Characterizing the Interactions between the Two Subunits of the p101/p110γ Phosphoinositide 3-Kinase and Their Role in the Activation of This Enzyme by Gβγ Subunits*

Recently, we have reported the purification and cloning of a novel G protein βγ subunit-activated phosphoinositide 3-kinase from pig neutrophils. The enzyme comprises a p110γ catalytic subunit and a p101 regulatory subunit. Now we have cloned the human ortholog of p101 and generated panels of p101 and p110γ truncations and deletions and used these in in vitro and in vivo assays to determine the protein domains responsible for subunit interaction and activation by βγ subunits. Our results suggest large areas of p101 including both N- and C-terminal portions interact with the N-terminal half of p110γ. While modifications of the N terminus of p110γ could modulate its intrinsic catalytic activity, binding to the N-terminal region of p101 was found to be indispensable for activation of heterodimers with Gβγ.

Recently, we have reported the purification and cloning of a novel G protein ␤␥ subunit-activated phosphoinositide 3-kinase from pig neutrophils. The enzyme comprises a p110␥ catalytic subunit and a p101 regulatory subunit. Now we have cloned the human ortholog of p101 and generated panels of p101 and p110␥ truncations and deletions and used these in in vitro and in vivo assays to determine the protein domains responsible for subunit interaction and activation by ␤␥ subunits. Our results suggest large areas of p101 including both N-and C-terminal portions interact with the N-terminal half of p110␥. While modifications of the N terminus of p110␥ could modulate its intrinsic catalytic activity, binding to the N-terminal region of p101 was found to be indispensable for activation of heterodimers with G ␤␥ .
Phosphoinositide 3-kinases (PI 3-kinases) 1 are responsible for the phosphorylation of inositol phospholipids in the D-3 position of the inositol ring. Their lipid products (PtdIns3P, PtdIns(3,4)P 2 , and PtdIns(3,4,5)P 3 ) function as second messengers in eukaryotic cells. Indeed, PI 3-kinases appear to be involved in the control of a host of cellular responses ranging from intracellular transport to cell motility and the suppression of apoptosis (see Refs. 1-3, for reviews).
Three classes of PI 3-kinases are distinguished (4). Type I PI 3-kinases can be rapidly activated by cell-surface receptors and in vivo make predominantly PtdIns(3,4,5)P 3 (5). They are heterodimeric enzymes comprising a 110-kDa catalytic subunit and a regulatory subunit. Type IA PI 3-kinases contain an ␣, ␤, or ␦ p110 catalytic subunit (6, 7) and a p50, p55, or p85 (␣ or ␤) regulatory subunit (8 -10). The regulatory subunits contain two SH2 domains which allow the enzyme to bind to, and be activated by key phosphotyrosine residues found in the cytoplasmic tails of growth factor receptors and various adapter proteins (11).
Type IB PI 3-kinase is made up of a p110␥ catalytic subunit and a p101 regulatory subunit (12). This PI 3-kinase seems to be specifically stimulated by receptors capable of activating heterotrimeric G proteins (12,13). It appears, that this effect is mediated by G protein ␤␥ subunits which can directly activate p101/p110␥ PI 3-kinase (12). Although several reports show that both the biological effects of p110␥ and its intrinsic sensitivity to G ␤␥ are substantially amplified by the presence of p101, some data suggest that p110␥ alone can be substantially catalytically activated by G ␤␥ and have clear biological effects (14,15). We have begun to address the issue of the role of p101, if any, in these events by analyzing the regions of p101 involved in interactions with p110␥ and furthermore, for those constructs that bind, how they affect the activation of the complex by G ␤␥ . We analyzed also the part played by p110␥ in the process of G ␤␥ activation and p101 binding.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis of p101 and p110␥-All point mutations, novel N or C termini, and internal deletions within porcine p101 cDNA were done using polymerase chain reaction-based strategies with mutagenic primers and Taq polymerase (Promega). Polymerase chain reaction-generated fragments were ligated back into pCMV3 with an N-terminal (EE)-or Myc-tag for expression in mammalian cells and into pAcOG3 with an N-terminal (EE)-tag for constructing baculovirus transfer vectors. All polymerase chain reaction-generated inserts were sequenced.
