Abstract

PIP3 works by recruiting downstream effectors containing lipid-binding domains

PIP3 was first identified by Cantley and colleagues in PDGF-stimulated fibroblasts [3]. The lipid was distinguished from other phosphoinositides by the mobility of its deacylated head group (gro-PIP3) on an anion-exchange HPLC column. The lipid can also be identified by its slow mobility, relative to phosphatidylinositol and mono- or bis-phosphoinositides, during thin layer chromatography [4]. More recent methods for the analysis of PIP3 have included displacement assays or blot-based assays with recombinant PIP3 binding proteins [5,6], and mass spectrometric methods [7].

Unlike PI[4,5]P2, which can be converted from a signaling lipid to the soluble second messengers IP3 and diacylglycerol by phospholipase C-mediated, 3-phosphoinositides are not substrates for phospholipase C [8]. Instead, they signal by binding to and recruiting effector proteins containing specific lipid binding domains. The first identified PIP3 effector was the protein kinase Akt/PKB, which contains a Pleckstrin Homology Domain (PH domain) that binds to PIP3 and PI[3,4]P2 and is required for Akt activation by PI 3- Additional PIP3-binding proteins include GRP-1, ARNO and centaurin-1, which are exchange factors for the ARF GTPases, cytohesin which regulates integrin signaling, and the Tec-family tyrosine kinase BTK[9]. The presence of a tandem Dbl-homology domain-PH domain was also identified as a common feature of guanine nucleotide exchange factors (GEFs) for Rho-family GTPases, including PIP3-regulated GEFs like Tiam1 and Vav [10,11]. In all of these cases, binding to PIP3 required an intact PH domain. It is important to note that only a subset of PH domains bind PIP3; others bind to different phosphoinositides, or do not bind lipids at all [9].

These studies led to a model in which production of PIP3 at the cytosolic face of cellular membranes leads to the recruitment and, in some cases, the allosteric activation of the effector proteins. The mechanism is most fully developed for Akt, where PIP3 binding to the Akt PH domain drives 3 related events: targeting of Akt to the cell membrane, which is coincident with membrane targeting of the upstream Akt activator PDK-1, followed by a conformational rearrangement of AKT that facilitates PDK-1 mediated phosphorylation on T308 in the Akt activation loop [12]. However, it must be noted that while lipid binding contributes to the membrane recruitment of some PH domain-containing proteins, protein-protein interactions may serve to further refine the site of recruitment, as well as the stability and duration of targeting. For example, the PH domains from Akt, ARNO, Btk and GRP1 all bind to PIP3, yet their expression in cells caused distinct inhibitory effects on cellular signaling, and mutation of residues not involved in PIP3 binding abolished these effects [13]. These data suggest that PIP3-independent interactions affect the location, stability or duration of PIP3-mediated targeting. In some cases, PIP3 binding to PH domains can regulate enzyme activity independently of membrane recruitment. For example, in the Arf GAP ARAP1, the PH domain is required for PIP3-stimulated GAP activity, yet does not bind tightly enough to PIP3 to mediate membrane binding [14]. In this case, membrane recruitment by a PIP3-independent mechanism, such as protein-protein interaction, facilitates allosteric regulation of the GAP by PIP3 binding to the PH domain. Overall, it seems likely that the specificity of PH domain-mediated signaling involves a combination of lipid mediated and protein mediated targeting.

In the repertoire of downstream signaling events initiated by production of PIP3, activation of the Akt kinase plays an outsized role [15]. The 3 isoforms of this serine/threonine protein kinase have major roles in the regulation of metabolism, and protein and lipid synthesis and degradation, through the FOXO family of transcription factors, the Glycogen Synthase Kinase 3 protein kinase, the AS160 regulator of glucose transport trafficking, and the mTOR protein kinase, among other downstream effectors. Akt also regulates survival pathways, inflammatory signaling, the cell cycle, and cardiovascular homeostasis. Other major effectors of PIP3 include guanine nucleotide exchange factors involved with vesicular trafficking and cytoskeletal regulation [16], and the Tec tyrosine kinases that regulate signaling in lymphocytes and mast cells [17]. The downstream effectors of PIP3 have been described in detail in recent reviews [18,19], and will not be discussed further here.

Class I PI 3-kinases are regulated by protein-protein interactions

The Class I PI 3-kinases are subdivided into the Class IA enzymes, in which the regulatory subunits have two SH2 domains that interact with tyrosine-phosphorylated proteins, and the Class IB enzymes, which interact with a distinct pair of structurally uncharacterized regulatory subunits (Figure 1). The Class IA heterodimer, which consists of an 85, 55 or 50 kDa regulatory subunit bound to one of three 110 kDa catalytic subunits, is extremely stable, and is essentially irreversible once formed [1]. The Class IB PI 3-kinase heterodimer is comprised of the p110γ catalytic subunit and either the p101 or p87 regulatory subunits [2]. We know of four types of regulatory interactions that activate heterodimeric PI 3-kinases.

