Lipid Raft Disruption Triggers Protein Kinase C and Src-dependent Protein Kinase D Activation and Kidins220 Phosphorylation in Neuronal Cells*

Kidins220 (kinase D-interacting substrate of 220 kDa) is a novel neurospecific protein recently cloned as the first substrate for the Ser/Thr kinase protein kinase D (PKD). Herein we report that Kidins220 is constitutively associated to lipid rafts in PC12 cells, rat primary cortical neurons, and brain synaptosomes. Immunocytochemistry and confocal microscopy together with sucrose gradient fractionation show co-localization of Kidins220 and lipid raft-associated proteins. In addition, cholesterol depletion of cell membranes with methyl-β-cyclodextrin dramatically alters Kidins220 localization and detergent solubility. By studying the putative involvement of lipid rafts in PKD activation and signaling we have found that active PKD partitions in lipid raft fractions after sucrose gradient centrifugation and that green fluorescent protein-PKD translocates to lipid raft microdomains at the plasma membrane after phorbol ester treatment. Strikingly, lipid rafts disruption by methyl-β-cyclodextrin delays green fluorescent protein-PKD translocation, as determined by live cell confocal microscopy, and activates PKD, increasing Kidins220 phosphorylation on Ser919 by a mechanism involving PKCϵ and the small soluble tyrosine kinase Src. Collectively, these results reveal the importance of lipid rafts on PKD activation, translocation, and downstream signaling to its substrate Kidins220.

PKD is mainly cytosolic, treatment with phorbol ester or receptor stimulation provokes a rapid recruitment of the enzyme to the plasma membrane (16, 38 -41). Although this translocation process has been well documented for PKD, very little is known regarding its plasma membrane microdomain distribution. In eukaryotic cells, the plasma membrane structure and composition are not homogeneous and present dynamic structures termed lipid rafts that are rich in low density lipids, such as cholesterol and sphingolipids (42,43). Caveolae represent a sub-population of lipid rafts characteristically containing the scaffolding protein caveolin-1 (44 -46). All lipid rafts including caveolae typically partition into low buoyant density detergentresistant membranes that float during sucrose gradient centrifugation (47). The lipid rafts contribute to keep surface molecules on specific locations of the cell membrane, concentrating a number of signaling molecules, including transmembrane and glycosylphosphatidylinositol (GPI)-anchored receptors as well as intracellular signaling intermediates, while excluding many others (44 -48). Rafts are believed to function as platforms specialized for signaling where, for example, kinases and their substrates could be in close proximity, facilitating signal transduction events to occur rapidly upon the appropriate signal (49).
Kidins220, which is predominantly expressed in neuroendocrine and neural tissues, is found concentrated in certain plasma membrane subdomains reminiscent of lipid rafts in PC12 cells (1). Here we report that Kidins220 is present in lipid rafts in brain and cells of nervous origin such as PC12 cells and primary neurons and that PKD translocates to these membrane microdomains after phorbol ester stimulation. Importantly, in cells where lipid rafts have been disrupted, PKD translocation is delayed, indicating that lipid microdomains are sites facilitating the recruitment of PKD to the plasma membrane. Furthermore, lipid raft integrity is important for keeping PKD activity at low levels since cholesterol depletion induces a significant increase on PKD autophosphorylation and Kidins220 phosphorylation by molecular mechanisms that involve PKC⑀ and the small tyrosine kinase Src. Collectively, this study integrates spatial and temporal data on PKD activation, translocation, and signaling to its substrate Kidins220 and defines the importance of lipid rafts in all these events.

EXPERIMENTAL PROCEDURES
Materials-Phorbol 12,13-dibutyrate (PDBu), methyl-␤-cyclodextrin (MCDX), filipin, poly-L-lysine, and laminin were from Sigma. GF I (GF 109203X or bisindolylmaleimide I) and PP2 were both purchased from Calbiochem. Nerve growth factor (NGF) was from Alexis Corp. (San Diego, CA). The polyclonal rabbit antisera against Kidins220 and phospho-Kidins220-Ser 919 were prepared as described before (1). Anti-Kidins220 monoclonal antibody was purified from supernatants of a hybridoma provided by the Cancer Research UK monoclonal antibody service. This hybridoma was obtained using as antigen a glutathione S-transferase-C terminus fusion protein of Kidins220 (comprising the last cytoplasmic C-terminal sequence of Kidins220, from threonine 1423 to leucine 1763). Phosphospecific antibodies recognizing Ser 916 and Ser 748 (20,25) were kindly provided by Dr. D. Cantrell (University of Dundee). The mouse monoclonal antibodies anti-Thy-1 and antisynaptotagmin I were purchased from AbCam (Cambridge, UK) and Calbiochem, respectively. The rabbit polyclonal antibodies against PKD/PKC, PKC⑀, Src, and caveolin-1 were from Santa Cruz Biotechnology Inc. (Santa Cruz, CA). The polyclonal antibody against PKD/ PKC used for immunoprecipitation and the phosphospecific antibody recognizing PKD-Ser 744 /Ser 748 (21) were purchased from Cell Signaling Technology Inc. (Beverly, MA). The anti-phosphotyrosine monoclonal antibody (4G10) was obtained from Upstate Biotech Inc. (Lake Placid, NY), and the phosphospecific polyclonal antibody recognizing PKD/ PKC phospho-Tyr 463 (23) was a generous gift from Dr. A. Toker and Dr. P. Storz (Harvard Medical School, Boston, MA). Texas Red-and Alexa 488-conjugated secondary antibodies as well as Texas Red-cholera toxin B were purchased from Molecular Probes (Eugene, OR). Horseradish peroxidase-conjugated anti-rabbit and anti-mouse second-ary antibodies and ECL were from Amersham Biosciences. All other reagents were from standard suppliers or as indicated in the text.