N-and C-terminal deletions of the human p110␥ were either done by digesting with the appropriate restriction enzymes at designated sites. The isolated DNA fragments were religated into pcDNA3 containing N-terminal Myc-tag linkers. The Ras-binding domain deletions were done by fusing the internal fragments in-frame to the N-terminal 169 amino acids, creating a three amino acid linker. Full-length human p101 was cloned into an N-terminal EE-tag containing pcDNA3 vector.
* 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.
The p101 purification was modified as follows, cells were sonicated into 0.15 M NaCl, 25 mM Hepes (pH 7.2, 4°C), 2 mM EGTA, 1 mM MgCl 2 plus antiproteases; cytoplasmic fractions were not pre-cleared with anti-Myc beads; washes after the anti-EE beads (Onyx Pharmaceuticals) were five times in 0.4 M NaCl, 20 mM Hepes (pH 7.4, 4°C), 1 mM EGTA, 1% Triton X-100, 0.4% cholate, and three further times in buffer H. To bind the subunits, p101 bound to packed anti-EE beads was mixed end on end for 2.5 h with a 25-fold molar excess of free p110␥ in a small volume of buffer H. Heterodimer on anti-EE beads was washed in buffer J (1% Triton X-100, 0.15 M NaCl, 20 mM Hepes (pH 7.4, 4°C), 1 mM EGTA) and in buffer FЈ. p101-p110␥ heterodimer was eluted in buffer FЈ supplemented with 125 g/ml EY peptide (EYMPTD). Typical yields were 2 mg of p110␥ and 25 g of p101/500 ml of Sf9 culture.
GST-p110␥ (the relevant recombinant baculovirus was a gift from R. Wetzker and encoded the human form of the protein) was purified from cytosolic fractions or Triton X-100 lysates of Sf9 cells as described by Cells from each electroporation were seeded into 175-ml tissue culture flasks in full growth medium. After 28 h, cells were harvested by trypsinization, washed once in phosphate-buffered saline, and cell pellets lysed in 1 ϫ phosphate-buffered saline, 1 mM EGTA, 1% Triton X-100. Cytoplasmic fractions were immunoprecipitated with anti-EE beads followed by washes in 2 ϫ phosphate-buffered saline, 1 mM EGTA, 1% Triton X-100. Samples for PI 3-kinase assays were washed further in 25 mM Hepes (pH 7.4, 4°C), 1 mM EGTA. Remaining samples were resolved by SDS-PAGE. Gels were either Coomassie stained or wet-blotted onto polyvinylidene fluoride membranes (Millipore) and immunoblotted according to their tags with anti-Myc antibody (obtained from Onyx Pharmaceuticals) and anti-EE ascites fluid (Babco) or with an anti-p101 antiserum (Microchemical Facilities, Babraham Institute).
PI 3-Kinase Assays-Free, purified protein from Sf9 cells or COS-7 derived protein on anti-EE beads was diluted in sample dilution buffer (2 mg/ml fatty acid free bovine serum albumin, 0.1 M KCl, 20 mM Hepes (pH 7.4, 4°C), 1 mM dithiothreitol). Lipid mixtures containing phosphatidylethanolamine and PtdIns(4,5)P 2 were dried down and sonicated into 0.1 M NaCl, 25 mM Hepes (pH 7.4, 4°C), 1 mM EGTA, 0.1% cholate for final concentrations of 50 M phosphatidylethanolamine (Sigma) and 5 M PtdIns(4,5)P 2 (13). They were supplemented with G protein ␤␥ subunits (prepared as in Ref. 16)  For the assay, 5 l of diluted protein were kept on ice until 20 l of lipid mixture with or without G ␤␥ was added and the mixture transferred to a 30°C waterbath. 4 min later, 50 l more sample dilution buffer supplemented with MgCl 2 to give a final concentration of 3.5 mM was added, another 4 min later, 10 l of diluted ATP was added. Assays were stopped after 12 min incorporation time by addition of 160 l of 1.25 M HCl. Assays were extracted with 800 l of CHCl 3 :MeOH (2:1) and then with CHCl 3 , MeOH, 1 M HCl (3:48:47). Lipids were deacylated and resolved on PEI TLC plates as described before (13). Alternatively, assays were conducted precisely as described by Leopoldt et al. (15) with PtdIns acting as the substrate.