Open in a separate window

Figure 1

Domain Organization and Activators of Class I PI 3-kinases

The Class IA PI 3-kinases consist of a p85 regulatory subunit containing SH3, BCR, proline-rich, SH2 and coiled-coil (iSH2) domains, and a p110 catalytic subunit (in yellow) containing an Adapter Binding Domain (ABD), a Ras Binding Domain (RBD), C2, helical and kinase domains. Shorter isoforms of p85 (p55α, p50α, p55γ) contain a truncated N-terminal segment linked to the first SH2 domain. The Class IB PI 3-kinase consists of structurally uncharacterized p87/84 or p101 regulatory subunits bound to a p110γ catalytic subunit (in yellow), whose domain organization is similar to that of Class IA PI 3-kinases. The mechanism of p87/p101 binding to p110γ is not yet known, but involves both the N- and C-terminal regions of p110γ [120]. Of the upstream activators shown, Ras binding regulates p85/p110α and p85/p110δ but not p85/p110β. Gγ regulates p85/p110β and p101/p110γ but not p85/p110α or p85/p110δ.

1. Disruption of SH2 domain-mediated inhibition. For the Class IA PI 3-kinases, displacement of inhibitory contacts between the p85 SH2 domains and the p110 catalytic subunit activate the enzyme [20–22]. Inhibitory interactions involving both the n-terminal and c-terminal SH2 domains have been demonstrated [23], although there is controversy as to whether the cSH2 domain is inhibitory for p110α [24]. The consensus is that SH2 domain binding to activated receptor tyrosine kinases or their substrates disrupt the inhibitory contacts and activates the p85/p110 dimer.

2. Membrane targeting. Class I PI 3-kinases interact with RTKs, activated Ras, activated Rac, Gβγ subunits, and SH3 domains from Src-family tyrosine kinases [1]. All of these proteins are integral membrane proteins or are membrane associated, through the covalent attachment of myristoyl, palmitoyl and isoprenoid lipids. Given that the substrate of the PI 3-kinases is a membrane lipid, increased membrane targeting is sufficient to increase the rate of PIP3 production by PI 3-kinases. This was most clearly shown by the hyperactive phenotype induced by direct membrane targeting of PI 3-kinase catalytic subunits by addition of a C-terminal CAAX motif [25]. Using deuterium exchange-mass spectrometry methods, Williams and coworkers have noted that the binding of tyrosine phosphorylated peptides to p85 SH2 domains leads to conformational changes that enhance p110α and p110δ interactions with membranes[26]. Both p110β and p110γ bind to Gβγ subunits, which are released downstream from activated G-protein-coupled receptors (GPCRs). For p110β, the binding site has been localized to the C2 domain-helical domain linker [27]. Activation by Gβγ binding is synergistic with activation of phosphotyrosine peptide binding to the p85 N-terminal SH2 domain, and may act in part by disrupting the inhibitory contact formed by this domain. However, activation by Gβγ is also seen with monomeric p110β or p110γ[28], suggesting that a large component of Gβγ-mediated activation is due to membrane targeting. A similar mechanism has been suggested for the activation of p101/p110γ by Gβγ[29]

3. Allosteric activation by small GTPases binding. As mentioned above, both small GTPases and trimeric G-proteins directly interact with Class I PI 3-kinase. All the p110 catalytic subunits contain a Ras-binding domain, although direct binding to Ras-GTP has only been demonstrated for p110α, p110δ and p110γ[30]. A crystal structure of the p110γ/Ras-GTP complex indicated conformational changes in the catalytic cleft, which presumably increases enzyme activity synergistically with membrane targeting [31]. Interestingly, it has been shown that activation by Ras preferentially occurs with the p87/p110γ dimer, as opposed to with the p101/ p110γ dimer [32], suggesting that the Class IB regulatory subunits influence the interactions of p110γ with Ras. Finally, binding of activated Rac or Cdc42 to the so-called BCR homology domain in the p85 subunit activates p85/p110α dimers in vitro[33], all this has not been demonstrated in intact cells. The mechanism of activation is not known, as there is little information on the conformation/structure of the N-terminal half of p85.

