Enveloped viruses fuse with host membranes without affecting cell integrity. Non-enveloped viruses and bacteria penetrate by rupturing endosomal membranes and thus expose complex-type carbohydrates from the endosome lumen to cytosolic proteins. Here we report on the dynamics and initial marker analyses of Galectin-3 (Gal3)-positive membranes triggered by incoming adenovirus species B/C in HeLa cells. Using mCherry-Gal3 reporter constructs, immunolabeling, confocal and electron microscopy, we detected robust signals from Gal3-containing, early endosomal antigen 1-positive membranes 1 h post-infection (pi). Adenoviruses penetrate from non-acidic endosomes with high efficiency, 15 min pi, and largely outnumbered the Gal3-positive membranes, suggesting that Gal3 recruitment to broken membranes is transient, or Gal3-positive membranes are rapidly turned-over. In support of rapid turn-over, Gal3 was found within single-membrane vesicles and degradative autophagosomes. The Gal3 membranes contained ubiquitin and the poly-ubiquitin binding protein p62/sequestosome-1, but only low amounts of virus, or membrane-lytic protein VI exposed from virions. Remarkably, the Gal3-positive membranes were cleared 3 h pi, slower than protein VI, which was cleared 30 min pi. The data show that broken early endosomes, but not virus particles, are rapidly removed by a process involving autophagy, which we term ‘endosomophagy’. We speculate that endosomophagy is pro-viral and attenuates innate immunity.

All viruses and many bacteria that enter eukaryotic cells activate a genetic program for replication and immune evasion^[2][3][4]. Invariably, host defense against the pathogens or host cell damage antagonizes the intruders^[5][6][7]. Damage to host cells occurs during invasion of bacteria and non-enveloped viruses, including Salmonella, Shigella, Listeria or Adenovirus, which disrupt phagocytic vacuoles and endosomes^[8][9]. Disrupted vacuoles recruit beta-galactoside binding lectins, such as galectin-3 (Gal3) and Gal8, both widely expressed in epithelial cells^{[10][11][12][13]}. Cytosolic galectins can also be subject to export by leader peptide-independent mechanisms^[14][15][16]. They can bind to enveloped viruses and bacteria, and act as pattern recognition receptors^[17]. Intracellular Gal8 coordinates the destruction of disrupted late endosomes and associated bacteria by autophagy, whereas Gal3 is involved in membrane sorting by clustering glycoproteins and glycolipids, and modulation of cell-cell contacts during organogenesis^[18][19][20].

Recently, Gal3 was reported to be recruited to endosomes disrupted by incoming human adenovirus type 5 (HAdV-C5)^[13][21]. Adenoviruses are non-enveloped human pathogens and widely used vectors in clinical gene therapy and vaccination^{[22][23][24][25]}. HAdV-C2 and C5 enter epithelial cells by receptor-mediated endocytosis and a stepwise uncoating program initiated at the plasma membrane by the differential movements of two receptors, integrin and coxsackievirus adenovirus receptor (CAR)^[26][27][28]. Virus penetration into the cytosol occurs from early endosomes in a pH-independent manner^[29][30]. It requires the membrane-lytic viral protein VI^[31][32][33], lysosomal secretion, the sphingolipid ceramide^[34][35], and gates the pathway for viral DNA genome separation from the capsid and nuclear delivery of the viral genome^[36][37]. Here, we describe a novel observation, the recruitment of Gal3, ubiquitin and the poly-ubiquitin binding protein p62/SQSTM1 to ruptured early endosomes, followed by clearance of disrupted membranes without clearance of virus particles from the infected cell. p62/SQSTM1 is an adaptor linking poly-ubiquitination to macroautophagy^[38][39]. We speculate that this mode of clearance is pro-viral, in contrast to the recruitment of Gal8 to ruptured phagosomal vacuoles in Salmonella infected cells, which presents a mode of pathogen restriction^[12].

Ruptured host membranes are danger signals, and enhance innate immunity against infections. The objective here was to analyze the composition and dynamics of ruptured endosomes induced upon virus entry into cells.

