Imaging ATG9A, a Multi-Spanning Membrane Protein

,


Introduction
ATG9A is the only transmembrane protein of the core autophagy machinery and is trafficked between the Golgi and a cytosolic ATG9A vesicle compartment, transiting through the endosomal compartment 1 . Having long been enigmatic, ATG9A has recently been described to function as a lipid scramblase, as it equilibrates lipids across membrane bilayers 2 , 3 . It is now clear that ATG9A resides at the top of the hierarchy in autophagosome formation, and its study is, thus, vital for understanding autophagy 4 , 5 . As such, ATG9A vesicles have been recently proposed as the "seed" of the autophagosome 6 , 7 . However, previous studies have demonstrated that ATG9A only transiently interacts with the forming autophagosome at different steps of its maturation and does not integrate into the autophagic membrane 6 , 8 , 9 , 10 , 11 . Thus, further investigations are needed to completely unravel the role and potential multiple functions of ATG9A in autophagosome formation. However, the discrepancy between the current models and the previous data can only be resolved through targeted experiments addressing the trafficking of ATG9A There are various tools in use to study ATG9A, each with advantages and disadvantages, and the use of these tools is complicated by the structure of ATG9A, its molecular function, and cellular trafficking 2 , 8 , 12 . ATG9A forms a homotrimer, is glycosylated, and is trafficked throughout the cell to compartments such as the Golgi, the endosomes, and the plasma membrane 13 , 14 . Given its complex itinerary, there are several challenges in interpreting readouts such as ATG9A dispersal from the Golgi upon specific treatments or stimuli (such as nutrient and serum starvation). ATG9A is extremely dynamic in terms of vesicular trafficking; indeed, ATG9A-containing vesicles have been defined as the ATG9A compartment in the context of starvation-induced autophagy. The ATG9A compartment, formed by these dynamic vesicles, transiently interacts with several intracellular organelles 8 , 15 , 16 , 17 . The techniques described here, including immunofluorescence, live imaging, and glycosylation assays, should aid in the detection and understanding of ATG9A biology. In particular, the approaches described in this article will help to address questions about localization to specific cellular compartments and interactions with specific protein partners and/or membrane compartments. As the ATG9A hydrophobic conserved core domain (PFAM domain PF04109) has a unique topology and ATG9A cycles between several membrane compartments, researchers should be aware of certain pitfalls and artifacts when transiently overexpressing ATG9A, including, but not restricted to, endoplasmic reticulum (ER) retention. Other possible issues may arise due to misfolding of the protein, artifactual aggregation in normal growing conditions, or insufficient detection of the vesicular compartment due to suboptimal permeabilization protocols for immunofluorescence.
When imaging endogenous ATG9A, care must be taken in the sample preparation and image acquisition to ensure the quality of the subsequent quantitative analysis and the correct interpretation of the data. Combining the techniques described in this article with standard biochemical approaches (such as immunoprecipitation or pull-down experiments not described here) should improve our understanding of ATG9A function. This experimental toolkit is intended to help new researchers navigate some of the assays required to determine the function of ATG9A in their biological system.

Protocol
All the reagents used in this study are commercially available, except for the ATG9A DNA constructs and homemade STO-215 antibody (see Table of

Image acquisition
1. Turn on the confocal microscope. Open the imaging software (see Table of

Image analysis of ATG9A dispersal
1. Download FIJI software from the internet (see Table   of Table of Materials), and transfer to a 1.5 mL microcentrifuge tube.  Table of Materials). Incubate as instructed by the PNGase F manufacturer.
7. Add a volume of 3x Laemmli sample buffer to achieve a 1x concentration, and incubate at 65 °C for 5 min before loading for electrophoresis using a Tris-acetate gel to maximize the separation of proteins. Transfer the proteins from the gel to a suitable membrane (i.e., PVDF [polyvinylidene difluoride]) using standard western blot protocols 21 (see Table of Materials).
NOTE: Boiling the samples at 95 °C will cause ATG9A to be aggregated, thus reducing the detection of ATG9A.
8. Perform the western blot using specific antibodies for ATG9A (STO-215 antibody, produced in-house 1 ) (see Table of Materials). Leave the higher-molecular weight section of the membrane uncut to visualize the higher ATG9A molecular weight species.

