Identification of Two SH3-binding Motifs in the Regulatory Subunit of Phosphatidylinositol 3-Kinase*

Src homology 3 (SH3) domains have been recently shown to bind to proline-rich sequences contained in 3BP1,3BP2, and SOS. In a recent study we demonstrated that phosphatidylinositol 3-kinase (PI 3-kinase) associates with the Fyn SH3 domain. Here we show that p86, the regulatory subunit of PI 3-kinase, binds directly to the SH3 domains of Abl, Lck, Fyn, and p86 itself. An examination of p86 amino acid sequence revealed two proline-rich sequences in its N-terminal region similar to those present in 3BP1,3BP2, and SOS. To test whether these sequences mediate the association of p86 with SH3 domains two peptides with amino acid composition cor- responding to the p86a proline-rich sequences were synthesized and used in competition assays. Both peptides worked equally well in inhibiting the binding of PI 3-ki-nase activity and p86a to Fyn SH3 domain, whereas a control peptide had no effect. These results indicate that, as in 3BP1 of Lck, Fyn, Abl, and p85 3-kinase. These data suggest that the peptides do not specifically affect SH3-containing proteins. These results indicate that PI 3-ki-matic activity of PI 3-kinas.e during precipitation, probably by blocking elements: one that is dependent on tyrosine phosphorylation However, these peptides had no effect in the activity of soluble PI itself, suggesting that PI 3-kinase may be a target Of the soluble enzymatic activity but that they rather preserve the enzy-naSe may use two modes for its interaction with upstream hydrophobic sites on the beads that cause inactivation.

Protein-protein interactions have been recently implicated as a major mechanism by which intracellular signaling pathways communicate with each other (1,2). In recent years two elements have been identified that play major roles in regulating protein-protein interactions: the Src homology 2 and 3 domains (SH2 and SH3 domains) (3). They are present in a wide variety of proteins, some that contain intrinsic enzymatic activity as in the members of the Src family of tyrosine kinases, phospholipase Cy and GTPase-activating protein (GAP),' or in adaptor proteins such as the regulatory subunit of PI 3-kinase (~851, c-Crk, Nck, and the Sem-5IGRB2 molecules (reviewed in Grants GM 36624 and GM 41890 (to L. C. C.) and by funds from the *This work was supported by the National Institutes of Health Lucille P. Markey Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ase; HPLC, high performance liquid chromatography; TBS, Tris-buffered saline; PI, phosphatidylinositol; gPI-3-P, glycerol-phosphorylinosi-to1 3-phosphate; gPI-4-P, glycerol-phosphorylinositol 4-phosphate; PI-3-P, phosphatidylinositol 3-phosphate; PIP, phosphatidylinositol phosphate. 3050; F a : 617-278-3033.
The function of SH3 domains has been less well characterized to date. These domains have been implicated in regulating cytoskeletal functions, because they have been identified in several proteins associated with the cytoskeleton in both yeast and higher eukaryotes (myosin, spectrin, BEM 1, ABP1, FUS 1, and in a ras guanine nucleotide exchange factor, CDC25) (3,8). I t has also been suggested that SH3 domains may play a negative regulatory function in the context of cytosolic protein-tyrosine kinases: mutations and deletion of the SH3 elements in Ab1 and Src lead to the activation of their transformation capacity (9)(10)(11)(12). Cicchetti et ul. (13) have cloned a protein by its ability to associate with the Ab1 SH3 domain. This protein, named 3BP1, contains sequence homology to rho-GAP, the Cterminal region of Bcr and n-chimerin, all proteins that contain intrinsic GTPase activity (181, indicating that 3BP1 is a GAP. Furthermore, the SH3 domains of Grb2/SEM-5/Drk have been found to associate with SOS, a guanine nucleotide exchange factor for ras (14-17). The interactions between Ab1 SH3 and 3BP1 have been mapped to a proline-rich 10 amino acid sequence (APTMPPPLPPVPP) in the C-terminal portion of the molecule (19). Amino acid substitutions have indicated that the proline residues are critical in the association of this sequence with the Ab1 SH3 domain. Four similar sequences have been identified in the C-terminal domain of SOS in the region that has been shown to associate with the SH3 domain of Grb2 (14, 15,17).
