Stathmin Family Proteins Display Specific Molecular and Tubulin Binding Properties*

Stathmin family phosphoproteins (stathmin, SCG10, SCLIP, and RB3/RB3 * /RB3 ( ) are involved in signal trans-duction and regulation of microtubule dynamics. With the exception of stathmin, they are expressed exclusively in the nervous system, where they display different spatio-temporal and functional regulations and hence play at least partially distinct and possibly complementary roles in relation to the control of development, plasticity, and neuronal activities. At the molecular level, each possesses a specific “stathmin-like domain” and, with the exception of stathmin, various combinations of N-terminal extensions involved in their association with intracellular membrane compartments. We show here that each stathmin-like domain also displays specific biochemical and tubulin interaction properties. They are all able to sequester two a / b tubulin heterodimers as revealed by their inhibitory action on tubulin polymerization and by gel filtration. However, they differ in the stabilities of the complexes formed as well as in their interaction kinetics with tubulin followed by surface plasmon resonance as follows: weak

various cell types and populations of the nervous system, and possessing a stathmin-like domain (SLD) 1 with various Nterminal extensions (3)(4)(5)(6). Their involvement in signal transduction and regulation of microtubule dynamics, in relation with their different spatio-temporal expression patterns, suggests that stathmin family proteins may play related but distinct and likely complementary roles in the regulation of differentiation, activities, and plasticity of the nervous system.
During rat development, expression of stathmin family proteins is highest from late embryogenesis until a week after birth (3,7). They are likely involved in neural differentiation: SCG10 was shown to be induced in neural crest cells when they differentiate into sympathetic neurons (7); in the PC12 cell model of neuron-like differentiation, SCG10, SCLIP, and to a lesser extent RB3Љ are induced in response to nerve growth factor (3,7,8), and overexpression of SCG10 potentiates neurite extension (9).
All proteins of the stathmin family are expressed in the adult brain, some of them at a reduced level (3,7), indicating that they also have a role in the mature nervous system. This role might be in relation with differentiated neural activities or with regeneration or plasticity, as suggested by the induction of SCG10 with neurite regrowth following corticostriatal deafferentiation (10), or the up-regulation of RB3/RB3Љ (but not SCG10) after neuronal activation in the hippocampus (8). Interestingly, the various stathmin family proteins are expressed in different but overlapping cell populations (3,(11)(12)(13)(14)(15). SCG10 and SCLIP are expressed only in neurons, whereas stathmin and RB3/RB3Ј/RB3Љ are also expressed in glial cells (3,7). Stathmin family proteins may thus play related but distinct roles in different physiological environments.
Stathmin (reviewed in Ref. 16) has been originally identified as a relay protein integrating diverse intracellular signaling pathways (17) through combinatorial phosphorylation of its four phosphorylation sites (18). Several target/partner candidates have been identified (19,20), among which tubulin (21) is the protein whose interaction with stathmin has been best characterized. Stathmin indeed interacts with two ␣ /␤ tubulin heterodimers in vitro to form a T 2 S complex (22,23). The two tubulin heterodimers bind mostly the predicted ␣-helical "interaction" domain of stathmin (24), likely forming a curved complex whose three-dimensional structure has been recently revealed with the stathmin-like domain of RB3 (25). Stathmin displays a microtubule destabilizing activity both in vitro and in vivo (21, 26 -29), which is stoichiometrically accounted for in vitro by a free tubulin sequestration mechanism (22,27), although it has been proposed to be also due in part to a direct catastrophe-promoting activity (21,30,31). Interestingly, the activity of stathmin toward tubulin and microtubules is diminished when it is phosphorylated on various site combinations (28,29,32). Being phosphorylated during mitosis and in response to numerous extracellular signals, stathmin appears as a phosphorylation-dependent microtubule destabilizing factor, which may play important roles in proliferating as well as postmitotic cell regulations, in particular in the nervous system (1).
The various stathmin-like domains (3,4) of the stathmin family proteins display 65-75% amino acid identity with stathmin, including the predicted ␣-helix, and several of the four stathmin phosphorylation sites. Besides their stathmin-like domain, the neural members of the stathmin family are characterized by additional N-terminal domains. The A domain is common to SCG10, SCLIP, and RB3/RB3Ј/RB3Љ with 56 -68% amino acid sequence identity. As demonstrated in the case of SCG10 and very likely in the case of SCLIP and RB3/RB3Ј/ RB3Љ, it is responsible for the membrane attachment and targeting of the proteins to the Golgi area (33); the presence of SCG10 is also demonstrated in neuritic processes and in the growth cone (3,6,28,34). RB3/RB3Ј/RB3Љ also possesses a specific additional AЈ domain between the A and the stathminlike domains. RB3Љ further possesses another additional AЉ domain, between A and AЈ. Finally, RB3Ј differs from RB3 within its stathmin-like domain by an alternative splicing thus resulting in a different C terminus. The roles of these specific domains are still unknown, but they likely participate in extending and specifying the properties and functions of the various stathmin family proteins.
