ALIX- and ESCRT-III-dependent sorting of tetraspanins to exosomes

Abstract

The intraluminal vesicles (ILVs) of endosomes mediate the delivery of activated signaling receptors and other proteins to lysosomes for degradation, but they also modulate intercellular communication when secreted as exosomes. The formation of ILVs requires four complexes, ESCRT-0, -I, -II, and -III, with ESCRT-0, -I, and -II presumably involved in cargo sorting and ESCRT-III in membrane deformation and fission. Here, we report that an active form of the ESCRT-associated protein ALIX efficiently recruits ESCRT-III proteins to endosomes. This recruitment occurs independently of other ESCRTs but requires lysobisphosphatidic acid (LBPA) in vivo, and can be reconstituted on supported bilayers in vitro. Our data indicate that this ALIX- and ESCRT-III-dependent pathway promotes the sorting and delivery of tetraspanins to exosomes. We conclude that ALIX provides an additional pathway of ILV formation, secondary to the canonical pathway, and that this pathway controls the targeting of exosomal proteins.

Graphical abstract

Figure 1.

ALIXΔPRR recruits the ESCRT-III protein CHMP4B. (A) Outline of ALIX organization into three structural domains: the BRO1 domain, the V-domain, and the autoinhibitory PRR. Experiments were performed using ALIXΔPRR, lacking PRR. (B) HeLa-MZ cells were transfected with ALIX-mCherry or ALIXΔPRR-mCherry and imaged by confocal microscopy. Scale bar: 10 µm. (C and D) As in B, but cells were labeled with antibodies against LBPA and EEA1 (C). The boxed areas in the merged images are shown in higher magnification. The fraction of endosomes containing ALIXΔPRR-mCherry and EEA1 or LBPA was quantified by calculating the Manders overlap coefficients (D). Box plot, median (box central line); 25% and 75% percentiles (box edges); 10% and 90% percentiles (whiskers); outliers (black circles). Mann–Whitney U test; ***, P < 0.001; n = 50 cells from three independent experiments; scale bar: 10 µm. (E and F) HeLa GFP-CHMP4B cells transfected with ALIXΔPRR-mCherry were imaged (E) by confocal microscopy. The number of CHMP4B granules was quantified (F) in cells expressing ALIXΔPRR-mCherry (ΔPRR) and untransfected cells (UT). Box plot as in D; n = 90 cells from three independent experiments. Scale bar: 10 µm. Pearson correlation coefficient for ALIX and CHMP4B in E: 0.72, n = 50 cells. (G and H) HeLa GFP-CHMP4B cells transfected for 18 h with ALIXΔPRR (ΔPRR) or an empty vector as control (ctrl) were fractionated by flotation in a sucrose density gradient. The postnuclear supernatant (PNS) and the light membranes (LM) were analyzed by Western blotting (G) using antibodies against CHMP4B, ALIX, and tubulin and RAB5 (equal loading controls). CHMP4B quantification in LM fractions by densitometry (H), relative to RAB5. Boxes, mean; error bars, ±SD (n = 3, from three independent experiments); t test, ctrl versus ΔPRR; *, P < 0.05.

Figure S1.

Stable cell line expressing GFP-CHMP4B and analysis of full-length ALIX versus ALIXΔPRR. (A and B) HeLa Kyoto cells stably expressing GFP-CHMP4B (HeLa GFP-CHMP4B cells) were imaged (A) by confocal microscopy and analyzed by Western blotting (B) using anti-CHMP4B antibody. Scale bar in A: 10 µm. (C and D) HeLa GFP-CHMP4B cells were transfected for 18 h with full-length ALIX (FL) or ALIXΔPRR (ΔPRR). The postnuclear supernatant (PNS) and the light membranes (LM) were analyzed by Western blotting using antibodies against CHMP4B and ALIX (C), as well as RAB5 (equal loading control). The relative amounts of CHMP4B in LM fractions was quantified by densitometry (D), using RAB5 intensity for the normalization of CHMP4B signal. Boxes, mean; error bars, ±SD (n = 3, from three independent experiments); **, P < 0.01.

Figure 2.

