Biochemical Journal

Research article

PPIP5K1 modulates ligand competition between diphosphoinositol polyphosphates and PtdIns(3,4,5)P3 for polyphosphoinositide-binding domains

Nikhil A. Gokhale , Angelika Zaremba , Agnes K. Janoshazi , Jeremy D. Weaver , Stephen B. Shears

Abstract

We describe new signalling consequences for PPIP5K1 (diphosphoinositol pentakisphosphate kinase type 1)-mediated phosphorylation of InsP6 and 5-InsP7 to 1-InsP7 and InsP8. In NIH 3T3 cells, either hyperosmotic stress or receptor activation by PDGF (platelet-derived growth factor) promoted translocation of PPIP5K1 from the cytoplasm to the plasma membrane. The PBD1 (polyphosphoinositide-binding domain) in PPIP5K1 recapitulated that translocation. Mutagenesis of PBD1 to reduce affinity for PtdIns(3,4,5)P3 prevented translocation. Using surface plasmon resonance, we found that PBD1 association with vesicular PtdIns(3,4,5)P3 was inhibited by InsP6 and diphosphoinositol polyphosphates. However, the inhibition by PPIP5K1 substrates (IC50: 5-InsP7=5 μM and InsP6=7 μM) was substantially more potent than that of the PPIP5K1 products (IC50: InsP8=32 μM and 1-InsP7=43 μM). This rank order of ligand competition with PtdIns(3,4,5)P3 was also exhibited by the PH (pleckstrin homology) domains of Akt (also known as protein kinase B), GRP1 (general receptor for phosphoinositides 1) and SIN1 (stress-activated protein kinase-interaction protein 1). We propose that, in vivo, PH domain binding of InsP6 and 5-InsP7 suppresses inappropriate signalling (‘noise’) from stochastic increases in PtdIns(3,4,5)P3. That restraint may be relieved by localized depletion of InsP6 and 5-InsP7 at the plasma membrane following PPIP5K1 recruitment. We tested this hypothesis in insulin-stimulated L6 myoblasts, using mTOR (mechanistic/mammalian target of rapamycin)-mediated phosphorylation of Akt on Ser473 as a readout for SIN1-mediated translocation of mTORC (mTOR complex) 2 to the plasma membrane [Zoncu, Efeyan and Sabatini (2011) Nat. Rev. Mol. Cell Biol. 12, 21–35]. Knockdown of PPIP5K1 expression was associated with a 40% reduction in Ser473 phosphorylation. A common feature of PtdIns(3,4,5)P3-based signalling cascades may be their regulation by PPIP5K1.

  • diphosphoinositol polyphosphate
  • inositol pyrophosphate
  • phosphoinositide 3-kinase (PI3K)
  • phosphoinositide

INTRODUCTION

The phosphate group is a ubiquitous signalling device that establishes specificity in ligand–protein and protein–protein interactions. The phosphate's bulk imposes geometric constraints on these interactions. The phosphate's negative charge also bestows specificity through ionic and hydrogen bonds with certain amino acid residues at physiological pH. An extreme example of these applications for the phosphate group is provided by the PP-InsPs (diphosphoinositol polyphosphates), also known as ‘inositol pyrophosphates’. These molecules exhibit Nature's most crowded 3D array of phosphate groups. However, there is much that we do not understand concerning the significance of specific proteins interacting with these multiple phosphate groups in PP-InsPs.

Three PP-InsPs are formed by two groups of enzymes; first, the 5-kinase activities of IP6K (inositol hexakisphosphate kinase) 1/2/3 [1,2] convert InsP6 and 1-InsP7 to 5-InsP7 and InsP8 respectively. Secondly, the 1-kinase activities of PPIP5K (diphosphoinositol pentakisphosphate kinase) 1/2 [35] phosphorylate InsP6 and 5-InsP7 to 1-InsP7 and InsP8 respectively. These PP-InsPs regulate many cellular processes in eukaryotes, including stress responses, apoptosis, vesicle trafficking, cytoskeletal dynamics, exocytosis, insulin signalling and neutrophil activation (for reviews, see [69]). It has been challenging to provide a mechanism of action of PP-InsPs that adequately explains how these molecules can regulate such a variety of biological processes.

Experiments with yeasts [10,11] have indicated that the Pho80/81/85 cyclin-dependent kinase complex is allosterically inhibited by 1-InsP7 (inadvertently thought to be 4/6-InsP7 in those reports, see [12]). However, those observations have not helped us understand the promiscuity of PP-InsP signalling in mammalian cells, in which 1-InsP7 is a minor constituent and 5-InsP7 is present at much higher levels [12,13]. An alternative mechanism of action of the PP-InsPs has been advanced by Snyder, Saiardi and co-workers that does offer multiple signalling opportunities: protein diphosphorylation [1416]. In vitro at least, PP-InsPs non-enzymically donate a phosphate to a pre-existing serine phosphate [1416]. However, that hypothesis does not rationalize why there are several PP-InsPs that are all equally capable of phosphorylating proteins (see [17]). In any case, techniques are not yet available to directly determine whether this protein diphosphorylation can occur in vivo.

The focus of the present study began with a separate mechanism of action that has been proposed just for 5-InsP7. Again, it is Snyder and co-workers who have been at the forefront of this work [1820]. This group has shown that, in vitro, 5-InsP7 can out-compete PtdIns(3,4,5)P3 for binding to certain PH (pleckstrin homology) domains, most notably that of the protein kinase Akt (or protein kinase B) [1820]. Moreover, 5-InsP7 was reported to have at least a 10-fold higher affinity for PH domains than did InsP6 [20]. Thus PtdIns(3,4,5)P3-mediated translocation of Akt and other PH domain proteins from the cytoplasm to the plasma membrane, a process that promotes the co-localization of multiprotein complexes and priming of kinase cascades [21], is viewed as being inhibited by the formation of 5-InsP7 from InsP6 [1820]. That hypothesis has been supported by experimental manipulation of IP6K expression in intact cells [1820]. Nevertheless, the latter phenomenon does not account for the significance of 1-InsP7 and InsP8. We considered that such information might be obtained by comparing the relative affinities of 1-InsP7, 5-InsP7 and InsP8 for PtdIns(3,4,5)P3-binding domains. We selected three PH domain proteins for these studies: (i) Akt, owing to the attention that particular kinase has received from Snyder and co-workers for its binding of 5-InsP7 [1820]; (ii) GRP1 (general receptor for phosphoinositides 1), one of several phosphoinositide-dependent guanine-nucleotide-exchange factors [22,23], which show relatively high selectivity for PtdIns(3,4,5)P3 over other inositol lipids [23,24]; and (iii) SIN1 (stress-activated protein kinase-interaction protein 1); the structure of its PH domain was solved recently (PDB codes 3ULB, 3ULC and 3VOQ) [25]. SIN1 mediates PtdIns(3,4,5)P3-dependent translocation of mTORC2 {mTOR [mammalian (or mechanistic) target of rapamycin] complex 2} to the plasma membrane [26]. The mTOR within mTORC2 phosphorylates Ser473 on Akt [26,27]. Thus genetic ablation of SIN1 eliminates stimulus-dependent Ser473 phosphorylation [27].

