The PIK3CA gene, encoding the p110α catalytic subunit of Class IA PI3Ks (phosphoinositide 3-kinases), is frequently mutated in many human tumours. The three most common tumour-derived alleles of p110α, H1047R, E542K and E545K, were shown to potently activate PI3K signalling in human epithelial cells. In the present study, we examine the biochemical activity of the recombinantly purified PI3K oncogenic mutants. The kinetic characterizations of the wt (wild-type) and the three ‘hot spot’ PI3K mutants show that the mutants all have approx. 2-fold increase in lipid kinase activities. Interestingly, the phosphorylated IRS-1 (insulin receptor substrate-1) protein shows activation of the lipid kinase activity for the wt and H1047R but not E542K and E545K PI3Kα, suggesting that these mutations represent different mechanisms of lipid kinase activation and hence transforming activity in cancer cells.
PI3Ks (phosphoinositide 3-kinases) are classified by their mode of regulation and substrate specificities. Class IA PI3Ks consist of a p110 catalytic subunit and a regulatory subunit. Currently, there are three known p110 isoforms (α, β and δ) and seven known regulatory subunit proteins: p85α, p85β and p55γ and their splicing variants. PI3Kα (p85α–p110α) belongs to Class I A heterodimeric PI3Ks [1,2]. The regulatory subunit p85α keeps p110α in a stable but repressed state in quiescent cells. Upon stimulation by growth factors, the two SH2 domains (Src homology 2 domains) on p85α interact with the phosphotyrosine motifs on activated RTKs (receptor tyrosine kinases) and recruit the catalytic subunit to the membrane, allowing the activated p110 catalytic subunit to convert PtdIns(4,5)P2 into PtdIns(3,4,5)P3. These membrane-bound second messengers in turn activate numerous downstream pathways involved in cell proliferation, survival, motility and growth.
The dysregulation of the PI3K signalling pathway is implicated in many human cancers [3–5]. In addition to inactivation of the PTEN (phosphatase and tensin homologue deleted on chromosome 10) tumour suppressor gene, mutations or amplifications of the PIK3CA gene (encoding the catalytic α-subunit of PI3K) have been recently reported in human cancers. In particular, a high frequency of ‘hot spot’ mutations have been observed in colorectal cancers, gastric cancers, brain cancers, glioblastomas, as well as in a smaller fraction (<10%) of breast and lung cancers [6–11]. These mutations are clustered in hot spots within the helical (exon 9) and kinase (exon 20) domains of p110α. Among these hot-spot mutations, the three most common tumour-derived alleles of p110α are E542K, E545K and H1047R. Whereas Glu542 and Glu545 are in the helical domain, His1047 is in the catalytic domain of p110α. Several studies have shown that these mutations elevate the lipid kinase activity of PI3Kα, resulting in potent activation of PI3K/Akt (also called protein kinase B) signalling [6,10,12,13]. However, little is known, to date, about the biochemical mechanism of this lipid kinase activity activation. In the present study, we constructed and purified the recombinant PI3K oncogenic mutants E542K, E545K and H1047R and have characterized their steady-state kinetic properties. In addition, we investigated the mechanism of mutant activation by examining their respective lipid kinase activities in the presence of activated RTKs in the form of either phosphorylated peptides or phosphorylated full-length protein. Our biochemical data suggest that the kinase domain mutant H1047R and the two helical domain mutants, E542K and E545K, are likely to augment the lipid kinase activity of PI3Kα through distinct mechanisms.
MATERIALS AND METHODS
QuikChange® site-directed mutagenesis kit was obtained from Stratagene. pENTR-TEV/D-topo vector, pDEST8 and LR clonase were from Invitrogen. PDGFR (platelet-derived growth factor receptor)-derived peptide was obtained from American Peptide with the amino acid sequence CGGY(PO3H2)MDMSKDESVDY(PO3H2)VPMLDM. IRS-1 (insulin receptor substrate-1)-derived peptide was obtained from American Peptide with amino acid sequence based on the sequence of human IRS-1-(607–636) HTDDGY(PO3H2)MPMSPGVAPVPSGRKGSGSY(PO3H2)MPMS. Synthetic diC8-PtdIns(4,5)P2 [1,2-dioctanoyl-sn-glycero-3-PtdIns(4,5)P2] was purchased from Cell Signaling Technology. All other laboratory chemicals and materials were of standard laboratory grade.
