Atg9 establishes Atg2-dependent contact sites between the endoplasmic reticulum and phagophores

Abstract

The autophagy-related (Atg) proteins play a key role in the formation of autophagosomes, the hallmark of autophagy. The function of the cluster composed by Atg2, Atg18, and transmembrane Atg9 is completely unknown despite their importance in autophagy. In this study, we provide insights into the molecular role of these proteins by identifying and characterizing Atg2 point mutants impaired in Atg9 binding. We show that Atg2 associates to autophagosomal membranes through lipid binding and independently from Atg9. Its interaction with Atg9, however, is key for Atg2 confinement to the growing phagophore extremities and subsequent association of Atg18. Assembly of the Atg9-Atg2-Atg18 complex is important to establish phagophore-endoplasmic reticulum (ER) contact sites. In turn, disruption of the Atg2-Atg9 interaction leads to an aberrant topological distribution of both Atg2 and ER contact sites on forming phagophores, which severely impairs autophagy. Altogether, our data shed light in the interrelationship between Atg9, Atg2, and Atg18 and highlight the possible functional relevance of the phagophore-ER contact sites in phagophore expansion.

Figure 1.

Atg2 and Atg9 directly interact. (A) Atg2–Atg9 interaction in different Y2H strains. Plasmids carrying the ATG2 or ATG9 gene fused with the BD or AD domains of the transcription factor Gal4, respectively, were transformed into Y2H WT (PJ69-4A) or atg18Δ (FRY382) strains. The pGBDU-C1 plasmid (empty) was used as a negative control. (B) Recapitulation of the Atg2–Atg9 interaction using the split-ubiquitin assay. All the split-ubiquitin constructs—pATG9_Cub_RURA3_Met313, pATG9_Nub_CUP_314, pATG18_Nub_CUP_314, and pATG2_Nub_Cub_314—were cotransformed into either WT (SEY6210) or atg2Δ (FRY383) cells. The pNub_CUP_314 plasmid was used as a negative control. (C) Atg18 is not required for the Atg2–Atg9 interaction. Cell extracts from atg2Δ atg9Δ (yDP29), atg2Δ Atg9-GFP (yDP191), and atg2Δ atg18Δ Atg9-GFP (yDP264) strains transformed with an empty vector (pRS315) or a plasmid expressing TAP-tagged Atg2 were subjected to pulldown experiments as described in Materials and methods. Immunoisolates were analyzed by Western blotting using anti-GFP and anti-TAP antibodies. (D) Model of the Atg9–Atg2–Atg18 complex. (E) A stretch of 34 amino acids between positions 1,232 and 1,268 of Atg2 is essential for the interaction with Atg9. Plasmids expressing the Atg2^1–1,302, Atg2^1–1,268, Atg2^1–1,204, Atg2^1–1,089, and Atg2^1–909 truncations were cotransformed with the vector carrying AD-Atg9 into the WT strain (PJ69-4A) before being assayed on the test plates. (F) Structural organization of Atg2 in domains as proposed (Kaminska et al., 2016). Through homology search (Finn et al., 2016), it appears that Atg2 possesses a Chorein-N domain (PF12624), a region with similarity to the mitochondrial protein FMP27 predicted to form a solenoid structure, an ATG2-CAD domain (PF13329) with unknown function, and a part similar to the Golgi APT1 protein of maize (PF10351). Additionally, the C terminus of Atg2 contains a region of high homology with the two mammalian Atg2 orthologues. It is composed of two ATG-C domains (PF09333) of unknown function. The first domain is truncated and lacks the distal part, whereas the second one is intact. The dashed lines indicate the identified region of Atg2 where the amino acids essential for its binding to Atg9 are localized. (G) Point mutants in Atg2. The Atg2 amino acid sequence between residues 1,232 and 1,271 is shown. The four Atg2 point mutants (PM1, PM2, PM3, and PM4) generated by replacing the charged and polar amino acids with alanines are indicated. The introduced alanines are in bold. (H) Interaction of point mutants with Atg9. BD-tagged Atg2 point mutants Atg2^PM1, Atg2^PM2, Atg2^PM3, and Atg2^PM4 were tested for their ability to bind AD-Atg9 in the WT strain (PJ69-4A) by Y2H assay. Only Atg2^PM3 was able to interact with Atg9.

Figure 2.

