The p85alpha subunit of phosphatidylinositol 3'-kinase binds to and stimulates the GTPase activity of Rab proteins.

Rab5 and Rab4 are small monomeric GTPases localized on early endosomes and function in vesicle fusion events. These Rab proteins regulate the endocytosis and recycling or degradation of activated receptor tyrosine kinases such as the platelet-derived growth factor receptor (PDGFR). The p85alpha subunit of phosphatidylinositol 3'-kinase contains a BH domain with sequence homology to GTPase activating proteins (GAPs), but has not previously been shown to possess GAP activity. In this report, we demonstrate that p85alpha has GAP activity toward Rab5, Rab4, Cdc42, Rac1 and to a lesser extent Rab6, with little GAP activity toward Rab11. Purified recombinant Rab5 and p85alpha can bind directly to each other and not surprisingly, the p85alpha-encoded GAP activity is present in the BH domain. Because p85alpha stays bound to the PDGFR during receptor endocytosis, p85alpha will also be localized to the same early endosomal compartment as Rab5 and Rab4. Taken together, the physical co-localization and the ability of p85alpha to preferentially stimulate the down-regulation of Rab5 and Rab4 GTPases suggests that p85alpha regulates how long Rab5 and Rab4 remain in their GTP-bound active state. Cells expressing BH domain mutants of p85 show a reduced rate of PDGFR degradation as compared with wild type p85 expressing cells. These cells also show sustained activation of the mitogen-activated protein kinase and Akt pathways. Thus, the p85alpha protein may play a role in the down-regulation of activated receptors through its temporal control of the GTPase cycles of Rab5 and Rab4.

Down-regulation of signal transduction pathways activated by receptor tyrosine kinases, such as the PDGFR, 1 includes endocytosis of the activated receptor complex (1,2). Receptormediated endocytosis involves multiple vesicle fusion events that effectively deliver the receptor-signaling complex to the early endosome. This complex is then disassembled and the receptor is either recycled back to the plasma membrane or sorted to the late endosome and lysosome for degradation (3,4).
Rab5 is a small monomeric GTPase involved in early endosomal fusion events such as the fusion of clathrin-coated vesicles (containing activated receptors undergoing endocytosis) with the early/sorting endosomes (reviewed in Refs. 5 and 6). GDP-bound Rab5 is inactive and bound to a guanine dissociation inhibitor (GDI) protein in the cytosol. GTP-bound Rab5 is active and localized on the cytoplasmic face of early/sorting endosomes where it is involved in binding specific effector proteins such as the early-endosomal autoantigen 1 (EEA1) (7,8). EEA1 is a cytosolic protein that is recruited to early endosomal membranes by binding to Rab5-GTP and the lipid product of the class III PI3K p150/hVPS34, phosphatidylinositol 3Ј-phosphate (7,9,10). EEA1 is a core component required for early endosomal fusion events (7,8,11,12). The half-life of phosphatidylinositol 3Ј-phosphate, and the duration of Rab5-GTP have been suggested to influence the rate and extent of the endosome fusion reaction (13). The low intrinsic GTPase activity of Rab5 hydrolyzes the bound GTP to GDP, and GAP proteins enhance this rate. When Rab5 is GDP-bound it is unable to interact with EEA1 and mediate membrane fusion events. Rab5-GDP binds GDI, dissociates from the early endosomes, and is again localized in the cytoplasm. Thus, Rab5 GTPase activity has been described as a critical and ratelimiting component of the docking and fusion process of the endocytic pathways (14). This suggests that proteins governing the nucleotide-bound state of Rab5 (GDIs, guanine nucleotide exchange factors, and GAPs) may have a regulatory role in endocytosis (15).
During endocytosis of the activated PDGFR, a class I PI3K remains associated with the receptor (2), consisting of an 85-kDa (p85␣) regulatory subunit and a 110-kDa (p110␣ or ␤) catalytic subunit. PI3K activity is important for generating specific phospholipids, phosphatidylinositol 3,4-bisphosphate and phosphatidylinositol 3,4,5-trisphosphate, involved in recruiting and activating downstream signaling proteins (reviewed in Ref. 16). Mutation of the PI3K binding sites on the PDGFR disrupts receptor trafficking, such that PDGFRs are initially internalized but undergo recycling back to the cell membrane instead of targeting to the lysosome (17,18). Thus, PI3K binding sites on the PDGFR are not required for initial receptor internalization events, or recycling, but are required to divert the PDGFR to a degradative pathway. This effect is specific for PI3K binding sites because mutant PDGFRs lacking binding sites for other signaling proteins (p120GAP, PLC␥1, and Syp) are endocytosed and degraded as wild type PDGFRs (18). The net effect is an increase in the number of tyrosine-phosphorylated PDGFRs at the membrane, which has been suggested to cause the sustained activation of downstream proliferative pathways (18). This has also led to the suggestion that class I PI3Ks (p85/p110) may function to regulate endocytosis and intracellular protein sorting.