Protein expression and purification
The full-length human His-p85α and each of the p110α variants [wt (wild-type), E542K, E545K and H1047R] were co-expressed in sf9 cells from modified Gateway pDEST8 baculovirus expression vectors containing the genes encoding p110α subunit and p85α subunit. The p110α (wt or mutants) in complex with His-tagged p85α was purified by capturing the complex on Ni-NTA (Ni2+-nitrilotriacetate)–agarose (Qiagen) through the His tag on the p85α protein. The p110α was then purified from contaminating p85α by using ion-exchange chromatography on MonoQ (GE Healthcare) and by size-exclusion chromatography on SDX200 (GE Healthcare). The resulting complex was approx. 95% pure without excess p85α.
Kinetic characterization of PI3Kα wt and mutants
PI3Kα (p85α–p110α) wt and mutant isoenzymes were characterized using an FP (fluorescence polarization) kit from Echelon Biosciences (K-1100) . Assays were performed at room temperature (25 °C) in 20 mM Mops (pH 6.5) and 25 mM MgCl2 with 50 μM PtdIns(4,5)P2 and 3 nM of p85α–p110α isoenzyme. Reactions were quenched at various time points using a potent PI3Kα inhibitor and detected using an LJL Biosciences Analyst instrument. The concentration of PtdIns(3,4,5)P3 product formed was calculated from change in FP signal using a PtdIns(3,4,5)P3 standard curve.
Activation of PI3Kα by PDGFR-derived peptide, IRS-1-derived peptide and phospho-IRS-1 protein
Activation of wt and mutant PI3Kα was measured in an SPA (scintillation proximity assay) format. Specific activity was determined in the presence of various concentrations of PDGFR-derived peptide, IRS-1-derived peptide or phospho-IRS-1 protein under the following assay conditions: 10 mM Mops (pH 6.5), 40 μM PtdIns(4,5)P2, 4 nM of PI3Kα isoenzyme, 25 mM MgCl2, 25 μM ATP and 0.25 μCi of [33P]ATP in a total volume of 25 μl. Reactions were quenched with 25 μl of 50 mM EDTA. Wheatgerm agglutinin-coated yttrium silicate SPA beads (GE Healthcare) were suspended in potassium phosphate buffer (pH 7.6) and added to the quenched reactions to a final concentration of 2 mg/ml in 17 mM phosphate. Percentage activation over control was calculated using eqn (1).
To activate the full-length rat IRS-1 protein, human GST (glutathione transferase)–IR (insulin receptor) was first activated by incubating at room temperature for 30 min with 0.5 mM ATP and 10 mM MgCl2 in 25 mM Hepes (pH 7.2). Activated GST–IR was then incubated with full-length rat IRS-1 (catalogue no. 13-124; Upstate) at equimolar concentration for 30 min at room temperature resulting in 600 nM of activated phospho-IRS-1 protein.
RESULTS AND DISCUSSION
Steady-state kinetic properties of PI3Kα mutants
High frequencies of somatic mutations in the PIK3CA gene have been identified in various human tumours. Among them, three hot-spot mutations, E542K, E545K and H1047R, account for more than 70% of all the mutations identified in human primary tumours to date [6,15]. Two of these hot-spot mutations, E542K and E545K, are located at the helical domain of p110α and the third hot-spot mutation, H1047R, is located in the C-terminal portion of the kinase domain. Several recent studies have shown that these mutations lead to elevated lipid kinase activity of PI3Kα and hence activate the Akt signalling pathway. To characterize more completely the kinetic mechanism of the oncogenic PI3K mutants, we expressed and purified the wt and all three hot-spot oncogenic mutants. The wt and mutant variants were co-expressed as p85α and p110α (or mutant p110α) in Sf9 cells and purified as p85α–p110α complexes. Densitometry analysis of SDS/PAGE gels of the purified wt and mutant variants showed approx. 