ALIXΔPRR recruits ESCRT-III onto late endosomes. (A) HeLa GFP-CHMP4B cells transfected with ALIXΔPRR-mCherry were stained with anti-LAMP1 antibodies (boxed area: higher-magnification views). Cells were permeabilized with saponin before fixation to reduce the cytosolic staining. Arrows point at endosomes containing ALIXΔPRR, CHMP4B, and LAMP1. (B) HeLa-MZ cells were transfected with ALIXΔPRR-mCherry and GFP-RILP (boxed area: higher-magnification view). (C) HeLa GFP-CHMP4B cells transfected with RILP and ALIXΔPRR-mCherry were labeled with antibodies against LAMP1. (D and E) HeLa GFP-CHMP4B cells transfected with ALIXΔPRR-mCherry were treated with 10 µM nocodazole for 2 h (to limit endosome movement), and endosomes containing both markers were analyzed by FRAP. White circles, photobleached regions (D); time 0, before photobleaching. Fluorescence recovery calculated for endosomes containing GFP-CHMP4B and ALIXΔPRR-mCherry (E). Dots, mean; shaded area, ±SD (n = 28 endosomes from three independent experiments). Scale bars (A–D): 10 µm.

Figure 3.

ALIXΔPRR recruits VPS22 and ESCRT-III proteins to endosomes. (A–D) HeLa cells stably expressing CHMP1B-Flag (A and C) or CHMP1A-V5 (B and D) were transfected (C and D) or not (A and B) with ALIXΔPRR-mCherry. Cells labeled with antibodies against Flag (A and C) and V5 (B and D) epitopes were analyzed by confocal microscopy (boxed areas: higher-magnification views). Scale bar: 10 µm. (E and F) HeLa GFP-CHMP4B cells transfected with ALIXΔPRR for 18 h were fractionated as in Fig. 1 G. The postnuclear supernatant (PNS) and light membranes (LM) fractions were analyzed by Western blotting with antibodies against the indicated proteins; RAB5, equal loading control. Quantification of each protein in LM fractions by densitometry (F), as in Fig. 1 H. Boxes, mean; error bars, ±SD (n = 3, from three independent experiments); t test; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.

Figure 4.

CHMP4B recruitment to endosomes depends on the CHMP4B and LBPA binding sites in ALIX BRO1 domain and can be recapitulated in vitro. (A and B) HeLa GFP-CHMP4B cells were transfected with ALIXΔPRR-mCherry (ΔPRR), ALIXΔPRR-I212D-mCherry (I212D), and ALIXΔPRR-QQ-mCherry (QQ); boxed areas: higher-magnification views. Scale bar: 10 µm. The number of CHMP4B granules was quantified and analyzed (B) as in Fig. 1 F. Mann–Whitney U test; *, P < 0.05; n = 90 cells from three independent experiments; UT, untransfected. (C–E) HeLa GFP-CHMP4B transfected with ALIXΔPRR (ΔPRR), ALIXΔPRR-I212D (I212D), ALIXΔPRR-QQ (QQ), or an empty vector (Ctrl) were as in Fig. 1 G. The PNS (C) and LM (D) fractions were analyzed by Western blotting with antibodies against CHMP4B and ALIX; tubulin and RAB5, loading controls. Quantification of each protein in LM fractions by densitometry (E), as in Fig. 1 H. Boxes, mean; error bars, ± SD (n = 3, from three independent experiments); t test; ***, P < 0.001. (F–H) Supported bilayers prepared with the lipid composition: PI (DOPC:DOPE:PI, 6.99:2:1 mol), PI + PS (DOPC:DOPE:PI:DOPS, 4.99:2:1:2 mol), and PI + LBPA (DOPC:DOPE:PI:LBPA, 4.99:2:1:2 mol) and labeled with N-Rhodamine PE (red) were preincubated (F, bottom row) or not (F, top row) with ALIX BRO1 domain for 40 min, and then for 25 min with CHMP4B-488 (green), and analyzed by time-lapse double-channel confocal microscopy. Each panel in F corresponds to a view in the green (CHMP4B-488) channel after 25 min; boxed areas, smaller view of the same field in the red channel (N-Rhodamine PE); dashed lines, areas covered with lipids. The time course of CHMP4B-488 association to LBPA-containing bilayers preincubated with ALIX is shown in G, corresponding to a magnified view of the boxed area in F. The intensity of CHMP4B-488 fluorescence was quantified on supported bilayers after 25 min (H). Box plot as in Fig. 1 D; Mann–Whitney U test; ***, P < 0.001; n = 20 lipid patches from two independent experiments; ns, not significant; scale bar: 30 µm. (I and J) The experiments were as in F using supported bilayers with the PI + LBPA composition only, except that I212D and QQ mutants in the ALIX BRO1 domain were used (I) instead of ALIX BRO1 domain. (J) Box plot as in Fig. 1 D; Mann–Whitney U test; ***, P < 0.001; n = 40 lipid patches from two independent experiments; scale bar: 30 µm.