The p85␣ protein has been described as a non-catalytic regulatory protein, however, it contains a GAP-like BH domain with sequence homology to RhoGAP proteins involved in regulating Rho GTPases, such as Cdc42 and Rac1. GAP domains are involved in stimulating the down-regulation of G proteins by accelerating their intrinsic GTPase activity. The p85␣ protein can bind to both Rac1-GTP and Rac1 (no nucleotide), but not Rac1-GDP (19). Whereas p85 binds preferentially to Cdc42-GTP, some binding to Cdc42-GDP was also noted (19). When tested at nanomolar concentrations, the p85␣ BH domain did not possess GAP activity toward Cdc42 and Rac1 (19,20). More recently, the GAP-like BH domain of p85␣ has been crystallized and its structure reveals a distinct folding pattern capable of binding a G protein (21). The ability of the folded BH domain of p85␣ to bind G proteins lends support to the suggestion that it may possess GAP activity toward a GTPase protein, which has yet to be identified.
In this report, we demonstrate that the BH domain of p85␣ has GAP activity toward several Rab GTPases, in particular Rab5 and Rab4. This is the first report of a catalytic activity for the regulatory subunit of PI3K, p85␣, and suggests that it may regulate vesicular membrane fusion events mediated by Rab5 and Rab4. We also find that p85 has GAP activity toward the Rho GTPases Cdc42 and Rac1, proteins involved in regulating actin structures. The ability of p85 to regulate G proteins involved in vesicle fusion events (Rabs) and vesicle movement (Rhos) suggest that p85 may play a central role in coordinating receptor endocytosis with cell signaling via activated receptors. In support of this suggestion, expression of p85 proteins defective for GAP activity show both reduced PDGFR degradation and sustained activation of pathways downstream of the receptor.

EXPERIMENTAL PROCEDURES
Expression and Purification of GST Fusion Proteins-The glutathione S-transferase (GST)-Rab5 (wild type and mutants S34N and Q79L), GST-Rab4 and GST-Rab6 contain the human Rab sequences and were a generous gift from G. Li (University of Oklahoma (22)). Human Rab11, Rac1, and Cdc42 were obtained by reverse transcription (Promega kit) from total RNA obtained from HeLa cells, followed by PCR with Pfu (Stratagene). Primers were used that incorporated a 5Ј BamHI and a 3Ј EcoRI site immediately upstream of the start, and downstream of the stop codon, respectively. The reverse transcriptase-PCR products were digested with BamHI and EcoRI and cloned into similarly digested pGEX6P1 (Amersham Biosciences). The integrity of the Rab11, Rac1, and Cdc42 sequences were completely verified by DNA sequencing. GST-p85(NϩC)SH2 was generated by PCR as described (23), and contains amino acids 314 -724 of bovine p85␣. The GST-p85⌬BH protein (amino acids 1-83 and 314 -724) was generated by subcloning the sequences encoding p85(NϩC)SH2 after those encoding GST-p85SH3. The remaining GST-p85 fusion proteins have been described previously (23). Site-directed mutagenesis of the GST-p85 plasmid, to introduce separate single amino acid changes of Arg 151 to Ala and Arg 274 to Ala, were carried out using the QuikChange method (Stratagene), according to the manufacturer's directions. DNA sequencing to ensure that no additional mutations had been introduced verified the entire p85 coding region.
GST fusion proteins were expressed and purified as described (23). The Rab5, Rab4, and Rab6 proteins were cleaved from GST using factor Xa (Amersham Biosciences) and the Rab11, Rac1, and Cdc42 proteins were cleaved from GST using PreScission protease (Amersham Biosciences), following the manufacturer's instructions. The majority of the full-length and fragments of p85 were generated by cleaving them from GST using thrombin (Sigma), following the manufacturer's instruc-tions. Proteins were concentrated using Amicon ultracentrifugal filter devices (Millipore). Protein concentrations were determined using a Bradford assay (Bio-Rad). The full-length p85 used in Fig. 1 was expressed after subcloning the p85 insert into the pGEX6P1 vector, and was liberated from GST using the PreScission protease as described above. A similar PreScission-cleaved p85 protein, with an additional six histidine residues at the carboxyl terminus, was used for the enzymelinked immunosorbent assays (ELISA).
GAP Assays-The majority of the Rab GAP assays were steady state assays carried out for 10 min as previously described (22). Briefly, Rab (200 ng) was preloaded with [␣-32 P]GTP (85 nM) for 30 min at 25°C in loading buffer (20 l; 20 mM Tris-HCl, pH 8, 2 mM EDTA, and 1 mM dithiothreitol). The reaction was started by the addition of MgCl 2 (to 10 mM), with and without added p85 protein. After 10 min, the reaction was stopped with the addition of stop solution (to 0.2% SDS, 5 mM EDTA, 5 mM GDP, 5 mM GTP) by heating at 65°C for 2 min. The GTP and GDP were separated by thin layer chromatography. Results were visualized using a phosphorimager and quantified using Quantity One software (Bio-Rad). The femtomoles of hydrolyzed GDP were calculated using the ratio of radiolabeled GDP to GTP on the thin layer chromatography plate and the amount of starting [␣-32 P]GTP. Results were graphed using Prism software (GraphPad Software, Inc., San Diego, CA) and are reported as the average of at least three independent determinations Ϯ S.E. Where error bars are not visible, they are smaller than the symbols on the graph. A single turnover GAP assay ( Fig. 1C only) was used to measure the catalytic rate of GTP hydrolysis by including a large excess of unlabeled GTP (to 1.7 mM) with the MgCl 2 step that initiates the GTP hydrolysis reaction. This assay used 10 M p85 protein (for samples with p85) and was carried out for the indicated times.
