Escape of HIV-1 Envelope Glycoprotein from Restriction of Infection by IFITM3

Interferon-induced transmembrane protein 3 (IFITM3) is a cellular factor that reduces HIV-1 infectivity by an incompletely understood mechanism. This study aimed to elucidate the role of the HIV-1 envelope glycoprotein (Env) in determining viral susceptibility to IFITM3.

ABSTRACT

Interferon-induced transmembrane protein 3 (IFITM3) is a cellular factor that reduces HIV-1 infectivity by an incompletely understood mechanism. We show here that viruses differing only in the envelope glycoprotein (Env) expressed on their surface have different sensitivities to IFITM3. Measurements of the sensitivity of viruses to neutralizing antibodies showed that IFITM3 increased the sensitivity of IFITM3-sensitive viruses to PG16, which targets the V1V2 loop, suggesting that IFITM3 promotes exposure of the PG16 epitope of IFITM3-sensitive viruses. Exchanges of V1V2 loops between the Env proteins of sensitive and resistant viruses revealed that V1V2 and V3 act together to modulate viral sensitivity to IFITM3. Coimmunoprecipitation experiments showed that IFITM3 interacted with both the precursor (gp160) and cleaved (gp120) forms of Env from IFITM3-sensitive viruses but with only the precursor (gp160) form of Env from IFITM3-resistant viruses. This finding suggests that the interaction between the Env protein of resistant viruses and IFITM3 was inhibited once Env had been processed in the Golgi apparatus. This hypothesis was supported by immunofluorescence experiments, which showed a strong colocalization of IFITM3 with Env of sensitive viruses, but only weak colocalization with Env of resistant viruses on the plasma membrane of virus-producing cells. Together, these results indicate that IFITM3 interacts with Env, inducing conformational changes that may decrease viral infectivity. This antiviral action is, nevertheless, modulated by the nature of Env, in particular its V1V2 and V3 loops, which after maturation may be able to escape this interaction.

IMPORTANCE Interferon-induced transmembrane protein 3 (IFITM3) is a cellular factor that reduces HIV-1 infectivity by an incompletely understood mechanism. This study aimed to elucidate the role of the HIV-1 envelope glycoprotein (Env) in determining viral susceptibility to IFITM3. We found that viruses differing only in Env expressed on their surface had different sensitivities to IFITM3. By comparing the Env proteins of viruses that were highly sensitive or resistant to IFITM3, we obtained new insight in the mechanisms by which HIV-1 escapes this protein. We showed that IFITM3 interacts with the Env protein of sensitive viruses in virion-producing cells, inducing conformational changes that may decrease viral infectivity. However, this antiviral action is modulated by the nature of Env, particularly the V1V2 and V3 loops, which may be able to escape this interaction after processing in the Golgi apparatus.

INTRODUCTION

The interferon-induced transmembrane (IFITM) proteins are a family of highly related proteins with variable N- and C-terminal domains and two hydrophobic membrane domains connected by a conserved intracellular loop. In humans, this family contains five expressed proteins, three of which (IFITM1, IFITM2, and IFITM3) are strongly upregulated by type I interferons and have been shown to inhibit diverse viruses, including influenza A virus (1, 2), hepatitis C virus (3), Ebola virus (4, 5), severe acute respiratory syndrome (SARS) coronavirus (4), Middle East respiratory syndrome (MERS) coronavirus (6), dengue virus (1, 7), Zika virus (8), Marburg virus (4), West Nile virus (1, 7), Rift Valley fever virus (9), Hantaan virus (9, 10) and HIV-1 (11–13) (reviewed in references 14 and 15).

In the case of HIV-1, IFITM proteins have been shown to target the viral life cycle at three different steps. The first antiviral mechanism of IFITM proteins to be described occurs in the target cells, where IFITM proteins act as a barrier against viral invasion by inhibiting viral fusion at the plasma or endosomal membranes of the cell (11, 16–22). IFITM proteins are thought to mediate changes in the physical properties of host cell membranes, but the mechanism involved remains a matter of debate (6, 19, 23, 24). It has been suggested that IFITM proteins prevent HIV-1 entry into cells by a mechanism dependent on virus coreceptor use and, thus, on the nature of the viral envelope glycoprotein (Env), with CCR5-tropic (R5) strains being more sensitive to IFITM1 and CXCR4-tropic (X4) strains being more sensitive to IFITM2 and IFITM3 (20).

The second antiviral mechanism involving IFITMs occurs in virus-producing cells. In these cells, IFITMs, particularly IFITM3, have been shown to colocalize with the structural proteins of developing virions, resulting in their incorporation into the nascent viral particles, leading to a decrease in the infectivity of these particles in target cells (12, 13). This mechanism, initially described for HIV-1, has since been extended to other viruses (25). It remains unclear whether the decrease in the infectivity of nascent viruses is due to the physical incorporation of IFITM proteins into viral particles or to a specific action of IFITM proteins in virion-producing cells during late events in the HIV-1 life cycle. The second of these hypotheses was supported by results from a study reporting that interactions between IFITM3 and Env in virus-producing cells inhibited the proteolytic cleavage of the gp160 Env precursor, thereby decreasing the amount of mature gp120/gp41 Env incorporated into virions (26). A second study reported that the defective proteolytic processing of Env was due to an IFITM3-driven rerouting of Env trafficking to the lysosomes for degradation (27). However, the processing defect of Env remains a matter of debate, as other studies observed no change in the amount of Env incorporated into virions following the overproduction of IFITM proteins (13, 20, 25, 28). Nevertheless, several HIV-1 strains appear to be refractory to IFITM3-mediated inhibition despite its incorporation into viral particles, and this resistance phenotype has been reported to be determined by Env, particularly by the V3 loop of the gp120 surface subunit (29). Thus, Env appears to be an important determinant of sensitivity to IFITM3, but the way in which IFITM3 affects the functional properties of Env remains poorly understood.

