A Small Molecule Inhibitor of PDK1/PLCγ1 Interaction Blocks Breast and Melanoma Cancer Cell Invasion

Strong evidence suggests that phospholipase Cγ1 (PLCγ1) is a suitable target to counteract tumourigenesis and metastasis dissemination. We recently identified a novel signalling pathway required for PLCγ1 activation which involves formation of a protein complex with 3-phosphoinositide-dependent protein kinase 1 (PDK1). In an effort to define novel strategies to inhibit PLCγ1-dependent signals we tested here whether a newly identified and highly specific PDK1 inhibitor, 2-O-benzyl-myo-inositol 1,3,4,5,6-pentakisphosphate (2-O-Bn-InsP5), could affect PDK1/PLCγ1 interaction and impair PLCγ1-dependent cellular functions in cancer cells. Here, we demonstrate that 2-O-Bn-InsP5 interacts specifically with the pleckstrin homology domain of PDK1 and impairs formation of a PDK1/PLCγ1 complex. 2-O-Bn-InsP5 is able to inhibit the epidermal growth factor-induced PLCγ1 phosphorylation and activity, ultimately resulting in impaired cancer cell migration and invasion. Importantly, we report that 2-O-Bn-InsP5 inhibits cancer cell dissemination in zebrafish xenotransplants. This work demonstrates that the PDK1/PLCγ1 complex is a potential therapeutic target to prevent metastasis and it identifies 2-O-Bn-InsP5 as a leading compound for development of anti-metastatic drugs.

1 possesses a pleckstrin homology (PH) domain that can bind to the phosphoinositide 3-kinase (PI3K) lipid product phosphatidylinositol 3,4,5-trisphosphate [PtdIns (3,4,5)P 3 ] and contribute to translocation of the enzyme to the plasma membrane 7 . Our previous work demonstrated that this binding is essential for PLCγ 1 activation 7 and it is required for EGF-induced migration of breast cancer cells 8 and basic fibroblast growth factor-mediated migration and remodelling of human umbilical vein endothelial cells 9 .
Deregulated PI3K signalling is very common in cancer and it leads to dysregulation of several intracellular processes, including cell survival, growth, proliferation and migration 10,11 . Among the downstream effectors of PI3Ks, 3-phosphoinositide-dependent protein kinase 1 (PDK1) and AKT have key roles in several cancer types 12 . PDK1 in particular is highly expressed in many human breast cancer cell lines and both PDK1 protein and mRNA are overexpressed in a majority of human breast cancers 13,14 . Overexpression of PDK1 is sufficient to transform mammary epithelial cells and, furthermore, PDK1 has an essential role in regulating cell migration 15,16 . Moreover, we have recently demonstrated that PDK1 can regulate PLCγ 1 activation in a mechanism involving assembly of a complex between the two enzymes upon growth factor stimulation 17 .
Here we investigated the effect of 2-O-Bn-InsP 5 on the novel identified PDK1/PLCγ 1 signalling pathway and its mechanism of action. We show that 2-O-Bn-InsP 5 binds to the PH domain of PDK1 with high affinity, as assessed by isothermal titration calorimetry (ITC). By binding PDK1 PH domain, 2-O-Bn-InsP 5 prevents the formation of the PDK1/PLCγ 1 complex. This in turn results in inhibition of cell migration, 3D Matrigel invasion of breast and melanoma cell lines in vitro, and ultimately tumour cell dissemination in zebrafish xenotransplants.
Taken together these data demonstrate that targeting the PDK1/PLCγ 1 is a promising strategy to prevent metastatic dissemination. Furthermore, these data identify 2-O-Bn-InsP 5 as a leading compound to develop anti-metastatic drugs.

2-O-Bn-InsP 5 is a specific PDK1 PH domain inhibitor.
