Functional Heterogeneity of Mammalian IFITM Proteins against HIV-1

ABSTRACT

Interferon-induced transmembrane proteins (IFITMs) are a family of interferon-inducible proteins that inhibit a broad range of viruses by interfering with viral-to-cellular membrane fusion. The antiviral activity of IFITMs is highly regulated by several posttranslational modifications and by a number of protein domains that modulate steady-state protein levels, trafficking, and antiviral effectiveness. Taking advantage of the natural diversity existing among IFITMs of different animal species, we have compared 21 IFITMs for their ability to inhibit HIV-1 at two steps, during virus entry into cells (target cell protection) and during the production of novel virion particles (negative imprinting of virion particles’ infectivity). We found a high functional heterogeneity among IFITM homologs with respect to both antiviral modalities, with IFITM members that exhibit enhanced viral inhibition, while others have no ability to block HIV-1. These differences could not be ascribed to known regulatory domains and could only be partially explained through differential protein stability, implying the existence of additional mechanisms. Through the use of chimeras between active and inactive IFITMs, we demonstrate that the cross talk between distinct domains of IFITMs is an important contributor of their antiviral potency. Finally, we identified murine IFITMs as natural variants competent for target cell protection, but not for negative imprinting of virion particles’ infectivity, suggesting that the two properties may, at least in principle, be uncoupled. Overall, our results shed new light on the complex relationship between IFITMs and viral infection and point to the cross talk between IFITM domains as a novel layer of regulation of their activity.

IMPORTANCE IFITMs are broad viral inhibitors capable of interfering with both early and late phases of the replicative cycle of many different viruses. By comparing 21 IFITM proteins issued from different animal species for their ability to inhibit HIV-1, we have identified several that exhibit either enhanced or impaired antiviral behavior. This functional diversity is not driven by differences in known domains and can only be partly explained through differential protein stability. Chimeras between active and inactive IFITMs point to the cross talk between individual IFITM domains as important for optimal antiviral activity. Finally, we show that murine IFITMs are not capable of decreasing the infectivity of newly produced HIV-1 virion particles, although they retain target cell protection abilities, suggesting that these properties may be, in principle, disconnected. Overall, our results shed new light on the complex layers of regulation of IFITM proteins and enrich our current understanding of these broad antiviral factors.

INTRODUCTION

The interferon-induced transmembrane proteins (IFITMs) are a family of membrane-bound proteins that play an important role in innate immune responses due to an exquisite ability to inhibit a large spectrum of viruses (1).

Members of this family present a highly similar structural organization characterized by an intramembrane domain (IMD, previously referred to as TM1), an intracellular loop (CIL), a transmembrane domain (TMD, previously referred to as TM2), and N and C termini of variable length and regulatory functions (2–4).

By virtue of this organization, IFITMs are localized in endo-lysosomal vesicles, plasma membrane, and Golgi, and they display a heterogeneous distribution influenced by both membrane dynamics and specific protein domains. As such, while the intracellular distribution of human IFITM1 is skewed toward the plasma membrane, the presence of endocytic signals at the longer N terminus of human IFITM2/3 confer them a higher endosomal/lysosomal localization (5–11).

IFITMs present at least two peculiar features that distinguish them from many antiviral factors. First, they can inhibit a broad spectrum of viruses. Second, by acting on viral-to-cellular membrane fusion, they can interfere with the following two distinct phases of the replicative cycle common to most viruses: (i) during the step of virus entry in target cells (property defined as target cell protection and historically the first associated with IFITM inhibition) (2; reviewed in reference 4), and (ii) by triggering the production of novel virion particles of decreased infectivity (property that we refer to as the negative imprinting of virion particles’ infectivity) (12–15). In target cells, very recent studies have visually shown how IFITMs lead to the sequestration of incoming virion particles in endosomes and enhance their trafficking to lysosomes for degradation (16–18).

Instead, in infected cells undergoing active virion production, IFITMs lead the production of novel virion particles that incorporate IFITMs and exhibit decreased infectivity (13–15, 19). Although whether the physical incorporation of IFITMs in virion particles is required for this effect remains to be formally demonstrated, the presence of IFITMs during the virion assembly process leads to virions with a reduced propensity to undergo membrane fusion, similar to what has been described during infection of target cells (14, 15, 18, 20–22), as well as in more recent functions associated with the biology of IFITMs (23–25). In the case of target cell protection, the membrane fusion defect has been proposed to be due to the direct rigidification of membranes in which IFITMs are inserted and oligomerize (20, 22, 26). Whether the same mechanism applies to membrane fusion inhibition in IFITM-containing virion particles remains unclear.

IFITMs are regulated by numerous posttranslational modifications (palmitoylation, methylation, phosphorylation, and ubiquitination) that control their intracellular levels, their pattern of distribution in cells, and, ultimately, their antiviral activities (5, 6, 10). In this respect, human IFITM2 and IFITM3 exhibit long N termini containing two juxtaposed regulatory domains, a PPxY domain that serves as a docking site for the E3-ubiquitin ligase neural precursor cell expressed developmentally downregulated protein 4 (NEDD4) and a YxxΦ domain (where Φ stands for bulky amino acid residues) that directs adaptor protein 2 (AP2) complex-mediated endocytosis (5, 6, 10), and these activities can be modulated by the kinase Fyn through phosphorylation of the tyrosine residue common to both domains (5, 27). In contrast, relatively little is known about regulatory domains in human IFITM1 that lack the above-mentioned domains due to a shorter N terminus and possesses instead a longer C-terminal tail. A single study reported the existence of a noncanonical dibasic sequence (KRxx) at the IFITM1 C terminus that triggers AP3-mediated endocytosis and lysosomal targeting (8). This domain appears, however, unique to human IFITM1.

