Cell Architecture: from Structure to Function
The regulation of membrane to cytosol partitioning of signalling proteins by phosphoinositides and their soluble headgroups C.P. Downes1 , A. Gray, A. Fairservice, S.T. Safrany2 , I.H. Batty and I. Fleming3 Division of Molecular Physiology, School of Life Sciences, MSI/WTB Complex, University of Dundee, Dundee DD1 5EH, U.K.
Abstract Inositol phospholipids [PIs (phosphoinositides)] represent a group of membrane-tethered signalling molecules which differ with respect to the number and distribution of monoester phosphate groups around the inositol ring. They function by binding to proteins which possess one of several domains that bind a particular PI species, often with high affinity and specificity. PH (pleckstrin homology) domains for example possess ligand-binding pockets that are often lined with positively charged residues and which bind PIs with varying degrees of specificity. Several PH domains bind not only PIs, but also their cognate headgroups, many of which occur naturally in cells as relatively abundant cytosolic inositol phosphates. The subcellular distributions of proteins possessing such PH domains are therefore determined by the relative levels of competing membrane-bound and soluble ligands. A classic example of the latter is the PH domain of phospholipase Cδ 1 , which binds both phosphatidylinositol 4,5-bisphosphate and inositol 1,4,5-trisphosphate. We have shown that the N-terminal PH domain of the Rho family guanine nucleotide-exchange factor, Tiam 1, binds PI ligands promiscuously allowing multiple modes of regulation. We also recently analysed the ligand-binding specificity of the PH domain of PI-dependent kinase 1 and found that it could bind abundant inositol polyphosphates such as inositol hexakisphosphate. This could explain the dual distribution of this key signalling component, which needs to access substrates at both the plasma membrane and in the cytosol.
Introduction Pleckstrin homology (PH) domains are protein modules of approx. 120 amino acids with approx. 250 representatives within the human genome [1–3]. Whilst many PH domains bind PIs (phosphoinositides), they often do so with relatively low affinity and/or specificity [1,2]. We have characterized the ligand-binding properties of several PH domains with the aim of defining a set of protein modules that can be used as probes for each of the currently known PI species. Specifically, we have made use of the PH domains of PLCδ 1 (phosphoinositide-specific phospholipase Cδ 1 ), general receptor for PI-1 and tandem PH-domain-containing protein 1 as specific reagents for the detection of PtdIns(4,5)P2 , PtdIns(3,4)P2 and PtdIns(3,4,5)P3 respectively in a variety of contexts. For example, they can be used as GFP (green fluorescent protein)-fusion proteins expressed in live cells to Key words: cytosol partitioning, phosphoinositide (PI), phosphoinositide-dependent kinase 1 (PDK1), pleckstrin homology domain (PH domain), signalling protein, Tiam1. Abbreviations used: AGC, cAMP-stimulated, cGMP-stimulated and C-kinase; APC, allophycocyanin; CamKII, calcium/calmodulin-dependent kinase II; DH, Dbl homology; FRET, fluorescence resonance energy transfer; GEF, guanine nucleotide-exchange factor; GFP, green fluorescent protein; GST, glutathione S-transferase; mTOR, mammalian target of rapamycin; NPH-Tiam1, N-terminal PH domain of Tiam1; PDK, phosphoinositide-dependent kinase; PH, pleckstrin homology; PI, phosphoinositide; PKB, protein kinase B; PI3K, phosphoinositide 3-kinase; PLC, phosphoinositide-specific phospholipase C; PH-PLCδ 1 , PH domain of PLCδ 1 . 1 To whom correspondence should be addressed (email
[email protected]). 2 Present address: Department of Pharmacy and Pharmacology, University of Bath, Claverton Road, Bath BA2 7AY, U.K. Present address: Cyclacel Ltd., James Lindsay Place, Dundee DD1 5JJ, U.K.
