Inhibition of Autophagy Alters Intracellular Transport of APP Resulting in Increased APP Processing

Alzheimer's disease (AD) pathology is characterized by amyloid beta (Aβ) plaques and dysfunctional autophagy. Aβ is generated by sequential proteolytic cleavage of amyloid precursor protein (APP), and the site of intracellular APP processing is highly debated, which may include autophagosomes. Here, we investigated the involvement of autophagy, including the role of ATG9 in APP intracellular trafficking and processing by applying the RUSH system, which allows studying the transport of fluorescently labeled mCherry‐APP‐EGFP in a systematic way, starting from the endoplasmic reticulum. HeLa cells, expressing the RUSH mCherry‐APP‐EGFP system, were investigated by live cell imaging, immunofluorescence, and Western blot. We found that mCherry‐APP‐EGFP passed through the Golgi faster in ATG9 knockout cells. Furthermore, ATG9 deletion shifted mCherry‐APP‐EGFP from early endosomes and lysosomes toward the plasma membrane concomitant with reduced endocytosis. Importantly, this alteration in mCherry‐APP‐EGFP transport resulted in increased secreted mCherry‐soluble APP and C‐terminal fragment‐EGFP. These effects were also phenocopied by pharmacological inhibition of ULK1, indicating that autophagy is regulating the intracellular trafficking and processing of APP. These findings contribute to the understanding of the role of autophagy in APP metabolism and could potentially have implications for new therapeutic approaches for AD.

APP is a transmembrane protein that is translated into the endoplasmic reticulum (ER) and then further transported along the Golgi apparatus and trans-Golgi network (TGN) toward the plasma membrane.Subsequently, it is rapidly internalized from the plasma membrane to the endosomal system [8,9].APP can cycle back to the TGN through the retromer complex or be routed to the lysosome for degradation.The site of intracellular APP processing is highly debated and might involve different cellular organelles: ER, Golgi, plasma membrane, endosomes, and potentially autophagosomes [3,[10][11][12][13][14].
Autophagy is critical to maintain protein homeostasis by degrading aggregated and potentially toxic proteins as well as dysfunctional organelles.Under starvation, autophagy is upregulated to provide the cell with nutrients.The autophagic process is highly regulated, and the conjugation of several autophagyrelated (ATG) protein complexes results in the formation of the double-membranous autophagosome, which delivers its content to the lysosome for degradation.ATG9 is the only transmembrane core autophagy protein residing in ER, Golgi, and a vesicular pool, therefore thought to deliver lipids to the growing phagophore [15,16].Upon initiation of autophagy, ATG9 is recruited and phosphorylated by the ULK1 complex, and thereby ATG9 and ULK1 kinase act far upstream of the autophagy cascade.The ULK1 complex acts first in the autophagy activation, and besides regulating ATG9, it recruits the vacuolar protein sorting 34 (VPS34) complex [17].The VPS34 complex generates phosphatidylinositol-3-phosphate (PI3P) at the phagophore [17].Notably, PI3P is deficient in AD brain tissue and PI3P was found to control APP processing depending on endosomal sorting complexes required for transport (ESCRT) [12].ATG9 was not only found to be closely intertwined and partially localized to endosomes [18,19] but also interacts with the PI3P-binding protein SNX4 [20].
The brains of AD patients show an accumulation of autophagosomes in dystrophic neurites, highlighting the importance of functional autophagy for brain homeostasis [21].The autophagic system is upregulated in early AD [22], followed by a downregulation in later disease stages [13,23].In addition to its degradative function, a role of autophagy is emerging in unconventional secretion [24].Autophagy mediates the unconventional secretion of Aβ in a mouse model of AD [25], and autophagosomes contain all necessary components for the processing of APP to generate Aβ [14].However, it is still unclear how autophagy is involved in APP transport and processing, and hence the generation and release of Aβ.
Most attempts that have been made to investigate the intracellular trafficking routes and the site of APP processing are based on steady-state conditions [26].Here, we present an approach to systematically follow the intracellular APP transport from the ER, by applying the "retention using selective hooks" (RUSH) system [27], to investigate the autophagy-mediated intracellular transport and processing of APP.Taking into account that ATG9 is involved in membrane trafficking and that APP is a membranebound protein, we were prompted to study APP processing in both ATG9 KO HeLa cells and through pharmacological manipulation of autophagy.
Using RUSH allows tracking of all stages of APP intracellular trafficking, including its journey from the ER, passage through the Golgi, transport to plasma membrane, endocytosis to early endosomes, and finally its arrival at lysosomes.We found that both ATG9 deficiency and pharmacological inhibition of autophagy shifted APP toward the plasma membrane, leading to increased processing of APP.This highlights the importance of autophagy, including ATG9 in APP processing, which is relevant for understanding the mechanism behind AD and to develop new therapeutic strategies for AD.

| Plasmids
The RUSH system consists of an ER hook protein fused to streptavidin (Str), which interacts with the streptavidin-binding peptide (SBP) linked to monomeric red fluorescent protein (mCherry) and enhanced green fluorescent protein (EGFP) tagged APP (mCherry-APP-EGFP).
The RUSH-Ii-Str_IRES_SBP-mCherry-APP-EGFP plasmid containing human invariant chain (Ii) of the major histocompatibility complex as an ER hook, an internal ribosomal entry site (IRES), and SBP-mCherry-APP-EGFP as a reporter have been cloned based on the Ii-Str_IRES_SBP-mCherry-E-cadherin construct, which was a gift from Franck Perez (Addgene plasmid #65289) [27].E-cadherin was cut out by FseI and PacI digest and was replaced with APP-EGFP amplified with the primers 5′-TTA AGG CCG GCC ACT GGA GGT ACC CAC TGA TGG and 5′-GCA TGG ACG AGC TGT ACA AGT AAT TAA TTA ACT GA from Addgene plasmid #69924 which was a gift from Zita Balklava and Thomas Wassmer [28].
Due to very low mCherry-APP-EGFP expression levels from the abovementioned plasmid, RUSH plasmids containing individually Ii-Str ER hook and SBP-mCherry-APP-EGFP were cloned based on the Gateway cloning system.The Ii-Str ER hook was amplified from the abovementioned RUSH-Ii-Str_IRES_SBP-mCherry-APP-EGFP plasmid with the primers 5′-CAA AAA AGC AGG CTC CGC CAC CAT GCA CAG GAG GAG AGC and 5′-CAA GAA AGC TGG GTC TCA CAT GGG GAC TGG GC and cloned into a Gateway pDONR207 vector.The SBP-mCherry-APP-EGFP was amplified from RUSH-Ii-Str_IRES_ SBP-mCherry-APP-EGFP plasmid using 5′-CAA AAA AGC AGG CTC CGC CAC CAT GTA CAG GAT GCA ACT CCT GTC and 5′-CAA GAA AGC TGG GTC GTG GTA TGG CTG ATT ATG ATC AGT TAT C primers and cloned into a Gateway pDONR207 vector.Both inserts were then shuttled into Gateway destination lentiviral plasmids pLenti or pHR with different antibiotic resistance genes.
Due to insufficient ER retention of mCherry-APP-EGFP by the Ii-Str hook, the lentiviral plasmid pCDH_Str-KDEL (from Franck Perez, Addgene plasmid #65307) was used as another ER hook in conjunction with the cloned pLenti_CMV_SBP-mCherry-APP-EGFP_blas vector.

| Cell Culture
HeLa and HEK293T cells were maintained in Dulbecco's modified Eagle medium (DMEM) with high glucose (D5796, Sigma) containing 10% fetal bovine serum (FBS, 10500064, Gibco) and were incubated at 37°C and 5% CO 2 .Cells were passaged at a confluency of 80%-90%.HeLa ATG9 wild-type (WT) and knockout (KO) cell lines were kindly provided by David Rubinsztein at the University of Cambridge, UK [29].Cell lines were regularly tested for mycoplasma contamination by Eurofins Genomics.

| Transient Plasmid and siRNA Transfection
HeLa cells were transfected using FuGene6 (E2691, Promega) at a confluency of 50%-70%, according to manufacturer's instructions, in the required cell culture vessel, and analysis was proceeded 24 h after transfection.During co-transfection with two individual plasmids encoding ER hook and mCherry-APP-EGFP, a ratio of 7:1 (hook:mCherry-APP-EGFP) was used.

| Generation of Stable Cell Lines Using Lentiviral Transduction
Cell lines with stable expression of the different RUSH mCherry-APP-EGFP plasmids described above were generated using lentiviral transduction.Lentivirus production was carried out using HEK293T packaging cells.The HEK293T cells were transfected with the desired lentiviral plasmids described above, pVSVG and pSPAX2 plasmids, using Lipofectamine 2000 transfection reagent (11668019, Thermo Fisher Scientific), according to the manufacturer's protocol.The cell culture medium was replaced with fresh medium 24 h after transfection.Forty-eight hours after transfection, the medium from the transfected HEK293T cells was collected, sterilized through a 0.45-μm filter, supplemented with polybrene (TR-1003-50U, Sigma), and added to the target HeLa cell line on the same day.Individual batches of virus were produced for the ER hook and the mCherry-APP-EGFP plasmids, mixed and applied with a ratio of 7:1 (hook:APP) to the target cell line.To ensure excess expression of the ER hook over mCherry-APP-EGFP, the infection procedure with the ER hook virus was repeated the next day.

