A Cryo-EM structure of human VPS34-CII with RAB5A showed a second RAB5 binding site on VPS15.

A. Design of VPS34 CII – RAB5 binding affinity measurement assay. RFP-trap beads were coated with mCherry-tagged VPS34-CII, incubated with various concentrations of AF488-labelled RAB protein and imaged by confocal microscopy. B. Quantification of RAB5-AF488 binding to VPS34-CII. Fluorescence intensity in the AF488 channel was normalised by the fluorescence intensity of mCherry channel. Plots are representative of three independent replicates for each experiment, and the reported Kd is average ± SD. Error bars in the plots are mean ± SD of the representative experiment. We verified that mCherry-tagging does not structurally compromise the VPS34 complexes (Fig. S1A) and that mCherry alone does not interact with RAB proteins (Fig. S1B). C-H. VPS34-CII has two RAB5-binding sites, one on VPS34 and the other on VPS15. The VPS34-CII REIE>ERIR reverse charge mutation eliminates RAB5A binding to the VPS34 site, but not the VPS15 RAB5-binding site. C. GUV-based assay with the wild-type VPS34-CII and VPS34-CII carrying a REIE>ERIR mutation in VPS34 C2 HH, in the presence and absence of membrane-tethered RAB5A. The REIE>ERIR mutation partially reduces VPS34-CII activation by RAB5A. Reaction progress curves are shown on the left, and initial rates on the right as bar graphs. All means are significantly different with p<0.05, except where not significantly different means are indicated by ‘n.s.’. Experiments were performed in triplicate; error bars show ±SD of the representative experiment. Micrographs show the AF647-PX signal at the end of reaction, scale bar = 5 µm. D. The REIE>ERIR mutation reduces RAB5A binding affinity compared to the wild-type VPS34-CII. The plot shows binding curves of RAB5A and VPS34-CII REIE>ERIR mutant together with the wild-type VPS34-CII + RAB5A binding curves. E. Composite cryo-EM map of VPS34-CII REIE>ERIR with RAB5A bound to VPS15 subunit. Icons of VPS34-CII in E-H are based on cartoons from BioRender F. Cartoon representation of the atomic model for VPS34-CII REIE>ERIR- RAB5A. G. A close-up view of the RAB5A binding site on VPS15. The RAB5A hydrophobic triad engages VPS15 helical solenoid domain. RAB5A residues F57 (yellow) and W74 (green) interact with VPS15 residue H580, RAB5A Y89 (orange) contacts VPS15 residues S579 and H580. H. Atomic models of four different VPS34-CII states obtained from a cryo-EM sample of the WT VPS34-CII + RAB5A. Number of particles used for reconstruction of each state is given below the models. VPS34-CII was observed in apo state, with RAB5A bound to VPS34 C2 HH, with RAB5A bound to VPS15, or with RAB5A molecules bound to each site at the same time.

The VPS15 RAB5 binding site of VPS34-CII is necessary for full activity and is likely the primordial RAB5-binding site.

