PI3Kβ—A Versatile Transducer for GPCR, RTK, and Small GTPase Signaling

Abstract

The phosphoinositide 3-kinase (PI3K) family includes eight distinct catalytic subunits and seven regulatory subunits. Only two PI3Ks are directly regulated downstream from G protein–coupled receptors (GPCRs): the class I enzymes PI3Kβ and PI3Kγ. Both enzymes produce phosphatidylinositol 3,4,5-trisposphate in vivo and are regulated by both heterotrimeric G proteins and small GTPases from the Ras or Rho families. However, PI3Kβ is also regulated by direct interactions with receptor tyrosine kinases (RTKs) and their tyrosine phosphorylated substrates, and similar to the class II and III PI3Ks, it binds activated Rab5. The unusually complex regulation of PI3Kβ by small and trimeric G proteins and RTKs leads to a rich landscape of signaling responses at the cellular and organismic levels. This review focuses first on the regulation of PI3Kβ activity in vitro and in cells, and then summarizes the biology of PI3Kβ signaling in distinct tissues and in human disease.

The ubiquitously expressed p110β was first cloned by homology to p110α (1) and is coded by the PIK3CB gene. Similar to other class IA catalytic subunits, p110β exists as an obligate heterodimer with a p85, p55, or p50 regulatory subunit. It was thought to be primarily regulated by receptor tyrosine kinases (RTKs) until the demonstration of its activation by Gβγ in the late 1990s (2). Subsequent studies confirmed its activation by a number of G protein–coupled receptor (GPCR) agonists (3–5). Interestingly, additional work suggested that phosphoinositide 3-kinase (PI3K) β was poorly activated by RTKs as compared with PI3Kα (6). Recent studies from neutrophils suggest that PI3Kβ is minimally responsive to individual RTK or GPCR ligands, and instead serves as a coincidence detector for combined GPCR/RTK stimuli (7). Thus, the regulation of PI3Kβ appears to be complicated and may vary between different cell types. The physiology of PI3Kβ signaling in animals is also complex. Whereas knockout of the p110β catalytic subunit leads to early embryonic lethality (8), its role in cell proliferation is most obvious in the context of tumor cells that have lost expression of the PTEN tumor suppressor (9, 10). In cancer, PI3Kβ also plays important roles in tumor cell invasion and metastasis (11). In normal tissues, PI3Kβ is critical for spermatogenesis and for macrophage, osteoclast, neutrophil, and platelet function, although the mechanisms involved are not yet clear (12–16). Given the clinical evaluation of PI3Kβ inhibitors for cancer and other illnesses (17), this unusual PI3K isoform is an important area of current research.

Structure of PI3Kβ

Structural studies on PI3Kβ have been discussed extensively in recent papers and reviews (18, 19). Similar to all the class IA PI3Ks (PI3Kα, PI3Kβ, and PI3Kδ), PI3Kβ is composed of a regulatory subunit (p85α, p85β, p55γ, p50α, or p55α) and a 110-kDa catalytic subunit (p110β) (Fig. 1). Key domains in the p85 regulatory subunit include an SH3 domain, which may form intramolecular contacts with the p85 proline-rich domains (20) and is critical for the formation of p85α homodimers (21); two proline-rich motifs that can bind to SH3 domains in Src family kinases and other proteins (22); a breakpoint cluster region (BCR) homology domain that binds to Rho family GTPases (23); two SH2 domains (nSH2 and cSH2), which recruit PI3Kβ to tyrosine-phosphorylated proteins containing pYXXM motifs (24); and a 100-Å antiparallel coiled coil, the iSH2 domain (25–27) (Fig. 1). In terms of interactions with p110β, the key region of p85 is the iSH2 domain, which forms an extremely tight interface with the adapter-binding domain (ABD) at the N terminus of p110β. p85/p110 Dimers are not thought to dissociate once formed in the cell, and p110 subunits are unstable in vitro and in vivo in the absence of p85 (28, 29). p110 Catalytic subunits additionally contain a Ras-binding domain (RBD) as well as C2, helical, and kinase domains (Fig. 1).

Figure 1.

Model of PI3Kβ and its interactions with tyrosyl phosphoproteins and Rho family GTPases. PI3Kβ is a heterodimer composed of a regulatory subunit and the p110β catalytic subunit. The structural domains of the p85 regulatory subunit [SH3, proline-rich (PPP), BCR, SH2, and iSH2 domains, shown in green] and the p110β catalytic subunit (ABD, RBD, C2, helical, and kinase domains) are shown. The model is based on the structure of p110α bound to the nSH2-iSH2 fragment of p85α, and p110β bound to the iSH2-cSH2 fragment of p85β. The ABD of p110β binds tightly to the iSH2 domain, which forms an antiparallel coiled coil. The C2 and kinase domains drape over the iSH2 domain, which makes regulatory contacts with the C2 domain (18). The nSH2 and cSH2 domains make inhibitory contacts with the helical, C2 and kinase domains (nSH2) or just the kinase domain (cSH2). The positions of the SH3, proline-rich, and BCR domains relative to the remainder of the molecule are not known. PI3Kβ is activated when phosphotyrosyl residues in RTKs or their substrates bind to the SH2 domains and disrupt the inhibitory contacts. PI3Kβ is also activated when GTP-bound Rac1 or Cdc42 binds to the RBD.

There are currently no structures of the full-length class IA PI3K heterodimer. However, structures of p110α and p110β bound to fragments of p85α or p85β (nSH2-iSH2 or iSH2-cSH2, respectively) have been solved (18, 30, 31). A structure of p110δ lacking the ABD has also been solved, but it is not informative with regard to regulation by p85, as it cannot bind to the iSH2 domain (32). However, deuterium exchange/mass spectrometry (DXMS) studies suggest that p85 regulates p110β and p110δ in a similar fashion (33).

In the X-ray structures of PI3Kα and PI3Kβ, the p110 ABD binds to the distal end of the iSH2 domain of p85, and the kinase and C2 domains drape over the proximal end of the iSH2 domain to make additional regulatory contacts (Fig. 1). As first predicted by biochemical studies from our group (27, 34), regulation of class IA PI3Ks by tyrosine-phosphorylated proteins involves the disruption of inhibitory contacts between the p85 SH2 domains and the catalytic subunit. In the nSH2-iSH2/p110α structure, the nSH2 domain of p85α contacts the helical, C2, and kinase domains of p110α (30). In the iSH2-cSH2/p110β structure, the cSH2 domain contacts only the kinase domain (18). In both structures, the so-called RBD is the only p110 domain that makes no direct contacts with the p85 regulatory subunits.

Regulation of PI3Kβ Activity

Activation by GPCRs

The binding site for Gβγ in the p110β catalytic subunit of PI3Kβ was defined by a combination of site-directed mutagenesis and mapping of contacts by DXMS (36). The binding site is a flexible loop, not visible in the X-ray structures, within the linker between the C2 and helical domains (Fig. 2). Although the sequences of the loops are divergent between p110β and p110γ, which also binds Gβγ, a pair of basic residues at the C-terminal end of the loops are critical for interactions of both kinases with Gβγ (36, 37). For PI3Kβ, mutation of these residues to Asp abolishes its activation by Gβγin vitro and in vivo. For PI3Kγ, the corresponding mutation abolishes Gβγ activation of p110γ monomers, but only partially reduces the sensitivity and magnitude of Gβγ activation of p101/p110γ heterodimers (37). This residual activation of PI3Kγ is due to the p101 regulatory subunit of PI3Kγ, which also contributes to Gβγ binding. The structure of p101 is not known, and its sequence contains no known domains. However, DXMS experiments show a decrease in solvent accessibility in the C terminus of p101 in the presence of Gβγ, suggesting that this region directly interacts with Gβγ (37). No interactions between Gβγ and the p85 regulatory subunit of PI3Kβ were detected by DXMS (36).