ELISA-based Binding Assays-Purified Rab5 and p85 proteins that had been cleaved from GST (see above) were used in an ELISA-based binding assay as described (24), with the following changes. Rab5 was stripped of nucleotide and loaded with, no nucleotide, GDP, or GTP␥S as described (25), prior to binding it to the wells. To detect bound p85 protein, an anti-p85 antibody (23) was used (0.25 g/ml), followed by an anti-rabbit horseradish peroxidase-conjugated secondary antibody (Santa Cruz Biotechnology; 0.125 g/ml). As a control, GST protein was also tested to determine nonspecific binding to the Rab5 protein and detected as described (24). At least two independent experiments, each containing duplicate determinations, were averaged and plotted using Prism software. Standard errors are indicated with error bars. Where no error bars are visible, the standard error is less than the size of the symbol on the graph.
Cell Culture, Transfections, and Mutant p85-expressing Cell Lines-NIH 3T3 and COS-1 cells were maintained as described previously in Dulbecco's modified Eagle's medium containing 100 units of penicillin G and 100 g of streptomycin per ml supplemented with 10% fetal bovine serum in a humidified incubator at 37°C with 5% CO 2 (23,26).
The HA3 vector (23), originally based on the Invitrogen vector (pRc/CMV2), was modified to a pFLAG3 vector that encoded three copies of the FLAG epitope (DYKDDDDK) upstream of the multiple cloning site. The small HindIII-BglII fragment from the HA3 vector was replaced by the following sequence in the new pFLAG3 vector: 5Ј-AA-GCTTCCACCATGGACTACAAGGACGACGATGACAAGGCTAGTGA-CTACAAGGACGACGACGATAAAGCGGCCGCTGATTACAAGGACG-ACGACGATAAAGCTAGCAGATCT-3Ј and was verified by DNA sequencing. Inserts encoding full-length wild type p85, p85⌬BH, p85R151A, and p85R274A were excised from the pGEX vectors described above, using BamHI and EcoRI and subcloned into BglII-and EcoRI-digested pFLAG3.
The pFLAG3 vector was modified to a pMyc3 vector that encoded three copies of the Myc epitope (EQKLISEEDL) upstream to the multiple cloning site. The small HindIII-NheI fragment from the pFLAG3 vector was then replaced by the following sequence in the new pMyc3 vector: 5Ј-AAGCTTCCACCATGGAACAGAAACTGATCAGCGAAGAG-GATCTGCTGAGCGAGCAGAAACTGATCAGCGAGGAAGAACTGGC-CGCGGAACAGAAACTGATCAGCGAAGAGGATCTGGCTAGC-3Ј. A pair of oligonucleotides (5Ј-AATTCGCGGCCGCGGGCC-3Ј and 5Ј-CG-CGGCCGCG-3Ј) were ligated into EcoRI-ApaI-digested FLAG-modified vector just described, to alter the multiple cloning site such that it now contained a new unique NotI site between these two restriction sites. This entire region of the new pMyc3 vector was verified by DNA sequencing. An insert encoding full-length wild type mouse p110␣ was amplified by PCR from the pCMV-p110myc plasmid (generously provided Dr. L. T. Williams (University of California, San Francisco, CA)) as a template. The BamHI-NotI-digested insert DNA was subcloned into the BglII-NotI-digested pMyc3 vector. The entire p110␣ coding sequence was verified by DNA sequencing. Transient transfections of FLAG-p85 Ϯ Myc-p110␣ plasmids into COS-1 cells were carried out using LipofectAMINE (Invitrogen) according to the manufacturer's directions as detailed previously (23).
NIH 3T3 cells were transfected (23) with the pFLAG3-p85 DNA and selected in G418 (400 g/ml). Drug-resistant clones were expanded and screened for FLAG-p85 protein expression, using a Western blot analysis with anti-FLAG M2 antibodies (Sigma; 5 g/ml) as previously described (23). Cells were stimulated with PDGF BB (50 ng/ml) for the indicated times, washed, and lysed as described previously (26).
PI3K Assay-Samples were immunoprecipitated using either anti-PDGFR antibodies (5 g, Santa Cruz Biotechnology) or anti-FLAG M2 (10 g, Sigma) as described previously (26). Immunoprecipitates were washed with each of the following: wash 1 (phosphate-buffered saline), wash 2 (100 mM Tris-HCl, pH 7.4, 50 mM LiCl), and wash 3 (10 mM Tris-HCl pH 7.4, 100 mM NaCl, 1 mM EDTA). Excess liquid was removed from the immunoprecipitates. Lipid micelles were generated by sonicating phosphatidylserine and phosphatidylinositol in PI3K assay buffer (25 mM Hepes, pH 7.4, 10 mM MgCl 2 ) in a sonicating water bath for 20 min. Each sample was incubated with lipid micelles (5 g of phosphatidylserine ϩ 2.5 g of phosphatidylinositol) in PI3K assay buffer and 10 Ci of [␥-32 P]ATP in a total volume of 50 l for 15 min at 20°C while gently rocking. The reaction was stopped by the addition of HCl (to 1.7 M). Lipids were extracted into chloroform:methanol (1:1) and further washed with methanol, 1 N HCl (1:1). Reaction products were dried down, resuspended in chloroform:methanol (1:1), and spotted onto a thin layer chromatography plate (Silica Gel 60; VWR Canlab). Samples were developed in 1-propanol, water, acetic acid (17.4:7.9:1) in a chromatography chamber for 4 h, dried, and exposed to a phosphorimager screen. Results were visualized and quantified using Quantity One software (Bio-Rad), and the statistical analysis was performed using Prism software (GraphPad Software, Inc., San Diego, CA).

