Natural history of Helicobacter pylori VacA toxin in human gastric epithelium in vivo: vacuoles and beyond

Uptake, intracellular trafficking and pathologic effects of VacA toxin from Helicobacter pylori have been widely investigated in vitro. However, no systematic analysis investigated VacA intracellular distribution and fate in H. pylori-infected human gastric epithelium in vivo, using ultrastructural immunocytochemistry that combines precise toxin localization with analysis of the overall cell ultrastructure and intercompartimental/interorganellar relationships. By immunogold procedure, in this study we investigated gastric biopsies taken from dyspeptic patients to characterize the overall toxin’s journey inside human gastric epithelial cells in vivo. Endocytic pits were found to take up VacA at sites of bacterial adhesion, leading to a population of peripheral endosomes, which in deeper (juxtanuclear) cytoplasm enlarged and fused each other to form large VacA-containing vacuoles (VCVs). These directly opened into endoplasmic reticulum (ER) cisternae, which in turn enveloped mitochondria and contacted the Golgi apparatus. In all such organelles we found toxin molecules, often coupled with structural damage. These findings suggest direct toxin transfer from VCVs to other target organelles such as ER/Golgi and mitochondria. VacA-induced cytotoxic changes were associated with the appearance of auto(phago)lysosomes containing VacA, polyubiquitinated proteins, p62/SQSTM1 protein, cathepsin D, damaged mitochondria and bacterial remnants, thus leading to persistent cell accumulation of degradative products.

internalization of glycosylphosphatidylinositol (GPI)-anchored proteins. VacA enters a tubulo-vesicular compartment, named GPI-anchored-protein-enriched early endosomal compartment (GEEC) or clathrin-independent carriers (CLICs), located close to the cell surface 2,9 . Delivered by a specific subset of early endosomes (motile because the formation of actin comet tails at their surface), most VacA then reaches late endosomes 5,10 . Here the pore-forming activity of VacA favors the accumulation of osmotically active ions (e.g., NH 4 + ) followed by swelling. This leads to the namesake effect of VacA, the development of a massive cytoplasmic vacuolation 2,5 . In addition, actin-driven motility of VacA-containing endosomes favors toxin delivery to mitochondria, where VacA translocates and causes apoptosis 2,11 . However, the exact mechanisms through which VacA reaches its mitochondrial target is still largely unknown.
Kern et al. 12 recently found that VacA reaches also the endoplasmic reticulum (ER) and the Golgi complex, thus identifying these organelles as novel target structures of the toxin. It remains however to be established how the toxin reaches ER and Golgi. Proteomic analysis of VacA-containing vacuoles (VCVs) purified from a T-cell line detected 122 VCV-specific proteins represented, in addition to typical endosomal/lysosomal proteins, by defined proteins from other organelles such as mitochondria, ER and Golgi 12 . The possibility thus arises that VCVs may exert a specific functional role in the intoxication process of the toxin, acting as a platform to trigger specific trafficking and signaling pathways exploited or influenced by VacA 8 .
VacA impairs immune responses, in particular by modulating the activity of immune effector cells like T lymphocytes, thus favoring the persistence of bacterial infection 13 . In addition, VacA has been found to inhibit antigen processing at endosomal level inside antigen-presenting cells (APCs) 14 . Cathepsin E has been identified as a crucial protease for antigen processing in APCs [15][16][17] and has also been detected in gastric epithelium 18 where its expression was markedly increased and expanded during H. pylori infection 19 . Thus, a possible VacA-sensitive role for cathepsin E in antigen processing inside gastric epithelial endosomes should be considered. Furthermore, VacA has been reported to specifically affect the host autophagic and lysosomal machinery, being also associated with autophagosomes, thus suggesting an additional role in the genesis of gastric epithelium pathologic changes 7,20,21 .
It must be underlined that most of the data reported derives from in vitro cellular models and that studies indicative of VacA interaction with, trafficking in, and action on the human stomach in vivo are still largely lacking. Thus, at least some of the results obtained so far should be interpreted with caution because, while they may make sense in terms of cell biology, it remains to be established whether, how and to what extent they have actual relevance in H. pylori-infected patients, especially considering the high cell, organ and species specificity of this bacterial infection.
