1. Evolutionary Biology
  2. Immunology and Inflammation
Download icon

Homology-guided identification of a conserved motif linking the antiviral functions of IFITM3 to its oligomeric state

  1. Kazi Rahman
  2. Charles A Coomer
  3. Saliha Majdoul
  4. Selena Y Ding
  5. Sergi Padilla-Parra
  6. Alex A Compton  Is a corresponding author
  1. HIV Dynamics and Replication Program, Center for Cancer Research, National Cancer Institute, United States
  2. Cellular Imaging Group, Wellcome Trust Centre for Human Genetics, University of Oxford, United Kingdom
  3. Department of Infectious Diseases, King’s College London, Faculty of Life Sciences & Medicine, United Kingdom
  4. Randall Division of Cell and Molecular Biophysics, King’s College London, United Kingdom
Research Article
  • Cited 0
  • Views 610
  • Annotations
Cite this article as: eLife 2020;9:e58537 doi: 10.7554/eLife.58537

Abstract

The interferon-inducible transmembrane (IFITM) proteins belong to the Dispanin/CD225 family and inhibit diverse virus infections. IFITM3 reduces membrane fusion between cells and virions through a poorly characterized mechanism. Mutation of proline-rich transmembrane protein 2 (PRRT2), a regulator of neurotransmitter release, at glycine-305 was previously linked to paroxysmal neurological disorders in humans. Here, we show that glycine-305 and the homologous site in IFITM3, glycine-95, drive protein oligomerization from within a GxxxG motif. Mutation of glycine-95 (and to a lesser extent, glycine-91) disrupted IFITM3 oligomerization and reduced its antiviral activity against Influenza A virus. An oligomerization-defective variant was used to reveal that IFITM3 promotes membrane rigidity in a glycine-95-dependent and amphipathic helix-dependent manner. Furthermore, a compound which counteracts virus inhibition by IFITM3, Amphotericin B, prevented the IFITM3-mediated rigidification of membranes. Overall, these data suggest that IFITM3 oligomers inhibit virus-cell fusion by promoting membrane rigidity.

Introduction

The intrinsic protection of cells from virus infection represents an early and essential aspect of antiviral innate immunity. Cytokines, including interferons, signal the presence of invading viruses and induce an ‘antiviral state’ via the expression of hundreds of antiviral genes (Yan and Chen, 2012; Bieniasz, 2004). This arsenal of antiviral proteins converges on many steps of the virus life cycle in order to collectively inhibit infection of cells and prevent virus spread. In addition, certain ‘front-line’ antiviral proteins impose a constant barrier to infection because they are expressed constitutively and are further upregulated by interferons. The interferon-induced transmembrane (IFITM) proteins are the earliest acting restriction factors known, inhibiting the entry of diverse viruses into cells by restricting fusion pore formation during virus-cell membrane fusion (Shi et al., 2017; Chesarino et al., 2014a; Perreira et al., 2013; Bailey et al., 2014). Among the growing list of viruses shown to be inhibited by IFITM proteins in cell culture and in vivo are orthomyxoviruses, flaviviruses, filoviruses, alphaviruses, and coronaviruses (Shi et al., 2017). IFITM3 is a potent inhibitor of Influenza A virus (IAV) infection in cell culture and in vivo, and consequently, it is the most studied member of the IFITM family (Bailey et al., 2012; Everitt et al., 2012; Allen et al., 2017). While the precise mechanism by which IFITM3 reduces virus-cell fusion remains unresolved, evidence suggests that it does so by altering the properties of lipid membranes.

Two models have been proposed to explain how IFITM3 inhibits virus fusion. In the first, IFITM3 plays an indirect role by interacting with VAMP-associated protein A (VAPA) and inhibiting lipid transport between the endoplasmic reticulum and endosomes, resulting in an accumulation of endosomal cholesterol (Amini-Bavil-Olyaee et al., 2013). High cholesterol content may inhibit the fusion of virus-containing vesicles with the limiting membrane of the late endosome, restricting virus release into the host cell cytoplasm. While inhibition of cholesterol trafficking has been shown to inhibit virus entry (Poh et al., 2012), this model of IFITM3 function has been subsequently supported by one study (Kühnl et al., 2018) but challenged by several others (Appourchaux et al., 2019; Lin et al., 2013; Desai et al., 2014; Wrensch et al., 2014). The second and currently favored model posits that IFITM3 directly inhibits fusion by locally reducing membrane fluidity (i.e. increasing membrane rigidity or order) (Lin et al., 2013; Li et al., 2013) and by inducing positive membrane curvature (Li et al., 2013). Membrane order and curvature influence one another (Vanni et al., 2014; Rangamani et al., 2014; Golani et al., 2019) and, together, they regulate numerous membrane fusion processes (Martens and McMahon, 2008). Consistent with the notion that IFITM3 induces local membrane order and curvature to disfavor virus-cell fusion, IFITM3 contains a juxtamembrane amphipathic helix which is essential for antiviral activity (Chesarino et al., 2017). Amphipathic helices have been identified in many eukaryotic, prokaryotic, and viral proteins and are well known for their ability to bind and bend membranes (Drin and Antonny, 2010; Giménez-Andrés et al., 2018). Another piece of evidence that supports the membrane deformation model of fusion inhibition is that IFITM3 incorporated into enveloped viruses also impairs virion fusion with target cells (Tartour et al., 2014; Yu et al., 2015; Tartour et al., 2017; Ahi et al., 2020; Sharma et al., 2019; Compton et al., 2016; Suddala et al., 2019; Compton et al., 2014). Despite this progress, a complete mechanistic view of IFITM3 is lacking because studies describing the impact of IFITM3 on membranes have not included mutants lacking antiviral function.

It was previously reported that IFITM3 forms clusters on virus-containing vesicles (Kummer et al., 2019) and that IFITM3 oligomerization promotes restriction of IAV (John et al., 2013). However, the determinants initially purported to mediate oligomerization (phenylalanines at residues 75 and 78 of the CD225 domain [John et al., 2013]) were later shown to be unnecessary when oligomerization was measured using a FRET-based approach in living cells (Winkler et al., 2019). Therefore, the oligomerization of IFITM3 appears to be influenced by unknown determinants and its importance to antiviral function is not established.

In the current study, we set out to identify a loss-of-function mutation in IFITM3 suitable for mechanistic studies by using a homology-guided approach. The IFITM genes (IFITM1, IFITM2, IFITM3, IFITM5, and IFITM10 in humans) are members of an extended gene family known as the Dispanin/CD225 family (hereafter referred to as CD225 proteins) (Sällman Almén et al., 2012; Zhang et al., 2012). Members of this group are characterized by the presence of a CD225 domain, but the functions of most remain unknown. However, one member is the subject of numerous studies because it is linked to neurological disorders. Mutations in proline-rich transmembrane protein 2 (PRRT2) result in conditions of involuntary movement, such as paroxysmal kinesigenic dyskinesia, benign familial infantile seizures, and episodic ataxia (Gardiner et al., 2015; Valtorta et al., 2016). PRRT2 is a neuron-specific protein that is localized to pre-synaptic terminals and which inhibits synaptic vesicle fusion (Meschia, 2018; Mo et al., 2019; Liu et al., 2016). Molecular studies of disease-associated missense mutations in PRRT2 (G305W/R) indicate that it causes loss-of-function, leading to unchecked neurotransmitter release (Valente et al., 2016; Coleman et al., 2018; Gardiner et al., 2012; van Vliet et al., 2012; Liu et al., 2012). Interestingly, the homologous residue in human IFITM3 is also subject to rare allelic variation in humans (G95W/R) and this mutation results in partial loss of activity against IAV infection (John et al., 2013). However, the reason why this site is essential for the respective functions of PRRT2 and IFITM3 was unknown.

Here, we demonstrate that glycine-95 of human IFITM3 resides within a GxxxG motif that is highly conserved among vertebrate IFITM3 orthologs as well as PRRT2. Mutation of glycine-91 or glycine-95 rendered IFITM3 less active against IAV (in target cells) and HIV-1 (in virus-producing cells). We found that the GxxxG motif mediates IFITM3 oligomerization in living cells, with glycine-95 playing a dominant role. An IFITM3 mutant (G95L) exhibiting loss of antiviral function was deficient for oligomerization, indicating that IFITM3 oligomerization and virus restriction are functionally associated. We leveraged this loss-of-function mutant to identify mechanistic correlates of antiviral function which are associated with IFITM3 oligomerization. We found that IFITM3 increased membrane order, as previously suggested, whereas IFITM3 encoding G95L or mutations within its amphipathic helix failed to do so. In an effort to further probe the importance of membrane order in the antiviral mechanism, we demonstrate that Amphotericin B (Ampho B) decreases the stiffness of IFITM3-containing membranes and rescues virus infection. These data indicate that promotion of membrane order by IFITM3 oligomers is required for its antiviral activity. Furthermore, we reveal that oligomerization is a shared requirement for the distinct anti-fusion functions performed by homologs IFITM3 and PRRT2.

Results

Homology-guided identification of a putative oligomerization motif within CD225 domains

While it has been suggested that IFITM3 may adopt multiple topologies (Li et al., 2013; Yount et al., 2012), experimental evidence indicates that IFITM3 is a type II transmembrane protein characterized by the presence of a cytoplasmic-facing amino terminus, a CD225 domain consisting of a hydrophobic intramembrane (IM) domain, a cytoplasmic intracellular loop (CIL), and a hydrophobic transmembrane (TM) domain, and a very short carboxy terminus facing the vesicle lumen or extracellular space (Bailey et al., 2013; Ling et al., 2016). PRRT2 is thought to adopt a similar topology in membranes (Rossi et al., 2016). We used Protter, an interactive application that maps annotated and predicted protein sequence features onto the transmembrane topology of proteins (Omasits et al., 2014) to visualize IFITM3 and PRRT2. The two proteins exhibit a similar predicted topology consisting of dual hydrophobic domains and a CIL, but the amino terminus is considerably longer in PRRT2 (Figure 1A and B). Given the association of G305W with loss-of-function of PRRT2, we wondered whether mutation of the homologous residue in IFITM3 (glycine-95) would compromise its respective functions as well. Upon comparing topologies and the protein alignment, we noticed that these glycines form part of a GxxxG motif in the CIL of IFITM3 and PRRT2, and this motif is intact in several other IFITM and CD225 proteins (Figure 1C). The GxxxG motif, also known generally as a (small)xxx(small) motif, is frequently associated with dimerization of membrane proteins (Teese and Langosch, 2015; Overton et al., 2003). Most often shown to mediate pairing of hydrophobic transmembrane helices within a bilayer, the motif has also been described to drive oligomerization from cytoplasmic loops or linkers (Lu et al., 2014). The GxxxG motif is conserved in IFITM3 of vertebrates, indicating that it may play an important functional role (Figure 1D).

Figure 1 with 1 supplement see all
Homology-guided identification of a putative oligomerization motif within CD225 domains.

(A) Schematic representation of the membrane topology of IFITM3 made with Protter. Residues corresponding to the amphipathic helix (yellow), palmitoylated cysteines (blue), phenylalanines purported to regulate oligomerization (red), and the glycines of the GxxxG motif (green) are indicated. (B) Schematic representation of the membrane topology of PRRT2 made with Protter. Residues corresponding to the glycines of the GxxxG motif (green) are indicated. (C) A partial amino acid alignment of CD225 domains from IFITM proteins, PRRT2, and TUSC5. Color codes are included as in (A and B). The position of the polymorphic glycine in PRRT2 associated with neurological disease (G305W) is underlined. (D) A partial amino acid alignment of IFITM3 orthologs in vertebrates. Conserved glycines in the GxxxG motif (green) are indicated.

91GxxxG95 is important for restriction of virus entry by IFITM3

We generated FLAG-tagged IFITM3 mutants in which glycine-91 and glycine-95 were changed to leucine (G91L and G95L), following an example set by characterization of the GxxxG motif in the human folate transporter (Wilson et al., 2015). We also produced a G95W mutant because a rare single-nucleotide polymorphism in IFITM3 known as rs779445843 gives rise to missense mutations (G95W/R) in human populations (Cunningham et al., 2019; Sherry et al., 2001). Furthermore, G95W in IFITM3 is analogous to the disease-associated polymorphism in PRRT2 (G305W) which results in loss of function (Coleman et al., 2018).

To begin the functional characterization of IFITM3 harboring mutations within its GxxxG motif, we assessed steady-state protein levels following transient and stable transfection into HEK293T cells. Transiently expressed IFITM3 mutants reached similar levels as wild-type (WT) as determined by flow cytometry (Figure 1—figure supplement 1A–B). However, western blot analysis revealed a significant expression defect for the G91L mutant, but not for the others (Figure 1—figure supplement 1C–D). In stably transfected cells, all IFITM3 constructs resulted in protein expression levels that exceeded those observed following transient transfection, and the only mutant exhibiting significantly reduced levels was G95W (Figure 2A and B).

Figure 2 with 1 supplement see all
91GxxxG95 is important for restriction of virus entry by IFITM3.

(A) HEK293T cells stably transfected with empty pQCXIP, IFITM3 WT-FLAG, or the indicated mutants were fixed, stained with anti-FLAG antibody and assessed by flow cytometry. FLAG levels are displayed as histograms. (B) Mean fluorescence intensity measurements of FLAG staining in (A) are shown for three independent experiments. (C) HEK293T cells stably transfected with empty pQCXIP, IFITM3 WT-FLAG, or the indicated mutants were challenged with IAV PR8 strain (MOI of 0.1), fixed at 18 hr post-infection, stained with an anti-nucleoprotein antibody, and assessed by flow cytometry. (D) Mean infection results representing 5–8 independent experiments are normalized to empty vector (set to 100%). (E) HEK293T cells stably transfected with empty pQCXIP, IFITM3 WT-FLAG, or the indicated mutants were challenged with replication-incompetent HIV-1 incorporating BlaM-Vpr and pseudotyped with VSV glycoprotein. Virus-cell fusion was assessed at 2.5 hr post-virus addition using the beta lactamase assay and flow cytometry. Results represent the mean of three independent experiments and are normalized to empty vector (set to 100%). Error bars indicate standard error. Statistical analysis was performed using one-way ANOVA. *, p<0.05; **, p<0.001. MFI, mean fluorescence intensity. Rel., relative.

