various conditions were statistically examined using the two-tailed, unpaired t test. (C) The R143A CA mutant virus was compared to the WT virus for their ability to induce type I IFN in THP-1 cells. THP-1 cells were infected with a GFP reporter virus carrying either the WT capsid or the capsid with the R143A mutation. Increasing amounts of virus input were used to achieve viral infectivity ranging from 10% to 60% GFP-positive, virus-infected cells. The results shown were compiled from four independent experiments. The amount of type I IFN was determined by using a bioassay that utilizes a reporter cell line stably encoding the luciferase gene under the control of the promoter of ISRE. (D) Shown is the amount of IFN in the culture supernatant normalized by the number of virus-infected cells. Since the WT virus did not induce IFN production at a lower range of virus infection (less than 30% GFP-positive cells), data points were separated into two groups. Each group was analyzed by the two-tailed, unpaired t test. (E) Induction of IP-10 mRNA was quantified using qRT-PCR. The copy numbers of IP-10 were quantified by qRT-PCR, and the values were normalized to those obtained in mock-infected cells. The results shown are the averages from two independent experiments, with error bars denoting SEM. The data were analyzed with the two-tailed, unpaired t test.

We next tested the ability of the R143A mutant to activate the cGAS-dependent DNA-sensing pathway. In contrast to infection with the WT virus, infection of THP-1 cells with the R143A mutant did not induce a strong type I IFN response (Fig. 9C). Comparison of the data points for infection resulting in more than 30% GFP-positive cells showed a statistically significant 3-log difference in IFN induction between these two viruses when the amount of type I IFN was normalized to the number of virus-infected cells (Fig. 9D). A very similar result was obtained when the levels of IP-10 expression were measured by quantitative reverse transcription-PCR (qRT-PCR); compared to mock-infected cells, a 20-fold increase in IP-10 mRNA induction was observed for cells infected with the WT virus, but there was no increase in IP-10 expression for cells infected with the R143A mutant virus (Fig. 9E).

The R143A mutant virus has been reported to be attenuated in infectivity (11, 74). In our hands, we found a 7- to 16-fold decrease in the quantity of virus particles produced by the R143A mutant, as determined by quantification of RT, and concomitant reduced infectivity, although when normalized by virus input, the infectivity of the R143A mutant virus was comparable to that of the WT virus (Fig. 10A). Thus, to achieve equivalent levels of infection with these two viruses, a larger inoculum size was required for the R143A virus than for the WT virus. To rule out the possibility that the difference in the inoculum size inadvertently affected innate immune responses, we utilized an LAI-based chimeric virus carrying CH040 CA, as its virus yield and infectivity were equivalent to those of the parental LAI virus (Fig. 10B). These observations were confirmed in a spreading infection assay (Fig. 10C). We infected THP-1 cells with a GFP reporter virus that encodes the capsid from the CH040 strain (Fig. 11). Similar to the observations made with the R143A mutant virus, the capsid of CH040 almost completely eliminated the ability of HIV-1 to induce the production of type I IFN and the expression of IP-10 in THP-1 cells (Fig. 11A). This observation was confirmed at a broad range of multiplicities of infection (MOI) (Fig. 11B). A dose-dependent increase in the production of type I IFN (Fig. 11B) was observed for the LAI strain but not the virus encoding CA from CH040. The differences between these two viruses were statistically significant even when data points for the low MOI were not excluded (Fig. 11B). The observed difference between the LAI and CH040 capsids (Fig. 11) was not due to the delayed type I IFN production by CH040 (Fig. 11C). The kinetics and extent of viral DNA synthesis by the CH040 virus were indistinguishable from those by the WT LAI virus (data not shown). Finally, a difference in the level of type I IFN production between the LAI and CH040 viruses was also observed in primary monocyte-derived macrophages (MDMs) (Fig. 12).

FIG 10.

The infectivity and replicative capacity of PF74-resistant viruses. Virion-associated RT amount (left), infectivity (center), and infectivity normalized by the RT amount (right) of the R143A mutant (A) or chimeric virus encoding the CH040 CA (B) were compared to those for the LAI WT virus. Infectious units (IU) were quantified based on the number of virus-infected cells, which were judged by GFP expression. One representative result from two independent experiments of transfection for virus production is shown. Each transfection generated three replicates of virus stock. The mean for triplicate values is shown, with error bars denoting standard deviations. (C) Spreading infection was examined by monitoring the number of GFP-expressing cells. Changes in the percentage of GFP-positive cells were plotted for replicates that showed a range of 0.3% to 1% GFP-expressing cells at 2 dpi. The results for at least six replicates are shown, with error bars denoting SEM.

FIG 11.