| Flow Cytometry to Sort Stable Cell Lines for Comparable APP Expression Levels
To ensure comparable mCherry-APP-EGFP expression levels between HeLa ATG9 WT and KO cells, they were sorted for similar EGFP intensities using flow cytometry.The lentiviral transduced cell lines were collected in Dulbecco's phosphate-buffered saline (DPBS, 14190144, Gibco) supplemented with 0.5% FBS at a density of 5 × 10 6 cells/mL.Cells were then sorted for equal levels of EGFP fluorescence using a BD FACSAria Fusion instrument with similar gates for ATG9 WT and KO cells (Figure S3).Flow cytometry experiments were performed with the assistance of the Biomedicum Flow Cytometry Core facility at Karolinska Institutet.

| Live Cell Imaging
For live cell imaging, 1.5 × 10 5 cells were seeded into 35-mm No. 1.5 cover slip 10-mm glass bottom dishes (P35G-1.5-10-C,MatTek) and incubated overnight.On the day of the imaging experiment, cells were washed once with DPBS and then kept in imaging medium consisting of FluoroBrite DMEM (A1896701, Gibco) supplemented with 10% FBS, 25 mM HEPES (15630106, Gibco), and 1× GlutaMAX (35050061, Gibco).Biotin (B4501, Sigma) was diluted in imaging medium and pre-warmed to 37°C.Microscopy was conducted with a Nikon CrEST X-Light V3 Spinning Disk microscope using a 60× objective.During live cell imaging, cells were incubated at 37°C and 5% CO 2 on the microscope stage.The cells were imaged at a rate of one frame per minute for five frames prior to adding biotin.The transport of fluorescently labeled APP was induced by adding a final concentration of 40 μM biotin in imaging medium to the cells on the microscope stage and imaging was continued for 180 min.The contrast and brightness of microscopy images were adjusted in Fiji.All image analysis was performed using Fiji software on stacks of individual cells outlined by a region of interest (ROI) covering the whole cell.
Quantification of the mCherry-APP-EGFP transport to the Golgi was conducted by obtaining an ROI around the Golgi as follows.Live cell imaging stacks were converted into 8-bit images, and the maximum intensity Z projection of images between 10 and 80 min of biotin treatment was performed.This time span was selected because it contained most of the Golgi-localized mCherry-APP-EGFP.After applying a median filter with a radius of 3, images were automatically thresholded for the Golgi region and manually adjusted if the Golgi area was not well covered.ROIs around the Golgi area were obtained using the analyze particle function.Subsequently, the Golgi ROI was applied to the mCherry and EGFP channels and fluorescence intensities were measured over the whole time series stack.Results were exported by "Read and Write Excel" (by Anthony Sinadinos, GitHub) macro for Fiji and additional calculations were carried out in Microsoft Excel.
To analyze vesicular structures, a published RUSH vesicle analysis Fiji macro was modified [30].During the modification of the published macro, a "Yen" auto threshold was applied to all images of a time series stack, and all particles between 4 and 200 pixels in size were included in the analysis.

| Immunofluorescence Staining and Microscopy
Stable RUSH mCherry-APP-EGFP expressing cell lines (0.35 × 10 5 cells per well) were seeded onto 13-mm No. 1.5 glass cover slips (631-0150, VWR) in 24-well plates and treated with a final concentration of 40 μM biotin in DMEM supplemented with 10% FBS for indicated durations.Afterward, cells were fixed in 4% formaldehyde solution in PBS (252549, Sigma) for 20 min at room temperature, washed three times with PBS, and then permeabilized in 0.4% CHAPSO (220202, Calbiochem) in PBS for 10 min.Cells were washed twice and subsequently blocked in 3% normal goat serum (NGS, 50062Z, Thermo Scientific) for at least 10 min.Cells were incubated with primary antibodies diluted in 3% NGS in PBS for 30 min.After washing three times in PBS, the cover slips were submerged in secondary antibodies diluted in 3% NGS in PBS for 30 min.Cells were washed three times in PBS and nuclei were stained in Hoechst 33342 (H3570, Thermo Scientific) diluted 1:2000 in PBS for 10 min.After washing twice in PBS and once in ddH 2 O, cells were mounted onto microscopy slides using ProLong Gold Antifade mounting media (P36930, Thermo Scientific).Microscope slides were imaged using a Zeiss LSM980-Airy2 confocal microscope equipped with a 40× objective and super-resolution mode.The contrast and brightness of microscopy images were adjusted in Fiji.All image analysis was performed using Fiji software [31] on images with subtracted background (rolling ball radius 50 pixels) of individual cells outlined by an ROI covering the whole cell.
Co-occurrence analysis (Figure S5A) of fluorescence signals was conducted on images with a median filter of 2 to reduce noise.The automated threshold "Yen" (for LAMP1, EEA1) or "Default" (for GM130, PDI) was applied to all individual channels followed by conversion to binary images.To create a third image containing only overlapping pixels, the image calculator operator "AND" was applied to two single-channel binary images.The area of positive pixels was measured for the individual channel images and the new overlap image.The co-occurrence factor was calculated as the relative area of the overlapping pixels in the third overlap image related to the area occupied in the single-channel images.A total overlap of the two images would result in a co-occurrence of 1 and a complete exclusion of the fluorescence signals would be 0.
In the second approach, considering vesicular structures of a defined size (Figure S5B), the organelle marker image was subjected to "Yen" auto thresholding, converted to a binary image, and punctae of 0.01-1 μm 2 in size were outlined with ROIs using the "analyze particles" function.Subsequently, those ROIs were applied to the other channels of the image, and the fluorescence intensity in each ROI was measured.The "Read and Write Excel" (by Anthony Sinadinos, GitHub) macro for Fiji was used to export the results and additional calculations were performed in Microsoft Excel.The ROIs with a fluorescence intensity twice the mean fluorescence of the whole cell were counted as positive.
Golgi fragmentation was analyzed by auto thresholding the image with "MaxEntropy," and the resulting binary image was subjected to "analyze particles" function of structures larger than 0.2 μm 2 to count the number of Golgi fragments per cell.

| Cell Surface Immunostaining and Microscopy
Cells stably expressing RUSH mCherry-APP-EGFP were grown on glass cover slips in 24-well plates (0.35 × 10 5 cells per well) and treated with 40 μM biotin in DMEM containing 10% FBS for indicated durations.Afterward, cells were incubated with primary antibodies diluted in DPBS (14040117, Gibco) for 30 min on ice, followed by three washes with ice-cold DPBS.Cells were fixed in 1% formaldehyde solution in PBS for 5 min on ice followed by 15 min at room temperature and washed three times with PBS.Next, the cells were placed in secondary antibodies diluted in PBS for 20 min.After washing the cells twice in PBS and once in ddH 2 O, the cover slips were mounted onto microscopy slides using ProLong Gold Antifade mounting media.Microscope slides were imaged using a Zeiss LSM980-Airy2 confocal microscope equipped with a 40× objective and superresolution mode.The contrast and brightness of microscopy images were adjusted in Fiji.All image analysis was performed using Fiji software [31] on images with subtracted background (rolling ball radius 50 pixels) of individual cells outlined by an ROI covering the whole cell.The mean gray value in the individual channels was measured over the whole cell.

| Endocytosis Antibody Uptake Assay
To evaluate endocytosis, cells were incubated with an antibody targeting the extracellular domain of APP on ice and shifted to 37°C to permit endocytosis for 10 min, as described by Lonic, Onglao, and Khew-Goodall [32].Cells stably expressing RUSH mCherry-APP-EGFP were seeded on glass cover slips in 24-well plates (0.35 × 10 5 cells per well) and treated with 40 μM biotin in DMEM containing 10% FBS for 2 h.After washing in ice-cold DPBS, cells were incubated with anti-APP 6E10 antibody diluted in DMEM supplemented with 10% FBS for 30 min on ice, followed by one wash with ice-cold DPBS.Cells were incubated for 10 min at 37°C and then washed once in ice-cold DPBS.To remove antibodies that remained at the cell surface, cells were incubated for 5 min in a solution of 0.2 M acetic acid and 0.5 M NaCl on ice.After washing the cells four times with ice-cold DPBS, they were fixed in 4% formaldehyde solution in PBS for 5 min on ice followed by 15 min at room temperature and washed three times with PBS.The subsequent permeabilization and immunofluorescence staining were performed as described above, and the resulting images were analyzed as those obtained by cell surface immunostaining.

