The G-Protein Rab5A Activates VPS34 Complex II, a Class III PI3K, by a Dual Regulatory Mechanism

Abstract

VPS34 complex II (VPS34CII) is a 386-kDa assembly of the lipid kinase subunit VPS34 and three regulatory subunits that altogether function as a prototypical class III phosphatidylinositol-3-kinase (PI3K). When the active VPS34CII complex is docked to the cytoplasmic surface of endosomal membranes, it phosphorylates its substrate lipid (phosphatidylinositol, PI) to generate the essential signaling lipid phosphatidylinositol-3-phosphate (PI3P). In turn, PI3P recruits an array of signaling proteins containing PI3P-specific targeting domains (including FYVE, PX, and PROPPINS) to the membrane surface, where they initiate key cell processes. In endocytosis and early endosome development, net VPS34CII-catalyzed PI3P production is greatly amplified by Rab5A, a small G protein of the Ras GTPase superfamily. Moreover, VPS34CII and Rab5A are each strongly linked to multiple human diseases. Thus, a molecular understanding of the mechanism by which Rab5A activates lipid kinase activity will have broad impacts in both signaling biology and medicine. Two general mechanistic models have been proposed for small G protein activation of PI3K lipid kinases. 1) In the membrane recruitment mechanism, G protein association increases the density of active kinase on the membrane. And 2) in the allosteric activation mechanism, G protein allosterically triggers an increase in the specific activity (turnover rate) of the membrane-bound kinase molecule. This study employs an in vitro single-molecule approach to elucidate the mechanism of GTP-Rab5A-associated VPS34CII kinase activation in a reconstituted GTP-Rab5A-VPS34CII-PI3P-PX signaling pathway on a target membrane surface. The findings reveal that both membrane recruitment and allosteric mechanisms make important contributions to the large increase in VPS34CII kinase activity and PI3P production triggered by membrane-anchored GTP-Rab5A. Notably, under near-physiological conditions in the absence of other activators, membrane-anchored GTP-Rab5A provides strong, virtually binary on-off switching and is required for VPS34CII membrane binding and PI3P production.

Significance

Class III phosphatidylinositol-3-kinases (PI3Ks) are lipid kinases localized to intracellular membranes, where they are activated by signals including small G proteins and produce the signaling lipid phosphatidylinositol-3-phosphate (PI3P). Subsequently, the lipid signal initiates key events during endocytosis and endosome development. This single-molecule study of a reconstituted class III PI3K pathway reveals that G protein activates the lipid kinase via a dual mechanism involving both membrane recruitment and allosteric activation. To our knowledge, this is the first mechanistic description of class III PI3K regulation by a G protein. The findings significantly extend our molecular understanding of a core regulatory network central to normal cell function and strongly linked to multiple human diseases, yielding impacts for signaling biology and medicine.

Introduction

A diverse array of cellular signaling pathways are regulated by small G proteins of the Ras GTPase superfamily that activate lipid kinases of the phosphatidylinositol-3-kinase (PI3K) family, thereby amplifying lipid kinase production of phosphatidylinositol (PIP) signaling lipids phosphorylated at the 3 position of the inositol headgroup (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12). Two mechanistic models have been proposed to explain how a membrane-anchored small G protein may activate its target PI3K (13, 14, 15, 16, 17, 18). 1) The membrane recruitment model postulates that membrane-anchored G protein drives a net increase in lipid kinase activity by recruiting more kinase molecules to the membrane surface where the lipid substrate resides. And 2) the allosteric activation model postulates that the G-protein-kinase interaction increases the specific activity (turnover rate) of each membrane-bound kinase molecule by triggering a change in its conformation or docking geometry on the target membrane surface. These activation mechanisms are not mutually exclusive; in principle, G protein stimulation of net PI3K activity could employ either mechanism alone or both together. A previous single-molecule analysis of H-Ras activation of PI3Kα, a class I PI3K, revealed that H-Ras association increases net PI3Kα lipid kinase activity and production of signaling lipid PI(3,4,5)P₃ via the membrane recruitment mechanism, with no allosteric increase in the turnover rate of the membrane-bound kinase molecule (13). Given the broad significance and diversity of G protein regulation of PI3K enzymes in normal cell function as well as many human disease states, further studies defining the range of G-protein-PI3K activation mechanisms may significantly advance both signaling biology and pharmaceutical development.

This study focuses on the activation of class III PI3K enzymes by the small G protein Rab, a member of the Ras superfamily (4,6,7,10, 11, 12,19, 20, 21). VPS34 complexes I and II are the two prototypical class III PI3K family members that phosphorylate the substrate lipid phosphatidylinositol (PI) to generate the signaling lipid phosphatidylinositol-3-phosphate (PI3P) (4,7,20,22, 23, 24, 25). The resulting PI3P lipid in turn recruits diverse signaling proteins possessing PI3P-specific targeting domains, including FYVE, PX, and PROPPINS, to the target membrane surface (4,26, 27, 28, 29, 30). VPS34 complexes I and II are both heterotetramers possessing the core lipid kinase subunit VPS34 and the modulator subunits VPS15 and Beclin1, as well as modulator subunit ATG14L (complex I) or UVRAG (complex II) (4,29). Overall, these complexes regulate a broad array of intracellular membrane remodeling and trafficking events (7,12,26,31, 32, 33, 34). Moreover, defective regulation of one or both complexes is strongly linked to human cardio, neuro, and hepatic diseases as well as RNA virus replication and multiple human cancers (25,27,35, 36, 37, 38). VPS34 complex I (VPS34CI) regulates the early stages of autophagy, including autophagosome formation, and is central to autophagic breakdown of cell components to regenerate amino acids, nucleotides, and lipids during metabolic stress. Such VPS34CI-regulated deconstruction of unnecessary or damaged components may also inhibit early events in tumorigenesis (4,39). VPS34 complex II (VPS34CII) is an important player in endocytosis, including phagocytosis, pinocytosis, and internalization of specific receptor-ligand complexes (34). VPS34CII also participates in other regulatory events in early endosome development, late autophagy, cytokinesis, and lysosome recycling (4).

Rab5A (Rab5) is an isoform of the Rab family of small G proteins (6,21,40). Like other Ras superfamily members, it possesses a conserved small GTPase domain serving as a GTP-regulated off-on switch and a C-terminal hypervariable region that tethers the GTPase domain to the target membrane via lipidation of specific Cys residues near the C-terminus (41,42). Rab5 is a key activator of phagocytosis and early endosome development through its ability to upregulate VPS34CII (4,20,34). At the same time, Rab5 activation of VPS34CII negatively regulates Rab5 and Rab7 by recruiting the GAP TBC1D2 to membranes (19). Although the Rab5 docking surfaces of VPS34CII have not yet been structurally defined (no co-complex structure is yet available), current evidence points to Rab5 contacts with the WD40 domain of the VPS15 subunit (4,10, 11, 12).

It is especially interesting to compare the small G protein regulation of class I and class III PI3K lipid kinases because of the major differences in their structural and regulatory organization (reviewed in (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11)). Class I PI3Ks possess autoinhibitory SH2 domains that block the membrane docking site until they are displaced by the binding of phospho-tyrosine (Pi-Tyr) residues displayed on the surface of membrane-associated or soluble activating proteins (43, 44, 45). Thus, the phosphorylation of Tyr residues on activator proteins by receptor (or other) tyrosine kinases, followed by binding of these Pi-Tyr residues to SH2 domains and the resulting allosteric activation, is essential for class I PI3K membrane targeting and lipid kinase activity. Such Pi-Tyr-SH2 allosteric activation alone is sufficient for modest lipid kinase activation, but full activation requires membrane-anchored G protein recruitment of the PI3K Ras-binding domain (RBD) yielding additional kinase molecules on the target membrane surface, thereby synergistically increasing the surface density of active kinase molecules and the net rate of PI(3,4,5)P₃ production. Thus, for class I PI3K activation, both tyrosine kinase-triggered allosteric mechanisms and small G-protein-triggered membrane recruitment mechanisms play distinct primary and secondary roles, respectively, in lipid kinase upregulation by Pi-Tyr-SH2 and G-protein-RBD interactions.

