|Home | About | Journals | Submit | Contact Us | Français|
Phosphoinositides (PtdInss) play key roles in cell polarization and motility. With a series of biosensors based on Förster resonance energy transfer, we examined the distribution and metabolism of PtdInss and diacylglycerol (DAG) in stochastically migrating Madin-Darby canine kidney (MDCK) cells. The concentrations of phosphatidylinositol (4,5)-bisphosphate, phosphatidylinositol (3,4,5)-trisphosphate (PIP3), phosphatidylinositol (3,4)-bisphosphate, and DAG were higher at the plasma membrane in the front of the cell than at the plasma membrane of the rear of the cell. The difference in the concentrations of PtdInss was estimated to be less than twofold between the front and rear of the migrating MDCK cells. To decode the spatial activities of PtdIns metabolic enzymes from the obtained concentration maps of PtdInss, we developed a one-dimensional reaction diffusion model of PtdIns metabolism. In this model, the activities of phosphatidylinositol monophosphate 5-kinase, phosphatidylinositol 3-kinase, phospholipase C, and PIP3 5-phosphatases were higher at the plasma membrane of the front than at the plasma membrane of the rear of the cell. This result suggests that, although the difference in the steady-state level of PtdInss is less than twofold, PtdInss were more rapidly turned over at the front than the rear of the migrating MDCK cells.
Cell migration is an important event during early development, inflammatory responses to infection, and wound healing, and it is an important pathological event during tumor invasion and metastasis. Despite morphological and functional differences, different migratory cells share a conserved set of polarity signals. Kinases for phosphoinositides (PtdInss), Rho GTPases, and the actin and microtubule cytoskeletons play key roles in signaling polarity in cells ranging from Dictyostelium discoideum (Charest and Firtel, 2006 ) to neurons (Luo, 2000 ; Aoki et al., 2007 ) and neutrophils (Van Keymeulen et al., 2006 ; Wong et al., 2006 ).
PtdInss are a family of phospholipids containing myo-inositol as their head group (reviewed in Takenawa and Itoh, 2001 ). Despite a relatively low abundance in biological membranes, PtdInss have been reported to regulate a myriad of cellular processes. Among them, phosphatidylinositol (4,5)-bisphosphate [PI(4,5)P2] is the major PtdInss at the inner leaflet of the plasma membrane. On growth factor stimulation, intracellular second messengers such as inositol (1,4,5)-trisphosphate (IP3), diacylglycerol (DAG), and phosphatidylinositol (3,4,5)-triphosphate (PIP3) are generated from PI(4,5)P2.
The discovery that the pleckstrin homology (PH) domain in the signaling proteins recognizes specific phosphoinositides revealed that stimulated proteins translocate to specific regions of the membrane to participate in signaling events via interaction with these lipids (reviewed in Di Paolo and De Camilli, 2006 ). A series of experiments indicated that the PH domains from different proteins recognize different PtdInss and led to the development of a novel tool for detecting the localization of these lipids in living and fixed cells by fusing the green fluorescent protein (GFP)-tag to the PH domain (reviewed in Di Paolo and De Camilli, 2006 ). The drawback of these GFP-tagged probes is that the images obtained with them are substantially affected by the ratio of the surface area of the plasma membrane to the cytosolic volume (surface-to-volume ratio [SVR]) (Craske et al., 2005 ). For example, by using a GFP-tagged PH domain, a location with a high SVR, such as the membrane ruffles, provides pseudopositive signals even when the PtdIns distribution is uniform. Therefore, GFP-tagged probes require appropriate negative controls and quantitative analysis (Yoshizaki et al., 2007 ).
Förster (or fluorescence) resonance energy transfer (FRET) is a process by which a fluorophore (donor) in an excited state nonradiatively transfers its energy to a neighboring fluorophore (acceptor), thereby causing the acceptor to emit fluorescence at its characteristic wavelength. Variants of GFP have provided genetically encoded fluorophores that serve as donors and/or acceptors in FRET (Pollok and Heim, 1999 ; Miyawaki et al., 2003 ). Using these GFP variants and FRET technology, FRET biosensors have been reported for PIP3, PI(3,4)P2, and DAG (Sato et al., 2003 ; Sato et al., 2006 ; Yoshizaki et al., 2007 ). Here, we extended this technique to develop FRET biosensors for PI(4,5)P2, PI(4)P, and DAG, and we analyzed the distribution of PtdInss in migrating MDCK cells. In addition, we developed a kinetic simulation model of PtdIns metabolism to clarify the mechanism of the polarized distribution of PtdInss in migrating MDCK cells.