Figure S2.

Recombinant CHMP4B and ALIX BRO1 domain protein purification. (A and B) A typical SDS-PAGE gel of recombinant human CHMP4B purification was stained with Coomassie brilliant blue (A). The gel compares aliquots taken at sequential steps of the purification: (1) Bacterial lysate expressing His-MBP-CHMP4B. (2) Protein elution from an MBPTrap column showing the His-MBP-CHMP4B purified protein. (3) Protein fragments obtained after incubation with TEV protease (His-MBP and CHMP4B). During the final step of the protein purification procedure, the protein was centrifuged to remove protein aggregates. Aliquots of the supernatant (4) and resuspended pellet (5) are shown. The same gel was exposed to UV light at 320-nm wavelength (B). Fluorescently labeled CHMP4B-488 is visible in wells 4 and 5. (C) A typical SDS-PAGE gel of the purification of human ALIX BRO1 domain was stained with Coomassie brilliant blue. The gel compares aliquots taken at sequential steps of the purification: (1) Bacterial lysate expressing GST-ALIX BRO1 domain. (2) GST-ALIX BRO1 domain purified with Glutathione Sepharose beads. (3) Protein fragments obtained after the incubation with PreScission protease (ALIX BRO1 domain and GST). During the final step of the protein purification procedure, the protein was centrifuged to remove protein aggregates. An aliquot of the supernatant (4) and the resuspended pellet (5) are shown. (D) A typical SDS-PAGE gel after purification of the human ALIX BRO1 I212D (lane 1) and QQ (lane 2) mutants, stained with Coomassie brilliant blue.

Figure 5.

CHMP4B recruitment by ALIXΔPRR is independent of other ESCRT proteins. (A and B) HeLa GFP-CHMP4B (green) cells were transfected with an siRNA-resistant mutant of ALIXΔPRR-mCherry (red), and treated with siRNAs against the indicated proteins; siCtrl 1, siRNA control pool used for siRNA pools against all targets except ALIX; siCtrl 2, single siRNA sequence used as control for the single siRNA against ALIX. Scale bar: 10 µm. The number of CHMP4B granules was quantified (B) as in Fig. 1 F. Box plot as in Fig. 1 D; Mann–Whitney U test; *, P < 0.05; n = 83 cells from three independent experiments; ns, not significant; UT, untransfected. (C) The KD efficiency of the different targets in A and B analyzed by Western blotting using antibodies against the indicated proteins. Tubulin was used as an equal loading control.

Figure 6.

ALIXΔPRR induces the endosomal accumulation of ubiquitinated proteins. (A) HeLa GFP-CHMP4B cells were transfected with ALIXΔPRR-mCherry, permeabilized with saponin 0.01% in PBS before fixation, and labeled with an antibody against conjugated ubiquitin (boxed area: higher-magnification view). Arrows point at endosomes containing ALIXΔPRR, CHMP4B, and conjugated ubiquitin. Scale bar: 10 µm. (B) HeLa-MZ cells processed as in A were labeled with antibodies against conjugated ubiquitin and LAMP1, and nuclei were stained with Hoechst. Scale bar: 10 µm. (C and D) The nuclei of HeLa cells treated as in A were stained with Hoechst to illustrate the accumulation of ubiquitinated proteins in cells expressing ALIXΔPRR-mCherry, compared with cells expressing ALIXΔPRR-I212D or ALIXΔPRR-QQ (C). Scale bar: 10 µm. The ubiquitin intensity per cell was quantified (D) in cells expressing ALIXΔPRR-mCherry (ΔPRR) and compared with either mutant. Box plot as in Fig. 1 D; Mann–Whitney U test; n = 135 cells from three independent experiments. ns, not significant. ***, P < 0.001. (E and F) HeLa GFP-CHMP4B cells transfected with ALIXΔPRR (ΔPRR) or an empty vector (Ctrl) were fractionated as in Fig. 1 G. The postnuclear supernatant (PNS) and light membranes (LM) fractions were analyzed by Western blotting (E) with antibodies against conjugated ubiquitin and ALIX; RAB5, equal loading control. The relative amounts of conjugated ubiquitin in PNS and LM were quantified by densitometry (F), using RAB5 intensity to normalize the signal. Boxes, mean; error bars, ±SD (n = 3, from three independent experiments); t test; **, P < 0.01.

