PI3K was first discovered in the mid-1980s as an enzymatic activity that coprecipitated with activated growth factor receptors and oncoproteins. It was soon appreciated that activation of PI3K and production of 3-PIs is a signaling response common to a great variety of receptor systems. The genes encoding PI3Ks were cloned starting in the early 1990s. Identification of PI3K inhibitors began at this time and continues to the present day, with a diverse set of potent and selective compounds produced by pharmaceutical companies worldwide (Rodon et al. 2013; Fruman and Rommel 2014). The reason for this intense activity is that hyperactive PI3K signaling is now recognized as a driving force in human cancer, inflammatory diseases, metabolic disorders, and in many other clinical conditions (Fruman and Rommel 2014; Hawkins and Stephens 2015). PI3K genes are bona fide oncogenes in that gain-of-function mutations are found in a large fraction of human cancers (Thorpe et al. 2015). Some of the most common tumor suppressor genes act by opposing PI3K signaling (Song et al. 2012). Gene targeting studies have confirmed the central role of PI3K enzymes and regulatory subunits in cell proliferation and in a diverse set of physiological functions (Vanhaesebroeck et al. 2010; Thorpe et al. 2015; Hawkins and Stephens 2015).
PI3K Enzymes and Products
Structural studies have provided considerable insight into the domain topology of class I PI3Ks, their mode of regulation by regulatory subunits and Ras, and their interaction with inhibitor compounds (Vadas et al. 2011). The p110γ isoform was the first to be crystallized, followed more recently by p110α, p110δ, and p110β. The studies of p110α have provided insight into the mechanisms by which oncogenic mutations in the genes encoding p110α (PIK3CA) and p85α (PIK3R1) cause elevated activity of p85α/p110α dimers.
There are three class II enzymes, each containing a C2 domain and termed PI3K-C2α, PI3K-C2β, and PI3K-C2γ. There is a single class III PI3K that is homologous to the yeast PI3K enzyme Vps34. Class III enzymes associate with the protein Vps15. Class II and III enzymes act on PtdIns to produce PtdIns-3-P, though the class II PI3Ks might also convert PtdIns-4-P to PtdIns-3,4-P2. It is clear that regulated production of PtdIns-3-P is essential for vesicle trafficking and for the initiation of autophagy. Compared to the class I PI3Ks, less is known about the function of class II and class III PI3Ks (Hawkins and Stephens 2016; Backer 2016; Vanhaesebroeck et al. 2010; Falasca and Maffucci 2012). However, recent gene targeting studies in mice have defined novel functions for individual class II isoforms (Braccini et al. 2015). Conditional knockouts as well as novel selective inhibitors of class III PI3K have helped clarify the roles of this enzyme.
PH domains selective for PIP3, or PtdIns-3,4-P2, are present in a number of important signaling enzymes and small G protein modifiers (Fig. 3). PDK-1 (phosphoinositide-dependent kinase-1) is a constitutively active serine/threonine kinase whose PH domain allows the enzyme to interact with membrane-associated substrates. Most notable of these are the AKT family of serine/threonine kinases (also known as protein kinase B). The three AKT proteins (AKT1, 2, 3) possess PH domains and are phosphorylated by PDK-1-on a threonine residue in the activation loop (T308 in AKT1). Further, AKT activation is achieved by phosphorylation of a C-terminal hydrophobic motif (S473 in AKT1). In many cellular contexts, the mechanistic target of rapamycin (mTOR) complex-2 (mTORC2) is responsible for AKT-S473 phosphorylation. Active AKT phosphorylates numerous substrates that regulate cell survival, proliferation, nutrient uptake, and metabolism, with some isoform selectivity (Toker and Marmiroli 2014).
PIP3-selective PH domains are also present in most tyrosine kinases of the Tec family, including ITK and BTK that control Ca2+ mobilization and other crucial signaling events downstream of antigen receptors in lymphocytes and mast cells. PH domains in guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs) link PI3K activation to regulation of some small GTPases. One well-known example is pREX-1, a GEF for Rac and Rho family GTPases that acts downstream of p110γ in neutrophils. PI3K promotes several Rac-mediated cellular processes (Fig. 3), including NADPH oxidase activation in neutrophils, and cytoskeletal reorganization in many cell types that promotes motility and releases aldolase to increase glycolytic flux (Hu et al. 2016).