Rab5 GAP Activity Is Encoded in the BH Domain of p85␣-
The GAP activity of p85 toward Rab5 was determined using a GAP assay in which Rab5 was preloaded with [␣-32 P]GTP and assayed alone, or with increasing concentrations of p85 protein. The reaction was initiated by the addition of MgCl 2 , and the [␣-32 P]GTP and [␣-32 P]GDP were separated by thin layer chromatography, prior to detection using a phosphorimager (Fig.  1A). The femtomoles of radiolabeled GDP generated by Rab5 GTPase activity were plotted as a function of p85 concentration (Fig. 1B). The GAP activity of p85 prepared using the PreScission cleavage method (maximum of 900 fmol of GDP produced in 10 min) was found to be higher than that of the p85 liberated by thrombin cleavage, used in all subsequent figures (maximum of 300 fmol of GDP produced in 10 min). The p85 protein stimulated the hydrolysis of GTP to GDP by Rab5 in a concentration-dependent manner with a maximum stimulation of ϳ570 -1700-fold, depending upon the preparation of p85 protein.
To ensure that p85 could accelerate the catalytic rate of Rab5 GTP hydrolysis, we used a single turnover GAP assay. Rab5 was preloaded with [␣-32 P]GTP and the reaction was carried out in the presence of excess unlabeled GTP, in the presence or absence of added p85. Aliquots were taken at various time points and the amount of hydrolyzed [␣-32 P]GDP generated was visualized and quantified using a phosphorimager (Fig.  1C). The addition of p85 to the assay did substantially accelerate the Rab5-mediated hydrolysis of the bound [␣-32 P]GTP to [␣-32 P]GDP, suggesting that p85 stimulates GTP hydrolysis by Rab5 and is not simply promoting nucleotide exchange.
Given the sequence similarity between the BH domain of p85 and several other GAP domains (19,21), we suspected that it was the BH domain of p85 that encoded the observed Rab5 GAP activity. We isolated polypeptides encompassing different p85 domains ( Fig. 2A), and tested each for Rab5 GAP activity (Fig. 2B). As expected, the BH domain of p85 contained substantial Rab5 GAP activity. The p85 SH3 domain, the (NϩC)SH2 domains (which included the p110-binding domain of p85), and a mutant of p85 lacking the BH domain (⌬BH; missing amino acids 84 -313), showed little or no GAP activity toward Rab5. Although the isolated BH domain was not as active as the full-length p85 protein, these results suggest that it is the BH domain of p85 that possesses Rab5 GAP activity and that it can adopt a suitable conformation to be at least partially active when expressed as an isolated domain.
GAP proteins typically act in two ways to accelerate GTP hydrolysis of the GTPase (28). First, many GAP proteins contain an arginine finger that is positioned in the active site and p85 Subunit of PI3K has Rab5/4 GAP Activity stabilizes the transition state during the hydrolysis reaction. Second, GAPs bind to the switch II region of the GTPase, which stabilizes and orients the critical glutamine residue within the active site of the GTPase. The literature on GAP proteins and their folded structures suggests that it is often difficult to predict from a sequence alignment which arginine residue corresponds to the arginine finger. The BH domain of human p85 has been crystallized and a surface of the protein containing a cluster of conserved residues has been proposed as a G protein binding site (21). There are two arginine residues within this proposed G protein binding site within the BH domain of p85, Arg 151 and Arg 274 . We have individually mutated each of these residues to alanine in the context of the full-length p85 protein, and tested each for Rab5 GAP activity (Fig. 2C). The p85R151A protein has about 65% of the GAP activity of wild type p85, whereas the p85R274A mutant has less than 5% GAP activity remaining. The GAP activity of the p85R274A mutant protein was also assayed using the single turnover GAP assay and was found to have little or no GAP activity, as compared with the wild type p85 protein (Fig. 1C). Thus, while both arginine residues are required for full p85 Rab5 GAP activity, Arg 274 appears to play a more critical role than Arg 151 .
Direct Binding Between p85␣ and Rab5-The ability of p85 to stimulate Rab5 GTP hydrolysis suggested that p85 should be able to bind directly to Rab5. To confirm this, we carried out an ELISA-based protein binding assay using purified Rab5 and p85 proteins. To ensure that the observed binding was in fact direct, these proteins did not contain any GST sequences. Rab5 was preloaded with the indicated nucleotide and immobilized on an ELISA plate. After blocking, samples were incubated with increasing concentrations of purified p85 protein, or an irrelevant protein (GST). Bound p85 protein was detected using anti-p85 antibodies, whereas bound GST protein was detected with anti-GST antibodies (Fig. 3A). The control GST protein was not able to bind to Rab5. The p85 protein bound directly to Rab5 in a concentration-dependent manner, and this interaction was similar in the absence of nucleotide, or in the presence of GDP or GTP␥S (a non-hydrolyzable form of GTP).
