Autophagy initiation by ULK complex assembly on ER tubulovesicular regions marked by ATG9 vesicles

Introduction

Autophagy is the membrane trafficking pathway that delivers intracellular material for degradation to lysosomes via de novoformation of double-membrane vesicles, the autophagosomes. Cells activate autophagy in response to nutrient limitation or accumulation of damaged proteins and organelles, and as a result they recycle building blocks generated in lysosomes into new macromolecules and sub-cellular structures1. Unsurprisingly, autophagy has implications for ageing and associated diseases, such as neurodegeneration, inflammation and cancer2,3,4,5. Autophagy also underpins physiological functions, such as development, cell differentiation and immunity6.

Autophagosome formation requires a protein machinery to act upon a membrane source, organize it into a flat sheet, expand it and finally fuse its extremities to enclose the cytosolic cargo. Yeast genetics have identified >30 autophagy-related (Atg) genes encoding the core autophagy machinery, most of which are conserved in mammals3,7. This protein machinery is organized into functional complexes that carry out the steps of autophagosome formation3. In brief, scarcity of amino acids inactivates mechanistic target of rapamycin (mTORC1) and releases the repression of a functional complex, including the protein kinase ULK1, which then translocates to membranes (initiation step). ULK1 activates the functional complex, including the lipid kinase VPS34, stimulating the synthesis of phosphatidylinositol 3-phosphate (PI3P) and the formation of an omegasome (nucleation step). The omegasomes are membrane platforms in contact with endoplasmic reticulum (ER), where the remaining core machinery is recruited. This includes vesicles of ATG9, the only transmembrane autophagy protein, and two conjugation systems, ultimately leading to the covalent attachment of the small ubiquitin-like protein LC3 to phosphatidylethanolamine (lipidation). LC3, the signature protein of autophagosomes, promotes the expansion of the autophagosomal membrane (also known as isolation membrane, elongation step), and its closure and fusion with the lysosome (maturation step).

The characterization of the membrane source that drives the nucleation and elongation of autophagosomes has proven so far to be more elusive. It is generally accepted that more than one membrane sources are likely to be involved in the different steps of autophagosome formation, including the ER, mitochondria, mitochondria-associated membranes, the Golgi, the plasma membrane and recycling endosomes8,9,10,11,12,13,14. Efforts to identify this membrane source have focused on two fronts: the co-localization of the autophagic machinery with pre-existing organelles and the characterization of the membrane compartment that hosts ATG9, the only transmembrane autophagy protein. Among the pre-existing organelles, autophagosomes induced by amino-acid starvation emerge adjacent to ER; however, the mechanistic contribution of this arrangement has remained unknown8,12,15,16,17,18. One possibility is that the autophagic machinery associates with two ER-associated membrane compartments: the ER exit sites (ERES) and the ER–Golgi intermediate compartment (ERGIC)19,20,21,22,23,24. The ERES are sites where proteins trafficked to Golgi are packaged into coat protein complex I (COPII)-coated carriers, creating an adjacent collection of vesicular–tubular structures that constitutes the ERGIC25. Of note, the Rab GTPase Rab1/Ypt1, which is required for the ER-to-Golgi trafficking, also promotes autophagy26,27,28,29,30. A second possibility is that the ER coordinates the redistribution of the ATG9 compartment during autophagy14,31,32. Interestingly, Ypt1 binds to Atg9 (refs 29, 33), one of the first proteins recruited at the pre-autophagosomal structure (PAS), promoting the recruitment of downstream proteins14,34. Moreover, the mammalian ATG9 colocalizes at the some point of its life cycle with the ULK1 complex13,32.