Like stathmin, the other members of its phosphoprotein family were also shown to destabilize the microtubule network when overexpressed in cultured cells (28,35). All these proteins thus form a family involved in the regulation of microtubule dynamics in the nervous system, in relation to differentiation, activities, and plasticity. Their diversity, partly originating in their different spatio-temporal expression and in the presence of specific combinations of N-terminal domains, suggests that they play related and likely complementary roles. The stathmin-like domains of stathmin family proteins might also have different contributions to the control of microtubules, especially as they show a molecular variability that could result in different tubulin binding and hence functional properties. This would be particularly relevant in cells of the nervous system and in their subcompartments such as axons, dendrites, or the growth cone, which display specific and distinctive microtubule organization and dynamics. To test this hypothesis, we examined the specific properties of the various stathmin-like domains. We show that they each interfere with microtubule polymerization and interact with two ␣ /␤ tubulin heterodimers. However, the T 2 S complexes formed have different stabilities and display distinct tubulin-SLD interaction kinetics, suggesting that each stathmin family protein is involved in microtubule dynamics regulation in a specific, distinctive fashion. The specific functional properties of the various proteins of the stathmin family in the nervous system thus appear to be determined not only by their characteristic N-terminal extensions but also through the fine-tuning of their stathmin-like domains.

Plasmid Constructs for Prokaryote Expression of Stathmin-like Domains
Standard recombinant DNA techniques were carried out as described (36). The full-length cDNA of human stathmin (37) within the pET-8c vector (38) was used for stathmin expression. Rat SCG10 (6), mouse SCLIP (3), rat RB3, and rat RB3Ј (4) cDNA clones were used for polymerase chain reaction amplification of the stathmin-like domain coding region. At the 5Ј end, the primers (Genset, Paris, France) were designed to introduce a NcoI site including an initiating ATG codon, a following alanine codon, and the appropriate sequence for each stathmin-like domain, starting at the residue corresponding to amino acid 5 in the stathmin sequence. At the 3Ј end, a BamHI site was introduced in the non-coding region. Vent DNA-polymerase and restriction enzymes were from Biolabs (Surrey, British Columbia, Canada). The resulting digested polymerase chain reaction fragments were subcloned into the corresponding sites of the pET-8c plasmid (Novagen, Madison WI) (39). For protein expression, the various plasmids were used to transform the Escherichia coli strain BL21(DE3) (Stratagene, La Jolla CA), which provides an inducible expression system suitable for the pET-8c vector.

Recombinant SLD Expression, Characterization, and Purification
Prokaryote Expression-An overnight preculture was used to seed 1 liter of Luria-Bertani medium containing 50 g/ml ampicillin, which was grown at 37°C. At exponential phase, recombinant protein expression was induced for 3 h by the addition of 0.4 mM isopropyl-␤-Dthiogalactopyranoside. Bacteria were then pelleted by sedimentation at 4°C, resuspended in 20 mM Tris-HCl, 1 mM EGTA, pH 8.0, containing the antiprotease mixture Complete (Roche Molecular Biochemicals), and sonicated three times for 1 min on ice.
Heat Stability and Protein Purification-Bacterial extracts were centrifuged at 100,000 rpm (Optima MAX ultracentrifuge, rotor TLA 100.2, Beckman Instruments, Fullerton, CA) for 6 min at 4°C to yield the S2 supernatants and P2 pellets. The S2 supernatants were then heated to 100°C for 5 min in the presence of 100 mM NaCl, and the samples were centrifuged again as above to yield the S3 supernatants and P3 pellets (17).
S3 extracts were adjusted to 20 mM Tris-HCl, 1 mM EGTA, pH 8.0, using Centriprep 10 (Millipore, Bedford, MA), loaded on a Q-Sepharose FF anion exchange column (Amersham Pharmacia Biotech), and eluted with a 0 -200 mM NaCl linear gradient in 20 mM Tris-HCl, 1 mM EGTA, pH 8.0. The eluted fractions were analyzed by SDS-polyacrylamide gel electrophoresis and Coomassie Blue staining. Stathmin-like domainpositive fractions were pooled, concentrated with Centriprep 10, and loaded on a Superose 12 HR 10/30 FPLC gel filtration column (Amersham Pharmacia Biotech) equilibrated with phosphate-buffered saline, 1 mM EGTA. Stathmin-like domain-positive fractions were pooled and concentrated as above. The precise masses of the purified protein products were checked by MALDI-TOF (see below), and their protein concentrations were accurately determined by amino acid analysis. In the case of RB3 SLD and RB3Ј SLD , purification and subsequent experiments were performed in the presence of 1 mM dithiothreitol to avoid disulfide bond formation.
Mass Spectrometric Analysis-Purified recombinant proteins were analyzed with a matrix-assisted laser-desorption/ionization time-offlight (MALDI-TOF) mass spectrometer (Voyager-DE STR Biospectrometry Work station, PE Biosystems Inc., Framingham, MA). The spectra of positive ions were recorded in linear mode with an accelerating voltage of 25 kV and a delayed extraction of 400 ns. The samples were mixed with a saturated solution of sinapinic acid (3,5-dimethoxy-4-hydroxicinnamic acid, Aldrich) in 30% acetonitrile and 0.1% aqueous trifluoroacetic acid. External calibration was performed with horse apomyoglobin using the monoprotonated and biprotonated ions with average mass-to-charge ratios of 16,952.56 and 8476.78, respectively.