A. The VPS15 SHMIT>DDMIE mutation partially reduced VPS34-CII activation by RAB5A. Bar graphs show initial rates from the GUV-based assay with the wild-type VPS34-CII and VPS34-CII carrying a 579-SHMIT-583>DDMIE mutation in the VPS15 helical solenoid (VPS15 mutant), in the presence and absence of membrane-tethered RAB5A. All means are significantly different with p<0.05, except where non significantly different means are indicated by ‘n.s.’. Error bars show ±SD of the representative experiment. Experiments were performed in triplicate with mCherry-tagged VPS34-CII. Reaction curves for the assay are shown in Figure S12B. B. Binding curves for RAB5A interaction with VPS34-CII VPS15 mutant and with the wild-type VPS34-CII + RAB5. Micrographs are in Figure S1G. C. The double mutation eliminates VPS34-CII activation by RAB5A, as shown by a GUV-based assay with the WT VPS34-CII and VPS34-CII carrying a double mutation (VPS34 C2 HH REIE>ERIR + VPS15 SHMIT>DDMIE). Reaction progress curves are shown on the left, and initial rates on the right as bar graphs. All means are significantly different with p<0.05, except where non significantly different means are indicated by ‘n.s.’. Error bars are ±SD of the representative experiment. Micrographs show the AF647-PX signal at the end of reaction, scale bar = 4 µm. Experiments were performed in triplicate with mCherry-tagged VPS34-CII. D. RAB5A does not bind VPS34-CII double mutant. Left: micrographs of RFP-trap beads coated with mCherry-VPS34-CII double mutant and incubated with increasing concentrations of RAB5A-AF488. Scale bar: 20µm. Right: binding curve for RAB5A-AF488 binding to mCherry-VPS34-CII double mutant measured from images shown on the left. E. The RAB5A binding site in VPS15, but not in VPS34 is conserved from humans to yeast. Left: multiple sequence alignment of VPS34 RAB binding site sequences. Right: multiple sequence alignment of VPS15 RAB5A binding site sequences. Identical residues are shown as white letters in red background. Red letters denote residues with similar properties. Structural elements shown above are based on VPS34 complex models from this study. F. Yeast Vps15 carrying a mutation in the RAB5 binding site does not show strong colocalisation with a GTP-locked form of the yeast RAB5A orthologue, Vps21 (Q66L). C-terminally EGFP-tagged Vps15 (WT or mutant) was transformed into a vps15Δ strain integrated with tandem-tomato-tagged Vps21 (Q66L) and subjected to light microscopy. Scale bars: 5 µm. G. Vps15 carrying a mutation in the RAB5 binding site is defective in CPY sorting. C-terminally 3 x FLAG tagged Vps15 (WT or mutant) was transformed into vps15Δ cells, and cellular and medium fractions (“cell” and “medium”, respectively) are subjected to western blotting. Left: Western blotting for CPY, FLAG for Vps15, and Pgk1. *: non-specific cross-react. Right: Quantification of mature (m) and precursor (p2) forms from triplicated experiments.

Comparison of VPS34-CI-RAB1A and VPS34-CII-RAB5A binding modes.

Left panels show features of VPS34-CI-RAB1A (PDB: 9MHG). Right panels show features of VPS34-CII-RAB5A. Icons of VPS34 complexes are based on cartoons from BioRender. A. RAB1A and RAB5A bind to the equivalent site in both complexes. B. RAB1A binds both helices of the VPS34 C2 HH. Switch 1 region (yellow) binds helix 2, while Switch 2 (orange) binds helix 1. RAB5A contacts only helix 2 from VPS34 C2 HH. The 9 Å distance between RAB1A and VPS15 SGD is comparable to the 7 Å distance between RAB5A and VPS15 SGD. The VPS15 WD40-RAB distance is larger in VPS34-CI-RAB1A (27 Å) than in VPS34-CII-RAB5A (9 Å). Distances were measured between the α-carbon atoms in selected VPS15 and RAB residues: VPS15 S776 and RAB1A T130 / RAB5A N139 for RAB-SGD distance; VPS15 G1298 and RAB1A T59 / RAB5A T68 for RAB-WD40 distance. The RAB residues are equivalent according to MSA. C. Hydrophobic triad residues in RAB1A and RAB5A bind VPS34 C2 HH in different modes. In RAB1A, residue F (yellow) contacts C2 HH helix 2, residue Y (orange) engages C2 HH helix 1, and residue W (green) is not involved in direct contact with the C2 HH. By contrast, in RAB5A, it is residue F (yellow) that is not involved in binding C2 HH, while both residues W (green) and Y (orange) engage the same VPS34 C2 HH helix 2. VPS34 C2 HH helix 2 residue R195 interacts with hydrophobic triad residues in both RAB1A and RAB5A. VPS34 residue R183 in helix 1 is involved in binding RAB1A, but not RAB5A. D. Switch 1 region (yellow) interacts with VPS34 C2 HH helix 1 in both VPS34 complex-RAB assemblies. RAB1A residue D47 forms a salt bridge with VPS34 residue R199. RAB5A Switch 1 has limited contact with VPS34 C2 HH helix 1. E. Switch 2 region (orange) adopts different conformations in VPS34-CI -RAB1A and VPS34-CII-RAB5A. In RAB1A, Switch 2 is in a loop conformation, embracing VPS34 residue R183 in VPS34 C2HH helix 1 that in return stabilizes the loop via polar contacts to RAB1A S78 and the backbone carbonyls of A68 and F73. In RAB5A, Switch 2 is in a helical conformation potentially stabilised by a hydrophobic interaction between RAB5A residues Y90 and P87. Switch 2 makes minimal contact with C2HH helix 2 through residue M88 that interacts with F198 in VPS34.

