Probing the Native Structure of Stathmin and Its Interaction Domains with Tubulin

Stathmin is a cytosoluble phosphoprotein proposed to be a regulatory relay integrating diverse intracellular signaling pathway. Its interaction with tubulin modulates microtubule dynamics by destabilization of assembled microtubules or inhibition of their polymerization from free tubulin. The aim of this study was to probe the native structure of stathmin and to delineate its minimal region able to interact with tubulin. Limited proteolysis of stathmin revealed four structured domains within the native protein, corresponding to amino acid sequences 22–81 (I), 95–113 (II), 113–128 (III), and 128–149 (IV), which allows us to propose stathmin folding hypotheses. Furthermore, stathmin proteolytic fragments were mixed to interact with tubulin, and those that retained affinity for tubulin were isolated by size exclusion chromatography and identified by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. The results indicate that, to interact with tubulin, a stathmin fragment must span a minimal core region from residues 42 to 126, which interestingly corresponds to the predicted α-helical “interaction region” of stathmin. In addition, an interacting stathmin fragment must include a short N- or C-terminal extension. The functional significance of these interaction constrains is further validated by tubulin polymerization inhibition assays with fragments designed on the basis of the tubulin binding results. The present results will help to optimize further stathmin structural studies and to develop molecular tools to target its interaction with tubulin.

Stathmin (␣, relay) (1), also referred to as Op 18 (2), is a small ubiquitous cytosolic phosphoprotein that has been proposed to be a relay integrating diverse intracellular signaling pathways (3). It is phosphorylated in parallel with the action of hormones (4 -6), growth factors (7,8), and neurotransmitters (9), and it interacts with putative downstream protein partners including tubulin (10 -16). Stathmin is also the generic element of a protein family including the neural proteins SCG10, SCLIP, RB3, and its two splice variants RB3Ј and RB3Љ (17). A great deal of progress has recently been made in under-standing the function of stathmin. It now appears that it may be one of the key regulators of microtubule dynamics, likely implicated in various microtubule-dependent cellular functions in interphase or mitosis (18 -21). Actually, it has been shown that stathmin destabilizes microtubules (15). Semi-quantitative in vitro studies (20,21) have revealed that stathmin prevents assembly and promotes disassembly of microtubules in a concentration-dependent manner and that inhibition of microtubule assembly is abolished by stathmin phosphorylation in vitro (21) or in vivo (20,22,23). It has been also shown that some of these properties are shared by the other stathmin family proteins (18,24,25).
However, although the destabilizing potential of stathmin toward microtubules is firmly established, the mechanism by which it fulfills this function is still not completely settled. Previous in vitro studies have shown that depolymerization of microtubules can be explained by stathmin sequestration of free tubulin in a T2S complex (16,26). We also have measured that stathmin alters neither the steady state GTPase activity of microtubules nor the critical concentration, all results consistent with a pure tubulin sequestering activity (26). On the other hand, Howell et al. (27) have proposed that stathmin may be a dual potential protein as follows: part of the N-terminal stathmin region could promote catastrophes, whereas tubulin sequestration could be supported by the rest of the molecule. These authors have proposed that sequestration predominates at low pH (6.8), whereas stathmin promotes catastrophes when rising the pH to 7.5. This duality could extend the properties of stathmin but, at this time, still has to be confirmed.
In the present work, we investigated the structural organization of stathmin and the mechanisms by which it interacts with its functional partners, such as tubulin. We therefore first demonstrated the existence and localization of structured domains within the stathmin molecule by means of limited proteolysis. The involvement of these domains in the interaction of stathmin with tubulin was examined by analyzing which of the smallest stathmin fragments can retain the capacity to interact with tubulin. For this study, we used a strategy based on proteolysis of stathmin, followed by sorting out the fragments still able to interact with tubulin using size exclusion chromatography. These polypeptides were then analyzed by Western blotting and identified by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS), 1 taking advantage of the high accuracy and sensitivity of this method (28,29). The functional value of the interaction results was validated by tubulin polymerization inhibition assays performed with three recombinant stathmin fragments constructed on the basis of the tubulin binding results. Altogether, our results reported here give clues for understanding the structural organization of stathmin as well as the mechanism of interaction of stathmin with tubulin. They may lead to the design of molecular tools tailored to mimic or block the microtubule depolymerization activity of stathmin.