We have recently shown that p54"" (T) and ~5 6 ' '~ associate with PI 3-kinase activity via their SH3 domains (20,27). PI 3-kinase phosphorylates the D-3 position of the inositol ring of phosphatidylinositol and has been shown to be activated by a wide range of growth factors and other cellular activators (1). PI 3-kinase is a heterodimer complex composed of an 85-kDa protein (regulatory subunit) and a 110-kDa protein (catalytic subunit) (21)(22)(23)(24)(25). The 85-kDa subunit has two SH2 domains, one SH3 domain and a domain with homology to BCR and various rho and rac GAPS. In this work, we show that PI 3-kinase interacts with the SH3 domains of Abl, Lck, Fyn, and p85 via its regulatory subunit (p85). Examination of the p85 amino acid sequence revealed two proline-rich sequences homologous to 3BP1 and SOS, suggesting that these may be the contact sites for the SH3 domains. Two synthetic peptides containing these sequences were able to block the association of PI 3-kinase and p85a to a GST-Fyn SH3 fusion protein, indicating that these proline-rich sequences in p85 are indeed SH3-binding motifs. These results suggest that SH3-containing proteins may be involved in linking tyrosine kinases not only to ras

1927
The PI 3-Kinase p85 Subunit Contains lluo SH3-binding Sites signaling pathways but to other signal transduction elements as well.

EXF'ERIMENTAL PROCEDURES
Reagents-Tissue culture media was purchased from Life Technologies, Inc. Acrylagel and Bis-acrylagel were obtained from National Diagnostics (Manville, NJ). Phosphatidylinositol, glutathione-Sepharose 4B beads, [y32-PlATP, and silica gel plates were purchased from Avanti (Alabaster, AL), Sigma, DuPont NEN, and E. Merck (Germany), respectively. The HPLC system used is from Hewlett-Packard (Germany), and the anion exchange Partisphere-Sax column is from Whatman (Clifton, NJ). Anti-p85 polyclonal antibodies used in this study were raised against a GST fusion protein containing the N-terminal SH2 domain of rat p85a. Antibodies against the C-terminal SH2 domain of p85a were obtained from Transduction (Lexington, KY).
Cell Culture-The T lymphoblastoid cell line HPB-ALL was cultured in RPMI 1640, 10% heat-inactivated fetal calf serum containing 2 xm L-glutamine at 37 "C and 5%C02. Spodoptera frugiperda (SF9) cells were cultured as described elsewhere (26). SF9 cells were plated in 60-mm plates and infected with baculovirus-p85a (multiplicity of infection = 10). The cells were harvested 40-60 h postinfection, washed twice with phosphate-buffered saline, and stored either on the plates or as cell pellets at -70 "C and used as a source of p85a.
Preparation of Cell Lysates-HPB-ALL cells were pelleted and washed twice with ice-cold phosphate-buffered saline. The cell pellets were incubated in ice-cold lysis buffer (137 xm NaCl, 20 xm Tris-HC1 (pH 8), 1 m~ MgCl,, 1 xm CaCl,, 10% glycerol, 1% Nonidet P-40, 1 m~ vanadate, 10 m~ NaF, 1 IIIM dithiothreitol, 1 xm phenylmethylsulfonyl fluoride, 1 pg/ml leupeptin) for 30 min at 4 "C with constant rocking. The lysates were cleared by centrifugation at 15,000 rpm for 10 min followed by passage through 0.2-pm filters. Protein concentration was then determined (Bio-Rad micro-assay), and the lysates were used at a concentration of 2 rng/dsample. Bovine serum albumin was added to the lysates at a final concentration of 2 mglml. 3 x lo6 SF9 cells overexpressing p85a (SF9lp85a) were lysed in 500 pl of lysis buffer (same as above) and subjected to the same clearing procedure. Bovine serum albumin was also added at a final concentration of 2 mg/ml. 100-pl aliquots were used per sample.
Synthetic Peptides-Peptides corresponding to the sequences SPPT-PKPRPPRPLP (residues 82-96 of p85a) and ERQPAPALPPKPPKP (residues 300-314 of p85a) were synthesized on a MILLIGEN EXCELL using a polystyrene support (PAL) which gives an amide on the Cterminal upon cleavage with trifluoroacetic acid. The peptides were HPLC-purified. The control peptide sequence is CELPELDLDLDLE. The peptides were dissolved in water at a final concentration of 10 xm with the exception of the control peptide (3 xm) and stored in aliquots at -20 "C.
The reaction was carried out for 10 min at 37 "C and the lipids were extracted with 160 pl of chlorofodmethanol (1:l). The organic layer was collected and the reaction products analyzed both by thin layer chromatography (TLC) and HPLC as described in detail elsewhere (30).