RESULTS
First, we simply reproduced the results of previous work showing purified p110␥ could bind p101 both in vitro and in vivo and that this association substantially increased the scale of activation of the PI 3-kinase with G ␤␥ from 1-2-fold to 50 -150-fold, but now using modified procedures and porcine and human versions of the proteins. Human p101 (GenBank accession number AF128881) was 88% identical to porcine p101 (c.f. human p110␥ is 94% idential to porcine p110␥). We found the species orthologs behaved interchangeably in these assays and gave results identical to those of the earlier work (Ref. 12, data not shown).
Analysis of p101 Structure/Function in Vitro-In order to understand the interactions between p101 and p110␥ and the ability of the heterodimer to respond to G ␤␥ subunits, we made panels of p101 and p110␥ constructs. Analysis of p101 structures involved in binding to p110␥ was first approached with the following assay. NH 2 -terminal (EE)-tagged p101 constructs were purified from Sf9 cells using anti-(EE)-beads. After washing, an excess of purified, NH 2 -terminal (6H)-tagged full-length p110␥ from Sf9 cells was mixed with the immobilized p101 constructs. The beads were washed again and eluted with an epitope (EY) peptide. The released proteins were analyzed by SDS-PAGE and PI 3-kinase assays with and without G ␤␥ subunits. This analysis revealed that all of the p101 constructs tested (Fig. 1A) had reduced, but still significant capacity to specifically bind p110␥ relative to full-length. Thus both the large N-and C-terminal truncations of p101 (⌬1-163 and 1-733) resulted in a 50% reduction in the recovery of p110␥, while ⌬283-581 reduced recovery by about 25% (Fig. 1B). Some p110␥ specifically bound to N-terminal (p101 1-163) and Cterminal fragments of p101 (p101⌬1-574) (about 10 and 30% of wild-type, respectively) (Fig. 1C).
All p101-p110␥ heterodimers were also assayed for PI 3-kinase activity in the presence or absence of G ␤␥ (Fig. 1D). All of the heterodimers created containing p101 truncations and deletions had reduced activation by G protein ␤␥ subunits. It is striking, however, that the N-terminal truncation (⌬1-163) completely abrogated activation by G ␤␥ , while the remaining truncations resulted merely in decreased ability of the respective heterodimers to be activated by G ␤␥ . Neither N-nor Cterminal p101 peptides bound to p110␥ yielded considerable activation when assayed in the presence of G ␤␥ subunits. Overall this suggests a number of regions in p101 are involved in its ability to interact with p110␥, particularly the N and C termini. Similarly, multiple areas in p101 are required for heterodimers to give maximal activation by G ␤␥ subunits, however, the N terminus is absolutely required for this process.
Analysis of p101 Structure/Function in Vivo-To address issues such as (a) the possibility that purification and our handling of the p101 constructs had resulted in varying levels of denaturation and that this effect generated the differential binding we observed, and (b) that some binding only occurs in vitro in the absence of competing proteins found in the cell, we examined the ability of p101 derivatives to bind to p110␥ when the proteins are transiently co-expressed in COS-7 cells. In addition, to allow for more detailed mapping of the impact of different areas of p101 on both binding to p110␥ and activation of the heterodimer with G ␤␥ , a number of further constructs were introduced ( Fig. 2A).