We reasoned further that insight into the relative affinities of the different PP-InsPs for PtdIns(3,4,5)P3-binding PH domains could also help explain the biological significance of receptor-mediated translocation of PPIP5K1 to the plasma membrane [28]. For example, that information could inform us whether the kinase activity of PPIP5K (phosphorylation of InsP6 and 5-InsP7 to 1-InsP7 and InsP8 respectively) might locally modulate the degree of ligand competition with PtdIns(3,4,5)P3 for PH domains. Moreover, recruitment of PPIP5K1 to the plasma membrane itself involves binding to PtdIns(3,4,5)P3 [28]. So we also determined the relative abilities of InsP6 and the PP-InsPs to compete with PtdIns(3,4,5)P3 for the PBD1 (polyphosphoinositide-binding domain) within PPIP5K1.

There has not previously been an analysis of PP-InsP specificity for any PtdIns(3,4,5)P3-binding domains. Indeed, none of the PP-InsPs are commercially available. Additionally, during most of the time that we were working on the present study, we were unaware of any laboratories that were chemically synthesizing these materials (only recently, one group announced that they now produce 5-InsP7 [29]). Instead, we prepared PP-InsPs enzymatically [12], and using SDS/PAGE methodology [30], our preparations of these molecules were 96–98% pure [31]. We used SPR (surface plasmon resonance) [32] to assay the relative affinities with which InsP6 and the PP-InsPs compete with PtdIns(3,4,5)P3. Intriguingly, Snyder's group have reported that the potency of 5-InsP7 binding to PH domains is complicated by an ‘order-of-addition’ effect. In their hands, 5-InsP7 bound more strongly to Akt when it was added before, rather than after, the PtdIns(3,4,5)P3 [19]. Thus, in our studies into ligand competition, we followed Snyder's protocol, by premixing either InsP6 or PP-InsP with the test protein construct, before its injection over immobilized PtdIns(3,4,5)P3.

We also considered that these studies might have a broader significance that goes beyond the consequences of receptor activation: cellular levels of PtdIns(3,4,5)P3 have been shown to increase after cells are subjected to a hyperosmotic challenge [33,34]. The newly formed PtdIns(3,4,5)P3 acts with other intracellular signals to protect against the potentially lethal consequences of osmotic imbalance, including perturbations to cellular hydration and cytoskeletal integrity, metabolic disruptions, and genomic instability [35,36]. Such stress is not simply restricted to those cell types that reside in anisosmotic environments, such as kidney cells, airway epithelial cells, lymphocytes and cells that make up bone and cartilage [35,3739]. Anisosmosis is also a potential hazard for any cell that transports metabolites and ions, or undertakes polymerization or depolymerization of cellular materials [40]. Yet, surprisingly, among the few studies that have considered whether hyperosmotic stress might promote translocation of PH domains to the plasma membrane, several have reported that it does not occur [34,4143]. Moreover, hyperosmotic stress dramatically activates the synthesis of InsP8 [44]. Thus a final goal of the present study was to investigate whether ligand competition between PtdIns(3,4,5)P3 and PP-InsPs might be relevant to cellular responses to hyperosmotic stress.

MATERIALS AND METHODS

Cell lines, cell culture and [3H]inositol radiolabelling

NIH 3T3 cells were cultured in six-well dishes in DMEM (Dulbecco's modified Eagle's medium) (11965–092, Invitrogen) containing 10% FBS (Invitrogen) and 1 mM sodium pyruvate (Sigma) at 37°C with 5% CO2. Cells were radiolabelled with 20 μCi/ml [3H]inositol as described previously [45]. Rat myoblast L6 cells were purchased from A.T.C.C. (Manassas, VA, U.S.A.) (CRL-1458™) and cultured in MEM (minimal essential medium) α (Invitrogen; A1049001) supplemented with 10% FBS at 37°C with 5% CO2. To create a stable L6K1-KD [L6 cell line with reduced (‘knocked-down’) expression of PPIP5K1], the shRNA construct GI716021 in the pGFP-V-RS vector (Origene Technologies) was used. The TR30013 non-effective shRNA cassette in the pGFP-V-RS plasmid was used to prepare L6CONTROL (control L6 cells). Puromycin resistance carried by the vector was used for selection. DNA constructs encoding retroviral vectors (12 μg), CMV gagpol (10 μg) and pMD2.G (1.2 μg) were cotransfected into HEK (human embryonic kidney)-293T cells on 100-mm dishes using Lipofectamine™ 2000 (Invitrogen). Culture medium was replaced the following day and 24 h later the virus in the supernatants was collected by centrifugation (2000 g for 10 min at 4°C) and used to infect L6 cells [MOI (multiplicity of infection)=40]. Polyclonal cell lines were selected using 4 μg/ml puromycin. All strains of L6 cells were limited to 12 passages. The L6 cells were radiolabelled for 5 days in 6-cm dishes with 23 μCi/ml [3H]inositol.

Experiments with cultured cells were performed after a 3 h period of serum starvation, in 2 ml medium containing 115 mM NaCl, 5 mM KCl, 1 mM NaH2PO4, 0.5 mM MgSO4, 10 mM glucose, 1.35 mM CaCl2 and 25 mM Hepes (pH 7.4 with NaOH).

The extraction and HPLC analysis of inositol polyphosphates and inositol lipids in [3H]inositol cells was performed as described [46] using a Partisphere SAX HPLC column [28,47].

DNA subcloning and protein purification

The full-length cDNA for human PPIP5K1 was used as a template for subcloning its PBD1 (amino acid residues 382–917 of PPIP5K1) into the vector pFLAG-CMV 5a (Sigma–Aldrich), with 5′-CGGAATTCATGCCCACCACATCTGGCACT-3′ as the EcoRI forward primer and 5′-GGGGTACCCTCAACACCTTTCACTCC-3′ as the KpnI reverse primer respectively. The nucleotides underlined represent the EcoRI and KpnI restriction sites respectively. The mutational forward primers 5′-ATTGCAATTATTGCTCATGGGGATCGT-3′ and 5′-GTGAAACACCCAGCTTTTTTTGCTCTG-3′ (the nucleotides in bold font represent the mutagenic codons), along with their corresponding reverse complements, were used to generate the double-mutant PBD1R399A/R417A for subcloning into the vector pFLAG-CMV5a, by overlap extension PCR for alanine residue. All constructs were subsequently transformed into DH5α cells for plasmid isolation and DNA sequencing.

