Requirement of Phosphatidylinositol(3,4,5)Trisphosphate in Phosphatidylinositol 3-Kinase-Induced Oncogenic Transformation

The family of phosphatidylinositol 3-kinases (PI3K) is grouped into three distinct classes based on lipid substrate specificity and mode of regulation (1). Class I PI3Ks can use the nonphosphorylated phosphatidylinositol (PI) and the monophosphate [PI(4)P] and the bisphosphate [PI(4,5)P₂] phosphatidylinositols as substrates, giving rise to PI(3)P, PI(3,4)P₂, and PI(3,4,5)P₃, which will be referred to as PIP, PIP₂, and PIP₃, respectively. Class II PI3Ks use PI and PI(4)P as substrates to produce PIP and PIP₂, respectively. Class III PI3Ks, however, can only use PI and convert it to PIP. In addition, class I PI3Ks have the capacity to catalyze the addition of a phosphate group to serine residues in proteins (2, 3). The protein substrates include MEK1 (4), the PI3K subunits p85 and p110, and tropomyosin (5). Structural similarities also place some protein kinases close to PI3K [PI3K-related kinases; e.g., target of rapamycin (TOR)].

In the PI3K pathway, signaling molecules downstream of PI3K, such as phosphoinositide-dependent kinase (PDK-1) and Akt (murine thymoma viral oncoprotein homologue), bind PIP₂ and PIP₃ with near equal affinity (6). The binding of these signaling molecules provides a mechanism for their recruitment to the plasma membrane and allows PDK-1 to phosphorylate Akt. Phosphorylation of Akt by PDK-1 enhances the kinase activity of Akt and results in the phosphorylation of downstream targets (7). The crystal structure of the pleckstrin homology domain of Akt in complex PIP₃ reveals significant ionic interactions between basic pleckstrin homology domain residues and the phosphates at the D3 and D4 positions of PI (6, 8, 9). No binding occurs between the pleckstrin homology domain and the phosphate at the D5 position, providing a structural basis for the lower binding affinity of Akt for PI(4,5)P₂.

The phosphorylation of lipids by PI3K is antagonized by the phosphatase PTEN (phosphatase and tensin homologue), which removes the phosphate group from the D3 position of the inositol ring of PIP₃ (10). PTEN negatively regulates the biological function of PI3K and inhibits the activation of signaling molecules downstream of PI3K (11). In cancer cells with loss-of-function mutations of PTEN, levels of PIP₃ are increased (12), and Akt and its downstream targets show enhanced activation (12). Lipid phosphatases that remove the phosphate group from the D5 position of the inositol ring, including SHIP (Src homology 2-containing inositol 5′-phosphatase) (13), SHIP2 (Src homology 2-containing inositol 5′-phosphatase 2) (14), and PIPP (proline-rich inositol polyphosphate 5-phosphatase) (15), also attenuate PI3K signaling (15-17). The fact that these phosphatases reduce Akt activation and signaling suggests that PIP₃ may be a critical mediator of PI3K signaling.

Here, we show that PIP₃ is required for the transforming properties of all oncogenic PI3Ks. We also show that class III PI3K, which can only produce PIP, cannot transform cells. In addition, the protein kinase activity of the γ isoform of class I PI3K by itself is insufficient to cause oncogenic transformation.

Previous studies from our laboratory showed that lipid kinase activity was required for transformation by p110γ and by the H1047R mutant of p110α (18). As PI3K produces several lipid products, we sought to identify the phospholipids that mediate the oncogenic signal of PI3K. We substituted a short sequence of the activation loop within the kinase domain of p110γ with the corresponding representative sequence from class II PI3K (PI3KC2α), class III PI3K (Vps34p), and the PI3K-related kinase, TOR, here called class IV (3, 4). This substitution changes the lipid substrate specificity and the subsequent lipid product specificity. Only the wild-type class I protein can produce PIP3 along with PIP2 and PIP (4). p110γ with the class II substitution can only produce PIP2 and PIP, whereas p110γ with the class III substitution can only produce PIP (4). These p110γ hybrids were expressed with the RCAS (avian retroviral vector replication-competent leukosis virus with splice acceptor) vector in chicken embryo fibroblasts (CEF). Of all the constructs evaluated, only the wild-type p110γ induced focus formation (Fig. 1A). Expression of wild-type p110γ and of the myristylated, constitutively active p110γ induced phosphorylation of Akt at both Ser⁴⁷³ and Thr³⁰⁸ (Fig. 1B). Expression of p110γ with the class II substitution caused a slight increase in Akt phosphorylation at Ser⁴⁷³, but no phosphorylation was detected at Thr³⁰⁸. Neither class III nor class IV substitutions of p110γ induced significant phosphorylation of Akt at Ser⁴⁷³ or Thr³⁰⁸ over background levels. All hybrid proteins were expressed at similar levels (Fig. 1B). These observations suggest that PIP₃ is required for oncogenic transformation. However, it is not known whether PIP₃ alone is sufficient for transformation or whether PIP₂ is also essential.