Figure 2.

Interactions with PI3Kβ with Gβγ and Rab5. The figure is a 180° rotation of Fig. 1. Gβγ binds to a flexible loop in the linker between the C2 and helical domains. The positions of PI3Kβ mutations that disrupt Gβγ binding K526K-DD are shown. In this view, the Rab5 binding site is located immediately behind the Gβγ loop, in the helical domain. The color scheme is as in Fig. 1, with p85 in green.

DXMS studies comparing PI3Kβ and p101/p110γ in the absence or presence of lipid vesicles show that for both enzymes, there is a striking increase in solvent accessibility of the Gβγ binding loop when lipids are present (36, 37). For PI3Kβ, interactions with Gβγ in the absence of lipids can only be detected when the two proteins are physically linked (e.g., in a chimera in which Gβγ is linked to the iSH2-cSH2 fragment of p85, which binds tightly to p110β). The finding that the interaction surface is buried in the absence of membranes may explain this observation, as well as anecdotal findings that Gβγ and PI3Kβ cannot be coimmunoprecipitated from detergent lysates that disrupt membranes.

The situation is more complicated for PI3Kγ. The Gβγ interacting region in p110γ shows a decrease in solvent exposure upon binding to the p101 regulatory subunit, which is reversed when the p101/p110γ dimer is incubated with lipid membranes (37). Thus, p101 masks the Gβγ binding site in the absence of membranes, preventing interactions between p110γ and Gβγ, but it allows Gβγ binding in the presence of membranes. Overall, p101 enhances the binding of PI3Kγ to Gβγ by contributing additional contacts to the complex. Mutation of p101, at residues that were identified by DXMS as binding to Gβγ, reduces the sensitivity and magnitude of Gβγ-mediated PI3Kγ activation in vitro, and mutation of the Gβγ binding sites in both p101 and p110γ abolishes activation (37). In intact cells, mutation of the Gβγ binding sites in either p110γ or p101 causes near total loss of PI3K activity in response to GPCR ligands (37).

For both PI3Kβ and PI3Kγ, membrane recruitment by interactions with Gβγ would increase the availability of lipid substrates. For PI3Kβ, the proximity of the Gβγ binding site to the nSH2 contact site in the p110β helical domain suggests that displacement of the inhibitory nSH2 domain may also be involved. Consistent with this, activation of PI3Kβ by Gβγ and tyrosine phosphopeptides is strongly synergistic (36). However, recombinant p110β is activated by Gβγ in the absence of p85, or in a complex with a fragment of p85 that lacks the nSH2 domain, suggesting that Gβγ directly affects the catalytic subunit (38). A better understanding of how Gβγ activates PI3Kβ may require a more complete structure of the nSH2/p110β interface.

Activation by small GTPases

The catalytic subunits of all class I PI3Ks contain a so-called RBD (Fig. 1). The RBDs of p110α, p110δ, and p110γ bind to Ras in a GTP-dependent manner. For p110β, the RBD binds to the Rho family GTPases Rac and Cdc42, and it does not interact with Ras (39). The dissociation constants for binding Rac and Cdc42 are 1.5 and 2.9 μM, respectively. Rac-GTPγS and Cdc42-GTPγS stimulate PI3Kβ lipid kinase activity by fivefold in vitro; in cells, activation by Rac and Cdc42 is additive with activation by Gβγ (40).

PI3Kβ is unique among class I PI3Ks in that it binds directly to activated Rab5. Early studies showed that Rab5-GTP binds to in vitro–translated p85α/p110β heterodimers or p110β monomers, but not to p85α monomers (41). Recent work from our laboratory also failed to detect Rab5 binding to isolated p85 (42). In contrast, Anderson and colleagues (43) have suggested that Rab5 binds in a GTP-independent manner to the BCR domain of p85. The Rab5 binding site in the p110β catalytic subunit was initially identified by our group, using a conservation-based scanning mutagenesis approach. We identified two residues whose mutation abolished PI3Kβ binding to Rab5-GTP (Q596 and I597; Fig. 2) (44). We recently mapped the Rab5 binding site in greater detail (42) and showed that it consists of two perpendicular α-helices (D509-A519 and K584-I597) in the helical domain, which correspond to α-helices Lα5 and Hα2B in the Zhang et al. (18) crystal structure (Protein Data Bank ID 2Y3A).

A recent paper from Whitecross and Anderson (45) has suggested that Rab5 also binds to the p110β RBD. However, the binding experiments in this study used a chimeric protein in which a fragment of the p85 iSH2 domain (residues 466 through 567) was fused to the N terminus of p110β; the goal was to produce a monomeric p110β that was stabilized through an intramolecular iSH2–ABD interaction. The iSH2 fragment used has not been structurally or biochemically characterized, and it deletes regions of the iSH2 that are known to make regulatory contacts with p110 catalytic subunits (31, 46–48). Heitz et al. (42) directly examined the RBD mutants that disrupted Rab5 binding to the iSH2^466–567-p110β chimera (45); none of them disrupted binding to full-length p85/p110β. Thus, there appears to be a single discrete PI3Kβ binding site for Rab5, located in the helical domain of p110β.

The data on Rab5 activation of PI3Kβin vitro has been inconsistent. An early study examining the binding of PI3Kβ to Rab5 suggested that the interaction activates PI3Kβ, but this conclusion was based on Akt phosphorylation in cells transfected with activated Rab5 (49). In more recent in vitro analyses, one group saw a twofold activation, whereas another group saw no activation (39, 50). Heitz et al. (42) also saw no effect of Rab5^GTP on the activity of recombinant PI3Kβ. Importantly, regulation of PI3Kβ by prenylated Rab5 has not been tested in vitro. Prenylated Rab5 is likely to enhance PI3Kβ binding to liposomes in vitro, which would increase substrate phosphorylation. Determination of the effects of prenylated Rab5 on PI3Kβ-specific activity will require single-molecule methods, analogous to recent studies on Ras activation of PI3Kα (51).

Activation of PI3Kβ by RTKs

The p85 regulatory subunits of the class IA PI3Ks contain two SH2 domains that make inhibitory contacts with the p110 catalytic subunits; these contacts are disrupted when tyrosine-phosphorylated proteins or peptides bind to the SH2 domains (Fig. 1). This model was first proposed based on biochemical studies with p110α and fragments of p85α (26), but is also applicable to PI3Kβ. In the PI3Kα heterodimer, the nSH2 domain of p85 contacts the helical, kinase, and C2 domains of p110α (30). The phosphotyrosine binding site is exactly coincident with the helical domain contact and is disrupted by phosphopeptide binding (27, 30). Moreover, at the corresponding contact site in the helical domain of p110α, E542K, E545K, and Q546K disrupt nSH2-mediated inhibition of p110α and these mutations are commonly found in patient tumors (27, 52). Although the X-ray structure of p110β does not include the p85 nSH2 domain (18), DXMS data show that phosphopeptide binding leads to a loss of nSH2 domain contacts with the p110β helical domain (53). However, the precise molecular contacts are not known. Although p110β N553 (corresponding to helical domain hotspot p110α-Q546) is mutated in advanced prostate cancer (54), introduction of a mutation corresponding to p110α-E545K into p110β led to only modest PI3Kβ activation (55), suggesting that the nSH2 may be oriented differently with regard to the p110β helical domain.