The cDNA encoding GRP1PH (PH domain of GRP1), subcloned into a modified pET-15 vector [48], was a gift from Dr D. Lambright (University of Massachusetts Medical School, MA, U.S.A.). The full-length SIN1 was a gift from Dr E. Jacinto (Robert Wood Johnson Medical School, NJ, U.S.A.). SIN1PH (PH domain of SIN1) (residues 358–522 [25]) was subcloned into pET-21a using 5′-CGGGATCCATGAGGGCAGACGGGGTTTTC-3′ as the BamHI forward primer and 5′-CCGCTCGAGCTGCTGCCCGGATTTCTT-3′ as the XhoI reverse primer. The bases underlined correspond to the restriction sites for BamHI and XhoI. These constructs for either GRP1PH or SIN1PH were each transformed into BL21(DE3) competent cells (Clontech), which were grown at 37°C in LB broth containing 50 μg/ml ampicillin. When the D600 was 0.6, the temperature was reduced to 25°C and expression was induced overnight with 1 mM IPTG. The next day, cells were collected by centrifugation (10 min, 2400 g, 4°C) and resuspended in ice-cold lysis buffer containing 25 mM Hepes (pH 8), 300 mM NaCl, 10 mM imidazole, Complete™ protease inhibitor cocktail tablets (1 tablet per 10 ml of the buffer; Roche, 11836170001) and 0.1% Triton X-100. The suspension was passed through a cooled continuous-flow cell disrupter (TS series; Constant Systems). The appropriate His6-tagged peptide was then purified from the cell lysate by FPLC (ÄKTA system, GE Healthcare) first using a 0.5 ml Ni-NTA (Ni2+-nitrilotriacetate) column eluted with an imidazole-based gradient (pH 8) and then a 5 ml HiTrap SP HP column (GE Healthcare) using an NaCl gradient (pH 7).

Quantitative real-time PCR

Total RNA was isolated from L6K1-KD and L6CONTROL cells using an RNeasy kit with DNase treatment (Qiagen). To quantify mRNA, cDNA was produced using MuLV reverse transcriptase with poly-dT and random hexamer primers. The cDNA was amplified using the Applied Biosystem Fast SYBR® Green Master Mix in an Applied Biosystem 7500 Fast Real-Time machine. The following primers were used: rat PPIP5K1 sense, 5′-ACTAAAGTCTGCCAGCGCCACACTGAT-3′ and rat PPIP5K1 antisense, 5′-GGCTTTTCAGAGCGATGTCGAAGTT-3′; and rat GAPDH sense, 5′-AGAGACAGCCGCATCTTCTTG-3′ and rat GAPDH antisense, 5′-GGTAACCAGGCGTCCGATAC-3′.

Cell proliferation

Cells (0.5×105) were seeded on to 3.5-cm dishes in MEM α containing 10% FBS. After 24 h, the FBS concentration was reduced to 2% FBS. At daily intervals, cells were released by trypsin-treatment and cell numbers were ascertained with a Countess cell counter (Life Technologies).

Cell lysis and immunoblotting

Cells (4×105) attached on to 6-cm dishes were washed twice with ice-cold PBS containing 1 mM Na3VO4, then lysed with RIPA buffer (Santa Cruz; 24948) (supplemented with 1 mM Na3VO4 and protease inhibitors according to the manufacturer's instructions) plus 25 μM okadaic acid (LC Laboratories). Soluble fractions of lysates were prepared by centrifugation (10 min, 10000 g, 4°C). Samples of the cellular lysates containing equal amounts of proteins were resolved by SDS/PAGE and transferred on to PVDF membrane. Proteins were then visualized by immunoblotting and detected with enhanced chemiluminescence (GE Healthcare; RPN2132), or the WesternBright kit from Bioexpress. The following primary antibodies were used: rabbit anti-phospho-Akt antibody (pSer473) (Cell Signaling; 9271, 1:200 dilution); rabbit anti-Akt antibody (Cell Signaling; 9272, 1:200 dilution); rabbit anti-β-actin antibody (Abcam; 1:1000 dilution); immunoaffinity-purified rabbit anti-PPIP5K1 antibodies (raised against a peptide comprising residues 917–1120 conjugated to keyhole-limpet haemocyanin) purchased from Covance (1:1000 dilution); and anti-SIN1 monoclonal antibody (Millipore; 05-1044, 1:1000 dilution), kindly provided by Dr D. Sarbassov (University of Texas M.D. Anderson Cancer Center, Houston, TX, U.S.A.). For detection, we used one of the following horseradish peroxidase-conjugated (1:5000 dilution) secondary antibodies, which were obtained from Cell Signaling Technology: anti-(mouse IgG) (7076) and anti-(rabbit IgG) (7074) antibodies.

Surface plasmon resonance

Unilamellar vesicles [77:20:3 (v/v) phosphatidylcholine/phosphatidylethanolamine/polyphosphoinositide] were immobilized on a Biacore Pioneer L1 sensor chip [32,49] [the dimensions of the dextran matrix were 2.4 mm (length)×0.5 mm (width)×100 nm (height)]. We injected 90 μl of the vesicles over the chip surface at a flow rate of 5 μl/min until the signal reached 5000 AU (arbitrary response units), which corresponds to approximately 50 pg of phospholipid/cm2 [50]. This is equivalent to a surface concentration of approximately 3×1014 phospholipid molecules/cm2, close to the value of 2×1014 phospholipid molecules/cm2 found in a previous SPR analysis of immobilized phospholipid vesicles [51]. As mentioned above, the polyphosphoinositide [usually PtdIns(3,4,5)P3] comprised 3% of the total phospholipid.

Next, the protein under study was injected at a flow rate of 5 μl/min in buffer containing 100 mM KCl, 10 mM NaH2PO4, 1 mM MgCl2 and 25 mM Hepes (pH 7.2). The resulting sensorgram was automatically corrected for any signal arising from protein interacting with control vesicles from which the test phosphoinositide was omitted. The saturating RU values (Req) were plotted against the protein concentrations (C) and the equilibrium binding affinity (Kd) was subsequently determined by performing a non-linear least squares analysis of the binding isotherm using the equation Req=Rmax/(1+Kd/C), where Rmax is the maximal Req value. The S.E.M. of individual Req determinations did not exceed 1% of the mean values.

Confocal immunofluorescence

Cells (1.5×105) were seeded on to 35-mm glass-bottom microwell dishes and transfected 24 h later with cDNA encoding either FLAG–PPIP5K1 (pDEST515 vector [4]), FLAG–PBD1 (pFLAG-CMV-5a vector [28]), FLAG–GRP1 (pFLAG vector kindly provided by Dr L.C. Santy, Pennsylvania State University, PA, U.S.A.) or EGFP–AktPH (pEGFP-N1 vector; SignaGen Laboratories), all using Lipofectamine™ 2000 (Invitrogen) according to the manufacturer's instructions. On the following day, cells were serum starved for 3 h and treated with the appropriate agonist as indicated in the Figure legends. For immunostaining, cells were rinsed twice with PBS, fixed with 1% paraformaldehyde in PBS containing 4% sucrose for 15 min at room temperature (20°C) and permeabilized with 0.2% Triton X-100 for 5 min, and washed with PBS. After blocking with 10% BSA in PBS for 30 min at 37°C, cells were incubated with the primary antibody (mouse monoclonal anti-FLAG antibody; Abcam, ab18230) for 2 h at 37°C and washed with PBS, followed by 1 h incubation at 37°C with the secondary antibody (Alexa Fluor® 488 goat anti-mouse; Invitrogen, A-11001) and washed with PBS before incubation with DAPI (Invitrogen, D1306) for 10 min at 37°C. After washing the cells twice with PBS, confocal images were taken on a Zeiss LSM 510 META using a Plan-Apochromat X 63/1.4 oil objective. The cellular distribution of the protein constructs was recorded by line intensity profiles [52] using LSM Image Examiner software. Data were exported into SigmaPlot where the different cell widths were normalized for unit length while the total fluorescence intensity of each cell was taken as 100%. The ratios of pixel intensity (plasma membrane/cytoplasm) were then calculated using an image segmentation protocol [53].

