PKCβ Phosphorylates PI3Kγ to Activate It and Release It from GPCR Control

The GPCR-activated PI3Kγ is also a key enzyme downstream of the IgE high affinity receptor FcεRI. PKCβ-dependent phosphorylation of PI3Kγ on Ser582 is the ‘missing link’ that functions as a molecular switch to divert PI3Kγ from GPCR inputs.

Mast cells do not express p101; however, they do express the homologous adaptor protein p84 ([PI3R6]) [6], which shares 30% sequence identity with p101. Both p101 and p84 potentiate the activation of p110c by Gbc, but the p110c-p101 complex is significantly more sensitive towards Gbc, and displays an enhanced translocation to the plasma membrane as compared with p110c-p84 [7]. Although p84 is absolutely required to relay GPCR signals to protein kinase B (PKB/Akt) phosphorylation and degranulation [6], its role is not completely understood: contrary to p110c-p101, p110c-p84 requires additionally the presence of the small G protein Ras, and might operate in distinct membrane micro-domains [6,7].
Interestingly, genetic ablation of p110c blocks high-affinity IgE receptor (FceRI)-dependent mast cell degranulation in vitro and in vivo [8]. In part this is due to the fact that initial IgE/antigenmediated mast cell stimulation triggers the release of adenosine and other GPCR ligands to feed an autocrine/paracrine activation of PI3Kc, which then functions as an amplifier of mast cell degranulation. Interestingly, a substantial part of the observed PI3Kc-dependent histamine-containing granule release (ca. 40%) was found to be resistant to Bordetella pertussis toxin (PTx) pretreatment [9,10]. Furthermore, although adenosine activates PI3Kc via the A3 adenosine receptor (A 3 AR; [ADORA3]), A 3 AR null mice are still sensitive to passive systemic anaphylaxis, and degranulation in A 3 AR 2/2 bone marrow-derived mast cells (BMMCs) upon antigen stimulation remains functional [11,12]. This and the strong degranulation phenotype of PI3Kc 2/2 BMMCs suggest that GPCR signaling does not generate the full input to PI3Kc-dependent degranulation, but a GPCR-independent activation mechanism for PI3Kc has yet to be defined.
Here we identify a mechanism that activates PI3Kc independently of GPCRs: we demonstrate that (i) IgE/antigen complexes and extracellular Ca 2+ influx activate PI3Kc, (ii) PI3Kc is operationally linked to the FceRI specifically by PKCb (PRKCB), (iii) and that the phosphorylation of Ser582 located in the helical domain of p110c by PKCb leads to the dissociation of the p84 adapter to decouple phosphorylated p110c from GPCR inputs. Further we characterize the p110c-p84 interface, and delineate an activation process that seems to be conserved among class I PI3Ks.

Thapsigargin-Induced Mast Cell Activation Needs PI3Kc
A committed step in mast cell activation is the influx of extracellular Ca 2+ by store-operated Ca 2+ entry (SOCE) [13]. Thapsigargin, which inhibits the sarco/endoplasmic reticulum Ca 2+ reuptake ATPase (SERCA), causes depletion of Ca 2+ stores, triggering SOCE. The latter achieves full-scale degranulation of BMMCs [14]. Surprisingly, BMMCs devoid of the p110c catalytic subunit of PI3Kc lost their responsiveness to thapsigargin and matched degranulation responses attained by wortmannin-pretreated cells ( Figure 1A). To investigate if thapsigargin-triggered, p110c-dependent degranulation involved release of adenosine, BMMCs were preincubated with adenosine deaminase (ADA) ( Figure 1B) to convert adenosine to inosine, which has a very low affinity for adenosine receptors. ADA attenuated degranulation induced by IgE/antigen but did not affect thapsigargin-stimulated degranulation in wild type cells and did not further attenuate γ β γ γ γ Figure 1. Thapsigargin-induced mast cell degranulation requires PI3Kc, but not GPCR signaling. (A) Granule release of wild type and p110c 2/2 BMMCs was determined detecting b-hexosaminidase (b-Hex) release into extracellular media. BMMC stimulation with IgE/antigen was initiated with the antigen (Ag, DNP-HSA at 10 ng/ml; 100 ng/ml IgE overnight). Alternatively, BMMCs were stimulated by the addition of thapsigargin (1 mM). Where indicated, BMMCs were preincubated for 15 min with 100 nM wortmannin. Released b-Hex was quantified 20 min after stimulation, and is expressed as mean 6 standard error of the mean (SEM) (n = 3; p-values in all figures are * or &: p,0.05, **: p,0.005; ***: p,0.0005; * depict here comparison with respective wild type control; & refer to comparison of untreated versus treated samples). (B) Granule release was assessed as above, but ADA (10 units/ml) was added to BMMC suspensions 1 min before stimulation where depicted. (C) Wild type or p110c 2/2 BMMCs were stimulated with adenosine (Ade; 1 mM) or thapsigargin (1 mM) for 2 min, and phosphorylation of PKB/Akt on Thr308 (pPKB), total PKB and p110c was analyzed by Western blotting. BMMCs were incubated in starving medium (2% FCS, without IL-3) for 3 h before stimulation, and pretreated with ADA where indicated. (D) Heterotrimeric Ga i proteins were inactivated by preincubation of wild type and p110c 2/2 BMMCs with 100 ng/ml PTx for 4 h, before thapsigargin (Tg) or adenosine was added as in (C). doi:10.1371/journal.pbio.1001587.g001