Interaction between Atg2 and Atg9 is essential for both bulk and selective autophagy. (A) Mutations in the putative Atg9-binding region of Atg2 lead to a severe block of bulk autophagy. The atg2Δ cells (FRY375) carrying both the pCuGFPATG8414 vector and a plasmid expressing Atg2 or the different Atg2 point mutants or the empty vector pRS416 were grown in SMD to an early log phase and transferred to the autophagy-inducing SD-N. Culture aliquots were collected 0, 1, 2, and 4 h after autophagy stimulation, and cell extracts were analyzed by Western blotting using an antibody against GFP. A graph representing the relative amount of the GFP-Atg8 chimera at each time point calculated from three independent experiments plus SD is shown on the right. Representative blots are shown on the left. (B) Defective autophagy caused by Atg2 mutations. The PHO8Δ60 atg2Δ strain (FRY388) was transformed with an empty vector (pRS416; atg2Δ) or plasmids expressing Atg2 or the different Atg2 point mutants. Transformed cells were cultured in SMD to early log phase and transferred into SD-N starvation medium for 4 h to induce autophagy. The Pho8Δ60 assay was performed as described in Materials and methods. (C) Mutations in the Atg9-binding region of Atg2 severely affect the cytosol-to-vacuole targeting pathway. Strains analyzed in A were cultured in SMD to early log phase. Samples were collected, and cell extracts were analyzed by Western blotting using the anti-Ape1 antiserum. The detected bands were then quantified as in A, and the percentages of precursor and mature Ape1 (prApe1 and mApe1, respectively) were plotted. The presented data represent the means of three independent experiments ± SD. (D) The identified Atg2^PM1, Atg2^PM2, and Atg2^PM4 mutants do not interact with Atg9 in vivo. Cell extracts from atg2Δ (yCK759) and atg2Δ Atg9-GFP (yDP191) strains transformed with an empty vector (pRS315) or plasmids expressing WT or point-mutated TAP-tagged Atg2, pATG2^PM1-TAP, pATG2^PM2-TAP, and pATG2^PM4-TAP were subjected to pulldown experiments and analyzed as in Fig. 1 C. (E) Atg18 interaction with Atg2 requires Atg2 binding to Atg9. Cell extracts from atg2Δ (FRY375) and atg2Δ Atg18-13×myc strains transformed with an integrative empty vector (RSGY015) or plasmids expressing TAP-tagged versions of Atg2, Atg2^PM1, or Atg2^PM4 (RSGY012, RSGY013, and RSGY014) were subjected to pulldown experiments and analyzed with anti-myc and anti-TAP antibodies.

Figure 3.

Atg2 binding to Atg9 promotes its direct interaction with Atg18. (A) Purified Atg9 and Atg2–18 complexes. Atg9-3×FLAG and Atg2–Atg18–TAP complexes were overproduced in yeast and purified as described in Materials and methods. Isolated proteins were separated by SDS-PAGE and visualized in gels with Coomassie staining. The asterisk indicates a degradation product. MW, molecular weight. (B) Schematic representation of liposome flotation assays. Liposomes containing or not containing Atg9 were incubated with purified Atg2–Atg18 complexes and mixed with 75% sucrose. Subsequent density centrifugation allowed separating unbound protein (bottom) from liposomes with bound protein (top). (C–F) Interaction of Atg2 and Atg18 with liposomes. Liposomes consisting of 69–72 mol% DOPC, 15 mol% DOPE, 12 mol% DOPS, 0.5 mol% Atto550-DPPE, and 0 or 3 mol% PtdIns3P were reconstituted with or without purified Atg9 in a 1:1,000 protein/lipid ratio. Top fractions of different liposome species incubated with purified Atg2–Atg18 (C), Atg2–Atg18^FAAG (D), Atg2^PM1–Atg18^FAAG (E), or Atg2^PM1–Atg18 (F) were TCA precipitated and loaded on SDS-PAGE gel. To analyze the amount of bound protein, gels were stained with Coomassie, and band intensities were quantified using ImageJ. The graphs show mean quantifications of three independent experiments ± SD.

Figure 4.

Atg2 requires PtdIns3P and lipid-packing defects to tightly associate with membranes in vitro. (A) GUVs with the same lipid composition as the liposomes used in Fig. 3 were incubated with either 400 nM purified Atg2-mGFP or an equal volume of buffer (control) for 5 min at room temperature before being imaged. Single focal plane (FP) images and maximum-intensity projections (MIPs) of 62 optical planes are shown. (B) Analysis of Atg2-mGFP binding to GUVs with different compositions. Where indicated, the lipid mixture used in A was altered by substituting 15 mol% DOPE(PE) with equal molarities of DOPC or ergosterol (erg), whereas 3 mol% PtdIns3P was replaced by an equal molarity of DOPC. Bars, 10 µm. (C) Quantification of Atg2-mGFP binding to GUVs of the experiment shown in B. At least 30 GUVs per sample were counted, and the graph represents means of three independent experiments ± SD.

Figure 5.