Cell Culture and Transfection-Primary dissociated E19 rat cortical primary cultures were prepared from the cerebral cortices of 19-day-old fetal Wistar rats as described (50). Briefly, meninges were removed from the embryonic brains, and cortices were dissected. Tissue was resuspended in minimal essential medium (Invitrogen) complemented with 10% fetal calf serum, 10% horse serum, 0.6% glucose, 16 g/ml gentamicin, and 2 mM glutamine. Cells were counted and seeded on laminin (4 g/ml)and poly-lysine (10 g/ml)-covered glasses in 35-mm dishes at a final concentration of 5 ϫ 10 5 and incubated at 37°C in an atmosphere of 5% CO 2 . Cells were used after 2-8 days in culture, as indicated in the text. Under these conditions Ͼ95% of the cells in culture were neurons as assessed by immunostaining with polyclonal antibodies against the neuronal-specific enolase and the glial fibrillary acidic protein (not shown). PC12 rat pheochromocytoma cells were cultured at 37°C in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 7.5% fetal calf serum, 7.5% horse serum, and 2 mM glutamine in a humidified atmosphere containing 5% CO 2 . When required, cells were treated with NGF (75 ng/ml) for 2 days, GF I (3.5 M), or PP2 (50 M) for 30 min, MCDX (15 mM) or filipin (10 g/ml) for 1 h, or PDBu (200 nM) for different times as specified in the text. For transfection, cells were seeded at 50 -60% confluence on poly-lysine (10 g/ml)-coated glass coverslips. Cells were transfected in serum-free medium 24 h after plating by using 1 g of DNA and 2.5 l of Lipo-fectAMINE2000® reagent (Invitrogen) per 35-mm dish according to the manufacturer specifications and, 24 h later cells were fixed and processed for immunofluorescence. The cDNA containing wild-type GFP-PKD and a translocation-deficient mutant GFP-PKD-P287G have been previously described (38).
Preparation of Synaptosomes from Rat Brain-Synaptosomal fractions were obtained from P15 Wistar rats by differential centrifugation of brain homogenates according to Gray and Whittaker (51) with slight modifications. Whole brains were placed in 40 volumes (w/v) of ice-cold phosphate buffer at pH 7.4 supplemented with sucrose (0.32 M) and homogenized in a Potter homogenizer in the presence of protease and phosphatase inhibitors (1 mM phenylmethylsulfonyl fluoride, 2 g/ml pepstatin, 1 mM dithiothreitol, 10 g/ml aprotinin, 10 g/ml leupeptin, 50 mM sodium fluoride, 5 mM sodium pyrophosphate, and 5 mM sodium orthovanadate) and then centrifuged at 5000 rpm in a SS34 rotor for 5 min at 4°C. The resultant supernatant was further centrifuged at 11,000 rpm in a JA20 rotor for 20 min at 4°C. The crude synaptosomal pellet was resuspended in 0.32 M sucrose and loaded onto a Ficoll gradient (4 ml of 13%, 1 ml of 9%, and 4 ml of 5% Ficoll prepared in 5 mM Hepes, pH 7.4, 0.32 M sucrose) and centrifuged at 22,500 rpm in a SW41 rotor for 35 min at 4°C. Bands at the interface between 5 and 9% Ficoll were isolated, diluted in sodium buffer (10 mM glucose, 5 mM KCl, 140 mM NaCl, 5 mM NaHCO 3 , 1 mM MgCl 2 , 1.2 mM Na 2 HPO 4 , and 20 mM Hepes, pH 7.4), and centrifuged in a microcentrifuge at 4°C and maximal speed for 12 min. The synaptosomal fractions were stored in the presence of protease inhibitors at a protein concentration of 1.5-2.0 mg/ml in sodium buffer.
Isolation of Lipid Rafts-Lipid rafts were isolated running sucrose gradients as described (52). Briefly, proliferating PC12 cells (seeded at 15 ϫ 10 6 cells/100-mm dish) or 10 ϫ 10 6 primary E19 rat cortical neurons cultured for 8 days were used. Cells were put in serum-free medium and pretreated or not for 1 h at 37°C with 15 mM MCDX. Trypan blue staining showed that MCDX treatment did not affect cell viability. Synaptosomal preparations were resuspended in sodium buffer and treated or not for 1 h at 37°C with 15 mM MCDX and centrifuged in a microcentrifuge at 4°C and maximal speed for 30 min. Cells or synaptosome pellets were resuspended in 1 ml of 1 or 0.25% Triton X-100 (depending on the protein subjected to study, as specified in the text) solubilization buffer (25 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1 or 0.25% TritonX-100) plus inhibitors and solubilized for 30 min at 4°C. Sucrose concentration of the lysates was adjusted to 41% before they were overlaid with 8.5 ml of 35% sucrose and 2.5 ml of 16% sucrose prepared in 10 mM Tris, pH 7.4. For lipid rafts isolation, tubes were ultracentrifuged (35,000 rpm in a SW41 rotor, 18 h, 4°C), and 10 fractions were collected from each gradient (from the top to the bottom, fractions 1-10), precipitated with 6.5% trichloroacetic acid in the presence of 0.05% sodium deoxycholate, washed with 80% cold acetone, dissolved, and boiled in 2ϫ Laemmli sample buffer. Samples were analyzed by SDS-PAGE followed by Western blotting and autoradiography.