Unlike class I PI3K enzymes, the class III PI3K complexes VPS34CI and VPS34CII lack autoinhibitory SH2 domains; thus, regulation of these complexes by small G proteins may play a primary, rather than secondary, regulatory role. This study modifies our previously described in vitro single-molecule approach (13,46, 47, 48, 49, 50) to elucidate the molecular mechanism by which GTP-Rab5 activates the VPS34CII lipid kinase and its PI3P production in a reconstituted Rab5-VPS34CII-PI3P-PX signaling pathway, schematically illustrated in Fig. 1. The single-molecule approach enables direct measurement of the effects of membrane-anchored GTP-Rab5 on both 1) the membrane recruitment of VPS34CII and 2) the specific activity (turnover rate) of the membrane-bound kinase molecule. Measurement of both parameters is required to resolve the contributions of membrane recruitment and allosteric regulation to the molecular mechanism of GTP-Rab5-VPS34CII regulation. Although the contribution of G-protein-triggered membrane recruitment can sometimes be quantified in live cells, quantification of the effect of G protein association on the turnover rate of membrane-bound kinase molecules is generally not possible in the cellular context. The findings reveal that Rab5 drives activation of the VPS34CII lipid kinase and PI3P production via a dual mechanism characterized by nearly equivalent, major contributions from both membrane recruitment and allosteric activation.

Figure 1.

Schematic illustration of the Rab5-VPS34CII-PI3P-PX signaling pathway. Shown is a schematic view of the Rab5-VPS34CII-PI3P-PX pathway that assembles on phagosome and early endosomal membranes, where the small G protein GTP-Rab5A (Rab5) regulates the lipid kinase VPS34CII via an unknown mechanism explored herein. The four subunits of VPS34CII are VPS34 (blue), VPS15 (rose), Beclin1 (gold), and UVRAG (green). VPS34CII phosphorylates the substrate lipid phosphatidylinositol (PI) to generate the signaling lipid PI-3-phosphate (PI3P). The resulting PI3P recruits multiple signaling proteins possessing a PI3P-specific membrane targeting domain such as FYVE, PX (shown), or PROPPINS to initiate key cellular processes. In the single-molecule membrane density and kinase activity assays, an Alexa-Fluor-labeled (gold symbol) nanobody or PX domain sensor is used to detect and count membrane-bound VPS34CII or PI3P product molecules, respectively. In addition, the reconstituted system employs an engineered Rab5 anchored to the membrane by reaction of native lipidation residue Cys212 (the more N-terminal site of the two native lipidation sites shown in the figure) with the maleimide headgroup of lipid PE-Mal. The illustrated structural model of human VPS34CII is based on the yeast VPS34 structure (Protein Data Bank, PDB: 5DFZ (23)) and on a recently reported model of membrane-bound human VPS34CII (12). The structure of the Rab5-VPS34CII complex is not yet known but is believed to involve an Rab5-VPS15(WD40) contact as indicated (4,10,11). The PX domain is from P40phox (PDB: 1H6H (51)). Shown are structures generated in MacPyMol. To see this figure in color, go online.

Materials and Methods

Reagents

Synthetic dioleoyl phospholipids PC (phosphatidylcholine; 1,2-dioleoyl-sn-glycero-3-phosphocholine), PS (phosphatidylserine; 1,2-dioleoyl-sn-glycero-3-phospho-L-serine), PE (1,2-dioleoyl-sn-glycero-3-phosphoethanolamine), PI (1,2-dioleoyl-sn-glycero-3-phosphoinositol), and the maleimide-modified headgroup phospholipid PE-Mal (1,2-dioleolyl-sn-glycero-3-phosphoenthanolamine-N-[4-(p-maleimidomethyl)cyclohexane-carboxamide]) were from Avanti Polar Lipids (Alabaster, AL). Alexa Fluor 555 C₂-maleimide (AF555), Alexa Fluor 488 C₅-maleimide (AF488) and CoverWell perfusion chambers were from Invitrogen (Carlsbad, CA). Glass supports were from Ted Pella (Redding, CA). Ultrapure (>99%) bovine serum albumin and ATP magnesium salt was from Sigma-Aldrich (St. Louis, MO). Other materials were obtained as previously described from the same suppliers (13,46,52).

Proteins

VPS34CII is a large, 386-kDa, heterotetrameric lipid kinase enzyme complex that we purified as previously described in detail with minor modifications (53). Briefly, Expi293f cells were transiently transfected with vectors for 48 h before harvesting. After lysis, purification was carried out on anti-protein-A IgG IgG Sepharose 6 Fast Flow beads (GE Healthcare, Chicago, IL). Afterwards, the double protein A (ZZ) affinity tag present on the VPS15 subunit for purification was cleaved by TEV protease. The resulting cleaved VPS34CII was concentrated and further purified by gel filtration then snap frozen in liquid N₂ in 12 μL, 2.9 μM aliquots and stored at −80°C. Aliquots were thawed only once.

Rab5A (Rab5) is a small, 23-kDa GTPase and member of the Ras superfamily. The protein was engineered to remove two surface Cys residues (C19S and C63S) and to introduce an active site mutation (Q79L) that inhibits its intrinsic GTPase activity by 100-fold relative to wild-type (54). In addition, the construct has a three-residue C-terminal truncation (Δ213–215) to ensure homogeneous membrane anchoring by one remaining native lipidation residue (C212).

Rab5A was purified using a 6×-His tag removed by SUMO protease cleavage. First, 2–4 L of Escherichia coli C41 (DE3) RIPL was transformed with plasmid pOP823 in 2xTY and grown until the optical density was 0.5–0.8 at 37°C. Isopropyl β-d-1-thiogalactopyranoside (IPTG) was then added to 0.3 mM, and overexpression was carried out at 18°C for 16 h. Cells were lysed by sonication in lysis buffer (25 mM HEPES (pH 8.0), 200 mM NaCl, 10 mM imidazole, 1 mM tris(2-carboxyethyl)phosphine (TCEP), 0.05 μL/mL nuclease, and 0.5 mg/mL lysozyme). Lysate was then clarified by centrifugation (Ti45 rotor, 35,000 rpm, 45 min). After passing through a 0.45-μm filter to remove aggregates and large particles, lysate was passed over a 5-mL bed volume HisTrap FF column (GE Healthcare) and washed with wash buffer (20 mM HEPES (pH 8.0), 300 mM NaCl, 5% glycerol, 10 mM imidazole, and 2 mM β-mercaptoethanol). The bound fraction was eluted with elution buffer (20 mM HEPES (pH 8.0), 100 mM NaCl, 5% glycerol, 200 mM imidazole, and 2 mM β-mercaptoethanol). After elution, fractions containing protein (as determined by polyacrylamide gel electrophoresis gel) were pooled, and ULP1 SUMO protease was added. The sample was dialyzed overnight in a presoaked 10,000 MWCO Snakeskin Dialysis Membrane during 6×-His tag cleavage in dialysis buffer (20 mM HEPES (pH 8.0), 150 mM NaCl, 5% glycerol, and 1 mM TCEP) at 4°C.

After dialysis, the sample was re-passed over a fresh 5-mL bed volume HisTrap Fast Flow column (GE Healthcare) and washed first with dialysis buffer, followed by wash buffer (20 mM HEPES (pH 8.0), 100 mM NaCl, 5% glycerol, 10 mM imidazole, and 2 mM β-mercaptoethanol) to remove cleaved 6×-His tag. Eluate was concentrated in a 10,000 MWCO centrifugal filter (4400 rpm) to a volume of 1 mL before nucleotide exchange. To exchange Rab5A into the desired nucleotide, ∼11-fold excess GTP or GDP (Jena Bioscience, Jena, Germany) was added to the concentrate as well as 10 mM ethylenediaminetetraacetic acid and allowed to incubate for 1.5 h at room temperature. Afterward, 20 mM MgCl₂ was added to replace the lost magnesium during ethylenediaminetetraacetic acid incubation and further incubated for 0.5 h.