The recruitment of the ULK1 complex confers to the earliest autophagy-specific structure forming de novo its autophagic character. Characterizing the provenance and mode of formation of this structure though has proven to be challenging: it is a short-lived intermediate that has not yet acquired the characteristic double-membrane crescent shape identified by electron microscopy. Herein, we take a comprehensive imaging-based approach to address where the ULK1 complex nucleates autophagosome formation. We find that both ERES/ERGIC and ATG9 are functionally important for the nucleation of autophagosomes, though ATG9 vesicles are more frequently observed in association with the earliest detectable autophagy structures. We develop for the first time a protocol for direct Stochastic Optical Reconstruction Microscopy (dSTORM) of autophagy-related structures and we observe the autophagic machinery in combination with the ERES/ERGIC and ATG9 compartments at sub-diffraction resolution. We find that ATG13 decorates a unique compartment that associates with small vesicles of ATG9 or ERGIC, but not with the conventional ERES. Using correlative light and electron microscopy (CLEM), we find that the ATG13 structures correspond to a tubulovesicular compartment surrounded by ER and mitochondria. We propose that new autophagosomes nucleate in regions, where the ULK1 complex coalesces with the ER and vesicles of ATG9.

Results

ERES/ERGIC contributes to autophagosome nucleation

Autophagosomes induced by amino-acid starvation in HEK293 cells form on ER-associated PI3P-rich structures called omegasomes8. It was recently proposed that the site of autophagosome nucleation on the ER is ERES22,35. We hypothesized that if ERES/ERGIC are the platform of autophagosome nucleation they should colocalize with puncta of the ULK1 complex, the earliest autophagy-specific structure. We used as a surrogate for the ULK1 complex ATG13, which is an important stable component of the complex involved exclusively in autophagy that has been extensively characterized (refs 16, 36, 37). Under autophagy-inducing conditions only a fraction of ATG13 puncta associated clearly with ERGIC53 (ERGIC marker; Fig. 1a) or SEC23 and SEC23IP (ERES markers; Supplementary Fig. 1). We reasoned that ATG13 might associate with ERES/ERGIC only during the nucleation step. However, in live cells only one third of the GFP-ATG13 particles emerged in association with mCherry-SEC16 (ERES marker38; Fig. 1b–d), suggesting that ERES/ERGIC can only partially account as platform for autophagosome nucleation.

Figure 1: ER exit promotes the formation of ATG13 puncta.

(a) HEK293 cells were fed or starved for 1 h, immunolabelled for ATG13 and ERGIC53, and imaged by wide-field microscopy. (b) Values are ATG13 particles in c,d associating with SEC16 particles in the first two frames from their emergence. From analysis of 73 montages. (c,d) Wide-field live-cell imaging of starved HEK293 cells expressing stably GFP-ATG13 and transiently mCherry-SEC16. Representative montages of ATG13 particles forming in association with SEC16 (c) or not (d) are shown. Arrowheads point at the ATG13 particles in the first two frames from their emergence, the same that were used for the analysis in b. (e) HEK293 cells were fed or starved for 1 h, treated with 50 μM H89 in the last 30 min, immunolabelled for ATG13 and imaged by wide-field microscopy. (f) HEK293 cells were starved for 1 h, immunolabelled for βCOP and ERGIC53 and imaged by wide-field microscopy. (g) HEK293 cells stably expressing GFP-DFCP1 were fed or starved, immunolabelled for δCOP and imaged by wide-field microscopy. Arrowheads in insets point at COPI particles adjacent to DFCP1 puncta or rings. (h) HEK293 cells were pre-treated with 3 μg ml−1 BFA for 3 h (long BFA) or for 30 min (short BFA) and for 30 min with 100 μM FLI06, then starved for 1 h in the presence or absence of BFA and FLI06, immunolabelled for ATG13 and imaged by confocal laser scanning microscopy. (i) Values are means±s.e.m. puncta of ATG13 per cell in e, for at least five different fields with 15–30 cells each. (j) Values are means±s.e.m. puncta of ATG13 per cell in h, for at least five different fields with 15–30 cells each. (k) Values are means±s.e.m. puncta of ATG13 per cell in an independent experiment, for at least five different fields with 15–30 cells each. Significance levels were determined with unpaired t-tests. ****P=0.0001%. Bar in a,e–h corresponds to 10 μm. Bar in c–e corresponds to 300 nm.