key: cord-0325914-0ma2oiez
authors: Rahman, Kazi; Coomer, Charles A.; Majdoul, Saliha; Ding, Selena; Padilla-Parra, Sergi; Compton, Alex A.
title: Homology-guided identification of a conserved motif linking the antiviral functions of IFITM3 to its oligomeric state
date: 2020-05-16
journal: bioRxiv
DOI: 10.1101/2020.05.14.096891
sha: 012d7a7b12e5293e1cc4ac26d0557067f14d8e52
doc_id: 325914
cord_uid: 0ma2oiez

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. We identified a GxxxG motif in many CD225 proteins, including IFITM3 and proline rich transmembrane protein 2 (PRRT2). Mutation of 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 in IFITM3 disrupted its oligomerization and reduced its antiviral activity against Influenza A and HIV-1. The oligomerization-defective variant was used to reveal that IFITM3 promotes membrane rigidity in a glycine-95-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.

amphipathic helix which is essential for antiviral activity [22] . Amphipathic helices have been 75 identified in many eukaryotic, prokaryotic, and viral proteins and are well known for their ability 76

to bind and bend membranes [23, 24] . Another piece of evidence that supports the membrane 77 deformation model of fusion inhibition is that IFITM3 incorporated into enveloped viruses also 78 impairs virion fusion with target cells [25] [26] [27] [28] [29] [30] [31] [32] . Despite this progress, a complete mechanistic 79 view of IFITM3 is lacking because studies describing the impact of IFITM3 on membranes have 80 not included mutants lacking antiviral function. 81

It was previously reported that IFITM3 forms clusters on virus-containing vesicles [33] 82 and that IFITM3 oligomerization promotes restriction of IAV [34] . However, the determinants 83 initially purported to mediate oligomerization (phenylalanines at residues 75 and 78 of the 84 CD225 domain [34] ) were later shown to be unnecessary when oligomerization was measured 85

using a FRET-based approach in living cells [35] . Therefore, the oligomerization of IFITM3 86 appears to be influenced by unknown determinants and its importance to antiviral function is not 87

established. 88

In the current study, we set out to identify a loss-of-function mutation in IFITM3 suitable 89 for mechanistic studies by using a homology-guided approach. The IFITM genes (IFITM1, 90

IFITM2, IFITM3, IFITM5, and IFITM10 in humans) are members of an extended gene family 91 known as the Dispanin/CD225 family (hereafter referred to as CD225 proteins) [36, 37] . 92

Members of this group are characterized by the presence of a CD225 domain but the functions of 93 most remain unknown. However, one member is the subject of numerous studies because it is 94 linked to neurological disorders. Mutations in proline-rich transmembrane protein 2 (PRRT2) 95 result in conditions of involuntary movement, such as paroxysmal kinesigenic dyskinesia, benign 96 familial infantile seizures, and episodic ataxia [38, 39] . PRRT2 is a neuron-specific protein that 97

is localized to pre-synaptic terminals and which inhibits synaptic vesicle fusion [40] [41] [42] . 98

Molecular studies of disease-associated missense mutations in PRRT2 (G305W/R) indicate that 99

it causes loss-of-function, leading to unchecked neurotransmitter release [43] [44] [45] [46] [47] . Interestingly, 100

the homologous residue in human IFITM3 is also subject to rare allelic variation in humans 101

(G95W/R) and this mutation results in partial loss of activity against IAV infection [34] . 102

However, the reason why this site is essential for the respective functions of PRRT2 and IFITM3 103 was unknown. 104

Here, we demonstrate that glycine-95 of human IFITM3 resides within a 91 GxxxG 95 105 motif that is highly conserved among vertebrate IFITM3 orthologs as well as PRRT2. Mutation 106 of glycine-95 rendered IFITM3 less active against entry of IAV and vesicular stomatitis virus 107 (VSV). Using fluorescence lifetime imaging microscopy (FLIM) and Förster resonance energy 108 transfer (FRET) of fluorophore-tagged IFITM3, we found that the 91 GxxxG 95 motif mediates 109 IFITM3 oligomerization in intact cells and that IFITM3 oligomers are mostly dimers. A mutant 110 exhibiting loss of antiviral function (G95L) was deficient for oligomerization, indicating that 111 IFITM3 oligomerization and restriction of virus entry are functionally associated. We leveraged 112 this loss-of-function mutant to identify mechanistic correlates of antiviral function which are 113 dependent upon IFITM3 oligomerization. Using a FLIM-based reporter, we found that IFITM3 114

increased membrane order, as previously suggested, while the G95L mutation failed to do so. 115

These data indicate that membrane stiffening by IFITM3 oligomers is required for antiviral 116 activity. In an effort to further probe the importance of membrane order in the antiviral 117 mechanism, we demonstrate that Amphotericin B (Ampho B) decreases the stiffness of IFITM3-118

containing membranes and prevents virus inhibition. These data provide significant new insight 119 into the mechanism by which IFITM3 inhibits the fusion of diverse pathogenic viruses and 120 reveal that oligomerization is a shared requirement for the distinct anti-fusion functions 121 performed by homologs IFITM3 and PRRT2. 122 123

