Coupling of Folding and Binding of Thymosin (cid:1) 4 upon Interaction with Monomeric Actin Monitored by Nuclear Magnetic Resonance*

Thymosin (cid:1) 4 is a major actin-sequestering protein, yet the structural basis for its biological function is still unknown. This study provides insight regarding the way this 43-amino acid peptide, mostly unstructured in solution, binds to monomeric actin and prevents its assembly in filaments. We show here that the whole backbone of thymosin (cid:1) 4 is highly affected upon binding to G-actin. The assignment of all amide protons and nitrogens of thymosin in the bound state, obtained using a combination of NMR experiments and selective labelings, shows that thymosin folds completely upon binding and displays a central extended region flanked by two N- and C-terminal helices. The cleavage of actin by subtilisin in the DNase I binding loop does not modify the structure of thymosin (cid:1) 4 in the complex, showing that the backbone of the peptide is not in close proximity

Cell locomotion or changes in cell shape are mediated by the rapid polymerization of actin in response to signaling. The source for this massive increase in F-actin is provided by a large reservoir of unassembled, "sequestered" actin, which is kept monomeric by interaction with G-actin-binding proteins (1). Thymosin ␤4 (T␤4), 1 first identified as the most abundant (300 M) actin-sequestering protein in human blood platelets (2,3) and neutrophils (4), was eventually shown, together with variants of the ␤-thymosin family, to bind G-actin in all vertebrates (5)(6)(7)(8) and some invertebrates (9). In contrast to profilin, another G-actin-binding protein, which can either sequester actin when filament barbed ends are blocked by capping proteins or participate in barbed end assembly when barbed ends are uncapped (10), T␤4 acts as a simple sequestering protein that prevents G-actin association to both ends of actin filaments. It reacts passively to the changes in critical concentration for actin assembly while amplifying them. T␤4 essentially binds G-actin with a high specificity for the ATP-bound form (11), an affinity in the 10 6 M Ϫ1 range (3,11), and relatively slow association-dissociation kinetics that are indicative of a structural change of the complex after rapid equilibrium binding (12). Finally, in binding to G-actin, T␤4 slows down metal ion/nucleotide dissociation (13,14). This protein is composed of a single WH2 domain, which has recently been identified in 37 other proteins belonging to various organisms, and was shown to be an evolutionarily conserved actin-monomer binding motif (15). Interestingly, different WH2 domains were found to perform different functions in the regulation of actin monomeric concentration and actin assembly processes. Whereas ␤-thymosins inhibit the assembly of actin filaments, the other WH2 domains characterized thus far display a profiling-like activity, promoting the assembly of actin filaments at the barbed end (16 -20). A fine structural and dynamic characterization of WH2 domain interfaces with actin is therefore essential to understand the functional variability of WH2 domains.
Whereas the structure and interface with actin of other G-actin binding proteins such as DNase I (21), gelsolin segment-1 (22), and profilin (23) are well known from crystallographic studies, T␤4 and, more generally, WH2 domains remain elusive. NMR studies have shown that T␤4 was mostly unstructured in aqueous solution, except for a short helical region (residues 5-16) at low temperature (24). Mutagenesis studies indicated that the ␣-helical structure of this region was required for binding to actin (25)(26)(27). Biochemical results based mainly on chemical cross-linking indicated that T␤4 bound G-actin in an extended conformation, with the N-terminal helical region (residues 1-19) making contacts with residues in subdomain 1 of actin, whereas the C-terminal region could be cross-linked to His 40 and Gln 41 of actin in subdomain 2 (12,28,29). According to these results, the ability of T␤4 to interfere with actin-actin contacts involved in filament assembly at both the barbed and pointed ends of the actin monomer was thought to account for its inhibition of polymerization via a simple steric mechanism and also for its competition with profilin and DNase I for binding to G-actin. Recent biochemical and thermodynamic studies of the binding of T␤4 to CaATP-actin and MgATP-actin, based mainly on fluorescence quenching of 5-(2acetylaminoethylamino)naphtalene-5-sulfonate conjugated to Cys 374 of actin upon T␤4 binding and on tritium exchange, also suggested that this small protein changed the conformation and structural dynamics of actin monomers (12).
Our study constitutes the first essential step for the under-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The atomic coordinates and structure factors ( standing of the molecular basis of G-actin sequestration by thymosin ␤4. A number of specificities of the biological system under study render its structural analysis especially challenging for both crystallography and NMR. The tendency of G-actin to self-assemble, even at low ionic strength, at concentrations greater than 50 -70 M, the high flexibility of T␤4, and the limited stability of the T␤4-G-actin complex at high concentrations have hampered crystallization of the complex thus far. These difficulties, combined with the large size of the complex (47 kDa), preclude 1 H, 15 N, 13 C NMR spectroscopy studies of the T␤4-actin complex that require long acquisition periods. Here we show that the assignment of the whole 15 N-1 H N correlation spectra of T␤4 bound to monomeric CaATP-actin and MgATP-actin is successfully performed using an alternative strategy based on appropriately chosen specific labels. Our results demonstrate that the whole T␤4 backbone is tightly bound to monomeric actin and that the structures of the peptide bound to CaATP-actin, MgATP-actin, and subtilin-cleaved actin are identical. The secondary structure of actin-bound thymosin ␤4 is obtained, and novel information is also provided about its fast dynamics using 1 H-15 N heteronuclear NOE experiments. A hypothetical model of the complex is proposed, based on biochemical data and our NMR results.
