Rac controls PIP5K localisation and PtdIns(4,5)P2 synthesis, which modulates vinculin localisation and neurite dynamics

In N1E-115 cells, neurite retraction induced by neurite remodelling factors such as lysophosphatidic acid, sphingosine 1-phosphate and semaphorin 3A require the activity of phosphatidylinositol 4-phosphate 5-kinases (PIP5Ks). PIP5Ks synthesise the phosphoinositide lipid second messenger phosphatidylinositol(4,5)bisphosphate [PtdIns(4,5)P2], and overexpression of active PIP5K is sufficient to induce neurite retraction in both N1E-115 cells and cerebellar granule neurones. However, how PIP5Ks are regulated or how they induce neurite retraction is not well defined. Here, we show that neurite retraction induced by PIP5Kβ is dependent on its interaction with the low molecular weight G protein Rac. We identified the interaction site between PIP5Kβ and Rac1 and generated a point mutant of PIP5Kβ that no longer interacts with endogenous Rac. Using this mutant, we show that Rac controls the plasma membrane localisation of PIP5Kβ and thereby the localised synthesis of PtdIns(4,5)P2 required to induce neurite retraction. Mutation of this residue in other PIP5K isoforms also attenuates their ability to induce neurite retraction and to localise at the membrane. To clarify how increased levels of PtdIns(4,5)P2 induce neurite retraction, we show that mutants of vinculin that are unable to interact with PtdIns(4,5)P2, attenuate PIP5K- and LPA-induced neurite retraction. Our findings support a role for PtdIns(4,5)P2 synthesis in the regulation of vinculin localisation at focal complexes and ultimately in the regulation of neurite dynamics.


Introduction
Neurite outgrowth is essential for the proper formation of neuronal networks in the developing nervous system and is ultimately dependent on cytoskeletal dynamics, which often requires the activity of the low molecular weight GTPases of the Rho family. Rho GTPases are molecular switches, which, upon binding GTP, undergo conformational changes that enable the regulation of downstream target proteins (Jaffe and Hall, 2005). Hydrolysis of GTP to GDP turns off and limits the extent of downstream signalling. Members of the Rho family of GTPases, which include RhoA, Rac1 and Cdc42 have been extensively studied in the context of neurite remodelling (Govek et al., 2005).
Neurite dynamics, controlled by Rho GTPase, have been widely studied in model systems such as the neuronal cell line N1E-115 (Gebbink et al., 1997;Hirose et al., 1998;Jalink et al., 1993;Jalink et al., 1994;Kozma et al., 1997;Kranenburg et al., 1999;Postma et al., 1996). These cells maintain a spherical morphology in the presence of serum; however, withdrawal from serum results in rapid cell flattening, differentiation and the formation of neurites. Stimulation of serum-starved cells with different repulsive agonists, such as lysophosphatidic acid (LPA), trigger a rapid collapse of neurites that is dependent upon the activation of RhoA. Using N1E-115 cells, we and others reported that activity of phosphatidylinositol 4-phosphate 5-kinases (PIP5Ks) is essential for neurite retraction in response to LPA, thrombin and sphingosine 1-phosphate (van Horck et al., 2002;Yamazaki et al., 2002). In isolated hippocampal neurones, overexpression of PIP5K also suppresses axonal growth induced by ARF6 inhibition (Hernandez-Deviez et al., 2004).
PIP5Ks are regulated by Rac (Hartwig et al., 1995;Tolias et al., 1998;Tolias et al., 2000;van Hennik et al., 2003), which prompted an exploration of a role for Rac in the regulation of PIP5Ks and PtdIns(4,5)P 2 in neuronal retraction. By defining the molecular interaction sites between the Rac and PIP5K, we have generated a PIP5K mutant that no longer interacts with endogenous Rac. We have used this and other mutants of Rac to demonstrate that the interaction between Rac and PIP5K is essential for the proper localisation of PIP5K and synthesis of phosphatidylinositol(4,5)bisphosphate [PtdIns(4,5)P 2 ]. Finally, we found that the interaction between PtdIns(4,5)P 2 and vinculin, a focaladhesion-stabilising protein, is required for PIP5K and LPAmediated neurite retraction.