Next, we used confocal immunofluorescence microscopy to address how modification of the GxxxG motif of IFITM3 influences its subcellular localization. In cells stably or transiently expressing IFITM3 constructs, we assessed the extent to which IFITM3 colocalizes with EEA1-GFP, a marker for early endosomes, and CD63, a marker for late endosomes/multivesicular bodies (Figure 2—figure supplement 1A–B). As shown previously (Shi et al., 2017; Huang et al., 2011; Feeley et al., 2011), IFITM3 WT was detected in early endosomes, late endosomes, and at the plasma membrane. Specifically, approximately 30% of IFITM3 WT was associated with early endosomes and 30% was associated with late endosomes. The G91L, G95L, and G95W mutations did not obviously alter colocalization between IFITM3 and endosomes in either stably or transiently transfected cells (Figure 2—figure supplement 1A–B). We noticed that stable expression of all IFITM3 constructs resulted in greater colocalization with CD63+ late endosomes. We also noticed that the G91L mutation resulted in an increased colocalization between IFITM3 and both early endosomes and late endosomes under transient conditions, but this difference was not statistically significant. Therefore, it is unlikely that mutations in the GxxxG motif impact the subcellular localization of IFITM3.

HEK293T cell lines stably expressing FLAG-tagged IFITM3 WT and mutants were then challenged with IAV to assess antiviral function. IFITM3 WT strongly protected cells from infection, but G91L resulted in a partial loss of virus restriction while G95L and G95W resulted in a more substantial loss of restriction (Figure 2C and D). The extent of restriction did not correlate with relative levels of IFITM3 protein in stable cell lines (Figure 2B), demonstrating that differential antiviral activity is not due to differential protein expression. To confirm that IFITM3 targets the virus entry step, we challenged cells with HIV-1 pseudotyped with VSV glycoprotein (VSV-G) in an assay for virus-cell fusion (Figure 1—figure supplement 1E and Figure 2E). Inhibition of HIV-VSV-G by the IFITM3 constructs resembled inhibition of IAV, in that G95L and G95W resulted in a substantial loss of restriction. G91L, however, only slightly impacted virus restriction and its activity was not significantly different from WT. Therefore, a rational approach identified glycine-95 to be a major determinant for the broad inhibition of virus entry by IFITM3. Since the G95L and G95W mutations were functionally redundant, we did not pursue the characterization of G95W further.

91GxxxG95 is important for restriction of HIV-1 virion infectivity by IFITM3

We and others have previously demonstrated that, in addition to preventing virus entry into naive target cells, IFITM3 performs another antiviral function in virus-producing (infected) cells by incorporating into virions, reducing viral glycoprotein abundance and function, and reducing the fusogenic potential of virions (Tartour et al., 2014; Yu et al., 2015; Tartour et al., 2017; Ahi et al., 2020; Sharma et al., 2019; Compton et al., 2014). Therefore, we tested the impact of mutations in the GxxxG motif on restriction of HIV-1 virion infectivity. IFITM3 WT reduced the infectivity of HIV-1, while G95L resulted in a partial loss of activity. Strikingly, G91L completely abrogated this antiviral function (Figure 3A and B).

91GxxxG95 is important for restriction of HIV-1 virion infectivity by IFITM3.

(A) HEK293T were co-transfected with HIV-1 molecular clone pNL4.3 and empty pQCXIP, IFITM3 WT-FLAG, or the indicated mutant. Virus-containing supernatants were harvested at 48 hr post-transfection and subjected to ultracentrifugation over sucrose pellets. Virion content was quantified by p24 CA ELISA and 50 ng p24 equivalent was added to TZM.bl cells for infectivity measurements. TZM.bl were fixed at 48 hr post-infection, stained with anti-CA antibody, and infection was assessed by flow cytometry. (B) Mean infection results of TZM.bl using virus derived from three independent transfections of HEK293T are shown and normalized to empty vector (set to 100%). (C) Whole cell lysates and (D) virus-containing supernatants were collected from HEK293T co-transfected with the HIV-1 molecular clone pNL4.3 and empty pQCXIP, IFITM3 WT-FLAG, or the indicated mutant at 48 hr post-transfection. Virus-containing supernatants were ultracentrifuged through sucrose cushions. Both lysates and concentrated, purified virus-containing supernatants (50 ng p24 equivalent) were subjected to SDS-PAGE and Western blot analysis. Immunoblotting was performed with anti-gp120, anti-gp41, anti-CA, anti-actin, and anti-FLAG. (E) Virion-associated levels of gp120 Env were quantified by measuring fluorescence of DyLight-conjugated secondary antibody and were normalized to levels of CA in three independent experiments. (F) Virion-associated levels of gp41 Env and (G) IFITM3-FLAG were quantified similarly. For anti-Env immunoblotting, the amount of gp120 or gp41 in virions was presented relative to empty vector (set to 100%). For anti-FLAG immunoblotting, the amount of IFITM3 WT in virions was set to 100%. Error bars indicate standard error. Statistical analysis was performed using one-way ANOVA. *, p<0.05; **, p<0.001. Rel., relative. Norm., normalized.

In order to identify correlates of antiviral function against HIV-1, we assessed viral Envelope (Env) levels expressed in virus-producing cells and levels incorporated into virions. The transient expression of IFITM3 WT in virus-producing cells was accompanied by a partial loss of Env gp120 and gp41 in those cells (Figure 3C) and in virions (Figure 3D) as previously reported (Yu et al., 2015; Ahi et al., 2020; Wang et al., 2017). This effect was measured in replicate experiments using quantitative immunoblotting (Figure 3E and F). In contrast, transient expression of the G91L or G95L mutants did not result in reduced Env levels (Figure 3C–F). Since the G91L and G95L mutations differentially impact the restriction of HIV-1 virion infectivity by IFITM3, it is therefore unlikely that Env quantity in virions fully accounts for this restriction. However, we observed that the G91L and G95L mutations strongly impaired the ability of IFITM3 itself to incorporate into virions (Figure 3G), and the extent of virion incorporation correlated with the measured impact on HIV-1 infectivity (Figure 3B). Together, our results demonstrate that the dual antiviral functions performed by IFITM3 (early-stage inhibition of virus entry and late-stage inhibition of virion infectivity) are critically regulated by the GxxxG motif. Interestingly, the former function depends primarily on glycine-95, while the latter function depends more on glycine-91.

91GxxxG95 regulates oligomerization of IFITM3 in living cells

Based on their position within a conserved GxxxG motif, we used Förster resonance energy transfer (FRET) to assess the roles played by glycine-91 and glycine-95 in the oligomerization of IFITM3. We constructed IFITM3 fused with yellow fluorescent protein (YFP) or mCherry at the amino terminus to create FRET pairs and to perform fluorescence lifetime imaging microscopy (FLIM). The measurement of fluorescence lifetimes allows for measurements that are independent of fluorophore expression level or diffusion rate (Day, 2014). In this framework, excitation of YFP (donor) results in energy transfer to mCherry (acceptor) when a molecular interaction brings the pair into close proximity. Co-transfection of mCherry and IFITM3 WT-YFP served to establish background FRET values. By comparison, IFITM3 WT-YFP and IFITM3 WT-mCherry co-transfection resulted in high levels of detectable FRET (Figure 4A and B). Relative to WT, introduction of the G91L mutation partially inhibited FRET, while the G95L mutation had a much stronger effect, reducing FRET to near background levels (Figure 4A and B). These data suggest that both glycine-91 and glycine-95 contribute to IFITM3 oligomerization, but that glycine-95 is the major determinant. FRET was also measured for heterologous pairs (combining IFITM3 WT-YFP and G95L-mCherry, and vice versa) and the results were indicative of an intermediate degree of oligomerization (Figure 4B). To complement our FRET analysis, we also analyzed fluorescence lifetimes of the donor (YFP). Co-transfection of IFITM3 WT-YFP and IFITM3 WT-mCherry resulted in significant decreases in YFP lifetimes (averaging about 400 picoseconds), consistent with IFITM3 oligomerization (Figure 4A and C). Meanwhile, co-transfection of G91L-YFP and G91L-mCherry resulted in partial decreases in YFP lifetimes (averaging about 225 picoseconds), while we observed only minor decreases in YFP lifetimes (averaging about 70 picoseconds) when G95L-YFP and G95L-mCherry were examined. Within this experimental framework, we confirmed that the F75A/F78A mutations did not affect IFITM3 oligomerization, nor did mutations within the amphipathic helix (S61A/N64A/T65A) (Figure 4—figure supplement 1). Since our experiments revealed that glycine-95 is the major determinant for IFITM3 oligomerization, we focused on the G95L mutant for further mechanistic characterization.

Figure 4 with 2 supplements see all
91GxxxG95 regulates oligomerization of IFITM3 in living cells.

(A) HEK293T were transiently co-transfected with IFITM3-YFP and mCherry or IFITM3-YFP and IFITM3-mCherry. Constructs encoded IFITM3 WT, IFITM3 G91L, or IFITM3 G95L. FRET-FLIM measurements were made, and images of FRET signal and YFP lifetimes are representative of 12–20 captured images per condition. (B) Whole-cell FRET analysis was performed on a minimum of 50 cells per condition and the results of three independent experiments were pooled. Dots correspond to individual cells. (C) Whole-cell YFP lifetimes were measured on a minimum of 50 cells per condition and the results of three independent experiments were pooled. Dots correspond to individual cells. A mean delta (Δ) value is indicated to represent the drop in YFP lifetime resulting from the pairing of IFITM3-YFP and mCherry versus the pairing of IFITM3-YFP and IFITM3-mCherry. Empty red circles are used to depict mCherry, filled red circles are used to depict IFITM3-mCherry (either WT, G91L, or G95L), and filled yellow circles are used to depict IFITM3-YFP (either WT, G91L, or G95L). Error bars indicate standard deviation. Statistical analysis was performed using one-way ANOVA. *, p<0.05; **, p<0.001. Scale bars, 10 μm. Ps, picoseconds. Au, arbitrary units.

To ensure that conclusions drawn from this imaging approach are relevant to IFITM3-mediated antiviral function, we tested the ability for our fluorescently-tagged IFITM3 constructs to inhibit IAV. The antiviral potency of IFITM3 WT tagged with mCherry or YFP at the amino terminus was decreased by 20% compared to IFITM3 WT-FLAG (Figure 4—figure supplement 2A–B). Nonetheless, IFITM3 G95L tagged with mCherry or YFP exhibited no antiviral activity whatsoever, validating the utility of our fluorescently-tagged IFITM3 constructs for mechanistic studies. During the preparation of this manuscript, Suddala et al. reported the construction and use of an IFITM3 fusion protein encoding a fluorescent protein placed internally, rather than terminally, after residue 40 of IFITM3 (Suddala et al., 2019). Since this fusion protein was purported to exhibit minimal loss to antiviral function, we decided to produce additional fusion proteins following a similar strategy. We introduced mCherry or YFP into IFITM3 after residue 40 as described in Suddala et al. and assessed these constructs for antiviral activity and oligomerization potential. We found that the internal placement of mCherry or YFP resulted in a greater loss of antiviral activity compared to the amino terminal placement of mCherry or YFP. Specifically, the constructs encoding mCherry or YFP internally lost more than 40% of their activity compared to a 20% loss exhibited by constructs encoding mCherry or YFP at the amino terminus (Figure 4—figure supplement 2A–B). For thoroughness, we tested the IFITM3 WT constructs encoding internal mCherry and YFP for FRET competence and found that oligomerization was apparent (Figure 4—figure supplement 2C). Furthermore, introduction of the G95L mutation into these constructs resulted in decreased oligomerization (Figure 4—figure supplement 2C). Therefore, these data demonstrate that our FRET-based approach to studying IFITM3 oligomerization is amenable to the placement of fluorophores at the amino terminus or after residue 40 of IFITM3. However, placement of YFP or mCherry at the amino terminus allows for better preservation of antiviral function.

Glycine-95 regulates oligomerization of IFITM3 in denaturing and native conditions

In parallel to our studies of IFITM3 oligomerization in single, living cells, we assayed the ability of IFITM3 pairs tagged with FLAG or myc to co-immunoprecipitate from bulk cell lysates. HEK293T were co-transfected with IFITM3-FLAG and IFITM3-myc followed by FLAG immunoprecipitation, SDS-PAGE, and quantitative immunoblotting. We found that IFITM3 WT-myc readily pulled down with IFITM3 WT-FLAG, while pull down of G95L-myc with G95L-FLAG was diminished by approximately 50% (Figure 5A and B). To address heteromultimerization between IFITM3 WT and IFITM3 G95L, we paired IFITM3 WT-FLAG with G95L-myc and the results were indicative of an intermediate degree of oligomerization (Figure 5—figure supplement 1A–B). Therefore, membrane-extracted IFITM3 forms oligomers, but the G95L mutation reduces oligomerization. We then performed blue native PAGE and immunoblotting to assess the oligomeric state of IFITM3 under non-denaturing conditions. Two populations of IFITM3 oligomers, exhibiting sizes of approximately 300 and 480 kilodaltons, were readily observed for IFITM3 WT-FLAG and, to a lesser extent, IFITM3 G95L-FLAG. The largest (480 kilodaltons) was reduced by approximately 50% for IFITM3 G95L (Figure 5C and D). A complete scan of the blue native PAGE result in Figure 5C is shown in Figure 5—figure supplement 1C. We did not detect a population of IFITM3 dimers (expected at approximately 30 kilodaltons), which may reflect the specific conditions under which blue native PAGE was performed here. The fact that IFITM3 G95L exhibited a reduced potential for higher order oligomer formation using this technique is consistent with our experiments using living cells or denatured bulk lysates. Therefore, glycine-95 is necessary for efficient IFITM3 oligomer formation.

Figure 5 with 1 supplement see all
Glycine-95 regulates oligomerization of IFITM3 in denaturing and non-denaturing conditions.

(A) HEK293T were transiently transfected with empty pQCXIP or the following pairs: IFITM3 WT-FLAG and IFITM3 WT-myc or G95L-FLAG and G95L-myc. Whole cell lysates were produced under mildly denaturing conditions and immunoprecipitation (IP) using anti-FLAG antibody was performed. IP fractions and volumes of whole cell lysates were subjected to SDS-PAGE and Western blot analysis. Immunoblotting was performed with anti-FLAG and anti-myc. Heavy chain IgG and Actin served as loading controls in the IP fraction and lysates fraction, respectively. Number and tick marks indicate size (kilodaltons) and position of protein standards in ladder. (B) Levels of IFITM3-myc (either WT or G95L) co-immunoprecipitated by anti-FLAG IP were quantified in (A) and represented as the mean of three independent experiments. Error bars indicate standard error. (C) HEK293T were transiently transfected with empty pQCXIP, IFITM3 WT-FLAG or G95L-FLAG. Cell lysates were produced with 1% digitonin and blue native PAGE was performed, followed by immunoblotting with anti-FLAG. Number and tick marks indicate size (kilodaltons) and position of protein standards in ladder. (D) Levels of IFITM3-FLAG (either WT or G95L) corresponding to ~480 kd and ~300 kD were quantified in (C) and represented as the mean of three independent experiments. Error bars indicate standard error. (D) Number and Brightness analysis was performed on monomeric mCherry and IFITM3-mCherry (either WT or G95L) as described in the Materials and methods. Statistical analysis was performed using student’s T test. *, p<0.05; **, p<0.001. Rel., relative.

