Scaffolding the cup-shaped double membrane in autophagy

Abstract

Autophagy is a physiological process for the recycling and degradation of cellular materials. Forming the autophagosome from the phagophore, a cup-shaped double-membrane vesicle, is a critical step in autophagy. The origin of the cup shape of the phagophore is poorly understood. In yeast, fusion of a small number of Atg9-containing vesicles is considered a key step in autophagosome biogenesis, aided by Atg1 complexes (ULK1 in mammals) localized at the preautophagosomal structure (PAS). In particular, the S-shaped Atg17-Atg31-Atg29 subcomplex of Atg1 is critical for phagophore nucleation at the PAS. To study this process, we simulated membrane remodeling processes in the presence and absence of membrane associated Atg17. We show that at least three vesicles need to fuse to induce the phagophore shape, consistent with experimental observations. However, fusion alone is not sufficient. Interactions with 34-nm long, S-shaped Atg17 complexes are required to overcome a substantial kinetic barrier in the transition to the cup-shaped phagophore. Our finding rationalizes the recruitment of Atg17 complexes to the yeast PAS, and their unusual shape. In control simulations without Atg17, with weakly binding Atg17, or with straight instead of S-shaped Atg17, the membrane shape transition did not occur. We confirm the critical role of Atg17-membrane interactions experimentally by showing that mutations of putative membrane interaction sites result in reduction or loss of autophagic activity in yeast. Fusion of a small number of vesicles followed by Atg17-guided membrane shape-remodeling thus emerges as a viable route to phagophore formation.

Fig 1. Stable vesicle shapes for different reduced volumes.

(A) Axisymmetric cup-shaped vesicle (phagophore) with reduced volume v = 0.45. (B) Disk-shaped (axisymmetric biconcave oblate) vesicle with reduced volume v = 0.594. (C) Axisymmetric prolate vesicle (tube) with reduced volume v = 0.75. The transitions occur at reduced volumes v ≈ 0.652 and v ≈ 0.592 from tube to disk, and from disk to phagophore, respectively [32, 33]. The energetics of the membrane tube-to-sheet transition in the regime of narrow tubes is explored in more detail in [14].

Fig 2. Membrane energy of different vesicle shapes at reduced volume v = v₃ = 0.577.

The elastic energy (blue dots) was determined by SA simulations as a function of the area differences Δa. Energies are reported in units of 8πκ, which is about 8πκ ≈ 250 to 500 k_BT for typical membrane rigidities of κ ≈ 10 to 20 k_BT. Arrows indicate the two energy barriers H₁ ≈ 0.069(8πκ) and H₂ ≈ 0.038(8πκ) for the tube-to-disk and disk-to-phagophore transitions, respectively. Also shown are characteristic membrane shapes (A) at the top of the barrier between disk and phagophore, (B) in the metastable disk minimum, (D) at the top of the tube-to-disk barrier, (F) in the metastable tube minimum, and (G) of three just-fused spheres for a large area difference. The free energy (red), obtained by combining umbrella sampling simulations using WHAM, indicates a slightly higher free energy barrier from entropic effects. See [14] for a more extensive exploration of the tube-to-sheet transition in membrane vesicles.

Fig 3. Coarse-grained MARTINI simulation of three fused vesicles.

(A) Three initial POPC vesicles each composed of 2048 POPC lipids just before fusion. (B) Post fusion structures of three vesicles fused together by gradually pulling lipids together from contacting vesicles to prevent water exchange between the inside and the outside of the vesicles. (C) Equilibrated tubular structure after 0.8 μs simulation. The post fusion structure composed of 6144 POPC lipids rapidly transitioned to a tubular vesicle which is stable during 1.2μs (see S1 Video). The 1.1 × 10⁶ water particles are not shown.

Fig 4. Atg17 dimer structure and membrane interaction.