PIP5K activity drives neurite retraction
We previously demonstrated that PIP5K activity was essential for neurite retraction in response to LPA and other agents in N1E-115 cells. To demonstrate that overexpression of PIP5K attenuates 3536 Journal of Cell Science 123 (20) neurite outgrowth in a more physiological setting, we examined the effects of PIP5K overexpression in primary murine cerebellar granule neurons (CGNs). Similarly to N1E-115 cells, transient expression of PIP5K potently inhibited neurite formation in CGNs (Fig. 1A). These data are in agreement with a previous study that , isolated from postnatal day 2 mice, were seeded onto laminin-and poly-L-lysine-coated dishes. CGNs were transfected with the constructs shown on the left and subsequently fixed and stained for the neuronal marker GAP-43 and imaged by confocal microscopy. Neurite length was quantified and is shown on right.
Scale bar: 50m. A total of 20 transfected cells were included for each condition. Error bars represent ± s.e.m. (B) N1E-115 cells were treated with the siRNA oligonucleotides as shown and after 36 hours cells were lysed and the RNA was extracted and used for qRT-PCR analysis to determine the levels of PIP5K, PIP5Kand PIP5K (left). After treatment with siRNA, N1E-115 cells were grown in serum and assessed for the number of cells with neurites. The first two bars show controls in which cells have been maintained in serum (control) or have been serum starved to induce neurites (-FCS). (C)Short hairpins targeting the murine PIP5K (herein defined using human nomenclature PIP5K) were cloned into pSuper and transiently transfected into HEK293 in the presence of each of the indicated isoforms of Myc-tagged PIP5K. After 72 hours, the cells were lysed and the lysates were separated by SDS-PAGE and immunoblotted using an antibody against the Myc tag. shRNA knockdown of murine PIP5K is sufficient to induce neurite outgrowth in the presence of serum in N2A cells. N2A cells were transfected with pSuper PIP5K RNAi or with control vector (pSuper) for 72 hours and cells were imaged by light microscopy. Shown are representative images of control cells (left) and cells expressing RNAi for murine PIP5K. Scale bar: 20m. demonstrated that PIP5K overexpression attenuated axonal growth induced by inhibition of Arf in hippocampal neurons (Hernandez-Deviez et al., 2004). To assess whether PIP5Ks are required for controlling neurite growth, we suppressed PIP5K expression in N1E-115 and N2A cells using RNAi and assessed neurite formation in the presence of serum. In N1E-115 cells, siRNA oligonucleotides abrogated the expression of their corresponding PIP5Ks (Fig. 1B) although suppression of PIP5K induced an increase in the mRNA encoding PIP5K. Suppression of PIP5K potently induced neurite formation whereas suppression of PIP5K or PIP5K had little effect (Fig. 1B). The lack of neurite outgrowth on suppression of PIP5K might be a result of upregulation of expression of PIP5K. The endogenous expression of PIP5K is already extremely low in N1E-115 cells and therefore, we tested whether the suppression of PIP5K in N2A cells, which express higher levels, could induce neurite outgrowth. The RNAi construct against PIP5K suppressed PIP5K expression, but did not suppress the expression of PIP5K or PIP5K (Fig. 1C). In N2A cells, siRNA-mediated knockdown of PIP5K induced neurite outgrowth in the presence of serum, whereas transfection with the control plasmid (pSuper) failed to induce neurites (Fig. 1C). siRNA-mediated suppression of PIP5K in N2A cells also induced neurite outgrowth (data not shown). These data support the concept that modulation of PIP5K activity can regulate neurite outgrowth or retraction in model systems such as N1E-115 and N2A cells. The data, however, do not support a role for a specific PIP5K isoform in the regulation of neurite outgrowth or retraction but rather for a role of PtdIns(4,5)P 2 synthesis in regulating this process. In support of this conclusion, we also show that overexpression of various PIP5K isoforms induced retraction in N1E-115 cells (see below).
The decrease in the number and length of neurites in N1E-115 cells induced by overexpression of PIP5K could be a result of enhanced retraction or attenuation of neurite outgrowth. To investigate this, we developed a drug-activatable PIP5K. Fusion of an estrogen-binding domain to kinases and transcription factors often generates an inactive protein (Barthel et al., 2002;Kohn et al., 1998;Pritchard et al., 1995), which can be activated by the addition of the estrogen antagonist 4-hydroxytamoxifen (4-OHT). The in vitro activity of PIP5K, when fused to an estrogen-binding domain (ER-PIP5K) was not significantly altered by 4-OHT treatment (supplementary material Fig. S1A,B). However, ER-PIP5K, unlike wild-type PIP5K, was predominantly localised in the cytosol and only translocated to the plasma membrane upon treatment with 4-OHT (supplementary material Fig. S1C). Overexpression of wild-type PIP5K in HeLa cells leads to dramatic vacuolisation in the cell; however, this did not occur with ER-PIP5K until after treatment with 4-OHT (data not shown). These data demonstrate that fusion of PIP5K to the ER domain prevents PIP5K from localising to the membrane and attenuates its activity in vivo until 4-OHT is added.
In contrast to the transfection of N1E-115 cells with wild-type PIP5K, cells expressing ER-PIP5K developed neurite-like structures that were comparable with those in serum-starved wildtype cells. As expected, ER-PIP5K localised in the cytosol and translocated to the plasma membrane only after treatment with 4-OHT (Fig. 2). 4-OHT treatment also induced retraction of preformed neurites, which eventually led to rounding of the cells (Fig. 2). Cells transfected with the ER-vector alone and treated with 4-OHT did not show neurite retraction (Fig. 2, right column). These data suggest that increased synthesis of PtdIns(4,5)P 2 can lead to the retraction of preformed neurites in N1E-115 cells.

Neurite retraction induced by PIP5K and LPA is dependent on the low molecular weight G protein Rac
Although PIP5K activity is essential for LPA-and Rho-driven neurite retraction in N1E-115 cells, it does not lead to the activation of Rho (supplementary material Fig. S3). Furthermore, it is unclear how these retraction agonists regulate PIP5K. Rac1, Cdc42 and Arf6 regulate PIP5K activity and localisation and therefore we investigated whether their activities are important in regulating PIP5K-mediated neurite retraction. Neither dominant-negative (DN)-Arf6 nor DN-cdc42 blocked the PIP5K-induced neurite retraction in N1E-115 cells (supplementary material Fig. S2). By contrast, expression of DN-Rac1 (RFP-DN-Rac1) blocked PIP5Kinduced neurite retraction (Fig. 3A). PIP5K interacts with the Cterminus of Rac1 (Rac1 CT ; Fig. 3A inset) and cell-permeable Rac1 CT peptides can perturb Rac-mediated actin polymerisation and migration (van Hennik et al., 2003). Overexpression of RFP-Rac1 CT also blocked PIP5K-mediated neurite retraction ( Fig. 3A; for quantification see Fig. 3B). Expression of Rac1 CT and DN-Rac1 did not alter the expression level of PIP5K in N1E-115 cells (Fig. 3C). Both Rac1 CT and DN-RacN17 inhibited LPA-mediated retraction (Fig. 3D), demonstrating that active Rac is also required in LPA-mediated neurite retraction in N1E-115 cells. However, these experiments do not indicate whether Rac acts upstream or downstream of PIP5K. To explore this further, we investigated how PIP5Ks interact with Rac.

