Characterization of Atg38 and NRBF2, a fifth subunit of the autophagic Vps34/PIK3C3 complex

Abstract

The phosphatidylinositol 3-kinase Vps34 is part of several protein complexes. The structural organization of heterotetrameric complexes is starting to emerge, but little is known about organization of additional accessory subunits that interact with these assemblies. Combining hydrogen-deuterium exchange mass spectrometry (HDX-MS), X-ray crystallography and electron microscopy (EM), we have characterized Atg38 and its human ortholog NRBF2, accessory components of complex I consisting of Vps15-Vps34-Vps30/Atg6-Atg14 (yeast) and PIK3R4/VPS15-PIK3C3/VPS34-BECN1/Beclin 1-ATG14 (human). HDX-MS shows that Atg38 binds the Vps30-Atg14 subcomplex of complex I, using mainly its N-terminal MIT domain and bridges the coiled-coil I regions of Atg14 and Vps30 in the base of complex I. The Atg38 C-terminal domain is important for localization to the phagophore assembly site (PAS) and homodimerization. Our 2.2 Å resolution crystal structure of the Atg38 C-terminal homodimerization domain shows 2 segments of α-helices assembling into a mushroom-like asymmetric homodimer with a 4-helix cap and a parallel coiled-coil stalk. One Atg38 homodimer engages a single complex I. This is in sharp contrast to human NRBF2, which also forms a homodimer, but this homodimer can bridge 2 complex I assemblies.

Keywords: Atg14; Atg38; Beclin 1; EM structure; NRBF2; Vps15; Vps30; Vps34; complex I; crystal structure.

Figure 1.

Stable heterotetrameric and heteropentameric assemblies of yeast complex I. (A) SEC-MALS analysis of purified recombinant heterotetrameric S. cerevisiae complex I on an S200 10/30 gel filtration column. The inset shows SDS-PAGE of the purified complex stained with InstantBlue. (B) SEC-MALS analysis of full-length Atg38 (red trace) and the Atg38 MIT domain (residues 1 to 78) (black trace) on a S75 10/30 column. The mass of full-length Atg38 is consistent with a dimer. The mass of the MIT domain alone is consistent with a monomer (calculated mass 9 kDa). (C) Purified heterotetrameric complex I was mixed with an excess of purified Atg38 and analyzed by gel filtration on a S200 10/30 column (red trace). For comparison, purified Atg38 alone was run on the same column (blue trace). The fractions were run on a SDS-PAGE and the gel was stained with InstantBlue (right). Note that the elution volumes for (A) and (C) differ since 2 different S200 columns were used. Inset: starting material of complex I+Atg38 (CI+Atg38). Also, complex I alone (CI) and Atg38 alone before mixing are loaded as indicated. (D) SEC-MALS analysis of complex I + Atg38 on a S200 10/30 column. a.u., arbitrary units; mAu, milli Absorbance units.

Figure 2.

Stable heterotetrameric and heteropentameric assemblies of human complex I. (A) SDS gels of purified human heterotetrameric (T) and heteropentameric (P) complexes coexpressed in HEK293T cells. (B) SEC-MALS profile of recombinantly expressed NRBF2, confirming it forms a dimer of 67 kDa. (C) IgG affinity isolation of complexes. The heteropentameric complex was reconstituted in vitro starting with a HEK293T lysate expressing the heterotetrameric complex (left) to which excess NRBF2 was added (right) (W: wash, IgG: beads, E1: elution fraction 1 after TEV cleavage from the IgG resin). (D) Glycerol gradient profiles of human complexes (left and middle gels). Both have comparable migration profiles, indicating a similar size. Right gel: glycerol gradient profiles of standard markers. 669 kDa: Thyroglobulin; 443 kDa: Apoferritin; 200 kDa: β-Amylase; 66 kDa: Bovine serum albumin. (E) Comparison of electron microscopy class averages of the heterotetrameric (top row) and heteropentameric complex (bottom row) embedded in negative stain show that the overall V-shaped architecture of both complexes is very similar. In particular views, there is additional density of NRBF2 detectable at the base of the complex, as confirmed by the difference map below. a.u., arbitrary units.

Figure 3.

Mapping the Atg38-complex I interface by HDX-MS. (A to E) Changes in HDX between heterotetrameric complex I, complex I+Atg38 and free Atg38. (A) Differences in HDX for Atg38 between free Atg38 and Atg38 in the heteropentameric complex I. A model for the N-terminal MIT domain is illustrated below with HDX differences mapped onto it. (B to E) Differences in HDX between heterotetrameric complex I and heteropentameric complex I for Atg14 (B), for Vps30 (C), for Vps15 (D) and for Vps34 (E). (F) GST affinity isolation experiments with various deletion variants of Atg38.