3
analyse spatial and temporal aspects of PI signalling processes. GST (glutathione S-transferase)-fusion proteins of these probes have also been used for ultrastructural analyses of their cognate lipid ligands by quantitative immunoelectron microscopy [4,5], revealing pools of each of these lipids at intracellular locations as well as at the plasma membrane. We also developed a FRET (fluorescence resonance energy transfer)based approach, which allows detection of PIs at picomolar sensitivity and can be used for mass assays of several lipid species and as a high throughput screening platform for PI kinases and phosphatases [6]. PH domains consist of two orthogonally arranged β-sheets with one end of this structure blocked off by an amphipathic α helix. They bind target ligands through electrostatic interactions involving positively charged protein residues and the monoester phosphate groups of the PI headgroup. Hence several PH domains have been found to interact quite effectively with soluble inositol phosphates that have the same or similar configuration to established lipid ligands [7–9]. This therefore raises the question of whether promiscuous ligand binding is a common feature of PH domains and how such promiscuity might contribute to the functions of the proteins with which they are associated. In this paper, we discuss three examples of PH domains which bind both PI and inositol phosphate ligands, each with a distinct physiological purpose. The PH domain of PLCδ 1 (PH-PLCδ 1 ) tethers the enzyme to substrate-bearing membranes, an effect that is antagonized by C 2005
Biochemical Society
1303
1304
Biochemical Society Transactions (2005) Volume 33, part 6
Ins(1,4,5)P3 , one product of PLC activity [9]. The N-terminal PH domain of the Rho family GEF (guanine nucleotideexchange factor), Tiam1 (NPH-Tiam1; where NPH stands for N-terminal PH), activates GTP exchange activity when bound to PI3K (phosphoinositide 3-kinase) lipid products and translocates to the cytosol in the presence of several inositol phosphate species [10]. Lastly, the PH domain of PDK1 (phosphoinositide-dependent kinase 1) has an unusual structure that allows binding of PIs as well as inositol polyphosphates such as InsP6 . This may account for the dual distribution of PDK1 between plasma membranes and cytosol, which is required for it to access the full range of its substrates [11].
PH-PLCδ 1 binds mutually antagonistic ligands Several studies have shown that the binding of PtdIns(4,5)P2 to PH-PLCδ 1 targets the holoprotein to the plasma membrane, its presumed site of action. The soluble product of PLC activity, Ins(1,4,5)P3 , has also been shown to have an affinity for this PH domain that is at least as great as that of the parent lipid. This appears to be physiologically significant because stimulation of Ins(1,4,5)P3 production following receptor-stimulated activation of PLC causes rapid translocation of GFP-tagged PH-PLCδ 1 from the plasma membrane to the cytosol resulting in feedback inhibition of PLC activity. This response has been utilized by several groups as a quantitative, spatial and temporal means of following PLC activity [9,12]. It remained unclear, however, whether the GFPtagged probe was detecting the increase in Ins(1,4,5)P3 concentration or the decline in PtdIns(4,5)P2 concentration that occurs concomitantly. To address this, we examined the cytosolic translocation of GFP-tagged PH-PLCδ 1 in 1321N1 astrocytoma cells stimulated with a thrombin receptor agonist and in parallel measured PtdIns(4,5)P2 and Ins(1,4,5)P3 levels biochemically [10]. Activation of thrombin receptors caused rapid loss and then recovery of PtdIns(4,5)P2 levels, a pattern mirrored by an increase and return to near basal levels of Ins(1,4,5)P3 and by the translocation of PH-PLCδ 1 from membranes to cytosol and back again. When this experiment was repeated in the presence of a high dose of wortmannin, to block PtdIns 4-kinase activity, the behaviours of Ins(1,4,5)P3 and PH-PLCδ 1 were similar, but the PtdIns(4,5)P2 concentration remained depressed over a prolonged period indicating the failure to resynthesize PtdIns(4,5)P2 under these conditions. This result implies that it is the production of Ins(1,4,5)P3 rather than the decline in PtdIns(4,5)P2 that has the major effect on PH-PLCδ 1 translocation in this system [10] emphasizing the feedback inhibitory role of Ins(1,4,5)P3 on PLC activity.
