Phosphatidic acid regulation of PIPKI is critical for actin cytoskeletal reorganization.

Type I phosphatidylinositol-4-phosphate 5-kinase (PIPKI) is the main enzyme generating the lipid second messenger phosphatidylinositol-4,5-bisphosphate [PI(4,5)P2], which has critical functions in many cellular processes, such as cytoskeletal reorganization, membrane trafficking, and signal transduction. All three members of the PIPKI family are activated by phosphatidic acid (PA). However, how PA regulates the activity and functions of PIPKI have not been fully elucidated. In this study, we identify a PA-binding site on PIPKIγ. Mutation of this site inhibited the PA-stimulated activity and membrane localization of PIPKIγ as well as the formation of actin comets and foci induced by PIPKIγ. We also demonstrate that phospholipase D (PLD) generates a pool of PA involved in PIPKIγ regulation by showing that PLD inhibitors blocked the membrane localization of PIPKIγ and its ability to induce actin cytoskeletal reorganization. Targeting the PIPKIγ PA-binding-deficient mutant to membranes by a membrane localization sequence failed to restore the actin reorganization activity of PIPKIγ, suggesting that PA binding is not only involved in recruiting PIPKIγ to membranes but also may induce a conformational change. Taken together, these results reveal a new molecular mechanism through which PA regulates PIPKI and provides direct evidence that PA is important for the localization and functions of PIPKI in intact cells.


Construction of plasmids
Construction of GFP-PIPKI ␥ wild-type (WT) has been previously described ( 31 ). All point mutants were generated by site-directed mutagenesis using the QuikChange kit from Agilent Technologies (Santa Clara, CA). The Mem-tagged GFP-PIPKI ␥ and -KRHH/A mutant were generated by adding the membrane-targeting signal from the tyrosine kinase Lyn (MGCIKSKRKD) ( 32 ) to the N-terminus of PIPKI ␥ protein. For protein purifi cation, PIPKI ␥ WT and mutants were cloned into the Eco RI and Hin dIII sites of the bacterial expression vector pET24 and contained a N-terminal T7 tag and a C-terminal 6xHis tag. All constructs were confi rmed by sequencing. To generate tetracycline-inducible stable cell lines, the coding sequences for GFP-PIPKI ␥ proteins were subcloned from the original pEGFP-C vector into the pcDNA5/FRT/TO vector from Invitrogen. All PIPKI ␥ WT and mutants are expressed at similar levels compared with WT (supplementary Figs. I-III).

Bacterial expression and purifi cation of recombinant PIPKI ␥ proteins
Recombinant WT and mutant PIPKI ␥ proteins were expressed in E. coli Rosetta 2 cells and purifi ed by Ni 2+ -chelate chromatography according to Qiagen's manual (Valencia, CA). Proteins were fi rst dialyzed against 20 mM Tris (pH 7.6), 200 mM NaCl, 5 mM ␤ -mercaptoethanol, and then 20 mM Tris (pH 7.6), 100 mM NaCl, and 5 mM ␤ -mercaptoethanol to exchange the buffer and remove the imidazole used in purifi cation. The concentrations of the proteins were measured by Coomassie Plus-200 protein assay reagent from Thermo Scientifi c (Rockford, IL). All purifi ed recombinant proteins used in our experiments were more than 95% pure, as judged by Coomassie Blue-stained SDS-PAGE. The purifi ed proteins were used within one week.

Liposome pulldown assay
The preparation of liposomes has been previously described (33)(34)(35). phosphatidylcholine (POPC), phosphatidic acid (POPA), and Brain PI4P were purchased from Avanti Polar Lipids. Large unilamellar vesicles (LUV) were generated by mixing POPC with PI4P (80: 20) or POPC and PI4P with POPA (60:20:20) dissolved in chloroform in the designated ratios. Mixed lipids were dried in a round-bottom fl ask under rotary evaporation placed under vacuum for 30 min, and then resuspended in 176 mM Sucrose, 20 mM Tris (pH 7.6) at 2 mM total phospholipid concentration. Hydrated lipids were subjected to at least six cycles of freeze thawing in liquid nitrogen and 37°C water bath before ten cycles of extrusion through a 100 nm membrane fi lter with a lipid extruder from Northern Lipids (Burnaby, BC, Canada). These 100 nm sucrose-loaded liposomes were collected after ultracentrifugation at 100,000 g , and then resuspended in 20 mM Tris (pH 7.6) and 100 mM NaCl.
Proteins were precleared at 100,000 g to remove any protein aggregates before adding it to the assay. Binding assays were performed by mixing 100 ng of the precleared protein and 400 M sucrose-loaded liposomes in 20 mM Tris (pH 7.6), 100 mM NaCl, 1 mM DTT, and 1 mM EDTA. After 30 min incubation at room temperature, protein bound to liposomes was recovered by ultracentrifugation at 100,000 g for 30 min. The pellets were then resuspended in SDS-PAGE buffer and resolved by SDS-PAGE. Western blot analysis was performed using primary mouse anti-T7 antibody and Alexa 680 goat anti-mouse secondary antibody. Band intensity was quantifi ed using the LI-COR Odyssey Infrared Imaging System from LI-COR Biotechnology (Lincoln, NE).