Plasmids for FRET-based monitors were constructed essentially as described previously (Mochizuki et al., 2001 ; Sato et al., 2003 ). The FRET biosensor for PI(4,5)P2, Phosphatidylinositol phosphate indicator (Pippi)-PI(4,5)P2, consisted of cyan fluorescent protein (CFP) (amino acids [aa] 1-237), a spacer (Glu-Ala-Ala-Ala-Arg)6, the PH domain of phospholipase C (PLC)δ (aa 18-130), a spacer (Glu-Ala-Ala-Ala-Arg)3-Gly-Gly-(Glu-Ala-Ala-Ala-Arg)3, yellow fluorescent protein (YFP) (aa 1-237), a spacer (Glu-Ala-Ala-Ala-Arg)7, and the C-terminal region of Ki-Ras4B (aa 169-188) or H-Ras (aa 170-189, called as HRas-CT, hereafter). For the FRET monitors for DAG and PI(4)P, the lipid-binding domain of Pippi-PI(4,5)P2 was replaced by the C1 domain of protein kinase C (PKC)βII (aa 37-140) and the PH domain of FAPP1 (aa 1-101), respectively. Other FRET biosensors, Pippi-PIP3 and Pippi-PI(3,4)P2, have been described previously (Sato et al., 2003 , 2006 ; Aoki et al., 2005 ). The plasmid coding mRFP-FKBP-5-Ptase-dom was a gift from Dr. Tamas Balla, and pERed-NLS-LDR (Lyn11-targeted FRB) was described previously (Aoki et al., 2007 ).
NIH 3T3 cells were purchased from the RIKEN Gene Bank (Wako-shi, Japan) and maintained in DMEM (Sigma-Aldrich, St. Louis, MO) supplemented with 10% fetal bovine serum (FBS). MDCK cells were from Dr. Yoshimi Takai (Kobe University, Kobe, Japan). For transient expression studies, cells were transfected using Polyfect (QIAGEN, Valencia, CA) or 293Fectin (Invitrogen, Carlsbad, CA). Cells were analyzed at 24 h after transfection. For immunoblotting, proteins were separated by SDS-polyacrylamide gel electrophoresis (PAGE) and transferred to a polyvinylidene difluoride membrane, followed by detection with the antibodies described below. The bound antibodies were detected by an ECL chemiluminescence detection system (GE Healthcare, Chalfont St. Giles, Buckinghamshire, United Kingdom), and binding was quantified with the aid of an LAS-1000 image analyzer (Fuji-Film, Tokyo, Japan). Anti-PLCγ and -phospho PLC (pTyr-783) antibodies were purchased from Cell Signaling Technology (Danvers, MA).
NIH3T3 cells were labeled with 32Pi (0.4 mCi/ml) for 4 h and stimulated with platelet derived growth factor (PDGF)-BB (50 ng/ml) for the indicated periods. Lipids were extracted with CHCl3:methanol:H2O (1:1:1, vol/vol), separated by TLC (separating solvent: CHCl3:methanol:ammonium:H2O [45:35:6:4, vol/vol]), and quantified by BAS2000 (Fuji-Film). To mark the region for phosphatidylcholine and phosphatidylinositol, the standard lipids were separated by the same TLC plate and chemically detected.
FRET imaging was performed essentially as described previously (Yoshizaki et al., 2003 ). Briefly, cells plated on a collagen- or fibronectin-coated 35-mm-diameter glass-base dish (Asahi Techno Glass, Chiba, Japan) were imaged every 1 min on an IX81 inverted microscope (Olympus, Tokyo, Japan) equipped with a laser-based autofocusing system, IX2-ZDC, and an automatically programmable XY stage, MD-XY30100T-Meta, which allowed us to obtain the time-lapse images of several view fields in a single experiment. For dual-emission ratio imaging of the intramolecular FRET biosensors, we used previously described filter sets and obtained images for CFP and FRET. After background subtraction was carried out, the FRET/CFP ratio was depicted using MetaMorph software (Molecular Devices, Sunnyvale, CA), and this image was used to represent FRET efficiency. The filters used for the dual-emission ratio imaging were purchased from Omega Optical (Brattleboro, VT): an XF1071 (440AF21) excitation filter, an XF2034 (455DRLP) dichroic mirror, and two emission filters, XF3075 (480AF30) for CFP and XF3079 (535AF26) for FRET. Cells were illuminated with a 75-W xenon lamp through a 6% ND filter (Olympus) and a 60× oil immersion objective lens (PlanApo 60×/1.4). The exposure time was 0.3 s when the binning of the cooled charge-coupled device camera Cool SNAP-HQ (Roper Scientific, Trenton, NJ) was set to 4 × 4. The ratio image of FRET/CFP was created with MetaMorph software and was used to represent the efficiency of the FRET.