Finally, a third antiviral mechanism by which IFITM proteins may restrict viral infection was recently discovered (30). IFITM proteins can inhibit HIV-1 protein synthesis by preferentially excluding viral mRNA transcripts from polysomes. Interestingly, the viral accessory protein Nef helps to overcome this restriction.

In this study, we investigated the disruption of Env function by IFITM3 further. We evaluated the impact of IFITM3 in virus-producing cells on the infectivity of HIV-1-pseudotyped viruses carrying various Env proteins derived from primary isolates or laboratory strains. We found that IFITM3 interacted with Env, causing conformational changes that might decrease viral infectivity. However, this antiviral action was governed by the nature of the Env protein, particularly by its V1V2 and V3 regions, which may be able to escape this interaction after maturation in the Golgi apparatus.

RESULTS

The sensitivity of Env-pseudotyped viruses to IFITM3 is dependent on Env.

We selected a panel of 11 X4, R5, and X4R5 Env clones from previous studies (31, 32). This selection included two Env clones from the laboratory strains NL4-3 (X4) and AD8 (R5) and nine Env clones (7 X5 and 2 X4R5) derived from primary isolates obtained from individuals with acute or chronic HIV-1 clade B infections (Table 1). Env-pseudotyped viruses expressing full-length Env clones were produced as previously described by transfecting HEK293T cells with an env deletion NL4-3 backbone plasmid (pNL4.3.LUC.R-E-) and an env expression vector (pCI-env) (31). We determined sensitivity to IFITM3-mediated inhibition under conditions in which this protein was present in virus-producing cells and incorporated into viral particles by ectopically overproducing this protein in HEK293T cells through cotransfection with a third plasmid encoding a Flag-tagged IFITM3 protein (pcDNA-ifitm3-Flag).

TABLE 1.

Env clones derived from primary isolates obtained from individuals with acute or chronic HIV-1 clade B infections

Patient ID	Stage of infection	Env clone	Tropism	Reference
60204	Acute	2c4	X4R5	31
750202	Acute	11c6	R5	31
750710	Acute	13c6	R5	31
330424	Acute	23c4	R5	31
770203	Acute	34c2	R5	31
840104	Acute	35c1	R5	31
590112	Acute	248-5	R5	32
590112	Chronic	249N3	R5	32
590112	Chronic	249V4	X4R5	32

The protein content of Env-pseudotyped viruses produced in the presence or absence of IFITM3 was evaluated by Western blotting after purification of viral supernatants of transfected HEK293T cells by centrifugation through a sucrose cushion (see Fig. 1A for a representative experiment). When IFITM3 was present in virus-producing cells, it was detected in the viral particles released, regardless of the Env clone used for pseudotyping, albeit at variable levels. The levels of gp160 Env precursor and gp120 mature Env subunit varied considerably between viruses, suggesting differences between viruses in their efficiencies of processing and incorporation of Env, but no major change was observed upon IFITM3 expression (Fig. 1A).

FIG 1.

Impact of IFITM3 on the infectivity of Env-pseudotyped viruses. (A) The viruses present in the cell supernatant were purified 72 h after the cotransfection of HEK293T cells with pNL4-3Δenv.luc, pCI-env, and pcDNA-ifitm3-Flag or empty pcDNA. They were concentrated by ultracentrifugation on a sucrose cushion, lysed, and examined by Western blotting. (B) Viral inputs were normalized according to reverse transcriptase activity, and infectivity was evaluated by measuring luciferase activity (RLU) 48 h after infection of 1 × 10⁴ TZM-bl cells. The impact of IFITM3 on the infectivity of each virus was evaluated by calculating the RLU ratio between IFITM3-bearing viruses and IFITM3-free (control) viruses, expressed as a percentage. The data shown are mean results obtained from at least three independent experiments performed in quadruplicate. (C) The intrinsic infectivity of viruses was evaluated by measuring RLU values 48 h after infection of 1 × 10⁴ TZM-bl cells with serial 5-fold dilutions of viral supernatants in quadruplicate. RLU values were then normalized for the supernatant RT amount. (D) Absence of correlation between relative infectivity (percentage of the control level) and intrinsic infectivity (RLU/ng RT) of viruses. The Spearman’s correlation coefficient ρ and the P value are indicated. (E) Absence of correlation between relative infectivity (percentage of the control level) and the amount of IFITM3 (results of densitometry analyses from Western blots of at least three independent experiments) incorporated into Env-pseudotyped viruses. Spearman’s correlation coefficient (ρ) and the P value are indicated.

We compared the capacities of each Env-pseudotyped virus produced in the presence and absence of IFITM3 to infect TZM-bl target cells (HeLa CD4⁺ CCR5⁺ CXCR4⁺) in a single round of infection. To that aim, IFITM3-free viruses were first normalized to 400 50% tissue culture infective doses (TCID₅₀) per milliliter. The input of IFITM3-bearing viruses was then normalized to obtain identical reverse transcriptase (RT) activities of their IFITM3-free counterparts, and infectivity was evaluated 48 h postinfection by measuring luciferase activity (relative luminescence units [RLU]). IFITM3 decreased the infectivity of seven Env-pseudotyped viruses (35c1, 34c2, 23c4, 248-5, NL4-3, 11c6, and 249N3) to 19% to 62% of their level of infectivity when produced in the absence of IFITM3. In contrast, IFITM3 did not affect the infectivities of the other four viruses (2c4, 13c6, AD8, and 249V4), which had infectivities similar to or higher than those of control viruses in the absence of IFITM3 (Fig. 1B). The observed differences in sensitivity/resistance to IFITM3 were not related to the intrinsic infectivity of the viruses (in the absence of IFITM3) (Fig. 1C) since no correlation was found between these two variables (P = 0.56, ρ = –0.02) (Fig. 1D). Similarly, no correlation was found between the levels of IFITM3 incorporation into viruses, quantified by densitometry, and their sensitivity/resistance to IFITM3 (P = 0.18, ρ = 0.44) (Fig. 1E). These findings suggest that the IFITM3 resistance/sensitivity of viruses was determined by the nature of the Env protein. However, no link could be found between this phenotype and the tropism of Env clones.