PIPKI kinase activity assay
Activity assays were performed as described ( 36 ) for 15 min at room temperature in 100 l reactions containing purifi ed with intracellular pathogens, such as Listeria , Vaccinia , and Shigella ( 11,12 ), and in cells stimulated with plateletderived growth factor (PDGF) and pervanadate ( 13 ). It has been demonstrated that intracellular pathogens use actin comets and the vesicles associated with them for intracellular propulsion ( 12,(14)(15)(16).
In contrast to the widely proposed function of PI(4,5)P2, the molecular pathways regulating PIP kinases as well as the precise mechanisms controlling their catalytic activity still remain largely unknown. A common enzymatic property of all three PIPKI members is that they are strongly stimulated by phosphatidic acid (PA) ( 17,18 ); this stimulation is specifi c to PA, as other acidic phospholipids, such as phosphatidylserine (PS), has no effect on the activity of PIPKIs ( 17,18 ). The best-known signaling PA is produced by the hydrolysis of phosphatidylcholine by phospholipase D (PLD) (19)(20)(21). Several cellular functions, such as actin cytoskeletal reorganization and vesicle traffi cking, are regulated by both PLD and PIPKI ( 22,23 ). Interestingly, PI(4,5)P2 is also an essential factor for PLD activity ( 22,23 ). It has been long proposed that this positive feedback regulation is important for cellular functions. However, this hypothesis is mainly supported by in vitro observations. It is unclear whether PA regulation of PIPKIs occurs in intact cells and whether it is physiologically relevant.
In the current study, we demonstrate that PA regulates the membrane targeting and activity of PIPKI through direct binding to specifi c basic amino acid residues on its membrane interacting surface. Using the PA-binding-defi cient mutant created in this study and pharmacological inhibitors, we further show that PLD-generated PA is required for the formation of actin comets and foci induced by PIPKI ␥ .

General reagents and antibodies
Cell culture media and sera were from Invitrogen (Carlsbad, CA). DAG kinase inhibitors were from Calbiochem (La Jolla, CA). All other reagents were of analytical grade unless otherwise specifi ed and were from Sigma (St. Louis, MO). All restriction enzymes were from New England Biolabs (Ipswich, MA). T7 antibody was from EMD Chemical (Cincinnati, OH). GFP antibody was from Abcam (Cambridge, MA). Rhodamine-phalloidin, goat anti-mouse IgG conjugated with Alexa 680, and goat anti-rabbit IgG conjugated with Alexa 680 were from Invitrogen. The PLD inhibitor FIPI (PLDi-FIPI) originally developed by Novartis (East Hanover, NJ) ( 24 ) was provided by Drs. Michael Gelb (University of Washington) and Michael Frohman (Stony Brook University) ( 25 ). The PLD inhibitor VU0155056 was from Avanti Polar Lipids (Alabaster, AL).

Sequence analysis and homology modeling
Primary sequence alignment of PIPKII ␤ and the ␣ , ␤ , and ␥ of PIPKI family members was performed using MultAlin ( 26 ). Secondary structure predictions were performed using the programs nnPredict (Expasy Tool) and Phyre ( 27 ). The three-dimensional model for a PIPKI ␥ monomer is modeled using homology modeling program MODELER ( 28 ) based on the structure of PIPKII ␤ (Protein Data Bank ID 1BO1) ( 29 ). The resulting structure is then manually docked on to a hypothetical membrane derived from a previously published POPC/POPG bilayer model ( 30 ).

