Structure of Lipid Kinase p110β/p85β Elucidates an Unusual SH2-Domain-Mediated Inhibitory Mechanism

Summary Phosphoinositide 3-kinases (PI3Ks) are essential for cell growth, migration, and survival. The structure of a p110β/p85β complex identifies an inhibitory function for the C-terminal SH2 domain (cSH2) of the p85 regulatory subunit. Mutagenesis of a cSH2 contact residue activates downstream signaling in cells. This inhibitory contact ties up the C-terminal region of the p110β catalytic subunit, which is essential for lipid kinase activity. In vitro, p110β basal activity is tightly restrained by contacts with three p85 domains: the cSH2, nSH2, and iSH2. RTK phosphopeptides relieve inhibition by nSH2 and cSH2 using completely different mechanisms. The binding site for the RTK's pYXXM motif is exposed on the cSH2, requiring an extended RTK motif to reach and disrupt the inhibitory contact with p110β. This contrasts with the nSH2 where the pY-binding site itself forms the inhibitory contact. This establishes an unusual mechanism by which p85 SH2 domains contribute to RTK signaling specificities.

human p110δ (UniProt O00329), human p110γ (UniProt P48736) and human p110α (UniProt P42336, bottom) were aligned using ClustalW (Thompson et al., 1994). The secondary structure elements of p110β and p110α are depicted above and below the alignment, respectively. Absolutely conserved amino acids are depicted as white letters with red background. Partially conserved residues are shown in red with blue boxes. The CBRs, Ploop, the catalytic loop and the activation loop are highlighted and labeled.
(B) Sequence alignment of murine p85β (UniProt O08908), human p85β (UniProt O00459), human p85α (UniProt P27986) and human p55γ (UniProt Q92569). The secondary structure elements of p85β are labeled on top of the alignment, whereas the elements of p85α are labeled underneath. The Tyr in the cSH2 that contacts the kinase domain of p110β is marked by a star (*). (A) Scan of the phosphor screen. The upper two rows are duplicates for the "brain lipids" vesicles. The lower two rows are duplicates for the diC8-PIP 2 /POPS vesicles (Invitrogen).
The reactions contained various PI3K constructs (at 5nM final concentration) in the presence or absence of 10 μM PDGF bis-phosphopeptide. Reactions were carried out with 1 μCi [γ 32 P]-ATP per reaction, with 100 μM ATP and 1 mg/ml total "brain lipids" or 1.5 mM diC8-PIP 2 /POPS vesicles for 60 min. The vesicle compositions are indicated.
The bars represent standard deviations of the mean for two measurements. The hatched bars are for reactions in the presence of PDGF bis-phosphopeptide.
(C) SDS PAGE analysis of the proteins used for the assays in (A and B). Samples of 10 μl of 1.7 nM protein solutions for each of the reaction were analyzed on a 4-12% NuPAGE Novex Bis-Tris gel in MES running buffer (Invitrogen). The upper band is the p110 subunit and the lower band is the regulatory subunit.
(D) Stereo view of the contacts between the p85β-cSH2 domain and the C-lobe of the kinase domain. The Tyr677-p85β (red) inserts into a hydrophobic groove between the two elbows k7/k8 (yellow) and k11/k12 (orange). Tyr677-p85 also makes a potential hydrogen bond with Ser1046-p110 (dashed black line).
Vesicles were "brain lipids" as in Figure S2. The protein concentration was 10 nM with 100 μM ATP and 1 mg/ml lipid. Reactions were carried out for 60 min. The bars show the integrated areas for the stearoyl-arachidonyl PtdIns (3,4,5) (F) Effect of p85 mutations on enzyme activity (Transcreener assay of ADP formation). Effect of mutations in the Ala-Glu-Pro-Tyr loop of the p85-cSH2 on p110β activity (normalized to the wild-type complex).
(G) Lipid kinase activity of the p110β-L1043H mutant in a complex with p85β-nicSH2 compared with the wild-type complex in the absence and presence of 10 μM PDGFR bisphosphopeptide (pY2). complexes. Protein kinase B (PKB) phosphorylation (pPKB) levels at residues Ser473 and Thr308 were quantified and normalized WT p110/p85 (WT).

