Oncogenic transformation induced by the p110β, -γ, and -δ isoforms of class I phosphoinositide 3-kinase

Results

The p110β, -γ, and -δ Isoforms of Class I PI3K Induce Oncogenic Transformation in Cell Culture. The β, γ, and δ isoforms of p110 were expressed with the avian retroviral vector replication-competent avian leukosis virus with splice acceptor (RCAS) in chicken embryo fibroblasts. All three isoforms induced the formation of transformed cell foci within 10 days (Fig. 1). These foci were particularly distinct in monolayer cultures expressing the γ or δ isoforms. The foci consisted of multiple layers of densely packed cells that contrasted sharply against the background of normal cells. The β isoform induced less distinct foci with fewer cell layers in each focus. The efficiency of transformation as measured by focus forming units per ng of DNA was 0.044 for the p110α isoform, 0.9 for the β isoform, 4.4 for the δ isoform, and 1.0 for the γ isoform. The fusion of a myristylation signal to the N terminus of the p110 isoforms enhanced oncogenic transforming ability. This enhancement was less pronounced with the γ and δ isoforms, which are potent transformers even without the added myristylation signal. In contrast, the transforming activity of the β isoform was greatly enhanced upon fusion to a myristylation signal. The average focus forming activity per ng of DNA of the myristylated isoforms was 6.9 for the α isoform, 4.3 for the β isoform, 7.3 for the δ isoform, and 1.9 for the γ isoform. These data show that in contrast to p110α, which in its wild-type version is not oncogenic, the β, γ, and δ isoforms have an inherent oncogenic potential that is revealed by overexpression.

Oncogenic Transformation by p110 Isoforms Requires Kinase Activity. PI3Ks are primarily lipid kinases that control diverse cellular processes by generating second messenger molecules, including phosphatidylinositol 3,4,5-trisphosphate (1, 2). However, PI3Ks also possess inherent protein kinase activity that can lead to autophosphorylation and phosphorylation of diverse target proteins (24–35). To determine the enzymatic activities required for oncogenic transformation, we generated kinase-inactive mutations and determined their transforming activity. The D910A mutation, which alters the aspartate residue located in the activation loop of the enzyme to an alanine, has been previously characterized in the δ isoform as a kinase-inactive mutant (36). The δ isoform harboring this mutation was not able to induce focus formation, suggesting that intact kinase activity is required for transformation (Fig. 2). The D910A mutation has not been studied in other isoforms, but the amino acid residue corresponding to D910 in p110δ and its surrounding sequences are highly conserved in all isoforms of p110. We introduced this mutation in the γ isoform (D964A) and in the highly oncogenic mutant isoform p110α H1047R (D933A), and we observed a complete inhibition of transforming activity. Similar to the D964A mutation, the K832R mutation, which has been previously characterized in the γ isoform to be kinase-inactive (27), completely blocked the transforming activity of p110γ. These results suggest a requirement for intact kinase activity of the p110 isoforms in the cellular transformation process.

We also tested a mutant that has lost lipid kinase activity but retains protein kinase activity. This lipid kinase-deficient p110 has previously been engineered and characterized in the α and γ isoforms by substituting the conserved catalytic core sequence with the corresponding sequence of the target of rapamycin (TOR) (37, 38). TOR belongs to the PI3K-related kinases; it has protein but not lipid kinase activity (39, 40). The mutant p110γ with only protein kinase activity failed to induce cellular transformation (Fig. 2). The transforming activity of the oncogenic H1047R mutant of p110α isoform was also completely abrogated by this lipid kinase-inactive mutation. These data show that lipid kinase activity is essential for oncogenic transformation and that protein kinase alone is insufficient.

The δ Isoform of p110 Constitutively Activates the Akt Signaling Cascade. We have previously shown that oncogenic versions of p110α induce a constitutive activation of the Akt pathway (41, 42). Here, we investigated downstream signaling of the p110β, -γ, and -δ isoforms. Chicken embryo fibroblasts transformed by these isoforms and positive and negative controls were examined by Western analysis with antibodies directed against Akt, phospho-Akt (Ser-473), p70 S6 kinase (S6K), and phospho-S6K (Thr-389) (Fig. 3A). Even under serum-starved conditions, cells expressing p110δ showed strong activation of the Akt pathway. The elevated levels of phospho-Akt and phospho-S6K present in p110δ-expressing cells were comparable to those detected in cells producing the oncogenic H1047R mutant of p110α. In contrast, serum-starved cells expressing the β or γ isoforms of p110 did not show constitutive activation of Akt. Only when cells were grown in the presence of 3% serum could activation of Akt by p110γ be detected (Fig. 3B). The p110γ mutant lacking total kinase activity or lipid kinase activity failed to induce an increase in the phosphorylation level of Akt. This correlation between transforming activity and activation of Akt suggests that p110γ-induced cellular transformation depends on Akt.