Gene targeting experiments have greatly added to the understanding of PI3K gene function in model organisms. The discussion that follows will focus on studies of class I PI3K genes in mice. For each of the catalytic subunits, null “knockout” (KO) alleles as well as kinase-dead “knock-in” (KI) alleles have been reported. p110α (Pik3ca) KO or KI results in an embryonic lethal phenotype in homozygotes. Mice heterozygous for p110αKI are viable but display severe defects in organismal metabolism, including insulin resistance, hyperglycemia, and adiposity. At the molecular level, p110α-deficient cells show reduced signaling downstream of receptors that couple to insulin receptor substrate (IRS) proteins. Together with pharmacological studies, this finding illustrates that p110α has a nonredundant function in cellular responses to insulin and insulin-like growth factor. Other studies of p110αKI mice have revealed a role for this isoform in angiogenesis. A distinct knockin strain carrying a mutation in the Ras-binding domain of p110α provided evidence that p110α is a Ras effector important for lymphangiogenesis and cellular transformation.
Mice with germline deletion of p110β (Pik3cb) die early in embryogenesis, but p110βKI mice are viable. Together with other results, this finding establishes kinase-independent functions of the p110β isoform. Kinase-dependent functions of p110β have also been identified through analysis of p110βKI mice as well as strains with conditional p110βKO in specific tissues. Key observations include: (1) p110β plays an important role in the malignant phenotype of a subset of PTEN-deficient cancer types, suggesting that p110β mediates growth factor-independent “basal” production of PIP3 that is normally hydrolyzed by PTEN; (2) p110β has a key function downstream of GPCRs including receptors for lysophosphatidic acid and chemokines, a finding supported by pharmacological approaches. This function is mediated by the direct binding of p110β to Gβγ subunits released by GPCR signaling (Dbouk et al. 2012).
Unlike p110α and p110β, p110δ and p110γ are not ubiquitously expressed and germline KO mice are viable. The most dramatic phenotypes of p110δ (Pik3cd) and p110γ (Pik3cg) KO mice are in the immune system. Notable examples include, for p110δKO and KI, impaired B cell development and function and defective T cell differentiation and trafficking. In p110γKO and KI mice, there are major defects in migration and function of innate immune cells (i.e., macrophages and dendritic cells). Combined inactivation of p110δ and p110γ produces an early block in T cell development and affects natural killer (NK) cell function. Mast cell function and allergic responses are impaired in the absence of either p110δ or p110γ.
There are three genes encoding class IA regulatory subunits: Pik3r1 encodes p85α, p55α, and p50α; Pik3r2 encodes p85β; and Pik3r3 encodes p55γ. Although inactivation of the Pik3r3 gene has not been reported, the Pik3r1 and Pik3r2 genes have been targeted in various ways and in combination. These investigations have shown complex biology of the class IA regulatory subunits. There are unique functions of individual isoforms in some contexts and redundant functions in others. The functions of class IA regulatory subunits in insulin signaling are particularly complex, in that individual deletion of p85α and p85β enhances insulin sensitivity whereas combined deletion reduces insulin sensitivity. p85α and p85β each have diverse functions in immune cells. Studies of insulin-responsive cells and lymphocytes have also provided evidence for “scaffolding” functions of p85α and p85β, independent of the associated catalytic subunit. There are two genes encoding class IB regulatory subunits (Pik3r5 encodes p101, Pik3r6 encodes p87). Knockout of Pik3r5 has shown a nonredundant role for p101 in neutrophil signaling and functional responses to inflammatory stimuli.
Gain-of-function alleles of class I catalytic subunits have been tested in transgenic models and in tissue culture. Increased p110 activity can be accomplished by overexpression of wild-type or oncogenic mutants, including naturally occurring p110α alleles (see below) and engineered mutations that target p110 proteins constitutively to the membrane. Cell culture experiments were the first to demonstrate transforming activity of p110 proteins and have shown that any of the four class I enzymes can transform fibroblasts when mutated and/or overexpressed. In vivo, activated alleles of p110α and p110β are oncogenic in mice. Analyses of mice with targeted deletion of PTEN have confirmed the tumor suppressor function of this lipid phosphatase and have shown gene dosage-dependent tumor penetrance. In addition, PTEN heterozygous mice develop lymphoproliferative disease. Similarly, mice expressing constitutively active AKT in T cells develop lymphoproliferation and thymic lymphomas. A truncation mutant of p85α (p65) that causes elevated PI3K signaling also disrupts T cell homeostasis and has oncogenic potential.