Identifi cation of potential PA-interacting residues in PIPKI ␥
PA-binding proteins typically interact with the negatively charged head group of PA through their positively charged amino acid residues, including lysine, arginine, and histidine ( 40,41 ). This interaction is either mediated by structurally well-defi ned protein modules in some proteins, such as the PH domain of Sos ( 35 ) and the PX domain of p47 phox ( 42 ), or by largely unstructured motifs harboring several basic amino acids in such proteins as Raf-1, mTOR, and DOCK2 ( 41,43,44 ). The crystal structure of PIPKII ␤ revealed a largely fl at surface that binds to the membrane via electrostatic interactions ( 29 ). A similar fl at surface on PIPKIs, for which structures do not yet exist, may be important for PA binding since the small headgroup of PA, i.e., a phosphate, may fi t well in a shallow pocket on a fl at surface. Because PA stimulates the activity of all PIPKI family members but not that of PIPKII or PIPKIII members ( 17,18 ), we hypothesized that the potential PA-interacting residues should fi t the following three criteria: i ) located close to or within the putative fl at membrane interacting surface, ii ) conserved in the three PIPKI members but not in PIPKII, and iii ) positively charged.
We chose to use PIPKI ␥ _i1 [Human Genome Variation Society nomenclature; also known as PIPKI ␥ -87 in previous literature ( 45 )] as the model to study the mechanism by which PA regulates PIPKIs. PA stimulation of PIPKI ␥ is the strongest among the three PIPKI family members ( 18 ), and PIPKI ␥ _i1 does not contain the C-terminal isoform-specifi c focal adhesion targeting tail found only in PIPKI ␥ -90 ( 31,46 ). Primary sequence alignment of PIPKII ␤ and all three members of PIPKI family ( ␣ , ␤ , and ␥ ) revealed that 11 basic amino acid residues on PIPKI ␥ _i1 (K97, R100, H126, H127, H192, K193, K200, R214, K219, K231, and R234), which align to the residues on the fl at membrane interacting surface of PIPKII ␤ , are unique to PIPKI ( Fig. 1A ). We hypothesized that some of these residues are responsible for the PA binding of PIPKI. To better confi rm our prediction based on the primary sequence analysis, we also performed structure modeling of PIPKI ␥ . In the absence of known PIPKI structure, we performed homology modeling based on the published PIPKII ␤ structure (Protein Data Bank ID 1BO1) ( 29 ), which has high homology with PIPKI ␥ ( Fig. 1A ). The structure modeling supported that PIPKI ␥ also has a fl at membrane interacting surface that is rich in basic amino acid residues, including all but two of those identifi ed by primary sequence alignment ( Fig. 1B ). Importantly, the two neighboring histidine residues on a fl exible loop, H126 and H127, lie very close to our hypothetical model membrane. Although this loop points away from the bilayer in our rigid-body model, its predicted fl exibility may allow it to adopt a different conformation upon protein motion and/or binding to a negatively charged bilayer. We reasoned that this loop might be useful for a regulated and specifi c binding of small membrane-bound molecules, such as PA.
protein, 400 M liposome, 20 mM Tris (pH 7.6), 100 mM NaCl, 50 M ATP, 10 mM MgCl 2 , and 5 Ci [ 32 P] ␥ ATP/reaction. Reactions were terminated using 1 M HCl, and the lipid products were extracted with 1:1 chloroform:methanol, washed with 1:1 1 M HCl:methanol, and separated on a 1% potassium oxalatepretreated Silica Gel H TLC plate (Analtech, Newark, DE) developed in chloroform:methanol:ammonium hydroxide:water (90:90:13:9). The air-dried TLC plate was exposed to autoradiography fi lm overnight. The corresponding product bands from the fi lm were scrapped from the TLC plate into scintillation vials, dissolved in EcoLume counting liquid, and quantifi ed on a Beckman Coulter LS 6500 Liquid Scintillation Counter (Indianapolis  IN). PI(4,5)P2 produced was calculated using the equation (cpm/fmol) = (Ci/mmol) × 2.22 × E to calculate fmol (cpm is counts per minute from the scintillation counter, Ci/mmol is the specifi c radioactivity on the day of use, and E is the counter efficiency) . This value was then divided by the amount of protein (ng) used in the assay.

Cell culture and transfection
Cos7 cells were kept in Dulbecco's modifi ed eagle medium (DMEM) containing 10% fetal bovine serum, 100 U/ml penicillin, and 100 mg/ml streptomycin. For transfection, cells grown in 6-well plates or 35 mm dishes (3-4 × 10 5 cells/dish) were transfected with 1 g of DNA/dish using the deacylated polyethylenimine reagent (PEI) ( 37 ). For inhibitor treatment, Cos7 cells were preincubated with the desired inhibitors or DMSO for 1 h before fi xation. To generate tetracycline-inducible stable cell lines, Flp-In T-Rex 293 cells cultured in DMEM containing 10% tetracycline-free fetal bovine serum were transfected with PIPKI ␥ WT and mutants cloned into pcDNA5/FRT/TO with pOG44 at a ratio of 9:1, and then selected with 100 g/ml hygromycin for targeted insertion for three weeks. The expression of PIPKI ␥ proteins in stable cell lines was induced with 1 g/ml doxycycline for one day. Western blotting analysis confi rmed that different PIPKI ␥ proteins expressed at similar levels (supplementary Fig. III).