Thymosin ␤4 and Thymosin ␤10 -Recombinant thymosins ␤4 and ␤10 were bacterially expressed in Escherichia coli using pET3d vectors kindly provided by Dr. Helen Lu Yin (University of Texas Southwestern Medical Center). Bacteria (BL21(DE3) strain) were grown at 28°C in diverse media depending on the labeling: (a) [U- 15  N]AA T␤4 (AA was alternatively Thr, Lys, Leu, and Ser), M9 medium implemented with amino acids (40 mg/liter), except the one to be labeled. 50 mg/liter of the labeled amino acid was added when the optical density was equal to 0.2.
The expression of all T␤4 variants was induced by addition of isopropyl-1-thio-␤-D-galactopyranoside when the optical density of the culture was 0.4. Bacteria were collected 2 h and 30 min later. The bacterial extract was treated with HClO 4 (up to 4%/v). The acid-soluble T␤4 solution was brought to pH 4, loaded on a 4 ϫ 25-cm RP 18 (Merck) column, eluted by 33% n-propanol in H 2 O, and further purified by high pressure liquid chromatography using two consecutive preparative columns (2.5 ϫ 25 cm, Hibar Lichrospher 100 RP-18E; Merck) and a 5-28% acetonitrile gradient.
Selective Labeling of Thymosin ␤4 -The selective labeling strategy was chosen to minimally perturb the culture conditions already used for the previous production of recombinant T␤4. In particular, the same bacteria strain was used. Because "pseudo-rich" conditions were used for bacteria growth, the protein yield was better than for uniform labeling, around 7 mg protein/culture liter.
Two main criteria were important for the choice of the labeled amino acid: (a) first, the frequency and position of the amino acid along the sequence; and (b) second, the possibility to specifically label the amino acid using the BL21(DE3) bacteria strain. This second criteria can be quite restrictive because a number of metabolic leakages and transaminase activities can lead to partial or complete dilution of the labeling for a number of amino acids (33,34). Considering the first criteria, the four amino acids Lys, Leu, Thr, and Ser were particularly interesting.
Lysine is the most frequent amino acid in the sequence (Lys 3 , Lys 11 , Lys 14 , Lys 16 , Lys 18 , Lys 19 , Lys 25 , Lys 31 , and Lys 38 ); it is nicely spread along the whole sequence and is part of the LKKTET fragment (residues 17-22 in T␤4), which is known to be a consensus recognition motif for actin-binding proteins. Threonine and leucine were also selected for their presence in the consensus motif (Thr 20 , Thr 22 , and Thr 33 ; Leu 17 and Leu 28 ). Moreover, the assignment of Leu 28 is particularly difficult due to its sandwiched position between two prolines. Finally, serine was initially selected due to its distribution along the sequence (Ser 1 , Ser 15 , Ser 30 , and Ser 43 ). However, the specific labeling of this residue could not be obtained using the BL21 bacteria strain, and a partial uniform labeling was obtained instead. This is probably due to a significant metabolic leakage from serine to glutamate, which itself leads to a metabolic leakage toward all other amino acids (33). In the case of leucines, leakages were also observed, but to a much smaller extent, and two main peaks corresponding to Leu 17 and Leu 28 were observed in the HSQC spectra of free [ 15 N]Leu T␤4. The most intense secondary peaks correspond to isoleucines Ile 9 and Ile 34 . These were, however, beyond detection in the complex.

NMR Samples and Experiments
NMR Samples-Thymosin NMR samples consisted of 15 N-labeled thymosins at a concentration of 1-2 mM in a 4 mM phosphate buffer (pH 6.9) containing 10 M ATP, 20 M Ca 2ϩ , 10% D 2 O, and 0.01% sodium azide. NMR samples of complexes between actin and thymosins were prepared by mixing 20 ml of 47 M actin with thymosins taken from a 2 mM stock solution up to a 1:1 molar ratio, occasionally with a small excess of thymosins. The proteins were then concentrated up to 350 -500 M complex using a Centriprep 30 device (Amicon) and centrifuged at 3000 rpm for 45 min. The complexes were stored in G buffer containing 5 mM Tris-Cl, 0.1 mM CaCl 2 , 0.2 mM ATP, 0.2 mM dithiothreitol, 5% D 2 O, and 0.01% sodium azide. The final pH value was 6.9.
NMR Experiments-NMR experiments were performed on Bruker Avance 600 MHz and 800 MHz spectrometers equipped with 5-mm triple resonance gradient probes. Quadrature phase detection in the indirectly detected dimension was obtained via States-TPPI mode (35). A Watergate sequence was used in all experiments to achieve water signal suppression (36).
Assignment of 1 H, 15 N, and 13 C resonances of T␤4 free in solution was obtained using standard procedures from two-dimensional 1 H-1 H NOESY and total correlation spectroscopy experiments, three-dimensional 1 H-15 N NOESY-HSQC, and three-dimensional 1 H-15 N total correlation spectroscopy-HSQC experiments and HNCA, HN(CO)CA, HNCO, HN(CA)CO, CBCANH, and CBCA(CO)NH experiments. Mixing time for NOESY experiments was set to 120 ms. Two mixing times of 50 and 80 ms were used for total correlation spectroscopy experiments.