1:1 (p85α/p110α) stoichiometry for all four purified proteins (results not shown). The kinetic characterization of the wt p85α–p110α and the mutant variants were performed using an FP activity assay. We used diC8-PtdIns(4,5)P2 in which the fatty acid moieties are eight carbons long and the CMC (critical micelle concentration) was determined to be higher than 100 μM. Using the water-soluble diC8-PtdIns(4,5)P2, it is possible to vary the concentration of PtdIns(4,5)P2 and attain saturating concentrations. [The Km for water-soluble diC8-PtdIns(4,5)P2 was estimated at 3.5 μM using the FP activity assay for the wt enzyme. For the FP assay, quantitative measurement of the lipid substrate Km was challenging because of the different background at various PtdIns(4,5)P2 concentrations. In the steady-state kinetic experiments, 40 μM PtdIns(4,5)P2 was used to ensure that the lipid substrate is saturating.] The kinetic parameters determined for the recombinantly purified wt and mutant variants of PI3Kα showed that the three oncogenic mutants have 2–3-fold increases in kcat relative to the wt enzyme (see Figure 1 and Table 1). The ATP Km, on the other hand, is largely unchanged for the E542K and H1047R mutants. For the E545K mutant, the ATP Km decreased 2-fold compared with that of the wt enzyme. These results are in agreement with the previous reports in which elevated lipid kinase activity for hot-spot mutants was observed with in vitro lipid kinase assays [6,8,10,13]. There are, however, a number of technical differences between these previous reports and the present results. First, the previous studies utilized phosphatidylinositol as the lipid substrate, while our work was performed with the more water-soluble PtdIns(4,5)P2 substrate. Secondly, immunoprecipitated protein from cell lysates was used as the enzyme source in the previous studies, whereas purified, recombinant enzymes were used for the present studies. The availability of purified, recombinant enzyme and water-soluble PtdIns(4,5)P2 substrate afforded us an opportunity to assess quantitatively the steady-state kinetic parameters for the wt and ‘hot spot’ mutant PI3Kα forms and thus define a mechanism for the transforming potential of the oncogenic mutations.
wt (a), E542K (b), E545K (c) and H1047R (d) PI3Kα lipid kinase activity in the presence of saturating PtdIns(4,5)P2 and varied concentrations of ATP
|PI3Kα||kcat (s−1)||Km ATP (μM)||kcat/Km ATP (s−1·μM−1)|
|PI3Kα||kcat (s−1)||Km ATP (μM)||kcat/Km ATP (s−1·μM−1)|
Activation of wt and mutant variants of PI3Kα by PDGFR-derived peptide and IRS-1-derived peptide
PI3Kα is activated in cells through the interaction of the p85 regulatory subunit with phosphorylated YXXM motifs on RTKs such as PDGFR and EGFRs (epidermal growth factor receptors), or receptor-associated molecules such as IRS-1 [16–19]. It is postulated that the binding of phosphorylated YXXM motif-containing proteins to the p85α SH2 domains de-represses the inhibitory effect of p85α on p110α and restores the catalytic activity of p110α [20–22]. One potential mechanism by which the oncogenic mutations of p110α elevate their lipid kinase activity and promote tumour growth may be through the loss of basal repression of the p110α catalytic activity by p85α. Indeed, for both the p85α C-terminal SH2 domain truncation mutant, isolated from murine lymphomas, and the small internal deletion (Met582–Asp605) mutant of p85α identified from human tumours, this mechanism is proposed to explain the elevation of PI3K activity [23,24]. In both cases, increased basal PI3K lipid kinase activity in the absence of growth factor stimulation was noticed. To test the hypothesis that the mutational activation by the ‘hot spot’ mutants is through loss of basal repression of p110α by p85α, we measured the lipid kinase activity of the wt and the mutant variants in the presence of doubly phosphorylated PDGFR and IRS-1 peptides. These studies were performed using an SPA developed in our laboratory (G. S. Van Aller, J. D. Carson, C. Fernandes, P. J. Tummino and L. Luo, unpublished work). This assay allowed us to directly monitor the 33P-labelled PtdIns(3,4,5)P3 product formation by wt and mutant PI3Kα catalysed reactions in the presence of doubly phosphorylated peptides