The CH040 capsid attenuates innate immune responses upon HIV-1 infection of THP-1 cells. (A) THP-1 cells were infected with a GFP reporter virus. The number of virus-infected cells was determined by counting the GFP-positive cells (left). IFN induction by mock-infected or virus-infected cells was examined by transferring culture supernatant onto HEK293 cell-based reporter cells that encode the luciferase gene under the control of the ISRE promoter. A standard curve was generated with known amounts of type I IFN and used for determining the amount of type I IFN for each condition (middle). IP-10 induction was quantitated using mRNA extracted from THP-1 cells 1 day after infection. The number of IP-10 RNA copies in the mock-infected sample was used as the basal level of IP-10 to calculate the magnitude of induction (shown as fold induction). Shown is the mean ± SEM (n = 2). Differences between LAI and CH040 were examined by the two-tailed, unpaired t test. (B) A dose-dependent increase in IFN production by LAI but not by CH040. The concentration of IFN in the culture supernatant was log₁₀ transformed and plotted against the number of GFP-positive cells (left). Linear regression was used to determine the correlation between the number of virus-infected cells and IFN production. Results from three independent experiments are shown, with each dot representing a single well containing THP-1 cells infected with different amounts of challenge virus. All the results were combined, and the difference between two viruses was examined using the two-tailed, unpaired t test (right). (C) A time course experiment for production of type I IFN. THP-1 cells were infected with VSV-G-pseudotyped GFP reporter virus encoding CA from LAI or CH040 using the same amount of virus input. At the indicated time points, culture medium was harvested from virus-infected cells, as well as mock-infected cells (mock), and used for quantification of type I interferon using the same bioassay. One representative result from two independent experiments is shown. The average for triplicate samples is used, with error bars denoting standard deviations.

FIG 12.

The CH040 capsid reduces the level of type I IFN production from primary macrophages infected with HIV-1. The production of type I IFN by MDMs infected with GFP reporter viruses is shown. MOIs were predetermined using THP-1 cells. A bioassay was used to quantify the amount of type I IFN in the culture medium from MDMs prepared from 2 blood donors. The mean from five independent experiments is shown, with error bars denoting SEM. P values were determined by the Mann-Whitney U test. **, P < 0.01; *, P < 0.05.

We showed that an amino acid substitution at position 216 in CA alters PF74 sensitivity (Fig. 5). We next asked whether the same position in CA also regulates innate immune activation. Indeed, this was the case (Fig. 13). Addition of the T216I substitution eliminated the ability of the WT LAI virus to induce an appreciated level of type I IFN production, whereas the reciprocal substitution (I216T) allowed the chimeric virus carrying the CH040 capsid to efficiently produce type I IFN (Fig. 13). Thus, the capsid from the CH040 strain was strikingly distinct from the one from the LAI strain in PF74 sensitivity and innate immune recognition of viral DNA, while both differences appeared to be regulated by the same amino acid substitution.

FIG 13.

Amino acids at position 216 in CA regulate HIV-induced innate activation of THP-1 cells. (A) THP-1 cells were infected with GFP-encoding reporter virus based on the LAI strain or the chimeric virus carrying CH040 Gag, with each strain encoding either a threonine or an isoleucine at position 216 in CA. The concentrations of IFN in the culture supernatant were measured using a reporter cell line together with known amounts of IFN for generating standard curves. Representative results from three independent experiments are shown. (B) Results were compiled from three experiments. IFN concentrations (units per milliliter) were normalized to the numbers of virus-infected cells. Data represent the mean ± SEM (n = 18) from 3 independent experiments with 6 different infectious doses. ****, P < 0.00001 (two-tailed, unpaired t test).

The unique ability of the CH040 strain to more effectively prevent cGAS recognition of viral DNA than a lab-adapted strain raised the possibility that this phenotype may be preferentially selected during HIV-1 transmission. Such a phenotype would be beneficial for HIV-1, which is susceptible to antiviral activity triggered by type I IFN and executed by a large repertoire of ISGs (50, 75, 76). We tested this possibility by studying the ability of a panel of eight additional T/F viruses to induce type I IFN in THP-1 cells. To specifically address the role of CA in our experimental setting, part of the gag gene encompassing the entire CA-coding sequence was transferred to an env-deficient GFP reporter clone to produce VSV-G-pseudotyped virus. Infection of THP-1 cells with increasing amounts of virus input showed that chimeric viruses, except for the one containing the capsid from the above-described CH040 strain, were able to induce the production of type I IFN at a level comparable to that by the LAI strain (Fig. 14). Note that none of these strains, except for CH040, carries the T216I substitution. Thus, HIV-1 strains harboring a capsid capable of effectively evading cytosolic DNA sensors in THP-1 cells were rare among the panel of T/F viruses examined in this work.

FIG 14.

A profile of type I IFN induction of THP-1 cells by HIV-1 reporter virus harboring different capsids from nine T/F strains. The effects of different CA sequences on type I IFN production were examined using a panel of LAI-based chimeric viruses in which part of the Gag-coding sequence, including the entire CA domain, was replaced with that from T/F strains. Activation of the type I IFN pathway in THP-1 cells infected with VSV-G-pseudotyped viruses was examined as described in the legend to Fig. 9. Results obtained in two independent experiments are shown, except for CH040, which was examined three times.