Confocal microscopy
Cos7 cells were cultured on cover slips and transfected. Twenty hours after transfection, cells were fi xed with 2% paraformaldehyde for 10 min, permeabilized with 0.1% Triton X-100, and stained with rhodamine-conjugated phalloidin (Invitrogen). Fluorescent images were captured using a Nikon A1 confocal microscope (Melville, NY) and processed using Adobe Photoshop CS 5 (San Jose, CA). All experiments were repeated at least three times. The localization of different PIPKI ␥ proteins was analyzed in more than 10 cells using NIH Image J program. For each cell, the ratio of plasma membrane and cytosolic PIPKI ␥ localization was the average of the ratios of pixel intensity on the plasma membrane and that in cytosol (10 pixels from the plasma membrane) from six random areas. For the analysis of actin cytoskeletal reorganization, the number of actin foci and comets was counted in 20 cells for each experiment and presented as the number of foci and comets per cell.

Phosphoinositide analysis
PI4P and PI(4,5)P2 were measured in the parental Flp-In T-Rex 293 cells or Flp-In T-Rex 293 cells expressing PIPKI ␥ WT and mutants using anion-exchange HPLC with suppressed conductivity detection as described before ( 38,39 ).

Statistics
Statistical signifi cance was evaluated by two-tailed Student t -test. All data are shown as mean ± SD. H127 (KRHH); mutant B contained H192, K193, and K200 (HKK); and mutant C contained R214, K219, K231, and R234 (RKKR). Although residues R214 and K219 were not contained within the potential membrane interacting surface in the modeled structure ( Fig. 1B ), they were included in mutant C in case our structure modeling The potential involvement of the above-mentioned basic residues in PA binding was tested by substituting them with uncharged alanine residues in PIPKI ␥ . To simplify mutagenesis and initial analysis, the constructs were divided into three groups based on proximity within the primary sequence: mutant A contained K97, R100, H126, and Conserved amino acids are highlighted in blue (75% conserved) and red (100% conserved). Green arrows and bars indicate secondary structure elements ( ␤ strands and ␣ helices, respectively) and corresponding amino acids for PIPKI (predicted using Expasy and Phyre prediction software) that also correspond to the same PIPKII ␤ secondary structure elements. (B) A homology model of PIPKI ␥ based on the structure of PIPKII ␤ (PDB code 1BO1). The secondary structure of the protein is yellow, and the catalytic residues are red. Groups of residues subjected to alanine substitution in the current study are shown in stick model and colored in dark green and blue (group A, residues K97, R100, H126, and H127); light green (group B, residues H192, K193, and K200); and cyan (group C, residues R214, R219, K231, and R234). A monolayer of hypothetical bilayer is shown as a surface colored from white (hydrophobic core) to blue (polar head group region). or absence of PA. The PA-binding-defi cient mutant A still maintained a comparable activity to the WT protein ( Fig. 3 ). In contrast, it lost the PA-stimulated activity ( Fig. 3 ), suggesting that direct PA binding is critical for the PA-stimulated activity of PIPKI ␥ .