Structure of the N-terminal region of BECLIN1 and UVRAG.

Upper: The N-terminal region of BECLIN1 (blue) is stabilized by a zinc finger (Zn2+ grey), and this region together with the long helical hairpin in the VPS15 solenoid region sandwich the UVRAG C2 domain. Two basic residues (green spheres) comprise a proposed PI3P-binding region (He et al., 2013). RAB5 on VPS15 is indicated with an orange molecular surface. The BH3 helix of BECLIN1 is colored magenta. Lower: linear domain compositions of VPS15, BECLIN1, and UVRAG. The colors correspond to equivalent domains in the cartoon.

RAB binding by wildtype and mutant VPS34 complexes

A. mCherry tagging does not compromise VPS34 complex structural integrity. nanoDSF melting curves for VPS34-CII WT, Tm=47.9°C (blue) and mCherry-tagged VPS34-CII Tm=48.1°C (red). B-G. Micrographs of RFP-trap bead based binding assays for RAB binding to mCherry-tagged VPS34 complexes. Top row shows mCherry channel fluorescence intensity, confirming mCherry-tagged protein immobilisation on RFP-trap beads. Middle row shows AF488 channel intensity from AF488-RAB proteins. Bottom row shows the brightfield images of beads. Scale bars in all micrographs are 20 µm. B. mCherry protein immobilised on RFP-trap beads does not bind RAB1A or RAB5A. C. RAB5A binding to WT VPS34-CII. D. RAB1A binding to WT VPS34-CII is too weak to be quantified. Left: micrographs of the binding assay. Right: fluorescence intensity of the AF488 channel shown in micrographs on the left was normalised by the intensity of the mCherry channel and plotted against the concentration of AF488-RAB1A. E. RAB5A binding to VPS34-CII carrying a 199-REIE-202 > 199-AAAA-202 mutation in VPS34. F. RAB5A binding to VPS34-CII carrying a 199-REIE-202 > 199-ERIR-202 mutation in VPS34. G. RAB5A binding to VPS34-CII carrying a 579-SHMIT-583 > 579-DDMIE-583 mutation in VPS15.

Analysis of the cryo-EM structure of VPS34-CII with the VPS34 199-REIE-202 > 199-AAAA-202 mutation bound to RAB5A.

Icons of VPS34-CII are based on cartoons from BioRender. A. A composite cryo-EM map of VPS34-CII REIE>AAAA with RAB5A bound to VPS34 C2 HH. B. Cartoon representation of the atomic model for VPS34-CII REIE>AAAA bound to RAB5A. Model was built using PDB depositions of VPS34 kinase domain (VPS34 residues 539-887; PDB: 3IHY) and RAB5A (PDB: 1R2Q), AlphaFold Multimer predictions of VPS15, VPS34 C2 domain (VPS34 residues 1-233), Beclin1 CC2-BARA domains (Beclin1 residues 172-269 for CC2 domain, 270-449 for BARA domain), UVRAG CC2-BARA2-CTD domains (UVRAG residues 240-333 for CC2 domain, 334-464 for BARA2 domain, and 465-699 for CTD), and AlphaFold3 prediction of Beclin1 CC1 domain (Beclin1 residues 139-171) and UVRAG CC1 domain (UVRAG residues 195-226). C. Overlay of single particle cryo-EM derived model of VPS34-CII REIE>AAAA-RAB5A (dark grey) and cryo-ET derived model of membrane bound VPS34-CII-BATS-RAB5A (light grey; PDB: 7BL1). Models are aligned on VPS34 helical and kinase domains (VPS34 residues 292-887). D. Tripartite site of RAB5A binding to VPS34-CII REIE>AAAA. RAB5A is situated close to VPS15 SGD (VPS15 residues 723-805) and WD40 domains (VPS15 residues 968-1358), and VPS34 C2 HH (VPS34 residues 170-210). RAB5A Switch 1 (yellow) and Switch 2 (orange) regions are oriented towards VPS34 C2 HH helix 2 (VPS34 residues 191-210) that carries the REIE>AAAA mutation (brown). VPS34 C2 HH helix 1 (VPS34 residues 170-184) does not contact RAB5A. E. Close-up of RAB5A Switch 1 (yellow) and VPS34 C2 HH helix 2 (blue) interaction, with other elements in the binding site coloured grey. RAB5A Switch 1 is rich in aliphatic and hydrophobic side chains that interact favourably with the charged to non-polar type REIE>AAAA mutation (brown) in VPS34 C2 HH helix 2.