MDCK cells were fixed with 3.7% paraformaldehyde, followed by permeabilization with 0.2% TritonX-100. After having been soaked for 30 min in phosphate-buffered saline (PBS) containing 10% FBS, cells were incubated for 1 h at room temperature with anti-Giantin antibody (Covance Research Products, Princeton, NJ), followed by incubation with Alexa-568 conjugated anti-rabbit IgG at 4°C for overnight. After washing with PBS, cells were imaged with an FV-1000 confocal microscope.
Microinjections were performed under an inverted microscope (Carl Zeiss, Jena, Germany) equipped with a micromanipulator 5171 (Eppendorf, Hamburg, Germany) by using an automated FemtoJet (Eppendorf) and Femtotip glass microcapillaries. Pippi-PIP3–expressing NIH3T3 cells were plated on fibronectin-coated 35-mm glass-bottomed dishes and microinjected with 10 μM PIP3-3-phosphothioate (Echelon, Salt Lake City, UT) and 0.1 μM Alexa568-hydrazide (Invitrogen). The injection parameters were kept as follows: Pi, 30 hPa; Ti, 0.1 s; and Pc, 30 hPa. Immediately after microinjection, cells were observed with a FV-1000 confocal microscope for the fluorescence spectrum analysis. We used an LD405 laser for the excitation, and acquired the emission profiles from 430 to 600 nm in a lamda scan mode. Similarly, the spectrometry of Digda-expressing NIH3T3 cells was obtained in presence and absence of 50 μM 1-oleoyl-2-acetylglycerol (OAG).
Kinetic reactions were described with mass action kinetics. All reactions and the kinetic constants used in this study are shown in Supplemental Material 2. For the modeling of FRET biosensor reactions, CellDesigner (version 4.0α; The Systems Biology Institute, Tokyo, Japan) was used for modeling and solving the ordinary differential equations (Kitano et al., 2005 ).
In the one-dimensional model, the reaction-diffusion problem was solved by numerical integration with a commercially available solver, COMSOL multiphysics version 3.4 with a Chemical-Engineering module and COMSOL Reaction Engineering Lab1.4 (Comsol, Los Angeles, CA) running on a Workstation PC (Dell Precision 390; Dell, Round Rock, TX). The reaction model generated by CellDesigner was imported into the one-dimensional model. See the Supplemental Materials 2 for details.
To investigate the spatiotemporal dynamics of PI(4,5)P2, we developed a FRET-based monitor for PI(4,5)P2, named Pippi-PI(4,5)P2, essentially as described previously (Figure 1A) (Sato et al., 2003 ; Aoki et al., 2005 ). In this biosensor, when the PH domain of PLC δ binds to PI(4,5)P2 on the cell membrane, the CFP is brought in proximity to the YFP, increasing the level of FRET. As an index of the level of FRET, we used the FRET/CFP ratio as described in the Materials and Methods. To evaluate Pippi-PI(4,5)P2, we visualized the FRET/CFP change in NIH3T3 cells, in which PDGF-BB (hereafter PDGF) induces rapid activation of PLC and concomitant decrease in PI(4,5)P2. As expected, in NIH3T3 cells expressing Pippi-PI(4,5)P2, the level of FRET/CFP was decreased rapidly at the plasma membrane (Figure 1, B and C). After this rapid and diffuse decrease in FRET/CFP, at 15–20 min after stimulation, we observed the restoration of FRET/CFP over a diffuse area. Interestingly, the increase was most prominent at the peripheral membrane protrusion. Such changes in FRET/CFP was not observed in cells expressing Pippi-PI(4,5)P2-R37D/R38E, which harbored two amino-acid substitutions within the critical lipid binding motif of the PH domain, demonstrating that the change in FRET/CFP depended on the PH domain binding to PI(4,5)P2. We also confirmed that mock stimulation did not change the FRET/CFP ratio (Supplemental Figure S1, A and B).