IFITM3 incorporation into viral particles increases their sensitivity to several neutralizing antibodies.

We explored the mechanism by which IFITM3 exerts its antiviral activity by measuring the effect of this protein on the sensitivities of viral particles to a panel of well-characterized human monoclonal broadly neutralizing antibodies (HuMobNAbs) targeting functional regions of Env, such as the V1V2-glycan epitope (33, 34), the V3-glycan epitope (34, 35), the CD4 binding site of gp120 (35, 36), the membrane-proximal external region (MPER) of gp41 (37–41), and the gp120/gp41 interface (42, 43). We found that IFITM3 incorporation sensitized pseudotyped viruses to the inhibitory actions of most HuMobNAbs, resulting in a statistically significant decrease in the half-maximal inhibitory concentrations (IC₅₀) for seven of these antibodies (Fig. 2). The antibodies for which sensitivity increased included PG9 and PG16, which target the V1V2-glycan epitope (Fig. 2A and B) (33, 34), PGT128 and 10-1074, directed against the V3-glycan epitope (Fig. 2D and E) (34, 35), 2F5 and 4E10, targeting MPER (Fig. 2G and H) (37–41), and 35O22, directed against the gp120/gp41 interface (Fig. 2K) (43). This effect concerned most of the epitopes tested, suggesting an overall effect of IFITM3 on the conformation of Env.

FIG 2.

Comparison of the neutralization sensitivities of Env-pseudotyped viruses produced in the presence (IFITM3) or absence (pcDNA) of IFITM3 by bNAbs targeting the V1V2-glycan epitope (A to C), the V3-glycan epitope (D and E), the CD4 binding site (F), the MPER (G to I), and the gp120/gp41 interface (K and L). The data shown are mean results obtained from three independent experiments performed in triplicate. Dot plots show the distribution of IC₅₀ (50% inhibitory concentrations, expressed in micrograms per milliliter) values for each bNAb within the two groups of viruses; the horizontal lines represent the 10th percentile, the median, and the 90th percentile. Each virus is represented by a different color, and distinct symbols were used to differentiate IFITM3-sensitive (circles) from -resistant (triangles) viruses. The significance of the observed differences was evaluated in a Wilcoxon test. ns, differences were not significant.

IFITM3-sensitive viruses are more sensitized to the inhibitory action of PG16 than IFITM3-resistant viruses, in the presence of IFITM3.

We investigated the properties differentiating IFITM3-sensitive from IFITM3-resistant viruses by determining whether the increase in sensitivity to neutralizing antibodies observed in the presence of IFITM3 was linked to viral sensitivity to IFITM3. So the impact of IFITM3 on the increase of sensitivity to neutralizing antibodies was quantified for each virus by calculating the ratio of IC₅₀ values between IFITM3-bearing viruses and IFITM3-free viruses. We then investigated whether these values correlated with the sensitivities of viruses to IFITM3. We found a positive statistically significant correlation only for the broadly neutralizing antibody (bNAb) PG16 (P = 0.01, ρ = 0.86) (Fig. 3), suggesting that IFITM3 promoted exposure of the glycan-V1V2 epitope in IFITM3-sensitive viruses.

FIG 3.

Correlation between the sensitivity of Env-pseudotyped viruses to IFITM3 and their increase in sensitivity to PG16 in the presence of IFITM3. The impact of IFITM3 on the increase in sensitivity to PG16 was evaluated by calculating the ratio of PG16 IC₅₀ values between IFITM3-bearing viruses and IFITM3-free viruses (control viruses), expressed as a percentage. Spearman’s correlation coefficient (ρ) and the P value are indicated.

The V1V2 and V3 loops of Env determine viral sensitivity to IFITM3.

We investigated the possible role of the V1V2 loop of Env targeted by PG16 in modulating the sensitivity of Env-pseudotyped viruses to IFITM3 by generating chimeric Env proteins in which the V1V2 loops of Env proteins from IFITM3-sensitive and IFITM3-resistant viruses were exchanged. We replaced the V1V2 loop of the IFITM3-resistant AD8 Env-pseudotyped virus with that of the sensitive NL4-3, 249N3, or 11c6 virus (Fig. 4A). The resulting chimeric Env-pseudotyped viruses were highly sensitive to IFITM3, which decreased their relative infectivity from 88% to 13% for AD8-V1V2(NL4-3), to 7% for AD8-V1V2(249N3), and to 12% for AD8-V1V2(11c6) (Fig. 4B). Conversely, the NL4-3 Env, conferring viral sensitivity to IFITM3, was modified to incorporate the V1V2 loop of resistant AD8, 2c4, or 249V4 virus, but unfortunately, the viruses containing these substitutions were noninfectious (data not shown). We therefore replaced the NL4-3 Env backbone with that of 249N3 Env to generate infectious chimeric Env-pseudotyped viruses (Fig. 4A). However, the impacts of the change differed considerably between the three V1V2 loops, with an increase in relative infectivity from 19% to 105% for 249N3-V1V2(AD8) but to only 33% for 249N3-V1V2(249V4) and, with a moderate decrease, to 8% for 249N3-V1V2(2c4) (Fig. 4C). These results suggested a possible role for other determinants, depending on the context of the Env clone considered.

FIG 4.

