Control of Intrinsically Disordered Stathmin by Multisite Phosphorylation*

Stathmin is an intrinsically disordered protein implicated in the regulation of microtubule dynamics and in the development of cancer. The microtubule destabilizing activity of stathmin is down-regulated by phosphorylation of four serine residues, Ser16, Ser25, Ser38, and Ser63. Here we have used calorimetric and spectroscopic methods, including nuclear magnetic resonance to analyze the properties of seven stathmin phosphoisoforms to bind tubulin and inhibit microtubule formation. We found that stathmin phosphorylation results in a substantial loss in hydration entropy upon tubulin-stathmin complex formation. Remarkably, a linear correlation between the free energy change of complex formation and the microtubule inhibition activities of stathmin phosphoisoforms was observed. This finding provides a biophysical basis for understanding the mechanism by which local stathmin activity gradients important for promoting localized microtubule growth are established. We further found that phosphorylation of Ser16 and Ser63 disrupts the formation of a tubulin-interacting β-hairpin and a helical segment, respectively, explaining the dominant role of these residues in regulating cell cycle progression. The insight into the tubulin-stathmin interaction offers a molecular basis for understanding the nature and the factors that control intrinsically disordered protein systems in general.

Intrinsically disordered proteins have gained enormous interest not only because they are recognized to play key roles in many central cellular processes including cell cycle control, signal transduction, and transcriptional regulation but also because of their particular importance for cancer development and protein deposition diseases (1)(2)(3)(4). Recent genome data base searches indicated that Ͼ30% of all eukaryotic proteins may be completely or partially disordered (5). This high frequency of occurrence has provoked a change of the paradigm that stable tertiary structure is necessary for protein function.
Intrinsically disordered polypeptide chain segments are primarily involved in molecular recognition and posttranslational modification including phosphorylation (6). Little is known, however, regarding the nature and mechanism of control of protein-protein interactions involving intrinsically disordered proteins. Stathmin is a key regulator of microtubule dynamics that lacks a stable three-dimensional structure (7,8). The soluble cytoplasmic protein destabilizes microtubules by binding tubulin dimers (7, 9 -12) and stimulating catastrophes (referred to as the transition of microtubule growth to shortening) (13)(14)(15), playing a central role for cell proliferation, cell migration, and mitotic spindle formation (reviewed by Refs. 16 -20). Interestingly, stathmin is expressed in high amounts in a wide variety of human malignancies, and its overexpression correlates with increased cell motility and invasion of human sarcomas in vivo (21). Moreover, stathmin is important in regulating innate and learned fear and is thus implicated in anxiety states of mental disorders (22). These findings underscore the crucial role of stathmin in a wide variety of central microtubule-dependent cellular processes.
Characteristic of intrinsically disordered proteins, stathmin is devoid of stable tertiary structure in isolation (7,8,23,24), whereas its N-terminal moiety adopts little regular secondary structure the C-terminal domain populates an ensemble of transient helical conformations (see Fig. 1A). Upon binding of stathmin to two head-to-tail aligned ␣/␤tubulin heterodimers, the N terminus folds into a ␤-hairpin, and the C-terminal helical domain becomes stabilized (11,12,25). The curved and capped structure of the ternary tubulin-stathmin complex (see Fig.  1B) provides a structural basis for understanding how stathmin family proteins destabilize microtubules.
In vivo, the activity of stathmin is down-regulated by posttranslational phosphorylation in response to a number of signals on four serine residues, Ser 16 , Ser 25 , Ser 38 , and Ser 63 (26 -29). In mitotic cells, for example, phosphorylation by an unknown kinase-phosphatase system allows creating local stathmin activity gradients, a process essential for regulating microtubule dynamics and spindle formation (30 -33). Phosphorylation of all four serine residues at the G 2 /M transition occurs sequentially; Ser 25 and Ser 38 are first phosphorylated by Cdk1, with subsequent phosphorylation of Ser 16 and Ser 63 by unknown kinase systems (26). Phosphorylation of Ser 16 and Ser 63 strongly down-regulates the microtubule destabilizing activity of stathmin (26 -29;34;35). In contrast, phosphorylation of Ser 25 and Ser 38 has only a moderate effect on down-regulation but is a prerequisite for allowing phosphorylation of Ser 16 and Ser 63 in vivo (26).