To confirm that the FRET/CFP ratio in Pippi-PI(4,5)P2–expressing cells correlated with the amount of PI(4,5)P2, we quantitated the amount of PI(4,5)P2 by TLC by using NIH3T3 cells labeled with 32Pi for 4 h and stimulated with PDGF for the indicated periods (Figure 1, D and E). Similarly to the change in FRET/CFP, the amount of PI(4,5)P2 was decreased rapidly and restored to the basal state within 20 min. To further demonstrate the correlation between the FRET/CFP ratio and amount of PI(4,5)P2 within the cells, we examined the activity of PLCγ1, which hydrolyzes PI(4,5)P2 to DAG and IP3 and plays the major role in decreasing the amount of PI(4,5)P2 in PDGF-stimulated cells (Berridge et al., 1984 ). Because the enzymatic activity of PLCγ1 correlates with phosphorylation at Tyr763 (Kim et al., 1991 ), the activity of PLCγ1 was monitored by Western blotting with anti-phosphorylated Tyr763 of PLCγ1 (Figure 1, F and G). Again, we observed rapid phosphorylation of PLCγ1 within 1–5 min, followed by dephosphorylation to the basal state in PDGF-stimulated NIH3T3 cells. These observations supported the idea that Pippi-PI(4,5)P2 could be used as a biosensor for PI(4,5)P2 on the plasma membrane.
Similarly to Pippi-PI(4,5)P2, we developed a FRET-based monitor for DAG named Digda (for Diacylglycerol indicator; Figure 2A). The PH domain of Pippi-PI(4,5)P2 was replaced with the C1 domain of PKCβII, which binds specifically to DAG. Thus, an elevated level of FRET is expected when DAG is increased at the cell membrane. In agreement with this prediction, when NIH3T3 cells expressing Digda were treated with 50 μM OAG, an analogue of DAG, the FRET/CFP value was elevated rapidly (Figure 2, B and C). Fluorescence spectra of Digda-expressing cells are shown in Supplemental Figure S1C. Digda-C67/132S having a two amino-acid substitution in the critical DAG-binding domain did not show such a response to OAG, demonstrating that the C1 domain binding to OAG was responsible for the increased FRET/CFP value. On PDGF stimulation of NIH3T3 cells, a transient increase in FRET/CFP was observed diffusely at the plasma membrane, suggesting that DAG was rapidly produced at the plasma membrane (Figure 2D). Notably, a remarkable increase in FRET/CFP was observed at the nascent lamellipodia induced by PDGF. Ten to 20 minutes after stimulation, the level of FRET/CFP gradually returned to the basal level. Again, we did not observe such changes in FRET/CFP in PDGF-stimulated cells expressing Digda-C67/132S or mock-stimulated cells expressing Digda (Supplemental Figure S1, D and E). Interestingly, there was a clear discrepancy in the time courses of PDGF-stimulated changes of PI(4,5)P2 and DAG. Although the concentrations of both PI(4,5)P2 and DAG reached their nadir or zenith within 5 min, the concentration of PI(4,5)P2 returned to the basal level significantly faster than did the concentration of DAG.