Results 124 125

Evolution-guided identification of a putative oligomerization motif within CD225 domains 126

A phylogenetic tree of CD225 domains found in the IFITM family (consisting of IFITM1, 127 IFITM2, IFITM3, IFITM5, and IFITM10 in humans) and other CD225 proteins (human PRRT2 128

and TUSC5) are indicative of common ancestry ( Figure 1A) . predicted topology consisting of dual hydrophobic domains and a CIL, but the amino terminus is 138

considerably longer in PRRT2 ( Figure 1B and Supplemental Figure 1A ). When comparing 139 topologies and the protein alignment ( Figure 1C) , we noticed the presence of a GxxxG motif in 140 the CIL of IFITM3 and PRRT2 ( Figure 1B and 1C) which is intact in several other IFITM and 141

CD225 proteins [5] . The GxxxG motif, also known as a glycine zipper, is frequently associated 142

with dimerization of membrane proteins [53, 54] . Most often shown to mediate pairing of 143 hydrophobic transmembrane helices within a bilayer, the motif has also been described to drive 144 oligomerization from cytoplasmic loops or linkers [55] . The GxxxG motif is conserved in 145

IFITM3 of vertebrates, indicating that it may play an important functional role ( Figure 1D ).

Glycine-95 is important for the antiviral functions of IFITM3 148

We generated FLAG-tagged IFITM3 mutants in which glycine-91 and glycine-95 were 149 changed to leucine (G91L and G95L), following an example set by characterization of the 150

GxxxG motif in the human folate transporter [56] . We also produced a G95W mutant because a 151 rare single nucleotide polymorphism known as rs779445843 gives rise to missense mutations 152 (G95W/R) in human populations [57, 58] . Furthermore, G95W in IFITM3 is analogous to a 153 disease-associated polymorphism in PRRT2 (G305W) which results in loss of function [44] .

Upon transfection into human cells, IFITM3 containing G95L or G95W exhibited steady-state 155

protein levels similar to wild-type (WT), but this was not the case for G91A (Supplemental 156 Figure 2A ). This finding may explain why a previous study failed to express an IFITM3 mutant 157 containing a stretch of alanines between positions 90 and 95 [13, 34] . Therefore, we focused on 158 G95L and G95W for further functional characterization. The subcellular localization of IFITM3 159

WT and G95L was found to be similar when confocal immunofluorescence microscopy was 160

performed. As shown previously [3, 59, 60] , protein was detected in early endosomes, late 161 endosomes, and at the plasma membrane (Supplemental Figure 2B) . We generated HEK293T 162 cell lines stably expressing FLAG-tagged IFITM3 WT and mutants to assess antiviral function 163 (Figure 2A ). Challenge of cell lines with IAV demonstrated that, while IFITM3 WT strongly 164

protected cells from infection, the G95L or G95W mutations resulted in a considerable loss of 165 virus restriction ( Figure 2B ). When we tested antiviral function following transient transfection 166 into HEK293T, we found that G95L or G95W resulted in a complete loss of IAV restriction 167 ( Figure 2C and Supplemental Figure 2A ). To confirm that IFITM3 targets the entry step of 168 IAV, we challenged cells with HIV-1 pseudotyped with hemagglutinin (HA, a type 1 viral 169 fusogen) and found that G95L results in a severe loss of pseudovirus restriction ( Figure 2D ).

The G95L and G95W mutations also abrogated restriction of HIV-1 pseudotyped with VSV 171 glycoprotein (a type 3 viral fusogen) in an assay for virus-cell fusion ( Figure 2E ). Therefore, a 172 rational approach identified glycine-95 to be essential for broad inhibition of virus entry by 173 IFITM3.

We and others have previously demonstrated that, in addition to preventing virus entry 175 into naïve target cells, IFITM3 performs another antiviral function in virus-producing (infected) 176 cells by incorporating into virions, reducing viral glycoprotein abundance and function, and 177

reducing the fusogenic potential of virions [25] [26] [27] [28] [29] 32] . Therefore, we also tested the relative 178

impact of IFITM3 WT and IFITM3 G95L on HIV-1 virion infectivity. IFITM3 WT reduced the 179 infectivity of HIV-1 and was accompanied by a defect in viral Envelope (Env) incorporation into 180

virions: processing of the Env gp160 precursor (as determined by gp120/gp160 ratio) was 181 decreased, resulting in relatively low levels of gp120 and gp41, consistent with previous work 182

