Gα_q binds to p110α/p85α phosphoinositide 3-kinase and displaces Ras

Materials

HA (haemagglutinin) antibody was obtained from Covance (Richmond, CA, U.S.A.). FLAG antibody was from Sigma (St. Louis, MO, U.S.A.). Phospho-Ser²¹⁷/Ser²²¹ MEK1/2 antibody (MEK is mitogen-activated protein kinase/extracellular-signal-regulated kinase kinase) was obtained from New England Biolabs (Beverly, MA, U.S.A.). Recombinant p110γ protein was purchased from Axxora (San Diego, CA, U.S.A.).

Constructs and recombinant proteins

The cDNA for human H-Ras was obtained by RT (reverse transcriptase)–PCR from HEK-293 (human embryonic kidney 293) cell RNA. It was used as a template to make V12Ras by PCR-based mutagenesis. FLAG–p110α was described previously [21]. FLAG–p110α amino acids 1–116 was made by mutating codon 117 of p110α to a stop codon. p110α amino acids 118–1068 was made by PCR using p110α as a template and the fragment was then subcloned into p3XFLAG-CMV10 (Sigma) to make FLAG–p110α amino acids 118–1068. FLAG–p110γ was constructed by subcloning the p110γ insert of IMAGE clone number 5749986 into p3XFLAG-CMV10. HA–Gα_qQ209L and Gα_qQ209L were described previously [21,22]. HA–MEK1 was a gift from N. G. Ahn (University of Colorado, Boulder, CO, U.S.A.). Akt–HA was obtained from R. Roth (Stanford University, Stanford, CA, U.S.A.).

Purification of the p110α/p85α PI3K complex [21] and Gα_q [23] from baculovirus-infected Sf9 cells was described previously. Purified recombinant H-Ras (amino acids 1–166) expressed in Escherichia coli was generously provided by Dr N. Nassar (Stony Brook University, Stony Brook, NY, U.S.A.). To make GST (glutathione S-transferase)–p110α fusion proteins, fragments encompassing the various p110α domains were made by PCR and then subcloned into pGEX-5X-3 (Amersham Biosciences, Piscataway, NJ, U.S.A.). GST and the GST–p110α fusion proteins were expressed in E. coli and purified using glutathione–Sepharose 4B (Amersham Biosciences).

Cell culture and lysate preparation

HEK-293 cells and COS7 cells were maintained in Dulbecco's modified Eagle's medium (Sigma) containing 10% (v/v) fetal bovine serum (Sigma) and antibiotics. COS7 cells were transfected using Lipofectamine™ (Invitrogen, Carlsbad, CA, U.S.A.). HEK-293 cells were transfected using TransIT-293 (Mirus, Madison, WI, U.S.A.) following standard methods. The FreeStyle 293 Expression System (Invitrogen) was used to express HA–Gα_qQ209L. Cells were rinsed with ice-cold PBS 2 days after transfection, and scraped into lysis buffer (50 mM Hepes, pH 7.5, 50 mM NaCl, 5 mM EDTA, 50 mM NaF, 10 mM sodium pyrophosphate, 1 mM sodium orthovanadate, 0.5 mM PMSF and 10 μg/ml aprotinin and leupeptin) with either 1% Triton X-100 or 1% Nonidet P40 plus 0.25% sodium deoxycholate. Homogenates were centrifuged at 15000 g for 15 min at 4 °C, and protein concentrations of the supernatants were determined using a Bradford assay (Bio-Rad, Hercules, CA, U.S.A.).

Kinase assays and Western blots

Akt activity was assayed following a method described previously [18]. PI3K was assayed as described previously [18] except that the PI3K assay buffer contained 40 mM Hepes (pH 7.5), 1 mg/ml fatty-acid-free BSA, 2 mM EGTA and 100 mM NaCl. Western blot signals were visualized using horseradish-peroxidase-linked secondary antibodies (Amersham Biosciences) and chemiluminescence reagents (PerkinElmer Life Sciences, Boston, MA, U.S.A.) [18].

Fluorescence measurements and data analysis

Purified Gα_q was stored in a solution containing 20 mM Hepes (pH 7.2), 1 mM EDTA, 3 mM MgCl₂, 400 mM NaCl, 0.7% CHAPS and 50 μM GDP (GDP buffer). To activate Gα_q, the protein was incubated for 1 h at 30 °C in GTP[S] (guanosine 5′-[γ-thio]triphosphate) buffer (50 mM Hepes, pH 7.2, 100 mM (NH₄)₂SO₄, 150 mM MgSO₄, 1 mM EDTA, 0.7% CHAPS and 100 μM GTP[S]) [24]. To activate Ras, the protein was incubated in a buffer containing 20 mM Hepes (pH 7.2), 200 μM (NH₄)₂SO₄, 5 mM EDTA and 5 μM GTP[S] for 1 h on ice. The reaction was stopped by adding 20 mM MgSO₄ and the protein was dialysed against buffer A (20 mM Hepes and 160 mM KCl, pH 7.2) plus 1 mM 2-mercaptoethanol for 30 min.