Author Summary
Phosphoinositide 3-kinases (PI3Ks) are involved in most essential cellular processes. Class I PI3Ks are heterodimers: class IA PI3Ks are made up of one of a group of regulatory p85-like subunits and one p110a, p110b, or p110d catalytic p110 subunit, and are activated via binding of their p85 subunit to phosphorylated tyrosine receptors or their substrates. The only, class IB PI3K member, PI3Kc, operates downstream of G protein-coupled receptors (GPCRs). Recent work suggested that PI3Kc also operates downstream of IgE-antigen complexes in mast cell activation, but no mechanism was provided. We show that clustering of the high-affinity IgE receptor FceRI triggers a massive calcium ion influx, which leads to PKCb activation. In turn, PKCb phosphorylates Ser582 of the PI3Kc catalytic p110c subunit's helical domain. Downstream of GPCRs, p110c requires a p84 adapter to be functional. Phosphomimicking mutations at Ser582 disrupt the p84-p110c interaction, and cellular Ser582 phosphorylation correlates with the loss of p84 from p110c. Thus our data suggest that PKCb phosphorylates and activates p110c downstream of calcium ion influx, while simultaneously disconnecting the phosphorylated p110c from GPCR signaling. Exploration of the p84-p110c interaction surface by hydrogen-deuterium exchange mass spectrometry confirmed that the p110c helical domain forms the main p84-p110c contact surface. Taken together, the results suggest an unprecedented mechanism of PI3Kc regulation. residual degranulation in p110c null BMMCs. Likewise, presence of ADA did not reduce the phosphorylation of PKB/Akt in response to thapsigargin via the p110c-dependent pathway ( Figure 1C) but did reduce phosphorylation of PKB/Akt in response to adenosine-illustrating that the added ADA removes adenosine quantitatively. PTx treatment of BMMCs ( Figure 1D) blocked adenosine-but not thapsigargin-stimulated PKB/Akt phosphorylation.
To exclude that autocrine/paracrine signaling to p110c occurred through PTx-insensitive Gaq subunits to phospholipase b (PLCb [PLCB2, PLCB3]), and a subsequent Ras activation by the Ras guanine nucleotide exchange factor RasGRP4 as described earlier in neutrophils [15], we used platelet activating factor (PAF) to trigger cyclic AMP-responsive element-binding protein (CREB) phosphorylation. PAF was reported earlier to trigger a PTx-insensitive Ca 2+ release from mast cells [9], and induced here a robust CREB phosphorylation comparable to adenosine and IgE/antigen. In contrast, PAF failed to trigger phosphorylation of PKB/Akt by itself, and did not enhance signaling of IgE/antigen to PKB/Akt ( Figure S1).
Altogether, these results clearly illustrate that thapsigargin stimulates BMMCs via a PI3Kc-dependent activation pathway, which operates separately from adenosine-induced activation of Gai/o trimeric G proteins.

PKCb Links Ca 2+ Mobilization to PI3Kc Activation
Protein kinase C (PKC) inhibitors (Ro318425, Gö6983, Gö6976) targeting classical and atypical PKCs, and the inhibitor PKC412, which mainly inhibits classical PKCs, all substantially blocked PKB/Akt phosphorylation in response to thapsigargin and phorbol 12-myristate 13-acetate (PMA) (Figures 3A and S2A). Rottlerin, with a limited selectivity for PKCd, had no effect on PKB/Akt activation. GPCR-dependent PI3Kc activation by adenosine was resistant to all tested PKC inhibitors ( Figure  S2B). The inhibitor profile suggested that a classical PKC activates PI3Kc. While PKB/Akt activation by PMA and thapsigargin was blocked in PKCb 2/2 BMMCs ( Figure 3B), signaling in PKCa 2/2 and PKCc 2/2 BMMCs remained intact ( Figure S2C). Deletion of PKCb eliminated phosphorylation of PKB/Akt on Thr308 and Ser473 completely, whereas a residual signal on Ser473 was observed after PI3K-inhibition by wortmannin. This may be explained by the observation that PKCb2 can function as a Ser473 kinase [18]. Adenosine, IL-3, and stem cell factor (SCF)induced PKB/Akt activation was not affected by elimination of PKCb ( Figure 3C), demonstrating that PKCb does not relay adenosine signals to PI3Kc, and is not required in cytokine and growth factor receptor-dependent activation of class IA PI3Ks in mast cells.