Mutation of Glu61 to leucine in PIP5K is sufficient to reduce its association with Rac in vivo
We first demonstrated that endogenous Rac interacts with endogenous PIP5K in neuronal cells. Rac was immunoprecipitated Fig. 2. Induction of PIP5K activity induces neurite retraction. N1E-115 cells were transiently transfected with ER-PIP5K and maintained as controls or stimulated with 4-OHT for the times indicated, after which the cells were fixed and immunostained using an isotype-specific antibody against PIP5K. Scale bar: 25m. The number and length of the neurites was assessed and is shown on the right. In the far right column, cells were transfected with ER alone and stimulated with 4-OHT for 4 hours as a control. The data are expressed as the mean ± s.d. and are representative of at least two different experiments. from N1E-115 cells and the amount of co-precipitating PIP5K activity was assessed. The Rac-specific antibody immunoprecipitated 3.5-times more endogenous PIP5K activity than was immunoprecipitated with a nonspecific antibody (Fig.  3E). Comparable data was obtained after immunoprecipitation of Rac from N2A cells (data not shown).
We identified a peptide region located at the N-terminus of PIP5K, designated peptide 1 (Fig. 4A), which interacted with both recombinant and endogenous Rac (Fig. 4B). The interaction of this peptide with Rac was specific because peptide 1 did not associate with RhoA, another GTPase regulator of PIP5K (Fig.  4C). Furthermore, a peptide corresponding to a C-terminal region of PIP5K (peptide 2), did not associate with either Rac or RhoA (Fig. 4C). We identified a point mutation in full-length PIP5K, 3538 Journal of Cell Science 123 (20) PIP5K E61L that, when compared with wild-type PIP5K, displayed reduced association with both the Rac CT peptide (Fig. 4D) and fulllength Rac1 (Fig. 4E).

PIP5K E61L localises to the cytosol but relocalises to the plasma membrane upon overexpression of Rac
Having established that PIP5K E61L has reduced affinity for Rac1, we characterised how this point mutation affects the enzyme's localisation and catalytic activity before examining its role in neurite retraction. Unlike wild-type GFP-PIP5K, which predominantly localised at the plasma membrane, GFP-PIP5K E61L was cytosolic when expressed in HeLa cells (Fig.  4F). GFP-PIP5K E61L also localised in the cytosol in HEK293 cells (data not shown) and in N1E-115 cells (Fig. 5B). Although these were serum starved overnight and imaged using confocal microscopy. Representative images are shown, from which neurite length and number were quantified (B). Scale bar: 10m. The inset demonstrates that Rac1 CT is able to interact with PIP5K. HeLa cells expressing Myc-tagged PIP5K were lysed before addition of either biotinylated control peptide or biotinylated Rac1 CT peptide. Peptide-protein complexes were purified using streptavidin-coated beads. Associated PIP5K was visualised by western blot (top) and associated PIP5K activity was determined (bottom). (C)Lysates from N1E-115 cells transiently transfected with the indicated constructs were lysed and western blotted using the antibodies shown. (D)DN-Rac1 attenuates LPA-mediated neurite retraction. N1E-115 cells, expressing RFP (control) or mRFP-DN-Rac1 (Rac N17 ) or RFP-Rac1 CT (Rac1 CT ), were serum starved overnight, treated with LPA (0.5M) fixed and the number of round cells was quantified. The data are from two experiments where at least 30 cells were assessed for each treatment. The data are shown as the mean ± the range from the two experiments. (E)Cells lysates from N1E-115 cells were immunoprecipitated using either a nonimmune (NI) or an anti-Rac antibody (Rac). Immunoprecipitates were then split and a portion was used to assess the presence of coprecipitating PIP5K activity (middle and quantified in the top panel) or used for western blotting to demonstrate the specific immunoprecipitation of Rac (bottom).
data are consistent with a role for Rac in the regulation of plasma membrane localisation of PIP5K, the E61L mutation could potentially affect PIP5K folding or inhibit its interaction with other unidentified interacting proteins. PIP5K E61L had the same in vitro PIP5K activity as the wild-type enzyme after immunoprecipitation [GFP-PIP5K synthesised 3250 U PtdIns(4,5)P 2 , whereas GFP-PIP5K E61L synthesised 3600 U], suggesting that it was unlikely that the E61L mutation affected protein folding. To demonstrate that the reduced affinity for endogenous Rac was primarily responsible for the delocalisation of PIP5K E61L , we overexpressed Rac, expecting that this might drive the interaction between Rac and PIP5K E61L in vivo. Although PIP5K E61L was cytosolic when overexpressed alone ( Fig. 4G, left panel), co-expression with active Rac led to the localisation of PIP5K E61L on the plasma membrane ( Fig. 4G, right panels) and colocalisation with Rac. The amino acid sequence of peptide 1 is shown in the grey box. This region is highly conserved among the three isoforms of PIP5K. Amino acid numbering corresponds to each separate murine PIP5K isoform. The arrow indicates the conserved glutamate, which when mutated to a leucine residue generates a mutant (PIP5K E61L ) that no longer interacts with Rac. (B)Endogenous Rac and recombinant GST-Rac1 bind to peptide 1 of murine PIP5K. Peptide 1 was coupled to activated NHS-Sepharose beads and incubated with either purified recombinant GST-Rac1 or HeLa cell lysates (where indicated) and associated Rac binding was determined by western blot. Duplicate samples for each condition are shown. (C)In a similar experiment, HeLa cell lysates were incubated with peptide 1 or peptide 2 coupled to NHS-Sepharose beads and interaction with endogenous Rac (top) or Rho (bottom) was assessed. Peptide 1 specifically binds endogenous Rac but not RhoA, whereas peptide 2 binds neither protein. (D)HeLa cells expressing GFP-PIP5K or GFP-PIP5K E61L were lysed before addition of biotinylated TAT or TAT-Rac1 peptide. Peptide-protein complexes were collected using streptavidin-coated beads. Western blotting, with an anti-PIP5K antibody determined the amount of associated PIP5K. The total expression levels for GFP-PIP5K and GFP-PIP5K E61L (top) or their interaction with the control peptide (middle) and with the Rac1 CT peptide (bottom) are shown. (E)HeLa cells expressing GFP-PIP5K or GFP-PIP5K E61L were lysed before addition of purified recombinant GST-Rac1. GST-Rac1-protein complexes were collected using glutathione Sepharose beads. Associated PIP5K was visualised by western blot. The expression levels of the total GFP-PIP5K and GFP-PIP5K E61L are shown (top), whereas the amount of each affinity-purified PIP5K is shown in the bottom. (F)Localisation of GFP-PIP5K and GFP-PIP5K E61L in HeLa cells was determined by confocal microscopy. In contrast to GFP-PIP5K, which localises to the plasma membrane, GFP-PIP5K E61L localises to the cytosol. Scale bar: 15m. (G)GFP-PIP5K E61L alone (left) or expressed with Myc-tagged L61Rac1 (right) were transfected into HeLa cells. After 24 hours the cells were fixed, stained with an anti-Myc antibody (bottom right) and analysed by confocal microscopy. When expressed alone, GFP-PIP5K E61L localises to the cytosol; however when co-expressed with Rac1, GFP-PIP5K E61L colocalises with Rac at the plasma membrane.