NPH-Tiam1 binds multiple ligands with distinct functions Tiam1 is a Rho family GEF, which activates Rac and/or Cdc42 by enhancing the release of GDP and hence increasing C 2005
Biochemical Society
the proportion of GTP-bound, active GTPase. Members of this family of GEFs contain at least one PH domain, closely associated with the active site DH (Dbl homology) domain. Tiam1 comprises a DH domain, consensus sequences for phosphorylation by several protein kinases, a PEST domain, a Discs-large homology region and two PH domains, one located on each side of the DH domain [13]. NPH-Tiam1 is required for constitutive targeting of Tiam1 to the plasma membrane and binds PIs in the rank order PtdIns(3,4,5)P3 > PtdIns(3,4)P2 > PtdIns(4,5)P2 , but only the 3-PIs are capable of stimulating its Rac-GEF activity [14]. Nevertheless, due to its much greater abundance in cells, PtdIns(4,5)P2 remained a candidate ligand for plasma membrane targeting. To address this, we performed experiments in 1321N1 astrocytoma cells similar to those described above for PHPLCδ 1 [10]. Indeed GFP-tagged C1199-Tiam1 (a truncated version of Tiam1 lacking the N-terminal PEST sequence) translocated from plasma membranes to cytosol upon stimulation of PLC-coupled thrombin receptors. The transient nature of this response was similar to that observed using PH-PLCδ 1 and was not affected by pretreatment with 10 µM wortmannin, indicating that it was due to PLC-mediated production of soluble inositol phosphates rather than the decline in PtdIns(4,5)P2 concentration. It was also possible to show cytosolic translocation of endogenous Tiam1 in response to activation of thrombin receptors [10]. Several inositol phosphate species are produced in cells following PLC activation due to the complex metabolism of Ins(1,4,5)P3 . Therefore we investigated the specificity of this translocation event by micro-injecting individual inositol phosphate species into GFP-tagged C1199-Tiam1 transfected cells. The naturally occurring inositol phosphates, Ins(1,4,5)P3 , Ins(1,3,4)P3 and Ins(1,3,4,5)P4 , were all able to cause translocation of C1199-Tiam1 when individually microinjected into astrocytoma cells, suggesting that these direct and indirect products of PLC activity are responsible for Tiam1 release from plasma membranes [10]. Several lines of evidence indicate that membrane localization of Tiam1 is required for the expression of Rac-GEF activity in vivo. Thus it might be expected that cytosolic translocation of Tiam1 would inhibit its function and hence reduce the level of cellular GTP-Rac. This was indeed found to be the case; when astrocytoma cells were stimulated with a thrombin agonist the GTP-Rac levels fell and then recovered in parallel with the cytosolic translocation of C1199-Tiam1. By contrast, when cellular PI3K activity was stimulated, GTP-Rac increased in agreement with the idea that PtdIns(3,4,5)P3 stimulates the GEF activity of Tiam1 [10]. Overall, the results suggest that inositol lipids play several important roles in regulating the catalytic activity and subcellular localization of Tiam1. PtdIns(3,4,5)P3 , PtdIns(3,4)P2 and PtdIns(4,5)P2 are candidate lipids that could account for the membrane localization of Tiam1. However, membrane association of Tiam1 was insensitive to wortmannin or stimuli that activate PI3K. This together with knowledge of the relative levels of these three lipids compared with their relative
Cell Architecture: from Structure to Function
affinities for Tiam1 suggests that PtdIns(4,5)P2 probably accounts for constitutive membrane association of Tiam1. This ligand binding may switch towards PtdIns(3,4,5)P3 when the concentration of the latter increases in response to growth factors and/or insulin, accounting for the PI3Kdependent stimulation of Tiam1 GEF activity towards Rac. Tiam1 catalytic activity is also regulated by phosphorylation mediated by CamKII (calcium/calmodulin-dependent kinase II) [15]. This appears to require PtdIns(4,5)P2 liganded and hence plasma membrane-targeted Tiam1. Hence, NPH-Tiam1 binds three types of ligands with distinct effects. Constitutive membrane association via PtdIns(4,5)P2 binding is necessary for CamKII-dependent activation downstream of Ca2+ signalling. PtdIns(3,4,5)P3 displaces PtdIns(4,5)P2 binding upon activation of PI3K and this directly activates Rac GEF activity. Finally, the generation of inositol phosphates via PLC activity competes with lipid ligands resulting in membrane to cytosol translocation and concomitant inhibition of Tiam1 function.