To quantitatively resolve the specific oligomeric state of IFITM3, we performed a Number and Brightness analysis (Digman et al., 2008; Nolan et al., 2017). This approach is a fluorescence fluctuation microscopy method capable of measuring the apparent average number of molecules and their brightness in each pixel over time, with brightness being proportional to oligomeric state. We restricted our analysis to brightness of IFITM3-mCherry as it trafficked in and out of the plasma membrane over time. On average, IFITM3 WT-mCherry was found to be 2.14 times brighter than mCherry monomers (Shaner et al., 2004; Figure 5E). In contrast, G95L-mCherry brightness was not significantly different than mCherry monomers (averaging 1.30 times brighter). Furthermore, relative to mCherry monomers, both IFITM3 WT-mCherry and G95L-mCherry formed assemblies that were up to five times brighter, but IFITM3 WT-mCherry demonstrated a greater propensity to form these higher order oligomers (Figure 5E). One caveat of this Number and Brightness analysis is that, in contrast to transmembrane proteins, mCherry monomers do not target membranes and are expressed mostly in the cytoplasm (Teese and Langosch, 2015). However, the fact that IFITM3 WT exhibits a mean brightness that is roughly twofold greater than IFITM3 G95L supports the notion that the former exists primarily as a dimer and the latter as a monomer. Together, these data suggest that IFITM3 forms dimers and higher order oligomers in membranes in a glycine-95-dependent fashion.

Membrane order is increased by IFITM3 oligomers and is a correlate of antiviral function

While the precise mechanism by which IFITM3 inhibits virus-cell fusion remains unresolved, a salient phenotype of IFITM protein expression is increased membrane order (reduced membrane fluidity) (Lin et al., 2013; Li et al., 2013; John et al., 2013). Therefore, we leveraged our loss-of-function mutant, G95L, to directly test whether membrane order is functionally associated with inhibition of virus entry by IFITM3. Previous reports of membrane order enhancement by IFITM family members involved the use of a cell-permeable dye known as Laurdan (Zhang et al., 2006). Here, we assessed membrane order using Laurdan and a recently described sensor known as fluorescent lipid tension reporter (FliptR) (Colom et al., 2018; Goujon et al., 2019; Coomer et al., 2020). FliptR is a planarizable push-pull probe that incorporates efficiently into artificial and living cell membranes and whose fluorescence parameters change upon alterations in local lipid packing (order). Specifically, FliptR responds to increasing membrane order at the plasma membrane and at endomembranes by planarization, leading to longer fluorescence lifetimes detected by FLIM (Colom et al., 2018; Dal Molin et al., 2015). Using FliptR in HEK293T stably expressing IFITM3-FLAG, we found that IFITM3 WT expression led to significantly increased membrane order (Figure 6A). In fact, the IFITM3-induced enhancement of membrane order was similar to that achieved by addition of soluble cholesterol, while cholesterol depletion with methyl-beta-cyclodextrin resulted in profound decreases in membrane order (Figure 6—figure supplement 1A–B). In contrast to WT, cells expressing IFITM3 G95L did not exhibit increased membrane order (Figure 6A). Similar results were obtained using Laurdan, in that IFITM3 WT expression resulted in significantly increased membrane order while G95L did not (Figure 6—figure supplement 1C–D). These data indicate that membrane order enhancement tracks with a functionally competent form of IFITM3 but not with a loss-of-function mutant. To further probe the functional importance of membrane order in the antiviral mechanism of IFITM3, we performed experiments in the presence of Ampho B. This antimycotic polyene compound was previously reported to overcome the antiviral activity of IFITM3, rendering cells stably expressing IFITM3 fully permissive to IAV (Lin et al., 2013). However, it was unknown how Ampho B counteracts the effects of IFITM3. When we added Ampho B to cells expressing IFITM3 WT, we no longer observed increased membrane order (Figure 6A). Furthermore, in identically treated cells, Ampho B prevented restriction by IFITM3 of HIV pseudotyped with hemagglutinin (HA) from IAV (Figure 6B). These findings show, for the first time, that the capacity for Ampho B to overcome the antiviral activity of IFITM3 is linked to its ability to decrease membrane order. Therefore, the use of Ampho B as an interrogative tool reinforced the role played by membrane order in the antiviral mechanism of IFITM3.

Figure 6 with 1 supplement see all
Membrane order is increased by IFITM3 oligomers in an Ampho B-sensitive manner.

(A) HEK293T cells stably transfected with empty pQCXIP, IFITM3 WT-FLAG, or G95L-FLAG were stained with 1 μM FliptR for 5 min and imaged by FLIM. In the condition indicated, 1 μM Amphotericin B was added to cells for 1 hr and washed away prior to addition of FliptR and imaging. The whole-cell mean fluorescence lifetime (τ), in addition to individual component lifetimes (long and short lifetimes, τ1 and τ2), was calculated using Symphotime for a minimum of 40 cells per condition and τ1 from three independent experiments were pooled and plotted. Dots correspond to individual cells. Error bars indicate standard deviation. (B) HEK293T cells stably transfected with empty pQCXIP, IFITM3 WT-FLAG, or G95L-FLAG were challenged with HIV pseudotyped with hemagglutinin (HA) from IAV WSN strain at a MOI of 1. In the condition indicated, 1 μM Amphotericin B was added to cells for 1 hr prior to virus addition. Cells were fixed at 48 hr post-infection, and infection was scored by GFP expression using flow cytometry. Statistical analysis was performed using one-way ANOVA. *, p<0.05; **, p<0.001. Ps, picoseconds. Rel., relative.

Since we previously reported a critical role for the amphipathic helix of IFITM3 in virus restriction (Chesarino et al., 2017), we tested whether the helix was also required for the enhanced membrane order observed in cells expressing IFITM3. Like the G95L mutant, a version of IFITM3 encoding S61A/N64A/T65A mutations in the amphipathic helix lost the ability to increase membrane order (Figure 6—figure supplement 1E). Since the S61A/N64A/T65A mutations did not inhibit oligomerization (Figure 4—figure supplement 1), our results identify oligomerization and the amphipathic helix as dual requirements for the antiviral functions of IFITM3. Overall, these data strongly suggest that virus restriction by IFITM3 occurs through oligomerization-dependent membrane stiffening induced by the amphipathic helix.

Disease-associated G305W impairs oligomerization of PRRT2

Since the GxxxG motif is a shared feature of IFITM3 and PRRT2 and that a naturally occurring mutant G305W is predictive of neurological disease, we assessed the oligomerization capacity of WT and mutant PRRT2. As in Figure 4, we constructed PRRT2 fused with YFP or mCherry to create FRET pairs. We observed that co-transfection of PRRT2 WT-YFP and PRRT2 WT-mCherry resulted in significant FRET, demonstrating that PRRT2 oligomerizes in living cell membranes (Figure 7A). However, FRET was significantly reduced for pairs containing G305W (G305W-YFP and G305W-mCherry) (Figure 7A). Furthermore, pairs containing G305W did not exhibit loss in YFP lifetimes relative to their WT counterparts (Figure 7B). These data suggest that mutation of glycine-305 in PRRT2 results in loss of protein oligomerization. Therefore, the divergent functions played by homologues IFITM3 and PRRT2 in the regulation of fusion processes are controlled by a common determinant.

Disease-associated G305W disrupts the oligomerization of PRRT2 in living cells.

(A) HEK293T were transiently co-transfected with PRRT2-YFP and mCherry or PRRT2-YFP and PRRT2-mCherry. Constructs encoded PRRT2 WT or PRRT2 G305W. Whole-cell FRET analysis was performed on a minimum of 50 cells per condition, and the results of three independent experiments were pooled. Dots correspond to individual cells. (B) Whole-cell YFP lifetimes were measured on a minimum of 50 cells per condition and the results of three independent experiments were pooled. Dots correspond to individual cells. A mean delta (Δ) value is indicated to represent the drop in YFP lifetime resulting from the pairing of PRRT2-YFP and mCherry versus the pairing of PRRT2-YFP and PRRT2-mCherry. Empty red circles are used to depict mCherry, filled red circles are used to depict PRRT2-mCherry (either WT or G305W), and filled yellow circles are used to depict PRRT2-YFP (either WT or G305W). Error bars indicate standard deviation. Statistical analysis was performed using one-way ANOVA. *, p<0.05; **, p<0.001. Au, arbitrary units. Ps, picoseconds.

Discussion

The physiological importance of IFITM3 in the control of many virus infections in vivo is becoming increasingly apparent (Zani and Yount, 2018; Kenney et al., 2017). While it has been proposed that membrane remodeling by IFITM3, at the level of membrane order and curvature, protects host cells from virus invasion, functional proof was lacking. However, two recent developments provided a glimpse of how (and when) IFITM3 inhibits virus-cell membrane fusion. First, the identification of an amphipathic helix located in the IM domain of IFITM3 provided a rational explanation for how membrane stiffening and/or bending may occur (Chesarino et al., 2017). Second, IFITM3 has been observed to intercept vesicles carrying inbound virions and to restrict their release into the cytoplasm, while viruses that are insensitive to IFITM3 evade its encounter (Suddala et al., 2019; Spence et al., 2019). Together with the previous demonstrations that the subcellular localization of IFITM3 determines its specificity and potency (Compton et al., 2016; Jia et al., 2012; Chesarino et al., 2014b), a ‘proximity-based’ mechanism presents itself in which IFITM3 interacts with and modifies host membranes needed by some viruses to fuse with cells. Importantly, this model accommodates the antiviral effect of IFITM3 inside viral lipid bilayers as well (Tartour et al., 2014; Yu et al., 2015; Ahi et al., 2020; Suddala et al., 2019; Compton et al., 2014).

Here, we provide evidence that an additional determinant (protein oligomerization) plays a crucial role within this mechanistic framework. Previously, an alanine scanning approach led to the identification of two residues (phenylalanine-75 and phenylalanine-78) that were important for IFITM3 oligomerization in cell lysates (John et al., 2013). This study did not functionally test a role for glycine-91 or glycine-95 in the oligomerization of IFITM3 because alanine mutagenesis overlapping these residues apparently led to loss of stable protein expression (Appourchaux et al., 2019; John et al., 2013). While it has been confirmed that F75A and F78A mutations disrupt antiviral activity (Suddala et al., 2019; Zhao et al., 2018), IFITM3 oligomerization was not impacted by these mutations when assessed by FRET in living cells (Winkler et al., 2019) and we confirm here that F75A and F78A mutations do not interfere with the measurement of IFITM3 oligomerization by FRET. Instead, we find that G91L partially abrogates oligomerization while G95L almost completely disrupts oligomerization. Moreover, our data shows that G95L strongly reduces the antiviral potential of IFITM3 as well as its capacity to increase membrane order (reduce membrane fluidity). Furthermore, we show that a compound previously found to abrogate the antiviral function of IFITM3, Ampho B (Lin et al., 2013; Suddala et al., 2019), decreases membrane order in IFITM3-expressing cells. Even though the exact mechanism by which Ampho B impacts mammalian membranes is unclear (Lin et al., 2013; Kamiński, 2014), these results identify that the membrane stiffening property of IFITM3 is a strict correlate of its antiviral functions in cells and, perhaps, in virions.

We found that glycine-91 and glycine-95 of the GxxxG motif were also important for restriction of HIV-1 virion infectivity by IFITM3. However, the G91L mutation had a stronger impact than G95L on this antiviral activity in virus-producing cells, suggesting that IFITM3 oligomerization is not as crucial for the restriction of HIV-1 infectivity as it is for the restriction of incoming virus in target cells. De novo HIV-1 assembly occurs at the plasma membrane of virus-producing cells, and this is the site where cellular IFITM3 is incorporated into virions (Compton et al., 2016). Importantly, we noticed that the majority of the FRET signal occurring between IFITM3-YFP and IFITM3-mCherry was detected in intracellular endosomes and other membrane vesicles, while FRET at the cell surface was relatively low. Therefore, IFITM3 oligomers may be more abundant in endosomal membranes and this agrees with our finding that viruses entering cells via fusion with endosomes (such as IAV) are strongly restricted by IFITM3 oligomers. Since the results presented in this manuscript strongly suggest that IFITM3-mediated antiviral activity occurs through oligomerization-dependent membrane stiffening, assessment of how virion-associated IFITM3 impacts the order of the virion membrane will be helpful in determining whether a distinct antiviral mechanism is at play there.

It will also be important to assess how oligomerization-defective mutants of IFITM3 (G95L) impact membrane curvature, another reported consequence of ectopic IFITM3 expression in cells (Li et al., 2013). It has been shown that protein oligomerization of the transmembrane protein Mic10 is essential for its capacity to induce curvature in mitochondrial membranes (Barbot et al., 2015). Furthermore, changes in membrane order are often accompanied by changes in membrane curvature (Vanni et al., 2014; Rangamani et al., 2014; Golani et al., 2019). It is possible that these two alterations to host membranes underlie the restriction of virus fusion by IFITM3. Moreover, amphipathic helices are characterized to interact with and bend membranes (Giménez-Andrés et al., 2018). Here, we show for the first time that the amphipathic helix of IFITM3 is critical for the membrane order enhancement by IFITM3. Since we also show that glycine-95 of the GxxxG motif is also required for membrane order enhancement, our data may suggest that oligomerization ‘activates’ the membrane deforming activity of the amphipathic helix, and as a result, its antiviral potential. It is possible that local insertion of multiple amphipathic helices into stretches of membrane is required for inhibition of virus fusion, and IFITM3 oligomers provide a means to fulfill that requirement.

In addition to mediating dimerization or oligomerization of transmembrane proteins (homomultimerization), GxxxG motifs have also been described to affect the propensity for interaction with other proteins (heteromultimerization) (Teese and Langosch, 2015; Faingold et al., 2012). Therefore, the GxxxG motif may also govern which proteins IFITM3 interacts with and to what extent. IFITM3 has been described to bind with IFITM1 and IFITM2, and it is interesting to consider how IFITM heteromultimers may contribute to antiviral protection of the cell (John et al., 2013). Furthermore, other host proteins have been described to interact, directly or indirectly, with IFITM3. This list includes cholesterol trafficking regulator VAPA and the metalloproteinase ZMPSTE24, two proteins that have been described as essential cofactors for the antiviral effects of IFITM3 (Amini-Bavil-Olyaee et al., 2013; Fu et al., 2017). Since the former is associated with the tendency for IFITM3 to cause cholesterol overload in endosomes, the G95L loss-of-function mutant could be used to rule in or rule out VAPA and cholesterol as players in the antiviral mechanism of IFITM3.