Immunoprecipitation and Western Blot Analysis-PC12 and 8-day rat primary cortical neurons were lysed in radioimmune precipitation assay buffer (25 mM Tris-HCl, pH 7.6, 1% Triton X-100, 1% sodium deoxy-cholate, 0.1% SDS, 150 mM NaCl, 2 mM EDTA, 2 mM dithiothreitol) with protease and phosphatase inhibitors for 30 min at 4°C, and lysates were then centrifuged for 10 min at 14,000 rpm. PKD was immunoprecipitated as described previously (1). For Western blot analysis, total cell lysates, immunoprecipitates, or different fractions of cells or synaptosomes were boiled for 5 min in 2ϫ Laemmli sample buffer and analyzed by SDS-PAGE followed by transfer (200 mM glycine, 25 mM Tris, 10% CH 3 OH) to nitrocellulose (Schleicher & Schuell) at 100 V for 1 h at 4°C. Membranes were blocked in TBST (20 mM Tris-HCl, pH 7.6, 137 mM NaCl , 0.05% Tween 20) plus 5% lowfat milk powder and incubated for 1 h at room temperature with the different primary antibodies in blocking solution. Membranes were incubated with the appropriate secondary antibodies conjugated to peroxidase as before, and immunoreactive bands were visualized by enhanced chemiluminescence.
In Vivo Cell Imaging, Immunofluorescence, and Confocal Microscopy-For live cell imaging cells were seeded on bottom glass 35-mm dishes coated with poly-lysine (MatTek Corp.) and placed inside a chamber at 37°C with a 5% CO 2 atmosphere on an inverted confocal microscope. For immunofluorescence cells grown on coverslips were fixed for 1 min in 4% paraformaldehyde in phosphate-buffered saline at 4°C followed by methanol for 5 min at Ϫ20°C. After blocking (1% bovine serum albumin for 15 min) cells were incubated with the corresponding primary antibodies for 1 h at room temperature, and immunoreactivity was detected with the suitable fluorophore-conjugated secondary antibody before mounting in slides with Mowiol 4 -88 (Harco). Before fixation and visualization some cells were preincubated with Texas Red-conjugated cholera-toxin B (50 g/ml) for 30 min at 4°C, which binds to the ganglioside type 1 (GM1) on the cell surface. All confocal images were acquired using a Leica TCS SP2 inverted confocal laser microscope with a 63ϫ Plan-Apochromatic oil immersion objective and were normalized for each color separately. GFP fluorescence was excited with an argon laser emitting at 488 nm. Confocal images presented are single sections of a series or two-dimensional maximal projections of a z-series through the cell depth, as specified in the text and figure legends. Images were processed for presentation with Leica Confocal Software Lite version and Adobe Photoshop 6.0 (Adobe Systems Inc, CA).

Kidins220 Extensively Localizes at Lipid Rafts in Rat PC12
Cells and Primary Cultures of Cortical Neurons-Kidins220 is an integral membrane protein that was originally identified in PC12 cells, where it presents an intense discontinuous punctate staining at the plasma membrane and is concentrated at the tip of extending neurites in NGF-differentiated PC12 cells (1). Because Kidins220 is enriched in brain and in the cells of nervous origin (1, 2), we wanted to investigate the cellular distribution of Kidins220 in E19 rat primary cortical neurons in culture and compare it to the one first observed in PC12 cells. Using anti-Kidins220 antibodies we examined Kidins220 localization by immunofluorescence in both cell types. As shown in Fig. 1A, immunostaining of cortical primary neurons revealed the presence of Kidins220 in the cell body and in the neuronal extensions, where it concentrated at the tip and showed a very similar punctate pattern to the one observed in NGF-treated PC12 cells, brighter in certain subdomains of the plasma membrane (see the arrows in Fig. 1A).
Some plasma membrane components, including various GPIanchored proteins, transmembrane proteins, and signaling molecules, together with cholesterol and other sphingolipids accumulate in microdomains of the plasma membrane that are resistant to non-ionic detergent solubilization, known as detergent-insoluble glycolipid-enriched domains or lipid rafts (44,47,53,54). To investigate whether the Kidins220 uneven plasma membrane distribution could be due to its association with these specific detergent-resistant membrane subdomains both cultures were double-stained with antibodies against Kidins220 and Thy-1, a GPI-anchored protein targeted to lipid rafts (55). Confocal microscopy analysis showed that, although there are some membrane regions that present differential staining between these two proteins, an extensive co-localization of Kidins220 with Thy-1 could be observed in discrete regions of the plasma membrane in primary cortical neurons and NGF-differentiated PC12 cells (Fig. 1B). This result shows that a significant amount of Kidins220 co-localizes with the lipid raft marker Thy-1 in primary cortical neurons as well as in NGF-treated PC12 cells.
Because lipid rafts are enriched in low density lipids (such as glycosphingolipids and cholesterol) that are insoluble in 1% Triton X-100, rafts components are easily separated by floating after detergent solubilization and sucrose gradient centrifugation (48). We took advantage of this widely used method to study in detail the association of Kidins220 to lipid rafts. To this end, PC12 cells were solubilized in 1% Triton X-100, and extracts were subjected to sucrose gradient centrifugation. Western blot analysis of fractions collected from the gradients detected most Kidins220 in the top fractions, where the lipid FIG. 1. Kidins220 localizes at lipid rafts. A, E19 rat primary cortical neurons in culture for 2 days (left panel) and PC12 cells treated with NGF for 2 days (right panel) were immunolabeled with the anti-Kidins220 rabbit polyclonal antibody. Kidins220 signal was brighter at some plasma membrane subdomains and at the tip of the neurites in both cultures (arrowheads). Confocal microscopy images correspond to the projection of a series of 0.4-m sections. B, E19 rat primary cortical neurons in culture for 2 days (upper panel) and PC12 cells treated with NGF for 2 days (lower panel) were processed for double-labeled immunofluorescence using Kidins220 polyclonal antibody (green) and with the GPI-anchored protein Thy-1 monoclonal antibody (red). Confocal microscopy images correspond to a single section of a series, except the right panel, which represents a two-dimensional projection of sections in z series taken across the depth of the cells at 0.4-m intervals. Calibration bar, 10 m. The yellow color in the merge image and the maximal projection (P max ) shows the extensive co-localization of these two proteins. C, biochemical identification of Kidins220 in detergentinsoluble fractions. PC12 cells were extracted with 1% Triton X-100 at 4°C and subjected to ultracentrifugation in sucrose gradients as described under "Experimental Procedures." Ten fractions were collected (from top to bottom, fraction 1-10), trichloroacetic acid-precipitated, and Western blot-analyzed with Kidins220, Thy1 and caveolin-1 (used as raft markers), synaptotagmin, PKC⑀, and PKD antibodies. Results are representative of three independent experiments. rafts proteins Thy-1 and caveolin-1 were also found ( Fig. 1C, fractions 2 and 3). Under the same conditions, the integral membrane protein synaptotagmin and cytosolic proteins, which can also be found as membrane-associated, such as PKC⑀ or PKD, remained concentrated at the bottom fractions of the gradient (Fig. 1C, fractions 9 and 10), which contain the soluble material and the majority of the cellular proteins. This result further proved Kidins220 association to lipid rafts.