EXPERIMENTAL PROCEDURES
Stathmin Limited Proteolysis-In order to obtain a good resolution of the stathmin structured domains, Lys-C and Glu-C endoproteases (sequencing grade proteases, Roche Molecular Biochemicals) were chosen because both have a high number of cleavage sites that are evenly distributed along the stathmin sequence except in its N terminus (23 and 30 sites, respectively (see Fig. 4B)). The concentration of the proteases is expressed for each proteolysis experiment throughout this paper as the enzyme:stathmin ratio (w/w). Lys-C and Glu-C were added to 9 g of stathmin (recombinant human stathmin (30)) in a 10-l final volume of AB or NH 4 HCO 3 buffer, respectively (AB buffer: 80 mM K-Pipes, pH 6.5, 1 mM EGTA, 5 mM MgCl 2 , 25 mM ammonium bicarbonate, pH 7.8). Different enzyme to stathmin ratios were used, 1:32,000, 1:16,000 (Lys-C and Glu-C), and 1:8000 (Lys-C), and incubated at 37°C (Lys-C) or at room temperature (Glu-C) for 15 min. Digestion was stopped by adding 90 l of 0.1% trifluoroacetic acid, 5% acetonitrile buffer. Evolution of the stathmin proteolysis was assayed by reverse phase HPLC and MALDI-TOF mass spectrometry.
Production and Sorting Out of Stathmin Fragments for Their Capacity to Interact with Tubulin-Stathmin fragments were produced by Lys-C or Glu-C proteolysis. The ability of the fragments to interact with tubulin was assessed by comparing their size exclusion chromatography elution volumes observed with or without prior incubation with tubulin. Fragments were then classified in two subtypes. The first comprises fragments that had a clear diminution of their elution volume in the presence of tubulin as compared with that observed in control conditions. They were considered to interact with tubulin. The second corresponds to fragments that eluted in the same position in the presence or absence of tubulin and that were considered as unable to interact with tubulin. Finally, a consensus region for stathmin interaction with tubulin was derived from the comparison of the sequences of the fragments from these two subtypes.
For each protease, two parallel stathmin digestion mixtures were generated. Proteolysis mixtures contained 125 g of stathmin and either Lys-C (1:12,000) or Glu-C (1:4000) enzyme:stathmin w/w ratio, in AB (Lys-C) or NH 4 HCO 3 (Glu-C) buffer. They were incubated at 37°C (Lys-C) or room temperature (Glu-C) for different times (0, 40, 80, 120 (Lys-C and Glu-C), and 160 min (Lys-C)) when an aliquot corresponding to 25 g of stathmin (1.5 nmol) was taken from each mixture. The aliquot for tubulin interaction received 25 l of 4ϫ AB buffer, water, and tubulin to reach 30 M tubulin (final volume 100 l), and the control aliquot received 25 l of 4ϫ AB buffer and water to 100 l. Samples were subjected to size exclusion chromatography on an FPLC Superose 12 HR 10/30 column equilibrated in AB buffer at 0.5 ml/min. Spectrophotometer recordings were performed at 280 nm. At this wavelength stathmin is not absorbing due to its lack of aromatic residues, and therefore only tubulin is detected. Fractions of 250 l were collected from 20 min after sample injection. Their stathmin or stathmin fragment content was analyzed by Western blotting and MALDI-TOF MS.
Size exclusion chromatography data are presented either as the elution volume or as Stokes radius. Stokes radii were determined as described previously (16). Briefly, a (Ϫlog K av ) 1/2 versus Stokes radius plot was constructed using a set of standard proteins of known Stokes radii (gel filtration calibration kit (Amersham Pharmacia Biotech)), where K av is a parameter calculated as where v e is the elution volume corresponding to the peak concentration of a protein; v o is the void volume of the column; and v t is the total volume of the gel bed). The elution volume of stathmin was determined precisely in a control run from a spectrophotometer recording. On the other hand, the elution volume of stathmin fragments was not directly accessible from spectrophotometer recordings. It was estimated to be that of the elution volume at which they were detected by MALDI-TOF MS if present in a single fraction or that of the mean elution volume if the fragment was detected in successive fractions.