PI3K in Disease
Inflammatory conditions and autoimmune diseases are associated with increased PI3K/AKT/mTOR signaling in immune cells (Hawkins and Stephens 2015). This phenomenon is generally secondary to other environmental or genetic conditions that maintain a heightened state of activation in specific immune cells. Nevertheless, the correlation between immune cell activation and PI3K/AKT/mTOR signaling has identified this pathway as a target for drug discovery efforts for inflammatory diseases.
Several other immune-related diseases are caused by naturally occurring mutations. A homozygous premature stop codon in exon 6 of PIK3R1 results in the absence of p85α and leads to a severe defect in B cell development and agammaglobulinemia. Conversely, a gain-of-function mutation in the p110δ protein in which the glutamic acid residue 1021 is replaced by a lysine (E1021K) causes elevated PI3K signaling in patients with activated PI3Kδ syndrome (APDS). Lymphocytes from these patients have increased levels of PIP3 and phosphorylated AKT protein and are prone to activation-induced cell death. This leads to primary immunodeficiency characterized by recurrent respiratory infections, airway damage, lymphopenia, increased circulating transitional B cells, altered immunoglobulin levels in serum, and impaired vaccine responses (Angulo et al. 2013; Lucas et al. 2016). Selective p110δ inhibitors IC87114 and GS-1101 are able to reduce the mutant PI3K activity in vitro which suggests that this may be a therapeutic approach for patients.
Increased PI3K-AKT signaling can also cause noncancerous somatic mosaicism (Vanhaesebroeck et al. 2016). The first overgrowth syndrome reported caused by an activating mutation in the PIK3CA gene was congenital lipomatous overgrowth, vascular malformations, and epidermal nevi syndrome (CLOVES syndrome). Others include germline loss-of-function mutations in PTEN which causes overgrowth while somatic loss of PTEN causes the type 2 segmental Cowden syndrome. A somatic AKT1 mutation in which a glutamine is replaced by a lysine residue (p.Glu17Lys) causes Proteus syndrome, a regional overgrowth syndrome that affects bones, adipose, and other mesenchymal tissues. De novo somatic mutations in PIK3CA, AKT3, and MTOR were also identified in patients with hemimegalencephaly (HME) which is characterized by overgrowth of either one or both cerebral hemispheres.
Insulin-resistance syndromes are a growing health epidemic. By definition, insulin resistance is associated with reduced signaling responses to insulin – including dampened PI3K/AKT activation. Such signaling defects can develop through environmental causes or direct genetic mechanisms. For example, mutations in PIK3R1 are found in patients with SHORT syndrome that is associated with insulin resistance (Winnay et al. 2016), whereas PTEN mutations can increase insulin sensitivity. Agents that increase PI3K activity or block hydrolysis of PIP3 could be useful therapies for insulin-resistance.