We have recently identified a synthetic derivative of the natural compound InsP 5 , named 2-O-Bn-InsP 5 , which potently and specifically inhibits PDK1 in vitro as well as the PDK1-dependent phosphorylation of AKT Thr308 in cancer cell lines and in vivo 21 . To further characterise the anti-PDK1 activity of this compound we first determined whether 2-O-Bn-InsP 5 was able to interfere with the platelet-derived growth factor (PDGF)-induced, PDK1-dependent AKT activation. To this end, NIH 3T3 cells were pre-treated for 10 min with 2-O-Bn-InsP 5 followed by 5 minutes stimulation with 30 ng/ml of PDGF (Fig. 1A) and changes in AKT conformation were monitored by a time-resolved FRET approach using an AKT activation sensor (eGFP-AKT-mRFP). Data revealed that PDGF treatment induced a conformational change in eGFP-AKT-mRFP and this was inhibited by pre-treating the cells with 2-O-Bn-InsP 5 (Fig. 1A). As a result of this impaired conformational change, 2-O-Bn-InsP 5 strongly inhibited the PDGF-induced phosphorylation of AKT at its residues Thr308 and Ser473 in NIH 3T3 transfected with eGFP-PDK1 and eGFP-AKT ( Fig. 1B-D). Taken together, these data demonstrate that 2-O-Bn-InsP 5 inhibits PDK1-dependent AKT phosphorylation and are consistent with our previous data in vitro and in vivo 21 .
Since it has been reported that InsP 5 can bind to PDK1 PH domain we hypothesised that 2-O-Bn-InsP 5 could inhibit PDK1 by binding to this domain and therefore interfere with its interaction with PtdIns(3,4,5)P 3 . To test this hypothesis, we determined the thermodynamic parameters of 2-O-Bn-InsP 5 binding to PDK1 PH domain by ITC (Table 1). The closely related PH domain of AKT2 was used as a control to monitor the specificity of binding of the inhibitor (Supplementary Figure 1A). From four independent experiments, performed at 25 °C, the binding affinity of 2-O-Bn-InsP 5 for the PH domain of PDK1 (K d ) was calculated to be 109 ± 44 nM with a 1:1 stoichiometry (n = 0.90 ± 0.04) ( Table 1 and Fig. 2A). Binding was driven by a favourable enthalpic contribution Δ H obs = − 4.1 ± 0.5 kcal/mol and an entropic (− TΔ S) contribution of − 0.291 ± 0.002 kcal/mol. However, no significant heat change was detected by titrating the inhibitor into the AKT2 PH domain performed at 25 °C (Fig. 2B) and at 15 °C indicating that the inhibitor is specific for the PDK1 PH domain. The specificity of 2-O-Bn-InsP 5 for PDK1 is consistent with our previous results from a kinase profiling assay that tested ~50 distinct kinases in vitro and that revealed a selective inhibition of PDK1 activity 21 . Importantly no direct inhibition of AKT activity in vitro was detected in this assay 21 , consistent with ITC results.
To further assess the specificity of 2-O-Bn-InsP 5 in a cellular context, MDA-MB-231 cells were treated with 2-O-Bn-InsP 5 50 µM or vehicle alone before stimulation with EGF and activation of several enzymes was assessed using a commercially available Proteome Profiler Human Phospho-Kinase Array Kit. Out of 45 proteins tested, 2-O-Bn-InsP 5 treatment specifically affected only the EGF-induced phosphorylation of AKT, ERK1/2 and PLCγ 1 (Supplementary Figure 1B), confirming the high specificity of this compound. Specifically, 2-O-Bn-InsP 5 reduced the PDK1-dependent phosphorylation of AKT at its residues Thr308 and Ser473 (Supplementary Figure 1B), consistent with our previous data ( Fig. 1B-D) and our previous report 21 . 2-O-Bn-InsP 5 reduced phosphorylation of PLCγ 1 at its residue Tyr783 (Supplementary Figure 1B), in agreement with our previous study demonstrating that chemical inhibition of PDK1 was able to affect the EGF-induced Tyr783 PLCγ 1 phosphorylation 17 Figure 1C). Taken together these data strongly suggest that 2-O-Bn-InsP 5 specifically inhibits PDK1 activation in a mechanism involving its binding to PDK1 PH domain.