Pulse sequences used to determine 15 N R 1 (N Z ), R 2 (N XY ), and heteronuclear 1 H-15 N NOE values were similar to those described previously (41)(42)(43). R 1 (N Z ) experiments were performed with 10 relaxation delays (0.012, 0.048, 0.072, 0.144, 0.300, 0.420, 0.600, 0.840, 1.200, and 2.400 s). For R 2 (N XY ) experiments, a set of 10 experiments was acquired, with relaxation delays of 0.008, 0.024, 0.040, 0.080, 0.136, 0.200, 0.304, 0.480, 0.640, and 0.800 s. In the two sets of experiments, the points corresponding to different relaxation delays were acquired in an interleaved manner to avoid any bias that could arise from progressive shim degradation. For all experiments, carrier frequencies were set at 117 ppm in the F1 dimension ( 15 N) and 7.6 ppm in the F2 dimension. Spectral widths were 23 ppm in the F1 ( 15 N) dimension and 3.5 ppm in the F2 ( 1 H N ) dimension. Cross-peak intensities were determined from peak heights using the SPARKY peak picking routine (Goddard, T. D., and Kneller, D. G. SPARKY 3 Software, University of California, San Francisco, CA). The relaxation rate constants R 1 (N Z ) and R 2 (N XY ) were obtained from nonlinear fits to mono-exponential functions using the Levenburg-Marquardt algorithm (45). The quality of the fits and error estimations were obtained using established Monte Carlo procedures, with an experimental gaussian error set to 4 times the rms base plane noise level and using 500 synthetic data for each NH vector. 15 N NOE enhancements were obtained as the ratio of the peak heights in the spectra recorded with and without saturation of protons during the relaxation delay.
For 15 N-1 H HSQC experiments carried on T␤4-actin complex, carrier frequencies were set at 117 ppm in the F1 dimension ( 15 N) and 7.6 ppm in the F2 dimension. Spectral widths were 28 ppm in the F1 ( 15 N) dimension and 5 ppm in the F2 ( 1 H N ) dimension. The three-dimensional 1 H-15 N NOESY-HSQC experiment used for the assignment of T␤4 bound to actin was performed with a mixing time equal to 40 ms.

Measurements of the Binding of T␤4 to G-actin and Subtilisin-cleaved Actin
The affinity of T␤4 for actin or subtilisin-cleaved actin (subactin) was derived from the increase in actin tryptophane fluorescence ( exc , 295 nm; em , 350 nm) linked to the binding of T␤4 (46). Experiments were carried out in a Spex spectrofluorimeter at 20°C, using 2 M CaATP-G-actin or subtilisin-cleaved actin in G buffer containing only 20 M ATP to minimize the screen effect.
The relative increase in fluorescence, (F Ϫ F 0 )/(F max Ϫ F 0 ), was expressed as function of the molar fraction of actin in complex with T␤4, (TA)/(A 0 ), according to the following equation.
Alternatively, the affinity of T␤4 for subtilisin-cleaved actin could be derived from its sequestering activity, by measuring the amount of F-actin at steady state in the presence of different concentrations of T␤4 (25,47). The concentration of F-actin, C w , was derived from the fluorescence of pyrenyl-actin. As the T␤4-actin complex formed, C w decreased linearly with the total concentration [T 0 ] of T␤4, according to the following equation: in which [TA] is the concentration of T␤4-actin complex; C w 0 and C w are the mass concentrations of F-actin in the absence and presence of T␤4, respectively; C C is the critical concentration for F-actin assembly; and K T is the equilibrium dissociation constant of the T␤4-actin complex. The values of C C for actin and subtilisin-cleaved actin were independently derived from critical concentration plots. The value of K T was derived from the slope C C /(C C ϩ K T ) of the lines describing the linear decrease of C w with [T 0 ].
Finally, the affinity of T␤4 for actin and subtilisin-cleaved actin could also be derived from the inhibition of nucleotide exchange linked to T␤4 binding. The CaATP-subtilisin-cleaved actin 1:1 complex was prepared by eliminating free ATP from the solution by Dowex-1 treatment (48) and diluted to 2.2 M in G buffer containing no nucleotide, 10 M CaCl 2 (final concentration), and variable amounts of T␤4. At time 0, 5.8 M ⑀ATP was added to the solution, and the rate of ⑀ATP exchange for bound ATP was derived from the increase in fluorescence of ⑀ATP associated to its binding to actin ( exc , 350 nm; em , 410 nm). The process was mono-exponential, and the observed first order rate constant varied with the concentration of T␤4, reflecting the binding of T␤4 to subtilisin-cleaved actin in rapid equilibrium, as follows where k Ϫ and k Ϫ Ј are the rate constants for dissociation of CaATP from subtilisin-cleaved actin and T␤4-subtilisin-cleaved actin complex, respectively; [T] is the concentration of free T␤4; and K T Ј is the equilibrium dissociation constant of the T␤4-subtilisin-cleaved actin complex.

Building of a Model of the Thymosin ␤4-actin G Complex
Software and Hardware-The X-PLOR (49) software was run on a Macintosh G4 PowerBook. The graphical analyses were performed using either MOLMOL (50) run on a G4 PowerBook or Insight (Accelerys) run on an O2 Silicon Graphics work station.
