SNX18 regulates ATG9A trafficking from recycling endosomes by recruiting Dynamin-2

Abstract

Trafficking of mammalian ATG9A between the Golgi apparatus, endosomes and peripheral ATG9A compartments is important for autophagosome biogenesis. Here, we show that the membrane remodelling protein SNX18, previously identified as a positive regulator of autophagy, regulates ATG9A trafficking from recycling endosomes. ATG9A is recruited to SNX18-induced tubules generated from recycling endosomes and accumulates in juxtanuclear recycling endosomes in cells lacking SNX18. Binding of SNX18 to Dynamin-2 is important for ATG9A trafficking from recycling endosomes and for formation of ATG16L1- and WIPI2-positive autophagosome precursor membranes. We propose a model where upon autophagy induction, SNX18 recruits Dynamin-2 to induce budding of ATG9A and ATG16L1 containing membranes from recycling endosomes that traffic to sites of autophagosome formation.

Keywords: ATG9; SNX18; autophagy; dynamin; recycling endosome.

Figure EV1. SNX18 regulation of autophagic proteins

1. HEK293A cells were fixed and immunostained with antibodies against endogenous SNX18 and ATG9A before confocal microscope analysis. Scale bar = 10 μm.

2. HEK293A control or SNX18 KO cells were transfected with GFP‐LC3B. Seventeen hours post‐transfection, the cells were starved or not for 2 h in EBSS ± 100 nM BafA1, before fixation and analysis by high‐throughput imaging (Zeiss AxioObserver, 20× objective). Scale bar = 10 μm.

3. The number of GFP‐LC3 spots observed in (A) was quantified from 40 fields of view (minimum 2,000 cells) per condition and quantitation of punctate objects using CellProfiler software (mean ± SEM from n = 3 independent experiments). Significance was determined by two‐way ANOVA and Bonferroni post‐tests where *P < 0.05.

4. HEK293A control or SNX18 KO cells were starved or not in EBSS for 4 h ± 100 nM BafA1, before cell lysis and immunoblotting of p62. Actin was used as loading control.

5. The p62 levels observed in (C) were quantified and normalised to fed within each cell line. The graph shows (mean ± SEM, n = 3), analysis by two‐way ANOVA and Bonferroni post‐test determined no significance between cell lines.

6. The level of mitophagy was determined by stable expression of a mitochondrial localised mCherry‐GFP tag. Mitophagy was induced by treatment of cells with 1 mM DFP for 24 h prior to fixation and high‐throughput analysis with a Zeiss AxioObserver widefield microscope (20×) to monitor for the formation of red only puncta. The number of red only puncta was determined by CellProfiler software from 30 fields of view and normalised to control cells with no treatment from n = 2 experiments. Each point represents a single replicate from a minimum of 1,000 cells per treatment.

7. The levels of ATG9 observed in Fig 1G were quantified relative to actin and normalised to fed control cells (mean ± SEM, n = 3). Analysis by two‐way ANOVA and Bonferroni post‐test determined no significance between cell lines.

8. Gene expression of SQSTM1, ATG9A, ATG16L1 and SNX9 was quantified by qPCR in HEK293A control or SNX18 KO cells. The graph shows the mean relative gene expression normalised to control cells from three independent experiments (mean ± SEM, n = 3). Analysis by two‐way ANOVA and Bonferroni post‐test determined no significance difference of targets between cell lines.

9. The levels of TfR observed in Fig 1G were quantified relative to actin and normalised to control fed cells (mean ± SEM, n = 3). Significance was determined by two‐way ANOVA and Bonferroni post‐tests where *P < 0.05.

10. HEK293A control or SNX18 KO cells were transfected with control siRNA or siRNA targeting ULK1 for 72 h, and cells were then starved or not for 2 h in EBSS ± 100 nM BafA1 before cell lysis and Western blot analysis. Actin was used as a loading control.

11. LC3 lipidation (LC3‐II) from (I) was quantified, and the graph shows the average level of LC3‐II relative to actin and normalised to Ctrl fed (mean ± SEM), n = 5. Significance was determined by two‐way ANOVA and Bonferroni post‐test where ***P < 0.001.

