Ultrastructural insights into pathogen clearance by autophagy

Abstract Autophagy defends cells against proliferation of bacteria such as Salmonella in the cytosol. After escape from a damaged Salmonella‐containing vacuole (SCV) exposing luminal glycans that bind to Galectin‐8, the host cell ubiquitination machinery deposits a dense layer of ubiquitin around the cytosolic bacteria. The nature and spatial distribution of this ubiquitin coat in relation to other autophagy‐related membranes are unknown. Using transmission electron microscopy, we determined the exact localisation of ubiquitin, the ruptured SCV membrane and phagophores around cytosolic Salmonella. Ubiquitin was not predominantly present on the Salmonella surface, but enriched on the fragmented SCV. Cytosolic bacteria without SCVs were less efficiently targeted by phagophores. Single bacteria were contained in single phagophores but multiple bacteria could be within large autophagic vacuoles reaching 30 μm in circumference. These large phagophores followed the contour of the engulfed bacteria, they were frequently in close association with endoplasmic reticulum membranes and, within them, remnants of the SCV were seen associated with each engulfed particle. Our data suggest that the Salmonella SCV has a major role in the formation of autophagic phagophores and highlight evolutionary conserved parallel mechanisms between xenophagy and mitophagy with the fragmented SCV and the damaged outer mitochondrial membrane serving similar functions.

4 and 8 hours after infection, a second longer-lasting ubiquitin signal appears in closer vicinity to the Salmonella surface if the bacteria are still not cleared by autophagy. 10 The bacterial-associated ubiquitin signal is recognised by selective autophagy receptors such as sequestosome 1 (SQSTM1/p62), 11 Tax1 binding protein 1 (TAX1BP1/ CALCOCO3) 12 and its paralogue nuclear domain 10 protein 52 (NDP52/CALCOCO2) 13 as well as optineurin (OPTN). 14 These selective autophagy receptors characteristically can bind to both ubiquitin on Salmonella and the LC3-positive autophagic membranes, thereby enabling assembly of the phagophore. In addition to the LC3-adaptor interaction, the way by which adaptors co-operate with the rest of the autophagic machinery to recognise and engulf Salmonella is an area of active investigation. In general terms, it appears that early autophagy proteins such as members of the ULK complex, the phosphatidylinositol 3-phosphate (PI3P) effectors WIPI proteins and the lipidation machinery component ATG16 all can recognise parts of the adaptor proteins. For example, it was recently shown that NDP52 forms a complex with FIP200 and SINTBAD/NAP1 leading to the recruitment of the autophagy machinery to Salmonella in the cytosol. The ULK complex localises to the Galectin-8-positive Salmonella surface, highlighting the importance of the damaged SCV for phagophore formation. 15,16 In addition, the WIPI2 PI3P effector promotes the localization of the TBK1 kinase to the invading Salmonella prior to autophagic engulfment (https://www.ncbi.nlm.nih.gov/ pubmed/27370208).
Salmonella first enters the cell interior through macropinocytosis and resides within a novel single membrane compartment termed SCV. 17 The SCV gradually acquires characteristics of the endocytic compartment as it matures. Within the SCV, Salmonella can replicate or, alternatively, it can escape into the cytosol where it is found either "naked" or still partially associated with ruptured SCVs. Autophagy appears to be triggered by bacteria in ruptured SCVs as an elimination mechanism although it also has been suggested to enable repair of the damaged membranes. [18][19][20] The sequence of ubiquitin-triggered and autophagy receptor-dependent recruitment of autophagosome membranes is well established, but the site of ubiquitination that initiates this sequence of events is less well-known. Equally perplexing are the topological relationships between the bacterial outer membrane, the SCV and the phagophore double-membrane as it is being formed through the action of adaptors and the autophagy core machinery. Finally, the stoichiometry of bacteria within the autophagic membranes is not entirely settled.
To establish the exact distribution of ubiquitin, autophagy receptors and LC3-positive membranes, we have performed a detailed ultrastructural analysis of the Salmonella-host membrane system. We show the distribution of ubiquitin, selective autophagy receptors and autophagosomal membranes using electron microscopy (EM). Our results demonstrate the exact topology of host and Salmonella membranes, ubiquitinated target proteins and components of the autophagy machinery and provide evidence that SCV membrane proteins are ubiquitination targets that are recognised by selective autophagy receptors, such as TAX1BP1 leading to the assembly of phagophores. Bacteria that have lost all remnants of SCV membranes are less likely to recruit autophagosomal membranes. These findings highlight the similarity between mitophagy and xenophagy; in the former, proteins associated with the outer mitochondrial membrane, which is reminiscent of the SCV, are ubiquitinated to promote clearance of mitochondria via autophagy.  Figure 1A and 2E) or enclosed fully or partially by phagophores (marked with a red star in Figure 1A and B and Figure 2C and E). Interestingly, phagophores not only capture single bacteria, but frequently we observed several Salmonellae engulfed by a single large phagophore, which appears to be zippered along the outline of the bacteria ( Figure 1A and B and Figure 2E). In places several single phagophores closely align along neighbouring Salmonellae, suggesting that the large phagophores may originate through fusion of several smaller phagophore membranes ( Figure 1B).
2.2 | Ruptured SCV membranes are present between the Salmonella surface and the phagophore We next used EM to perform a detailed analysis of the Salmonellahost membrane topology. The bacterium itself is surrounded by a double membrane, an outer and an inner membrane separated by periplasm. The detailed topology of the Salmonella surface is shown in Figure S1A.
After invasion, the pathogen resides in the SCV, which allows replication of the bacteria. At the ultrastructural level, the SCV appears as a continuous single membrane often containing numerous very small vesicles in the space between the Salmonella surface (SS) and the SCV membrane (Figure 2A,B). SCV damage exposes the bacteria to the cytosol, which triggers the formation of a dense ubiquitin coat and the assembly of the phagophore. We next used conventional EM to visualise ultrastructural detail of phagophores on the bacterial surface, and found that fragments of the SCV membrane are often present between the Salmonella surface and the phagophore ( Figure 2C,D,E and G). In contrast, "naked" cytosolic bacteria without an SCV membrane (green stars) are not surrounded by a phagophore (Figure 2E and F).  Our quantitation shows that for both TAX1BP1 and p62, the highest amount of signal is present on the SCV membrane and on the inner membrane of the phagophore but not directly on the Salmonella surface ( Figure 6C and G). TAX1BP1 is also a cargo adaptor protein for MYO6, a myosin motor that has previously been shown to be recruited to ubiquitin-positive Salmonella that also contain TAX1BP1, p62 as well as LC3. 12 The ultrastructural analysis shows that MYO6 displays a very similar distribution as TAX1BP1 on the SCV membrane and on the cytosolic outer membrane of the phagophore ( Figure S2).  Figure 7A). The results were plotted as the percentage of Galectin-8-positive bacteria that also contain ubiquitin ( Figure 7B) or the percentage of ubiquitinpositive bacteria that carry a ubiquitin coat ( Figure 7C). Our data show that more than 80% of Salmonellae that are positive for Galectin-8 also contain a ubiquitin signal and more than 70% of Salmonellae that are ubiquitin-positive contain Galectin-8-positive SCVs. These results were confirmed by EM, which demonstrates that the majority of the ubiquitin signal is not on the Salmonella surface but associated with the SCV irrespective of whether the SCV was already associated with an autophagic phagophore ( Figures 7D,E and 8).