Impact of the V1V2 and V3 loops of Env proteins on viral sensitivity to IFITM3. (A) Alignment of the amino acid sequences of the V1V2 and V3 loops of the Env proteins of IFITM3-sensitive (249N3, 11c6, and NL4-3) and IFITM3-resistant (AD8, 2c4, and 249V4) viruses. Positively charged amino acids are shown in red, negatively charged residues in yellow, and potential N-glycosylation sites in green. (B to G) Env-pseudotyped viruses expressing parental or chimeric Env were produced in the presence or absence of IFITM3 by cotransfecting HEK293T cells with pNL4-3Δenv.luc, pCI-env (parental or chimeric), and pcDNA-ifitm3-Flag or empty pcDNA. The viruses present in cell supernatants were harvested, their levels were normalized on the basis of reverse transcriptase (RT) activity, and their infectivity was evaluated 48 h postinfection by measuring luciferase activity (RLU). The impact of IFITM3 on the infectivity of each virus was evaluated by calculating the RLU ratio between IFITM3-bearing viruses and IFITM3-free viruses (control viruses), expressed as a percentage. The data shown are mean results obtained from at least three independent experiments performed in quadruplicate. Chimeric Env proteins were produced by exchanging V1V2 and/or V3 loops as follows. (B) V1V2 loops from Env proteins from IFITM3-sensitive viruses were incorporated into the Env protein of the resistant AD8 virus. (C) V1V2 loops from the Env proteins of IFITM3-resistant viruses were incorporated into the Env protein of the sensitive 249N3 virus. (D) V3 loops from the Env proteins of IFITM3-sensitive viruses were incorporated into the Env protein of the resistant AD8 virus. (E) V3 loops from Env proteins of IFITM3-resistant viruses were incorporated into the Env protein of the sensitive NL4-3 virus. (F) V1V2 and V3 loops from the Env proteins of sensitive viruses were incorporated into the Env protein of the resistant AD8 virus. (G) V1V2 and V3 loops from the Env proteins of resistant viruses were incorporated into the Env protein of the sensitive NL4-3 virus. The statistical significance of differences in sensitivity to IFITM3 between parental and chimeric Env-pseudotyped viruses was evaluated in Student’s t tests. **, P ≤ 0.01; ***, P ≤ 0.001.

We hypothesized that the V3 loop of Env might play a role complementary to that of V1V2 in determining the sensitivities of viruses to IFITM3 for several reasons. The V3 loop is located close to the V1V2 domain at the apex of the Env trimer (44, 45), and certain V3 loop substitutions have been shown to significantly increase viral resistance to PG16 (33). In addition, the V3 loop was recently reported to be a major determinant of the modulation, by Env, of the sensitivity of the NL4-3 laboratory strain to IFITM3 (29). We addressed this question by generating chimeric Env proteins in which the V3 loops of IFITM3-sensitive and IFITM3-resistant viruses were exchanged. We replaced the V3 loop of the resistant AD8 Env-pseudotyped virus with that of the sensitive NL4-3, 249N3, or 11c6 virus (Fig. 4A). The NL4-3 V3 loop substitution resulted in a noninfectious virus. We therefore evaluated the impact of the 249N3 and 11c6 substitutions only. Both viruses were more sensitive to IFITM3 than the parental AD8 virus, with a decrease in relative infectivity from 88% to 41% for both AD8-V3(11c6) and AD8-V3(249N3) (Fig. 4D). Conversely, NL4-3 Env, conferring viral sensitivity to IFITM3, was modified to contain the V3 loop of the resistant AD8, 2c4, or 249V4 virus (Fig. 4A). The AD8 V3 loop increased the resistance of the NL4-3 virus to that of the AD8 virus, as previously reported (29), with an increase in relative infectivity from 30% to 118% for NL4-3-V3(AD8), whereas the 2c4 and 249V4 V3 loops had a much more moderate impact on the sensitivity of the NL4-3 virus, increasing relative infectivity to only 60% for NL4-3-V3(2c4) and to 51% for NL4-3-V3(249V4) (Fig. 4E). Together, these data indicate that both the V1V2 and V3 loops are important determinants of the modulation of viral sensitivity to IFITM3 by Env.

We then combined substitutions of V1V2 and V3 loops, based on the hypothesis that the partial impact observed for certain Env proteins during the independent replacement of these loops might be increased by replacing both loops simultaneously. We therefore replaced the V1V2 and V3 loops of AD8 with those of NL4-3, 249N3, or 11c6 and, conversely, the V1V2 and V3 loops of NL4-3 with those of AD8, 2c4, or 249V4. For the three combined substitutions in AD8, relative infectivity decreased considerably, from 88% to levels similar to those in the parental sensitive viruses: 23% for AD8-V1V2V3(NL4-3), 25% for AD8-V1V2V3(249N3), and 35% for AD8-V1V2V3(11c6) (Fig. 4F). Similarly, two of the three combined substitutions in NL4-3 increased the resistance of the NL4-3 virus from 30% to levels similar to those in parental resistant viruses: 120% for NL4-3-V1V2V3(AD8) and 87% for NL4-3-V1V2V3(249V4) (Fig. 4G). However, the last substitution in NL4-3, involving the V1V2 and V3 loops of 2c4, had a modest, statistically nonsignificant impact on the sensitivity to IFITM3 of NL4-3, suggesting that the resistance of 2c4 involves other determinants. Together, these data confirmed that the V1V2 and V3 domains have complementary effects on the modulation of viral sensitivity to IFITM3. This is highlighted, particularly, by the V1V2 and V3 domains of the 249V4 virus, resistant to IFITM3, which increased the resistance of a sensitive virus only moderately when introduced separately but to a much greater extent, reaching levels similar to those in the parental 249V4 virus, when introduced together.

As replacements of V1V2 and/or V3 domains have varied impacts on the intrinsic infectivities of viruses (in the absence of IFITM3) (Fig. 5A to F), we determined if a link between the intrinsic infectivity of viruses and their phenotype of sensitivity/resistance to IFITM3 could be found. As observed above with parental viruses, no correlation was observed between these two variables (P = 0.32, ρ = 0.25) (Fig. 5G), confirming that changes in sensitivity to IFITM3 are due to specific replacements of the V1V2 and V3 domains and not to modifications of the intrinsic infectivities of viruses. However, close inspections of the amino acid sequences of the V1V2 and V3 loops did not identify specific sequence signatures that could be used to distinguish between IFITM3-sensitive and IFITM3-resistant viruses. The numbers of both potential N-glycosylation sites and of charged residues were similar in the two groups (Fig. 4A).