The present study was specifically aimed to investigate the natural history of VacA in H. pylori-colonized human gastric epithelium in vivo by means of transmission electron microscopy (TEM) and ultrastructural immunocytochemistry, the optimal technology to investigate membrane-limited compartments like those involved in VacA intracellular trafficking and action 22,23 . In particular, we analyzed: a) H. pylori interaction with the gastric epithelium in delivering VacA toxin to host cells, b) VacA uptake at plasma membrane level, c) VacA intracellular trafficking and vacuole development, d) interorganellar interactions and presence of VacA in structurally normal or pathologically altered organelles other than endosomes/vacuoles, such as mitochondria, ER, Golgi, and cellular degradative structures like those of the autophagic-lysosomal pathway or the ubiquitin-proteasome system (UPS).

Results
VacA uptake and accumulation in endocytic-endosomal vesicles of H. pylori-infected human gastric epithelium. In infecting H. pylori, VacA immunoreactivity was mostly concentrated on bacterial outer membrane and its inner and outer leaflets (Fig. 1A,C-F). In addition, in agreement with previous findings on bacterial liquid culture 22 , VacA was found in some outer membrane vesicles (OMVs), ranging from 50 to 300 nm in size, detaching from bacteria and sometimes adhering to the epithelial membrane ( Fig. 1A,G,H). Inside H. pylori-infected gastric epithelium, some clear vesicles filling subluminal cytoplasm and infiltrating in between mucin-filled secretory granules also showed VacA reactivity (Fig. 1A,D-F). VacA-reactive bacteria found in the same preparation served as positive controls for the reactivity of such clear vesicles (Fig. 1A,C,E). Neither such clear vesicles nor VacA reactivity were observed in the gastric epithelium from H. pylori-negative patients taken as a control (Fig. 1B). Bacterial adhesion to the epithelium was characterized either by direct contact, sometimes with apparent fusion, of the respective membranes (Fig. 1A,F), or by connecting thin fimbria-type filaments immunoreactive for LPS O-antigen and VacA (Fig. 1C,D). Both H. pylori products were seen to accumulate on the confronting luminal epithelial membrane (Fig. 1C,D) and to be taken up by subluminal clear vesicles (Fig. 1D,F).
VacA-containing tubular invaginations of the luminal epithelial membrane of H. pylori-infected epithelium were found to penetrate cell cytoplasm and to interact with VacA-positive subluminal clear vesicles (Fig. 1F,G). These tubular invaginations were apparently lacking clathrin coat. They are likely to represent the human in vivo equivalents of the non-coated endocytic pits taking up VacA and generating VacA-positive superficial (early) endosomes (i.e., the aforementioned GEEC/CLICs) in experiments on cell lines in vitro 2,9 . In addition, clear endosomal vesicles were frequently found to directly contact each other and with larger vacuoles, showing focal loss of their limiting membrane at contact site, a pattern highly suggestive for vesicle fusion (Figs 1A and 2A,B). This may provide a pathway for VacA trafficking to juxtanuclear (late) endosomes and their VCVs derivatives 2,10 .
VCV role as VacA-distributing platform. VacA-accumulating VCVs were frequently found to directly contact ER, whose cisternae were also observed to open into VCVs, thus allowing free communication between respective lumina and possible transfer of VacA molecules from VCVs to ER (Figs 1J and 2A,B). As in turn ER cisternae frequently enveloped mitochondria (Figs 1J and 2A), this may well generate a pathway for VacA  transfer also to mitochondria, in addition to the direct VCV-mitochondria communications we occasionally found (Fig. 1J). Of interest was also the finding of VacA reactivity in some Golgi cisternae and adjacent ER cisternae (Fig. 2C). In parallel tests, both the Golgi complex ( Fig. 2D) and ER (Fig. 2E) of control gastric epithelium from uninfected human biopsies failed to show any VacA reactivity.

Cathepsin E in endosomal vesicles and vacuoles.
In normal, non-infected gastric epithelium, cathepsin E was found to be closely restricted to the rough ER (RER) (Fig. 2F). However, in H. pylori-infected gastric foveolar epithelium, cathepsin E was also detected in endosomes, both peripheral and juxtanuclear, and related VCVs (Fig. 2G,H).