Pharmacological strategies to target the PI3K/AKT/mTOR pathway
Pan-class I PI3K inhibition
Prevent compensation by other PI3K isoforms. Disadvantage: potential toxicity, inability to fully inhibit target at tolerated doses
Pictilisib, Taselisib, Buparlisib, Copanlisib
Isoform-selective PI3K inhibition
Less toxicity than pan-class I inhibition; may provide better target inhibition and efficacy in diseases driven by a dysregulation of a specific isoform of PI3K (Example = Idelalisib, approved for B cell malignancies). Potential for cell-extrinsic antitumor effects. Disadvantage: potential for compensation by other isoforms
Idelalisib (p110δ), Duvelisib (p110δ/γ), Alpelisib (p110α), MLN1117/TAK-117 (p110α), AZD8186 (p110β)
May provide efficacy in cancers driven by elevated AKT activity. Therapeutic window is questionable due to central role of AKT in whole-body glucose homeostasis
Allosteric mTOR inhibition
Currently approved for immunosuppression (rapamycin) and a small subset of cancers (rapalogs). Disadvantages: incomplete mTOR inhibition, loss of feedback inhibition
Sirolimus, Everolimus (RAD-001), Temsirolimus (CCI-779), Ridaforolimus (AP-23573)
ATP-competitive mTOR inhibition
Dual PI3K/mTOR (not selective for mTOR)
Multinode inhibition of the pathway can overcome feedback and crosstalk. Disadvantage: toxicity
NVP-BEZ235 (Dactolisib), GDC-0980, XL765
mTOR-selective (TOR-KI; also known as TORKi, asTORi)
Potential for strong antitumor activity with improved tolerability. Disadvantage: therapeutic window is not yet established
MLN0128/TAK-228, AZD2014, CC-223
Linkage of rapamycin with TOR-KI can overcome mutations that confer resistance to either class
RapaLink-1, 2, 3
One approach is to develop inhibitors of all class I PI3Ks. Compounds are available for both preclinical and clinical studies that are highly selective for class I enzymes (“pan-class I inhibitors”). Examples include the Genentech compounds GDC-0941 (Pictilisib) and GDC-0032 (Taselisib), the Novartis compound BKM120 (Buparlisib) and the Bayer compound BAY 80-6946 (Copanlisib) that are in clinical trials for cancer. Other companies and academic laboratories have synthesized compounds with a comparable target profile. A second approach is to develop isoform-selective PI3K inhibitors. An inhibitor of p110δ, GS-1101 (idelalisib), has demonstrated efficacy in several B cell malignancies (Yang et al. 2015). Idelalisib was approved in 2014 for relapsed/refractory chronic lymphocytic leukemia (CLL), relapsed follicular lymphoma, and relapsed small lymphocytic lymphoma. In vitro and ex vivo studies showed that idelalisib (formerly CAL-101) reduced survival of CLL primary cells in coculture with stromal cell lines and also reduced CLL migration and chemotaxis. This suggests that PI3Kδ inhibition targets malignant B cells by inhibiting their ability to respond to stromal factors for survival, proliferation, and homing signals produced by the tumor microenvironment.
Besides the success of idelalisib in certain B cell malignancies, the relative benefits of pan-class I inhibitors versus isoform-selective inhibitors in cancer have not yet been established and might vary among tumor subtypes. Inhibitors selective for p110α might be particularly effective against cancers with PIK3CA mutations while avoiding toxicities associated with pan-PI3K inhibition. Likewise, inhibitors of p110β might find utility in cancers with loss of PTEN. However, there is evidence that inhibition of p110α can result in compensatory activation of p110β and vice versa. One of the more interesting discoveries in recent years is that inhibitors of p110δ or p110γ can have antitumor activity via cell-extrinsic mechanisms that involve modulating immune cells in the tumor microenvironment (Ali et al. 2014; Kaneda et al. 2016a, b; De Henau et al. 2016).
A third approach is to inhibit AKT. Various compounds have been developed that are either ATP-competitive or act through allosteric inhibition of enzyme activity or membrane localization. An example in clinical trials is GDC-0068 (Ipatasertib). Considering that AKT inactivation has more profound effects on cellular and organismal metabolism than does PI3K inactivation, there is some concern about the therapeutic window for AKT inhibitors. Since there are three AKT isoforms with distinct functions depending on cancer context, it might be possible to achieve a better therapeutic window in some tumors using isoform-selective compounds. Another concern about both PI3K and AKT inhibitors in general is that they will not prevent mTORC1 activation through other mechanisms.
A fourth approach is to use allosteric inhibitors of mTOR, exemplified by rapamycin and its analogs RAD001 and CCI-779. Rapamycin (sirolimus) has been a clinically approved immunosuppressant since 1997 and has antiproliferative effects in many cancer cell lines. RAD001 (everolimus) and CCI-779 (temsirolimus) are approved for the treatment of advanced renal cell carcinoma. Everolimus is also approved for pancreatic neuro-endocrine tumors and in combination with exemestane for hormone-dependent breast cancer. However, rapamycin and rapalogs have mechanistic drawbacks. They do not completely inhibit mTORC1 activity and do not bind to mTORC2 or inhibit mTORC2 activity acutely in cells. Furthermore, rapamycin disables a negative feedback loop by which S6K1 dampens upstream PI3K/AKT and Ras signaling. Therefore, cancer cells treated with rapalogs generally display elevated PI3K/AKT and Ras activity, enhancing survival even when proliferation is suppressed.