[ 26, 28, 61] . By comparison, IFITM3 G95L exhibited reduced activity against HIV-1 infectivity 183

and impacted Env processing/incorporation to a lesser extent ( Figure 2F and Supplemental 184 Figure 2C ). Furthermore, IFITM3 G95L itself was incorporated into HIV-1 virions about 50% 185 less than WT IFITM3 (Supplemental Figure 2C) . Therefore, the dual antiviral functions ( Figure 3A and 3C). However, we did not observe decreases in YFP lifetimes when G95L-YFP 206

and G95L-mCherry were examined. To ensure that conclusions drawn from this imaging 207

approach are relevant to IFITM3-mediated antiviral function, we tested the ability for 208

fluorescently-tagged IFITM3 to inhibit IAV. WT IFITM3-mCherry exhibited nearly the same 209 antiviral potency as WT IFITM3-FLAG while WT IFITM3-YFP was slightly less potent. 210

Moreover, versions of fluorescently-tagged IFITM3 encoding G95L exhibited significantly less 211 antiviral activity than their WT counterparts (Supplemental Figure 3A ). Using the same FRET-212 based approach, we found that the G91L mutation partially resulted in higher YFP lifetimes 213 relative to WT IFITM3, indicating that G91L also impairs oligomerization of IFITM3 214 (Supplemental Figure 3B) . Overall, these data provide the first indication that glycine-91 and 215

glycine-95 are important for IFITM3 oligomerization. To quantitatively resolve the specific 216 oligomeric state of IFITM3 under these conditions, we performed a Number and Brightness 217

analysis [63, 64] . This approach is a fluorescence microscopy method capable of measuring the 218 apparent average number of molecules and their brightness in each pixel over time, with 219 brightness being proportional to oligomeric state. Analysis was restricted to IFITM3-mCherry 220 during its diffusion in and out of a static membranous compartment (the plasma membrane). On 221 average, WT IFITM3-mCherry was found to be 2.14 times brighter than mCherry monomers 222

[65] ( Figure 3D ). In contrast, G95L-mCherry brightness was not significantly different than 223 mCherry monomers (averaging 1.30 times brighter). Furthermore, relative to mCherry 224 monomers, both WT IFITM3-mCherry and G95L-mCherry formed assemblies that were up to 225 five times brighter, but WT IFITM3-mCherry demonstrated a greater propensity to form these 226

higher-order oligomers ( Figure 3D ). These data suggest that fluorescently-tagged IFITM3 forms 227 dimers and higher-order oligomers in membranes in a glycine-91-and glycine-95-dependent 228

fashion.

In parallel to our studies of IFITM3 oligomerization in single, living cells, we assayed the 230 ability of IFITM3 pairs tagged with FLAG or myc to co-immunoprecipitate from bulk cell 231

lysates. HEK293T were co-transfected with IFITM3-FLAG and IFITM3-myc followed by 232 FLAG immunoprecipitation, SDS-PAGE, and quantitative immunoblotting. We found that WT 233 IFITM3-myc readily pulled down with WT IFITM3-FLAG, while pull down of G95L-myc with 234

G95L-FLAG was diminished by approximately 50% (Figure 4A and 4B). Therefore, 235

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 237 IFITM3 under non-denaturing conditions. Two populations of IFITM3 oligomers, exhibiting 238

sizes of approximately 300 and 480 kilodaltons, were readily observed for WT and, to a lesser 239 extent, IFITM3 G95L. The largest (480 kilodaltons) was nearly absent for IFITM3 G95L 240 ( Figure 4C ). In contrast to the Number and Brightness analysis ( Figure 3D ), we did not observe 241 dimers, which may result from the conditions under which blue native PAGE was performed 242

here (Supplemental Figure 4A) . Nonetheless, the fact that IFITM3 G95L exhibited a reduced 243 potential for higher-order oligomer formation using this technique is consistent with our 244 experiments using living cells or denaturing conditions. Therefore, glycine-95 is necessary for 245 efficient IFITM3 oligomer formation and oligomerization is necessary for its antiviral function. 246 247 248

While the precise mechanism by which IFITM3 inhibits virus-cell fusion remains 250 unresolved, a salient phenotype of IFITM protein expression is increased membrane order 251

(reduced membrane fluidity) [17, 34] . Therefore, we leveraged our loss-of-function mutant to 252 directly test whether membrane order is functionally associated with inhibition of virus entry by 253

IFITM3. Previous reports of membrane order enhancement by IFITM family members involved 254

the use of a cell-permeable dye known as Laurdan [66] . Here, we assessed membrane order 255 using a recently described sensor known as fluorescent lipid tension reporter (FliptR) [67] [68] [69] .