VacA, intracellular pathologic changes and cellular degradative structures. In addition to cytoplasmic vacuoles, several other pathologic changes were found in H. pylori-infected gastric epithelium, among which mitochondrial lesions, with loss of cristae or matrix lysis, and increased autophagy of damaged mitochondria (Fig. 2I,J). The close direct or ER-mediated interaction found between VCVs and mitochondria is likely to account for the focal VacA reactivity found in the latter organelles, which sometimes was directly coupled with pathologic changes (Fig. 2I,J).
Particle-rich cytoplasmic structures (PaCSs) 24 , characterized by a collection of barrel-like particles heavily reactive for 19 S and 20 S proteasome, were detected in the cytoplasm below the nucleus (Fig. 3G), especially in cells showing less prominent cytotoxic changes and scarce or no auto(phago)lysosomes. Such structures were also intensely positive for FK1 antibody-reactive polyubiquitinated proteins (Fig. 3G) and, moderately/focally, for VacA (Fig. 3H), while they were unreactive for K63-linked pUb chains (Fig. 3I) and p62/SQSTM1 protein (not shown). They thus reproduced structural and cytochemical patterns of the PaCSs previously seen in H. pylori gastritis, gastric cancer and several cell lines [24][25][26] , including their close topographic relationship with surrounding ribosomes and RER cisternae (Fig. 3G-I). Unfortunately, despite testing different antibodies raised against the K48-linked pUb chains (i.e., the type of pUb chains known to be selectively associated with proteasomal degradation 27 , we did not find any antibody which worked in our TEM experimental conditions (i.e., aldehyde-osmium fixed resin-embedded specimens). This prevented a direct proof of K48-linked pUb nature of the FK1-positive polyubiquitinated proteins stored by PaCSs.

Discussion
A prominent finding of this investigation of H. pylori-infected human gastric epithelium in vivo was the detection of a population of small subluminal vesicles (and/or tubulovesicles), interposed with or overlying mucin granules, which were essentially lacking in non-infected epithelium. The actual presence of VacA immunoreactivity in a substantial fraction of such vesicles as well as in endocytic pits of the luminal plasma membrane strongly supports the endocytic-endosomal nature of such vesicles and their likely induction by bacterial infection, as previously suggested by in vitro experiments 22,[28][29][30] . The non-coated tubular nature of these endocytic pits suggests that the clathrin-independent non-caveolar pinocytic mechanism 31 of VacA internalization documented in experimental models in vitro 9,32 may have an in vivo counterpart in H. pylori-infected patients. Given their abundance in the infected epithelium, especially in association with H. pylori intimately adhering to surface epithelial cells, it seems likely that the endocytic-endosomal vesicles represent the main route of VacA cellular uptake in vivo. Although VacA-containing OMVs as well as whole H. pylori bodies were also found to intimately adhere with and enter epithelial cell lines in vitro and the gastric epithelium in vivo 22,23,33,34 (the present study), these remained less common findings than VacA-containing endocytic-endosomal vesicles inside the cells of H. pylori-infected gastric biopsy samples here studied. Our immunocytochemical findings also suggest that endosomal vesicles represent the main route of VacA intracellular trafficking, from subluminal early to juxtanuclear late endosomes, where most VCVs accumulate in vivo.
Of special interest is our observation of VCVs frequently contacting and also opening into adjacent ER cisternae. Of high interest is also the observation of close simultaneous interactions of VCVs with ER cisternae and mitochondria, of which here we provide the first in vivo demonstration as VacA targets. Our findings thus provide the in vivo counterpart in H. pylori-infected human epithelium of the in vitro observations by Kern et al. 12 in non-gastric cell lines (i.e., Jurkat T-cell and epithelial HeLa cell lines) showing that VacA also targets ER and Golgi and suggesting that VCVs may have a key role in VacA intoxication processes beyond the vacuoles. Indeed, VCVs seem to act as a platform to trigger specific trafficking pathways exploited by the toxin. Our direct immunocytochemical detection of VacA inside Golgi cisternae further confirms Kern et al. 's data 12 , although the exact route of VacA transport to the Golgi (from ER to Golgi?) remains unknown. Our in vivo findings also support Calore et al. 's in vitro observations 35 suggesting that VacA might be transferred to mitochondria by endosomalmitochondrial juxtapositional exchange.