All p101 constructs were (EE)-tagged and co-expressed with (Myc)-tagged p110␥ in COS-7 cells in transient transfection assays. Cytosolic fractions of harvested cells were subjected to anti-(EE) immunoprecipitations, half of which were used to estimate binding stoichiometries on gels or blots (see Fig. 2B, for an example) while the other half was assayed for PI 3-kinase activity in the presence or absence of G protein ␤␥ subunits.
The data describing binding of p101 derivatives to p110␥ in this assay is summarized in Fig. 2C. Binding to p110␥ was affected in all p101 constructs except for the point mutations and DE328 -341STP constructs, supporting the conclusion from the in vitro studies indicating that more than one area of p101 contributes to binding p110␥. It appears, however, that in the COS-7 cell assays compared with the in vitro assays, the N-terminal deletions lead to a larger decrease in p110␥ binding ability than C-terminal deletions (25 versus 70% binding) and N-terminal peptides can rescue more binding than C-terminal peptides (50 versus 12% binding).
We also assayed all constructs immunoprecipitated from COS-7 cells for PI 3-kinase activity. The resulting data (Fig.  2D) underlines the data already obtained from the in vitro assays from Sf9-produced protein. Again, we found that both C and N termini of p101 contribute to full activation of the complex by G protein ␤␥ subunits. Any deletion within p101 interfered with the ability of the heterodimer to be completely activated. Interestingly, deletion of the C-terminal 150 amino acids (1-733 and p84) and deletion of much larger C-terminal portions (1-314, 1-265, and 1-163) lead to a similar decrease in activation by G ␤␥ to less than 25% of that of full-length p101, stressing the role for the very C-terminal part. The most dramatic result, however, is induced by deletion of the very Nterminal portion of p101 (constructs ⌬1-153 and ⌬1-263) which lead to complete loss of activation, confirming the con-clusions from the in vitro assays that this region of p101 is essential for activation of the complex by G protein ␤␥ subunits.
Analysis of p110␥ Structure/Function-To define the sites in p110␥ involved in both interaction with p101 and activation by G ␤␥ , we prepared a set of constructs shown in Fig. 3 for expression in COS-7 and Sf9 cells. These constructs were used in assays aimed at defining the regions that interacted with p101 that were relevant to the G ␤␥ activation of p101-p110␥ heterodimers and were required for basal PI 3-kinase activity.
PI 3-Kinase Activity of p110␥ Constructs-Increasing N-terminal truncation of p110␥ systematically reduced its basal catalytic activity (Fig. 4A). Deletions beyond residue 369 had no detectable PI 3-kinase activity. N-terminal tags seemed to increase the basal catalytic activity of p110␥. The activity of the N-terminal (6H)-tagged p110␥ (6H-p110␥) was reduced 2-fold by removal of the tag with thrombin and the activity of defined p110␥ constructs was reduced by half by switching from an FIG. 1. In vitro study of p101 association with p110␥ and activation of complexes by G ␤␥ . A, panel of p101 constructs used for expression and purification from Sf9 cells. All p101 proteins were N-terminal (EE)-tagged (E). B and C, in vitro association of the (EE)-p101 proteins to independently purified, N-terminal (6H)-tagged p110␥ as detailed under "Experimental Procedures." Fractions of eluted heterodimer were visualized by SDS-PAGE and Coomassie staining. p101b was prepared from an equivalent extract of Sf9 cells infected with wild-type baculovirus and thus represents a control for nonspecific binding of p120 to the anti-(EE) beads. All data shown is representative of at least four independent experiments. D, PI 3-kinase assays of fractions of the heterodimers shown in B and C using PtdIns(4,5)P 2 as substrate, with and without addition of G ␤␥ as indicated. The data are mean Ϯ range (n ϭ 2-5) drawn from independent experiments. The activity in the absence of G ␤␥ was normalized to the amount of p110␥ protein in each assay (estimated from the equivalent SDS-PAGE gels).