Other materials and methods

Human recombinant AKT1 (124006) was purchased from EMD Millipore, insulin was bought from Sigma–Aldrich and PDGF (platelet-derived growth factor) was supplied by R&D Systems. The 1-InsP7, 5-InsP7 and InsP8 were prepared enzymatically as described previously [31]. Cytochalasin B was purchased from Sigma–Aldrich. Sources of other materials are as described previously [28]. Statistical analysis was performed with paired t tests (P<0.05 was considered significant). Densitometric analysis of Western blots was performed with ImageJ, and data were normalized to the signal obtained from loading controls.

RESULTS

PBD1 recapitulates receptor-dependent translocation of PPIP5K1 to the plasma membrane in vivo

We used confocal microscopy to analyse the impact of receptor activation by PDGF on the cellular distribution of FLAG-tagged PtdIns(3,4,5)P3-binding domains in NIH 3T3 cells (Figure 1). The effect of receptor activation on the intracellular distribution of our constructs was quantified using an image segmentation protocol (see the Materials and methods section). In control experiments, FLAG–GRP1 was distributed throughout the cell until PDGF was added, whereupon it translocated to the plasma membrane (Figures 1A and 1B), as in a previous study [54].

Figure 1 Effect of PDGF on the intracellular distribution of PPIP5K1 and PBD1 in NIH 3T3 cells

After 1 day in which NIH 3T3 cells were transfected with cDNA encoding FLAG-tagged constructs of either GRP1, PPIP5K1, PBD1 or PBD1R399A/R417A, cells were serum starved for 3 h (see the Materials and methods section) and then either vehicle (control) or 50 ng/ml PDGF were added for 5 min. Cells were then fixed and the distribution of each construct between the plasma membrane and the cytoplasm (ratio PM/C) determined by confocal microscopy and an image segmentation protocol [28]. The distribution of the FLAG constructs are shown in green, and blue indicates the DAPI-stained nucleus. The scale bar is 10 μm (A). The insets show cells from which the nuclear staining was digitally removed. (BE) The data are represented as means±S.E.M. **P<0.01 compared with control.

In unstimulated cells, FLAG–PPIP5K1 was distributed throughout the cytoplasm, but it was excluded from the nucleus (Figure 1A). We found that 5 min stimulation of NIH 3T3 cells with PDGF caused PPI5K1 to translocate from the cytoplasm to the plasma membrane (Figures 1A and 1C). Similar data were obtained in a previous study [28].

A modular view of PPIP5K1 [55] led us to previously propose that receptor-dependent translocation of PPIP5K1 is mediated solely by its PBD1 [28]. We have now tested that idea for the first time in intact cells. For these experiments, we transfected NIH 3T3 cells with FLAG–PBD1. Although some of the PBD1 was present in the nucleus, much of this protein was distributed throughout the cytoplasm (Figure 1A). The PBD1 translocated to the plasma membrane in response to PDGF (Figures 1A and 1D). A PBD1R399A/R417A mutant that has reduced binding to PtdIns(3,4,5)P3 [28] did not translocate to the plasma membrane after the addition of PDGF (Figure 1). These data led us to conclude that PBD1 is sufficient to recapitulate receptor-dependent translocation of PPIP5K1 to the plasma membrane in vivo.

Effects of PDGF on cellular levels of inositol lipids and PP-InsPs in NIH 3T3 cells

To gain insight into the biological relevance of ligand competition between PtdIns(3,4,5)P3 and PP-InsPs, we next measured the levels of inositol lipids, as well as InsP6 and PP-InsPs, under the same conditions that we used in the translocation experiments (see above). We found that 5 min stimulation with 50 ng/ml PDGF increased levels of PtdIns(3,4,5)P3 2-fold (Figure 2A). There was no significant effect on the levels of either PtdIns(3,4)P2 or PtdIns(4,5)P2 (Figures 2B and 2C). We also found that 5 min stimulation with PDGF did not alter the levels of either InsP6, InsP7 or InsP8 (Figures 2D, 2E and 2F). Some of these observations contrast sharply with the results of a previous study [19] with mouse embryo fibroblasts, in which levels of InsP6 and InsP7 increased 2-fold following 5 min stimulation with IGF-1 (insulin-like growth factor 1) (the levels of InsP8 were not described). In that study [19], cells were serum starved overnight before the addition of IGF-1. In contrast, we removed serum from our cells only 3 h before addition of PDGF. However, we must note that InsP6 and InsP7 may perhaps differentially respond to receptor activation in a cell-type and/or agonist-specific manner.

Figure 2 Effects of PDGF on the levels of inositol lipids InsP6 and PP-InsPs in NIH 3T3 cells

Either vehicle (control) or 50 ng/ml PDGF were added for 5 min to NIH 3T3 cells containing [3H]inositol that had been serum starved for 3 h (see the Materials and methods section). The incubations were then quenched, and either the [3H]inositol lipids (AC) or the [3H]inositol phosphates (DF) were extracted and quantified by HPLC. The data (either percentage of total inositol lipids or percentage of total inositol phosphates, as appropriate) are represented as means±S.E.M. for three experiments. *P<0.05 compared with control.

Effects of hyperosmotic stress on cellular levels of inositol lipids and PP-InsPs in NIH 3T3 cells

We next found that 30 min exposure of NIH 3T3 cells to hyperosmotic stress (by treatment with 0.2 M sorbitol) elevated PtdIns(3,4,5)P3 levels 4.5-fold (Figure 3A). Levels of PtdIns(3,4)P2 were unaffected by the addition of sorbitol (Figure 3B), but PtdIns(4,5)P2 levels fell slightly (Figure 3C), perhaps in part reflecting its increased hydrolysis due to PLC (phospholipase C) activation [56]. We also found that hyperosmotic stress slightly reduced InsP6 levels (Figure 3D). This effect has been observed previously in a study with HEK cells [4]. Levels of 5-InsP7 also decreased following hyperosmotic stress (Figure 3E), whereas InsP8 concentration increased 4.5-fold (Figure 3F), apparently reflecting an activation of PPIP5K [4]. Thus the PP-InsP profile in osmotically stressed cells (Figure 3) differs significantly from that in PDGF-stimulated cells (Figure 2). Could that impact the spatial dynamics of PtdIns(3,4,5)P3-binding domains in the osmotically challenged cells? That was the next question that we addressed.

Figure 3 Effects of hyperosmotic stress on the levels of inositol lipids InsP6 and PP-InsPs in NIH 3T3 cells

The serum-free medium bathing NIH 3T3 cells containing [3H]inositol (see the Materials and methods section) was replaced with 2 ml of fresh medium with or without the addition of 0.2 M sorbitol to induce hyperosmotic stress. After 30 min, the incubations were then quenched, and either the [3H]inositol lipids (AC) or the [3H]inositol phosphates (DF) were extracted and quantified by HPLC. The data (either percentage of total inositol lipids or percentage of total inositol phosphates, as appropriate) are represented as means±S.E.M. for five experiments. *P<0.05, **P<0.01 compared with control.

