Phosphorylation of NG2 Proteoglycan by Protein Kinase C-α Regulates Polarized Membrane Distribution and Cell Motility*

Protein kinase C (PKC)-α phosphorylation of recombinant NG2 cytoplasmic domain and phorbol ester-induced PKC-dependent phosphorylation of full-length NG2 expressed in U251 cells are both blocked by mutation of Thr2256, identifying this residue as a primary phosphorylation site. In untreated U251/NG2 cells, NG2 is present along with ezrin and α3β1 integrin in apical cell surface protrusions. Phorbol ester treatment causes redistribution of all three components to lamellipodia, accompanied by increased cell motility. U251 cells expressing NG2 with a valine substitution at position 2256 are resistant to phorbol ester treatment: NG2 remains in membrane protrusions and cell motility is unchanged. In contrast, NG2 with a glutamic acid substitution at position 2256 redistributes to lamellipodia even without phorbol ester treatment, rendering transfected U251 cells spontaneously motile. PKC-α-mediated NG2 phosphorylation at Thr2256 is therefore a key step for initiating cell polarization and motility.


From the Cancer Research Center, The Burnham Institute, La Jolla, California 92037
Protein kinase C (PKC)-␣ phosphorylation of recombinant NG2 cytoplasmic domain and phorbol ester-induced PKC-dependent phosphorylation of full-length NG2 expressed in U251 cells are both blocked by mutation of Thr 2256 , identifying this residue as a primary phosphorylation site. In untreated U251/NG2 cells, NG2 is present along with ezrin and ␣ 3 ␤ 1 integrin in apical cell surface protrusions. Phorbol ester treatment causes redistribution of all three components to lamellipodia, accompanied by increased cell motility. U251 cells expressing NG2 with a valine substitution at position 2256 are resistant to phorbol ester treatment: NG2 remains in membrane protrusions and cell motility is unchanged. In contrast, NG2 with a glutamic acid substitution at position 2256 redistributes to lamellipodia even without phorbol ester treatment, rendering transfected U251 cells spontaneously motile. PKC-␣-mediated NG2 phosphorylation at Thr 2256 is therefore a key step for initiating cell polarization and motility.
Transmembrane proteoglycans such as CD44 and syndecans make important contributions to communication between the exterior and interior of the cell (1,2). We are elucidating specific signaling functions for NG2, a membrane-spanning chondroitin sulfate proteoglycan found on several types of immature progenitor cells and on a variety of tumor types (3). Two hallmarks of both progenitor and tumor cells are increased motility and proliferation, both of which are influenced by NG2 (4 -7).
Several mechanisms have been suggested to account for the contribution of NG2 to these processes. These include sequestration of growth factors (5,8), modulation of the activity of kringle domain proteins (9,10) and matrix metalloproteinases (11), and interaction with other cell surface molecules or with extracellular matrix components that regulate signaling pathways involved in cell proliferation and motility (12)(13)(14)(15). NG2 engagement has been shown to result in activation of the small GTPases cdc42 and rac (15,16) and in ␤ 1 integrin-dependent activation of focal adhesion kinase and ERK-1/2 1 (17,18). The fundamental importance of these pathways in cell physiology underscores the potential significance of elucidating the specific contributions of NG2 to transmembrane communication.
Protein phosphorylation and dephosphorylation are recognized as critical aspects of intracellular signaling. By altering protein conformation and by creating docking sites for proteinprotein interaction, phosphorylation and dephosphorylation of cytoplasmic tyrosine, serine, and threonine residues provide an extremely versatile means of regulating signaling pathways (19,20). Although the NG2 cytoplasmic domain contains several threonine residues that might serve as phosphorylation sites (21), to date we have had no experimental verification of this type of modification in NG2 and no information about potential functional consequences. The current work sheds initial light on these questions by identifying Thr 2256 as the primary site for NG2 modification by PKC-␣. In addition, we show that phosphorylation at this site changes the distribution of NG2 on the cell surface. Whereas non-phosphorylated NG2 is localized to small membrane protrusions distributed over most of the cell surface, Thr 2256 -phosphorylated NG2 is largely associated with extensive lamellipodia at the cell periphery. This redistribution of NG2 and the resulting polarization of cells are accompanied by significant increases in cell motility.
