Bidirectional Control of Autophagy by BECN1 BARA Domain Dynamics

Abstract

Membrane targeting of the BECN1-containing class III PI 3-kinase (PI3KC3) complexes is pivotal to the regulation of autophagy. The interaction of PI3KC3 complex II and its ubiquitously expressed inhibitor, Rubicon, was mapped to the first β sheet of the BECN1 BARA domain and the UVRAG BARA2 domain by hydrogen-deuterium exchange and cryo-EM. These data suggest that the BARA β sheet 1 unfolds to directly engage the membrane. This mechanism was confirmed using protein engineering, giant unilamellar vesicle assays, and molecular simulations. Using this mechanism, a BECN1 β sheet-1 derived peptide activates both PI3KC3 complexes I and II, while HIV-1 Nef inhibits complex II. These data reveal how BECN1 switches on and off PI3KC3 binding to membranes. The observations explain how PI3KC3 inhibition by Rubicon, activation by autophagy-inducing BECN1 peptides, and inhibition by HIV-1 Nef are mediated by the switchable ability of the BECN1 BARA domain to partially unfold and insert into membranes.

Fig. 1.. Mapping the Rubicon PIKBD.

(A) Schematic diagram of the predicted secondary structure of Rubicon (upper). MBP tagged full length Rubicon or deletion mutants were co-expressed with the subunits of PI3KC3-C2 (GST tagged UVRAG; Strep tagged VPS34, VPS15 and BECN1) in HEK293 cells. Both MBP and GST pull down were performed. The resulting precipitates were visualized by SDS-PAGE (lower). GST pull-down precipitates were represented as input. FL, full length; MR, middle region; NT, N terminus; CT, C terminus; MRN, middle region N terminus; MRC, middle region C terminus; PIKBD, PI3KC3-binding domain. (B) Size exclusion chromatography of PI3KC3-C2 complexes. Different complexes are indicated with color codes at right. The peak fraction of each complex was used for subsequent activity assays. UV, ultraviolet. (C) Activities of PI3KC3-C2 complexes (12.5 nM) on SUVs containing PI. Different complexes are indicated with color codes. RLU, relative light units. (D) Confocal images of GUVs showing the binding of FYVE domain (green) and different PI3KC3 complexes (red). GUVs were incubated with wild type C2, Rubicon FL-C2, or Rubicon PIKBD-C2 complex respectively. Scale bars, 5 μm. (E) MBP tagged full length Rubicon or PIKBD were co-expressed with the subunits of PI3KC3-C1 or C2 in HEK293 cells. The MBP pull-down precipitates were visualized by SDS-PAGE.

Fig. 2.. Mapping binding sites of Rubicon and PI3KC3-C2 by HDX-MS.

(A) Difference plot of percent deuteron incorporation in BECN1 of PI3KC3-C2 versus deuterons incorporated in the presence of Rubicon, 30s in D₂O. (B) Difference plot of percent deuteron incorporation of UVRAG of PI3KC3-C2 versus deuterons incorporated in the presence of Rubicon, 30s in D₂O. (C) HDX-MS difference map onto the crystal structure of human BECN 1 BARA domain (wheat) and crystal structure of yeast PI3KC3-C2 (grey), BECN1 BARA domain was aligned to the yeast crystal structure. Regions showing a decrease in HDX are depicted in blue (10% decrease) and lavender (5% decrease) in the presence of Rubicon. (D) BECN1 or mutants were co-expressed with full length MBP-tagged Rubicon and the other subunits of PI3KC3-C2. Both MBP and GST pull-down precipitates were visualized by SDS-PAGE. (E) Difference plot of %D incorporated into Rubicon PIKBD within the PI3KC3-C2 complex versus deuterons incorporated into Rubicon PIKBD alone after 10 s in D₂O. The Rubicon PIKBD secondary structure drawing (upper) illustrates the heat map of apo state. Three predicted α-helices (α1-α3) are colored according to %D exchange. Data are presented as Mean ± Stddev, n=3. (F) MBP tagged Rubicon PIKBD or PIKBD mutants were co-expressed with the subunits of PI3KC3-C2 in HEK293 cells. Both MBP and GST pull-down precipitates were visualized by SDS-PAGE. See also Figure S1 and S2.