While glycine-95 is critical for the oligomerization and anti-fusion activity of IFITM3, we show that the homologous site in PRRT2, glycine-305, regulates its oligomerization as well. The naturally occurring G305W/R mutations found in patients with neurological dysfunction are known to disrupt PRRT2 activity, and our results here provide novel insight into how loss-of-function occurs. Therefore, the presence of a shared GxxxG motif in IFITM3, PRRT2, and some other CD225 family members suggests that an ancestral CD225-containing protein performed an unknown function that required oligomerization. Interestingly, the GxxxG motif is not intact in human IFITM5 and IFITM10. IFITM5 is involved in bone formation and exhibits some antiviral activity when expressed ectopically (Huang et al., 2011), while the function of IFITM10 is unknown (Smith et al., 2014). We wonder whether all CD225 proteins play roles in regulating membrane fusion processes in cells—only time and further experiments will tell. However, just as the GxxxG motif in IFITM3 may mediate both homo- and heteromultimerization, it has been reported that PRRT2 interacts with cellular fusogens known as soluble N-ethylmaleimide-sensitive factor attachment protein receptors (SNAREs) in a glycine-305-dependent manner (Coleman et al., 2018). Our findings raise the possibility that oligomerization of CD225 proteins results in alteration of protein architecture and the display of novel docking sites for protein-protein interactions, allowing for expansion of their functional repertoire.

Materials and methods

Key resources table
Reagent type
(species) or
resource
DesignationSource or
reference
IdentifiersAdditional
information
Cell line
(Homo-sapiens)
HEK293TATCCCRL-3216
RRID:CVCL_0063
Cell line
(Homo-sapiens)
TZM-blNIH AIDS Reagent Resource8129–442
RRID:CVCL_B478
Strain, strain background (Influenza A Virus)A/PR/8/34 (H1N1)Charles River Laboratories10100781Clarified allantoic fluid
Recombinant DNA reagentpNL4-3
(plasmid)
NIH AIDS Reagent Resource114
Recombinant DNA reagentpNL4-3ΔEnv
(plasmid)
Eric O. Freed
Recombinant DNA reagentpMD2.G (VSV-G)
(plasmid)
Addgene12259
RRID:Addgene_12259
Recombinant DNA reagentpCMV4-BlaM-Vpr
(plasmid)
Addgene21950
RRID:Addgene_21950
Recombinant DNA reagentpReceiver-YFP-IFITM3
and mutants
(plasmid)
This paperYFP fused to IFITM3 at amino terminus
Recombinant DNA reagentpReceiver-mCherry-IFITM3 and mutants
(plasmid)
This papermCherry fused to IFITM3 at amino terminus
Recombinant DNA reagentpReceiver-PRRT2-YFP and mutants
(plasmid)
This paperYFP fused to PRRT2 at carboxy terminus
Recombinant DNA reagentpReceiver-PRRT2-mCherry and mutants
(plasmid)
This papermCherry fused to PRRT2 at carboxy terminus
Recombinant DNA reagentpQCXIP-FLAG-IFITM3
and mutants
(plasmid)
Compton et al., 2014 and this paperFLAG fused to IFITM3 at amino terminus
Recombinant
DNA reagent
EEA1-GFP
(plasmid)
Addgene42307
RRID:Addgene_42307
Commercial assay or kitLiveBLazer FRET-B/G Loading Kit with CCF2-AMThermo FisherK1032
Chemicalcompound, drugAmphotericin BSigmaC4951
Chemical compound, drugFliptRSpirochromeCY-SC020
Chemical compound, drugLaurdanInvitrogenD250
Chemical compound, drugCholesterol, water-solubleSigmaA2942
Chemical compound, drugMethyl-beta-cyclo-dextrinSigmaC4555
AntibodyAnti-IAV NP mouse monoclonalAbcamAA5H1:500 (flow)
AntibodyAnti-p24 CA mouse monoclonalNIH AIDS Reagent Resource35371:1000 (WB)
AntibodyAnti-p24 CA
[KC57-FITC] mouse monoclonal
BDCO66046651:500 (flow)
AntibodyAnti-CD63
[MX-49.129.5] mouse monoclonal
Santa Cruz Biotechnologysc-52751:400 (IF)
AntibodyAnti-FLAG
[M2] mouse monoclonal
SigmaF18041:1000 (WB)
1:500 (flow)
AntibodyAnti-IFITM3
[EPR5242] rabbit monoclonal
Abcamab1094291:1000 (WB)
1:200 (IF)
AntibodyAnti-Env gp120b sheep polyclonalNIH AIDS Reagent Resource2881:1000 (WB)
AntibodyAnti-Env gp41
[2F5] human monoclonal
NIH AIDS Reagent Resource14751:1000 (WB)
AntibodyAnti-Actin
[C4] mouse monoclonal
Santa Cruz Biotechnologysc-477781:1000 (WB)
AntibodyAnti-c-Myc rabbit monoclonalSigmaC39561:1000 (WB)

Sequence retrieval and alignments

Request a detailed protocol

Protein sequences for CD225 proteins (including IFITM proteins, PRRT2, and TUSC5) were retrieved from UniProt and multi-sequence alignments were performed with ClustalX.

Cell lines and plasmids

Request a detailed protocol

HEK293T (ATCC: CRL-3216) and TZM-bl (NIH AIDS Reagent Resource: 8129) and any derivatives produced in this study were cultivated at 37°C and 5% CO2 in DMEM (Gibco) complemented with 10% fetal bovine serum (Hyclone) and 1% penicillin-streptomycin (Gibco) and regularly passaged with the aid of Trypsin-EDTA 0.05% (Gibco). All cell lines tested negative for mycoplasma. Retroviral pQCXIP vectors encoding IFITM3 fused with amino-terminal FLAG were previously described (Compton et al., 2016; Jia et al., 2012). Retroviral pQCXIP vectors encoding IFITM3 fused with amino-terminal myc were produced by appending myc by PCR and cloning into BamH1/EcoR1 sites. pReceiver constructs encoding IFITM3 tagged with YFP or mCherry at the amino terminus were produced by Genecopeia. IFITM3 tagged with YFP or mCherry internally after residue 40 were produced by Integrated DNA Technologies (in these constructs, YFP or mCherry flanked on both sides by a flexible linker (GGGSGG) was inserted after residue 40 and residues 41 and 42 of IFITM3 were deleted, as performed in Suddala et al., 2019). PRRT2 tagged with carboxy-terminal YFP or mCherry were produced by Integrated DNA Technologies. Mutations in IFITM3 and the G305W mutation in PRRT2 were introduced by site-directed mutagenesis (QuikChange Lightning) or by ligation-independent cloning. HEK293T cell lines stably expressing pQCXIP plasmids were produced by transfecting 250,000 cells in a 12-well plate with 0.8 μg DNA using Lipofectamine2000 (Invitrogen) and selecting with puromycin at a concentration of 10 μg/mL for at least 2 weeks.

Virus productions, infections, and virus-cell fusion assay

Request a detailed protocol

Influenza A Virus [A/PR/8/34 (PR8), H1N1] supplied as clarified allantoic fluid was purchased from Charles River Laboratories. Infectious virus titers were calculated using a flow cytometry-based method in HEK293T cells (Grigorov et al., 2011), and infections were performed as follows: HEK293T cells (either stably expressing or transiently transfected with 1.5 μg empty pQCXIP or IFITM3 WT or the indicated mutant fused with amino terminal FLAG) were seeded in 24-well plates (50,000 per well) overnight and overlaid with indicated amounts of virus diluted in 225 μL of complete DMEM for approximately 18 hr. Cells were washed with 1X PBS, detached with Trypsin-EDTA, fixed/permeabilized with Cytofix/Cytoperm (BD), immunostained with anti-IAV NP (AA5H; Abcam), and analyzed on a LSRFortessa flow cytometer (BD). Replication-incompetent HIV-1 pseudotyped with IAV WSN HA and NA was produced by transfecting HEK293T with 2 μg pR8ΔEnv, 1 μg pcRev (NIH AIDS Reagent Resource: 11415), 3 μg Gag-GFP, 1.5 μg of hemagglutinin, and 1.5 μg of neuraminidase from IAV WSN strain, H1N1 (gifts from G. Melikyan). Replication-incompetent HIV-1 pseudotyped with VSV-G for virus-cell fusion assays was produced by transfecting HEK293T with 15 μg pNL4-3ΔEnv, 5 μg pCMV4-BlaM-Vpr, and 5 μg pMD2.G (VSV-G). Replication-competent HIV-1 was produced by transfecting HEK293T with 15 μg pNL4-3 and 5 μg empty pQCXIP or pQCXIP encoding IFITM3 WT or the indicated mutant fused with amino-terminal FLAG. Transfections were performed using the calcium-phosphate method. Briefly, 6 million HEK293T were seeded in a T75 flask. Plasmid DNA was mixed with sterile H2O, CaCl2, and Tris-EDTA (TE) buffer, and the totality was combined with Hepes-buffered saline (HBS). The transfection volume was added dropwise, and cells were incubated at 37°C for 48 hr. Supernatants were clarified by centrifugation, passed through a 0.45 μm filter, and concentrated by ultracentrifugation through a 20% sucrose cushion at 25,000 x g for 1 hr at 4°C. Lentivirus titers were measured using an HIV-1 p24 ELISA kit (XpressBio). To measure infectivity of virus supernatants, 50 ng p24 equivalent volumes were added to TZM-bl cells and cells were fixed/permeabilized with Cytofix/Cytoperm (BD) at 48 hr post-infection, immunostained with anti-Gag KC57-FITC (BD), and analyzed by flow cytometry. To measure virus-cell fusion, 100–300 ng p24 equivalent HIV-1-VSV-G produced with pCMV4-BlaM-Vpr was added to HEK293T cells stably expressing empty pQCXIP, IFITM3 WT or the indicated mutant for 2.5 hr at 37°C. Cells were washed with CO2-independent medium and incubated with CCF2/AM mix containing probenecid for one hour at room temperature in the dark. Cells were washed with cold 1X PBS, fixed/permeabilized with Cytofix/Cytoperm (BD), and analyzed on a LSRII flow cytometer.

Confocal immunofluorescence microscopy

Request a detailed protocol

HEK293T cells were seeded in μ-slide eight-well chambers (Ibidi) (30,000 per well) overnight and transfected with 0.2 μg EEA1-GFP (Addgene: 42307) and 0.2 μg empty pQCXIP or pQCXIP-WT IFITM3 or the indicated mutant fused with amino-terminal FLAG using Lipofectamine2000 (Invitrogen). At 48 hr post-transfection, cells were fixed/permeabilized with Cytofix/Cytoperm and immunostained with anti-CD63 antibody (MX-49, sc-5275; Santa Cruz Biotechnology) and anti-IFITM3 antibody (EPR5242, ab109429; Abcam). Cells were imaged using the Leica TCS SP8 confocal microscope with a 63X objective and oil immersion and analysis was performed in Fiji (ImageJ). Colocalization analysis was performed with the Colocalization tool in Imaris (Bitplane) as follows: a region of interest was manually created to single out cells which were positive for EEA1-GFP and colocalization between IFITM3/EEA1-GFP and IFITM3/CD63 was measured in 3D using confocal stacks of 20–30 optical sections.

Immunoprecipitation, SDS-PAGE, and western blot analysis

Request a detailed protocol

HEK293T cells were transfected with 1.5 μg empty pQCXIP or a combination of 0.75 μg pQCXIP-FLAG-IFITM3 and 0.75 μg pQCXIP-myc-IFITM3 (encoding either WT or G95L). At 48 hr post-transfection, cells were lysed in a buffer containing 0.5% IPEGAL (Sigma), 50 mM Tris (pH 7.4), 150 mM NaCl, and 1 mM EDTA. Immunoprecipitation was performed using anti-FLAG M2 magnetic beads (Sigma) for a period of 3 hr with rotation at 4°C. Magnetic beads were isolated with a DynaMag-2 magnet (Thermo Fisher) and washed prior to addition of 1X NuPAGE LDS sample buffer and 1X NuPAGE Sample Reducing Reagent (Invitrogen). The non-immunoprecipitated whole cell lysates and immunoprecipitated fractions were heat denatured at 90°C for 15 min and 12 μL and 5 μL, respectively, were loaded into Criterion XT 12% Bis-Tris polyacrylamide gels (Bio-Rad) for SDS-PAGE using NuPAGE MES SDS Running Buffer (Invitrogen). Proteins were transferred Amersham Protran Premium Nitrocellulose Membrane, pore size 0.20 μm (GE Healthcare). Membranes were blocked with Odyssey blocking buffer in PBS (Li-COR) and incubated with anti-FLAG M2 (F1804; Sigma) and anti-c-Myc (C3956; Sigma). Secondary antibodies conjugated to DyLight 800 or 680 (Li-COR) and the Li-COR Odyssey imaging system were used to reveal specific protein detection. Images were analyzed and assembled using ImageStudioLite (Li-COR). To compare the steady-state expression levels of WT IFITM3 and mutants, HEK293T cells were transfected with 1.5 μg of empty pQCXIP or pQCXIP-FLAG-IFITM3 (encoding WT or the indicated mutants) and expression was measured using anti-FLAG M2 and analysis by flow cytometry or western blot analysis. For the former, cells were fixed/permeabilized with Cytofix/Cytoperm (BD) at 48 hr post-transfection, immunostained with anti-FLAG M2 (F1804; Sigma) and analyzed on a LSRFortessa (BD). For the latter, cells were lysed and separated by SDS-PAGE as indicated above and immunoblotting was performed with anti-FLAG M2 (F1804; Sigma) and anti-actin (C4, sc-47778; Santa Cruz Biotechnology). For blotting of HIV-1 proteins in cell lysates and concentrated virus supernatants, cell/virion lysis was performed with radioimmunoprecipitation (RIPA) buffer (Thermo Fisher) supplemented with Halt Protease Inhibitor mixture EDTA-free (Thermo Fisher) or 0.01% Triton-X (Sigma), respectively, and lysates were complemented with 1X NuPAGE LDS sample buffer and 1X NuPAGE Sample Reducing Reagent (Invitrogen) prior to heat denaturation at 90°C for 15 min. Samples were migrated and transferred as indicated above and immunoblotting was performed with the following antibodies: anti-p24 CA (NIH AIDS Reagent Resource: 3537), anti-gp120b (NIH AIDS Reagent Resource: 288), anti-gp41 2F5 (NIH AIDS Reagent Resource: 1475), anti-actin (C4, sc-47778; Santa Cruz Biotechnology), and anti-FLAG M2 (F1804; Sigma).

Blue native PAGE

Request a detailed protocol

HEK293T cells were seeded in 6-well plates (750,000 per well) overnight and transfected with 1 μg empty pQCXIP or pQCXIP-FLAG-IFITM3 (either WT or G95L). At 48 hr post-transfection, cells were lysed with 1X NativePAGE Sample Buffer (Invitrogen) containing 1% digitonin. Lysates were mixed with 5% Coomassie G-250 at a volume to volume ratio of 20:1. Approximately 20 μg of protein per sample was loaded into a NativePAGE Novex 4–16% Bis-Tris polyacrylamide gel (Invitrogen) according to manufacturer’s instructions. A NativeMark Unstained Protein Standard was loaded as reference ladder. Following PAGE, proteins were transferred to Immobile FL PVDF membrane (EMD Millipore) and immunoblotting was performed with anti-FLAG M2 (F1804; Sigma). A secondary antibody conjugated to DyLight 800 or 680 (Li-COR) and the Li-COR Odyssey imaging system were used to reveal specific protein detection. Images were analyzed and assembled using ImageStudioLite (Li-COR).