FIG 5.

Impact of replacements of V1V2 and V3 domains on the intrinsic infectivities of viruses. (A to F) The intrinsic infectivities of parental and chimeric viruses contained in HEK293T cell supernatants were evaluated by measuring RLU values 48 h after infection of 1 × 10⁴ TZM-bl cells with serial 5-fold dilutions of the supernatants in quadruplicate. RLU values were then normalized for the supernatant RT amount (RLU per nanogram of RT). (G) Absence of correlation between relative infectivity (percent of the control level) and intrinsic infectivity (RLU per nanogram of RT). Spearman’s correlation coefficient (ρ) and the P value are indicated.

We extended our analysis of the correlation between viral sensitivity to IFITM3 and the increase in sensitivity to PG16 to the chimeric viruses by assessing the sensitivity of all these viruses to PG16 in the presence and absence of IFITM3. After we discarded two constructs highly resistant to PG16 (IC₅₀ above the highest concentration of PG16 used), the analysis confirmed that the incorporation of IFITM3 into the viral particle sensitized the virus to the inhibitory action of PG16, with decreases in the IC₅₀ of PG16 correlated with sensitivity to IFITM3 (P = 0.042, ρ = 0.42) (Fig. 6).

FIG 6.

Correlation between the sensitivities of chimeric and parental Env-pseudotyped viruses to IFITM3 and their increase in sensitivity to PG16 in the presence of IFITM3. Spearman’s correlation coefficient (ρ) and the P value are indicated.

Finally, we addressed the question of a putative link between the IFITM3 sensitivity phenotype of viruses and their tropism by determining the coreceptor use of chimeric Env-pseudotyped viruses by measuring their ability to infect CD4⁺ U373 MAGI cells expressing either the CXCR4 or the CCR5 coreceptor. However, as with the parental viruses, we found no link between their IFITM3 sensitivity phenotype and the tropism of chimeric Env proteins (Fig. 4B to G).

IFITM3 interacts with the gp120 surface subunit of IFITM3-sensitive viruses.

Previous studies have reported the existence of a physical interaction between Env and IFITM3 in virion-producing cells (26). We sought to confirm the existence of this interaction and to explore whether differences in interactions could be observed between IFITM3-resistant and IFITM3-sensitive viruses, with three IFITM3-resistant Env proteins (AD8, 2c4, and 249V4) and three IFITM3-sensitive Env proteins (NL4-3, 11c6, and 249N3). HEK293T cells were cotransfected as described above with the pNL4.3.LUC.R-E- vector and each pCI-env vector in the presence or absence of IFITM3 (pcDNA-ifitm3-Flag or empty pcDNA). Cell lysates were obtained 32 h after transfection and used for Western blotting to evaluate the production and processing of Env and the production of IFITM3 (Fig. 7A). As observed in viral particles, IFITM3 expression levels varied slightly across experiments in cells transfected with pCDNA-ifitm3-Flag. However, IFITM3 levels had no significant effect on the production and processing of Env. Cells lysates were then subjected to immunoprecipitation with an anti-gp120 serum or with an anti-Flag antibody, and the immunoprecipitated proteins were subjected to Western blotting (Fig. 7B). We found that, regardless of the Env protein tested, the anti-gp120 serum precipitated gp120 and its precursor gp160 and coimmunoprecipitated IFITM3. In contrast, when immunoprecipitation was performed with the anti-Flag antibody, differences were observed between IFITM3-sensitive and IFITM3-resistant Env proteins. For IFITM3-resistant Env proteins, only the gp160 precursor was coimmunoprecipitated with IFITM3, whereas for IFITM3-sensitive Env proteins, both the gp160 precursor and the gp120 mature subunit were coimmunoprecipitated. This suggests that, for IFITM3-resistant Env proteins, the interaction between Env and IFITM3 in virus-producing cells may be impaired following the processing of Env in the Golgi apparatus.

FIG 7.

Interaction of IFITM3 and Env in virion-producing cells. (A) HEK293T cells were cotransfected with pNL4-3Δenv-luc, pCI-env, and pcDNA-ifitm3-Flag or empty pcDNA. Cells were lysed 32 h posttransfection by incubation in membrane-disrupting RIPA buffer. The levels of HIV-1 Env and IFITM3 in cell lysates were assessed by Western blotting with anti-gp120 and anti-Flag antibodies, respectively. β-Actin levels were used as a loading control. IFITM3/actin ratios were determined by quantifying the intensities of the IFITM3 and actin bands with ImageQuant TL8.1 software (GE Healthcare). (B) Lysates were immunoprecipitated (IP) with an antiserum against HIV-1 gp120 or a sheep IgG isotype control (top panel) and with an anti-Flag antibody or a rabbit IgG isotype control (bottom panel). Immunoprecipitated proteins were subjected to Western blotting, as in panel A.