The direct measurement of phosphoinositides in BMMCs confirmed that ablation of PKCb or its inhibition eliminated production of PtdIns(3,4,5)P 3 triggered by thapsigargin and PMA, but not adenosine ( Figure 3D-3F). Interestingly, the link between PKCb and PI3Kc seems to be transient in nature, as PMA stimulation triggers short lived PtdIns(3,4,5)P 3 peaks ( Figure 3F). Impaired FceRI-triggered degranulation has been reported in both p110c 2/2 [9] and PKCb 2/2 BMMCs [19], and the sensitivity of degranulation to PKC inhibition fits the phospho-PKB/Akt output ( Figure 3G). This, combined with the similarity of p110c and PKCb null phenotypes in IgE/antigen-induced degranulation ( Figure 3H), suggests a direct link of PKCb and PI3Kc downstream of FceRI.

PKCb Binds and Phosphorylates PI3Kc
Co-expression of p110c with tagged full length or truncated PKCb2 ( Figure 4A) revealed that only the catalytic domain fragment and a pseudo-substrate deletion mutant of PKCb2 formed complexes with p110c ( Figure 4B), suggesting that the presence of the pseudo-substrate in PKCb results in a closed conformation that is unable to interact with p110c. An in vitro protein kinase assay with recombinant PKCb2 and glutathione Stransferase (GST)-tagged wild type p110c or catalytically inactive p110c (KR; Lys833Arg mutant) as substrate, showed that PKCb robustly phosphorylated p110c ( Figure 4C). The capability of p110c to auto-phosphorylate [20] was not required in the process.
Analysis of phosphorylated, catalytically inactive p110c by liquid chromatography tandem mass spectrometry (LC-MS/MS) identified Ser582 as a target residue of PKCb (YES P [582]LKHPK; spectra in Figure S3). Mass spectrometric multiple reaction monitoring (MRM) ( Figure S3C) showed that Ser582 phosphorylation was absent in assays lacking PKCb, or when PKC-inhibitor was added ( Figure 4D). Ser582 phosphorylation was also detected by MRM in PMA and IgE/antigen stimulated BMMCs ( Figure 4D, lower panel).
Although the extended peptide around Ser582 scores as a PKC substrate site, the core Arg-X-X-Ser582 sequence is a putative recognition site for several protein kinases (scores are PKC.protein kinase A.calcium/calmodulin-dependent kinases [CAMK]). As mast cell activation is accompanied by a massive influx of extracellular Ca 2+ , we assessed if CAMK could phosphorylate p110c directly. In the presence of 32 P-c-ATP, recombinant CAMKII (CAMK2) incorporated equal amounts of phosphate into free and p84-bound p110c. In the same experiment, PKCb preferentially phosphorylated free p110c, illustrating a preference of PKCb for p110c surfaces obscured in the p84-p110c complex. CAMK also substantially phosphorylated p84, which was borderline in in vitro assays with PKC (12 versus 3 mol %) ( Figure S5A). Probing the phosphorylation of Ser582 in vitro demonstrated that access to this site is blocked when p84 is bound to p110c ( Figure  S5B).
In cellular assays stimulating BMMCs with thapsigargin, CAMK activation could be monitored using anti-phospho-CAMKII antibodies. While PKC inhibitors left CAMKII phosphorylation .50% intact, Ser582 and PKB/Akt phosphorylation were both reduced to background levels ( Figure S6), even in a context favoring Ca 2+ -triggered responses. The above, and the fact that PKCb 2/2 BMMCs showed a major reduction in phospho-Ser582 after PMA or IgE/antigen stimulation ( Figure 5C and 5D), illustrate that phosphorylation of p110c is mainly mediated by PKCb, and that other PKC isoforms (see also Figure S2) and Ca 2+ -dependent kinases attribute to less than 18% of the observed overall signal.

