Specific Serine-Proline Phosphorylation and Glycogen Synthase Kinase 3β-directed Subcellular Targeting of Stathmin 3/Sclip in Neurons*

Background: Stathmins are involved in the control of neuronal differentiation. Results: Stathmin 3 displays specific phosphorylation patterns by ERK2 and CDK5 and is a unique stathmin substrate for glycogen synthase kinase 3β mostly at neurite tips. Conclusion: Stathmin proteins can be specifically and locally regulated by proline-directed phosphorylation. Significance: Molecular and subcellular regulation of stathmins may contribute to the control of the development and plasticity of the nervous system. During nervous system development, neuronal growth, migration, and functional morphogenesis rely on the appropriate control of the subcellular cytoskeleton including microtubule dynamics. Stathmin family proteins play major roles during the various stages of neuronal differentiation, including axonal growth and branching, or dendritic development. We have shown previously that stathmins 2 (SCG10) and 3 (SCLIP) fulfill distinct, independent and complementary regulatory roles in axonal morphogenesis. Although the two proteins have been proposed to display the four conserved phosphorylation sites originally identified in stathmin 1, we show here that they possess distinct phosphorylation sites within their specific proline-rich domains (PRDs) that are differentially regulated by phosphorylation by proline-directed kinases involved in the control of neuronal differentiation. ERK2 or CDK5 phosphorylate the two proteins but with different site specificities. We also show for the first time that, unlike stathmin 2, stathmin 3 is a substrate for glycogen synthase kinase (GSK) 3β both in vitro and in vivo. Interestingly, stathmin 3 phosphorylated at its GSK-3β target site displays a specific subcellular localization at neuritic tips and within the actin-rich peripheral zone of the growth cone of differentiating hippocampal neurons in culture. Finally, pharmacological inhibition of GSK-3β induces a redistribution of stathmin 3, but not stathmin 2, from the periphery toward the Golgi region of neurons. Stathmin proteins can thus be either regulated locally or locally targeted by specific phosphorylation, each phosphoprotein of the stathmin family fulfilling distinct and specific roles in the control of neuronal differentiation.

During development and plasticity of the nervous system, neuronal growth, migration, and functional morphogenesis are controlled by numerous extracellular signals and the resulting activation of intracellular signaling pathways. Many of them target directly or indirectly the subcellular cytoskeleton, including microtubules whose dynamics are essential for many cellular processes such as neuritic growth, branching, and guidance.
Stathmins have been suggested to play a role in the control of diverse stages of neuronal differentiation, including neuronal polarity, axonal growth and branching, or dendritic development (2, 20, 26 -30). They are essentially expressed in the nervous system during embryonic and postnatal development with, however, at least partially distinct expression profiles for the various members of the family, suggesting distinct functions in the control of neural patterning and morphogenesis (2,5,6,31,32). Although stathmins 2 and 3 display identical expression and subcellular localization in hippocampal neurons in culture, RNAi inhibition of stathmin 2 or 3 expression results in growth cone expansion or increased axonal branching, respectively (26). Also, inhibition of stathmin 3 expression in Purkinje cells of cerebellar slices in culture impairs the formation of the dendritic tree (27), indicating that the major role of any given stathmin may depend on its specific, local, and temporal subcellular and cellular environment.
An important issue is thus deciphering specificities among the various members of the stathmin family, their roles, and the underlying molecular and regulatory features (2,13). Within their highly similar SLDs, stathmins possess a short divergent proline-rich domain (PRD) (6), linking their tubulin binding cap with their duplicate tubulin-binding regions (16,33). In stathmins 1-3, this region is flanked by two proline-serine phosphorylation sites that have been shown for stathmins 1 and 2 to be targets for proline-directed kinases such as ERKs and CDKs (7,34). The phosphorylation of stathmin 3 has, however, not been previously characterized. We hypothesized that functional differences between stathmins 2 and 3 might be due in part to differences in their regulation by phosphorylation (20,26). We reveal here at least one additional phosphorylation site within the PRD of stathmin 3 and show that stathmins 2 and 3 are phosphorylated differently by proline-directed kinases involved in the control of neuronal differentiation. ERK2 and CDK5 display different site specificities on the two proteins. Furthermore, we show for the first time that stathmin 3, unlike stathmin 2, is a substrate for GSK-3␤, a proline-directed kinase involved in the control of neuronal morphogenesis (35)(36)(37)(38)(39), which phosphorylates a serine residue within its PRD. Finally, we reveal a specific subcellular localization of GSK-3␤-phosphorylated stathmin 3 at neuritic tips and within the actin-rich peripheral zone of the axonal growth cone of differentiating hippocampal neurons in culture. Altogether, our results provide molecular clues to the specific regulation and regulatory roles of stathmin 3, and substantiate our hypothesis that each phosphoprotein of the stathmin family fulfils distinct and specific roles in the control of neuronal differentiation.