Figure 7.

ALIXΔPRR does not affect EGFR transport to lysosome and degradation, lysosomal cathepsin D maturation, and endolysosomal pH. (A and B) HeLa GFP-CHMP4B cells transfected for 18 h with ALIXΔPRR (ΔPRR) or an empty vector (Ctrl) were fractionated as in Fig. 1 G (≈70% cells were transfected). The total cell lysate and LM (A) were analyzed by Western blotting with antibodies against EGFR and ALIX; RAB5, equal loading control. The RAB5 and ALIXΔPRR panels (A) are the same as in Fig. 6 E, because the same membrane was used for blots against conjugated ubiquitin (Fig. 6 E) and EGFR (A). The relative amounts of EGFR in total cell lysate and light membranes (LM) was quantified by densitometry (B), using RAB5 intensity to normalize the EGFR signal. Boxes, mean; error bars, ±SD (n = 3, from three independent experiments); t test; ns, not significant. (C and D) HeLa GFP-CHMP4B cells were transfected for 18 h with ALIXΔPRR (ΔPRR) or an empty vector (Ctrl; ≈70% cells were transfected). Cells were sequentially incubated at 37°C for 3 h in serum-free medium, for 1 h with 10 µg/ml cycloheximide, and finally for the indicated time with 100 ng/ml EGF. Samples were collected and analyzed by Western blotting with antibodies against EGFR and ALIX; tubulin, equal loading control (C). The relative amount of EGFR was quantified by densitometry (D), using tubulin intensity to normalize the EGFR signal. Circles, mean; error bars, ±SD (n = 3, from three independent experiments). t test. There is no significant difference between Ctrl and ΔPRR for the different time points. (E and F) The experiments were as in C and D, except that cells were transfected with ALIXΔPRR-mCherry and analyzed by automated triple-channel fluorescence microscopy. In contrast to A–D, 30–40% cells were transfected to facilitate the compared analysis of transfected versus untransfected cells. Each row represents the imaged area obtained from one 96-well plate (E). Scale bar: 1 mm. For each time point, >12,000 cells were imaged, and ALIXΔPRR-mCherry positive (ΔPRR) or negative (no ΔPRR) cells were segmented using mCherry signal (insert in F). The endosomal EGFR was also segmented (EGFR granules), and the average EGFR granule intensity per cell was plotted for each time point (F). Because at time 0 min, EGFR is mostly at the plasma membrane, 15 min was used as first time point. Circles, mean; error bars, ± SD; t test. There is no significant difference between untransfected and ALIXΔPRR-mCherry transfected cells. (G and H) HeLa GFP-CHMP4B were untransfected (UT) or transfected for 18 h with an empty vector (Ctrl) or ALIXΔPRR (ΔPRR). Total cell lysates were analyzed by Western blotting with antibodies against ALIX or cathepsin D (G); tubulin, equal loading control. The relative amounts of cathepsin D were quantified by densitometry (H), using tubulin intensity to normalize the signal. Boxes, mean; error bars, ±SD (n = 3, from three independent experiments); t test; ns, not significant. (I and J) Cells transfected with ALIXΔPRR (ΔPRR) as in E and F were treated with EGF-647 for 45 min at 37°C to label endosomes, and then with Lysosensor to label acidic compartments. Cells were then analyzed by automated microscopy as in E and F, and the intensity of the lysotracker signal per cell (J) was quantified. Box plot as in Fig. 1 D; n = 1,000 endosomes from four independent experiments. AU, arbitrary units. Scale bar: 10 μm.