The interaction between PIP5K and Rac is required to induce neurite retraction in N1E-115 cells
Unlike the overexpression of wild-type PIP5K, PIP5K E61L did not induce neurite retraction in N1E-115 cells after serum starvation (Fig. 5A). PIP5K E61L localised in the cytosol in N1E-115 cells (Fig. 5B, top panel), but translocated to the plasma membrane when co-expressed with wild-type Rac (Fig. 5B, bottom panels). Furthermore, PIP5K E61L induced neurite retraction when coexpressed with wild-type Rac (Fig. 5B, bottom panels and quantification in Fig. 5C). These data show that Rac controls the plasma membrane localisation of PIP5K and that the interaction between Rac and PIP5K is required for neurite retraction. In HeLa cells, expression of PIP5K induces rounding and vacuolisation. Expression of PIP5K E61 was cytosolic and did not induce these phenotypes. However, when co-expressed with wild-type Rac, PIP5K E61 was present on the membrane and induced phenotypes that were similar to wild-type PIP5K. PIP5K E61 did not relocalise to the membrane nor did it induce rounding or vacuolisation when it was co-expressed with a mutant of Rac that does not interact with PIP5K (supplementary material Fig. S4). These data show that PIP5K E61L localises in the cytosol because of reduced association with endogenous Rac. Peptide 1 and Glu61, which mediate the interaction between PIP5K and Rac are both highly conserved in all PIP5K isoforms. The corresponding glutamate in the PIP5K isoforms (PIP5K, Glu117; PIP5K, Glu61; PIP5K, Glu111) was mutated to leucine and we analysed their ability to induce neurite retraction in N1E-115 cells. As expected, PIP5K, PIP5K and PIP5K all induced neurite retraction when overexpressed, whereas the corresponding glutamate mutants did not (Fig. 5D). The glutamate mutants in all PIP5K isoforms no longer localised at the plasma membrane whereas their wild-type counterparts did (data not shown).
To demonstrate the requirement for endogenous Rac in the localisation and activity of PIP5K, we suppressed the expression of endogenous Rac and assessed PIP5K localisation and ability to induce neurite retraction. When GFP-PIP5K was coexpressed with an RNAi construct that targeted the expression of both Rac1 and Rac3 (Hajdo-Milasinovic et al., 2007), GFP-PIP5K plasma membrane staining was strongly reduced (Fig. 5E). The suppression of Rac also attenuated PIP5K-induced neurite retraction (Fig. 5E, bottom panels). Previous studies in N1E-115 cells showed that Rac3 was required for LPA-induced rounding and we found that specific suppression of Rac3 levels (Hajdo-Milasinovic et al., 2007), also attenuated neurite retraction induced by PIP5K (Fig.  5F). These data demonstrate that Rac is required to localise PIP5K at the membrane and that Rac is required for PIP5K-induced neurite retraction.
As PIP5K E61L did not itself induce neurite retraction, we assessed its role in LPA-induced neurite retraction. In control cells, LPA stimulation triggered neurite retraction and cell rounding, which as previously demonstrated, was blocked by the overexpression of a kinase-inactive PIP5K (PIP5K KD ) (Fig. 5G). Surprisingly, overexpression of the PIP5K E61L (Fig. 5G) also attenuated LPA-induced neurite retraction. In vitro, PIP5K E61L is active, and it is therefore surprising that it acts as a dominant negative. To further explore this, we assessed the levels of PtdIns(4,5)P 2 in cells overexpressing either wild-type PIP5K or PIP5K E61L . Surprisingly, we found that PIP5K E61L elevated PtdIns(4,5)P 2 levels to the same extent as wild-type PIP5K in HeLa, HEK293 and N1E-115 cells (Fig. 5H). We hypothesised that uncoupling PIP5K from Rac (PIP5K E61L mutant) might lead to aberrantly localised synthesis of PtdIns(4,5)P 2 , which inhibits LPA-driven retraction. To test this concept, we introduced a point mutation into the PIP5K E61L that inactivated its PIP5kinase activity (PIP5K KDE61L ). In accordance with our hypothesis, PIP5K KDE61L no longer attenuated LPA-induced neurite retraction in N1E-115 cells (Fig. 5G, bottom panel). Together, these data suggest that the interaction between Rac and PIP5K is required to drive the localised synthesis of PtdIns(4,5)P 2 , which is required for neurite retraction.