A fifth approach is to use ATP-competitive mTOR inhibitors. These bind in the active site of the mTOR enzyme in both complexes and fully inhibit phosphorylation of all known mTORC1 and mTORC2 substrates. There are two classes of competitive mTOR inhibitors. One class is not mTOR-selective and inhibits PI3K enzymes (which are structurally related to mTOR) at equivalent concentrations. These are termed dual PI3K/mTOR inhibitors, exemplified by the compounds BEZ235 (Dactolisib), GDC-0980, and XL675. The other groups are highly selective mTOR kinase inhibitors (TOR-KI) that do not directly inhibit PI3K. Examples of first generation TOR-KIs are PP242, Torin-1, AZD8055, and WYE-354; clinical candidates include AZD2014, TAK-228 and CC-223. The evidence thus far indicates that in many tumor models, TOR-KIs provide anticancer efficacy that is equivalent to dual PI3K/mTOR inhibitors and better than rapamycin and rapalogs. A third generation of mTOR inhibitors was described recently, termed RapaLinks, which can overcome mTOR drug resistance mutations through chemical linkage of rapamycin to a TOR-KI moiety (Rodrik-Outmezguine et al. 2016). Further preclinical and clinical studies are needed to identify the optimal target profile that balances efficacy and tolerability in different cancer patient populations.
Although some genetic events that activate PI3K/AKT are sufficient to transform cells in vitro and promote tumorigenesis in animal models, patient-derived human cancers generally carry multiple additional genetic lesions. Thus, it is likely that inhibitors of this pathway will be most effective when combined with other targeted therapies directed to upstream or parallel pathways. Several reviews have discussed rational combinations (Fruman and Rommel 2014; Rodon et al. 2013).
PI3K enzymes generate 3-phosphorylated inositol lipids that coordinate many aspects of cellular physiology. In response to extracellular signals, class I PI3K produces the second messenger PIP3 that recruits numerous cytoplasmic proteins to the plasma membrane to propagate key signaling pathways. In general, PIP3 production promotes cell proliferation and survival and also plays a key role in cellular and organismal metabolism. Most cancer cells carry mutations that trigger constitutive elevation of PIP3 levels, and activated immune cells depend on PI3K signaling to carry out specialized functions. Gene targeting in mice has revealed unique and complex functions of individual PI3K catalytic and regulatory isoforms. Based on the role of aberrant PI3K activation in human disease, PI3K drug discovery has been a priority and will continue to produce inhibitor molecules of interest to basic and translational researchers.
- Ali K, Soond DR, Piñeiro R, Hagemann T, Pearce W, Lim EL, Bouabe H, Scudamore CL, Hancox T, Maecker H, Friedman L, Turner M, Okkenhaug K, Vanhaesebroeck B. Inactivation of PI(3)K p110δ breaks regulatory T-cell-mediated immune tolerance to cancer. Nature. 2014;510(7505):407–11.PubMedPubMedCentralCrossRefGoogle Scholar