As a complementary approach, we also used a pull-down assay with immobilized GST-Rab5 fusion proteins preloaded with different nucleotides (Fig. 3B). Purified cleaved p85 pro- Rab5-GDP (‚), or Rab5-GTP␥S (Ⅺ) were blocked and incubated with increasing concentrations of purified p85 protein (E, ‚, and Ⅺ) or control GST protein (•). Bound p85 (or GST) protein was detected using anti-p85 (or anti-GST) antibodies, followed by a secondary antibody conjugated to horseradish peroxidase and quantified by measuring the absorbance at 450 nm of the acidified product as described (24). B, GST and GST-Rab5 were immobilized on glutathione-Sepharose beads and loaded with the indicated nucleotide. Purified PreScission-cleaved p85 was added and bound p85 was detected after washing, using an immunoblot analysis with anti-p85 antibodies. C, GST and GST-Rab5 mutants known to selectively bind GDP (S34N) or GTP (Q79L) were used in a pull-down assay with wild type p85 protein, as described for panel B. D, the indicated p85 mutants were tested for their abilities to bind to GST (control), GST-Rab5S34N-GDP, and GST-Rab5Q79L-GTP as described above.
p85 Subunit of PI3K has Rab5/4 GAP Activity tein bound best to Rab5 (no nucleotide), followed by Rab5-GTP␥S. Significant amounts of p85 also bound to the transition state analogue conformation of Rab5 (GDP-AlF 4 ), and to Rab5-GDP. We found similar results using a pull-down assay with two Rab5 mutants. The Rab5S34N mutant preferentially binds GDP, whereas the Rab5Q79L mutant is unable to hydrolyze bound GTP. Purified cleaved p85 bound to both GST-Rab5S34N and GST-RabQ79L preloaded with GDP and GTP, respectively, which were immobilized on glutathione-Sepharose beads (Fig. 3C). This result indicated a direct interaction between p85 and Rab5, which was not significantly altered by the nucleotide-bound state of Rab5.
The ability of the p85 mutants deficient for GAP activity (⌬BH, R151A and R274A) to bind to Rab5S34N-GDP and Rab5Q79L-GTP was also determined (Fig. 3D). All three mutants, including the p85 protein lacking the entire BH domain, still bound some form of Rab5 suggesting that domains other than the BH domain of p85 participate in Rab5 binding. The p85R151A protein, which retains 65% of its GAP activity, bound to the GDP-and GTP-bound forms of Rab5, much as the wild type p85 protein. In contrast, the two BH mutations (⌬BH and R274A) that knockout the GAP activity of p85 have large changes in their abilities to bind Rab5. The p85⌬BH mutant only bound Rab5Q79L-GTP, whereas the p85R274A mutant only bound Rab5S34N-GDP.
The p85 Protein Has GAP Activity toward Some Rab and Rho Family Proteins-To test whether the GAP activity of p85 was specific for the Rab5 GTPase, or would also stimulate the GTPase activities of other Rab proteins, p85 was added to GAP assays containing each of Rab4, Rab6, and Rab11 (Fig. 4A). Rab4 and Rab11, like Rab5, are involved in endocytic vesicle fusion events, whereas Rab6 is involved in membrane fusion events important during retrograde transport from the Golgi back to the endoplasmic reticulum. The p85 protein displayed similar GAP activity toward Rab5 and Rab4, and of the Rab proteins tested, these were the GTPases most stimulated by p85 (Fig. 4A). The GTPase activity of Rab6 was also stimulated by p85, but to a lesser degree, whereas Rab11 GTPase activity was not significantly altered by p85 addition. These results suggest that p85 is a selective GAP protein for several different Rab family GTPases; in particular, it can stimulate the downregulation of Rab5 and Rab4.
Previous studies have tested the ability of p85 to act as a GAP protein toward Rac1 and Cdc42, two Rho GTPases involved in the regulation of the actin structures (19,20). At concentrations of p85 from 0.1 to 1 M, p85 was not found to possess GAP activity toward Rac1 or Cdc42. There have, however, been reports of GAP proteins (e.g. ArfGAP; see "Discussion") that require micromolar concentrations to stimulate GTP hydrolysis toward their cognate G protein. Because the concentration of p85 used in our Rab GAP assays was typically in the 10 -20 M range, and given the precedence for some GAP proteins requiring higher concentrations, we retested p85 for GAP activity toward Rac1 and Cdc42 (Fig. 4B). We observed that p85 did have GAP activity toward both Rac1 and Cdc42, to a similar extent as toward Rab5 and Rab4. Thus, while p85 does not act as a general GAP toward all Rab proteins (i.e. not Rab11, and to lesser extent toward Rab6), it is able to regulate the GTP hydrolysis of some Rab proteins and also some Rho proteins.