EXPERIMENTAL PROCEDURES
Plasmid Constructs and Mutations-For bacterial expression, SLD of stathmin 3 (WT and mutants) and stathmin 2 were expressed in the pET-8C vector as previously described (13).
For eukaryote expression, St3-myc and mutants were expressed in the pcDNA3-myc vector as previously described (40). Mutations to alanines of serines 60, 68, or 73 of St3-SLD were generated by site-directed mutagenesis using the QuikChange mutagenesis kit (Stratagene) with pairs of complementary mutagenic primers. All constructs were confirmed by DNA sequencing.
Protein Expression and Purification-Recombinant SLD expressions were performed as previously described (13). Briefly, plasmids encoding the respective SLDs were used to transform the Escherichia coli BL-21 (Stratagene). An overnight preculture, grown at 37°C, was used to seed 100 ml of LB medium. At exponential phase, recombinant protein expression was induced for 3 h by adding 0.4 mM isopropyl thio-␤-Dgalactoside. Cells were then pelleted by centrifugation, resuspended in 20 mM Tris-HCl, 1 mM EGTA, pH 8.0, containing Complete antiprotease mixture (Roche Diagnostics) and heated to 100°C for 15 min. The pellets were sonicated three times for 1 min on ice before ultracentrifugation at 100,000 ϫ g (Optima MAX ultracentrifuge, rotor TLA 100.2, Beckman Instruments) for 6 min at 4°C to yield the S2 supernatants and P2 pellets. S2 supernatants were loaded on a Q-Sepharose FF anion exchange column (GE Healthcare) and eluted with a 0 -200 mM NaCl linear gradient in 20 mM Tris-HCl, 1 mM EGTA, pH 8.0. The eluted SLD positive fractions were pooled and concentrated with Centriprep 10 (Millipore). Protein concentrations were determined by optical density at 206 nm.
HeLa Cell Culture and Transfection-HeLa cells were grown in DMEM (Dulbecco's modified Eagle's medium)-Glutamax with 10% decomplemented fetal calf serum and 1% penicillin/ streptomycin (Invitrogen) at 37°C in a humidified atmosphere containing 5% CO 2 . Twice a week, cells were dissociated in 0.25% trypsin and 1 mM EDTA (Invitrogen), and then diluted 1:10 before replating. For transfection, HeLa cells were seeded at 12.5 ϫ 10 3 cells/cm 2 in 35-mm dishes. They were transfected 16 h after plating, using the liposoluble reagent Lipofectamine TM 2000 (Invitrogen), according to the manufacturer's instructions: 3 l of reagent and 1 g of plasmid DNA were used for one 35-mm dish.
Primary Hippocampal Neurons-Primary cultures of hippocampal neurons from E18 Sprague-Dawley or Wistar rats were prepared as described (41). Briefly, hippocampi were dissociated by trypsinization and trituration and plated for immunofluorescence on poly-L-lysine-coated glass coverslips at 2.4 ϫ 10 3 cells/cm 2 or for biochemical analysis in poly-L-lysinecoated 35-mm culture dishes at 3 ϫ 10 5 cells per dish. After cells were allowed to attach for 3 h in attachment medium (10% horse serum, 2 mM L-glutamine, 1 mM sodium pyruvate, minimal essential medium (Invitrogen)), the medium was changed for Neurobasal TM (Invitrogen) supplemented with 2% B27 (Invitrogen) and 2 mM L-glutamine (Invitrogen).
Transfected HeLa cells cultured in 35-mm dishes were washed with cold PBS and lysed in 50 l of HeLa lysis buffer (50 mM Tris-HCl, pH 7.5, 125 mM NaCl, 10% glycerol, 1% Triton X-100) containing Complete protease inhibitor (Roche Diagnostics) and phosphatase inhibitor (Thermo Scientific) mixtures. The rest of the protocol was identical to that for hippocampal neurons except that only 1 ⁄ 2 of proteins per culture dish were loaded for polyacrylamide gel electrophoresis.
Immunofluorescence Microscopy and Image Analysis-Images were acquired with a ϫ40 objective mounted on a LEICA DM 6000 microscope (Leica Microsystems), equipped with a Micromax camera (Roper Scientific) controlled with Metamorph (Universal Imaging) or a LEICA SP2 confocal laser scanning microscope and a ϫ63 objective (Leica microsystems).