Binding to Ras Is Essential for Transformation by p110β and p110γ. The four isoforms of p110 have Ras-binding domains in the N-terminal portion of their sequence. Previous studies have shown that a point mutation in this region, K227E, blocks the interaction of p110α with Ras and the ability of Ras to activate PI3K (43). This mutation, when introduced into the myristylated form of p110α (41), or into one of the oncogenic mutations of p110α, had no effect on the transforming potential of the protein (Fig. 4). The lysine residue that corresponds to K227 of the α isoform is highly conserved in other p110 isoforms. Earlier studies had shown that the corresponding point mutation in the γ isoform (K255E) also interferes with the interaction with Ras (37), and we generated that mutation in p110γ. In contrast to the α isoform, the K255E-mutated p110γ was significantly compromised in its oncogenic potential (Fig. 4). Similarly, the transforming activity of the β isoform was inhibited when the corresponding mutation (K230E) was introduced in p110β. The transforming potential of p110δ, however, was not affected by this mutation (K223E). These data suggest that the oncogenic activity of p110β and p110γ depends on upstream input from Ras, but for p110α and p110δ such input is not essential for transformation.

Inhibitory Effects of Rapamycin on Oncogenic Transformation Induced by p110 Isoforms. We have previously shown that the transformation induced by oncogenic forms of p110α is mediated by TOR and can be effectively blocked by rapamycin, a specific inhibitor of TOR (42, 44). To determine whether the oncogenicity induced by other members of class I PI3K is also TOR-dependent, the effect of rapamycin on the transforming activity of p110β, -δ, and -γ was investigated (Fig. 5). Focus formation by the H1047R mutant of p110α was strongly inhibited at the concentration of 1 ng/ml of rapamycin, as previously reported (42). Similarly, cellular transformation induced by other isoforms of class I PI3K was substantially inhibited by the same concentration of rapamycin. However, rapamycin had no effect on the transforming activity of the v-Jun oncoprotein. These results suggest that the oncogenic signal of p110β, p110γ, and p110δ is mediated by a TOR-dependent process.

Discussion

Overexpression of the wild-type catalytic subunits p110β, -γ, or -δ of class I PI3K is sufficient to induce an oncogenic phenotype in cultured cells. In contrast, wild-type p110α lacks this transforming potential but can acquire it by point mutations or by the addition of a myristylation or farnesylation signal (41, 42). The potential role of the non-α isoforms of class I PI3K in cancer has not been fully explored. However, there are reports of elevated expression of p110β and p110δ in various human cancers (17, 18). In contrast to the prevalence of p110α mutations detected in various tumor types, there have been no reports of cancer-specific mutations in p110β, -γ, or -δ (15). It is possible that this absence of mutations in the non-α isoforms reflects their oncogenic potential as wild-type proteins. Differential expression of wild-type p110β, -γ, or -δ could conceivably make a contribution to the oncogenic phenotype of the cancer cell. Reexamining expression profiles of various cancers for possible up-regulation of non-α isoforms at the RNA and protein levels appears worthwhile.

The reasons for the oncogenic potential of the non-α isoforms of p110 are not known. Some mutated residues that are linked to a gain of function in p110α are found in wild-type non-α isoforms. For instance, the H1047R mutation of p110α is highly oncogenic; the p110γ wild-type protein carries an arginine residue at the corresponding site (R1076). Yet the arginine to histidine mutation at this site in p110γ does not affect oncogenicity (Fig. 2). The N345K mutation in the C2 domain also makes p110α oncogenic (M. Gymnopoulos and P.K.V., unpublished data). The corresponding residue of the wild-type p110β (K342) and p110δ (K332) already contains lysine at this site. The effect of mutating these residues in p110β and p110δ to the corresponding asparagine residue of p110α remains to be determined. Introducing the cancer-specific mutations of p110α into the non-α isoforms may provide information on oncogenic mechanisms.