Model Building-The initial actin structure was extracted from the Protein Data Bank entry 1atn.pdb (51). A thymosin initial structure was built manually using the Builder module of Insight software. According to NMR results, a helical geometry was imposed to the Met 6 -Asp 13 and Ile 34 -Ala 40 segments, whereas the Lys 18 -Ser 30 region was constrained to adopt an extended conformation. The and angles of residues 5, 18, 31, and 32 were randomized, and the structure was submitted to a short torsion dynamic to eliminate eventual bad van der Waals interactions. Distances of 30 Ϯ 10, 50 Ϯ 10, and 55 Ϯ 10 Å between residues 3 and 18, 18 and 38, and 3 and 38 were imposed during this first dynamic to obtain an initial structure already compatible with the previously identified cross-links between the peptide and G-actin (12,29). The two files were merged, and the thymosin was aligned and translated such that its main axis lies parallel to the actin plane at a distance of 30 Å above it.
Three sets of constraints were designed to drive the formation of the complex. The first set was introduced to maintain the helical geometry of the thymosin 6 -13 and 34 -40 segments. It was composed of 12 H N -O (1.8 -2.2Å) and N-O (1.8 -3.2Å) distance restraints, enforcing the ␣-helices hydrogen bonds, and of 12 (47 Ϯ 5°) and (57 Ϯ 5°) dihedral angle restraints. The second set was derived from cross-linking experiments. The N⑀ atoms of the thymosin Lys 3 , Lys 18 , and Lys 38 were constrained to be in close vicinity (Ͻ5 Å) of the carboxylic functions of the actin Glu 167 (Lys 3 ), Asp 1 , or Glu 4 (Lys 18 ) and Glu 41 (Lys 38 ). In the case of Lys 18 of thymosin, an ambiguous restraint was used to allow both possibilities. Finally, the last set was designed to impose close packing between the thymosin and the actin. More precisely, each C␣ of the thymosin Thr 20 -Ser 30 segment was constrained to be in proximity (Ͻ9 Å) of an actin C␣ atom through the application of an ambiguous restraint. Similarly, the side chains of Met 6 , Ile 9 , and Phe 12 were constrained to be in contact with the protein by introducing a restraint between their C␤ atoms and the actin C␣ atoms (Ͻ10 Å). These last constraints allowed the positioning of hydrophobic side chains of the N-terminal amphiphilic helix of thymosin toward actin, as shown by previously published mutagenesis data (27).
The system was optimized in three steps. First, thymosin was simply rotated above actin to minimize the distance between the thymosin Lys 3 and Lys 18 and actin Asp 1 , Glu 4 , and Glu 167 . Second, the system was submitted to a torsion dynamic, in presence of the cross-linking-derived and C␤-C␣ restraints, followed by a cartesian dynamic in presence of all restraints. The two dynamics (5000 steps of 1 fs) were realized at a temperature of 5000 K. The structure was finally minimized to rectify the geometry and optimize the van der Waals interactions. The topallhsa and parallhsa parameter and topology files of the X-PLOR software were used in conjunction with an energy function in which the electrostatic term was turned off and the van der Waals interaction was modeled by a simple quadratic repulsion term (except for the final minimization, which was realized with a true van der Waals potential). A square NOE potential was used for the distance restraints of the first set, and a soft potential was used for the others. In both cases, the R-6 mean was used to allow the application of the ambiguous restraints.

RESULTS
Thymosin ␤4 Is Mainly Unfolded in Solution at 25°C, Except for a Small Helical Content in Its N-terminal Region-The assignment of the proton, nitrogen, and carbon resonances of T␤4 free in solution were obtained using standard procedures. The assignments of the proton resonances of T␤4 in solution at 2°C are in agreement with those reported previously (24). The HSQC spectrum of T␤4 in solution is typical of an unfolded protein, with a very narrow range of amide proton resonances (Fig. 1A). However, structural and dynamic data clearly show the partial formation of secondary structures in the N-terminal and C-terminal segments of the peptide. At 2°C, the presence of NH-NH(i, iϩ1), H␣-NH(i, iϩ2), H␣-NH(i, iϩ3), H␣-NH(i, iϩ4), and H␣-H␤(i, iϩ3) NOE correlations for all residues along segment 5-16 indicates that this part of the protein folds as a ␣-helix. This conclusion is corroborated by the CSI of H␣ and C␣ nuclei (Fig. 1B) (37)(38)(39). The concomitant presence of strong H␣-NH(i, iϩ1) correlations in this region, however, shows that the helix form is in fast exchange with an extended strand conformation. No NOEs characteristic of the existence of secondary structures were found in any other region of the protein. However, the small negative H␣ and positive C␣ shifts relative to the random coil shifts observed for residues 31-37 suggest that this segment has a weak tendency to fold into an ␣-helix. The tendency of segment 31-37 to form a helix at 2°C in water is also in agreement with the folding of this segment into an ␣-helix in trifluoroethanol (52). The ␣-helical fold of the N-terminal part of T␤4 is anticipated from the analysis of primary sequence. The N-terminal sequence Pro 4 -Asp 5 -Met 6 -Ala 7 -Glu 8 -Ile 9 indeed corresponds to an N-capping box motif: h-(D,N,T,S)-N1-N2-(E,Q)-h, where N stands for any amino acid, and h is a hydrophobic residue (53)(54)(55). This motif is known to initiate the first turns and the propagation of ␣-helices in the C-terminal direction and to strongly stabilize their N-terminal extremity. In contrast with data obtained at 2°C, no NOE correlation indicative of any secondary structure was found at 25°C. The 15 N relaxation data obtained on T␤4 free in solution at 2°C and 25°C reinforces the conclusion obtained from structural data. The low values of heteronuclear NOEs, which are all below 0.5 at 2°C and below 0 at 25°C, provide clear evidence of the high flexibility of T␤4 in aqueous solution (Fig. 2). In a fully structured, globular protein of this size, these values are expected to lie around 0.7 at 2°C and 0.4 at 25°C. Heteronuclear NOEs found here are thus compatible with a globally unfolded protein. The spectral densities for the N-H vectors of the backbone at the three frequencies 0, N (60 MHz), and around H -N (540 MHz) were calculated from R 1 (N z ) and R 2 (N x,y ) relaxation time constants and heteronuclear 1 H-15 N NOEs measured at 2°C and 25°C using a reduced matrix approach (56, 57) (data not shown). Both absolute and relative values of these spectral densities at the three frequencies again emphasize the internal flexibility along the whole polypeptide chain. However, their profiles along the sequence are not completely homogeneous. At both temperatures, higher values of J(0) compensated by lower values of (J H ) in the segment 10 -16 demonstrate a restriction of internal flexibility in this fragment typical of a nascent helix. This restriction of motions in the hundred of picoseconds time scale can be also seen in the C-terminal segment 31-36 of the protein, although to a smaller degree.