Neurons showing a healthy morphology based on tubulin labeling were chosen randomly on each slide. All acquisition features were kept constant for the whole study. Image names were randomized for blind analysis. Stathmin 2 and stathmin 3 labeling intensities were quantified in the Golgi and the rest of the cytoplasm automatically, using Metamorph software and a home written journal. Tubulin and Golgi (CTR 433)-labeled images were segmented to obtain the cytoplasm and Golgi areas. The cytoplasm area with the exclusion of the Golgi was obtained by logical operations on the above defined areas. Stathmin 2 and 3 average intensities in the Golgi and cytoplasm excluding the Golgi were computed for each cell, from which the ratio between the average intensity in the Golgi and the rest of the cytoplasm was deduced. Quantification was made on 20 cells in three different experiments. Histograms presented in Fig. 7 correspond to the mean of this ratio for one representative experiment. Analysis was performed using Metamorph, Microsoft Excel, and GraphPad Prism5. All statistical comparisons of immunofluorescence intensities were made using the Mann-Whitney test. Images were prepared for printing using Adobe Photoshop. hippocampal neurons in culture, RNAi invalidation revealed very different and specific roles for stathmin 2 and 3 in axon morphogenesis (26). Although they have been proposed to display the characteristic four conserved phosphorylation sites originally identified in stathmin 1 (5-7, 34), we reasoned that differences in their roles and regulation might result from differences in their phosphorylation. We therefore re-examined their amino acid sequences with an emphasis on differences instead of identities ( Fig. 1 and supplemental Fig. S1). The overall highly conserved stathminlike SLD domain of all four stathmins in vertebrates includes a short diverging proline-rich stretch of amino acids (PRD) (2, 4 -6) (Fig. 1B). Indeed, although the PRD of each stathmin is highly conserved through evolution in vertebrates (supplemental Fig. S1), its sequence is characteristic of the corresponding member of the stathmin family. In stathmin 1, this region (residues 25-40) is bordered by two serineproline phosphorylation sites (serines 25 and 38) that are substrates for ERK2 and CDK kinases, two kinase families known to be involved in the control of neural differentiation (43,44), with a preference of serine 25 for ERKs and 38 for CDKs (7,9). The PRD of stathmin 2 (residues 59 -74) harbors corresponding serine-proline sites at positions 62 and 73 that are also substrates for the same kinases, with a more exclusive specificity of CDKs for serine 73 (34). The phosphorylation of stathmin 3 had not been characterized so far. Its PRD (residues 59 -75) also displays corresponding serineprolines at flanking positions 60 and 73, but also two addi- B. tional ones, at positions 65 and 68 (Fig. 1C), which may be additional phosphorylation sites for regulating the functions of stathmin 3. Specific Phosphorylation of Stathmin 3 by CDK5, ERK2, and GSK-3␤ in Vitro-We therefore decided to examine and characterize the phosphorylation of stathmin 3 by three classical serine-proline-directed protein kinases involved in signaling in the nervous system, namely ERK2, CDK5, a neuronal CDK previously shown to phosphorylate stathmin 1 (45), and GSK-3␤, a serine-proline directed kinase involved in axon formation and branching (35)(36)(37)(38)(39) and neurogenesis (37,46), and not previously reported to phosphorylate stathmin proteins.
We first compared the phosphorylation of recombinant SLDs of stathmins 2 and 3 in vitro in the presence of ␥-[ 32 P]ATP and the respective kinases ( Fig. 2A). As revealed by SDS-PAGE analysis of [ 32 P]phosphate incorporated in each of the proteins, ERK2 and CDK5 promoted the phosphorylation of the SLDs of both stathmins 2 and 3 to similar levels, although ERK2 was somewhat more efficient to phosphorylate stathmin 2. Remarkably, stathmin 3 appeared to be a substrate also for GSK-3␤, whereas this kinase did not promote any detectable phosphate incorporation into stathmin 2.
To identify the stathmin 3 site(s) phosphorylated by each of the three proline-directed kinases tested, we generated St3-SLD single alanine mutants of the two flanking serines 60 and 73, and of the two specific stathmin 3 PRD serines 65 and 68 (Fig. 1D), and probed them for in vitro phosphorylation kinetics (0 -120 min) in comparison with the wild type (WT) protein (Fig. 2, B and C). Alanine mutation of serine 60 did not inhibit phosphorylation of St3-SLD by CDK5 or ERK2, but it did inhibit phosphorylation by GSK-3␤ by at least 70%. On the other hand, mutation of serine 73 did not inhibit phosphorylation by ERK2 but did inhibit phosphorylation by CDK5 by about 50%. Among the two additional serine-proline sites in stathmin 3, mutation of serine 65 did not inhibit phosphorylation by any of the three kinases tested (not shown), whereas mutation of serine 68 inhibited phosphorylation of St3-SLD by both CDK5 (ϳ40% inhibition) and ERK2 (ϳ80% inhibition). Altogether, these results indicate that CDK5 phosphorylates similarly serines 68 and 73, whereas ERK2 targets mostly serine 68 and GSK-3␤ mostly serine 60. These results show that phosphorylation of stathmin 3 by the three proline-directed kinases tested is indeed specific, differing from sites predicted by homology with the previously described phosphorylation sites of stathmins 1 and 2 (serine 73 for CDKs and serines 60 and to a lesser extent 73 for ERK2). Altogether, our in vitro phosphorylation results unravel novel kinase specificities and phosphorylation sites unique to stathmin 3 among stathmin proteins, and likely contributing to its specific biological regulation and properties.