DISCUSSION

The present work used both cell-free and cell-based assays to reveal a novel phenotype of the viral capsid. A unique profile of resistance to high doses of PF74 in an infectivity assay was associated with reduced sensitivity to PF74-induced core destabilization in an in vitro assay. The capsids of PF74-resistant viruses allowed the more effective escape of viral DNA from cGAS recognition than those of other strains. Phenotypic variations in PF74 resistance and innate sensing between the T/F strains examined in this work were linked to a genetic polymorphism at the trimeric interface between CA hexamers, which mediates key interactions that influence capsid stability. Finally, imaging-based assays revealed that these unique PF74-resistant variants contain stable cores. These observations suggest that capsid stabilization is an underlying mechanism of the new phenotype, which couples capsid stability to HIV-1 DNA sensing.

The three new PF74-resistant viruses identified in this work consisted of two CA mutants (the R143A and T216I mutants) and a naturally occurring strain (CH040). Their dose-response curves for PF74 were different from those of previously described resistant variants (58, 64, 77–79), pointing to a novel resistance mechanism. Notably, these new resistant variants counteracted the inhibitory effects of PF74 rather specifically at high PF74 concentrations. This is in contrast to the drug resistance caused by blocking capsid binding to the two well-characterized host factors CPSF6 and CypA, in which PF74 antiviral activity was neutralized at low drug doses (64, 77). These observations suggest that the resistance to high doses of PF74 is mechanistically unique and not caused by differential capsid binding to these cellular proteins. It remains unclear whether any other extrinsic host factors are involved in the unique resistance profile of these drug-resistant viruses.

High doses of PF74 have been shown to block reverse transcription in studies described in the literature, but their effects on the capsid vary between studies (26, 27, 54, 55, 57–64). PF74 was shown to stabilize or destabilize the capsid, depending on the method used. A possible explanation to reconcile these seemingly conflicting observations was proposed in a recent work in which high doses of PF74 had two distinct effects: facilitating capsid opening while preventing lattice disassembly (80). It seems reasonable to assume that the newly identified PF74-resistant viruses counteract one or both effects. Given that these resistant viruses contain cores with elevated stability and delayed uncoating (Fig. 7 and 8), it is likely that they counteract the capsid-opening activity of PF74.

In a centrifugation-based assay, PF74 displayed core-destabilizing activity (Fig. 2). This core-destabilizing activity was partially neutralized by new PF74-resistant viruses. The neutralizing effects were modest but statistically significant (Fig. 3, 4, and 6). This finding appears to contradict a reported ability of PF74 to prevent capsid disassembly (27, 54, 62, 80). A possible explanation for this disagreement is a dilution effect, wherein a host factor(s) that stabilizes the capsid is reduced or lost during core isolation, which would lead to capsid destabilization. In this scenario, PF74 triggers initial capsid opening and thereby exacerbates capsid destabilization. We note that the fate-of-capsid assay, another method that found PF74-mediated capsid destabilization, also involves a procedure in which subviral complexes are diluted. Further studies are needed to confirm this idea.

The critical role played by cGAS in cytosolic DNA sensing has been firmly established (81). This protein appears to be essential for innate recognition of HIV-1 DNA in multiple experimental systems, while other host proteins participate in this sensing pathway (45, 48, 49, 51, 52, 82–85). In this work, we asked if a new PF74 resistance phenotype, which appears to be associated with capsid stabilization, influences innate sensing of HIV-1 DNA. To this end, we used monocytic THP-1 cells, which were shown to activate IFN signaling upon HIV-1 infection through cGAS-dependent recognition of viral DNA (45, 48, 82). Our results showed that genetic changes in CA almost completely eliminated the ability of HIV-1 to induce type I IFN in THP-1 cells. Thus, these viruses are distinct from previously described CA mutants, which displayed elevated levels of innate immune activation in myeloid cells (51, 52, 86, 87). Our observations were made with virus whose infectivity was comparable to that of a widely used lab-adapted strain and thus argue against the possibility that differences in the inoculum size may inadvertently affect innate responses. Overall, our results support a general model for capsid-mediated regulation of HIV-1 DNA sensing (51, 52, 88) and are in line with previous work demonstrating a similar role of capsid stability in regulating DNA sensing for murine leukemia virus (89).

HIV-1 uncoating can influence the access to viral DNA by the cytosolic sensor cGAS (51, 52). The precise nature of uncoating remains controversial, but it appears that a certain quantity of CA molecules remains associated with the cytoplasmic RTCs (reviewed in references 5 to 7). The simplest scenario is that stable cores disassemble slowly, such that the excess CA molecules that remain bound to subviral complexes limit the exposure of viral DNA to cytoplasm DNA sensors. This model accords with the observation that PF74-resistant viruses appear to have stable cores and display delayed kinetics of disassembly in cell-based assays (Fig. 7 and 8). However, it is also possible that these PF74-resistant variants utilize distinct innate immune evasion mechanisms through different capsid functions.