| Western Blotting
Cell culture medium and cell lysates were subjected to Western blotting to analyze the content of proteins of interest.Stable RUSH mCherry-APP-EGFP expressing cell lines were seeded into six-well plates (3.5 × 10 5 cells per well).On the next day, they were washed with DPBS, and then treated with a final concentration of 40 μM biotin in DMEM without FBS to release mCherry-APP-EGFP from the ER.After different time points, cell culture medium was collected and supplemented with protease (031M-C, GBiosciences) and phosphatase (P0044, Sigma) inhibitor cocktails, centrifuged at 1500 rpm at 4°C for 10 min, and the supernatant was transferred to a new Eppendorf tube.Cells were washed once with DPBS and lyzed in an adequate amount of RIPA buffer (89900, Thermo Scientific) supplemented with protease and phosphatase inhibitors.Cell lysates were incubated on ice for 10 min, sonicated for 5 s, and centrifuged at 14 000 rpm for 15 min at 4°C.Supernatants were transferred to new Eppendorf tubes, and Western blot samples were prepared by mixing similar protein concentrations with 4× Laemmli sample buffer (161-0747, Bio-Rad) supplemented with 2-mercaptoethanol (M6250, Sigma).Before loading to the SDS gels, samples were boiled at 95°C for 5 min.Samples for blotting ATG9 were not boiled but kept at room temperature for 10 min before loading onto the gel.
Samples were loaded to 4%-20% Mini-PROTEAN TGX Precast Protein Gels (4561096, 4561095, Bio-Rad), and after running, proteins were transferred to 0.2-μm nitrocellulose membranes (1704270, Bio-Rad) using a Trans-Blot Turbo system (Bio-Rad).As a loading control, total protein was visualized using the Revert 700 Total Protein Stain Kit (926-11021, LI-COR), according to the manufacturer's instructions.Afterward, the membrane was blocked in 5% skim milk in Tris-buffered saline with 0.05% Tween 20 (TBS-T) for 1 h at room temperature.Primary and secondary antibodies were diluted in 5% bovine serum albumin in TBS-T, and between incubations, the membranes were washed three times for 7 min in TBS-T.Membranes were imaged using a LI-COR Odyssey DLx system and the ImageStudio software.To detect another target protein, membranes were incubated for 15 min in Restore Plus Western Blot Stripping Buffer (46430, Thermo Scientific), and after washing three times in TBS-T, the above-described protocol was repeated from blocking onwards.Western blot band intensity was quantified using ImageStudio.

| Immunoprecipitation Using GFP-and RFP-Trap Agarose
In transient transfection experiments (Figure S2), individual cells expressing RUSH mCherry-APP-EGFP were detectable by fluorescence microscopy; however, in Western blot analysis, it was difficult to detect mCherry-APP-EGFP processing bands due to low transfection efficiency of mCherry-APP-EGFP.Therefore, GFP-trap (gtma, Chromotek) or RFP-trap (rta, Chromotek) was used to enrich for EGFP-or mCherry-tagged fragments, respectively.RFP-trap was performed with conditioned medium to detect mCherry-sAPP and GFP-trap with the cell lysates to detect CTFs-EGFP, as outlined in the manufacturer's protocol.

| Statistics
Statistical analysis was performed in GraphPad Prism and all values are shown as mean ± SEM, unless indicated otherwise.Data were tested for normal distribution by Shapiro-Wilk test and were subsequently analyzed for statistical differences using two-tailed unpaired t-test.If data points were not following a normal distribution, statistical differences were assessed by Mann-Whitney test.To compare more than two independent groups, statistical differences were assessed by one-way ANOVA or Kruskal-Wallis test.The respective statistical test and the number of independent experiments are stated in the figure legends.The considered significance levels were as follows: *p < 0.05; **p < 0.01; ***p < 0.001.

| ATG9-Deficient Cells and Pharmacological Modulation of Autophagy as Tools to Study the Role of Autophagy in APP Transport and Processing
ATG9 is a core autophagy protein that delivers phospholipids to the autophagic membrane and also plays a key role in intracellular trafficking.To investigate the role of autophagy and its key protein ATG9, in APP metabolism, we used two approaches.First, we took advantage of an already existing ATG9 KO HeLa cell line [29], and second, we pharmacologically modulated different stages of autophagy.
The absence of ATG9 protein in the KO cells was confirmed by immunostaining (Figure 1A) and Western blot (Figure 1C).Autophagy is impaired in ATG9 KO cells, as shown by a decrease in LC3 lipidation upon Bafilomycin A1 treatment, even though a slight LC3-II production remained, which may be the result of noncanonical LC3 lipidation that has been previously reported [33][34][35] (Figure 1C).Because we are interested in APP processing in AD, we used the main brain-expressed human isoform APP 695 in our study [36].To visualize APP transport and processing, we fused mCherry to the N-terminus and EGFP to the C-terminus of APP (mCherry-APP-EGFP).In all figures and descriptions, EGFP signal is derived from the C-terminus of full-length APP, as well as the APP processing fragments CTFα, CTFβ, and AICD.The mCherry tag visualizes the N-terminus of full-length APP as well as sAPPα and sAPPβ fragments.Full-length mCherry-APP-EGFP is characterized by an equal ratio of colocalizing EGFP and mCherry fluorescence, and after APP processing, the EGFP and mCherry signals will diverge.
Under steady-state conditions, a large fraction of EGFP and mCherry fluorescent signals overlaps, representing full-length mCherry-APP-EGFP (Figure 1A).Separated red and green punctae are visible, indicating cleaved mCherry-APP-EGFP fragments.In addition, red punctae might also result from quenching of the EGFP fluorescence due to the low pH in lysosomes.The localization of mCherry-APP-EGFP partially overlaps with ATG9A (Figure 1B), which is found in TGN and endosomes, and thereby shares its subcellular localization with APP [16].We further employed pharmacological modulation of autophagy to investigate whether the effects observed in the ATG9 KO cells could be extended to autophagy, thus enabling us to separate ATG9 and autophagy-induced effects (Figure 1D).Bafilomycin A1 blocks the lysosomal proton pump V-ATPase, thereby preventing the acidification of the lysosome and finally mitigating the autophagosome-lysosome fusion [34].In contrast to the mechanism of action of Bafilomycin A1, the ULK inhibitor MRT68921 attenuates autophagic activity in an early stage.ULK inhibitor treatment resulted in reduced LC3-II levels under Bafilomycin A1 treatment as compared with the control [35].Amino acid and serum starvation by replacing the nutrient-rich cell culture medium with HBSS is a common practice to stimulate autophagic flux, as shown by decreased LC3-II levels in the absence of Bafilomycin A1 (Figure 1D).

| Optimization of the RUSH System to Study the Transport and Processing of mCherry-APP-EGFP
Given the complexity and overlap of the intracellular APP transport routes, we sought to apply a system to synchronize its secretory transport.Therefore, we adopted the RUSH system [27] to mCherry-APP-EGFP as a cargo protein, as depicted in Figure 2A.RUSH makes use of an ER-resident hook protein fused to streptavidin, which binds to SBP fused to the Nterminus of mCherry-APP-EGFP.Biotin interferes with that interaction and thereby releases a wave of fluorescently labeled APP from the ER.mCherry-APP-EGFP was robustly expressed when using two individual plasmids containing the ER hook and mCherry-APP-EGFP as compared with the low expression levels when using a single plasmid encoding both mCherry-APP-EGFP and the ER hook (Figure S1A).Next, different ER hook proteins were evaluated for their ability to hook up mCherry-APP-EGFP in the ER.This revealed that the soluble streptavidin and KDEL amino acid motif containing ER and potentially saturate the transport machinery [37], interfering with the analysis of mCherry-APP-EGFP trafficking in ATG9 KO compared with WT cells.Therefore, we generated stable HeLa cell lines expressing the Str-KDEL ER hook and mCherry-APP-EGFP linked to SBP and sorted both ATG9 WT and KO cell lines for similar EGFP fluorescence intensity using flow cytometry (Figure S3A).
The newly synthesized full-length mCherry-APP-EGFP was retained in the ER, as confirmed by the colocalization of mCherry and EGFP with protein disulfide isomerase (PDI), an ER marker, at the 0-min time point, that is, before addition of biotin (Figure 2B).mCherry-APP-EGFP exiting the ER could be observed through a change in mCherry and EGFP intracellular localization at 60 and 120 min of biotin treatment, accompanied by a decreasing overlap with the ER marker PDI over time (Figure 2B,C).Full-length mCherry-APP-EGFP was similarly retained in the ER in ATG9 WT and KO cells and followed a similar decrease in mCherry-APP-EGFP overlap with the ER marker PDI over time, except at 60 min after biotin release (Figure 2C).Residual ER localization of mCherry-APP-EGFP at 120-min post-biotin treatment was unaffected by cycloheximide co-treatment starting 60 min before biotin addition to block the new synthesis of mCherry-APP-EGFP (Figure S4A).Therefore, it likely represents ER-residing mCherry-APP-EGFP independent of its interaction with the hook or insufficient mCherry-APP-EGFP release.However, also processed AICD-EGFP is localized to the cytosol and therefore might contribute to the overlap with the ER marker PDI.Due to the ER spanning through the entire cell, other subcellular localizations of mCherry-APP-EGFP, such as Golgi, will also contribute to a co-occurrence with PDI.
After leaving the ER, mCherry-APP-EGFP was transported to the Golgi apparatus, which is confirmed by the colocalization of mCherry and EGFP with the cis-Golgi marker GM130 at 60 and 120 min after addition of biotin (Figure 3A).In immunostaining experiments of fixed cells, mCherry and EGFP fluorescence signals showed the greatest colocalization with GM130 at 60 min after ER export in both ATG9 WT and KO cells (Figure 3B).ATG9 deficiency caused a reduced localization of mCherry-APP-EGFP to the cis-Golgi at 120 and 180 min after ER release as compared with WT (Figure 3B).In contrast, ULK inhibition increased the colocalization of EGFP and mCherry signals with the cis-Golgi marker GM130 at 120 min of biotin treatment.The occurrence of mCherry-tagged APP in the cis-Golgi was slightly increased by BafA1 and HBSS incubation (Figure 3D,E).In addition, a significant fragmentation of the Golgi apparatus was observed in ATG9 KO cells as well as in WT cells treated with BafA1 and starved in HBSS but not in WT cells treated with ULK inhibitor (Figure 3C,F).ATG9 KO cells not expressing the RUSH mCherry-APP-EGFP constructs showed a similar morphology in Golgi structure as ATG9 KO cells stably expressing the RUSH system (Figure S4B).
Taken together, co-staining of mCherry-APP-EGFP with ER and Golgi markers at different time points after biotin treatment showed that the RUSH system could be successfully implemented for mCherry-APP-EGFP, and the addition of biotin triggered the ER export and subsequent arrival of mCherry-APP-EGFP at the Golgi in ATG9 WT and KO cells.ATG9 deficiency and pharmacological inhibition of autophagy differentially affected the localization of mCherry-APP-EGFP to the cis-Golgi.