FliptR is a planarizable push-pull probe that incorporates efficiently into artificial and living cell 257 membranes and whose fluorescence parameters change upon alterations in local lipid packing 258 (order). Specifically, FliptR responds to increasing membrane order at the plasma membrane and 259

at endomembranes by planarization, leading to longer fluorescence lifetimes detected by FLIM 260 [67, 70] . Using HEK293T stably expressing FLAG-IFITM3, we found that WT IFITM3 261 expression led to significantly increased membrane order, consistent with previous reports [14, 262 17] ( Figure 5A ). In fact, the IFITM3-induced enhancement of membrane order was similar to 263

that achieved by addition of soluble cholesterol, while cholesterol depletion with methyl-beta-264

cyclodextrin resulted in profound decreases in membrane order (Supplemental Figure 5A ). In 265 contrast to WT, cells expressing IFITM3 G95L did not exhibit increased membrane order 266 ( Figure 5A ). These data indicate that membrane order enhancement tracks with a functionally-267 competent form of IFITM3 but not with a loss-of-function mutant. To further probe the 268 functional importance of membrane order in the antiviral mechanism of IFITM3, we performed 269 experiments in the presence of Ampho B. This antimycotic polyene compound was previously 270

reported to overcome the antiviral activity of IFITM3, rendering cells stably expressing IFITM3 271

fully permissive to IAV [14] . However, it was unknown how Ampho B counteracts the effects of 272

IFITM3. When we added Ampho B to cells expressing WT IFITM3, we no longer observed 273

increased membrane order ( Figure 5A ). Furthermore, in identically-treated cells, Ampho B 274

prevented restriction of HA-mediated virus entry by IFITM3 ( Figure 5B ). These findings show, 275

for the first time, that the capacity for Ampho B to overcome the antiviral activity of IFITM3 is 276 linked to its ability to decrease membrane order. Therefore, the use of Ampho B as an 277

interrogative tool reinforced the role played by membrane order in the antiviral mechanism of 278 IFITM3. Overall, these data strongly suggest that IFITM3-mediated antiviral activity occurs 279 through oligomerization-dependent membrane stiffening.

Disease-associated G305W impairs oligomerization of PRRT2 282

Since the GxxxG motif is a shared feature of IFITM3 and PRRT2 and that naturally-283 occurring variation at glycine-305 of PRRT2 is predictive of neurological disease, we assessed 284 the oligomerization capacity of WT and mutant PRRT2. As in Figure 3 , we constructed PRRT2 285 fused with YFP or mCherry to create FRET pairs. We observed that co-transfection of PRRT2-286 YFP and PRRT2-mCherry resulted in significant FRET, demonstrating that PRRT2 oligomerizes 287 in living cell membranes ( Figure 6A ). However, FRET was significantly reduced for pairs 288 containing G305W (G305W-YFP and G305W-mCherry) ( Figure 6A ). Furthermore, pairs 289

containing G305W did not exhibit loss in YFP lifetimes relative to their WT counterparts 290 ( Figure 6B ). These data suggest that mutation of glycine-305 in PRRT2 results in loss of protein 291

oligomerization. Therefore, the divergent functions played by homologues IFITM3 and PRRT2 292

in the regulation of fusion processes are controlled by a common determinant. 293 294 295

Discussion 296 297

The physiological importance of IFITM3 in the control of many virus infections in vivo 298

is becoming increasingly apparent [71, 72] . While it has been proposed that membrane 299 remodeling by IFITM3, at the level of membrane order and curvature, protects host cells from 300 virus invasion, functional proof was lacking. However, two recent developments provided a 301 glimpse of how (and when) IFITM3 inhibits virus-cell membrane fusion. First, the identification 302

of an amphipathic helix located in the IM domain of IFITM3 provided a rational explanation for 303

how membrane stiffening and/or bending may occur [22] . Second, IFITM3 has been observed to 304 intercept vesicles carrying inbound virions and to restrict their release into the cytoplasm, while 305

viruses that are insensitive to IFITM3 evade its encounter [31, 73] . Together with the previous 306 demonstrations that the subcellular localization of IFITM3 determines its specificity and potency 307

[30, 74, 75], a "proximity-based" mechanism presents itself in which IFITM3 interacts with and 308 modifies host membranes needed by some viruses to fuse with cells. Importantly, this model 309

accommodates the antiviral effect of IFITM3 inside viral lipid bilayers as well [25, 26, 28, 31, 310 32]. 311

Here, we provide evidence that an additional determinant (protein oligomerization) plays 312 a crucial role within this mechanistic framework. Previously, an alanine scanning approach led to 313 the identification of two residues (phenylalanine-75 and phenylalanine-78) that were important 314

for IFITM3 oligomerization in cell lysates [34] . Importantly, this study did not functionally test a 315 role for glycine-91 or glycine-95 in the oligomerization of IFITM3 because alanine mutagenesis 316 overlapping these residues led to loss of stable protein expression [13, 34] . While it has been 317 confirmed that F75A and F78A mutations disrupt antiviral activity [31, 76] , IFITM3 318 oligomerization was not impacted by these mutations when assessed by FRET in living cells 319

[35]. It remains possible that these two phenylalanines cooperate with the GxxxG motif to 320 mediate IFITM3 oligomerization. However, our data shows that G95L reduces the antiviral 321 potential of IFITM3 as well as its capacity to increase membrane order (reduce membrane 322 fluidity). Furthermore, we show that a compound previously found to abrogate the antiviral 323 function of IFITM3, Ampho B [14, 31], decreases membrane order in IFITM3-expressing cells.