Phosphorylation of Ser582 Positively Regulates p110c's Activity and Displaces p84
To evaluate if the Ser582 phosphorylation affected the intrinsic activity of p110c, a phosphorylation-mimicking mutant (Ser582-Glu) was produced. The activity of the p110c Ser582Glu mutant was enhanced approximately 2-fold independently of the substrate used (PtdIns(4,5)P 2 , PtdIns, or auto-phosphorylation) ( Figure 6A-6C). Ser582 is localized in the helical domain of p110c, and it is interesting to note that helical domain mutants of p110a found in tumors display a similar increase in enzyme activity [21][22][23][24][25].
As mutations in the helical domain of p110a attenuate contacts with the p85 regulatory subunit [21], we examined binding of p110c mutants to the PI3Kc adaptor subunit p84: the substitution of Ser582 with Glu and Asp, abrogated p110c-p84 interactions in HEK293 cells and BMMCs, while the Ser582Ala replacement favored p110c-p84 complex formation ( Figure 6D-6F). In line with this, PMA-induced phosphorylation of Ser582 in BMMCs was suppressed by the overexpression of p84 ( Figure 6G), which fits the very limited access of PKCbII to in vitro phosphorylate Ser582 in the p110c-p84 complex (reduced to 20% of phosphorylation of free p110c) ( Figure S5B).
Most importantly, the correlation of phosphorylation of Ser582 on p110c and the release of p84 could also be established in wild type BMMCs: when stimulated with PMA, the amount of p84 that could be co-precipitated with p110c was reduced significantly, and was linked to Ser582 phosphorylation of p110c. In the inverse coimmunoprecipitation, anti-p84-associated p110c was reduced, and phosphorylation was below detection levels in the remaining p84associated p110c ( Figure S7). The collected results are in agreement with a mechanism in which PKCbII-mediated phosphorylation of Ser582 and the interaction of p84 and p110c are exclusive events, and in which PKCb action displaces p84 from p110c.

The Helical Domain of p110c Binds and Is Stabilized by p84
In order to understand how p84 could mask Ser582 phosphorylation, and to map the p110c-p84 contact interface, hydrogen deuterium exchange mass spectrometry (HDX-MS) was used. HDX-MS elucidated contacts of class IA p110d with its p85 regulatory subunit, and the mechanism of action of cancer-linked mutations in p110a [25,26]. HDX-MS relies on amide hydrogen exchange with solvent at a rate dependent on their involvement in secondary structure and solvent accessibility. Following proteolysis, location and extent of deuterium uptake are analyzed by peptide mass determination. The primary sequence of p110c was covered .90% by 202 peptide fragments ( Figure S8; Table S1).
Deuterium ( 2 H) incorporation into free and p84-complexed p110c was analyzed at seven time points (3 to 3,000 s). Differences in 2 H-exchange of free and complexed p110c were mapped onto the crystal structure of p110c lacking the N-terminal domain (PDB ID:2CHX, residues 144-1,093), to visualize conformational changes induced by p84 ( Figures 7A, S9, and S10). Peptides with highest decrease in 2 H incorporation (.1.0 Da) clustered to the RBD-C2 linker, the C2-helical domain linker, and the helical domain. The 2 H-incorporation in the presence of p84 is visualized as integrated average difference in exchange at all seven time points in Figure 7B, illustrating that the helical domain provides the dominant interface with p84. Due to difficulties in producing free p84, the contacts on p84 with p110c could not be mapped. Interestingly, in the absence of the p84 subunit, the majority of peptides in the helical domain exhibited broad isotopic profiles (HDX of peptide 623-630, which is representative of peptides in the helical domain, is shown in Figures 7C and S11). This type of profile known as type 1 exchange (EX1) kinetics is indicative of concerted dynamic motions of a substructure in a protein, rather than the local fluctuations characteristic of EX2 kinetics [27].
The N-terminus of p110c was shown to stabilize the p110c-p101 heterodimer [28]. Expression of p110c mutants lacking the first 130 amino acids (D130-p110c) seemed to support this view, as association with p84 was lost ( Figure 7D). However, when truncated p110c was N-terminally tagged with GST (GST-D130-p110c), binding of p84 was restored. Although we detected a small decrease in the 2  interaction seems to be dispensable. The N-terminus of p110c instead has a role in stabilizing the intact catalytic subunit. The helical domain is the main location of interaction with p84; however, it appears that this interaction is vulnerable and easily broken, as a single phosphorylation at Ser582 is able to disrupt the contact.