2-O-Bn-InsP 5 inhibits PLCγ1-PDK1 complex formation in MDA-MB-231 breast cancer cells. It
is well established that PDK1 is required for activation of several AGC kinases including AKT. More recently we demonstrated that PDK1 can also regulate activation of PLCγ 1 in a mechanism involving formation of a specific complex between the two enzymes 17 . We therefore decided to investigate whether inhibition of PDK1 by 2-O-Bn-InsP 5 might affect the formation of PLCγ 1/PDK1 protein complex. Stimulation of MDA-MB-231 overexpressing PLCγ 1 and PDK1 with EGF for 3 minutes induced major reorganisation of the plasma membrane ( Fig. 3A white arrows) and formation of a PDK1/PLCγ 1 complex, as demonstrated by increased steady-state FRET signal (Fig. 3B), as previously reported 17 . Treatment of MDA-MB-231 with 2-O-Bn-InsP 5 inhibited reorganisation of the plasma membrane (Fig. 3A) and completely blocked the EGF-induced FRET signal (Fig. 3B) The increase in FRET efficiency upon PDGF treatment indicates a change in conformation of AKT that is prevented upon pre-treatment with the inhibitor. Box and whiskers plots of FRET efficiencies are displayed for the indicated conditions. Each cell is represented by a red symbol. A Mann-Whitney test was used to calculate the P values shown on the graph (n = 3 experiments; **p = 0.0016, ***p = 0.0001). (B) Cells were left untreated or treated with 2-O-Bn-InsP 5 prior to PDGF stimulation and phosphorylation of AKT at residues Threonine 308 (pT308) and Serine 473 (pS473) was assessed by Western blotting. (C,D) Results from densitometry analysis of Western blotting showing significant decrease in the normalized phosphorylation of AKT at T308 and S473 from 4 independent experiments (**p = 0.002, *p = 0.007 respectively). indicating that 2-O-Bn-InsP 5 prevents the formation of the PLCγ 1/PDK1 complex (Fig. 3B). In agreement with results from FRET analysis, treatment with 2-O-Bn-InsP 5 also reduced the EGF-induced endogenous association between PLCγ 1 and PDK1 detected by immunoprecipitation analysis in MDA-MB-231 (Fig. 3C). Quantification of the number of cells showing major membrane rearrangements and localisation of both PLCγ 1 and PDK1 at the plasma membrane confirmed that treatment with 2-O-Bn-InsP 5 strongly inhibited the EGF-induced translocation of both enzymes (Fig. 3D). Taken together these data demonstrate for the first time that 2-O-Bn-InsP 5 inhibits association of PLCγ 1 and PDK1 at the plasma membrane.

2-O-Bn-InsP 5 inhibits the EGF-induced PLCγ1 tyrosine phosphorylation.
We previously reported that the formation of a PDK1/PLCγ 1 complex is involved in regulation of PLCγ 1 activation 17 . Importantly, data from the phosphokinase array indicated that treatment with 2-O-Bn-InsP 5 reduced the EGF-induced phosphorylation of PLCγ 1 at its residue Tyr783 (Supplementary Figure 1B), which is required for the enzyme activation. To better investigate whether the effect of 2-O-Bn-InsP 5 on PDK1/PLCγ 1 complex formation results in inhibition of PLCγ 1 activity, MDA-MB-231 cells were left untreated or treated with 2-O-Bn-InsP 5 before stimulation with EGF ( Fig. 3E,F). The PDK1 inhibitor GSK2334470 was used in parallel in these experiments. Consistent with our previous results 17 , treatment with GSK2334470 strongly reduced the EGF-induced PLCγ 1 phosphorylation (Fig. 3E,F). Importantly, treatment with 2-O-Bn-InsP 5 also significantly inhibited PLCγ 1 phosphorylation upon EGF stimulation (Fig. 3E,F, Supplementary Figure 1C). These data indicate for the first time that blockade of the PDK1/PLCγ 1 complex formation by 2-O-Bn-InsP 5 is able to inhibit Tyr783 PLCγ 1 phosphorylation.

2-O-Bn-InsP 5 inhibits fibronectin-induced cell migration in MDA-MB-231.