During the course of experiments that will be described later, it became necessary to determine the level of phosphatidylinositol (4)-monophosphate [PI(4)P] in the plasma membrane as a precursor of PI(4,5)P2. Although PI(4)P has been reported to localize mainly at the Golgi apparatus, a small amount of PI(4)P localizes at cytoplasmic vesicles and the plasma membrane (Balla et al., 2005 ). Thus, we developed another biosensor for PI(4)P at the plasma membrane, named Pippi-PI(4)P, by using the PH domain of the FAPP1 protein (Figure 2F). For the validation of this biosensor, we used phenylarsine oxide (PAO) at a concentration of 0.1 μM, which inhibits type-IIIα phosphatidylinositol 4-kinase (PI4K) and thereby reduces the amount of PI(4)P at the plasma membrane. As expected, treatment with PAO rapidly and diffusely reduced the FRET/CFP level of MDCK cells expressing Pippi-PI(4)P (Figure 2G). In one experiment, we used a negative control biosensor, the Ras and interacting protein chimeric unit (Raichu)-PAK-RhoA, in which the Rac-binding domain of PAK and RhoA were inserted between YFP and CFP. Because PAK does not bind to RhoA, this biosensor is used to negate the nonspecific change in FRET/CFP (Yoshizaki et al., 2003 ). As expected, the change in the FRET/CFP ratio was not observed in cells expressing Raichu-PAK-RhoA (Supplemental Figure S1F). Next, we examined whether Pippi-PI(4)P could monitor the increase in PI(4)P by using a rapamycin-inducible membrane translocation system (Varnai et al., 2006 ; Aoki et al., 2007 ). FKBP-5ptase-dom is a fusion protein consisting of the catalytic domain of phosphoinositide 5-phosphatase (5ptase) and the FK506-binding protein (FKBP), a ligand for rapamycin. On treatment with rapamycin, FKBP-5ptase-dom is recruited to the plasma membrane, where another rapamycin ligand, a fragment of the mammalian target of rapamycin (FRB), is localized by the attached myristylation-moiety (Figure 2H). The translocation of FKBP-5ptase-dom dephosphorylates PI(4,5)P2 to yield PI(4)P at the plasma membrane (Varnai et al., 2006 ). Intriguingly, the increase in PI(4)P was not apparent at the plasma membrane; however, the biosensor was accumulated at perinuclear vesicular structures concomitant with an increase in FRET (Figure 2H). We interpreted this result as follows: On FKBP-5ptase-dom translocation to the plasma membrane and the resulting increase in PI(4)P, the excess amount of PI(4)P might have accelerated the rate of endocytosis. The Pippi-PI(4)P bound to PI(4)P might have been cotransported to these endosomes and exhibited a high FRET/CFP area.
To confirm that the high FRET/CFP ratio at the perinuclear region was not due to the aggregation of the FRET biosensor, we constructed biosensors fused to HRas-CT, which locates the probes not only at the plasma membrane but also at the Golgi apparatus via the exocytic pathway (Apolloni et al., 2000 ). The Golgi localization of the probes was confirmed by coimmunostaining with a Golgi protein, Giantin (Supplemental Figure S1G). As shown in Figure 2I, the FRET/CFP ratio at the Golgi area was lower than that at the plasma membrane in cells expressing Pippi-PI(4,5)P2-HRas-CT. In contrast, in cells expressing Pippi-PI(4)P-HRas-CT, the FRET/CFP ratio at the Golgi area was higher than that at the plasma membrane, negating the possibility that the high FRET/CFP ratio at Golgi area was caused by the accumulation of the probe. Notably, this observation is consistent with the previous report that the PI(4)P is enriched at the intracellular compartment including the Golgi apparatus (Balla et al., 2005 ).
Many biochemical events, including cell–matrix contact turnover and cytoskeletal restructuring during cell migration, are under the control of PtdInss. Using the FRET biosensors described above and reported previously, we examined the distribution of phosphoinositides and DAG at the plasma membrane of migrating MDCK cells. Typically, the migrating MDCK cells consist of a lamellar front half and a dome-shaped back half due to the presence of the nucleus and perinuclear organelles. The level of PI(4,5)P2 was significantly higher at the plasma membrane in the front than in the rear of the cell (Figure 3A; Supplemental Movie 1). The FRET/CFP values ranged from 1.2 to 1.6 in most of the observed cells, which indicated that the concentration of PI(4,5)P2 ranged from 4 to 8 μM, as discussed in Supplemental Material 1. Such a gradient in the level of FRET/CFP was not observed in cells expressing the mutant Pippi-PI(4,5)P2-R37D/R38E (Supplemental Figure S3A). Similarly to PI(4,5)P2, DAG also exhibited a polarized distribution, with a higher concentration in the front, particularly at lamellipod, than in the back (Figure 3B). Again, such a gradient was not observed in cells expressing the Digda-C67/132S mutant (Supplemental Figure S3B). For the comparison, we also examined the distribution of products of phosphatidylinositol 3-kinase (PI3-kinase), PIP3, and PI(3,4)P2, in migrating MDCK cells by using the Pippi-type FRET biosensors Pippi-PIP3 and -PI(3,4)P2, having the PH domains of GRP and TAPP1, respectively (Aoki et al., 2007 ; Yoshizaki et al., 2007 ). As reported previously (Parent et al., 1998 ; Servant et al., 2000 ), both PtdInss were enriched in the front with an increasing gradient toward the leading edge (Figure 3, C and D). The specificity was confirmed by FRET imaging of cells expressing the Pippi-PIP3-R284C mutant (Supplemental Figure S3C) and by the fluorescence spectrography of cells expressing Pippi-PIP3 in the presence or absence of a hydrolysis-resistant analogue of PIP3 (Supplemental Figure S3D). Finally, we examined the level of PI(4)P with Pippi-PI(4)P. In contrast to the other PtdInss examined in this study, PI(4)P was distributed uniformly at the entire plasma membrane (Figure 3E).