Most viruses engage in receptor-mediated endocytosis and extensive membrane trafficking using cellular mechanisms^[40][41]. Enveloped viruses encode membrane proteins, which fuse the viral and endosomal membranes, or, in rare cases, the plasma membrane. Non-enveloped viruses are membrane-free, except for certain picornaviruses which occur in both enveloped and non-enveloped forms^[42]. They all encode membrane-interacting proteins, which are typically encased in a capsid, and can be activated or exposed by cues from the host cell during entry^[35][43]. Activation of membrane-active proteins leads to pore formation, piercing or rupture of the host membrane^[44][45][46]. Rupture of internal membranes by HAdV-C5 exposes sugar epitopes to cytosolic proteins, and recruits cytosolic Gal3 to the broken membranes^[13].

To track broken membranes, we expressed mCherry-Gal3 fusion protein in HeLa cells, followed by continuous infection with HAdV-C5 or HAdV-B3 for 1h. HAdV-C5 enters epithelial cells through CAR and integrin receptors, and HAdV-B3 through CD46 and desmoglein-2 receptors^{[3][47][48][49]}. Both viruses penetrate HeLa cells independent of endosomal acidification with an efficiency higher than 70% (HAdV-C5) or about 40% (HAdV-B3) within 30 min of cold-synchronized infection^[30]. Uninfected cells had diffuse cytoplasmic red fluorescence and variable levels of nuclear red fluorescence, in contrast to the prominent cytoplasmic mCherry-Gal3 foci in both types of infected cells, indicating that membrane rupture was independent of the nature of entry receptors (Fig. 1A, and for additional uninfected cells, see panels 1C, F, H). Most of the mCherry-Gal3 foci were devoid of HAdV-C5 particles labeled with Alexa488 fluorophore, as indicated by quantitative image analyses of single confocal slices (Fig. 1B, shown are maximal projections of confocal stacked images). Note that only two of the overlapping puncta co-localized in any of the single confocal slices (indicated by gray arrow heads; single slices not shown). Poor co-localization of HAdV-C2 and mCherry-Gal3 foci was confirmed by a low overlap of mCherry-Gal3 with the incoming membrane-lytic viral protein VI, detected by an affinity-purified polyclonal antibody (Fig. 1C, two arrow heads point to the only co-localization events in these cells by single slice analyses). In contrast, a large fraction of the mCherry-Gal3 foci was robustly positive for the early endosomal antigen 1 (EEA1), indicating that the broken membranes had features of early endosomes (Fig. 1D). These results were in agreement with earlier notions that HAdV escapes quickly from endosomes and is transported by dynein/dynactin mediated processes toward the nucleus^[50][51][52]. Protein VI on the other hand likely remains membrane associated, and may be sorted separately for degradation^[26][53].

Remarkably, visual inspection of the images suggested that the number of HAdV-C5 particles, or protein VI puncta, exceeded the number of mCherry-Gal3 foci (Fig. 1B,C). The same observation was made with HAdV-B3-infected cells (not shown). Although the resolution of virus structures and Gal3-positive membranes was diffraction-limited in our confocal micrographs, the number of mCherry-Gal3 foci correlated with virus dose (data not shown). We also noted that Gal3 foci robustly emerged with 1 h continuous infection schemes, but were rarely visible with shorter infection pulses or at 20 min post-warming after cold-synchronized infections (data not shown). Since virus penetration from endosomes is asynchronous, and may involve different types of endosomes (clathrin-derived endosomes and macropinosomes), but not late endosomes and lysosomes^[29], we tested if Gal3-positive membranes were turned-over. Ultrastructural analyses of Gal3 by immuno-electron microscopy in cryo-sections of normal HeLa cells (not transfected with mCherry-Gal3 constructs) showed that most Gal3 immuno-gold signal was in the cytosol, as expected, although infected cells tended to have less cytosolic Gal3 than uninfected cells (Fig. 1E). Infected cells, however, had more Gal3 within vesicles and autophagosomes than uninfected cells. We tentatively classified the Gal3-positive autophagosomes as degradative autophagosomes, based on their partly visible double bilayer membranes and on the contents comprising electron-dense organelle-like material^[1] (Fig. 1E, structures labeled Aph). The structures described here are different from amphisomes, which are formed when an endosome fuses with an autophagosome^[1]. Our Gal3-positive autophagosomes appear to contain internal membrane remnants (Fig. 1E) and are cleared with faster kinetics than calcium-phosphate-damaged endosomes, and silica- or LLOMe-damaged lysosomes, which are cleared by amphisomes^[54][55].