PIP5K induced neurite retraction is dependent on the interaction between PtdIns(4,5)P 2 and vinculin
Neurite extension is facilitated by interactions with the extracellular matrix through the formation of focal contacts (Suter and Forscher, 1998;Suter and Forscher, 2000), which can be stabilised by recruitment of vinculin (Xu et al., 1998). Vinculin is maintained in an inactive conformation by the interaction of its globular head domain with its tail domain (Ziegler et al., 2006) and can be The length of the neurites was quantified (C). The mean neurite length of cells expressing PIP5K E61L is significantly higher (P<0.005) than those coexpressing PIP5K E61L and Rac. (D)N1E-115 cells transfected with GFPlabelled wild-type or glutamate mutants of PIP5K isoforms (,  or ). Cells were serum starved and the numbers of round cells were assessed and presented graphically. All wild-type isoforms of PIP5K induce neurite retraction and rounding, which is severely attenuated in the corresponding glutamate mutants. (E)N1E-115 cells co-transfected with a control siRNA (left) or an siRNA targeting the expression of Rac (right), with GFP-PIP5K. Cells were fixed and GFP expression was analysed by confocal microscopy (top). A line was drawn across the cells (as indicted) and the mean pixel intensity was plotted (middle). The membrane to cytosol ratio was determined (bottom left) and demonstrated a loss of membrane staining upon suppression of endogenous Rac. Neurite length (bottom middle) or neurite numbers (bottom right) were also determined and demonstrated that suppression of endogenous Rac expression attenuated PIP5K-induced neurite retraction. (F)N1E-115 cells were transfected as shown and then serum starved overnight. The cells were fixed and the number of cells with neurites was assessed. The data show that GFP-PIP5K-induced neurite retraction requires the expression of endogenous Rac3. (G)PIP5K KD and PIP5K E61L block LPA-induced neurite retraction in N1E-115 neuroblastoma cells. After transfection with the constructs shown, cells were serum starved for 3 hours to induce neurite outgrowth before stimulation with 0.5M LPA. Representative images are shown of fluorescence images (left) before LPA stimulation and phase-contrast images, before LPA stimulation and 15 minutes after LPA stimulation (middle and right, respectively). Scale bar: 20m. The data were quantified and shown in the graph below. In this experiment, an additional point in which a mutation has been introduced into GFP-PIP5K E61L to render it kinase inactive (GFP-PIP5K KDE61L ) is shown and demonstrates that the kinase activity is required for the dominant-negative effect of GFP-PIP5K E61L . (H)HeLa (left), HEK293 (middle) and N1E-115 (right) cells were transfected with the constructs shown and were labelled with [ 32 P]orthophosphate for 2 hours after which phosphoinositides were extracted and the levels of PtdIns(4,5)P 2 were analysed and quantified. A duplicate transfection was used to determine the expression levels of GFP-PIP5K (bottom). Error bars represent s.e.m. activated by numerous mechanisms, including its interaction with actin-binding proteins and acidic phospholipids such as PtdIns(4,5)P 2 . Activation stimulates vinculin recruitment to focal adhesions (Chen et al., 2005;Chen et al., 2006). Interestingly, the role of PtdIns(4,5)P 2 in the activation of vinculin has recently been challenged, and PtdIns(4,5)P 2 also appears to stimulate the dissolution of vinculin from the focal adhesion complex (Chandrasekar et al., 2005;Saunders et al., 2006). This led us to investigate whether vinculin has a role in PIP5K-induced neurite retraction.
Expression of PIP5K in HeLa cells led to a decrease in the average number of focal adhesions in a cell staining positive for vinculin. (Fig. 6A,B). In contrast to PIP5K, expression of either PIP5K KD or PIP5K E61L did not disrupt vinculin staining in HeLa cells, nor did they change the average number of vinculinpositive structures per cell (Fig. 6A, quantified in Fig. 6B). It should be noted that endogenous vinculin levels were unaffected by expression of the constructs used (Fig. 6C). To study vinculin dissolution in focal contacts in N1E-115 cells, serum-deprived cells were treated with a sub-maximal concentration of LPA that was insufficient to cause complete retraction of all neurites, and vinculin localisation was examined in the remaining population of neurites. Before stimulation, the endogenous vinculin staining in control neurites was similar to that found in cells expressing mRFP-PIP5K E61L or mRFP-PIP5K KD , although neurites expressing the kinase-inactive mutant often showed stronger vinculin staining. After LPA stimulation, the number of vinculinpositive focal contacts decreased significantly in the tips of the remaining neurites in control cells. By contrast, the number of vinculin-positive focal contacts remained unchanged in the tips of neurites expressing mRFP-PIP5K E61L or mRFP-PIP5K KD (Fig. 6D, quantified in Fig. 6E). These data suggest that there is a link between vinculin dissolution and PIP5K and LPA-driven neurite retraction. However, vinculin might be lost from focal adhesions as a consequence of the retraction process. To investigate a causal role for PtdIns(4,5)P 2 -mediated regulation of vinculin in focal adhesion dissolution and neurite retraction, we used a mutant of vinculin (vinculin lipid deficient, VincLD) that has reduced affinity for PtdIns(4,5)P 2 (Chandrasekar et al., 2005;Ziegler et al., 2006). Similarly to wild-type vinculin, VincLD interacts with actin and talin and also localises to focal adhesions, suggesting that interaction with PtdIns(4,5)P 2 is not required for the activation of vinculin. Compared with cells expressing wildtype vinculin, focal adhesions in cells overexpressing VincLD are more stable because of a reduced turnover and reduced retrograde sliding (Chandrasekar et al., 2005). If the interaction between PtdIns(4,5)P 2 and vinculin is causal in inducing destabilisation of focal contacts, then overexpression of VincLD would be expected to suppress PIP5K-induced retraction. NIE-115 cells expressed both vinculin and VincLD to approximately the same levels, and both proteins localised in focal contacts and in the cytosol (Fig. 7A). Expression of VincLD or wild-type vinculin alone did not change the average neurite length after serum starvation. However, overexpression of VincLD abrogated PIP5K-induced neurite retraction, whereas wild-type vinculin had little effect (Fig. 7A). To confirm PtdIns(4,5)P 2 -dependent regulation of vinculin in LPA-induced neurite retraction, we overexpressed wild-type vinculin or VincLD in N1E-115 cells, treated the cells with LPA and assessed neurite retraction using videomicroscopy. Compared with cells expressing wild-type vinculin, expression of VincLD significantly delayed neurite retraction induced by LPA treatment (Fig. 7B). These data suggest that the synthesis of PtdIns(4,5)P 2 , either in response to LPA stimulation or PIP5K overexpression, regulates vinculin interaction at focal contacts, which in turn affects contact stability.

