Introduction

Extracellular vesicles (EVs) are phospholipid bilayer enclosed structures^1–3, which have important roles in cellular homeostasis and intercellular communication. Exosomes have been defined as small (∼ 50-200 nm) EVs (sEVs) of endosomal origin^{1, 3}. Although autophagy is a major cellular homeostatic mechanism, and is implicated in a broad spectrum of human diseases, the intersection of autophagy and exosome secretion remains poorly understood. Recently, regulatory interactions have been shown between autophagy-related molecules and EV biogenesis^{4, 5}. Furthermore, the LC3-conjugation machinery was demonstrated to specify the cargo packaged into EVs⁶. Importantly, both others and we reported the secretion of LC3-carrying exosomes^{6, 7}. Particularly relevant to the findings presented here, is the implication of amphisomes (hybrid organelles formed by the fusion of late endosomes/multivesicular bodies (MVBs) with autophagosomes^{8, 9}) in EV biogenesis. It was suggested that fusion of the limiting membrane of amphisomes with the plasma membrane of cells results in a subsequent release of exosomes by exocytosis^{1, 3, 10}. The current study was prompted by our recent data showing the in vivo en bloc release of large, MVB-like sEV clusters by human colorectal cancer cells¹¹. Here we investigated if this was a colorectal cancer cell-specific phenomenon. Unexpectedly, we found that it was a general, autophagy-related mechanism of sEV release that we designated as “torn bag mechanism”.

Results and Discussion

In this study, we analyzed in situ fixed, cultured cells with the released EVs preserved in their original microenvironment on a gelatin/fibronectin coated surface, and we detected large multivesicular EVs (MV-lEVs) in sections of different immersion fixed organs. We tested tumorous HT29, HepG2, and non-tumorous HEK293, HEK293T-PalmGFP, HL1 cell lines, as well as primary suspension type bone marrow derived mast cells (BMMCs). In addition, we studied ultrathin sections of mouse kidney and liver.

By the analysis of transmission electron micrographs of all tested cell types, we identified budding (Fig.1A-G) and secretion (Fig.1H-N) of MV-lEVs carrying intraluminal vesicles (ILVs). Importantly, in all cases we found evidence for the extracellular rupture of the limiting membrane of MV-lEVs and the release of ILVs (Fig.1O-U). For this novel type of sEV release, we suggest the designation “torn bag mechanism” which is distinct from the exocytosis of MVBs and amphisomes^{1, 3, 10} and from the release of plasma membrane derived small EV (sEVs) by ectocytosis¹².

Figure 1.

Most relevant to the in vivo conditions, we also observed the same phenomenon within the blood vessels of ultrathin sections of both murine kidney (Fig.1V) and liver (Fig.1W,X). In these cases, both the intact MV-lEVs (Fig.1V-X) and the “torn bag release” of sEVs (Fig.1V) were detected. Fig.1X shows that a circulating leukocyte releases MV-lEVs by ectocytosis.

Next, we decided to further investigate the subcellular origin of the ILVs within the secreted MV-lEVs. We analyzed the microenvironment of in situ fixed HEK293T-PalmGFP cells by confocal microscopy. The PalmGFP signal of HEK293T-PalmGFP cells principally associates with the plasma membrane¹³ (Fig.2_S1C-E). The green fluorescence helped us to identify the plasma membrane-derived limiting membrane of MV-lEVs. In agreement with our previous findings on HT29 colorectal cancer cells, within the MV-lEVs, we found CD63/ALIX (Fig.2A,G) CD81/ALIX (Fig.2B,H), CD63/TSG101 (Fig.2.C,I) and CD81/TSG101 (Fig.2D,J) double positive ILVs or ILV clusters. In line with a recent finding¹², CD81 was present in the plasma membrane-derived limiting membrane (Fig.2B,D,F), while CD63 was only found inside the MV-lEVs (Fig.2A,C,E).

Figure 2.

Fig.2_S1.