Further immuno-fluorescence microscopy analyses indicated that the HAdV-C5 induced mCherry-Gal3 foci were strongly positive for the pan-ubiquitin antibody FK2 (Fig. 1F), and a fraction of these foci was positive for the poly-ubiquitin-binding protein p62/SQSTM1 (Fig. 1G). Interestingly, the autophagic marker LC3B was not enriched on mCherry-Gal3 or EEA1 positive membranes of infected cells, but remained diffuse in the cytoplasm (data not shown). This suggested that ubiquitin coordinates LC3B-independent steps for recruiting the autophagic machinery to clear the broken early endosomes. In agreement with this notion, we found that the mCherry-Gal3 foci were readily cleared from the infected cells within 3 h of infection (Fig. 1H). The clearance kinetics of the broken early endosomes was considerably slower than the disappearance of protein VI, which was nearly undetectable in western blots 30 min pi (Fig. 1I).

In sum, the data are reminiscent of a previous observation that ruptured phagocytic vacuoles in Shigella-infected cells are poly-ubiquitinated, p62/SQSTM1 positive and targeted for autophagic destruction, with concomitant reduction of early inflammatory and cytokine responses^[10]. Our data support the earlier finding that the levels of Gal3 and p62/SQSTM1 are reduced in HAdV-B3- or C5-infected A549 human epithelial cells compared to uninfected cells^[56]. We suggest the term ‘endosomophagy’ for the clearance of broken early endosomes, and speculate that endosomophagy suppresses danger signals and cell death pathways, and enhances virus infection. This is distinct from lysophagy, which can be induced by chemicals, such as the lysosomotropic reagent L-Leucyl-L-leucine methyl ester (LLOMe), which oligomerizes, forms toxic products in the lysosomal lumen, leads to lysosomal lysis, and thus triggers necrosis^[57][58][59].

The observation here shows that adenovirus-ruptured early endosomal membranes are targeted for clearance. These membranes are Gal3-positive, contain ubiquitin and the polyubiquitin-binding protein p62/SQSTM1, and are associated with degradative autophagosomes. This endosomal clearance is termed here ‘endosomophagy’. It is distinct from lysosomal rupture, which leads to inflammation and necrotic cell death.

Our observation has implications for innate immunity, inflammation, and infectious disease. Notably, in spite of the robust cross-reacting cellular and humoral immune responses against HAdV infections, latent or persistent adenovirus infections in humans occur, and may last for years^[60]. The report here also impinges on gene therapy and vaccination protocols, considering the prominent roles of adenoviruses in clinical gene therapy, and as vaccination adjuvants^[61]. Follow-up analyses of this observation can be directed to the identification of host and viral targets for ubiquitin ligases and deubiquitinases, host adaptor proteins, or the role of virus-induced sphingolipids in tuning endosomophagy^[34]. Notably, sphingolipids have key roles in the regulation of autophagy at the levels of transcription, translation, and morphogenesis of autophagosomes^[62]. Another emerging question is how Gal3 and ubiquitin turnover are orchestrated on the ruptured membranes. Further studies can also be directed toward professional antigen-presenting cells, such as macrophages and dendritic cells, which play a major role in inducing and controlling local and global immune responses^[63]. Finally, specific experiments can address the question if the clearance of the membrane-lytic viral protein VI is related to or distinct from the mechanism of endosomophagy.