Reports that GST-p110␥ can be substantially activated by G ␤␥ in vitro (15) lead us to test the effect of manipulating p110␥ structure on its responsiveness to G ␤␥ , despite the fact that we have never detected significant effects of G ␤␥ on porcine p110␥ activity in the absence of p101. Testing a range of catalytically active porcine and human p110␥ constructs with or without N-terminal or C-terminal (6H) tags in the presence and absence of G ␤␥ showed up to 2-fold activations of human (but not porcine) proteins with either our standard assay procedure or that of Leopoldt et al. (15; see below). However, with human GST-p110␥ we found that although our assay procedure (using PtdIns(4,5)P 2 and 3.5 mM MgCl 2 ) failed to reveal any activation by G ␤␥ , using PtdIns and 10 mM MgCl 2 in the assay (15), we could reproducibly detect 6 -7-fold activations by G ␤␥ (Fig. 4B); this was unaffected if the constructs were purified from membrane or cytosolic fractions (data not shown).
Interaction of p110␥ Constructs with p101 and Activation of Heterodimers by G ␤␥ -Experiments performed with proteins expressed in either COS-7 or Sf9 cells indicated that the N terminus of p110␥ was critical for binding to p101. p110␥ deletions through the series ⌬1-122, ⌬1-133, and ⌬1-144 (Fig.  5A, for examples) showed decreasing ability to bind p101, a ⌬1-169 construct showed no detectable binding (Fig. 5B). However, other regions were clearly involved in binding p101 (see Fig. 5C for an overview). Hence, N-terminal peptides (e.g. 1-169) had low p101 binding potential and further deletion of either the Ras-binding domain (178 -330) or the central regions (330 -775) of p110␥ also significantly reduced p101 binding. It is significant that those p110␥ constructs capable of binding p101, once bound, were all very similarly activated by G ␤␥ (Fig.  5D). Catalytically active p110␥ constructs incapable of binding p101 remained insensitive to G ␤␥ when assayed in the presence of p101 (Fig. 5C).

DISCUSSION
It is clear from the above results that multiple regions of p101 are involved in either direct (i.e. physical contact) or indirect interactions with p110␥. Interestingly, the G ␤␥ sensitivity of the p110␥ heterodimers formed by using various p101   FIG. 2. In vivo association of p101 constructs with p110␥ and activation of resulting heterodimers by G ␤␥ . A, further p101 constructs for use in COS-7 transient transfection assays included further N-and C-terminal peptides and larger deletions. Furthermore, an unusual acidic region was replaced by a stretch of serines, threonines, and prolines (DE328 -341STP); a region bearing a vague similarity to PH domains of known signaling proteins as well as potential key residues therein were mutated (⌬161-263, Y193C, W252A); a possible spliced variant of p101 is also included (p84, this sequence diverges from p101 at residue 733 resulting in a truncated version of the protein; P. Hawkins and A. Eguinoa, unpublished data). Again, all p101 constructs included N-terminal (EE)-tags (E). B, COS-7 cells were transiently transfected with mammalian expression vectors encoding a (EE)-p101 construct and a (Myc)-p110␥ (for controls, the total amount of DNA was made up to with irrelevant DNA). Cytoplasmic fractions of transfected cells were subjected to immunoprecipitations with (EE)-beads. The amount of co-immunoprecipitated (Myc)-p110␥ was visualized on Coomassie-stained protein gels. Binding ratios were estimated by eye with a minimum of three independent transfections being taken into account for each p101 construct. C, graphical illustration of all binding data obtained from the COS-7 transient transfection assays. D, aliquots of the COS-7-derived p101-p110␥ heterodimers (on the beads) were assayed for PI 3-kinase activity in the presence or absence of G ␤␥ . Fold activation obtained with G ␤␥ for the different constructs is detailed comparatively in this graph. The legend on the y axis indicates with which construct p110␥ had been expressed. constructs clearly correlated with how tightly they formed heterodimers, but absolutely depended on the presence of the N terminus of p101. We have previously published some binding data showing G ␤␥ can apparently bind about five times more