Figure 4.

Crystal structure of the C-terminal domain of Atg38. (A) Overall structure of the homodimer. (B) Close-up view showing the interactions of helix A with the straight helix B in the cap region (upper panel) and the asymmetric A-B contacts in the cap (lower panel). (C) Close-up view of the intersubunit interactions in the stalk. (D) Helix-helix interactions between subunits result in decreased global HDX-MS, while the linker connecting the MIT domain to the C-terminal domain shows rapid exchange consistent with intrinsic disorder.

Figure 5.

Effect of Atg38 C-terminal truncations on cellular localization. (A) Schematic representation of C-terminal truncations of Atg38. Genomic DNA fragments containing the ATG38 promoter region (-997 base pairs) and a deletion series of ATG38 were C-terminally fused to an EGFPx2 coding fragment on a CEN plasmid (pRS416). (B) Micrographs of GFP and differential interference contrast (DIC) channels of the Atg38 truncations. The plasmids carrying the truncations described in (A) were transformed into atg38Δ cells. Cells were grown to mid-log phase (OD₆₀₀ = 0.8–1.0), then shifted to SD(-N) for 4 h, then subjected to microscopy. Scale bars: 5 µm. (C) Numbers of dots (%) were counted from triplicates of 100 cells/construct ± SD. The graph is a representative of 2 independent experiments. ****, P < 0.0001 compared to full length (FL); NS, Not Significant (P > 0.05) compared to FL.

Figure 6.

Roles of the Atg38 C-terminal domain in dimerization, interaction with complex I and cellular fractionation. (A) All Atg38 C-terminal truncations cause overexpression of the Atg38 constructs. Total cell lysates from growing and nitrogen starvation conditions were subjected to western blotting. The same C-terminal deletion fragments as used in Fig. 5 were C-terminally tagged with 3X FLAG on a CEN plasmid (pRS416). The plasmids were transformed into atg38Δ cells, and grown to mid-log phase (OD₆₀₀ = 0.8–1.0; Growing), then shifted to SD(-N) for 4 h. Total cell lysates were prepared as described in Materials and Methods. Asterisks indicate proteins of interest. Pgk1 expression was used as a loading control. (B) Atg38 C-terminal truncations cause dissociation of the proteins from the P15 pellet. Subcellular fractionation of the same proteins as in (A). Cells were grown to mid-log phase (OD₆₀₀ = 0.8–1.0; Growing), then shifted to SD(-N) medium and the total cell lysates (T), and supernatant (S15) and pellet (P15) fractions were analyzed. Pho8 was used as a membrane fraction (P15) control. (C) The C-terminal, crystallized region is important for homodimerization of Atg38. Immunoprecipitation between Atg38-ZZ and the same Atg38 truncations used above during nitrogen starvation. Yeast strain with chromosomally ZZ-tagged ATG38 was transformed with the same plasmids as (A) and (B). Cells were grown to mid-log phase, then shifted to SD(-N) for 4 h. The truncated proteins were affinity isolated with anti-Flag antibody (IP) then, the interacting Atg38-ZZ was detected with anti-ZZ antibody). The Atg38 (Full length, FL, and deletion series) proteins were detected with anti-Flag antibody. (D) SDS PAGE analysis of Atg38 deletion variants expressed in E. coli. (E) SEC-MALS analysis of the deletion variants shown in (D) carried out on a Superdex 75 10/30 gel filtration column. Variants with a single segment deleted have masses that are consistent with dimers (capΔ, stalkΔ and tailΔ). Constructs with 2 or 3 segments deleted have monomeric masses (stalkΔ+tailΔ, capΔ+stalkΔ and C terminusΔ). The C terminusΔ construct consists of residues 123 to 226, corresponding to a calculated mass of 14 kDa. Both SEC-MALS and intact mass spectrometry (Fig. S17) are consistent with a mass of approximately 11 kDa, suggesting that there has been some proteolysis of the construct during expression or purification from E. coli. (F) The crystallized region is important for the binding to Atg14. Immunoprecipitation between Atg14-ZZ and the Atg38 deletion series during nitrogen starvation. The atg38Δ strain, in which the ATG14 gene was chromosomally integrated with ZZ-tag at its C terminus, was transformed with the same plasmids as above. Cells were grown to mid-log phase, then shifted to SD(-N) for 4 h. The Flag-tagged truncations were affinity isolated with anti-Flag antibody (IP) and the interacting ZZ-tagged protein in the immunoprecipitates was detected with an anti-ZZ antibody, while Atg38 was detected with anti-Flag antibody. (G) Summary of in vivo experiments with Atg38 truncation variants.