Gel filtration purification took place in a Superdex 75 16/60 (GE Healthcare) equilibrated with two-column volumes of gel filtration buffer (25 mM HEPES (pH 7), 150 mM NaCl, and 0.5 mM TCEP). Fractions were pooled and concentrated using a 10K filter (Amicon Ultra-15; Sigma-Aldrich). Sample was then frozen in liquid N₂ in aliquots ranging from 200 to 1000 μM and 10–100 μL before storage at −80°C. Some aliquots were subaliquoted and again snap frozen and stored at −80°C. No protein used in experiments was thawed more than twice.

p40^phox-PX domain (PX) is a 17-kDa, PI3P-specific, lipid-targeting domain and was expressed in E. coli C41 (DE3) RIPL cells. After culture in 2xTY medium at 37°C to an optical density of 0.6, cells were overexpressed with 0.3 μM IPTG at 30°C for ∼16 h. Cells were harvested by centrifugation (20 min, 4000 × g), and pellets were resuspended in lysis buffer (20 mM HEPES (pH 8.0), 200 mM NaCl, 1 mM TCEP, 0.05 μL/mL nuclease (Thermo Fisher Scientific, Waltham, MA), and 0.5 mg/mL lysozyme (MP Biomedicals, Santa Ana, CA)). The resuspension was lysed via sonication (6-min process time, 10 s on and 10 s off, and 60% power). Lysate was clarified by further centrifugation (45 min, 30,000 × g). Supernatant was filtered through a 0.45-μm filter (MilliporeSigma, Burlington, MA). 2.5 mL of Glutathione Sepharose 4B resin (GE Healthcare) was added and allowed to incubate for ∼45 min while mixing at 4°C. The slurry was then added to a gravity flow column (Bio-Rad Laboratories, Hercules, CA) and washed with 100 mL lysis buffer, 100 mL wash buffer (20 mM HEPES (pH 8.0), 300 mM NaCl, and 1 mM TCEP), and 100 mL TEV cleavage buffer (20 mM HEPES (pH 8.0), 200 mM NaCl, and 1 mM TCEP). TEV protease was added to cleave the GST affinity tag at the N-terminus of the construct. Cleavage was carried out overnight while mixing at 4°C. Afterwards, fractions were collected and concentrated in an Amicon Ultra-15 (10,000 kDa MWCO; MilliporeSigma). Gel filtration purification took place on a Superdex 75 16/60 (GE Healthcare) equilibrated with gel filtration buffer (20 mM HEPES (pH 8.0), 200 mM KCl, and 1 mM TCEP). Peak fractions were selected via SDS-PAGE gel and further concentrated to 1.3 mM. Aliquots of 180 μL were snap frozen in liquid N₂ and stored at −80°C. Some aliquots were subaliquoted and again snap frozen and stored at −80°C. No protein used in experiments was thawed more than twice.

Nanobodies CA12588 and CA12588-Cys are both 13-kDa nanobodies that bind VPS34 complexes I and II (Fig. S1). Nanobody CA12588 was created as described previously with minor modifications (23). Briefly, purified human VPS34 complexes I and II (see above for their purification) were chemically cross-linked using a CovalX K200 stabilization kit (CovalX, Zurich, Switzerland) by following the manufacture’s protocol. The antigens were shipped to the Steyaert laboratory for nanobody discovery. Nanobodies were generated following the previously described protocol (55). Briefly, two llamas were immunized six times with in total 1 mg of either cross-linked complex. From each animal, peripheral blood lymphocytes were collected 4 days after the final antigen boost. Total RNA was purified from these cells and converted into cDNA via reverse transcriptase polymerase chain reaction (PCR). The nanobody repertoire was amplified and cloned into the phage display vector pMESy4 containing a C-terminal 6× His tag and an EPEA tag (Patent WO2011147890 Pardon, E., Steyaert, J., W. L. (2011) epitope tag for affinity-based applications.). To identify complex specific nanobodies, complex I and complex II, both cross-linked and native, were solid phase immobilized in a 96-well MaxiSorp plate (Nunc, Rochester, NY). After incubation with the nanobody-phage library and several washes, antigen-bound phages were recovered by proteolysis with trypsin. After two selection rounds, individual clones from each selection were screened, and 21 families and 19 families were discovered for complex I and complex II, respectively, using an enzyme-linked immunosorbent assay (ELISA). For hydrogen deuterium exchange-mass spectrometry (HDX-MS) experiments (Fig. S1; Table S1), the CA12588 nanobody was expressed by the PMESy4 plasmid in the E. coli periplasm and purified as previously described (23). Briefly, affinity purification on Ni-NTA beads was followed by gel filtration on an S75 10/30 column. Peak fractions were collected and concentrated to 440 μM, then 10 μL aliquots were snap frozen in liquid N₂ and stored at −80°C.

Nanobody CA12588-Cys (NB-Cys) is a variant of CA12588 engineered to possess a single surface-exposed Cys residue at its C-terminal position (C121) for use as a fluor labeling site for AF555 (or AF488 for polarization assay in Fig. S1). To generate NB-Cys, the CA12588 gene coding for the residues 1–118 was cloned between the BamH1 and HindIII sites of pOPE6 (Addgene, Watertown, MA). The NB-Cys from the resulting plasmid (pYO1214) was purified using a His14-NEDD8-tag and its specific protease, Brachypodium distachyon (bd)NEDP1 (56). The same protein expression and purification procedures were followed as in (57), yielding the following amino acid sequence for the nanobody after protease cleavage:

QVQLVESGGGLVQAGGSLRLSCAASGSIFNVVGWYRQAPGKQRELVADITGGGSTRYVDSVKGRFTISRDNAKNTVVLQMNSLKPEDTAVYYCNNRPYRSLSVIGGYWGQGTQVTVSSGGC.

In brief, a plasmid carrying the CA12588-Cys coding sequence was transformed into NEBExpress Iq cells (C3037I; New England Biolabs, Ipswich, MA). The transformed bacteria were grown in a 225-mL 2TY medium preculture with 50 μg/mL kanamycin at 24°C for 18 h. The preculture was added to a 900-mL 2TY culture containing 50 μg/mL kanamycin, grown at 24°C for 1 h before induction with 0.2 mM IPTG for 5 h at 24°C. The induced bacteria were harvested at 6800 × g for 25 min, flash frozen in liquid nitrogen, and stored at −80°C until use.

The bacteria pellet was suspended in 30 mL lysis buffer (50 mM Tris (pH 8.0), 300 mM NaCl, 25 mM imidazole (pH 8.0), 10 mM dithiothreitol, 250 mM sucrose, and 1 mM PMSF), sonicated with 2-s-on-3-s-off cycles at 60% power for 5 min. The lysate was centrifuged at 18,000 × g for 30 min at 4°C. The supernatant was mixed with a 1.5-mL set volume of Ni-NTA beads (30210; QIAGEN, Hilden, Germany) rotated at 13 rpm for 1.5 h at 4°C. The lysate-beads mixture was transferred to an empty gravity flow column and washed once with 100 mL lysis buffer, then eluted four times with 5 mL elution buffer (50 mM Tris (pH 8.0), 300 mM NaCl, 500 mM imidazole (pH 8.0), and 5 mM dithiothreitol). The elution fractions were combined and concentrated down to 2.5 mL using a 3-kDa cut-off concentrator (UFC9003; MilliporeSigma) then buffer-exchanged with maleimide-labeling buffer (100 mM potassium phosphate (pH 7.2), 150 mM NaCl, 250 mM sucrose, and 5 mM TCEP) using a PD-10 desalting column (17-0851-01; GE Healthcare). The NEDD8 tag was cleaved with 4 μM of purified 14×His-MBP-SUMO-NEDP1 (house made) at 4°C for overnight. The NEDD8 tag and 14×His-MBP-SUMO-NEDP1 were removed by Ni-NTA beads, and the flowthrough fraction was concentrated down to 0.8 mL using a 3-kDa cut-off concentrator. The concentrated sample was subjected to gel filtration on an S75 10/30 column (17-5174-01; GE Healthcare), which had been equilibrated with gel filtration buffer (maleimide-labeling buffer supplemented with 5 mM TCEP). The peak fractions were combined and concentrated down to 314 μM using a 3-kDa cut-off concentrator. Five aliquots of 200 μL were snap frozen in liquid N₂ and stored at −80°C. Subsequently subaliquots snap frozen in liquid N₂, stored at −80°C, and thawed only once before use.

Labeling protein constructs for use as fluorescent sensors

PX domain and nanobody were labeled with AF555 (or AF488 in Fig. S1) maleimide using a standard procedure. Briefly, 1–5 μM of target protein was combined with twofold excess dye in activity buffer (pH 7) containing ∼5-fold excess TCEP. The reaction was allowed to proceed at 4°C for 4 h with gentle agitation. After labeling, excess free dye was quenched with glutathione (5 mM) and then removed using equilibrated Bio-Spin P-6 gel columns (Bio-Rad Laboratories). Concentration and labeling efficiency were determined by spectrophotometer. Aliquots of 10 μL at 5–10 μM were snap frozen in liquid N₂, stored at −80°C, and thawed only once before use.