Expression of p85 BH Mutants in Cells-To determine the effects of mutations within the BH domain of p85 on cellular functions, we first needed to establish whether these mutations also compromised other known functions of p85 such as binding to the catalytic subunit of PI3K, p110, and growth factor-dependent receptor binding. COS-1 cells were cotransfected with a Myc epitope-tagged p110␣ (Myc-p110␣), and either a wild type or mutant version of a FLAG epitope-tagged p85 (FLAG-p85). Lysates from these transfected cells were confirmed to express both Myc-p110␣ and FLAG-p85 fusion proteins, as determined using an immunoblot analysis with anti-Myc and anti-FLAG antibodies (Fig. 5A). Anti-FLAG immunoprecipitates of these cell lysates were divided in half, with half of the samples immunoblotted to test for the presence of Myc-p110 protein, followed by FLAG-p85 protein (Fig. 5B), and the other half subjected to a PI3K assay (Fig. 5C). As a control, anti-p85 immunoprecipitates from untransfected cells were also assayed for PI3K activity. These results indicated that all of the FLAG-p85 proteins (wild type, ⌬BH, R274A, and R151A mutants) can associate with the Myc-p110␣ protein and its encoded PI3K activity.
NIH 3T3 cells were also transfected with wild type FLAG-p85 or mutants lacking the GAP activity encoded by the BH domain (FLAG-p85⌬BH and FLAG-p85R274A). Cell lines stably expressing these FLAG-p85 proteins were treated with PDGF BB for various times. The PDGFR typically binds p85 in a PDGF-dependent manner, as shown by the presence of p85 in anti-PDGFR immunoprecipitates from PDGF-treated NIH 3T3 cell lysates (Fig. 6A). Similarly, we find that anti-PDGFR immunoprecipitates contain FLAG-p85 proteins (both wild type and BH mutants ⌬BH and R274A) after PDGF stimulation. PDGFR immunoprecipitates from all of these cell lines also contain substantial amounts of associated PI3K activity after PDGF treatment (Fig. 6B). Thus, whereas p85⌬BH and p85R274A have severely decreased Rab5 GAP activity, they are still able to bind the catalytic subunit of PI3K, p110, and associate with the PDGFR in a PDGF-dependent manner as the wild type p85 protein can.
Expression of the p85 BH Mutants in Cells Alters PDGFR Down-regulation, as well as MAPK and Akt Signaling-To determine the effects of deletion of the p85 BH domain on PDGFR down-regulation and cell signaling pathways, we characterized the NIH 3T3 cells stably expressing FLAG-p85, FLAG-p85⌬BH, and FLAG-p85R274A (Fig. 7A). Cells were treated with PDGF for various times and the amount of p85 Subunit of PI3K has Rab5/4 GAP Activity PDGFR protein was analyzed using Western blot analysis (Fig.  7B). Untransfected NIH 3T3 cells that are stimulated with PDGF for increasing times typically show a rapid reduction in PDGFR protein levels because of the degradation of receptor protein as part of the down-regulation mechanism. By 60 min the majority of the receptor has been degraded (Fig. 7B, lower). Similar results are obtained for NIH 3T3 cells expressing wild type FLAG-p85. In contrast, cells expressing the FLAG-p85⌬BH or FLAG-p85R274A mutants show a delay in the PDGFR degradation profile with the majority of the receptor still present at 60 min or longer (Fig. 7B, lower). The corresponding anti-phosphotyrosine blots (Fig. 7B, upper) show enhanced tyrosine phosphorylation of a protein that likely corresponds to the PDGFR, most markedly in cells expressing p85R274A. Thus, mutations of the p85 BH domain impaired the ability of the receptor to be degraded and increased the half-life of the activated PDGFR.
To assess if the enhanced activated PDGFR levels at the later time points gave rise to enhanced receptor signaling, the levels of activated Akt and activated MAPK were determined (Fig. 7, C and D). Cells expressing FLAG-p85⌬BH or FLAG-p85R274A showed minor increases in the generation of phospho-Akt during the PDGF stimulation time course, when compared with the parental NIH 3T3 or FLAG-p85 expressing cells (Fig. 7C). However, striking differences were observed for activated MAPK (Fig. 7D). Cell expressing FLAG-p85⌬BH or FLAG-p85R274A showed enhanced MAPK activation even prior to PDGF treatment, and maintained high levels of phospho-MAPK (pMAPK) for 60 min or longer (Fig. 7D, upper). This result suggests that MAPK signaling pathways remain active while PDGFR levels are high in FLAG-p85⌬BH-or FLAG-p85R274A-expressing cells. These results are consistent with a previous report that suggested that MAPK signaling, but not PI3K signaling, is dependent upon internalized PDGFR (29). DISCUSSION Whereas the BH domain of the p85 protein has been known to have sequence homology to the GAP domains of other proteins, this is the first report demonstrating that this domain possesses GAP activity. The fact that an isolated p85 BH domain has significant Rab5 GAP activity and that the p85⌬BH protein, which lacks the BH domain and also lacks activity, firmly establishes that the p85 BH domain encodes Rab5 GAP activity. We have also generated a point mutant of p85 within the BH domain (R274A) with severely reduced GAP activity. Our results further show that p85␣ has the ability to stimulate the GTPase activity of several, but not all of the Rab family members tested. This suggests that the GAP activity of p85 is specific to a subset of Rab proteins. Of the Rab GTPases tested is this study, p85 displays the highest GAP activity toward Rab5 Ͼ Rab4 Ͼ Ͼ Rab6, with little or no activity toward Rab11. Rab5 has been shown to function in membrane fusion events between clathrin-coated vesicles and early endosomes, as well as in homotypic fusions between early endosomes (30,31). Rab4 is suggested to play a role in fast recycling (2-5 min) from early endosomes back to the plasma membrane, whereas Rab11 is believed to be important for slow recycling (15-30 min) of receptors, via intermediate recycling endosomes (32). In contrast, Rab6 is associated with transport events in Golgi and trans-Golgi network membranes (33). FIG. 5. The p85 BH domain mutants retain their ability to bind to p110 and associate with PI3K activity. COS-1 cells were cotransfected with the indicated FLAG-p85 encoding plasmid together with one encoding Myc-p110␣, or left untransfected as a control. A, lysates (20 g of total protein) from each were immunoblotted with anti-myc antibodies, stripped, and reprobed with anti-FLAG antibodies. B, lysates (200 g of total protein) were immunoprecipitated (IP) with anti-FLAG antibodies and immunoblotted with anti-myc antibodies, stripped and reprobed with anti-FLAG antibodies. C, anti-FLAG immunoprecipitates were assayed for associated PI3K activity. Radiolabeled PI3K lipid products were resolved by thin layer chromatography and visualized using a phosphorimager. As a positive control, an anti-p85 immunoprecipitate of lysates from untransfected NIH 3T3 cells was also assayed for associated PI3K activity.