Cholesterol Depletion Alters Kidins220 Membrane Localization and Solubility-Although some lipid raft components are resistant to cholesterol-sequestering drugs, the use of these agents to disrupt membrane rafts is well established (for review, see Ref. 47). Studies with these drugs led to the conclusion that cholesterol plays a key structural role in the lipid microdomain architecture (56). To analyze the importance of cholesterol in the maintenance of Kidins220 at lipid rafts we used MCDX, a drug that extracts cholesterol from the membranes (57). We first investigated how cholesterol depletion could affect Kidins220 membrane localization by confocal microscopy. Rat PC12 cells and primary cortical neurons (Fig. 2, A and B, respectively) showed a completely different immunostaining and hardly any co-localization of Kidins220 with Thy-1 when treated with MCDX compared with untreated cells, indicating the importance of cholesterol on Kidins220 membrane distribution.
To further analyze if the changes in Kidins220 localization followed by MCDX treatment were accompanied by alterations on Kidins220 detergent solubility, 1% Triton X-100 extracts were prepared from both cell types pretreated or not with MCDX before isolation of rafts by sucrose gradient centrifugation. Gradient fractionation and Western blot analysis of extracts obtained from untreated PC12 cells showed the majority of Kidins220 and Thy-1 floating in fractions 2 and 3 (Fig. 3A). In cortical neurons the bulk of Kidins220 is present in the same fractions; however, Thy-1 is predominantly present in fraction 1, being less abundant in fraction 2 (Fig. 3B). This result is in agreement with the lower degree of co-localization of Kidins220 and Thy-1 in cortical neurons when compared with PC12 cells presented in Fig. 1 and further indicates that Kidins220 and Thy-1 may co-exist in a smaller sub-population of membranes of low buoyancy in neurons than in PC12 cells. Preincubation with MCDX induced a consistent shift of Kidins220 and Thy-1 to higher buoyant density fractions, appearing in more soluble fractions of the gradient both in PC12 cells and cortical neurons (Fig. 3, A and B, respectively).
To demonstrate that the Kidins220 presence in detergentinsoluble glycosphingolipid domains was not an exclusive characteristic of cells in culture, we also examined Kidins220 solubility in brain membrane preparations. Extracts from postnatal rat brain synaptosomes were prepared with 1% Triton X-100 and loaded on sucrose gradients. After Western blot , and P15 rat brain synaptosomes (C) were untreated (ϪMCDX) or pretreated (ϩMCDX) for 1 h with MCDX (15 mM), extracted with 1% Triton X-100 at 4°C, and ultracentrifuged in sucrose gradients as described under "Experimental Procedures." Ten fractions were collected (from top to bottom, fraction 1-10) and Western blot-analyzed for Kidins220 and Thy-1. Kidins220 is solubilized from lipid rafts in the top fractions and is shifted to more soluble fractions at the bottom of the gradient after MCDX treatment. Results are representative of three independent experiments. analysis, Kidins220 and Thy-1 were mainly found in fraction 1 (Fig. 3C). Pretreatment of brain synaptosomes with MCDX completely shifted the Kidins220 signal from fraction 1 to the bottom fractions of the gradient, mainly to fraction 10 (Fig. 3C). However, the drug was much less effective in removing Thy-1 from the top fraction to more soluble fractions, indicating that distinct raft components in brain synaptosomes can display different sensitivities to MCDX and suggesting that Kidins220 and Thy-1 may be found together in a lipid raft sub-pool also in the brain.
Active PKD Translocates to Detergent-insoluble Membranes in PC12 Cells Stimulated with Phorbol Esters-The presence of Kidins220 in lipid microdomains prompted us to investigate whether PKD, the kinase that phosphorylates Kidins220 (1), could be associated to lipid rafts in neuronal cells. In other cell systems, it has been described that treatment with phorbol esters or receptor stimulation provokes the redistribution of PKD from the cytosolic compartment to the plasma membrane (16, 38 -41), but no studies have been published on PKD translocation in cells of nervous origin. Additionally, a detailed anal-ysis of PKD recruitment to differential plasma membrane microdomains has not been reported so far. To approach this issue, we transfected PC12 cells with GFP-PKD, stimulated them with PDBu, and visualized by confocal microscopy the distribution of GFP-PKD compared with the staining pattern of several lipid rafts markers. In resting PC12 cells GFP-PKD is mainly cytosolic (Fig. 4, A-C, upper panels). Stimulation with PDBu for 15 min provokes the recruitment of GFP-PKD to the plasma membrane where it highly co-localizes with Kidins220 as well as with Thy-1 (Fig. 4, A and B). Furthermore, in vivo binding of Texas Red-conjugated cholera toxin B to the lipid raft-enriched ganglioside GM1 at the surface of PC12 cells overlapped significantly with translocated GFP-PKD (Fig. 4C).