Antibodies-Affinity-purified rabbit and guinea pig antibodies against rat NG2 were prepared in our laboratory (22). Rabbit antibodies specific for the XTpX(K/R) phosphothreonine motif and for phosphorylated ezrin (Thr 567 ) were obtained from Cell Signaling Technology. Monoclonal antibody against PKC-␣ was from BD Biosciences. Monoclonal antibody against the human ␣ 3 integrin subunit was obtained from Chemicon (Temecula, CA). Rabbit antibodies against ezrin and glutathione S-transferase (GST) were generous gifts from Dr. Heinz Furthmayr (Stanford University) and Dr. Elena Pasquale (The Burnham Institute), respectively. Fluorochome-coupled second antibodies were purchased from BIOSOURCE International (Camarillo, CA). Goat anti-rabbit immunoglobulin coupled to 6-nm colloidal gold was obtained from Jackson ImmunoResearch Laboratories (West Grove, PA). Rhodamine-labeled phalloidin was obtained from Molecular Probes (Eugene, OR).
Cell Culture and Stable Transfections-U251MG human astrocytoma cells (23) transfected with cDNA for rat NG2 have been previously described (24). Cells were grown at 37°C in 5% CO 2 in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Tissue Culture Biologicals, Tulare, CA).
Mutant NG2 constructs (in pcDNA/ampI, Invitrogen) in which specific threonine residues were replaced by either valine or glutamic acid were made using Stratagene's QuikChange mutagenesis kit. For this purpose, overlapping pairs of oligonucleotide primers containing appropriately altered codons were purchased from FisherOligos. The resulting mutations were confirmed by DNA sequencing. Stable transfection of U251MG cells with these mutant NG2 cDNAs was accomplished as described by Stallcup and Dahlin-Huppe (25). Unless otherwise specified, cell populations used for our experiments contained Ͼ85% NG2positive cells.
Immunoprecipitation and Western Blotting-Immunoprecipitation and chondroitinase treatment of NG2 were performed according to . Precipitates were fractionated by SDS-PAGE on 4 -12% Novex gradient gels (Invitrogen) and transferred to Immobilon-P membranes (Millipore, Bedford, MA). The membranes were immunoblotted with antibodies against NG2, GST, PKC-␣, or the XTpX(K/R) phosphothreonine motif. After washing, membranes were treated with horseradish peroxidase-labeled goat anti-mouse or antirabbit IgG (Bio-Rad), and immunoreactive proteins were detected with a chemiluminescence kit (ECL, Amersham Biosciences).
NG2 Cytoplasmic Fusion Proteins-The cytoplasmic region of NG2 was amplified by PCR from a rat NG2 cDNA template, cloned into the BamHI and XhoI sites of the pGEX-4T-1 vector (Amersham Biosciences), and used to transform Escherichia coli BL21. The resulting fusion protein was purified from bacterial extracts using glutathioneagarose beads (Sigma). Mutant forms of the NG2 cytoplasmic fusion protein, in which individual threonines were replaced by glutamic acid, were constructed using the strategy described for full-length NG2 mutants.
In Vitro Phosphorylation of NG2 Cytoplasmic Domain by PKC-␣-10 g of wild type or mutant fusion protein bound to glutathione-agarose beads was incubated with 500 ng of recombinant PKC-␣ (Upstate, NY) for 30 min using the [␥-32 P]ATP assay described by Ng et al. (26). Following this reaction, the protein beads were washed twice with phosphate-buffered saline, and bound material was subjected to SDS-PAGE on 4 -20% gradient gels. Proteins were transferred to Immobilon-P membranes, and labeled components were detected by autoradiography using BioMax MR film (Kodak). Metabolic 32 P Labeling, Phosphoamino Acid Analysis, and Tryptic Peptide Mapping-Cells were incubated in phosphate-free Dulbecco's modified Eagle's medium for 1 h at 37°C and then labeled for 4 h with 4 mCi of 32 P i /ml in phosphate-free medium. Cells were incubated with or without 30 nM calyculin A or 1 M PMA within the last 15 min of labeling. NG2 immunoprecipitates were resolved by SDS-PAGE on 4 -12% gels, transferred to Immobilon P membranes, localized by autoradiography, and excised for further analysis. Two-dimensional phosphoamino acid analysis of excised NG2 was performed as described previously (27). The 32 P-labeled amino acids were detected by autoradiography and identified on the basis of co-migration with ninhydrinstained standards. Two-dimensional tryptic peptide mapping was carried out according to the method of Luo et al. (28).
