Autophosphorylation of serine 608 in the p85 regulatory subunit of wild type or cancer-associated mutants of phosphoinositide 3-kinase does not affect its lipid kinase activity

Background The α-isoform of the Type 1A Phosphoinositide 3-kinases (PI3Kα) has protein kinase activity as well as phosphoinositide lipid kinase activity. The best described substrate for its protein kinase activity is its regulatory subunit, p85α, which becomes phosphorylated on Serine 608. Phosphorylation of Serine 608 has been reported to down-regulate its lipid kinase activity. Results We have assessed whether oncogenic mutants of PI3Kα, which have up-regulated lipid kinase activity, have altered levels of Serine 608 phosphorylation compared to wild type PI3Kα, and whether differential phosphorylation of Serine 608 contributes to increased activity of oncogenic forms of PI3Kα with point mutations in the helical or the kinase domains. Despite markedly increased lipid kinase activity, protein kinase activity was not altered in oncogenic compared to wild type forms of PI3Kα. By manipulating levels of phosphorylation of Serine 608 in vitro, we found no evidence that the protein kinase activity of PI3Kα affects its phosphoinositide lipid kinase activity in either wild-type or oncogenic mutants of PI3Kα. Conclusions Phosphorylation of p85α S608 is not a significant regulator of wild-type or oncogenic PI3Kα lipid kinase activity.


Background
Phosphoinositide 3-kinases (PI3Ks) are a ubiquitous family of lipid kinases that catalyse the phosphorylation of phosphoinositide lipids at the 3' position on the inositide ring [1]. The Class 1A sub-group exists as heterodimers of a p110 catalytic subunit and a p85 regulatory subunit that can phosphorylate three phosphoinositides (PI, PI-(4)-P and PI-(4,5)-P 2 ) in vitro, but are thought to primarily phosphorylate PI-(4,5)-P 2 in vivo to form PI-(3,4,5)-P 3 (PIP 3 ) [1]. PIP 3 resides almost exclusively in the plasma membrane and propagates signals by recruiting and activating a variety of downstream proteins that contain PIP 3 binding domains (primarily Pleckstrin Homology (PH) domains) [2]. PI3K/PIP 3 signalling regulates a wide range of cellular processes including cell growth, survival, glucose metabolism and migration [1,3].
Class 1A PI3Ks also have protein kinase activity. The p110α catalytic subunit can phosphorylate its regulatory subunit, p85α, at Serine 608 (S608). Phosphorylation of this site has been reported to result in feed-back inhibition by down-modulating the lipid kinase activity of p110α [15][16][17], however the role of phosphorylation of S608 in signalling by endogenous PI3Kα and the structural mechanism of down-modulation of lipid kinase activity by S608 phosphorylation are not well described. We have assessed the possibility that the increased lipid kinase activities of oncogenic mutants of PI3Kα could be partly due to alterations in the phosphorylation of S608. Rather than make point mutations in S608, which can potentially subtly alter the structure and thus the activity of p110α/p85α heterodimers, we have manipulated the levels of phosphorylation of S608 of highly-purified, recombinant p110α/p85α in vitro and tested the effect on lipid kinase activity. Neither complete dephosphorylation nor a high percentage occupancy of S608 by a phosphate group significantly altered the lipid kinase activity of wild-type PI3Kα. The levels and kinetics of S608 phosphorylation in two oncogenic mutants, E545K and H1047R, were not significantly different to that of wildtype p110α/p85α and the lipid kinase activities of mutant PI3Kα were similarly unaffected by phosphorylation of S608. This suggests that phosphorylation of S608 is not a significant regulator of PI3Kα lipid kinase activity.

