SNARE mimicry by the CD225 domain of IFITM3 enables regulation of homotypic late endosome fusion

Abstract

The CD225/Dispanins superfamily consists of membrane proteins that regulate vesicular transport and membrane fusion events driving neurotransmission, glucose transport, and antiviral immunity. However, how the CD225 domain controls membrane trafficking was unknown. We reveal that the CD225 domain contains a SNARE-like motif that enables interaction with cellular SNARE fusogens. Proline rich transmembrane protein 2 (PRRT2) encodes a SNARE-like motif that enables interaction with neuronal SNARE proteins, and mutations therein disrupt SNARE binding and are linked to neurological disease. Another CD225 member, interferon-induced transmembrane protein 3 (IFITM3), protects cells against Influenza A virus infection. IFITM3 interacts with SNARE proteins that mediate late endosome-late endosome (homotypic) fusion and late endosome-lysosome (heterotypic) fusion. IFITM3 binds to syntaxin 7 (STX7) in cells and in vitro, and mutations that abrogate STX7 binding cause loss of antiviral activity against Influenza A virus. Mechanistically, IFITM3 disrupts assembly of the SNARE complex controlling homotypic fusion and accelerates the trafficking of endosomal cargo to lysosomes. Our results suggest that SNARE modulation plays a previously unrecognized role in the diverse functions performed by CD225 proteins.

Figure 1:. CD225 proteins contain a motif resembling an R-SNARE motif.

(A) A selection of SNARE proteins from yeast and mammals were aligned with proteins with SNARE-like motifs, and a 53-residue stretch of sequence is shown. Proteins are grouped by Q-SNARE, R-SNARE, or R-SNARE-like, and CD225 proteins are labeled as a subset of the R-SNARE-like proteins. Residues in red text are the central polar arginine or glutamine residues characteristic of canonical SNARE proteins and are labeled as the central “0” layer. Surrounding heptad repeats of hydrophobic residues are labeled from “−7” to “+8” in relation to the “0” layer, and sites highlighted in yellow correspond to hydrophobic residues. Underlined residues are those found in PRRT2, which when mutated, are associated with neurological dysfunction. (B) Left: HEK293T cells were co-transfected with SNAP25-HA and either PRRT2-FLAG (WT, R295Q, G305W, or A320V) or Empty Vector. SDS-PAGE and immunoblotting were performed with anti-HA and anti-FLAG in whole cell lysates. Anti-actin was used as loading control. Right: From co-transfected cells, SNAP25-HA was immunoprecipitated with anti-HA followed by SDS-PAGE and immunoblotting with anti-HA and anti-FLAG. Immunoglobulin chain was used as loading control. Numbers and tick marks left of blots indicate position and size (in kilodaltons) of protein standard in ladder. The FLAG/HA ratio was calculated for the indicated lanes and shown as mean and standard error (normalized relative to WT, which was set to 100%). Differences that were statistically significant from WT as determined by one-way ANOVA are indicated by (*). Exact p values are as follows (from left to right): p < 0.0001, p < 0.0001, p < 0.0001. Immunoblots were performed independently three times (biological replicates), and a representative example is shown. Sc; Saccharomyces cerevisiae. Ig; immunoglobulin. IP; immunoprecipitation. EV; empty vector. WT; wild-type.

Figure 2:. IFITM3 contains an R-SNARE-like motif and interacts with ectopic endosomal Q-SNARE STX7.