While several studies have examined the antiviral properties of individual animal IFITM molecules (albeit essentially during target cell protection) (11, 28–34), none has, so far, compared the antiviral properties of a large number of animal IFITMs in a single homogeneous setup, a comparison that may reveal differences in terms of species specificity and that may point to IFITM domains involved in protein stability, intracellular distribution, antiviral potency, and so forth.

Using a homogeneous setting, we have compared here 21 different animal IFITM proteins for their antiviral activities against HIV-1 in human cells in terms of target cell protection and negative imprinting of virion particles. Our findings highlight a remarkable heterogeneity in the action of these proteins against HIV-1, with homologs that have either completely lost their ability to inhibit the virus, while others, on the contrary, can do it more efficiently. The antiviral activities of the different IFITMs cannot simply be ascribed to the presence or absence of known regulatory domains in the N terminus of IFITMs that clearly distinguish IFITM2/3 molecules from IFITM1s, and it can only be partially explained through protein stability, suggesting the existence of additional layers of regulation. The phenotype of chimeras obtained between antiviral and nonantiviral IFITMs highlights the importance of a cross talk between individual domains for optimal antiviral activity. Finally, while past results from our lab failed to dissociate the two antiviral properties ascribed to IFITMs (19), we herein identify murine IFITMs as molecules intrinsically deficient in their ability to mediate the negative imprinting of HIV-1 virions while maintaining their effect during target cell protection. Thus, these results indicate, for the first time, that these two properties may be, at least in principle, uncoupled.

RESULTS

Description of the mammalian IFITMs used in this study.

To benefit from the natural variability existing among IFITMs, 20 different IFITM proteins derived from a wide range of mammalian species were selected, based on cells’ availability and successful cloning and compared to an identical hemagglutinin (HA)-tagged human IFITM3 (H3), the antiviral functions of which have been well characterized with respect to HIV-1 (Fig. 1A) (14, 15, 35–37). For clarity’s purposes, the IFITM genome reference nomenclature was maintained, despite the caveat that this may be misleading, because some are from poorly annotated genomes, and IFITM gene family annotations do not follow its evolutionary history. Two referenced IFITM1-like genes were also included in our analysis (Canis lupus [D; dog], D1La, and D1Lb), along with murine (Mus musculus; M) IFITM6 and IFITM7 (see Fig. 1A for a schematic representation of the genomic organization of IFITMs used here). At a general level, the percentage identity across the mammal IFITMs analyzed here was high, with a minimum of 50% for the most divergent members (M6 and M7) (Fig. 1B; for alignment purposes, human IFITM1 and IFITM2 were also included). A major difference between human IFITM1 and IFITM2/3 is the presence in the former of a shorter N terminus devoid of regulatory sequences that play important roles in IFITM trafficking and stability and of a longer C terminus that contributes to its antiviral properties, at least in human IFITM1.

FIG 1.

Presentation of the animal IFITMs used here. (A) Schematic genomic organization of IFITMs. Black boxes indicate IFITMs that have been functionally analyzed here. Given that the anti-HIV-1 activities of human IFITM1, −2, and −3 have been well characterized, only human IFITM3 was used here as functional standard for the remaining IFITMs. (B) Identity matrix of the different IFITMs. (C) Comparison of the N and C termini of animal IFITMs. (C, Left) Species cladogram. (C, Right) Comparison of N- and C-terminal sequences of selected IFITMs. N-terminal sequences were aligned as described in Materials and Methods, while the C-terminal sequences have little sequence homology and are shown here as sequence list.

A simple survey of the lengths of the N and C termini among the mammalian IFITMs tested here indicates that while the N terminus appears homogeneous in IFITM2/3-like and IFITM1-like members (53 to 55 and 31 to 32 amino acids, respectively), the C terminus varies to a greater extent from 1 amino acid in M1 to 18 in M2 (Fig. 1C). In addition, a number of IFITM molecules present combinations of either long N and C termini (B3, P2 and P3, M2, D3, C3, and G3) or short N and C ones (M1 and R1).

All IFITM2/3-like members exhibited the adjacent PPxY and YxxΦ domains important for NEDD4 and AP2 recruitment, respectively (6, 27), with the sole exception of the rabbit IFITM3 protein (R3) that presented only the former. Lastly, while the N-terminal regions of mammalian IFITMs present enough homology to allow their good alignment, the C-terminal regions exhibit very little sequence homology between the selected species (Fig. 1C).

Next, the intracellular distribution of the different IFITMs was examined by confocal microscopy upon ectopic transfection in HEK293T cells (Fig. 2). Several IFITMs exhibited a punctuate intracellular staining similar to H3 (B3, P2, P3, M2, M3, M6, G3, and D3), in some cases marked by accrued perinuclear accumulation (C3, R3, and, to a lower measure, M7) compatible with their accumulation in the Golgi. Instead, higher plasma membrane distribution was observed for the remaining members (B1, B2, P1, M1, D1 and D1La, D1Lb, and C1 and R1). Overall, the different IFITM proteins display a heterogeneous distribution skewed toward either the plasma membrane or internal membranes, as described for their human counterparts (4, 19). The only exception was B2 that, despite its nomenclature, displays both an IFITM1-like plasma membrane distribution and sequence features.