Non-GPCR-mediated activation of PI3Kc has not been reported so far, but it has been shown that phorbol esters and Ca 2+ ionophores can modulate phosphoinositide levels in a variety of cells, including platelets [37], adipocites [38], fibroblasts [39], and hematopoietic cells [40]. The proposed mechanisms have Where indicated (IgE), BMMCs were loaded with IgE (100 ng/ml) overnight. PI3Kc was immunoprecipitated from cell lysates with an anti-PI3Kc antibody, before precipitated protein was probed for phosphorylated p110c (pp110c) with a phospho-specific anti-pSer582 antibody (validation of antibody see Figure S4). PI3Kc phosphorylation is shown normalized to total PI3Kc levels (mean 6 standard error of the mean [SEM], n = 3; * depict analysis using unstimulated control. & reference point is IgE only). (B) IgE/antigen-induced Ser582 phosphorylation of p110c requires Ca 2+ influx. Cells were stimulated as in (A), but exposed to EDTA, EGTA, or loaded with BAPTA/AM where indicated (see Figure 2). Phosphorylated p110c was detected as in ( been diverse and involved protein tyrosine kinases and GPCR signaling. A recent finding that protein kinase D (PKD) can phosphorylate two distinct sites on the p85 regulatory subunit to control class IA PI3K activity [41] is an indication that PI3K control is more complex than anticipated.
PI3Kc has been shown to be a key element in enhancing IgE/ antigen output by the release of adenosine. This process involves signaling downstream of Gai-coupled A 3 AR, and is sensitive to PTx and ADA [9]. The resistance of thapsigargin-induced degranulation to ADA shown here, and the fact that PTx did not diminish the PI3Kc-dependent, thapsigargin-induced phosphorylation of PKB/Akt, points to a novel mechanism of PI3Kc activation, which is clearly distinct from GPCR action. This Ca 2+mediated PI3Kc activation requires SOCE and [Ca 2+ ] i .600 nM. In contrast, GPCRs yield phosphorylation of PKB/ Akt in mast cells even in the absence of a change in [Ca 2+ ] i . Furthermore, increases in [Ca 2+ ] i triggered via GPCRs remain at levels incapable of engaging a Ca 2+ -dependent activation of PI3Kc.
Thapsigargin bypasses the signaling chain from IgE/antigenclustered FceRI to the activation of phospholipase Cc (PLCc) and inositol(1,4,5)-trisphosphate (Ins(1,4,5)P 3 ) production, and triggers SOCE by the depletion of Ca 2+ stores. That thapsigargin requires functional PI3K activity to induce mast cell degranulation was first demonstrated using the PI3K inhibitors wortmannin and LY294002 [14], but no link between [Ca 2+ ] i rise and PI3Kc activity was established previously. The fact that inhibitors for classical PKCs only prevented the thapsigargin-and PMAinduced phosphorylation of PKB/Akt, while the GPCR-mediated phosphorylation of PKB/Akt remained intact, points to a link between classical PKCs and PI3Kc. Experiments on BMMCs lacking PKCa, PKCb, and PKCc, showed that only the PKCb null cells lost the ability to activate PKB/Akt in response to thapsigargin or PMA stimulation. As the PMA-and thapsigargininduced phosphorylation of PKB/Akt on Ser473 showed a partial resistance to wortmannin, and as it has been reported that PKCb can directly phosphorylate Ser473 in the hydrophobic motif of PKB/Akt [18], the effect of genetic and pharmacological targeting of PKCb was also validated measuring PtdIns(3,4,5)P 3 production directly. The lack of PtdIns(3,4,5)P 3 production in thapsigargin or PMA-stimulated BMMCs treated with the PKC inhibitor PKC412, and in cells devoid of PKCb, is in agreement with a requirement of PKCb upstream of PI3Kc. That signaling from PKC to PI3K plays a role in mast cell degranulation is further supported by the close correlation of PKC inhibitor sensitivity of phosphorylated PKB/Akt and degranulation responses. Moreover, the loss of PKCb or PI3Kc results in a similar reduction of degranulation over a wide range of IgE/antigen concentrations. The results obtained here are in agreement with previous findings that mast cells and mice devoid of p85a/p55a/p50a [42,43] and p85b [44] remain fully responsive to IgE/antigen complexes. A previous report showed a biphasic activation of PI3K with PI3Kc having an early role and PI3Kd a later role downstream of FceRI in murine mast cells [45,46].