or the phosphorylated IRS-1 protein. In contrast, the presence of the activating reagents (peptides or full-length protein) seemed to interfere with the FP activity assay, presumably through affecting the interaction between PtdIns(3,4,5)P3 and the PH domain (pleckstrin homology domain). As shown in Figure 2(a), concentration-dependent activation of lipid kinase activity was observed for the wt and H1047R PI3Kα in the presence of doubly phosphorylated PDGFR peptide. The maximal activation for wt PI3Kα was 1.6-fold and the half-maximal effect peptide concentration was estimated at 200 nM. This result agreed well with the previously reported activation effect of doubly phosphorylated PDGFR peptide using purified PI3Kα, where an approx. 2-fold activation was reported . For the H1047R mutant, the maximal activation was decreased to 1.2-fold and the half-maximal effect peptide concentration was 130 nM. There was only slight activation for E545K at high concentration of the doubly phosphorylated peptide. No activation of lipid kinase activity by the doubly phosphorylated PDGFR peptide was observed for the E542K mutant. These results indicate that for the E545K and E542K mutants, the lipid kinase activity cannot be further activated by RTK binding. While the activation of the H1047R mutant by doubly phosphorylated PDGFR peptide is observable, the extent of activation is decreased compared with that of the wt PI3Kα. In addition to the PDGFR peptide, the doubly phosphorylated IRS-1 peptide was evaluated (Figure 2b). Both wt and H1047R PI3Kα showed concentration-dependent increases in lipid kinase activity in the presence of IRS-1 peptide with maximal activation approx. 2-fold. The half-maximal effect peptide concentration was 6.3 μM for wt and 4.0 μM for H1047R. In contrast, much weaker lipid kinase activity activation (<1.2-fold) was observed for E545K and E542K mutants. Altogether, these results suggest that the E542K and E545K mutants cannot be further activated by doubly phosphorylated peptides, whereas the H1047R mutant behaves similar to the wt enzyme and can be activated by the doubly phosphorylated peptides that mimic phosphorylated RTK protein.
wt (○), E542K (Δ), E545K (◇) and H1047R (□) PI3Kα lipid kinase activity in the presence of (a) PDGFR-derived peptide or (b) IRS-1-derived peptide
Effect of IRS-1-derived peptide on ATP binding of wt and mutant variants of PI3Kα
It is of great interest to understand the effect of the binding of the activating RTK agents on the kinetic parameters of wt and mutant variants of PI3Kα. In particular, deconvolution of the effect of the binding of doubly phosphorylated peptide on catalysis (kcat effect) versus substrate binding (Km effect) can help us to better understand the mechanism of lipid kinase activity activation. We first attempted to examine the effects by FP activity assay. Unfortunately, the presence of the doubly phosphorylated peptides seemed to have adverse effect on the FP assay, rendering this assay unsuitable for the study. We then used the SPA format to measure the kinetic parameters for wt and mutant variants of PI3Kα in the presence and absence of IRS-1 peptide. The SPA format could not be used to quantitatively measure the turnover rate (kcat) because of the lack of a 33P-labelled PtdIns(3,4,5)P3 product standard curve. Nevertheless, this assay can be used to measure the ATP Km for wt and mutant variants of PI3Kα in the presence and absence of the doubly phosphorylated peptide. As shown in Table 2, the ATP Km values determined for the wt and three mutant variants of PI3Kα are very similar to each other without IRS-1 peptide. These values agree well with the results obtained by the FP activity assay. Interestingly, no observable changes in ATP Km were seen in any of the four enzymes in the presence of 15 μM IRS-1 peptide, suggesting that the binding of doubly phosphorylated peptide does not affect ATP binding for the wt and the three mutant variants.