FIG 2.

Confocal microscopy analysis of animal IFITMs. HEK293T cells ectopically transfected with DNAs coding the indicated IFITMs were fixed and analyzed by immunofluorescence and confocal microscopy analyses. The pictures are representative of the major distribution pattern observed by examining >50 positive cells per construct.

Mammalian IFITMs display a heterogeneous behavior with respect to their ability to protect target cells from infection and to negatively imprint HIV-1 virion particles infectivity.

To date, animal IFITM antiviral activities have been mostly evaluated in homologous host-species-virus settings, i.e., in cells of the same species and against viruses specific for that species (28, 29, 38–41), making a direct comparison of their antiviral potency difficult. Here, we aimed at comparing the antiviral behavior of the different IFITMs in a single cellular setting, and to this end, both target cell protection and negative imprinting of virion’s infectivity abilities of the different IFITMs were assessed against HIV-1. Target cell protection was evaluated by challenging CD4/CXCR4-HEK293T cells expressing the different IFITMs with an NL4-3-envelope HIV-1 green fluorescent protein (GFP)-coding virus prior to flow cytometry analysis 2 days postinfection (Fig. 3A). Negative imprinting of virion particles was determined by producing HIV-1 viruses in cells expressing the different IFITMs and by using purified and normalized amounts of virion particles to challenge target cells (Fig. 3A). Transfections were carried out with DNA levels that allowed a comparable IFITM expression with interferon (IFN)-stimulated primary cells, at least as appreciated for human IFITMs, as precedingly described in references 14 and 19. The same DNA levels were then used for the remaining IFITM orthologs. Upon ectopic expression, the steady-state levels of the different IFITMs varied considerably following Western blot (WB) analyses with few members either barely detectable or detectable only upon overexposure (as M6, M7, and D3; Fig. 3B and see below Fig. 4 for quantification by intracellular flow cytometry). In the case of HIV-1 virions produced in the presence of IFITMs, the levels of virion-associated IFITM proteins mirrored their intracellular levels of expression, in agreement with our previous mutagenesis study conducted on human IFITM3 (14, 19). As we and others already documented (14, 15, 19, 35), IFITM expression exerted a minor but detectable effect on the extent of virion released (Fig. 3B) so that subsequent experiments made use of exogenous reverse transcriptase (exo-RT)-normalized virions.

FIG 3.

Mammalian IFITMs exhibit a broad functional heterogeneity in their ability to inhibit HIV-1 during either target cell protection or negative imprinting of virion particles infectivity. (A) Schematic representation of the experimental system used here. To assess the activity of IFITMs in mediating the protection of target cells from infection, HEK293T cells were transiently transfected with plasmids coding CD4, CXCR4, and the indicated IFITMs. Cells were challenged 36 h later with a GFP-coding NL4-3-bearing HIV-1 vector for a single round of infection assay. The extent of infection was assessed by flow cytometry. To determine the ability of IFITMs to affect the infectivity of newly produced virion particles, HEK293T cells were instead transfected with DNAs coding the HIV-1 virus in addition to IFITMs. The amount of virion particles released in the supernatant was determined by exo-RT assay after ultracentrifugation through a 25% sucrose cushion. Virion infectivity was assessed after exo-RT normalization on HeLa P4 cells that stably express both CD4 receptor and CXCR4 coreceptor. The extent of infection was again assessed by flow cytometry. (B) WB panels display representative results obtained when IFITMs were expressed in target cells (cell lysates, top panels) or in virion-producing cells (purified virion particles, bottom panels). (C) Infectivities measured during target cell protection and negative imprinting of virion particles (average, SEM; n ≥ 4; one-way ANOVA tests between the indicated conditions; only statistically significant P values are color-coded). (D) Correlation between the two antiviral properties for each IFITM molecule. *, P < 0.0001, following a Pearson’s correlation coefficient analysis.

FIG 4.

Steady-state levels of the different IFITM proteins indicate a correlative, not absolute, trend between antiviral phenotype and levels of expression of IFITMs. Intracellular staining was used to measure the intracellular levels of IFITMs by flow cytometry in function of their MFI. (A and B) Representative histograms and cumulative variations in the MFIs, in this case normalized to human IFITM3 (gray in panel A represents staining of negative controls). Graphs present average and SEM obtained from 3 independent experiments. (C) Correlations between the antiviral effects of IFITMs and their intracellular accumulation levels. *, P = 0.0001 (left) and P = 0.01 (right), following a Pearson’s correlation coefficient analysis between the indicated conditions.

Under the conditions used here, H3 expression in target cells induced a 2-fold protection of cells from HIV-1 infection (Fig. 3C, white bars for target cell protection results). Several mammalian IFITMs exhibited antiviral activities comparable to the ones of H3 (in particular, B3, P1, M2, M3, G3, and R3), while others displayed enhanced inhibitory capacities (10-fold reduction for B1, B2, D1La, and R1; 4- to 8-fold reduction for P2, P3, M1, and D1Lb). Interestingly, several IFITMs were found to display no discernible antiviral effect during target cell protection (C1, C3, D1, D3, M6, and M7). While this could be foreseen for the most divergent IFITM members (specifically M6 and M7 that no data have so far linked to IFN responses), the lack of antiviral activities in the remaining IFITM members was unexpected.