(A) Human endosomal SNARE proteins were aligned with aligned with human IFITM proteins, and a 53-residue stretch of sequence is shown. Residues in red text are the central polar arginine or glutamine residues characteristic of canonical SNARE proteins and are labeled as the central “0” layer. Surrounding heptad repeats of hydrophobic residues are labeled from “−7” to “+8” in relation to the “0” layer, and sites highlighted in yellow correspond to hydrophobic residues. Black underlined residues in IFITM3 which were mutated for functional experiments. Green underlined residues in IFITM3 correspond to the previously described amphipathic alpha helix. (B) HEK293T cells were transfected with IFITM3-FLAG (WT, F75/78A, R85Q, or G95L) and, 48 hours later, challenged with IAV (A/PR/8/34 (PR8), H1N1) at a multiplicity of infection of 0.10. 18 hours post-infection, cells were fixed, permeabilized, immunostained with anti-FLAG and anti-NP, and analyzed by flow cytometry. The percentage of NP+ cells was measured as a fraction of FLAG+ cells and shown as mean and standard error (normalized relative to Empty Vector, which was set to 100%). Infections were performed independently three times (biological replicates). Differences that were statistically significant from WT as determined by one-way ANOVA are indicated by (*). Exact p values are as follows (from left to right): p < 0.0001, p < 0.0001, p < 0.0001. (C) Structural prediction of the R-SNARE-like motif of IFITM3 was performed with Alphafold and FATCAT. Residues 55–113 of IFITM3 were modeled against a template of the R-SNARE motif of VAMP8 which was previously crystallized as part of the STX7-STX8-Vti1b-VAMP8 trans-SNARE complex (PDB: 1GL2). (D) Left: HEK293T cells were co-transfected with IFITM3-FLAG or Empty Vector and a construct encoding a single SNARE protein: STX7-HA, STX8-HA, Vti1b-HA, VAMP8-HA, or VAMP7-myc. SDS-PAGE and immunoblotting were performed with anti-HA, anti-myc, and anti-FLAG in whole cell lysates. Anti-actin was used as loading control. Right: From co-transfected cells, IFITM3-FLAG was immunoprecipitated with anti-FLAG followed by SDS-PAGE and immunoblotting with anti-HA, anti-myc, and anti-FLAG. Heavy chain immunoglobulin chain was used as loading control. Co-transfection of Empty Vector and STX7-HA was used as a negative control for immunoprecipitation. (E) Left: HEK293T cells were co-transfected with STX7-HA and either IFITM3-FLAG (WT, F75/78A, or R85Q) or Empty Vector. SDS-PAGE and immunoblotting were performed with anti-HA and anti-FLAG in whole cell lysates. Anti-actin was used as loading control. Right: From co-transfected cells, IFITM3-FLAG was immunoprecipitated with anti-FLAG followed by SDS-PAGE and immunoblotting with anti-FLAG and anti-HA. Light chain immunoglobulin was used as loading control. Co-transfection of Empty Vector and STX7-HA was used as a negative control for immunoprecipitation. The HA/FLAG ratio was calculated for the indicated lanes and shown as mean and standard error (normalized relative to WT, which was set to 100%). Differences that were statistically significant from WT as determined by one-way ANOVA are indicated by (*). Exact p values are as follows (from left to right): p < 0.0001, p < 0.0001. Numbers and tick marks left of blots indicate position and size (in kilodaltons) of protein standard in ladder. Immunoblots were performed independently three times, and a representative example is shown. IAV; Influenza A virus. Ig; immunoglobulin. IP; immunoprecipitation. EV; Empty Vector. WT; wild-type.

Figure 3:. IFITM3 interacts with endosomal Q-SNAREs in a G95-dependent manner

(A) Top: HEK293T cells were co-transfected with STX8-HA, STX7-HA, or Vti1b-HA and either Empty Vector or IFITM3-FLAG (WT, F75/78A, or G95L). SDS-PAGE and immunoblotting were performed with anti-HA and anti-FLAG in whole cell lysates. Anti-tubulin was used as loading control. Bottom: from co-transfected cells, STX8-HA, STX7-HA, or Vti1b-HA were immunoprecipitated with anti-HA followed by SDS-PAGE and immunoblotting with anti-HA and anti-FLAG. Heavy chain immunoglobulin was used as loading control. Co-transfection of Empty Vector and STX8-HA was used as a negative control for immunoprecipitation. The FLAG/HA ratio was calculated for the indicated lanes and shown as mean and standard error (normalized relative to WT with STX7-HA, which was set to 100%). Differences that were statistically significant from WT with STX7-HA as determined by one-way ANOVA are indicated by (*). Exact p values are as follows (from left to right): p < 0.0001, p < 0.0001, p < 0.0001, p < 0.0001, p < 0.0001. Numbers and tick marks left of blots indicate position and size (in kilodaltons) of protein standard in ladder. Immunoblots were performed independently three times, and a representative example is shown. (B) HEK293T cells stably expressing Empty Vector, IFITM3 WT-FLAG, or IFITM3 G95L-FLAG were transfected with STX7-GFP, fixed, immunostained with anti-FLAG, and analyzed by confocal immunofluorescence microscopy. Colocalization was measured between FLAG and STX7-GFP by calculating the Pearson’s Correlation Coefficient using Fiji software. Coefficients were calculated from medial Z-slices from three fields of view containing 5–15 cells per condition and presented as means and standard error. Scale bar = 15 microns. Ig; immunoglobulin. IP; immunoprecipitation. EV; Empty Vector. WT; wild-type. PCC; Pearson’s Correlation Coefficient.

Figure 4:. Endogenous IFITM3 interacts with endosomal Q-SNAREs and R-SNAREs via an R-SNARE-like motif.