Figure 1 Ligand-binding characteristics of PKBα PH domain analysed by time-resolved FRET A FRET signal was generated by incubation of the FRET fluorophores [6] with 20 nM PH domain and 80 nM biotinylated PtdIns(3,4,5)P3 . Increasing concentrations of unlabelled PIs (A) or inositol polyphosphates (B) were used to displace the PH domain leading to a loss of the FRET signal. See text for further details.
Inositol lipids and phosphates maintain the distinct membrane-bound and cytosolic pools of PDK1 required to access its diverse substrates In each of the above examples, soluble inositol phosphates can antagonize the positive roles played by lipid ligands in the regulation of PH domain-containing proteins. The situation differs with PDK1, which needs to access substrate proteins some of which become membrane-bound and others which are cytosolic [16,17]. PDK1 was first identified as the kinase responsible for PtdIns(3,4,5)P3 -dependent phosphorylation of Thr308 within the activation loop of PKB (protein kinase B)/Akt [18,19]. PDK1 possesses a C-terminal PH domain that binds PIs and a kinase domain that phosphorylates and activates many other kinases, in addition to PKB, which belong to the AGC (cAMP-stimulated, cGMP-stimulated and C-kinase) subfamily of kinases. A remarkable feature of PDK1 is that insulin and growth factors control its activity by regulating its interaction with substrates. In the case of PKB, this is achieved by a small pool of PDK1 that is constitutively anchored to plasma membranes via its PH domain. PKB is normally cytosolic, but translocates to the plasma membrane upon activation of type I PI3Ks, which generate PtdIns(3,4,5)P3 and/or PtdIns(3,4)P2 , both of which are ligands for PKB’s PH domain (Figure 1). Co-localization of PDK1 and PKB in this manner appears sufficient to account for PI3K-dependent phosphorylation of Thr308 in the latter enzyme [20]. Full activation of PKB requires a second kinase that phosphorylates Ser473 within the hydrophobic motif and was previously termed PDK2, since phosphorylation of this site could also be blocked by inhibitors of PI3K. Recent compelling evidence suggests that Ser473 is phosphorylated by the mTOR (mammalian target of rapamycin) kinase in complex with a protein called Rictor [21]. Ironically, this complex, unlike the mTOR–Raptor
complex, is not sensitive to inhibition by rapamycin. PDK1 also activates p70S6K (p70 ribosomal S6-kinase), SGK (serum and glucocorticoid responsive kinase), RSK (p90 ribosomal S6-kinase) and protein kinase C isoforms by phosphorylating these enzymes within their activation loops. These enzymes do not possess PH domains and are localized within the cytosol. Growth factor and/or insulin stimulation causes the PI3K-dependent phosphorylation of these AGC kinases within their hydrophobic motifs, resulting in the formation of a docking motif that binds cytosolic PDK1 through a pocket on the latter’s catalytic domain. This binding is required for PDK1-dependent phosphorylation of the activation loop of the cytosolic substrates [22]. Hence, it is necessary to maintain both cytosolic and membrane-bound pools of PDK1 for efficient activation of its full range of substrates.
PtdIns(4,5)P2 binding may account for the constitutive membrane-associated pool of PDK1 In order to investigate how these two major pools of PDK1 arise, we performed a detailed analysis of the ligandbinding characteristics of both full-length PDK1 and its isolated PH domain [11]. To do this, we made use of a novel approach for determining the ligand-binding affinities of lipid-binding protein domains. This employs a sensor complex, comprising biotinylated lipids bound to streptavidincoupled APC (allophycocyanin; a fluorescent protein), a C 2005
Biochemical Society
1305
1306
Biochemical Society Transactions (2005) Volume 33, part 6
GST-fusion protein of the lipid-binding protein under study, and anti-GST labelled with Lance chelate (fluorescence donor). When the components of this complex assemble and the sample is excited at 340 nm, FRET occurs between the Lance chelate and APC and can be detected as a timeresolved emission signal at 665 nm. The addition of exogenous lipids or inositol phosphates capable of binding the domain under study displaces the complex and reduces the time-resolved FRET signal. This reduction in signal is the basis for determining the affinities of a wide range of competing ligands. In agreement with previous studies, fulllength PDK1 was found to have a very high affinity for PtdIns(3,4,5)P3 , much greater than that for the PH domain of PKB (see Figure 1), but with a significant affinity for PtdIns(4,5)P2 (0.7 µM). Given the high concentration of the latter lipid [103 -fold greater than PtdIns(3,4,5)P3 ] [23] and that inhibition of PI3K did not cause displacement of PDK1 from membranes, we suggest that PtdIns(4,5)P2 binding may account for the plasma membrane pool of PDK1.