| Live Cell Imaging Reveals Altered RUSH mCherry-APP-EGFP Transport in ATG9 KO Cells
To get a more detailed insight into the dynamics of APP transport and how it is influenced by the loss of ATG9, we next performed live cell imaging with one frame per minute over a total duration of 3 h (Videos S1 and S2).Like in fixed cells, live cell imaging revealed that mCherry-APP-EGFP is retained in the reticular structures of the ER before being released by biotin.Biotin induced ER export and transport of mCherry-APP-EGFP to the Golgi in ATG9 WT and KO cells after 30-60 min (Figure 4A, Videos S1 and S2).The EGFP and mCherry signals of APP reached their peak fluorescence intensity in the Golgi region at slightly earlier time points in ATG9 KO compared with WT cells (Figure 4B,C).Those results could indicate a slightly faster mCherry-APP-EGFP ER-to-Golgi transport, a faster passage through Golgi or post-Golgi transport, as well as a combination of the aforementioned, in ATG9-deficient cells.
After Golgi export, mCherry-APP-EGFP reaches the endosomal system directly or by passing through the plasma membrane.Sixty minutes after releasing mCherry-APP-EGFP from the ER, intracellular vesicular structures appeared progressively, indicative of mCherry-APP-EGFP localization to the endosomal-lysosomal system (Figure 4A,D,E, Videos S1 and S2).Statistical analysis was performed on data obtained at 60-, 120-, and 180-min biotin treatment.The number of mCherry-and EGFP-positive structures was assessed over time by thresholding and counting the number of particles in Fiji.The number of mCherry punctae was higher in ATG9 KO cells compared with WT at all time points, and ATG9 KO cells contained more EGFP-positive vesicular structures at 60 min after  Similarly to ATG9 KO cells, autophagy inhibition by ULK inhibitor increased the number of EGFP-and mCherry-positive vesicles while reducing their size (Figure 5D,E).Blockage of acidification of lysosomes by BafA1 also significantly reduced the size of mCherry-and EGFP-loaded vesicular structures, whereas starvation by HBSS did not affect the number and size of RUSH mCherry-APP-EGFP punctae (Figure 5D,E).
In summary, fluorescence microscopy data from both fixed and live cell imaging showed consistent mCherry-APP-EGFP transport from ER to Golgi and a slight shift in the time to reach the maximum of mCherry-APP-EGFP localized to the Golgi in ATG9 KO cells.After Golgi export, mCherry-APP-EGFP appeared in vesicular structures, and autophagy inhibition by ATG9 KO or ULK inhibitor treatment caused an increase in the number of RUSH mCherry-APP-EGFP containing vesicular structures, yet reduced their size.

| Reduced Localization of mCherry-APP-EGFP to Early Endosomes in ATG9 KO Cells
Having found that ATG9 KO cells contained a greater number of mCherry and EGFP vesicular structures prompted us to investigate the identity of those vesicles in more detail.Therefore, we performed immunofluorescence staining for markers of the endosomal-lysosomal system known to take part in intracellular APP trafficking.APP contains a C-terminal YENPTY motif that promotes its clathrin-mediated endocytosis from the plasma membrane into early endosomes [38].
Thus, we first stained for early-endosome antigen 1 (EEA1) and measured the overlap of mCherry-APP-EGFP fragments with EEA1 in two different ways using self-made Fiji macros.First, images were thresholded and EEA1 and mCherry-APP-EGFP co-occurrence was analyzed over the whole cell by generating a third image only displaying pixels that were present in both thresholded single-channel images (Figure S5A).The cooccurrence factor was calculated as the relative area of the overlapping pixels in the third image related to the area occupied in the single-channel image (Figure 5F).In some cells, EEA1 clustered around the Golgi apparatus, which is heavily loaded with mCherry-and EGFP-tagged APP, especially after 60 min of ER export, and might result in false-positive early-endosome localization (Figure 5A).Therefore, in the second approach to avoid including clustered EEA1 vesicles around the Golgi, thresholded EEA1-positive spots with a size between 0.01 and 1 μm 2 were outlined with ROIs using the "analyze particles" function (Figure S5B).Subsequently, the ROIs of the EEA1 staining were applied to the mCherry and EGFP images, and the fluorescence intensities inside the ROIs were measured.All EEA1 ROIs with a fluorescence intensity twice the mean whole cell fluorescence intensity were considered positive and displayed as the percentage of mCherry-APP-EGFP-positive structures relative to the total number of EEA1 vesicles inside a cell (Figure 5G).Based on the co-occurrence analysis, the localization of EGFP and mCherry to EEA1-positive vesicles was significantly reduced in ATG9 KO cells at 60 and 120 min (Figure 5F).Consistently, EEA1 vesicles of 0.01-1 μm 2 size in the ATG9 KO cells contained significantly less mCherry-APP and APP-EGFP at 60 and 180 min (Figure 5G), whereas the number and size of EEA1 vesicles were slightly increased in ATG9 KO cells (Figure 5H).Moreover, most of the vesicles in ATG9 WT and KO cells are positive for both mCherry and EGFP fluorescence signals, indicative of the presence of full-length mCherry-APP-EGFP or both mCherry-and EGFP-APP cleavage fragments being present inside the same vesicle (Figure 5A).Together these data suggest that proportionally less early endosomes contain mCherry-APP-EGFP.

| Less mCherry-APP-EGFP Reaches the Lysosome in ATG9-Deficient Cells
Despite autophagic delivery of cargo to the lysosome through autophagosomes is defective in ATG9 KO cells, APP can enter the lysosome via different routes, including the endocytic pathway, in which the early endosome matures to a late endosome and finally fuses with a lysosome [39].To follow the intracellular route of APP and its divergence with the lysosomal system, we next examined the colocalization of mCherry-APP-EGFP and the lysosomal marker LAMP1 (Figure 6A).
Even though there was no difference between WT and ATG9 KO cells in the co-occurrence of EGFP and mCherry with LAMP1 analyzing the whole cell area (Figure 6B), the percentage of EGFP-as well as mCherry-positive LAMP1 punctae of 0.01-1 μm 2 in size was significantly decreased in ATG9 KO cells at 60 and 120 min of biotin treatment (Figure 6C).Interestingly, ATG9 KO cells contained an increased number of lysosomes per cell, which were smaller in size compared with WT cells (Figure 6D).Furthermore, the staining of LAMP1 appeared  B,E) Co-occurrence analysis of GM130 and EGFP (left) or mCherry (right) signal was performed using automated thresholding and the image calculator operator "AND" in Fiji to generate an image with only co-occurring pixels.(C,F) Analysis of the number of Golgi fragments per cell using automated thresholding and the "analyze particle" function in Fiji.Grubbs' test was performed to remove significant outliers (α = 0.05).(D) HeLa WT cells stably expressing the RUSH constructs Str-KDEL ER hook and mCherry-APP-EGFP were treated for 180 min with 500 nM Bafilomycin A1, 1 μM ULK inhibitor MRT68921, or starved in HBSS.During the last 120 min of incubation, 40 μM biotin was added to the cells.Statistical analysis comparing ATG9 WT and KO cells was performed using Mann-Whitney test on data from three independent experiments with n > 97 cells per condition.The statistical effects of pharmacological modulation of autophagy compared with the nontreated control were analyzed by Kruskal-Wallis test on data from three independent experiments with a total of n > 107 cells per condition (significance levels: *p < 0.05; **p < 0.01; ***p < 0.001).
overall weaker and more dispersed in ATG9 KO cells, indicative of reduced LAMP1 content of the lysosomes (Figure 6A).
Co-occurrence analysis of mCherry-and EGFP-tagged APP revealed a slight decrease in overlapping signals representing fulllength APP or cleaved APP in the same subcellular structure in ATG9 KO cells at 60 min after ER export (Figure 6E).At 180 min after biotin addition, the co-occurrence of mCherry and EGFP was instead slightly increased in the ATG9 KO cells, indicating a change in APP processing dynamics over time or differential contribution of EGFP quenching in the lysosome in ATG9 KO cells as compared with WT cells (Figure 6E).reduced localization of mCherry-APP-EGFP to LAMP1-positive lysosomes could result from a decreased proportion of EEA1positive early endosomes containing mCherry-APP-EGFP.