Representative Results
ATG9A is a transmembrane protein associated with several intracellular membrane compartments 8 , 17 , 22 , 23 , 24 . In basal conditions, ATG9A is mainly localized at the trans-Golgi network (TGN), as indicated by the immunofluorescence of the endogenous protein and the overlaps with GM130, a cis-Golgi marker (Figure 1A), as well as in small vesicles that partially overlap with the endocytic recycling compartment (ERC) 23 . ATG9A localization at the Golgi can be detected using different immunofluorescence protocols. However, the vesicular fraction of ATG9A, as well as its change of localization, in particular the increase in the vesicular pool, in response to specific stimuli such as nutrient and serum starvation, can be quite variable in intensity and difficult to visualize with conventional imaging approaches. The ratio between ATG9A localized at the Golgi and ATG9A localized to a vesicular fraction is termed the ATG9A dispersal rate.
To detect changes in the ATG9A dispersal rate, for instance upon EBSS treatment, which is used to deplete both serum and amino acids, a Golgi marker such as GM130 or TGN46 and a cytoskeleton marker such as Phalloidin, which stains the cell contour 25 , are useful to readily quantify the ATG9A dispersal ( Figure 1B). Importantly, the mean fluorescence ratio analysis can only be interpreted as a comparative measure between conditions rather than as a fixed rate of dispersal. The ratio between compartments is highly dependent on biological and non-biological factors such as the cell line used, the staining quality, or the thresholding methods applied (Figure 1B). For this reason, the researcher needs to set up a pipeline that is able to detect ATG9A Golgi enrichment in their specific experimental conditions and then extend the analysis with the same parameters to all the images in the set to be analyzed. Representative binary images and areas selected for the analysis of ATG9A mean fluorescence are shown as a guide in Figure 1B. domains, of which the C-terminal sequence encompasses almost half the protein 12 . Importantly, the localization pattern of overexpressed ATG9A can be influenced by which protein end is tagged (Figure 2A). In particular, when using transient expression systems and tagging ATG9A directly on its N-terminus with a fluorescent tag (e.g., eGFP, mRFP, or derivatives), its Golgi localization can be partially compromised, with less enrichment seen in basal (i.e., fed) conditions, while the ATG9A vesicles are still readily visible (Figure 2A). Tagging ATG9A on its C-terminus seems to slightly induce larger GFP positive clusters that could be aggregated. Finally, a monomeric version of mRFP-ATG9A also shows similar fluorescent clusters of vesicles and little Golgi staining in overexpressing cells (Figure 2A).
ATG9A folds in the ER membrane before being trafficked to the Golgi and ATG9A vesicles. During its residence in the ER, ATG9A becomes modified by N-linked glycans on Asparagine 99, and then upon reaching the Golgi, it acquires complex, mature N-linked glycans 1 , 14 . This modification by glycosylation can be detected through western blot by the appearance of a double band 14 . Consistent with its intracellular localization, most endogenous ATG9A harbors complex N-linked glycans, and, therefore, the highermolecular weight band is predominant, with a faint lowermolecular weight band also visible ( Figure 2B). The presence of a double band is most readily seen when using Tris-acetate gels to improve the resolution of higher-molecular weight proteins (Figure 2B, control, t = 0). When the endogenous protein is subjected to PNGase F (Peptide:N-glycosidase F) treatment, which removes most of the complex N-linked glycans, the protein runs as a single band (Figure 2B, PNGase F, t = 0). Therefore, the N-linked glycosylation status of ATG9A can be used as a proxy to monitor the exiting of ATG9A from the ER to the Golgi, which is reflected by the relative ratio between the two bands.
When transfecting mRFP-ATG9A constructs transiently, the overexpressed protein initially accumulates in the ER, potentially because the trafficking machinery is unable to fold and traffic all the ATG9A, and the lower molecular weight band is predominant (Figure 2C, control t = 0). between an N-terminal fluorophore and ATG9A helps the overexpressed protein behave similarly to the endogenous one (Figure 3). Indeed, overexpressed mCherry-3xFLAG-ATG9A colocalizes with the Golgi marker GM130 in fed conditions ( Figure 3A). Importantly, this localization and the ATG9A vesicular compartment are preserved over time, allowing the spatiotemporal study of the trafficking of ATG9A ( Figure 3B).

Discussion
This study illustrates the various tools that can be used to investigate ATG9A localization. Firstly, this study describes how ATG9A can be visualized by immunofluorescence and how this can be quantified. Secondly, strategies that can be the Golgi staining of ATG9A, which eventually compromises the detection of ATG9A redistribution to other membrane compartments. Additionally, since ATG9A is present in many intracellular compartments 1 , 13 , 17 , 22 , 23 , 24 , 27 , 28 ,it is important to use specific membrane markers, together with ATG9A, to identify where ATG9A is located. Several approaches have been used in the past to quantify ATG9A localization, including Pearson's correlation coefficient for colocalization 29 . However, the partial overlap of ATG9A with the Golgi and the distinct vesicular compartment leads to a high number of pixel outliers, which may bias the interpretation of the correlation coefficient. For this reason, a more simplistic approach based on the ratio of the mean fluorescence in the two compartments to be analyzed is preferred, and this approach is less sensitive to cell-by-cell variability. For further information on image analysis through microscopy, readers are directed to this book chapter 30 .
When investigating the glycosylation status of ATG9A, the selection of gels for running the western blots is important.
For this protocol, 3%-8% Tris-acetate gels are preferred because they offer the highest resolution for larger proteins, but alternative gel compositions or running buffers that offer a good separation of high-molecular weight proteins can also be used. The experimenter can ensure the For the live-cell imaging of ATG9A, an Airyscan microscope, relying on the fast Airyscan function, provides optimal resolution of typically about 120 nm. For localization accuracy, frame rates of around 1-2 frames per second (fps) in super-resolution mode are optimal depending on how many channels are imaged. Similar confocal microscopes that can image at high speed can also be used for the imaging of ATG9A vesicles; however, it should be noted that the imaging speed can directly affect the detection of events and, therefore, affect the interpretation of the data.
In summary, the presented protocols describe ways to quantify and characterize ATG9A localization by immunofluorescence, live-cell microscopy, and its glycosylation status. These protocols can aid researchers working with ATG9A and help avoid some pitfalls.