Transmission electron microscopic detection of the release and extracellular fate of large, multivesicular EVs secreted by different cell lines and cells in mouse organs

Major steps of the release of large, multivesicular EVs (MV-lEVs) were detected in the case of all tested cell lines including the immortal, non-tumorous HEK293T-PalmGFP (A,H,O), HEK293 (B,I,P), the tumorous cell lines HepG2 (C,J,Q) and HT29 (D,K,R), the beating cardiomyocyte cell line HL1 (E,L,S) and the primary suspension of bone marrow derived mast cells (BMMCs) (F,M,T). The different phases of EV secretion were also captured in the circulation of mouse kidney (V) and liver (W,X). According to the electron micrographs, we found evidence for the budding (A-G,X) and secretion (H-N,V,W) of the MV-lEVs. We also detected the extracellular rupture of the limiting membrane of the released MV-lEVs with the escape of the intraluminal vesicles (ILVs) by a “torn bag mechanism” (O-U,V). Although it is not always clear whether the secreted MV-lEVs have a single or double limiting membrane, several micrographs suggest the presence of the double membrane (Y-AF) in the secreted MV-lEVs. In the case of BMMCs (Y), the release phase of a multivesicular structure is captured. The bottom portion of this structure embedded in the cytoplasm is surrounded by a single membrane (white arrow head) while the upper (budding) portion is covered by double membrane (asterisk). In the shematic figures (G, N, U) the limiting membrane of MV-lEV presumably with plasma-membrane origin was indicated by red, the original limiting membrane of intracellular amphisomes, which may fragmented during the release process was indicated by blue while the ILVs of the MV-lEV were shown by gray color.

Detection of conventional sEV markers and the LC3 protein in HEK293T-PalmGFP cell-derived EVs

Widely used sEV markers (CD63, CD81, ALIX and TSG101) and LC3B were tested in MV-lEVs found in the microenvironment of the releasing cells by confocal microscopy after in situ fixation (A-F). Normalized fluorescence intensities were calculated in order to determine the relative localization of the limiting membrane (PalmGFP), the conventional sEV markers and the LC3B signal (G-L). Fluorescence intensity peaks of sEV markers were largely overlapping with each other, while the LC3B signal and the sEV markers showed separation. Co-localization rates were also calculated (M). The sEV markers co-localized with one another as no significant difference was found among them. In contrast, low co-localization rates were detected between the “classical” sEV markers and LC3B (one-way ANOVA, p<0.0001, n=8-26 confocal images). Real time release of LC3 positive sEVs by the “torn bag mechanism” was studied in the case of HEK293T-PalmGFP-LC3RFP cells by Elyra7 SIM2 super-resolution live cell imaging (N,O). Images were recorded continuously, and selected serial time points are shown. LC3 positive, red fluorescent small particles were released within a 5 min timeframe (O) and are indicated by white arrows. Presence of CD63 and LC3B were detected in the case of a sEV fraction separated from serum-free condition medium using immunogold transmission electronmicroscopy (TEM). HEK293T-PalmGFP derived sEV fraction is shown by negative-positive contrast without immune labelling (P). In double-labelled immunogold TEM images (Q,R), distinct LC3B positive (Q) and CD63 positive (R) sEVs were found. However, CD63-LC3B double positive EVs were not detected. Black arrowheads indicate 10 nm gold particles identifying LC3B, while white arrowheads show 5 nm gold particles corresponding to the presence of CD63. Quantitative analysis of TEM images was performed (S), and the diameters of different EV populations were determined. The LC3B negative population was significantly smaller than the LC3B positive one (p<0.0001, t-test; n=79-100). No difference was detected when the immunogold labelled sEV fraction (either LC3B positive or negative, LC3B+/-) and the unlabeled sEV fraction (sEV) were compared (p<0.05, t-test, n=112-179).