To understand how such gradients of PtdInss concentration are generated, we examined the effect of inhibitors for enzymes that metabolize PtdInss. First, to examine the role of PI(4)P as a precursor of PI(4,5)P2, MDCK cells expressing Pippi-PI(4,5)P2 were treated with PAO. As shown in Figure 4A, PAO treatment rapidly and strongly decreased the FRET level in Pippi-PI(4,5)P2-expressing MDCK cells, suggesting that PI(4)P is the primary source for PI(4,5)P2 at the plasma membrane. The half-life of PI(4,5)P2 was estimated to be ≈2 min (Figure 4, right). Next, to examine the catabolism of PI(4,5)P2, MDCK cells were treated with LY294002 or U73122, which inhibit PI3-kinase and PLC, respectively. As shown in Figure 4, B and D, treatment with U73122, but not LY294002, induced a rapid increase in the FRET efficiency of Pippi-PI(4,5)P2 in the front half of the cell, indicating that PI(4,5)P2 was primarily catabolized by PLC in migrating MDCK cells. LY294002 was found to decrease the level of FRET in Pippi-PIP3–expressing cells (Figure 4C); therefore, the failure to detect an increase in PI(4,5)P2 after treatment with LY294002 indicates that PI3-kinase does not contribute to the catabolism of PI(4,5)P2 to a detectable level. Intriguingly, we noticed that the U73122-induced increase in PI(4,5)P2 was most prominent at the lamellipod (Figure 4, E and F), suggesting the presence of an increasing gradient of PLC activity in the direction of the leading edge.
The finding that the concentrations of PI(4,5)P2, PI(3,4)P2, PIP3, and DAG were higher in the front than in the rear of the cell could be explained simply by the high activity of type-IIIα PI4-kinase in the front. However, the difference in the gradients of these PtdInss and DAG should provide a clue to help in deciphering the distribution of the other PtdIns-metabolizing enzymes. To decode such activity maps from the concentration maps of PtdInss and DAG, we developed a one-dimensional reaction-diffusion model of PtdIns metabolism at the plasma membrane. To constrain the kinetic parameters, we first examined the gradients of PtdInss and DAG that were plotted along the direction of migration (Figure 5). As already described, each of PI(4,5)P2, DAG, PIP3, and PI(3,4)P2 was higher in the front than in the rear of the cell. As described in Supplemental Material 1, the difference in the concentrations between the front and back edges was estimated to be 33, 34, 57, and 39% for PI(4,5)P2, DAG, PIP3, and PI(3,4)P2, respectively. Of note, PI(4,5)P2 was distributed uniformly within the lamellipod, whereas the other lipids showed an increasing gradient toward the leading edge.
A block diagram of the kinetic simulation model of PtdIns metabolism is shown in Figure 6A. PI(4)P, the concentration of which was assumed to be constant, is the inflow, and phosphatidic acid (PA) and IP3 are the outflows of this network. Three kinases, phosphatidylinositol (4)-monophosphate 5-kinase [PI(4)P 5-kinase], PI3-kinase, and DAG-kinase; three phosphatases, PIP2 5-phosphatase (PIP2 5-Pase), phosphatase and tensin homologue (PTEN), and PIP3 5-phosphatase (PIP3 5-Pase); and one phospholipase, PLC, were integrated into this model. In the one-dimensional reaction-diffusion model along the direction of migration, we set the origin of the abscissa at the border between the front and back halves and normalized the lengths of these halves to 1.0, respectively (Figure 6B). A detailed kinetic description of this model is given in the supplemental materials part 2. With this information in hand, we used the data shown in Figure 4 to constrain the distribution of enzymes that regulate the metabolism of PtdInss and DAG at the plasma membrane. Because we cannot decisively determine the distribution of the enzymes based on the present data, we will present the most likely distribution of the enzymes (Figure 6, C and D). The rationale behind this distribution is also described in the supplemental materials part 2. First, phosphatidylinositol monophosphate (PIP) 5-kinase activity is slightly higher in the front than in the back half of the cell. Second, PLC activity is set to zero in the back half of the cell, with an increasing gradient toward the leading edge. This observation is based on the data obtained in U73122-treated cells (Figure 4, D–F). Third, PI 3-kinase activity is concentrated at the area close to the leading edge. Fourth, PIP3 5-Pase activity is slightly higher in the front than in the back half of the cell. This is predicted because the slope of PIP3 is steeper than that of PI(3,4)P2 in the front half of the cell (Figure 5, F and H). Finally, DAG-kinase PTEN and PIP2 5-Pase were set to be constant, simply because we do not have data to support the uneven distribution of these enzymes.
