What’s in an E3: role of highly curved membranes in facilitating LC3–phosphatidylethanolamine conjugation during autophagy

ABSTRACT

During autophagosome formation, ATG3, an E2-like enzyme, catalyzes the transfer of LC3-family proteins (including Atg8 in yeast and LC3- and GABARAP-subfamily members in more complex eukaryotes) from the covalent conjugated ATG3–LC3 intermediate to PE lipids in targeted membranes. A recent study has shown that the catalytically important regions of human ATG3 (hereafter referred to as ATG3), including residues 262 to 277 and 291 to 300, in cooperation with its N-terminal curvature-sensing amphipathic helix (NAH), directly interact with the membrane. These membrane interactions are functionally necessary for in vitro conjugation and in vivo cellular assays. They provide a molecular mechanism for how the membrane curvature-sensitive interaction of the NAH of ATG3 is closely coupled to its conjugase activity. Together, the data are consistent with a model in which the highly curved phagophore rims facilitate the recruitment of the ATG3–LC3 complex and promote the conjugation of LC3 to PE lipids. Mechanistically, the highly curved membranes of the phagophore rims act in much the same manner as classical E3 enzymes in the sumo/ubiquitin system, bringing substrates into proximity and rearranging the catalytic center of ATG3. Future studies will investigate how this multifaceted membrane interaction of ATG3 works with the putative E3 complex, ATG12–ATG5-ATG16L1, to promote LC3–PE conjugation.

During macroautophagy/autophagy, a key step is the covalent conjugation of LC3-family proteins to phosphatidylethanolamine (PE) lipids in the phagophore membrane. The resulting LC3–PE conjugate regulates phagophore expansion and provides a docking site for autophagic cargoes. Three ubiquitin-like conjugation enzymes (E1-, E2-, and E3-like) mediate the conjugation. In the current model, ATG7 (E1-like) catalyzes the ATP-dependent conjugation of the carboxyl-terminal glycine residue in LC3 (which becomes exposed after ATG4-dependent proteolytic removal of a few C-terminal residues) to the catalytic cysteine residue of ATG3 (E2-like). The ATG3–LC3 intermediate is then targeted to the tip of the phagophore, where LC3 is transferred to PE lipids facilitated by coordinated protein-protein and protein-lipid interactions in ATG3 and the ATG12–ATG5-ATG16L1 complex (E3-like) (Figure 1). In vitro, however, LC3-conjugated ATG3 itself can catalyze the transfer of LC3 to PE lipids upon the binding of its N-terminal amphipathic helix (NAH) to small liposomes with highly curved membranes that model the phagophore tip. Our recent study identifies additional membrane interactions in human ATG3 (hereafter referred to as ATG3) essential for conjugase activity [1]. Specifically, interactions of the catalytically important regions of ATG3 with the membrane, in conjunction with the curvature-sensitive interaction of its NAH, target the C-terminal located catalytic residue Cys264 to the bilayer surface where the substrate PE lipids are anchored and promotes LC3B–PE conjugation.

Figure 1.

Comparison of the classical ubiquitination process and LC3B–PE conjugation pathway. (A) In the classical ubiquitination process, ubiquitin (Ub, yellow) is attached to the E1-ubiquitin activating enzyme (green) in an ATP-dependent manner before transferred to an E2 conjugating enzyme (purple). Ubiquitin in the Ub–E2 complex is then transferred to a lysine residue in the substrate (red) mediated by a RING E3 (cyan). (B) By comparison, in the LC3B–PE conjugation process, LC3B (yellow) is attached to the E1-like ATG7 (green) in an ATP-dependent manner before transferred to an E2-like ATG3 (purple). LC3B in the ATG3–LC3B complex is transferred to the amino headgroup of phosphatidylethanolamine (PE, red) substrates in the target membrane (cyan head groups with gray aliphatic chains and red PEs), which is mediated by the highly curved membrane geometry (E3-like?) alone or with the putative E3-like ATG12–ATG5-ATG16L1 (gray). The N-terminal amphipathic helix (NAH, yellow) and helices F (green) and G (blue) of ATG3 are shown. Whether LC3B is involved in membrane interaction and how the putative E3-like ATG12–ATG5-ATG16L1 promotes the transfer of LC3B from ATG3–LC3B to the membrane remain to be determined.

ATG3 is an intrinsically disordered protein with more than one-third of its 314 residues unstructured. Some disordered regions can be deleted without markedly affecting its conjugase activity in vitro; for example, deleting residues 90–190 (ATG3[∆90–190]) results in a construct that can bind to curved membranes and supports LC3B–PE lipid conjugation in vitro. The structure of a deletion construct of ATG3[∆N25, ∆90–190], lacking residues 1–25 (including the NAH) and thus membrane-binding activity, is similar (RMSD C_α ~1Å) to those of yeast and Arabidopsis ATG3. The most significant structural differences occur in helix F (residues 268–278) adjacent to Cys264. Notably, helix F in ATG3[∆N25, ∆90–190] is highly dynamic in solution, as revealed by NMR hydrogen/deuterium exchange experiments. We speculate that the high degree of conformational flexibility in the catalytic regions of ATG3 may be a critical functional feature: providing access to ATG7 and LC3B for ATG3–LC3B conjugation, facilitating subsequent association of the ATG3–LC3B complex with the membrane and the ATG12–ATG5-ATG16L1 complex, and promoting the transfer of LC3B from ATG3–LC3B to PE lipids at the phagophore tip in vivo.