FRET and FLIM for oligomerization studies

Request a detailed protocol

HEK293T cells were seeded in μ-slide eight well chambers (Ibidi) (50,000 per well) overnight and transiently co-transfected with 0.25 μg IFITM3-YFP and 0.25 μg mCherry or 0.25 μg IFITM3-YFP and 0.25 μg IFITM3-mCherry using TransIT-293 (Mirus). In parallel experiments, pairs of plasmids encoding PRRT2-YFP and PRRT2-mCherry were co-transfected. Living cells in Fluorobrite DMEM (Gibco) were imaged with a Zeiss LSM 780 confocal microscope using a 63X objective and oil immersion. To assess FRET, donor YFP fluorescence was detected with a gallium arsenide phosphide photomultiplier tube (GaAsP PMT) with a 520–550 nm emission window following excitation by a 514 nm laser. Acceptor mCherry fluorescence was detected with a GaAsP PMT detector with a 570–615 nm emission window following excitation by a 561 nm laser. FRET signal (acceptor mCherry fluorescence triggered by excitation of donor YFP) was collected with a GaAsP PMT detector with 570–615 nm emission window after excitation with a 514 nm laser. At least 50 cells per condition were examined in each experiment. Each cell was assigned a FRET index calculated using the FRET and colocalization analyzer plugin for Fiji (ImageJ). FLIM analysis of donor YFP was performed by excitation with a 950 nm two-photon, pulsed laser (Coherent) tuned at 80 MHz with single photon counting electronics (Becker Hickl) and detection with a HPM-100–40 module GaAsP hybrid PMT (Becker Hickl). Analysis was limited to cells exhibiting 250–1000 photons per pixel to mitigate the effects of photobleaching and low signal to noise ratio. SPCImage NG software (Becker Hickl) was used to acquire the fluorescence decay of each pixel, which was deconvoluted with the instrument response function and fitted to a Marquandt nonlinear least-square algorithm with two exponential models. The mean fluorescence lifetime was calculated as previously described (Sun et al., 2011) using SPCImage NG. At least 30 cells per condition were analyzed in each experiment.

Number and brightness analysis

Request a detailed protocol

HEK293T cells were seeded in μ-slide eight-well chambers (Ibidi) (50,000 per well) overnight and transiently with 0.50 μg IFITM3-mCherry using TransIT-293 (Mirus). Living cells in Fluorobrite DMEM (Gibco) were imaged with a Zeiss LSM 780 confocal microscope using a 63X objective and oil immersion. Regions of interest were limited to portions of cells which were immobile and which focused on plasma membrane fluorescence (intracellular signal from vesicular membranes was excluded). The axial position of a specimen during acquisition was stabilized using the Adaptive Focus Control module. mCherry was excited with a 561 nm laser and detected with a 570–615 emission window. For each cell, 100 frames were acquired at a frame rate of 0.385 frames per second with a 9.75 μs pixel dwell time and pixel size of 151.38 nm. Images were always acquired at 256 × 256 pixels such that the pixel size remained three to four times smaller than the volume of the point-spread function. Photobleaching of fluorescent proteins during data acquisition was corrected using a detrending algorithm (Nolan et al., 2017). Twenty cellular regions were examined per condition. Pixel-by-pixel brightness values were calculated in Fiji (ImageJ).

FLIM for study of membrane order with FliptR

Request a detailed protocol

HEK293T stably expressing empty pQCXIP or pQCXIP-WT IFITM3 or the indicated mutant were seeded in μ-slide eight-well chambers (Ibidi) (50,000 per well) overnight and stained with 1 μM FliptR (Spirochrome) for 5 min according to the manufacturer’s protocol. Imaging was performed with a 63X objective under oil immersion on a Leica SP8-X-SMD confocal microscope. When indicated, cells were treated with 100 μg/mL soluble cholesterol (C4951, Sigma), 5 mM methyl-cyclo-beta-dextrin (MßCD) (C4555, Sigma), or 1 μM Amphotericin B (A2942, Sigma) for 1 hr prior to addition of FliptR and imaging. Fluorescence was detected by hybrid external detectors in photon counting mode following excitation by a 488 nm pulsed laser turned to 20 MHz with single photon counting electronics (PicoHarp 300). Analysis was limited to cells with at least 250–1000 photons per pixel to mitigate the effects of photobleaching and low signal-to-noise ratio. Fluorescence decay of each pixel in FliptR-stained cells was acquired by Symphotime 64 software (Picoquant), and deconvoluted with the instrument response function and fitted to a Marquandt nonlinear least-square algorithm with two exponential models. The mean fluorescence lifetime (τ), in addition to individual component lifetimes (long and short lifetimes, τ1 and τ2), were calculated using Symphotime. At least 30 cells per condition were analyzed in each experiment.

Laurdan staining for study of membrane order

Request a detailed protocol

The membrane probe Laurdan (6-dodecanoyl-2-dimethylamino naphthalene, D250, Invitrogen) was dissolved in DMSO to create a stock solution of 9.19 mM. HEK293T cells were incubated with 1.8 µM Laurdan for 1 hr at 37°C. Generalized polarization (GP) is a ratiometric method which is used to quantitatively report membrane order in living cells. The GP value is calculated as follows:

Generalizedpolarization=IblueIgreenIblue+Igreen

where Iblue and Igreen are the fluorescence intensities emitted at 440 nm and 490 nm, respectively. Conventionally, 440 nm and 490 nm are the emission maximums for ordered lipid and disordered lipid bilayers, respectively. Images were acquired with a LSM780 (Zeiss) laser scanning microscope using a 63X oil immersion objective, coupled to a two-photon Ti:Sapphire laser (Coherent) tuned to 780 nm and 80 MHz. An SP 760 nm dichroic filter was used to separate laser light from fluorescence signal. The fluorescence signal was acquired from 416 nm to 474 nm for the blue channel and from 475 nm to 532 nm for the green channel using the GaAsP PMT detectors of the LSM780. GP values were calculated in Fiji (ImageJ) as previously described (Rentero et al., 2019). GP values were calculated for each cell of interest and the numeric difference in GP values for the whole cell are presented normalized to empty vector (set to 0).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83
  84. 84

Decision letter

  1. Olga Boudker
    Senior Editor; Weill Cornell Medicine, United States
  2. Mark Marsh
    Reviewing Editor; University College London, United Kingdom
  3. Mark Marsh
    Reviewer; University College London, United Kingdom
  4. Camilla Benfield
    Reviewer

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Interferon induced transmembrane (IFITM) proteins are a subgroup of the dispanin (CD225) domain-containing protein family. IFITMs inhibit the replication of a broad range of enveloped viruses, primarily by inhibiting membrane fusion. The authors identify a motif in the dispanin domain that mediates IFITM3 oligomerization and show that oligomerization correlates with IFITM3-mediated reduction in membrane fluidity and viral restriction. Significantly, a naturally occurring mutation in a similar motif in another dispanin family protein, proline rich transmembrane protein 2 (PRRT2), a neuron-specific regulator of neurotransmitter release, is linked to paroxysmal neurological disorders.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Homology-guided identification of a conserved motif linking the antiviral functions of IFITM3 to its oligomeric state" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Mark Marsh as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by a Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Camilla Benfield (Reviewer #3).

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

While the reviewers all acknowledge that the mechanisms underlying IFITM inhibition of virus membrane fusion remain unclear, the reviewers point out fundamental concerns with the constructs you have used, i.e. the miss-sorting and loss of functional activity seen by other labs using FP-tagged IFITM proteins. The concerns of the reviewers are detailed below and I hope you find these useful as you develop the work.

Reviewer #1:

Interferon induced transmembrane (IFITM) proteins inhibit the entry of a broad range of enveloped viruses. Focusing mainly on IFITM3, work from a number of labs has suggested that IFITM expression influences membrane stiffness/order, thereby inhibiting viral fusion. Although generally accepted, the mechanism through which IFITMs in general, and IFITM3 in particular, influence membrane stiffness is not fully understood. The corresponding author has previously shown that an amphipathic helix in the conserved CD225/dispannin domain, which is believed to insert into the cytoplasmic leaflet of membranes containing IFITM3, is required for antiviral activity. In this paper, the authors show that a CD225 domain associated GxxxG motif, that in other proteins is referred to as a “glycine zipper”, is required for IFITM3 oligomerization and antiviral activity. Moreover, they show that expression of IFITM3 increases membrane order and that IFITM3 proteins containing mutations of the glycines in the GxxxG motif show reduced oligomer formation and reduced impact on membrane order. Interestingly, they find that another CD225 domain containing protein (PRRT2), that regulates synaptic vesicles, also appears to form oligomers and that disease associated mutations of the PRRT2 GxxxG motif also impact on oligomer formation.

Overall this is an interesting study that supports the notion that IFITM3 inhibits virus fusion by modulating the stiffness of membranes, and that this modulation is associated with the oligomerization mediated at least in part through the GxxxG motif. Overall, I am supportive of publication though I have some concerns with aspects of the data that the authors need to consider.

Figure 2: All data is expressed as relative numbers; the authors should give actual numbers for at least some of the infectivity experiments. Do the authors know why IFITM3 restriction of flu is less effective in stable transfectants that transients? Is there a difference in the level of IFITM3 expression? If so, how does this impact on experiments measuring IFITM3 oligomerization and membrane order? These latter experiments are all done with transiently expressed IFITM3.

Figure 2SB; Previous published work has indicated the bulk of IFITM3 is associated with endosomes. Here it seems that most of the expressed IFITM3-FP is at the cell surface, and very little is in endosomal compartments. Why is this? Weston et al. (PLoS One 2014) previously showed that C terminal tags on IFITM proteins, which in the case of IFITM3 would be exposed to the endosomal lumen, could be proteolytically cleaved. Could the FP tags used here be subject to cleavage in endosomes?

Subsection “Glycine-95 is important for the antiviral functions of IFITM3”; Can the authors explain why the use of an IAV/HIV pseudotype virus reports on early entry events?

Figure 3; In contrast to the distribution of IFITM3 shown in Figure 3, most of the FRET signal reporting oligomerization appears to be on intracellular membranes. Are the FP-tagged proteins trafficking differently, or is oligomerization associated with specific sub-cellular compartments?

Subsection “GxxxG regulates oligomerization of IFITM3 in living cells and bulk lysates”; Can the authors explain what they mean by “diffusion in and out of a static membranous compartment (the plasma membrane)”. The plasma membrane is not a static compartment and how would IFITM3-mCherry be diffusing in and out of this compartment?

Is it valid to compare the intensity of pixels in a 2D array of membrane-associated IFITM3-mCherry with cytoplasmic mCherry?

Reviewer #2:

This is a beautifully written manuscript that describes a functional oligomerization motif within the important innate antiviral immunity protein IFITM3. Despite the excellent writing, the study is generally limited in scope as it essentially boils down to characterization of a single point mutant of IFITM3. Novelty of the paper is also low as the residue in question (G95) has already been shown to be important for IFITM3 activity (John et al., 2013). Furthermore, most of the conclusions of the manuscript are also already known, including that IFITM3 dimerization is essential for activity (John et al., 2013), that IFITM3 increases membrane stiffness (Li, PLOS Path, 2013), and that antiviral function of IFITM3 is countered by Amphotericin (Lin, Cell Reports, 2013). There are also concerns with some of the reagents and much of the data presentation.

1) Fluorescent protein fusions of IFITM3 have been generated and tested by multiple laboratories and none were found to be active regardless of the termini at which the protein was fused or the fluorescent protein that was used. This was such an impediment to live cell imaging of IFITM3 that the top labs in the field have gone to extraordinary measures to visualize IFITM3 in live cells (Spence et al., 2019; Peng, J Am Chem Soc, 2016; Suddala et al., 2019). It is not possible that the fluorescent fusion proteins used in this paper are antivirally active as presented in Figure 3.

2) All of the virus infectivity data is presented as normalized infectivity, and no raw data flow cytometry plots are shown. For influenza infections, an MOI of 0.1 is used. Thus, the maximum % infection would be at most ~7%, providing a minimal dynamic range for observing antiviral effects of IFITM3 or for comparing IFITM3 mutants. This minimal dynamic range, along with the strange results with fluorescent fusions, coupled with unnecessary obfuscation of the primary data, overall decreases confidence in some of the results.

Reviewer #3:

Previously, F75 and F78 were thought to link oligomerisation to antiviral restriction, but recently this was refuted by Winkler et al., 2019, who found no effect of these residues on IFITM3-IFITM3 interactions. Since most previous work used these F75/78 mutants, discovering the determinants of IFITM3 oligomerisation and its relation to antiviral activity remains important and not yet answered.

The authors report that Gly-95 of human IFITM3 reduces restriction, oligomerisation and membrane rigidification by IFITM3.

However, aspects of these conclusions need further support, as I have detailed below. I also feel that the novelty and conceptual advance is likely not sufficient for publication by eLife, but rather incremental. IFITM3 has already been shown to increase membrane order (Li et al., 2013) and G95R was previously shown to attenuate restriction of IAV (John et al., 2013).

1) Do the authors, as per the title, link the motif or AA reside 95 to oligomerisation/restriction?, vast majority of data pertains to G95L – see below.

2) The Introduction states that the GXXXG motif is also known as a glycine zipper whereas my understanding is that the most significant glycine zipper sequence patterns are (G,A,S)XXXGXXXG and GXXXGXXX(G,S,T), which contain a GXXXG motif (https://www.pnas.org/content/102/40/14278, not referenced in the current paper)

3) I do not feel this is an evolution-guided analysis (title of the first Results section), which to me would imply analysis of evolutionary selection pressures. I am more comfortable with “homology-guided”, as per the title. I also do not follow the logic of the phylogenetic tree of CD225 domains in 1A “indicative of a common ancestry”, all CD225 superfamily proteins have this domain, by definition. Not sure what the tree adds. Authors also say that the GXXXG motif is conserved in vertebrate IFITM3s but (i) So are many other residues/ AA motifs conserved in the highly conserved CD225 domain of vertebrate IFITM3, as the alignment Figure 1D shows, and (ii) surely to make a point about conservation among vertebrates, an alignment where 5 of the 7 mammalian IFITM3 sequences shown are from primates, i.e. the same mammal order, is less informative than an alignment with wider range of mammal or vertebrate species whose IFITM3 has been sequenced. Do IFITM5 and IFITM10 oligomerise, as they lack an intact GXXXG? (not commented on)

4) Although authors say they focused on G95L and G95W mutants for further functional characterisation (subsection “Glycine-95 is important for the antiviral functions of IFITM3”), no localisation data is shown for G95W and for some reason, G95W is omitted from panels D and F in Figure 2, which shows effects on antiviral restriction. Later, oligomerisation is not assessed for G95W (Figures 3 and 4) and yet despite the lower expression levels noted for G91L by transient transfection (S2A), this G91L is reinserted into the analysis of oligomerisation (Figure supp 3B and interpretation lines subsection “GxxxG regulates oligomerization of IFITM3 in living cells and bulk lysates”). The authors do not explain the logic for this “pick and choose” approach.