Next, the different IFITMs were tested for their ability to alter the infectivity of normalized HIV-1 virions (Fig. 3C, blue bars). As expected, H3 drove the production of virions exhibiting a 3-fold decrease in infectivity over a single round of infection assay. In addition, several IFITMs behaved as H3 (B2, B3, P1, P2, P3, D1La, R1, and R3). In stark contrast, a more drastic phenotype (6- to 10-fold reduction in infectivity) was observed for B1, G3, and D1Lb, while no antiviral activity was detected for C1, C3, D1, D3, M1, M2, M3, M6, and M7. Lastly, the expression of M2, M6, and D3 exerted a positive effect on the infectivity of virion particles.

At a general level, a strong positive correlation was observed between the ability of individual IFITMs to interfere with the infectivities measured in target cell protection and negative imprinting of virions (Fig. 3D; r = 0.77, P < 0.0001), suggesting a strong interdependence between these two activities, as previously suggested (19). The only exceptions were noted in murine IFITMs M1, M2, and M3 proficient in target cell protection but lacking detectable effects on the infectivity of newly formed HIV-1 virions. Overall, the results obtained here highlight the heterogeneous behavior of animal IFITMs with respect to their antiviral properties, as assessed in a unique cellular and viral setting.

The intracellular levels of the different mammalian IFITMs is an important, but not unique, parameter of their antiviral properties.

The antiviral potency of IFITM proteins is influenced by a number of factors and, among them, those governing the protein’s steady-state levels (6, 11, 19, 35). To determine whether protein stability (at least at steady state) could relate to the antiviral behavior of the IFITMs examined here, we used intracellular staining and flow cytometry analysis (Fig. 4A for representative panels and Fig. 4B for cumulative data). When thus analyzed and compared to human IFITM3, the intracellular levels of accumulation of IFITMs varied from 2-fold higher to 4-fold lower. Variations spanned the animal clade division in that IFITMs within one species could present increased or decreased steady-state levels. The intracellular levels of IFITM expression (as assessed by the mean fluorescent intensity [MFI]) were inversely correlated with the infectivities measured in both target cell and negative imprinting, indicating that IFITMs present at higher levels were generally more antiviral (Fig. 4C; r = −0.7601, P = 0.0001 and r = −0.54, P = 0.01, respectively). This correlation was, however, lower for the negative imprinting of virion particles defect, indicating that this property is less reliant on the levels of IFITMs in cells than target cell protection. Overall, this analysis indicates that protein stability at steady state is an important parameter of the antiviral properties of IFITMs, in agreement with data present in the literature.

To assess the contribution of protein levels to the lack of antiviral activity during target cell protection, property more influenced by the intracellular levels of IFITMs, increasing doses of D1 and C1 were used in a comparison with other IFITMs exhibiting antiviral activity (D1Lb and R1, respectively). To this end, target cells transfected with different amounts of IFITM-coding DNAs were challenged with wild-type (WT) HIV-1 prior to flow cytometry analysis 2 days later (Fig. 5A). Protein levels were quantified by densitometry following WB, as this yielded similar results than intracellular staining and flow cytometry analyses under the conditions used here.

FIG 5.

Phenotypic differences observed among selected IFITM molecules are also governed by their intrinsic behavior and not only by protein stability. (A) Increasing doses of the nonfunctional D1 and C1 IFITMs were compared to a fixed dose (starting from equimolar to a 4-fold excess) of antiviral D1Lb and R1, respectively, in target cell protection. IFITM expressing cells were either analyzed by WB or challenged prior to flow cytometry analysis 3 days later. IFITM levels were, in this case, measured by densitometry following WB, as this yielded comparable results than intracellular staining and fluorescence-activated cell sorter (FACS). (B) As in panel A, but using decreasing doses of the antiviral R1 molecule over a fixed dose of the nonfunctional C1. Graphs present average and SEM of at least 4 independent experiments, and WB panels present typical results obtained. *, P < 0.05, following a one-tail Student's t test between the indicated conditions.

Increasing doses of D1 did not significantly alter virion infectivity, and when D1 was expressed at similar levels than D1Lb, the different antiviral behavior was still significant (Fig. 5A). A similar behavior was observed upon the expression of increasing doses of C1, although, in this case, we were unable to achieve a notable increase in C1 protein levels. To circumvent this issue, we used a reciprocal setup in which a fixed dose of C1 was compared to decreasing levels of R1 (Fig. 5B). In this case, despite the fact that the antiviral activity of R1 was gradually lost upon dilution, R1 was still able to inhibit viral infection when expressed at similar levels than C1. Altogether, these results indicate that protein stability is important but is not the sole contributor of the observed differences in the antiviral behavior of IFITMs.

Elements described as crucial for IFITM antiviral activities are largely conserved among mammalian IFITMs.