Hyperosmotic stress promotes translocation of GRP1 and PPIP5K1 to the plasma membrane

Several studies have reported that two proteins that contain PH domains, Akt and PDK1 (phosphoinositide-dependent kinase 1), do not translocate to the plasma membrane following a hyperosmotic challenge [34,4143]. In contrast, we found that GRP1 translocated to the plasma membrane after cells were subjected to hyperosmotic stress (Figures 4A and 4B), to a similar degree to that observed after PDGF treatment (Figures 1A and 1B). To our knowledge, this recruitment of GRP1 reveals a novel cellular response to hyperosmotic stress, and may reflect adaptive responses of membrane trafficking, which GRP1 can regulate [22,23]. The observation that GRP1PH translocated to the plasma membrane (Figures 4A and 4B), despite the accompanying increase in InsP8 levels (Figure 3F), does not support speculation [20] that InsP8 might be an especially potent inhibitor of PH domain association with PtdIns(3,4,5)P3.

Figure 4 Effect of hyperosmotic stress on the intracellular distribution of PPIP5K1 and PBD1 in NIH 3T3 cells

At 1 day after NIH 3T3 cells were transfected with cDNA encoding FLAG-tagged constructs of either GRP1, PPIP5K1, PBD1 or PBD1R399A/R417A, cells were serum starved for 3 h (see the Materials and methods section). Then, the medium was replaced with 2 ml of fresh medium with or without 0.2 M sorbitol added to induce hyperosmotic stress. Where indicated, cells were treated with 5 μg/ml cytochalasin, 30 min before initiation of osmotic stress. Cells were then fixed and the distribution of each construct between the plasma membrane and the cytoplasm (ratio PM/C) was determined by confocal microscopy and an image segmentation protocol [28]. The distribution of the FLAG epitope is shown in green, and blue indicates the DAPI-stained nucleus. The scale bar is 10 μm (A). (BD) The data are represented as means±S.E.M. **P<0.01 compared with control.

We found further that PPIP5K1 translocated to the plasma membrane after hyperosmotic stress (Figures 4A and 4C). The degree of that translocation (Figures 4A and 4C) was similar to that which was observed after PDGF treatment (Figures 1A and 1C). Control experiments with cytochalasin (Figure 4A) indicated that the translocation of PPIP5K was not secondary to cytoskeletal rearrangement, as has been reported for some PH-domain proteins [21]. Stress-induced translocation of PPIP5K was recapitulated by the PBD1, but not by the R399A/R417A mutant (Figures 4A and 4D) that had lost its high affinity for PtdIns(3,4,5)P3 [28].

InsP6 and PP-InsPs compete with PtdIns(3,4,5)P3 for binding to the PBD1

Snyder and co-workers [20] have reported that 5-InsP7 competes with PtdIns(3,4,5)P3 for certain PH domains. In the latter experiments, ligand competition was determined in co-sedimentation assays with PtdIns(3,4,5)P3 in lipid vesicles. We have studied the relative affinities of PBD1 for 5-InsP7 and the other PP-InsPs. We used SPR, which, as far as we are aware, has not been applied previously to the study of ligand competition between PP-InsPs and PtdIns(3,4,5)P3 for any protein.

We first added the PBD1 of PPIP5K to immobilized lipid vesicles that contained 3% (v/v) PtdIns(3,4,5)P3. For biological relevance, the mobile phase contained 100 mM KCl, 10 mM NaH2PO4 and 1 mM MgCl2. In these experiments, the size of the sensorgram signal (the broken line labelled ‘control’ in Supplementary Figure S1 at http://www.biochemj.org/bj/453/bj4530413add.htm) corresponds to the degree of association of PBD1 to the lipid vesicles. No signal was observed in the absence of PtdIns(3,4,5)P3 (not shown; see also [28]). We found that as little as 1 μM 5-InsP7 (a physiological concentration; [57]) reduced (by approximately 15%) the extent to which PBD1 associated with vesicle-bound PtdIns(3,4,5)P3 (Supplementary Figure S1). We further found that 1 μM InsP6 was approximately half as effective a ligand as 5-InsP7 (Supplementary Figure S1).

Snyder and co-workers [20] used Mg2+-free medium in their experiments with PtdIns(3,4,5)P3-binding domains. We found that the removal of Mg2+ from our buffers did not influence PBD1 binding to PtdIns(3,4,5)P3 (Supplementary Figure S1). The absence of Mg2+ did slightly increase the potency with which 5-InsP7 and InsP6 competed with PtdIns(3,4,5)P3, but without affecting the marginal preference for 5-InsP7 as a ligand (Supplementary Figure S1). The presence of Mg2+ is more physiological, so we retained it in our further experiments described below.

More detailed ligand competition experiments (Figure 5) were used to determine the IC50 values for InsP6 and 5-InsP7 competing with PBD1 binding to PtdIns(3,4,5)P3 (14 μM and 7 μM respectively). What could be the physiological consequence of ligand competition between PtdIns(3,4,5)P3 and InsP6 or 5-InsP7? In resting cells, in which PPIP5K1 is present in the cytoplasm (Figure 1B), we can predict from the data described in Figure 5 that InsP6 and/or 5-InsP7 would likely be associated with the PBD1. We propose that these inositol phosphates would tend to elevate the threshold concentration of PtdIns(3,4,5)P3 that would need to be achieved before it would recruit PPIP5K to the plasma membrane. This might be a means by which the cell could ensure that such translocation of PPIP5K1 only occurred in response to a sustained, stimulus-dependent elevation in PtdIns(3,4,5)P3, rather than to random fluctuations (‘noise’) in lipid levels.

Figure 5 SPR-based analysis of the effects of InsP6 and PP-InsPs on binding of the PBD1 of PPIP5K1 to PtdIns(3,4,5)P3

(AD) The PBD1 (100 nM), either alone or after 15 min pre-incubation with the designated inositol phosphate (up to 50 μM as indicated), was injected at a flow rate of 5 μl/min on to immobilized lipid vesicles containing 3% PtdIns(3,4,5)P3. The mobile phase contained 25 mM Hepes (pH 7.2), 100 mM KCl, 10 mM NaH2PO4 and 1 mM MgCl2. Representative sensograms are plotted in arbitrary units (AU). In (A) and (B), data were also obtained with 0.1 μM inositol phosphate, but the sensorgrams were not significantly different from those obtained in the absence of ligand, and so are omitted for clarity. In additional experiments (results not shown), the effects of chemically synthesized 5-InsP7 [kindly provided by Dr D. Fiedler (University of Princeton, Princeton, NJ, U.S.A.)] were found to be identical to our enzymatically synthesized 5-InsP7 (see the Materials and methods section). (EH) End points are plotted against inositol phosphate concentrations.

Once concentrated at the plasma membrane, the kinase activity of PPIP5K1 could deplete the levels of InsP6 and 5-InsP7 and elevate levels of 1-InsP7 and InsP8. These metabolic effects could be localized because of the reduced rates of diffusion in the narrow, subplasmalemmal ‘unstirred’ space [58]. To explore the possible consequences of such compartmentalization, we next examined the extent to which 1-InsP7 and InsP8 might inhibit binding of PBD1 to PtdIns(3,4,5)P3. Such experiments represent a new development in inositol pyrophosphate research. To date, 5-InsP7 is the only member of the PP-InsP family that has been investigated as a possible ligand for any PtdIns(3,4,5)P3-binding domains, although there has been speculation of the potential biological significance if InsP8 were to be a more potent ligand [7,9,20].