HDX-MS shows protection in the two RAB5A binding regions in WT VPS34-CII.

Icons of VPS34-CII are based on cartoons from BioRender. A. RAB5A binding strongly protects the two RAB5A binding sites observed in cryo-EM structure: the VPS34 C2 HH helix 2 (highlighted in blue) and VPS15 helix 570-585 (highlighted in pink). Changes in deuteration are displayed as a relative change between VPS34-CII-RAB5A and apo VPS34-CII. The changes are mapped on a model of WT VPS34-CII apo. B. A close-up of the VPS34 binding site for RAB5A on VPS34-CII. In addition to the VPS34 C2 HH helix 2, RAB5A binding protects other nearby elements: VPS15 SGD, VPS15 linker, UVRAG CC2, and VPS34 C2 domain. Deuterium uptake plots are shown for these regions. C. A close-up of the VPS15 binding site for RAB5A on VPS34-CII. RAB5A binding strongly protects VPS15 helix 570-585 and VPS15 helical solenoid residues 446-462 and 643-655. Beclin1 CC1 is also protected. Deuterium uptake plots are shown for these regions.

nanoDSF experiments show that VPS34-CII mutants that disrupt RAB5 binding are structurally intact.

A. Mutations to modify RAB5A binding to VPS34-CII either do not affect or increase the stability of VPS34-CII. The VPS34 199-REIE-202 > 199-AAAA-202 and VPS15 579-SHMIT-583 > 579-DDMIE-583 mutation in VPS34-CII did not affect the melting temperature of the complex. The VPS34 199-REIE-202 > 199-ERIR-202 mutation increased the melting temperature of VPS34-CII by 2.6°C. B. Plots shown in A with a smaller temperature range on the x-axis for closer curve inspection.

The RAB5 binding mutation in Vps15 in yeast VPS34-CII abolishes binding to Vps21, a yeast RAB5 orthologue.

A. Micrographs of RFP-trap bead-based binding assays for Vps21 Q66L-EGFP binding to mCherry-tagged yeast VPS34-CII (ScVPS34-CII). Top two rows: Vps21 Q66L-EGFP recruitment to WT ScVPS34-CII. Bottom two rows: Vps21 Q66L-EGFP recruitment to ScVPS34-CII carrying RAB5 binding mutant. Micrographs are representatives from three independent experiments. Scale bars: 100 μm. B. Quantification of A. Representative quantifications from three independent experiments are shown. KD: Average ± SD from three independent experiments.

VPS34-CII with and without RAB5A has minimal conformational differences compared to other VPS34-CII structures.

Icons of VPS34-CII are based on cartoons from BioRender. A. Cryo-EM reconstruction of WT VPS34-CII apo form at 3.9 Å resolution. Particles contributing to this map do not have RAB5A bound and are a subset of all particles in the VPS34-CII + RAB5A sample (see materials and methods and Figure S11C). B. Cryo-EM reconstruction of WT VPS34-CII-RAB5A at 4.0 Å resolution. RAB5A is bound to VPS34 at the same site as in VPS34-CII REIE>AAAA mutant (see Figure S2A). C. Overlay of atomic models of WT VPS34-CII apo (light grey), and WT VPS34-CII-RAB5A (dark grey). Models are aligned on the VPS15 helical solenoid (VPS15 residues 315-802), with RMSD of 3.3 Å across whole models. The conformation of the complex does not change substantially upon RAB5A binding. D. Overlay of the atomic models of WT VPS34-CII-RAB5A (light grey) and VPS34-CII REIE>AAAA-RAB5A (dark grey). Models are aligned on the VPS15 helical solenoid domain (VPS15 residues 315-802), with RMSD of 2.5 Å across whole models. E. Overlay of the atomic models of WT VPS34-CII-RAB5A (light grey), and cryo-ET derived model of VPS34-CII-BATS-RAB5A on the membrane (dark grey). Models were aligned on VPS34 kinase domain (VPS34 residues 539-887), and the backbone RMSD between VPS34, BECLIN1, UVRAG, and VPS15 WD40 domains from the two complexes is 7.0 Å. RMSD for VPS15 pseudokinase, helical solenoid, and SGD domains could not be measured accurately due to corrected VPS15 sequence register in the single particle EM-derived model of WT VPS34-CII-RAB5A.