Discussion
In N1E-115 cells, PIP5K-or LPA-induced neurite retraction, requires active Rac to regulate the synthesis of PtdIns(4,5)P 2 . Increased PtdIns(4,5)P 2 synthesis is sensed by the phosphoinositidebinding domain of vinculin, leading to less vinculin in focal contacts and ultimately to focal adhesion disassembly and neurite retraction. Our data are consistent with PtdIns(4,5)P 2 synthesis driving neurite retraction in N1E-115 cells and in primary neurons. However, they are not consistent with any specific PIP5K isoform controlling neurite retraction. Rather, we suggest that PIP5K isoforms are regulated by specific inputs to control PtdIns(4,5)P 2 synthesis and ultimately drive focal adhesion disassembly and neurite retraction. It should be noted that unlike previous data showing that overexpression of the talin-binding splice variant of PIP5K, PIP5K661, or its kinase-inactive counterpart can destabilise focal adhesions (Ling et al., 2002), PIP5K-mediated loss of focal adhesions in N1E-115 cells is dependent on PtdIns(4,5)P 2 synthesis.
PIP5K interacts with many low molecular weight G proteins, including Arf (Honda et al., 1999), Rho (Chong et al., 1994;Ren et al., 1996) and Rac (Tolias et al., 1998;Tolias et al., 2000). PIP5K-induced neurite retraction is not dependent on either Arf or Rho, but is dependent on active Rac. Rac interacts with PIP5K in a GTP-independent manner and the differential cellular localisation of activated Rac, compared with its inactive counterpart, might regulate the localisation and activity of PIP5K at the plasma membrane. In our study, dominant-negative Rac (RacN17) attenuated PIP5K-induced neurite retraction, although surprisingly, it did not prevent PIP5K localisation at the membrane. However, immunostaining revealed that RacN17 was localised at the plasma membrane. RacN17 also interacts with PIP5K and this is probably