Membrane localization of PIPKI ␥ and its actin cytoskeletal reorganization activity are dependent on PA binding
The PIPKI members are reported to localize to the plasma membrane and/or intracellular vesicles ( 2,5,6 ). When overexpressed, these enzymes induce drastic reorganization of the actin cytoskeleton ( 13,18,49 ). To assess the contribution of PA binding to PIPKI membrane localization and actin cytoskeletal reorganization activity, we utilized a GFP-tagged PIPKI ␥ construct ( 31 ) and subcloned all mutants described above into this plasmid. Western blotting analysis confi rmed that WT and all mutants used in the current study were expressed at similar levels (supplementary Fig. II), suggesting that these point mutations do not induce deleterious effects on protein structure. Similar to previous reports, transiently transfected GFP-PIPKI ␥ localized to the plasma membranes as well as intracellular vesicles in Cos7 cells and induced the formation of two kinds of fi lamentous actin (F-actin) structures resembling those found in pathogen-infected cells (12)(13)(14)(15): comets (phalloidin-stained F-actin structure with a head and a tail) and foci (F-actin-labeled round spots) ( Fig. 4 ). The formation of actin comets and foci has been known to be dependent on PI(4,5)P2 and PI(4,5)P2-regulated actin regulatory proteins ( 50,51 ). Although localized to the same membrane structures as WT protein, the kinase-inactive PIPKI ␥ mutant D253A ( 46 ) failed to induce the formation of actin comets and foci, and the number of comets and foci were similar to those in nontransfected cells ( Fig. 4 ), confi rming their formation is dependent on PI(4,5)P2 production ( Fig. 4A-C ). The PAbinding mutant A (KRHH/A) was diffused in the cytosol and failed to induce the formation of either comets or foci ( Fig. 4A-C ). Corresponding to their PA-binding capacity, the HH/A and KR/A double mutants also showed decreased membrane localization and comet and foci formation ( Fig. 4 ). In contrast, mutants B and C, which bind PA normally, retained membrane localization and similar quantities of comets and foci compared with the WT protein ( Fig. 4A-C ). Taken together, our results suggest that PA binding is critical for the membrane localization of PIPKI and its ability to induce the formation of actin comets and foci. We next measured PI4P and PI(4,5)P2 levels using HPLC with suppressed conductivity detection ( 38,39 ). To better compare the contribution of exogenous PIPKI ␥ proteins on the background of other PI(4,5)P2-generating enzymes in cells and to minimize the toxicity of long-term PIPKI ␥ overexpression, we established several stable cell lines expressing PIPKI ␥ WT and mutants under the control of a tetracycline-inducible promoter using the Flp-In T-Rex system from Invitrogen. Western blotting analysis confi rmed similar expression levels of PIPKI ␥ proteins in prediction did not refl ect the true structure of the protein. Western blotting analysis of these mutants showed that they express at the level similar to WT protein (supplementary Fig. I), suggesting that introducing these point mutations were not likely to cause a structural change. The ability of both WT and mutant PIPKI ␥ _i1 (hereinafter PIPKI ␥ ) to bind PA was then determined by a sedimentation assay using recombinant proteins purifi ed from E. coli and synthetic liposomes. As expected, we found that PIPKI ␥ WT binds to PA-containing liposomes. As predicted by our model, PA binding was dramatically reduced by the mutations in mutant A (KRHH/A) (only around 14% of liposome binding was left) but was not affected by those in mutant B or C, suggesting that the residues in mutant A are important for PA binding ( Fig. 2A ). To further defi ne the key residues in mutant A (KRHH/A), proteins with two amino acid mutations, K97A/R100A (KR/A) and H126A/H127A (HH/A), were tested for their PA binding. Both mutants showed a reduction in PA binding (74% and 38% liposome binding, respectively); however, the reduction in binding was less than with mutant A ( Fig. 2B ). These results suggest that effi cient PA binding requires the coordination of all four residues in group A (KRHH). This result is in line with previous the conclusion that many PA-binding proteins bind to PA through similar multivalent interactions ( 41,43,44 ).
To further confi rm the specifi city of the PA binding through the KRHH residues, we examined whether PIPKI ␥ binds to PA in a concentration-dependent manner and whether the KRHH residues are mainly responsible for PA binding rather than other acidic phospholipids, such as PS, and the catalytic substrate of PIPKI ␥ , PI4P. The purifi ed PIPKI ␥ WT and mutant A were tested for their interaction with different concentrations of sucrose-loaded liposomes containing PA, PS, or PI4P. Similar to the fi ndings above, the binding of mutant A to PA liposomes was signifi cantly decreased compared with WT ( Fig. 2C ). On the other hand, both WT and the mutant A bound to PS and PI4P liposomes at similar levels ( Fig. 2C ). These data support our hypothesis that PA binding of PIPKI is mediated by specifi c amino acids, which are independent of or less critical for residues mediating the interactions with other acidic phospholipids and substrate binding.

PA binding is required for PA-stimulated but not basal PIPKI activity
PA regulation is a unique property of PIPKI ( 17,18 ). Furthermore, the substrate specifi city of PIPKI and PIPKII can be switched by swapping one amino acid in the activation loop (E411 in PIPKI ␥ ) ( 47,48 ). The activation loop is located approximately 280 residues downstream of the PAbinding site in the primary sequences ( Fig. 1A ) and at the opposite side of the PA-binding site on the modeled threedimensional structure of PIPKI ␥ (more than 30 Å away) ( Fig. 1B ). These fi ndings suggest a regulatory role for PA binding of PIPKI rather than direct roles in catalytic specifi city and basal activity of PIPKIs. To investigate the role of PA binding in the regulation of PIPKI activity, PI(4,5)P2 production was measured using liposomes in the presence