Antiphosphostathmin 3 Antibodies Reveal in Vitro and in Vivo Phosphorylation of Stathmin 3 at Serines 60 and 73-To further characterize the phosphorylation of stathmin 3 in vitro and in vivo, we immunized rabbits with peptides including the phosphorylated forms of the previously characterized phosphorylation sites of stathmin 3 at serines 60, 68, and 73. Whereas the ELISA titer of the antiserum directed to phosphoserine 68 was unfortunately too low, we obtained potent and specific purified polyclonal antibodies directed to phosphoserines 60 (anti-St3-60P) and 73 (anti-St3-73P), as revealed by ELISA and cross-specific phosphopeptide dot blots (not shown). To further characterize their specificity for the phosphorylated residues within the entire SLD of stathmin 3, we expressed various myc-tagged St3-SLD forms, either wild type (WT) or mutated to alanine in positions 60 (St3/60A) or 73 (St3/73A) in HeLa cells, to allow their phosphorylation by endogenous kinases. The corresponding cell extracts were analyzed by Western blotting with monoclonal anti-myc for total transfected protein expression, and with purified anti-St3-60P and anti-St3-73P antibodies to evaluate their phosphorylation on the corresponding sites (Fig. 3A). As expected, whereas all three proteins were expressed to similar levels (anti-myc labeling), anti-St3-60P recognized the WT protein and St3/73A, but not St3/60A. Conversely, anti-St3-73P recognized the WT protein and St3/60A but not St3/73A. The two antibodies are hence efficient to recognize the whole protein in its phosphorylated states, and fully specific for their target phosphosites.
The specific antibodies were then used to identify directly the sites phosphorylated by the three proline-directed kinases, whose target sites were inferred from recombinant mutation in our in vitro phosphorylation studies. Recombinant St3-SLD was phosphorylated in vitro in the presence of GSK-3␤, CDK5, or ERK2 as above but with only trace amounts of radioactive ␥-[ 32 P]ATP to control for actual phosphate incorporation. The control and phosphorylated proteins were then analyzed by Western blot with anti-St3, anti-St3-60P, and anti-St3-73P purified antibodies (Fig. 3B). As a further control for the phosphospecificity of our novel antibodies, unphosphorylated stathmin 3 was as expected not recognized by the phosphostathmin 3-directed antibodies. Anti-St3-60P recognized St3-SLD phosphorylated in the presence of GSK-3␤ and, to a lesser extent, ERK2, whereas anti-St3-73P recognized only the protein phosphorylated by CDK5. These results, in agreement with the in vitro phosphorylation assay with the various serine to alanine mutated St3-SLDs shown in Fig. 2, strongly suggest that GSK-3␤ phosphorylates stathmin 3 essentially at serine 60, and that this same serine could also be phosphorylated likely as a secondary site by ERK2. In contrast, CDK5 phosphorylates stathmin 3 at serine 73, which might also be poorly phosphorylated by GSK-3␤.
Taking advantage of the good titer and specificity of the antiphosphostathmin 3 antibodies, we examined endogenous phosphorylation of stathmin 3 in neurons. Western blots of extracts from hippocampal neurons in culture reveal that both anti-St3-60P and anti-St3-73P recognize a band at the level of the endogenous stathmin 3 protein revealed by the anti-stathmin 3 antibody (Fig. 3C). This result is a good indication that stathmin 3 phosphorylation on serines 60 and 73 occurs in vivo.