Even though the exact mechanism by which Ampho B impacts mammalian membranes is 325

unclear [14, 77] , these results identify that the membrane stiffening property of IFITM3 is a strict 326

correlate of its antiviral functions in cells and, most likely, in virions. We found that the 327 oligomerization-defective G95L mutant lessened the impact of virion-associated IFITM3 on 328

HIV-1, suggesting that IFITM3 oligomers are needed to maximally reduce virion infectivity. 329

It will be important to assess how oligomerization-defective mutants of IFITM3 (G95L) 330

impact membrane curvature, another reported consequence of ectopic IFITM3 expression in cells 331

[17]. It has been shown that protein oligomerization of the transmembrane protein Mic10 is 332 essential for its capacity to induce curvature in mitochondrial membranes [78] . Furthermore, 333 changes in membrane order are often accompanied by changes in membrane curvature [18] [19] [20] . It 334 is possible that these two alterations to host membranes underlie the restriction of virus fusion by 335 IFITM3. Moreover, amphipathic helices are characterized to interact with and bend membranes 336

[24]. However, the impact that the amphipathic helix of IFITM3 has on membrane order and 337 curvature has not been assessed. Regardless of how the amphipathic helix is responsible for 338

inhibition of virus fusion, our data raise the possibility that oligomerization "activates" the 339 antiviral potential of the amphipathic helix. It is possible that local insertion of multiple 340 amphipathic helices into stretches of membrane is required for inhibition of virus fusion, and 341

IFITM3 oligomers provide a means to fulfill that requirement. 342

In addition to mediating dimerization or oligomerization of transmembrane proteins 343

(homomultimerization), GxxxG motifs have also been described to affect the propensity for 344

interaction with other proteins (heteromultimerization) [53, 79] . Therefore, glycine-95 may also 345 govern which proteins IFITM3 interacts with and to what extent. IFITM3 has been described to 346

bind with IFITM1 and IFITM2, and it is interesting to consider how IFITM heteromultimers may 347

contribute to antiviral protection of the cell [34] . Furthermore, other host proteins have been 348 described to interact, directly or indirectly, with IFITM3. This list includes cholesterol 349

trafficking regulator VAPA and the metalloproteinase ZMPSTE24, two proteins that have been 350 described as essential cofactors for the antiviral effects of IFITM3 [10, 80] . Since the former is 351 associated with the tendency for IFITM3 to cause cholesterol overload in endosomes, the G95L 352 loss-of-function mutant could be used to rule in or rule out VAPA and cholesterol as players in 353 the antiviral mechanism of IFITM3. 354

While glycine-95 is critical for the oligomerization and anti-fusion activity of IFITM3, 355

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 357 known to disrupt PRRT2 activity, and our results here provide novel insight into how loss-of-358 function occurs. Therefore, the presence of a shared GxxxG motif in IFITM3, PRRT2, and some 359 other CD225 family members suggests that an ancestral CD225-containing protein performed an 360 unknown function that required oligomerization. We wonder whether all CD225 proteins play 361 roles in regulating membrane fusion processes in cells-only time and further experiments will 362

tell. However, just as glycine-95 in IFITM3 may mediate both homo-and heteromultimerization, 363

it has been reported that PRRT2 interacts with cellular fusogens known as soluble N-364 ethylmaleimide-sensitive factor attachment protein receptors (SNAREs) Cytofix/Cytoperm (BD), immunostained with anti-IAV NP (AA5H; Abcam), and analyzed on a 400

LSRFortessa flow cytometer (BD). Replication-incompetent HIV-1 pseudotyped with IAV WSN 401

HA and NA was produced by transfecting HEK293T with 2 μg pR8ΔEnv, 1 μg pcRev (NIH 402 AIDS Reagent Resource: 11415), 3 μg Gag-GFP, 1.5 μg of hemagglutinin, and 1.5 μg of 403 neuraminidase from IAV WSN strain, H1N1 (gifts from G. Melikyan). Replication-competent 404

HIV-1 was produced by transfecting HEK293T with 15 μg pNL4-3 and 5 μg empty pQCXIP or 405 pQCXIP encoding IFITM3 WT or the indicated mutant fused with amino-terminal FLAG. 406

Replication-incompetent HIV-1 pseudotyped with VSV-G for virus-cell fusion assays was 407 produced by transfecting HEK293T with 15 μg pNL4-3ΔEnv, 5 μg pCMV4-BlaM-Vpr, and 5 μg 408

pMD2.G (VSV-G). Transfections were performed using the calcium-phosphate method. Briefly, 409

six million HEK293T were seeded in a T75 flask. Plasmid DNA was mixed with sterile H 2 0, 410