Atg2 and Atg2^PM1 bind membranes similarly. (A) Atg2-mGFP isolated from WT or atg18Δ cells was incubated with GUVs as in Fig. 4 A. GUVs without PtdIns3P were used as controls. Binding to GUVs was quantified as described in Materials and methods. (B) Purified and DY-647–labeled Atg2 or Atg2^PM1 were incubated with GUVs, and binding was quantified and controlled as in A. At least 30 GUVs per sample were counted, and graphs represent means of three independent experiments ± SD. Bars, 10 µm. MW, molecular weight.

Figure 6.

Atg2^PM1 and Atg2^PM4 mutants are not normally distributed to the PAS. (A) Cellular distribution of Atg2-GFP variants in atg2Δcells (FRY375) transformed with plasmids expressing Atg2-GFP, Atg2^PM1-GFP, or Atg2^PM4-GFP. Strains were grown to an early log phase before being nitrogen starved for 3 h. Cells were imaged by fluorescence microscopy before and after nitrogen starvation. (B) Quantification of the percentage of cells with one or more Atg2-GFP–positive dot in the experiment presented in A. (C) Atg2 binding to Atg9 is required for Atg18 recruitment to the PAS. Cellular distribution of endogenous Atg18-GFP in WT (RSGY017) or atg2Δ (RSGY018) carrying mCherryV5-Atg8 fusion protein and transformed with integrative plasmids expressing TAP-tagged versions of Atg2 (pATG2-TAP(405); RSGY019), Atg2^PM1 (pATG2^PM1-TAP(405); RSGY020), or Atg2^PM4 (pATG2^PM4-TAP(405); RSGY021) strains. Strains were grown to an early log phase before being nitrogen starved for 3 h. Cells were imaged by fluorescence microscopy before and after nitrogen starvation. DIC, differential interference contrast. Bars, 5 µm. (D) Quantification of the percentage of cells with colocalizing puncta presented in C. Graphs represent means of three experiments ± SD. Asterisks highlight significant differences with the strain carrying WT Atg2.

Figure 7.

Atg9 interaction with Atg2 is required for Atg9 normal subcellular distribution. (A) Localization of endogenous Atg9-GFP in WT (KTY97) or atg2Δ (SAY118) cells transformed with integrative plasmids expressing TAP-tagged versions of Atg2 (pATG2-TAP(405); RSGY003), Atg2^PM1 (pATG2^PM1-TAP(405); RSGY004), or Atg2^PM4 (pATG2^PM4-TAP(405); RSGY005) strains was analyzed. DIC, differential interference contrast. (B) Quantification of the percentage of cells displaying a single Atg9-GFP punctum in the experiment shown in A. (C) Examination of Atg9-GFP distribution on the phagophores adjacent to giant Ape1 by fluorescence microscopy. The atg2Δ mutant expressing Atg9-GFP and mCherry-Atg8 (CUY10934) was transformed with the pDP105 plasmid and analyzed as described in Materials and methods. Bars: (main images) 5 µm; (insets) 1 µm. (D) Statistical evaluation of phagophores displaying Atg9-GFP at their extremities. Graphs represent means of three experiments ± SD. Asterisks highlight significant differences with the strain carrying WT Atg2.

Figure 8.

Atg2^PM1 and Atg2^PM4 are recruited to the PAS, but they have altered distribution on the phagophore. (A) Atg2 localization at the PAS was visualized by BiFC. Strains (RSGY087, RSGY089, and RSGY090) expressing both endogenous Atg1-VC and Atg2-VN, Atg2^PM1-VN, or Atg2^PM4-VN and carrying a mCherryV5-Atg8 construct were grown to an early log phase in YPD before being nitrogen starved for 3 h and imaged. Cells (RSGY088) expressing only Atg1-VC and mCherryV5-Atg8 were used as controls. (B) Quantification of the percentage of BiFC puncta colocalizing with mCherry-Atg8 in the experiment shown in A. (C) Quantification of percentage of cells that present at least one mCherryV5-Atg8 punctum in A. (D) Quantification of the mean size in nm² and intensity of the fluorescent signal in a.u. of mCherryV5-Atg8 puncta depicted in A. Data analysis was performed as described in Materials and methods. (E) Atg2–Atg9 interaction at the PAS was visualized by BiFC. Strains (RHY031, RHY032, and RHY033) expressing both endogenous Atg9-VN and Atg2-VC, Atg2^PM1-VC, or Atg2^PM4-VC and carrying a pCumCherryV5ATG8 construct were processed as in A. Cells (RHY030) expressing only Atg9-VN and mCherryV5-Atg8 were used as a control. DIC, differential interference contrast. Bars, 5 µm. (F) Quantification of the percentage of BiFC puncta colocalizing with mCherry-Atg8 in the experiment shown in E. Graphs represent means of three independent experiments ± SD. Asterisks highlight significant differences with the strains expressing WT Atg2 (B and F) or atg2Δ (C and D) cells.