Thymosin ␤4 Forms a Tight 1:1 Complex with Monomeric CaATP-actin at 25°C- Fig. 3A shows the HSQC spectra recorded at 25°C for a sample containing equimolar amounts of [U- 15  residues. Use of TROSY versions of these experiments was unsuccessful. The three-dimensional 1 H-15 N NOESY-HSQC was the only experiment that provided reliable results. Assuming inter-residual and sequential amide/side chain protons proximities, the analysis of H N -H N and H N -H␣ connectivities allowed a possible assignment of the N-terminal segment Met 6 -Ser 15 and the C-terminal fragment Glu 35 -Ser 43 . However, no correlation could be found allowing the assignment of amino acids in the central fragment 16 -34. A method based on selective labeling was then considered.
The strategy for selecting Leu, Thr, and Lys as 15 N-labeled amino acids is described under "Experimental Procedures." All the results obtained for selective labeling of T␤4 were in good agreement with amino acid evolution in the nitrogen cycle of E. coli (33,34). 15 N HSQC spectra of [ 15 N]AA T␤4 (where AA is Leu, Thr, or Lys) bound to G-actin are shown in Fig. 3B. Resonances obtained for the selectively labeled T␤4 perfectly superimposed those obtained for the uniformly labeled protein.
Finally, the three successful selective labelings, in conjunction with the 15 N NOESY-HSQC experiment, allowed the complete assignment of nitrogen and amide protons of T␤4 bound to G-actin. All amide protons strips could be assigned by researching similarity between strips, with the additional knowledge of the type of the amino acids that had been selectively labeled. The amino side chain undergoing the larger shift upon binding to G-actin could be assigned to Asn 26 , suggesting its location at the interface of the complex. Assignment of H␣ protons can also be suggested from the cross peaks between amide and H␣ protons in the 15 N NOESY-HSQC. When H N -H N correlations indicative of a helical folding were present, the most intense H N -H␣ correlation was considered to be intraresidual. In the other cases, the most intense H N -H␣ correlation was considered to be sequential (H N (i)-H␣(i-1)).
Besides this strategy, an HSQC spectrum of the complex between T␤10 and monomeric CaATP-actin was performed. A large spreading of the resonances upon binding was observed, as for T␤4, demonstrating the formation of an analogous tight complex of T␤10 with actin. The central segment 16 KLKK-TETQEKN is strictly conserved between T␤4 and T␤10. The absence of chemical shift variations in the HSQC spectra of the two thymosins for the peaks previously assigned to 18 LKK-TETQEK in the T␤4-CaATP-actin complex confirmed the assignment of these 9 residues (data not shown).

T␤4 Bound to G-actin Is Composed of an Extended Central Segment Flanked by Two ␣-Helices, and Its Dynamic Behavior
Demonstrates the Binding of the Whole Peptide's Backbone to G-actin-The very large variations of chemical shifts observed for amide nitrogens and protons along the whole backbone of free and actin-bound T␤4 demonstrate that the polypeptide chain undergoes a complete reorganization upon binding (Fig.  3C). The CSI calculated on H N, 15  converges on the existence of two ␣-helices, Asp 5 -Leu 17 and Lys 31 -Ala 40 , linked by an extended fragment, Lys 18 -Asn 26 (Fig. 3D).
Heteronuclear 1 H-15 N NOEs could be measured on T␤4 bound to G-actin at 25°C. The histogram of resulting NOEs is shifted from a value centered around Ϫ0.2 for free T␤4 to 0.7 for bound T␤4 (Fig. 2B). Due to the low concentration of the complex and the restricted recording time imposed by the limited stability of the sample, a low signal to noise ratio and therefore imprecise values of heteronuclear NOEs were obtained. However, the large positive shift of NOE values upon T␤4 binding to actin clearly indicates that T␤4 adopts the correlation time of a fully bound peptide. The smaller values of heteronuclear NOEs in the center of the sequence indicate that the corresponding N-H bonds of T␤4 in the complex retain some internal flexibility in the hundred of picoseconds range ( Fig. 2A).