A mechanistic link between PKCb and the catalytic subunit of PI3Kc was initially difficult to establish, as the direct PtdIns(3,4,5)P 3 response to PMA-stimulation was transient (main peak half life ,1 min), and the two full-length enzymes interacted only weakly. The observation that truncated, activated forms of PKCb formed stable complexes with p110c, suggested that PKCb must attain an open conformation to interact with p110c, and that PKCb binds to p110c via its catalytic domain. This contact resulted in phosphorylation of Ser582 on p110c, which could be detected both in vitro and in PMA or IgE/antigen-stimulated BMMCs by mass spectrometry. The PKCb-mediated phosphorylation of p110c was confirmed using site-specific anti-phospho-Ser582 antibodies. Stimuli like PMA, thapsigargin, and IgE/ antigen complexes all required PKC to signal to PKB/Akt, which correlated with the phosphorylation of Ser582 on p110c. Moreover, the phosphorylation of Ser582 on p110c was sensitive to removal of extracellular Ca 2+ , buffering of [Ca 2+ ] i , and the genetic deletion of PKCb. Adenosine stimulates PI3Kc via GPCRs and PTx-sensitive trimeric G proteins [9] in a Ca 2+independent process, and did not yield a detectable phosphorylation of Ser582.
The increased turnover of PtdIns and PtdIns(4,5)P 2 , and the increased rate of auto-phosphorylation displayed by p110c with a phosphate-mimicking mutation (Ser582Glu), suggests that a structural change in the helical domain of p110c is sufficient to increase the catalytic activity independent of the presence of the p84 subunit. Previous work examining the activation of the class IA p110a, p110b, and p110d catalytic subunits has shown that part of the activation mechanism occurs through a conformational change from a closed cytosolic form to an open form on membranes [25,26]. The helical domain of p110c is exquisitely well placed to propagate conformational changes due to the fact that it is in contact with every other domain in p110c. In the crystal structure of N-terminally truncated (D144) p110c, the side chain of Ser582 points inward [47,48], and has to rotate to accommodate a phosphate. Our HDX-MS results showed a dynamic ''breathing'' motion in the helical domain in the free p110c catalytic subunit that may allow for temporary exposure of Ser582, enabling modification by PKC. HDX results showed that the p84 subunit slowed or prevented this dynamic motion, and this correlated with a decreased efficiency of phosphorylation by PKC in cells in the presence of p84.
Although Ser582 is not in a direct contact with the kinase domain, it is structurally linked to it: the heat repeat HA1/HB1 housing Ser582, and the connecting intra-helical loop (residues 560 to 570), along with helix A3 (624-631) are in contact with helices ka9 and ka10 in the C-lobe of the kinase domain (known as the regulatory arch [49,50]) and could transduce a conformational change to the catalytic center of p110c ( Figure 7C). The phosphorylated Ser582 and phosphorylation-mimicking mutants may activate lipid kinase activity by causing a conformational shift at this interface. It has been shown recently that this region of the p110c kinase domain is critical in regulating lipid kinase activity, as phosphorylation of Thr1024 in the ka9 by protein kinase A (PKA) negatively regulates p110c activity in vitro and in cardiomyocytes ( Figure 7C) [51].
In contrast to the p110a-p85 heterodimer, stabilized by the Nterminal adaptor-binding domain (ABD)/inter SH2 domain interaction, the association of the p110c subunit with its adaptor subunit is quite vulnerable, and the Ser582Glu mutant, but not Ser582Ala, abrogated the formation of a p110c-p84 complex. HDX-MS revealed that the helical domain of p110c was stabilized by p84. Ser582 is located in the center of the p110c-p84 contact surface, which explains how a change in charge (Ser582Glu) breaks the interaction with p84, either by direct contact or by destabilization of the helical domain. Overexpression of p84 shields p110c from a PMA-induced phosphorylation, and suggests that binding of p84 to p110c, and Ser582 phosphorylation by PKCb are mutually exclusive. This implies that the two activation modes of PI3Kc-by GPCRs or PKCb-are completely separated. At low [Ca 2+ ] i , PI3Kc is exclusively activated by Gbc subunits. It has been demonstrated that a PI3Kc adapter protein is absolutely needed for functional GPCR inputs to p110c [6]. If the interaction of p84 with p110c is blocked by the phosphorylation of Ser582, p110c is decoupled from its GPCR input (for a schematic view of the process see Figure 8). The PKCb-mediated activation and phosphorylation of p110c constitutes therefore an unprecedented PI3K molecular switch, which enables the operation of p110c downstream of FceRI signaling, and will elucidate cell typespecific activation processes in allergy and chronic inflammation.