FIG. 6. The p85 BH domain mutants retain their ability to bind to activated PDGFRs and do not prevent PDGF-dependent association of PI3K activity with the receptor. NIH 3T3 cells stably expressing the indicated FLAG-p85 proteins were stimulated for various times with PDGF BB. A, lysates (200 g) from each cell line were immunoprecipitated (IP) with anti-PDGFR antibodies and immunoblotted with anti-FLAG antibodies (for cells expressing FLAG-p85, FLAG-p85⌬BH, and FLAG-p85R274A), or with anti-p85 antibodies (for untransfected NIH 3T3 cells). B, lysates from cells that had been stimulated with PDGF BB (5 min; ϩ) or not (Ϫ) were immunoprecipitated with anti-PDGFR antibodies and were assayed for PDGFR-associated PI3K activity.

p85 Subunit of PI3K has Rab5/4 GAP Activity
We also find that p85 has GAP activity toward Cdc42 and Rac1, important regulators of actin structures. Cdc42 in particular has been suggested to play a role in regulating the movement of vesicles by stimulating actin comet-tail formation (34). It has been hypothesized previously (35) that there may be multifunctional scaffolding proteins that can link endocytosis to the actin cytoskeleton. We suggest that p85 may be an example of this because its GAP activity toward these Rho family GTPases could allow the coordination of cell signaling by activated receptors with vesicle movement during receptormediated endocytosis.
Many GAP proteins bind preferentially to the GTP-bound form of their respective G protein as would be expected, because GAP proteins act on the GTP-bound forms. There are also GAP proteins that bind well to GDP-bound, GTP-bound, or G protein in the absence of nucleotide. Some examples of these latter GAP proteins include BNIP-2 binding to Cdc42 (36) and the RhoGAP protein binding to GTP-bound and GDP-bound forms of Cdc42, Rac1, and RhoA (37). Thus, whereas some GAP proteins bind preferentially to the GTP-bound form of a G protein, clearly this is not always the case. Perhaps the low affinity interactions between some G proteins and their respective GAP proteins are subject to additional regulation by subcellular location, or additional cellular factors.
Certainly this appears to be the case for ADP-ribosylation factor (Arf) and ArfGAP, which bind with low affinity and exhibit modest GAP activity in the absence of phospholipid or coatomer protein. Arf GTPases regulate coatomer (nonclathrin) recruitment during vesicle transport and so functionally they are very similar to the Rab GTPases (38). Arf1-GTP forms a stoichiometric complex with the coatomer protein and this interaction is required to maintain the stability of the vesicle coat (39). The relatively weak GAP activity of ArfGAP toward Arf1-GTP is enhanced 10 -1000-fold by the presence of coatomer protein or phospholipids (39,40). This has been suggested to be a proximity effect because the tripartate ArfGAP⅐Arf1-GTP⅐coatomer complex, or colocalization of ArfGAP and Arf1-GTP to the same phospholipid vesicles, effectively creates a high local concentration of both proteins. The requirement for high concentrations of both ArfGAP and Arf1-GTP ensures that Arf1-GTP levels are sustained during the assembly of vesicle coats (38,39).
Our results suggest that Rab5 binds p85 directly in a low affinity interaction independent of its nucleotide-bound state. A previous report has shown that p110␤ can bind selectively to GTP-bound Rab5 in the absence of p85␣ (41). Binding between p85␣ and Rab5-GTP was only detected in the presence of p110␤, consistent with a role for p110␤ in promoting the formation of a tripartate complex p85␣⅐Rab5-GTP⅐p110␤ (41). In conjunction with our p85␣⅐Rab5 binding results, these results suggest that either p110␤ increases the affinity of p85␣ for Rab5-GTP through a conformational change in p85␣ or that the additional contacts between p110␤ and Rab5-GTP help to bring p85␣ and Rab5-GTP together. These possibilities are currently under investigation. Regardless of the outcome of these studies, there is a significant pool (at least 30%) of p85 in cells that is not associated with p110 (42,43). This so-called "monomeric" p85 has been suggested to compete with p85/p110 for binding to activated receptors as a mechanism to attenuate activation of the PI3K pathway. Thus, the GAP activity of monomeric p85 may be sufficient to regulate the Rab GTPases in the absence of associated p110 in cells.