5) I do not find that the localisation data which is presented for WT and G95L to be convincing (i.e. Figure S2B; IF performed for transient IFITM3 expression in EEA1-GFP overexpressing cells). The authors state their IF shows IFITM3 distributed in early and late endosomes and the PM (subsection “Glycine-95 is important for the antiviral functions of IFITM3”), whereas the staining to me seems very dominantly PM with minimal endosomal staining? Why is there no IF done on the stable cell lines which much of the restriction data in Figure 2 pertain to (panels B, D, E). Further, since numerous studies have shown localisation to be a key feature/correlate of IFITM3 restriction, stronger data confirming “similar subcellular localisation” of the Glycine mutants should be presented (both G95L and G95W).

6) It is known that expression levels of IFITM3 affect its antiviral restriction. For transient transfection experiments, authors provide a single western blot panel (Figure supplementary 2A) which indicates the two Gly-95 mutants express similarly to WT when transiently expressed. However, it would be better to perform densitometric quantification of replicate independent blots, and/ or provide western panels to shown expression in the same experiments as restriction is assessed following transient transfection. For the stable cell lines, FACS analysis is shown (Figure 2A) which indicates that the mutants are less well expressed in this system. First, I think that replicates should have been analysed in order to show whether there is any statistically significant difference between the cell lines. (Currently only the MFI for a single replicate for each cell line is given: WT=814, G95W=590 , G95L=605). Secondly, the authors need to comment on whether this expression level difference might influence the interpretation of the restriction data obtained for the cell lines, and indeed the other phenotypes seen.

7) Authors state that late stage inhibition of HIV-1 entry is regulated by glycine 95 (subsection “Glycine-95 is important for the antiviral functions of IFITM3”), these data are shown in a single bar chart in Figure 2F and a composite supplementary Figure 2C. In supplementary Figure 2C, the authors should explain how the quantification relates to the Western blot panels- presumably all the quantification graphs refer to the right-hand panels of the westerns labelled “virus supernatants”, but this needs to be clarified. Furthermore, could the (arguably small) differences seen for wt vs. glycine mutant in right-hand panel (supernatants) not be a consequence of the differences seen in the left-hand panel (i.e. expression levels in whole cell lysates), the latter potentially due to the effects shown for IFITM3 on HIV protein synthesis (Lee et al., 2018, “IFITM proteins inhibit HIV-1 protein synthesis”)? It is not clear to me that these data wholly support the authors interpretation that late stage inhibition of HIV-1 entry is regulated by glycine 95 and the conclusion that “oligomerization-defective G95L mutant lessened the impact of virion-associated IFITM3 on HIV-1, suggesting that IFITM3 oligomers are needed to maximally reduce virion infectivity”.

8) Figure 3 (on oligomerisation) is confusing, the circular symbols are not explained and I was unclear what the 3 columns represented and whether the labels at the top (mCherry, IFITm3-YFP and IFITM3-mCherry) pertained to the columns (in which case they need to be centered over the columns) or were a legend. Accordingly, I could not interpret these data or the bar charts below the IF which, again, used these undefined circular symbols. The figure legend refers to filled red and filled yellow circles, which I could not see on the figure. I have the same problem interpreting Figure 6 (analagous experiments with PRRT2).

9) The co-IP and native PAGE data lack loading controls (Figure 4) and I was also unclear why i. the heterologous pairs had not been tried for co-IPs experiments ie WT with G95L, ii. there was no replicates or quantitation for the native PAGE (i.e. an equivalent panel as for 4B).

10) Using the FliptR system, authors report that “membrane order enhancement tracks with a functionally competent form of IFITM3 but not with a loss-of-function mutant”. It would have been useful to assess one of the many other loss of function IFITM mutants that have been described in the literature, e.g. the S-palmitoylation defective mutants or an amphipathic helix-lacking mutant (especially since some of the same authors discovered IFITM3's amphipathic helix). Further, while positive and negative controls for membrane order were used (Figure 6—figure supplement 1), it would help strengthen the data on membrane rigidity if authors had used an additional method, especially one that has been used previously to study IFITM3 e.g. Laurdan.

https://doi.org/10.7554/eLife.58537.sa1

Author response

[Editors’ note: The authors appealed the original decision. What follows is the authors’ response to the first round of review.]

While the reviewers all acknowledge that the mechanisms underlying IFITM inhibition of virus membrane fusion remain unclear, the reviewers point out fundamental concerns with the constructs you have used, i.e. the miss-sorting and loss of functional activity seen by other labs using FP-tagged IFITM proteins. The concerns of the reviewers are detailed below and I hope you find these useful as you develop the work.

Reviewer #1:

Overall this is an interesting study that supports the notion that IFITM3 inhibits virus fusion by modulating the stiffness of membranes, and that this modulation is associated with the oligomerization mediated at least in part through the GxxxG motif. Overall, I am supportive of publication though I have some concerns with aspects of the data that the authors need to consider.

Figure 2; All data is expressed as relative numbers; the authors should give actual numbers for at least some of the infectivity experiments. Do the authors know why IFITM3 restriction of flu is less effective in stable transfectants that transients? Is there a difference in the level of IFITM3 expression? If so, how does this impact on experiments measuring IFITM3 oligomerization and membrane order? These latter experiments are all done with transiently expressed IFITM3.

In order to render the results more transparent, we have incorporated many examples of raw data throughout the manuscript. Regarding IAV infection, we have included raw infection data as measured by FACS in Figure 2C. For clarity, we removed the panel from Figure 2 showing IAV infection in cells transiently expressed IFITM3 constructs. Now, only the results in stably expressing cell lines are shown there, and these results are more than adequate to show that mutation at G91 and G95 impair anti-IAV activity. To compare the expression levels in transiently-transfected and stably-transfected HEK293T cells, please see Figure 2A and 2B and Figure 1—figure supplement 1A and 1B. In the stable cell lines, IFITM3 constructs are expressed uniformly and at higher levels relative to transient transfections. To address the question of how transient vs stable expression of IFITM3 impact oligomerization and membrane order, we have introduced new data measuring oligomerization and membrane order following transient AND stable expression of IFITM3. The antiviral activity of IFITM3 against IAV is found in Figure 2D (stable) and Figure 4—figure supplement 2A and B (transient). The effect of IFITM3 on membrane order is found in Figure 6B (stable) and Figure 6—figure supplement 1E (transient). Importantly, whether IFITM3 was expressed transiently or stably did not impact the antiviral activity/membrane rigidification by IFITM3 WT nor did it impact the loss of those effects by mutation of the GxxxG motif.

Figure 2SB; Previous published work has indicated the bulk of IFITM3 is associated with endosomes. Here it seems that most of the expressed IFITM3-FP is at the cell surface, and very little is in endosomal compartments. Why is this? Weston et al. (PLoS One 2014) previously showed that C terminal tags on IFITM proteins, which in the case of IFITM3 would be exposed to the endosomal lumen, could be proteolytically cleaved. Could the FP tags used here be subject to cleavage in endosomes?

Thank you for raising this important issue regarding the subcellular localization of IFITM3 and its mutants. The data presented in former Supplemental Figure 2B was actually that of IFITM3 tagged with FLAG at the N-terminus, not the C-terminus. Furthermore, the fluorescent protein constructs (mCherry and YFP) were also placed at the N-terminus of IFITM3. As the reviewer points out, Weston et al. previously showed that C-terminal tags could be cleaved and we specifically avoided C-terminal tagging for that reason. Anyway, we have now exhaustively tested whether our FLAG-tagged IFITM3 constructs differ with regards to subcellular localization. We have included new immunofluorescence data in Figure 2—figure supplement 1A (stable expression) and 2B (transient expression) and performed quantitative colocalization analysis between IFITM3 and early endosomal marker EEA1 or late endosomal/MVB marker CD63. This analysis shows that a significant amount (50-60%) of IFITM3 stably expressed in HEK293T is present in early and late endosomes, and this proportion was not significantly altered by mutations in the GxxxG motif. In stable cell lines, IFITM3 WT and mutants were more abundant in CD63+ MVB than they were under transient transfection conditions (Figure 2—figure supplement 1A (stable expression) and 2B (transient expression)). Therefore, these data demonstrate that a general loss of endosomal localization is not responsible for the differential antiviral functions exhibited by IFITM3 WT and GxxxG mutants.

Subsection “Glycine-95 is important for the antiviral functions of IFITM3”; Can the authors explain why the use of an IAV/HIV pseudotype virus reports on early entry events?

We have removed these data from Figure 2 as it was unnecessary. Instead, we highlight our data using HIV-VSV-G incorporating BlaM-Vpr in Figure 2E as a readout for viral access to the cytoplasm (entry/fusion).

Figure 3; In contrast to the distribution of IFITM3 shown in Figure 3, most of the FRET signal reporting oligomerization appears to be on intracellular membranes. Are the FP-tagged proteins trafficking differently, or is oligomerization associated with specific sub-cellular compartments?

The reviewer raises an interesting observation. The data reporting FRET and YFP lifetimes of FP-tagged IFITM3 are now shown in Figure 4A. The distribution of YFP lifetimes (lower row) indicate that YFP signal is present at the cell surface and intracellular compartments (endosomes/other vesicles), suggesting that FP-tagged IFITM3 are not trafficking differently compared to the FLAG-tagged constructs analyzed in Figure 2—figure supplement 1. However, the lowest YFP lifetimes in cells, indicative of sites where FP oligomerization is occurring, are those present in endosomes/vesicles. This is most apparent in the WT condition, where the largest drop in YFP lifetime per cell occurs. Accordingly, the FRET signal is also concentrated in intracellular compartments and less FRET is detectable on the cell surface. These data may suggest that IFITM3 oligomers are mostly localized to intracellular membranes such as endosomes while fewer IFITM3 oligomers are present on the cell surface. Since this point may be relevant to the various functions performed by IFITM3 in health and disease, we now discuss these results; “Importantly, we noticed that the majority of the FRET signal occurring between IFITM3-YFP and IFITM3-mCherry was detected in intracellular endosomes and other membrane vesicles (Figure 4). In contrast, FRET at the cell surface was relatively low. Therefore, IFITM3 oligomers may be more abundant in endosomal membranes and this agrees with our finding that viruses entering cells via fusion with endosomes (such as IAV) are strongly restricted by IFITM3 oligomers.”

Subsection “GxxxG regulates oligomerization of IFITM3 in living cells and bulk lysates”; Can the authors explain what they mean by “diffusion in and out of a static membranous compartment (the plasma membrane)”. The plasma membrane is not a static compartment and how would IFITM3-mCherry be diffusing in and out of this compartment?

We thank the reviewer for pointing out this oversight. We changed the text to read “We restricted our analysis to brightness of IFITM3-mCherry as it trafficked in and out of the plasma membrane over time.”

Is it valid to compare the intensity of pixels in a 2D array of membrane-associated IFITM3-mCherry with cytoplasmic mCherry?

The reviewer brings up a good point. Our “empty mCherry” is cytoplasmic and does not contain transmembrane domains, meaning that it does not represent the ideal negative control for studying the specific interaction between proteins in membranes using Number and Brightness. We added text to this section to underline this caveat to the analysis: “One caveat of this Number and Brightness analysis is that, in contrast to transmembrane proteins, mCherry monomers do not target membranes and are expressed mostly in the cytoplasm. Nonetheless, the fact that IFITM3 WT exhibits a brightness that is roughly two-fold greater than IFITM3 G95L supports the notion that the former exists primarily as a dimer and the latter as a monomer.”

Reviewer #2:

This is a beautifully written manuscript that describes a functional oligomerization motif within the important innate antiviral immunity protein IFITM3. Despite the excellent writing, the study is generally limited in scope as it essentially boils down to characterization of a single point mutant of IFITM3. Novelty of the paper is also low as the residue in question (G95) has already been shown to be important for IFITM3 activity (John et al., 2013). Furthermore, most of the conclusions of the manuscript are also already known, including that IFITM3 dimerization is essential for activity (John et al., 2013), that IFITM3 increases membrane stiffness (Li, PLOS Path, 2013), and that antiviral function of IFITM3 is countered by Amphotericin (Lin, Cell Reports, 2013). There are also concerns with some of the reagents and much of the data presentation.

We thank the reviewer for their praise. Concerning the statement that this article “boils down to a single point mutant,” we have added significant amounts of new data to highlight the effects of both G91 and G95 of the GxxxG motif. As a result, data on G91 is now found in Figures 1-4 and the data are less focused on a single mutation. Our new findings indicate that mutation of G95 has a larger impact on protein oligomerization than G91, which justifies our focus on G95 in the last two figures of the paper. We tried to study a double mutant whereby the G91L and G95L mutations were introduced simultaneously into IFITM3, but the polypeptide product was not expressed at detectable levels following transfection into HEK293T (not shown). Overall, our paper describes the functional impact of the GxxxG motif rather than a single point mutation, and this underlines the novelty of our study because the GxxxG motif in IFITM3 has not previously been reported or functionally described.

Additionally, the reviewer has failed to appreciate the other points of novelty in our manuscript:

1) While mutation of G95 was previously shown to impact anti-IAV activity, the reason for loss of function was unknown. Here, we provide the mechanism for loss-of-function.

2) While a previous publication suggested that IFITM3 dimerizes (John et al., 2013), that study implicated F75 and F78 as drivers for oligomerization. Here, we added fresh data confirming that F75/F78 do not play a critical role in oligomerization as measured by our FRET approach (Figure 4—figure supplement 1), which agree with another study utilizing FRET (Winkler et al., 2019). Even though John et al. showed that F75 and F78 are important for antiviral function, the fact that F75 and F78 are dispensable for oligomerization means that the authors did not show that “dimerization is essential for activity.” On the contrary, our results leveraging mutation of the GxxxG motif are the first to indicate that oligomerization is essential for antiviral activity.

3) While we openly agree that previous literature demonstrated that IFITM3 increases membrane stiffness and that Amphotericin B overcomes the antiviral activity of IFITM3, we are the first to directly show that membrane stiffness is directly linked to antiviral function. By using a loss-of-function mutation (G95L), we are the first to show that membrane stiffening is required for antiviral activity. Furthermore, by incorporating Amphotericin B into the same assay, we are the first to show that Amphotericin B negates the impact of IFITM3 on membrane stiffness. Together, these data provide significant novel information about how IFITM3 performs its antiviral activities AND how a drug overcomes these antiviral activities.