Statistical Analysis
Numeric results were tested for significance using a two-tailed Student's t test, (paired or unpaired, as imposed by datasets). * or & , ** or && , and *** or &&& refer to p-values p,0.05, p,0.005, and p,0.0005. * and & were used for comparison of different genotypes, stimuli an conditions as indicated. Calculations were carried out using Graph Pad Prism, Microsoft Excel, or Kaleidagraph software.

Deuterium Exchange Reactions
Protein stock solutions (5 ml; Hsp110c-C-His 6 : 30 mM; Hsp110c-C-His 6 /Mmp84-C-His 6 : 35 mM) were prepared in 20 mM Tris [pH 7.5], 100 mM NaCl, 1 mM ammonium sulfate, γ γ β α γ β γ γ Figure 8. Phosphorylation of Ser582-loss of GPCR coupling of p110c. In a resting mast cell, the PI3Kc complex is responsive to GPCRmediated dissociation of trimeric G proteins. An adapter protein (here p84) is required for a productive relay of the GPCR signal to PI3Kc. When FceRI receptors are clustered via IgE/antigen complexes, a signaling cascade is initiated, which triggers the depletion of intracellular Ca 2+ stores and the opening of store-operated Ca 2+ channels. The resulting increase in [Ca 2+ ] i and PLCc-derived diacylglycerol activate PKCb, which binds to p110c and subsequently phosphorylates Ser582 (Rpp110c). Phosphorylated p110c cannot interact with p84, and is therefore unresponsive to GPCR inputs. GPCR input to PI3Kc coincides with migration and adhesion, while Ca 2+ /PKCb activation of p110c occurs when mast cells degranulate. The phosphorylation of PKB/Akt occurs downstream of PtdIns(3,4,5)P 3 , which originates from G protein-activated p84-p110c complex or PKCb-activated pp110c. The phosphorylated residues Thr308 and Ser473 of PKB/Akt are used to monitor PI3K activation. More detailed effector signaling event schemes can be found in [52]. doi:10.1371/journal.pbio.1001587.g008 and 5 mM DTT. Exchange reactions were initiated by addition of 25 ml of a 98% D 2 O solution containing 10 mM HEPES (pH 7.2), 50 mM NaCl, and 2 mM DTT, giving a final concentration of 82% D 2 O. Deuterium exchange reactions were allowed to carry on for seven time periods, 3, 10, 30, 100, 300, 1,000, and 3,000 s of on-exchange at 23uC, before addition of quench buffer. Onexchange was stopped by the addition of 40 ml of a quench buffer containing 1.2% formic acid and 0.833 M guanidine-HCl, which lowered the pH to 2.6. Samples were then immediately frozen in liquid nitrogen until mass analysis. The full HDX-MS protocol can be found in Text S1. Figure S1 PAF-mediated signaling does not activate PI3K, and does not synergize with FceRI co-stimulation (related to Figure 1). IgE-sensitized (100 ng/ml IgE, overnight) or non-sensitized wild type BMMCs were IL-3 depleted for 3 h and stimulated with either 1 mM adenosine (Ade), 1 mM PAF, or 5 ng/ml antigen (post IgE-sensitization) 6 PAF for 2 min. Subsequently, cell lysates were subjected to SDS-PAGE. Phosphorylation of Ser473 in PKB/Akt (A), (B) Ser133 in cyclic AMPresponsive element-binding protein (CREB) and Ser660 in PKCbII was monitored by immunodetection with phosphositespecific antibodies (n = 3, *: p,0.05; * refers to unstimulated control). (EPS) Figure S2 Effect of PKC-inhibitors on PMA-or adenosine-induced PKB phosphorylation (S473) (related to Alignments were done by inspection of the crystal structures of PI3Kc (1E8Y), PI3Ka (3HHM), PI3Kb (2Y3A), and PI3Kd (2WXR). Secondary structure elements are labeled as indicated in the legend. S582 is colored red, while cancer-associated PI3Ka mutations are marked as blue. (TIF) Figure S4 Anti-phospho-Ser582 antibody validation (related to Figure 5). (A) Wild type BMMCs were transfected with empty vector, expression plasmid for GFP-PI3Kc wild type or the GFP-PI3Kc S582A mutant. On the next day, cells were stimulated with 200 nM PMA for 45 s, and PI3Kc was immunoprecipitated from cell lysates. Specificity of the anti-S582 antibodies was validated by Western blotting. (EPS) Figure S5 Global and site-specific in vitro phosphorylation of monomeric p110c and p110c-p84 complexes by PKCb and CamKII (related to Figure 5). (A) Equal amounts of purified recombinant p110c-His 6 (2.5 pmole) or p110c-His 6 / EE-p84 complexes were incubated with 20 pmole wortmannin to eliminate auto-phosphorylation signals. Free or complexed p110c (with p84 protein) was incubated with 10 mM ATP and 5 mCi of [ 32 P]-c-ATP, and equal specific activity of recombinant PKCbII and CamKII (Life Technologies assays: 30 pmol phosphate incorporation/min.) for 30 min at 30uC. Subsequently, proteins were denatured and separated by SDS-PAGE followed by Coomassie staining (lower panel). 