Uncovering the PI3Ksome: Phosphoinositide 3-Kinases and Counteracting PTEN Form a Signaling Complex with Intrinsic Regulatory Properties

Abstract

Production of the phosphoinositide lipid phosphatidylinositol (3,4,5)trisphosphate [PI(3,4,5)P₃, or PIP₃] by class I phosphoinositide 3-kinases (PI3Ks) is a major signaling mechanism whose deregulation contributes to serious diseases, including cancer. New findings suggest that tyrosine kinase receptor engagement results in the assembly of hetero-oligomeric PI3K complexes in which PI3Kα first activates PI3Kβ, and PI3K catalytic activity then promotes recruitment and activation of the PIP₃-removing tumor suppressor PTEN. Thus, PIP₃ production is fine-tuned through formation of an intrinsically regulated “PI3Ksome.”

TEXT

The phosphoinositide 3-kinase (PI3K) signaling pathway is activated by many cellular receptors and controls numerous processes, including the cell cycle, metabolism, survival, migration, and genome stability. It is the most frequently deregulated signaling pathway in human cancers. Not surprisingly, PI3Ks are being pursued as therapeutic targets for cancers, inflammatory diseases, and other indications (1).

Among three mammalian PI3K classes, the most cancer-relevant class I PI3Ks α to δ phosphorylate the inositol ring 3-position of the membrane lipid phosphatidylinositol (4,5)bisphosphate [PI(4,5)P₂)] to generate phosphatidylinositol (3,4,5)trisphosphate [PI(3,4,5)P₃]. PI(3,4,5)P₃ binds and recruits key signaling effectors, such as Tec, Pdk1, and Akt family protein kinases to cellular membranes, resulting in their incorporation into “signalosome” complexes, activation, and engagement of downstream pathways (1, 2). PI(3,4,5)P₃ cellular hyperactivity severely dysregulates signaling and can transform cells. To avoid this, PI(3,4,5)P₃ levels are tightly regulated through receptor-controlled PI3K activation and PI(3,4,5)P₃ removal by lipid phosphatases, such as the tumor suppressor PTEN, which reverses the PI3K reaction (1–3) (Fig. 1A). Despite their importance, it has remained unclear whether and how PI3K and PTEN interact molecularly in vivo. In this issue of Molecular and Cellular Biology, Pérez-García et al. (4) present results from orthogonal genetic and pharmacologic studies to address this important question. They show that tyrosine kinase receptor engagement results in the assembly of hetero-oligomeric PI3K complexes in which PI3Kα first activates PI3Kβ and then the PI3K catalytic activity promotes recruitment and activation of PTEN (Fig. 1B). Thus, rather than being regulated by independent enzymes, PI(3,4,5)P₃ production is fine-tuned through formation of an intrinsically regulated signaling module, for which we propose the term “PI3Ksome.” The initial feed-forward amplification of PI3K activity followed by PTEN negative feedback shapes a biphasic response which augments PI(3,4,5)P₃ levels and signaling early after receptor stimulation but later restricts them.

FIG 1.

(A) PI3K and the lipid phosphatases PTEN and SHIP antagonistically control cellular levels of PI(3,4,5)P₃, which recruits and activates signaling effector proteins by binding to their pleckstrin homology (PH) domain and other domains. (B) A new study by Pérez-García and colleagues (4) suggests that PI3K signaling involves multimeric “PI3Ksome” complexes whose activity is intrinsically regulated through p110α activation of p110β and p110 activation of PTEN.

The PI3Ks α, β, and δ are obligate heterodimers of a 110-kDa catalytic subunit (p110α, p110β, and p110δ, respectively) and a regulatory subunit (p85α, p85β, p55α, p55γ, and p50α) which stabilizes and inhibits p110 in resting cells (5). Upon cell stimulation, the SH2 domains of the regulatory subunits bind to tyrosine-phosphorylated membrane receptors, recruiting p85/p110 into proximity with PI(4,5)P₂ and causing p110 activation (5). Pérez-García et al. showed that the p85 subunits also mediate the assembly of the PI3Ksome, an ∼440-kDa hetero-oligomeric complex that contains p85, p110α, and p110β. Their findings expand those of a previous study that found that p85 subunits can oligomerize p110α (6). p85 molecules dimerize through their SH3-BcR/BH domains and bind p110 through their two C-terminal SH2 domains and intervening region (7). Based on the new study, both these interactions may be required for p110α/p110β cooligomerization, likely resulting in a p110α/p85/p85/p110β “sandwich” arrangement (Fig. 1B).

Pérez-García and colleagues also found that interfering with p110α function reduced p110β activity in serum-stimulated cells (4), but p110β inactivation had little effect on p110α activity. Moreover, p110β but not p110α had higher activity in the heterodimeric ∼440-kDa PI3K complex than in monomeric fractions. Disruption of p110α/p110β heteromers reduced p110β-dependent PCNA loading onto chromatin and also p110β but not p110α activity in serum-stimulated cells. Consistent with p85 involvement in p110β activation by p110α, p85-independent G protein-coupled receptor (GPCR)-induced p110β activation appeared to be less p110α dependent than p85-dependent platelet-derived growth factor receptor activation of p110β. Based on these data, assembly of the PI3Ksome allows PI3Kα to activate PI3Kβ (Fig. 1B). This could explain the sequential activation of p110α and -β in the cell cycle, although this was not studied in detail by Pérez-García et al.