CaCl 2 , and Tris-EDTA (TE) buffer, and the totality was combined with Hepes-buffered saline 411

(HBS Odyssey blocking buffer in PBS (Li-COR) and incubated with anti-FLAG M2 (F1804; Sigma) 448

and anti-c-Myc (C3956; Sigma). Secondary antibodies conjugated to DyLight 800 or 680 COR) and the Li-COR Odyssey imaging system were used to reveal specific protein detection. 450

Images were analyzed and assembled using ImageStudioLite (Li-COR immersion. To assess FRET, donor YFP fluorescence was detected with a gallium arsenide 491 phosphide photomultiplier tube (GaAsP PMT) with a 520-550 nm emission window following 492 excitation by a 514 nm laser. Acceptor mCherry fluorescence was detected with a GaAsP PMT 493

detector with a 570-615 nm emission window following excitation by a 561 nm laser. FRET 494 signal (acceptor mCherry fluorescence triggered by excitation of donor YFP) was collected with 495

a GaAsP PMT detector with 570-615 nm emission window after excitation with a 514 nm laser. 496

At least 50 cells per condition were examined in each experiment. Each cell was assigned a 497

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 499 laser (Coherent) tuned at 80 MHz with single photon counting electronics (Becker Hickl) and 500 detection with a HPM-100-40 module GaAsP hybrid PMT (Becker Hickl). Analysis was limited 501

to cells exhibiting 250-1000 photons per pixel to mitigate the effects of photobleaching and low 502 signal to noise ratio. SPCImage NG software (Becker Hickl) was used to acquire the 503 fluorescence decay of each pixel, which was deconvoluted with the instrument response function 504

and fitted to a Marquandt nonlinear least-square algorithm with two exponential models. The 505

mean fluorescence lifetime was calculated as previously described [82] using SPCImage NG. At 506 least 30 cells per condition were analyzed in each experiment. 507 508

Number and Brightness analysis. HEK293T cells were seeded in μ-slide 8 well chambers 509

(Ibidi) (50,000 per well) overnight and transiently with 0.50 μg IFITM3-mCherry using TransIT-510 293 (Mirus). Living cells in Fluorobrite DMEM (Gibco) were imaged with a Zeiss LSM 780 511 confocal microscope using a 63X objective and oil immersion. Regions of interest were limited 512 to portions of cells which were immobile and which focused on plasma membrane fluorescence 513 (intracellular signal from vesicular membranes was excluded). The axial position of a specimen 514 during acquisition was stabilized using the Adaptive Focus Control module. mCherry was 515 excited with a 561 nm laser and detected with a 570-615 emission window. For each cell, 100 516

frames were acquired at a frame rate of 0.385 frames per second with a 9.75 μs pixel dwell time 517

and pixel size of 151.38 nm. Images were always acquired at 256 x 256 pixels such that the pixel 518 size remained 3-4 times smaller than the volume of the point-spread function. Photobleaching of 519 fluorescent proteins during data acquisition was corrected using a detrending algorithm [64] . 520

Twenty cellular regions were examined per condition. Pixel-by-pixel brightness values were 521 calculated in Fiji (ImageJ). 522 523

FLIM for study of membrane order with FliptR. HEK293T stably expressing empty pQCXIP 524 or pQCXIP-WT IFITM3 or the indicated mutant were seeded in μ-slide 8 well chambers (Ibidi) 525

(50,000 per well) overnight and stained with 1 μM FliptR (Spirochrome) for 5 mins according to 526 the manufacturer's protocol. Imaging was performed with a 63X objective under oil immersion 527 on a Leica SP8-X-SMD confocal microscope. When indicated, cells were treated with 100 528 μg/mL soluble cholesterol (C4951, Sigma), 5 mM methyl-cyclo-beta-dextrin (MßCD) (C4555, 529

Sigma), or 1 μM Amphotericin B (A2942, Sigma) for one hour prior to addition of FliptR and 530

imaging. Fluorescence was detected by hybrid external detectors in photon counting mode 531

following excitation by a 488 nm pulsed laser turned to 20 MHz with single photon counting 532 electronics (PicoHarp 300). Analysis was limited to cells with at least 250-1000 photons per 533 pixel to mitigate the effects of photobleaching and low signal to noise ratio. Fluorescence decay 534 of each pixel in FliptR-stained cells was acquired by Symphotime 64 software (Picoquant), and 535 deconvoluted with the instrument response function and fitted to a Marquandt nonlinear least-536 square algorithm with two exponential models. The mean fluorescence lifetime (τ), in addition to 537

individual component lifetimes (long and short lifetimes, τ 1 and τ 2 ), were calculated using 538

Symphotime. At least 30 cells per condition were analyzed in each experiment. co-transfected with EEA1-GFP and empty pQCXIP or IFITM3-FLAG (encoding WT or G95L). 591