Supported target lipid bilayers and Rab5A coupling

The supported bilayer lipid mixture employed was DOPC/DOPE/DOPS/DOPI/DOPE-Mal in a mole ratio of either 49:20:25:5:1 (superactivating) or 54:25:10:10:1 (physiological). Previously described procedures (13,46, 47, 48, 49, 50,52) were utilized to deposit these lipids as a homogeneous bilayer on ultraclean glass slides. Subsequently, membranes possessing or lacking anchored Rab5A were generated using a modification of the procedure previously employed to anchor H-Ras to supported lipid bilayers (13). GTP- or GDP-loaded Rab5A was covalently linked via its native lipidation Cys residue C212 to the maleimide-derivatized headgroup of DOPE-Mal lipids. Rab5A (or, for Rab5A-free membranes, the equivalent volume of buffer) was added to a final protein concentration of 6 μM in the chamber above the supported bilayer, then reacted for 24 h at 4°C in coupling buffer (25 mM HEPES (pH 7.0), 150 mM NaCl, 2.5 mM EGTA, and either 5 mM Mg²⁺ or Mn²⁺, in which this EGTA-Mg²⁺-Mn²⁺ buffering system yields ∼2.5 mM free Mg or Mn²⁺). Subsequently unreacted DOPE-Mal lipid was quenched, and free Rab5A was removed by washing with kinase activity buffer (25 mM HEPES (pH 8.0), 150 mM NaCl, 5 mM glutathione, 2.5 mM EGTA, and 5 mM MnCl₂ for superactivating conditions, or same except pH 7.4 and 5 mM MgCl₂ as the divalent metal for near-physiological conditions). Care was taken to carry out the Rab5A coupling reaction in a reproducible fashion to ensure consistent membrane density of anchored Rab5A, as confirmed by its consistent effects on VPS34CII membrane binding and kinase activity.

Single-molecule experiments

Total internal reflection fluorescence microscopy (TIRFM) experiments were carried out at 21.5 ± 0.5°C on an objective-based Nikon TE2000U inverted microscope (Tokyo, Japan) in a previously described home-built TIRFM system (13). The microscope is equipped with a Nikon Apochromat, numerical aperture 1.49 60× TIRF oil immersion objective. The excitation source (532 nm) is a diode-pumped, solid-state CNI-Laser model MGL-III-532 300 mW (Changchun New Industries Optoelectronics Technology, Changchun, China). Total excitation power leaving the microscope and striking the sample is lowered via neutral density filters to 2.3 mW for observation (measured as epifluorescence by a power meter). Fluorescence emerging from the sample passes through multiple short-pass and long-pass emission filters to eliminate source excitation contamination and sample noise, then is imaged by an Andor iXon Life 897 EMCCD camera (Andor, Belfast, UK). The field of the camera EMCCD chip is cropped from 512 × 512 to 256 × 256 pixels, and images are captured at 50 Hz, yielding an exposure time of 20 ms per frame. The physical size of each borderless pixel is 16 × 16 μm², and the calibrated sample image scaling is 4.2 pixels μm⁻¹ (0.238 μm pixel⁻¹), as measured using a TGG01 silicon calibration grating (NanoAndMore, Watsonville, CA), yielding a net optical magnification of 67×.

After creation and washing of supported lipid bilayers lacking or possessing anchored Rab5 (see above), the surface of the bilayer was blocked using concentrated ultrapure bovine serum albumin to a final concentration of 100 μg/mL to prevent or reduce target protein nonspecific binding to surfaces in the system as previously described (13,46,52). Subsequently, activity buffer containing proteins, nucleotides, and reagents were added to the system and were equilibrated for 3 min. These include 1 mM GTP or GDP for anchored GTP-Rab5 or GDP-Rab5, respectively, ATP (1 mM) for phosphorylation of PI by VPS34, and PX domain (2 μM total) for PIP binding and counting. VPS34CII surface density measurements also included anti-VPS34CII nanobody (∼1.7 μM total). Finally, aliquots of VPS34CII were thawed on ice and diluted via a minimal factor into activity buffer (extensive dilution at this state reduces activity) and stored on ice. Just before use, that initial VPS34CII stock was further diluted to its final concentration (9 nM total) in the reaction mixture above the membrane.

To minimize contributions from small numbers of immobile unfolded proteins bound to a low density of membrane defects, a bleach pulse of ∼5-fold higher power than that used for imaging was applied for ∼15 s; then fluorescence was allowed to return to a steady state for at least 60 s before data acquisition. For each sample, a set of two to four video streams were acquired at a frame rate of 20 frames/s and a spatial resolution of 4.2 pixels/μm using NIS-Elements Basic Research (Nikon). After this, a background video was taken to later subtract the background from experimental totals during data analysis.

Single-molecule diffusion tracking

As previously published, single-molecule tracks were analyzed using the ImageJ plugin ParticleTracker (13,46,52,58), giving an x-y coordinate of each track per frame as well as a brightness value. The resulting text files containing coordinates and brightness were then further analyzed in Mathematica software (Wolfram Research, Champaign, IL). Data were gated by removing excessively bright and faint tracks to remove tracks due to aggregates or contaminants. Data were additionally gated by diffusion coefficient to remove stationary particles and very quickly diffusing particles, both of which are uncharacteristic of a stable, mobile protein. These gatings were described in detail and validated previously (49,50,59).

Single-molecule VPS34CII membrane binding densities

To probe lipid binding capabilities of VPS34CII and its modulation by membrane-anchored Rab5A, we employed a VPS34-specific NB-Cys labeled at its C-terminal Cys residue with AF555 maleimide (Thermo Fisher Scientific). Binding measurements focused on VPS34CII tracks bound to the membrane for at least 100 ms to eliminate transiently bound protein and contaminants; these tracks are operationally defined as stable membrane-bound VPS34CII (49,50,59).

To calculate the density of stable VPS34CII complexes on the membrane surface in a given video, the average number of single-molecule tracks present in the observation field (60 × 60 μm²) was determined. Because our standard video analysis identifies all single-molecule tracks and determines the number of frames spanned by each track, the simplest calculation is to sum the frame numbers of all tracks and then divide the sum by the number of frames in the full video to yield the average density of molecules per frame. When the track density or background was too high to resolve individual tracks, fluor-labeled nanobody was diluted with unlabeled (dark) nanobody in a known ratio to optimize track resolution while retaining the ability to calculate the total number of membrane-bound VPS34CII complexes from the observed number fluorescent tracks. Photobleaching and photoblinking of tracks was not an issue because the average residence time on the bilayer per track before dissociation (<0.5 s) was significantly shorter than the average bleach or blinking time of AF555 under our standard experimental conditions (>20 s). It follows that bleach or blinking effects could be neglected as noted previously (49,50). Notably, the nanobody yielded no detected effect on VPS34CII lipid kinase activity, and Rab5-triggered changes in VPS34CII membrane surface density were quantitatively explained within error by measured changes in the membrane association and dissociation kinetics of the nanobody-VPS34CII complex (see Results). Moreover, nanobody binding provides stable protection of its docking surface in HDX-MS experiments (see Fig. S1; Table S1). These observations indicate that although the nanobody is not covalently coupled to VPS34CII, its kinetic stability is adequate for effective use as a VPS34CII density sensor.

Single-molecule VPS34CII membrane association and dissociation kinetics

The kinetics of VPS34CII association and dissociation were obtained from single-molecule surface diffusion tracks as previously described (13). Briefly, the apparent on-rate (operationally defined as the rate of appearance of stable, membrane-bound VPS34CII complexes) was measured for each video by dividing the total number of VPS34CII diffusion tracks appearing in the video (excepting tracks already present in the first frame) by the total video timespan. This on-rate was divided by the field area and the bulk VPS34CII concentration to generate a pseudo-first-order, on-rate constant. The off-rate constant of the VPS34CII membrane-bound state was measured by plotting a frequency distribution of the single-molecule bound state lifetimes, as defined by the timespans of their diffusion tracks (excepting incomplete tracks present in the first or last frame), and fitting the resulting distribution with a single exponential decay for a first-order dissociation reaction.