To investigate potential amino acid differences or evolutionary paths between the mammalian IFITMs with and without antiviral activities, we performed comparative genetic and phylogenetic analyses. Benefiting from the natural variations existing among mammalian IFITMs, we grouped members according to their nonantiviral (C1, D1, C3, and D3, only members devoid of both antiviral activities were grouped here) and antiviral (the remaining ones) behavior, omitting M6 and M7 from the analysis due to their higher divergence (Fig. 6A and B). No single common amino acid change was able to discriminate IFITMs according to their antiviral behavior. Indeed, most IFITM members displayed three conserved cysteines (C71, 72, 105; amino acid positions relative to H3) important for palmitoylation and membrane association (7, 42), with the exception of C71 replaced by phenylalanine in antiviral D1La and D1Lb IFITMs. Most IFITMs also displayed four conserved phenylalanines (F63, F67, F75, and F78) reported as relevant for antiviral functions (43, 44). Residues F75 and F78 were previously proposed to mediate IFITM3 oligomerization, but this was not the case when examined in living cells (26, 45). Therefore, these four phenylalanines are important functional determinants for unknown reasons. The only exceptions found in our analysis were the F75 and F78 residues that were replaced by isoleucine and tyrosine in C1 and M1, M2, and M3, respectively. Moreover, IFITMs possess several lysines that can be ubiquitinated, one specific to the long N terminus of IFITM2/3-like proteins and three others shared by all IFITMs (K83, K88, and K104) (6, 46). Of these, K83 and K88, but not K104, were strictly conserved among mammalian IFITMs, despite the fact that ubiquitination of K104, together with K83, is required for a recently identified scaffolding property of IFITM3 involved in B cell signaling (25). Remarkably, in none of the positions mentioned above and described in the literature to modulate the antiviral properties of IFITMs, a strict correlation existed between antiviral and nonantiviral IFITMs. Furthermore, residues described as forming an amphipathic helix in the IMD of human IFITMs (VWSLFNTL/I/VF) (47), as well as a very recently identified GxxxG motif with crucial relevance in membrane rigidification (26), were also conserved among all mammalian IFITMs tested. Finally, the “nonantiviral” IFITMs did not group together on a phylogenetic analysis (Fig. 6C), suggesting that they do not share an ancestor and that the absence of antiviral activity is not due to an ancient single event.

FIG 6.

Antiviral and nonantiviral IFITMs do not present obvious distinguishing marks. (A and B) Amino acid alignment of the tested IFITMs (A) and corresponding logo plots (B) for the antiviral and nonantiviral (D3, C3, C1, and D1; highlighted in gray) IFITMs, as described in this study. Gray underlines strictly nonantiviral IFITMs. Visualization was performed with Geneious (Biomatters, Inc.), with color-coding according to polarity. (A) H3 is set as the reference sequence; dots correspond to residues similar to the reference sequence, while hyphens correspond to gaps. Numbering is according to H3 (human IFITM3) in panel A and to the consensus sequence of the alignment in panel B. (C) Phylogenetic tree of the IFITMs tested here. Node supports are aLRT. The scale bar represents number of amino acid substitutions per site.

Therefore, the main elements previously described as crucial for the antiviral activity of IFITMs are largely conserved among mammalian IFITMs, indicating that additional ones are likely to contribute to the fine regulation of their overall antiviral behavior.

Chimeras between active and inactive IFITM homologues highlight the importance of the cross talk between individual domains to finely tune antiviral behavior.

To identify the domain(s) at the basis of the distinct behavior of the different IFITMs, we generated chimeric IFITM proteins in which domains of the antiviral R1 were inserted into the nonantiviral C1 either individually or in combination (Fig. 7A). Contrarily to other individual domains, the C terminus of R1 (C1-C_R1) promoted a substantial increase in protein accumulation at steady state, and its combination with the TMD domain [C1-(TMD+C)_R1] further increased it (Fig. 7B and C). The different chimeras did not exhibit gross changes in their intracellular distribution after confocal microscopy analysis, with the exception of the C1-TMD_R1 chimera that appeared to have lost the plasma membrane distribution typical of IFITM1-like proteins (Fig. 7D). Interestingly, when cells expressing the different chimeras were challenged with HIV-1 in target cell protection, a single chimera was able to exhibit significant antiviral activity [C1-(TMD+C)_R1] (Fig. 7E), indicating the importance of the cross talk between the TMD and the C-terminal domains for the antiviral activity.

FIG 7.

A chimera approach between C1 and R1 IFITM proteins identifies the importance of the cross talk between the TMD and the C-terminal domains for optimal antiviral activity. (A) Schematic representation of the chimeras used. (B and C) Western blot analysis and densitometric quantification following ectopic DNA transfection in HEK293T cells. (D) Representative images of the intracellular distribution of the different chimeras obtained by confocal microscopy. (E) Chimeras were tested in target cell protection by challenging HEK293T cells expressing the indicated chimeras 36 hours post-DNA transfection with GFP-coding NL4-3-bearing HIV-1. The extent of infection was measured 2 days later by flow cytometry. The graphs present average and SEM of at least 4 independent experiments, and WB panels present typical results obtained. *, P < 0.05, following a one-way ANOVA test between the indicated conditions.

A similar approach was used to engineer chimeras between D1 and D1Lb, respectively, inactive and active against HIV-1 (Fig. 8A). Also in this case, the C terminus of D1Lb (D1-C_b) was able to substantially rescue the low protein expression levels of D1. However, in this case, it was the combination of the C-terminal with the N-terminal domains [D1-(N+C)_b] that yielded the highest increase in protein levels (Fig. 8B and C). No gross modifications of the intracellular distributions of the different chimeras were observed, with the exception of the D1-C_b and the D1-(TMD+C)_b chimeras that exhibited higher perinuclear accumulation than WT (Fig. 8D). When the chimeras were tested for their ability to mediate target cell protection against HIV-1, we again found that the cross talk between the TMD and the C terminus was important for the antiviral activity (Fig. 8E). Interestingly, in this case, the combination between the N terminus and the IMD domains was also important to restore full antiviral activities. These results indicate that cross talks between individual domains are important for optimal antiviral activity, with the combination between the TMD and the C terminus being important for both chimeric configurations examined, while the one between N terminus with IMD and C terminus with TMD was important only in the case of the D1/D1Lb chimera.