The cSH2 domain of p85β makes an inhibitory contact with the C terminus of p110β (Fig. 1) (18). Unlike the nSH2–helical domain contact, the phosphotyrosine binding site of the cSH2 domain is slightly displaced from the inhibitory contact. As a consequence, displacement of the cSH2, as well as activation of the enzyme, requires longer peptides than the minimal four–amino acid peptides that bind with high affinity to the p85 nSH2 domains (56). DXMS experiments suggest that a similar mode of cSH2 inhibition applies to p110δ but not p110α (18). Conversely, earlier studies showed that tyrosine-phosphorylated peptides do in fact activate PI3Kα by binding to the cSH2 domain (57, 58). The reason for the discrepancy between these studies is not clear, but it could involve a distinct mechanism for cSH2 interactions with p110α (58).

As discussed above, despite the relative proximity of the Gβγ binding site to the nSH2 contact site in the helical domain (Fig. 2), Gβγ activation is not mediated solely by nSH2 displacement. This is most clearly demonstrated by the synergy between tyrosine phosphopeptides and Gβγ for PI3Kβ activation (2, 36). Interestingly, the role of the cSH2 domain in this synergy has not been tested; it would be interesting to compare Gβγ and phosphopeptide activation of PI3Kβ containing a p85 regulatory subunit with disabling mutations in either the nSH2 or the cSH2 phosphopeptide binding sites.

Activation of PI3Kβ by mutation

PIK3CB is mutated infrequently in human cancers, particularly as compared with the commonly mutated PIK3CA (52). In advanced prostate cancer, mutations were found in regions near mutational hotspots in PIK3CA: Arg48, homologous to Arg38 in p110α; Asn553, homologous to Gln546 in p110α; and E1051K, which corresponds to Gly1049 in p110α and is next to the H1047R hotspot (54). A large-scale sequencing analysis of human tumors identified the D1067V mutation in five distinct tumor types (59). Mutation of D1067 has also been detected in PTEN-deficient breast cancer cells treated with the pan-PI3K inhibitor GDC-0941 (D1067Y) (60), in a patient with non–small cell lung cancer (D1067V) (61), and in gastrointestinal stromal tumors expressing activated KIT (D1067V) (62). Mutation of L35V in intraductal tubulopapillary pancreatic neoplasms (63) and E633K in an HER2-positive breast tumor (64) have been identified. Additional recurrent mutations are listed in the COSMIC database (https://cancer.sanger.ac.uk/cosmic). So far, the D1067V and E1051K mutations have been shown to be activating in in vitro and in cell-based studies (60, 65). The constitutive activation of PI3Kβ by mutation could provide a mechanism for PI3Kβ-dependent tumorigenesis in some genetic backgrounds.

Activation of PI3Kβ in cells: coincidence detection and membrane targeting

Recent studies suggest that PI3Kβ plays a unique role as an integrator of simultaneous RTK and GPCR activation (Fig. 3A). Analysis of phosphatidylinositol 3,4,5-trisphosphate (PIP₃) production in intact neutrophils, expressing mutated forms of the different class I isoforms and stimulated with GPCR and RTK ligands, showed a striking pattern (7). PIP₃ production in response to RTK ligands was partially dependent on all of the p85-coupled class IA enzymes. Surprisingly, PIP₃ production in response to GPCR ligands was dependent on PI3Kγ but not PI3Kβ. However, PIP₃ production in cells treated simultaneously with RTK and GPCR ligands was exclusively dependent on PI3Kβ. These data help explain a previous observation in neutrophils that reactive oxygen species (ROS) production in response to FcγR activation required PI3Kβ but not PI3Kγ (15). FcRγ ligation causes the autocrine production of the GPCR ligand leukotriene B₄. The combined signaling by FcRγ (which couples to tyrosine kinases) and the LTB4 receptor (BLT1, a GPCR) preferentially required PI3Kβ, despite that stimulation with LTB4 alone activated PI3Kγ. These studies suggest that PI3Kβ serves a special function as a coincidence detector for signaling from RTKs and GPCRs. A recent study suggests that coincidence detection is also required for RTK activation of PI3Kβ; in mouse embryonic fibroblasts (MEFs), PI3Kβ is recruited to activated PDGF receptors, but PI3Kβ-mediated PIP₃ production in PDGF-stimulated cells requires an intact RBD (66). This suggests that PI3Kβ activation by PDGF requires dual inputs from the tyrosyl-phosphorylated PDGF receptor and activated Rac1 or Cdc42.

Figure 3.

Coincidence detection and targeting in Gβγ activation of PI3Kβ. (A) In neutrophils, PI3Kβ is specifically activated by the combination of an RTK and a GPCR ligand. (B) In MEFs, targeting of PI3Kβ to lipid rafts is required for its activation by Gβγ. (C) In triple-negative breast cancer cells, PI3Kβ activation by Gβγ requires the formation of macropinocytotic cups.

Studies in hematopoietic cells suggest a similar convergence of GPCR and Rac1 signaling (Fig. 3B). In PTEN-null hematopoietic cells, knockdown of p110β reduces Akt and Rac1 activation (67). These defects are rescued by wild-type p110β but not by an RBD mutant that is defective for Rac1 binding. The finding that Rac1 binding to PI3Kβ is required for Rac1 activity suggests a positive feedback loop between these two proteins. Subsequent work from the same group showed that in p110α/p110β double-knockout MEFs rescued with p110β, an intact RBD was required for GPCR-mediated activation of Akt (68). The mechanism involves the targeting of PI3Kβ to lipid rafts through its binding to Rac1; wild-type p110β, but not an RBD mutant, was detected in detergent-resistant membranes, and LPA-stimulated Akt activation was rescued by a raft-targeted p110β RBD mutant. Surprisingly, general plasma membrane targeting of the p110β RBD mutant, by addition of the Kras C-terminal CAAX motif, did not activate Akt under basal conditions, and it did not support LPA-stimulated Akt activation. Taken together, these studies describe a different form of coincidence detection, in which activation of PI3Kβ by GPCRs in some circumstances requires simultaneous engagement of both Rac1 and Gβγ, either through a positive feedback loop or via membrane targeting.

A distinct form of membrane targeting is required for GPCR stimulation of PI3Kβ in MDA-MB-231 breast cancer cells (40). In contrast to MEFs, activation of PI3Kβ by GPCRs did not require an intact Rac1 binding domain. Instead, it was shown that GPCR activation of PI3Kβ required the formation of macropinosomes (Fig. 3C). These clathrin-independent endocytic vesicles are formed by the maturation of circular dorsal ruffles into macropinocytotic cups (69). Macropinocytotic cups have been shown to amplify PI3K signaling in macrophages (70), presumably by limiting diffusion of 3-phosphoinositides in a manner analogous to integrin-mediated diffusion barriers in phagocytic cups (71). In MDA-MB-231 cells, inhibition of macropinocytosis blocked activation of PI3Kβ by both constitutively active Rac1 and GPCR ligands. The requirement for a specialized membrane domain for PI3Kβ signaling appears to be a common theme in the MEF and breast cancer cell studies.