The current model for the regulation of Rab5 suggests that the delivery of Rab5-GDP to early endosomal membranes and the displacement of GDI are mediated by a GDI displacement factor in conjunction with unidentified targeting factors that may aid in Rab5-GDP recruitment (44). A guanine nucleotide exchange factor then stimulates the activation of Rab5 by promoting the exchange of GDP for GTP. Then Rab5-GTP recruits effector proteins such as EEA1 to the membrane and vesicle fusion events are initiated. Rab5 hydrolyzes the bound GTP to GDP with the aid of GAP proteins, returning Rab5 to an inactive state and resulting in the re-extraction of Rab5 from the membrane by binding to GDI.
The expression of FLAG-p85R274A has a more pronounced effect in reducing PDGFR degradation, as well as sustaining receptor activation (i.e. tyrosine phosphorylation) and downstream signaling, when compared with expression of FLAG-p85⌬BH. Whereas both of these mutant proteins have about 5-8% GAP activity remaining, they display activation statespecific binding preferences for Rab5 that are different from each other, and distinct from the wild type p85 protein.
Whereas wild type p85 binds to both the inactive GDP-bound Rab5 and the active GTP-bound Rab5, FLAG-p85R274A binds only Rab5-GDP and FLAG-p85⌬BH binds only Rab5-GTP.
We have incorporated our results regarding the binding and regulation of Rab5 by p85, into the current model for the regulation of Rab5 and endosomal membrane fusion events to explain these effects on PDGFR signaling and degradation. We speculate that the receptor-associated wild type p85 may bind p85 Subunit of PI3K has Rab5/4 GAP Activity Rab5-GDP to act as a targeting factor for membranes containing activated receptors undergoing endocytosis. Additionally, wild type p85 may bind Rab5-GTP to act as a GAP protein to limit the time that Rab5 is in an active GTP-bound state able to promote membrane fusion. The FLAG-p85⌬BH protein does not bind Rab5-GDP, so we suggest that the assistance provided by the p85 protein in recruiting Rab5-GDP to endosomal membranes may be lost in cells expressing this mutant. Because Rab5-GDP can still be recruited to endosomal membranes by binding to the GDI displacement factor, this may have a relatively small effect in reducing the efficiency of Rab5-GDP membrane recruitment. After Rab5 nucleotide exchange, the FLAG-p85⌬BH protein can bind to Rab5-GTP, but can only weakly stimulate Rab5-mediated hydrolysis of GTP to GDP. Thus Rab5 remains in its active GTP-bound state for a longer time, promoting membrane fusion events.
In contrast, the FLAG-p85R274A binds only Rab5-GDP and could therefore still effectively act together with the GDI displacement factor to efficiently target Rab5-GDP to membranes containing activated receptor-p85 complexes. However, because this mutant does not bind Rab5-GTP, it would fail to stimulate Rab5-mediated hydrolysis of GTP to GDP even with its residual low GAP activity. Rab5 would thus remain active even longer because the intrinsic rate of Rab5 GTP hydrolysis is minimal. If this model is correct, the net effect would be more membrane fusion events during receptor endocytosis (Rab5mediated), and likely also more membrane fusion events during rapid recycling (Rab4-mediated). The net effect would be less opportunities for receptor complexes to be sorted to a degradative pathway.
Rab5-GTP regulates vesicle fusion events during receptormediated endocytosis by binding the EEA1 effector protein, which is necessary to tether vesicles during fusion events (7,8). Thus, similar to Arf1-GTP, Rab5-GTP is a stoichiometric component required for early endosomal membrane fusion events. By requiring high concentrations of the p85 GAP protein to stimulate Rab5 GTP hydrolysis, the levels of Rab5-GTP can be sustained during endocytic membrane fusion events. Because p85 remains associated with activated receptors during endocytosis, a high local concentration of receptor-associated p85 and vesicle-associated Rab5-GTP would accumulate in the early endosome. At this point active Rab5-GTP would no longer be necessary (i.e. after delivery of the receptor to the early endosome) and after p85 GAP-stimulated GTP hydrolysis, Rab5-GDP could be released from the early endosomal membrane, and bind GDI.
Active Rab4-GTP promotes recycling of receptors from the early endosome back to the plasma membrane, so the Rab4 GAP activity of p85 could slow this process and increase the time receptors spend in early endosomes. Cells expressing the p85 BH mutants (i.e. p85⌬BH or p85R274A) lacking Rab5 and Rab4 GAP activity would be expected to undergo more rapid Rab5-mediated endocytosis and more rapid Rab4-mediated recycling, providing a shortened time for PDGFRs within early endosomes. It has been suggested that the sorting of endocytosed receptors is the result of repeated low efficiency sorting events rather than from a single sorting event in which all receptors are diverted to the lysosome (45). Thus, the length of time a receptor is present in the early endosome could influence whether receptor targeting and ultimately degradation take place. It then follows that regulatory proteins controlling early endosomal membrane fusion or fission events would have the ability to alter the proportion of receptor that is degraded or recycled. In support of this hypothesis, we observed a decreased rate of PDGFR degradation in cells expressing p85 BH mutants lacking Rab5/4 GAP activity. Thus, p85␣ may play a role in the regulation of receptor sorting functions within early endosomes through its temporal control of the GTPase cycles of Rab5 and Rab4.