Expression, purification and characterisation of recombinant wild-type and mutant PI3Kα
The same strategy that was used to express and purify active, recombinant bovine p110α/p85α [18] was used to generate human, C-terminally EE-tagged, full length, wild-type p110α/p85α in Sf9 insect cells using the Bac-N-Blue baculovirus system. When Sf9 cells were coinfected with baculoviruses encoding both subunits, p85 was expressed at higher levels than p110αEE, therefore to obtain purified p110αEE/p85α, affinity chromatography using an antibody directed against the epitope tag (EE mAb) was used to capture the enzyme complex but not excess p85α. Competitive elution with an EE-tag peptide (EYMPME), followed by anion exchange chromatography were performed as described [18]. PI3K purified in this way comprises a homogeneous 1:1 complex of the p110α and p85α subunits as assessed by both SDS-PAGE ( Figure 1A) and analytical size exclusion chromatography [18]. Recombinant p110αEE/p85α was estimated to be > 95% pure based on densitometry of the relative levels of all Coomassie blue-stained bands in each lane and no peptides other than those derived from p110αEE or p85α were detected by LC-MS/MS of tryptic digests of the purified recombinant complex. We have also expressed and purified to homogeneity EE-tagged complexes of p85α with p110α containing two of the most common cancer-associated mutations from two different regions of p110α (E545K in exon 9 and H1047R in exon 20) ( Figure 1A).
As shown previously [15,19], recombinant forms of the p110αEE E545K /p85α and p110αEE H1047R /p85α complexes have higher lipid kinase activities than p110αEE WT / p85α. The H1047R mutant had higher activity than the E545K mutant ( Figure 1B). Kinetic parameters were derived for recombinant wild type and mutant p110αEE/ p85α complexes by varying the concentration of ATP and measuring initial reaction rates (Table 1), and showed that, when PI was used as the substrate, the K m for H1047R and E545K were similar to wild type. When PI-(4,5)-P 2 was used as a substrate, the K m for H1047R was decreased approximately 2-fold compared to wild type, but that for E545K was not significantly different. Both mutants showed a significant 3 to 4-fold increase in V max when either PI or PI-(4,5)-P 2 was used as a substrate, with the V max for H1047R being significantly higher than that for E545K ( Table 1). The kinetic characteristics of purified, recombinant, EE-tagged forms of wild-type and mutant PI3Kα are therefore essentially the same as those previously reported for purified, recombinant His-tagged forms of PI3Kα [15,19] and endogenous wild-type and mutant PI3K [4,5].
Effect of dephosphorylation of p85α S608 on PI3Kα lipid kinase activity S608 in the p85 subunit is a substrate for the protein kinase activity of PI3Kα [17,20]. Phosphorylation of S608 by p110 is reported to down-regulate PI3Kα lipid kinase activity [17,20]. This raises the possibility that higher lipid kinase activities of oncogenic mutants of PI3Kα could be partly due to alterations in the levels of phosphorylation of S608.
To better compare lipid kinase activities of wild-type and mutant PI3Kα that were phosphorylated or unphosphorylated at S608, we opted to manipulate phosphorylation levels of recombinant p110αEE/p85α complexes in vitro rather than to make point mutations in S608. To perform lipid kinase assays with dephosphorylated PI3Kα, it was necessary to find a phosphatase that dephosphorylated p85α but not inositol lipids. We compared three different protein phosphatases (Figure 2A). We have previously used Antarctic Phosphatase to dephosphorylate affinity-precipitated proteins for characterisation on 2D gels [21], however Antarctic Phosphatase was inactive at temperatures >4°C that are required for significant lipid and protein kinase activity of PI3Kα, and so had no effect on levels of phosphorylation of p85α or PI at room temperature ( Figure 2A). Alkaline phosphatase is a broad spectrum phosphatase that was able to dephosphorylate both S608 and PI-3-P. In contrast, recombinant λ phosphatase dephosphorylated S608 but not PI-3-P ( Figure 2A). λ phosphatase was therefore added to PI kinase assays to continually remove phosphate groups from S608 and allow comparison of the lipid kinase activity of phosphorylated and dephosphorylated PI3Kα.
As expected, inclusion of λ phosphatase in kinase assays resulted in no significant incorporation of 32 P-labelled phosphate groups into p85α, demonstrating that levels of p85α phosphorylation were low in the presence of λ phosphatase ( Figure 2B, upper panel). Dephosphorylated p110αEE WT /p85α did not have a significantly different PI kinase activity to p110αEE WT /p85α that became phosphorylated on p85α over the course of the assay ( Figure 2B, lower panel). Similarly, the E545K and H1047R mutants did not have significantly different PI kinase activities when p85α was either dephosphorylated or phosphorylated, although the PI kinase activities of p110αEE E545K /p85α and p110αEE H1047R /p85α were higher than that of p110αEE WT /p85α as expected ( Figure 2B, lower panel). Interestingly, as observed previously [15], the protein kinase activities of p110αEE E545K /p85α and p110αEE H1047R /p85α were not significantly different to that of p110αEE WT /p85α ( Figure 2B, upper panel).