Stathmin 3 Phosphorylated on Serine 60 or 73 Displays Divergent Subcellular Localization during Neuronal Differentiation-We have shown previously that stathmin 3 plays essential roles in axonal morphogenesis of embryonic rat hippocampal neurons in culture (26), a powerful and widely used model for studying neuronal differentiation (47). As a member of the stathmin phosphoprotein family, stathmin 3 was predicted to be regulated by phosphorylation (5), and the detection of stath-  Their relative phosphorylation by each kinase is revealed by comparing the ratio of the 32 P over the Coomassie Blue signals, the resulting ratio being arbitrarily set to 100% for stathmin 3 (right panel), revealing comparable phosphorylation of stathmins 2 and 3 by CDK5 and ERK2, as well as the unique phosphorylation of stathmin 3 by GSK-3␤. B, wild type (WT) stathmin 3 SLD recombinant protein and its alanine mutated forms (see Fig. 1) at the indicated serine residues were phosphorylated by CDK5, ERK2, or GSK-3␤ for the indicated times. Reaction products were analyzed by SDS-PAGE followed by Coomassie Blue staining and phosphorimaging for 32 P incorporation. C, the ratio of the 32 P incorporated at 60 min over the Coomassie Blue signal from at least three independent experiments such as in B was quantified, the resulting value being arbitrarily set at 100% for the wild type protein.
Mean Ϯ S.E. min 3 phosphorylated on serines 60 and 73 in hippocampal neurons in culture (Fig. 3) indicates that this is actually the case in vivo. We therefore examined closely the subcellular localization of stathmin 3 phosphoforms in hippocampal neurons during their differentiation in culture.
As we have shown previously (26), stathmin 3 is essentially localized at the Golgi and vesicle-like structures throughout the cell body and along neurites, and presents a specific accumulation at the axonal growth cone of differentiating neurons (Fig.  4), as visualized by immunocytochemistry with a polyclonal antibody recognizing all, nonphosphorylated and phosphorylated forms of stathmin 3. In contrast to total stathmin 3 label-ing, anti-phosphostathmin 3 antibodies revealed distinct subcellular localizations of stathmin 3 phosphorylated on serines 60 and 73, respectively. Both anti-St3-63P and anti-St3-73P revealed a diffuse labeling in the cell body, not restricted to the Golgi and often including the nuclear region, mostly in the case of St3-73P. In the growing process shafts, the two antibodies also revealed a diffuse, less punctuate and vesicle-like staining than the total stathmin 3 labeling. This diffuse and more extended labeling with the anti-phosphostathmin 3 antibodies suggests that the corresponding phosphoforms may be partially soluble and have specific local functions within the neuron. The low level of soluble labeling with the anti-stathmin 3 antibody, which also recognizes the phosphoforms, suggests that the latter, although strongly labeled by the corresponding specific antiphosphostathmin antibodies, represent a minor proportion of total stathmin 3. Interestingly, the immunostaining within the neurite tips and growth cones strongly diverged between the two phosphoforms: only anti-St3-60P displayed a labeling at neurite tips and within the peripheral, actin-rich domain of the growth cone (Figs. 4). Moreover, anti-St3-60P seemed to stain more strongly the growth cone than other neurite tips. The similar labeling with the anti-stathmin 3 and anti-St3-60P antibodies in growing lamellipodia suggests that a significant proportion of stathmin 3 is phosphorylated on serine 60 in this cellular region. In growth cone peripheral regions the anti-St3-60P labeling seems relatively more intense than the anti-stathmin 3 one, which suggests that stathmin 3 in these regions is mostly phosphorylated on serine 60. Altogether, these observations reveal for the first time distinct subcellular localizations of stathmin 3 depending on its phosphorylation at different sites, which might reflect either its specific phosphorylation at corresponding locations within the cell or relocalization of the protein according to its phosphorylation state, most likely in relationship with its functional regulation. A, extracts from HeLa cells transfected with various stathmin 3 constructs (St3-myc, St3/60A-myc, and St3/73A-myc) were analyzed by Western blot using anti-myc for total and the indicated stathmin 3 phosphospecific antibodies, demonstrating their actual specificities for the corresponding phosphorylation sites. B, recombinant WT St3-SLD protein was phosphorylated in vitro by GSK-3␤, CDK5, or ERK2, respectively, using cold (upper panels) or radioactive ATP (bottom panel, [ 32 P]) in parallel experiments, and run on SDS-PAGE. The nonradioactive samples were analyzed by Western blot with the indicated antibodies for total (St3) and the indicated stathmin 3 phosphospecific forms, allowing the identification of the sites phosphorylated by the various kinases. The radioactive reaction products were analyzed by phosphorimaging to ascertain phosphate incorporation with each kinase tested. C, total extracts of hippocampal neurons in culture were analyzed by Western blot for total and phosphorylated stathmin 3 using the indicated antibodies, demonstrating the endogenous neuronal phosphorylation of stathmin 3 on serines 60 and 73.