Precipitation of p85a from SF91p85a Cell Lysates-SFWp85a cell lysates (one-fifth of a 60-mm plate-100 p l ) were incubated with immobilized GST and GST-SH3 fusion proteins (10 pg) for 15 min at 4 "C with constant rotation. An anti-p85 immunoprecipitation was performed in parallel as positive control. The beads were then washed three times with lysis buffer followed by the addition of 20 pl of 2 x loading buffer containing P-mercaptoethanol (31) and boiled for 5 min. The samples were loaded on a 10% acrylamide gel and upon completion of the run, electrophoretically transferred onto a nitrocellulose membrane. The membrane was blocked in TBS, 5% non-fat dry milk followed by 1-h incubation with an anti-p85 antibody (1:lOOO dilution) obtained from Transduction (generated against a fragment from the C-terminal region of p85a that does not cross-react with GST). The membrane was then washed three times (5 min) with TBST (TBS, 0.2% Tween 20) and incubated for another hour with goat anti-rabbit horseradish peroxidase (1:10,000 dilution). The membrane was then washed for at least 1 h with TBST and developed using the ECL kit from Amersham Corp.
Peptide Competition Assays-The peptides were preincubated with immobilized GST or GST-Fyn SH3 (10 pg) in a 100-pl volume for 5 min at 4 "C with constant rocking. 100-pl aliquots of the cell lysates (HPB-ALL or SF91p85a) were then added for a further 10-min incubation at 4 "C with rocking. The beads incubated with HPB-ALL cell lysates were washed and subjected to lipid kinase assay as described above, whereas the beads incubated with SF9lp85a cell lysates were harvested and analyzed by immunoblot for the presence of p85a in the precipitates as above. The nitrocellulose membranes were stained with Ponceau Red prior to the immunoblot procedure to verify that the amounts of GST-Fyn SH3 fusion protein was the same in all samples. The final peptide concentration in the final 200-1.11 volume of the assays is indicated in the figures. These conditions were determined by time course studies that revealed that preincubating the GST fusion protein beads with the peptides for longer periods of time or at room temperature did not increase the amount of inhibition of baculovirus p85a or HPB-ALLderived PI 3-kinase binding to GST-Fyn SH3 fusion proteins.

RESULTS
Association of PI 3-Kinase with SH3 Domains-GST fusion proteins containing the SH3 domains of p85 (regulatory subunit of PI 3-kinase), Abl, Lck, and Fyn were bound to glutathione-Sepharose beads and incubated with HPB-ALL cell lysates.
Lipid kinase activity was assayed on the precipitates, and the reaction products were analyzed by TLC and HPLC (see "Experimental Procedures"). Fig. L4 depicts a TLC analysis of one of these experiments which shows that GST-SH3 fusion proteins precipitated considerable amounts of PI kinase activity (lanes 2 6 ) , whereas GST alone (lane 1 ) precipitated negligible amounts of PI kinase activity. Although the same amounts of GST-SH3 fusion proteins were used (10 pg), GST-85SH3 was the most efficient in precipitating PI kinase activity followed in order of efficiency by GST-Ab1 SH3, GST-Fyn SH3, and GST-Lck SH3. Further analysis of the reaction product by HPLC of the deacylated lipid demonstrated one single peak that co-migrated with glycerol-phosphorylinositol 3-phosphate (gPI-3-P) ( Fig. 1, B and C), confirming the identity of the SH3 precipitable lipid kinase activity as PI 3-kinase.
The Regulatory Subunit of PI 3-Kinase (~85) Mediates the Direct Association of PI 3-Kinase with SH3 Domains-It was next of interest to determine whether the association of PI 3-kinase with the SH3 domains was direct or mediated by an adaptor molecule. We first examined the available amino acid sequences of p85 and pllO (the catalytic subunit of PI 3-kinase) for motifs homologous to the proline-rich sequences identified in 3BP1,3BP2, and SOS that have been implicated in the direct association of these molecules to SH3 domains (13,15,19).