Our binding data (Figure 5) show that 1-InsP7 (IC50=43 μM) and InsP8 (IC50=32 μM) exhibit 6–7-fold lower affinity for PBD1 than InsP6 and 5-InsP7. That is to say, the products of PPIP5K kinase activity, 1-InsP7 and InsP8, were considerably weaker ligands than the substrates InsP6 and 5-InsP7 respectively. These data raise the possibility that, following its translocation to the plasma membrane, the kinase activity of PPIP5K might generate a microenvironment in which there is a reduction in the degree of competition of inositol phosphates with PtdIns(3,4,5)P3 for the PBD1.

InsP6 and PP-InsPs compete with PtdIns(3,4,5)P3 for the PH domains of GRP1 and Akt

Snyder and co-workers [20] have studied competition of both InsP6 and 5-InsP7 for PtdIns(3,4,5)P3 binding by a range of PH domains; in each case, 5-InsP7 was reported to be 10–20-fold more potent at competing with PtdIns(3,4,5)P3 than InsP6. The PBD1 of PPIP5K1 did not exhibit that high degree of ligand specificity (Figure 5). So we next studied binding of InsP6 and PP-InsPs to some PH domains. Both of these inositol phosphates inhibited GRP1PH association with PtdIns(3,4,5)P3; the effects of 10 μM were particularly dramatic (Supplementary Figure S2 at http://www.biochemj.org/bj/453/bj4530413add.htm). Nevertheless, 5-InsP7 was only marginally more potent than InsP6 (Supplementary Figure S2). Moreover, as was the case with PBD1, we found that the products of PPIP5K kinase activity, 1-InsP7 and InsP8, were considerably weaker ligands for GRP1PH than InsP6 and 5-InsP7 substrates (Supplementary Figure S2). These data led us to hypothesize that the kinase activity of PPIP5K1 at the plasma membrane might generate a localized environment that favours GRP1 association with PtdIns(3,4,5)P3. We next considered whether this proposed function of PPIP5K1 might be relevant to other proteins with PtdIns(3,4,5)P3-selective PH domains.

It has been Akt that has received most attention with regard to competition between PtdIns(3,4,5)P3 and 5-InsP7 [1820]. Thus we next studied the interaction of InsP6 and the PP-InsPs with Akt (Figures 6A–6D) from which we obtained IC50 data (Figures 6E–6H). In these experiments, InsP6 (IC50=14 μM) was half as potent as 5-InsP7 (IC50=7 μM) at competing with PtdIns(3,4,5)P3. That difference in relative affinities of InsP6 and 5-InsP7 for Akt is much lower than the 10–20-fold difference reported previously [20]. We further found that the affinities of Akt for 1-InsP7 (IC50=45 μM) and InsP8 (IC50>50 μM) were severalfold lower than those for InsP6 and 5-InsP7 (Figure 6). These results with Akt mirror those that were obtained with GRP1PH (Supplementary Figure S2) and PBD1 (Figure 5): the products of PPIP5K kinase activity, 1-InsP7 and InsP8, were weaker ligands than the substrates InsP6 and 5-InsP7.

Figure 6 SPR-based analysis of the effects of InsP6 and PP-InsPs on binding of AktPH to PtdIns(3,4,5)P3

(AD) The AktPH (100 nM), either alone or after 15 min pre-incubation with the designated inositol phosphate (0–50 μM as indicated), was injected at a flow rate of 5 μl/min on to immobilized lipid vesicles containing 3% PtdIns(3,4,5)P3. The mobile phase contained 25 mM Hepes (pH 7.2), 100 mM KCl, 10 mM NaH2PO4 and 1 mM MgCl2. Representative sensorgrams are plotted in arbitrary units (AU). In (A) and (B), data were also obtained with 1 μM inositol phosphate, but the sensorgrams were not significantly different from those obtained in the absence of ligand, and so are omitted for clarity. (EH) End points are plotted against inositol phosphate concentrations.

Competitive binding of the PH domain of SIN1 to PtdIns(3,4,5)P3, InsP6 and PP-InsPs

We next studied the interactions of inositol pyrophosphates with SIN1PH; its crystal structure was published recently [25]. We found that SIN1PH bound to PtdIns(3,4,5)P3 in lipid vesicles with an affinity (Kd=141 nM; Figures 7A and 7B) that was similar to the values determined previously by SPR, under comparable experimental conditions, for PBD1 (Kd=96 nM [28]) and GRP1PH (Kd=59–170 nM [28,59]); AktPH has been reported to have a 4-fold higher Kd valve for PtdIns(3,4,5)P3 (35 nM [60]). Some binding of SIN1PH to PtdIns(4,5)P2 was also detected, but with severalfold lower affinity (Kd=629 nM; Figures 7B and 7D) than that for PtdIns(3,4,5)P3.

Figure 7 SPR-based equilibrium-binding analysis of the affinity of SIN1PH for PtdIns(3,4,5)P3 and PtdIns(4,5)P2

SIN1PH was injected at 5 μl/min over immobilized lipid vesicles containing 3% of either (A and B) PtdIns(3,4,5)P3 (25, 50, 90, 125, 200, 375, 600 or 700 nM protein) or (C and D) PtdIns(4,5)P2 (100, 200, 250, 500, 1200 or 1800 nM protein). (A) and (C) show representative sensorgrams, with the same y-axis scale for comparative purposes. (B) and (D) depict the Kd determinations.

In ligand competition experiments (Figure 8), we found that the IC50 values for InsP6 and 5-InsP7 were very similar at 6 and 3.7 μM respectively. Limitations of ligand supply prevented us from precisely determining the much weaker affinities for 1-InsP7 and InsP8, but by extrapolation we estimate the IC50 values to exceed 100 μM (Figure 8). Compared with the other PH domains that we have tested, SIN1PH exhibited the greatest degree of discrimination between the substrates and the products of PPIP5K1 activity. Nevertheless, the identical rank order of affinities for InsP6 and PP-InsPs exhibited by PBD1, GRP1, Akt and SIN1 suggest that this could be a property of other PtdIns(3,4,5)P3-selective PH domains.

Figure 8 SPR-based analysis of the effects of InsP6 and PP-InsPs on binding of SIN1PH to PtdIns(3,4,5)P3

(AD) The SIN1PH (100 nM), either alone or after 15 min pre-incubation with the designated inositol phosphate (0–50 μM as indicated), was injected at a flow rate of 5 μl/min on to immobilized lipid vesicles containing 3% PtdIns(3,4,5)P3. The mobile phase contained 25 mM Hepes (pH 7.2), 100 mM KCl, 10 mM NaH2PO4 and 1 mM MgCl2. Representative sensograms are plotted in arbitrary units (AU). (EH) End points are plotted against inositol phosphate concentrations.