Why does activation of PI3Kβ in cells require coincident signaling and/or specific localization? For activation downstream from GPCRs, this may be due to the weak binding affinity of the monovalent Gβγ–PI3Kβ binding interaction, as compared with the multivalent binding of Gβγ to both p110γ and its p101 regulatory subunit. In this case, activation of PI3Kβ might be enhanced by diffusion barriers that locally elevate Gβγ concentrations and lead to more effective coupling to PI3Kβ. Alternatively, the dramatic synergistic activation of PI3Kβ by tyrosine phosphopeptides and Gβγin vitro may also occur in cells, with coincident activation leading to greatly increased activity. Teleologically, regulation of PI3Kβ by coincident activation would provide additional signaling specificity, leading to PIP₃ production under conditions that would not preferentially activate other class I PI3Ks.

Activation of PI3Kβ in cells: comparisons with PI3Kα

There has been disagreement as to the relative activity of the two ubiquitously expressed class IA PI3Ks, PI3Kα and PI3Kβ (72). Based on a DXMS study that detected contacts between the cSH2 domain of p85 and the kinase domain of p110β but not p110α, it was proposed that the basal activity of PI3Kβ is inhibited relative to that of PI3Kα (18). The low basal activity of PI3Kβ was also used to explain the enhanced ductal branching and tumorigenesis caused by deletion of p110β in mouse models of breast cancer (73); loss of p110β was proposed to increase the pool of p85 available to stabilize the more active p110α, which was observed as an increase in PI3K activity in polyoma middle T antigen or HER3 immunoprecipitates (in MMTV-PYMT and MMTV-NeuT mice, respectively).

In contrast to studies suggesting that PI3Kβ has relatively low activity as compared with PI3Kα, overexpression of p110β, p110δ, or p110γ in chick embryo fibroblasts leads to transformation, whereas wild-type p110α does not (74). This finding provides an explanation for the fact that p110α is frequently mutated in human cancers, whereas mutation of the other isoforms is rare (75). The enhanced transforming capacity of p110β relative to p110α may be due in part to the absence of an inhibitory contact between the C2 domain of p110β and the iSH2 coiled coil domain of p85; this contact is present in p110α, and p110α mutants that mimic the sequence of p110β in this region are transforming (76).

How can these two views on the relative activity of PI3Kα and PI3Kβ be reconciled? In vitro studies on the relative basal specific activities of PI3Kα and PI3Kβ, and their Michaelis constant values for substrate, are not consistent [discussed in Salamon and Backer (77)]. An interesting complication is that in cells expressing similar levels of PI3Kα and PI3Kβ, the binding of PI3Kβ to tyrosine-phosphorylated proteins is reduced relative to PI3Kα [e.g., see Knight et al. (6)]. This may be due to the binding of PI3Kβ to Rab5 or Gβγ, which may limit its accessibility to RTKs and their substrates. Thus, the enhancement of Akt activation in mouse models of breast cancer upon knockdown of p110β (73) may not be due to a difference in PI3Kβ vs PI3Kα activity, but rather a difference in their ability to associate with the oncogenic drivers of PYMT- and HER2/Neu-dependent tumors, which are tyrosine-phosphorylated proteins.

A second complication is that in vitro measurements of lipid kinase activity do not take into account the activity that occurs in response to multiple, potentially synergistic activators of PI3Ks. For example, activation of PI3Kα and PI3Kβ by tyrosine-phosphorylated peptides (pY) is similar, in the range of 3- to 5-fold up to 20-fold, depending on the publication (36, 38). PI3Kβ activation by Gβγ is similar to that seen with pY. However, Gβγ in combination with pY leads to nearly 50- to 125-fold activation of PI3Kβ (36, 38). Thus, the net signaling output (i.e., the levels of PIP₃) from synergistic activation of PI3Kβ by GPCRs and RTKs could be higher than for RTK activation of PI3Kα. Consistent with this model, PIP₃ in neutrophils stimulated with C5a and CSF-1 is 2.5-fold greater that in cells stimulated with either ligand alone. This increase is abolished by loss of PI3Kβ activity (7).

PI3Kβ inhibitors

Isoform-selective inhibitors against the PI3Ks have taken advantage of variations within the ATP binding site of these kinases. Unlike the early pan-PI3K inhibitor wortmannin, which covalently modifies the active site of PI3Ks (78), the isoform-selective inhibitors act in an ATP-competitive manner. The PI3Kβ-selective inhibitor TGX-221 has been a mainstay of cell-based experiments on this enzyme since is description in 2005 (14). The PI3Kβ-selective inhibitor KIN-195 was shown to preferentially block the growth of PTEN-null cells from a panel of 422 human tumor cell lines (10). More recent PI3Kβ-selective inhibitors include SAR260301, which synergizes with BRAF and MEK inhibitors to inhibit the growth of PTEN-deficient/mutant BRAF human melanoma cells (79), but which is cleared too rapidly in humans to reach clinically useful levels of inhibition (80); GSK2636771, which blocks DU145 prostate cancer cell invasion and bone metastasis in mice (81), and is currently in clinical trials for treatment of PTEN-deficient solid tumors (82); BL140, which blocks the growth of prostate tumor cell lines that are resistant to androgen receptor inhibition (83); AZD6482, which has mild antiplatelet activity in humans (84); and (P)28, which shows enhanced metabolic stability (85). A second strategy has been the development of combined PI3Kβ/PI3Kδ inhibitors, based on the relatively high conservation between these two isoforms. New dual-selectivity compounds with efficacy in PTEN-deficient LNCaP and PC3 prostate tumor xenografts have been described (86, 87).

Cell Biological Functions of PI3Kβ

Canonical downstream signaling by PI3Kβ

The class I PI3Ks vary in their regulation and targeting, but the core mechanisms of their downstream signaling activity are shared (Fig. 4). Class I enzymes predominantly produce PIP₃ in cells. PIP₃ signals both directly through the recruitment of proteins containing PIP₃ binding domains [Pleckstrin homology (PH) and Phox homology domains] or indirectly through dephosphorylation to phosphatidylinositol 3,4-bisphosphate and phosphatidylinositol 3-phosphate, which bind to distinct effectors (88, 89). A major effector of class I PI3Ks is the Akt protein kinase, which regulates a remarkably wide network of cellular activities, including survival, metabolism, protein and lipid biosynthesis, and proliferation (90). Akt-independent pathways downstream from production of PIP₃ include activation of Tec family tyrosine kinases (91) and the guanine nucleotide exchange factors for small GTPases in the Rho and Arf families (92, 93), both of which contain PH domains and are regulated by PIP₃. The canonical PI3K signaling pathways have been reviewed extensively (94–96) and are not discussed here. We instead focus on aspects of PI3K signaling that are unique to PI3Kβ.

Figure 4.