Next we studied the possible autophagy related aspects of the secreted MV-lEVs. ILVs were tested for the autophagy marker LC3B in parallel with CD63 and CD81. Although LC3B, CD63 and CD81 were all present in association with the ILVs (Fig.2E,F), the LC3B and CD63 (Fig.2K) and the LC3B and CD81 (Fig.2L) signals did not overlap. Fig.2M shows that while the known sEV markers (CD63, CD81, TSG101 and ALIX) strongly co-localized with each other, LC3B positivity hardly showed co-localization with CD63 or CD81. Immunocytochemistry results of HT29 and HepG2 cells further validated the findings obtained with the HEK293T-PalmGFP cells (Fig.2_S2). The ILVs of HEK293T-PalmGFP and HepG2 cell lines were Rab7 positive (Fig.2_S1A,B) suggesting a late endosomal origin. Western blot validation of the applied antibodies is summarized in Fig.2_S3. The sEV markers were also tested on separated sEVs of HEK293T-PalmGFP cells by transmission electron microscopy (TEM). Fig.2P confirms the typical sEV morphology. With TEM double immunogold labelling, using anti-LC3B and anti-CD63 antibodies simultaneously, we found distinct LC3B positive (Fig.2Q, Fig.2_S5O) and CD63 positive (Fig.2R, Fig.2_S5O) sEVs. Based on the analysis of TEM images, the diameters of unlabeled, and LC3B positive and negative sEVs were determined (Fig.2S). The LC3B positive sEVs had a significantly larger diameter as compared to the LC3B negative ones. The simultaneous presence of CD63, CD81, TSG101, ALIX and the autophagosome marker LC3B within the MV-lEVs (rather than on their limiting membrane) identified the extracellular MV-lEVs as en bloc released amphisomes.

Fig.2_S2

Fig.2_S3.

Fig.2_S4.

Fig.2_S5.

The “torn bag mechanism” was also monitored by live cell SIM² super-resolution microscopy analysis of HEK293T-PalmGFP-LC3RFP cells (Fig.2N,O). The release of the LC3 positive red fluorescent signal was detected within a relatively short period of time (260 sec) (Fig.2O). We could rule out the possibility that rupture of the limiting membrane detected by TEM (Fig.1O-T,V) was a fixation artefact by showing the release of LC3 positive sEVs from amphiectosomes with live-cell imaging. Characterization of the in-house developed HEK293T-PalmGFP-LC3RFP cell line is shown in Fig.2_S4.

In the following step, we addressed the question whether LC3, associated with the ILVs of MV-lEVs, indeed reflected autophagy origin. We tested MVBs (Fig.2_S5A,F,K), autophagosomes (Fig.2_S5B,G,L), amphisomes (Fig.2_S5C,H,M), amphiectosomes (Fig.2_S5D,I,N) and isolated sEV fractions of the same cells (Fig.2_S5E,J,O). Using immune electron microscopy, as expected, we found CD63 single positivity in MVBs (Fig.2_S5K). In autophagosomes, the limiting phagophore membrane was LC3B positive, and CD63 positivity was also present (Fig.2_S5L). The limiting membrane of amphisomes was LC3B negative, and the internal membranous structures were either LC3B or CD63 positive (Fig.2_S5M). The same immunostaining was also observed in the ILVs of the released amphiectosomes (Fig.2_S5N). Importantly, the separated sEVs of HEK293T-PalmGFP cells were either LC3B or CD63 positive (Fig.2_S5O). Thus, we confirmed our confocal microscopy results at the ultrastructural level. Using immunogold TEM, we provided further evidence for the budding/ectocytosis mechanism of amphiectosome release (Fig.2_S5P). The diameters of ILVs within MVBs, amphisomes and amphiectosomes were compared (Fig.2_S5Q), and the differences were likely due to the different pH and osmotic conditions within these organelles. In agreement with our observations with separated sEVs, LC3B positive ILVs had a significantly larger diameter than the LC3B negative ones (Fig.2_S5R) possibly indicating their different intracellular origin. Based on all the above findings, we propose the following model (Fig.3A): autophagosomes and MVBs fuse to form amphisomes, and the inner, LC3 positive membrane of autophagosomes disintegrates¹⁴. It curls up and forms LC3 positive ILVs. Therefore, amphisomes contain both MVB-derived and autophagosome-derived, LC3 negative and positive ILVs, respectively. The amphisome is next released from the cell by ectocytosis. Finally, the plasma membrane-derived limiting membrane ruptures enabling the ILVs escape to the extracellular space by a “torn bag mechanism”.

Figure 3.