The tryptic peptide digests of RB3 SLD and RB3Ј SLD were desalted and mixed with a saturated solution of 2,5-dihydroxybenzoic acid in 0.1% aqueous trifluoroacetic acid. MALDI-TOF mass spectra of the peptide mixture were performed in positive ion reflector mode with an accelerating voltage of 20 kV and a delayed extraction of 200 ns. External calibration was performed using the protonated ions of des-Arg bradykinin, angiotensin, and ACTH-(clip 1-17) with monoisotopic mass-tocharge ratios of 904.468, 1296.685, and 2465.199 respectively.
After desalting of the tryptic digests, half of the sample was dried and dissolved in 3 l of a mixture of water/formic acid/methanol (49:1: 50) for tandem mass spectrometry analyses using a nano-electrospray ionization Q-TOF (Q-TOF II, Micromass, Manchester, UK). The samples were loaded into nanoelectrospray capillaries (Protana, Odense, Denmark). The capillary voltage was set at 1000 V, the cone voltage at 45 V, and collision energies were about 20 -30 eV. Argon was used as the collision gas. The instrument was calibrated between 150 and 2000 atomic mass units with a solution of NaI.
Stokes Radius Measurement-The Stokes radii (R S ) and apparent molecular masses of the various proteins were measured by gel filtration on a Superose 12 HR 10/30 FPLC column (Amersham Pharmacia Biotech) equilibrated with 80 mM Pipes-KOH, 5 mM MgCl 2 , 1 mM EGTA, pH 6.5, at a 0.5 ml/min flow rate. The standard proteins used to calibrate the column are as follows: ribonuclease A (R S ϭ 16.4 Å), chymotrypsinogen A (R S ϭ 20.9 Å), ovalbumin (R S ϭ 30.5 Å), bovine serum albumin (R S ϭ 35.5 Å), aldolase (R S ϭ 48.1 Å), catalase (R S ϭ 52.2 Å), and ferritin (R S ϭ 61 Å). The void volume V 0 was measured as the blue dextran elution volume and the total volume of the gel bed V t as the acetone elution volume. The data were plotted according to Siegel and Monty (40). At least two runs were performed for each protein.

Bovine Brain Tubulin Preparation
Tubulin was purified from bovine brain crude extracts by two cycles of polymerization (41), followed by phosphocellulose chromatography (42). Tubulin was stored at Ϫ80°C in either 25 mM Mes-KOH, 0.25 mM dithiothreitol, 0.25 mM EGTA, 0.125 mM MgCl 2 , 0.025 mM EDTA, pH 6.8, for short term use, or in 50 mM Mes-KOH, 0.5 mM dithiothreitol, 0.5 mM EGTA, 0.25 mM MgCl 2 , 0.05 mM EDTA, 3.4 M glycerol, 0.1 M GTP, pH 6.8, for long term storage. In the latter case, an additional cycle of polymerization was performed before use, at the end of which tubulin was resuspended in 12.5 mM Mes-KOH, 0.25 mM EGTA, 0.25 mM MgCl 2 , pH 6.8. Tubulin concentration was determined by amino acid analysis.

In Vitro Tubulin Polymerization Assay
Samples containing varying concentrations of either tubulin alone or tubulin plus stathmin-like domain in 50 mM Mes-KOH, 30% glycerol, 6 mM MgCl 2 , 0.5 mM GTP, 1 mM EGTA, pH 6.8, were loaded in 100-l quartz cuvettes with a 1-cm light path. Tubulin polymerization was turbidimetrically monitored at 350 nm (43) in an Ultrospec 3000 thermostated spectrophotometer (Amersham Pharmacia Biotech). The temperature was raised from 3 to 37°C, and the increase in turbidity was recorded until a plateau was reached. The temperature was then set back at 3°C, and the decrease in turbidity recorded until depolymerization was complete. The polymerized stationary state level was defined as the difference between the plateau value at 37°C and the base line after return at 3°C.

Gel Filtration Assay
100-l samples containing 10 M tubulin, either alone or with varying concentrations of each stathmin-like domain, were analyzed by gel filtration on a Superose 12 HR 10/30 FPLC column (Amersham Pharmacia Biotech) equilibrated with 80 mM Pipes-KOH, 1 mM EGTA, 5 mM MgCl 2 , pH 6.5, at a 0.5 ml/min flow rate. The elution profile of the samples was recorded either at 278 nm (only tubulin is visible, as the stathmin-like domains have very few aromatic amino acids) or at 226 nm (both species monitored). When necessary, fractions were collected and analyzed by Western blot with an anti-␤-tubulin monoclonal antibody (Amersham Pharmacia Biotech) and a rabbit polyclonal serum against SCG10 (4).