effectively to p101 than p110␥ (12), but we cannot discount important binding of G ␤␥ to p110␥, as others have identified clear effects of G ␤␥ on GST-p110␥ (13, 15; see also Fig. 4B). There are, therefore, two types of explanation for this data. First, that a lot of the contacts between p101 and p110␥ are required for G ␤␥ to have their full effects on p110␥ catalytic activity (whether the G ␤␥ bind to p101, p110␥, or both). Second, that the primary sites of G ␤␥ binding that influence the activity of the heterodimer are located on p101 and are "co-incidentally" disrupted by changes which impinge on p101-p110␥ interaction. This apparently highly interwoven structure-function relationship between p101, G ␤␥ , and p110␥ is not surprising if it is accepted that binding of p101 and G ␤␥ can have such a profound effect on the catalytic site in p110␥. The binding site for p101 in p110␥, although primarily involving the N terminus of the protein, is a far larger region extending deeper into the protein. Considering the fact that several parts of p101 are involved in binding to p110␥, it would seem reasonable that a substantial piece of p110␥ is involved in binding p101 (note that in p110␣ a relatively short region of the N terminus is thought to bind to a short inter-SH2 domain segment of p85 (17,18)). In contrast to the situation with p101, where deletions affecting the total contact area and strength of binding to p110␥ also affected the sensitivity of the heterodimer to G ␤␥ , deletions through the N terminus of p110␥, that reduced the efficiency of binding to p101, had no effect on the activation of heterodimeric enzyme that could be recovered. Clearly, this analysis cannot be complete because deletions removing more than 200 N-terminal residues begin to strongly reduce the intrinsic catalytic activity of p110␥, such that deletions beyond residue 369 activity is completely lost (see below), yet these deeper regions are clearly involved in binding to p101. How-ever, the implication of this data is that some of the contacts between p101 and p110␥, involved in their binding, can be eliminated by disruptions on the p110␥ side without effect on G ␤␥ activation of the heterodimer, but that contacts broken by p101 disruptions reduce the ability of G ␤␥ to activate the enzyme. The simplest explanation for these observations is that p101 disruptions are also interfering with G ␤␥ -binding sites that are most important for activation of the heterodimer; i.e. that the latter of our proposed models (see above) is most likely to be correct.
In the process of engineering and assaying a range of p110␥ constructs it became clear that the N terminus of the protein had significant potential to influence catalytic activity. Hence the addition of N-terminal GST and (6H)-tags apparently increased the activity of the enzyme. In contrast, binding of p101, All proteins were derived from Sf9 cells, and purified according to their tags. The basal catalytic activity was compared with that of N-terminal (6H)-tagged p110␥ for each individual construct. Note that thrombin cleavage of the (6H)-tag results in loss of 26 N-terminal amino acids due to an internal thrombin cleavage site (R. Williams, personal communication). GST-p110␥ could not be successfully cleaved by thrombin. B, porcine p101, porcine (6H)-p110␥, porcine p101/(6H)-p110␥, and human GST-p110␥ were purified from Sf9 cells as described under "Experimental Procedures" and aliquots were assayed for PI 3-kinase activity in the presence or absence of G ␤␥ according to the protocol of Leopoldt et al. (15). Samples were normalized for their basal activity (this is not equivalent to amount of p110␥ in the assay, see A for this comparison). which appears to interact most strongly with the N terminus of p110␥ lead to a 5-fold inhibition of basal catalytic activity. This effect of p101 appears to have some parallels with observations that p85 binding to p110␣ suppresses its PI 3-kinase catalytic activity (19). Clearly the scale of activation of the p101/p110␥ heterodimer by G ␤␥ in vitro indicates this "simple model" cannot apply here, however, this phenomenon lead us to test whether N-terminal tagging also influenced the sensitivity of p110␥ alone to G ␤␥ subunits.