Analysis of pllO did not reveal any sequences especially rich in proline residues; however, we identified two proline-rich sequences in p85a as depicted in Fig. 2. The first sequence is located just downstream of the SH3 domain (residues 82-96 in p85a), and the second sequence is just upstream of the Nterminal SH2 domain (residues 300-314 in p85a). These proline-rich sequences flank the region of p85 that is homologous No lipid is detected in the GST precipitate, whereas one single peak that co-migrates with gPI-3-P standard is detected on the GST-85SH3 precipitate, indicating that the 85SH3 domain associates with PI 3-kinase. All the other reactions produced similar products (data not shown). C, for PI-3-P standard we used the lipid produced on a lipid kinase reaction performed on anti-p85 immunoprecipitate. a-p85 (0 ) and r3H1gPI to the C-terminal region of BCR. The lbwerpart of Fig. 2 depicts the proline-rich sequences of 3BP1,3BP2, SOS, and p85a. Mutation analysis of the 3BP1 proline-rich sequence has pointed out the critical residues involved in the interaction with the Ab1 SH3 domain (19). Mutation of residues Ala', Pro', Pro7, and Prolo abolish binding to Ab1 SH3, whereas changing Prog decreases dramatically the binding affinity (residues in boldface). The critical proline residues (Pro', Pro7, and Prolo) are present in both p85a sequences, and they are indicated in boldface. The presence of these proline-rich sequences suggested that p85 may be sufficient to provide the direct association of PI 3-kinase with SH3 domains.
To study SH3/p85 association we opted for a baculovirus expression system where human p85a is overexpressed in a background of insect proteins (see "Experimental Procedures"). In this system the possibility that another protein may be mediating the binding of p85 to the SH3 domains is virtually eliminated allowing us to determine whether p85 binds directly to the SH3 domains. GST-SH3 fusion proteins on beads were incubated with lysates derived from SF9 cells that express p85a (SF9/p85a). The beads were precipitated and the associated proteins analyzed by Western blot for the presence of p85a (Fig. 3) as described under "Experimental Procedures." Lane 1 depicts p85a immunoprecipitated by an anti-p85 antibody (positive control). GST alone failed to precipitate p85a from cell lysates (lane 2), whereas all GST-SH3 fusion proteins precipitated p85a (lanes 3 4 , indicating that a direct association between p85 and SH3 domains do occur. The ability of the differ-  -85SH3 (lane 3 ) . GSTwere incubated with SF9/p85a lysates for 15 min at 4 "C with constant rocking. The beads were washed twice with buffer containing 1% NoNdet P-40 and 2 x loading buffer was added. The samples were boiled for 5 min, separated on a 10% SDS-polyacrylamide gel electrophoresis, and transferred to nitrocellulose. An anti-p85 immunoblot is shown. The position of p85a is indicated. p85a was detected in the precipitates of GST-85SH3, GST-Ab1 SH3, GST-Lck SH3, and GST- Fyn SH3 (lanes  3-6). It was also detected in the anti-p85 immunoprecipitate (lane 1, positive control). No p85a was detected in the GST lane (lane 2 ) , indicating that p85a interacts directly with the SH3 moiety of the GST fusion proteins. ent SH3 domains to bind p85a paralleled their efficiency in precipitating PI 3-kinase activity from HPB-ALL cell lysates (compare Fig. 1 with Fig. 3).
p85 Contains l h o SH3-binding Motifs-& mentioned above, examination of the p85 amino acid sequence revealed two regions homologous to sequences identified in 3BP1 and SOS that have been implicated in direct binding to the SH3 domains of Ab1 and Grb2, respectively (13,15,19) (Fig. 2). To investigate whether these sequences can bind to SH3 domains two peptides with amino acid composition corresponding to residues 82-96 (SPPTPKPRPPRPLP) and 300314 (ERQPAPALPPKPPKP) of p85a (see Fig. 2) were synthesized (see "Experimental Procedures"). The control peptide sequence contained residues CELEPELDLDLDLE. These peptides were used in competition studies as follow. In these experiments, we focused on the association of p85a to the Fyn SH3 domain. Immobilized GST-Fyn SH3 fusion proteins were preincubated for 5 min at 4 "C with