Sensorchip Coupling-The first flow cell of each CM5 sensor chip was coupled with bovine serum albumin to be used as a reference flow cell, and the three others were coupled with various stathmin-like domains to allow direct comparison. The coupling was performed using the Amine Coupling Kit and the automated Immobilization Application Wizard included in the BIAcore software. The proteins to be coupled were in 10 mM sodium acetate, pH 5.0, and the coupling level was aimed at 1500 resonance units, except for RB3Ј SLD for which we coupled two flow cells at 1500 and 2500 resonance units.
Interaction Cycles-To study the interaction of tubulin on the flow cells coupled with the various stathmin-like domains, SPR experiments were performed at 20 l/min in 80 mM Pipes-KOH, 1 mM EGTA, 5 mM MgCl 2 , pH 6.5. In a single interaction run, a given concentration of tubulin was injected for 720 s to record the association phase on the four flow cells, after which buffer was supplied for 720 s to record the dissociation phase. At last, 10 l of 20 mM NaOH, 2.5 M NaCl were injected to regenerate the flow cells. The whole process was automated, and we performed multiple interaction runs with tubulin concentrations ranging from 0.6 to 20 M. We checked the absence of mass transfer. For the analysis, the sensorgrams were processed by subtracting the corresponding reference flow cell sensorgram, in order to abolish base-line drift, bulk, and nonspecific interaction contributions. The half-association (dissociation) time was determined as follows: if R M is the maximal signal reached during the association phase, then the half-association (dissociation) time is the time elapsed from the start of the association (dissociation) phase to reach the R M /2 level.

Biochemical Characterization of Recombinant Stathmin-like Domains
All stathmin family proteins possess a stathmin-like domain displaying a high (65-75%) amino acid sequence identity with stathmin ( Fig. 1). The aim of the present study was to determine how this strong but partial sequence identity is translated into similar or different functional properties, by comparing the biochemical properties of stathmin and stathmin-like The four stathmin phosphorylation sites, identified as phosphorylated in vivo (black dots), and the corresponding conserved putative phosphorylation sites (crosses) for the other family members are indicated. The characteristic predicted ␣-helix region with high coiled-coil probability is hatched. B, amino acid sequence of stathmin-like domains produced in bacteria. The generic term "stathmin-like domains" includes stathmin and the related stathmin-like domains. Stathmin was produced full length, whereas the other stathmin-like domains started at the amino acid corresponding to amino acid 5 of stathmin, preceded by a methionine (cleaved) and an alanine (bold italic). For all proteins, the amino acid number is always the number of the corresponding residue in stathmin. The C-terminal sequence divergence between RB3 and RB3Ј is boxed. Amino acid differences between the various stathmin-like domains are highlighted. Double arrows indicate the variable regions flanking the predicted ␣-helix. The sequences in gray are the two duplicated stretches in the ␣-helix, the first being more conserved among the stathmin family proteins than the second. domains, as well as their activities toward tubulin and microtubules. For clarity, we use the generic term "stathmin-like domains" that includes stathmin and the related stathmin-like domains, unless specified otherwise. Stathmin was produced in bacteria as described previously (38), and four other constructs were generated to produce the recombinant stathmin-like domains of SCG10, SCLIP, RB3, and RB3Ј, respectively, designated SCG10 SLD , SCLIP SLD , RB3 SLD , and RB3Ј SLD . RB3Љ SLD was not studied separately because it is identical to RB3 SLD (Fig. 1).
Solubility and Heat Stability of Stathmin-like Domains-To examine the solubility of the various recombinant proteins, the corresponding bacterial extracts were submitted to high speed centrifugation. As shown for RB3 SLD in Fig. 2, all stathmin-like domains were recovered, like stathmin, in the soluble "S2" fraction. Furthermore, after heat treatment at 100°C in the presence of 100 mM NaCl, the recombinant stathmin-like domains were also recovered in the subsequent high speed soluble fraction "S3," which indicates that they all possess the characteristic heat stability property of stathmin (17). Interestingly, the presence of the additional AЈ domain common to all RB3 proteins ( Fig. 1) resulted in the production of an insoluble RB3 SLD -AЈ protein, essentially recovered in the insoluble "P2" fraction ( Fig. 2). Due to the low yield of soluble RB3 SLD -AЈ, its biochemical and functional properties were not further investigated in this study.
The molecular masses of the recombinant proteins were checked by MALDI-TOF MS analysis. For stathmin, SCG10 SLD , and SCLIP SLD , the measured masses correspond to the calculated masses of the proteins without N-terminal methionine. For RB3 SLD and RB3Ј SLD , they correspond to the masses of the proteins without N-terminal methionine and a mass increment of about 42 Ϯ 3 Da. Since RB3 SLD and RB3Ј SLD are identical except for their C-terminal amino acid sequence, we assumed that both could bear the same post-translational acetylation that would account for a 42-Da mass increment. A tryptic digestion followed by MALDI-TOF MS analysis of the generated peptides showed that the N-terminal peptide of both RB3 SLD and RB3Ј SLD was indeed acetylated. This result was confirmed by the fact that their N-terminal sequencing was 80 -90% blocked. Furthermore, tandem mass spectrometric fragmentation of the N-terminal peptides using a Q-TOF mass spectrometer revealed that N-terminal peptides of both RB3 SLD and RB3Ј SLD were N-␣-acetylated on their N-terminal alanine.