To investigate the process of amphiectosome release, we exposed the MV-lEV releasing cells to different in vitro treatments (Fig.3B). The release of MV-lEVs was monitored by confocal microscopy of in situ fixed cell cultures. Optimal test conditions were determined (Fig.3_S1A-F) and the results are summarized in Fig.3C. Cytochalasin B did not have any effect on the discharge of MV-lEVs suggesting that the release did not involve a major actin-dependent mechanism. In contrast, there was a significant reduction of the MV-lEV secretion upon exposure of the cells to Colchicine indicating a role of microtubules in the release of the MV-lEVs. While Rapamycin significantly reduced the discharge of MV-lEVs, Chloroquine and Bafilomycin induced an enhanced MV-lEV secretion. Rapamycin activates autophagic degradation¹⁵, therefore, it induces a shift towards degradation as opposed to secretion. The lysosomotropic agents Chloroquine and Bafilomycin are known to interfere with the acidification of lysosomes^{16, 17}. By blocking the degradation pathway of MVBs/amphisomes (Fig.3B), an enhanced sEV secretion is observed. This effect is well known for exosome secretion from MVBs^{18, 19}. The diameters of the released MV-lEVs were determined based on confocal images (Fig.3_S1G). Metabolic activity of the cells was determined by a Resazurin assay, and a significant reduction was detected upon exposure of the cells to Rapamycin (Fig.3_S1D) in line with previously published data²⁰. LC3II is the membrane-associated, lipidated autophagic form of LC3²¹ and it is the hallmark of autophagy related membranes¹⁴. Importantly, by Western blot, we not only showed the presence of the membrane-bound LC3II in serum-free, lEV-depleted (sEV containing) conditioned medium of both HEK293T-PalmGFP and HEK293T-PalmGFP-LC3RFP cells, but the amount of LC3II substantially increased upon Chloroquine treatment (Fig.3D). Raw data of Western blots are available in Fig.3_S2.

Fig.3_S1.

Fig.3_S2.

Recent advances in the EV field shed light on migrasomes, a special type of MV-lEVs^{22, 23}. With their pomegranate-like ultrastructure, migrasomes resemble amphiectosomes. Therefore, we tested the presence of TSPAN4, a migrasome limiting membrane marker²³, in amphiectosomes. Fig.4A,B,G,H show that although TSPAN4 was present intraluminally in the HEK293T-PalmGFP-derived MV-lEVs, it was clearly absent from their limiting membrane. Surprisingly, we identified two different HT29 cell-derived MV-lEV populations: one with intraluminal TSPAN4 only (Fig.4C,E,I,K), and another one with a TSPAN4 positive limiting membrane (Fig.4D,F,J,L). This raised the possibility that the latter population corresponded to migrasomes. Our co-localization analysis also confirmed the existence of two distinct MV-lEV populations (Fig.4M). Next, we carried out live-cell imaging on HEK293T-PalmGFP-LC3RFP cells. The released MV-lEVs were either LC3 positive or negative intraluminally (Fig.4N,O). Our TEM images confirmed that certain cell types can release both migrasome-like structures and amphiectosomes. MV-lEVs with typical migrasome-associated retraction fiber(s) were detected in the case of HL1 (Fig.4P), HEK293T-PalmGFP (Fig.4Q) and BMMC cells (Fig.4R). Importantly, the same cells also released amphiectosomes by budding from the cell surface (Fig.4S-U). Taken together, based on the absence of TSPAN4 in their limiting membrane and their lack of association with retraction fibers, amphiectosomes appear to be distinct from migrasomes. Besides migrasomes, another MV-lEV type was described in the case of gastrointestinal tumors and low-grade glioblastoma cells referred to as spheresome^{24, 25}. However, there is no data on a relationship of spheresome release and autophagy. Recently, endothelial cell derived, multicompartmented microvesicles (MCMVs) were shown to protrude and pinch-off from the cell surface releasing ILVs by a mechanism similar to exocytosis²⁶. The absence of protrusion clusters described for MCMV²⁶ distinguishes amphiectosomes from MCMVs.

Figure 4.

Our approach, involving in situ fixation of cultures and tissues, made it possible to recognize sEV release from amphiectosomes by the “torn bag mechanism”. We propose that this mechanism can be easily missed if conditioned medium is subjected to centrifugation, SEC purification or even to simple pipetting, which may rupture of the limiting membrane of amphiectosomes. This aligns with our observation that the spontaneous escape of ILVs from untouched amphiectosomes can occur as early as 5 minutes after amphiectosome release. Our data presented here suggest that amphiectosome secretion and the “torn bag mechanism” may have a significant, however, previously unrecognized role in sEV biogenesis.