FIG 8.

D1/D1Lb chimeras further stresses the importance of the cross talks between individual IFITM domains for optimal antiviral activity. Panels are as in the legend to Fig. 7. (A) Schematic representation of the chimeras used. (B and C) Western blot analysis and densitometric quantification following ectopic DNA transfection in HEK293T cells. (D) Representative images of the intracellular distribution of the different chimeras obtained by confocal microscopy. (E) Chimeras were tested in target cell protection by challenging HEK293T cells expressing the indicated chimeras 36 hours post-DNA transfection with GFP-coding NL4-3-bearing HIV-1. The extent of infection was measured 2 days later by flow cytometry. The graphs present average and SEM of at least 4 independent experiments, and WB panels present typical results obtained. *, P < 0.05, following a one-way ANOVA test between the indicated conditions.

Finally, the antiviral chimeras were also tested for their ability to mediate the negative imprinting of virion particles infectivity. For this, HIV-1 virions produced in cells ectopically expressing the IFITMs were used to challenge target cells (Fig. 9A for a schematic presentation of the chimeras and the experimental setup). All chimeric forms of IFITMs decreased HIV-1 virions’ infectivity, with C1-(TMD+C)_R1 exerting an antiviral effect largely comparable to the parental R1 homologue and the D1-(N+IMD)_b and D1-(TMD+C)_b chimeras still antiviral despite the fact that the latter appeared less potent than D1Lb (Fig. 9B). Hence, the domain cross talks showed above as relevant for target cell protection activity play an important role also during negative imprinting of virion infectivity.

FIG 9.

Chimeras between C1/R1 and D1/D1Lb that are active for target cell protection conserve their ability to negatively imprint HIV-1 virions infectivity. (A) Experimental scheme and chimeras tested here. (B) Virions were produced in the presence of the indicated chimeras, normalized, and used to challenge target cells. Infectivity was determined 48 hours later by flow cytometry. The graph presents average and SEM of 3 independent experiments. *, P < 0.05, following a one-way ANOVA test between the indicated conditions.

Murine IFITM3 is a natural IFITM variant devoid of negative imprinting of HIV-1 virion particles’ ability.

An important result highlighted from our study is the fact that despite being proficient in target cell protein, murine IFITM1, −2, and −3 seem not to affect the infectivity of newly produced HIV-1 virion particles. To determine whether this could be dependent on insufficient levels of expression, HIV-1 virions were produced in the presence of increasing doses of M3, prior to virion normalization and challenge of target cells. Even under conditions of increased doses of M3 (Fig. 10A), a consistent lack of antiviral effects was noted, suggesting a true deficiency in negative imprinting of HIV-1 virion’s infectivity, at least under the experimental conditions used here.

FIG 10.

Murine IFITM3 intrinsically lacks the ability to negatively modulate the infectivity of newly produced HIV-1 virion particles. (A) Increasing doses of human and murine IFITM3s (H3 and M3, respectively) were used to produce HIV-1 particles and to determine whether M3 would hinge on HIV-1 virions infectivity. The panels and the graph present typical results obtained (n = 3, average and SEM). *, P < 0.05, following a one-way ANOVA test between the indicated conditions. (B) The Y78F substitution was introduced in M3, and the mutant was tested for its ability to negatively imprint newly produced HIV-1 virions. The panels and the graph present typical results obtained (n = 3, average and SEM). No statistically significant difference was observed between control and mutant.

As mentioned above, we noticed that F78 (numbering with respect to human IFITM3; see Fig. 6) is commonly substituted by tyrosine in murine IFITMs (M1, M2, and M3) that lack negative imprinting of virion particles’ abilities. As such, we substituted this residue back to phenylalanine (Y78F) in the context of M3 and tested its ability to impair the infectivity of newly produced virion particles, using the same experimental setup. However, no change in infectivity could be measured when virion particles were produced in the presence of increasing levels of the M3 Y78F mutant (Fig. 10B), indicating that this residue does not play an important role in this setting.

Overall, these results indicate that murine IFITM3 is a natural variant impaired in its ability to negatively imprint the infectivity of newly produced virion particles, at least in the case of HIV-1, and also further highlight that protein stability is not the sole contributor of the observed differences in the antiviral behavior of IFITMs.

DISCUSSION

In this study, we have taken advantage of the natural variability existing between a large panel of mammalian IFITMs to compare their potency against HIV-1 in a single experimental system through expression in human cells. In doing so, we have uncovered IFITM homologs that exhibit either enhanced, or completely absent, ability to affect both target cell protection and negative imprinting of HIV-1 virions infectivity. Interestingly, despite few exceptions, the domains described in previous studies as important for either posttranslational modifications, multimerization, or overall antiviral properties are conserved across all the IFITMs tested here independently of their antiviral behavior, pointing to the existence of additional determinants for the antiviral activities of IFITMs. In this respect, the results obtained with our chimeras indicate that one such determinant may be the cross talk between individual IFITM domains.