| DISCUSSION
The correct deposition of a ubiquitin coat is crucial to create a docking space for the autophagy machinery but also a signalling platform for local nuclear factor kappa B (NFkB) activation. Recent results from a number of labs obtained using super-resolution microscopy, revealed the uneven distribution of ubiquitin around cytosolic Salmonella with different ubiquitin chains segregated into distinct patches. 21,22 Although these high-resolution light microscopy approaches allow the visualisation of ubiquitin-enriched and less dense regions around the bacteria, they do not provide the optical resolution to identify the exact nature and position of the ubiquitin target. We therefore used EM in this study to determine the exact distribution of the ubiquitin and also about 10% of Salmonellae are positive for FK2-positive ubiquitin signal. 9 Therefore, the early response not only involves the Galectin-8 "eat-me" signal but also uses ubiquitin as an amplification signal to recruit autophagy receptors leading to phagophore formation around fragments of the SCV. This hypothesis is supported by our finding that "naked" Salmonella particles without any SCV fragments detectable at the ultrastructural level, are less frequently surrounded by a phagophore.
Our findings suggest that multiple Salmonella particles can exist within one very large phagophore raise some interesting points. It is likely that such membrane nucleation and expansion must depend on tight interactions between selective autophagy adaptors recognising the Salmonella eat-me signals (eg, ubiquitination) and the autophagy machinery, and this probably explains why the phagophore membranes follow tightly the contour of the bacterial outer membrane.
Recent work has discovered interactions between NDP52 and the ULK complex as well as between TBK1 and WIPI2. 15,23 Given the tightness of these engulfing membranes with their targets, additional such interactions will undoubtedly be discovered between adaptors and early autophagy proteins. Of note, the Atg19 cargo receptor in yeast contains multiple Atg8 binding sites, and this facilitates tight wrapping of the autophagic membrane around its cargo. 24 Similarly, the presence of multiple adaptors on Salmonella (including p62, optineurin, TAX1BP1) each with its own LC3 interacting region and the ability to multimerize may create a multivalent interaction with the autophagosomal membrane that would enable tight engulfment.
What is less clear is the mechanism by which these large phagophores are formed. In some of our images, we can clearly see what appear to be smaller phagophores fusing together whereas in others there appears to be a continuous phagophore making its way around a cluster of Salmonella particles from both directions. The former type of event would predict that there must be some activity within the autophagic components capable of fusing double membrane phagophores, perhaps akin to the final step of phagophore closure. Such activity would also be relevant for cases where a large mitochondrial fragment is enclosed by multiple phagophores fusing together at some point. 25 The latter type of event (a seemingly continuous phagophore enclosing a very large target) is characterised by autophagosomes beyond the usual size limit that has been reported to range from 0.15 μm diameter for the CVT pathway in yeast to 1 μm diameter for autophagy and mitophagy in mammalian cells. 26 What mechanisms ensure that the autophagic machinery can stay engaged continuously on structures up to 30 μm in circumference? Are there feedback controls operating for such unusual autophagosomal sizes? If this is in principle possible, why do we also see multi-phagophore membranes engulfing similarly-sized Salmonella clusters? What determines these two modes of selective autophagy? Answering these questions must be one of the objectives of future work as the precise mechanisms of Salmonella autophagy are established.