Figure 9.

The PAS and the ER are in close association in Atg2^PM1- and Atg2^PM4-expressing cells. (A) Strains analyzed in Fig. 8 A (RSGY087, RSGY088, RSGY089, and RSGY090) were transformed with the pDP245 plasmid and grown to an exponential phase before adding 250 µM of CuSO₄ 4 h before reaching 0.6 OD₆₀₀. At that point, 400 nM of rapamycin was added, and incubation was continued for an additional 3 h. Bars: (main images) 1 µm; (insets) 300 nm. (B) Quantification of the percentage of BiFC signal detected on the entire surface of the mCherry-Atg8–positive phagophore and not on its extremities in the experiment shown in A. The graph represents the mean of three experiments ± SD. Asterisks indicate significant differences with the strain carrying WT Atg2. (C) Cryosections of 250–300 nm from either the atg2Δ mutant or cells expressing Atg2^PM1 or Atg2^PM4 were labeled with an anti-Ape1 antibody (10 nm gold; indicated with red spheres in videos). Using a conventional electron microscope, the areas of interest were selected based on the immunogold labeling, and dual-tilt series were recorded using a 200-kV transmission electron microscope. Tomographic slices (inverted grayscale) extracted from different tomograms illustrating different types of association between the Ape1 oligomer and the ER. Single- and double-direction arrows indicate the region of contact and the distance, respectively, between the Ape1 oligomer and the ER. The contours of the Ape1 oligomer (white) and of the ER (yellow) are shown in the middle panels. V, vacuole. Asterisks indicate Ape1 oligomers. Bars: (adjacent ER) 156 nm; (tethered ER) 184 nm. Representative examples of types of associations are also shown as 3D reconstructions in Videos 1 (Ape1 oligomer in the atg2Δ mutant with adjacent ER at a distance between 30 and 150 nm), 2 (Ape1 oligomer in the atg2Δ strain with an ER tethered with a single point of contact), 3 (Ape1 oligomer in Atg2^PM1-expressing cells with an ER tethered with a surface contact), and 4 (Ape1 oligomer in Atg2^PM1-expressing cells with an ER tethered with enwrapping). (D) Quantification of the different Ape1–ER contacts profiles described in the text in the three analyzed strains in C.

Figure 10.

Atg2 determines the contact sites between the phagophore and the ER. (A) Analysis of the ER–phagophore connection in cells generating giant Ape1 by fluorescence microscopy. The atg2Δ mutant expressing Sec63-GFP and mCherry-Atg8 (CUY10935) was transformed with both pDP105 and the pRS416 empty vector or a plasmid expressing Atg2 (pYCG_YNL242w), Atg2^PM1 (pYCG_YNL242w_PM1), or Atg2^PM4 (pYCG_YNL242w_PM4). The resulting strains were grown in SMD to an early log phase before to induce the formation of giant Ape1 as described in Materials and methods and to image the cells. Bars: (main images) 5 µm; (insets) 1 µm. (B) Quantification of the type of ER association to the mCherry-Atg8–positive phagophore in the experiment shown in C. Enwrapped defines all those situations when the ER was tethered to almost the entire surface of the phagophore, and Connected is when there was at least one point of contact between the ER and the phagophore. The graph represents the mean of three experiments ± SD. Asterisks indicate significant differences with cells expressing WT Atg2. (C) Conservation among species of the Atg2 residues involved in Atg9 binding. The amino acid sequence of S. cerevisiae (S.c.) Atg2 between residues 1,232 and 1,271 was aligned with that of Homo sapiens (H.s.) ATG2A and ATG2B, Mus musculus (M.m.) ATG2A and ATG2B, Drosophila melanogaster (D.m.), and Schizosaccharomyces pombe (S.p.) Atg2 using the Clustal Omega program (http://www.ebi.ac.uk/Tools/msa/clustalo/). The amino acids mutated in Atg2^PM1 and Atg2^PM4 are in bold. Asterisks indicate conservation of the residue, and colons designate similarity. (D) Left: Atg9 is confined at the extremities of the phagophore, where Atg2 also gets specifically concentrated by binding to this transmembrane protein. Atg9–Atg2 association also promotes the Atg18 recruitment, and collectively, these three factors play a key role in generating phagophore–ER contact sites at this location, although those appear to be preferentially generated at one of the two edges of the phagophore. Right: Inability of Atg2^PM1 and Atg2^PM4 to bind Atg9 impairs their targeting at the ends of the growing phagophore and Atg18 recruitment to this precursor structure. Redistribution of Atg2^PM1 and Atg2^PM4 on the phagophore surface leads to the formation of more extensive, wrongly positioned, and likely nonfunctional contact sites with the ER.