Figure S7. Posttranslational Modifications of the Regulatory Subunit
The residues undergoing phosphorylation are highlighted as colored spheres (http://www.uniprot.org/uniprot/). The phosphorylated residues from all three isoforms are mapped on the structure of Mmp110β/p85β-icSH2 complex. The residue numbering for yellow and red spheres refers to p85α. The dashed line (blue) is the disordered iSH2/cSH2 linker.
The phosphorylated residues could affect the activity of the complex. For example, a conserved tyrosine in Iα1 of the iSH2 undergoes phosphorylation in all three regulatory subunits (Tyr467-Hsp85α, Tyr464-Hsp85β, Tyr199-Hsp55γ, Tyr458-Mmp85β) (as listed in UniProt entries P27986, O08908, O00459, Q92569). This tyrosine is in close proximity to both the activation loop and the C2 β7/β8 loop and its phosphorylation could potentially affect the kinase activity. Another conserved Tyr modified by phosphorylation is in the iSH2-cSH2 linker (Tyr607-Hsp85α, Tyr605-Hsp85β, Tyr341-Hsp55γ, Tyr599-Mmp85β). Interestingly, this Tyr is just before Ser608 in Hsp85α, which is known to be auto-phosphorylated by the p110 catalytic subunit, resulting in lower enzymatic activity (Foukas et al., 2004). Tyr phosphorylation may have a similar effect as Ser608 or it may influence the status of Ser608 phosphorylation.
Besides the phosphorylation sites shared by all regulatory subunits, there are several phosphorylated sites unique to p85β relative to p85α, one being Tyr467-Hsp85β in Iα1 of iSH2 that contacts the C2 β7/β8 loop. The modification of this tyrosine could potentially influence the interaction between p85-iSH2 and the catalytic domain. It will be interesting to see whether the p85 phosphorylation sites are targeted by PTEN (phosphatase and tensin homolog), a tumor suppressor commonly mutated in cancers that was shown to associate with p85 during kinase activation (Barber et al., 2006;Rabinovsky et al., 2009) and proposed to dephosphorylate p85β (He et al., 2010).

Table S1. Comparison of GDC0941 Contacts with p110 Subunits
The contacts between the bound GDC0941 and catalytic subunits are summarized for the p110β, p110δ and p110γ isotypes. An "x" marks the presence of a contact (distance less than 4 Å). Potential hydrogen bonds are also indicated.

Mmp110
Hsp110 Hsp110γ  Model of nSH2 and cSH2 release from p110 by pY phosphopeptides (pY shown as yellow spheres) . The pY binding site on the nSH2 is at the interface with p110, whereas the pY binding site on the cSH2 is exposed. Phosphopeptides having at least 5 residues following the pY are necessary to break the contact of the cSH2 with p110.

Protein Expression and Purification
Recombinant baculoviruses were generated and propagated using the Bac-to- The heterodimer was concentrated to 9 mg/ml, frozen in liquid nitrogen and stored at -80˚C.

Cloning and Protein Purification for ΔABD-p110β
The construct of an ABD-truncated version of p110β (ΔABD-p110β) was made using a strategy described previously for generation of ΔABD-p110δ, in which a TEV protease cleavage site is inserted in the linker region between ABD and RBD (Berndt et al., 2010). The correct insertion of the TEV site was confirmed by DNA sequencing (amino acid sequence:
The crystal structure was solved by molecular replacement using PHASER (McCoy, 2007) with the previously published p110δ structure (PDB ID: 2WXG) and the iSH2 structure (PDB ID: 2RD0) as the search models and subsequently refined using REFMAC (Murshudov et al., 1997). COOT (Emsley and Cowtan, 2004) was used to manually place the known cSH2 structure (PDB ID: 1QAD) in the 2mFo-DFc electron density of the map. Refinement using BUSTER (Blanc et al., 2004) was iterated with manual re-building using COOT until the structure converged.
Final statistics for the 3.3 Å resolution model are given in Table 1. The model has three residues in the disallowed regions of the Ramachandran plot and 81.9% of residues in the most favored regions as defined by PROCHECK (Laskowski et al., 1993). The structure of p110 has residues 1-12, 228-234, 299-319, 402-431, 514-

Preparation of Liposomes
Liposomes containing a defined mixture of brain lipids (we refer to these liposomes as "brain lipids") contained 5% brain phosphatidylinositol-4,5-bisphosphate (Avanti 840046), 45% brain phosphatidylethanolamine (Avanti 830022), 15% brain phosphatidylcholine (Avanti 840053), 20% brain phosphatidylserine (Sigma P6641), 5% sphingomyelin (Sigma S0756) and 10% cholesterol (Avanti 700000). They were prepared in 2 mg batches by adding the lipid solutions into a 4 ml glass vial containing 130 μl chloroform/methanol (3:1) solution. The lipid solution was then dried to a film using an argon stream and desiccated under vacuum for 30 min. Lipids were then resuspended in 1 ml of a solution containing 20 mM Tris, 100 mM KCl, 1 mM EGTA, at a concentration of 2 mg/ml. The lipid suspension was sonicated in the glass vial with a bath sonicator for 5 min, before transferring it to a 1.5 ml Eppendorf tube and sonicating it for another 10 min. Subsequently, the liposomes were subjected to 10 cycles of freezing in liquid nitrogen followed by thawing at 42 °C for 2 min. Finally, the preparation was extruded with an Avanti Mini-Extruder 10 times through a 0.1 μm polycarbonate membrane (Nucleopore).