We found that of a large range of constructs GST-p110␥ alone was clearly activated by G ␤␥ , but only under the conditions previously reported to show this effect (15). The very fact that G ␤␥ can have some significant effect on GST-p110␥ in the absence of p101 may be taken to suggest one primary point of interactions of G ␤␥ with p101/p110␥ is through direct binding to the p110␥ subunit. This gains some support from work indicating G ␤␥ can bind to p110␥ alone. We have previously found that although five times more G ␤␥ could be rescued from in vitro binding assays associated with p101 than with p110␥ we could detect above background binding to p110␥ (12). Fur-thermore, Leopoldt et al. (15) have shown G ␤␥ can be recovered with p110␥ immunoprecipitated from Sf9 cells, and they went on to map this binding to two distinct regions within p110␥. However, there are major problems with these approaches (15). First, neither directly demonstrated that the G ␤␥ -binding sites were relevant to the activation of the PI 3-kinases and second as both analyses depended on an "immunoprecipitation wash" protocol they are subject to selective recovery of interactions with appropriate affinities and on/off rates; meaning much more physiologically important binding sites could be missed. As G ␤␥ are notoriously prone to nonspecific hydrophobic interaction and for binding to effectors with low affinity but high on/off rates this means these assays are of very limited value in defining the regions that bind G ␤␥ that lead to activation of the enzyme.
As a consequence of these considerations and in the light of (i) lack of effect of G ␤␥ on other, non-GST-tagged p110␥ constructs, (ii) the unphysiological nature of the substrate and ionic conditions required to see G ␤␥ activation of GST-p110␥, (iii) the inevitable problem that in using p110␥ alone rather than in a com- FIG. 5. Interaction of p110␥ constructs with p101. A, in vitro interaction assays to assay binding ability of free (6H)-p110␥ constructs to immobilized (EE)-p101 as described under "Experimental Procedures" are shown here with full-length p110␥ and two N-terminal truncations in comparison. Left-hand lanes show p110␥ derivatives after two different purification steps and right-hand lanes show the p110␥ derivatives obtained via binding to the (EE)-p101 beads. Binding of ⌬1-144 could not be detected on the gel, since the proteins co-migrated, however, activity assays showed that a small degree of binding did take place (see below). B, Myc-tagged human p110␥ and EE-tagged human p101 were transiently co-transfected into COS-7 cells as indicated and the protein complexes were co-immunoprecipitated and immunoblotted. I, p101 expression: anti-p101 Western blot of an anti-EE immunoprecipitation; II, p101 binding to p110␥: anti-p101 Western blot of an anti-Myc immunoprecipitation; III, p110␥ expression: anti-Myc Western blot of the cell lysates. C, overview of the ability of all p110␥ constructs to bind p101 as estimated by eye from Coomassie-stained gels or Western blottings derived from in vitro interaction assays and COS-7 transient transfection assays. It is indicated also, which constructs are catalytically active and where complexes are activated by G ␤␥ . D, extent of fold-activation of catalytic activity by G ␤␥ for aliquots of the three p101-p110␥ heterodimers prepared in vitro from Sf9-derived proteins and shown in Fig. 2A. plex with p101 further hydrophobic regions of protein interaction may be exposed, and (iv) that the presence of the GST fusion will drive homodimerization of p110␥, we consider the evidence that the effects of G ␤␥ on p101/p110␥ PI 3-kinase are only via direct interaction with p110␥ to be very weak.
Overall, our data suggests that p101 and p110␥ interact primarily through large areas covering the N and C termini of p101 and the N-terminal half of p110␥ and that the areas which bind G ␤␥ giving the major effect on PI 3-kinase activity are probably located on p101. Given the substantial difficulties encountered in studying specific, low affinity interactions of G ␤␥ subunits with various effectors in vitro (e.g. studying G ␤␥ activation of PLC␤s can be seen as an analogous problem), it is likely that a combination of further technologies, including detailed structural information, will be required to yield further insight into the mechanism of action of G ␤␥ on p101/p110␥ PI 3-kinase.