Fig. 3.. Rubicon PIKBD inhibits PI(3)P production and autophagic flux in vivo.

(A) Distribution of GFP-FYVE (green) and Rubicon fragment (red) in U2OS cells. Cells transfected with Flag vector, Flag-Rubicon, Flag-Rubicon-PIKBD, or Flag-Rubicon-PIKBD mutant were starved with EBSS for 2 h. Scale bars, 10 μm. (B) Quantification of the number of GFP-FYVE puncta per cell in cells treated as in D. The data are quantified with Image J, and analyses were performed using GraphPad Prism 5 software. Data are presented as Mean ± Stddev, n = 50. ***p < 0.001. (C) The representative Western Blot for autophagic flux assay in Rubicon-KO cells with anti-LC3 antibody. After the transfection of mStrawberry (mSt), mSt-Rubicon, or mSt-Rubicon mutant, cells were starved with EBSS in the presence or absence of Bafilomycin A1(Baf A1). (D) Quantification of autophagic flux in (C). The data are quantified using Image J, and analyses were performed using GraphPad Prism 5 software. Comparisons between groups were performed using the one-way analysis of variance. Data are denoted as Mean ± Stddev, n=3. p values ≤ 0.05 were considered significant (*). Anti-RFP antibody and ponceau S (Po-S) were used to confirm the transfection efficacies of each plasmid and the loading protein amounts, respectively.

Fig. 4.. Cryo-electron microscopy on PI3KC3-C2:Rubicon PIKBD.

(A) Negative stain on PI3KC3-C2 MBP-Rubicon PIKBD and model for domain placement. (B) 2D-class averages of PI3KC3-C2:MBP-Rubicon PIKBD by cryo-EM. (C) Model fit of 6.8 Å map, BECN1 in lime green, helix corresponding to Rubicon PIKBD, UVRAG in pink, VPS15 in magenta, and VPS34 in forest green. (D) Inset of the BECN1 BARA domain where Rubicon PIKBD binds. (E) MBP tagged Rubicon PIKBD α1 or α2&3 were co-expressed with the subunits of PI3KC3-C2 in HEK293 cells. Both MBP and GST pull-down precipitates were visualized by SDS-PAGE. (F) Confocal images of GUVs showing the binding of FYVE domain (green) and different PI3KC3 complexes (red). GUVs were incubated with C2, PIKBD-C2, or PIKBD α1-C2 complex respectively. Scale bars, 5 μm. See also Figure S3-S6, Table S1 and Movie S1.

Fig. 5.. Membrane docking by BECN1 BARA β-sheet 1.