Figure 8.

ALIXΔPRR stimulates the release of exosomes containing tetraspanins. (A and B) HeLa GFP-CHMP4B cells transfected for 18 h with ALIXΔPRR (ΔPRR) or GFP as a control and incubated with exosome-free medium for 24 h. The cell medium was collected, and exosomes were isolated by differential centrifugation. The cell lysate and the exosome samples were analyzed by Western blotting using antibodies against the indicated proteins. The same amount of protein was loaded for each condition (A). The amounts of protein were quantified by densitometry and are expressed relative to free GFP. In B, boxes, mean; error bars, ±SD (n = 3, from three independent experiments); t test; *, P < 0.05, ***, P < 0.001; ns, not significant. (C and D) HeLa GFP-CHMP4B cells were treated with siRNA against ALIX or nontarget siRNAs (siCtrl) and incubated with exosome-free medium for 24 h. Exosomes were isolated as in A and B, and then cell lysates and exosomes (C) were analyzed by Western blotting as in A and B. The relative amount of protein in exosome fractions was quantified by densitometry (D). Boxes, mean; error bars, ±SD (n = 3, from three independent experiments); t test; *, P < 0.05, ***, P < 0.001; ns, not significant. (E and F) The experiment was as in C and D except that siRNAs against CHMP6 and not against ALIX were used. The upper band in the syntenin blot of siCtrl is CHMP6, because the same membrane was reused for incubation with both antibodies. The relative amounts of protein were then quantified by densitometry (F). Boxes, mean; error bars, ±SD (n = 3, from three independent experiments); t test; *, P < 0.05, **, P < 0.01. (G and H) HeLa GFP-CHMP4B cells were transfected with WT CD9 (WT) tagged with V5 or with the V5-CD9/3R mutant (3R). Total cell lysates and exosomes prepared as in A and B were analyzed by Western blotting using antibodies against V5, ALIX, flotillin-1, and syntenin (G). The V5 blot is shown after short and long exposure, so that WT V5-CD9 and V5-CD9/3R can be better compared. The relative amounts of CD9 were quantified by densitometry (H). Boxes, mean; error bars, ±SD (n = 3, from three independent experiments); t test; *, P < 0.05.

Figure S3.

The QQ mutant of ALIXΔPRR does not stimulate the release of exosomes containing tetraspanins. (A) HeLa GFP-CHMP4B cells transfected for 18 h with ALIXΔPRR (ΔPRR) or GFP as a control and incubated with exosome-free medium for 24 h. The cell medium was collected, and exosomes were isolated by differential centrifugation. The cell lysates and the exosome fractions were analyzed by Western blotting using antibodies against CD63, calnexin ALIX, CHMP4, and GFP. Calnexin was not found in the exosome fractions prepared from cells expressing ALIXΔPRR; neither were GFP-CHMP4B and GFP. The same amount of protein was loaded for each condition. UT, untransfected cells. (B–D) HeLa GFP-CHMP4B cells were transfected for 18 h with ALIXΔPRR (ΔPRR) or with the QQ mutant of ALIXΔPRR (QQ). After processing as in A, the cell lysates and exosome fractions were analyzed by Western blotting using antibodies against the indicated proteins (B). The relative amount of protein in exosome fractions was quantified by densitometry (C and D). C shows the intensity of the QQ mutant relative to ALIXΔPRR, and D shows the relative intensities of the indicated markers in the ALIXΔPRR fractions relative to the fractions prepared from cells expressing the QQ mutant. Boxes, mean; error bars, ±SD (n = 3, from three independent experiments); t test; *, P < 0.05, **, P < 0.01, ***, P < 0.001; ns, not significant.

Figure 9.

