The term autophagy comprises three mechanisms of degradation (macroautophagy, microautophagy, and chaperone-mediated autophagy). While all of these converge in the lysosomal compartment, the structures involved as well as their degradation capacities differ significantly. The process described originally, which dominates in quantitative terms, is macroautophagy (usually termed autophagy for short). Besides its vital role in nutrient starvation, macroautophagy is now considered the most important cellular degradation pathway for elimination of long-lived macromolecules and surplus organelles. In this process, cargo is sequestered by double-membrane vesicles (autophagosomes), which later fuse with the lysosomal compartment. Formation of these organelles critically depends on the Atg8 family of small ubiquitin-like proteins, which are enzymatically conjugated to membrane lipids. Whereas the yeast S. cerevisiae expresses only a single member of this family (Atg8 itself), human cells contain several homologues, including GABAA receptor-associated protein (GABARAP).
Fig. 1 NMR ensembles of GABARAP (left) and Atg8 (right). α-helices are highlighted in orange, β-sheets in blue.
We were among the first to determine the solution structures of human GABARAP and yeast Atg8 by NMR spectroscopy (Fig. 1). Both molecules feature a structural motiv called the β-grasp fold, which is characteristic of the ubiquitin superfamily of proteins; it consists of a central four-stranded β-sheet with two α-helices on its concave face. A distinctive feature of the Atg8 family is the N-terminal extension containing two additional α-helices, which are attached to the convex side of the β-sheet. In the case of Atg8, this N-terminal subdomain shows large conformational mobility on a micro- to millisecond time scale (Fig. 1, right). In contrast, the N-terminus of GABARAP is relatively well-defined (Fig. 1, left).
Insight into the ligand specificity of GABARAP was first obtained by screening phage-displayed peptide libraries. Database searches with a consensus motif yielded several potential native binding partners, including calreticulin and the heavy chain of clathrin. We have characterized the ligand binding mode of GABARAP extensively, e.g. by determining the crystal structures of complexes with artificial as well as calreticulin-derived peptides (Fig. 2). In all cases, the interaction was mediated by two hydrophobic pockets on the GABARAP surface (hp1, hp2) accommodating a WxxΦ motif (Φ = hydrophobic side chain) in the ligand.
Fig. 2 Crystal structures of GABARAP complexed with an artificial ligand (K1, left) and a calreticulin peptide (CRT(178-188), right). In the ribbon representation of GABARAP, the β-grasp motif and the N-terminal extension are coloured dark blue and light blue, respectively. Hydrophobic side chains of the ligands contacting the apolar patches of GABARAP are drawn explicitly. The two binding modes differ significantly; only the second one has been found in physiological interactions.
The hexameric ATPase NSF (N-ethylmaleimide-sensitive factor), which is part of the minimal vesicle fusion machinery, has been reported by others to interact with GABARAP and its paralogue GATE-16. We have proposed a model of the NSF-GABARAP complex (Fig. 3). By virtue of their C-terminal lipidation, GABARAP or GATE-16 may serve to anchor the NSF complex to membranes.
Fig. 3 Model of the NSF-GABARAP complex (surface representation) in axial (left) and equatorial view (right). The identical subunits of the hexameric ATPase are coloured blue and light grey, the GABARAP molecules are orange.
Recently, we were able to show that GABARAP interacts with the pro-apoptotic Bcl-2 family member Nix (Nip-like protein x). NMR spectroscopy revealed that binding is again mediated by hp1 and hp2 on the GABARAP surface. Bcl-2 itself also displays significant affinity for GABARAP, which probably contributes to the well-known anti-autophagic character of the Bcl-2 protein. In this case, however, GABARAP contributes only hp1 residues to the interaction. Notably, the three-dimensional structure of lipid-conjugated GABARAP, i.e. the biologically active species, has not been determined experimentally. By using a thiol-reactive lipid compound, we have been able to anchor the GABARAP C-terminus to nanodisc particles, thus mimicking the membrane-attached state. NMR studies indicated that this molecule largely retains its native fold, with its hydrophobic pockets accessible for interaction with target proteins. Finally, we have provided the first quantitative account of GABARAP oligomerization, based on DOSY-HSQC measurements accessing the diffusion properties of the species in solution.
Our current research focuses on the structure and function of established as well as candidate players in macroautophagy. Using X-ray crystallography, we could determine the structure of human ATG101; the latter is an essential component of the ULK complex, which is responsible for the initiation of macroautophagy. The protein displays a so-called HORMA domain fold, which is named after the prototype family members Hop1p, Rev7p and MAD2 (Fig. 4). Whereas MAD2 has been found to undergo equilibration between an open and a closed conformation, ATG101 appears to be locked in an open state. In vivo, ATG101 interacts with ATG13, which also contains a HORMA domain (albeit in a closed conformation). Based on our ATG101 structure as well as available data for ATG13, we have established a model for this complex.
Fig. 4 Crystal structure of human ATG101. The ribbon model (left) reveals the HORMA domain architecture, which contains a central helical scaffold flanked by β-structure elements. Three insertions (ext1, ext2, ext3) relative to the MAD2 structure are highlighted; these are likely to represent adaptation of the HORMA fold to the specific functionality of ATG101. In the topology diagram (right), these insertions are marked in yellow.
Further projects include LC3C, a poorly characterized member of the Atg8 family with remarkable properties, Atg9, the only core autophagy protein containing multiple membrane-spanning helices, and the membrane-anchored immunophilin FKBP38, which modulates Bcl-2 function via a hitherto unknown mechanism.