Shape Characterization by Stokes Radius Measurement-We analyzed the various stathmin-like domains by gel filtration, and their elution volumes were used to calculate their Stokes radii and apparent molecular masses (Table I). As expected, stathmin displayed an atypical behavior, since it eluted like a globular protein of about 110 kDa, i.e. about 6-fold its actual molecular mass. This high apparent molecular mass reveals an asymmetrical shape (44). The other stathmin-like domains also displayed abnormally high Stokes radii (Table I), implying that they are asymmetrical proteins as well. However, by using the MM app /MM MS ratio (apparent molecular mass determined by gel filtration/molecular mass measured by MALDI-TOF MS) as a measure of this asymmetry, it appears that stathmin is more elongated than the other stathmin-like domains, followed by SCG10 SLD and SCLIP SLD and then by RB3 SLD and RB3Ј SLD (Table I). Altogether, all stathmin-like domains share an elongated form but to significantly different extents, suggesting that they possess specific structural features that might be functionally relevant.

Activity on Tubulin Polymerization in Vitro
In order to compare the functional properties of the various stathmin-like domains, we assessed their activity on tubulin polymerization in vitro. The addition of any stathmin-like domain to the tubulin polymerization reaction resulted in a decreased microtubule amount at steady state. For example, 4 M RB3 SLD lowered the amount of microtubules normally formed with 20 M tubulin to that obtained with 12 M tubulin, as if 4 M RB3 SLD prevented 8 M tubulin to enter the polymerization reaction (Fig. 3A). The effects of increasing concentrations of stathmin-like domain on microtubule assembly with 20 M tubulin revealed that the presence of any stathmin-like domain at a concentration C induced the same effect as a 2C tubulin decrease (Fig. 3B). Therefore, everything occurs as if each stathmin-like domain molecule were able to sequester two ␣ /␤ tubulin heterodimers. It thus seemed very likely that they might interact directly with tubulin, like stathmin which forms a T 2 S complex preventing the corresponding tubulin to enter the polymerization reaction (22,27).

Interaction of Stathmin-like Domains with Tubulin
Characterization of the Tubulin-Stathmin-like Domain Complexes-In order to assess the existence of a direct tubulin-SLD interaction, samples containing 10 M bovine brain tubulin and a concentration range of each stathmin-like domain (1, 2.5, 5, 10, and 20 M) were loaded on an FPLC gel filtration column. Monitoring at 278 nm was used to follow tubulin elution only, whereas both tubulin and stathmin-like domains were monitored either at 226 nm or in outflow fractions by Western blotting. As illustrated in Fig. 4A for SCG10 SLD , we observed for each stathmin-like domain a second tubulin peak with a smaller elution volume, which probably corresponds to tubulin within a larger molecular complex. The latter is probably a tubulin-SLD complex because its amount increased with stathmin-like domain concentration. Moreover, the gel elution profiles obtained with 5, 10, and 20 M stathmin-like domain are superimposable and do not reveal any free tubulin. Thus, 5 M stathmin-like domain is sufficient to complex 10 M tubulin, which is consistent, as in the case of stathmin, with a 2:1 tubulin:SLD stoichiometry.
Western blot analysis (Fig. 4B) confirmed the stoichiome- Finally, monitoring the elution outflow at 226 nm (which allows the detection of both tubulin and stathmin-like domains) revealed a small peak corresponding to free stathminlike domain in excess, as shown for RB3Ј SLD in Fig. 4C. This peak was present when the tubulin:SLD ratio was lower than 2:1, i.e. when there was more than 5 M stathmin-like domain for 10 M tubulin. Altogether, our results are consistent with a 2:1 stoichiometry of the tubulin-SLD complexes.
Different Stabilities of the Tubulin-Stathmin-like Domain Complexes-Besides the overall similarities of the various stathmin-like domains in their interference with microtubule assembly and their interaction with tubulin, the comparison of the gel filtration elution profiles points out that the various tubulin-SLD complexes have actually distinct properties. Indeed, the resulting shifted elution peaks had different shapes, as seen, for example, for 10 M tubulin in the presence of 10 M of any stathmin-like domain (Fig. 5A). The shifted peaks can be sorted in three types as follows: tubulin peaks obtained with SCG10 SLD and RB3 SLD were the most shifted toward the smaller elution volumes and displayed a narrow and relatively symmetrical shape. The peaks with stathmin and SCLIP SLD were less shifted and had an asymmetrical shape. This is probably due to the dissociation along the column of some originally complexed molecules, as they underwent dilution during their migration and were separated from the slower   migrating monomers. Finally, the peak obtained with RB3Ј SLD was eluted at an intermediary position between the other complexes and free tubulin; this late elution position likely results from a more effective dissociation. Indeed, if the dissociation on the column is extensive from the beginning of the elution, it is possible that no tubulin molecule originally complexed in the sample remains associated, all being eluted at an average intermediate position between the volumes corresponding to the free and complexed forms of tubulin.