It is possible that the stoichiometry of phosphorylation of S608 remains low after the 40 minute kinase reaction used, and thus lipid kinase activities of dephosphorylated and only partially phosphorylated p110αEE/p85α were compared in Figure 2B. We therefore performed a timecourse to determine whether a difference in activity of phosphorylated and dephosphorylated p110αEE/p85α became apparent with increasing levels of phosphorylation of p85α. Levels of phosphorylation of S608 became saturated after approximately 30 minutes ( Figure 2C, upper panel). In some assays, the level of phosphorylation of the E545K mutant was apparently higher than wild-type PI3Kα or the H1047R mutant, although this was an inconsistent observation and was not significantly different in most assays. When λ phosphatase was added, lack of incorporation of 32 P into p85α demonstrated that λ phosphatase continued to dephosphorylate S608 throughout the assay. ( Figure 2C, lower panel, closed symbols). Despite apparently saturating phosphorylation of S608 between 40 and 120 minutes, the difference in the PI kinase activity of phosphorylated Michaelis-Menten kinetic parameters (substrate affinity (K m ) and maximum reaction velocity (V max )) for the production of PI-3-P or PI-(3,4,5)-P 3 were calculated from assays in which the concentration of ATP was varied. c K m and V max are shown as mean ± SEM, n = 4. and dephosphorylated wild-type, E545K or H1047R PI3Kα did not increase over time ( Figure 2C, lower panel), suggesting that phosphorylation of S608 does not regulate the PI kinase activity in either wild-type or mutant PI3Kα.
Effect of phosphorylation of p85α S608 on PI3Kα lipid kinase activity To ensure that the apparent lack of difference in PI kinase activity of phosphorylated and dephosphorylated PI3Kα was not due to a low stoichiometry of phosphorylation, even when levels were apparently saturated, we allowed the phosphorylation reaction to proceed for 16-24 hours using unlabelled ATP or no ATP for the mock phosphorylation reaction. Phosphorylated and mock-phosphorylated PI3Kα were buffer exchanged to remove excess cold ATP, then the extent of phosphorylation of S608 was measured using MALDI-MS of tryptic digests of wild-type and mutant PI3Kα. Overlapping tryptic peptides containing S608 were observed at m/z 3483.4 and m/z 3611.5 in mock or unphosphorylated PI3Kα ( Figure 3A, lower spectrum in each panel). Phosphorylation of PI3Kα led to the appearance of peptides at m/z 3563.5 and m/z 3691.6 ( Figure 3A, upper spectrum in each panel), which are 80 mass units greater than the masses of the peptides containing nonphosphorylated S608 and suggest that these peptides incorporated one phosphate group per peptide. There was no evidence for phosphorylation of peptides other than those containing S608. To estimate the extent of phosphorylation of S608, we normalised the spectra to the height of a peptide at m/z 3426.6 that is not altered when PI3Kα is phosphorylated, and compared the heights of the peaks of non-phosphorylated S608-containing peptides. Phosphorylation of p110αEE WT /p85α led to almost complete disappearance of the peaks at m/z 3483.4 and m/z 3611.5 corresponding to nonphosphorylated S608, suggesting that at least 90% of p110αEE WT /p85α was phosphorylated on S608 ( Figure 3A, upper spectrum of panel 1). In support of this, the amount of 32 P incorporated into p85α in a subsequent protein kinase time-course assay was negligible, suggesting that nearly all S608 residues were already occupied by an unlabelled phosphate group ( Figure 3B, upper panel). The extent of phosphorylation of p85α S608 in p110αEE E545K /p85α and p110αEE H1047R /p85α was less than that observed for p110αEE WT /p85α, as the heights of the peaks at m/z 3483.4 and m/z 3611.5 were less decreased by phosphorylation compared to the height of the same peaks in mock-phosphorylated p110αEE E545K / p85α and p110αEE H1047R /p85α ( Figure 3A, panels 2 & 3), suggesting that a small proportion of mutant PI3Kα remained unphosphorylated after 16 hours. In agreement, low levels of 32 P were incorporated into p85α in subsequent protein kinase time-course assays ( Figure 3B, upper panel), suggesting that a small proportion of S608 in the E545K and H1047R mutants was not occupied by an unlabelled phosphate group. Nevertheless, the majority of p110αEE E545K /p85α and p110αEE H1047R /p85α was phosphorylated on S608 under these conditions. Figure 3 Phosphorylation of p85α S608 does not alter PI kinase activity. 