The oncogenicity of all isoforms of class I p110 depends on kinase activity. For mutant p110α and wild-type p110γ, lipid kinase is essential (Fig. 2). Protein kinase may also be required but is not sufficient. For p110β and -δ, the relative roles of lipid versus protein kinase activities remain to be investigated. However, the tumor-suppressive effect of the lipid phosphatase PTEN strongly argues in favor of a dominant, if not exclusive, role of lipid kinase activities in the oncogenic transformation induced by p110 isoforms.

The requirements for upstream and downstream signaling in the transformation process induced by the p110 isoforms are in accord with published literature. These requirements divide the isoforms into two groups: one consisting of p110α and -δ, the other encompassing p110β and -γ. Both p110α and -δ are linked to upstream receptor tyrosine kinases (reviewed in refs. 45 and 46). Although p110α and -δ have Ras-binding domains and can bind Ras, a mutation that abolishes Ras binding of oncogenic p110α does not interfere with transformation. The analogous mutation in p110δ also retains transforming potential, but the effect of this mutation on Ras binding remains to be determined (Fig. 4). Cancer-specific mutations of p110α and the wild-type protein p110δ are strong stimulators of Akt signaling, which is activated constitutively under serum-starved conditions. In contrast to the α and δ isoforms, upstream signaling to the γ isoform originates with G protein-coupled receptors (2, 7, 8). The β isoform can be activated by G protein-coupled receptors and by receptor tyrosine kinases (9–11, 45). The γ isoform depends on interaction with Ras for oncogenic transformation, and the same appears to be true of p110β, although additional data on Ras binding are needed to confirm this conclusion (Fig. 4). The upstream signaling requirements of the p110β and -γ isoforms are therefore distinct from those of the α and δ isoforms. The β and γ isoforms are also set apart by their effect on Akt. Under conditions of serum starvation, they do not induce a detectable activation of Akt (Fig. 3A). These results are consistent with the recent findings showing that the wild-type p110β by itself does not induce activation of Akt in human mammary epithelial cells under serum-starved conditions (47). However, adding a myristylation signal to the N terminus of p110β results in a constitutive, serum-independent activation of Akt (47, 48). Similarly, the myristylated form of p110γ also induces constitutive activation of Akt in Rat1 fibroblasts (48), but the stimulating effect of the wild-type p110γ on Akt is revealed only in the presence of serum (Fig. 3B), and that effect is weak compared to that of the α and δ isoforms. Yet oncogenic transformation by all four isoforms of class I PI3K is highly sensitive to rapamycin (Fig. 5). The Akt-TOR axis is therefore an essential and common feature of the transforming signal emanating from the isoforms of p110.

The relative enzymatic activity of p110α is higher than that of other isoforms (49, 50), suggesting that the inability of the wild type p110α to induce transformation is not due to low specific activity. Western analyses suggest that the non-α isoforms of p110 are expressed at higher levels than p110α, assuming that the sensitivities of isoform detection by the different antibodies are roughly comparable (Fig. 3). Overexpression of p110α but not of p110β may be toxic (51), and the inability of wild-type p110α to induce oncogenic transformation could be attributed to low levels of expression.

Expression of p110α, -β, and -δ (class IA PI3K) depends on the availability of the corresponding regulatory subunits, which may stabilize the catalytic subunits (49, 52, 53). In contrast, p110γ of class IB PI3K is stable as a monomer without its regulatory subunit p101 (54, 55). Although we have not cotransfected the regulatory subunits with p110 isoforms, stable overexpression of p110β, -δ, and -γ was achieved (Fig. 3). This observation suggests that endogenous p85 expressed in CEF may function as a stabilizing factor (Fig. 3). Expression of a particular p110 isoform of class IA tends to affect the expression levels of other isoforms that share the same regulatory subunit. For example, the endogenous levels of p110α are down-regulated in cells overexpressing the β or δ isoform. Expression of the δ isoform also leads to reduced endogenous levels of p110β (Fig. 3). Overexpression of a particular p110 isoform may therefore titrate the corresponding regulatory subunit, limiting the availability to other isoforms and, thus, cause down-regulation.

The unexpected oncogenic potential of the non-α isoforms of p110 places these proteins in the same category with oncoproteins like Myc, which also induces oncogenic transformation as a wild-type protein if overexpressed (56–59). The extent to which the non-α isoforms of p110 contribute to human cancer needs to be investigated in future studies.