The ATG3[∆90–190]–LC3B intermediate supports the transfer of LC3B to PE lipids in bicelles, an established model for highly curved membranes. To dissect the underlying molecular mechanism that activates the conjugase activity of ATG3, we used an optimized version of the ATG3[∆90–190] construct that stabilizes its interaction with bicelles while retaining function in vitro. The resulting construct, ATG3[∆90–190, 4 M] (H240Y, V241A, P263G, and H266L), allows the assignment of all but 13 residues in the NMR spectra of ATG3[∆90–190, 4 M] bound to bicelles. Importantly, residues adjacent to Cys264, including 244–263 and 266–310, are assigned in the membrane-bound state. The new assignments facilitate the use of NMR cross-saturation experiments to identify residues that directly interact with the membrane. In addition to the NAH, residues 262–277 (region I) and 291–300 (region II) show pronounced cross-saturation effects, implying their membrane proximity in the bicelle-bound ATG3[∆90–190, 4 M] state.

Residues 265–277, within region I, adopt an amphipathic helical conformation in the bicelle-bound ATG3[∆90–190, 4 M] based on analyses of their ¹³C_α and ¹³C_β shifts. Intriguingly, some hydrophobic residues are solvent-exposed in the apo state and, thus, presumably primed to interact with the membrane. Substituting hydrophobic residues in this region with aspartic acid abrogates (H266D, A267D, or I273D) or markedly reduces (V269D or M270D) LC3B–PE formation in vitro.

Residues 291–300 in region II, like region I, adopt a helical conformation with an extensive nonpolar face and a small polar face. Four conserved sequential residues (F₂₉₃LKF₂₉₆) experience the most cross-saturation effects upon membrane binding. Phe293 and Leu294 interact with strands in the central β-sheet, stabilizing the fold, whereas residues Lys295 and Phe296 are solvent-exposed in the apo structure. Substitution of Lys295 with leucine (ATG3^K295L) remains fully functional. By contrast, the replacement of Lys295 with glutamic acid (ATG3^K295E) abolishes conjugase activity, suggesting that Lys295 interacts with the membrane by “snorkeling” its positively charged amino group to interact with the negatively charged phospholipid head group. Similarly, replacing Phe296 with glutamic acid, serine, or alanine nearly abolishes in vitro function. In mouse embryonic fibroblast cells, the F296L, but not F296S, mutation catalyzes LC3B–PE formation and supports autophagic flux, suggesting that hydrophobic interactions of Phe296 with the membrane are necessary in vivo.

Our in vitro conjugation reactions do not possess a classical E3 (the ATG12–ATG5-ATG16L1 complex). Instead, transferring LC3B from the ATG3–LC3B intermediate to PE lipids needs only highly curved liposomes containing acidic lipids and the substrate PE lipids. This reaction involves targeting of ATG3 to the highly curved phagophore membrane via the NAH and the interaction of C-terminal regions I and II in ATG3 with the membrane. In doing so, ATG3 places Cys264 near the lipid head groups and rearranges its catalytic center. In this system, the highly curved membrane itself serves the role of the classical E3 by bringing ATG3–LC3B and PE lipids into proximity and priming the active site of ATG3 for catalysis. Our work leaves several unanswered questions, including does LC3B interact with the membrane upon ATG3–LC3B binding and how these multifaceted membrane interactions of ATG3 integrate into the function of the ATG12–ATG5-ATG16L1 complex in vivo. The contribution of LC3B to membrane binding could further enhance the binding and stability of the active conformation of ATG3–LC3B. At the same time, ATG12–ATG5-ATG16L1 participation could provide additional selectivity for the phagophore tip and/or further stabilize the active state for LC3B transfer. However, answers to these questions must await further detailed investigations of the reconstituted system.

Funding Statement

This work was supported by the National Institutes of Health through R01 GM127730 (FT), R01 GM127954 (HGW), R01 CA222349 (HGW) and Four Diamonds through 4D21_2024_1001 (FT) and 4DIA_153697 (JMF). The NMR instruments used in this project (RRID:SCR_023244) were funded, in part, by the Pennsylvania State University College of Medicine via the Office of the Vice Dean of Research and Graduate Students and the Pennsylvania Department of Health using Tobacco Settlement Funds (CURE). The content is solely the responsibility of the authors and does not necessarily represent the official views of the University or College of Medicine. The Pennsylvania Department of Health specifically disclaims responsibility for any analyses, interpretations, or conclusions.

Disclosure statement

No potential conflict of interest was reported by the author(s).