Effects of PPIP5K1 knockdown on a PtdIns(3,4,5)P3-signalling cascade in intact cells: SIN1-dependent Akt phosphorylation

The data described above led us to hypothesize that, following translocation to the plasma membrane of PPIP5K1, its kinase activity may generate a localized environment of inositol phosphate levels that favours the association of PH domain proteins with PtdIns(3,4,5)P3. We next interrogated this concept in intact cells, by investigating whether a reduction in PPIP5K1 expression would inhibit the activity of an endogenously expressed PtdIns(3,4,5)P3-dependent kinase cascade. Among the proteins that we have studied (see above), it is the PH domain in SIN1 that shows the greatest degree of difference in relative affinities for the substrates versus the products of PPIP5K. We therefore considered that an assay for SIN1 function might provide the more sensitive readout of the biological consequences of PPIP5K1 knockdown.

SIN1 mediates PtdIns(3,4,5)P3-dependent translocation of the mTORC2 to the plasma membrane [26]. The mTOR within mTORC2 phosphorylates Ser473 on Akt [26,27]. Thus genetic ablation of SIN1 eliminated stimulus-dependent Ser473 phosphorylation and, significantly, without impacting the degree of phosphorylation of Thr308 by PDK1 [27]. We therefore used Ser473 phosphorylation of Akt as a biological readout in our search for possible changes in SIN1 function following knockdown of PPIP5K1 expression. We used the L6 skeletal myoblast cell line for these experiments, since previous work [3] has shown that PPIP5K1 is relatively highly expressed in skeletal muscle; we posited, therefore, that the consequences of PPIP5K1 knockdown might be particularly significant in a skeletal muscle cell model. L6 cells were also considered suitable for these experiments because, similar to NIH 3T3 cells (see above), there was a persistent translocation of FLAG–PPIP5K1 to the plasma membrane when cells were stimulated with an agonist (insulin) that promotes PtdIns(3,4,5)P3 accumulation (Supplementary Figure S3 at http://www.biochemj.org/bj/453/bj4530413add.htm).

By using retroviral delivery of shRNA against PPIP5K1, we created a new L6 cell line (L6K1-KD) in which the levels of PPIP5K protein were 70% lower than those in L6CONTROL that was infected with a non-effective shRNA (Figures 9A–9C). Importantly, the degree of expression of SIN1 protein was similar in these two cell lines (Figure 9A). The PPIP5K1 knockdown did not affect the levels of InsP6 (Figure 9D), but the levels of InsP7 (Figure 9E) and InsP8 (Figure 9F) were 17% (P<0.02) and 37% (P<0.001) lower in L6K1-KD cells compared with L6CONTROL cells. These data are consistent with PPIP5K1 contributing to both 1-InsP7 and InsP8 synthesis in intact cells. The fact that the diminution in InsP8 was less than the degree of the knockdown may reflect the presence of PPIP5K2 in these cells.

Figure 9 Cellular levels of inositol pyrophosphates and insulin-mediated Akt phosphorylation on Ser473 in L6K1-KD and L6CONTROL cells

L6K1-KD (KD) and L6CONTROL (Con) cells were prepared as described in the Materials and methods section. (A) Representative Western blot analysis of the degree of expression of PPIP5K1 and SIN1, with β-actin loading controls. (B) Densitometric analysis of PPIP5K1 protein expression, means±S.E.M. for three experiments (expression in L6CONTROL cells was normalized to unity). (C) Quantitative real-time PCR analysis of PPIP5K1 mRNA expression (means±S.E.M. for four experiments; expression in L6CONTROL cells was normalized to unity). (D) Total [3H]InsP6. (E) Total [3H]InsP7. (F) Total [3H]InsP8. The data in (D), (E) and (F) represent means±S.E.M. for six experiments with serum-starved cells. (G) Representative Western blot analysis of Akt phosphorylation on Ser473, and total Akt loading controls, after cells were treated with 10 nM insulin for 0, 5, 15 and 30 min. (H) Densitometric analysis of Akt phosphorylation on Ser473, as a ratio to total Akt (means±S.E.M. for five experiments, performed as in G; all values were normalized to the signal in L6CONTROL cells after 30 min insulin treatment, since no Ser473 phosphorylation was detected at zero min). (I) Proliferation assay. *P<0.05, **P<0.01, ***P<0.001 (compared with control).

Knockdown of PPIP5K1 expression did not affect insulin-stimulated PDK1-mediated [26] phosphorylation of Thr308 on Akt (Supplementary Figure S4 at http://www.biochemj.org/bj/453/bj4530413add.htm), indicating that there was near-normal recruitment of Akt to the plasma membrane. Indeed, we observed similar rates of insulin-dependent translocation of exogenous GFP–AktPH to the plasma membrane in L6K1-KD and L6CONTROL cells (Supplementary Movie 1 at http://www.biochemj.org/bj/453/bj4530413add.htm). On the other hand, we found that at 5 and 15 min after the addition of insulin, the degree of phosphorylation of Ser473 on Akt was 40% lower in L6K1-KD cells than in L6CONTROL cells (Figures 9G and 9H). These data are consistent with knockdown of PPIP5K1 impairing PtdIns(3,4,5)P3-mediated SIN1-dependent recruitment of mTORC2 to the plasma membrane, and hence being responsible for a reduction in the extent to which mTOR phosphorylates Akt. Finally, since the mTORC2 complex to which SIN1 belongs is believed to contribute to the regulation of cell growth, we measured the rates of proliferation of the L6K1-KD and L6CONTROL cells (Figure 9I). The knockdown of PPIP5K1 expression reduced the rate of cell growth by approximately 30% (Figure 9I).

DISCUSSION

The present study provides new information concerning the participation of PPIP5K in cell signalling. For example, we show that the stimulus-dependent PtdIns(3,4,5)P3-mediated translocation of PPIP5K1 to the plasma membrane can be fully recapitulated by its PBD1 (Figure 1). We further show that this translocation of PPIP5K1 is more widespread than was originally appreciated, since it occurs following the increases in PtdIns(3,4,5)P3 that result from either receptor activation or hyperosmotic stress (Figures 14). We have also conducted the first analysis of the specificity with which PP-InsPs compete for PtdIns(3,4,5)P3-binding domains (Figures 5, 6 and 8, and Supplementary Figure S2). Consequently, we have discovered that binding of PtdIns(3,4,5)P3 is more strongly competed by the substrates of PPIP5K (InsP6 and 5-InsP7) than by the products (1-InsP7 and InsP8). We propose that, in resting cells, ligand competition from cytoplasmic InsP6 and PP-InsPs will restrain non-sustained stochastic increases in PtdIns(3,4,5)P3 from recruiting signalling proteins to the plasma membrane. However, once more prolonged stimulus-dependent increases in PtdIns(3,4,5)P3 outcompete binding of inositol phosphates, PPIP5K1 will then translocate to the plasma membrane (Figures 1 and 4). The ensuing local increase in PPIP5K1 catalytic activity may sustain a subplasmalemmal microenvironment in which the concentrations of InsP6 and 5-InsP7 are lower than those in the cytosol. This, in turn, would support further translocation of both PPIP5K1 and PtdIns(3,4,5)P3-binding PH domains (Figure 10). That represents a potential positive feedback mechanism; many signalling cascades rely on such amplification in order to prevent mechanisms of noise suppression constraining response sensitivity [61]. Thus optimum signalling by PtdIns(3,4,5)P3 may depend on coincidence detection: the translocation of both PPIP5K and a PH-domain protein. We obtained data in support of this model by showing decreased activation of an endogenous PtdIns(3,4,5)P3 signalling cascade in a cell line in which PPIP5K1 expression was knocked down (Figure 9).