The dissociation of the complex is even more striking in more dilute concentration conditions, such as 10 M tubulin ϩ 1 M stathmin-like domain, for example (Fig. 5B). In the case of stathmin and SCLIP SLD , there was no more individualized complex peak but only a shoulder in front of the free tubulin peak. On the other hand, the tubulin-RB3 SLD complex peak remained individualized and was eluted at the same position as observed with 10 M RB3 SLD (about 10.5 ml). Therefore, it appears that the tubulin-RB3 SLD complex is far less sensitive to dissociation along the column than tubulin-stathmin and tubulin-SCLIP SLD . The tubulin-SCG10 SLD complex seems to have an intermediate stability, as its elution began at an early position but was spread out toward the free tubulin peak. As mentioned above, the tubulin-RB3Ј SLD complex is the least stable, since the tubulin peak was even less shifted at 1 M than at 10 M RB3Ј SLD .
As the tubulin-RB3 SLD complex remained totally associated at 1 M RB3 SLD , we tried even more dilute concentration conditions to test to what extent the stability of this complex was different from the others. The elution profile of a stoichiometric mixture of 1 M tubulin ϩ 0.5 M RB3 SLD (Fig. 5C) reveals clearly that the complex did not dissociate during migration, as the peak position was not delayed and no trail appeared. We did not try lower concentrations, as the signal:noise ratio was too low.
In conclusion, we found that the complexes of tubulin with the various stathmin-like domains display different stabilities in the following order: RB3 SLD , SCG10 SLD , stathmin and SCLIP SLD , and finally RB3Ј SLD .

Different Interaction Kinetics of Tubulin with the Various Stathmin-like Domains
We further characterized the kinetics of the interaction between the various stathmin-like domains and tubulin by using the SPR BIAcore technology, which monitors the mass concentration variation at the vicinity of a surface. A constant tubulin concentration flow was applied (association phase) on flow cells coupled with the various stathmin-like domains, followed by a flow of buffer containing no tubulin (dissociation phase). The SPR signals were then corrected for nonspecific signal using a reference flow cell coupled with bovine serum albumin.
An example of the corresponding interaction kinetics is given in Fig. 6A, where the signals corresponding to a 4 M tubulin run have been normalized (the maximum signal corresponding to 100%), to allow their more direct comparison. We see clearly that the various tubulin-SLD complexes displayed distinct association and dissociation kinetics. As a measure of the kinetics differences, we represented the half-association and dissociation times, which vary non-linearly with tubulin concentration (Fig. 6, B and C).
The association kinetics of tubulin to stathmin, SCG10 SLD , and SCLIP SLD could not be distinguished, whereas association to RB3 SLD was much slower and to RB3Ј SLD was much faster (this latter observation does not appear so clearly on the halfassociation plot (Fig. 6B), as the measured times are very close to the minimal measurable value, but it is striking on the SPR curve in Fig. 6A). For their dissociation kinetics, stathmin and SCLIP SLD were not significantly different, although tubulin-SCLIP SLD always dissociated slightly more slowly. Tubulin-SCG10 SLD dissociated significantly slower. Moreover, tubulin-RB3 SLD dissociated even slower, whereas tubulin- RB3Ј SLD dissociated extremely fast as compared with all the other stathmin-like domains. Altogether the various kinetics of tubulin association and dissociation with stathmin-like domains are in good agreement with the stabilities of the formed complexes as revealed by gel filtration chromatography: intermediate stability for stathmin and SCLIP SLD ; slightly higher stability for SCG10 SLD due to slower dissociation; and strong stability for RB3 SLD with both slow association and dissociation, as opposed to weak stability with RB3Ј SLD with both fast association and dissociation. DISCUSSION The biological expression and regulation of phosphoproteins of the stathmin family suggest that they play different, possibly complementary roles in development, plasticity, and activities of the nervous system. In the case of stathmin, its proposed signal integration and relay functions are at least in part mediated by its interaction with tubulin and hence its involvement in the control of microtubule dynamics (reviewed in Ref. 1). Interestingly, the other phosphoproteins of the stathmin family share a similar but distinct stathmin-like domain and have been shown to interfere also with microtubule assembly in vivo and in vitro (9,28,35). We demonstrate here that all known stathmin-like domains interact with tubulin and inhibit microtubule assembly in vitro through tubulin sequestration. We further report that each stathmin-like domain interacts with tubulin in a distinctive way, most likely contributing to the specific biological roles of the various stathmin family phosphoproteins, particularly in the differentially regulated control of microtubule assembly in the diverse cell and subcellular compartments of the nervous system, at the various stages of development and adult life.
Biochemical Properties of Stathmin-like Domains-All recombinant stathmin-like domains display characteristic stathmin-like biochemical properties, such as high solubility and heat stability, as well as high Stokes radius indicating an elongated shape. This illustrates their genuine stathmin-like character, consistent with their extensive amino acid sequence identity (65-75%), and their similar predicted secondary structure including a long ␣-helix (3,4,45). However, beside these overall similarities, the various stathmin-like domains display distinctive structural features as suggested by differences in their measured Stokes radii, which most likely reveal different conformations possibly related to specific biological roles.