Intriguingly, the existence of distinct interaction domains between the ER and other organelles (such as mitochondria and endosomes), known as membrane contact sites (MCSs), has been recently demonstrated (reviewed in 36 ). At MCSs, organelle membranes are closely apposed and tethered (but apparently do not fuse), and here various protein complexes might work in concert to perform specialized functions as binding, sensing and transferring molecules, as well as engaging in organelle biogenesis and dynamics 36 . Through the establishment of such physical contacts with different cell organelles, ER seems thus emerging as a key player in spaziotemporal control of organellar dynamics inside the cell. It has been speculated that these interorganellar "synapses" might serve as a direct delivery route between compartments, bypassing usual trafficking pathways known so far 37 . Whether canonical MCSs may have a role in VacA trafficking it remains to be investigated.
The endosomal accumulation of VacA toxin and other H. pylori antigens, like LPS O-antigen, is especially interesting considering 1) the well-known role of endosomes in taking up, storing and processing antigens to be membrane-presented in HLA molecules background 38,39 , 2) the high de novo expression of HLA-DR by H. pylori-infected gastric epithelium 19 , 3) the capacity of VacA to interfere with antigen processing by professional APCs at endosomal level 14 , as well as 4) our present finding of endosomal localization of cathepsin E, an aspartic protease crucial for antigen processing in several APCs [15][16][17]40,41 . A VacA-sensitive role for cathepsin E in H. pylori antigen processing at endosomal level would account for the expression increase and expansion (including endosomal involvement in addition to RER) of this protease we found in the infected gastric epithelium 19 (this study). This would also suggest the VacA-and cathepsin E-storing endosomal compartment as an appropriate site for the inhibitory action of VacA on cell processing of H. pylori antigens.
Several pathologic changes have been reported in cell lines incubated with H. pylori or its VacA toxin including, besides VCVs [28][29][30]42,43 , mitochondrial lesions 11,44,45 , and altered autophagic/lysosomal processes 7,20,21 . We found that all such changes occurred also in vivo in the infected human gastric epithelium, although cell vacuolation was less prominent in vivo than that observed in non-confluent cell monolayers in vitro, while being more akin to that seen in confluent cell monolayers 46 (our unpublished data). In addition, we provided direct immunocytochemical evidence for in vivo accumulation of VacA toxin inside swelled endosomes (i.e., VCVs) as well as mitochondria, which in turn also showed structural damage. The latter finding seems especially relevant as VacA has been shown to damage mitochondria in vitro causing mitochondrial membrane depolarization and cytochrome c release 11,45 . This in turn may potentially trigger mitophagy, the selective form of autophagy targeting damaged mitochondria to limit cell damage and prevent cell death 7,47 . In this respect, here we provide direct in vivo evidence of autophagic degradation of damaged mitochondria containing VacA.
VacA activity was found to persist for a long time inside cultured gastric epithelial cells 48 whose vacuoles showed ultrastructural and cytochemical patterns of both late endosomes and lysosomes 28,29,49 and, especially after prolonged incubation time (i.e., 24 hours or more), also showed autolysosomal features 28,50 . These in vitro findings well fit with present in vivo detection of close topographical and cytochemical correlations between VCVs and supranuclear auto(phago)lysosomal bodies.
Aggresome-like induced structure (ALIS)-type bodies reactive for polyubiquitinated proteins and p62/SQSTM1 51 have been reported by confocal microscopy in gastric epithelium of mice infected with a mouse-adapted H. pylori strain 52 . Apparently similar pUb-storing bodies have been characterized ultrastructurally as autophagosomes/autolysosomes in LPS-matured dendritic cells (DCs) 53,54 and macrophages 55 , and suggested to be storage site of potentially antigenic polyubiquitinated proteins to be processed and membrane-presented by such cells 54,56,57 . In the present study, we ascertained at ultrastructural level the development in H. pylori-infected human gastric epithelium of cytoplasmic bodies reactive for K63-linked pUb chains, p62/SQSTM1 protein, and LC3 protein and obtained direct evidence of a H. pylori-related origin of such bodies by detecting bacterial remnants, toxins, and antigens inside them. The presence of vesicular membranes, LC3 and p62/SQSTM1 proteins, and K63-linked pUb chains is highly suggestive for an autophagic component 53,[58][59][60] . Furthermore, we also obtained ultrastructural and cytochemical (e.g., cathepsin D reactivity) evidence for a lysosomal contribution to their genesis. Taken together, our findings support an auto(phago)lysosomal nature of VacA-containing ALIS-type bodies developing in H. pylori-infected human gastric epithelium. A role for VacA itself in the origin and persistence of auto(phago)lysosomal bodies may be considered, given the toxin-induced loss of cathepsin D activity, crucial for lysosomal function 21,61 . Our detection of H. pylori LPS in ALIS-type auto(phago)lysosomal bodies is also worth noting, given the role played by bacterial LPS in their genesis in human DCs 54 .