. A mutant of vinculin unable to interact with PtdIns(4,5)P 2 blocks PIP5Kinduced neurite retraction and significantly delays LPA-induced neurite retraction.
(A)N1E-115 cells transfected with either wildtype GFP-vinculin or GFP-VincLD (which no longer interacts with phosphoinositides). Cells were fixed and analysed by confocal microscopy. Scale bar: 15m. Cell lysates were analysed by western blotting with an antibody against GFP. The western blot shows that both vinculin constructs were expressed to approximately the same level (middle). N1E-115 cells were transfected with the indicated constructs and were serum starved overnight to induce neurites. The numbers of neurites were determined and represented graphically (right). (B)Cells were transfected with GFP (white bars), mutant GFP-VincLD (grey bars) and GFP-vinculin (black bars). The cells were then serum starved, stimulated with LPA and imaged by real-time microscopy. At the time points indicated, the neurite length of individual cells was determined. Error bars represent s.e.m. responsible for maintaining PIP5K at the membrane. The fact that RacN17 attenuates PIP5K-induced neurite retraction suggests either that active Rac is required for PIP5K mediated PtdIns(4,5)P 2 synthesis as well as regulating PIP5K localisation, or that active Rac is also required for neurite retraction downstream of PIP5K.
We have identified a point mutant of PIP5K (PIP5K E61L ) that has decreased affinity for Rac. PIP5K E61L is catalytically active, predominantly cytosolic and unlike its wild-type counterpart, does not induce neurite retraction upon overexpression. The use of PIP5K E61L to study the role of Rac in regulating PIP5Ks gives clearer and more interpretable results than either suppression of Rac expression or the dominant-negative versions of Rac, which both can inhibit or activate many other upstream and downstream Rac targets. Importantly, we show that co-overexpression of PIP5K E61 with wild-type Rac targeted PIP5K E61 back to the membrane and induced neurite retraction. These data are consistent with a decrease in the affinity for Rac being the primary defect in PIP5K E61 localisation. All PIP5Ks tested interact with Rac in a GTP-independent manner and all have the conserved glutamate residue. Expression of all three isoforms of PIP5Ks induced neurite retraction in N1E-115 cells and mutation of the equivalent Glu61 residue to leucine blocked PIP5K-induced neurite retraction (Fig.  5D) and led to their localisation in the cytosol (data not shown). These data suggest that the regulation of PIP5K localisation by Rac is conserved in PIP5K isoforms. It also suggests that mutation of Glu61, or its equivalent residue in PIP5Ks, will be a useful tool in determining the role of Rac in PIP5K-regulated processes. PIP5Ks phosphorylate PtdIns4P at the plasma membrane, and in many lipid kinases, the ability to interact with the membrane is a key regulatory step in controlling their activity. Interaction with PtdIns4P mediates PIP5K membrane interaction (Kunz et al., 2000), as does an electrostatic interaction with the negative charge of the inner surface of the plasma membrane (Fairn et al., 2009). Our data show that interaction with Rac also controls membrane localisation of PIP5K. How Rac regulates membrane localisation of PIP5K is not completely clear. It might purely be a consequence of Rac tethering PIP5K at the membrane, or the interaction with Rac might lead to post-translational modifications of PIP5K, which ultimately control localisation. How these various mechanisms that control localisation at the membrane are co-ordinated to regulate PtdIns(4,5)P 2 synthesis is not clear.
Suppression of Rac expression also leads to cytosolic localisation of GFP-PIP5K and reduced neurite retraction. Of the three isoforms of Rac, Rac1 is ubiquitously expressed, whereas Rac2 is limited to haematopoietic cells and Rac3 is expressed in the brain. Although Rac1 and Rac3, share 92% overall homology and have identical switch loop regions, they do have antagonistic functions when overexpressed or silenced in N1E-115 cells (Hajdo-Milasinovic et al., 2007). Rac1 overexpression induces neurite outgrowth, whereas Rac3 expression induces cell rounding. Interestingly, Rac3 and PIP5K-mediated neurite retraction show similar characteristics: they are independent of Rho signalling; they do not require myosin light chain kinase activity and they are dependent on the loss of adhesion complexes (Hajdo-Milasinovic et al., 2007;Hajdo-Milasinovic et al., 2009). Furthermore, both PIP5K and Rac3 modulate LPA-induced retraction. These data suggest that Rac3 and PIP5K work in the same pathway. Indeed, overexpression of a kinase-inactive PIP5K attenuates Rac3-induced neurite retraction (our unpublished data). The induction of neurite retraction by Rac3 is determined by the polybasic region (PBR) in the C-terminus of the protein, which is different to the PBR in Rac1. As the PBR mediates the interaction with PIP5K, we assessed the interaction of Rac3 with PIP5K and demonstrated that Rac3 interacts better than Rac1 with PIP5K (supplementary material Fig. S5).
Do PIP5Ks and PtdIns(4,5)P 2 synthesis have a physiological role in the determination of neurite dynamics in vivo? PIP5K (Sasaki et al., 2005) or PIP5K (Wang et al., 2008) knockout mice do not show an overt neuronal phenotype, whereas PIP5Kknockout mice are embryonic lethal. However, our data in N1E-115 cells suggest that the level of PtdIns(4,5)P 2 is important in controlling neurite outgrowth. In HeLa cells, knockdown of any one single isoform does not lead to a significant decrease in PtdIns(4,5)P 2 levels, and knockout of PIP5K, which is a highly expressed isoform in the brain, decreases PIP5K activity in an embryonic brain lysate by only 50% (Wang et al., 2008). These data suggest that double or triple knockouts of the PIP5K isoforms might be required to reveal a neuronal phenotype. In this study, to demonstrate physiological importance, we show that overexpression of PIP5K in primary cerebellar granular neurones also induces neurite retraction, which is similar to a previous study showing that axonal extension of hippocampal neurones is attenuated by PIP5K overexpression (Hernandez-Deviez et al., 2004). Furthermore, we show that siRNA-mediated suppression of PIP5K, or the overexpression of a kinase-inactive PIP5K, leads to neurite induction in N1E-115 cells, implicating the importance of local levels of PtdIns(4,5)P 2 . Previous studies have also shown that neuronal outgrowth, in response to guidance cues such as brainderived growth factor and Netrin-1, is dependent on phospholipase-C-mediated PtdIns(4,5)P 2 hydrolysis (Xie et al., 2005). How decreases in PtdIns(4,5)P 2 lead to neurite extension is not clear. A reduction in PtdIns(4,5)P 2 could lead to a decrease in membrane tension (Niebuhr et al., 2002), which might enable cytoskeletal invasion and the initiation of neurite outgrowth PIP5K-mediated retraction can be attenuated by overexpression of a mutant vinculin that is unable to interact with PtdIns(4,5)P 2 . Furthermore, LPA-mediated loss of focal contacts and retraction are attenuated by overexpression of a kinase-inactive PIP5K. In contrast to previous results (Gilmore and Burridge, 1996), our data suggest that enhanced PtdIns(4,5)P 2 synthesis leads to vinculin dissolution from focal adhesions and stimulates focal adhesion turnover. How might increased PtdIns(4,5)P 2 stimulate the loss of focal adhesions? The interaction site between PtdIns(4,5)P 2 and vinculin overlaps with the interaction site between actin and vinculin, and PtdIns(4,5)P 2 can inhibit the actin-vinculin interaction (Steimle et al., 1999). The activation of vinculin and its localisation at focal adhesions depends on its interaction with both actin and talin (Chen et al., 2006). Increased PtdIns(4,5)P 2 synthesis at focal adhesions might decrease the interaction of vinculin with the actin cytoskeleton and thereby favour vinculin inactivation, leading to its dissolution from focal adhesions. The role of vinculin in stabilising focal adhesions is well established. Focal adhesions in vinculin-knockout mouse embryonic fibroblasts are less stable and are generally smaller (Xu et al., 1998). Overexpression of activating mutants of vinculin stimulates vinculin interaction with focal adhesions and enhances the number and size of focal adhesions (Humphries et al., 2007). Furthermore, and relevant to this study, embryos derived from vinculin-knockout mice show poorly developed neuronal circuitry, and suppression of vinculin expression in PC12 cells leads to fewer extended neurites, filopodia and lamellipodia (Varnum-Finney and Reichardt, 1994).
At the rear of a migrating cell, the disassembly of focal adhesions is required for the release of the cell from the extracellular matrix to enable directional migration. We suggest that activation of PIP5K and increased local synthesis of PtdIns(4,5)P 2 at the trailing edge of a migrating cell regulates vinculin interactions, to decrease focal adhesion stability. Interestingly, a recent study has demonstrated that PIP5K is localised to the uropod and that PtdIns(4,5)P 2 synthesis is required for uropod retraction during neutrophil migration (Lokuta et al., 2007).