| ATG9 Deficiency Increases the Localization of mCherry-APP-EGFP to the Plasma Membrane
Having shown that inhibited autophagy increased the number while decreasing the size of mCherry-APP-EGFP containing vesicular structures and that less mCherry-APP-EGFP resides in early endosomes in ATG9 KO cells, we next asked whether the localization of mCherry-APP-EGFP at the plasma membrane was altered.We therefore applied a protocol to specifically stain cellsurface-localized mCherry-APP-EGFP by incubating the cells with an anti-APP 6E10 antibody on ice before fixation and by omitting permeabilization.Before the addition of biotin (0 min), when mCherry-APP-EGFP was retained in the ER, no surface localized mCherry-APP-EGFP was detected (Figure 7A).Upon biotin treatment, mCherry-APP-EGFP arrived at the plasma membrane (Figure 7A).Interestingly, ATG9-deficient cells entailed significantly increased levels of mCherry-APP-EGFP at the plasma membrane (Figure 7C).Notably, in all tested autophagy-modulating conditions, that is, autophagy inhibition by ULK inhibitor and BafA1, as well as autophagy activation by HBSS starvation, the levels of cell-surface-localized mCherry-APP-EGFP were increased (Figure 7B,D).This suggests that either more mCherry-APP-EGFP is transported to the plasma membrane potentially due to altered post-Golgi transport or that less mCherry-APP-EGFP is endocytosed as it has been shown that APP is rapidly internalized from the plasma membrane to lysosomes [9].To test this hypothesis, we performed an endocytosis antibody uptake assay by incubating cells on ice with the APP 6E10 antibody targeting the extracellular domain of APP.Afterward, cells were shifted to 37°C in a complete cell culture medium for 10 min to permit endocytosis of bound antibodies.Then the cells were stripped with acid to remove remaining plasma membrane-bound antibodies and intracellular APP was visualized.Endocytosed APP was only detectable when cells were shifted to 37°C for 10 min but not when fixing cells directly after incubation with the APP 6E10 antibody on ice (0 min, Figure 7E).While an increased amount of mCherry-APP-EGFP was localized to the plasma membrane in ATG9 KO cells, less mCherry-APP-EGFP was endocytosed in ATG9 KO cells (Figure 7F).Reduced mCherry-APP-EGFP endocytosis in ATG9 KO cells represents a direct link between increased plasma membrane localization and reduced occurrence of mCherry-APP-EGFP in the endolysosomal system.

| ATG9-Deficient Cells Generate Increased Levels of mCherry-sAPP and CTFs-EGFP
The altered transport of mCherry-APP-EGFP in ATG9 KO cells prompted us to investigate the processing of mCherry-APP-EGFP in more detail.We therefore performed Western blots to analyze the levels of full-length mCherry-APP-EGFP and its different cleavage fragments in ATG9 WT and KO RUSH mCherry-APP-EGFP samples at different time points after ER release with biotin.Non-transduced HeLa cells (empty) were included as a negative control to distinguish endogenous APP and nonspecific background.APP 22C11 N-terminal antibody detected multiple high molecular weight bands around 160 kDa, which were absent in non-transduced cells, and which most likely correspond to immature and mature mCherry-APP-EGFP, with lower and higher molecular weight, respectively (Figure 8A).Immature mCherry-APP-EGFP is synthesized into the ER and retained by the ER hook, corresponding to the lower migrating band being strongest at 0 min.APP acquires its full glycosylation pattern while passing through Golgi [40], as seen by the appearance of a higher molecular weight mCherry-APP-EGFP band peaking 60 min after biotin release, which is in line with the time of the peak localization of mCherry-APP-EGFP to the Golgi, as shown by the live cell imaging experiments (Figure 4A-C).After 60 min of biotin release, the intensity of the full-length mCherry-APP-EGFP bands progressively decreased, indicating that mCherry-APP-EGFP is being processed into smaller fragments and/or degraded (Figure 8A,C).At 180 min after ER release, the full-length mCherry-APP-EGFP protein levels in ATG9 KO cells were 1.44 times greater than in WT cells (Figure 8C).
Upon encountering α-secretase, full-length mCherry-APP-EGFP is cleaved into mCherry-sAPP (145 kDa) and CTF-EGFP (35 kDa), which could be detected 60 min after biotin treatment and increased progressively (Figure 8A,D,E).After 180-min treatment with biotin, the amount of CTF-EGFP in ATG9 KO cells was significantly higher compared with WT cells (Figure 8D).The identity of the CTFs-EGFP bands detected by the GFP antibody was further confirmed in GFP-trap  experiments and subsequent probing of the membrane with an antibody recognizing the C-terminus of APP (Figure S2D).
Upon α-secretase cleavage, the ectodomain sAPPα is either directly shed from the plasma membrane or released intraluminally and secreted into the extracellular space.Therefore, conditioned medium was probed after different times of biotin incubation with the APP 22C11 N-terminal antibody to detect mCherry-sAPP (Figure 8A).We could observe a 145-kDa mCherry-sAPPα band, and the identity was further confirmed by treatment with ADAM10 inhibitor, which drastically reduced mCherry-sAPPα production (Figure 8B).Notably, the amount of extracellular mCherry-sAPPα increased over time, and after 180 min, the conditioned medium of ATG9 KO cells contained two times the amount of mCherry-sAPPα as compared with WT cells (Figure 8E).Similarly, increased Altogether, our data show unambiguously an enhanced processing of mCherry-APP-EGFP, leading to increased secretion of mCherry-sAPPα and production of CTFs-EGFP in ATG9 KO cells, in ATG9A siRNA-mediated knockdown cells, and in ULK inhibitor-treated cells, implicating an inhibitory role of autophagy in APP metabolism.