The current knowledge of the tubulin-stathmin interaction provides a unique basis to gain detailed insight into factors regulating intrinsically disordered protein systems. To define how multiple phosphorylation sites control stathmin function, we here have explored seven stathmin phosphoisoforms by using biochemical and biophysical methods.
Recombinant unlabeled and 15 N uniformly labeled stathmin proteins were bacterially expressed, purified, and processed as described (7,8). Specific phosphorylation by protein kinase A (for phosphorylation of Ser 16 and Ser 63 ) and a mixture of MAPK 2 and CdC 2 (for phosphorylation of Ser25 and Ser38) was achieved by incubating the proteins with the respective kinases (2.8, 2.0, and 0.5 units of protein kinase A, MAPK, and/or CdC 2 , respectively, per g of stathmin) in 50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 5 mM EGTA, 2 mM dithiothreitol, 500 M ATP for 6 -8 h at 30°C. Product formation was assessed by native-PAGE. Kinases were heat-inactivated at 75°C for 10 min. Phosphoisoforms were purified to high homogeneity (Ͼ93%) by anion exchange chromatography. Highly pure bovine brain GTP-tubulin was obtained from Cytoskeleton Inc.
The identities of stathmin proteins were assessed by mass spectral analyses. Concentrations of protein samples were determined by the Advanced Protein Assay (Cytoskeleton Inc.).
Tubulin-polymerization Assay-In vitro polymerization of tubulin was performed according to Ref. 26. Briefly, 4 M tubulin in G buffer (80 mM PIPES-KOH, pH 6.8, 1 mM MgCl 2 , 1 mM EGTA, 1 mM GTP) supplemented with 4 mM MgCl 2 was preincubated with 4 M of stathmin (in the same buffer) for 30 min at room temperature in a total reaction volume of 100 l. Polymerization was initiated by adding 1 l of a 400 M taxol stock solution and incubating at 37°C for 1.5 h. Microtubules were separated from tubulin-stathmin oligomers by sedimentation at 300,000 ϫ g for 15 min at 37°C in an Optima TLX ultracentrifuge (Beckman Instruments). The protein contents of supernatants and pellets were analyzed with the bicinchoninic acid protein assay reagent (Pierce).
Protein samples (0.35 mg/ml) for CD were in phosphate-buffered saline (10 mM sodium phosphate, pH 7.4, 150 mM NaCl). Far-ultraviolet CD spectra and thermal unfolding profiles were recorded on a Jasco J-810 spectropolarimeter (Jasco Inc.) equipped with a temperature-controlled quartz cell of 0.1-cm path length. A ramping rate of 1°C⅐min Ϫ1 was used to record the thermal unfolding profiles. 15 N, 1 H-HSQC NMR experiments of 19 mg/ml protein samples in G buffer were carried out at 25°C on a Varian UnityPlus 600 spectrometer operating at 600 MHz proton frequency. For resonance assignment, three-dimensional 15 N-edited TOCSY-HSQC using a clean DIPSI-2 mixing sequence and three-dimensional 15 N-HSQC-TOCSY-NOESY-HSQC were recorded.
Modeling-The 3.5 Å resolution x-ray crystal structure of the tubulin-RB3 stathmin-like domain (SLD) complex (PDB entry 1SA0), and the PyMol (De-Lano Scientific LLC, San Carlos, CA) and Moloc (40) software packages were used for modeling studies. The accuracy of the conformations of key residue side chains was verified by inspecting the electron density map for PDB entry 1SA0. Solvent-accessible area calculations were carried out with the program NACCESS V2.1.1 with the default set of atomic radii and parameters. The stathmin residues 29 -45 were taken from PDB entry 1SA1. Tyr 63 of the RB3-SLD was replaced by a serine to reflect the human stathmin sequence. Stathmin and RB3-SLD share 72% sequence identity, and 82 and 92% of all tubulin-contacting RB3-SLD residues are invariant or similar, respectively, in stathmin.