Prior to labelling with 7-(dimethylamino)coumarin-4-acetic acid succinimidyl ester (referred to as coumarin; Molecular Probes, Eugene, OR, U.S.A.), both Gα_q and Ras were dialysed overnight against buffer A. The pH of the Gα_q and Ras solutions was raised by adding 1 μl of 2 M K₂HPO₄ (pH 8.5) per 100 μl before labelling with a 4-fold molar excess of coumarin. The reaction mixtures were incubated on ice for 1 h and then free probe was removed by dialysis against buffer A at 4 °C.

LUVs (large unilamellar vesicles), 0.1 μm in diameter, were formed from a 2:1 mixture of POPC (1-palmitoyl-2-oleoyl phosphatidylcholine) and PS (1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine; Avanti Polar Lipids, Alabaster, AL, U.S.A.). Lipids were dissolved in chloroform, dried under vacuum and resuspended in buffer A at room temperature (22 °C). The lipids were then subjected to five cycles of freeze–thawing to form multilamellar vesicles and then extruded through a polycarbonate filter (0.1 μm diameter) for ten cycles to produce LUVs.

Fluorescence measurements were performed on an ISS spectrofluorimeter (ISS, Champaign, IL, U.S.A.) in 3 mm pathlength cuvettes that were filled with 120 μl of sample. All samples contained 80 μM LUVs. Buffer controls used buffer A plus 1 mM dithiothreitol. Coumarin-labelled proteins were excited at 340 nm and fluorescence emission intensity was scanned from 380 to 580 nm. The integrated area was calculated for each point of the titration curve. Binding to either coumarin–Gα_q or coumarin–Ras resulted in a maximal 20% increase in fluorescence. The spectra were corrected for loss of fluorescence caused by dilution from addition of buffer alone. These background-corrected spectra were then normalized to the last point on the titration curve to give a final value of 1.0. This process assumes complete binding and allows us to determine apparent dissociation constants (K_d) in nanomolar units when the data are fitted to a bimolecular association curve. Experiments were performed in triplicate and results shown are the means±S.E.M.

Gα_q inhibits p110α/p85α PI3K

We used recombinant proteins purified from baculovirus-infected insect cells to test if Gα_q inhibits the lipid kinase activity of the p110α/p85α PI3K complex in vitro. Addition of Gα_q·GDP resulted in a 46±17% (mean±S.E.M., n=4) decrease in PI3K activity as compared with the buffer control, whereas the same amount of Gα_q·GTP[S] caused a 92±1% (mean±S.E.M., n=4) decrease in activity (Figure 1). These results show that Gα_q directly inhibits the p110α/p85α PI3K complex and activated Gα_q is a more potent inhibitor than inactive Gα_q.

Gα_q binds to p110α/p85α PI3K

Next, we used a fluorescence spectroscopy technique to measure the K_d between p110α/p85α PI3K and Gα_q. Since Gα_q is a membrane-bound protein, these measurements were performed in the presence of lipid vesicles consisting of POPC/PS (2:1) to mimic the membrane environment. Gα_q was labelled with coumarin and reconstituted into the vesicles, and then PI3K was titrated into the solution. The emission intensity of the coumarin-labelled Gα_q, which increases upon binding to other proteins [25], was then measured. Coumarin labelling does not impair Gα_q function, since coumarin–Gα_q activates PLCβ isoenzymes as efficiently as unlabelled Gα_q [25]. Control experiments showed that PI3K also binds to the lipid vesicles with high affinity, with a membrane partition coefficient of approx. 10 μM (results not shown). Since we carried out these studies at lipid concentrations (80 μM) where both proteins are bound, the interactions monitored here occur between proteins confined to vesicle surfaces.

Figure 2 shows that addition of increasing amounts of p110α/p85α PI3K into a solution containing active Gα_q·GTP[S] caused an increase in fluorescence that reached a plateau at approx. 100 nM PI3K. The K_d of PI3K binding to Gα_q·GTP[S] was calculated to be 11±1 nM. The binding curve using inactive Gα_q·GDP yielded a K_d of 77±25 nM. The binding curve for Gα_q·GDP appears to consist of two exponential curves, suggesting that two binding species might be present, which may be due to different conformational or oligomeric states. Additional biophysical analysis would be needed to clarify these possibilities. By comparison, K_d values for coumarin-labelled Gα_q·GTP[S] binding to three isotypes of PLCβ ranged from 0.3 to 2.8 nM, as determined by fluorescence resonance energy transfer (data in [25] were recalculated to give K_d values at the same lipid concentration used in these studies). The increase in fluorescence was reversed when trypsin was added to the solution (results not shown). In addition, the binding curves shifted to the right with increasing amounts of Gα_q·GTP[S], but the K_d values remained the same, indicative of a true protein–protein equilibrium association (results not shown). Gα_q·GTP[S] did not bind to the p110γ PI3K under these conditions (K_d>1800 nM; Figure 2), suggesting that Gα_q might selectively inhibit only certain PI3K isoforms. These results demonstrate that Gα_q binds directly to the p110α/p85α PI3K, and active Gα_q binds with a higher affinity than inactive Gα_q.