4. Disruption of SH3/proline rich domain binding. The SH3 domain of p85, which resides at the extreme N-terminus of the protein, has been shown to bind to peptides derived from the internal proline-rich domains (PRDs) in p85. These data suggest that the N-terminus of p85 may exist in a closed conformation mediated by intramolecular SH3-PRD interactions. Binding events that would be predicted to disrupt such a conformation activate p85/p110 dimers. For example, p85/p110 dimers are activated by SH3 domains Src-family kinases [34], which can bind to p85 PRDs. While these data suggest a model in which exogenous proline-rich peptides or SH3 domains disrupt an inhibitory intramolecular SH3-PRD interaction, there is no direct evidence supporting this model. Given that the BCR domain lies between the two proline-rich domains, it is possible that Rac/Cdc42 binding to the BCR domain also functions by disrupting intramolecular SH3-propline-rich domain interactions.

Activating oncogenic mutations are commonly found in the p110α catalytic subunit [35], as well as in the p85 regulatory subunit [36,37]. The mechanisms by which these mutations activate PI 3-kinase are in some cases known. For example, the frequently mutated E545K residue in the p110α helical domain disrupts the inhibitory contact made by the N-terminal SH2 domain [20], and thus mimics phosphotyrosine-mediated activation. In contrast, a second commonly mutated site in p110α, H1047R, acts by increasing the basal association of p85/p1107aga; dimers with membranes [21]. The net effect of most activating mutants of p110α appears to be to induce conformational changes that increase membrane binding [26]. Mutants of both p85 and p110α in human tumors have been described; these mutants occur at an inhibitory contact between the C2 domain of p110α and the iSH2 domain of p85 [36,38,39]. It appears that these mutations mimic a loosening of the C2 domain-iSH2 domain interface that occurs in response to phosphotyrosine-mediated displacement of the inhibitory nSH2 domain [26].

PIP3 phosphatases shape the kinetics of PIP3 signaling

The importance of the timely elimination of PIP3-mediated signals, once they are initiated, is illustrated by the fact that the PIP3 3-phophatase, Phosphatase and Tensin Homology (PTEN), is frequently mutated or eliminated in human cancers [40,41]. The normal functioning of metazoan cells requires that the PIP3 signal be sharply constrained in both time and space.

The half-life of PIP3 in cells has been difficult to measure, but appears to be quite short. Using a PIP3 antibody to measure the decay in EGF-stimulated plasma membrane PIP3 induced by acute treatment with the PI 3-kinase inhibitor LY294002, the half-life for PIP3 was estimated to be less than 5 sec [42]. Of note, this study also found that a commonly used method to measure PIP3 levels in live cells, expression of GFP-labeled PH domains, led to an approximately 5-fold increase in the half-life of PIP3. This effect was presumably due to the sequestration of PIP3 by PH domain overexpression. This analysis suggests that measurements of PIP3 dynamics using PH domain probes, which have estimated the half-life of PIP3 to be approximately 1 min [43], likely overestimate the duration of the PIP3 signaling.

The overall dynamics of the cellular PIP3 response is stimulation dependent. In fMLP-stimulated neutrophils, PIP3 levels peaked by 20 sec and declined to close to basal levels by 2 min [44]. In PDGF-stimulated fibroblasts, PIP3 peaked by 40 sec but remained elevated at 5 min [45]. Other studies have suggested that sustained elevations of PIP3 occurs in response to acute stimulation of receptor tyrosine kinases [3,46]. The lifetime of the PIP3 signal in specialized cases is more transient. For example, in macrophages, there is an accumulation of PIP3 in the phagocytic cup during FcRγ-mediated phagocytosis; PIP3 disappears during closure of the phagosome, over a period of 2–3 min[47,48]. In Dictyostelium acutely exposed to a chemoattractant gradient, PIP3 accumulates at the leading edge in a biphasic manner, with a transient peak at 20 sec followed by a sustained elevation at 2 min[49].

The rapid turnover of newly synthesized PIP3 occurs either through removal of the 3-phosphate by PTEN, returning the lipid to the PI[4,5]P2 precursor, or through removal of the 5-phosphate by phosphatases such as SHIP or synaptojanin [50]. PTEN-mediated hydrolysis clearly terminates the PIP3 signal, and numerous studies have shown that inactivation or deletion of PTEN leads to increased levels of PIP3 [40]. The downstream effects of the activity of the 5-phosphatases is less clear, since their product is a lipid that can still signal via PH domains that bind to PI(3,4)P2, including those of Akt or Tapp1 [51,52]. Nonetheless, knockout of SHIP enhances the inflammatory response of myeloid and mast cells [53,54], clearly indicating its role as a negative regulators of PI 3-kinase signaling.

Interestingly, studies in Dictyostelium have shown that PTEN may contribute to certain PIP3-mediated cellular responses, particularly those that require an asymmetric spatial distribution of the lipid. cAMP-stimulation of Dictyostelium leads to an acute and spatially restricted accumulation of PIP3 at the leading edge of the cell [55]. This is accomplished at least in part by PTEN localization to the sides and rear of the cell, and its exclusion from the leading edge [56]. The positive role of PTEN in the chemotactic response is demonstrated by the impaired directionality of PTEN null cells, which produce poorly coordinated protrusions that reduce movement toward the chemoattractant source [57].

