http://orcid.org/0000-0002-7828-8118
ABSTRACT
The sorting nexin Atg20 interacts with the selective macroautophagy/autophagy scaffolding protein Atg11, suggesting an important role for Atg20 in the initiation of selective autophagy. To explore this possibility, we recently investigated the structure and function of Atg20 using a variety of biophysical and yeast genetic approaches. Our data demonstrate that the BAR domain of Atg20 interacts with Snx4/Atg24 to form an asymmetric heterodimeric BAR domain complex. Atg20 also contains a long intrinsically disordered N terminus that facilitates binding to Atg11 and a large 89-amino acid insertion in its BAR domain, which we have termed the BAR-GAP. This BAR-GAP region is a unique feature of Atg20 and has not been observed in other BAR domains. Furthermore, the BAR-GAP of Atg20 contains an amphipathic helix which is required for membrane binding, tubulation and autophagy. Our findings demonstrate the important role of this novel region in autophagy.
Autophagy in Saccharomyces cerevisiae is initiated at the phagophore assembly site (PAS), a punctate structure that is adjacent to the vacuole. The PAS is formed through interactions between the Atg17-Atg31-Atg29 complex and Atg11. Upon autophagy induction, the kinase-containing Atg1-Atg13 complex is recruited to the PAS through the interaction of Atg13 with either Atg17 or Atg11. If Atg13 binds to Atg17 it leads to the initiation of nonselective (bulk) autophagy. If Atg13 instead binds to Atg11 it leads to the initiation of selective types of autophagy, including mitophagy and the cytoplasm-to-vacuole targeting (Cvt) pathway. Atg11 is also able to recruit several other autophagy proteins to the PAS, including the sorting nexin Atg20. Whereas structural and functional studies have shed light on many of the proteins in these autophagy initiation complexes, Atg20 has remained largely unexplored.
To gain insight into the potential role of Atg20 in autophagy initiation we recently explored the structure and function of this protein. We initially performed a computational prediction to better understand the domain architecture of Atg20. As expected, Atg20 is predicted to contain a PX domain for recognizing phosphatidylinositol 3-phosphate (PtdIns3P) and a BAR domain to potentially aid in shaping the phagophore membrane. Atg20 is also predicted to contain 2 large disordered regions. The first is at the N terminus of Atg20, comprising the first 160 amino acids of the protein. The second is an 89-amino acid region located within the BAR domain, which we termed the BAR-GAP. We purified the N-terminal domain of Atg20 and confirmed that it is disordered in solution. Surprisingly, removal of either the N-terminal disordered region or the BAR domain of Atg20 leads to a reduction, but not complete loss, of Atg11 binding, demonstrating that Atg20 contains 2 distinct Atg11 binding regions.
We hypothesized that the disordered regions in Atg20 may be regulated by post-translational modifications. Using mass spectrometry, we showed that Atg20 contains 10 acetylated lysines and 12 phosphorylated residues. The majority of these post-translational modification sites do in fact reside in the disordered regions of Atg20, supporting our hypothesis. Furthermore, mutation of these sites leads to a decrease in the efficiency of the Cvt pathway, suggesting that post-translational modifications regulate the function of the disordered regions in Atg20.
Most BAR domain proteins form homodimers via their BAR domains. However, immunoprecipitation studies suggested that Atg20 does not homodimerize, but rather interacts with Snx4/Atg24, another PX and BAR domain-containing protein. This led us to hypothesize that Atg20 forms a heterodimeric complex with Snx4 via their BAR domains. To test this, we generated different truncations of Atg20 and monitored Snx4 binding. These data revealed that Atg20 binds to Snx4 only through its BAR domain. In addition, using recombinant expression in E. coli, we observed that Atg20 is insoluble when expressed alone but forms a soluble heterodimer when expressed with Snx4. These results confirm that Atg20 forms a heterodimer with Snx4 via its BAR domain but is incapable of forming a homodimer. To gain insight into the overall architecture of the Atg20-Snx4 complex we utilized several biophysical approaches, including small angle X-ray scattering, analytical ultracentrifugation, and molecular dynamics to generate a structural model of the Atg20-Snx4 complex. This model demonstrated that the Atg20-Snx4 complex is an asymmetric heterodimer, with the dynamic BAR-GAP region of Atg20 being located at one side of the BAR domain complex. These unique features suggest that the Atg20-Snx4 complex is structurally unlike any other BAR domain complex.
The BAR-GAP region of Atg20 is predicted to contain an amphipathic helix, a common feature of BAR domains which is often implicated in membrane binding. However, the location of the predicted amphipathic helix is unique to Atg20 and has not been observed in other BAR domains. To test the importance of this region in membrane binding we mutated 2 hydrophobic residues in the predicted amphipathic helix to glutamic acid. This mutation dramatically impairs Atg20 binding to liposomes, and leads to a reduction of autophagy in cells. To visualize the effect of the Atg20-Snx4 complex on membranes, we monitored liposomes incubated with the Atg20-Snx4 complex using electron microscopy. The Atg20-Snx4 complex tubulates liposomes as has been observed for other BAR domains. However, this tubulation is almost completely lost when the predicted amphipathic helix in the BAR-GAP is mutated. Taken together, these findings suggest that the BAR-GAP region mediates Atg20-membrane interactions, which are important for the function of Atg20 in autophagy.
Our recent work highlights the importance of an asymmetric BAR domain complex in autophagy. It is interesting to note that the asymmetry of this complex is further exaggerated by Atg11, which binds to Atg20 but not to Snx4. As Atg11 also interacts with autophagy receptors that recognize cargo during selective autophagy, this likely positions Atg20 closer to the cargo than Snx4. This overall arrangement of the Atg20-Snx4 complex could suggest a role for the Atg20-Snx4 complex in influencing the overall shape of the phagophore membrane during autophagy. Our data also demonstrate that the Atg20-Snx4 complex prefers to bind larger vesicles, suggesting a potential role in recognizing flat membrane sheets rather than highly curved membranes. Because autophagy is initiated with highly curved Atg9-containing vesicles, it is possible that the Atg20-Snx4 complex may not play a role until after these vesicles undergo fusion to form a double-membrane sheet. As such, Atg11-mediated early recruitment of the Atg20-Snx4 complex to the PAS may enable the complex to act on the earliest membrane sheets formed during autophagy.
Disclosure of potential conflicts of interest
No potential conflicts of interest were disclosed.