Gel Electrophoresis and Western Blots-One-dimensional electro-phoresis was performed on 13% sodium dodecyl sulfate-polyacrylamide gels. Proteins were either silver-stained on fixed gels as described previously (16)  1% Triton X-100). Membranes were probed with different stathmin antisera directed against either the region between residues 15 and 27 at 1:10,000 dilution (31), an epitope located in the region encompassing residues 44 -125 at 1:5000, 2 or the C terminus (amino acid residues 134 -149) at 1:20,000 (31). Bound antibodies were detected by chemiluminescence (ECL, Amersham Pharmacia Biotech) with the appropriate secondary antibodies, the membranes being exposed to XAR5 film (Eastman Kodak Co.). MALDI-TOF Mass Spectrometry-Stathmin fragments were analyzed by MALDI-TOF mass spectrometry. The cleaved sites on stathmin were identified after determination of the molecular masses of the polypeptides fragments, using the stathmin sequence and the known cleavage specificity of the endoproteinases. Stathmin amino acid residues were numbered from 2 to 149, residue 1 representing the Nterminal cleaved methionine encoded by stathmin mRNA. The sample solution was mixed with a saturated solution of sinapinic acid (3,5dimethoxy-4-hydroxycinnamic acid, Aldrich) in 30% acetonitrile, 0.1% trifluoroacetic acid. The spectra of positive ions were recorded in linear mode on a MALDI-TOF mass spectrometer (Voyager Elite, Perseptive Biosystem, Inc. Framingham, MA) equipped with a delayed extraction device. For laser desorption, a nitrogen laser beam ( ϭ 337 nm) was focused on the target. Delayed extraction time was set at 150 ns. About 200 shots were averaged for each acquired spectrum. External calibration was performed with apomyoglobin using the monoprotonated ion of the dimer, and of the monomer, and the biprotonated ion of the monomer with average mass to charge (m/z) ratios corresponding to 33,904, 16,952.5, and 8476.75 respectively. The difference between the calculated average mass and the experimental mass determination (0.05% to 0.1%) is consistent with the accuracy of MALDI-TOF mass spectrometry in linear mode.
Preparation of Recombinant Stathmin Fragments-DNA manipulations were carried out using standard recombinant techniques (32). cDNA encoding stathmin fragments were constructed by the "megaprimer" polymerase chain reaction technique (33) with oligonucleotides primers from Genset (France) and the human stathmin cDNA as a 2 P. A. Curmi, unpublished data. substrate. Three different stathmin fragments were generated, whose limits were chosen for being as close as possible to either "the core" (fragment 44 -125) (see under "Results" for a definition of the "stathmin core"), the core extended on its N-side to the Lys-C site Lys-13 (fragment 12-125), and the core with a C-extension to Glu-138 (fragment 44 -138), an amino acid residue that represents the point of sequence divergence in the stathmin family proteins (17). The polymerase chain reaction fragments were cloned in the Escherichia coli expression vector pET-8c, expressed in the E. coli strain BL21, and purified as described (30). Fragment concentrations were determined by amino acid analysis.

FIG. 2. MALDI-TOF MS analysis of stathmin limited proteolysis with Lys-C.