25 μg purified, recombinant p110αEE WT /p85α, p110αEE E545K /p85α or p110αEE H1047R /p85α was incubated with (phosphorylated) or without (mock) 1 mM ATP for 16 hr. Excess ATP was removed by buffer exchange into TBS containing 10 mM 2-mercaptoethanol then phosphorylated or mock phosphorylated PI3Kα was concentrated to approximately 0.5 mg/ml. A. MALDI-MS spectrum from m/z 3400-3750 of a Fe 3+ IMAC-enriched tryptic digest from phosphorylated (upper spectrum) or mock phosphorylated (lower spectrum) p110αEE WT /p85α, p110αEE E545K /p85α and p110αEE H1047R /p85α. The intensity scale for these spectra were normalised to the peak at m/z 3424.5, which corresponds to a peptide encompassing residues 35-66 from p85α ( 35 GSLVALGFSDGQEAKPEEIGWLNGYNETTGER 66 ) that is not phosphorylated by the protein kinase activity of p110αEE/p85α. Serine 608 from p85α was observed within 2 overlapping peptides of m/z 3483.4 corresponding to residues 594-622 from p85α ( 594 LNEWLGNENTEDQYSLVEDDEDLPHHDEK 662 ) and m/z 3611.5 corresponding to residues 593-622 from p85α ( 593 KLNEWLGNENTEDQYSLVEDDEDLPHHDEK 662 ). Incubation with ATP resulted in the appearance of peptides of masses m/z 3563.5 and m/z 3691.6 which are 80 mass units (the mass of a phosphoryl-group) greater than m/z 3483.4 and m/z 3611.5. B. 250 ng of phosphorylated (closed symbols) or mock phosphorylated (open symbols) p110αEE WT /p85α (circles), p110αEE E545K /p85α (triangles) or p110αEE H1047R /p85α (squares) were assayed for PI (lower panel) or protein (upper panel) kinase activities as described above.
The PI kinase activities of phosphorylated and mockphosphorylated wild-type PI3Kα were not significantly different over a 2 hour time-course. Similarly, the PI kinase activities of phosphorylated and mock-phosphorylated E545K and H1047R PI3Kα were not significantly different, although the activities of H1047R, and to a lesser extent E545K, were higher than that of wild-type PI3Kα. Nearly stoichiometric phosphorylation of S608 therefore did not affect the PI kinase activity of wild-type or mutant PI3Kα, further suggesting that phosphorylation of S608 does not regulate PI kinase activity.

Discussion
As reported previously for bovine PI3Kα [18], we have successfully expressed and purified recombinant complexes of full-length human p110α and p85α, the prototypic form of Class 1A PI3K using the strategy of selecting for the of p110αEE/p85α complex by placing a 6 amino acid epitope tag at the C-terminus of the p110α, the subunit with limiting expression levels. The addition of C-terminal EE tag to p110α did not appear to affect PI3Kα, as recombinant p110αEE/p85α had high stability and the expected activity and substrate specificity. High purity recombinant PI3Kα allowed us to assess the relative phosphoinositide lipid and protein kinase activities of wild-type and two common, tumourassociated mutants of PI3Kα.
Purified, recombinant oncogenic mutant forms of PI3Kα had higher lipid kinase activities compared to wild-type PI3Kα ( Figure 1B) as expected [4][5][6]13]. The ATP K m was not significantly different between wildtype and mutant forms of PI3Kα (Table 1) as previously reported [19]. However, there was a significant difference in ATP V max when either PI or PI-(4,5)-P 2 was used as a substrate, in agreement with a previous study [15]. This suggests that mutant forms of PI3Kα phosphorylate and turn over inositol lipids more rapidly, rather than having inherent higher substrate affinities, thus generating increased levels of PIP 3 and potentially partially explaining increased downstream signalling when the PI3K pathway is activated.