(A) Atg17 dimer in ribbon representation. The circle indicates the crescent curvature. Mutant clusters in one of the two copies of Atg17 are indicated with a space-filling atomic model. M1, blue; M2 unique residues, cyan; M3, green; M4, yellow; and M5 orange. (B) Coarse-grained representation of Atg17 dimer. (C) Side and (D) top views of three just-fused vesicles tethered by two groups of three Atg17 dimers. The snapshot shows the very first moment after fusion when two narrow catenoid membranes connect the vesicle pairs. (E) The Atg17 dimer matches the profile of the bowl membrane at the top of the second barrier (see point A in Fig 2) between disk and phagophore conformations.

Fig 5. Phagophore induction by Atg17 dimers in the intermediate binding regime: The paddle pathway.

(A-H) Snapshots taken along a simulation of Atg17-induced phagophore induction with an Atg17-membrane binding strength of u = 0.12 (membrane: transparent blue; Atg17 dimers: red). (A) Vesicle after fusion (with v = v₃ = 0.577) and opening of the two catenoid shaped necks, interacting with six Atg17 dimers. The initial vesicle transforms to the bowl (F,G) and the phagophore (H) through intermediate paddle (D) conformations. (H) Phagophore shape shown in transparent side view (left) and opaque top view (right).

Fig 6. Strong binding regime: The starfish pathway.

(A-F) Snapshots taken along a simulation of Atg17-induced phagophore induction with an Atg17-membrane binding strength of u = 0.17 (membrane: transparent blue; Atg17 dimers: red). The initial vesicle (A) with v = v₃ = 0.577 transforms to the bowl (F) through intermediate starfish conformations (B,C), induced by strong lateral binding of Atg17 dimers (three right strands in (D,E)) and their cooperative interactions. The transitions from bowl toward the phagophore is similar to that in the paddle pathway of Fig 5.

Fig 7. Tubular trap in the weak binding regime.

(A-F) cooperative binding of six single-stranded Atg17 dimers in two groups of three. Although a paddle-like shape starts to form in snapshot C, the weak binding strength u = 0.05 of the dimers, is not strong enough to stabilize the paddle. Eventually the vesicle moves back to the unfavorable tube shape (D-F) and gets trapped there.

Fig 8. Membrane remodeling with straight proteins.

(A-D) Snapshots taken along a simulation of attempted phagophore induction with a straight variant of an Atg17-like protein and a membrane binding strength of u = 0.1 (membrane: transparent blue; Atg17 dimers: red).

Fig 9. Phagophore formed by flexible Atg17 dimers.

(A) Structure from a simulation of phagophore induction by six flexible Atg17 dimers with a stiffness K_ang = 200 k_BT, and a membrane binding strength of u = 0.1 (membrane: transparent blue; Atg17 dimers: red). (B) One of the Atg17 dimers shows different curvatures of its two crescents to accommodate better to the phagophore shape.

Fig 10. Atg17 mutations do not affect the complex assembly in vitro.

The Atg17 mutations were introduced into the Atg17-Atg31-Atg29 construct for E coli expression. (A) Size exclusion chromatography of the recombinant ternary complex showed that the wild-type and mutants eluted in similar positions on a superpose 6 column, suggesting that Atg17 mutations do not interfere with dimerization. (B) SDS-PAGE confirmed that the peak fractions on superpose 6 column contain Atg17, 31, 29. (C) Atg17 mutations do not affect Atg13 binding. GST-Atg13 (351–458) was able to pull down the wild type or mutants of Atg17-Atg31-Atg29 complex with similar affinity.

Fig 11. Atg17 crescent surface mutations impair autophagy.

(A) Pho8Δ60 assay for monitoring autophagy was performed following growth in rich (white) or nitrogen starvation (grey) media. Error bars represent S.D. of triplicate experiments. (B) The expression of Atg17-GFP was monitored by western blot against GFP. Western blot against 3-phosphoglycerate kinase (PGK1) was used as the loading control for yeast cell extracts.