Figure 1. SNX18 regulates ATG9A trafficking from recycling endosomes

1. HEK293A cells were transfected with myc‐SNX18 for 17 h, then fixed and immunostained against myc and ATG9A before analysis by confocal microscopy. Scale bar = 10 μm.

2. HEK293A cells were starved or not for 2 h in EBSS before fixation and immunostaining against ATG9A and transferrin receptor (TfR). The cells were analysed by confocal microscopy. Arrowheads mark ATG9A‐ and TfR‐positive structures. Scale bar = 10 μm.

3. The colocalisation of ATG9A and TfR from (B) was quantified from > 100 cells per condition with Zen software (Zeiss) and normalised to fed state (mean ± SEM, n = 6). *P < 0.05, by Student's t‐test.

4. HEK293A Ctrl or SNX18 KO cells were starved for 2 h in EBSS, fixed and immunostained against ATG9A and TfR, before analysis by confocal microscopy. Scale bar = 10 μm.

5. The colocalisation of ATG9A and TfR observed in (D) was quantified from more than 100 cells per cell line with Zen software (Zeiss) and normalised to control cells (mean ± SEM, n = 7). *P < 0.05, by Student's t‐test.

6. Cells were treated as in (D) and the juxtanuclear localisation of TfR quantified in WT and SNX18 KO cells from > 100 cells per cell line (mean ± SEM, n = 6). ***P < 0.001, by Student's t‐test.

7. HEK293A SNX18 Ctrl or KO cells were starved or not in EBSS for 4 h ± 100 nM Bafilomycin A1 (BafA1) before cell lysis and Western blot analysis.

8. LC3 lipidation from (G) was quantified, and the graph shows the average level of LC3‐II relative to actin normalised to Ctrl fed (mean ± SEM, n = 6). *P < 0.05 as determined by two‐way ANOVA and Bonferroni post‐test.

9. Long‐lived protein degradation was measured in HEK293A SNX18 Ctrl or KO cells as the release of ¹⁴C‐valine after 4 h of starvation ± 3‐methyladenine (3MA). The starvation‐induced autophagic degradation is quantified as the difference in proteolysis in starved cells ± 3MA and normalised to the degradation of control cells (mean ± SEM, n = 3). *P < 0.05, by Student's t‐test.

Figure 2. SNX18 is required for recruitment of ATG16L1 to WIPI2 structures

1. HEK293A Ctrl or SNX18 KO cells were starved or not in EBSS for 2 h before fixation and immunostaining with antibodies against endogenous WIPI2 and ATG16L1. The cells were analysed by confocal microscopy. Arrowheads mark WIPI2‐ and ATG16L1‐positive structures. Scale bar = 10 μm.

2. Cells were treated as in (A), and colocalisation between WIPI2 and ATG16L1 was quantified with Zen software (Zeiss) using 10 images per condition, including > 100 cells from three independent experiments. The graph shows the average colocalisation of WIPI2 and ATG16L1 (mean ± SEM, n = 3). Statistical significance was determined by one‐way ANOVA and Bonferroni's multiple comparison test where *P < 0.05.

3. The number of ATG16L1 spots observed in (A) was quantified using CellProfiler software, and the graph shows the number of ATG16L1 spots per cell (mean ± SEM, n = 3). Significance was determined by one‐way ANOVA and Bonferroni's multiple comparison test where **P < 0.01, *P < 0.05.

4. The number of WIPI2 spots observed in (A) was quantified as in (C).

5. HEK293A SNX18 KO cells were fixed and immunostained with antibodies against ATG16L1, ATG9A and TfR. Images were obtained by confocal microscopy. Scale bar = 10 μm.