VPS34CII net and specific lipid kinase activity

A single-molecule kinase assay was employed to quantify net VPS34CII lipid kinase activity and PI3P production adapted from a previously described methodology for studying a class I PI3Ks (13,46). The method counts every molecule of product lipid generated by VPS34CII by adding a saturating concentration of AF555 fluor-labeled PX40^phox domain (2 μM, a level ≥16-fold higher than the reported K_D ≤ 120 ± 20 nM (60)) to bind and detect each PI3P molecule created by the complex. When the track density or fluorescence background was too high to resolve individual tracks, the fluor-labeled PX domain was diluted with unlabeled (dark) PX domain in a known ratio to provide adequate track resolution while retaining the ability to calculate the total number of membrane-bound PX-PI3P complexes from the observed number fluorescent tracks. Finally, VPS34CII was added to the chamber (9 nM, final concentration) from a working stock diluted in kinase activity buffer (see above) to start the kinase reaction. Fluorescent PX domain tracks were captured at five time points—five times per minute for membranes(+)Rab5A under superactivating conditions and once every minute for 5 min for all other conditions. Finally, the rate of the reaction was determined as the slope of activity over time.

The lipid kinase time courses and reaction rates include a small correction for nonspecific PX domain binding to the target membrane. Each time point datum was corrected by subtracting out the small, constant number of background PX domain tracks observed on the substrate membrane before the addition of VPS34CII. Even for the basal VPS34CII reaction in the absence of membrane-anchored Rab5, this correction is minimal. Specifically, on the substrate bilayer lacking Rab5 and PI3P, the background density of PX domain (45 ± 39 molecules per field, 60 × 60 μm²) is approximately two orders of magnitude smaller than the PX density observed after 2 min of basal VPS34CII lipid kinase reaction, generating PI3P (4000 ± 2000 molecules per field), which in turn is approximately two orders of magnitude larger than the density of VPS34CII molecules on the membrane (52 ± 3 molecules per field). It follows that the basal activity measured for the VPS34CII lipid kinase reaction in the absence of Rab5 is real and is not an artifact of background PX binding to the membrane.

The VPS34CII-specific activity was also determined, representing the turnover rate of the average, membrane-bound VPS34CII single molecule. This specific activity was calculated as the ratio of the net VPS34CII lipid kinase activity (PI3P molecules field⁻¹ min⁻¹) to the density of membrane-bound VPS34CII (molecules field⁻¹), in which the net lipid kinase activity and the membrane-bound kinase density were measured independently as described above.

Statistics

For each parameter measured in Figs. 3, 4, 5, 6, 7, and 8, a total of n = 3–5 means were determined on different days, in which each mean averaged data for three or more replicate videos or time courses carried out on the same day. The final measured parameter represents the average of these n means, and its error bar is the standard error of the mean. Moreover, each measured parameter reflects the contributions of hundreds to thousands of single molecules because the total number of replicate videos or time courses analyzed per parameter ranged from 9 to 15, and each video or time course contributed tens to hundreds of single molecules.

Figure 3.

Representative TIRFM single-particle tracks of freely diffusing fluorescent proteins. (A) The AF555-labeled PX domain of human P40phox bound to the PI3P lipid, whereas the protein-lipid complex freely diffuses on the target bilayer surface. (B) Shown is VPS34CII tagged with AF555-labeled nanobody and simultaneously bound to both the target bilayer and membrane-anchored GTP-Rab5. (C) Shown is VPS34CII tagged with AF555-labeled nanobody and bound to the target bilayer in the absence of membrane-anchored Rab5. Shown are trajectories composed of 20-ms single steps as captured by the 50 s⁻¹ frame rate at 21 ± 0.5°C under superactivating conditions on PE/PC/PS/PI/Mal-PE membranes (49:20:25:5:1 mol %) in 25 mM HEPES (pH 8.0), 150 mM NaCl, 5 mM glutathione, 2.5 mM EGTA, 5 mM Mn²⁺, 1 mM GTP, and 1 mM ATP. Shown is the global averaging of the mean surface diffusion constant (D) over 9–15 videos for each protein, in which each video contributes tens to hundreds of single-molecule tracks and yields D = 0.9 ± 0.1, 0.4 ± 0.1, and 0.6 ± 0.2 μm² s⁻¹ for the PX-PI3P complex, VPS34CII (+) Rab5, and VPS34CII (−) Rab5, respectively.

Figure 4.

Effect of membrane-anchored Rab5A on net VPS34CII lipid kinase activity: superactivating conditions. (A) Single-molecule TIRFM images show VPS34CII lipid kinase reactions initiated at t = 0 while monitoring the product lipid PI3P using a saturating level of fluor-labeled PX domain to bind and detect individual product PI3P molecules. The findings show that under superactivating conditions, VPS34CII lipid kinase activity can be detected in the absence of Rab5 but increases dramatically in the presence of membrane-anchored GTP-Rab5. (B) Shown is the single-molecule time course of the VPS34CII lipid kinase reaction plotting the number of product molecules generated per field as a function of reaction time, illustrating stimulation by membrane-anchored GTP-Rab5. (C) Given is the rate of the VPS34CII lipid kinase reaction defined by the slope of the single-molecule time course (previous panel), illustrating the significant, 33 ± 12-fold (p < 0.001) rate enhancement by membrane-anchored GTP-loaded Rab5 ((+) GTP-Rab5) relative to membranes lacking the G protein ((−) GTP-Rab5). Also shown are controls demonstrating the lack of stimulation by membrane-anchored, GDP-loaded Rab5 ((+) GDP-Rab5) or by membranes pretreated with a maleimide quencher to block GTP-Rab5 anchoring to the membrane ((+) quenched Rab5) before the anchoring reaction and washing to remove free GTP-Rab5. Single-molecule kinase reactions under were carried out at 21 ± 0.5°C under superactivating conditions: the supported lipid bilayer PC/PE/PS/PI/Mal-PE was 49:20:25:5:1 mol %, and reaction buffer 25 mM HEPES (pH 8.0), 150 mM NaCl, 5 mM glutathione, 2.5 mM EGTA, 5 mM Mn²⁺, 1 mM GTP, and 1 mM ATP. The observation field was 60 × 60 μm².

Figure 5.

Effect of membrane-anchored GTP-Rab5A on VPS34CII membrane surface density: superactivating conditions. (A) Shown is the single-molecule surface density of VPS34CII molecules stably bound to the target supported lipid bilayer, showing the significant, 6.6 ± 2.5-fold (p < 0.001) enhancement of VPS34CII density on membranes possessing anchored GTP-Rab5A relative to membranes lacking Rab5 (338.0 ± 52 and 51.6 ± 28 molecules field⁻¹, respectively). Single VPS34CII molecules were detected and counted using a fluorescent, monoclonal sensor nanobody to tag individual kinase molecules. (B) Controls were carried out to ascertain whether the sensor nanobody perturbed the membrane binding or lipid kinase activities of VPS34CII. Shown are single-molecule kinase assay data (Fig. 4) revealing that nanobody has no significant effect, within error, on the net rate of VPS34CII production of PI3P either in the absence or presence of membrane-anchored GTP-Rab5A. These findings provide strong evidence that nanobody binding has a negligible effect on VPS34CII membrane association and specific kinase activity, consistent with the HDX-MS data, revealing that the nanobody docking surface is distal from the kinase membrane docking and active sites (Fig. S1; Table 1). For both panels, superactivating conditions is the same as above (see Fig. 4 legend). The observation field was 60 × 60 μm².

Figure 6.

Effect of membrane-anchored GTP-Rab5A on the kinetics of VPS34CII membrane association and dissociation: superactivating conditions. (A) The membrane-anchored GTP-Rab5 significantly increases by 3.6 ± 1.5-fold (p = 0.01), the pseudo-first-order, on-rate constant (k’_on) for the appearance of stably bound VPS34CII single molecules on the target membrane surface, from k’_on = 4.7 × 10⁶ ± 1.6 × 10⁶ events per (μm² × [VPS34CII, M] × s) in the absence of Rab5 to 1.7 × 10⁷ ± 4.2 × 10⁶ events per (μm² × [VPS34CII, M] × s) in the presence of anchored GTP-Rab5. (B) The membrane-anchored GTP-Rab5 significantly decreases by 1.9 ± 0.1-fold (p = 0.002), the first-order off-rate constant for VPS34CII membrane dissociation (k_off), from 4.2 ± 0.3 s⁻¹ in the absence of Rab5A to 8.2 ± 0.7 s⁻¹ in the presence of anchored GTP-Rab5. Dissociation rates were obtained from the indicated bound state lifetime distributions for populations of single VPS34CII molecules. In both experiments, single molecules of VPS34CII were detected by a fluorescent nanobody sensor under superactivating conditions as above (see Figs. 4 and 5 legends).