| Discussion
The dysregulation of autophagy in AD has been established from studies of cell lines, mouse models, and patients.One of the most striking features is the accumulation of autophagic vacuoles in dystrophic neurites.Autophagosomes contain APP, its cleavage products Aβ and C-terminal fragments, as well as the secretase enzymes.Autophagy could hence contribute to Aβ pathology in addition to the known Aβ degradative function of the autophagosomal-lysosomal pathway.However, it remains to be investigated in detail how and at which stages autophagy is contributing to the transport and processing of APP under physiological conditions, and how autophagy impairment would affect APP trafficking.The intracellular transport of APP involves cycling between numerous organelles, while the site of APP processing is still under debate.Therefore, in this study, we took advantage of the RUSH system and optimized it for a mCherry-APP-EGFP construct to follow the APP transport in a systematic way, starting from the ER.Thus, it is possible to uncover mechanisms that might be masked under steady-state conditions.Here, the RUSH system could be successfully adopted for mCherry-APP-EGFP as a cargo protein and optimal ER retention was achieved using the Str-KDEL hook as confirmed by costaining with PDI.By addition of biotin, mCherry-APP-EGFP was released from the hook and exported from the ER, followed by the arrival of mCherry-APP-EGFP at the Golgi, and finally localized to the plasma membrane and endolysosomal system.The observed mCherry-APP-EGFP intracellular localization is in accordance with endogenous APP and similar fluorescently tagged APP variants used in previous studies [8,41].
Using the RUSH mCherry-APP-EGFP system in ATG9 KO cells and cells treated with autophagy modulators, we could identify the involvement of autophagy including ATG9 in different stages of APP intracellular transport (Figure 9).ATG9 deficiency shortened the time for mCherry-APP-EGFP to reach its peak in Golgi localization; subsequently, mCherry-APP-EGFP was redirected intracellularly to the plasma membrane, away from early endosomes and lysosomes.Even though ATG9 KO cells contained a greater number of mCherry-APP-EGFP-positive vesicular structures, its colocalization with early endosomes (EEA1) and lysosomes (LAMP1) was reduced, whereas more mCherry-APP-EGFP was localized to the plasma membrane.When the complex intracellular route of APP is disturbed, it encounters its cleavage enzymes differently, and indeed the amount of mCherry-sAPPα and CTFs-EGFP was increased in ATG9 KO cells, as confirmed by Western blot analysis.Notably, the sum of altered intracellular localization of mCherry-APP-EGFP to the Golgi, early endosomes, lysosomes, and plasma membrane in ATG9 KO cells compared with WT ultimately gives rise to a two-fold increase in secreted mCherry-sAPP fragments.
The first step in the intracellular trafficking route of mCherry-APP-EGFP after ER release is the transport to the Golgi apparatus.
The slightly earlier peak in Golgi localization of mCherry-APP-EGFP in ATG9 KO cells as seen by live cell imaging could result from a faster ER-to-Golgi transport or faster Golgi export of mCherry-APP-EGFP upon ATG9 deficiency.Because the ER and, in particular, ER exit sites, omegasomes and COPII vesicles, are all potential structures contributing to autophagosomes [42][43][44], it is plausible that APP can be directly routed for degradation via autophagy from the ER without passing the Golgi.This mechanism is most likely disrupted in ATG9 KO cells, giving rise to more mCherry-APP-EGFP arriving early at the Golgi.Because autophagy inhibition by the ULK inhibitor and Bafilomycin A1 both increased the colocalization of mCherry-APP-EGFP with the Golgi, the observed alteration in Golgi passage in ATG9 KO cells is specifically attributed to the loss of ATG9 rather than impairment of autophagy.We further observed an altered Golgi morphology represented by an increased number of Golgi fragments in ATG9 KO cells.Interestingly, Golgi fragmentation has been found in both AD mouse models and AD brains, triggered by Aβ and is associated with accelerated trafficking of APP [45].This is in agreement with the observed faster mCherry-APP-EGFP transport to the plasma membrane in ATG9 KO cells.On the other hand, ULK inhibition had no effect on Golgi morphology and subsequently increased the localization of mCherry-APP-EGFP to the Golgi as opposed to ATG9 deficiency, indicating a specific role of ATG9 in Golgi regulation.Bafilomycin A1 and starvation both increased the number of GM130-positive Golgi fragments per cell, which could be a result of stress-induced change in Golgi morphology [46,47].Both treatments also increased the colocalization of mCherry-APP-EGFP with the Golgi marker GM130 similar to ULK inhibition, for which the underlying mechanisms are yet to be investigated.
Interestingly, the alteration in cis-Golgi morphology in ATG9 KO cells is in contrast with the described role of ATG9 in preventing heat-induced Golgi fragmentation [48].Luo et al. identified an ATG9-GRASP55 Golgi fragmentation axis, which is distinct from the degradation of GM130 under heat stress.Notably, GRASP55 is a medial-trans-Golgi localized protein, whereas we analyzed the cis-Golgi protein GM130 to assess Golgi morphology, which could result in a different interpretation.Interestingly, GM130 has also been identified as a negative regulator of early autophagy [49].
Several previous studies highlight the importance of precise sorting of APP and its secretases for the post-Golgi transport [50][51][52].Heterotetrameric adapter protein 4 (AP-4) has not only been shown to mediate the export of ATG9A from the Golgi [53] but has also been implemented in sorting APP toward endosomes [54].In this delicate process, AP-4 could be a potential link connecting ATG9 and APP Golgi export.In the absence of ATG9A, the AP-4-mediated sorting of APP for its post-Golgi transport might be disturbed, resulting in reduced APP localization to early endosomes and lysosomes, as observed using the RUSH mCherry-APP-EGFP system.However, this hypothesis remains to be experimentally proven and additional sorting mechanisms in post-Golgi transport might be involved.
We propose that ATG9 deficiency caused an intracellular rerouting of mCherry-APP-EGFP where more mCherry-APP-EGFP reached the plasma membrane, facilitating its rapid cleavage by α-secretase resulting in the observed increased secreted mCherry-sAPPα levels.Indeed, inhibition of endocytosis sheds APP and leads to increased secreted sAPPα [55].Elevated levels of plasma membrane-localized APP could result from the observed reduced endocytosis of mCherry-APP-EGFP from the plasma membrane in ATG9 KO cells.Subsequently, less mCherry-APP-EGFP reaches early endosomes and lysosomes in ATG9 KO cells.As ATG9 itself can be localized to the plasma membrane and is internalized into endosomes and finally recruited to omegasomes upon autophagy initiation [56,57], it is possible that APP and ATG9 share similar endocytic carriers, and in the absence of ATG9, the internalization of APP is disturbed.
As described above, APP can either be routed directly from the Golgi to early endosomes [52] or pass through the plasma membrane where it is internalized via clathrin-mediated endocytosis to early endosomes [58].Notably, ATG9A does not only act in autophagy but also has additional roles in endosomal trafficking [57], and indeed we observed decreased endocytosis of mCherry-APP-EGFP in ATG9 KO cells.ATG9 deficiency subsequently reduced the localization of mCherry-APP-EGFP to early endosomes, which would result in reduced mCherry-APP-EGFP processing in a PI3P/ESCRT-dependent manner at the endosomal surface [12].However, we observed increased levels of mCherry-APP-EGFP at the plasma membrane, which resulted in increased processing, potentially due to encountering α-secretase at the plasma membrane.Early endosomes can mature to late endosomes and eventually fuse with lysosomes.As a transmembrane protein, APP is a target of lysosomal degradation and can reach the lysosome via autophagic or endocytic pathways [9,59,60].In ATG9 KO cells, Western blot revealed a reduced degradation of full-length mCherry-APP-EGFP over time and less mCherry-APP-EGFP was localized to peripheral LAMP1-positive lysosomes in immunostaining.Altogether, this indicates a reduced delivery of mCherry-APP-EGFP to the lysosome in the ATG9 KO cells.Besides routing APP to the lysosome via autophagosomes, ATG9 has been shown to facilitate the post-Golgi transport of hydrolytic enzymes to lysosomes, and the absence of ATG9 impaired lysosomal degradation [61].Furthermore, a blockage of lysosomal acidification by Bafilomycin A1 has been shown to increase cell surface expression of APP, stabilize full-length APP, and promote sAPP and Aβ secretion [62].This is concordant with our observation using the RUSH mCherry-APP-EGFP system under Bafilomycin A1 treatment.Therefore, the increased cell surface localization of mCherry-APP-EGFP, full-length APP levels, and secretion of mCherry-sAPP in ATG9 KO cells could potentially also be explained by dysfunctional lysosomes in ATG9 KO cells as phenocopied by Bafilomycin A1 treatment.
Even though we observed an increased number of vesicular structures in ATG9 KO cells compared with WT cells, there was a decreased localization of mCherry-APP-EGFP to early endosomes and lysosomes.Instead, the increased number of mCherry-APP-EGFP punctae might be attributed to an elevated number of post-Golgi vesicles on route to the plasma membrane, giving rise to elevated processing of mCherry-APP-EGFP at the plasma membrane, but these remain to be identified.Inhibition of autophagy with ULK inhibitor caused a similar phenotype of increased mCherry-APP-EGFP-positive vesicles of smaller size linking it to loss of autophagy.The reduced overlap of mCherry-APP-EGFP with early endosomes and lysosomes upon deletion of autophagy is in accordance with another study using tandem-fluorescently labeled APP showing starvation-induced autophagy activation increased the localization of APP to endosomes and lysosomes [63].
The pharmacological modulation of autophagy by ULK inhibition largely phenocopied the observed effects in ATG9 KO cells, which is in agreement with the fact that ULK recruits and phosphorylates ATG9, and hence both act in the early step of the autophagy pathway.On the other hand, Bafilomycin A1 does not block the formation of autophagosomes, but their clearance, and acts on lysosomes.Since Bafilomycin A1 increased cell surface localization of mCherry-APP-EGFP and mCherry-sAPP secretion similar to the effects observed in ATG9 KO cells, this effect is most likely ascribed to autolysosomal changes.Starvation with HBSS likewise slightly increased the plasma membrane localization of mCherry-APP-EGFP relative to intracellular mCherry and EGFP levels, and while more fulllength mCherry-APP-EGFP was present, reduced CTFs-EGFP were observed.The partially overlapping effects of autophagy induction and inhibition could be attributed to the complexity of targeted pathways by starvation as well as indirect effects on trafficking, activity, and expression levels of secretases.
Despite RUSH being a useful tool to study the transport and processing of mCherry-APP-EGFP in a synchronized way from the ER, it comes with limitations.Two fluorescent proteins on the N-and C-terminus are insufficient to visualize the multitude of possible cleavage fragments of APP, not only including products of the different secretases, but APP is also a known substrate of caspases.This is especially limiting in live cell imaging microscopy studies because additional staining of specific APP residues would be required.Therefore, the imaging results will not distinguish whether an overlap of EGFP and mCherry signal corresponds to full-length mCherry-APP-EGFP or cleaved mCherry and EGFP fragments in the same subcellular localization.Furthermore, mCherry fluorescence is resistant to low pH, whereas EGFP will be quenched, and hence structures with low pH might only display red fluorescence signals even though fulllength mCherry-APP-EGFP is present.The use of additional biochemical methods, including Western blot, yields more detailed information about the nature of the APP processing species.
In this study, we analyzed the intracellular transport and processing of WT APP 695 using the RUSH system and observed elevated levels of mCherry-APP-EGFP processing fragments, including mCherry-sAPP and CTFs-EGFP.WT APP expressed in HeLa cells is mainly subjected to the α-secretase pathway, as ADAM10 inhibition largely abolished mCherry-sAPP levels in the conditioned medium, whereas in neurons, β-secretase contributes to APP processing and generation of Aβ.Previous studies highlight that α-secretase processing of APP is taking place at the plasma membrane or the TGN [8,51].As we do not see a divergence of mCherry and EGFP signals of APP in the Golgi and very little APP processing fragments are present in Western blot when mCherry-APP-EGFP is mainly localized to the Golgi as determined from imaging experiments at 60 min after ER release, our data do not support a pronounced processing of RUSH mCherry-APP-EGFP in the Golgi.
It remains to be investigated whether the autophagy deficiencyinduced altered trafficking and processing of APP would also hold true in a disease setting, for example, when studying an APP containing AD-causing mutations in a neuronal model system.There is evidence that APP carrying familial AD-associated mutations behaves substantially different than WT APP under autophagy modulation [62,64].Autophagy is dysfunctional in AD and might contribute to altered APP intracellular transport and processing, and thereby potentially facilitating AD pathology.This study reinforces to exploit the modulation of the autophagic system to tune the trafficking and processing of APP, and thereby contribute to the development of new treatments to prevent AD.Our findings further support the importance of functional autophagy for physiological APP intracellular transport and processing.