Immunofluorescence-In a few cases live cells were immunostained at 4°C to evaluate the presence of NG2 on the cell surface. In most cases cells were fixed with 4% paraformaldehyde for 10 min at room temperature prior to immunofluorescence labeling. Labeling was carried out as described by Stallcup and Dahlin-Huppe (25). All specimens were viewed using a Nikon Optiphot microscope (Garden City, NY) equipped for epifluorescence and phase contrast. A Nikon 40ϫ PlanApo objective (oil immersion, 1.0 numerical aperture) was used for all imaging. Images were acquired using Kodak TMAX 400 film and printed on Kodak F5 paper.
Electron Microcopy-Cells were prepared and processed as described for immunofluorescence, except that fluorochrome-labeled second antibodies were replaced by goat anti-rabbit immunoglobulin conjugated to 6-nm colloidal gold. Following the labeling procedure, cells were processed as described by Garrigues et al. (29) and embedded in Pelco Eponate 12T (Ted Pella, Inc., Redding, CA). Ultrathin sections (50 -70 nm) were cut on an Ultracut E ultramicrotome (Reichert-Jung) and double stained with uranyl acetate and lead citrate. Samples were examined at 75 kV using a Hitachi-600A electron microscope.
Cell Motility Assays-Confluent cell monolayers were established in 35-mm tissue culture dishes and "wounded" with a series of parallel scratches using a blue Eppendorf tip (26). Old medium and loose cells were aspirated and replaced with fresh medium, either with or without 20 nM PMA. After a 20-h incubation, cultures were fixed with 4% paraformaldehyde and double stained for NG2 and rhodamine phalloidin. Motility was quantified by counting NG2-positive cells that had clearly migrated into the wound, as judged by separation from the bulk of phalloidin-positive cells in the monolayer. The total number of migratory cells in a given culture was divided by the total length of the wounds to obtain a value for motile cells per millimeter. At least 60 mm of total wound length in each of three replicate cultures was examined for each cell type under each experimental condition.

NG2 Is Phosphorylated on Threonine-
The cytoplasmic domain of rat NG2 contains one serine and seven threonine residues (21). To determine possible utilization of these residues as phosphorylation sites, phosphoamino acid analysis was performed on NG2 immunoprecipitated from transfected U251 cells (U251/NG2) metabolically 32 P i -labeled in the presence of the serine/threonine kinase inhibitor calyculin A (30 nM). Under these conditions NG2 contains only phosphothreonine (Fig. 1A).
Analysis of the cytoplasmic sequence of rat NG2 using a phosphorylation prediction program (PhosNet version 2.0 (30)) identifies Thr 2256 , Thr 2265 , and Thr 2278 as potential phosphorylation sites, based on the presence of the consensus sequence XTX(R/K). An algorithm for sequences involved in proteinprotein interaction (Scansite ver 2.0; 31) further identifies the 2252 KRNKTGK 2258 sequence (consensus (K/R)XXTX(K/R)) as a potential recognition site for PKC. This sequence surrounding Thr 2256 is conserved in the human (32), mouse (33), chicken (GenBank TM accession number XP_425058), and pufferfish (GenBank TM accession number CAF98402) homologs of the proteoglycan, consistent with the idea that the motif may have functional importance. The Thr 2256 site therefore appears to be a prime candidate for cytoplasmic phosphorylation of NG2.
Protein Kinase C Is a Potential Mediator of NG2 Phosphorylation-To further characterize threonine phosphorylation of NG2, the proteoglycan was immunoprecipitated from U251/ NG2 cells and immunoblotted using a phosphothreonine antibody specific for the XTX(K/R) motif. Treatment of cells with PMA (Fig. 1B, lane 2) significantly increased the phosphorylation level of NG2 over the basal level (CT, Fig. 1B, lane 1), suggesting that activated PKC could be responsible for NG2 phosphorylation. Because serine/threonine phosphatases can inactivate PKC (34) and may possibly dephosphorylate NG2, an additional experiment was performed in the presence of the PP1/PP2A phosphatase inhibitor calyculin A (35). This treatment resulted in highly elevated threonine phosphorylation of NG2 (Fig. 1B, lane 3).