Both PA-dependent and PA-independent electrostatic interactions may regulate the functions of PIPKI ␥ in intact cells
A recent study has shown that nonspecifi c electrostatic interaction between acidic phospholipids and PIPKI is important for membrane association of the protein ( 52 ). Based on our fi ndings above, charge alone does not seem to play a signifi cant role in PIPKI regulation since the reduction of charges did not correspond to the reduction of PA binding, kinase activity, membrane targeting, or formation of actin comets or foci. The mutants in groups A (KRHH/A), HH/A, KR/A, B (HKK/A), and C (RKKR/A), had a charge reduction of 4, 2, 2, 3, and 4, respectively. However, PA regulation of PIPKI ␥ was mainly mediated by the residues K97, R100, H126, and H127 ( Figs. 2-4 ). The similarly decreased charge in mutants B and C did not different cell lines with doxycycline induction (supplementary Fig. III). As expected, overexpression of PIPKI ␥ WT led to a signifi cant increase of PI(4,5)P2 level compared with that in the parental cells (about 220% change) and to a reduction of its enzymatic substrate PI4P level (about 50% reduction) ( Fig. 4D ). In contrast, PI(4,5)P2 and PI4P levels in the mutant A-expressing cells were only raised slightly (about 110%) ( Fig. 4D ). Consistent with their abilities of binding to PA and inducing the formation of actin comets and foci, PI(4,5)P2 and PI4P levels in the mutant B-and C-expressing cells were comparable to those in WT-expressing cells ( Fig. 4D ). This result provides evidence that PA regulation is also important for PIPKI ␥ activity in intact cells and that the change in PI(4,5)P2 levels is responsible for actin reorganization in cells that express different PIPKI ␥ s.  Fig. 1 , were mutated to alanine. The purifi ed recombinant WT or mutant protein was incubated with sucrose-loaded liposomes containing 10% PA. After centrifugation, bound proteins in the pellets were detected by Western blot using a T7-tag antibody. The inputs are 10% of proteins used for binding. Quantifi cation of relative binding of the mutants to PA-liposomes compared with WT is shown below the blot. N = 3; ** P < 0.0005 versus WT. (B) Binding between PA and PIPKI ␥ is through multivalent interaction. Mutations of HH and KR residues in group A to alanine also reduced PA binding of PIPKI ␥ . Quantifi cation of relative binding of the mutants to PA-liposomes compared with WT is shown below the blot. N = 3; * P < 0.05 versus WT; ** P < 0.004 versus WT. (C) Specifi c PA binding of PIPKI ␥ . The purifi ed recombinant WT or mutant A was incubated with different concentrations of sucrose-loaded liposomes containing either 10% PA (left), 10% PS (middle), or 10% PI4P (right). After centrifugation, bound proteins in the pellets were detected by Western blot. The Western blot is a representative of three experiments. The dose curves are quantifi cation of relative binding of the mutants to liposomes compared with WT from three independent experiments. membrane localization and actin reorganization activity of mutant B+C in intact cells were most likely caused by the reduction of general electrostatic interaction between membranes and PIPKI ␥ . Taken together with our fi nding that the PA-binding-defi cient mutant A binds to PS and PI4P normally ( Fig. 2C ), our results suggest that both PAdependent and PA-independent electrostatic interactions regulate PIPKI ␥ in intact cells.

PIPKI ␥ activity is regulated by PA generated from PLD
The best-known source of signaling PA is generated by PLD, which is involved in the regulation of actin cytoskeletal reorganization ( 21 ). To investigate whether PLD activity regulates PIPKI functions, we treated cells expressing GFP-PIPKI ␥ with PLD inhibitors. Consistent with the fi ndings using the PA-binding-defi cient PIPKI mutants, PLD inhibitors developed by two different sources, PLDi-FIPI ( 24,25 ) and PLDi-Avanti (purchased from Avanti Polar Lipids) ( 53 ), prevented the membrane association of PIPKI ␥ and signifi cantly blocked the formation of actin comets and foci induced by PIPKI ␥ ( Fig. 7 ). Thus our data provide the fi rst evidence that PLD-generated PA is required for membrane localization and functions of PIPKI ␥ in intact cells.