Materials and Methods

Cell Culture and Transformation Assays. Primary cultures of CEF were prepared from White Leghorn embryos obtained from Charles River Breeding Laboratories (Preston, CT). For transformation assays, DNA was transfected into CEF by using either the Lipofectamine reagent (Invitrogen, Carlsbad, CA) or the DMSO/Polybrene method (60). After incubation of cells under nutrient overlay, foci of transformed cells were counted on day 10 after transfection. Assays with infectious retroviral vectors were performed as described in ref. 13 and 61. Transfected or infected cells were fed every other day with nutrient agar and then stained with crystal violet. For serum starvation, subconfluent cultures were maintained in Ham's F-10 medium with 0.5% FCS and 0.1% chicken serum. After 40–44 h, the cells were harvested for protein analysis.

Plasmid Construction. Wild-type and myristylated chicken p110α expression constructs were described in ref. 41. The human wild-type p110α expression construct was generated by PCR amplification of the original cDNA clone DKFZp686D20244 (RZPD, Berlin) by using the primers Hu-α-koz-F/Hu-α-rev (see Table 1, which is published as supporting information on the PNAS web site, for the primer sequences). The PCR product was MluI- and ClaI-digested and cloned into MluI and ClaI cloning sites of pBSFI adaptor plasmid (62). To generate human wild-type p110β, p110δ, and p110γ constructs, the same cloning procedures were used with the following exceptions: p110β was PCR amplified from the “pBS-SK(+)-hump110β” plasmid (B.V., unpublished data) by using the primers β-koz-for/β-rev and digested with EcoRI and HindIII. p110δ was PCR-amplified from the “pMT2SM-oligo+hp110δWTnotag3UTR plasmid” (B.V., unpublished data) by using the primers δ-koz-for/δ-rev and digested with BamHI and ClaI. p110γ was PCR-amplified from the “pcDNA3-p110γ-VSV” plasmid (63), a generous gift from Reinhard Wetzker (Friedrich Schiller-University, Jena, Germany), with the primers γ-koz-for/γ-rev and digested with BamHI and MluI. Digested PCR products were cloned into corresponding sites of the pBSFI cloning vector, and then subcloned into the avian retroviral vector RCAS.Sfi, a modified version of RCAS (62).

For myristylated p110 constructs, similar cloning procedures were used with the following modifications: p110β was PCR-amplified by using primers β-5′-Mlu and β-3′-Hind and cloned into MluI and HindIII sites of the Myr-pBSFI vector, which contains the amino-terminal 14 amino acids of c-Src. The Myr-pBSFI plasmid was previously constructed by cloning the annealed oligonucleotide strands (Myr-5′B and Myr-3′M) into the BamHI and MluI sites of the pBSFI vector. p110δ and p110γ were PCR-amplified by using primers δ-5′-Mlu/δ-3′-Cla and γ-5′-Mlu/γ-3′-Sal, respectively, and cloned into MluI, ClaI and MluI, SalI sites of the Myr-pBSFI vector.

The following point mutations were generated by using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) with the indicated sets of primers: human p110α H1047R (primers Hu-α-H1047R1/Hu-α-H1047R2), p110α H1047R/K227E (primers αH1047R-K227E1/αH1047R-K227E2), p110α H1047R/D933A (primers αH1047R-D933A1/αH1047R-D933A2), p110β-K230E (primers β-K230E1/β-K230E2); p110δ-K223E (primers δ-K223E1/δ-K223E2), p110δ-D910A (primers δ-D910A1/δ-D910A 2); p110γ-K255E(primers γ-K255E1/γ-K255E2), p110γ-D964A (primers γ-D964A1/γ-D964A2), and p110γ-K832R(primers γ-K832R1/γ-K832R2); p110γ-R1076H (primers γ-R1076H1/γ-R1076H2). Lipid kinase-inactive mutants of p110αH1047R and p110γ were constructed as described in refs. 37 and 38.

Western Analysis. Western blotting was performed as described in refs. 41 and 42. In addition to previously described antibodies, anti-phospho-Akt (T308) antibody and anti-p85 antibody were obtained from Cell Signaling Technology (Beverly, MA), anti-p110β antibody (S-19) was purchased from Santa Cruz Biotechnology (Santa Cruz, CA), and VSV-G polyclonal antibody (PRB-192P) was purchased from Covance Research Products (Berkeley, CA) to detect VSV-tagged p110γ proteins. Anti-p110δ antibody has been described in ref. 3.

Supplementary Material