PKC comprises an 11-member family of phospholipid-dependent serine-threonine kinases with important roles in cell proliferation, differentiation, motility, and apoptosis (36,37). We utilized two PKC inhibitors to implicate specific PKC isoforms in NG2 phosphorylation. GF109203X is an inhibitor of PKC isoforms ␣, ␤1, ␤2, ␥, ␦, and ⑀, whereas Gö6976 is a highly selective inhibitor of the classic PKC isoforms ␣, ␤1, ␤2, and ␥ (38). Both basal and stimulated levels of NG2 threonine phosphorylation were significantly reduced or abolished when U251/NG2 cells were treated with these agents (Fig. 1C), suggesting mediation of NG2 phosphorylation by a classic PKC isoform. Significantly, PKC-␣ is the sole member of the classic PKC family expressed in U251 cells (39). In support of the possibility that PKC-␣ is responsible for NG2 phosphorylation, we used co-immunoprecipitation to demonstrate the existence of an NG2/PKC-␣ complex during the first 20 min after treatment of U251/NG2 cells with PMA (Fig. 1D).
In Vitro Phosphorylation of the NG2 Cytoplasmic Domain by PKC-␣-To obtain more direct evidence that PKC-␣ mediates NG2 phosphorylation, a GST fusion protein containing the full cytoplasmic region of NG2 (GST-NG2c, Fig. 2A), was used as the substrate for in vitro [␥-32 P]ATP kinase modification by recombinant PKC-␣. SDS-PAGE analysis of phosphorylated products from this reaction reveals a radiolabeled 34-kDa band corresponding to GST-NG2c (Fig. 2B). In contrast, GST itself is FIG. 1. Involvement of classic PKC isoforms in NG2 phosphorylation. A, phosphoamino acid analysis of NG2 immunoprecipitated from 30 nM calyculin Atreated U251/NG2 cells revealed the presence of phosphothreonine. B, U251/NG2 cells were grown under control conditions (CT), or exposed to 100 nM PMA or 30 nM calyculin A (CA) for 20 min. Chondroitinase-treated NG2 immunoprecipitates were analyzed by immunoblotting with an NG2 antibody and with a phosphothreonine antibody that recognizes the XTpX(K/R) motif (Tp). C, basal, PMA-induced, and calyculin-induced threonine phosphorylation in cells pretreated for 1 h with (ϩ) or without (Ϫ) PKC inhibitors GF109203x and Gö6976 at 1 M and 100 nM, respectively, were examined by immunoblotting with the XTpX(K/R)-specific antibody (Tp). D, NG2 was immunoprecipitated following treatment of U251/ NG2 cells with 1 M PMA for 10, 20, 30, and 60 min. Immunoblotting shows that levels of PKC-␣ association with NG2 are substantially increased at 20 -30 min following PMA stimulation, returning to basal levels after 60 min.

FIG. 2. In vitro phosphorylation of the NG2 cytoplasmic domain by PKC-␣.
A, Coomassie Blue-stained electrophoretogram comparing the GST-NG2 cytoplasmic fusion protein (GST-NG2c) and GST itself. B, products from PKC-␣mediated in vitro kinase reactions were fractionated by SDS-PAGE. Autoradiography reveals extensive 32 P incorporation from [␥-32 P]ATP into a 34-kDa band matching the mobility of the GST-NG2c fusion protein. C, non-radioisotopic PKC-␣-mediated kinase reactions were carried out with GST and GST-NG2c coupled to glutathione-Sepharose beads. After washing, components bound to the beads were analyzed by immunoblotting with the XTX(R/K)-specific phosphothreonine antibody (Tp). The mobility of phosphorylated bands match those of GST-NG2c and PKC-␣. Immunoblotting for GST confirmed the presence of GST and the GST-NG2c fusion protein on the blot. not a substrate for PKC-␣ phosphorylation under identical conditions. In addition, immunoblotting of reaction products with the XTX(R/K)-specific antibody reveals phosphorylated components at 80 and 34 kDa (Fig. 2C). The 34-kDa band corresponds to GST-NG2c, confirming that threonine is phosphorylated by PKC-␣ in the in vitro system. The 80-kDa band is PKC-␣ itself, autophosphorylated at Thr 631 within the kinase domain (40). Along with the data on NG2 phosphorylation in U251/NG2 cells, these in vitro results lend support to the idea that PKC-␣ mediates threonine phosphorylation in the NG2 cytoplasmic domain. The co-elution of phosphorylated PKC-␣ and GST-NG2c from glutathione-agarose beads corroborates our earlier observation that activated PKC-␣ directly binds to NG2.