A general distinction between human IFITM1 and IFITM2/3 proteins is the reciprocal presence of long C and N termini, respectively. Long N termini are indeed well maintained in mammalian IFITM2/3 that also exhibit a strict conservation of the NEDD4 and AP2 recruitment domains with one sole exception (R3). However, the C terminus appears more heterogeneous, and long C termini are, in some cases, present in combination with long N termini in IFITM2/3 molecules (B3, P1, P2, M3, D3, C3, and G3). Whether the latter endows such molecules with additional regulatory features with respect to those described for their human counterparts remains unknown. Relatively few studies have examined the functions of the C terminal of IFITM proteins and have been essentially focused on human IFITM1 in which a regulatory dibasic KRxx domain has been shown to regulate association to AP3 (8). However, this domain is absent in all the remaining IFITMs so that how and whether the C terminus regulates more commonly the behavior and properties of IFITM1-like proteins remain unclear. In this respect, it is of interest that M1 and R1 present very short N and C termini, which may indicate them as core minimal IFITM homologs devoid of some of the multiple regulatory possibilities described for other IFITMs.

It is evident from our comparative analysis that no common specific residue or described domain can discriminate active from inactive IFITMs. A recent report highlighted the importance of the polymorphism of residues (P, W, T, F, and Y) in the rates of palmitoylation on adjacent cysteines (C71, C72) and in the restriction capacity of bat IFITMs (32), and another determined the functional relevance of a polymorphism within the AP2 binding site of African green monkeys IFITM3 alleles (I22) (48). While variations at these residues do not explain the antiviral behavior of the IFITM tested here, these findings raise the possibility that additional polymorphism may finely tune IFITM functions, possibly in a species-specific context.

Previous work from a number of laboratories, including ours, has highlighted the importance of the intracellular levels of accumulation for the antiviral effects of human IFITM3 (14, 15, 19, 35, 36). However, the results obtained with two nonantiviral IFITM proteins (C1 and D1), as well as with chimeras, clearly indicate that protein levels are not the sole determinant of antiviral behavior, pointing to the existence of additional layers of regulation.

With the caveat that we have not examined the dynamic distribution of the different IFITMs and that for a given mutation, it is difficult to completely separate effects on protein levels from those on antiviral behavior, the results we have obtained with two sets of chimeras indicate that the cross talk between two distinct domains (in particular, the TMD and the C terminus) is an important modulator of antiviral functions. While the C-terminal and TMD domains in R1 are very different from those in C1, they exhibit only four and five amino acid changes between D1 and D1Lb, respectively. Yet the transfer of the C-terminal and TMD domains is sufficient to restore antiviral functions. The fact that the combination of N terminus and C terminus can also exert similar effects in the context of the D1/D1Lb but not of the C1/R1 chimera suggests that such cross talks are likely to be different according to the specific IFITM examined, probably as a result of species-specific adaptations.

At present, the reasons for the absence of antiviral activities by some IFITM molecules remain unclear, and, in terms of evolution, it is entirely possible that this has resulted from a loss of antiviral activity that has been compensated by other IFITM members of a given species not tested here. However, an alternative, and not mutually exclusive, possibility we favor is that certain IFITMs may be subjected to host species-specific regulation and/or exhibit optimal activity against viruses naturally encountered in the same species. In this respect, the feline C1 and C3 IFITMs examined here are poorly expressed in human cells and inactive against HIV-1, but they may exhibit better stability and antiviral properties in a feline cell-virus context.

To support this contention, it is interesting to note that murine IFITMs (M1, M2, and M3) were found in this study to inhibit HIV-1 during target cell protection but not to alter the infectivity of newly formed HIV-1 virion particles. While on one hand, this finding is important because it suggests for the first time that these two properties may be dissociated, on the other, it raises the possibility that murine IFITM3 may have been optimized for activity against viruses of its own species. The latter would be in agreement with a recent study reporting that murine IFITM3 is capable of interfering with the production of infectious virion particles of the murine leukemia virus (MLV) (49). Thus, these results highlight the possibility that species-specific adaptations may have shaped the optimal antiviral activity of a given IFITM against viruses of the same species.

In conclusion, our study highlights a broad functional heterogeneity among mammalians IFITMs against HIV-1 and indicates that the cross talk between IFITM domains finely tunes the antiviral properties of the different IFITMs, adding a novel layer of complexity to the regulation of the activities of these broad antiviral factors.

MATERIALS AND METHODS

Cell culture, plasmids, and antibodies.

HEK293T epithelial cell line ATCC (CRL-3216) and HeLa P4/P5 expressing the CD4 receptor along with the CXCR4 and CCR5 coreceptors were cultured in complete Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% fetal calf serum (FCS). The following IFITMs were amplified by PCR from cDNAs obtained from relevant animal cell lines (cell lysates were kind gifts of the CelluloNet facility of the SFR-Biosciences in Lyon Gerland or of Frederick Arnaud at the Infections Virales et Pathologie Comparée [IVPC], Lyon, France): MBDK from Bos Taurus cell line, PK (15) from Sus scrofa, RK15 from Oryctolagus cuniculus, TIGMEC from Capra hircus, and CRFK from Felis catus. Murine and canine IFITMs were obtained by gene synthesis (Genewiz). All IFITMs were cloned as BamHI/NotI fragments into a pcDNA-HA vector. The species abbreviations used in the manuscript are H, Homo sapiens; M, Mus musculus; G, Capra hircus; D, Canis lupus familiaris; C, Felis catus; P, Sus scrofa; B, Bos Taurus; and R, Oryctolagus cuniculus. Retrieved sequences were strictly identical to the gene ID numbers as follows: H3, gene ID 10410; M1, 68713; M2, 80876; M3, 66141; M6, 213002; M7, 74482; G3, 102180655; D1, 483397; D1La, 475935; D1Lb, 483396; D3, 606890; C1, 101086846; C3, 101100838; P1, 100127358; P2, 100620056; P3, 100518544; B1, 353510; B2, 615833; B3, 777594; R1, 103347593; and R3, 100353245. D1/D1Lb and C1/R1 chimeras were obtained by gene synthesis (Genewiz). CD4-, CCR5-, and CXCR4-coding plasmids were obtained from the AIDS Reagent and Reference Program of the NIH (catalog nos. ARP-158, ARP-3325, and ARP-3326, respectively). The anti-HA (catalog no. H3663) and anti-α-tubulin (catalog no. T8203) antibodies were purchased from Sigma. The anti-HIV-1 Gag and anti-HIV-1 Env antibodies were similarly obtained from the AIDS Reagent and Reference Program of the NIH (clone 183-H5C and catalog no. 288, respectively).