RESULTS AND DISCUSSION
For the following studies, milligram amounts of pure and specific single (denoted p16 and p63), double (denoted p25,38 and p16,63), triple (denoted p16,25,38 and p25,38,63), and quadruple (denoted p16,25,38,63) stathmin phosphoisoforms were produced ( Fig. 2A). The activities of the proteins were assessed in vitro by a microtubule polymerization assay. Under the experimental conditions applied the efficiency to inhibit microtubule formation decreased from 90 to 0% with a differential combination of stathmin phosphorylation (Fig. 2B). These findings are consistent with stathmin sequestering tubulin dimers into assembly-incompetent complexes (9, 10), a process controlled by phosphorylation. In agreement with the microtubule polymerization data obtained in vivo (26 -29), phosphorylation of Ser 16 and Ser 63 contributes most to stathmin inactivation.
The thermodynamics of the tubulin-stathmin interaction was assessed by isothermal titration calorimetry (supplemental Fig. 1). Between 6 and 25°C, stathmin binds two tubulin subunits and all thermodynamic parameters are thus referred to the ternary tubulin-stathmin T 2 S complex (supplemental Table 1). As shown in Fig. 2C, the binding reaction is predicted to be driven by both enthalpy and entropy at physiological temperatures. The large apparent negative heat capacity change of ⌬C p,T 2 S,obs ϭ Ϫ1504 Ϯ 206 cal mol Ϫ1 K Ϫ1 suggests that the hydrophobic effect (removal of non-polar surface from water) promotes T 2 S complex formation (36). As a consequence, the apparent entropic and enthalpic contributions to the free energy change of complex formation vary with temperature in a linear and nearly parallel manner, changing sign at ϳ28 and ϳ44°C, respectively.
Empirical studies on proteins showed that the removal of hydrophobic and polar surface from water contributes Ϫ45 and 26 cal mol Ϫ1 /100 Å 2 , respectively, to ⌬C p (37). We have calculated the total buried hydrophobic and polar surface areas from the 3.5 Å resolution x-ray crystal structure of the ternary complex formed between the stathmin homologue RB3 and tubulin (denoted T 2 R, see Ref. 11) as 5316 and 3074 Å 2 , respectively. Accordingly, the estimated heat capacity change ⌬C p,T 2 S,cal amounts Ϫ1593 cal mol Ϫ1 K Ϫ1 , in good agreement with the experimentally obtained value for the tubulin-stathmin complex (see above). These findings are consistent with a mechanism in which dehydration of the protein-protein interface is the major driving force of T 2 S complex formation.
Isothermal titration calorimetry showed that each stathmin phosphoisoform bound two tubulin dimers as observed for unmodified stathmin (Table 1 and supplemental Fig. 1). The equilibrium dissociation constant, K D,T 2 S , of T 2 S for unmodified stathmin is 5.3 ϫ 10 Ϫ13 M 2 under the conditions applied. This relatively high value underscores the dynamic nature of the tubulin-stathmin equilibrium observed in cellu-

TABLE 1 Thermodynamic binding parameters derived from the titration of tubulin with stathmin variants and phosphoisoforms
Isothermal titration calorimetry measurements were carried out at 6°C in 80 mM PIPES, pH 6.8, 1 mM EGTA, 1 mM MgCl 2 , 1 mM GTP. All thermodynamic parameters are referred to the ternary T 2 S complex. ND, not determined. could not be evaluated because binding was too weak. For all measured stathmin phosphoisoforms, a reduced binding entropy that is partially offset by an increased binding enthalpy is observed (Fig. 2D).

K D,T 2 S ⌬H ⌻⌬S ⌬G
The finding that phosphorylation of Ser 16 and Ser 63 contributes most to the reduced binding of stathmin explains the dominant role of these residues for in vivo inactivation (26 -29). The moderate effect obtained with phosphorylation of Ser 25 and Ser 38 correlates with their location in the proline/serine-rich loop segment of stathmin (Fig. 1), which is poorly ordered in the T 2 R complex (11,12). The local perturbation caused by phosphorylated Ser 25 and Ser 38 , however, is expected to facilitate phosphorylation of the adjacent Ser 16 and Ser 63 residues as suggested from in vivo data (26).