Thr 2256 Is the Major PKC␣ Phosphorylation Site in NG2-To determine the site of NG2 threonine phosphorylation by PKC-␣, we prepared three additional GST fusion proteins containing NG2 cytoplasmic domains in which Thr 2256 , Thr 2265 , and Thr 2278 were mutated to glutamic acid residues. These mutant fusion proteins were used along with the wild type fusion protein GST-NG2c as substrates for in vitro [␥-32 P]ATP phosphorylation by recombinant PKC-␣. Tryptic phosphopeptide mapping reveals that phosphorylation of the T2256E mutant differs significantly from the other three species (Fig. 3). The T2256E map lacks the major phosphopeptide (spot 1) present in the other species. Persistence of a lesser phosphopeptide (spot 2) suggests that the cytoplasmic domain may contain a secondary atypical PKC-␣ phosphorylation site. Nevertheless, the dramatic reduction of phosphorylation in the T2256E mutant demonstrates that Thr 2256 is the major site for PKC-␣ phosphorylation of NG2.
As an additional means of evaluating Thr 2256 as the site of NG2 phosphorylation, we also performed phosphopeptide mapping of full-length NG2 before and after treatment of U251/ NG2 cells with PMA in the presence of 32 P i (Fig. 4). These phosphopeptide maps are more complex than those obtained with the isolated NG2 cytoplasmic domain (Fig. 3). This may be due in part to incomplete tryptic digestion of the full-length protein, leading to the occurrence of different sized fragments containing the same phosphorylated residue. For example, the immediate juxtaposition of the basic cytoplasmic sequence 2251 RKR 2253 to the hydrophobic transmembrane domain (residues 2226 -2250) might adversely affect trypsin cleavage at these sites (41). In the absence of PMA treatment several tryptic phosphopeptides are visible on the chromatogram of wild type NG2. PMA treatment leads to the appearance of two new phosphopeptides (1 and 2), indicative of PMA-dependent phosphorylation at a new site (or sites) in the proteoglycan. In contrast, PMA treatment of U251 cells expressing an NG2 variant in which the threonine at position 2256 is replaced by valine (U251/NG2T2256V) did not lead to appearance of these new phosphopeptides (arrows), consistent with the idea that Thr 2256 is the key site of threonine phosphorylation in NG2.
PMA-dependent Alteration of Cell Morphology and Redistribution of NG2-In untreated U251/NG2 cells, NG2 is associated with a meshwork of plasma membrane protrusions widely distributed over the apical cell surface (Fig. 5a). NG2 is highly co-localized in these protrusions with the ERM protein ezrin (Fig. 5b) and with ␣ 3 ␤ 1 integrin (data not shown). The localization of this meshwork of processes to the cell surface is confirmed by their accessibility to NG2 antibody when immunostaining is performed at 4°C on living cells (Fig. 5c). The presence of NG2 on cell surface membrane protrusions can be clearly seen in electron micrographs of immunogold-labeled  (1 and 2). These peptides were not present in material from PMA-treated U251/NG2-T2256V cells (arrows). U251/NG2 cells (Fig. 5d). Extensive arrays of phalloidinstained stress fibers (Fig. 5f) suggest that these cells with abundant NG2-positive membrane protrusions (Fig. 5e) are quiescent and firmly attached to the substratum.