(A) Left: HeLa cells (WT or IFITM3 KO) were transfected with STX7-HA. SDS-PAGE and immunoblotting were performed with anti-HA and anti-IFITM3 in whole cell lysates. Anti-actin was used as loading control. Right: STX7-HA was immunoprecipitated with anti-HA antibody followed by SDS-PAGE and immunoblotting with anti-HA and anti-IFITM3. Heavy chain immunoglobulin was used as loading control. An isotype matched antibody was used as a control for immunoprecipitation. (B) Endogenous STX8 was immunoprecipitated from HeLa (WT or IFITM3 KO) with anti-STX8 followed by SDS-PAGE and immunoblotting with anti-STX8 and anti-IFITM3. Heavy chain immunoglobulin was used as loading control. An isotype matched antibody was used as a control for immunoprecipitation. The (#) symbol denotes the presence of light chain immunoglobulin. (C) Endogenous Vti1b was immunoprecipitated from HeLa (WT or IFITM3 KO) with anti-Vti1b followed by SDS-PAGE and immunoblotting with anti-Vti1b and anti-IFITM3. Light chain immunoglobulin was used as loading control. (D) Endogenous VAMP8 was immunoprecipitated from HeLa (WT or IFITM3 KO) with anti-VAMP8 followed by SDS-PAGE and immunoblotting with anti-VAMP8 and anti- IFITM3. Light chain immunoglobulin was used as loading control. (E) Endogenous VAMP7 was immunoprecipitated from HeLa (WT or IFITM3 KO) with anti-VAMP7 followed by SDS-PAGE and immunoblotting with anti-VAMP7 and anti-IFITM3. Two forms of VAMP7 were detected following immunoprecipitation: one form of approximately 25 kD and another of approximately 15 kD (Wojnacki et al, 2021). Light chain immunoglobulin was used as loading control. (F) Left: Recombinant STX7 and recombinant IFITM3 (WT, F75/78A, or G95L) were mixed together in vitro and reaction inputs were visualized by SDS-PAGE and immunoblotting with anti-STX7 and anti-IFITM3. Right: recombinant STX7 was immunoprecipitated with anti-STX7 followed by SDS-PAGE and immunoblotting with anti-STX7 and anti-IFITM3. Heavy chain immunoglobulin was used as loading control. Numbers and tick marks left of blots indicate position and size (in kilodaltons) of protein standard in ladder. Immunoblots were performed independently two times and a representative example is shown. Ig; immunoglobulin. IP; immunoprecipitation. KO; knockout. WT; wild-type. r; recombinant.

Figure 5:. IFITM3 inhibits the interaction between STX7/8 and VAMP8.

(A) HEK293T cells stably expressing Empty Vector, IFITM3 WT-FLAG, or IFITM3 G95L-FLAG were co-transfected with STX7-HA and VAMP8-Myc. Whole cell lysates were subjected to SDS-PAGE and immunoblotting with anti-HA, anti-FLAG, and anti-Myc. Actin was used as a loading control. Right: STX7-HA was immunoprecipitated with anti-HA followed by SDS-PAGE and immunoblotting with anti-HA, anti-FLAG, and anti-Myc. The Myc/HA ratio was calculated for the indicated lanes and shown as mean and standard error (normalized relative to Empty Vector, which was set to 100%). Differences that were statistically significant from Empty Vector as determined by one-way ANOVA are indicated by (*). Exact p values are as follows (from left to right): p = 0.0016, p = 0.4318. (B) HEK293T cells stably expressing Empty Vector, IFITM3 WT-FLAG, or IFITM3 G95L-FLAG were transfected with STX7-GFP and VAMP8-mCherry, fixed, and analyzed by confocal immunofluorescence microscopy. Colocalization was measured between STX7-GFP and VAMP8-mCherry by calculating the Pearson’s Correlation Coefficient using Fiji software. Coefficients were calculated from medial Z-slices from three fields of view containing 5–15 cells per condition and presented as means and standard error. Scale bar = 15 microns. (C) HEK293T cells stably expressing Empty Vector, IFITM3 WT-FLAG, or IFITM3 G95L-FLAG were inoculated with IAV (+) or medium (−) for 18 hours and subjected to whole cell lysis. Endogenous STX8 was immunoprecipitated with anti-STX8 followed by SDS-PAGE and immunoblotting with anti-STX8, anti-FLAG, anti-VAMP8, and anti-VAMP7. The VAMP8/STX8 and VAMP7/STX8 ratios were calculated for the indicated lanes and shown as mean and standard error (normalized relative to No Virus Empty Vector, which was set to 100%). Differences that were statistically significant from No Virus Empty Vector as determined by one-way ANOVA are indicated by (*). Exact p values are as follows (from left to right, top to bottom): p = 0.0007, p = 0.6986, p = 0.0093, p = 0.4523, p < 0.0001, p = 0.7960. (D) Endogenous STX8 was immunoprecipitated from HeLa cells (WT or IFITM3 KO) with anti-STX8 followed by SDS-PAGE and immunoblotting with anti-STX8, anti-IFITM3, and anti-VAMP8. Light chain immunoglobulin was used as loading control. An isotype matched antibody was used as a control for immunoprecipitation. The VAMP8/STX8 ratio was calculated for the indicated lanes and shown as mean and standard error (normalized relative to WT, which was set to 100%). Differences that were statistically significant from WT as determined by student’s T test are indicated by (*). Exact p value: p = 0.0009. Numbers and tick marks left of blots indicate position and size (in kilodaltons) of protein standard in ladder. The (#) symbol denotes the presence of light chain immunoglobulin. Immunoblots were performed independently three times, and a representative example is shown. Ig; immunoglobulin. IP; immunoprecipitation. EV; Empty Vector. WT; wild-type. PCC; Pearson’s Correlation Coefficient.