To ascertain the nature of PKD membrane association to lipid rafts we further examined its detergent solubility by biochemical fractionation of extracts subjected to sucrose gradient centrifugation. In the case of membrane-associated proteins, for example the members of the Ras family of small GTPases, the differential compartmentalization in membrane microdomains has been widely studied by solubility test assays performed with 0.25% instead of 1% Triton X-100 (58,59). As shown in Fig. 5A, after partitioning of 0.25% Triton X-100 extracts from unstimulated PC12 cells in sucrose gradients, the majority of endogenous PKD appeared in the most soluble material at the bottom of the gradient in fraction 10, although a slight signal of the kinase distributed along the gradient up to lipid raft fractions. After PDBu stimulation a significant amount of total PKD was shifted to fractions 2 and 3, where Kidins220 also floated (Fig. 5B). To follow PKD activation state in the fractions we used a phosphospecific antibody that detects active PKD autophosphorylated at Ser 916 (25). In unstimulated cells, Ser 916 phosphorylation was not detectable (not shown). Immunoblotting of fractions obtained from PDBu-treated cells with the phospho-Ser 916 antibody showed that fractions 2 and 3 were the ones containing the bulk of active PKD (Fig. 5B, PKD-pSer916), thereby demonstrating a prominent association of active PKD to detergent-insoluble membranes after PDBu stimulation.
We next investigated whether cholesterol plays a major role on PKD recruitment to these low buoyancy membrane domains. Incubation of control PC12 cells with MCDX and extraction with 0.25% Triton X-100 completely shifted total PKD to fraction 10, only partially affecting Kidins220 fraction localization (Fig. 5A). The lack of Kidins220 extraction from the raft fractions by pretreatment with MCDX in this case, when compared with previous Figs. 1 and 3, might be due to the use of 0.25% instead of 1% Triton X-100 buffer. In PDBu-stimulated cells, pretreatment with MCDX efficiently eliminated total (PKD) and active PKD (PKD-pSer916) from the raft compartment to the soluble fractions at the bottom of the gradient, where it remained phosphorylated at Ser 916 (Fig. 5B, see the PKD-pSer916 signal in fraction 10). These results show the importance of cholesterol in maintaining active PKD associated to detergent-insoluble fractions and strongly indicate that PDBu-induced PKD activation is not blocked by lipid raft disruption.
Delayed Translocation of PKD after Cholesterol Depletion-The previous results obtained by sucrose gradient centrifugation showed that active PKD translocation could be affected by disruption of lipid rafts with MCDX. To further investigate this notion, we carried out confocal microscopy analysis of PC12 cells transiently expressing GFP-PKD. A comparison of the distribution of GFP-PKD in untreated and MCDX-pretreated cells and the ability of GFP-PKD to translocate to the plasma membrane by PDBu stimulation under both conditions was made by real time confocal imaging of live cells. As shown in Fig. 6A, at time 0 GFP-PKD was distributed throughout the cytosol of untreated PC12 cells. After PDBu stimulation, membrane targeting of GFP-PKD started to be appreciable within 5-7 min and was maximal within 15-17 min (Fig. 6A, panels B-E). Image analysis of cells treated with MCDX up to 60 min (Fig. 6B, panels F-K) showed no major changes on GFP-PKD distribution. After PDBu addition to these cells, the recruitment of GFP-PKD to the plasma membrane only started to be detectable at later time points (15-20 min) and did not reach the maximum within the 25 min recorded (Fig. 6B, with PDBu, panels L-S). The same results were obtained when cells were pretreated with filipin, another cholesterol-sequestering drug (not shown). These images clearly demonstrate that pretreatment with MCDX delays the starting point of GFP-PKD targeting to the plasma membrane after PDBu addition Ͼ2-3 times, clearly impairing the PKD translocation process.
MCDX Treatment Enhances PKD Activity and Kidins220 Phosphorylation without Predominantly Increasing Their Colocalization-Our results from in vivo time-lapse confocal microscopy show that in MCDX-treated cells GFP-PKD hardly starts to translocate after 15 min of PDBu stimulation (Fig. 6B,  panel O). If PKD recruitment to plasma membrane microdomains is a prerequisite for enzymatic activation, the delay on PKD translocation suffered after lipid raft disruption may be accompanied by an impairment of its activation time course. To analyze in detail PKD activation and signaling after cholesterol depletion and to correlate these data with the state of PKD translocation observed by in vivo imaging, we measured PKD autophosphorylation at Ser 916 and PKD-mediated phosphorylation of Kidins220 at Ser 919 . We have previously published that PDBu activation of PKD provokes Kidins220 phosphorylation at this residue (1). Cell extracts from PC12 cells pretreated with MCDX and stimulated with PDBu for 15 or 25 min, 2 time points in which we had detected a lack or partial translocation of PKD, were Western-blotted with antibodies specifically recognizing each phosphoserine residue. Unexpectedly, both PKD-Ser 916 and Kidins220-Ser 919 phosphorylation were consistently increased by MCDX pretreatment in unstimulated cells (Fig.  7A), indicating that disruption of lipid rafts by cholesterol depletion results in a significant increase of PKD activity, signaling, and the ability to autophosphorylate and transphosphorylate its downstream substrate. This effect on PKD activity was also observed when filipin was used (results not shown). Treatment with phorbol esters for 15 or 25 min induced PKD maximal activation, making differences among cells exposed or not to MCDX not appreciable (Fig. 7A). Together with the in vivo confocal images, these results indicate that PKD can be fully active before a complete translocation to the plasma membrane occurs.