Material and methods

Cell lines

The HEK293 human embryonic kidney, the HepG2 human hepatocyte carcinoma cell line and the HT29 human colon adenocarcinoma cell lines were purchased from the European Collection of Authenticated Cell Cultures (ECACC) through their distributor (Sigma). The HL1 cell line was purchased from Millipore. The HEK293T-PalmGFP human embryonic kidney cells was kindly provided by Charles P. Lai²⁷. Mouse bone marrow-derived mast cells (BMMCs) were differentiated and expanded as we described previously²⁸. The HEK293, HEK293T-PalmGFP and HepG2 cell lines were grown in DMEM (Gibco)^{29, 30}, the HT29 cells were cultured in RPMI 1640 (Gibco)¹¹, while the HL1 cells were grown in Claycomb medium³¹. All cells were cultured with 10 % fetal bovine serum (FBS, BioSera) in the presence of 100 U/mL of Penicillin and 100 µg/mL Streptomycin (Sigma).

Before analysis by confocal microscopy, the cells were cultured on the surface of gelatin-fibronectin coated glass coverslips (VWR). The coating solution contained 0.02 % gelatin (Sigma) and 5 mg/mL fibronectin (Invitrogen). Coverslips were coated overnight (O/N) at 37 °C.

For transmission electron microscopy (TEM), the adherent cells (HEK293, HEK293T-PalmGFP, HepG2, HT29 and HL1) were grown in gelatin-fibronectin coated 8-well Flux Cell Culture Slides (SPL).

Cell cultures were tested regularly for Mycoplasma infection by PCR, with the following PCR primers: GAAGAWATGCCWTATTTAGAAGATGG and CCRTTTTGACTYTTWCCAC-CMAGTGGTTGTTG ³¹.

Generation of HEK293T-PalmGFP-LC3RFP cell line

For generation of a stable HEK293T-PalmGFP-LC3RFP cell line, HEK293T-PalmGFP cells were transfected by LentiBrite RFP-LC3 Lentiviral particles (Merck) according to the instructions of the manufacturer. The GFP-RFP double positive cells were sorted by a HS800 Cell Sorter (SONY) and cell banks (master and working cell banks) were prepared. The success of the stable transfection was analyzed by immunocytochemistry and Western blotting. Results are shown in Fig.2_S3.

Confocal microscopy

Confocal microscopy was carried out as we described earlier^{28, 31} with some modifications. Briefly, unlike the majority of the studies in the field of EVs, here we analyzed untouched, in situ fixed and cultured cells and their microenvironment. Since centrifugation may disrupt the limiting membrane of amphiectosomes, the in situ fixation made it possible to observe them in their intact form. The culture medium was gently removed by pipetting from above the cells leaving a thin medium layer only (approximately 150 µL of liquid on the cells). Without any further washing, cells were in situ fixed by 4 % paraformaldehyde (PFA) in phosphate buffer saline (PBS) for 20 min at room temperature (RT). The released lEVs were either fixed and captured during the release or were preserved on the gelatin/fibronectin surface coating. After fixation, 3x 5 min washes with 50 mM glycine in PBS were carried out. In the case of the non-fluorescent HepG2 and HT29 cells, a lactadherin-based plasma membrane staining was performed^{13, 28, 31}. Lactadherin (Haematologic Technologies) was conjugated to ATTO488 fluorophore (abcam) according to the instructions of the manufacturer. The lactadherin-ATTO488 conjugate was added to the cells in 1:100 dilution in PBS (for 1 h, RT). The unbound lactadherin was removed by washing with PBS (3 times, 5 min, RT) and post-fixation was carried out by 4 % PFA (20 min, RT). PFA was removed by washes with 50 mM glycine in PBS (3 times, 5 min, RT). Blocking and permeabilization of the cells were performed by 10 % FBS with 0.1 % TritonX-100 (Sigma) in PBS (1 h, RT). In general, primary antibodies were applied in 1:200 dilution O/N at 4 °C in the above blocking and permeabilization solution. Excess primary antibodies were eliminated by washing with the blocking and permeabilization solution (3 times, 5 min RT). The secondary antibodies were applied in 1:1000 dilution in 1 % FBS in PBS (1 h, RT). Unbound secondary antibodies were eliminated by washing (1 % FBS, in PBS, 2 times, 5 min; PBS 2 times, 5 min; water 2 times, 5 min) and the samples were mounted in ProLong Diamond with DAPI (Invitrogen). Slides were examined by Leica SP8 Lightning confocal microscope with adaptive lightning mode using a HC PL APO CS2 63x/1.40 OIL objective with hybrid detector. The applied lookup tables (LUT) were linear during this study. For image analysis and co-localization studies, we applied Leica LASX software using the unprocessed raw images. In case of co-localization studies, 20 % threshold and 10 % background settings were applied.