Supranuclear auto(phago)lysosomal bodies were more frequently found in cells showing prominent cytotoxic lesions, while being scarce or absent in those lacking such changes. This supports a role of VacA-dependent cytotoxicity, with special reference to damaged organelles, in the genesis of auto(phago)lysosomes, in keeping with previous observations in vitro concerning the fate of VacA-induced vacuoles 28,50 .
The UPS-rich PaCS is a focal collection of proteasome barrel-like particles, polyubiquitinated proteins and heat-shock proteins identified in some neoplastic or fetal tissues and cell lines [24][25][26]54,55 . This structure, specifically arising inside ribosome-rich cytoplasm below the nucleus, is likely to have a role in quality control and degradation of misfolded, mutated or anyway damaged cytosolic proteins, known to be produced in excess in neoplastic or fetal cells as well as in hematopoietic cells specifically stimulated by trophic factors and interleukins 25,26,54 . Our detection of PaCSs in H. pylori-infected epithelium might be suggestive of an altered protein turnover resulting in UPS stress.
In H. pylori-infected human gastric epithelium, unlike auto(phago)lysosomal bodies, PaCSs were usually found in the cytoplasm below the nucleus and, preferentially, in cells showing limited cytotoxic changes. Concerning the polyubiquitinated proteins they store, PaCSs differed from autophagosomes/autolysosomes in being unreactive for K63-linked pUb-directed antibodies, while being reactive for the FK1 antibody, known by free ribosomes (asterisk) and RER cisternae (arrow). Also note in (G) selective PaCS reactivity for the polyubiquitinated protein-specific FK1 antibody (20-nm gold). (H) Focal PaCS reactivity for VacA (enlarged in H1). Note VacA reactivity (white arrow) also at the level of a mitochondrion (white asterisk) enveloped by an ER cisterna (black asterisk). (I)The unreactivity of PaCS for K63-linked pUb chains is shown; also note direct merging of several RER cisternae with the PaCS (arrows).
SCIEntIFIC REpoRtS | 7: 14526 | DOI:10.1038/s41598-017-15204-z to recognize in vitro polyubiquitinated proteins exhibiting either K63-or K48-linked pUb chains 62 . As only K48-linked polyubiquitinated proteins are known to selectively associate with proteasome 27 , it seems likely that the FK1-reactive and K63-linkage-unreactive polyubiquitinated proteins associated with proteasome particles inside PaCSs are to be interpreted as K48-linked proteins. Interestingly, the direct opening we found of some ER cisternae into PaCSs may indicate a pathway through which VacA reaches this essentially cytosolic UPS-rich structure and, more in general, might also suggest a way for endocytosed exogenous antigens to reach cytosolic proteasome for class I cross-presentation 54,63 .
Our present findings show that, notwithstanding their common association with H. pylori infection and common storage of polyubiquitinated proteins, ALIS-type auto(phago)lysosomal bodies and PaCSs are cytochemically linked to two different protein-degradative pathways, namely: 1) the K63-linked pUb, p62/SQSTM1 and LC3 positive autophagic-endolysosomal system, and 2) the proteasome and K48-linked pUb chain positive UPS, respectively. In keeping with our recent findings in human DCs in vitro 54 , PaCSs may represent an early, chaperon protein-promoted and ubiquitin/proteasome-mediated cellular attempt to repair or degrade H. pylori-induced misfolded proteins, whereas the ALIS bodies may result from cellular activation of the autophagic/lysosomal pathway by severe cytotoxic lesions affecting cytoplasmic organelles.