the indicated concentrations of peptides. The beads were then further incubated for 10 min a t 4 "C with cell lysates from HPB-ALL cells or SF9/p85a cells (see "Experimental Procedures"). The GST-Fyn SH3 precipitates from HPB-ALL cells were assayed for PI kinase activity, whereas the GST-Fyn SH3 precipitates from SF9/p85a cells were analyzed for the presence of p85a by immunoblot (see Fig. 4). Both peptides were equally capable of inhibiting binding of p85a (Fig. 4A) and PI 3-kinase activity (Fig. 4B) to GST-Fyn SH3. The 50% maximal inhibition for both peptides occurred a t a concentration between 200 and 400 p~ peptide. In addition, an 85-kDa protein was the most abundant protein observed in GST-Fyn SH3 precipitates from SF9/p85a cells by Coomassie Blue staining and was the only one that disappeared when GST-Fyn SH3 beads were preincubated with proline-rich peptides, indicating that this interaction was not mediated by another protein endogenous to SF9 cells (data not shown). The control peptide had no effect on the binding of p85a to the Fyn SH3 domain; however, in the PI kinase assays it somewhat increased the total precipitable activity. Two other highly charged but unrelated peptides also caused a similar slight increase in precipitable PI 3-kinase activity when added at millimolar concentration (not shown), suggesting that this is a nonspecific effect.' We also The same effect was observed on immunoprecipitable PI 3-kinase activity when protein A-Sepharose beads were incubated with multiple control peptides, including polylysine, prior to immunoprecipitation. immobilized GST-Fyn SH3 fusion proteins were incubated with 10, 100, or 1000 p~ of peptides P82-96 (lunes 44, P300-314 (lunes [7][8][9], and P control (lunes [10][11][12] or no peptide (lune 3 ) for 5 min a t 4 "C with constant rotation. 100 pl of SF9/p85a cell lysates were added per sample, and the beads were incubated for another 10 min a t 4 "C with constant rocking. An anti-p85 immunoprecipitation (lune 1, positive control) and a GST precipitation (lune 2, negative control) were carried out in parallel. The beads were washed and processed as described (see "Experimental Procedures"). An anti-p85 immunoblot is shown. The position of p85 is indicated. The immunoblot was quantitated with a Bio-Rad imaging system. Normalizing to the amount of p85 detected in lane 3 (GST-Fyn SH3, no peptide) as loo%, lune 4 (10 p~ P82-96) is also loo%, lune 5 (100 p~ P82-96) is 77%, lune 6 (1000 p~ P82-96) is 20%, lune 7 (10 p~ P300-314) is 91%, lune 8 (100 p~ P300-314) is 65%, and lune 9 (1000 p~ P300-314) is 17%. The amounts of p85a detected in lunes [10][11][12] corresponding to the control peptide remained constant a t about 95%, even at the highest concentration of peptide used. No p85a was detected in the GST lane (lune 2). B, immobilized GST-Fyn SH3 fusion proteins were incubated for 5 min a t 4 "C with constant rocking with different concentrations of peptides P82-96, P300-314, and P control or no peptide, as indicated in the figure. 100-pl HPB-ALL cell lysates were added to individual samples, and they were further incubated for 10 min a t 4 "C with rocking. A GST precipitation was carried out in parallel. The beads were washed and subjected to lipid kinase assay. Lipids were extracted and analyzed as described. Following separation by TLC, individual PIP spots were quantitated by a Bio-Rad imaging system. The data are expressed as percent of control, i.e. 100% corresponds to the amount of PI-3-P detected in the GST-Fyn SH3 precipitates in the absence of peptide. The data presented investigated if the proline-rich peptides had any effect on the PI DISCUSSION 3-kinase activity of purified enzyme. The peptides (P82-96, p300-314, and control) were incubated for min at room In this report we investigate the mechanism of association of temperature with 3-kinase prior to lipid kinase PI 3-kinase with SH3 domains. We have shown previously that assay. me peptides at concentrations ranging from to l ooo PI 3-kinase associates with the SH3 domains of the Src-like had no apparent effect on the activity of purified PI 3-kinase kinases ~5 6 ' '~ (27) and P59h.n (T) (20). In this study, we Provide (data not shown).