Figure 2 PDK1 phosphorylates membrane-bound and cytosolic substrates PDK1 occurs constitutively in distinct plasma membrane-bound and cytosolic pools. PI binding to the PH domain is required for membrane localization. The model proposes that PtdIns(4,5)P2 is likely to be the physiologically relevant ligand for this purpose. PDK1 is not regulated directly, but phosphorylates Thr308 in PKBα when the latter translocates to plasma membranes following activation of PI3K. It is also proposed that the cytosolic pool of PDK1 is maintained by InsP6 competing with PIs for binding to its PH domain. This pool phosphorylates the cytosolic substrates illustrated by binding to their phosphorylated hydrophobic motifs. Further details are given in the text. ECM, extracellular matrix; GSK, glycogen synthase kinase; P70S6K, p70 S6 kinase; PKC, protein kinase C.
Soluble inositol polyphosphates are potent ligands for the PH domain of PDK1 The PH domain of PDK1 has an unusual ‘budded’ structure with an N-terminal extension that is part of the overall polypeptide fold and a particularly spacious ligand-binding site. This probably accounts for the unexpected finding that InsP5 and InsP6 bind strongly to full-length PDK1 [11]. This contrasts starkly with our findings for PKB, which, in our hands, bound 3-PIs, but did not bind any soluble inositol phosphates to a significant degree (Figure 1; but compare with contradictory data in [24,25]). This raised the question of whether InsP6 , which occurs naturally in the majority of cells at levels which exceed its K d for binding PDK1 [26], could function to maintain the physiologically significant, cytosolic pool of PDK1. Evidence to support this idea was obtained by comparing the behaviour of full-length PDK1 with that of its isolated PH domain. The latter has a much lower affinity for inositol polyphosphates, and to a lesser extent for PIs, than the full-length protein, implying that there may be interactions of some ligands with non-PH domain regions of the holoprotein. Whereas full-length GFP-tagged PDK1 was mainly cytosolic when expressed in HEK-293 (human embryonic kidney 293) cells and did not translocate in response to stimulation with insulin-like growth factor 1, the isolated PH domain did translocate, as did the PH domain of PKB. These data are compatible with the idea that the high affinity of PDK1 for soluble inositol polyphosphates may function to retain a pool of this kinase in the cytosol [11].
Conclusions The ligand-binding characteristics of many PH domains are complex and often include both membrane-bound and soluble components. The subcellular localization of proteins containing apparently promiscuous PH domains is therefore likely to reflect the balance of competing ligands and this C 2005
Biochemical Society
in turn enhances the versatility of control mechanisms that function through dynamic changes in the concentrations of PH domain-binding second messengers. For PDK1, the situation is summarized in Figure 2. We propose that the membrane-associated pool of this lipid is constitutively bound to PtdIns(4,5)P2 allowing it to phosphorylate PKB only when the latter translocates to membranes in response to activation of PI3K. This mechanism therefore depends upon the distinct ligand-binding characteristics displayed by the PH domains of PDK1 and PKB (Figure 1) respectively. On the contrary, we have suggested that the cytosolic pool of PDK1 is maintained by the constitutively high cytosolic concentration of InsP6 and/or InsP5 and that this is necessary for efficient phosphorylation of its soluble substrate proteins.