To support this hypothesis, we used confocal immunofluorescence microscopy to track the colocalization of IFITM3 and Env in cells producing two IFITM3-resistant viruses (AD8 and 249V4) and two IFITM3-sensitive viruses (NL4-3 and 249N3). HEK293T cells were transfected with each pCI-env, the pNL4-3Δenv.luc vector, and the pcDNA vector containing the ifitm3 gene. Thirty-two hours posttransfection, the cells were labeled with an anti-IFITM3 antibody, an anti-Env antibody, and an antibody targeting the endoplasmic reticulum, the Golgi apparatus, or the plasma membrane. Antibody binding was detected by incubation with secondary antibodies conjugated to different fluorochromes: Alexa Fluor 488 (green) for the IFITM3 protein, Alexa Fluor 594 (red) for Env, and Alexa Fluor 647 (purple) for cellular compartments. Under each set of conditions, we obtained about 30 z-stack acquisitions, from which Imaris software calculated the mean Pearson colocalization coefficient. We observed strong colocalization of Env and IFITM3 in the endoplasmic reticulum (Fig. 8A) and Golgi apparatus (Fig. 8C), regardless of the Env protein studied. An analysis of Pearson colocalization coefficients showed no major difference in the degrees of Env-IFITM3 colocalization in these two compartments, except for the IFITM3-resistant 249V4 virus, which displayed higher levels of colocalization than the IFITM3-sensitive 249N3 virus in the Golgi apparatus, although colocalization remained strong even for the sensitive virus (Fig. 8B and D). Interestingly, a major difference in Env-IFITM3 colocalization was observed at the plasma membrane, where strong Env-IFITM3 colocalization was observed for the two IFITM3-sensitive viruses (NL4-3 and 249N3), but very weak colocalization was observed for the two resistant viruses (AD8 and 249V4) (Fig. 8E). This observation was confirmed by a comparison of mean Pearson's colocalization coefficients: 0.64 and 0.60 for NL4-3 and 249N3, respectively, but only 0.38 and 0.36 for AD8 and 249V4, respectively (Fig. 8F). Together, these results suggest that IFITM3 interacts with IFITM3-sensitive Env proteins from their synthesis in the endoplasmic reticulum and during their transport to the Golgi complex, right up to their arrival on the plasma membrane at viral particle assembly sites, whereas the interaction between IFITM3 and resistant Env proteins is impaired or less frequent after Env processing in the Golgi apparatus.

FIG 8.

Confocal microscopy analysis of IFITM3 and Env expression in cells producing IFITM3-resistant and IFITM3-sensitive viruses. HEK293T cells were transiently cotransfected with pNL4-3Δenv.luc, pCI-env, and pcDNA-ifitm3-Flag. The cells were fixed 32 h posttransfection, permeabilized, and labeled with an anti-IFITM3 antibody, an anti-Env antibody, and an antibody targeting the endoplasmic reticulum (ER) (A), the Golgi apparatus (C), or the plasma membrane (PM) (E). Antibody binding was detected by incubation with secondary antibodies conjugated to different fluorochromes: Alexa Fluor 488 (green) for IFITM3, Alexa Fluor 594 (red) for Env, and Alexa Fluor 647 (purple) for the cellular compartments. Images were acquired with a Leica TCS SP8 g-STED confocal microscope. The entire cell was analyzed by z-stack acquisition; ±30 z-stacks were acquired for each set of conditions. The histograms show mean Pearson colocalization coefficients calculated with Imaris software, quantifying the triple colocalization of Env, IFITM3, and cell components: the endoplasmic reticulum (ER) (B), the Golgi apparatus (D), or the plasma membrane (PM) (F). The statistical significance of the mean Pearson colocalization coefficients was evaluated in Student’s t tests. ***, P ≤ 0.001.

DISCUSSION

Several reports have suggested that Env may be an important determinant of the sensitivity of HIV-1 to IFITM3. However, it remains unclear how IFITM3 affects the functionality of Env. We addressed this question here by comparing Env-pseudotyped viruses with different sensitivities to IFITM3 incorporation into the virion. Using viruses differing solely in terms of the Env clone expressed at their surface, we showed that the nature of the Env protein determined viral sensitivity to IFITM3 independently of the level of IFITM3 incorporation into virions. Previous studies have highlighted the broad spectrum of viruses targeted by IFITM proteins, suggesting that IFITM proteins act by altering the fusogenic properties of cellular or viral membranes through a nonspecific generic mechanism. However, the protection provided by specific Env proteins suggests that, in HIV, this mechanism may be overcome or may be insufficient to prevent infection. Consistently with some (13, 20, 25, 28), but not all (26, 27), previous reports, we detected no major differences in Env levels or processes between virions that did and did not contain IFITM3. We therefore investigated possible changes to the conformation of Env. Using neutralizing antibodies targeting various conformational epitopes, we found that, regardless of their sensitivity to IFITM3, Env-pseudotyped viruses had a moderately higher sensitivity to most of these antibodies in the presence of IFITM3. This result may reflect subtle changes to the conformation of Env, possibly induced by changes in the physical properties of membranes, as previously suggested (23). However, despite these changes, some Env-pseudotyped viruses were resistant to the IFITM3-mediated inhibition of infection. Interestingly, the increase in viral sensitivity to one of these antibodies, PG16, could be used to differentiate between IFITM3-sensitive and IFITM3-resistant viruses, as viruses sensitive to IFITM3 displayed a greater increase in sensitivity to this antibody in the presence of IFITM3. This finding suggests that IFITM3 may promote the exposure of the N-glycan V1V2 epitope targeted by PG16 at the surface of viruses sensitive to IFITM3. By constructing chimeric Env proteins, we showed that the V1V2 loop encompassing the PG16 epitope and the V3 loop were important molecular determinants of viral sensitivity to IFITM3 and sensitization to PG16. These data confirm previous studies identifying the V3 loop as one of the determinants of viral sensitivity to the inhibition of infection by IFITM3 (29). The complementary role of these two highly glycosylated regions is not surprising, given that the assembly of gp120 subunits into trimers involves interactions at the apex between the V1V2 and V3 loops (44, 45), and this organization exposes several basic residues at the trimer apex that are important targets for PG16 (45). V3 determines the choice of coreceptor, CCR5 or CXCR4, but no direct link was found between viral coreceptor use and sensitivity to IFITM3, with a wide range of sensitivities observed in both R5 and dually tropic X4R5 viruses. However, given the conflicting reports concerning this point (20, 29, 46), a study including a larger panel of X4 coreceptor-using viruses is required to determine whether sensitivity to IFITM3 is affected by viral coreceptor use. Likewise it would be interesting to include a larger panel of viruses from chronically infected patients to more precisely assess whether the sensitivity of viruses to IFITM3 is affected by the stage of infection.