Phenotypic changes in sensitivity to the antiviral activity of high PF74 doses were perfectly matched to those in type I IFN induction in THP-1 cells. We envision that these two phenotypic outcomes can be used as indirect indicators for capsid stability. Infection-based approaches are amenable to large-scale studies as primary screening assays. A few caveats inherent to these experimental systems include a lack of the applicability of the PF74-based assay to viruses that bind to PF74 with a reduced affinity. Additionally, normalization in THP-1 cells would be difficult for viruses that impair viral DNA synthesis. Nonetheless, we expect that these new tools will complement existing assays for capsid stability and disassembly (53, 54, 56, 90–92).

The CH040 strain was more effective than other strains in the ability to evade cGAS-mediated sensing of viral DNA in the assay system used in this work. This property, which seems beneficial for virus propagation, was rare and found only in one of nine strains, although the number of T/F viruses assessed in this work is limited. One account to explain the rarity of such a phenotype is that other evading mechanisms, such as degradation of viral DNA by TREX1 (47, 88) and shielding of viral DNA by cofactor recruitment (51, 52), may be sufficient to prevent HIV-1 DNA sensing under physiologically relevant conditions.

In this work, we identified one naturally occurring HIV-1 variant (CH040) with a capsid that is phenotypically distinct from the capsids in the other tested strains. As described above, our results suggest that cores from this variant are more stable than those from the others. Interestingly, the CH040 strain does not have any obvious defect in infectivity or replicative capacity (68, 76). Therefore, it appears that stable cores do not necessarily impair viral fitness in vivo, although this likely depends on the degree of stability. This observation was somewhat unexpected, because the conventional wisdom holds that optimal capsid stability is fine-tuned. However, recent work has shown that certain HIV-1 CA mutants with hyperstable capsids can replicate efficiently in T cell lines (58, 70, 78). It should also be noted that among a panel of HIV-1 CA mutants that exhibited abnormal capsid stability, the R143A mutant was unique in its ability to replicate in primary T cells (11).

Among the ten T/F viruses examined in this study, only the CH040 strain had a distinct capsid. However, this phenotype may not be unique in nature. The T216I substitution responsible for the capsid phenotype of CH040 is present in 0.15% of 9,056 HIV-1 group M strains in the Los Alamos HIV-1 databases. Furthermore, the amino acids at position 216 are not a threonine in 1.5% of the predicted CA sequences. Whether amino acid residues other than an isoleucine at position 216 in CA change the capsid function requires further investigation. It will also be interesting to determine whether the genetic polymorphism at position 216 in CA affects clinical outcomes, including CD4⁺ T cell decline and immune activation levels, thereby influencing HIV-1 pathogenicity.

Our study adds to a growing body of evidence for phenotypic variations in postentry steps mediated by the HIV-1 capsid. It has been shown that deviations from the normal course of uncoating by artificial perturbations often do not show strong impacts on viral infectivity (92–97). HIV-1 CA can alter nuclear entry pathways and specific preferences for integration site selection without critically modulating viral replicative capacity (31, 98). These observations suggest that the viral capsid, which is known to be genetically fragile (99), is unexpectedly tolerant of phenotypic changes during early steps of the viral replication cycle. This may be relevant to therapeutic strategies for CA-targeting antivirals, an avenue that has been actively pursued as a next-generation drug, as the malleability of the viral capsid may govern the capacity of HIV-1 to acquire resistance to such drugs.

MATERIALS AND METHODS

Plasmid DNA.

Env-defective reporter virus clones encoding GFP or luciferase (Luc) in place of the nef open reading frame are based on the LAI strain of HIV-1. The CA mutants (E45A, R143A, and Q219A mutants) have been described previously (100). A BssHII-ApaI fragment containing the entire capsid-encoding segment was PCR amplified using a panel of full-length T/F HIV-1 infectious molecular clones obtained through the NIH AIDS Reagent Program as the template and cloned into the LAI-based molecular infectious clone (101). The following molecular clones for T/F viruses were used: RHPA.c/2635, WITO.c/2474, CH040.c/2625, CH058.c/2960, CH077.t/2627, CH106.c/2633, THRO.c/2626, REJO.c2864, RTRJO.c2851, and SUMA.c/2821 (68). Other point mutations were introduced into the GFP reporter clones by using standard cloning procedures. Plasmid DNAs for expression of HIV-1 Gag-Pol (pCRV1-Gag-Pol), Vpr-INmNeonGreen fusion protein, and VSV-G glycoprotein (pHCMV-G) have been described previously (71, 102, 103). CypA-DsRed Express2 (CypA-DsRed2) was cloned by replacing the DsRed in the CypA-DsRed plasmid (54) with the brighter and more photostable version DsRed Express2 by molecular cloning using BamHI and NotI sites. Gene depletion vectors were generated based on the pLKO.1-TRC control plasmid DNA (plasmid number 10879; Addgene) by cloning short hairpin RNA (shRNA)-coding oligonucleotides into the EcoRI and AgeI sites. The targeted sequences, which were described previously (45), were as follows (only sense strands are listed here): for luciferase, 5′-AACTTACGCTGAGTACTTCGA-3′, and for cGAS, 5′-GGAAGGAAATGGTTTCCAA-3′.