Figure EV2. The effect of SNX18 and TBC1D14 on recycling endosome function

1. HEK293A cells were transfected with GFP, GFP‐SNX18 or GFP‐TBC1D14. The cells were incubated with Alexa‐647 transferrin for 15 min in complete medium, followed by a chase for the indicated periods. At the end of each time point, the cells were trypsinised, fixed and analysed by flow cytometry. The line graph shows the percentage of retained Alexa‐647 transferrin in GFP‐positive cells, normalised to t = 0 (mean ± SEM, n = 3). *P < 0.05 by Student's t‐test.

2. HEK293A cells were transfected with control, SNX18 or RAB11 siRNA for 3 days, then incubated with Alexa‐647 transferrin, treated and analysed as in (A). Knockdown efficiency was verified by immunoblotting using the indicated antibodies.

3. HEK293A cells were transfected with GFP‐TBC1D14, fixed and immunostained with an antibody against endogenous SNX18, followed by confocal microscope analysis. Scale bar = 10 μm.

Figure 3. SNX18 recruits Dynamin‐2 to facilitate ATG9A traffic from recycling endosomes

1. The domain structure of SNX18 consists of an N‐terminal SH3 domain followed by an unstructured LC region and a C‐terminal PX‐BAR domain responsible for membrane binding and tubulation. The SH3 domain binds the GTPase Dynamin‐2, while the LC region binds to AP‐1 and LC3/GABARAP.

2. HEK293A cells transfected with mCherry‐SNX18 wild type (WT) were fixed and immunostained against Dynamin 17 h post‐transfection. Images were obtained by confocal microscopy. Scale bar = 10 μm.

3. HEK293A cells were transfected with myc control, myc‐tagged SNX18 WT or mutant SNX18 W38K before cell lysis. The lysates were incubated with magnetic anti‐myc microbeads before immunoblotting of the cell lysate (input) and the immunoprecipitate (myc‐IP) with antibodies against Dynamin‐2, ATG16L1 and myc.

4. HEK293A cells transfected with mCherry‐SNX18 W38K were fixed and immunostained against myc and Dynamin 17 h post‐transfection and imaged by confocal microscopy. Scale bar = 10 μm.

5. The colocalisation of Dynamin and mCherry‐SNX18 WT or W38K observed in (B) and (D) was quantified with Zen software (Zeiss) using 10 images per condition. The graph shows fold change colocalisation of Dynamin and mCherry‐SNX18, normalised to cells expressing mCherry‐SNX18 WT (mean ± SEM, n = 3). **P < 0.01 by Student's t‐test.

6. HEK293A SNX18 KO cells were cotransfected with YFP1‐SNX18 WT or W38K and YFP2‐Dynamin‐2 or vice versa. Seventeen hours post‐transfection, the cells were fixed and immunostained against TfR or ATG16L1 before analysis by confocal microscopy. The number of YFP spots was counted using CellProfiler software from 25 images per condition and normalised to SNX18 WT control (mean ± SEM, n = 3). Significance was determined by Student's t‐test where *P < 0.05 and ***P < 0.001.

7. The colocalisation of YFP puncta and TfR observed in the cells from (F) was analysed by confocal microscopy. Arrowheads mark YFP‐ and TfR‐positive structures. Scale bar = 10 μm.

8. The colocalisation of YFP puncta and ATG16L1 observed in the cells from (F) was analysed by confocal microscopy. Arrowheads mark YFP‐ and ATG16L1‐positive structures. Scale bar = 10 μm.

Figure EV3. SNX9 knockdown does not exacerbate SNX18 KO phenotype

1. The number of Dynamin puncta observed in Fig 3B and D was quantified using CellProfiler software. The graphs show average object number, object area and intensity plotted per field of view (11 per condition) from n = 1 (bar represents mean value).

2. HEK293A control or SNX18 KO cells were transfected with control siRNA or siRNA targeting SNX9 for 72 h, and cells were then starved or not for 2 h in EBSS ± 100 nM BafA1 before cell lysis and immunoblotting using the indicated antibodies. Actin was used as a loading control.

3. LC3‐II levels were quantified from each condition relative to actin and normalised to fed control cells (mean ± SEM, n = 3). Analysis by two‐way ANOVA and Bonferroni post‐test determined no significance difference between cell lines or treatment with siRNA.