To confirm that PIP₃ is an indispensable component of the oncogenic signal of all class I PI3K isoforms, the effect of the 5′-phosphatase PIPP on transformation was studied (15). PIPP is a lipid phosphatase that dephosphorylates inositol lipids at the D5 position of the inositol head group of either PIP₃ or PI(4,5)P₂ and consequently has been shown to reduce the level of PIP₃ at the plasma membrane (15). Expression of PIPP or of the related 5′-phosphatase, SHIP, reduces the level of Akt phosphorylation at Ser⁴⁷³ (15, 19). The catalytic subunits of class I PI3K, p110β, p110γ, and p110δ, and the H1047R mutant of p110α were expressed in CEF along with PIPP or vector alone. PIPP significantly reduced the number of foci formed by all p110 isoforms but had no effect on focus formation by the v-Jun oncoprotein, which was used as control (Fig. 2A). The expression of PIPP resulted in some reduction in the levels of Myc-P3K and of p110αH1047R but did not significantly affect the levels of p110β, p110γ, or p110δ (Fig. 2B). Although we cannot rule out that the reduced abundance of Myc-P3K and of p110αH1047R contributed to lower focus counts, we consider this possibility unlikely because, even at these reduced expression levels, Myc-P3K and p110αH1047R were efficient focus formers in control experiments.

The level of Akt phosphorylation at Thr³⁰⁸ and Ser⁴⁷³ mediated by all PI3K isoforms was also reduced in CEF expressing PIPP (Fig. 2C). As a measure of Akt activity, we determined the levels of the Akt substrates in PIPP-transfected CEF. FoxO1 (forkhead box O transcription factor) is a transcriptional regulator that controls the expression of several genes involved in cell proliferation including p21^Cip1 (20) and cyclin D2 (21). FoxO1 is phosphorylated on Thr²⁴ and Ser²⁵³ by Akt, which then results in the nuclear exclusion and proteolytic degradation of FoxO1 (22, 23). In CEF transformed by PI3K or Akt, FoxO1 is undetectable (24). Expression of PIPP attenuated the degradation of FoxO1 induced by oncogenic PI3K.

The transformation by class I PI3K isoforms is sensitive to inhibition by rapamycin, suggesting that the activity of the TOR kinase is essential for transformation (18). We analyzed the phosphorylation state of two substrates of TOR that are involved in translational control: p70 S6 kinase and 4E-BP1 (eukaryotic initiation factor 4E-binding protein 1). p70 S6 kinase is phosphorylated by TOR on Thr³⁸⁹, and this activation is required for the phosphorylation ribosomal protein S6 by p70 S6 kinase (25). 4E-BP1 binds to eukaryotic initiation factor 4E and thereby prevents assembly of the cap-binding complex that initiates translation of mRNAs. 4E-BP1 is phosphorylated by TOR on Ser⁶⁵, and this phosphorylation releases eukaryotic initiation factor 4E from 4E-BP1, allowing assembly of the initiation complex (26). Expression of PIPP reduced the phosphorylation of both p70 S6 kinase and 4E-BP1. This observation suggests that reduced levels of PIP₃ result in lower activity of TOR. To control for a possible effect of PIPP downstream of Akt or directly on Akt itself, we repeated the above analysis with constitutively active, myristylated Akt (Myr-Akt). PIPP expression did not affect Myr-Akt expression or signaling. We conclude that the effect of PIPP on PI3K signaling is due to a reduction in the levels of PIP₃. In the absence of a myristylation signal, Akt depends on PIP₃ for activation, which in turn mediates the oncogenic signaling of PI3K.