The Structures of T␤4 Bound to CaATP-actin and MgATPactin Are Identical-De La Cruz et al. (12) recently observed significant differences in the binding characteristics of T␤4 to CaATP-actin and MgATP-actin. 15 N-1 H HSQC spectra of T␤4 bound to CaATP-actin or MgATP-actin are identical (data not shown). Amide nitrogen and proton chemical shifts are very sensitive to local electronic environment, and the identical 1 H N -15 N chemical shifts found for the bound state of T␤4 bound to either CaATP-actin or MgATP-actin demonstrate that the fold of T␤4 is identical in both cases. The only difference was the systematic presence of a higher amount of free T␤4 when MgATP-actin was used. This is likely due to the higher efficiency of polymerization of MgATP-actin as compared with CaATP-actin, leading to a higher excess of free T␤4 in the NMR sample.
Cleavage of G-actin by Subtilisin Does Not Alter the T␤4actin Complex-T␤4 binding to subtilisin-cleaved actin (subactin) was linked to a 9% increase in the actin tryptophane fluorescence, as compared with an 11.6% increase upon binding to unmodified G-actin (Fig. 4A). Binding constants of 1.5 Ϯ 0.2 and 2.7 Ϯ 0.4 M were found for the complexes of T␤4 with CaATP-bound actin and subactin, respectively. The actin-and subactin-sequestering activities of T␤4 were compared under different ionic conditions. Within low ionic strength polymerization buffer (1 mM MgCl 2 and no KCl), the critical concentrations for filament assembly were 0.35 M for actin and 1.7 M for subactin (Fig. 4B, inset). The slopes of the linear decrease in F-actin versus the total concentration of T␤4 yielded values of 0.4 M for K T and 1 M for K T Ј (Fig. 4B, a). These values refer to binding of T␤4 to the MgATP-bound form of G-actin. At physiological ionic strength (1 mM MgCl 2 and 0.1 M KCl), the critical concentrations were 0.1 and 0.5 ⌴ for actin and subactin, respectively (Fig. 4B, a, inset). When increasing amounts of T␤4 were added to a 16 M unmodified F-actin solution in Mg/KCl polymerization buffer, the decrease in F-actin was not a linear function of T␤4 concentration, as observed previously (11), which reflected the interaction of T␤4 with F-actin at high concentration (Fig. 4B, b). From the linear portion of the curve at low T␤4 concentration, a K T value of 1.2 M was derived. In contrast, the concentration of subtilisin-cleaved F-actin decreased perfectly linearly with increasing T␤4, demonstrating that when loop 38 -52 is cleaved, T␤4 no longer interacts with F-actin at high ionic strength (Fig. 4B, b). A K T Ј value of 2.7 M was derived from these experiments.
Nucleotide exchange is known to be strongly inhibited on G-actin upon T␤4 binding (13). We find that this property is conserved for subtilisin-cleaved actin. The rate of nucleotide exchange was 1.6-fold higher on subtilisin-cleaved actin than on unmodified actin. Binding of T␤4 to subactin lowered the exchange rate by 100-fold, as for unmodified actin. The T␤4 concentration dependence of the observed first order rate constant was consistent with a K T Ј value of 1.5 M for binding of T␤4 to subactin (data not shown).
In conclusion, cleavage of loop 38 -52 on G-actin, which leads to higher critical concentrations for actin filament assembly, causes only a 2-fold decrease in the affinity of T␤4 and does not affect the functional properties of T␤ 4 , except for the loss in ability of T␤4 to weakly interact with subtilisin-cleaved F-actin.
The 15 N-1 H HSQC spectrum of T␤4-subactin at 0.5 mM at 25°C was fully superimposable over the one obtained with unmodified actin (data not shown), except for slightly narrower linewidths in the case of subactin, due to the lower concentration of polymerized actin. NMR data provide information on the localization of T␤4 on actin. The structural integrity of actin is known to be conserved after subtilisin cleavage of segment Val 43 -Met 47 in the DNase I binding loop, which is locked up in the three-dimensional structure of G-actin by a ␤-sheet itself stabilized by an ␣-helix (Ref. 58 and this study). The lack of difference in the chemical shifts between T␤4 bound to actin and subactin demonstrates that all amide nitrogens and protons of T␤4 are in the same environment in both complexes and that they are thus not likely to be in the vicinity of the cleaved segment Val 43 -Met 47 in subactin.

DISCUSSION
Most structural studies of protein-protein complexes address the interaction of two folded partners with defined three-dimensional structures (59 -63). However, a number of proteins are physiologically unstructured and fold only upon binding to their biological target (64). Many of these proteins are involved in cell cycle regulation processes in which their unstructured state provides significant advantages such as a fast turnover, relatively low binding constants, and the ability to bind several targets (64,65). The unfavorable binding entropy linked to the induced folding process requires a precise control over the thermodynamics of the binding. A structural characterization of the proteins, both alone and bound to their target, is crucial to understand how these polypeptides selectively recognize their targets and perform specific functions. Nuclear magnetic resonance is uniquely well suited to provide detailed structural and dynamic information about unstructured proteins and the level of structure of each partner upon binding and thus to describe the local and global thermodynamics governing the binding process of unfolded states (43,66,67).
The present work constitutes the first structural study of the interaction between a major actin-sequestering protein, thymosin ␤4, and unmodified G-actin at concentrations (0.3-0.5 mM) that are physiological.
Strategies for the Assignment of Thymosin ␤4 Bound to Gactin-Despite the small size of T␤4, its assignment when bound to G-actin presents major difficulties due to the low stability and large size of the complex. The impossibility of growing recombinant actin at a significant level precludes the opportunity to obtain large quantities of deuterated actin that could drastically reduce the relaxation of the system and allow the use of classical three-dimensional NMR experiments or their TROSY counterparts for the assignment. We show here that this drawback can be partly bypassed by using a selective labeling strategy. In the case of a relatively small polypeptide such as thymosin, the combined use of the three selective labelings of threonines, lysines, and leucines and a three-dimensional 1 H-15 N NOESY-HSQC experiment was sufficient to obtain the complete assignment of amide protons and nitrogen resonances.