Cells and viruses

HeLa cervical carcinoma cells, subline Ohio (from L. Kaiser, University Hospital, Geneva, Switzerland) were grown at 37°C under 5% CO2 in Dulbecco's modified Eagle's medium (DMEM; Sigma) supplemented with 7.5% fetal calf serum (FCS; Life Technologies) and 1% nonessential amino acids (Sigma). HAdV-C2, HAdV-C5 and HAdV-B3 were grown in human bronchial epithelial A549 cells (American Type Culture Collection), isolated, and labeled with Alexa Fluor 488 (Alexa488; Life Technologies) as previously described^[30].

Immunofluorescence analysis of Gal3 foci

The pmCherry-Gal3 construct was generated by PCR amplification of Gal3 coding sequence from U2OS cDNA flanked by HindIII and EcoRI sites and cloned into pmCherry-C1 (Clontech). HeLa-Ohio cells (0.5×106) were transfected with 8 µg of pmCherry-Gal3 using Neon technology (100 µl tip, 1005 V, 35 ms, 2 pulses; Life Technologies). Cells expressing Gal3 with an N-terminal mCherry tag (mCherry-Gal3) were grown in 24 well glass coverslips for 24 h and then were exposed to continuous infection with 1–5 µg/ml virus for the indicated time points. Cells were subsequently fixed in 4% PFA and stained with an anti-EEA1 antibody (mouse, clone 14; Transduction Laboratories), anti-p62/SQSTM1 antibody (mouse, clone 5F2; MBL) or the anti-ubiquitin antibody FK2 (Life Sensors) that detects K29-, K48-, and K63-linked mono- and poly-ubiquitinated proteins.

Imaging was performed with a Leica SP5 confocal microscope equipped with a 40x objective (oil immersion, numerical aperture 1.25) and a 63x objective (oil immersion, numerical aperture 1.25). Z-stacks composed of 8×0.5 μm steps were acquired at a frequency of 8000 Hz applying bidirectional scan, line averaging 32x and minimized acquisition time. Maximum projections of Z-stacks were analyzed using a customized Matlab routine (Matlab 2009b, available upon request). Fluorescence intensity of either virus labeling or antibody staining on the position of mCherry-Gal3 foci was determined and mean values per cell are shown. To evaluate background values in uninfected cells without mCherry-Gal3 foci, randomly generated and cytosolic foci of pixel size equivalent to mCherry-Gal3 foci localized were quantified. Further details are described in^[30][26]. Statistical analyses were performed using GraphPad Prism software (Version 5, GraphPad Software, Inc. La Jolla). Single cell-based assays are represented as scatter dot plots, where the horizontal bars indicate the mean value and the vertical bars the standard deviation. Two-tailed p-values were calculated by unpaired t-tests with Welch's correction and confidence interval 95%.

Co-localization of mCherry-Gal3 and protein VI

pmCherry-Gal3-transfected HeLa-Ohio cells grown on coverslips in 24 well dish were infected with 0.8 µg atto647-labeled HAdV-C2 (kindly provided by I-Hsuan Wang) for 1 h at 37°C. Control cells did not receive any virus. Cells were stained with affinity-purified rabbit anti-protein VI antibodies^[26] and secondary AlexaFluor 488-conjugated anti-rabbit antibodies. Samples were imaged with Leica SP5 confocal laser scanning microscope using a 63x objective (oil immersion, numerical aperture 1.4) and zoom factor 4. Stacks were recorded at 0.5 µm intervals using 4x averaging, between frames sequential method and a frequency of 1000 Hz. Shown are maximal projections of Z-stacks, but co-localization of protein VI and mCherry-Gal3 dots were checked also on individual confocal sections.