Analysis of the cryo-EM structure of VPS34-CI- RAB1A (PDB 9MHG).

Icons of VPS34-CI are based on cartoons from BioRender. A. Close-up of RAB1A hydrophobic triad (F48 from Switch 1, W65 from Interswitch, and Y80 from Switch 2) with associated cryo-EM density. RAB1A residue W65 does not interact with VPS34-CI and instead stabilises the structure of RAB1A hydrophobic triad via interactions with RAB1A residues Q63 (from Interswitch region) and K13 (from outside the Switch regions). B. VPS34 residue R183 potentially makes hydrogen bonds with RAB1A residue S78 and the backbone carbonyls from RAB1A A68 and F73. Distances between hydrogen bond donor and acceptor atoms were measured with UCSF ChimeraX and fall within the range of distances of 2.2-4.0 Å compatible with hydrogen bonding (Jekrey, 1997).

Multiple sequence alignment of selected human RAB GTPases and their isoforms.

Top two rows show RAB5A and RAB1A sequences for ease of comparison. Switch 1 and Switch 2 regions are highlighted with light blue background. Hydrophobic triad residues are indicated with arrows and framed in matching colour backgrounds. In Switch 1, there is an A > D change between RAB5A and RAB1A. In Switch 2, there is a P > S change between RAB5A and RAB1A. Residues in red background and white letters are identical; residues in red letters have conserved properties. The secondary structure elements depicted above the MSA are based on the crystal structure of RAB5A (PDB: 1N6H).

VPS34-CII has a zinc finger at BECLIN1 N-terminus.

Cryo-EM density of WT VPS34-CII-RAB5A at the BECLIN1 N-terminus. The zinc ion is coordinated by BECLIN1 residues C18 and C21, and C137 and C140. Created with BioRender.

Cryo-EM densities show that the N-terminus of VPS15 is myristoylated and VPS15 is bound to a GDP nucleotide.

Icon of VPS34 complex is based on a cartoon from BioRender. A. VPS15 residue G2 is myristoylated (magenta). The myristoyl is inserted into a pocket formed of VPS15 N-lobe residues F55, A99, L101, L64, Y67 (purple), as well as P-loop residues F37 and F38 (green), and activation loop residues F186 and Y187 (blue). GDP nucleotide is close to the myristoyl pocket. VPS15 activation and P-loops are close to VPS34 helix kα10 (dark blue), but there is no contact with VPS34 catalytic loop (orange) or activation loop (light blue). The model is based on the high-resolution structure of VPS34-CII REIE>AAAA-RAB5A. B. Extra density in the VPS15 pseudokinase domain (VPS15 residues 24-315) is consistent with presence of a GDP nucleotide and a Mg2+ ion. VPS15 R103 contacts the GDP, and D166 coordinates the Mg2+ ion.

Single particle cryo-EM data processing workflow for VPS34-CII REIE>AAAA-RAB5A.

Single particle cryo-EM data processing workflow for VPS34-CII REIE>ERIR-RAB5A.

Single particle cryo-EM data processing workflow for WT VPS34-CII - RAB5A.

GUV-based kinase assay reaction curves with error bars for results shown in main figures.

A. Reaction curves for the assay shown in Fig. 1C. B. Reaction curves for the assay shown in Fig. 2A. C. Reaction curves for the initial rate quantification shown in Fig. 2C.