The biochemical characterization of recombinant stathminlike domains revealed an unexpected feature, the N-␣-acetylation of RB3 SLD and RB3Ј SLD . Although N-␣-acetylation is rarely observed in prokaryote systems, it was reported previously for the endogenous ribosomal proteins L12, S5, and S18 in E. coli, as well as for some recombinant proteins (46). SCG10 SLD and SCLIP SLD did not undergo this modification, although they share the same (M)ADMEV N-terminal sequence as RB3 SLD , the first difference arising only at the sixth residue (a lysine for SCG10 SLD and SCLIP SLD instead of an isoleucine for RB3 SLD and RB3Ј SLD ). This further indicates that the determinants for N-␣-acetylation are far broader than the very first N-terminal residues (47).
Microtubule Destabilizing Activity of Stathmin Family Proteins-Our present observation that all stathmin-like domains of the family inhibit tubulin polymerization in vitro, in a way similar to stathmin (21,27) and SCG10 (9), confirms their stathmin-like character at the functional level. These results demonstrate in vitro an action on microtubule assembly for all members of the family, in agreement with in vivo observations of microtubule interference of entire proteins of the family when overexpressed in HeLa cells (28). Moreover, this is the first demonstration of microtubule assembly inhibition by RB3Ј, whose activity could not be evidenced in vivo.
Quantitative analysis of the activities of stathmin-like domains toward tubulin polymerization reveals that the presence of any stathmin-like domain at a concentration C inhibits the polymerization of a concentration 2C of tubulin. In the case of stathmin, the existence of a T 2 S complex (22,23) is sufficient to account for the stoichiometric effect of stathmin on tubulin polymerization in vitro, if one considers that one stathmin sequesters two ␣ /␤ tubulin heterodimers and prevents them from entering polymerization (27). As we demonstrate here the FIG. 6. Different interaction kinetics of tubulin with stathminlike domains. A, surface plasmon resonance "net" sensorgrams (bovine serum albumin reference signal subtracted) revealing an interaction between soluble tubulin and the various stathmin-like domains immobilized on CM5 sensorchips. The association phase (from 0 to 720 s) corresponds to a constant 4 M concentration of free tubulin in the flow, whereas the dissociation phase (from 720 s to the end) corresponds to buffer without tubulin. The signals were normalized (100% ϭ maximal signal) in order to highlight the kinetics differences. B and C, the mean half-association times (B) and mean half-dissociation times (C) were measured for each stathmin-like domain at various tubulin concentrations. Error bars correspond to the standard deviation obtained for two to seven different measurements, whereas the absence of an error bar indicates a single measurement. existence of complexes between tubulin and stathmin-like domains, it is likely that all stathmin-like domains inhibit tubulin polymerization in vitro by tubulin sequestration. Recently, the tubulin-RB3 SLD complex was actually crystallized, and xray diffraction analysis revealed a three-dimensional structure compatible with tubulin sequestration (25).
Stathmin has been shown also to increase microtubule catastrophe frequencies in some conditions (21,30,31). Although such an effect is expected to some extent as a consequence of tubulin sequestration, it has been proposed that it might also result from a direct interaction of stathmin with microtubule ends, which has not been observed so far. Quantitative comparison of the tubulin binding and catastrophe-promoting activities of stathmin-like domains would give clues to determine to what extent tubulin sequestration and catastrophe promotion are two independent actions of these proteins on microtubule stability.
Distinct Tubulin Interaction Properties of Stathmin-like Domains-The interaction between stathmin-like domains and tubulin has been characterized by means of gel filtration, which informs on the complex formed and its stability, and by surface plasmon resonance, which allows us to follow the kinetics of the interaction, as done previously for stathmin (22).
Some limitations preclude the interpretation of SPR results in terms of actual kinetic rate constants. Indeed, the reaction pathway for the formation of the T 2 S complexes is not known and very probably involves second-order kinetics. Moreover, heterogeneities of two kinds are likely to result in additional complexity of the interaction kinetics. Brain tubulin is heterogeneous due to the existence of several genes and post-translational modifications of the gene products; a differential interaction of stathmin-like domains with the diverse tubulin isoforms might be actually of physiological significance. A more technical heterogeneity is that of the stathmin-like domains, which may be coupled to the sensorchip through one or several random lysines. In any case, the SPR technology allowed us to compare the relative interaction potencies of the various stathmin-like domains with tubulin and to reveal, at least qualitatively, a true diversity of the tubulin-SLD interactions. Indeed, the interaction of SCLIP SLD with tubulin is very close to that of stathmin: SCG10 SLD associates similarly but dissociates more slowly from tubulin; RB3 SLD displays both the slowest association and dissociation phases; RB3Ј SLD displays the fastest ones.