32 P-incorporation was visualized by autoradiography and quantified on a phospho-imager (Typhoon 9400, middle panel). Band intensities were quantified with ImageQuant TL Software (Amersham Biosciences, top panel; n = 4, * p,0.05). Insert: quantification of p84 phosphorylation from the reactions shown in (A), (open bars, n = 4, * p,0.05). Phospho-PKCbII levels were subtracted from phospho-p84 signals (due to identical apparent Mr on SDS-PAGE). The filled bar represents phosphorylated p84-His 6 after Ni 2+ -NTA pull-down to minimize contaminating PKCbII autophosphorylation signals (n = 3). (B) Equal amounts of purified recombinant p110c-His 6 (2.5 pmole) or p110c-His 6 /EE-p84 complexes were incubated as indicated with ATP (100 mM), recombinant PKCbII and CamKII for 30 min at 30uC. Proteins were denatured, separated by SDS-PAGE followed by immunological detection of p110c and p84 protein, as well as site-specific phosphorylation of residue Ser582 (pp110c S582) in p110c. PKCbII and CamKII input was adjusted to equal protein kinase activity (30 pmol of transferred phosphate/ min). For quantification, Ser582 phosphorylation levels were normalized (n = 4, *: p,0.05, ***: p,0.0005). (EPS) Figure S6 Phosphorylation of p110c requires active PKCb (related to Figure 5). (A) Effect of PKC inhibitors on thapsigargin induced p110c Ser582 phosphorylation. IL-3 deprived BMMCs were preincubated with PKC inhibitors (1 mM Ro318425 or 1 mM AEB071) for 20 min before stimulation with 1 mM thapsigargin for 2 min. PI3Kc was immunoprecipitated from cell lysates with anti-p110c antibody (see Methods). Precipitated protein was then probed for phosphorylated p110c (pp110c) using phospho-specific anti-pSer582 antibodies. PI3Kc phosphorylation is shown normalized to total p110c levels (mean 6 standard error of the mean [SEM], n = 3; * depict comparison with stimulated control). (B) Cells were stimulated as in (A). PKB (T308) and CamKII (T286) phosphorylation was determined by Western blotting. Data are the average of three independent experiments 6 SEM. (EPS) Figure S7 Ser582 phosphorylation releases p84 from p110c (related to Figure 6). IL-3 deprived BMMCs were stimulated with 100 nM PMA for 2 min. PI3Kc complex was coimmunoprecipitated from cell lysates using either (A) anti-p110c or (B) anti-p84 antibodies. The amount of p84 co-immunoprecipitated with p110c (A) or p110c co-immunoprecipitated with p84 (B) was normalized to the total amount of p110c or p84, respectively. Data are the average of five independent experiments  Figure 7). Changes in deuteration levels were mapped onto the crystal structure of PI3Kc (PDB ID: 2CHX) as in Figure 7.The isotopic profiles of two selected peptides (579-592, 623-630) from the helical domain are shown at three or four time points of H/D on exchange +/2 the p84 subunit. In the absence of the p84 adaptor the majority of peptides in the helical domain showed broadening of the isotopic profiles indicative of EX1 kinetics (see 30 s of HDX in free p110c). The helices HB1, HA2 (579-592), and HA3 (624-631) selected are all structurally linked, with HA3 located at the interface of the helical domain with the C-lobe. Ser582 (red) and Thr1024 (yellow) have been highlighted as a reference. (TIF)

Supporting Information
Table S1 Deuterium exchange data of all analyzed peptides of PI3Kc in the absence or presence of p84 are summarized in tabular form (related to Figure 7). Percent hydrogen deuterium exchange was calculated for each of the seven time points and colored according to the legend. Data show the mean of two independent experiments. The charge state (Z), maximal number of exchangeable amides (#D), starting residue number (S), and ending residue number (E) are displayed for every peptide. (XLSX) Text S1 Extended experimental procedures, reference to animals and plasmids. Detailed description of experimental procedures, materials, and further reference to the origin of genetically modified mice used here, and a primer to the determination of deuterium incorporation (HDX_MS). (DOCX)