Figure 7.

Effect of membrane-anchored GTP-Rab5A on the specific activity (turnover rate) of membrane-bound VPS34CII: superactivating conditions. (A) The membrane-anchored GTP-Rab5 significantly increases by 5.2 ± 1.8-fold (p = 0.003), the turnover rate of the average membrane-bound VPS34CII molecule from 27 ± 8 PI3P products min⁻¹ in the absence of Rab5A to 141 ± 32 PI3P products min⁻¹ in the presence of anchored GTP-Rab5A. The superactivating conditions are the same as above (see Figs. 4 and 5 legends).

Figure 8.

Effect of membrane-anchored GTP-Rab5 on three VPS34CII activity parameters: comparison of superactivating and near-physiological conditions. (A) Under near-physiological conditions, VPS34CII binding to membranes lacking Rab5 was not measurable, and VPS34CII binding to membranes possessing membrane-anchored GTP-Rab5 was significantly lower by 10.2 ± 6.3-fold (p < 0.001) relative to superactivating conditions. (B) Under near-physiological conditions, net VPS34CII lipid kinase activity on membranes lacking Rab5 was not measurable, whereas net VPS34CII lipid kinase activity on membranes possessing membrane-anchored GTP-Rab5 was easily measured but significantly lower by 9.0 ± 4.0-fold (p = 0.001), relative to superactivating conditions. (C) Under near-physiological conditions, the turnover rate of VPS34CII molecule was not measurable on membranes lacking Rab5. However, the turnover rate of the average, membrane-bound VPS34CII molecule on membranes possessing anchored GTP-Rab5 was the same, within error, when measured under near-physiological and superactivating conditions. The superactivating conditions are the same as above (see Figs. 3, 4, 5, 6, and 7 legends). Near-physiological conditions were 21 ± 0.5°C, supported lipid bilayer PC/PE/PS/PI/Mal-PE was 54:25:10:10:1 mol %, and the reaction buffer was 25 mM HEPES (pH 7.4), 150 mM NaCl, 5 mM glutathione, 2.5 mM EGTA, and 5 mM Mg²⁺. The observation field was 60 × 60 μm².

Results

Pathway components employed in reconstitution: VPS34CII, Rab5A, P40^phox PX domain, and the target lipid bilayer

Three recently developed protein constructs (12) were employed to reconstitute G-protein-regulated PI3K signaling in this study. Fig. 2 presents the schematic domain structure of each component. The human PI3K VPS34CII was expressed in mammalian cells and purified as the native, untagged heterotetramer of subunits VPS34, VPS15, Beclin1, and UVRAG. The human, small G protein Rab5A that regulates VPS34CII and other membrane-associated signaling proteins was engineered with mutations (see Materials and Methods) that block hydrolysis of the activating GTP ligand and retain a single native Cys lipidation residue at the disordered C-terminus for membrane anchoring. The resulting mutant Rab5A protein (Rab5) was expressed and isolated from E. coli, then loaded with GTP or GDP. Where indicated, the untagged, purified protein was anchored to the target membrane via covalent reaction with maleimide-derivatized phospholipid. The isolated PX membrane targeting domain of human P40^phox protein (PX) was engineered to possess a single surface Cys for fluorescent labeling and expression in E. coli, then isolated and labeled with AF555. The resulting PX-AF555 domain was used as a single-molecule sensor-effector of its target lipid PI3P. Published biophysical studies have measured K_D ≤ 120 nM for PX binding to PI3P on target membrane surfaces (60); thus, the 2-μM PX concentration employed herein yields ≥94% occupancy of PI3P and ensures saturation of the PI3P product molecule population with the sensor.

Figure 2.

Subunit and domain organization of proteins employed in this study. (A) Human VPS34CII is a class III PI3K composed of four multidomain subunits as shown. The protein employed in this study is the full-length, tag-less heterotetramer purified from mammalian cells via a double protein A purification tag (ZZ tag) at the C-terminus of VPS15 that was proteolytically removed by TEV protease to give the construct shown. (B) Human Rab5A (Rab5) is a small, monomeric G protein. The construct employed herein is purified from bacterial cells via its 6×-His purification tag and then loaded with GTP. The construct possesses an active site mutation (Q79L) that blocks GTP hydrolysis and retains a single surface Cys residue for membrane anchoring at position 212. The C212 residue near the native C-terminus is a native lipidation site, and in this construct, C212 is placed at the C-terminus by truncation Δ213–215, whereas the 6×-His tag is removed by SUMO proteolysis during purification. (C) The isolated PX domain of human P40phox is purified from bacterial cells via its GST purification tag. The construct is mutated to possess a single-surface Cys residue near the N-terminus for labeling with fluorophore for use as a sensor for its target PI3P lipid. The GST tag is removed by proteolysis during purification via TEV proteolysis. See Materials and Methods and (12).

Supported target lipid bilayers for single-molecule TIRFM studies contained a mixture of the major background and substrate lipids found in the cytoplasmic leaflet of endosomal membranes (Table 1). PC, PE, and PS served as background lipids, whereas PI served as the target lipid kinase substrate. The supported membranes also contained a small, fixed mole fraction (1%) of maleimide-PE possessing the maleimide moiety on its headgroup. After supported membrane deposition, the maleimide-PE was reacted with Rab5 (or buffer with no protein), then unreacted maleimide-PE was quenched with excess reducing reagent. Finally, free Rab5 and other reaction products not anchored to the membrane were removed by washing with kinase activity buffer, yielding target lipid bilayers possessing (or lacking) membrane-anchored Rab5. Fig. 3 illustrates representative single-molecule tracks of the PX-PI3P complex, VPS34CII, and the VPS34CII-Rab5 complex on the supported lipid bilayer, showing that the reconstituted protein-membrane complexes are freely diffusing and suitable for quantitative biochemical and biophysical measurements.

Table 1.

Conditions Employed and Comparison with Cellular Conditions

	Superactivating	Physiological	Cellular
PC (mol %)	49	54	49
PE	20	25	26
PS	25	10	7
PI	5	10	6
PE-Mal	1	1	0
Mg (mM, free)	0	2.5	0.5–1
Mn (mM, free)	2.5	0	<<0.2
pH	8	7.4	7.4
VPS34CII (nM)	9	9	<19

Shown are levels of key components in experiments carried out under superactivating and near-physiological conditions, as well as the corresponding cellular estimates. VPS34CII and substrate PI levels were adjusted to ensure that VPS34CII kinase activity stayed within a measurable, linear initial rate range. Cellular mole fractions of lipids are estimates for intracellular endosomal membranes (51,67), whereas cytoplasmic concentration of VPS34CII is an upper limit defined by the cytoplasmic concentration of its limiting subunit UVRAG, ∼19 nM (68). Lipids are 1,2-dioleoyl-sn-glycero-3 derivatives of phosphocholine (PC), phosphoethanolamine (PE), phosphoserine (PS), phosphoinositol (PI), and phosphoethanolamine-maleimide (PE-Mal).

Strategy for probing the mechanism of VPS34CII activation by Rab5

Two models have been proposed as the primary mechanisms by which small G proteins activate PI3K isoforms. The membrane recruitment model hypothesizes that GTP-activated, membrane-anchored G protein recruits PI3K to the membrane, thereby increasing the surface density of the membrane-active lipid kinase. The allosteric activation model hypothesizes that the docking of PI3K to the GTP-activated G protein on the membrane triggers an allosteric, activating conformational change within the PI3K that is transmitted to its lipid kinase active site.

To resolve these models, we reconstituted the Rab5-VPS34CII-PI3P-PX signaling pathway on target supported lipid bilayers. Then, we utilized single-molecule methods to measure the effect of membrane-anchored GTP-Rab5A on both 1) the membrane surface density of VPS34CII to detect the contributions of membrane recruitment and 2) the specific lipid kinase activity (turnover rate) of membrane-bound VPS34CII to detect the contributions of allosteric activation.