In PMA-treated cells NG2 localization is shifted to prominent newly formed lamellipodia (Fig. 6a). If treated cells are returned to PMA-free culture conditions for 24 h, lamellipodia disappear and NG2-positive membrane protrusions reappear, demonstrating the reversibility of the process. In PMA-induced lamellipodia there is excellent co-localization of NG2 with ezrin ( Fig. 6b) and ␣ 3 ␤ 1 integrin (not shown). Immunostaining with a phosphospecific antibody (26) shows that phosphorylated ezrin (Thr 567 ) is present in the lamellipodia of PMA-treated cells, as well as in the membrane protrusions of untreated cells (not shown). Phalloidin staining also demonstrates abundant filamentous actin in PMA-induced lamellipodia and reveals significantly reduced levels of actin-containing stress fibers in cells with polarized morphologies (Fig. 6d). Cells in the transfected U251 population routinely lose NG2 expression over a period of several weeks in culture. These NG2-negative cells are much less likely to form lamellipodia in response to PMA treatment, often retaining their arrays of stress fibers under these conditions (Fig. 6, c and d). A quantitative assessment of PMA-induced lamellipodia formation was made using a culture in which 55% of the cells had lost NG2 expression. Extensive phalloidin-positive lamellipodia were observed in 79% of the NG2-positive cells. In contrast, only 12% of the NG2-negative cells exhibited lamellipodia, suggesting a functional role for NG2 during lamellipodia extension in response to PMA.
PMA-induced redistribution of NG2 to lamellipodia is blocked by the PKC inhibitor GF109203X (Fig. 6, e and f). The ability of this inhibitor to block PKC-stimulated phosphorylation of NG2 (Fig. 1B) suggests that phosphorylation of the proteoglycan could be a factor in its redistribution to lamellipodia. This possibility was further examined utilizing U251 cells expressing mutant versions of NG2 in which Thr 2256 , Thr 2265 , and Thr 2278 were replaced by valine residues. Under control conditions the localization of these three NG2 species to cell surface protrusions is indistinguishable from that seen with wild type NG2 (Fig. 7A, panels a, c, and e). The response of the T2265V and T2278V mutants to PMA treatment also resembles that of wild type U251/NG2 cells (Fig. 7A, panels b and f): in each case NG2 is redistributed to prominent lamellipodia. In contrast, T2256V transfectants are relatively resistant to the effects of PMA (Fig. 7A, panel d). Although some formation of NG2-positive lamellipodia occurs, NG2 remains associated with apical protrusions in most cells. These data are quantified in Fig. 7B by determining the percentage of cells in each population that display NG2 in lamellipodia as opposed to apical membrane protrusions. PMA treatment shifts NG2 expression from membrane protrusions to lamellipodia in each case except for the T2256V mutant, consistent with the idea that PKC-mediated phosphorylation of Thr 2256 is required for this molecular redistribution. Like wild type NG2, the T2256V species is co-localized with ezrin and ␣ 3 ␤ 1 integrin in membrane protrusions. Phosphorylation of ezrin appears to be unaffected in these sites, as judged by immunostaining with a phosphospecific antibody.
Because substitution of a negatively charged amino acid can mimic the conformational effects of phosphorylation (42), we transfected U251 cells with NG2 constructs in which glutamic acid residues replaced threonines 2256, 2265, and 2278. The T2265E and T2278E transfectants (Fig. 7A, panels i and j) proved to be indistinguishable from the valine mutants and the wild type transfectants. These mutant NG2 species were primarily localized to apical protrusions in untreated cells and were redistributed to lamellipodia following PMA treatment. In contrast, T2256E transfectants prominently displayed NG2 in lamellipodia both in the absence and presence of PMA (Fig. 7A,  panels g and h). Replacement of Thr 2256 with glutamic acid thus mimics the effects of NG2 phosphorylation. Ezrin and PMA-dependent Increases in Cell Motility-The polarized morphology, ruffling lamellipodial membranes, and lack of stress fibers seen in PMA-treated U251/NG2 cells are reminiscent of characteristics exhibited by motile cells. We used monolayer scratch assays to quantify PMA-dependent changes in cell motility. Panels a-c of Fig. 8 show, respectively, a fresh scratch (a) and the ability of U251/NG2 cells to migrate into the scratch under control (b) and PMA-stimulated (c) conditions. Although cell motility is apparent under control conditions, the enhanced motility induced by PMA treatment is quite dra-matic. When these cultures are double stained for NG2 and phalloidin, the presence of NG2-positive, phalloidin-positive lamellipodia in the motile cells is evident. Moreover, although the culture contains both NG2-positive (arrows in Fig. 8d) and NG2-negative cells (asterisks in Fig. 8e), the motile cells are almost invariably NG2-positive. NG2-negative cells remain predominantly within the monolayer and are infrequently seen in the scratch. With a U251/NG2 population in which 55% of the cells had lost NG2 expression, quantitation of migration revealed that 93% of the motile cells expressed NG2.