Amphiectosome release and its modulation

Based on our data, a model of amphiectosome release was generated (A). According to this model, the fusion of MVBs and autophagosomes forms amphisomes. The LC3B positive membrane layer (indicated in cyan) undergoes disintegration and forms LC3B positive ILVs inside the amphisome. Later, the amphisome is released into the extracellular space by ectocytosis and can be identified extracellularly as an amphiectosome. Finally, the limiting membrane(s) of the amphiectosome is ruptured and the ILVs are released as sEVs into the extracellular space by a “torn bag mechanism”. Steps of amphisome formation including LC3 positive ILV formation in 30 µM Chloroquine-treated HEK293T-PalmGFP cells was followed by super-resolution (STED) microscopy (B-F). The super-resolution STED channels were LC3B (cyan) and CD63 (magenta), while yellow indicates the confocal PalmGFP signal. Intracellular vesicular structures (such as endosomes, MVBs and amphisomes) may recieve Palm-GFP from the plasma membrane. An MVB (B), an autophagosome with Palm-GFP negative membrane (D), fusion of an autophagosome and an MVB (C), formation of LC3B positive ILVs in an amphisome (F) and a mature amphisome (E) were detected. To confirm the origin of the external membrane layer of amphiectosomes, fluorescently labelled WGA was applied. The plasma membrane of the living non-fluorescent HEK293 cells was labelled. As the external membrane of the budding amphiectosome was WGA positive, its plasma membrane origin is confirmed (G). In order to further support our model on amphiectosome release and “torn bag” EV secretion, different in vitro treatments were applied. Cytochalasin B, Colchicine, Chloroquine, Bafilomycin A1 and Rapamycin were used to modulate amphiectosome release. Targeted molecular processes are summarized (H). While Cytochalasin B inhibits actin-dependent membrane budding and cell migration, Colchicine blocks the microtubule-dependent intracellular trafficking. While Chloroquine and Bafilomycin have similar, Rapamycin has opposite effect on lysosome-autophagosome or lysosome-amphisome fusion. Chloroquine and Bafilomycin inhibit lysosomal degradation while Rapamycin accelerates it. Based on confocal microscopy, Cytochalasin B (CytoB) did not alter the dynamics of amphiectosome release (I). In contrast, both Colchicine (Colch) and Rapamycin (Rapa) significantly inhibited the release of amphiectosomes, while Chloroquine (Chloro) and Bafilomycin (Bafilo) increased the release frequency. There was no difference between the effect of Chloroquine and Bafilomycin (I). Results are shown as mean ± SD of 3-4 independent biological replicates, analyzed by one-way ANOVA and Student’s t test, *: p<0.05, **: p<0.01, ns: non-significant. Original LASX files, which served as a basis of our quantification, are publicly available (doi: 10.6019/S-BIAD1456). Example for the calculation is shown on Fig3_S1H. Presence of membrane-bound (lipidated) LC3II was tested by Western blotting. The total protein content of serum-, cell- and large EV-depleted conditioned medium of HEK293T-PalmGFP (PalmGFP) and HEK293T-PalmGFP-LC3RFP (PalmGFP-LC3RFP) cells was precipitated by TCA and 20 µg of the protein samples were loaded on the gel (J). The lipidated LC3II band was detected in all cases. Relative expression of control (Ctrl) and Chloroquine (Chloro)-treated samples were determined by densitometry. Chloroquine treatment increased the LC3II level by approximately two fold. Results are shown as mean ± SD of n=6 biological replicates.

Comparison of amphiectosomes and migrasomes

Commonly used sEV markers (CD63, CD81) and TSPAN4, a suggested migrasome marker, were tested in in situ fixed intact MV-lEVs of HEK293T-PalmGFP (A,B) and HT29 (C-F) cells by confocal microscopy. Normalized fluorescence intensities were calculated in order to determine the relative localization of the limiting membrane (with PalmGFP or lactadherin labelling) and the CD63/TSPAN4 and CD81/TSPAN4 markers (G-L). In the case of HEK293T-PalmGFP-derived EVs, we did not find migrasomes with TSPAN4 in their limiting membrane. The TSPAN4 signal was only detected intraluminally in the MV-lEVs. The limiting membranes of HT29-derived MV-lEVs were either TSPAN4 positive and negative. The co-localization rate between the limiting membrane and TSPAN4 was low in case of HEK293T-PalmGFP-derived EVs. In the case of HT29 cells, two MV-lEV populations were identified: one with low and one with high co-localization rates (M). Live cell imaging of HEK293T-PalmGFP-LC3RFP cells showed retraction fiberassociated MV-lEVs with or without intraluminal LC3 positivity (N,O). Using TEM, we could identify structures with retraction fiber-associated morphology in the case of HL1 (P), HEK293T-PalmGFP (Q) and BMMC (R) cells. For comparison, budding of amphiectosomes of the same HL1 (S), HEK293TPalmGFP (T) and BMMC cells (U) are shown (without being associated with long retractions fibers).