Transmission electron microscopy

Adherent cells (HEK293, HEK293T-PalmGFP, HepG2, HT29 and HL1), as well as mouse (C57BL/6) kidney and liver tissues pieces (approx. 1.5 mm x 1.5 mm) were immersed in and fixed by 4 % glutaraldehyde (48 h/4 °C), post-fixed by 1% osmium tetroxide (2 h, RT) and were embedded into Epon resin (Electron Microscopy Sciences) as described previously³². In the case of BMMCs, 920 μL cell suspension was complemented with 80 μL 50% glutaraldehyde to reach the final 4 % glutaraldehyde concentration. Cells were fixed for 48h at 4°C and were post-fixed by 1% osmium tetroxide (2 h, RT). During sample preparation, BMMCs were collected by gravity-based sedimentation. Due to the high viscosity of Epon resin, BMMCs were embedded in LR White low viscosity resin (SPI Supplies) according to the instructions of the manufacturer. Ultrathin sections (60 nm) were contrasted by uranyl acetate (3.75 %, 10 min, RT) and lead citrate (12 min, RT).

For immunogold TEM studies, cells and tissues were fixed by 4 % PFA with 0.1% glutaraldehyde (48 h, 4 °C) and were post-fixed by 0.5 % osmium tetroxide (30 min, RT). Samples were embedded into LR White and were immunogold labelled as described previously¹¹. The contrast was enhanced by uranyl acetate (3.75 %, 1 min, RT) and lead citrate (2 min, RT). Purified HEK293T-PalmGFP-derived sEV-s were detected by negative-positive contrasting and immunogold labelling as described previously³¹. Antibodies were used in 1:50 dilution. Detailed list of the used antibodies is available in the Key Resources Table.

For all electron microscopic studies, a JEOL 1011 transmission electron microscope was used. Images were captured with the help of Olympus iTEM software and for image analysis, Image J software was used.

Live cell imaging

The HEK293T-PalmGFP-LC3RFP stable cell line was cultured the same way as HEK293T-PalmGFP cells. Before the experiments, gelatin-fibronectin coated 10-well coverslip bottom chamber slide (Greiner-BioOne) was seeded and treated by 30 μM Chloroquine O/N. Release of migrasomes, amphiectosomes and sEVs were followed by the Leica SP8 Lightning confocal microscope equipped with an Okolab environmental chamber and a Zeiss ELYRA 7 with Lattice SIM² super-resolution fluorescent microscope with the help of 63x/1.4 plan-apochromat Oil objective. For image analysis, we applied Leica LASX, Zeiss ZEN Blue and Image J software.

Modulation of amphiectosome release

In order to test the release mechanism of amphiectosomes and to distinguish them from migrasomes, different treatments were applied overnight (O/N) in fresh, serum containing cell culture medium except for Colchicine, where 1 h treatment was selected. Maturation and fusion of endosomes and lysosomes were inhibited by 15.5 µg/mL (30 μM) Chloroquine (Invitrogen) and 10 nM BafilomycinA1 (Sigma). Actin polymerization was inhibited by 125 ng/mL Cytochalasin B (Sigma). Tubulin polymerization and function were inhibited by 250 pg/mL Colchicine, while an autophagy-related degradation was induced by 50 ng/mL Rapamycin. The selected concentrations were determined based both on literature data and our preliminary experiments (Fig.3_S1). Cellular metabolic activity was determined by a metabolic activity-based Resazurin assay³¹. Fresh cell culture medium was added to control cultures a day before the in situ fixation. Reagents were diluted in fresh cell culture medium. Leica TCS SP8 Lightning confocal microscope was used for detection of amphiectosome release. A few hundred µm² sized area with 15-20 µm in height was tile-scanned with a few hundred cells. The MV-lEVs were recognized as CD63 positive EVs surrounded by GFP positive membrane. They were counted and were normalized to the number of nuclei.