The local concentrations of PtdInss and DAG calculated by the simulation model and those estimated by the FRET images are overlaid in Figure 6, E–H, showing a reasonable correlation. Finally, the distribution and time-course of the level of PI(4,5)P2 were simulated and shown using a kymograph and a line plot (Figure 6, I and J), which were very similar to those shown in Figure 4, E and F, demonstrating that the present model could reliably recapitulate the kinetics of PtdInss in migrating MDCK cells.
Uneven or polarized localization of PtdInss serves as a compass for various migrating cells, such as D. discoideum and neutrophils (reviewed in Bourne and Weiner, 2002 ). In the gradient sensing D. discoideum, PI 3-kinase and PTEN are accumulated in the leading edge and uropod, respectively (Funamoto et al., 2002 ; Iijima and Devreotes, 2002 ). Recently, it has been reported that SHIP1 (SH2-containing inositol phosphatase 1), rather than PTEN, plays an important role in dephosphorylation of PIP3 in chemotaxic neutrophils (Nishio et al., 2007 ). All these reports demonstrate that the asymmetric distribution of PIP3 and its metabolic enzymes plays critical roles in cell migration. Importantly, however, these studies do not necessarily negate the possibility that the PI(4,5)P2 gradient shown in our study also contributes to establishment of the cell polarity directly by regulating the cytoskeleton and indirectly by increasing the concentration of the substrate for PI 3-kinase. Meanwhile, our simulation analysis implied that a simple increase in the most upstream component of this cascade, PI(4,5)P2, cannot explain the elevation of the other components at the lamellipod, i.e., not only PIP 5-kinase, but also PI 3-kinase, PLC, and PI 5-Pase are up-regulated in the front of the migrating cells (Figure 6). Thus, we need to take not only the amount but also the turnover rate of PtdInss into consideration of the role of PtdInss in cell migration. In support this, it has been reported that both PLCγ1 and PIPKIγ661, a variant of type I γ PIP 5-kinase, form a complex and both are localized at the leading edge of migrating cells (Piccolo et al., 2002 ; Sun et al., 2007 ).
PI(4)P and PI(4,5)P2 are the two major PtdInss, although they constitute only 0.5% of the total lipids in the eukaryotic cell membrane. It is classically admitted that the concentrations of PtdIns, PI(4)P and PI(4,5)P2 remain at the steady-state levels in the inner leaflet of the plasma membrane, due to the continuous phosphorylation/dephosphorylation reactions by specific kinases and phosphatases (Michell, 1975 ; Payrastre et al., 2001 ). This so-called “futile cycle” accounts for ~7% of the basal ATP consumption in human blood platelets (Verhoeven et al., 1987 ; Payrastre et al., 2001 ). In this study, we showed that the PtdInss turnover rate in the front is faster than that of the rear of the migrating MDCK cells. Then, for what purpose do cells further accelerate the “futile cycle” at the front of the cells? We speculate that such local high turnover rate increases the sensitivity to the external cues. To address this question, we need to integrate external cues into in the kinetic simulation model of the PtdInss. Further on this line, this kinetic simulation model will predict the effect of drugs targeting phosphoinositide-metabolizing enzymes on cell migration.
As a possible precursor of PI(4,5)P2, we focused on PI(4)P in the plasma membrane because the quantity of PI(5)P was much lower than the level of PI(4)P in eukaryotes (see above) (Rameh et al., 1997 ). Although the majority of PI(4)P and its precursor PtdIns localized to the Golgi apparatus, and although it is known that type II (wortmannin-insensitive) and type III (wortmannin-sensitive) PI 4-kinases (α and β) are primarily localized to endomembranes such as the Golgi apparatus and endoplasmic reticulum, a small amount of PI(4)P also locates to the plasma membrane, due to the activity of type-IIIα PI 4-kinase (Balla et al., 2005 ). By using a biosensor localized at the plasma membrane, we found that the concentration of PI(4)P was uniform over the entire plasma membrane. Meanwhile, the concentrations of its derivative, PI(4,5)P2, and other PtdInss were higher in the plasma membrane of the front than that of the back half of the cell. This observation suggested two possibilities: first, production of PI(4,5)P2 from PI(4)P is higher at the front than at the back of the cell; or second, PI(4,5)P2 accumulates preferentially at the front of the cell. Although the second possibility is not negated by any data, there are many studies supporting the former possibility, as are discussed in the next paragraph.