Motility assays were also conducted with the various mutant NG2 transfectants. Like U251/NG2 cells (Fig. 9A, a and b), the motility of U251 cells expressing the T2265V and T2278V mutants is potentiated by PMA treatment (Fig. 9A, panels e and f). In contrast, PMA treatment has little effect on the motility of cells expressing the T2256V mutant (Fig. 9A, c and d). Conversely, T2256E transfectants are spontaneously motile in the scratch assay, consistent with their morphology and localization of this NG2 species to lamellipodia (Fig. 9A, panels g and  h). As expected, the T2265E and T2278E transfectants do not behave differently from wild type transfectants or the T2265V and T2278V mutants (Fig. 9A, panels i and j). These results are quantified in Fig. 9B. We propose that these observations are explained by the inability of PMA-activated PKC-␣ to phosphorylate NG2 in the T2256V mutant and by the ability of glutamic acid substitution to mimic phosphorylation at Thr 2256 .

DISCUSSION
In vitro we have been able to demonstrate directly that PKC-␣ phosphorylates the NG2 cytoplasmic domain. In cell culture work we have used PMA as a means of stimulating PKC-␣-mediated phosphorylation of NG2. Due to its ability to activate both conventional and novel PKC isoforms (39), as well as other intracellular targets (43), PMA can exert widespread effects on protein phosphorylation and on cellular signaling processes. In future work it will therefore be important to investigate NG2 phosphorylation under more physiological circumstances (for example, growth factor stimulation, and engagement of extracellular matrix components) in which intracellular signaling pathways are activated in a less global fashion. Nevertheless, even against the complex background of PMA-induced effects, our results provide evidence that the phosphorylation of NG2 is a functionally significant event. Phosphorylation of NG2 at Thr 2256 is accompanied by redistribution of the proteoglycan on the cell surface, by polarization of the cell and several key molecular components, and by significant increases in cell motility. That NG2 phosphorylation is a cause, rather than an effect, of these processes is demonstrated by the behavior of the T2256V and T2256E transfectants. These mutations are incapable of affecting the full spectrum of PMA/PKC activities but can only affect NG2 and any molecules that serve as downstream targets of the proteoglycan. The inability of the T2256V transfectants to undergo molecular and morphological polarization (and ensuing increases in cell motility) in response to PMA treatment argues that phosphoryla-tion at Thr 2256 is at least a partial requirement in order for these processes to occur. This idea is supported by the observation that molecular and cellular polarization, along with increased motility, occur spontaneously in T2256E transfectants, where the negatively charged glutamic acid residue mimics threonine phosphorylation.
The localization of NG2 to different cell surface structures under different sets of conditions identifies NG2 as a potential regulator of membrane microdomains, a role also attributed to members of the syndecan family (2). In well spread, quiescent cells NG2 is localized to short membrane protrusions from the cell surface, reminiscent of NG2-rich microspikes found on the surface of human melanoma cells (29). Ezrin and PKC-␣ are co-localized within very similar processes in fibrosarcoma cells (26), consistent with our observation that NG2, ezrin, and ␣ 3 ␤ 1 integrin all appear to be present in membrane protrusions on the surface of U251/NG2 cells. It has been proposed that membrane protrusions of this sort serve as specialized scaffolds for presentation of molecules that mediate cell-cell and cell-matrix interactions (29). These protrusions appear to be distinct from retraction fibers, another type of microdomain that contains NG2 under certain conditions (25). Retraction fibers are associated with the trailing edges of some motile cells and can be significantly longer than apical membrane protrusions.