We hypothesized that the exposure of the PG16 epitope of IFITM3-sensitive viruses might be favored by a specific interaction between IFITM3 and the mature trimer. The existence of a physical interaction between Env and IFITM3 has been reported in virion-producing cells (26). We therefore explored whether this interaction differed between IFITM3-resistant and IFITM3-sensitive viruses. In coimmunoprecipitation experiments on virion-producing cells, we found that IFITM3 interacted with both the precursor and mature forms of Env of IFITM3-sensitive viruses, but with only the precursor form of IFITM3-resistant viruses. These results were confirmed by tracking the colocalization of IFITM3 and Env in virion-producing cells. We observed a colocalization of IFITM3 with Env proteins from sensitive viruses, from the synthesis of these proteins in the endoplasmic reticulum, and during their trafficking through the Golgi complex, right up to their arrival at the plasma membrane. In contrast, the colocalization of IFITM3 and Env proteins from IFITM3-resistant viruses decreased after the passage of Env through the Golgi apparatus.

Together, these findings suggest that IFITM3 interacts with the mature Env trimer of IFITM3-sensitive viruses, stabilizing a conformational state of the trimer in which the PG16 epitope is exposed, thereby preventing the conformational changes required for receptor or coreceptor engagement and fusion between the viral and cell membranes. In contrast, the posttranslational modifications undergone by the Env trimer of viruses resistant to IFITM3 would block the access of IFITM3 to its binding site, rendering the Env trimer functional. The access of IFITM3 to its binding site on the Env trimer may be modulated by the nature of the V1V2 and V3 loops, known to undergo important posttranslational modifications. However, given the high diversity of Env sequences, a close inspection of the amino acid sequences of the V1V2 and V3 loops did not reveal specific sequence signatures distinguishing between IFITM3-sensitive and IFITM3-resistant viruses. Several different changes in these regions may, therefore, be involved. The localization of the IFITM3 binding site on the Env trimer has yet to be determined. It may be present in conserved regions within the V1V2 and/or V3 loops, but in this case, as IFITM3 sensitizes viruses to the inhibitory action of PG16, the binding of IFITM3 cannot impede the binding of PG16. IFITM3 binding therefore probably occurs outside the V1V2 and V3 loops, but the access of IFITM3 to the binding site may be regulated by these highly glycosylated domains, as previously reported for the access of some neutralizing antibodies to their epitopes (47).

It is interesting to note that for two viruses resistant to IFITM3, an increase in infection is even observed in the presence of this protein. How IFITM3 might enhance viral infection remains unanswered, but the enhancing activities of IFITM proteins on entry into target cells of other RNA enveloped viruses, such as human coronaviruses, have already been documented (48–50). For these viruses, it has been recently proposed that the effect of IFITM3 on infection may be a result of opposing pro- and antiviral mechanisms and that shifts in the balance of these activities may allow viruses to escape this innate defense mechanism. We can therefore hypothesize that depending on the fine interaction of IFITM3 with the entry machinery of a given virus, which consists of envelope glycoproteins, as shown here, but maybe also viral receptors and/or other host entry factors at the site of membrane fusion, this protein may either block or promote viral infection.

n conclusion, the data presented here highlight the important role of Env in modulating the sensitivity of HIV-1 to the inhibition of infection by IFITM3. They provide new information about the mechanism by which Env may escape this restriction factor.

MATERIALS AND METHODS

Production of Env-pseudotyped viruses.

Env-pseudotyped viruses were produced by cotransfecting 5 × 10⁶ HEK293T cells with 4 µg of one of the various pCI-env plasmids, 8 µg of pNL4.3Δenv.luc (51), and either 1 µg of empty pcDNA (control virus) or 1 µg of pcDNA-ifitm3-Flag in the presence of FuGene-6 transfection reagent (Promega). The viruses present in the supernatants of the cells were quantified after 72 h of culture by measuring the amount of reverse transcriptase (RT) using the reverse transcriptase assay colorimetric kit (Roche) according to the manufacturer’s recommendations. Viruses were harvested, purified by filtration (with a filter with 0.45-µm pores), and stored as aliquots at –80°C. Titration of infectious viruses was performed by infecting 1 × 10⁴ TZM-bl cells, with serial 5-fold dilutions of viral supernatants in quadruplicate, in the presence of 30 µg/ml DEAE-dextran. Infection levels were determined after 48 h by measuring the luciferase activity of cell lysates using the Bright-Glo luciferase assay (Promega) and a Centro LB 960 luminometer (Berthold Technologies). Wells producing relative luminescence units (RLU) >2.5 times the background were scored as positive and used to calculate both the intrinsic infectivities of viruses (RLU per nanogram of RT) and the 50% tissue culture infectious doses (TCID₅₀) in supernatants. The TCID₅₀ was calculated according the Reed and Muench formula (52) using the Excel macro found at https://www.hiv.lanl.gov/content/nab-reference-strains/html/home.htm. For the analysis of viral protein content, the supernatants of 293T cell cultures were overlaid on a 20% sucrose cushion and centrifuged at 87,000 × g for 1.5 h at 4°C. The viral pellets obtained were solubilized by overnight incubation at 4°C in 100 µl phosphate-buffered saline (PBS) supplemented with 1% Triton X-100 and protease inhibitor cocktail. The resulting suspension was then subjected to Western blotting.

Viral infectivity.

We assessed the infectivities of Env-pseudotyped viruses produced in the presence or absence of IFITM3 in quadruplicate by infecting 1 × 10⁴ TZM-bl target cells. First, IFITM3-free viruses were normalized to 400 TCID₅₀/ml. The input of IFITM3-bearing viruses was then normalized to obtain the identical reverse transcriptase activities of their IFITM3-free counterparts, and infectivity was evaluated 48 h after infection by measuring the luciferase activity (RLU) of cell lysates in Bright-Glo luciferase assays (Promega) with a Centro LB 960 luminometer (Berthold Technologies).

Determination of coreceptor usage.