Transcreener assay of ADP production
The lipid kinase activity was determined using the Transcreener ADP2 FP Assay (BellBrook Labs, Madison, WI) according to manufacturer"s specifications. This non-radioactive assay measures activity by quantitating the amount of ADP formed. The ADP formed by PI3K competes with fluorescent ADP tracer for binding to an anti-ADP antibody, thereby decreasing fluorescence polarization. Because of the competitive nature of this assay, the change in fluorescence polarization signal is not a linear function of enzyme activity, and a convenient measure of enzyme activity is the EC50, or the concentration of enzyme necessary to give half-maximal inhibition of labeled ADP binding to the antibody, for a given concentration of substrate and a fixed time of reaction at initial rate. While activity of two different enzyme constructs can be compared at a fixed concentration of protein and a fixed concentration of substrate, the assay is most reliable when about 2%-10% of the ATP is converted to substrate. This means that when comparing the most active constructs with the least active, the dynamic range is so large that these limits are quickly exceeded.
Consequently, the EC50 provides a convenient way to compare the activities of various constructs. Briefly, reactions were performed in 10 μl volume in 384-well black plates  Figure 1D) or 75 nM ( Figure S3B). The reaction was carried out for 1 h at room temperature.
To determine the effect of receptor tyrosine kinases (RTK) on PI3K activity, the assays were performed in absence or presence of 10 μM pY peptides: equilibration. Fluorescence polarization was measured on a PHERAStar plus HTS microplate reader (BMG Labtech), using a fluorescence polarization module with excitation centered at 633 nm and emission at 650 nm. The data were fit in Prism (GraphPad) using a threeparameter exponential decay model. [γ

P]-ATP assays of PIP 3 production
In order to directly measure the PIP 3 production by the enzyme, we used the nitrocellulosebinding assay (Knight et al., 2007). Basically, 15 μl reactions having 0.1 μCi/μl [γ32P]-ATP, 100 μM ATP, 1 mg/ml lipid and 5 nM PI3K in a buffer containing 50 mM Hepes pH 7.5, 100 mM NaCl, 3 mM MgCl2 and 1 mM EGTA were incubated for 60 min at room temperature. A 2 μl aliquot of each reaction was then spotted onto a nitrocellulose filter. After spotting, the membrane was air-dried for 5 min and washed in 200 ml wash solution (1M NaCl, 1% phosphoric acid) for 30 sec. This was followed by five more washes with 200 ml wash buffer for 5 min each. The membrane was air-dried for 1 h, and then wrapped in a plastic wrap and exposed to a phosphoimager plate (Molecular Dynamics). The plate was exposed for 5 min and scanned by a Typhoon Scanner (GE Healthcare).

Mass Spectrometry Assays of PIP 3 Production
The PIP 3 lipids produced in PI3K in vitro assays as described above were also measured by a mass-spectrometry based assay (Clark et al.).

Cell Culture, Transfection and Cell Lysis
Phoenix cells (human embryonic kidney 293T-derived cells, Orbigen) were used to examine PKB phosphorylation levels in cells over-expressing wild-type or mutated p110/p85 complexes. Cells were cultured at 37 °C, 5% CO 2 in RPMI-1640 GlutaMAX (Invitrogen) complemented with 10% heat inactivated foetal bovine serum, 100 U/ml Penicillin, 100 μg/ml Streptomycin. Transfection was done in 6-well plates using GeneJuice transfection reagent (Novagen) as described by the manufacturer. DNA (2 μg total) was used for each transfection, always using equal amounts of the catalytic and regulatory subunit DNA when was performed using GeneTools software (Syngene) and data analyzed in GraphPad Prism.

Phospholipids Extraction
To get a more direct measurement of PI3K activity in cells, we quantified total PIP 3 levels. To extract phospholipids, 1 ml of ice-cold 1N HCl was added to cells in a 3.5 cm well. Cells were scraped, transferred to a 1.5 ml tube and centrifuged at 13000 rpm for 5 min at 4°C. Supernatant (SN) was discarded and the pellet was resuspended in 750 μl methanolchloroform-1N HCl (484:242:23.55), followed by 170 μl water. Samples were vortexed and allowed to stand for 5 min at room temperature. Phases were then split by addition of 725 μl chloroform and 170 μl of 2M HCl. Samples were vortexed and centrifuged at 5000 rpm for 5min. Lower organic phases were collected and transferred to 1.5 ml tubes containing 708 μl fresh "upper phase" (upper phase taken from of a chloroform-methanol-0.01N HCl (240:120:90) solution. Samples were vortexed, spun down and the lower phase was collected into clean 1.5 ml tubes to be used for PIP 3 and PIP 2 quantification by mass spectrometry assay (manuscript under review).

Differential Scanning Fluorimetry Assays
Differential scanning fluorimetry measurement was performed to test for conditions that would stabilize p110/p85 heterodimers (Niesen et al., 2007 Melting temperatures (Tm) for each condition were derived from melting curves.

Binding Assays of Phosphopeptides to p110β/p85α-icSH2 Complex
In order to verify that all phosphopeptides (wild-type and mutant) used in the activity assays bind to the enzyme, we carried out competition binding assays. Binding of a fluoresceinlabeled phosphopeptide derived from human PDGFR (739-744) (Fluor-GpYMDMS) to the p110β/p85α-icSH2 was first performed to determine the Kd of the fluorescent peptide. Twofold serial dilutions of the p110β/p85α-icSH2 complex were prepared in binding buffer (50 mM Tris pH7.5, 100 mM NaCl and 1 mM TCEP) with highest protein concentration of 86 nM.