PI3Kα also has protein kinase activity and has been reported to phosphorylate a number of protein substrates [22,23] as well as its own p85α regulatory subunit on S608 [16,17,20]. The role of S608 phosphorylation in the regulation of PI3Kα lipid kinase activity remains controversial. Despite previous reports that phosphorylation of S608 down-regulates the lipid kinase activity of the p110α catalytic subunit, we did not find any evidence to support this by in vitro manipulation of the levels of phosphorylation of S608 of purified, recombinant p110αEE/p85α using either a protein phosphatase or saturating pre-phosphorylation. The reason for this discrepancy is not clear. In two studies, semi-purified recombinant PI3Kα was used to show phosphorylation of S608 reduces the lipid kinase activity of p110α [17,20]. Semi-purified PI3Kα can be unstable, thus it is possible that the apparent decrease in the lipid kinase activity of semi-purified PI3Kα over time was due to p110α degradation rather than S608 phosphorylation. A number of other studies have not observed a decrease in lipid kinase activity of recombinant, purified or immunoprecipitated PI3Kα in time-course assays [24][25][26][27].
The time-course of PIP 3 formation by endogenous PI3K in cells shows an initial increase, followed by a decrease in PIP 3 levels [28,29], but this is largely due to the action of inositol 3-and 5-phosphatases that directly dephosphorylate PIP 3 to form PI-(4,5)-P 2 or PI-(3.4)-P 2 respectively [30,31]. The phosphorylation of S608 in cells increases upon stimulation with insulin or PDGF [16], which mirrors activation of the lipid kinase activity of PI3Kα, thus it is not clear whether phosphorylation of S608 down-regulates endogenous PI3Kα activity. Mutation of S608 to a non-phosphorylatable residue has been shown to decrease lipid kinase activity rather than increase it, as would be expected if mutation to alanine prevented phospho-S608-induced down regulation of PI kinase activity [16], suggesting that mutation of this residue, which resides in the interSH2 domain of p85α, affects PI3Kα lipid kinase activity by altering the structure of PI3Kα rather than by affecting the phosphorylation status.
Phosphorylation of S608 has also been reported to lead to dissociation of the p85α and p110α subunits [16]. Free p110α has higher activity than p110α in complex with p85α but is very unstable and is quickly degraded [32], thus it is possible that the decrease in lipid kinase activity when p85α S608 is phosphorylated could be explained by dissociation of p110α and p85α and rapid degradation of free p110α. However, our highly purified recombinant forms of PI3Kα, when phosphorylated to saturation on p85α S608, showed no evidence of subunit dissociation by size-exclusion chromatography and no evidence of p110α degradation by SDS-PAGE (data not shown). The lack of observed dissociation, degradation or decrease in lipid kinase activity concomitant with increasing p85α S608 phosphorylation in our highly purified, recombinant PI3Kα suggests that previous reports of S608 phosphorylation resulting in down-regulation of lipid kinase activity were due to the use semi-purified recombinant PI3Kα which is non-specifically degraded over time.
A number of mechanisms that regulate PI3Kα enzymatic activity have been described, but no single event has been demonstrated to result in full activation of this enzyme. It is likely that full activation of PI3Kα is a multi-step process. Within the p110α/p85α complex, p110α is both inhibited and structurally stabilised by tight binding to the p85α subunit [32]. Activation of the PI3Kα involves protein-protein interactions that relieve the inhibition of the p110α kinase activity that is due to inter-subunit interactions of p110α with the N-terminal SH2 and inter-SH2 domains of p85α [33]. One mechanism of activation of PI3Kα is binding to specific phosphotyrosine-containing motifs (pYXXM) present in receptor tyrosine kinases (RTKs) and cytoplasmic signalling proteins, such as IRS-1, to the p85α N-and Cterminal SH2 domains [34][35][36], which disrupts the inhibitory contact between the p85α N-terminal SH2 and the p110α catalytic domain [37,38]. The E545K mutation is thought to increase activity by disrupting the p85α N-terminal SH2/p110α interface, similar to binding of RTKs [37,38]. Binding of a range of other intracellular proteins, such as activated Ras, SH3 domain-containing proteins and small GTPases, have also been reported to activate PI3Kα [39][40][41] but it is not clear whether binding of these ligands directly activates enzyme activity or whether binding results in activation by translocating PI3Kα to the plasma membrane, where its lipid substrate is located [42,43]. Tyrosine phosphorylation of p85α Y688 by Src family tyrosine kinases has also been shown to increase PI3Kα activity [44], but the mechanism is not known.