Cell culture.

Adherent cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Corning) supplemented with 10% fetal bovine serum (FBS; Sigma) and 1× penicillin-streptomycin (P-S; Corning). THP-1 and MT4 cell lines were cultured in RPMI 1640 medium (Corning) supplemented with 10% FBS, 1× P-S, and 2 mM l-glutamine (Corning). The following reporter cell lines were described previously: TZM-bl (NIH AIDS Reagent Program) and HEK293-ISRE-Luc (a gift from Xuguang Li) (104). To generate THP-1 cell lines stably expressing shRNA targeting luciferase or cGAS, THP-1 cells were plated at a half million cells per well in 24-well plates and infected with 0.3 ml of 10-fold-concentrated virus. Puromycin (Invitrogen) was added at 0.5 μg per ml to the infected cells at 2 days after infection. Puromycin-resistant cells were expanded for 2 weeks before using infection and mRNA extraction. Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood from anonymous blood donors (New York Blood Center) by density gradient centrifugation. CD14-positive monocytes were purified from PBMCs using an EasySep human monocyte isolation kit (Stemcell). CD14⁺ monocytes were differentiated for 6 days in RPMI 1640 supplemented with 10% FBS, 1× P-S, 2 mM l-glutamine, and 100 ng per ml of granulocyte-macrophage colony-stimulating factor (PeproTech) or macrophage colony-stimulating factor (PeproTech) to generate MDMs.

Viruses.

Virus stocks were generated by transient transfection of HEK293T (293T) cells using polyethylenimine (PolySciences). Virus stocks generated with env-deficient infectious clones used for infection assays were pseudotyped with the VSV-G protein by cotransfection with the plasmid pHCMV-G. Virus stocks for the core isolation assay were made without the VSV-G expression vector. Lentivirus vectors encoding shRNA targeting cellular genes were cotransfected with the Gag-Pol expression vector pCRV1-Gag-Pol into 5 million 293T cells in a 100-mm tissue plate, and virus was concentrated with polyethylene glycol (PEG; Sigma). Briefly, 8 ml of filtered supernatant was mixed with 2 ml of 30% (wt/vol) PEG in 1× phosphate-buffered saline (PBS) and stored at 4°C overnight. Virus was pelleted by centrifuging the mixture at 1,200 × g for at least 45 min at 4°C and resuspended with 0.8 ml of DMEM with 10% FBS. For generating high-titer virus stocks, several 100-mm plates were seeded with 5 million 293T cells per plate. Culture supernatant harvested at 2 days after transfection was filtrated with a Steriflip filter unit (pore size, 0.45 μm; polyvinylidene difluoride; Millipore), carefully layered onto 5 ml of a 20% (wt/vol) sucrose solution made with 1× PBS in a Beckman polyallomer tube (Beckman Coulter), and spun down for 2 h at 24,000 rpm and 4°C using an SW28 rotor. Virus pellets were resuspended with 3.2 ml of culture medium, and aliquots were stored at −80°C until use.

Infection and replication.

Adherent cells were plated at 5,000 cells per well in 96-well plates. Infection was enhanced by addition of DEAE-dextran (20 μg/ml) and spinoculation (1,200 × g for 30 min). Suspension cells were plated at 30,000 cells per well in 96-well plates. MDMs were plated for infection at 5 × 10⁵ cells per ml in 96-well plates. For infection experiments that examined the sensitivity of VSV-G-pseudotyped viruses to PF74 (Sigma and Aobious), cells were infected at an MOI of 0.1. To quantify the induction of type I IFN, THP-1 cells or MDMs were infected with various amounts of inoculum. Culture supernatant was harvested at 1 day after infection for IFN bioassay (see below). The number of GFP-positive cells was counted at 2 or 3 days after infection using an LSR II flow cytometer (BD Biosciences) or a Guava easyCyte flow cytometer (Millipore). Virus-infected TZM-bl cells were lysed with 25 μl of 1× luciferase cell culture lysis reagent (Promega) for 5 min, and one-fifth of the cell lysates were mixed with 25 μl of the luciferase assay reagent (Promega) before measuring luciferase activity on a luminometer. Nevirapine (NIH AIDS Reagent Program), raltegravir (Aobious), and dolutegravir (Aobious) were used at 50 μM, 10 μM, and 10 μM, respectively. In the spreading infection assay, MT4 cells were used as target cells together with replication-competent GFP reporter viruses. Virus stocks generated with intact provirus clones carrying different CA-encoding sequences were used to infect 20,000 MT4 cells per well in a 96-well plate with virus inputs that yielded approximately 1% GFP-positive cells at 2 days postinfection (dpi). A fraction of the virus-infected cells was harvested from 2 to 4 dpi for counting the number of GFP-expressing cells by flow cytometry.