Our previous study demonstrated that the PDK1/PLCγ 1 complex is required for cancer cell migration and invasion 17 . Therefore, we next investigated whether inhibition of the complex formation and PLCγ 1 activation by 2-O-Bn-InsP 5 was  (Fig. 5A). Dose response analysis showed that 2-O-Bn-InsP 5 was able to inhibit cell migration by 80% compared to untreated cells at a concentration as low as 10 μ M (Fig. 5B). These data demonstrate that 2-O-Bn-InsP 5 inhibits cancer cell migration.  in cells lacking PLCγ 1. Taken together these data demonstrate that 2-O-Bn-InsP 5 is able to inhibit invasion of several cancer cell lines more efficiently than the parental molecule InsP 5 .  Then, we injected highly metastatic MDA-MB-231 breast cancer cells stably expressing GFP into the heart 22 of 48h post fertilization Tg(kdrl:HsHRAS-mCherry) s896 zebrafish embryos, which express Cherry fluorescent protein specifically in endothelial cells. To assess the correct injection of tumour cells into the heart and/or cardiac chamber, zebrafish embryos were live-imaged by confocal microscopy (Fig. 6C) immediately after the injection. Embryos displaying a similar number and distribution of injected tumour cells were selected and randomly divided into a group that was left untreated and a group that was treated with 2-O-Bn-InsP 5 100 μM. Then, after three days, zebrafish embryos were fixed, immunostained using anti-Cherry and anti-GFP antibodies and imaged by high resolution confocal microscopy. Embryos untreated or treated with 2-O-Bn-InsP 5 showed a similar amount of GFP-derived fluorescence (Fig. 6D). Because GFP-derived fluorescence is proportional to the number of GFP-expressing MDA-MB-231 cells injected, this data indicates that untreated and treated embryos had a similar number of surviving cancer cells. Injected embryos untreated or treated with 2-O-Bn-InsP 5 were imaged by high resolution confocal microscopy and Z-stack scans were used to visualise cherry-positive blood vessels and GFP-positive MDA-MB-231 ( Fig. 6E top and middle panels). Then, we used Imaris software to generate 3D-rendered projections of the embryos vasculature and to exclude background green fluorescence observed in the yolk sack and gut before quantifying metastases number and volume (Fig. E bottom panels). Implantation of MDA-MB-231 breast cancer cell in zebrafish embryos resulted in widespread tumour cell dissemination and  (Fig. 6H). These results demonstrate for the first time that 2-O-Bn-InsP 5 prevents MDA-MB-231 cells from disseminating and support the conclusion that inhibition of the novel PDK1/PLCγ 1 pathway may represent a novel anti-metastatic strategy.

Discussion
Several lines of evidence from different groups including our own have indicated that inhibition of PLCγ 1 may represent a promising strategy to block metastasis spread. In particular, we previously demonstrated that downregulation of PLCγ 1 expression was able to revert metastasis formation in nude mice 4 . These data strongly supported the conclusion that PLCγ 1 inhibition has therapeutic potential to counteract metastasis dissemination and growth 4 . Despite this evidence and further data indicating that PLCγ 1 has a key role in tumourigenesis 3,6,12 , the development of selective inhibitors has proven problematic since phospholipases in general are not very good pharmacological targets 6,23 . In our quest for inhibitors of PLCγ 1 signalling we devised a strategy based on defining the signalling pathway and the interacting partners of PLCγ 1 rather than specifically targeting this phospholipase. This led us to the discovery for PDK1 as a novel PLCγ 1 interacting protein 17 involved in the activation of the phospholipase and in the regulation of PLCγ 1-dependent cellular functions, including cell invasion. Consistent with a role for PDK1 in tumorigenesis, PDK1 overexpression and increased copy number in breast cancer has been shown to correlate with upstream lesions such as PTEN loss, PIK3CA mutation and EGFR amplification, and resulted in increased tumour growth, cell motility and poor prognosis 13 . Our previous study further suggests a major role for PDK1 in controlling metastasis development and progression by promoting PLCγ 1 activation 17 , strengthening the notion that PDK1 is an important potential target to develop novel anti-cancer strategies 12,14 . In this scenario the novel PDK1/PLCγ 1 interaction might play an important role and could be an important therapeutic target for metastasis.
We anticipated the targeting of PDK1/PLCγ 1 interaction at the plasma membrane to be an efficient strategy to counteract PLCγ 1 activation. Therefore, we aimed at designing an allosteric inhibitor that would specifically impair PDK1 plasma membrane localisation (rather than targeting its kinase domain) by interfering with its PH domain binding. We predicted that this approach would also lead to the finding of more tumour-specific and less toxic inhibitors. Structural analysis determined that PDK1 PH domain has an unusually spacious ligand binding site compared to other PtdIns(3,4,5)P 3 -binding PH domains 18 . In particular, PDK1 PH domain presents additional space around the D2-and D6-hydroxyl groups, which potentially could accommodate further phosphate groups. Thus, in addition to binding the PtdIns(3,4,5)P 3 head group, PDK1 PH domain could also bind to the soluble inositols InsP 5 and InsP 6 . 2-O-Bn-InsP 5 was designed to exploit the peculiar PDK1-PH domain structure in order to reach high specificity. ITC experiments show that 2-O-Bn-InsP 5 binds to PDK1-PH domain but not to AKT2-PH domain suggesting that 2-O-Bn-InsP 5 inhibits specifically PDK1 activation by direct binding to its PH domain (Fig. 2). In agreement, PDK1 is inhibited with high selectivity by 2-O-Bn-InsP 5 in a SelectScreen kinase profiling service 21 .