T␤4 Undergoes a Complete Folding Process upon Binding to G-actin-The large spreading of all resonances in the HSQC of T␤4 upon binding to G-actin clearly demonstrates that all amide protons and nitrogens of T␤4 are in a structured environment in the complex. When two folded proteins interact, differences of chemical shifts between the free and bound state can be exploited to delineate the interface of the complex using the chemical shift mapping technique. When the folding is induced by the binding, the resonance shifts are due to the organization of both intra-and inter-molecular protein interfaces, and it is not possible to differentiate between them. However, because the physical principles of protein folding and protein-protein association are similar, the way chemical shifts vary between the bound, folded state and the free, unfolded state is directly connected to the CSI and can be related to the type of local folding in the bound state. Its analysis demonstrates that the binding of T␤4 to actin is linked to the stabilization of the nascent N-terminal helix (residues 6 -13), the formation of a C-terminal helix (residues [31][32][33][34][35][36][37][38][39][40], and the organization of an extended interface with actin of the central segment (residues [17][18][19][20][21][22][23][24][25][26]. Implications for the Binding Site of Thymosin ␤4 on Gactin-Although the assignment and secondary structure characterization of the bound form of T␤4 do not allow us to characterize the organization of these secondary structures on G-actin, our results, combined with those of previous studies, give clear indications about the T␤4 binding site on actin. Mutagenesis and cross-linking studies have shown that the correct formation of the N-terminal ␣-helix (residues 6 -16) was crucial for the binding of T␤4 to G-actin and that this amphiphilic helix very likely made hydrophobic contacts with domain 1 of G-actin via the side chains of Met 6 , Ile 9 , Phe 12 , and Lys 16 (25,(27)(28)(29). Cross-links were obtained between Lys 38 of T␤4 and Gln 41 of G-actin and between Cys 43 of the mutant S43C-T␤4 and His 40 of G-actin, suggesting that the C-terminal segment of T␤4 was located near domain 2 of actin (12,29). Additional cross-links between Lys 3 /Lys 18 of T␤4 and Glu 167 / Asp 1 or Glu 4 of G-actin located the N-terminal segment of T␤4 near the pointed end of actin (29). Safer et al. (29) proposed from these cross-links that the central segment containing the consensus sequence 17 LKKTET adopts an extended conformation allowing T␤4 to run from the pointed end to the barbed end of G-actin.
Our results confirm the stabilization of the N-terminal helix upon binding to actin and are compatible with the hypothesis of an extended central segment, 17 LKKTET, possibly extended to 23 QEKNPLP. The very large variations of chemical shifts in this region are indeed indicative of the formation of an ordered, extended structure and suggest the formation of hydrogen bonds with residues of actin. The large chemical shift variations of the amino nitrogen and protons of Asn 26 is a strong indication for the direct involvement of this side chain in the binding. In an extended structure, a maximal translation of 3.4 Å/residue is allowed; hence an extended segment of 14 residues extends over 47 Å, allowing thymosin ␤4 to extend from the pointed end to the barbed end of actin. The fact that the modified F-actin (q) or subtilisin-cleaved actin (E). The fluorescence of pyrenyl-actin was measured after a 16-h incubation at room temperature. a, measurements in low ionic strength assembly buffer (1 mM MgCl 2 added to G buffer). Actin concentration, 1.5 M; subtilisincleaved actin concentration, 3 M. Inset, critical concentration plots for actin (f and q) and subtilisin-cleaved actin (Ⅺ and E) in low ionic strength (f and Ⅺ) and physiological ionic strength (q and E) buffers. Note that the specific fluorescence of subtilisin-cleaved F-actin is 28% lower than that of unmodified F-actin. b, measurements at physiological ionic strength (1 mM MgCl 2 , 0.1 M KCl). Actin and subtilisin-cleaved actin concentrations ϭ 16 M .   FIG. 4. Cleavage of G-actin by subtilisin does not alter the thymosin ␤4-actin complex. A, fluorescence titration curves of unmodified and subtilisin-cleaved actin by thymosin ␤4. T␤4 at the indicated concentrations was added to 2 M CaATP-G-actin (q) or subtilisin-cleaved actin (E). The percentage increase in tryptophane fluorescence was measured. Solid lines are binding curves calculated according to Eq. 1 and using maximal relative increases in fluorescence of 11.6% and 9% for actin and subtilisin-cleaved actin, respectively, and values of 1.5 and 2.7 M for K T and K T Ј . B, thymosin ␤4 sequesters subtilisin-cleaved actin as well as unmodified actin. T␤4 at the indicated concentrations was added to solutions of 2% pyrenyl-labeled un-cleavage of actin by subtilisin does not modify the electronic environment of T␤4 amide protons demonstrates that the backbone of the C-terminal end of the peptide does not make direct contact with segment 42-47 of actin, in agreement with the accessibility of Gly 46 of actin in T␤4-actin complex to proteolytic enzymes (12). However, it seems at first sight less compatible with the formation of cross-links between C-terminal residues of T␤4 and His 40 -Gln 41 of G-actin. All results, however, can be accommodated by the high flexibility of the DNase I binding loop that allows the cross-link of Lys 38 and Ser 43 of T␤4 with actin. Finally, the results are in favor of an induced fit between the C-terminal end of T␤4 and actin, resulting in a locked conformation of the beginning of the DNase I binding loop that is not perturbed by the cleavage of peptide 43-47 by subtilisin.