A, the indicated enzyme to stathmin ratio mixtures (w/w) were incubated at 37°C for 15 min except for the control (upper panel) where digestion (1:8000, w/w) was stopped immediately after addition of the enzyme Lys-C. Stathmin, 17,172 Da calculated average mass for its protonated ion, appeared as a singly protonated ion, labeled stathmin or (ϩ1), and as bi-and tri-protonated ions labeled (ϩ2) and (ϩ3), respectively. Each of the other peaks corresponds to a stathmin fragment, and, for the sake of clarity, only the singly charged fragments are labeled. Stathmin peptides produced by Lys-C digestion were identified with their average molecular masses using the stathmin sequence and the known cleavage specificity of the endoproteinase. Some ions may correspond to either a small protonated stathmin peptide or a larger biprotonated peptide (ϩ2). This was the case for peptide 43-104 (marked with an asterisk). Its mass cannot be distinguished from the biprotonated ion of fragment 2-128. Since we did not observe any other peptide that begins or ends at position 43/42 or 105/104, respectively, and its intensity is correlated to that of fragment 2-128, we did not consider sites 42  Tubulin Polymerization Assay-Tubulin was purified from bovine brain by two cycles of polymerization and depolymerization followed by phosphocellulose chromatography (34). Tubulin concentration was determined by amino acid analysis, and the protein was stored at Ϫ80°C in 50 mM MES-KOH, pH 6.8, 0.5 mM EGTA, 0.25 mM MgCl 2 , 0.1 mM GTP until use. Tubulin polymerization was monitored turbidimetrically at 350 nm in an Ultrospec 3000 spectrophotometer (Amersham Pharmacia Biotech) thermostated at 37°C (1 cm light path). Experiments were carried out in 50 mM MES-KOH, pH 6.8, 30% glycerol, 0.5 mM EGTA, 6 mM MgCl 2 , and 0.5 mM GTP (buffer M). Critical concentration plots representing the amount of polymerized tubulin observed at steady state versus the total concentration of tubulin, measured in the absence or presence of stathmin and stathmin fragments, were constructed as described (26).

Stathmin Presents Four Structured Domains in Solution-
Limited proteolysis is a classical strategy to isolate discretely folded domains of proteins taking advantage that the exposed regions between folded domains are the most sensitive protease cleavage sites. Therefore, we performed limited proteolysis of stathmin, using Lys-C and Glu-C, two proteases exhibiting a high number of cleavage sites randomly distributed along the stathmin sequence. Conditions of limited proteolysis with these two enzymes were determined by analyzing the progress of stathmin proteolysis by reverse phase HPLC and MALDI-TOF MS. Stathmin digestion is actually a progressive process as observed on Fig. 1. To obtain reproducible limited proteolysis we used a fixed digestion time of 15 min with varying relative enzyme concentrations. Figs. 2A and 3A show, respectively, the MALDI mass spectra of stathmin before and after 15 min digestion with Lys-C at 1:32,000, 1:16,000, and 1:8000 and with Glu-C at 1:32,000 and 1:16,000. The delimitation of stathmin structured domains was based on the analysis of initial cleavage site distribution patterns. Stathmin secondary structure predictions were used to assist interpretation of the results. Scores of ␣-helix and limits of stathmin secondary patterns predicted by the Chou and Fasman method were used for reference as they are representative of results obtained with various different secondary structure prediction algorithms (35)(36)(37)(38).
At the lowest protease:stathmin ratio (1:32,000), despite the relatively even distribution of theoretical cleavage sites, both Lys-C and Glu-C cleave stathmin only at discrete sites unevenly distributed along the sequence. Two clusters of cleavage sites lie in the regions between amino acid residues 10 -21 and 81-95. In addition, isolated cleavages occurred with Lys-C at residue Lys-128 and with Glu-C at residue Glu-113. The mass profiles for each enzyme looked very similar at 1:32,000 and at 1:16,000 with, however, an increased intensity of the peaks at 1:16,000. New cleavage sites appeared either immediately next to the initial sites (Lys-80 and -126 for Lys-C) or located at the beginning of a predicted ␣-helix (Lys-42, Lys-C) and in a region where the ␣-helix score is lower (Lys-104 and Lys-62 and Glu-65). Interestingly, the latter region surrounds Ser-63 which is an important stathmin phosphorylation site. At 1:8000, Lys-C proteolytic cleavage was more extensive and probably reflected the beginning of stathmin denaturation. The superimposed Lys-C and Glu-C concentration-dependent data with ␣-helix secondary structure prediction (bottom of Figs. 2B and 3B) reveals the uneven localization of initial protease attack. Fig. 4 gives a synthetic view of initial cleavage site distribution together with a sequence map of stathmin showing the delimitation of the deduced stathmin structured domains. Four structured domains can be identified on the basis of our limited proteolysis conditions. Domain I extends from amino acid residue 22 to 81. Secondary structure predictions show that it is probably built from its N-terminal end of about 20 amino acid residues folded in random coil followed by an ␣-helix stretch of about 39 amino acids. This domain contains three of the four stathmin phosphorylation sites observed in vitro and in vivo (Ser-25, -38, and -63). Domains II and III cover roughly the two halves of a predicted ␣-helix from amino acid residues 95-113 and 113-128, respectively. Finally domain IV covers the Cterminal stretch of stathmin beyond amino acid 128, a region predicted as random coil. The N-terminal end of stathmin (residues 2-21) was not considered as a stable structured domain since it contains two pairs of potential Lys and Glu sites, and each was cut at 1:32,000.