evidence that the 85-kDa subunit of PI 3-kinase mediates the association with the SH3 domains of Lck, Fyn, Abl, and p85 3-kinase. These data suggest that the peptides do not specifically affect SH3-containing proteins. These results indicate that PI 3-kimatic activity of PI 3-kinas.e during precipitation, probably by blocking elements: one that is dependent on tyrosine phosphorylation However, these peptides had no effect in the activity of soluble PI itself, suggesting that PI 3-kinase may be a target Of the soluble enzymatic activity but that they rather preserve the enzy-naSe may use two modes for its interaction with upstream hydrophobic sites on the beads that cause inactivation. and is mediated by its SH2 domains as in the case of its association to activated receptor-tyrosine kinases and one that is phosphorylation-independent, in which p85 associates with the SH3 domains of upstream molecules. Both interactions may occur in parallel in the case of activated cytosolic-tyrosine kinases that contain SH2 and SH3 domains. Examination of the p85 amino acid sequence revealed the presence of two proline-rich motifs highly homologous to the ones identified in 3BP1,3BP2, and SOS. These sequences have been implicated in mediating the interaction of 3BP1 and 3BP2 to the Ab1 SH3 domain and SOS to the SH3 domains of Grb2/ SEM-5/Drk (13)(14)(15)(16)(17)19). In this work, we show that the p85 proline-rich sequences are indeed SH3-binding motifs. Both sequences (P82-96 and P300-314, synthetic peptides) were able to inhibit binding of PI 3-kinase activity and p85a to a GST-Fyn SH3 fusion protein, whereas an unrelated sequence had no effect. The peptide concentrations required for 50% inhibition of SH3 binding to p85 (-200 p~) were quite high compared with concentration of phosphopeptides that inhibit SH2 domain interactions (-10 nM (4,7,32,33)). However, they were similar to the concentrations of dynamin peptides that bind p85 SH3 domain (34). The ability of the GST-SH3 domains to retain association with baculovirus-expressed p85 despite relative harsh washing conditions suggests that the affinity of p85 for SH3 domains is much higher than is the affinity of the individual proline-rich peptides. This could be explained by the ability of multiple SH3 domains on the same bead to simultaneously bind a single p85 protein via interaction with both proline-rich domains. The actual affinity of the proline-rich peptides for immobilized SH3 domains are probably underestimated in the competition experiment (Fig. 4) because of the entropic advantage of binding the bidentate sites in p85 over the individual peptides. Preliminary experiments using fulllength p85 to block binding of lymphocyte-derived PI 3-kinase to SH3 domains support this model (data not shown). The presence of two proline-rich sequences in p85 suggests a model for recruitment of PI 3-kinase to cross-linked CD4/~56''~ and TCFU CD3 complexes. It is possible that two SH3 domains from adjacent ~5 6 ' '~ molecules simultaneously grab a single p85, thereby increasing the effective affinity.
The finding that the SH3 domains of Ab1 and Grb2ISEM-5/ Drk interact with 3BP1 and SOS, respectively, has revealed the mechanism by which tyrosine kinases may communicate with the ras signaling pathway (13)(14)(15)(16)(17). 3BP1 has sequence homology to rho-GAP, n-chimerin, and the C-terminal portion of BCR (13), all of which have been shown to contain GTPase activities (18), whereas SOS is a nucleotide exchange factor (GDP to GTP) for ras (14,15,17,35). It is interesting to note that the two proline-rich sequences in p85 flank a region that has sequence homology to the C-terminal portion of BCR, 3BP1, nchimerin, and rho-GAP (18,22,24,25). Although no GTPase activity has been recovered from recombinant p85 or purified PI 3-kinase to date3 on the basis of homology, it is possible that p85 may be a GAP for an as yet unidentified low molecular weight GTP-binding protein. If that is the case the association of the p85 SH3 domain with its own proline-rich sequences may regulate this activity, i.e. when the SH3 domain is bound to one of these sequences the intrinsic GTPase activity is inhibited and when the p85SH3 domain is in the "open" conformation p85 may have GTPase activity. Further studies are necessary to address this possibility. The alternative possibility that selfassociation of the SH3 domain of p85 regulates PI 3-kinase activity was tested by investigating the effect of the proline-rich peptides on the activity of purified PI 3-kinase. No effect was L. Feig, R. Cerione, and L. C. Cantley, unpublished results.
observed. This experiment does not rule out the possibility that under some conditions in vivo regulation of PI 3-kinase activity could occur through this domain.
The different proline-rich sequences may confer SH3 domain target specificity as in the case of phosphotyrosine-containing motifs and SH2 domains (Ref. 7 and references therein). Ciccheti et al. (13) have shown that 3BP1 binding to SH3 domains is highly selective. It binds best to the Ab1 SH3 domain followed by Src SH3 and has extremely low affinity for the SH3 domains of n-Src and Crk. Booker et al. (34) have shown that the SH3 domain of p85 binds to dynamin, a GTP-hydrolyzing microtubule-associated enzyme. A 21-residue dynamin peptide was found to bind strongly to the SH3 domains of p85a and phospholipase Cy, weakly to Src SH3 and p67C-terminal SH3, and not at all to the N-terminal SH3 domains of p67 and p47, n-Src SH3, GAP SH3, spc SH3. These results indicate that these proline-rich sequences may confer target specificity. In this study we observed that the 85-kDa subunit of PI 3-kinase binds best to its own SH3 domain, followed by Ab1 SH3, Fyn SH3, and Lck SH3. These results suggest that the p85SH3 domain selfassociates to regulate interactions with other proteins.