Cells were fixed, permeabilized, and immunostained at 48 hours post-transfection with anti-592 IFITM3 and anti-CD63 and cells were analyzed by immunofluorescence confocal microscopy. 593

All images are average Z-stacks from 10 consecutive medial sections. Scale bars, 10 μm. (C) 594

Whole cell lysates and virus-containing supernatants were collected from HEK293T co-595

transfected with the HIV-1 molecular clone NL4.3 and empty pQCXIP, WT IFITM3-FLAG, or 596

the indicated mutant at 48 hours post-transfection. Virus-containing supernatants were 597 ultracentrifuged through sucrose cushions. Both lysates and concentrated, purified virus-598

containing supernatants were subjected to SDS-PAGE and Western blot analysis.

Immunoblotting was performed with anti-gp120, anti-gp41, anti-CA, anti-actin, and anti-FLAG. 600

Levels of IFITM3 (FLAG), gp120, and gp41 were quantified and normalized to levels of CA. 601

For anti-FLAG immunoblotting, the amount of WT IFITM3 in virions was set to 100%. For anti-602 gp120 and anti-gp41, levels observed in the pQCXIP empty vector were set to 100%. used to depict IFITM3-mCherry (either WT or G95L), and filled yellow circles are used to depict 635 IFITM3-YFP (either WT or G95L). Statistical analysis was performed using one-way ANOVA. 636 *, P < 0.05; **, P < 0.001. 637 

(A) As in Figure 5A , except that soluble cholesterol (100 μg/mL) or methyl-beta-cyclo-dextrin (5 677 mM) were added to untransfected HEK293T cells for one hour and washed away prior to 678 addition of 1 μM FliptR and cells were imaged by FLIM. τ 1 was measured for a minimum of 40 679 cells per condition and the results of three independent experiments were pooled. Dots 680 correspond to individual cells. Error bars indicate standard deviation. Statistical analysis was 681 performed using one-way ANOVA. *, P < 0.05; **, P < 0.001. 682 used to depict mCherry, filled red circles are used to depict PRRT2-mCherry (either WT or 691 G305W), and filled yellow circles are used to depict PRRT2-YFP (either WT or G305W). Error 692 bars indicate standard deviation. Statistical analysis was performed using one-way ANOVA. *, P 693 < 0.05; **, P < 0.001 .  694  695  696  697  698  699  700  701  702  703  704  705  706 707 708 

Whole cell lysates were produced 642 under mildly denaturing conditions and immunoprecipitation (IP) using anti-FLAG antibody was 643 performed. IP fractions and volumes of whole cell lysates were subjected to SDS-PAGE and 644 Western blot analysis. Immunoblotting was performed with anti-FLAG and anti-myc. Number 645 and tick marks indicate size (kilodaltons) and position of protein standards in ladder. (B) Levels 646 of IFITM3-myc (either WT or G95L) co-immunoprecipitated by anti-FLAG IP were quantified 647 from conditions in (A)

Work in the lab of AAC is supported by the Intramural Research Program of the National 717 Institutes of Health, National Cancer Institute, Center for Cancer Research

Bieniasz PD: Intrinsic immunity: a front-line defense against viral attack

More than meets the I: the diverse antiviral and 730 cellular functions of interferon-induced transmembrane proteins

Regulation of the trafficking and antiviral 733 activity of IFITM3 by post-translational modifications

IFITMs Restrict the Replication of Multiple 736 Pathogenic Viruses

IFITM-Family Proteins: The Cell's First Line of 738

Annual Review of Virology

Ifitm3 limits the severity of acute influenza in 740 mice

IFITM3 restricts the morbidity and mortality associated with influenza. 743 Nature

SNP-mediated disruption of CTCF binding at 746 the IFITM3 promoter is associated with risk of severe influenza in humans

Effector IFITM3 Disrupts Intracellular Cholesterol Homeostasis to Block Viral Entry. 750 CHOM 2013

U18666A, an intra-cellular cholesterol 752 transport inhibitor, inhibits dengue virus entry and replication

Late Endosomal/Lysosomal Cholesterol Accumulation Is a Host Cell-756

Protective Mechanism Inhibiting Endosomal Escape of Influenza A Virus

Functional Mapping of Regions Involved in the 760 Negative Imprinting of Virion Particle Infectivity and in Target Cell Protection by

Interferon-Induced Transmembrane Protein 3 against HIV-1

Amphotericin B Increases Influenza A Virus Infection by Preventing 765 IFITM3-Mediated Restriction

IFITM3 restricts influenza 767 A virus entry by blocking the formation of fusion pores following virus-endosome 768 hemifusion

Coronavirus Spike Protein: Evidence for Cholesterol-Independent Mechanisms

IFITM proteins restrict viral membrane hemifusion

A sub-nanometre view of how 776 membrane curvature and composition modulate lipid packing and protein 777 recruitment