Western blotting

Presence of proteins and specificity of the used primary antibodies were confirmed by Western blotting as described previously³¹. For accurate quantification (free of variations potentially caused by EV purification), we analyzed cell-, serum- and lEV-free conditioned medium. The cells were cultured O/N in a serum-free culture medium. After harvesting, the cells were eliminated by centrifugation (300 g, 10 min at 4 °C) followed by a 2.000g centrifugation (30 min at 4 °C) to eliminate lEVs. Total protein content of the conditioned, serum-, cell- and lEV-free medium was precipitated by trichloroacetic acid as described previously^{31, 33}. The protein pellets were suspended in in cOmplete Protease Inhibitor Cocktail (Roche) containing radio-immunoprecipitation assay (RIPA) buffer.

When whole cell lysate was tested for validation of antibodies and the HEK293T-PalmGFP-LC3RFP cell line, cells were lysed in cOmplete Protease Inhibitor Cocktail (Roche) containing RIPA buffer.

Polyacrylamide gel electrophoresis was carried out using 10 % gels (acrylamide/bis-acrylamide ratio 37.5:1) or any kDa precast gels (Biorad) and a MiniProtean (BioRad) gel running system. For better solubilization of membrane proteins, equal volumes of 0.1 % TritonX-100, Laemmli buffer and samples were mixed as described previously³⁴. Approximately 10-30 µg protein were loaded into each well. Following electrophoretic separation, proteins were transferred to PVDF membranes (Serva). Membranes were blocked with 5 % skimmed milk powder or 5 % BSA in washing buffer for 1 h. Primary antibodies were applied in 1:1000 dilution with the exception of the anti-CD63 (rabbit), anti-CD81 (rabbit) and anti-CD81 (mouse) antibodies where 1:500, 1:2500 and 1:100 dilutions were used, respectively. Peroxidase-labelled secondary antibodies were applied in 1:10 000 dilution. The signals were detected by ECL Western Blotting Substrate (Thermo Scientific) with an Imager CHEMI Premium (VWR) image analyzer system. In case of quantification, equal protein amounts were loaded to the gels. Within a biological replicate, the control and Chloroquine-treated samples were run on the same gels. To enable comparison, the relative expression of control and Chloroquine-treated samples were determined and compared.

Software and statistical analysis

For image capturing, analysis and co-localization studies, Leica LAS X, Zeiss ZEN Blue, Olympus iTEM and Image J software were used. Figures and graphs were generated using GraphPad Prism 9.4.1 and Biorender (BioRender.com). For statistical analysis, standard deviation was calculated. Unpaired two tailed student t-tests and one-way ANOVA were used (* p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001).

Acknowledgements

This research was funded by the NVKP_16-1-2016-0004 grant of the Hungarian National Research, Development and Innovation Office (NKFIH), ÚNKP-23-3-I-SE-2, the Semmelweis Innovation Fund (STIA 2020 KFI), Hungarian Scientific Research Fund (OTKA K120237 and FK 138851), Eötvös Loránd University Excellence Fund (EKA 2022/045-P101), Hungary Academy of Sciences, LP2022-13/2022, VEKOP-2.3.2-16-2016-00002, VEKOP-2.3.3-15-2017-00016, the Higher Education Excellence Program (FIKP) and the Therapeutic Thematic Programme TKP2021-EGA-23. This study was also supported by the grants RRF-2.3.1-21-2022-00003 (National Cardiovascular Laboratory Program) and 2019-2.1.7-ERA-NET-2021-00015. The project has received funding from the EU’s Horizon 2020 Research and Innovation Programme under grant agreement No. 739593.

The authors would like to thank Gábor Valcz for for his invaluable support throughout the research process, for critically reviewing the manuscript and for his insightful comments. The authors would like to thank the ZEISS Microscopy Customer Center Team for the collaboration, and Dr. Abel Pereira da Graça for the possibility to use the Elyra7 SIM² super-resolution microscope. The authors are grateful to Györgyné Vidra, Györgyi Balogh and Andrea Orbán for their technical help and advises.

Conflict of interest

EIB is a member of the Advisory Board of Sphere Gene Therapeutics Inc. (Boston, MA, USA) and ReNeuron (UK).