In conclusion, our in vivo study shows that VacA mainly enters H. pylori-infected human gastric epithelium by endocytosis and accumulates into endosomes and endosome-derived VCVs, which directly communicate with ER cisternae and ER-enveloped mitochondria. This latter finding supports toxin trafficking from VCVs to other organelles such as ER/Golgi and mitochondria, as previously suggested by in vitro experiments. De novo endosomal expression by infected gastric epithelium of the antigen-processing proteinase cathepsin E may have a role in the complex VacA-associated host immune-inflammatory response which characterizes H. pylori infection. VacA-induced cytotoxic effects on cell organelles and protein turnover is associated with activation of main cellular degradative systems with persistent accumulation of degradative products inside auto(phago)lysosomes.

Materials and Methods
Human biopsy samples. We reinvestigated biopsy samples of gastric mucosa taken in the period 1981-95 from 26 patients (15 males and 11 females, aged between 26 and 79 years) undergoing routine endoscopic and histologic examination for dyspepsia as requested by the physician in charge of the patient and with the written consent of the patient 24 . The study has been approved by the Ethics Committee of Fondazione IRCCS Policlinico San Matteo (Pavia, Italy) as a reinvestigation of archival material along the same line (i.e., diagnosis of H. pylori-dependent gastritis) as for the original written consensus. All the methods were performed in accordance with the relevant guidelines and regulations.
Six biopsy specimens (3 from the antrum and 3 from the corpus of the stomach) were taken from each patient. From each biopsy site, 2 samples were processed for light microscopy and 1 sample for TEM. Fifteen patients resulted H. pylori-positive in all biopsies at both light microscopy (Giemsa staining and histochemistry for H. pylori LPS) and TEM (detection of characteristic bacterial ultrastructure coupled with VacA and H. pylori LPS cytochemistry) investigation. Their biopsy specimens were thus judged as suitable for present investigation on VacA interaction with human gastric mucosa. Four patients resulted H. pylori-negative in all biopsies from the antrum and corpus, extensively investigated at both light microscopy and TEM as above. The eight TEM-processed biopsy specimens from these 4 patients were thus taken as negative controls in the present study. Biopsy specimens from the remaining 7 patients showed more limited H. pylori colonization, often unequally distributed among different specimens. These cases were not further investigated in the present study.
TEM and ultrastructural immunocytochemistry. For TEM investigation, biopsy samples were fixed for 4 hours with 2% formaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.3), followed by 1% osmium tetroxide for 1 hour, and then embedded in Epon-Araldite resin 24 . Thin (~70 nm) sections were stained with uranyl-lead or underwent the immunogold procedure followed by uranyl-lead counterstaining, as previously detailed 22,23,64 . Specimens were analyzed by a Jeol JEM-1200 EX II transmission electron microscope equipped with an Olympus CCD camera (Mega View III). Images were processed and assembled by using the Adobe Photoshop CS5 software.
As secondary Abs, anti-rabbit or anti-mouse immunoglobulins labeled with 10, 15 or 20 nm gold particles (British Bio Cell, Cardiff, UK, and Aurion, Wageningen, The Netherlands) were used 23 Tests to evaluate the specificity of immunogold labeling were carried out using antibodies absorbed with excess antigen and omitting or substituting the specific antibodies in the first layer of the immunogold procedure. Positive and negative controls were obtained by parallel investigation of H. pylori cultures, epithelial cell cultures, and H. pylori-positive or -negative gastric mucosa specimens as in previous studies 22,23 . In particular, both anti-VacA antibodies used were tested using parallel TEM investigation on well-characterized bacterial cultures either VacA-producing (H. pylori strains 60190, ATCC 49503, and CCUG 17874, from Culture Collection University of Göteborg, Sweden) or not producing the toxin (H. pylori strain 60190:v1, the isogenic mutant of the 60190 strain in which the vacA gene was disrupted by insertional mutagenesis, kindly provided by T.L. Cover, Nashville, TN). We also assessed the specificity of these antibodies by means of SDS-PAGE, followed by Western blotting, on bacterial lysates and broth culture filtrates of the aforementioned H. pylori strains.
Data availability. No datasets were generated or analysed during the current study.