Canonical and isoform-specific signaling by PI3Kβ. All class IA PI3Ks produce PIP₃, which binds to PH domains in Akt, GEFs for Rho-family GTPases, and the Tec tyrosine kinases, leading to a complex array of downstream signals. Additionally, PI3Kβ has been specifically implicated in vesicular trafficking in the endocytic and autophagic systems. PI3Kβ plays a role in a variety of nuclear functions, including DNA repair and replication and nuclear envelope maintenance, as well as chromosome segregation during mitosis.

Vesicular trafficking and macroautophagy

PI3Kβ binding to the endosomal GTPase Rab5 suggests a role in vesicular trafficking in the endocytic system (Fig. 4) (41). Two studies have suggested a kinase-independent role for PI3Kβ during receptor-mediated endocytosis: transferrin and EGF uptake was abolished in p110β knockout MEFs but restored in MEFs expressing kinase-dead p110β (35, 97). A caveat to these studies is that endocytosis was measured using fluorescent ligands at concentrations that greatly exceeded the dissociation constant for receptor binding, making it difficult to distinguish receptor-mediated vs fluid phase uptake. However, p110β knockdown but not treatment with TGX211 inhibited ligand-induced IGF-1 receptor downregulation, consistent with a kinase-independent role for PI3Kβ in endocytic trafficking (98).

PI3Kβ also plays a poorly understood role in macroautophagy induced by serum deprivation: macroautophagy is inhibited in p110β KO MEFs and is rescued by expression of kinase-dead p110β but not a mutant deficient for binding to Rab5 (99). Rab5 has been suggested to be a positive regulator of macroautophagic clearance of Huntington aggregates in flies (100). Rab5 activation levels were reduced in p110β knockout MEFs, and expression of an activated Rab5 restored macroautophagy (99).

The authors proposed that p110β maintains Rab5 activity by sequestering it from Rab5 GTPase-activating proteins (GAPs), including p85, which was previously suggested to possess Rab5, Cdc42, and Rac-1 GAP activity (43). This model is problematic for several reasons. First, overexpression of activated Rab5 inhibits mTORC1 (101), which could enhance autophagy independently of PI3Kβ. Second, mass spectrometry studies of fibroblasts show that Rab5 is present in a 150-fold excess over p110β (102). Given these data, p110β is unlikely to have an appreciable effect on global Rab5 activity levels through stoichiometric binding to Rab5. Finally, it is unlikely that p85 is a Rab5 GAP. Studies claiming to show p85 GAP activity toward Rab5 and Rac1 used a 50-fold molar excess of p85 over the GTPase (99); this reverses the usual relationship between GAP and substrate. Under standard GAP activity assay conditions in which the GTPase is in excess, p85α has no detectable GAP activity toward Rac1 (103) or Cdc42 (42, 104). Another study also could not detect p85α GAP activity toward Rho family GTPases, including Rac1 and Cdc42, even when p85 was present in 50-fold excess. An analysis of conserved residues in known Rho GAPs suggested that p85 was a GAP-like protein without GAP activity (105). Given these data, the mechanism by which p110β knockout regulates Rab5 activity, and whether this is related to its effects on autophagy, remain important unanswered questions.

Finally, PI3Kβ plays a kinase-independent role in glucose-stimulated insulin secretion. Studies in insulinoma cells and human islets show decreased exocytosis after treatment with a short hairpin RNA targeting p110β, but not after treatment with the PI3Kβ inhibitor TGX-211 (106). Interesting, knockdown of p110α has opposite effects, leading to increased exocytosis and insulin secretion.

A unifying feature of all of these studies is that PI3Kβ seems to function in a kinase-independent manner in the context of vesicular trafficking. This is consistent with the observation that p110β knockout produces embryonic lethality (8), whereas a kinase-dead knock-in is viable (35). This suggests that p110β has independent scaffolding functions mediated by its interactions with other proteins, including Rab5, Rac, Gβγ, and possibly the wide range of proteins that interact with the p85 regulatory subunit. Interestingly, the other Gβγ-regulated PI3K, PI3Kγ, also has well-documented kinase-independent functions (107). It is intriguing to speculate that kinase-independent signaling and Gβγ binding are somehow related.

Nuclear functions and mitosis

A specific role for PI3Kβ in proliferation was first demonstrated by the microinjection of PI3Kβ-specific inhibitory antibodies, which block DNA synthesis in fibroblasts in response to insulin and LPA but not PDGF (108). A subsequent study showed inhibition of DNA synthesis by PI3Kβ-specific antibody microinjection in a number adenocarcinoma cell lines. However, cell survival was unaffected by inhibition of PI3Kβ but was reduced by inhibition of PI3Kα (109). Consistent with distinct roles for PI3Kα and PI3Kβ in proliferation, cell cycle analysis of PI3K activity in synchronized cells showed that PI3Kα is activated at G₀/G­₁ and during G₁. In contrast, PI3Kβ activity is maximal during late G₁ (110).

Despite its interactions with cytosolic regulators such as Gβγ and Rab5, PI3Kβ can also be detected in the nucleus (Fig. 4) and has a nuclear localization signal in its C2 domain (111). Surprisingly, nuclear translocation of overexpressed PI3Kβ is only seen when coexpressed with the p85β regulatory subunit, suggesting that differences between p85α and p85β regulate accessibility of the nuclear localization signal; the structural basis for this finding is not clear. Nuclear PI3Kβ has been implicated in a variety of functions, including resistance to apoptosis (111), regulation of DNA replication by lipid kinase-dependent activation of Akt, which phosphorylates proliferating cell nuclear antigen (PCNA), and kinase-independent loading of PCNA onto chromatin (112). PI3Kβ also binds to the PCNA regulatory replication factor C (113) and is required for maintenance of the nuclear envelope and assembly of nuclear pore complexes (114). Finally, both PI3Kα and PI3Kβ have been implicated in mitosis at early and late stages, respectively (115).

PI3Kβ in Cancer

PTEN-deficient tumors

PI3Kβ has been implicated in a variety of cancer types [Table 1 (35, 54, 116–130)], in some cases due to amplification or overexpression (117, 120, 127, 131). However, PI3Kβ has been most strongly linked to tumorigenesis involving deletion or mutation of the PTEN tumor suppressor (Fig. 5). An initial study compared the effect of p110α vs p110β knockdown on the growth of PTEN-replete or PTEN-null cell lines. Whereas knockdown of p110α inhibited the growth of cell lines expressing mutant PIK3CA, knockdown of p110β but not p110α inhibited the growth of PTEN-deficient lines (9). Subsequent studies confirmed a specific requirement for PI3Kβ in the growth of PTEN-deficient glioma, glioblastoma, prostate, breast, and endometrial cancer cell lines (132–137). Deletion or inhibition of PI3Kβ suppresses prostate cancer in mice lacking PTEN in the anterior prostate, as well as leukemia in mice lacking PTEN in hematopoietic stem cells (67, 97). Studies using PI3Kβ-selective inhibitors similarly showed enhanced sensitivity to PI3Kβ inhibition in PTEN-deficient tumor cell lines (10, 138, 139). Combined inhibition of PI3Kα and PI3Kβ was required for tumor regression in a PTEN-deficient estrogen receptor–positive breast cancer xenograft model (140). The requirement for PI3Kβ is not observed in all PTEN loss phenotypes, as development of prostate cancer but not glomerulonephritis, pheochromocytoma, or thyroid cancer was inhibited in mice with a heterozygous p110β kinase-dead knock-in (141).