Stathmin 3 Phosphorylated on Serine 60 Is Localized Partially within Actin-rich Regions of Growth Cones, Neurite Tips, and
Branch Sites-The growth cone can be divided into three regions: the central domain containing splayed microtubules, the transition domain where actin filaments anchor into a loose MT network, and the peripheral domain containing a dense meshwork of actin and few exploratory microtubules (48). Because we noticed differences in St3-60P versus total stathmin 3 immunolabeling between the peripheral and central domains of the growth cone (Fig. 4B), we decided to examine more closely the localization of St3-60P in the growth cone in relationship to the cytoskeleton. We performed triple labeling of stathmin 3 or St3-60P with both actin and tubulin on hippocampal neurons after 3 days in culture (Fig. 5), a stage when growth cones are fully mature and active. As previously described (26), confocal analysis revealed that stathmin 3 is abundant in the growth cone central domain, where it partially follows the dense dynamic microtubule network, whereas it does not appear significantly colocalized with peripheral actin except along some exploratory microtubules penetrating the peripheral region (Fig. 5, A and B). Immunoreactivity of its St3-60P phosphoform is also present in the growth cone central domain, possibly at a relatively lower density than total stathmin 3. However, unlike total stathmin 3, St3-60P immunoreactivity also labels the peripheral actin-rich domain and its filopodia, as ascertained by the yellow/orange color in the corresponding merged images of Fig. 5, A and B. It is also strongly detected along actin-rich filopodial extensions stemming from the distal axon shaft (Fig. 5, A and B), as well as in other neurite tips and likely neuritic branch sites that appear also almost devoid of microtubules (Fig. 5C). The lack of visible labeling for total stathmin 3 in the actin-rich extending regions reveals that, although it represents a minor fraction of total stathmin 3 in neurons, the majority of stathmin 3 present in these regions is phosphorylated on serine 60.

GSK-3␤ Inhibitors Specifically Inhibit Phosphorylation of Stathmin 3 Serine 60 in Neurons in Vivo-
We have shown that stathmin 3 can be phosphorylated in vitro on serine 60 by GSK-3␤ and that, unlike other phosphorylated forms of stathmin 3 such as St3-73P, St3-60P has a specific and particular distribution within the growth cone of differentiating hippocampal neurons in culture. To assess whether stathmin 3 is phosphorylated by GSK-3␤ in vivo, we examined the effect of pharmacological inhibition of the kinase in neurons in culture with two distinct GSK-3␤-specific inhibitors: AR-A014418 (49) and inhibitor XV (50). We checked that AR-A014418 efficiently inhibited GSK-3␤-mediated 32 P incorporation into recombinant St3-SLD in the nanomolar range in vitro (Fig. 6A). To test the efficacy of the two inhibitors on GSK-3␤ activity in vivo, we monitored the amount of phosphorylation of Tau, a well known GSK-3␤ substrate (51), by immunoblotting treated cell extracts with a specific antibody directed to Tau phosphorylated at serine 199. As expected, the phosphorylation of Tau decreased upon pharmacological treatment with the GSK-3␤ inhibitors, dose dependently in the 10 M range in vivo (Fig. 6B). Interestingly, both AR-A014418 and inhibitor XV treatments resulted, at the same concentrations, in a decrease also of stathmin 3 phosphorylated on serine 60, as revealed by immunoblotting of the same extracts with the purified anti-St3-60P antibody (Fig.  6C). On the other hand, no significant reduction of St3-73P was observed (Fig. 6D). The in vivo phosphorylation of the in vitro identified GSK-3␤ phosphorylation site of stathmin 3 being inhibited by well characterized GSK-3␤ pharmacological inhibitors, it is likely that stathmin 3 is indeed a substrate for GSK-3␤ in live neurons.
Inhibition of GSK-3␤ Induces Stathmin 3 Relocalization in Hippocampal Neurons in Culture-The specific subcellular localization of St3-60P in neurons in culture prompted us to ask whether the overall subcellular localization of stathmin 3 is dependent on its phosphorylation level by GSK-3␤. To answer this question, we treated hippocampal neurons in culture with the GSK-3␤ inhibitor AR-A014418. We did observe, as it has been described (37,46,52), that GSK-3␤ inhibition led to abnormal neuronal morphology, with neurons appearing more ramified and less differentiated in the presence of the inhibitor at 20 M (data not shown). Interestingly, the treated cells seemed to also display a modified subcellular stathmin 3 immunoreactivity distribution, in particular as compared with stathmin 2, whose distribution did not seem to be affected. Indeed, in control neurons, both stathmin 2 and 3 were immunodetected as expected at the Golgi, along the neurite shafts, and in neurite tips and growth cones. In contrast, in GSK-3␤-inhibited neurons, whereas no significant difference was apparent in stathmin 2 distribution, stathmin 3 appeared somewhat more diffuse and was mainly detected at the Golgi, with a clear decrease and lack of punctuate staining in neurite tips and the growth cone (Fig. 7A). To quantify this observation, we performed triple labeling of rat hippocampal neurons with anti-tyrosinated tubulin, anti-stathmin 2 or 3, and anti-Golgi (CTR-433) antibodies. The Golgi and tubulin labelings were used to define "masks" covering either the Golgi area of each cell considered or the entire cell minus the Golgi, respectively. These two masks were used to quantify the mean intensity of stathmin 2 or 3 labeling in the two corresponding areas of each cell. The ratio between the intensity per surface area in the Golgi and the rest of the cells was calculated for at least 20 cells in each condition in three separate experiments (Fig. 7B). Quantifications reveal that there was indeed a significant increase of this ratio for stathmin 3 in the treated cells compared with control cells. This effect is dose-dependent and is not observed for stathmin 2. Stathmin 3 localization is thus specifically more restricted to the Golgi in neurons where GSK-3␤ is inhibited. This observation may support the hypothesis of a specific localization of stathmin 3 depending on its phosphorylation by GSK-3␤, although it may also more indirectly result from the pleiotropic actions of GSK-3␤ during neuronal differentiation.