| Conclusion
The secretory pathway and autophagy are interconnected at multiple stages and share several core proteins.The residing time of APP in a certain cellular compartment dictates its processing fate.A change in intracellular transport dynamics and localization of APP could alter its likelihood of being processed by secretases releasing different processing fragments.In our study, ATG9 deficiency or pharmacological inhibition of autophagy causing dysfunctional macroautophagy leads to an intracellular rerouting of mCherry-APP-EGFP and shifting the intracellular homeostasis away from degradation toward increased secretion of APP processing fragments.An altered APP intracellular transport route could affect its processing fate, potentially leading to downstream effects on its processed fragments and their functions.Altogether, this study helps to understand the role of autophagy in APP metabolism and will potentially contribute to the identification of new treatment strategies for AD.

FIGURE 1 |
FIGURE 1 | Inhibited autophagic flux in ATG9 KO HeLa cells.(A) Representative images of WT and ATG9 KO HeLa cells stably expressing mCherry-APP-EGFP and immunostained for ATG9A (magenta).(B) Intensity profile plot of the fluorescence intensities of EGFP (green), mCherry (red), and ATG9A (magenta) along the white line shown in (A).(C) Western blot analysis of autophagic flux in WT and ATG9-deficient HeLa cells by assessing LC3-II levels in cell extracts upon treatment with 100 nM Bafilomycin A1 (BafA1) for 24 h.(D) Pharmacological modulation of autophagic flux in HeLa WT cells by 1 μM ULK inhibitor MRT68921 (ULKinh) or starvation in HBSS with or without 500 nM BafA1 for 4 h.LC3 and total protein levels were detected by Western blotting.

FIGURE 2 |
FIGURE 2| Design of retention using selective hooks (RUSH) system to track mCherry-APP-EGFP transport and processing.(A) Schematic representation of the RUSH mCherry-APP-EGFP system.The mCherry-APP-EGFP fusion protein interacts with the ER hook containing a KDEL ER retention motif and streptavidin, thereby retaining mCherry-APP-EGFP in the ER.Upon addition of biotin, the fluorescent APP reporter is released from the ER hook and is further transported along the secretory pathway.After cleavage of mCherry-APP-EGFP by secretases, the EGFP and mCherry fluorescence signals diverge, thereby indicating the site of intracellular mCherry-APP-EGFP processing.(B) Representative images of ATG9 WT (left) and KO (right) HeLa cells stably expressing the ER hook and mCherry-APP-EGFP constructs depicted in (A).Cells were treated for 0, 60, 120 or 180 min with 40 μM biotin to release mCherry-APP-EGFP from the ER and were subsequently processed for immunofluorescence staining of PDI (magenta) as an ER marker confirming the ER retention of mCherry-APP-EGFP at 0 min.(C) Co-occurrence analysis of PDI and EGFP (left) or mCherry (right) signal was performed using automated thresholding and the image calculator operator "AND" in Fiji to generate an image with only co-occurring pixels.All statistical analysis was performed using Mann-Whitney test on data from three independent experiments with n > 106 cells per condition (significance level: *p < 0.05).
releasing mCherry-APP-EGFP from the ER.Similarly, the number of mCherry-and EGFP-positive punctae was increased at 60, 120, and 180 min in fixed ATG9 KO cells (Figure5A-C).In addition, fixed cells showed a decreased size of mCherry-and EGFP-positive vesicular structures in ATG9 KO cells, particularly at 120 min of biotin treatment (Figure5B,C).Differences in absolute numbers of vesicles per μm 2 as well as vesicle size in live cell imaging compared with fixed cells likely result from fragments/cell sample preparation, fixation, and permeabilization as well as image acquisition with different confocal microscopes and custom-made image analysis pipelines.

FIGURE 3 |
FIGURE3 | Transport of RUSH mCherry-APP-EGFP to Golgi after release from ER. (A) ATG9 WT (left) and KO (right) HeLa cells stably expressing the RUSH constructs Str-KDEL ER hook and mCherry-APP-EGFP reporter were fixed and immunostained for the cis-Golgi marker GM130 (magenta) after 40 μM biotin treatment for 0, 60, 120 or 180 min.(B,E) Co-occurrence analysis of GM130 and EGFP (left) or mCherry (right) signal was performed using automated thresholding and the image calculator operator "AND" in Fiji to generate an image with only co-occurring pixels.(C,F) Analysis of the number of Golgi fragments per cell using automated thresholding and the "analyze particle" function in Fiji.Grubbs' test was performed to remove significant outliers (α = 0.05).(D) HeLa WT cells stably expressing the RUSH constructs Str-KDEL ER hook and mCherry-APP-EGFP were treated for 180 min with 500 nM Bafilomycin A1, 1 μM ULK inhibitor MRT68921, or starved in HBSS.During the last 120 min of incubation, 40 μM biotin was added to the cells.Statistical analysis comparing ATG9 WT and KO cells was performed using Mann-Whitney test on data from three independent experiments with n > 97 cells per condition.The statistical effects of pharmacological modulation of autophagy compared with the nontreated control were analyzed by Kruskal-Wallis test on data from three independent experiments with a total of n > 107 cells per condition (significance levels: *p < 0.05; **p < 0.01; ***p < 0.001).

FIGURE 4 |
FIGURE 4 | Live cell imaging of RUSH mCherry-APP-EGFP comparing ATG9 WT and KO cells.(A) Live cell imaging of ATG9 WT (left) and KO (right) cells stably expressing the RUSH constructs Str-KDEL ER hook and mCherry-APP-EGFP reporter.The cells were maintained in a humidified microscopy chamber at 37°C and 5% CO 2 and treated with 40 μM biotin on the stage.Imaging was performed with a speed of one frame per minute over a total period of 180 min.(B,C) The left graph shows the quantification of the fluorescence intensity of EGFP (B) or mCherry (C) in the Golgi area over time adjusted to the initial fluorescence intensity at time point 0 min.The connecting line represents the mean fluorescence intensity ± SEM.The graph to the right indicates the resulting time required to reach the peak normalized fluorescence intensity in the Golgi area.(D,E) Quantification of the number of EGFP-(D) or mCherry-containing (E) vesicular structures between 45 and 180 min.Both graphs display the number of vesicular particles per cell area.The connecting line in the left graph represents the mean ± SEM.Statistical analysis was carried out on data obtained from 60-, 120-, and 180-min time points as shown in the graph on the right.Statistical analysis was performed using Mann-Whitney test on data from three independent experiments with n > 215 cells per genotype (significance levels: *p < 0.05; **p < 0.01; ***p < 0.001).

FIGURE 5 |
FIGURE 5 | Legend on next page.