Table 1.

PI3Kβ in Cancer

Tissue	Role of PI3Kβ	References
Breast	Expression correlates with decreased survival in humans	Carvalho et al. (116)
	Overexpression correlates with increased proliferation in ER-positive breast cancer cells	Crowder et al. (117)
	KO reduces tumors in HER2/Neu mouse model	Ciraolo et al. (35)
	Enhances tumorigenesis via YAP/TAZ signaling	Zhao et al. (118)
Endometrial	Knockdown inhibits proliferation	An et al. (119)
Ovarian	Knockdown increases sensitivity to paclitaxel	Jeong et al. (120)
Colorectal	Knockdown inhibits proliferation	Liu et al. (121)
	Upregulated in colorectal cancer cells treated with radiation	Yu et al. (122)
	Knockdown increases sensitivity to oxaliplatin in colon cancer cells	Liu et al. (121)
	Both PI3Kα and PI3Kβ contribute to colorectal cancer	Wen et al. (123) and Wu et al. (124, 125)
Gastric	Expression correlates with shorter survival and poor prognosis	Kim et al. (126)
Prostate	Mutated in advanced prostate cancer	Robinson et al. (54)
	Required for androgen receptor signaling; overexpression causes androgen-independent signaling	Zhu et al. (127)
Glioblastoma	Expression linked to tumor recurrence and poor prognosis	Pridham et al. (128) and Varghese et al. (129)
	PI3Kβ and c-Jun N-terminal kinase inhibition blocks U87-MG tumor growth	Zhao et al. (130)

Abbreviations: ER, estrogen receptor; KO, knockout.

Figure 5.

PI3Kβ in cancer. PI3Kβ is amplified and activated by mutations in some tumors. It plays a prominent role in the proliferation of tumors that are deficient for the PTEN tumor suppressor. Additionally, PI3Kβ is critical for invasion and metastasis in breast cancer cells, as well as in the development of resistance to PI3Kα inhibitors.

The situation is more complicated in cells lines in which PTEN loss is coupled with activation of an RTK pathway. Development of endometrioid adenocarcinoma due to loss of PTEN and expression of activated KRas was sensitive to loss of PI3Kα but not PI3Kβ, whereas growth of ovarian surface epithelial cells deficient for PTEN required PI3Kβ and p53. However, these cells became PI3Kα-dependent upon expression of activated KRas (142). Similarly, PTEN-deficient cell lines from endometrioid endometrial cancers, which frequently harbor additional mutations in KRas, were resistant to the PI3Kβ inhibitors GSK2636771 and AZD6482 but sensitive to combinations that included the PI3Kα inhibitor A66 (143). In a mouse model of HER2-positive, PTEN-deficient breast cancer, deletion of p110α but not p110β impaired tumor development. However, although inhibition of PI3Kα and HER2 inhibited the growth of HER2-positive, PTEN-deficient xenografts, addition of a PI3Kβ inhibitor was required for tumor regression (144). A similar result was seen in a panel of PTEN-deficient breast, prostate, and renal tumors models, where combined PI3Kβ and mTOR inhibitors reduced tumor growth (145). These studies make the case for simultaneously targeting multiple PI3K isoforms, or using PI3K inhibitors in conjunction with oncogenic kinase inhibitors.

The mechanism that defines the differential requirement for PI3Kβ vs other isoforms in PTEN-deficient cells and tumors is not clear. The distinction does not appear to be due to tissue-specific factors, as the PI3Kβ dependence seen in prostate cancer driven by PTEN loss (97) is not seen in prostate cancer driven by expression of the polyoma middle T antigen (146). One possibility is that GPCR inputs, which regulate PI3Kβ but not PI3Kα or PI3Kδ, are particularly important in the setting of PTEN loss (97). This could occur due to feedback inhibition of RTK-driven PI3K signaling, through mTORC1-mediated inhibitory phosphorylation of IRS-1 or GRB-10 (147–150).

Alternatively, studies in prostate cancer cells and myotubes suggest that PI3Kβ regulates basal levels of PIP₃, but not acute changes in PIP₃ in response to RTK stimulation (6, 133). In myotubes, PI3Kβ inhibitors had no effect on insulin-stimulated Akt phosphorylation but increased sensitivity to inhibition of PI3Kα (6). Furthermore, basal PIP₃ levels were acutely increased by inhibition of PTEN, and the increase was reversed by inhibition of PI3Kβ (6). Based on these studies, the effect of p110β knockout on prostate cancer driven by PTEN loss was suggested to be due to increased basal PIP₃ levels (97). A third alternative is suggested by a study comparing the effects of PI3Kβ inhibition vs p110β knockdown in bladder carcinoma xenografts in mice; knockdown was more effective at causing regression, suggesting kinase-independent functions for PI3Kβ in the context of PTEN loss. Finally, dependence on PI3Kβ in PTEN-null tumor cells was linked to reductions in E-cadherin and increases in N-cadherin (151), suggesting a potential link to the endothelial–mesenchymal transition.

Metastasis, resistance, and immune surveillance

In addition to its role in tumorigenesis, PI3Kβ plays an independent role in tumor metastasis (Fig. 5). Replacement of wild-type p110β with a Gβγ-unresponsive mutant in human triple-negative breast cancer lines had no effect on proliferation or orthotopic tumor growth in SCID mice (11). However, in vitro tumor cell invasion and experimental metastasis were both markedly inhibited. The loss of invasive capacity correlated with defects in the formation of invadopodia, actin-based protrusions that are critical for matrix metalloprotease secretion and matrix degradation by tumor cells.

A recent study has identified PI3Kβ as important in the resistance of tumor cells to treatment with inhibitors of PI3Kα (Fig. 5) (152). In a large panel of human breast cancer cell lines, cells treated with the PI3Kα-specific inhibitor BYL-719 showed an initial reduction in PIP₃ levels, which returned to normal after 24 hours. The recovery of PIP₃ levels was dependent on PI3Kβ; in HER2-overexpressing cell lines treated with BYL-719, PI3Kβ activation was mediated by recruitment to HER2/ErbB3 dimers, rather than through GPCR-mediated mechanisms. In the HER2/Neu mammary tumor model, treatment of animals with inhibitors of both PI3Kα and PI3Kβ led to tumor shrinkage, whereas treatment with either inhibitor alone only reduced the rate of tumor growth. A similar result was obtained with BYL-719–resistant breast cancer cells, but through coupling of PI3Kβ to the IGF-1 receptor: inhibition of the IGF-1 receptor or PI3Kβ restored sensitivity to the PIKα inhibitor (153). In mice expressing the HER2/Neu oncogene in the mammary epithelium, concurrent deletion of p110α delayed tumorigenesis, but tumors that did arise showed a loss of PTEN and a reliance on PI3Kβ in 23% of cases (154). These studies suggest that inhibition of PI3Kβ may be efficacious in overcoming resistance to PI3Kα inhibitors.