DISCUSSION
During neuronal differentiation, growth cone progression, neurite elongation, and axonal branching rely on essential cytoskeleton rearrangements. Stathmin family phosphoproteins are involved in the regulation of such developmental processes as they are microtubule regulating proteins highly expressed in the nervous system (2,5,13,53). Although similar in many respects, stathmin 2 (SCG10) and stathmin 3 (SCLIP) fulfill distinct, independent and complementary regulatory roles in axonal morphogenesis (26). We show here for the first time that, in contrast to the other stathmins, stathmin 3 is a specific substrate for GSK-3␤, a proline-dependent protein kinase playing prominent roles in neuronal differentiation. We further show that stathmins 2 and 3 possess distinct phosphorylation sites differentially regulated by proline-directed kinases within their specific proline-rich domains. Finally, we demonstrate the specific localization of stathmin 3 in actin-rich regions at neurite tips when phosphorylated at serine 60 likely by GSK-3␤. Stathmin proteins can thus be either regulated locally or locally targeted by specific phosphorylation, which likely participates in their specific roles in neuronal differentiation.

Stathmin 3 Possesses Specific Phosphorylation Sites within Its PRD Region and Is a Target for GSK-3␤ Both in Vitro and in
Vivo-Stathmins display a high degree of sequence identity within their SLDs, with the exception of their 14 -16-residue proline-rich PRD region (2,6). Interestingly, stathmins 2 and 3 did not display homologous patterns of phosphorylation sites for the two classical proline-directed kinases ERK2 and CDK5, both in vitro and in transfected cells in culture. In particular, serine 68 of stathmin 3 is a target for ERK2 instead of serine 60 as would be expected by comparison with serine 62 of stathmin 2, whereas serine 60 is the main site for GSK-3␤, which does not phosphorylate stathmin 2.
It is interesting to note that the diverging PRD sequence was considered so far as a mere unstructured linker between the two well defined tubulin-interacting regions of stathmins as revealed by x-ray crystallography (16). However, as the PRD region is conserved through evolution within each member of the stathmin family and includes specific proline-serine phosphorylation sites, it is likely that it also fulfills specific and functional regulatory roles. One possibility is that the resulting regulation could apply indirectly on the tubulin interacting properties of the given stathmins. However, phosphorylation of the proline-dependent sites within the PRD of stathmin 1 is not very effective at inhibiting its interaction with tubulin (54). It is thus more likely that phosphorylation within their PRD regulates some other yet unidentified properties of stathmins and/or their interaction with other cellular partners.
The specific phosphorylation of stathmin 3 by GSK-3␤ is of particular interest in the context of neuronal differentiation, as this proline-directed kinase is known to play a role in the control of axon formation and morphogenesis (reviewed in Refs. [35][36][37][38][39]46), and to interact with microtubule regulating proteins such as CRMP2 (36), MAP2 (55), MAP1B (56), or Tau (57). The main phosphorylation target of GSK-3␤ on stathmin 3 is serine 60. This phosphorylation site might be primed by the possible phosphorylation site at position 65, 5 residues downstream as described for some GSK-3␤ substrates (58), or it may belong to the nonprimed category of GSK-3␤ sites such as, for example, in MAP1b (59).  As we were able to show that stathmin 3 is indeed phosphorylated on serine 60 and that this phosphorylation is reduced by GSK-3␤ inhibitors in differentiating neurons in culture, it is very likely that stathmin 3 is a direct substrate for this kinase in vivo. GSK-3␤ is also known to be involved in the phosphorylation of Tau and APP, and to be overactivated in brains of Alzheimer patients (60). It would thus be of interest to examine if the phosphorylation of stathmin 3 at serine 60 is affected in this pathology.