FIGURE 5 |
FIGURE 5 | Less RUSH mCherry-APP-EGFP localizes to early endosomes upon deletion of ATG9.(A) ATG9 WT (left) and KO (right) HeLa cells stably expressing the Str-KDEL ER hook and mCherry-APP-EGFP were fixed after 60, 120, and 180 min of treatment with 40 μM biotin and immunostained for the early-endosome marker EEA1 (magenta).(B,C) Confocal microscopy images as shown in (A) were analyzed for vesicular structures containing EGFP (B) or mCherry (C) fluorescence signal.The left graph displays the number of vesicular particles per μm 2 , and the size of these particles is depicted in the right graph at 60, 120, and 180 min after releasing APP from the ER.(D,E) HeLa WT cells stably expressing Str-KDEL ER hook and mCherry-APP-EGFP were treated for 180 min with 500 nM Bafilomycin A1, 1 μM ULK inhibitor MRT68921, or starved in HBSS.During the last 120 min of incubation, 40 μM biotin was added to the cells.Vesicular structures positive for EGFP (D) or mCherry (E) were analyzed as described in (B) and (C).(F)Co-occurrence analysis of EEA1 and EGFP (left) or mCherry (right) signal in ATG9 WT and KO cells shown in (A) was performed over the whole cell using automated thresholding and the image calculator operator "AND" in Fiji to generate an image with only co-occurring pixels.(G) Colocalization analysis of EEA1-positive structures between 0.01 and 1 μm 2 in size was performed using the "analyze particles" function in Fiji.(H) Vesicular structures positive for EEA1 were analyzed as described in (B) and (C).Statistical analysis comparing ATG9 WT and KO cells was performed using Mann-Whitney test on data from three independent experiments with a total of n > 132 cells per condition.The statistical effects of pharmacological modulation of autophagy compared with a nontreated control were analyzed by Kruskal-Wallis test on data from three independent experiments with n > 107 cells per condition (significance levels: *p < 0.05; **p < 0.01; ***p < 0.001).

FIGURE 6 |
FIGURE 6 | Reduced localization of RUSH mCherry-APP-EGFP to LAMP1-positive vesicles.(A) ATG9 WT (left) and KO (right) HeLa cells stably expressing the Str-KDEL ER hook and mCherry-APP-EGFP were fixed after 60, 120, and 180 min of treatment with 40 μM biotin and were immunostained for the lysosome marker LAMP1 (magenta).(B) Co-occurrence analysis of LAMP1 and EGFP (left) or mCherry (right) signal was performed over the whole cell using automated thresholding and the image calculator operator "AND" in Fiji to generate an image with only cooccurring pixels.(C) Colocalization analysis of mCherry-APP-EGFP and LAMP1-positive structures between 0.01 and 1 μm 2 in size was performed using the "analyze particles" function in Fiji.(D) Confocal microscopy images as shown in (A) were analyzed for LAMP1 vesicular structures.The left graph displays the number of LAMP1 vesicular particles per μm 2 , and the size of those particles is depicted in the right graph at 60, 120, and 180 min after releasing mCherry-APP-EGFP from the ER.(E) Co-occurrence analysis of mCherry and EGFP signal was conducted as described in (B).Statistical analysis was carried out using Mann-Whitney test on data from three independent experiments with n > 130 cells per condition (significance levels: *p < 0.05; **p < 0.01; ***p < 0.001).

FIGURE 7 |
FIGURE 7 | Legend on next page.

FIGURE 7 |
FIGURE7 | ATG9 deficiency increases the localization of RUSH mCherry-APP-EGFP to the plasma membrane while endocytosis is reduced.(A) ATG9 WT (left) and KO (right) HeLa cells stably expressing the Str-KDEL ER hook and mCherry-APP-EGFP were treated for 0 or 120 min with 40 μM biotin and subjected to cell surface immunostaining.Plasma membrane-localized "surface APP" (magenta) was visualized using the anti-APP 6E10 antibody.(B) HeLa WT cells stably expressing Str-KDEL ER hook and mCherry-APP-EGFP were treated for 180 min with 500 nM Bafilomycin A1, 1 μM ULK inhibitor MRT68921, or starved in HBSS.During the last 120 min of incubation, 40 μM biotin was added to the cells.The cells were processed for immunostaining of cell-surface-localized APP as described in (A).(C,D) Fluorescence intensity analysis of plasma membrane-localized APP normalized to the mCherry and EGFP content of the same cell comparing HeLa ATG9 WT and KO cells (C) or HeLa WT cells treated with autophagy-modulating compounds (D).(E) ATG9 WT (left) and KO (right) HeLa cells stably expressing the Str-KDEL ER hook and mCherry-APP-EGFP were treated for 120 min, with 40 μM biotin and an endocytosis antibody uptake assay was performed for 10 min at 37°C using the anti-APP 6E10 antibody (magenta).(F) Fluorescence intensity analysis of endocytosed APP normalized to the mCherry and EGFP intensity of the same cell.Statistical analysis comparing ATG9 WT and KO cells was performed using Mann-Whitney test on data from three independent experiments with a total of n > 100 cells per condition.The statistical effects of pharmacological modulation of autophagy compared with an untreated control were analyzed by Kruskal-Wallis test on data from three independent experiments with n > 84 cells per condition (significance levels: *p < 0.05; **p < 0.01; ***p < 0.001).

FIGURE 8 |
FIGURE 8| Increased processing of RUSH mCherry-APP-EGFP in ATG9 KO cells and upon autophagy inhibition.(A) Representative Western blots of lysates from ATG9 WT and KO HeLa cells stably expressing the Str-KDEL ER hook and mCherry-APP-EGFP treated with 40 μM biotin for different durations.The negative control (empty) consists of WT cells that do not express the RUSH mCherry-APP-EGFP construct to visualize endogenous APP or unspecific bands on the Western blots.Twenty micrograms of cell lysate was analyzed for full-length mCherry-APP-EGFP isoforms using the APP N-terminal 22C11 antibody and CTFs-EGFP using a GFP-detecting antibody.The loaded volume of the cell culture supernatant was adjusted based on the total protein concentration of the corresponding cell lysate.(B) Cells were pretreated for 60 min with 10 μM α-secretase ADAM10 inhibitor GI254023X, followed by an additional 180-min treatment with 40 μM biotin to release mCherry-APP-EGFP from the ER.The loading of cell culture medium was adjusted as described in (A).(C,D) Quantification of full-length mCherry-APP-EGFP (C) and CTFs-EGFP (D) normalized to total protein load and adjusted to the full-length mCherry-APP-EGFP levels at the 0-min time point of the corresponding genotype.(E) Quantification of mCherry-sAPP in the cell culture medium adjusted to the corresponding intracellular full-length mCherry-APP-EGFP levels at the 0-min time point.(F) Representative Western blots of lysates from HeLa WT cells stably expressing the Str-KDEL ER hook and mCherry-APP-EGFP transfected with Ctrl siRNA or ATG9A siRNA for 3 days followed by incubation with 40 μM biotin for 180 min.The knockdown was confirmed with an antibody targeting ATG9A and APP processing fragments were detected as described in (A).(G,H) Expression levels of fulllength mCherry-APP-EGFP (G) and CTFs-EGFP (H) were normalized to total protein load and adjusted to the mean of the corresponding experiment to account for batch effects.(I) Ratio of mCherry-APP-EGFP and CTFs-EGFP expression levels.(J) Quantification of mCherry-sAPP in the cell culture medium adjusted to the mean of the experiment batch.(K) Representative Western blots of lysates from HeLa WT cells stably expressing the Str-KDEL ER hook and mCherry-APP-EGFP treated for 240 min with 500 nM Bafilomycin A1, 1 μM ULK inhibitor MRT68921, or starved in HBSS.During the last 180 min of incubation, 40 μM biotin was added to the cells.(L-O) Quantification of expression levels as described in (G)-(J).Data points are shown as mean ± SEM.Statistical analysis was conducted by two-tailed unpaired t-test comparing ATG9 WT and KO in each time point or comparing Ctrl siRNA treated cells with ATG9A siRNA treated cells.Statistical effects for pharmacological modulation of autophagy compared with the control were assessed by one-way ANOVA (n = 3 independent experiments, significance level: *p < 0.05; **p < 0.01; ***p < 0.001).

FIGURE 9 |
FIGURE9 | Model of shifted mCherry-APP-EGFP intracellular transport and increased processing upon deletion of ATG9 or ULK inhibition.As a transmembrane protein, mCherry-APP-EGFP is transported along the secretory pathway, from ER to Golgi to the plasma membrane, where it is rapidly internalized into endosomes and finally degraded in lysosomes.Under physiological conditions (upper panel), autophagy plays an important part in homeostatic turnover of mCherry-APP-EGFP and in the delivery to the lysosome for degradation.During autophagy impairment in ATG9 KO cells or by ULK inhibition (lower panel), the turnover of mCherry-APP-EGFP via autophagy is impaired.This results in mCherry-APP-EGFP rerouting toward the plasma membrane, resulting in increased APP processing to mCherry-sAPP and CTFs-EGFP.Increased localization of mCherry-APP-EGFP at the plasma membrane could be a direct consequence of the impaired mCherry-APP-EGFP endocytosis in ATG9 KO cells, and subsequently, proportionally less mCherry-APP-EGFP was observed in early endosomes and lysosomes.