Phosphorylation at Thr 2256 results in translocation of NG2 to yet another type of membrane microdomain, the lamellipodia. The one or more mechanisms underlying redistribution of phosphorylated NG2 are not currently understood. We previously reported that PMA treatment of U251/NG2 cells results in proteolytic cleavage of NG2 (24), suggesting the possibility of a causal relationship between NG2 phosphorylation and NG2 cleavage. However, these two sets of events occur on very different time scales. PMA-induced phosphorylation of NG2 occurs within 20 -30 min, and redistribution of NG2 from membrane protrusions to lamellipodia can be observed within 1 h. In contrast, proteolytic generation of significant quantities of truncated NG2 is not detectable for 10-12 h after addition of PMA. The most plausible mechanism for PMA-induced cleavage of NG2 remains that PKC-dependent protein synthesis is required to generate the enzyme responsible for NG2 proteolysis (24).
An alternative explanation for the effect of NG2 phosphorylation on its membrane localization may be induction of a conformational change that alters the interaction of the proteoglycan with a cytoplasmic or extracellular ligand, resulting in a shift in its site of membrane targeting. Our results on colocalization of NG2 with ezrin and ␣ 3 ␤ 1 integrin, in conjunction with the report of an interaction between ezrin, PKC-␣, and ␤ 1 integrins (26), suggest the existence of a complex between NG2, ezrin, ␣ 3 ␤ 1 , and PKC-␣. We have recently established the ability of NG2 to interact with ␣ 3 ␤ 1 (18). Although an interaction between NG2 and ezrin has not been previously reported, the existence of such an interaction would support the apparent linkage of NG2 with the actin cytoskeleton (44). There are currently no data concerning the effect of NG2 phosphorylation on these putative interactions. However, because NG2 is colocalized with ezrin and ␣ 3 ␤ 1 in both membrane protrusions and lamellipodia, it is possible that phosphorylation does not affect interaction with either molecule. Instead, phosphorylation and dephosphorylation of NG2 may dictate its interaction with another yet-to-be-defined molecule. In this context it would be of interest to know the spatial relationship between NG2 and its putative PDZ scaffolding partners MUPP1 and GRIP1 (45,46) under control and PMA-stimulated conditions. This determination awaits development of MUPP1 and GRIP1 antibodies suitable for immunostaining.
Our studies with both the NG2 cytoplasmic domain and the full-length proteoglycan demonstrate a physical interaction between NG2 and PKC-␣. The proteoglycan may therefore participate in recruiting the kinase to key submembranous sites where its activity can influence signaling pathways that affect cell morphology and motility. Such a role has been previously proposed for the transmembrane proteoglycan syndecan-4 (47). Intriguingly, in addition to its role in phosphorylating NG2, PKC-␣ is also responsible for the phosphorylation of ezrin at Thr 567 (26). This phosphorylation leads to the dissociation of ezrin oligomers into monomers, a requirement for the ability of ezrin to effect changes in cell morphology and motility (48). Immunostaining with a phosphospecific (Thr 567 ) antibody reveals the presence of phosphorylated ezrin along with NG2 in both membrane protrusions and lamellipodia. Based on the behavior of the T2256V and T2256E transfectants, we suggest that phosphorylation of NG2 is an important trigger for cell polarization and translocation of the putative NG2/ezrin/␣ 3 ␤ 1 complex to ruffling lamellipodia. Because PKC-␣-mediated phosphorylation of ezrin still occurs in the membrane protrusions of T2256V transfectants, PKC-␣-mediated phosphorylation of NG2 would appear to be a key step needed to initiate the translocation process.
PKC-␣ has been implicated by a number of studies as a key activator of morphological change and cell motility (49 -51). The fact that some of these studies have been performed with NG2-negative cells makes it clear that targets other than NG2 play key roles in PKC-␣-induced cell migration. Among proteoglycans, for example, PKC-mediated phosphorylation of CD44 (1) and syndecans (52)(53)(54) plays essential roles in regulating changes in cell morphology. In the case of U251 cells, PMA is capable of inducing polarization and motility in the NG2-negative parental cells. In the absence of NG2 these PMA-induced effects on U251 cells are mediated by other PKC isoforms such as PKC-⑀ (55). Our current results demonstrate that expression of NG2 in U251 cells has the effect of potentiating the effects of PMA on molecular rearrangement, cell polarization, and cell motility via a mechanism that is dependent on PKC-␣-mediated phosphorylation of NG2 and redistribution of this proteoglycan on the cell surface. Further work will be aimed at elucidating additional details of the molecular mechanisms that underlie these phenomena.