ALIX controls tetraspanin levels in endosomes and exosomes. (A and B) HeLa-MZ cells were transfected for 18 h with ALIXΔPRR-mCherry (A) or mCherry (not shown) as control: only 30–40% cells were transfected to facilitate the analysis of transfected versus untransfected cells. After fixation, cells were labeled with antibodies against LAMP1 as well as antibodies against either CD81 or CD63, GM130, and calnexin (A and B), as well as EGFR, syntenin, or flotillin-1 (Fig. S4). Samples were analyzed by automated triple-channel fluorescence microscopy (A). Cells were segmented using the ALIXΔPRR-mCherry or the mCherry signal to identify transfected and untransfected cells. To quantify the distribution of each marker within LAMP1-positive endolysosomes, a mask of the LAMP1 staining pattern was created by segmentation, and this mask was applied onto the staining of the given marker (CD81, CD63, EGFR, syntenin, flotillin-1, calnexin, and GM130), so that the integrated intensity per cell of the marker present in LAMP1-positive endolysosomes could be quantified in the mask and plotted (B). To illustrate the image analysis process, the presence of each marker in the LAMP1 mask only was visualized on the computer screen (reflecting exactly what was being quantified), and the corresponding image of the mask was captured. The righthand panels in A show these screen captures illustrating the staining of each marker within LAMP1-endosomes only (LAMP1 mask). UT, untransfected control cells. Boxes, mean; error bars, ±SD (n = >5,000 cells); t test; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant. Scale bar: 10 µm. (C) HeLa-MZ cells were treated with siRNA against ALIX or nontarget siRNAs (siCtrl), as in Fig. 8 (C and D) and were processed for automated fluorescence microscopy as in A and B. ALIX KD efficiency was analyzed by Western blotting (insert in C). The corresponding micrographs are shown in Fig. S5 (A and B). The integrated intensity of each marker per cell was quantified and normalized to the corresponding control showing cells treated with nontarget siRNAs (C). Boxes, mean; error bars, ±SD (n = >5,000 cells); t test; *, P < 0.05, **, P < 0.01; ns, not significant.

Figure S4.

EGFR, syntenin, and flotillin-1 in endosomes of cells expressing ALIXΔPRR. HeLa-MZ cells were transfected for 18 h with ALIXΔPRR-mCherry (A) or mCherry (not shown) as a control; only 30–40% of the cells were transfected to facilitate the compared analysis of transfected versus untransfected cells. After fixation, cells were labeled with antibodies against LAMP1 as well as antibodies against EGFR, syntenin, or flotillin-1 (panels showing CD81 or CD63, GM130, and calnexin are shown in Fig. 9). Samples were then processed for high-throughput automated triple-channel fluorescence microscopy. Cells were segmented using the ALIXΔPRR-mCherry or the mCherry signal to identify transfected and untransfected cells. To quantify the distribution of each marker within LAMP1-positive endolysosomes, a mask of the LAMP1 staining pattern was created by segmentation, and this mask was applied onto the staining of the given marker (EGFR, syntenin, flotillin-1) so that the integrated intensity per cell of the marker present in LAMP1-positive endolysosomes could be quantified in the mask (Fig. 9 B). To illustrate the image analysis process, the presence of each marker in the LAMP1 mask only was visualized on the computer screen (reflecting exactly what was being quantified), and the corresponding image of the mask was captured. The righthand panels show these screen captures illustrating the staining of each marker within LAMP1-endosomes only (LAMP1 mask). Scale bar: 10 µm.

Figure S5.

ALIX depletion does not affect the endosomal distribution of the various markers. (A and B) HeLa-MZ cells were treated with siRNA against ALIX or nontarget siRNAs (siCtrl) and processed for automated fluorescence microscopy using the same antibodies as in Fig. 9, A and B; and Fig. S4 to detect CD63, CD81, EGFR, and syntenin (A), as well as flotillin-1, GM130, and calnexin (B). In cells depleted of ALIX with siRNAs, the intensity of the granules per cell was quantified and compared with mock-treated control cells. To quantify the distribution of each marker within LAMP1-positive endolysosomes, a mask of the LAMP1 staining pattern was created by segmentation, and this mask was applied onto the staining of the given marker (A: CD63, CD81, EGFR, and syntenin; B: flotillin-1, GM130, and calnexin) so that the integrated intensity per cell of the marker present in LAMP1-positive endolysosomes could be quantified (Fig. 9 C). To illustrate the image analysis process, the presence of each marker in the LAMP1 mask only was visualized on the computer screen (reflecting exactly what was being quantified), and the corresponding image of the mask was captured. These screen captures illustrate the staining of each marker within LAMP1-endosomes only (LAMP1 mask). Scale bar: 10 µm.