Regarding Kidins220, in agreement with our previously published data (1), a 15-min incubation with PDBu raised Ki-dins220-Ser 919 phosphorylation (Fig. 7A). Importantly, MCDX treatment further enhanced this phosphorylation. Maximal phosphorylation of Kidins220 at Ser 919 was obtained after stimulation with PDBu for 25 min independently of MCDX pretreatment (Fig. 7A). The increase in Kidins220 phosphorylation could be just a consequence of PKD activation by MCDX treatment. Alternatively, a more complex situation would result from the combination of this fact together with the dramatic effects of cholesterol depletion on both Kidins220 cellular localization (shown in Fig. 2) and PKD translocation (detailed in Fig. 6) that could affect their co-localization pattern (presented in Fig. 4A). To test this last hypothesis PC12 cells were transfected with GFP-PKD, and we made a comparative analysis of Kidins220 immunostaining and GFP-PKD distribution in fixed cells by confocal microscopy, studying how their colocalization varied after MCDX treatment followed by 15 or 25 min of PDBu stimulation. When the cells were treated with MCDX, GFP-PKD, and Kidins220 co-localization was not significantly increased (Fig. 7B, ϩ MCDX, see Merge and Zoom images). As observed in vivo (Fig. 6), MCDX treatment strongly blocked GFP-PKD translocation after 15 min of PDBu stimulation (Fig. 7B, ϩ MCDX ϩ PDBu, 15 min. Compare these images with Fig. 4, showing GFP-PKD translocation after 15 min of PDBu treatment). Here again, no major changes on their co-localization pattern were detected. Only after incubating the cells 25 min with PDBu, an increase of both proteins co-localizing at the plasma membrane was evident even though GFP-PKD was not fully translocated and considerable amounts of GFP-PKD were still in the cytosol (Fig. 7B, ϩ MCDX ϩ PDBu, 25 min). From all the data, one can speculate that cholesterol depletion, by delaying the time course of PKD translocation, favors a situation in which active PKD phosphorylates Kidins220 in a common intracellular compartment. Because PKD translocation is achieved after longer times of PDBu incubation, active PKD would be able to fully phosphorylate Kidins220 at the plasma membrane. activated by PDBu as wild type PKD (38) is able to increase Kidins220 phosphorylation although not to the same extent as wild type PKD (data not shown). These results support the notion that PKD, once activated, can phosphorylate Kidins220 in an intracellular compartment and that it will further phosphorylate its substrate at the plasma membrane once translocation has occurred. Taken together, these results demonstrate that the integrity of lipid rafts is crucial for the correct localization and signaling of these two molecules.
MCDX Treatment Increases PKD-Ser 748 and Tyrosine Phosphorylation and Its Association to PKC⑀ and Src-We next wanted to analyze the molecular mechanisms participating in PKD activation by MCDX treatment. Phosphorylation is a key event in the control of PKD activation. Two major pathways controlling PKD phosphorylation state and activity have been described. The first one implicates the phosphorylation of Ser 744 /Ser 748 at the activation loop by the novel PKCs, PKC⑀ and PKC (18 -21). The second is mediated by Src/Abl signaling and activates PKD through tyrosine phosphorylation of the pleckstrin homology domain in response to oxidative stress (22)(23)(24). In an effort to study the participation of the PKC pathway on MCDX-induced activation of PKD, we first examined the levels of PKC⑀ and PKC in PC12 cells. As shown in Fig. 8A, both PKC isoforms are expressed in these cells; PKC⑀ was very abundant when compared with PKC. More importantly, Western blot analysis of PKD immunoprecipitates prepared from PC12 cells showed that PKC⑀ (Fig. 8A) but not PKC (data not shown) interacts with PKD in PC12 cells and that MCDX treatment enhanced this association. This result is the first demonstration of the existence of endogenous PKC⑀/ PKD complexes in vivo. Next, PKD activation loop phosphorylation was analyzed by Western blotting the same immunoprecipitates with a phosphospecific antibody recognizing the activating residue Ser 748 (20,21). A significant increase on Ser 748 phosphorylation was detected in MCDX-treated cells (Fig. 8A). This result suggested that PKC⑀, by transphosphorylating Ser 744 /Ser 748 at the activation loop of PKD, participates in the catalytic activation of PKD triggered by MCDX treatment. Tyrosine phosphorylation was also examined in PKD immunoprecipitates from cholesterol-depleted cells. Although MCDX did not modify the pattern of tyrosine-phosphorylated proteins in total lysates, it provoked a significant increase in endogenous PKD-phosphotyrosine levels that was accompanied by the presence of the upstream tyrosine kinase Src co-immunoprecipitating with PKD (Fig. 8B). These results indicate that both PKC⑀ and Src interact with PKD, participating in PKD phosphorylation and activation induced by MCDX.