The SPR data can be compared with results following gel filtration, which sort the various complexes according to their stability along the column: interestingly, RB3 SLD generates the most stable complex, followed by SCG10 SLD , stathmin, and SCLIP SLD , and finally RB3Ј SLD . These results are highly consistent with the SPR results, which mutually strengthen their validity and significance. The fact that the tubulin-SLD stability differences were not revealed by following the action of stathmin-like domains on tubulin polymerization is not surprising because of the high tubulin concentrations necessarily used in this assay, thus leading to full association of low as well as high affinity stathmin-like domains with tubulin.
The predicted ␣-helix encompassing the "core" region (residues 42-126) (24) is essential for the interaction with tubulin and corresponds most likely to the 91-amino acid ␣-helix interacting with two ␣ /␤ tubulin heterodimers in the tubulin-RB3 SLD complex (25). This ␣-helix is made of two duplicated stretches (30 -40% identity) of 35 residues (25, 37) (Fig. 1B), whose spacing is consistent with that of the two ␣ /␤ tubulin heterodimers in the tubulin-RB3 SLD complex (25). As the amino acid sequence of the first stretch is significantly more conserved between all stathmin-like domains than that of the second, the two stretches might have different contributions to the binding of the two tubulins.
The characteristic tubulin interaction differences among the various stathmin-like domains result from their 25-35% differences in primary sequences. The two regions (residues 29 -39 and 138 -149) flanking the interacting ␣-helix are the most divergent (Fig. 1B) and could contribute to the observed tubulin interaction differences if they were involved in stabilizing the interaction. However, as these regions are also phylogenetically variable, this would not be the case if the characteristic interaction properties of the various members of the stathmin family with tubulin were, as expected, conserved through evolution. It is thus likely that more subtle sequence differences within the more conserved regions of the various stathmin-like domains are responsible for the observed tubulin interaction differences.
The two splice variants RB3 SLD and RB3Ј SLD have opposite tubulin interaction properties, although their sequences are identical up to residue 124, but are totally divergent on their C termini. This suggests that either the C-terminal part of RB3 SLD (which is absent from RB3Ј SLD ) is very important to stabilize the interaction or the C-terminal part of RB3Ј SLD has a strong destabilizing effect. It is interesting that differential splicing can direct the expression of the rb3 gene toward proteins forming a highly stable or unstable complex with tubulin, which might be of physiological importance regarding the regulation of microtubule dynamics in the corresponding cells.
Biological Significance-Microtubule dynamics are regulated by many stabilizing and destabilizing factors, such as microtubule-associated proteins or Kin1, whose opposite activities establish a finely tuned balance between microtubule polymerization and depolymerization. The diversity of stathmin family proteins, together with the diversity of MAPs in the nervous system, might enhance the flexibility of this regulation, each protein contributing to it differently with specific tubulin interacting properties and microtubule destabilizing activity. In particular, different tubulin interaction kinetics and complex stability might result in different contributions to microtubule assembly dynamics, allowing a fine-tuning of its regulation according to the local tubulin/microtubule status within the cell, as well as to specific needs in various cells and cell compartments within the nervous system. Moreover, the different intracellular distribution of stathmin family proteins might allow the local regulation of the microtubule network. Indeed, stathmin is cytosolic whereas stathmin family proteins are membrane-bound, with a punctate localization at the level of the Golgi apparatus, neuronal processes, and growth cones (6,34). The concentration of stathmin-related proteins on some membrane compartments may thus create a local change in microtubule dynamics, which could be important for process elongation or organelle transport. This could explain why most membrane-bound members of the family are expressed in neural cells where stathmin itself is abundant. Furthermore, the activity of stathmin family proteins toward microtubules might be regulated by their local release from the membrane compartment, as it has been suggested in the case of SCG10 (35), this release being possibly controlled differently for the various proteins of the family. It is also possible that the various stathmin family proteins interact differently with the various tubulin isoforms, thus modifying the microtubule composition and stability.
An additional complexity in the regulation of the neural microtubule network might come from the differential phosphorylation of stathmin family proteins. Indeed, stathmin is known to integrate intracellular signaling pathways through combinatorial phosphorylation, and phosphorylation has been shown to regulate the microtubule destabilizing activity of stathmin and SCG10 (35,48). As several but not all the stathmin phosphorylation sites are present in the various stathminlike domains, their phosphorylation may occur under different conditions; each protein would thus be able to modulate the microtubule network differently, possibly locally, according to its phosphorylation state. The presence of other additional domains AЈ and AЉ in the RB3 proteins may further broaden the diversity of stathmin family functions, in part through alternative splicing.
Altogether, we show that in addition to their overall structural and functional stathmin-like properties, the various stathmin-like domains display specific molecular and tubulin binding properties most likely of physiological relevance. The stathmin family phosphoproteins thus form a set of microtubule regulators with diverse properties, which may participate in cytoskeleton reorganization in the developing and the mature nervous system. For a better understanding of the role of these proteins, it will be important to continue assessing their diversity, particularly by investigating the role of their Nterminal domains, their phosphorylation, and also their differential expression and localization in tissues and cells. All these properties may participate in defining the specific physiological role of each stathmin family protein.