| Reagents and antibodies
The following commercial antibodies were used in this study; mono-

| Plasmids
In this study, we used the following plasmids: human TAX1BP1 in pEGFP, 27 human MYO6 without any splicing inserts in pEGFP, 28 Galectin-8 in pEGFP was a kind gift from Felix Randow. 9

| Cell culture
Primary mouse embryonic fibroblasts (MEFs) from wild-type mice were prepared as previously described 12 and immortalised using the SV40 large T-antigen (pEF321-T). The MEF cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% foetal bovine serum, 2 mM Glutamine and 100 U/mL penicillin and 100 μg/mL streptomycin under 5% CO 2 and expanded over several weeks to make mixed immortalised MEF cultures.

| Immunofluorescence microscopy
MEFs were plated on glass coverslips, fixed in 4% formaldehyde, permeabilised in 0.02% Triton X-100 in phosphate-buffered saline (PBS) and blocked in 1% BSA in PBS. The fixed cells were incubated with primary antibodies at room temperature for 2 hours followed by secondary antibodies conjugated to Alexa fluor 488, 568 or 647. Images were taken on a Zeiss LSM710 confocal microscope with ZeissZEN software. Manual quantitation was performed to determine the % of Galectin-8-positive Salmonella that were also LC3-or ubiquitin-positive for more than 100 bacteria from multiple cells from three independent experiments. F I G U R E 8 Cartoon illustrating the suggested role of the Salmonellacontaining vacuole (SCV) membrane in phagophore formation. Ubiquitin is not only present on the Salmonella surface, but also enriched on the fragmented SCV nanogold particles (Nanoprobes, 2004). After washing, the gold labelling was intensified by using a Gold Enhance EM kit (Nanoprobes, 2113). After stopping the gold enhancement reaction in 1% aqueous sodium thiosulfate solution, the cells were washed in distilled water and post fixed in 1% OsO4 containing 1.5% K 4 [Fe (CN) 6 ] in 0.1 M phosphate buffer for 1 hour. After washing in distilled water, the cells were dehydrated with a graded series of ethanol, infiltrated with resin (Agar Scientific, Agar 100 resin kit, R1031), which was polymerised at 60 C for 2 days. Ultrathin sections were collected onto grids, post stained with uranyl acetate and lead citrate and viewed using an FEI Technical Spirit TEM.

| Quantitative analysis of immunogold labelling
Electron micrographs were taken from single ultrathin sections from randomly-selected fields that included at least one immunolabelling gold particle with an identifiable autophagosome or/and salmonella. Micrographs were analysed with ImageJ-Win64 software. The Gold particles were counted to be associated with each membrane.