Remarkably, a linear correlation between the free energy of T 2 S complex formation and the tubulin polymerization inhibition activities of stathmin phosphoisoforms is observed (Fig. 3). In agreement with cell biological data (26 -29), this correlation suggests that already moderate changes in the tubulin-stathmin equilibrium significantly influence microtubule dynamics and, as a consequence, microtubule function in a particular in vivo situation. This conclusion provides a biophysical basis for understanding how spatial gradients of differentially inactive stathmin molecules promote localized microtubule growth, a process essential for, e.g. mitotic spindle assembly (30 -33).
The secondary structures and thermal stabilities of stathmin phosphoisoforms were probed by CD spectroscopy. CD recorded at a low temperature from unmodified stathmin revealed a spectrum with ϳ45% helical content (Fig. 4A). Characteristic of proteins lacking stable tertiary structure, a fully reversible, broad unfolding transition to a random coil structure was observed upon thermal denaturation (Fig. 4B). The phosphorylation of Ser 63 reduces both the helical content (20 -30% in the 5-30°C temperature range) and the thermal stability of stathmin. In contrast, phosphorylation of Ser 16 , Ser 25 , and Ser 38 affects only moderately its secondary structure throughout the 5-80°C temperature range.
Our biophysical and biochemical studies in combination with structural information on the ternary T 2 R complex provide a strong basis for understanding the mechanism underlying Ser 63 and Ser 16 phosphorylation. A correlation between secondary structure and markedly reduced tubulin-binding affinity is apparent with stathmin isoforms phosphorylated at Ser 63 . Remarkably, in T 2 S Ser 63 (tyrosine in T 2 R) projects into the solvent and is not involved at the protein-protein interface (Fig. 4C). However, the residue is embedded within the major helix nucleation site of stathmin, Glu 55 -Ala 73 (23), which drives helix formation of the C-ter-  minal domain in isolation (Fig. 1). NMR experiments of peptides encompassing Glu 55 -Ala 73 demonstrated that the presence of the phosphoryl group on Ser 63 introduces a kink in the helical backbone leading to the dispersion of the peptide sequence Glu 55 -Arg 61 (23). The driving force of this distortion can be explained by the strong propensity of phosphoserine to interact with the main chain (38). Mutating Ser 63 to glutamic acid (denoted S63E) only moderately affects the secondary structure, thermal stability, and tubulin binding affinity of stathmin (Table 1 and supplemental Fig. 2), underscoring the unique properties of the bulky dianionic phosphoryl group to disrupt the helical conformation of the helix nucleation site. This local effect explains the reduced tubulin binding activities of stathmin isoforms phosphorylated at Ser 63 . The phosphoryl group hinders the alignment of residues Lys 53 , Leu 54 , Ala 57 , Arg 60 , and Arg 61 , which tightly interact with the ␣1-tubulin monomer (Fig. 4C).
As shown in Fig. 5A, in T 2 R Ser 16 is located within the tight turn connecting the two ␤-strands of the ␤-hairpin. The residue is stabilized by an intermolecular hydrogen bond formed between its main chain oxygen atom and the side chain of ␣1Asn356. As a consequence, Ser 16 is oriented toward the ␣1-tubulin surface, and an introduction of a phosphoryl group is expected to result in a steric clash of its bulky phosphorylated side chain. This prediction was tested by NMR experiments. The 15 N, 1 H-HSQC measurements of 15 N-labeled stathmin and p16 proteins revealed spectra with limited chemical shift dispersion (Fig. 5B), characteristic for intrinsically disordered proteins populating an ensemble of helical secondary structures. Comparison of these two spectra, however, reveals three prominent differences. One new peak (at 7.55/118.2 ppm), which originates from an arginine side chain forming a hydrogen bond (most probably Arg 14 ), and two prominent N-H backbone reso-nance shifts were found in the 15 N-p16 HSQC spectrum. The first peak shift (from 8.18/115.1 to 8.79/117.5 ppm) is in a spectral region typical for serine residues and is likely to originate from Ser 16 . The second peak shift (from 8.25/127.5 to 8.45/127.1 ppm) most probably stems from a residue close in sequence to Ser 16 .