Interestingly, knockdown of p110β in several hematopoietic cell lines [IM9 (B-lymphoblast), K562 (erythroleukemia), and U937 (lymphoma)] enhances their susceptibility to killing by natural killer cells. This increase correlates with increased tumor cell expression of natural killer cell stimulatory receptors in the NKG2D, DNAM-1, or 2B4 families, and is reversed by blocking antibodies to these receptors (155). This response would appear to dovetail with the decrease in immune suppressor functions caused by loss of PI3Kδ (156), suggesting that combined PI3Kβ/PI3Kδ inhibition might lead to increased efficacy against some tumor types.

PI3Kβ Signaling at the Organismic Level

PI3Kβ knockout or inhibition disrupts normal physiological functions, as well as some pathological phenotypes, in a variety of tissues [Table 2 (8, 35, 97, 157–172)]. We focus on a few examples in which the role of PI3Kβ has been explored in some depth.

Table 2.

PI3Kβ in Physiology and Pathophysiology

Tissue	Role of PI3Kβ	References
Whole animal	Global KO, early embryonic lethality	Bi et al. (8)
	Global KD knock-in: peripheral insulin resistance, β-cell hyperplasia, decreased duration of hepatic insulin signaling	Ciraolo et al. (35) and Jia et al. (97)
Hypothalamus	KO in anorexogenic POMC neurons led to central leptin resistance, increased adiposity, and diet-induced obesity	Al-Qassab et al. (157)
	KO in orexogenic ARGP neurons led to resistance to diet-induced obesity	Al-Qassab et al. (157)
	PIK3CA and PIK3CB KO in AGRP neurons blocked leptin and insulin signaling, leading to weight gain and impaired glucose metabolism	Huang et al. (158)
	Intraventicular PI3Kα/β inhibition did not block leptin-induced anorexia	Tups et al. (159)
	KO of PIK3CA but not PIK3CB in POMC neurons blocked leptin-stimulated sympathetic outputs	Bell et al. (160)
	PIK3CA/PIK3CB double KO, but not PIK3CA KO, inhibited LPS-stimulated hypophagia	Borges et al. (161)
Adipocytes	PI3Kβ inhibition blocked augmentation of insulin signaling by gastric inhibitory peptide	Mohammad et al. (162)
Muscle	Increased p110β expression in troglitazone-treated patients with diabetes	Kim et al. (163)
	Skeletal muscle KO caused late-onset glucose intolerance	Matheny et al. (164)
	Knockdown in C2C12 myoblasts enhanced expression of myogenic makers and also enhanced IGF-1 signaling	Matheny et al. (165)
	PI3Kβ inhibitors delayed C2C12 differentiation	Matheny et al. (164)
	Skeletal muscle KO reduced quadriceps mass and muscle strength	Matheny et al. (164)
	Knockdown (but not inhibition) increased IGF-1 and AMPK signaling in C2C12	Matheny et al. (98) Yu et al. (166)
Brain	Increased p110β mRNA and protein levels in cortical neurons from FMRP KO mice (fragile X model)	Ascano et al. (167), Gross et al. (168), and Sharma et al. (169)
	Increased colocalization of PI3Kβ with the synaptic marker synaptophysin in FMRP KO neurons	Gross et al. (168)
	p110β Upregulated in cells from fragile X patients; PI3Kβ inhibition blocked constitutively elevated protein synthesis	Gross et al. (170) and Kumari et al. (171)
	Viral p110β knockdown in prefrontal cortex of FMRP KO mice, or heterozygous deletion of PIK3CB in FRMP−/− with mice, led to improved cognition	Gross et al. (172)

Abbreviations: KD, kinase dead; KO, knockout.

Immune system

PI3Kβ plays important roles in a number of distinct immune system cell types. Although these cells express the RTK-regulated PI3Ks (PI3Kα and PI3Kδ) as well as the GPCR-regulated PI3Ks (PI3Kβ and PI3Kγ), it is clear that the distinct PI3Ks play nonredundant roles, particularly in macrophages and neutrophils. No PI3K isoform selectivity was seen with regard to activation/differentiation, apoptosis, or cytokine secretion in T-cells stimulated with anti-CD3 and anti-CD28 antibodies (173), and both PI3Kβ and PI3Kδ are involved in calcium responses caused by stimulation of FcεRI receptors in mast cells (174).

In macrophages, published reports do not present a consistent picture. Microinjection of inhibitory antibodies to p110β blocked uptake of apoptotic cells and FcRγ-mediated phagocytosis in primary mouse macrophages (175). In contrast, short hairpin RNA knockdown of p110β did not affect either phagocytosis or pinocytosis in the Raw 264.7 macrophage cell line (176); this discrepancy could reflect either the difference between primary cells vs cell lines or the inhibitory methods used. Differences between primary macrophages and Raw 264.7 cells were also observed during LPS-stimulated M1 polarization of macrophages; knockdown of p110β inhibited LPS-stimulated IL-12 production in monocyte-derived macrophages and dendritic cells (177), whereas it enhanced LPS-stimulated IL-12 expression and nitrous oxide synthase activity in Raw 264.7 cells (178). Knockdown of p110β in Raw 263.7 cells also blocked the induction of TNFα by ceramic and titanium particles, which mimic the stimulation of osteolysis by the deterioration of surgically revised joints (179). In osteoclasts, which are derived from the same myeloid precursors as macrophages, knockout of PI3Kβ did not affect development but was required for the actin-rich sealing rings that mediate bone resorption, leading to increased trabecular bone density (16).

In neutrophils, activation of surface receptors by microbes or foreign bodes stimulates production of reactive oxygen species by the NAD phosphate oxidase complex in a PI3Kβ-dependent manner. As discussed above in the section on coincidence detection, PI3Kβ is required for ROS production in neutrophils stimulated simultaneously with IgG–antigen complexes (which bind to Fcγ receptors) and leukotriene B₄ (which binds to BLT1 receptors) (15). Responses to fungal hyphae also required PI3Kβ, along with PI3Kγ, in this case downstream of β₂-linked integrins (180). Whereas ROS production in response to the GPCR ligand N-formyl-Met-Leu-Phe was largely due to PI3Kγ (181), PI3Kβ was required for ROS production in N-formyl-Met-Leu-Phe–stimulated p110γ^−/− neutrophils (182). G protein–mediated activation of PI3Kβ was also required for ROS production in response to integrin stimulation of neutrophils (7).

Platelets

Platelet adhesion to sites of endothelial damage is mediated by both adhesion receptors and integrins [the collagen receptors glycoprotein VI (GPVI) and α₂β₁ integrin, as well as α_IIbβ₃ integrin, which binds to von Willebrand factor, fibrinogen, and fibronectin (183, 184)]. Subsequent platelet activation involves signaling by both integrins and GPCRs (thrombin and purinergic receptors) (185). Given the role of PI3Kβ in both GPCR and integrin signaling (7), it is not surprising that PI3Kβ is a key signaling molecule in platelets (Fig. 6).

Figure 6.

PI3Kβ signaling in platelets. Platelet adhesion and activation is mediated through a variety of cell surface receptors and integrins. PI3Kβ acts downstream of outside-in signaling by integrins that bind to collagen, laminin, fibronectin, vitronectin, and von Willebrand factor (vWF), as well as GPCRs activated by ADP or thrombin. PI3Kβ is also required for inside-out signaling, which increases the binding affinity of cell surface integrins for the extracellular matrix through the activation of the small GTPase Rap1b. Pharmacological inhibition of PI3Kβ inhibits platelet adhesion and thrombus formation, and it also inhibits thrombus stabilization.