Hydrodynamic Properties of Stathmin Proteolytic Fragments Generated for Tubulin Interaction-In order to generate both a great diversity and reasonable amounts of stathmin fragments, stathmin was subjected to Lys-C or Glu-C digestion using en- zyme concentrations (1:4000) that lead to advanced but not complete proteolysis. After digestion, stathmin fragments were first chromatographed on a size exclusion column to assess their hydrodynamic parameters. In a control run, we found that the peak of intact recombinant stathmin elutes with a 39-Å Stokes radius that corresponds to an approximately 70-kDa globular protein, as described previously (16). This stathmin feature is currently attributed to the fact that stathmin is an asymmetrical shaped monomer (39) rather than to its existence as a dimer or multimer. Western blot analyses (Fig. 5A) revealed that stathmin fragments generated during proteolysis are eluted in an ordered fashion with a continuum decrease of Stokes radii (for Stokes radius evaluation of the various fragments, see "Experimental Procedures") and no abrupt step. Identification of the fragments was achieved by MALDI-TOF MS. By using this technique, we noticed that the continuous decrease of Stokes radii was indeed proportional to fragment size reduction (Fig. 5B). These results suggested that the stathmin proteolytic fragments most probably retain their native structure. In addition, this indicated that stathmin fragments migrated free of any residual small peptide and that they do not interact either with each other or with intact stathmin. All these results further support the conclusion that stathmin migrates as a monomer peak. Stokes radii distribution ranged from 39 Å for stathmin to about 28 Å for the smallest detectable fragments that were about 60 -70 amino acids in length (fragments 14 -80, 14 -85, 81-143, and 86 -149). Again, it is worthy of note that a Stokes radius of 28 Å corresponds to a globular protein of about 34 kDa, when the corresponding stathmin fragments have an actual molecular mass of 7-10 kDa. This clearly showed that the stathmin asymmetry trait is retained even by the smallest fragments analyzed.
Delimitation of Stathmin Interacting Region with Tubulin-When the stathmin proteolysis products were mixed to interact with tubulin prior to chromatography, we observed a multimodal shift of the tubulin peak that ranged from 12.5 ml (tubulin control elution volume) to 10 Ϯ 0.1 ml (Fig. 6). The relative importance and the intensity of the shifted peaks were only dependent on the completion of stathmin digestion. The further stathmin was digested the less the tubulin peak was displaced. Fig. 6A shows a representative elution profile obtained after addition of an 80-min Lys-C (1:12,000) stathmin digest to tubulin. Western blot analyses of fractions from such runs (Fig.  6B) showed that stathmin or some of its fragments were present from the leading edge of the shifted peak to the end of the peak (elution was monitored at 280 nm, a wavelength where only tubulin absorbs due to the absence of aromatic residues in stathmin). Stathmin fragments displaced in this assay (reduction of their elution volume) were identified by MALDI-TOF MS analysis of the corresponding fractions (Fig. 6C). These polypeptides presented a reduction of their elution volume and correspond to fractions 2-8.