PA binding is required for both membrane localization and actin reorganization activity of PIPKI
Phospholipid binding is important for both membrane recruitment and activation of many proteins ( 40 ). Alternatively, phospholipid binding may also regulate the activity of proteins through a membrane targeting-independent mechanism ( 40,54 ). To determine whether PA-mediated membrane targeting alone is suffi cient for PIPKI activation, PIPKI ␥ proteins were targeted to membranes through a PA-independent mechanism by adding the membrane targeting motif of the tyrosine-protein kinase Lyn to their N-termini (Mem-PIPKI ␥ ) ( 32 ). Both Mem-PIPKI ␥ and Mem-KRHH/A displayed subcellular localizations to plasma membranes and vesicles similar to the WT protein.
The addition of the membrane targeting motif did not affect the actin cytoskeletal reorganization activity of WT protein or the PI(4,5)P2 and PI4P levels ( Fig. 8A -C ). In contrast, although Mem-KRHH/A was successfully localized to membranes, it failed to induce the formation of actin comets and foci ( Fig. 8A-C ). Measurement of PI(4,5) P2 and PI4P levels confi rmed that targeting KRHH/A to membranes was unable to increase its enzymatic activity ( Fig. 8D ). This result suggests that PA binding is critical for both membrane targeting and activation of PIPKI ␥ . The exact mechanism through which PA regulates PIPKI ␥ still remains to be determined. It is likely that PA binding induces a conformational change in PIPKI ␥ or enhances its binding to other regulators in addition to targeting the protein to membranes.