Figure 6:. IFITM3 inhibits assembly of late endosome homotypic fusion machinery and promotes the delivery of endosomal cargo to lysosomes.

(A) HEK293T stably expressing Empty Vector, IFITM3 WT-FLAG, or IFITM3 G95L-FLAG were co-transfected with STX7-HA, STX8-HA, Vti1b-HA, and VAMP8-Myc. Following whole cell lysis, samples were either boiled at 100°C (+) or not (−). SDS-PAGE and immunoblotting was performed with anti-Myc and anti-FLAG. Actin was used as loading control. Numbers and tick marks left of blots indicate position and size (in kilodaltons) of protein standard in ladder. Immunoblots were performed independently twice, and a representative example is shown. (B) HEK293T cells stably expressing Empty Vector, IFITM3 WT-FLAG, or IFITM3 G95L-FLAG were pulsed with Dextran Alexa Fluor 488 for 2 hrs followed by addition of Magic Red for 5 mins. Living cells were then analyzed immediately by confocal immunofluorescence microscopy. Colocalization between Dextran and Magic Red was measured by calculating the Pearson’s Correlation Coefficient using Fiji software. Coefficients were calculated from medial Z-slices from three fields of view containing 5–15 cells per condition and presented as means and standard error. Scale bar = 15 microns. PCC; Pearson’s Correlation Coefficient.

Figure 7:. IFITM3 traffics incoming IAV to lysosomes in a G95-dependent manner.

(A) HEK293T cells stably expressing Empty Vector, IFITM3 WT-FLAG, or IFITM3 G95L-FLAG inoculated with IAV at a multiplicity of infection of 1 and incubated on ice for 40 minutes. Cells were washed, fresh medium containing Lysotracker Deep Red (50 nM) was added, and cells were placed at 37°C for 4 hours. Four hours later, cells were fixed and immunostained with anti-NP antibody and Hoechst and analyzed by confocal immunofluorescence microscopy. Mock-inoculated (PBS) Empty Vector cells were used as a negative control. Scale bar = 15 microns. Medial Z-slices representative of each condition are shown. (B) The percentage of NP+ nuclei was quantified in Empty Vector, IFITM3 WT-FLAG, or IFITM3 G95L-FLAG cells. Symbols represent fields of view containing approximately 15–30 total cells and mean percentage plus standard error is shown. Differences that were statistically significant between the indicated conditions as determined by one-way ANOVA are indicated by (*). Exact p values are as follows (from left to right): p < 0.0001, p = 0.0028. (C) The Corrected Total Cell Fluorescence (CTCF) intensity of Lysotracker staining was quantified in Empty Vector, IFITM3 WT-FLAG, or IFITM3 G95L-FLAG cells. Symbols represent fields of view containing approximately 15–30 total cells and mean CTCF intensity plus standard error is shown. Differences that were statistically significant between the indicated conditions as determined by one-way ANOVA are indicated by (*). Exact p values are as follows (from left to right): p = 0.0005, p = 0.0001. (D) Examination of NP signal in relation to Lysotracker staining in HEK293T stabling expressing IFITM3 WT-FLAG or IFITM3 G95L-FLAG. White boxes indicate fields of view selected for detailed analysis. Right: detailed view of three different Z slices from confocal stack. White boxes indicate fields of view selected for further detailed analysis. White arrows indicate bright NP puncta that colocalize with Lysotracker or which tightly appose a Lysotracker+ compartment. Colocalization between NP and Lysotracker was measured by calculating the Pearson’s Correlation Coefficient using Fiji software. Coefficients were calculated from medial Z-slices from three fields of view containing 5–15 cells per condition and presented as means plus standard deviation. Scale bar = 15 microns. EV; Empty Vector. WT; wild-type. hpi; hours post-inoculation. CTCF; Corrected Total Cell Fluorescence. PCC; Pearson’s Correlation Coefficient.

Figure 8:. A model of inhibition of homotypic late endosome fusion by IFITM3.

IFITM3 selectively regulates the trans-SNARE assembly driving late endosome-late endosome fusion (STX7-STX8-Vti1b-VAMP8), while enabling late endosome-lysosome fusion. As a result, endocytic cargos (including viruses) are more efficiently trafficked to lysosomes in IFITM3-expressing cells. LE; late endosome. Lys; lysosome.