Stathmin 3 Phosphorylated at Serine 60 Is Present at the Growth Cone-We have shown that stathmin 3 phosphorylated at serine 60 (St3-60P) but not stathmin 3 phosphorylated at serine 73 (St3-73P) is located at neurite tips and in particular within the growth cone. It is interesting to note that, by contrast, phosphorylated stathmin 2 is absent from the growth cone of retinal ganglion neurons (61). Moreover, inhibition of GSK-3␤, which reduces the phosphorylation level of stathmin 3 at serine 60, induced a relocalization of stathmin 3 toward the cell body in hippocampal neurons in culture. These results suggest that serine 60 phosphorylation of stathmin 3 is important for its specific localization at the leading edge of the cell. Interestingly, a similar DOCK7-mediated phosphorylation-related subcellular localization along the axon was previously described for stathmin 1: in early stage hippocampal neurons, local phosphorylation of stathmin 1 at serine 16 is higher in the nearly formed axon than in the future dendrites, and increases toward the distal edge of growing axons (28). Other microtubule-associated proteins have been shown to have their spatiotemporal localization regulated by phosphorylation, such as for example, doublecortin (DCX) which, like stathmin, possesses a tubulin-binding site and a proline-rich domain (62), or MAP1B whose phosphorylation on nonprimed GSK-3␤ sites is restricted to growing axons (59).
How St3-60P is selectively localized at neurite tips is presently unknown. One possibility is that stathmin 3 is specifically phosphorylated locally at the growth cone. The overall redistribution of stathmin 3 protein toward the soma upon GSK-3 inhibition could reveal that, conversely, its phosphorylation by GSK-3␤ would result in the migration of additional stathmin 3 toward neurites and their tips. This migration could either provide more substrate for the kinase or compensate for the unphosphorylated protein that has been phosphorylated locally. Alternatively, stathmin 3 may be partially targeted to the leading edge of neurites upon phosphorylation at serine 60. Such a process has been recently shown for stathmin 1 through its association to a specific kinesin complex when phosphorylated at serine 38 (63,64).

Stathmin 3 Phosphorylated at Serine 60 Colocalizes with Actin in the Growth Cone Peripheral Domain and Axonal
Filopodia-We observed previously the presence of stathmins 2 and 3 along dynamic exploratory microtubules within the actin-rich peripheral zone of the growth cone (26). We show here that stathmin 3 mostly in its serine 60-phosphorylated form colocalizes with the actin-rich structures at neurite extremities, as well as in filopodial extensions along neurites. Although a direct interaction of stathmin 3 with actin has not been demonstrated so far, stathmin 3 could be another protein, regulated by phosphorylation, contributing to the interactions between the actin and tubulin networks.
Indeed, neurite outgrowth necessitates coordinated events between the two major cytoskeletal components actin and tubulin (65). This integration could be mediated by molecules such as Pod-1, which contains molecular microtubule and actin-binding domains (66). However, there are also few examples of molecular complexes that can bridge between these cytoskeletal components, such as for example, the Spn-DCX interaction (67), DCX being known to interact with microtubules (68), whereas Spn is able to cross-link F-actin into bundles (69). These two proteins display maximal overlap in the proximal part of the growing neurite tip. Interestingly, DCX colocalization with actin is mediated by phosphorylation by JNK. Under normal activity, DCX invades the actin-rich region of the growth cone, but this invasion decreases when JNK activity is inhibited (62).
In the case of stathmin 3, it could either interact directly with actin, for example, through its PRD region, which does not interact with tubulin, or through another intermediate partner to be identified. It is interesting to note that, although not formally recognized as a stathmin partner at the time, actin from a brain homogenate was retained, like tubulin, on a stathmin affinity column (70). A direct or indirect interaction with actin dynamics could also explain the previously described involvement of stathmin 3 in the control of axonal branching (26), which takes place at the growth cone and along the shaft when microtubules splay apart, allowing shorter microtubules to invade the nascent actin-rich branches (71,72).
In conclusion, our results provide new insight into the molecular bases and pathways allowing distinct and specific functions and regulation of phosphoproteins of the stathmin family. We showed for the first time that stathmins 2 and 3 are regulated differently by phosphorylation, that stathmin 3 is a unique stathmin substrate for GSK-3␤, which phosphorylates it on a site related to its subcellular localization. Such specific, potential local phosphorylation in response to various extracellular or intracellular cues may direct the specific, diverse functions of various members of the stathmin family in neurons. Moreover, phosphorylation of stathmin 3 could contribute also to the integration between the actin and tubulin networks at the dynamic edge of growing neurites. Elucidating these pathways and understanding the molecular mechanisms relating phosphorylation and subcellular localization of stathmins will be of high interest to better understand the process involved in the development and maturation of the nervous system.