Analysis of incoming protein VI

HeLa cells (1.5×105) were seeded in a 12 well plate the day before the infection. HAdV-C5 (16.8 µg) was added to the cells and allowed to bind and internalize at 37°C for 30 min in RPMI 1640 medium supplemented with 0.2% bovine serum albumin and 20 mM HEPES-NaOH, pH 7.4. Free virus was removed by washing the cells, which were further incubated at 37°C for the indicated times in DMEM medium supplemented with 10% fetal calf serum, and thereafter lysed in lysis buffer (0.2 ml of 200 mM Tris, pH 8.8, 20% glycerol, 5 mM EDTA, 50 mM DTT, 5% SDS, 0.02% bromophenol blue). Lysates were boiled at 95°C for 5 min and centrifuged at 16,000x g for 5 min. Proteins were resolved by SDS-PAGE and hexon and protein VI were detected by western blotting using anti-hexon (Abcam, ab6982) and anti-protein VI antibodies^[26], and anti-β-tubulin (Amersham) as a loading control.

Electron microscopy

Cryo-ultramicrotomy and immunocytochemistry was performed based on the protocol of Tokuyasu^[64][65][66]. The procedure was similar to one described earlier^[26]. About 3×106 subconfluent cells were fixed in PBS containing 2% pFA and 0.2% glutaraldehyde for 1 h, scraped off the dish, pelleted at about 500x g for 10 min, washed several times in PBS, embedded in a small volume of 10% gelatine (Sigma, G6650) in PBS at 37°C, pelleted and solidified in a thin-walled Eppendorf tube on ice o/n. The cells embedded in gelatine were removed by slicing the tube wall and then were infiltrated with 2.3 M sucrose in PBS at 4°C for 2 days. The gelatine cell block was mounted onto a metal plate, snap frozen in liquid nitrogen, and placed into a Leica EM Ultra Cut UC6 / FC6 machine. After trimming to about 0.25 mm2 surface area, ultrathin 80 nm thick cryo-sections were obtained with a diamond knife at -120°C. Frozen sections were collected onto a drop of cold sucrose on a wire loop, brought to room temperature, and transferred from the loop onto a Formvar-coated Ni-EM grid (125 µm mesh size). Grids were washed several times in PBS at room temperature, and at 40°C for 10 min to remove the gelatine. Aldehydes were blocked by 3 short incubations in PBS containing 0.15% glycine, pH 7.5, and 2 washes in PBS. Samples were blocked in PBG buffer consisting of PBS, 0.2% gelatine, 0.5% BSA (AppliChem, A6588, 0050) and 0.01% Tween20 (Thermo Scientific, 28320) for 10 min, and then with the primary rabbit immunoglobulin (IgG) anti-Gal3 antibody (PeproTech) at 1:20 in PBG for 1 h, room temperature, washed 4 times for 3 min in PBG, blocked again for 3 min, incubated with the secondary goat anti-rabbit IgG conjugated to 10 nm gold (BBI Solutions) at 1:50 in PBG, and washed several times in PBG, PBS, and H2O. Samples were fixed in 0.5% glutaraldehyde in H2O for 20 min, washed 5 times in H2O, and stained in 1.8% methylcellulose, 0.3% uranyl acetate on ice for 5 min. Excess liquid was blotted off, the sample dried on ice for several minutes, and analyzed in a Philips CM100 (100 kV at 46000 magnification using a digital CCD camera Gatan Orius 1000, 4kx2.6k pixels) or a ZEISS TEM10 (80 kV, 50000 magnification, using digital CCD camera Gatan Erlangshen ES500W, Model 782).

The work was supported by a research grant from the Swiss National Science Foundation (310030B_160316), and a Medical Research and Development project from the Swiss initiative for systems biology SystemsX.ch (2014/264 Project VirX, evaluated by the Swiss National Science Foundation) to UFG, and DFG grant Me1626/5-1 to HM.

The authors would like to thank Gery Barmettler, Center for Microscopy and Image Analysis of the University of Zurich, for advice and support in immunocytochemistry experiments.

The authors declare no conflicts of interest.

Not applicable.

No fraudulence is committed in performing these experiments or during processing of the data. We understand that in the case of fraudulence, the study can be retracted by ScienceMatters.