Besides the ∼440-kDa PI3Ksome, serum stimulation also induced the formation of an ∼660-kDa complex containing both p110α and p110β. Previous studies had shown that p85α, p85β, p110α, p110β, and PTEN can form an ∼660-kDa PTEN-associated complex (PAC) and that p85 subunits are needed for PTEN interactions with p110 in vitro (8, 9). Pérez-García et al. showed that PTEN is not required for p110α/p110β heterodimerization but that p110 subunits are required for PTEN incorporation into the PAC. Moreover, serum stimulation increased PAC-associated PTEN activity more than monomeric PTEN activity. Knockdown of p85α, p85β, p110α, or p110β, or pharmacologic inhibition of p110α or p110β, reduced receptor-induced PTEN activity and increased late (but not early) Akt activation. This suggests that the p110α and p110β catalytic activities are required for full PTEN activation (Fig. 1B).

Although some details remain to be elucidated, the new data add significant new insights to our understanding of how PI3Ks and PTEN interact in cells and raise several important questions. They suggest that assembly of the ∼440-kDa and ∼660-kDa PI3Ksomes generates intrinsically regulated catalytic modules where positive and negative circuits control PI(3,4,5)P₃ production (Fig. 1B). Receptor engagement first activates PI3Kα, which produces some PI(3,4,5)P₃ and activates PI3Kβ. PI3Kβ then augments PI(3,4,5)P₃ production. Next, the complex recruits and activates PTEN, which counteracts the PI3Ks and removes PI(3,4,5)P₃ to limit its accumulation or reverse it to prestimulation levels, essentially turning off PI3K downstream signaling. What is the purpose of organizing these events in intrinsically regulated modules? Combining positive and negative feedback can (i) render signaling robust against cell-to-cell variations in reactant levels or the microenvironment, (ii) tune signaling sensitivity, amplitude, and kinetics, or (iii) distinguish between gradual and digital responses (10, 11). Clarifying whether the PI3Ksome serves such purposes will require a combination of mathematical simulations with experiments in single cells. Another plausible possibility is that topologically constraining PI3K and PTEN function into modules limits PI(3,4,5)P₃ accumulation to specific membrane domains, avoiding its production in subcellular regions where it would be detrimental. In two examples, this could be critical for the localized PI(3,4,5)P₃ function at adherens junctions or T cell immunological synapses (12, 13). Addressing these possibilities will provide leads toward elucidating the biological significance of the PI3Ksome, which is clearly a major open question.

A perhaps even more important question raised by the new study is whether reduced PTEN activation by PI3Kα/β could limit the therapeutic efficacy of PI3Kα/β inhibitors or contribute to drug resistance, particularly in cancers. Related to this question, it will be important to determine if PI3Ksome abundance and composition differ between cell types, stimuli, context, or subcellular compartments, or are altered by disease-associated mutations of PI3Ksome components. In particular, do some cancer-associated PTEN mutations disrupt its tumor suppressor function by impairing PI3Ksome recruitment? With only ∼50% of PI3K oligomerizing (according to Pérez-García et al. [4]), it will also be important to find out what determines which PI3K molecules oligomerize and whether the complex also recruits PTEN.

Besides PTEN, the lipid phosphatase SHIP can also remove PI(3,4,5)P₃ by dephosphorylating it into PI(3,4)P₂ (1, 2, 14) (Fig. 1A). SHIP can bind p85 (5). Thus, it will be interesting to find out whether SHIP is also part of the PI3Ksome or if it associates with PI3K in a distinct complex.

Beyond these broad implications, the new study by Pérez-García et al. also raises some interesting questions about mechanistic details. For example, how does PI3Kα promote PI3Kβ activation, and why does PI3Kβ not activate PI3Kα? Or, how do PI3Ksome interactions activate PTEN? Although monomeric p85 can bind and activate PTEN in vitro, cellular p85 is not monomeric, and studies of an endometrial cancer-specific p85α mutant suggest that PTEN may be stabilized by dimeric p85 (5, 9, 15). So, PI3Ksome assembly by p85 and direct p85-dimer binding to PTEN may both be important, but more detailed studies are required. Whatever the precise mechanism, reduced PTEN activity due to PI3Ksome disruption may provide a novel explanation for the puzzlingly reduced PTEN activity and elevated PI(3,4,5)P₃ signaling in p85α-deficient insulin-responsive liver cells, changes that were previously hypothesized to reflect functional p85 replacement by p50 for activation of p110 but not PTEN, or reduced p85 competition with p85/p110 heterodimers for receptor/adaptor binding (3, 5, 9, 16, 17).

Functional genomics and proteomics have transformed our understanding of signaling from simple cascades into complex networks organized into topologically distinct signaling “hubs” (10). The data reported by Pérez-García et al. suggest that the “hub” concept includes intrinsically regulated PI3Ksome modules (4). Their further study promises exciting novel insights into one of the most important cellular signaling pathways.