(A) The OH overlap helix (purple) of BECN1, which is incompatible with the full-length BECN1 subunit, is occupied by a longer helix corresponding to the Rubicon PIKBD (blue). BECN1 BARA domain has three β-α repeats (orange, lime green, and yellow), in the second β-α repeat (lime green) there are three aromatic residues (red) that constitute the aromatic finger necessary for membrane binding. Rubicon-PIKBD (blue) binds to amino acids Trp277 and His278 (orange) located on the first β-α repeat in the BARA domain. Residue register comes from a 1.4 Å crystal structure of the BECN1 BARA domain (4DDP). (B) Model for membrane docking by BECN1 BARA domain. The BECN1 BARA domain is located at the tip of the regulatory “left hand” arm of the PI3KC3-C2 complex, where it docks onto membranes, promoting access of the kinase to its lipid substrate. One mode of membrane insertion is through the aromatic finger, the second occurs when the β-sheet of the first β-α repeat flips from its hydrophobic pocket in order bind the membrane. Beta sheet 1 mutants are incapable of binding membrane, leading to reduction in PI3KC3 activity. Representative snapshots of MD simulations of membrane-anchored Beclin BARA domain with locked beta-sheet 1 (C) or unlocked beta-sheet 1 (D). BECN1 BARA domain in cyan, locked (C) or unlocked (D) beta-sheet 1 (residues 265–287) in red, phosphate atoms of the upper membrane leaflet depicted as golden spheres. (E) Confocal images of GUVs showing the binding of FYVE domain (green) and different PI3KC3 complexes (red). GUVs were incubated with C2, BECN1^BS1PL-C2, or BECN1^DDD-C2 complex respectively. (F) Quantitation of the reaction kinetics on the membrane from individual GUV tracing in (C) (see Methods, Mean ± Stddev; C2 N = 13, BECN1^BS1PL-C2 N = 7, BECN1^DDD-C2 N = 8). (G) Confocal images of GUVs showing the binding of FYVE domain (green) and different PI3KC3 complexes (red). GUVs were incubated with C2, BECN1^FF→SS-C2, or BECN1^WH→DD-C2 complex respectively. (H) Quantitation of the reaction kinetics on the membrane from individual GUV tracing in (G) (Mean ± Stddev; C2 N = 20, BECN1^FF→SS-C2 N = 12, BECN1^WH→DD-C2 N = 20). Scale bars, 5 μm. See also Movie S2.

Fig. 6.. BECN1 autophagy-activating peptide promotes PI3KC3 membrane binding in vitro.

(A and B) Activities of 25 nM PI3KC3-C2 (A) or PI3KC3-C1 (B) on SUVs in the absence and presence of Tat BECN1 peptide. Data are presented as Mean ± Stddev, n = 3. ***p < 0.001. (C) Activities of 25 nM PI3KC3-C2 or BECN1^BS1PL-C2 on SUVs in the absence and presence of Tat BECN1 peptide. Data are presented as Mean ± Stddev, n = 3. ***p < 0.001. (D and E) Confocal images of GUVs showing the binding of FYVE domain (green) and PI3KC3-C2 or PI3KC3-C1 (red) in the absence and presence of BECN1 peptide. Scale bars, 10 μm. (F) Model for PI3KC3 activation by BECN1 peptide. An allosteric switch in beta sheet 1 facilitates this sequence to flip out and bind membranes. This equilibrium favors beta sheet 1 to occupy its hydrophobic pocket over the flipped out state, however, in the presence of T-BP, beta sheet 1 flips out, binds membrane and then activates PI3KC3, as T-BP is present to occupy this hydrophobic pocket.

Fig. 7.. HIV-1 Nef inhibits PI3KC3-C2 in vitro.

(A) Sequence alignment for HIV-1 Nef, Rubicon and Pacer. Completely conserved regions among the proteins are colored bright and predominantly conserved regions are enclosed in blue boxes. Residues that were included in the Rubicon mutant that knocked out binding and was used in the cell studies are underlined. (B and C) Activities of 25 nM PI3KC3-C2 (B) or PI3KC3-C1 (C) on SUVs in the absence and presence of HIV-1 Nef. Data are presented as Mean ± Stddev, n = 3. ***p < 0.001. (D) Confocal images of GUVs showing the binding of FYVE domain (green) and PI3KC3-C2 (red) in the presence of wild type Nef and a Nef mutant in the putative C2 binding site (“Nef^Mu”, conserved DLEK to AAAA). Scale bars, 10 μm. (E) Quantitation of the reaction kinetics on the membrane from individual GUV tracing in (D) (Mean ± Stddev; C2 N = 12, C2 + Nef N = 7, C2 + Nef^Mu N = 8). (F) Schematic for bidirectional regulation of PI3KC3 in normal physiology, infection, and therapeutic intervention.