4. HEK293A control or SNX18 KO cells were transfected with either control siRNA or SNX9 siRNA for 72 h before starvation in EBSS for 2 h followed by fixation and immunostaining against endogenous ATG16L1 before analysis with high‐throughput imaging (Zeiss AxioObserver, 20× objective). Scale bar = 10 μm.

5. The number of ATG16L1 spots observed in (D) was quantified using CellProfiler software from 25 fields of view per condition and normalised to Ctrl fed siScr (mean ± SEM, n = 3). Significance was determined between cell lines by two‐way ANOVA and Bonferroni post‐test where *P < 0.05. Comparison of siScr starved to siSNX9 starved by one‐way ANOVA and Bonferroni post‐test was not significant (ns) in either Ctrl or SNX18 KO cells.

6. HEK293A cells were cotransfected with myc control, myc‐SNX18 WT or myc‐SNX18 W38K together with flag‐ATG16L1 before cell lysis. The lysates were incubated with magnetic anti‐flag microbeads before immunoblotting of the cell lysate (input) and the immunoprecipitate (flag‐IP) with antibodies against flag and myc.

Figure 4. SNX18‐mediated recruitment of Dynamin‐2 is important for its function in autophagy

1. HEK293A SNX18 KO cells transfected with myc control, myc‐SNX18 WT or myc‐SNX18 W38K were starved in EBSS for 2 h before fixation and immunostaining with antibodies against myc, ATG9A and TfR and analysed with confocal microscopy. The images show representative observations from SNX18 KO cells. Scale bar = 10 μm.

2. The colocalisation of ATG9 and TfR in the SNX18 KO cells from (A) was quantified with Zen software (Zeiss), from > 10 images per condition. The graph shows fold change colocalisation normalised to SNX18 KO cells transfected with myc control (mean ± SEM, n = 3). Significance was determined by one‐way ANOVA and Bonferroni's multiple comparison test where **P < 0.01, ***P < 0.001.

3. Immunoblotting of SNX18 in cell lysates from HEK293A control, SNX18 KO and KO cells with stable expression of mCherry‐SNX18 WT or ‐W38K mutant. Immunoblotting of actin serves as a loading control. Arrowheads indicate endogenous SNX18 (in WT cells) and mCherry‐SNX18 WT and W38K.

4. Long‐lived protein degradation was measured as release of ¹⁴C‐valine after 4 h of starvation ± 3‐methyladenine (3MA) in the cells shown in (C). The autophagic flux is quantified as the difference in proteolysis in starved cells ± 3MA and normalised to the flux in control cells (mean ± SEM, n = 3). Significance was determined by one‐way ANOVA and Bonferroni's multiple comparison test where *P < 0.05 and **P < 0.01.

5. The cells shown in (C) were starved or not for 2 h with EBSS before fixation and immunostaining with antibodies against endogenous WIPI2 and ATG16L1. The cells were then analysed by confocal microscopy. The arrowheads mark WIPI2‐ and ATG16L1‐positive structures. Scale bar = 10 μm.

6. Quantification of the number of ATG16L1 puncta from images represented in (E). Punctate structures were quantified using ImageJ and represented as the mean ATG16L1 puncta per cell ± SEM from n = 3 independent experiments. Significance was determined by one‐way ANOVA and Bonferroni's multiple comparison test where **P < 0.01.

7. Model for the role of SNX18 in ATG9A and ATG16L1 traffic. Trafficking of ATG9A and ATG16L1 through recycling endosomes is important for their function in autophagosome biogenesis. SNX18 promotes formation of ATG9A‐ and ATG16L1‐positive vesicles from recycling endosomes, and its binding to Dynamin‐2 is required for this. In cells lacking SNX18, ATG9A accumulates in juxtanuclear recycling endosomes and autophagy is inhibited. This can be rescued by WT SNX18, but not by a Dynamin‐2 binding‐deficient mutant SNX18, indicating that Dynamin‐2 is involved in formation of ATG9A vesicles destined for the autophagosomes.