MCDX-induced PKD Activation Is Mediated by PKC⑀ and Src Signaling Pathways-We finally investigated whether MCDX-mediated PKD activation would be decreased by preincubating the cells with the specific PKC inhibitor GF I (Fig. 9A) or with the Src family of kinases inhibitor PP2 (Fig. 9B). To analyze these results, we used the phosphospecific antibody PKD-Ser 916 to determine the PKD activation state as above and new phosphoantibodies recognizing two other residues within the PKD sequence shown to be transphosphorylated FIG. 8. PKC⑀ and Src associate with PKD in MCDX-treated PC12 cells. Total lysates or PKD immunoprecipitates were prepared from PC12 cells untreated (Ϫ) or pretreated (ϩ) for 1 h with MCDX (15 mM). A, PKC⑀ interacts with PKD in PC12 cells, and MCDX treatment increases both PKC⑀ association and PKD-Ser 748 phosphorylation. The levels of PKC⑀ and PKC in the total lysates or in PKD immunoprecipitates (IP) were determined by Western blot. PKD-Ser 748 phosphorylation was also detected in PKD immunoprecipitates by Western blotting with a phospho-specific antibody (PKD-pSer748). B, PKD is tyrosine-phosphorylated and associates with Src in MCDX-treated PC12 cells. Tyrosine phosphorylation was determined by Western blotting total cell extracts or PKD immunoprecipitates using anti-phosphotyrosine antibodies (PY). Blots were stripped and re-probed with anti-PKD and anti-Src antibodies, showing that both tyrosine-phosphorylated bands (arrow heads) correspond to PKD and Src. Results shown are typical of three independent experiments. specifically by PKC⑀ (21) and Src/Abl (23,24); they are Ser 744 at the activation loop and Tyr 463 at the pleckstrin homology domain, respectively. Pretreatment of PC12 cells with GF I (3.5 M for 30 min) reduced PKD autophosphorylation on Ser 916 and transphosphorylation by PKC⑀ on Ser 744 /Ser 748 , indicating that PKC activity was required for MCDX-induced PKD activation (Fig. 9A). The participation of the Src/Abl pathway on PKD activation by cholesterol depletion was also shown after incubating the cells with the Src inhibitor PP2 (50 M for 30 min), which resulted in a decrease on PKD-Ser 916 autophosphorylation and Tyr 463 transphosphorylation (Fig. 9B). PKD activation by filipin was also inhibited by these two drugs (results not shown). To demonstrate that Kidins220 phosphorylation was dependent on PKD activity under these conditions, we carried out a Western blot of the same cell extracts, which showed a reduction of Kidins220-Ser 919 phosphorylation after GF I or PP2 treatment (Fig. 9C). Altogether, these results demonstrate that both PKC and Src/Abl pathways participate in the observed activation of PKD after lipid rafts disruption.

DISCUSSION
In this study we show by confocal microscopy analysis and biochemical fractionation that Kidins220 is constitutively associated with membrane rafts in brain synaptosomes and in cells of nervous origin, primary cortical neurons and PC12 cells. More importantly, we demonstrate that active PKD is inducibly recruited to lipid rafts after PDBu stimulation and that cholesterol depletion delays the kinetics of translocation to the plasma membrane. Finally, MCDX treatment results in the activation of the endogenous PKD, evidenced by increased Ser 916 , Ser 744 /Ser 748 , and Tyr 463 and the phosphorylation of the downstream substrate Kidins220 at Ser 919 by mechanisms involving PKC⑀ and Src/Abl signaling pathways.
Kidins220 associates constitutively to rafts, and one obvious question that remains to be answered deals with the mechanism by which this occurs. Kidins220 could directly target to or interact with the raft lipids or it could bind raft-associated proteins. A number of receptors and intracellular signaling molecules have been localized to lipid rafts in neurons (60,61). Most membrane raft-associated proteins are constitutive residents because of GPI modification or through palmitoylation at membrane-proximal cysteine residues (43,62). Kidins220 contains several cysteine residues in its very C-terminal tail susceptible to palmitoylation (Cys-1639, Cys-1649, Cys-1663) 2 that could participate in the anchoring of Kidins220 to lipid rafts. An alternative mechanism may involve Kidins220 binding to other detergent-resistant membrane-associated proteins. The best candidates are ephrin and neurotrophin receptors, shown to interact with Kidins220 (2) and whose association with low density lipid rafts has been well documented (61,(63)(64)(65)(66)(67)(68)(69)(70)(71). Further studies will need to examine the structural features that target Kidins220 to lipid rafts, and mutagenesis experiments will determine the raft association signal/s within Kidins220 sequence.
Regarding PKD, many factors may contribute to its inducible recruitment to lipid rafts. PKD translocation to the plasma membrane depends on the C1 domain (16,38). This segment is known to be a high affinity binding site for DAG and phorbol ester (6,10) and plays a crucial role on PKD plasma membrane targeting (17,38,40). PKD-C1 domain would be strongly attracted toward lipid rafts since they cluster many G-proteincoupled and tyrosine kinase receptors and phospholipase C proteins, which convert these in hot spots for DAG production under physiological stimulation conditions. Additionally, pro-tein-protein interactions could also play an important role in this process. Not only Kidins220 but several PKD-interacting proteins such as the B-cell receptor complex, Syk and phospholipase C␥1 (34), PKC⑀ (19), Src (24), and G␣ q (40) together with other lipids involved in PKD signal transduction, such as phosphatidylinositol-4,5-bisphosphate (72), have been shown to concentrate at rafts constitutively or after receptor stimulation (47,(73)(74)(75). Thus, PKD association to lipid rafts can be a multifactorial process where lipid and protein interactions may play a key role, and in which not only the C1 domain but the different domains of the kinase may be involved.
MCDX treatment causes a major delay on PDBu-induced PKD translocation, indicating an essential role of lipid raft integrity in the plasma membrane recruitment of the kinase. After lipid rafts disruption, the enzyme may need more time to find the mislocalized interacting partners, lipids or proteins, at the plasma membrane. Although alternative explanations are possible, these findings could now further explain some unique physiological features of PKD translocation. One of them is the apparent irreversibility of PKD translocation to the plasma membrane, recently proposed to be mediated by its interaction with G␣ q (40). The association of PKD to lipid rafts could also contribute to the persistent nature of PKD translocation, which would be characterized by a very low dissociation constant due to the multivalent binding nature of both lipid raft resident proteins and lipids and PKD. An interesting issue to investigate would be whether the persistent translocation and association of PKD with G␣ q occurs in these detergent-insoluble membranes.