PI3Kβ was first associated with platelet function in a study that also described the first PI3Kβ-specific inhibitor, TGX-221 (14). Inhibition of PI3Kβ activity blocked thrombus formation in mice, platelet adhesion to von Willebrand factor–coated surfaces (presumably via α_IIbβ₃ integrins), ADP-stimulated activation of Rap1b, as well as calcium signaling and α_IIbβ₃ integrin activation initiated by shear stress. Subsequent studies implicated PI3Kβ as important for signaling by numerous platelet activators: inhibition or knockout of p110β disrupted platelet activation downstream from both the GPVI and α₂β₁ collagen receptors (186–189), as well as from the thrombin and P2Y12 purinergic receptors. A recent mass spectrometry study showed that PI3Kβ was the major isoform responsible for PIP₃ production in response to thrombin or collagen-related peptide stimulation (190). In the apo3^−/− mouse model of atherosclerotic disease, PI3Kβ was required for stimulation of thrombus formation by the TNF family receptor CD40 (191). Although not normally required for platelet priming by IGF-1, PI3Kβ was required for platelet priming in PIK3CA knockout platelets (192). Both PI3Kα and PI3Kβ had a role in thrombosis due to autoimmune antiphospholipid antibodies (193), but PI3Kα was a negative regulator of responses to thrombopoietin (194).

The precise role of PI3Kβ in integrin signaling in platelets is complex (Fig. 6). Several groups have reported that PI3Kβ is activated downstream from both α_IIbβ₃ integrins (195) and α₂β₁ integrins, the latter by a calcium/Pyk2-mediated pathway (196, 197). In contrast, PI3Kβ activation by GPVI is upstream rather than downstream from calcium/Pyk2 (197). However, PI3Kβ (in some cases, along with PI3Kα) has also been implicated in inside-out signaling by Rap1b, leading to activation of α_IIbβ₃ integrins in response to ADP or collagen (14, 186, 187, 196, 198). Thus, PI3Kβ seems to be involved in feed-forward activation of integrin-mediated adhesion by both inside-out and outside-in mechanisms.

PI3Kβ is also critical for thrombus stabilization, particularly under conditions of high shear stress (14, 199). However, initial thrombus formation was not affected, such that pharmacological inhibition of PI3Kβ did not prolong bleeding time (14, 200). This suggests that inhibition of PI3Kβ may be a useful antithrombotic therapy; preliminary studies show efficacy in humans (84, 200), and in reducing clotting complications during extracorporeal circulation (201).

Spermatogenesis and fertility

Homozygous knock-in of kinase-dead p110β leads to a complete loss of male fertility as well as a decrease in female fertility (12, 13). In male mice expressing kinase-dead p110β, there was testicular atrophy with loss of spermatogenic cells but not Sertoli cells or Leydig cells; testosterone was normal and FSH was increased, suggesting normal upstream endocrine function. Specific inactivation of PI3Kβ in Sertoli cells also caused sterility, but it did not lead to a loss of germ cells (13). PI3Kβ seems to act in part as a regulator of androgen receptor–mediated transcription in Sertoli cells (13), as well as in LnCAP prostate cancer cells (127). Consistent with this hypothesis, inhibition of PI3Kβ blocked androgen-stimulated increases in transcriptional activation, as measured by H3K4me2 histone methylation (202). PI3Kβ may also act by signaling downstream from the RTK c-Kit (12), although there is disagreement as to whether PI3Kβ rather than PI3Kα is the predominant PI3K recruited to activated c-Kit receptors in the testes (13). In female mice, p110β expression and Cdc42-PI3Kβ coimmunoprecipitation correlated with increased activation of primordial follicles in cultured mouse ovaries (203).

Conclusions and Unanswered Questions

As the clinical evaluation of PI3Kβ and dual PI3Kβ/PI3Kδ selective inhibitors proceeds, we will obtain more information about the utility of therapies targeting this interesting PI3K isoform. Regardless of its potential for clinical intervention, the enzymology and cell biology of PI3Kβ offer rich areas of investigation. Although nuclear localization of the p85β/p110β heterodimer has been documented (111), we know very little about the localization of PI3Kβ in the cytosol. Its ability to bind to activated RTKs and to Gβγ would suggest acute recruitment to the plasma membrane, but this has not been demonstrated by microscopy [although localization to lipid rafts has been demonstrated biochemically (68)]. Similarly, PI3Kβ binding to Rab5 suggests localization to early endosomes, and PI3Kβ–Rab5 complexes were initially detected in clathrin-coated vesicle preparations (41). However, the distribution of PI3Kβ between distinct cellular compartments, as well as the dynamics and regulation of this distribution, is not known. The kinase-independent functions of PI3Kβ in endocytic trafficking are also intriguing, and hopefully experiments with mutants that selectively disrupt PI3Kβ interactions with known binding partners (Rac1, Cdc42, Rab5, Gβγ, RTKs) will lead to mechanistic insights into the scaffolding functions of PI3Kβ.

The regulation of PI3Kβ activity is also not fully understood. A detailed biochemical analysis of how PI3Kβ is activated by different combinations of binding partners (Rac1, Cdc42, Rab5, Gβγ, RTKs) is needed. A major complication is that all of these interactors are integral or peripheral membrane proteins, and the substrates of PI3Kβ are of course lipids. Sorting out the contributions of these interactions to membrane recruitment vs allosteric activation of PI3Kβ lipid kinase activity may require the use of soluble substrates, such as short-chain phosphatidylinositol 4,5-bisphosphate, or single-molecule methods such as those recently used to define the mechanism of PI3Kα activation by Ras (51). We also note that although the RBD of p110β binds to Rac1 and Cdc42, it also binds to the poorly understood G proteins DIRAS1 and 2 (39); we have no idea how this interaction regulates PI3Kβ in cells.

The role of PI3Kβ in human cancer is only superficially understood and, in particular, the mechanism for its importance in the proliferation of PTEN-null tumors is still unclear. Work in this area, as well as studies addressing its role in resistance to inhibitors of PI3Kα and other PI3K isoforms, is likely to be clinically important. Of particular interest is the mechanistically obscure activity of PI3Kβ in androgen receptor signaling, which could be important in understanding its role in prostate cancer. Additionally, the role of PI3Kβ in metastatic disease, both in tumor cells and in stromal cells, will likely become a key area of investigation in coming years.

Hopefully, the development of better inhibitors of PI3Kβ will identify this enzyme as a useful clinical target for other indications—treatment of fragile X syndrome, antithrombolytic therapy, perhaps even male contraception. In the meantime, we can look forward to more surprises from this highly idiosyncratic member of the PI3K family.

Glossary

Abbreviations:

ABD

adapter-binding domain

BCR

breakpoint cluster region

DXMS

deuterium exchange/mass spectrometry

GAP

GTPase-activating protein

GPCR

G protein–coupled receptor

GPVI

glycoprotein VI

MEF

mouse embryonic fibroblast

PCNA

proliferating cell nuclear antigen

PH

Pleckstrin homology

PI3K

phosphoinositide 3-kinase

PIP₃

phosphatidylinositol 3,4,5-trisphosphate

pY

tyrosine-phosphorylated peptides

RBD

Ras-binding domain

ROS

reactive oxygen species

RTK

receptor tyrosine kinase