PF74 binding.

The affinity of PF74 for recombinant WT and R143A CA hexamers was determined as described previously (64). Briefly, CA hexamers of the WT virus and R143A mutant (containing a substitution of arginine to alanine at position 143 in CA) were generated by in vitro assembly of purified recombinant CA carrying four amino acid substitutions that stabilize the hexamer through disulfide bond formation (105). [³H]PF74, produced from triiodo-PF74 (78), was used in equilibrium dialysis. Various concentrations (0.10 to 1.0 μM) of [³H]PF74 in 500 μl solution were added in duplicate into the buffer chamber, and 300 μl of 1.0 μM CA hexameric protein was added into the sample chamber of the rapid equilibrium dialysis plate (Thermo Fisher Scientific). The plate was rotated at 100 rpm during the 24 h of incubation at 37°C. The ³H in each sample removed from each side of the chamber was quantified by liquid scintillation counting in a TopCount (PerkinElmer) scintillation counter. The [³H]PF74 concentrations were determined using a reference sample containing a known quantity of [³H]PF74. Calculation of dissociation constant (K_d) values was performed using a one-site binding model in GraphPad Prism software (GraphPad Software).

Isolation of HIV-1 cores.

The effects of PF74 on HIV-1 cores were assessed using a previously described procedure (11, 53, 58) with several modifications. Virus particles pretreated with or without 10 μM PF74 at 37°C for 1 h were subjected to the spin-through method to isolate cores. A sucrose gradient tube consists of a layer of 15% sucrose containing 0.5% Triton X-100 (Sigma), which is placed between a top barrier fraction of 7.5% sucrose and a linear discontinuous 30 to 70% sucrose gradient (made with 30, 40, 50, 60, and 70% sucrose). Sucrose solutions were made with STE buffer (10 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM EDTA). Pretreated virion particles were carefully loaded on the top of the gradient and spun using ultracentrifugation at 100,000 × g for 16 to 18 h at 4°C. After the centrifugation, fractions of 700 μl were harvested from the top of the gradient, and the pellet was resuspended with 700 μl STE buffer. The density of the fractions was measured by weighing 100 μl of each fraction (except for the pellet fraction) using 1.7-ml microcentrifuge tubes.

RT activity assay.

RT activity in fractions harvested after ultracentrifugation was quantitated using a SYBR green I-based product-enhanced RT (SG-PERT) assay with several modifications (106). Five μmicroliters of each fraction was lysed for 10 min at room temperature with an equal amount of 2× lysis buffer (0.25% Triton X-100, 50 mM KCl, 100 mM Tris-HCl, pH 7.4, 40% glycerol), which was supplemented with 0.4 units of RiboLock RNase inhibitor (Thermo Fisher Scientific) per μl before use. The lysed materials were diluted 10-fold before use in reverse transcription and quantitative PCR on an Applied Biosystems 7500 fast real-time PCR system (Thermo Fisher Scientific). Ten microliters of the samples was mixed with equal amounts of 2× reaction buffer. The composition of the 2× reaction buffer was the following: 5 mM (NH₄)₂SO₄, 20 mM KCl, 20 mM Tris-Cl (pH 8.3), 10 mM MgCl₂, 0.2 mg/ml bovine serum albumin (BSA; New England BioLabs), 1× SYBR green I nucleic acid gel stain (Thermo Fisher Scientific), 400 μM deoxynucleoside triphosphates (Bioline), 1 μM forward primer, 1 μM reverse primer, 7 nM MS2 RNA (Sigma), 50 nM carboxy-X-rhodamine reference dye (Thermo Fisher Scientific), 40 units per ml of RiboLock RNase inhibitor, and 25 units per ml of Hot Diamond Taq DNA polymerase (Eurogentec). The following primers were used: MS2fwd (5′-TCCTGCTCAACTTCCTGTCGAG-3′) and MS2rev (5′-CACAGGTCAAACCTCCTAGGAATG-3′). The real-time PCR conditions were as follows: 42°C for 20 min, 95°C for 2 min, and 40 cycles of 95°C for 5 s, 60°C for 5 s, 72°C for 15 s, and 80°C for 7 s. The PCR was run with serially diluted samples of known RT amount, which were used to generate a standard curve for absolute quantification.

p24 CA ELISA.