Upon the addition of unlabeled tubulin to [ 15 N]stathmin, WT or S25A,S38A,S63A, all except the last eight C-terminal stathmin residues broaden beyond distinction in the HSQC spectrum because of the large ϳ200-kDa size of the complex. This finding demonstrates that most stathmin residues become tightly bound in T 2 S (Fig. 5C, black). In contrast, ϳ40 strong and several weaker resonances are visible in the spectrum of the complex formed with 15 N-p16 (Fig. 5C, red). Most of these resonances are broad showing limited chemical shift dispersion, characteristic for residues that are in rapid exchange between weakly bound and unbound random coil states. Some peaks most likely originate from the N-terminal domain of stathmin. First, both resonances shifting upon phosphorylation of unbound stathmin and which most likely stem from Ser 16 and Arg 14 (see above) are visible in the HSQC spectrum of the complex formed with [ 15 N]p16, indicating that they are not tightly bound to tubulin. Second, a stathmin fragment lacking the first 40 N-terminal residues (denoted ⌬N) showed similar tubulin-binding properties as p16 (Table 1). Third, substituting glutamic acid for Ser 16 (denoted S16E) partially mimics the effect of phosphorylation on T 2 S complex formation (Table 1 and supplemental Fig. 2). This finding is consistent with the hypothesis that steric clash with ␣1-tubulin and intramolecular interaction of the bulky phosphoryl group with the backbone (38) and/or Arg 14 side chain (25) is the underlying mechanism. These data demonstrate that phosphorylation of Ser 16 strongly impairs binding of the ␤-hairpin to ␣1-tubulin and explain the reduced tubulin-  (Fig. 4C). B, superposition of 15 N, 1 H HSQC spectra of 15 N-WT (black) and [ 15 N]p16 (red). Spectral changes that most likely originate from Ser 16 and adjacent residues are indicated by blue arrows. The new signal at 7.55/118.2 ppm that is because of an arginine side chain forming a hydrogen bond (most likely Arg 14 ) is indicated by a blue circle. The peak is invisible in 15 N-WT because of fast solvent exchange, but visible in [ 15 N]p16, most likely because of interaction of the arginine side chain with the phosphorylated Ser 16 residue, which slows down its solvent exchange rate. C, 15 N, 1 H-HSQC spectra of tubulin-bound 15 N-WT (in black with assignments) and 15 N-p16 (in red). The two resonances that have shifted upon phosphorylation of unbound stathmin (see A) and most likely originate from Ser 16 and an adjacent residue are highlighted by circles.
binding activities of stathmin isoforms phosphorylated at Ser 16 . As Ser 16 is conserved throughout stathmin family proteins (16), this mechanism is expected to apply to all stathmin homologues.
From a thermodynamic point of view the down-regulating effect of stathmin phosphorylation can be explained by the substantial loss in hydration entropy upon T 2 S complex formation, which is larger than the gain in enthalpy of the system. Consistent with this conclusion, phosphorylation of Ser 16 and Ser 63 disrupts the formation of the ␤-hairpin and the helix nucleation site, respectively, impairing binding of these two key secondary structure elements to ␣1-tubulin and leading to the exposure of non-polar surface to water. This finding opens an avenue to design strategies to interfere with abnormal microtubule dynamics observed in many human malignancies displaying high levels of stathmin expression (16). Targeting the ␤-hairpin and/or helix nucleation site of stathmin is expected to perturb the dynamic equilibrium of the microtubule filament system possibly inhibiting tumor invasion in vivo (21) or even inducing apoptosis of transformed cells (39).
Taken together, our findings provide new mechanistic insight into the control of stathmin function by multisite phosphorylation. They further offer a molecular basis for understanding the nature and modes of regulation of intrinsically disordered protein systems in general.