4) The study is not “limited in scope,” since our identification of a GxxxG in IFITM3 and other CD225 proteins is indicative of a shared requirement for oligomerization in the diverse membrane fusion processes controlled by the CD225 protein family. We show that mutation of G305W in PRRT2 disrupts its ability to oligomerize, adding fresh insight this mutation that is directly linked to neurological disease in humans.

1) Fluorescent protein fusions of IFITM3 have been generated and tested by multiple laboratories and none were found to be active regardless of the termini at which the protein was fused or the fluorescent protein that was used. This was such an impediment to live cell imaging of IFITM3 that the top labs in the field have gone to extraordinary measures to visualize IFITM3 in live cells (Spence et al., 2019; Peng, J Am Chem Soc, 2016; Suddala et al., 2019). It is not possible that the fluorescent fusion proteins used in this paper are antivirally active as presented in Figure 3.

We understand that fluorescent tags can impede the behavior of membrane proteins in live cell imaging. We are not ignorant of the results achieved by Spence et al. and Suddala et al., which were published during the preparation of our manuscript. We have included new data outlining the oligomerization potential and antiviral activity of IFITM3 constructs madae a la Suddala et al.—that is, we introduced the YFP/mCherry tags internally into IFITM3, after codon 40. We then compared the anti-IAV activity of these internally-tagged IFITM3 constructs with our IFITM3 constructs containing N-terminal tags. Our results showed that the IFITM3 constructs containing YFP/mCherry at the N-terminus exhibit more potent antiviral activity than those encoding YFP/mCherry internally. Specifically, our YFP/mCherry placed at the N-terminus resulted in a 20% loss of antiviral function compared to FLAGIFITM3, while YFP/mCherry placed internally at residue 40 resulted in a 40% loss of antiviral function.

An example of raw data and the combined results from three infection experiments is now shown in Figure 4—figure supplement 2A and B. We also went ahead and assessed whether the internally-tagged constructs were competent for FRET as measured by FLIM microscopy. We found the WT versions of these constructs produced FRET and that the G95L mutation abrogated FRET (Figure 4—figure supplement 2C), mirroring our original results obtained using IFITM3 tagged with YFP/mCherry at the N-terminus (Figure 4B and 4C). Therefore, mutation of G95 results in loss of oligomerization in all fluorescently tagged constructs. Our focus on the use of IFITM3 fluorescently tagged with YFP and mCherry at the Nterminus is justified by their more potent antiviral activity. While these results may seem at odds with Suddala et al., they tagged IFITM3 internally with different fluorophores (mNeonGreen and mTFP1) than we did (YFP and mCherry) and the antiviral potential of these different constructs may be impacted by fluorophore choice.

An equally important point to stress here is that we utilized FLAG-IFITM3 and myc-IFITM3 to address oligomerization as well (in Figure 5), in order to confirm our results using fluorescently-tagged IFITM3. Since the reviewer does not question the antiviral function performed by FLAG-IFITM3, our findings with FLAG-IFITM3 and myc-IFITM3 further support that oligomerization is critical for antiviral function.

2) All of the virus infectivity data are presented as normalized infectivity, and no raw data flow cytometry plots are shown. For influenza infections, an MOI of 0.1 is used. Thus, the maximum % infection would be at most ~7%, providing a minimal dynamic range for observing antiviral effects of IFITM3 or for comparing IFITM3 mutants. This minimal dynamic range, along with the strange results with fluorescent fusions, coupled with unnecessary obfuscation of the primary data, overall decreases confidence in some of the results.

We agree that our manuscript would be better appreciated if data obfuscation was minimized. In order to render the results more transparent, we have incorporated many examples of raw data throughout the manuscript. Raw data of flow cytometry results can now be found in Figure 2C (IAV infection of HEK293T stably expressing IFITM3 constructs), Figure 3A (HIV infection of TZMbl cells as a readout of virion infectivity), Figure 1—figure supplement 1E (fusion of HIV-BlaM-Vpr-VSV-G in HEK293T stably expressing IFITM3 constructs), and Figure 4—figure supplement 2A (IAV infection of HEK293T transiently expressing FLAG- and fluorescent protein-tagged IFITM3).

Figure 2C shows raw data for IAV infections performed at an MOI of 0.1. We observe about ~12% infection occurring in HEK293T cells at this MOI, which is similar to the ~7% predicted by the reviewer. However, we have also performed other experiments using higher and lower MOIs, and our conclusions are not impacted. For example, please consider the raw FACS data in Author response image 1:

Author response image 1

These data show that varying the MOI from 0.05 to 0.5 does not influence our measurements of antiviral function by IFITM3 WT or mutants. Therefore, we believe our choice to report infection data using an MOI of 0.1 is appropriate.

Reviewer #3:

However, aspects of these conclusions need further support, as I have detailed below. I also feel that the novelty and conceptual advance is likely not sufficient for publication by eLife, but rather incremental. IFITM3 has already been shown to increase membrane order (Li et al., 2013) and G95R was previously shown to attenuate restriction of IAV (John et al., 2013).

We thank the reviewer for appreciating the importance of our results with regards to IFITM3 oligomerization. However, the reviewer has failed to appreciate the other points of novelty in our manuscript:

1) While mutation of G95 was previously shown to impact anti-IAV activity, the reason for loss of function was unknown. Here, we provide the mechanism for loss-of-function.

2) While we openly agree that previous literature demonstrated that IFITM3 increases membrane order, we are the first to directly show that membrane order is directly linked to antiviral function. By using a loss-of-function mutation (G95L), we are the first to show that membrane stiffening is required for antiviral activity. Furthermore, by incorporating Amphotericin B into the same assay, we are the first to show that Amphotericin B negates the impact of IFITM3 on membrane stiffness. Together, these data provide significant novel information about how IFITM3 performs its antiviral activities AND how a drug overcomes these antiviral activities.

3) We believe this manuscript is appropriate for publication in eLife because of the cross-cutting nature of the data—it is not merely an incremental advance on the function of IFITM3. The basis for our study was actually gleaned from the fact that a non-synonymous SNP (G305W) in PRRT2 had been associated with neurological disease in humans. We decided to study the homologous residue in IFITM3 (G95) and identified a GxxxG motif in IFITM3 and as well as other CD225 proteins, indicating a shared requirement for oligomerization in the diverse membrane fusion processes controlled by the CD225

protein family. Importantly, we show that mutation of G305W in PRRT2 disrupts its ability to oligomerize, adding fresh insight to our understanding of how this mutation results in disease.

1) Do the authors, as per the title, link the motif or AA reside 95 to oligomerisation/restriction?, vast majority of data pertains to G95L – see below.

We have added significant amounts of new data to highlight the effects of both G91 and G95 of the GxxxG motif. As a result, data on G91 is now found in Figures 1-4 and the data are less focused on a single mutation. Our new findings indicate that mutation of G95 has a larger impact on protein oligomerization than G91, which justifies our focus on G95 in the latter figures of the paper. We tried to study a double mutant whereby the G91L and G95L mutations were introduced simultaneously into IFITM3, but the polypeptide product was not expressed at detectable levels following transfection into HEK293T (not shown). Overall, our paper describes the functional impact of the GxxxG motif rather than a single point mutation, and this underlines the novelty of our study because the GxxxG motif in IFITM3 has not previously been reported or functionally described.

2) The Introduction states that the GXXXG motif is also known as a glycine zipper whereas my understanding is that the most significant glycine zipper sequence patterns are (G,A,S)XXXGXXXG and GXXXGXXX(G,S,T), which contain a GXXXG motif (https://www.pnas.org/content/102/40/14278, not referenced in the current paper)

We thank the reviewer for raising this point. Since the GxxxG sequence in IFITM3 does not clearly belong to an extended motif resembling a glycine zipper (GxxxGxxxG or similar), we removed “glycine zipper” from the text. Instead we chose to refer to the motif as a “GxxxG motif, also known as a (small)xxx(small) motif” as described by one of our references (Teese et al., 2015).

3) I do not feel this is an evolution-guided analysis (title of the first Results section), which to me would imply analysis of evolutionary selection pressures. I am more comfortable with “homology-guided”, as per the title. I also do not follow the logic of the phylogenetic tree of CD225 domains in 1A “indicative of a common ancestry”, all CD225 superfamily proteins have this domain, by definition. Not sure what the tree adds. Authors also say that the GXXXG motif is conserved in vertebrate IFITM3s but (i) So are many other residues/ AA motifs conserved in the highly conserved CD225 domain of vertebrate IFITM3, as the alignment Figure 1D shows, and (ii) surely to make a point about conservation among vertebrates, an alignment where 5 of the 7 mammalian IFITM3 sequences shown are from primates, i.e. the same mammal order, is less informative than an alignment with wider range of mammal or vertebrate species whose IFITM3 has been sequenced. Do IFITM5 and IFITM10 oligomerise, as they lack an intact GXXXG? (not commented on)

We agree that the term “evolution-guided” is inappropriate. This statement has been corrected to the preferred language “homology-directed.” We also removed the phylogenetic tree since the common ancestry of the CD225 proteins, specifically that of IFITM3 and PRRT2, is clearly described elsewhere. To strengthen the notion that GxxxG is conserved among vertebrates, we updated Figure 1D to include the additional species: cow, cat, microbat, and sperm whale. With this expanded list of species, the conservation of the GxxxG motif is even more apparent.

Regarding IFITM5 and IFITM10, we introduced additional text in the Discussion stating “Interestingly, the GxxxG motif is not intact in human IFITM5 and IFITM10. It is tempting to speculate that decay of the GxxxG motif is one reason why an antiviral function for these IFITM members is yet to be reported. It is also possible that oligomerization is unimportant for their respective non-antiviral roles in cells: IFITM5 is involved in bone formation, while the function of IFITM10 is unknown.”

4) Although authors say they focused on G95L and G95W mutants for further functional characterisation (subsection “Glycine-95 is important for the antiviral functions of IFITM3”), no localisation data are shown for G95W and for some reason, G95W is omitted from panels D and F in Figure 2, which shows effects on antiviral restriction. Later, oligomerisation is not assessed for G95W (Figures 3 and 4) and yet despite the lower expression levels noted for G91L by transient transfection (S2A), this G91L is reinserted into the analysis of oligomerisation (Figure supp 3B and interpretation lines subsection “GxxxG regulates oligomerization of IFITM3 in living cells and bulk lysates”). The authors do not explain the logic for this “pick and choose” approach.

We agree that the absence of G95W from certain figures made the data appear imbalanced. We have now included additional data on IFITM3 G95W, including results on its expression level, subcellular localization, and antiviral activity. These data can now be found in Figure 2, Figure 1—figure supplement 1, and Figure 2—figure supplement 1. We quantified the protein expression of IFITM3 WT and mutants, including G95W, by flow cytometry and western blotting and performed quantitative and statistical analysis for both methods. Here are some highlights of what we found:

1) G95W lost activity against IAV and VSV-G-driven virus-cell fusion to approximately the same degree as G95L;

2) G95W and G95L exhibit a similar subcellular localization relative to IFITM3 WT, in that all colocalize with the plasma membrane, early endosomes, and late endosomes to similar extents following transient and stable transfection in HEK293T. Our quantitative analysis revealed that about ~50% of IFITM3 resides in endolysosomes, irrespective of the mutations in the GxxxG motif.

Due to the fact that G95W behaves similarly to G95L in terms of its antiviral activity, we did not think it was necessary to explore the effect of both G95L and G95W on oligomerization and membrane order. Overall, we are hopeful that the reviewer will be satisfied with the fact that results for G91L, G95L, and G95W are now presented side-by-side in Figure 2, Figure 1—figure supplement 1, and Figure 2—figure supplement 1.

5) I do not find that the localisation data which is presented for WT and G95L to be convincing ( i.e. Figure S2B; IF performed for transient IFITM3 expression in EEA1-GFP overexpressing cells). The authors state their IF shows IFITM3 distributed in early and late endosomes and the PM (subsection “Glycine-95 is important for the antiviral functions of IFITM3”), whereas the staining to me seems very dominantly PM with minimal endosomal staining? Why is there no IF done on the stable cell lines which much of the restriction data in Figure 2 pertain to (panels B, D, E). Further, since numerous studies have shown localisation to be a key feature/correlate of IFITM3 restriction, stronger data confirming “similar subcellular localisation” of the Glycine mutants should be presented (both G95L and G95W).

In response to this criticism, we completely overhauled our immunofluorescence data. We performed additional experiments whereby IFITM3 and all mutants (G91L, G95L, and G95W) in both transiently and stably transfected HEK293T cells were localized relative to EEA1-GFP and CD63. We performed transfections and stainings in three independent experiments and performed colocalization analysis on Zstacks using Imaris software. These new results can now be found in Figure 2—figure supplement 1. We found that approximately 50% of IFITM3 is present in early/late endosomes following ectopic expression, which is broadly consistent with our previous work studying FLAG-IFITM3 (Compton et al., 2016). We also found that cells stably expressing IFITM3 contain higher levels of IFITM3 in CD63+ late endosomes/MVB, which likely explains why stably expressing cells exhibit more potent antiIAV activity than cells which were transiently transfected. Importantly, there were no statistically significant differences between the degree to which WT, G91L, G95L, and G95W colocalized with early/late endosomes (Figure 2—figure supplement 1). These results suggest that mutations in the GxxxG motif result in loss of function for reasons other than altered subcellular localization.

6) It is known that expression levels of IFITM3 affect its antiviral restriction. For transient transfection experiments, authors provide a single western blot panel (Figure supplementary2 A) which indicates the two Gly-95 mutants express similarly to WT when transiently expressed. However, it would be better to perform densitometric quantification of replicate independent blots, and/ or provide western panels to shown expression in the same experiments as restriction is assessed following transient transfection. For the stable cell lines, FACS analysis is shown (Figure 2 A) which indicates that the mutants are less well expressed in this system. First, I think that replicates should have been analysed in order to show whether there is any statistically significant difference between the cell lines. (Currently only the MFI for a single replicate for each cell line is given: WT=814, G95W=590 , G95L=605). Secondly, the authors need to comment on whether this expression level difference might influence the interpretation of the restriction data obtained for the cell lines, and indeed the other phenotypes seen.