An antigen-capture enzyme-linked immunosorbent assay (ELISA) was used to quantify the amount of p24 CA in the fractions after the spin-through procedure using a published protocol with minor modifications (53, 107). Monoclonal anti-HIV-1 p24 antibody (monoclonal antibody 183-H12-5C; NIH AIDS Reagent Program) was used to coat Immulon 2HB 96-well plates (Thermo Fisher Scientific) at 1.3 μg/ml in PBS and 37°C overnight. The coated plate was blocked for 1 h with 0.25 ml of blocking solution, which consists of 5% milk and 0.5% BSA (Sigma) in PBS. Samples were added to the plates together with serial dilutions of a standard sample that contained a predetermined amount of p24 CA. Several dilutions were made from the original fraction so that the values for the p24 amount fell within a linear range of the standard curve. After 2 h of incubation, hyperimmune human patient serum (HIV immunoglobulin; NIH AIDS Research and Reference Reagent Program) was diluted 1:20,000 and added before 1 h of incubation. Peroxidase-conjugated goat anti-human immunoglobulin G (Pierce) was diluted at 20 pg per ml in a PBS-based ELISA sample diluent buffer containing 10% FBS and 0.5% Triton X-100 (Sigma). The plates were washed six times after each process with 0.2% Tween 20 (Sigma) in PBS, except for the washing step before blocking, where PBS was used. After the final wash, the enzymatic activity of the peroxidase was detected by addition of 3,3′,5,5′-tetramethylbenzidine (TMB) solution using a TMB microwell peroxidase substrate kit (KPL Inc). The reaction was stopped by adding an equal volume of 0.5 M H₂SO₄ (EMD Millipore). The optical density at 450 nm was measured on a microplate reader. A standard curve generated with serially diluted virus stocks of known p24 amount was used to interpolate the concentration of the samples.

In vitro imaging assay for capsid stability.

The in vitro single-virus uncoating assay was performed as described previously (54, 71). Briefly, viruses bearing CA from WT or mutant strains were labeled with INmNeonGreen (INmNG) expressed in producer cells. Viruses were bound on a cover glass, and the viral membrane was permeabilized with 100 μg/ml of saponin for 30 s to expose the cores. The viral cores were then treated with 0.1 mg/ml of cytosolic extract from 293T cells containing 200 nM CypA-DsRed for 90 s to allow CypA-DsRed binding to the cores. After removal of the cytosolic extract, CypA-DsRed-bound cores were washed once with PBS to remove any unbound CypA-DsRed, and cores colabeled with INmNG and CypA-DsRed were imaged at room temperature. Images were taken from four fields of view every 30 s. CsA (5 μM) was added at 30 min postpermeabilization to displace CypA-DsRed from cores that had not completed uncoating. The number of INmNG and CypA-DsRed viral punctae was calculated using a spot detector in the Icy image analysis platform (http://icy.bioimageanalysis.org/).

Live-cell imaging of postfusion cores.

A live cell-based assay for capsid uncoating was performed as previously described (54, 71). VSV-G-pseudotyped virus particles were colabeled with INmNeonGreen and CypA-DsRed in 293T producer cells. Virus inoculum containing 10 pg of p24 was allowed to bind to 5 × 10⁵ TZM-bl cells (equivalent to an MOI of 0.008) by spinoculation at 1,500 × g at 12°C for 30 min. Prior to virus binding, cell nuclei were stained for 10 min with 2 μg per ml of Hoechst 33342 dye. The cells were washed twice, and virus entry was synchronously initiated on a temperature- and CO₂-controlled microscope stage by adding prewarmed FluoroBrite (Gibco) imaging medium supplemented with 10% FBS. Imaging was performed for a period of 2 h with 20-s time-lapse acquisition of the whole-cell volume. At 90 min postinfection, CsA (10 μM) was added and imaging was continued. Addition of CsA displaces CypA-DsRed from postfusion cores that have not yet completed uncoating, without affecting the CypA-DsRed signal from intact viruses residing in the endosome. The fraction of postfusion cores per cell was calculated by dividing the number of cytosolic cores that lost CypA-DsRed upon CsA application at 90 min postinfection by the total number of CypA-DsRed-positive cores at 0 min.

Capsid permeabilization assay.

293T cells (ATCC) were transfected with (i) pBru3ori-ΔEnv-luc2, (ii) pVpr-mRuby3-IN (72), and (iii) pCMV-VSV-G plasmids using the Lipofectamine 2000 reagent (Thermo Fisher Scientific) in the presence of 5-ethynyl uridine (EU), as previously described (55, 73). After 48 h, the cell supernatant was collected and filtered through a 0.45-μm-pore-size polyethersulfone syringe filter (Millipore). Excess EU was removed, and virus was concentrated via a Lenti-X concentrator (Clontech). HeLa cells were synchronously infected with EU-labeled HIV-1 (20 ng of p24, as determined by ELISA; XpressBio). Cells were fixed at 15 to 90 min postinfection, permeabilized, stained for viral RNA (catalog number EU-AF647 Click-iT; Invitrogen) and cell nuclei (Hoechst 33342), and mounted with coverslips. Confocal imaging was performed on a Nikon A1 laser scanning confocal microscope, with 7 to 9 z-stacks (0.5-μm spacing) taken per sample. RNA punctae and mRuby3-IN-labeled particles were enumerated via Imaris software. The average number of RNA punctae in the uninfected negative-control samples for each experiment was subtracted from each imaging field. Grubbs’ extreme Studentized deviate test was used to identify and exclude statistical outliers. The capsid permeabilization kinetics were analyzed by multiple linear regression modeling using SAS software to test if the slopes were equivalent. Using the F statistic with the Bonferroni correction for pairwise comparisons, P values of less than 0.0167 were considered statistically significant.