Figure 10 Proposed biological significance of stimulus-dependent PtdIns(3,4,5)P3-mediated translocation of PPIP5K1 to the plasma membrane

Recruitment of PPIP5K1 to the plasma membrane driven by stimulus-dependent synthesis of PtdIns(3,4,5)P3 by PI3K (phosphoinositide 3-kinase). The catalytic activity of PPIP5K1 is hypothesized to locally decrease subplasmalemmal levels of InsP6 and 5-InsP7 through their phosphorylation to 1-InsP7 and InsP8 respectively. This, in turn, is predicted to relieve inhibition of the binding to PtdIns(3,4,5)P3 of signalling proteins with PH domains. Thus PPIP5K1 recruitment to the plasma membrane is considered to activate PtdIns(3,4,5)P3 cascades.

We determined that there are approximately 2-fold differences in the affinities with which InsP6 and 5-InsP7 bind to Akt and certain other PH domains; this is a less dramatic discrimination than the 10- to 20-fold differences reported by others [9,20]. Methodological differences may be responsible; we used SPR, a favoured technique for quantitative analysis of ligand–protein interactions [32], and we employed buffers that approximated physiological conditions. Additionally, in our hands, the rank order of relative affinities of InsP6 and PP-InsPs for PtdIns(3,4,5)P3-selective protein domains (5-InsP7>InsP6>1-InsP7 and InsP8) did not correlate with phosphate number. This new information concerning the interactions of PP-InsPs with PH domains not only helps to determine its biological significance, but could also assist the goal of developing therapeutically relevant drugs that target PH domain–PtdIns(3,4,5)P3 interactions [62].

Our hypothesis that a localized catalytic activity of PPIP5K1 might regulate PtdIns(3,4,5)P3 signalling cascades is strengthened by a body of evidence describing reduced rates of diffusion in the narrow, subplasmalemmal ‘unstirred’ space [58]. For example, in the subplasmalemmal space the diffusion coefficient of cAMP has been estimated to be seven orders of magnitude slower than that in the bulk of the cytoplasm [58]. Also of relevance is an emerging opinion [7,16,63] that the free cytoplasmic concentration ratio of InsP6/InsP7 is lower than that suggested by assays of total cellular InsP6, due to the latter's compartmentalization. For example, InsP6 is divided between metabolically inert and (smaller) receptor-regulated pools [63]; the latter description of metabolic compartmentalization implies some physical separation of these different InsP6 pools. Indeed, the mammalian Ins(1,3,4,5,6)P5 2-kinase that produces InsP6 is predominantly a nuclear enzyme [64]. Furthermore, the only known mammalian InsP6 phosphatase (MINPP; multiple inositol polyphosphate phosphatase) is restricted to the lumen of endoplasmic reticulum and Golgi [65,66]. Perhaps InsP6 enters a luminal pool through a transmembrane InsP6 transport protein such as that which is found in plants [67]. Other evidence of a cellular pool of InsP6 that does not freely exchange with the bulk cytoplasm comes from the demonstration that this polyphosphate is a structural cofactor that is deeply buried within adenosine deaminase [68]. There are other known examples of InsP6 fulfilling this structural function in plants [69], and so its wider applicability to proteins in mammalian cells can be anticipated. Even electrostatic binding of InsP6 to the surface of proteins may significantly reduce its soluble concentration, especially as so many proteins can bind to this inositol phosphate [70].

We have also used a PPIP5K1 knockdown cell model to directly challenge our new hypothesis that PtdIns(3,4,5)P3 signalling cascades might be regulated by PPIP5K1 (Figure 9). To maximize biological relevance, we looked for an effect of this knockdown on an endogenously expressed PtdIns(3,4,5)P3 signalling cascade, rather than one in which the component under study was overexpressed. We also anticipated that another important factor might be the relative affinities of a PtdIns(3,4,5)P3-binding protein for the substrates and products of PPIP5K1. The latter consideration led us to focus on SIN1; 1-InsP7 and InsP8 bound to SIN1 with almost two orders of magnitude lower affinity than InsP6 and 5-InsP7 (Figure 8). This difference in relative affinities was considerably larger than those of the other PtdIns(3,4,5)P3-binding domains that we investigated in the present study (Figures 5 and 6, and Supplementary Figure S2).

SIN1 mediates PtdIns(3,4,5)P3-dependent translocation of the mTORC2 multiprotein complex to the plasma membrane [26]. The mTOR within mTORC2 phosphorylates Ser473 on Akt [26,27]. Using the degree of Ser473 phosphorylation as a readout for SIN1-mediated PtdIns(3,4,5)P3-dependent recruitment of mTORC2, we recorded up to 40% inhibition of this signalling cascade in the L6K1-KD cells (Figure 9). We interpret these data as evidence in support of our new hypothesis concerning the biological significance of PPIP5K1 translocation. Further work will be necessary to determine whether other PtdIns(3,4,5)P3-binding PH domains might be regulated by PPIP5K1 in vivo; our SPR data offer Akt (Figure 6) and GRP1 (Supplementary Figure S2) as candidates. If this new action of PPIP5K does regulate many physiological aspects of PtdIns(3,4,5)P3 signalling, it would represent a somewhat ironic demonstration that the promiscuity of inositol pyrophosphate signalling does not arise from a multiplicity of biological actions of these molecules, but instead depends on both 1-InsP7 and InsP8 being relatively inactive.

AUTHOR CONTRIBUTION

All authors designed and performed the experiments, analysed the data, and contributed to writing the paper.

FUNDING

This research was supported by the Intramural Research Program of the National Institutes of Health (NIH)/NIEHS.

Acknowledgments

We are grateful to the laboratory of Dr D. Sarbassov for providing reagents and experimental protocols that greatly assisted the immunodetection of SIN1. We also acknowledge the considerable help that we received from the viral core and the protein expression core at the National Institute of Environmental Health Sciences (NIEHS).

Abbreviations: GRP1, general receptor for phosphoinositides-1; GRP1PH, PH domain of GRP1; HEK, human embryonic kidney; IGF-1, insulin-like growth factor 1; IP6K, inositol hexakisphosphate kinase; L6CONTROL, control L6 cells; L6K1-KD, L6 cell line with reduced (‘knocked-down’) expression of PPIP5K1; MEM, minimal essential medium; mTOR, mammalian (or mechanistic) target of rapamycin; mTORC, mTOR complex; PBD1, polyphosphoinositide-binding domain; PDGF, platelet-derived growth factor; PDK1, phosphoinositide-dependent kinase 1; PH, pleckstrin homology; PP-InsP, diphosphoinositol polyphosphate; PPIP5K1, diphosphoinositol pentakisphosphate kinase type 1; SIN1, stress-activated protein kinase-interaction protein-1; SIN1PH, PH domain of SIN1; SPR, surface plasmon resonance

References

View Abstract