The single-molecule measurements were carried out under two contrasting experimental conditions: superactivating and near physiological. Superactivating conditions were developed to generate high levels of VPS34CII lipid kinase activity on target membrane surfaces and included the use of Mn²⁺ (rather than physiological Mg²⁺) as the dominant divalent metal ion, as well as a higher-than-physiological mole fraction of anionic lipids in the target bilayer (12). These superactivating conditions enabled measurement of VPS34CII surface density and specific activity in both the absence and presence of membrane-anchored Rab5A. For comparison, complementary studies were also carried out under near-physiological conditions, including use of Mg²⁺ as the dominant divalent metal ion and major lipids better approximating the composition of the endosomal cytoplasmic leaflet. These conditions enabled analysis of Rab5-regulated VPS34CII lipid kinase activity under conditions more closely resembling the cellular context. Table 1 shows the protein, buffer, and lipid compositions of the superactivating and near-physiological conditions and compares them with cellular conditions.

Membrane-anchored GTP-Rab5 increases net VPS34CII lipid kinase activity: superactivating conditions

Our previously described methods for reconstitution and single-molecule analysis of PI3K lipid kinase signaling pathways (13,46, 47, 48, 49, 50) were adapted here for the Rab5-VPS34CII-PI3P-PX signaling pathway. Briefly, activity buffer was added to the supported target membrane containing substrate lipid PI and either lacking or possessing membrane-anchored Rab5. Then, VPS34CII was added to start the lipid kinase reaction, and a saturating concentration of fluorescent PX domain was employed to detect and count the newly formed PI3P product lipid molecules on the supported bilayer via single-molecule TIRFM. This method yields the total number of product PI3P molecules generated per unit membrane area as a function of time, yielding the net rate of product formation. Fig. 4 shows that under superactivating conditions, membrane-anchored Rab5 significantly increases the net rate of VPS34CII-catalyzed PI3P production by 33 ± 12-fold (p < 0.001) relative to the same reaction lacking Rab5, up to a maximal level of 48,000 ± 11,000 PI3P molecules field⁻¹ min⁻¹. Experimental conditions were adjusted to ensure that the kinase activity measurement remained in the linear, initial rate phase of the kinase reaction throughout the measurement.

Controls tested the specificity of GTP-Rab5 stimulation of VPS34CII lipid kinase activity. Fig. 4 shows that when the maleimide-derivatized headgroups on the target membrane are blocked with excess glutathione before the reaction with GTP-Rab5 to eliminate membrane-anchored GTP-Rab5 or when GDP-Rab5 is anchored to the bilayer rather than GTP-Rab5, no significant stimulation of VPS34CII lipid kinase activity is observed. These controls show that the observed dramatic activation of VPS34CII by anchored GTP-Rab5 on a target membrane surface is specific for the GTP-activated state of Rab5.

Membrane-anchored GTP-Rab5 increases VPS34CII membrane surface density: superactivating conditions

To measure the density of VPS34CII on the surface of the supported target membrane, a fluorescent, monoclonal nanobody that specifically binds VPS34 complexes I and II was employed to count the number of membrane-bound VPS34CII molecules via single-molecule TIRFM. Fig. 5 shows that under superactivating conditions, membrane-anchored GTP-Rab5A increases the density of VPS34CII on the target membrane surface by 6.6 ± 2.5-fold (p < 0.001). Thus, membrane recruitment of VPS34CII is one element of the mechanism by which Rab5 activates lipid kinase activity and PI3P production.

Further kinetic analysis sheds light on the kinetic mechanism of the observed Rab5-triggered, VPS34CII density increase on the target membrane. Fig. 6 shows this density increase arises from both a 3.6 ± 1.5-fold increase (p = 0.01) in rate of appearance (on-rate) of stably membrane-bound VPS34CII molecules in the presence of membrane-anchored GTP-Rab5 coupled with a 1.9 ± 0.1-fold decrease (p = 0.002) in the VPS34CII off-rate from the target membrane. The Rab5-associated slowing of VPS34CII dissociation from the membrane is also evident in the representative single-molecule tracks and average diffusion constants presented in Fig. 3, in which binding to GTP-Rab5 increases the number of VPS34CII diffusion steps before dissociation and modestly reduces the VPS34CII diffusion constant as expected for complex formation (59). Overall, the findings indicate that GTP-Rab5 significantly speeds the formation of stable VPS34CII membrane complexes while modestly slowing both their dissociation and surface diffusion. Notably, the combined effects of GTP-Rab5 on the rate of VPS34CII stable membrane association and the rate of membrane dissociation quantitatively account, within error, for the Rab5-triggered increase in VPS34CII surface density.

A control tested whether the nanobody employed to measure VPS34CII surface density perturbed the membrane binding or activity of the lipid kinase. Addition of excess nanobody to the single-molecule assay of VPS34CII lipid kinase activity had no significant effect on the VPS34CII catalyzed rate of PI3P production on the target membrane surface, either in the absence or presence of membrane-anchored GTP-Rab5 (Fig. 5). It follows that nanobody binding to VPS34CII did not significantly alter its membrane binding or specific kinase activity. This finding is consistent with an HDX-MS analysis that maps nanobody docking to a VPS34CII surface distal to the membrane binding surface and the kinase active site (Fig. S1; Table S1; (12)).

Membrane-anchored GTP-Rab5 increases the specific activity (turnover rate) of membrane-bound VPS34CII: superactivating conditions

The average specific activity, or turnover rate, of a single VPS34CII molecule on the membrane surface can be calculated from the newly measured single-molecule parameters as the ratio of the net rate of PI3P production per unit membrane area to the VPS34CII membrane surface density. The calculation reveals that membrane-anchored GTP-Rab5 increases the turnover rate of the average membrane-bound VPS34CII molecule 5.2 ± 1.8-fold (p = 0.003) from 27 ± 8 PI3P molecules min⁻¹ in the absence of Rab5 to 141 ± 32 PI3P molecules min⁻¹ in the presence of Rab5, as shown in Fig. 7. Thus, in addition to the membrane recruitment noted above, Rab5 association also increases the specific activity of the membrane-bound VPS34CII molecule.

Rab5 regulation of VPS34CII membrane surface density and specific activity (turnover rate): physiological conditions

The superactivating conditions employed for the above measurements enabled quantitative analysis of VPS34CII surface density and lipid kinase activity in both the absence and presence of membrane-anchored GTP-Rab5, providing key mechanistic insights. In the cell, however, significant VPS34CII activity in the absence of Rab5 would reduce the efficacy of GTP-Rab5 as a tight off-on switch. When the single-molecule measurements were repeated under more physiological divalent metal, anionic lipid, and pH conditions (Table 1), the VPS34CII membrane density and net lipid kinase activity were still measurable on membranes possessing anchored GTP-Rab5 but not on membranes lacking Rab5. On GTP-Rab5 membranes Fig. 8 shows that relative to superactivating conditions, the near-physiological conditions significantly decrease the VPS34CII surface density by 10.2 ± 6.3-fold (p < 0.001) and significantly decrease the VPS34CII net lipid kinase activity by 9.0 ± 4.0-fold (p = 0.001). Notably, however, the specific activity of VPS34CII was the same, within error, for the average membrane-bound, GTP-Rab5-VPS34CII complex under superactivating and near-physiological conditions: the turnover rate per complex was 141 ± 32 and 160 ± 39 min⁻¹, respectively. It follows that superactivating conditions increase VPS34CII binding to Rab5 membranes ∼10-fold but have little or no effect on the lipid kinase specific activity once the GTP-Rab5-VPS34CII complex is bound to the target membrane surface.

Discussion

The findings indicate that the mechanism of Rab5 activation of VPS34CII lipid kinase activity includes both a membrane recruitment component that significantly increases the kinase surface density, and an allosteric activation component that significantly increases the specific activity (turnover rate) of individual, membrane-bound kinase molecules. Single-molecule measurements of VPS34CII membrane surface density reveal membrane recruitment by anchored GTP-Rab5 under both superactivating conditions designed to maximize net VPS34CII kinase activity, and under near-physiological conditions that better mimic the cellular environment (see Results and Table 1). Specifically, under superactivating conditions, target membranes possessing anchored GTP-Rab5 (but not GDP-Rab5) yield a 6.6 ± 2.5-fold increase (p < 0.001) in VPS34CII single-molecule density on the target membrane surface relative to membranes lacking anchored Rab5. Notably, under near-physiological conditions, little or no VPS34CII membrane binding is detected on target membranes lacking Rab5, whereas substantial VPS34CII density is detected on target membranes possessing anchored GTP-Rab5. These observations suggest that under cellular conditions, VPS34CII binding to the target membrane is tightly regulated by membrane-anchored GTP-Rab5 (or some other membrane-associated activator), which acts as a strong, virtually binary off-on switch essential for VPS34CII membrane recruitment.