The activation loop of the catalytic domain of PI3K is critical in determining lipid substrate specificity and overall activity. In an effort to determine whether the activation loop defines the different oncogenic potencies of the class I PI3K isoforms, we replaced the activation loop of the wild-type p110α with the activation loop of p110δ or p110γ. Wild-type p110α is not oncogenic and does not transform CEF, whereas both p110δ and p110γ potently induce focus formation in CEF (18). p110α with the activation loop of p110δ (p110α-δ) and p110α with the activation loop of p110γ (p110α-γ) did not acquire the ability to transform CEF and in this respect behaved like wild-type p110α (Fig. 3). In addition, the lipid kinase activity of p110α-δ and p110α-γ was similar to wild-type p110α (Supplementary Fig. S1). These data show that oncogenic activity inherent in the wild-type non-α isoforms of class I p110 cannot be transferred solely with the activation loop.

All class I p110 isoforms exhibit intrinsic protein serine kinase activity as well as lipid kinase activity. p110γ autophosphorylates and phosphorylates several other proteins including the PI3K regulatory subunit p85, MEK1 (4), and nonmuscle tropomyosin (5). Mutants of p110αH1047R and p110γ that retain only the ability to phosphorylate proteins (lipid kinase inactive) cannot induce transformation in CEF (18). As myristylation increases the oncogenicity of p110γ and p110β and renders these proteins to constitutively activate Akt and several substrates downstream of Akt, a myristylated protein kinase-only version of p110γ was produced. This construct did not induce focus formation in CEF and did not activate Akt phosphorylation (data not shown). Therefore, the protein kinase activity alone cannot mediate the oncogenic properties of PI3K, even when p110 is constitutively localized to the membrane.

hVps34 is the human homologue of yeast Vps34p and the only member of class III PI3K in mammals (27). hVps34 can use PI in a phosphorylation reaction to produce PIP (27). We cloned hVps34 into the RCAS expression vector and transfected the construct into CEF. Focus formation was not observed in CEF expressing either wild-type or a myristylated form of hVps34 (Fig. 4A). Expression of the wild-type or the myristylated form of hVps34 did not increase the phosphorylation status of Akt, p70 S6 kinase, 4E-BP1, and glycogen synthase kinase-3β or cause a change in the level of FoxO1 (Fig. 4B). The signaling of hVps34 did not change in the presence or absence of serum. Both the wild-type and the myristylated Vps34 exhibited the ability to phosphorylate PI in an invitro lipid kinase assay (Supplementary Fig. S1).

The three classes PI3Ks are characterized by the specific sets of phospholipids, which they can use as substrates. Consequently, each class of PI3K generates a characteristic set of products, dictated by substrate preferences. There are three PI products with a phosphate at the D3 position, PIP, PIP₂, and PIP₃. Here, we present evidence suggesting that PIP₃ is an essential component of the oncogenic signal that originates in PI3K. It is not known, however, whether PIP₃ is the only PI3K product required for oncogenicity. A possible participation of PIP₂ and PIP cannot be excluded based on the available data, but these latter products of PI3K are clearly not sufficient for oncogenicity.

The activation loop hybrids of p110γ with the ability to produce a restricted set of phospholipids link oncogenic activity to the ability to produce PIP₃. The PIPP phosphatase, which reduces intracellular levels of PIP₃, effectively interferes with the PI3K-induced oncogenic transformation but fails to affect an oncogenic signal that is not dependent on the PI3K-Akt pathway.

The phospholipid products of PI3K recruit Akt and its activating kinase PDK-1, allowing PDK-1 to phosphorylate Akt on Thr³⁰⁸. PDK-1 and Akt bind to PIP₃ with high affinity. Both also bind to PI(4,5)P₂, but weakly, suggesting that the most efficient activation of Akt is mediated by PIP₃ (28, 29). Our results suggest that phosphorylation of Akt on both Thr³⁰⁸ and Ser⁴⁷³ is dependent on PIP₃.

It is conceivable that other proteins, which preferentially bind PIP₃ over PIP₂, are involved in PI3K-induced oncogenic transformation. Some of these PIP₃-binding proteins are the tyrosine kinase Btk (Bruton's tyrosine kinase; 30), the cytohensin 2 and GRP1 (general receptor for phosphoinositides 1) guanine triphosphate exchange factors (31), and the Sos (son of sevenless) and Gab1 (Grb2 associated binder) adaptor proteins (32). The tyrosine kinase Btk has been shown to cause oncogenic transformation by acquiring gain-of-function mutations (33, 34). In addition, Btk appears to be a target of the BCR-Abl fusion protein, as inhibition of Btk by a small-molecule inhibitor induces apoptosis of BCR-Abl-expressing leukemia cells as effectively as inhibition of BCR-Abl itself (35). The oncoprotein Ras can be activated by Shc (Src homology 2 domain containing protein) via the adaptors Sos and Gab1, which has revealed these adaptors as novel targets for anticancer therapy (36). Possible roles of these PIP₃-binding proteins in PI3K signaling remain to be explored.