Models of the complex were built that are compatible with our NMR results and with the previously mentioned biochemical data (see "Experimental Procedures"). The problem of the docking of two proteins is particularly arduous when one of the proteins undergoes large conformational changes upon binding, as is the case for thymosin here. The use of experimentally derived constraints allowed a delimiting of the interaction surface (from the cross-links and mutagenesis results) and a restriction of thymosin ␤4-bound conformations (from our NMR results) but was not sufficient to converge toward a unique cluster of complexes. In particular, the lack of orientational restraints for the two helices of T␤4 toward actin leads to a large uncertainty in the position of the extended central segment. However, the models clearly show that thymosin ␤4 can cross the whole surface of G-actin, with the N-terminal and C-terminal helices making contacts with the pointed and barbed end of G-actin, respectively (Fig. 5).
HSQC spectra unambiguously demonstrate that the structure of T␤4 is the same when bound to CaATP-actin or MgATPactin. De La Cruz et al. (12) reported significant differences in the thermodynamic and kinetics parameters for the binding of T␤4 to CaATP-actin and MgATP-actin. This suggests that the peptide is able to drive the two different initial states of CaATP-actin and MgATP-actin toward a unique Ca/MgATPactin-T␤4 bound state.
Implication for the Sequestering Activity of Thymosin ␤4 -Thymosin ␤4 belongs to the class of WH2 domain proteins that contain an evolutionarily conserved actin monomer-binding mo-tif originally found in ␤-thymosins. They are all thought to play a role in the regulation of actin dynamics through their binding to actin monomers. The few structural studies of WH2 domains show that they are mostly unfolded alone under physiological conditions. Remarkably, they perform different functions in the regulation of actin polymerization, with some preventing and others promoting actin filament assembly, and the structural basis of these differences has not yet been elucidated.
The extended conformation of T␤4 on G-actin leads to a broad interface, which is compatible with the wide conformational changes undergone by the protein upon binding. The large entropy loss generated by the folding of intrinsically unfolded proteins upon binding to their target is usually compensated by broad interfaces, allowing the formation of numerous hydrogen bonds and the expulsion of a large number of water molecules. In the case of thymosin, the helical propensity of the N-terminal segment also leads to a favored entropy of binding and most probably constitutes the driving force for binding actin. Moreover, it is known that protein-protein binding sites often contain hydrophobic patches. The hydrophobic face of the N-terminal amphiphilic helix of T␤4, consisting of Met 6 , Ile 9 , Phe 12 , and the side chain of Lys 16 , constitutes a unique favorable recognition hydrophobic patch along the T␤4 primary sequence. This hydrophobic N-terminal surface, which is present in most WH2 domains (15), probably forms a standard interface with actin that suffices to bring a stable association, but not to bring the function specificity of these proteins. The unique sequestering activity of ␤ thymosins among the family of WH2 domains characterized thus far is then very likely due to a higher specificity of the C-terminal segment for actin or to a specific orientation of this C-terminal segment that prevents the nucleotide exchange and impedes actin-actin interaction at the pointed end of the filament. To investigate the specificity of the sequestering function of thymosins, some mutations that could modify this function can be proposed from our NMR results. Two regions can be probed first. The central region containing the so-called "actin binding motif" LKKTET is highly specific of the thymosin family. Residues of this region undergo the largest chemical shift variations upon binding to actin, indicative of a tight implication in the interaction with actin, most probably through the formation of intermolecular hydrogen bounds. Some of these residues are not strictly conserved in other WH2 domains, especially Thr 20 and Asn 26 , which are replaced by hydrophobic residues (Ala or Val for Thr 20 and Ile for Asn 26 ) in a number of other WH2 domains such as the domain 1 of ciboulot, which was recently shown to exhibit a profilin-like function (19). Subtle variations of the interactions in this region could lead to a variation of the orientation or the dynamics of the C-terminal segment of the WH2 domain. The location and stability of this C-terminal segment, which extends to the pointed end of actin in the case of T␤4, are likely to be crucial for the sequestering activity (29). In this segment, the acidic residue Glu 35 , which also undergoes a large chemical shift variation upon actin binding, is replaced by non-acidic, hydrophobic residues in a number of other WH2 domains. The analysis of the activity and structural features of mutants suggested from this study should bring additional information on the key interactions that govern the specificity of function of T␤4 and, more generally, actin-binding WH2 domains.
Acknowledgments-We thank Nicolas Birlirakis for help in the recording of NMR experiments. The 800 MHz spectra were run on the spectrometer installed at Institut de Chimie des Substances Naturelles-Centre National de la Recherche Scientifique, procured with the help of Région Ile de France and the Association pour la Recherche contre le Cancer.  (49) using structural constraints derived from our NMR results and previously published biochemical data. The two different orientations shown were generated using MOLMOL software (50). G-actin is depicted in gray, whereas thymosin is colored in blue and red (helices). Side chains of residues Lys 3 , Lys 18 , Lys 38 , and Ser 43 of thymosin and Asp 1 , Glu 4 , His 40 , Gln 41 , and Gln 167 of actin were shown to form cross-links and are shown in yellow. The cross-link constraints are all satisfied and are visualized as green dotted lines.