Virus production, purification, normalization, and infections.

Standard single rounds of infection vectors were produced by transient calcium phosphate transfection with DNAs coding for a mini-viral genome bearing GFP (pRRL-CMV-GFP), the Gag and Gag-Pol structural proteins (8.2), and the NL4-3 Env and the HIV-1 Rev protein (Rev) (4, 4, 1, and 0.5 μg, respectively, for a 10-cm plate), as described in references 13 and 14.

The target cell protection properties of the different IFITMs were assessed by using the above-mentioned virus at a multiplicity of infection (MOI) comprised between 0.1 and 0.2 on HEK293T cells ectopically transfected 36 hours earlier with DNAs coding CD4, CXCR4, and the different IFITMs (0.2, 0.2, and 1 μg, respectively).

To assess the negative imprinting of virion infectivity abilities of IFITMs, HIV-1-derived lentiviral particles were produced in the presence of IFITMs by calcium phosphate DNA cotransfection of HEK293T cells using Gag-Pro-Pol, viral GFP genome, Env, Rev, and IFITMs or pcDNA control plasmid (using 4, 4, 0.5, 0.5, 3 μg each for a 10-cm dish plate) as previously described (13, 14, 19). We have already shown that in the case of human IFITMs, these levels were similar to those observed for the endogenous form in primary dendritic cells stimulated with type I interferon.

In this case, 48 hours after transfection, cell-free supernatants were syringe filtered (0.45 μm), and virion particles were purified by ultracentrifugation through a 25% sucrose cushion (wt/vol) for 2 h at 110,000 × g. Virion pellets were then resuspended in phosphate-buffered saline (PBS) for further analyses and normalized by exogenous reverse transcriptase (exo-RT) as described. Infections were carried out on HeLa P4/P5 cells with normalized amounts of virions for 2 to 4 h, and the extent of infection was assessed 3 days later by flow cytometry.

Immunofluorescence and confocal microscopy.

HEK293T cells growing on glass coverslips coated with poly-l-lysine 0.01% were transfected with plasmids coding animal IFITMs with Lipofectamine 3000 according to the manufacturers’ instructions (Invitrogen, Life Technologies). Twenty-four hours later, cells were fixed with paraformaldehyde (PFA) (3.7%), permeabilized with Triton 0.1% in PBS, and incubated first with the primary anti-HA antibody overnight (at a 1:100 dilution) and then with a secondary antibody conjugated to Alexa Fluor 488 (Molecular Probes, Life Technologies; catalog no. A21202; 1:1,000 dilution). Coverslips were stained with a DAPI (4′,6-diamidino-2-phenylindole)-containing solution (1:10,000 dilution in PBS) and mounted using the antiquenching solution Fluoromount-G (Southern Biotech). Fluorescent confocal images were collected using a Zeiss LSM 880 AiryScan confocal microscope and pictures were analyzed with the ImageJ software.

Intracellular staining and flow cytometry analysis.

HEK293T cells expressing the different IFITMs were permeabilized with the Cytofix/Cytoperm Plus kit (catalog no. 554715; BD), according to the manufacturer’s instructions before incubation with a phycoerythrin-conjugated anti-HA antibody (Miltenyi Biotec; catalog no. 130-092-257) and flow cytometry analysis. The mean fluorescent intensity of HA-positive cells was then determined using the FlowJo software.

Sequence analyses and phylogenetics.

Sequence alignment of the various IFITMs was performed using amino acid sequences with PRANK-F with a gap rate of 0.005 and minor adjustments (50). The N and C termini of the selected mammalian IFITMs are highly variable (low homology) and are the result of complex evolutionary histories. They aligned very differently depending on the alignment method and gap parameters. All in all, only the central portion of IFITMs is enough homologous and can be correctly aligned. Visualization in Fig. 6 was produced using Geneious (Biomatters, Inc.), with color-coding according to polarity, and logo plots. Phylogenetics in Fig. 6 were performed using PhyML with JTT + G + I as a model and aLRT for node support (51, 52). The sequence alignment in FASTA format and the phylogenetic tree in nwk format are available at https://doi.org/10.6084/m9.figshare.13365893 and https://doi.org/10.6084/m9.figshare.13365893, respectively.

Statistical analysis.

One-way analysis of variance (ANOVA), one-tailed Student t tests, or Pearson’s correlation coefficient were used, as indicated in the legends to figures.

Data availability.

All relevant data are within the paper.