The central role of PIP₃ in PI3K-induced oncogenic transformation is also reflected by the elevated levels of PIP₃ seen in many cancers. This fact highlights the potential tumor suppressor nature of 5′-phosphatases and suggests that loss-of-function mutations or down-regulation of these phosphatases may be a marker of increased PI3K signaling in tumors. This suggestion is supported by the finding that another 5′-phosphatase, SHIP, is inactivated in adult T-cell leukemia, which displays high levels of phosphorylated Akt (37). Our results with PIPP are in accord with previous work, which shows that the 5′-phosphatases SHIP-1 and SHIP-2 can attenuate Akt signaling (38).

Our studies failed to detect an oncogenic potential in class III PI3K. Class III PI3K is represented by one protein in humans, hVps34. This homologue of the yeast enzyme Vps34p can only use PI to produce PIP (39). One of the physiologic roles of hVps34 is to regulate trafficking of lysosomal enzymes through the trans-Golgi network (40). PIP interacts with FYVE domains of proteins (41), and our results suggest that FYVE domain-containing proteins are not sufficient for oncogenic transformation.

Our results document an essential role of PIP₃ in the oncogenic functions of PI3Ks.

All wild-type and myristylated class I PI3K isoforms constructs were described previously (18). p110γ constructs with replaced catalytic core of class II, III, and IV enzymes as well as p110α-δ and p110α-γ hybrid constructs are designed as described in ref. 4.

All PI3K, hVps34 coding sequences were ultimately cloned into the avian retroviral vector RCAS.Sfi, a modified version of RCAS (42).

In vitro PI3K activity was analyzed as described (43), with minor modifications. The immune complexes were prepared by incubating 400 μg protein with 5 μL anti-p110α antibody for 2 h at 4°C followed by 2 h of incubation with protein A-agarose. The beads were washed three times with lysis buffer and twice with TNE [10 mmol/L Tris (pH 7.5), 100 mmol/L NaCl, 1 mmol/L EDTA]. Subsequently, the immune complexes were incubated with 50 μL kinase reaction buffer containing 20 mmol/L HEPES (pH 7.5), 10 mmol/L MgCl₂, 200 μg/mL PI (sonicated), 60 μmol/L ATP, and 200 μCi/mL (1 Ci/37 GBq) [γ-³²P]ATP for 5 min at room temperature. The reaction was terminated by adding 80 μL of 1 N HCl, and the phosphorylated lipids were extracted with 160 μL chloroform/methanol (1:1). Samples were dried down and dissolved in chloroform and spotted onto Silica Gel 60 TLC plates (EMD). The plates were developed in a borate buffer system (44) and visualized by autoradiography.

Primary cultures of CEF were prepared from white leghorn embryos purchased from Charles River Breeding Laboratories. Each transfection with DNA or infection with avian retrovirus was done with newly prepared CEF. CEF were routinely cultured in F-10.

For focus assays, CEF were infected with RCAS virus containing the relevant oncogene as described previously (45). Cells were fed nutrient agar every other day and then stained with crystal violet to visualize the foci of transformed cells. Inhibitors were prepared in 100% DMSO and added at the desired concentration to the nutrient agar overlays. An equivalent amount of 100% DMSO was added to cells as a control.

Cells were transfected with different RCAS vectors expressing PIPP and oncogenic p110 constructs. PIPP was expressed by RCAS carrying the subgroup B envelope protein [RCAS(B)] and the p110 proteins were expressed with RCAS carrying the subgroup A envelope protein [RCAS(A)]. For efficient expression of two proteins in the same cell, it is essential to use RCAS vectors with different envelope proteins to avoid competition for the same cell surface receptor. A v-jun-expressing vector was used as a control.

Cell lysates were prepared from CEF that had been transfected with DNA using the DMSO/polybrene method (46). For serum starvation, subconfluent cultures were kept in Ham's F-10 medium with 0.5% FCS and 0.1% chicken serum for 40 h before 2 h of starvation in Ham's F-10 medium alone. Inhibitors or DMSO were added at the appropriate concentration during the final 2 h in serum-free Ham's F-10 medium. Western blotting was done as described previously in refs. 18, 43, 47.

No potential conflicts of interest were disclosed.