Here we show for the first time that 2-O-Bn-InsP 5 targets the recently discovered PDK1-dependent PLCγ 1 activation 17 by impairing PDK1 and PLCγ 1 plasma membrane localisation and the formation of the PDK1/PLCγ 1 complex required for the PDK1-dependent PLCγ 1 phosphorylation and activation (Figs 3 and 7). 2-O-Bn-InsP 5 treatment reduced PDK1/PLCγ 1 complex formation as shown by FRET and co-immunoprecipitation analyses (Fig. 3). More importantly, we show that the 2-O-Bn-InsP 5 -mediated inhibition of PDK1/PLCγ 1 complex assembly results in inhibition of cancer cell migration, invasion and in vivo dissemination using zebrafish xenotransplants (Fig. 6). Together these results strongly suggest that the blockade of PDK1/PLCγ 1 interaction by 2-O-Bn-InsP 5 is sufficient to inhibit cell invasion (Fig. 7). However, our results do not rule out the possibility that 2-O-Bn-InsP 5 might act via other PDK1 targets beside PLCγ 1. Therefore, it remains to be determined whether 2-O-Bn-InsP 5 could affect these or other biological functions also by targeting additional effectors in addition to PDK1-or PLCγ 1-dependent effectors. Altogether our data show for the first time that the PDK1/PLCγ 1 interaction is pharmacologically targetable and that its inhibition may have anti-metastatic effects in vivo. Therefore, 2-O-Bn-InsP 5 could potentially be a promising anti-cancer lead compound to design novel anti-metastatic drugs. Allosteric compounds have emerged as an attractive new class of small molecule inhibitors in cancer. Due to their specific mode of interaction that does not involve competition with ATP at the active site, they have been shown to be highly selective. We were among the first to suggest that activation of proteins involved in cell growth and tumourigenesis could be inhibited by interfering with their translocation to the plasma membrane 24 . In particular we showed that cytoplasmatic inositol phosphates could compete with PtdIns(3,4,5)P 3 for the binding to AKT PH domain preventing its translocation to the plasma membrane and activation 24 thus representing an important alternative to the use of inhibitors directly targeting the catalytic domain 24 . Recent work has reinforced the idea that small molecule inhibitors can act by interfering with the localization of proteins with key roles in cancer progression 25,26 . For instance, although the cancer-associated protein KRAS had long been considered undruggable, a novel strategy was recently developed based on the indirect inhibition of its membrane localization 26,27 . In this respect results from our current work provide further support to the conclusion that inhibition of protein membrane translocation can represent a useful alternative strategy to block protein activation and ultimately processes associated with tumorigenesis. By binding to PDK1 PH domain, the allosteric inhibitor 2-O-Bn-InsP 5 inhibits PDK1 and PLCγ 1 membrane localisation and the PDK1/PLCγ 1 complex formation required for full PLCγ 1 activation. Thus, the allosteric PDK1 inhibitor 2-O-Bn-InsP 5 targets PDK1 activity with a different mechanism of action than currently available ATP competitive inhibitors like GSK2334470. Because 2-O-Bn-InsP 5 interferes with membrane localisation of PDK1 and PLCγ 1 (Fig. 3) it could be speculated that 2-O-Bn-InsP 5 could selectively inhibit activation of PDK1 substrates by interfering with PDK1 (and its substrates) localisation to the plasma membrane. Since PDK1 has been shown to be a master regulator at the hub of many downstream signalling pathways, the effect of an allosteric inhibitor that would interfere with specific downstream substrates activation by modulating PDK1 spatial localization would be of great value allowing for the selective inhibition of some of the PDK1 signalling over others. Taken together, our findings show that 2-O-Bn-InsP 5 could lead to the development of distinct more selective (hence less toxic) therapeutic strategies by targeting specifically the PDK1/PLCγ 1 complex formation at the plasma membrane and its downstream pathways that may lead to the development of novel drugs capable of selectively inhibiting the invasive potential of cancer cells.