Phosphorylation of p85α S652 by PKC [45] (as well as p85α S608 by p110α) has been reported to decrease PI3Kα activity. S608 is in the inter-SH2 domain of p85α, but is not within the section that has been observed to contact p110α [37,38,46], therefore it is not clear whether phosphorylation of S608 could influence the inter-subunit interactions of p110α and p85α. A model of the structure of p110 with the N-terminal SH2, Cterminal SH2 and inter-SH2 domains of p85 [47] suggests that the C-terminal part of the inter-SH2 domain (containing S608) and the C-terminal SH2 domain has the potential to contact the catalytic domain of p110. Activation of PI3Kα due to binding of the p85α Cterminal SH2 domain to phosphotyrosine-containing proteins would be expected to be due to a conformational change that disrupts the interface between the p85α C-terminal SH2 domain and the p110α catalytic domain, analogous to the disruption of the inhibitory contact between the p85α N-terminal SH2 and the p110α catalytic domain. In contrast, for phosphorylation of S608 to decrease PI3Kα enzymatic activity, it would have to stabilise this interface and oppose the phosphotyrosine binding-induced disruption. Regulation of PI3Kα by stabilisation of inter-subunit interactions has not yet been demonstrated.

Conclusions
Studies showing a decrease in PI kinase activity of wildtype PI3Kα that correlated with increasing p85α S608 phosphorylation suggested the possibility that decreased S608 phosphorylation of oncogenic PI3Kα could be a mechanism underlying its increased PI kinase activity. This study, in which highly purified, recombinant wildtype and oncogenic, mutant PI3Kα was dephosphorylated or fully phosphorylated on serine 608 in vitro (thus avoiding point mutations which can potentially alter protein structure and activity), found no evidence that phosphorylation of S608 influenced PI kinase activity. In addition, levels of phosphorylation of oncogenic PI3Kα were not different to that of wild-type PI3Kα.

Gel electrophoresis
Recombinant p110αEE/p85α complexes were separated by SDS-PAGE using 10% Tris-glycine gels. Fast Coomassie Blue staining and destaining was carried out using a microwave oven as described [51].

Protein kinase assays
Protein kinase assays were carried out in TBS containing 2 mM MgCl 2 , 2 mM MnCl 2 , 0.2 mM ATP and 5-10 μCi [ 32 P]γATP. Phosphorylated proteins were separated from free [ 32 P]γATP by SDS-PAGE as described above. Gels were fixed and stained with Coomassie Blue, then dried, exposed to a phosphor screen and analysed as described above.
Phosphatase treatment 500 ng recombinant PI3K was mixed with 2-10 U recombinant Calf Intestinal Alkaline Phosphatase, 0.5-2 U Antarctic Phosphatase or 8-40 U Lambda (λ) Protein Phosphatase (NEB) and assayed for PI kinase activity or protein kinase activity as described above.
In vitro phosphorylation of recombinant p110αEE/p85α 25 μg purified wild type or mutant p110αEE/p85α complex was incubated with or without 1 mM ATP in the presence of 20 mM Tris pH 7.5, 150 mM NaCl, 5 mM 2-mercaptoethanol, 2 mM MgCl 2 and 2 mM MnCl 2 at room temperature for 16 hr. Phosphorylated or mock phosphorylated PI3Kα was buffer exchanged into TBS containing 10 mM 2-mercaptoethanol using Sephadex G-25 (GE Life Sciences) and concentrated using centrifugal filters (Amicon Ultra15 10,000 NMWL, Millipore). 250 ng of phosphorylated or mock phosphorylated (control) PI3Kα was assayed for PI kinase activity or protein kinase activity as described above.