GTPases of the Rho family and Arf6 stimulate the production of PI(4,5)P2 via type Iα PIP 5-kinase (reviewed in Ling et al., 2006 and Santarius et al., 2006 ). Using FRET-based biosensors for GTPases, we and others have found that the activities of RhoA, Rac1, and Cdc42 are increased toward the leading edge of migrating cells (Kraynov et al., 2000 ; Itoh et al., 2002 ; Nalbant et al., 2004 ; Kurokawa and Matsuda, 2005 ). In addition to the GTPases, Ajuba, which is a component of integrin-mediated adhesive complexes and regulates localization of the p130Cas/Crk/DOCK/Rac signaling cascade (Pratt et al., 2005 ), interacts with and activates type Iα PIPK (Kisseleva et al., 2005 ). Therefore, although the mechanisms could be divergent, there is ample evidence that activation of type Iα PIP 5-kinase causes accumulation of PI(4,5)P2 in the lamellipod of the front of the cell. In agreement with these observations, when PI(4,5)P2 was rapidly reduced by plasma membrane targeting of PI 5-Pase, the cells stopped migrating and PI(4)P increased in the peri-nuclear region (Figure 2H).
For analysis of the spatio-temporal regulation of PtdInss, three methods have been used successfully. The first method involves PtdInss-specific antibodies (Gascard et al., 1991 ). For the immunostaining, however, cells have to be fixed. Because lipids are not fixed by paraformaldehyde, permeabilization of the plasma membrane by detergent may misconduct the distribution. The second method is to use GFP-tagged PH domains, which are translocated from the cytoplasmic pool to membranes upon the increase in PtdInss (reviewed in Balla, 2005 ). This method has enabled us for the first time to view the dynamics of PtdIns metabolism in living cells. However, the method has drawbacks in terms of quantifying the membrane translocation of the probes. For example, this method is sensitive to the thickness of the cells, i.e., when cells increase their thickness during migration or upon stimulation, particularly at the membrane ruffles, accumulation of GFP-tagged proteins tends to be erroneously scored as membrane-targeting. In addition, this method usually cannot detect the membrane targeting of GFP-tagged proteins to the plasma membrane below or above the perinuclear area, due to the bright background from cytoplasmic GFP proteins. Therefore, the GFP-tagged PH domain usually cannot be used to examine the intracellular gradient of the PtdInss. The third method, which is based on FRET, was originally developed by Umezawa and colleagues to detect the level of PIP3 (Sato et al., 2003 ). One drawback, or maybe a merit, of this method is that the biosensor has to be anchored to the membranes of interest. Here, we used biosensors fused with the C-terminal region of K-Ras proteins, which brings the proteins preferentially to the plasma membrane. Therefore, we neglected the events in the endomembrane compartments. This problem can be overcome by using different membrane-targeting signals. For example, by using the carboxyl-terminus of the H-Ras protein, we could compare the concentration of PI(4,5)P2 between the plasma membrane and Golgi apparatus (Figure 2I).
In summary, using FRET-based biosensors, we visualized the gradient of PtdInss and DAG in the plasma membrane of migrating MDCK cells and predicted the activity map of PtdIns-metabolizing enzymes. We concluded that, although the difference in the steady-state level of PtdInss is less than twofold, PtdInss were more rapidly turned over at the front than the rear of the migrating MDCK cells. The physiological significance of this high turnover rate should be addressed in a subsequent study.
We thank Dr. Tamas Balla (National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD) for the generous gift of the plasmid coding for mRFP-FKBP-5ptase-dom. We are grateful to N. Yoshida, N. Fujimoto, A. Abe, K. Fukuhara, and Y. Kasakawa for technical assistance. We also thank the staff of the Matsuda laboratory for technical advice and helpful input. This work was supported by grants from the Ministry of Education, Culture, Sports, Science, and Technology of Japan; the New Energy and Industrial Technology Development Organization; the Mitsubishi Foundation; and the Japan Health Science Foundation.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E08-03-0315) on August 6, 2008.