Single-molecule measurements of both VPS34CII membrane surface density and the net rate of substrate lipid PI3P production together yield the specific activity (turnover rate) of a single, average, membrane-bound kinase molecule. Under superactivating conditions in the absence of Rab5, the turnover rate of a VPS34CII molecule bound to the target membrane is 27 ± 8 PI3P product molecules min⁻¹. Under the same conditions on a target membrane possessing anchored GTP-Rab5 (but not GDP-Rab5), the turnover rate of the GTP-Rab5-VPS34CII complex is increased 5.2 ± 1.8-fold (p = 0.003) to 141 ± 32 PI3P product molecules min⁻¹. This substantial allosteric activation of triggered by membrane-anchored GTP-Rab5 could arise from a conformational change transmitted from the Rab5 binding site through the VPS34CII molecule to its kinase active site. Alternatively, the activation could arise from an altered membrane docking geometry triggered by the association of VPS34CII with anchored GTP-Rab5 that alters the interaction of VPS34CII with the bilayer to enhance kinase activity. These allosteric activation mechanisms are not mutually exclusive and could operate together.

Strikingly, the single-molecule turnover rate of the GTP-Rab5-associated VPS34CII complex is the same, within error, under superactivating and near-physiological conditions: 141 ± 32 PI3P product molecules min⁻¹ and 160 ± 39 PI3P product molecules min⁻¹ (p = 0.7), respectively. The simplest explanation for this notable similarity is that membrane-anchored GTP-Rab5 provides the same level of allosteric VPS34CII activation under both conditions. Further, these findings indicate that superactivating conditions drive increased activation of VPS34CII, relative to near-physiological conditions, by increasing the density of lipid kinase on the target membrane, rather than by increasing its turnover rate. Supporting this picture, the VPS34CII single-molecule surface density on membranes possessing anchored GTP-Rab5 is 10.2 ± 6.3-fold higher (p < 0.001) under superactivating conditions than under near-physiological conditions. In the absence of Rab5, the fold increase in kinase surface density observed for superactivating versus near-physiological conditions is substantial but cannot be quantified because there is no measurable VPS34CII binding under the latter conditions.

The increase in VPS34CII surface density triggered by superactivating conditions, relative to near-physiological conditions, likely stems from several factors. Superactivating conditions provide greater membrane negative surface charge (because of higher anionic lipid density and pH) and a different major divalent metal (Mn²⁺ instead of Mg²⁺). Increased kinase recruitment to a membrane with greater negative surface charge would be consistent with the typical preference of PI3K lipid kinases, including VPS34CII, for negatively charged membranes (12,61). The contribution of replacing Mg²⁺ with Mn²⁺ could also be significant if Mn²⁺ has a lower affinity for anionic lipid headgroups and thus provides less screening of the membrane negative charges. However, the available evidence suggests that a screening difference may not be a major contributor because both divalent metals are believed to have similar affinity for the dominant anionic lipid PS (62,63). Another possibility is that Mn²⁺ replaces Mg²⁺ in the kinase substrate binding site and yields a higher affinity for the PI substrate lipid, which could also increase target membrane affinity.

It is interesting to compare the contrasting regulation of class I and class III PI3K enzymes by small G proteins as revealed by single-molecule studies of their membrane recruitment and specific lipid kinase activities. Our previous single-molecule analysis of H-Ras regulation of the class I PI3Kα (13) supported our models (44,45,64,65) for the initial activation step in which Pi-Tyr residues bind to the autoinhibitory SH2 domains of PI3K, thereby displacing the SH2s from their inhibitory docking sites on the membrane docking face of PI3K. The resulting allosteric activation enables PI3K docking to the target membrane surface and initiation of lipid kinase activity (44,45,65). The single-molecule findings also revealed that membrane-anchored H-Ras provides secondary, synergistic activation of PI3Kα by increasing the formation rate of its stable, membrane-bound state, thereby increasing net PIP₃ production because of higher kinase density on the membrane with no allosteric activation of the kinase turnover rate (13,15,66). More generally, it appears likely that the same combination of primary allosteric SH2 regulation by Pi-Tyr activators and secondary membrane recruitment activation by Ras pertains to other class I PI3Ks regulated by Pi-Tyr and Ras isoforms (15).

In contrast to the class I PI3Ks, the class III VPS34CI and VPS34CII enzymes lack SH2 domains and need no Pi-Tyr activation (4). The present single-molecule studies reveal that under near-physiological conditions, membrane-anchored GTP-Rab5 regulates tight, binary VPS34CII off-on switching by a dual molecular mechanism involving both membrane recruitment and allosteric activation of the lipid kinase. As observed for H-Ras activation of PI3Kα, the membrane recruitment component of Rab5-triggered VPS34CII activation is dominated by a Rab5-stimulated increase in the formation rate of the stable membrane-bound kinase rather than by slower dissociation of the kinase from membrane. One advantage of this kinetic scheme is that strong stimulation the formation rate of the kinase membrane-bound state, unlike strong slowing of its dissociation rate, yields increased kinase membrane density while retaining the ability of the activated lipid kinase population to rapidly switch off via dissociation at the end of a signaling event. Other membrane-anchored regulatory proteins that activate VPS34CI or VPS34CII may well, like GTP-Rab5, employ both recruitment and allosteric mechanisms. Alternatively, two separate pathways could, in principle, synergistically regulate VPS34 complexes using different mechanisms, analogous to the synergistic regulation of class I PI3Ks by both an allosteric mechanism (via RTK and Pi-Tyr modulation of SH2) and a membrane recruitment mechanism (via membrane-anchored GTP-Ras) (13, 14, 15,18).

For both class I and III PI3Ks, our single-molecule studies suggest that the lifetime of the fully activated kinase on the membrane, as well as the turnover rate of the membrane-bound kinase molecule, are together tuned to generate 0–1 product signaling lipids (PI(3,4,5)P₃ and PI(3)P, respectively) per membrane binding event (see both Figs. 6 and 7; (13) herein). Although this limited activity may appear quite low, it is notable that each molecule of signaling lipid produced, unlike the lipid kinase that created it, remains membrane-associated and active until it is degraded by a lipid phosphatase or lipase. Thus, each product lipid molecule yields an extended period of downstream signaling during which it can initiate multiple rounds of downstream signaling events, including recruitment of protein kinases that initiate phosphorylation cascades. In short, each product lipid yields a large biological effect, and the number of product lipids must be carefully constrained to avoid excessive downstream signaling.

This finding that Rab5 tightly regulates net VPS34CII lipid kinase activity under near-physiological conditions, along with the established importance of VPS34CII in human disease states, together suggest that a subset of the extensive list of VPS34CII disease-linked mutations (4,7,22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33,35, 36, 37, 38) will be linked to mutations at the Rab5-VPS34CII interface. Disease-linked mutations that decrease Rab5-VPS34CII affinity are predicted to decrease PI3P production and downregulate downstream signaling. Alternatively, mutations that increase Rab5-VPS34CII affinity are predicted to increase PI3P production and superactivate downstream pathways. These hypothesized mechanisms will become testable after the Rab5-VPS34CII interface is defined by structural studies.

These findings suggest a number of directions for future research. One unresolved question is whether the newly observed allosteric component of GTP-Rab5 activation of VPS34CII arises from a GTP-Rab5-triggered change in VPS34CII conformation or rather from altered membrane docking geometry. It will also be important to ascertain how VPS34CII activation by Rab5 (and other Rab isoforms) complements or synergizes with other activating modalities such as local membrane curvature, anionic lipid density, and divalent cation fluctuations (12). Finally, the single-molecule approach is ideally suited to elucidate the molecular mechanisms of disease-linked mutations and therapeutic drugs that alter Rab-VPS34CII binding interactions, membrane docking, PI3P production, and recruitment of downstream signaling proteins.

Author Contributions

J.J.F. performed the conception; T.C.B. and J.J.F. performed the experimental design; T.C.B. performed the data collection; T.C.B. and J.J.F. performed the data analysis; J.J.F. and T.C.B. performed the data interpretation; J.J.F. and T.C.B. prepared the manuscript; M.T.G contributed Fig. 3 and edited the manuscript; R.L.W., Y.O., and S.T. contributed essential materials and advice; Y.O. and S.H.M. contributed Fig. S1 and Table S1; and J.S. and E.P performed the nanobody discovery.

Editor: Kalina Hristova.