IFN bioassay.

We examined the production of type I IFN by using the reporter cell line HEK293-ISRE-Luc (104). This reporter cell line carries an IFN-stimulated response element (ISRE) within the promoter region driving the expression of luciferase. One day prior to use, HEK293-ISRE-Luc cells were plated on a 96-well flat-bottom plate at 30,000 cells per well with 100 μl of cell suspension. At 1 day after infection of THP-1 cells or MDMs, 50 μl of culture supernatant was harvested and transferred into a 96-well flat-bottom plate containing HEK293-ISRE-Luc cells. In each experiment, the IFN standard curve was generated using a stock solution of recombinant interferon alphaA/D (Sigma). A working solution of IFN at 10,000 units per ml was used to generate 10-fold serial dilutions. These standard dilutions were added at 10 μl per well in duplicate. The cells were placed in a tissue culture incubator after addition of supernatant samples and IFN standards. On the following day, the cells were lysed with 20 μl of 1× buffer prepared from luciferase cell culture lysis 5× reagent (Promega) for 5 min at 37°C. Five microliters of cell lysates was transferred to an opaque 96-well plate, and 25 μl of the luciferase assay reagent (Promega) was added, followed by reading the relative luciferase units in a luminometer.

Quantification of cellular RNA by qRT-PCR.

The knockdown efficiency of mRNA encoding cGAS was assessed using quantitative reverse transcription-PCR (qRT-PCR). Total RNA was isolated from cell pellets collected on three different dates using the TRIzol reagent (Thermo Fisher Scientific), based on the manufacturer’s protocol. The extracted RNA was used for cDNA synthesis with random hexamers and a high-capacity cDNA reverse transcription kit (Applied Biosystems). The expression levels of the respective genes were determined by a SYBR green-based quantitative PCR using gene-specific primers and normalized to GAPDH (glyceraldehyde-3-phosphate dehydrogenase) mRNA levels using the 2^−ΔΔCT threshold cycle (C_T) method. The primers used were hcGAS-F (5′-CCTGCTGTAACACTTCTTAT-3′), hcGAS-R (5′-TTAGTCGTAGTTGCTTCCT-3′), GAPDH-F (5′-GGCTGAGAACGGGAAGCTT-3′), and GAPDH-R (5′-AGGGATCTCGCTCCTGGAA-3′).

Expression of C-X-C motif chemokine ligand 10 (CXCL10; also known as IP-10) was assessed by a TaqMan-based real-time PCR. RNA was extracted from virus-infected THP-1 cells at 1 day after infection by using a NucleoSpin 8 RNA kit (Macherey-Nagel) and used for cDNA synthesis as described above. The cDNA sample was used for real-time PCR with TaqMan universal master mix II (Applied Biosystems). The following primers were used as the forward and reverse primers, respectively: IP10f (5′-TGAAATTATTCCTGCAAGCCAATT-3′) and IP10r (5′-CAGACATCTCTTCTCACCCTTCTTT-3′). The sequence of the TaqMan probe (IP10p) was as follows: 5′-FAM-TGTCCACGTGTTGAGATCATTGCTACAATG-TAMRA-3′ (where FAM is 6-carboxyfluorescein and TAMRA is 6-carboxytetramethylrhodamine). A standard curve ranging from 9 × 10⁷ to 90 copies per reaction mixture was generated by making 10-fold serial dilutions from the plasmid carrying the amplicon (pGEM-T Easy-huIP10).

Quantitative PCR for viral DNA.

THP-1 cells were infected with virus stocks that were treated with 80 units per ml of Turbo DNase (Thermo Fisher Scientific) at 37°C for 1 h before infection. Virus-infected cells were harvested at different time points and stored at −80°C. Cell pellets were processed for DNA extraction using a NucleoSpin tissue kit (Macherey-Nagel). The concentrations of genomic DNA were measured using a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific). Seminested quantitative PCR was performed to quantify the early and late products of reverse transcription. Amplification of early RT products used primers ME31 (5′-AGACCAGATTTGAGCCTG-3′) and uoLTRr (5′-CCACACTGACTAAAAGGGTCTGA-3′) in the first PCR and primers MH531 (5′-TGTGTGCCCGTCTGTTGTGT-3′) and uoLTRr in the second PCR. PCR for the late PCR products were performed with primers MH535 (5′-AACTAGGGAACCCACTGCTTAAG-3′) and MH532 (5′-GAGTCCTGCGTCGAGAGAGC-3′) in the first reaction and primers MH531 (5′-TGTGTGCCCGTCTGTTGTGT-3′) and MH532 in the second reaction. The second PCR was performed using PowerUp SYBR green master mix (Thermo Fisher Scientific) on a 7500 Fast real-time PCR system (Applied Biosystems). A standard curve was generated using known amounts of template DNA, plasmid pLai3-GFP3 DNA, by serially diluting by 10-fold. Results were normalized to the concentration of DNA.