The biology of PTEN deletion involves phenotypes that are more complex than can be explained simply by an increase in PIP3 levels or lifetime. PTEN has a protein phosphatase activity whose disruption leads to distinct phenotypes [58,59]. PTEN also has signaling functions in the nucleus that do not appear to involve dephosphorylation of PIP3 [41,60].

How do different Class I PI 3-kinase isoforms achieve signaling specificity?

The distinct functions of the Class I PI 3-kinases have been studied in vitro, in cells and in animal models. Abundant evidence has accumulated showing that these lipid kinases perform quite different functions [2]. While isoform specific signaling by a protein kinase could easily be due to distinct substrate specificities, all the Class I PI 3-kinases make the same product, PIP3. In that case, how is isoform-specific signaling achieved? The question of how different p110 isoforms perform distinct tasks is especially interesting for the Class IA catalytic subunits, which share the same regulatory subunits and therefore should have similar interactions with tyrosine phosphorylated proteins.

Isoform specific signaling by PI 3-kinases requires combinatorial inputs

Taken together, it does not seem as if any one activation or targeting mechanism can produce a specific or preferential activation of any given PI 3-kinase isoforms (Figure 2). In some cases, the predominance of one isoform over another may be explained by differences in the kinetic properties of the isoforms; this seems most likely to be the case for p110α, given its higher basal activity [66,67]. Alternatively, isoform-specific PI 3-kinase signaling, in a cell that contains multiple isoforms, may require a combination of localization constraints and synergistic activation by multiple activators (Table 1). For example, in a growth factor-stimulated cell, p85/p110α and p85/p110β might both reside in an endosome, via receptor tyrosine kinase binding for the former and Rab5 binding for the latter. If the growth factor were a potent activator of endosomal Ras, this would synergistically activate p85/p110α but not p85/p110β, and lead to a signal that was primarily p110α-dependent. A combined RTK/GPCR signal might preferentially activate p85/p110β, whereas a combined GPCR/Ras signal might preferentially activate p87/p110γ. Of all the isoforms, differential signaling by p110α and p110δ was the first to be described [108] and remains the hardest to understand. Both isoforms bind RTKs and Ras, and both undergo autoinhibitory autophosphorylation (although by different mechanisms [109,110]. The ability of RTKs to distinguish between these isoforms may reside in the still mysterious cellular code for the matching of p110 isoforms (p110α, p110β, and p110δ) with p85 isoforms (p85α, p55α, p50α, p85β and p55γ).

Open in a separate window

Figure 2

Pathways to isoform-specific PI 3-kinase activation

While no one PI 3-kinase activator yields specific or preferential activation of any one PI 3-kinase isoform, combinations of activators and distinct activator locations could allow differential activation of distinct PI 3-kinases.

Finally, it should be noted that PI 3-kinases have activities not directly related to PIP3 production, which may affect signaling output. All of the Class I PI 3-kinases have protein kinase activity. Although in most cases the activity seems to be limited to autophosphorylation [109–111], a protein kinase-only mutant of p110γ was able to signal to Erk [112] and a limited number of the protein substrates have been described [113,114]. In addition, PI 3-kinases have been shown to have biological activities that do not require an active kinase domain [115]. For example, mice expressing kinase dead p110β are viable, although infertile, whereas homozygous knockout of p110β produces embryonic lethality [97,98,116]. p110γ, in particular, has been shown to regulate contractility in the heart by serving as a scaffold for phosphodiesterase 3B and G-protein-coupled receptors kinase 2, which regulates the β2-adrenergic receptor [117]. The scaffolding activity of the PI 3-kinase regulatory and catalytic subunits is likely to contribute to the spectrum of downstream responses to distinct PI 3-kinase isoforms.

Conclusions and outlook

The focus of much of current research on PIP3-mediated signaling is on the identification of novel downstream effectors, but there is still a great deal to be learned about the activation, targeting and specificity of the enzymes producing and regulating the production of PIP3 itself. A review of our knowledge of the activation mechanisms for Class I PI 3-kinasese suggests that differential outputs from the distinct PI 3-kinase isoforms depends on different combinations of simultaneous upstream inputs, different sites of action, as well as contributions from kinase-independent scaffolding functions. Given the compelling data linking aberrant regulation of Class I PI 3-kinase to pathological outcomes, a more precise understanding of how these enzymes are regulated in cells should hold considerable promise for novel interventions into human disease.

Acknowledgments

List of Abbreviations

References