- Angulo I, Vadas O, Garçon F, Banham-Hall E, Plagnol V, Leahy TR, Baxendale H, Coulter T, Curtis J, Wu C, Blake-Palmer K, Perisic O, Smyth D, Maes M, Fiddler C, Juss J, Cilliers D, Markelj G, Chandra A, Farmer G, Kielkowska A, Clark J, Kracker S, Debré M, Picard C, Pellier I, Jabado N, Morris JA, Barcenas-Morales G, Fischer A, Stephens L, Hawkins P, Barrett JC, Abinun M, Clatworthy M, Durandy A, Doffinger R, Chilvers ER, Cant AJ, Kumararatne D, Okkenhaug K, Williams RL, Condliffe A, Nejentsev S. Phosphoinositide 3-kinase δ gene mutation predisposes to respiratory infection and airway damage. Science. 2013;342(6160):866–71.PubMedPubMedCentralCrossRefGoogle Scholar
- Backer JM. The intricate regulation and complex functions of the Class III phosphoinositide 3-kinase Vps34. Biochem J. 2016;473(15):2251–71. PMID: 27470591.Google Scholar
- Braccini L, Ciraolo E, Campa CC, Perino A, Longo DL, Tibolla G, Pregnolato M, Cao Y, Tassone B, Damilano F, Laffargue M, Calautti E, Falasca M, Norata GD, Backer JM, Hirsch E. PI3K-C2γ is a Rab5 effector selectively controlling endosomal Akt2 activation downstream of insulin signalling. Nat Commun. 2015;6:7400.Google Scholar
- Dbouk HA, Vadas O, Shymanets A, Burke JE, Salamon RS, Khalil BD, Barrett MO, Waldo GL, Surve C, Hsueh C, Perisic O, Harteneck C, Shepherd PR, Harden TK, Smrcka AV, Taussig R, Bresnick AR, Nürnberg B, Williams RL, Backer JM. G protein-coupled receptor-mediated activation of p110β by Gβγ is required for cellular transformation and invasiveness. Sci Signal. 2012;5(253):ra89.Google Scholar
- De Henau O, Rausch M, Winkler D, Campesato LF, Liu C, Cymerman DH, Budhu S, Ghosh A, Pink M, Tchaicha J, Douglas M, Tibbitts T, Sharma S, Proctor J, Kosmider N, White K, Stern H, Soglia J, Adams J, Palombella VJ, McGovern K, Kutok JL, Wolchok JD, Merghoub T. Overcoming resistance to checkpoint blockade therapy by targeting PI3Kγ in myeloid cells. Nature. 2016;539(7629):443–7.PMID: 27828943.Google Scholar
- Hawkins PT, Stephens LR. Emerging evidence of signalling roles for PI(3,4)P2 in Class I and II PI3K-regulated pathways. Biochem Soc Trans. 2016;44(1):307–14.PMID: 26862220.Google Scholar
- Hu H, Juvekar A, Lyssiotis CA, Lien EC, Albeck JG, Oh D, Varma G, Hung YP, Ullas S, Lauring J, Seth P, Lundquist MR, Tolan DR, Grant AK, Needleman DJ, Asara JM, Cantley LC, Wulf GM. Phosphoinositide 3-kinase regulates glycolysis through mobilization of aldolase from the actin cytoskeleton. Cell. 2016;164(3):433–46.PubMedPubMedCentralCrossRefGoogle Scholar
- Kaneda MM, Cappello P, Nguyen AV, Ralainirina N, Hardamon CR, Foubert P, Schmid MC, Sun P, Mose E, Bouvet M, Lowy AM, Valasek MA, Sasik R, Novelli F, Hirsch E, Varner JA. Macrophage pI3Kγ drives pancreatic ductal adenocarcinoma progression. Cancer Discov. 2016a;6(8):870–85.Google Scholar
- Kaneda MM, Messer KS, Ralainirina N, Li H, Leem CJ, Gorjestani S, Woo G, Nguyen AV, Figueiredo CC, Foubert P, Schmid MC, Pink M, Winkler DG, Rausch M, Palombella VJ, Kutok J, McGovern K, Frazer KA, Wu X, Karin M, Sasik R, Cohen EE, Varner JA. PI3Kγ is a molecular switch that controls immune suppression. Nature. 2016b;539(7629):437–42.PMID: 27642729.Google Scholar
- Lucas CL, Chandra A, Nejentsev S, Condliffe AM, Okkenhaug K. PI3Kδ and primary immunodeficiencies. Nat Rev Immunol. 2016;16(11):702–14.PMID: 27616589.Google Scholar
- Rodrik-Outmezguine VS, Okaniwa M, Yao Z, Novotny CJ, McWhirter C, Banaji A, Won H, Wong W, Berger M, de Stanchina E, Barratt DG, Cosulich S, Klinowska T, Rosen N, Shokat KM. Overcoming mTOR resistance mutations with a new-generation mTOR inhibitor. Nature. 2016;534(7606):272–6.PubMedPubMedCentralCrossRefGoogle Scholar