Stathmin fragments were classified according to their shift in the presence of tubulin, and a shifted fragment was considered to interact with tubulin. Fig. 7 summarizes the results obtained with Lys-C and Glu-C digestions of stathmin using this assay for different proteolysis times. This classification shows that the stathmin potential for interaction with tubulin may be lost in two ways as follows: first by an important loss of its N-terminal region as seen with the 50 -149, 63-149, or 81-149 fragments and second by the cleavage of its C-terminal region as revealed by the 2-113, 2-109, or 2-85 fragments. On the other hand, it appears that fragments still interacting with tubulin may have a limited N-terminal shortening as seen with the 31-149 and the 42-149 fragments or a partial deletion of the C-terminal region as observed with the 2-126 and 2-135 fragments. These observations allow us to propose a minimal consensus stathmin fragment necessary for tubulin interaction. It is composed of a core region from amino acids 42 to 126 which must have an N-or a C-terminal extension. The minimal limits of these extensions are not definitively positioned, but they should be between amino acids 14 and 42 in the Nterminal region and cannot be predicted for the C-terminal region.
The Stathmin Core with an Extension on Its N-or C-side Is Necessary for Inhibition of Tubulin Polymerization-Stathmin interaction with tubulin has been shown to inhibit tubulin polymerization into microtubules. To assess the functional value of the delimited stathmin regions for interaction with tubulin, we examined the effects on tubulin polymerization of three recombinant stathmin fragments representative of either the core region (fragment 44 -125), the core region extended on its N-side (fragment 12-125), or on its C-side (fragment 44 -138), and we compared their efficiency to that of wild type stathmin (Fig. 8). In the presence of 2.5 M stathmin, spontaneous polymerization of tubulin into microtubules was inhib-  149)) region of stathmin. For each fraction, the elution volume and the corresponding Stokes radius (R S ) is indicated. Proteolysis generated both a wide variety and detectable amounts of stathmin fragments. B, the Stokes radius of stathmin was measured precisely in a control run (f), whereas Stokes radii of stathmin fragments (q) were estimated as described under "Experimental Procedures" using MALDI-TOF MS (Horizontal bars reflect the uncertainty in this determination). Stokes radii decreased regularly with stathmin fragment length from 39 Å for stathmin (f) to about 28 Å for the smallest fragment detected in fraction 16. ited at all tubulin concentrations leading to a corresponding shift of the critical concentration plot parallel to the control plot. The apparent critical concentration shifted from about 4 to 9 M, consistently with the formation of a T2S complex as previously reported (26). Under the same conditions, 2.5 M fragments 12-125 and 44 -138 had similar effects on the steady state microtubule levels, whereas 2.5 M fragment 44 -125 did not inhibit microtubule assembly at any tubulin concentration tested (Fig. 8). The effects of the recombinant stathmin fragments tested is thus in good agreement with the stathmin interaction experiments, as the extended cores are efficient to inhibit tubulin polymerization, whereas the core itself is not. DISCUSSION In order to progress in the understanding of the molecular domain organization of stathmin as well as the stathmin-tubulin interaction and its regulation, we investigated the structural properties of stathmin in solution by means of limited proteolysis, and we proposed an approach to solve the stathmin interaction regions with tubulin.