Materials and Methods
Chemicals GAP-43, NeuN, and GFAP antibodies were purchased from Chemicon and DMEM medium and supplements were obtained from Gibco (Karlsruhe, Germany).

Cell lines and cultures, protein preparation and immunoblotting
N1E-115 cells were cultured in DMEM (Gibco) medium containing 10% FCS (Biochrom, Berlin, Germany) and antibiotics (Gibco) in a humidified atmosphere at 37°C and 5% CO 2 . Cell lines were transfected using FuGene6 (Roche) according to the manufacturer's protocol. For total protein extraction, cell pellets were resuspended in lysis buffer containing 50 mM Tris-HCl (pH 8), 150 mM NaCl, 1% Triton X-100, 0.1% SDS, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride with the inclusion of a protease inhibitor cocktail (Roche). After centrifugation at 15,000 g for 10 minutes at 4°C, 20 g of the supernatant was subjected to SDS-polyacrylamide gel electrophoresis, electroblotted onto a nitrocellulose membrane and immunostained as indicated in the results. Equal loading amounts of the probes were estimated using immunostaining with anti-mouse -tubulin or actin monoclonal antibodies (Sigma).

Dissociated cerebellar granule neuron cultures
The whole cerebellum was isolated from postnatal FVB mice (P2-P4) and meninges were carefully removed. Tissues were transferred to culture medium [Neurobasal medium supplemented with 30 nM sodium selenite (Sigma), B27 (Gibco), penicillin and streptomycin] and gently triturated with a fire-polished Pasteur pipette and trypsinised for 5 minutes at 37°C. Cells were seeded on glass coverslips pre-coated with laminin and poly-L-lysine (20 g/ml and 0.2 mg/ml, respectively) with a density of 200,000 cm 2 and cultured at 37°C under 5% CO 2 . Cytosine arabinofuranoside (10 M, Sigma) was added after 16 hours to inhibit replication of non-neuronal cells. FuGene6 was used for transfection. Cultured cells were fixed with 4% paraformaldehyde, stained for GAP-43 (Chemicon) and processed for confocal microscopy.

Statistical significance
Data from all experiments were obtained from at least three independent experiments. Data from immunoblots shown in the figures are selected representatives from performed experiments. Analysis was performed using the Mann-Whitney U-test, or as described (StatviewII, Abacus). The level of significance was set at *P<0.05, **P<0.01, ***P<0.001. Error bars represent s.d.

Metabolic labelling with [ 32 P]orthophosphate
Transfected or non-transfected cells were orthophosphate labelled as previously described (Halstead et al., 2001). After extraction, phospholipids were treated with monomethylamine and the deacylated products were analysed using PEI-cellulose plates (Whatman) as described previously (Halstead et al., 2001).

Peptide and GST Rac1 binding
The 20-residue peptide designated petide1 was used to pull down either purified GST Rac1 or endogenous Rac1 from HeLa cell lysates. Briefly, the peptide was coupled with NHS-activated Sepharose beads before incubation with HeLa cell lysates. Cells were lysed in Lysis Buffer B (50 mM Tris-HCl, pH 7.5, 400 mM NaCl, 10 mM MgCl 2 , 10% glycerol, 1% Nonidet P-40 and 0.1% SDS), centrifuged for 10 minutes at 14,000 r.p.m. at 4°C. The supernatant was then incubated with approximately 10 g of one of the coupled peptides at 4°C for 1 hour while rotating. After incubation, samples were washed five times in cell lysis buffer and resuspended in 25 l SDS sample buffer. Rac1 or RhoA binding was determined by western blot analysis. For the peptide binding experiments with the Rac1 C-terminus peptide, TAT fusion peptides (van Hennik et al., 2003) were used. Cells were transfected with either wild-type PIP5K or PIP5K E61L before lysis. Bound PIP5K was detected by western blotting using a GFP antibody that recognised the GFP fusion proteins of PIP5K. For experiments with full-length GST-Rac1, before incubation with cell lysates, GST-Rac1 was coupled to glutathione Sepharose beads. After incubation with cell lysates at 4°C for 1 hour while rotating, glutathione Sepharose pellets were washed five times with lysis buffer and bound PIP5K was detected by western blotting using a GFP antibody that recognised the GFP fusion proteins of PIP5K.

Immunoprecipitation
Cells were lysed by the addition of lysis buffer (50 mM Tris-HCl, pH 8.0, 50 mM KCl, 10 mM EDTA, 1% NP-40), left on ice for 15 minutes and the nuclei and cell debris were removed by centrifugation (14,000 r.p.m. in a microcentrifuge, 4°C).
Lysates were incubated with appropriate antibodies overnight and were captured using Protein-G Sepharose (1 hour 4°C). Immunoprecipitates were washed three times in the above buffer and were then resuspended in SDS loading buffer, heated to 73°C for 10 minutes and separated by SDS-PAGE (Invitrogen pre-cast gels). The separated proteins were transferred to nitrocellulose and then immunoblotted using the appropriate antibodies. Detection was carried out with either ECL or Super Signal (Pierce).