DISCUSSION
On the basis of sequence analysis, structure modeling, and liposome pulldown assays, we demonstrate that protein residues H126 and H127, which are likely located on affect the membrane localization and actin cytoskeletal reorganization activity, supporting the important role of specifi c PA binding in PIPKI regulation. However, these observations do not rule out the possibility that both specifi c and nonspecifi c electrostatic interactions contribute to the regulation of PIPKI. Membrane targeting of many proteins often requires coincidence detection of different lipids and proteins by individual domains and specifi c and nonspecifi c electrostatic interactions ( 40 ). To determine whether overall protein charge was also responsible for the regulation of PIPKI ␥ in our experimental system, three additional mutants were generated, containing different combinations of mutants A, B, and C, to further reduce the positive charges on the predicted membrane interacting surface of PIPKI ␥ . The new "charge mutants," A+B, A+C, and B+C, resulted in decreases of positive charge of 7, 8, and 7, respectively. The in vitro liposome pulldown experiment revealed that both charge mutants containing A (A+B and A+C) had similar PA binding compared with mutant A, suggesting that removal of additional positive charges from these two mutants did not further reduce the PA binding of mutant A (KRHH) ( Fig. 5 ). Furthermore, a decrease of 7 positive charges at sites away from group A residues did not affect the PA binding of mutant B+C ( Fig. 5 ). These results support our fi nding that PIPKI ␥ specifi cally binds to PA through the KRHH residues in group A.
When expressed in Cos7 cells, the GFP-tagged mutants A+B and A+C localized to the cytosol and failed to induce the formation of actin comets and foci, similar to mutant A ( Fig. 6 ). The GFP-B+C mutant behaved slightly differently from mutant B and mutant C, separately. While a portion of proteins still localized to membranes, there was a decrease of plasma membrane localization ( Fig. 6 ). Consistent with membrane localization, the number of actin comets was also slightly decreased in GFP-B+C-expressing cells ( Fig. 6 ). Because the level of PA binding of mutant B+C was comparable to the WT protein in the in vitro reconstituted liposome binding assay, the decreases in the Fig. 3. Mutant A (KRHH/A) retains the normal basal activity but has lost the PA-stimulated activity. Purifi ed recombinant PIPKI ␥ WT and mutant A proteins (0.5 g or 2 g) were assayed for their kinase activity to convert PI4P to PI(4,5)P2 in the absence and presence of PA using restituted liposomes. Data shown represent one of two experiments with similar results. and C) did not correspond to their PA binding ability, suggesting that charge alone is not the determining factor for PA binding of PIPKI. Two previous studies also reported different PA-binding sites on PIPKIs. Using a 96-well plate coated with phospholipids, the purifi ed mouse PIPKI ␤ (human PIPKI ␣ ) was reported to bind PA through its C terminus ( 55 ). However, this result did not explain why a fl exible loop, along with the nearby K197 and R100 are key residues for specifi c PA binding by PIPKI ␥ . This conclusion is also supported by two other evidences: i ) mutant A (KRHH/A) retained its basal catalytic activity even when PA-stimulated activity was lost, suggesting that its substrate binding ability is not affected by these mutations; and ii ) charge changes on the other PIPKI ␥ mutants (groups B For example, both histone fold and PH domains on Sos bind to PA ( 54 ).
It has become increasingly apparent that most proteins are directed to their functional destination by more than one targeting determinant ( 40 ). The coexistence of multiple signals ensures the specifi city of protein targeting within the cell. The membrane localization and activity of PIPKI may depend on several factors, including substrate, regulatory proteins, and phospholipids. A single mutation (E362A) in the activation loop of the PIPKI ␤ altered its substrate selectivity and caused it to be displaced from the plasma membrane ( 47,48 ). A point mutation on the Rac interaction site also disrupted the membrane localization of PIPKI ␤ ( 57 ). Less is known about how the localization and function of PIPKI are regulated by phospholipids. It has been reported that a nonspecifi c electrostatic interaction with the inner surface of the plasma membrane contributes to the membrane binding of PIPKI ( 52 ). Our data now show that PA binding plays a critical role in the membrane localization and activity of PIPKI. How PA and other PIPKI regulators are coordinated to control the localization of enzymes at the membranes and regulate PI(4,5)P2 synthesis is unclear. Different combinations of regulators may target PIPKI members to various subcellular localizations, where they perform member-specifi c cellular functions. The PA-binding-defi cient mutant identifi ed in the current study, mutant A (KRHH/A), will be a useful tool to dissect the diverse regulation and functions of PIPKI proteins.
A rather surprising fi nding was that targeting the PAbinding-defi cient mutant KRHH/A to the membrane did not rescue one of the most known cellular functions deleting the C terminus of the three PIPKI members did not change the PA stimulation of the enzymes in an earlier report ( 18 ). In another study, by assessing the membrane translocation of human PIPKI ␤ in Cos7 cells, a PA-binding site was mapped to amino acid residues 209-215 ( 56 ), which are very close in proximity to the substrate (PI4P) binding site. The different PA-binding sites identifi ed by us and other groups might have resulted from the use of different methods to study PA-PIPKI interaction. Although it is believe that the liposome system used in our study is the best way to mimic cell membranes in vitro, it is also possible that there are multiple PA-binding sites on PIPKI.   ( 11,12,(14)(15)(16). Previous studies have demonstrated that actin polymerization regulated by PI(4,5)P2, Nck, and WASP are key in the formation of actin comets and foci ( 13,50,51,60 ). Our current fi ndings have demonstrated that PA generated from PLD regulates the activity of PIPKI in intact cells and that this signaling pathway is critical for the PI(4,5)P2mediated formation of comets and foci. PI(4,5)P2 is also an essential factor for the activation of PLD ( 34,61 ). The existence of the PLD (PA)-PIPKI [PI(4,5)P2] positive signaling feedback loop may provide cells a quick way to generate suffi cient PI(4,5)P2 at the right subcellular locations when rapid actin dynamics are crucial under acute conditions, such as bacterial infection. Furthermore, phospholipid signals generated on moving vesicles might provide a precise control of the site of actin polymerization, which is critical for directional vesicle movement in the actin comet-mediated cellular processes.
In summary, we have identifi ed a PA-specifi c interacting site on PIPKI ␥ . Our data also demonstrate that PLDgenerated PA is required not only for the membrane targeting of PIPKI ␥ but also for its activation. These results provide direct evidence that the PLD (PA)-PIPKI [PI(4,5)P2] signaling feedback loop plays a central role in cytoskeletal dynamics in intact cells. Future studies of these phospholipid signals could provide great insight and understanding of fundamental cytoskeletal dynamics and vesicle traffi cking.
The authors thank Dr. Michael Frohman for his suggestions and the members of his laboratory for their support. mediated by PIPKI, induction of actin comets and foci. This differs from many other proteins whose ability to interact with membranes is a key regulatory step in controlling their activity. It is likely that PA binding does not only target PIPKI to membranes but also induces a conformation change required for the activation of the enzyme. Interestingly, two key PA-binding residues, H126 and H127, are located in a rather fl exible loop on the modeled PIPKI ␥ structure, implying a capacity for potential conformational changes upon PA binding. A similar PA regulation of protein activity has been recently reported for Sos ( 54 ). While the binding of the PH domain of Sos to PA is critical for the ligand-induced membrane recruitment of Sos and, hence, for Sos-mediated Ras activation ( 35 ), the binding of the histone fold domain to PA activates Sos through a conformational switch independent of its membrane binding ( 54 ). It is possible that both PA and nonspecifi c electrostatic interaction are required for the initial membrane recruitment of PIPKI ␥ and that only PA binding is required for the stabilization of membrane localization and activation of PIPKI ␥ .
The formation of actin comets and foci may represent a mechanism utilized by cells to transport membranebound organelles under both physiological and pathological conditions. For example, Way and colleagues observed the propulsion of Golgi-derived endosomes by spontaneously formed ''little actin tails'' in HeLa cells ( 11,12 ). The formation of PI(4,5)P2-dependent actin comets and foci were also reported in cells stimulated with PDGF and pervanadate ( 13 ), in the polarized biosynthetic traffi c in epithelial cells ( 58 ), and in the internalization of macropinosomes forming from membrane ruffl es ( 59 ). Many pathogens are known to hijack the host's cytoskeletal machinery and use actin comets for