Stathmin Displays Several Structured Domains in Solution-Proteolytic cleavage is governed by solvent accessibility and protein flexibility. Proteolytic protection is conferred to regions of the protein that are either buried within a rigid structure or interacting with other parts of the molecule. In contrast, proteolytic sensitivity is localized in regions that are solvent-accessible, unstructured, or flexible (40 -44). Some structural information may thus be deduced from determination of protection against enzyme proteolysis. By using this method, we have found that stathmin displays four structured domains, and we have shown that the cleavage accessibility map fits reasonably with secondary structure prediction given by several methods. Domain I (residues 22-81) is schematically divided in two types of predicted secondary structures as follows: random coil from residues 22 to 45 (domain Ia) and ␣-helix from residues 45 to 81 (domain Ib). Its resistance to limited proteolysis suggests that its two halves may be folded over each other to confer this self-protection. Another possibility could be that part Ia is compact with no accessible cleavage site between it and part Ib. Interestingly, domain I harbors three of the four stathmin phosphorylation sites, whose phosphorylation status might lead to major changes in stathmin folding. Domains II (residues 95-113) and III (residues 113-126) fit reasonably within the limits of the second predicted ␣-helical region of stathmin. Domain IV (from residues 128 to 149) has a high content of Lys-C and Glu-C potential cleavage sites. Its resistance toward proteolysis was ascertained by the observation that it behaved as a proteolysis end product isolated by reverse phase HPLC. Furthermore, the limits of this domain are identical to those of the region corresponding to exon V of stathmin (45), which reinforces its individuality. FIG. 6. Interaction assay for the elucidation of stathmin interaction regions with tubulin. A, stathmin fragments generated in the same conditions as for Fig. 5A were incubated with tubulin before being chromatographed on a size exclusion column. Under these conditions, a multimodal shift of the tubulin peak was observed from 12.5 (tubulin control elution volume) to about 10 ml (leading edge of the shifted peak). B, Western blot analyses of the tubulin peak showed that stathmin and some of its proteolytic products eluted with tubulin, whereas others retained their control elution volumes. Fractions 2-8 were considered to contain displaced stathmin fragments. C, identification of stathmin fragments that were displaced in the presence of tubulin was achieved using MALDI-TOF MS.
Stathmin Folding Hypotheses-Delimitation of structured stathmin domains gives a new insight of stathmin folding and helpful data for the construction of stathmin folding models. Fig. 9 presents two stathmin folding hypotheses matching the existence of these domains. We have imagined that stathmin could adopt schematically an extended or a hairpin turn conformation that in both cases agrees with stathmin being an asymmetrically shaped monomer (16,39). In the two hypotheses, we propose either that the two halves of domain I are packed over each other or that the first half presents a compact structure similar to that proposed for domain IV. In hypothesis A, domains II and III are presented end to end with domain I, whereas in hypothesis B, domains II and III are bent back toward the ␣-helix of domain I in an antiparallel fashion. Computer analysis of stathmin sequence revealed the existence of an internal repeated sequence with 40% identity between amino acid residues 48 -82 and 99 -134 (46). Interestingly, both regions have a high probability for coiled-coil formation that may be involved in internal interaction as represented in hypothesis B or for stathmin partner recognition. Finally, it is known that phosphorylation may regulate the secondary, tertiary, and even quaternary structure of proteins (47,48). Thus we propose that phosphorylation of domain I may promote localized changes in the folding of this domain, which may modify the stathmin ability to interact with protein partners such as tubulin.
A Stathmin Core Extended on Either Its N-or C-side Is Required for Interaction with Tubulin-In order to understand better the molecular mechanisms by which stathmin interacts with tubulin, we examined which of the smallest stathmin fragments can retain the capacity to interact with tubulin. The strategy used to solve the stathmin interaction regions with tubulin combined proteolysis of stathmin, selection of fragments that retain affinity for tubulin by size exclusion chromatography, and their identification by MALDI-TOF mass spectrometry.