References 1 Ferguson, K.M., Kavran, J.M., Sankaran, V.G., Fournier, E., Isakoff, S.J., Skolnik, E.Y. and Lemmon, M.A. (2000) Mol. Cell 6, 373–384 2 Lietzke, S.E., Bose, S., Cronin, T., Klarlund, J., Chawla, A., Czech, M.P. and Lambright, D.G. (2000) Mol. Cell 6, 384–394 3 Ferguson, K.M. and Lemmon, M.A. (2000) Biochem. J. 350, 1–18 4 Watt, S.A., Kular, G., Fleming, I.N., Downes, C.P. and Lucocq, J.M. (2002) Biochem. J. 363, 657–666 5 Watt, S.A., Kimber, W.A., Fleming, I.N., Leslie, N.R., Downes, C.P. and Lucocq, J.M. (2004) Biochem. J. 377, 653–663 6 Gray, A., Olsson, H., Batty, I.H., Priganica, L. and Downes, C.P. (2003) Anal. Biochem. 313, 234–245 7 Kojima, T., Fukuda, M., Watanabe, Y., Hamazato, F. and Mikoshiba, K. (1997) Biochem. Biophys. Res. Commun. 236, 333–339 8 Kavran, J.M., Klein, D.E., Lee, A., Falasca, M., Isakoff, S.J., Skolnik, E.Y. and Lemmon, M.A. (1998) J. Biol. Chem. 273, 30497–30508
Cell Architecture: from Structure to Function
9 Hirose, K., Kadowaki, S., Tanabe, M., Takeshima, H. and Iino, M. (1999) Science 284, 1527–1530 10 Fleming, I.N., Batty, I.H., Prescott, A.R., Gray, A., Kular, G.S., Stewart, H. and Downes, C.P. (2004) Biochem. J. 382, 857–865 11 Komander, D., Fairservice, A., Deak, M., Kular, G.S., Prescott, A.R., Downes, C.P., Safrany, S.T., Alessi, D.R. and van Aalten, D.M.F. (2004) EMBO J. 23, 3918–3928 12 Nash, M.S., Young, K.W., Willars, G.B., Challiss, R.A.J. and Nahorski, S.R. (2001) Biochem. J. 356, 137–142 13 Minard, M.E., Kim, L., Price, J.E. and Gallick, J.E. (2004) Breast Cancer Res. Treat. 84, 21–32 14 Fleming, I.N., Gray, A. and Downes, C.P. (2000) Biochem. J. 351, 173–182 15 Fleming, I.N., Elliott, C.M., Buchanan, F.G., Downes, C.P. and Exton, J.H. (1999) J. Biol. Chem. 274, 12753–12758 16 Toker, A. and Newton, A.C. (2000) Cell (Cambridge, Mass.) 103, 185–188 17 Mora, A., Komander, D., van Aalten, D.M.F. and Alessi, D.R. (2004) Cell Dev. Biol. 15, 161–170 18 Alessi, D.R., James, S.R., Downes, C.P., Holmes, A.B., Gaffney, P.R., Reese, C.B. and Cohen, P. (1997) Curr. Biol. 7, 261–269
19 Stokoe, D., Stephens, L.R., Copeland, T., Gaffney, P.R., Reese, C.B., Painter, G.F., Holmes, A.B., McCormick, F. and Hawkins, P.T. (1997) Science 277, 567–570 20 McManus, E.J., Collins, B.J., Ashby, P.R., Prescott, A.R., Murray-Tait, V., Armit, L.J., Arthur, J.S. and Alessi, D.R. (2004) EMBO J. 23, 2071–2082 21 Sarbassov, D.D., Guertin, D.A., Ali, S.M. and Sabatini, D.M. (2005) Science 307, 1098–1101 22 Collins, B.J., Deak, M., Arthur, J.S., Armit, L.J. and Alessi, D.R. (2003) EMBO J. 22, 4202–4211 23 Batty, I.H. and Downes, C.P. (1996) Biochem. J. 317, 347–351 24 Frech, M., Andjelkovic, M., Ingley, E., Reddy, K.K., Falck, J.R. and Hemmings, B.A. (1997) J. Biol. Chem. 272, 8474–8481 25 Piccolo, E., Vignati, S., Maffucci, T., Innominato, P.F., Riley, A.M., Potter, B.V., Pandolfi, P.P., Broggini, M., Iacobelli, S., Innocenti, P. et al. (2004) Oncogene 23, 1754–1765 26 Irvine, R.F. and Schell, M.J. (2001) Nat. Rev. Mol. Cell Biol. 2, 327–338
Received 22 June 2005
C 2005
Biochemical Society
1307