Gα_q interferes with Ras binding to PI3K

The data in Figure 1 indicate that activated Gα_q inhibits intrinsic PI3K activity. We wondered if the interaction with Gα_q might also interfere with binding of PI3K to an activator such as Ras. We used fluorescence spectroscopy to first measure the K_d of p110α/p85α complex binding to Ras. Fluorescence intensity measurements were made after adding increasing amounts of PI3K to lipid vesicles preincubated with coumarin-labelled Ras. As expected, PI3K bound to Ras·GTP[S] with a K_d of 31±12 nM, while the K_d with inactive Ras·GDP was estimated to be at least 30 times higher (Figure 3). This correlates well with a value of approx. 150 nM for the interaction between an activated Ras mutant and p110α/p85α measured in solution [10]. When the lipid vesicles were reconstituted with coumarin–Ras in the presence of unlabelled Gα_q·GTP[S], specific binding between Ras and PI3K was greatly reduced (Figure 3). This result indicates that Gα_q·GTP[S] interferes with Ras binding to p110α/p85α.

We also investigated if Gα_q interferes with PI3K binding to tyrosine-phosphorylated proteins that are thought to contribute to receptor-mediated PI3K activation. Using fluorescence spectroscopy, we found that binding between coumarin-labelled tyrosine-phosphorylated IRS-1 (insulin receptor substrate-1) and p110α/p85α was not affected by Gα_q·GTP[S] (results not shown).

Gα_q binds to the p85-binding domain of p110α

In vitro binding assays using PI3K deletion mutants identified an RBD located in the N-terminal region of p110α and p110β [10] (see Figure 4A). In addition, X-ray crystallography of a Ras·p110γ complex revealed that Ras establishes contacts with the RBD and the catalytic domain [13]. Since Gα_q blocks Ras binding to PI3K, we asked whether Gα_q also binds to the RBD of p110α. Figure 4(A) illustrates GST fusion proteins encompassing the various domains of p110α that were purified from bacteria and mixed with lysates of mammalian cells expressing HA-tagged Gα_qQ209L. The GST fusion proteins were pulled down with glutathione–Sepharose, and the presence of HA–Gα_qQ209L was assessed by Western blotting. Surprisingly, we found that the p85-binding domain of p110α co-precipitated with HA–Gα_qQ209L, but the RBD did not (Figure 4A).

We also tested whether a FLAG-tagged p110α fragment (amino acids 118–1068) that does not contain the p85-binding domain co-immunoprecipitates with HA–Gα_qQ209L from extracts of COS7 cells expressing both proteins. FLAG–p110α amino acids 118–1068 exhibits substantial lipid kinase activity in an in vitro assay, indicating that the protein is folded correctly (results not shown). Consistent with our findings above using GST fusion proteins, FLAG–p110α amino acids 1–116 co-precipitated with HA–Gα_qQ209L, but FLAG–p110α amino acids 118–1068 did not (Figure 4B). FLAG-tagged p110γ, which contains an RBD but lacks a p85-binding domain, also did not co-immunoprecipitate with co-transfected HA–Gα_qQ209L (Figure 4B). We showed earlier that the amount of p85α bound to p110α does not change in the presence of HA–Gα_qQ209L [21]. Therefore, even though activated Gα_q binds to the p85α-binding site on p110α, it does not inhibit PI3K activity by preventing formation of the p110α/p85α complex.

Gα_q interferes with Ras/PI3K signalling

In addition to PI3K, Ras has a number of other direct targets, including the protein kinase Raf. We reasoned that Gα_q should specifically inhibit the Ras/PI3K pathway and not the Ras/Raf pathway if Gα_q competes with Ras for binding to PI3K. To measure the effect of Gα_q on Ras/PI3K signalling, we assayed Akt activity in HEK-293 cells transfected with Akt–HA in the presence or absence of V12Ras and Gα_qQ209L. The protein kinase Akt is activated by the lipids produced by PI3K and its activity serves as an indirect measure of PI3K activity. Figure 5(A) shows that Akt activity was almost three times higher in cells expressing V12Ras than in vector-transfected control cells, and this increase was completely blocked by Gα_qQ209L. Bommakanti et al. [19] also observed that activated Gα_q exerts a strong inhibitory effect on Ras activation of Akt. Western blotting showed that the amount of Akt–HA in each sample was similar (Figure 5A).

To assess the effect of Gα_q on Ras/Raf signalling, we examined the phosphorylation of MEK1 at Ser²¹⁷/Ser²²¹. Binding of activated Ras to Raf leads to the phosphorylation and activation of MEK1. HEK-293 cells were transfected with HA–MEK1 in the presence or absence of V12Ras and Gα_qQ209L. Western blotting with a phospho-specific antibody showed that V12Ras activates MEK1, but Gα_qQ209L did not block this response (Figure 5B). Reprobing the blot with HA antibody showed that the amount of HA–MEK1 was similar in each lane (Figure 5B). These results indicate that active Gα_q does not block all Ras effector pathways.