Activation of wt and mutant variants of PI3Kα by phosphorylated IRS-1
It was previously reported that multiply phosphorylated IRS-1 protein has a more prominent PI3K activation effect than that of monophosphorylated or doubly phosphorylated YXXM peptides . To evaluate the activation effects of phosphorylated IRS-1 protein on wt and mutant variants, we used full-length rat IRS-1 protein. The rat IRS-1 protein is 88% identical with the human IRS-1, with residues surrounding the YXXM motifs mostly conserved. The rat-IRS-1 was phosphorylated in vitro by human IR (see the Materials and methods section). The lipid kinase activity for wt, E542K, E545K and H1047R PI3Kα were measured in the presence of 0, 20 and 200 nM of phosphorylated IRS-1 protein. As shown in Figure 3, both wt and H1047R showed a ≥2-fold increase in lipid kinase activity in the presence of 200 nM of phosphorylated IRS-1 protein. In contrast, no significant activity increase was observed for the E542K and E545K mutants at any phospho-IRS-1 concentration tested. To investigate further the activation effects of phosphorylated IRS-1 protein on wt and H1047R PI3Kα, we measured the lipid kinase activity of wt and H1047R in the presence of six different concentrations of phosphorylated IRS-1 protein. The titration curves of the IRS-1 protein showed a clear IRS-1 concentration-dependent elevation of the lipid kinase activity for both wt and H1047R PI3Kα (Figure 4). At 300 nM of phosphorylated IRS-1 protein, the highest concentration achievable under the assay conditions, wt PI3Kα had a 2.3-fold increase in lipid kinase activity, while H1047R PI3Kα had an approx. 4-fold increase. Interestingly, the titration curves for IRS-1 suggested that the activation effect was saturated for wt PI3Kα at 300 nM but not for H1047R PI3Kα. The calculated half-maximal effect IRS-1 concentration was 125±20 nM for the wt reaction. Taken together, these results and the results from the doubly phosphorylated PDGFR and IRS-1 peptide studies suggest that RTK binding exerts an activation effect on the wt and H1047R, but not on E542K and E545K PI3Kα, indicating that the two helical domain mutants and the kinase domain mutant have different mechanisms of mutational activation. In a recent publication, Gymnopoulos et al.  reported the mapping of the hot-spot PI3Kα mutants along with 15 rare cancer-derived mutants of PIK3CA on to a structural model of p110α. They proposed that the mutants can be divided into three groups on the basis of their location in distinct functional domains of p110α and each group induces a gain of PI3K function through a different molecular mechanism. In particular, the kinase domain mutations, including H1047R, are located near the hinge binding region of the activation loop and are postulated to enhance the catalytic activity of the enzyme by locking the loop in the ‘on’ position. On the other hand, the helical domain mutations, including E542K and E545K, are clustered into an exposed region and might interact with the cellular membrane. The switch in charge for these residues might be a means of enhancing protein–protein interaction or protein–membrane interaction. Most recently, Miled et al.  reported their studies on the mechanism of activating mutations in the helical domain. On the basis of the crystallographic and biochemical data, they proposed that the helical domain mutations such as E545K disrupt the interaction between p110α and p85 nSH2 (N-terminal SH2 domain). The authors suggested that the interaction between p85 iSH2 (inter-SH2 domain) and p110α ABD (adaptor-binding domain) provides the major stabilizing interaction in PI3K. The interaction between the p85 nSH2 and p110α helical domain, on the other hand, plays an important role in mediating the inhibition of p110α lipid kinase activity [20,22]. They showed that the p85 nSH2–p110α helical domain interaction is mostly charge-charge interaction. Reverse charge mutations such as glutamic acid to lysine mutation can disrupt the interaction and release p110α from this inhibitory interaction with p85 nSH2. Indeed, the results from our studies show that the E542K and E545K mutants in the p110α helical domain cannot be further activated by RTK binding, supporting the model that the reverse charge mutations have interrupted the inhibitory p85 nSH2–p110α helical domain interaction.
wt, E542K, E545K and H1047R PI3Kα lipid kinase activity in the presence of 0 nM (white bars), 20 nM (grey bars) and 200 nM (black bars) phosphorylated IRS-1 protein
wt (a) and H1047R (b) PI3Kα lipid kinase activity in the presence of various concentrations of phosphorylated IRS-1 protein
In summary, our results suggested that all three ‘hot spot’ mutants, E542K, E545K and H1047R, have a reproducible approx. 2-fold increase in lipid kinase activity. However, the E542K and E545K mutants cannot be further activated by RTK binding, suggesting that these glutamic acid to lysine mutations might activate the p110α through the same mechanism as the binding of the RTKs. In contrast, the H1047R mutant is still sensitive to RTK activation, agreeing with the hypothesis that the kinase domain mutations affect the activation loop and that this mode of activation is distinct from the activation by growth factor binding. These biochemical data are in good agreement with the crystallographic data recently reported by Miled et al.  and shed new light on the oncogenic PI3K mutations. These findings will help us to better understand the mechanism of mutational activation of PI3K p110α mutants and develop strategies to target the oncogenic mutants specifically for the treatment of cancer.
The authors are employed by GlaxoSmithKline for Biomedical Research.
platelet-derived growth factor receptor
gene encoding the catalytic α-subunit of PI3K
receptor tyrosine kinase
- SH2 domain
Src homology 2 domain
N-terminal SH2 domain
scintillation proximity assay