We agree that we did not objectively characterize the relative expression levels of IFITM3 WT and mutants. Therefore, we quantitatively assessed levels of IFITM3 WT, G91L, G95L, and G95W by both flow cytometry and western blotting and under conditions of both transient and stable transfection. In both cases, quantitation and statistical analysis from multiple experiments were performed. For the flow cytometry data, we show an example of flow histograms as well as a combined bar graph representing the mean fluorescence intensities from 3-4 experiments (Figure 2A and 2B for stable transfection, Figure 1—figure supplement 1A and 1B for transient transfection). For the western blotting data, we show an example of a membrane scan of cell lysates produced from a single transfection as well as a combined bar graph representing the mean band intensity from 3 experiments as measured by Odyssey Li-Cor technology (which quantitatively measures secondary antibodies coupled to fluorophores emitting in the infrared range, and fluorescence is directed measured; this is superior to traditional densiometry making use of HRP-coupled secondary antibodies, which is a semi-quantitative technique that requires enhanced chemiluminescence) (Figure 1—figure supplement 1C and D for transient transfection). These data allow the comparison of steady-state protein levels between IFITM3 WT and the mutants in a more convincing and transparent manner. This analysis revealed that G95W was expressed at significantly lower levels than WT in stable cell lines (Figure 2B) and G91L was expressed at significantly lower levels than WT in transient transfected cells (Figure 2—figure supplement 1D). Since, in stable cell lines, G95W is just as antiviral as G95L (Figure 2D and 2E), this difference does not seem to impact function. However, the decreased expression of G91L under transient conditions probably does impact our interpretation of Figure 3B, which has been updated in this revision. Here, the G91L mutation results in a complete loss of activity against HIV-1 in producer cells (manifesting in loss of virion infectivity), and we believe this is due to the combined effects of 1) partial loss of oligomerization for IFITM3 containing G91L, and 2) lower steady state levels of IFITM3 containing G91L. Importantly, we did not detect any significantly different levels of expression between WT and G95L irrespective of the method used (flow cytometry, western blot) or the transfection (transient, stable). Therefore, any functional difference observed between WT and G95L is very unlikely to be the result of differential subcellular localization or steady-state protein level. This later point justifies our focus on G95L in Figures 5 and 6.

7) Authors state that late stage inhibition of HIV-1 entry is regulated by glycine 95 (subsection “Glycine-95 is important for the antiviral functions of IFITM3”), these data are shown in a single bar chart in Figure 2F and a composite supplementary Figure 2C. In supplementary Figure 2C, the authors should explain how the quantification relates to the Western blot panels, presumably all the quantification graphs refer to the right-hand panels of the westerns labelled “virus supernatants”, but this needs to be clarified. Furthermore, could the (arguably small) differences seen for wt vs. glycine mutant in right-hand panel (supernatants) not be a consequence of the differences seen in the left-hand panel (i.e. expression levels in whole cell lysates), the latter potentially due to the effects shown for IFITM3 on HIV protein synthesis (Lee et al., 2018, “IFITM proteins inhibit HIV-1 protein synthesis”)? It is not clear to me that these data wholly support the authors interpretation that late stage inhibition of HIV-1 entry is regulated by glycine 95 and the conclusion that “oligomerization-defective G95L mutant lessened the impact of virion-associated IFITM3 on HIV-1, suggesting that IFITM3 oligomers are needed to maximally reduce virion infectivity”.

We have moved all of the data pertaining to the effect of IFITM3 on HIV virion infectivity to the main figures (now Figure 3). We also included the effect of the G91L mutation to this set of the results, such that a more general conclusion can be made about the role of the GxxxG motif in this activity. We find that G91L completely abrogated this anti-HIV activity while G95L inhibited it partially. To better understand the mechanistic reason for loss of anti-HIV function by G91L/G95L, we examined HIV Env levels in virus producing cells (Figure 3C) and virus-containing supernatants (i.e. virions) (Figure 3D). It appears that the reviewer misunderstood these data. It was previously shown by us and others (Compton et al., 2014, Tartour et al., 2014, Compton et al., 2016, Yu et al., 2015, Ahi et al., 2020) that IFITM3 impairs HIV Env processing in virus-producing cells, resulting in decreased levels of Env incorporated into virions, and IFITM3 itself incorporates into virions. However, there is also evidence that IFITM3 impairs Env function in a qualitative manner (Compton et al., 2014 and Ahi et al., 2020), and this seems to correlate with the extent of IFITM3 incorporation into virions. Here, we perform quantitative measurements of Env and IFITM3 in virions/virus supernatants. An example of one virion blot is seen in Figure 3D, and virus producing cells from the same experiment are provided in Figure 3C, to confirm that the impact of IFITM3 on Env originates in virus-producing cells (as the reviewer suggests). We then show composite data from multiple blots of virions/virus supernatants in order to glean a mechanistic understanding of why G91L and G95L result in loss of impairment of virion infectivity. We quantified levels of Env gp120 (Figure 3E), Env gp41 (Figure 3F), and IFITM3 (Figure 3G) in virions/virus supernatants in order to identify correlates of antiviral function which have been lost by G91L and G95L mutations. We observe that G91L and G95L rescue gp120 and gp41 levels in HIV-1 virions, suggesting that the GxxxG motif must be important for the reduction of Env in virions. However, G91L and G95L restore Env levels to a similar extent yet the same mutations differentially impact HIV infectivity, suggesting that Env levels do not completely explain this antiviral function. However, we observe that IFITM3 encoding G91L is less incorporated into virions relative to G95L and WT, suggesting that our data allows us to tease apart the contributions made by Env and IFITM3 in the loss of virion infectivity (Figure 3G). We now discuss this in detail in the Results: “Since the G91L and G95L mutations differentially impact the restriction of HIV-1 virion infectivity by IFITM3, it is therefore unlikely that Env quantity in virions fully accounts for this restriction. However, we observed that the G91L and G95L mutations strongly impaired the ability of IFITM3 itself to incorporate into virions (Figure 3G), and the extent of virion incorporation correlated with the measured impact on HIV-1 infectivity (Figure 3B). Together, our results demonstrate that the dual antiviral functions performed by IFITM3 (early-stage inhibition of virus entry and late-stage inhibition of virion infectivity) are critically regulated by the GxxxG motif. Interestingly, the former function depends on primarily on glycine-95, while the latter function depends more on glycine-91.”

It is unlikely that the process outlined by Lee et al., 2018 is at play here, because we are studying virion infectivity which has been normalized to levels of p24 Gag in the supernatant. That means that we can exclude the impact of IFITM3 on viral protein translation and focus solely on infectivity per ng amount of p24 Gag. It remains possible that IFITM3 WT leads to loss of Env in virus-producing cells due to a process outlined by Lee et al. However, here we are not concerned with why Env levels are lower in producer cells, but rather, whether differences in Env quantity account for the antiviral activity of IFITM3 WT and the lack thereof of G91L/G95L. Our data suggest that Env quantity in virions only partially explains the full antiviral potential of IFITM3 when expressed in virus-producing cells, which has been recently outlined and discussed in a recent article from our lab (Ahi et al., 2020).

8) Figure 3 (on oligomerisation) is confusing- the circular symbols are not explained and I was unclear what the 3 columns represented and whether the labels at the top (mCherry, IFITm3-YFP and IFITM3-mCherry) pertained to the columns (in which case they need to be centered over the columns) or were a legend. Accordingly, I could not interpret these data or the bar charts below the IF which, again, used these undefined circular symbols. The figure legend refers to filled red and filled yellow circles, which I could not see on the figure. I have the same problem interpreting Figure 6 (analagous experiments with PRRT2).

We apologize for the confusion. It appears that the automatic formatting performed by eLife’s manuscript submission portal prevented some of the colored circles from showing correctly. We have ensured that this will not happen again in the revised figure, which is now Figure 4. The labels at the top of the figure represent the legend. To make this more apparent, we drew a black box around those labels, and moved them to the bottom of the panels. We are confident that this figure, and the others containing similar data, will be entirely interpretable now.

9) The co-IP and native PAGE data lack loading controls (Figure 4) and I was also unclear why i. the heterologous pairs had not been tried for co-IPs experiments ie WT with G95L, ii. there was no replicates or quantitation for the native PAGE (i.e. an equivalent panel as for 4B).

To satisfy this request, we added additional loading controls (heavy chain for the IP fraction and actin for the whole cell lysates fraction) to the co-IP data which now appears in Figure 5.

We appreciate the reviewer’s demand that heterologous pairs (IFITM3 WT + IFITM3 G95L) be examined by co-IP. We have included these novel results into Figure 5—figure supplement 1B and C. We found that the G95L mutation introduced into one member of the pair resulted in an intermediate loss of oligomerization, mirroring our results using the FRET-based assay (Figure 4B and 4C). Furthermore, we added the quantitation of three replicates of the blue native PAGE, which now appears in Figure 5D. We thank the reviewer for pointing out these omissions.

10) Using the FliptR system, authors report that “membrane order enhancement tracks with a functionally competent form of IFITM3 but not with a loss-of-function mutant”. It would have been useful to assess one of the many other loss of function IFITM mutants that have been described in the literature, e.g. the S-palmitoylation defective mutants or an amphipathic helix-lacking mutant (especially since some of the same authors discovered IFITM3's amphipathic helix). Further, while positive and negative controls for membrane order were used (supp Figure 5), it would help strengthen the data on membrane rigidity if authors had used an additional method, especially one that has been used previously to study IFITM3 e.g. Laurdan.

We agree that testing additional IFITM3 mutants for their capacity to promote membrane order would contribution additional novel insight and strengthen the appropriateness of this manuscript for eLife. To that end, we introduced new data into Figure 6—figure supplement 1E. Here, we show that IFITM3 lacking a functional amphipathic helix (S61A, N64A, T65A) is incapable of increasing membrane order. Since both G95L and S61A/N64A/T65A exhibit a loss of antiviral activity (as shown in this manuscript and in Chesarino et al., 2017, respectively), this further confirms the functional link between virus restriction and elevated membrane order. However, unlike G95L, we show that S61A/N64A/T65A does not result in a loss of oligomerization (this result added to Figure 4—figure supplement 1). These novel data support our hypothesis that the antiviral functions of IFITM3 require both an amphipathic helix AND the oligomerization determinants conferred by the GxxxG motif. This significant finding is discussed in the Discussion: “Here, we show for the first time that the amphipathic helix of IFITM3 is critical for the membrane order enhancement by IFITM3. Since we also show that glycine-95 of the GxxxG motif is also required for membrane order enhancement, our data suggest that oligomerization “activates” the membrane deforming activity of the amphipathic helix, and as a result, its antiviral potential. It is possible that local insertion of multiple amphipathic helices into stretches of membrane is required for inhibition of virus fusion, and IFITM3 oligomers provide a means to fulfil that requirement.”

Lastly, as the reviewer recommended, we performed Laurdan staining as a complimentary method to study the impact of IFITM3 on membrane order. We found that the results with Laurdan mirrored our results using the FliptR approach, such that IFITM3 WT increased order while G95L did not. These new data are presented in Figure 6—figure supplement 1C and D.

https://doi.org/10.7554/eLife.58537.sa2

Article and author information

Author details

  1. Kazi Rahman

    HIV Dynamics and Replication Program, Center for Cancer Research, National Cancer Institute, Frederick, United States
    Contribution
    Conceptualization, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing - review and editing
    Contributed equally with
    Charles A Coomer
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2986-0007
  2. Charles A Coomer

    1. HIV Dynamics and Replication Program, Center for Cancer Research, National Cancer Institute, Frederick, United States
    2. Cellular Imaging Group, Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, United Kingdom
    Contribution
    Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - review and editing
    Contributed equally with
    Kazi Rahman
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2523-6485
  3. Saliha Majdoul

    HIV Dynamics and Replication Program, Center for Cancer Research, National Cancer Institute, Frederick, United States
    Contribution
    Formal analysis, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0530-6354
  4. Selena Y Ding

    HIV Dynamics and Replication Program, Center for Cancer Research, National Cancer Institute, Frederick, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4413-644X
  5. Sergi Padilla-Parra

    1. Cellular Imaging Group, Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, United Kingdom
    2. Department of Infectious Diseases, King’s College London, Faculty of Life Sciences & Medicine, London, United Kingdom
    3. Randall Division of Cell and Molecular Biophysics, King’s College London, London, United Kingdom
    Contribution
    Software, Supervision, Funding acquisition, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8010-9481
  6. Alex A Compton

    HIV Dynamics and Replication Program, Center for Cancer Research, National Cancer Institute, Frederick, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    For correspondence
    alex.compton@nih.gov
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7508-4953

Funding

National Institutes of Health (Intramural Research Program)

  • Kazi Rahman
  • Charles A Coomer
  • Saliha Majdoul
  • Selena Y Ding
  • Alex A Compton

European Research Council (ERC-2019-CoG-863869 FUSION)

  • Sergi Padilla-Parra

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

Acknowledgements

We thank Guoli Shi, Stephen Lockett, and the Optical Microscopy and Image Analysis Laboratory of the National Cancer Institute, Center for Cancer Research for providing technical support during the acquisition and analysis of microscopy data.

Senior Editor

  1. Olga Boudker, Weill Cornell Medicine, United States

Reviewing Editor

  1. Mark Marsh, University College London, United Kingdom

Reviewers

  1. Mark Marsh, University College London, United Kingdom
  2. Camilla Benfield

Publication history

  1. Received: May 4, 2020
  2. Accepted: October 27, 2020
  3. Accepted Manuscript published: October 28, 2020 (version 1)
  4. Version of Record published: November 13, 2020 (version 2)

Copyright

This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.

Metrics

  • 610
    Page views
  • 109
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Evolutionary Biology
    2. Genetics and Genomics
    Tom Hill, Robert L Unckless
    Research Article Updated

    Hosts and viruses are constantly evolving in response to each other: as a host attempts to suppress a virus, the virus attempts to evade and suppress the host’s immune system. Here, we describe the recurrent evolution of a virulent strain of a DNA virus, which infects multiple Drosophila species. Specifically, we identified two distinct viral types that differ 100-fold in viral titer in infected individuals, with similar differences observed in multiple species. Our analysis suggests that one of the viral types recurrently evolved at least four times in the past ~30,000 years, three times in Arizona and once in another geographically distinct species. This recurrent evolution may be facilitated by an effective mutation rate which increases as each prior mutation increases viral titer and effective population size. The higher titer viral type suppresses the host-immune system and an increased virulence compared to the low viral titer type.

    1. Ecology
    2. Evolutionary Biology
    Naïma Jesse Madi et al.
    Research Article

    Microbes are embedded in complex communities where they engage in a wide array of intra- and inter-specific interactions. The extent to which these interactions drive or impede microbiome diversity is not well understood. Historically, two contrasting hypotheses have been suggested to explain how species interactions could influence diversity. 'Ecological Controls' (EC) predicts a negative relationship, where the evolution or migration of novel types is constrained as niches become filled. In contrast, 'Diversity Begets Diversity' (DBD) predicts a positive relationship, with existing diversity promoting the accumulation of further diversity via niche construction and other interactions. Using high-throughput amplicon sequencing data from the Earth Microbiome Project, we provide evidence that DBD is strongest in low-diversity biomes, but weaker in more diverse biomes, consistent with biotic interactions initially favoring the accumulation of diversity (as predicted by DBD). However, as niches become increasingly filled, diversity hits a plateau (as predicted by EC).