Protein-Induced Membrane Curvature Alters 779 Local Membrane Tension

Membrane Curvature and Tension Control 781 the Formation and Collapse of Caveolar Superstructures

Mechanisms of membrane fusion: disparate players and 783 common principles

IFITM3 requires an amphipathic helix for antiviral activity

Amphipathic helices and membrane curvature

IFITM proteins are incorporated onto HIV-1 virion 793 particles and negatively imprint their infectivity

IFITM Proteins Restrict HIV-1 Infection by Antagonizing the Envelope 796 Glycoprotein. CellReports

Interference with the production of 799 infectious viral particles and bimodal inhibition of replication are broadly conserved 800 antiviral properties of IFITMs

Retroviral Envelope Abundance and Function and Is Counteracted by glycoGag

Macaque interferon-induced transmembrane proteins limit replication of 806 SHIV strains in an Envelope-dependent manner

809 Natural mutations in IFITM3 modulate post-translational regulation and toggle 810 antiviral specificity

Interferon-induced transmembrane protein 3 blocks fusion of sensitive 813 but not resistant viruses by partitioning into virus-carrying endosomes

IFITM Proteins Incorporated into HIV-1 Virions Impair Viral Fusion and 817 Spread

IFITM3 Clusters on Virus Containing Endosomes 819 and Lysosomes Early in the Influenza A Infection of Human Airway Epithelial Cells. 820 Viruses

The CD225 domain of IFITM3 is required for both IFITM protein 823 association and inhibition of influenza A virus and dengue virus replication

Analysis 826 of IFITM-IFITM Interactions by a Flow Cytometry-Based FRET Assay

The Dispanins: A Novel 829

Gene Family of Ancient Origin That Contains 14 Human Members

Transmembrane Gene Family in Vertebrates

The clinical and genetic heterogeneity of paroxysmal 835 dyskinesias

PRRT2: from Paroxysmal Disorders to 837 Regulation of Synaptic Function

Synaptic or ion channel modifier? PRRT2 is a chameleon-like regulator of 839 neuronal excitability

PRRT2 deficiency induces paroxysmal kinesigenic 841 dyskinesia by influencing synaptic function in the primary motor cortex of rats

PRRT2 mutations lead to neuronal dysfunction and 845 neurodevelopmental defects

PRRT2 Regulates Synaptic Fusion by Directly Modulating 851 SNARE Complex Assembly. CellReports

PRRT2 gene mutations: from paroxysmal 854 dyskinesia to episodic ataxia and hemiplegic migraine

PRRT2 phenotypes and 857 penetrance of paroxysmal kinesigenic dyskinesia and infantile convulsions

860 Mutations in PRRT2 result in paroxysmal dyskinesias with marked variability in clinical 861 expression

S-palmitoylation and ubiquitination differentially 863 regulate interferon-induced transmembrane protein 3 (IFITM3)-mediated resistance 864 to influenza virus

Interferon-induced transmembrane protein 866 3 is a type II transmembrane protein

Combined approaches of 869 EPR and NMR illustrate only one transmembrane helix in the human IFITM3

A Novel Topology of Proline-rich Transmembrane Protein 2 (PRRT2)

Protter: interactive protein feature 875 visualization and integration with experimental proteomic data

Role of GxxxG Motifs in Transmembrane Domain Interactions

Single-cell glycolytic activity regulates membrane tension and HIV-1 925 fusion

Fluorescent flippers for mechanosensitive membrane 928 probes

Antiviral Protection by IFITM3 In Vivo

Human Genetic Determinants of Viral 933

935 IFITM3 directly engages and shuttles incoming virus particles to lysosomes

The N-Terminal Region of 938 IFITM3 Modulates Its Antiviral Activity by Regulating IFITM3 Cellular Localization. 939 Journal of Virology

Phosphorylation of the antiviral 941 protein interferon-inducible transmembrane protein 3 (IFITM3) dually regulates its 942 endocytosis and ubiquitination

Identification of Residues Controlling Restriction versus Enhancing Activities 945 of IFITM Proteins on Entry of Human Coronaviruses

Recent progress in the study of the interactions of amphotericin B with 948 cholesterol and ergosterol in lipid environments

Mic10 Oligomerizes to Bend Mitochondrial Inner Membranes at Cristae Junctions

A GxxxG-like motif within HIV-1 fusion peptide is critical to 954 its immunosuppressant activity, structure, and interaction with the transmembrane 955 domain of the T-cell receptor

ZMPSTE24 defends against influenza and other pathogenic 958 viruses

Rapid Titration of Measles and Other 960 Viruses: Optimization with Determination of Replication Cycle Length

Investigating protein-protein interactions in living cells 963 using fluorescence lifetime imaging microscopy

We thank Stephen Lockett and the Optical Microscopy and Image Analysis Laboratory of the 712 National Cancer Institute, Center for Cancer Research, for technical support.