In the present work we have found that a minimal stathmin consensus region from amino acid residues 42 to 126, referred to as the "core," is necessary but not sufficient for interaction with tubulin. Interestingly this core corresponds to the entire predicted ␣-helix of domains I, II, and III, previously predicted to be the core of the interaction domain of stathmin with its functional targets and/or partners (10). For its interaction with tubulin, we have found that this core must possess an extension at least of its N-or C-terminal region, the lengths of which are not yet precisely determined. One possibility is that to interact significantly with tubulin, the core needs to be extended with a few additional amino acids on either or both extremities. A more seducing hypothesis could be that tubulin binding to stathmin may occur in two different ways, an "N way" and a "C way," in which either an N-or a C-terminal extension of the core would allow binding of stathmin to tubulin. The N way and the C way may thus reflect the existence on stathmin of two binding sites for tubulin, which could be related to the existence of a stathmin internal repeat (46) and which could account for the stoichiometry of the stathmintubulin complex of one stathmin molecule for two tubulins (␣␤-heterodimers) (16,26). It should be noted that each of the folding hypotheses presented above displays a good accessibility of the core region for the interaction with tubulin, and each is compatible with the binding of two tubulin dimers per stathmin. The necessity of either an additional N-or C-terminal extension for the binding of the core to tubulin may also indicate the existence of a symmetry in the folding of part Ia and domain IV. Although we were not able in the present work to access directly the stathmin fragment to tubulin stoichiometry for all the interacting fragments, results obtained with specific stathmin fragments on tubulin polymerization suggest that it FIG. 7. Stathmin minimal consensus region for interaction with tubulin. A, using a size exclusion chromatography interaction assay, stathmin fragments, generated either with Lys-C or Glu-C, were classified in two categories relative to their capacity to interact with tubulin. Fragments that presented a similar elution volume in the presence or absence of tubulin are displayed in the "no interaction" group. Those that had a clear diminution of their elution volume in the presence of tubulin were classified in the "interaction" group. B, by comparing the stathmin fragment length in the two groups, we deduced the existence of a stathmin core that is necessary for interaction with tubulin, as it is common to all fragments of the interaction group. C, tubulin interaction is predicted for stathmin fragments that possess a "stathmin core" together with an N-and/or C-terminal extension, the minimal lengths of which are not exactly determined. is as in the case of stathmin itself, of one fragment molecule for two tubulins (see below).
The Potential of Interaction of a Stathmin Fragment Is of Predictive Value for Inhibition of Microtubule Assembly-To evaluate the functional relevance of the stathmin-tubulin interaction results, we studied the effects on tubulin polymerization of three specific stathmin fragments representative of either the stathmin core, the N-way or the C-way. As expected, polymerization inhibition of tubulin was observed with the two fragments fitting the tubulin interaction requirements (fragments 12-125 and 44 -138), whereas no effect on polymerization was noted with the fragment representative of the sole stathmin core (fragment 44 -125). These observations suggest that the interaction constrains deduced from the present study should help to predict the effect of various stathmin fragments on tubulin polymerization. This hypothesis is in fact in line with previous results obtained by other investigators as follows: Marklund et al. (19) have shown that deletion of stathmin amino acid residues 4 -55 resulted in a loss of its in vivo microtubule depolymerization activity, and Howell et al. (27) have shown that the ⌬100 -147 mutant has lost its in vitro sequestering activity, whereas the ⌬5-25 still retained it. Finally the coherence between results from our interaction and functional studies reinforces the idea that stathmin inhibits tubulin polymerization mainly by tubulin sequestration. In particular the prerequisite of a tubulin interaction potential for a stathmin fragment to inhibit microtubule assembly argues against the idea of a tubulin-directed regulatory activity of stathmin distinct from tubulin complex formation as proposed by Larsson et al. (49).
In conclusion, the delimitation of structured domains in stathmin should help to understand its interactions with protein partners and the regulatory role of phosphorylation in these processes. Furthermore, it should serve to optimize protein constructs for detailed structural analysis using x-ray diffraction or NMR spectroscopy. The identification of minimal stathmin regions still able to interact with tubulin will provide the basis to understand some of the functional properties of stathmin. It also represents a necessary step to design molecular tools targeted to disturb the stathmin-tubulin equilibrium that could in turn alter some of the tubulin-dependent processes, notably during the cell cycle, all these being under progress. Finally, we speculate that similar regions may be relevant in the other members of the stathmin family for their interaction with tubulin. FIG. 9. Hypotheses about stathmin folding. A schematic diagram of stathmin derived from domain delimitation by limited proteolysis combined with secondary structure predictions is used to present alternative hypotheses about stathmin domain folding. Predicted ␣-helices are represented as cylinders. The four stathmin structured domains (see Fig. 4) are indicated in roman numerals, and cleavage positions observed with limited proteolysis are numbered. In hypothesis A, domains II and III are presented in line with domain I, whereas in hypothesis B, domains II and III are bent back toward the domain I ␣-helix in an antiparallel manner. Regarding domain I folding, its protection toward proteolysis may result from the packing of its two halves (Ia and Ib) over each other or from the presence of a compact structure of region Ia symmetrical to that predicted for domain IV (light gray drawing).