Materials and Methods
Reagents and antibodies. Chemicals were purchased from Sigma-Aldrich, UK. Human PDGF was purchased from R&D Systems. Recombinant EGF was purchased from Peprotech, UK. Antibodies were purchased as follows: pAKT (Thr308), pAKT (Ser473), pPLCγ 1 (Y783), total AKT and PDK1, pERK1/2 T202/Y204 and total ERK1/2 from Cell Signaling Technology, GAPDH from Abcam, PLCγ 1 from Santa Cruz Biotechnology. Glassbottomed 35-mm dishes were obtained from MatTek Corporation. InsP 5 and 2-O-Bn-InsP 5 were synthesised as previously reported 21,28 . Each compound was purified to homogeneity by ion-exchange chromatography on Q-Sepharose Fast Flow resin and used as the triethylammonium salt, which was fully characterized by 31  lines were serum starved overnight before the experiment. Cell migration and invasion assays were performed as described 17,30 . Briefly, cells were left untreated or pre-treated with 50 μ M of 2-O-Bn-InsP 5 for 30 minutes before being detached, counted and plated on inserts. For migration assays, a suspension of 10,000 cells in 150 μ l was homogenously added in the upper chamber and the lower chamber was filled with RPMI or DMEM containing 0.5% BSA. Cells were allowed to migrate for 4 hours at 37 °C, 5% CO 2 in the absence or presence of the inhibitor. For invasion assay, Matrigel pre-coated inserts (8.0 μ m pores, 10 mm diameter, BD Bioscience, UK) were used. Inserts were re-hydrated for 1 hour in DMEM or RPMI 1640 according to the cell type before the experiment. A suspension of 10,000 cells (60,000 for MDA-MB-435) in 500 μ l was homogenously added in the upper chamber and the lower chamber was filled with RPMI or DMEM containing 10% FBS. Cells were allowed to invade for 36 hours at 37 °C, 5% CO 2 in the absence or presence of the inhibitor. Cells that did not migrate or invade were removed using a cotton bud whereas cells that had migrated or invaded were fixed with paraformaldehyde and stained with 0.1% crystal violet solution for 10 minutes. A Leica phase-light microscope using a 10X magnitude Scientific RepoRts | 6:26142 | DOI: 10.1038/srep26142 objective was used for manual counting. A minimum of five fields was counted per insert. Each experiment was performed in duplicate and the average of cells/fields was calculated. Intracellular Calcium Measurement. MDA-MB-231 and MDA-MB-435 were seeded on bottom-glassed chambers (Labtek, UK) and serum starved overnight. Cells were then incubated with HBSS (Life Technologies) containing 0.5% BSA, 2 mM CaCl 2 , 4 μ M Fluo-4-AM (Life Technologies) for 45 minutes at 37 °C, washed twice in HBSS 0.5% BSA 2mM CaCl 2 and left in the same solution for 30 minutes for de-esterification of the Fluo-4-AM dye. Where indicated 50 μ M InsP 5 or 50 μ M 2-O-Bn-InsP 5 were added during the de-esterification step for 30 minutes. Fluorescence was measured using the LSM 510 inverted confocal microscope equipped with a chamber for live imaging at 37 °C supplied with 5% CO 2 using a 20X objective. After recording basal fluorescence, cells were then stimulated with EGF (20 ng/ml) and fluorescence was measured for 6 minutes before further stimulation with 1 mM ATP (Sigma, UK). Thirty cells were selected and variation of fluorescence intensity was analysed for each sample.

Lysis conditions and Western blotting analysis.
After stimulation or treatment, NIH 3T3 were lysed for 5 min on ice in Lysis buffer [20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 100 mM NaF, 10 mM Na 4 P 2 O 7 , and 10 mM EDTA supplemented with one complete protease inhibitor cocktail tablet (Roche) and 1% Triton X-100]. To terminate the reaction, SDS Sample buffer [125 mM Tris-HCl (pH 6.8), 6% SDS, 20% glycerol, and 0.02% bromophenol blue supplemented with 10% β-mercaptoethanol] was added and the samples were boiled for 5 min. The proteins were separated on a NuPAGE 4 to 12% Bis-Tris Gel (Life Technologies) and transferred to a polyvinylidene difluoride (PVDF) membrane (Immobilon FL, Millipore). For protein detection with the Odyssey Infrared Reader (LI-COR), membranes were incubated in blocking buffer (LI-COR) for 1 h, and incubated for 24 h at 4 °C with antibodies against pAKT (Thr308) together with total AKT, or pAKT (Ser473) together with total AKT (each antibody diluted in blocking buffer at 1:1000). The infrared dye-conjugated secondary antibodies IRDye 800CW and 680LT (Rockland) were used at a 1:5000 dilution in blocking buffer for 1 h.