What might be the function of segregating Kidins220 and active PKD into lipid rafts? Compartmentalization of proteins is needed to organize and regulate cellular activities, and although lipid rafts are known to play a key role in polarized membrane trafficking and proteolytic processing, their best established role is signal transduction (43,45,47,62). Many reports implicate neuronal rafts in the compartmentalization of the multitude of signals impinging on the cell surface into distinct signaling cascades by including or excluding key signaling molecules, like cell-surface receptors, intracellular signal-transduction molecules, and transmembrane ligands (60,61,66,76), and regulating in this way cellular functions such as cell adhesion, axon guidance, or synaptic transmission (45,62,77). Whereas it is tempting to hypothesize a possible function for rafts in the polarized sorting of Kidins220 and PKD in the brain and in neurons, as appears to be the case for other neuronal proteins (78,79), their raft association could serve primarily to cluster these two proteins with other signaling molecules.
In this study we show that the majority of Kidins220 partitions into detergent-resistant membrane fractions, and given the established role of lipid rafts and caveolae in signal transduction, this probably reflects a physiologically relevant pool of Kidins220 involved in signal transduction through these microdomains. The differential targeting of Kidins220 to various lipid raft subpopulations, observed by the different co-localization pattern and detergent solubility in PC12 cells, neurons, and brain, may segregate Kidins220 access to different signaling cascades emerging from these microdomains. In agreement to this, Kidins220 has so far been found downstream of two different signaling pathways. After Kidins220 identification as a PKD substrate, the same protein was cloned and named ARMS (ankyrin repeat membrane spanning) as a downstream effector of ephrin and neurotrophin tyrosine kinase receptors (2). Neurotrophins form a family of neurotrophic factors, including NGF, brain-derived neurotrophic factor, NT3, and NT4/5, whose signaling is mediated by two kinds of surface receptors, the Trk family of tyrosine kinase receptors and p75 neurotrophin receptor. Kong et al. (2) demonstrate ARMS interaction with Trks and p75 and its tyrosine phosphorylation upon the activation of Trk and ephrin receptors. As previously mentioned, ephrins and their receptors Trks and p75 are enriched in low buoyant density membranes (61, 63, 65, 66, 68 -71). It has been reported that the signaling cascades triggered by NGF through p75 and TrkA involve lipid rafts (47,65). The fact that drugs such as MCDX decrease NGF-stimulated TrkA autophosphorylation and inhibit signal transduction pathways downstream of TrkA such as the mitogen-activated protein kinase pathway (68) indicates that localization to lipid rafts enhances the signal output of TrkA. Thus, activation of these receptors within lipid rafts could result in Kidins220 tyrosine phosphorylation and in signal propagation through this membrane compartment. Additionally, mechanisms regulating the activity of lipid raft-associated PKD and the total amount of PKD in rafts are also expected to affect Kidins220 signaling. The observation that cholesterol depletion increases basal phosphorylation of Kidins220-Ser 919 , eventually affecting its yet unknown function, provides a mechanism of lipidmediated control of Kidins220 signaling also by regulating PKD activity and localization. Whether the modulation of the physiological properties of Kidins220 by its location to lipid rafts is caused by lipid-protein or by protein-protein interactions or reflects post-translational modifications (e.g. phosphorylation) by proteins concentrated in rafts has yet to be resolved since Kidins220 function and participation in the different cellular processes has to be established.
The selective recruitment of signaling molecules to the plasma membrane is an essential prerequisite for the functioning of various signaling cascades. This is, indeed, the case for PKD, which translocates from the cytosol to the plasma membrane after in vivo stimulation. We suggest a physiological model in which the hydrolysis of phosphatidylinositol-4,5bisphosphate by phospholipase C after receptor stimulation will take place mainly at lipid rafts, increasing the DAG levels in these lipid microdomains and triggering the recruitment of activated PKD to these areas. Active PKD at rafts will be in close proximity to other lipid raft-contained proteins (i.e. active receptors, Ser/Thr or tyrosine kinases as well as G proteins) that may further regulate its activity and to its substrates (Kidins220 and other still unidentified targets), increasing their phosphorylation and regulating their downstream signaling. This model would also integrate and explain the increase on PKD basal activity after MCDX treatment. On one hand, lipid rafts disruption modifies the steady-state activation of the enzyme, as evidenced by an increase on its activation loop and tyrosine phosphorylation, an enhancement of its autophosphorylation, and the phosphorylation of its downstream substrate Kidins220. This can be easily explained assuming that cholesterol depletion can release the raft components normally confined to this specific compartment, including PKD-activating molecules such as PKC⑀ and Src, which would favor their interaction with PKD and subsequent phosphorylation and activation of the enzyme in cellular localizations different from the plasma membrane. On the other hand, the observed delay in its translocation rate produced by MCDX treatment provokes a large shift in the duration time of active PKD outside the plasma membrane. This could lead to dramatic changes in the pattern of proteins that PKD can phosphorylate in a nonnatural localization, explaining the increase in Kidins220-Ser 919 phosphorylation and indicating that altering raft structure may significantly affect PKD functions. These results also suggest that PKD activity, signaling, and function might be affected by the distinct lipid composition of membranes in different cell types or the alteration of membrane lipids either by the clinical use of lipid-lowering drugs or associated to several diseases (80).
In conclusion, this report is the first one identifying Kidins220 and active PKD localization at lipid raft microdomains and demonstrates that cholesterol depletion causes PKD activation through PKC and Src-dependent mechanisms. Based on our results, we propose that protein-lipid interactions should be considered as a mechanism controlling PKD and Kidins220 localization, phosphorylation, signaling, and function. Considering the high molecular weight of Kidins220 and the variety of interacting domains contained within its sequence, it could serve as a scaffold for the assembly and organization of complexes involved in PKD and neurotrophin and ephrin signaling within lipid rafts. Whether there is any connection between ephrins, neurotrophins, and PKD pathways and the role of lipid rafts in segregating these cascades constitutes an interesting issue to pursue in future studies.