Coreceptor usage was determined using U373 MAGI cells stably expressing CD4 and either CCR5 (CD4/U373/CCR5) or CXCR4 (CD4/U373/CXCR4). Each well of a 96-well plate was seeded with an aliquot of 1.5 × 10⁴ cells the day before infection. Cells were infected with 25 µl of pseudotyped virus corresponding to an amount normalized according to reverse transcriptase activity (1,000 ng) for 2 h at 37°C. We then added 175 µl Dulbecco modified Eagle medium (DMEM) supplemented with 20 µg/ml DEAE-dextran and 5% fetal bovine serum (FBS); 48 h after infection, luciferase activity was measured and viral tropism determined.

Neutralization assay.

The sensitivity of Env-pseudotyped viruses to neutralization was assessed in TZM-bl cells with the HuMobNAbs PG9, PG16, PGT145, PGT128, 10-1074, 3BNC117, 2F5, 4E10, 10E8, 8ANC195, and 35O22 (NIH AIDS Reagent Program), as previously described (31, 32).

Construction of chimeric env genes with V1V2 and/or V3 exchanges.

Chimeric env genes were constructed by exchanging the regions of env genes encoding V1V2 and/or the V3 loops conferring viral sensitivity (NL4-3, 11c6, 249N3) or resistance (AD8, 2c4, 249V4) to IFITM3. We first amplified pCI-envΔV1V2 or pCI-envΔV3 and V1V2 or V3 sequences by PCR with overlapping primers and Q5 high-fidelity DNA polymerase (New England Biolabs, USA). We checked, by electrophoresis in agarose gels, that the amplified fragments were the correct size. We then treated the PCR products with the restriction endonuclease DpnI to digest the methylated parental DNA and assembled them with the Gibson assembly cloning kit according to the manufacturer’s instructions (New England Biolabs, USA). We checked that the sequence of the env constructs was correct by Sanger sequencing.

Immunoprecipitation assay.

We investigated interactions between IFITM3 and HIV-1 Env proteins in virus-producing cells by cotransfecting 5 × 10⁶ HEK293T cells with 8 µg of pNL43Δenv-luc, 4 µg of pCI-env, and 1 µg of pcDNA-ifitm3-Flag. Thirty-two hours posttransfection, the cells were washed in phosphate-buffered saline (PBS) and lysed in RIPA buffer (1× PBS, 0.5% sodium deoxycholate, 1% Nonidet P-40, and 0.05% SDS) on ice. Clarified cell lysates were then incubated with Sepharose beads (rec-protein G-Sepharose 4B conjugate; Invitrogen, Courtaboeuf, France) conjugated with either an anti-Flag antibody (F7425; Sigma) or an antiserum against HIV-1 gp120 (NIH AIDS Reagent Program) for the immunoprecipitation of IFITM3 and Env, respectively. The beads were incubated overnight at 4°C with gentle shaking, washed five times in PBS, and then heated in Laemmli buffer supplemented with 1% β-mercaptoethanol to separate the beads and immunoprecipitated proteins, which were examined by Western blotting.

Western blotting.

Immunoprecipitated proteins or viral particles (obtained by ultracentrifugation) were denatured by heating them and fractionated by SDS-polyacrylamide gel electrophoresis (8 to 16% gradient; Eurogentec, Liège, Belgium). The separated proteins were then transferred to nitrocellulose membranes by semidry blotting (Eurogentec). The membranes were incubated with anti-HIV-1 p24 (ab53841; Abcam), anti-gp120 (ab21179; Abcam), or anti-FLAG-M2 (F1804; Sigma) primary antibodies, thoroughly washed, and incubated with rabbit anti-goat (ab6741; Abcam) or rat anti-mouse (1144-05; Southern Biotech) horseradish peroxidase (HRP)-conjugated secondary antibodies. Finally, membranes were incubated with HRP substrates for chemiluminescence (Pierce ECL Plus or Super Signal West Femto; Thermo Fisher Scientific, Waltham, MA, USA). Signals were acquired with the ImageQuant LAS 500 system and analyzed with ImageQuant TL8.1 (GE Healthcare, Chicago, IL, USA) or ImageJ software.

Immunofluorescence staining.

About 3.5 × 10⁶ HEK293T cells were used to seed 90-mm-diameter petri dishes containing coverslips precoated with 0.1% (wt/vol) poly-l-lysine (Sigma). The cells were incubated for 24 h and were then transfected as described above. After incubation for 32 h, cells were fixed in 4% paraformaldehyde in PBS, washed with 10 mM glycine in PBS, and permeabilized in 0.1% Triton X-100 in PBS for 15 min. The cells were then incubated with antibodies targeting HIV-1 Env (2G12; NIH AIDS Reagent Program) (37), IFITM3 (11714-1-AP or 66081-1-Ig; Proteintech), and one of the following cell markers: the sodium/potassium ATPase for the plasma membrane (ab76020; Abcam), the GolgB1 protein for the Golgi apparatus (HPA011008; Sigma), or calnexin for the endoplasmic reticulum (sc70481, mouse; Santa Cruz Biotechnology). Primary antibody binding was detected by incubation with appropriate secondary antibodies conjugated to Alexa Fluor 488 for IFITM3 (green), Alexa Fluor 594 for Env (red), or Alexa Fluor 647 for cell markers (purple) (Molecular Probes, Invitrogen). Cells were then mounted on glass slides in an aqueous mounting medium for fluorescence staining (BrightMount, ab103746; Abcam).

Confocal microscopy.

Images were captured with a Leica TCS SP8 g-STED inverted confocal microscope equipped with a 63× 1.4-numerical-aperture (NA) oil immersion objective (HXC PL 13 APO 63×/1.40 OIL CS) (Leica, Wetzlar, Germany). The entire cell was analyzed by z-stack acquisition, with 30 z-stacks acquired for each set of conditions. Before colocalization analysis, a single cell was cut off for the quantification, with Imaris software, of colocalization between the various labeled proteins.

Statistical analysis.

All statistical analyses were performed in GraphPad Prism 5 (La Jolla, CA, USA), with Wilcoxon’s tests, Student’s t tests, and Spearman’s correlation tests. Data from at least three independent experiments were used for the analysis.