MDA-MB-231 were lysed in lysis buffer containing 50 mM Tris-HCl (pH 7.4), 5 mM EDTA, 0.1% NP40, 250 mM NaCl, and proteases inhibitors cocktail 2 (Sigma-Aldrich, UK). Lysates were then centrifuged at 13,000 rpm for 10 min at 4 °C. Proteins were separated by SDS/PAGE and transferred to nitrocellulose. Membranes were incubated with a solution of 5% milk (w/v) and then incubated overnight with the diluted primary antibodies [pPLCγ 1(Y783) or GAPDH]. The appropriate peroxidase-conjugated secondary antibodies (Sigma-Aldrich, UK) were used and proteins were detected by enhanced chemi-luminescence reaction (GE Healthcare, UK).

Phosphokinase antibody array. Following overnight starvation, MDA-MB-231 cells were left untreated
or pre-treated with 50 μ M 2-O-Bn-InsP 5 and then stimulated with 50 ng/ml EGF for 10 min in the presence or absence of the inhibitor. Cells were lysed as described above and phosphokinase antibody array was performed according to the manufacturer's instructions (R&D Systems). Steady state FRET by Acceptor Photo-bleaching. Cells were treated as for confocal microscopy analysis. Cells were imaged with excitation λ = 488 nm and λ = 543 laser line and emission spectra were collected respectively in two different channels of the PMT detector. Briefly, cells overexpressing PLCγ 1 or PDK1 or co-expressing PLCγ 1 and PDK1 were co-stained with Alexa-488 and Alexa-555 conjugated antibodies. The FRET efficiency was measured by acceptor photo-bleaching, as previously described 17,32 . A selected area of the cell was repeatedly photobleached with λ = 543 nm laser line at full power for 1 minute and FRET efficiency was measured as the increase (or dequenching) of donor fluorescence after photo-bleaching in the selected area. All fluorescence measurements were performed in MetaMorph software (Molecular Devices Inc., USA). Reference spectra were generated from coverslips incubated separately with the two Alexa-conjugated secondary antibodies alone in order to generate single positive sample and all data were corrected for cross-talk and background fluorescence. Percentage of FRET was calculated measuring donor fluorescence of 10 cells before and after acceptor bleaching.
Preparation of the recombinant PH domains. BL21 (DE3) (100 μ l) were electroporated with 500 ng of either pTriEX6-GST-PH AKT2 or pTriEX6-GST-PH PDK1 plasmid DNA, recovered in 200 μ l of SOC medium and plated on ampicillin plates. A colony was picked and grown overnight in 5 ml L-Broth supplemented with 100 μ g/ml of ampicillin at 37 °C in a shaking incubator. This starter culture was used to inoculate 2 × 200 ml L-Broth/ampicillin and allowed to grow at 37 °C with shaking. When the OD 600nm reached approximately 0.5 the expression of the proteins was induced by addition of 1 mM IPTG. Cultures were allowed to grow overnight at 20 °C in a shaking incubator.
The bacteria from each culture were pelleted by centrifugation and then lysed at 4 °C in 2 × 10 ml of lysis buffer comprised of Buffer A (50 mM Tris-HCl pH7.5; 150 mM NaCl; 1 mM EDTA; 5% glycerol) supplemented with 1% Triton X-100, 1 mM DTT, 10 mM β − Glycerophosphate, 1 mM NaF, 10 mM Benzamidine and complete protease inhibitor cocktail (Roche). The re-suspended bacteria were sonicated (3 × 10 s at an amplitude of 10 microns at 4 °C) and centrifuged at 20,000 g for 10 min at 4 °C. The supernatant was subsequently added to glutathione-sepharose beads (1 ml bed volume) and incubated for 2 h. After extensive washing with Buffer A + 1 mM DTT the beads were re-suspended in 500 μ l of the same buffer and the PH domains cleaved from the beads by addition of 3C protease. Cleavage was allowed to occur overnight at 4 °C. The 1 ml total volume of cleaved PH domains were concentrated to 500 μ l using a Vivaspin 5,000 Da cut off concentrator and applied to a S75 size exclusion column equilibrated in the ITC Buffer comprising of Buffer A + 0.5 mM TCEP.
Isothermal Titration Calorimetry Experiments. The purified PH domains of PDK1 and AKT2 were adjusted to 45 μ M and 48 μ M respectively for the ITC experiments. The inhibitor 2-O-Bn-InsP 5 was diluted with ITC buffer (see above) to a 10 fold higher concentration than the individual PH domains. ITC titrations were performed on a VP-ITC200 MicroCalorimeter (MicroCal) at 25 °C (unless otherwise stated) with 15 × 3 μ L injections of 2-O-Bn-IP5 into PH domain located in the cell. Subtraction of the heats of dilution, integration of the raw data and subsequent fitting to a one site binding model was carried out using the Origin (version 7) software supplied with the instrument. and anesthetized with 0.04 mg/ml of tricaine (MS-222, Sigma). Anesthetized embryos were then transferred into a 1% agarose gel for microinjection.
For injection into the heart 22 , 50-100 cells, manually counted, were injected above the ventral end of the duct of Cuvier where it opens into the heart using a manual injector (Picospritzer III, Parker Hannifin Instruments). After injection, the zebrafish embryos were briefly imaged using a Leica SPE confocal microscope. Embryos displaying a similar number of injected cells were selected and divided into a control (untreated) group and treated group (incubated in marine salt water supplemented with 100 μ M 2-O-Bn-InsP 5 ). Embryos were then immediately transferred into an incubator set at 35 °C to compromise between the optimal temperature requirements for fish and mammalian cells. Three days after injection zebrafish embryos were immunostained with rabbit anti-GFP (MBL) and chicken anti-RFP (MBL) followed by staining with Alexa-conjugated goat anti rabbit and goat anti chicken antibodies (Life Technologies) and imaged with a Zeiss LSM 710 Scanning Confocal Microscope using a 20x objective. 3D-rendered models of zebrafish embryos were generated with Imaris (Bitplan AG) using tiled z-stacks of 15-20 confocal slices. Dimensional analysis of metastases was carried out using the surface rendering function of Imaris based on the GFP signal throughout the tiled z-scan series. GFP signal detected within the vessel was excluded by generating a mask based on the mCherry signal within the blood vessels. Autofluorescence from the yolk sac and signal from the intestine was excluded. The number and the volume of metastasis were measured in n = 6 control and n = 10 treated embryos from 2 independent experiments. Experiments were discarded when the survival rate of the control group was < 90%. Alternatively ~300 cells were resuspended in PBS and 1-5 nL of tumour cells solution were injected into the perivitelline cavity of each embryo 33 using a manual injector (Picospritzer III, Parker Hannifin Instruments). After injection, the fish embryos were immediately transferred into 35 °C incubator and examined every day for monitoring tumor dissemination using a fluorescent microscope. Three separate experiments were performed. Counting of disseminated cells was done using a Zeiss Axioplan epifluorescence microscope and disseminated cells counted under high magnification.
Zebrafish locomotor activity assay. We assayed 6dpf untreated embryos (n = 15), or treated with 2-O-Bn-InsP 5 (n = 16) for their locomotor activity. Larvae were raised at 28 °C on a 14/10 h light/dark cycle. Embryos were fed on day 5, and morning of day 6. Two hours before the behavioural procedure, embryos were transferred to a clean plate with no food, and were then moved to the behavioural room. For each trial, 4 embryos were transferred to a tank (11.2 × 7.5 × 2.0 cm) containing 4 insets (6.8 × 2.2 × 2.0 cm) with opaque walls, each with 15 mL of water. The tank was placed inside a Zebrabox (ViewPoint Life Sciences) that was continuously illuminated with infrared and white lights. The swimming behaviour of each fish was monitored for 10 minutes using an automated video-tracking system (Zebralab, ViewPoint Life Sciences), and the movement of each larva was recorded using the Zebralab tracking quantization mode. The Zebralab tracking quantification thresholds were set in the following way: detection threshold, 120 (color transparent); movement threshold: high-speed, 6.0 mm/sec and inactivity, 2.0 mm/sec; bin size, 60 s. Red and green tracking lines in Fig. 6G correspond to high-speed and slow-speed trajectories, respectively. Locomotor activity was quantified based on the total distance covered by each fish.

Statistical analysis.
A nonparametric Mann-Whitney test was used to compare the medians of the two data sets for the FLIM data with GraphPad Prism software. To interpret the distribution of data, we used box and whiskers plots. The box and whiskers plot is a histogram-like method for displaying upper and lower quartiles and maximum and minimum values in addition to the median. Each red symbol represents the FRET efficiency of one cell. For the Zebrafish tumor experiments and the locomotor activity assay a two-tailed, unequal variance Student's t-test was used.
In all the other experiments statistical analyses was performed using the one-tailed, paired Student's t-test.