The role of backbone motions in ligand binding to the c-Src SH3 domain1

doi:10.1006/jmbi.2001.5083 available online at http://www.idealibrary.com on

J. Mol. Biol. (2001) 313, 873±887

The Role of Backbone Motions in Ligand Binding to the c-Src SH3 Domain Chunyu Wang, Norma H. Pawley and Linda K. Nicholson* Department of Molecular Biology and Genetics, Cornell University, Ithaca, NY 14853, USA

The Src homology 3 (SH3) domain of pp60c-src (Src) plays dual roles in signal transduction, through stabilizing the repressed form of the Src kinase and through mediating the formation of activated signaling complexes. Transition of the Src SH3 domain between a variety of binding partners during progression through the cell cycle requires adjustment of a delicate free energy balance. Although numerous structural and functional studies of SH3 have provided an in-depth understanding of structural determinants for binding, the origins of binding energy in SH3ligand interactions are not fully understood. Considering only the protein-ligand interface, the observed favorable change in standard enthalpy (
H ˆ ÿ9.1 kcal/mol) and unfavorable change in standard entropy (T
S ˆ ÿ2.7 kcal/mol) upon binding the proline-rich ligand RLP2 (RALPPLPRY) are inconsistent with the predominantly hydrophobic interaction surface. To investigate possible origins of ligand binding energy, backbone dynamics of free and RLP2-bound SH3 were performed via 15N NMR relaxation and hydrogen-deuterium (H/2H) exchange measurements. On the ps-ns time scale, assuming uncorrelated motions, ligand binding results in a signi®cant reduction in backbone entropy (ÿ1.5(0.6) kcal/mol). Binding also suppresses motions on the ms-ms time scale, which may additionally contribute to an unfavorable change in entropy. A large increase in protection from H/2H exchange is observed upon ligand binding, providing evidence for entropy loss due to motions on longer time scales, and supporting the notion that stabilization of pre-existing conformations within a native state ensemble is a fundamental paradigm for ligand binding. Observed changes in motion on all three time scales occur at locations both near and remote from the protein-ligand interface. The propagation of ligand binding interactions across the SH3 domain has potential consequences in target selection through altering both free energy and geometry in intact Src, and suggests that looking beyond the protein-ligand interface is essential in understanding ligand binding energetics. # 2001 Academic Press

*Corresponding author

Keywords: SH3 domain; backbone dynamics; thermodynamics; ligand binding; NMR

Introduction Src homology 3 (SH3) domains are found in many key signal transduction proteins. These Present address: C. Wang, Department of Biochemistry and Molecular Biophysics, Columbia University, New York, NY 10032, USA. Abbreviations used: SH3, Src homology 3; Src, pp60c-src; ITC, isothermal titration microcalorimetry; NOE, nuclear Overhasuer effect. E-mail address of the corresponding author: [email protected] 0022-2836/01/040873±15 $35.00/0

small, modular binding domains mediate intracellular signaling through transient interactions with speci®c proline-rich sequence motifs. Their biological importance is underscored by the fact that mutations in SH3 domains are often associated with malignancy.1,2 The SH3 domain of the cellular proto-oncoprotein pp60c-Src (Src) plays two distinct roles in regulating the participation of this nonreceptor tyrosine kinase in signal transduction. Through intramolecular interactions it stabilizes the inactive form of the kinase,3 and through intermolecular interactions it facilitates the assembly of multiprotein complexes in the active form.4 # 2001 Academic Press

874 Dynamic cycling between SH3 and this myriad of partners during progression through the cell cycle requires these interactions to be transient and exquisitely regulated.5 In order to understand the ligand binding properties of the canonical SH3 domain, extensive ligand screenings with peptides derived from cDNA libraries, combinatorial peptide libraries and phage display experiments have been carried out,6 and numerous structures of SH3 domains both with and without bound ligand have been solved.5 Elegant solution NMR studies have established two classes of Src SH3 ligands: class I with consensus sequence RXLPPLP (X ˆ any amino acid), with RLP2 (comprised of residues RALPPLPRY) serving as a paradigm, and class II with consensus sequence XPPLPXR.7 Both classes of ligands contain a core hPPhP motif (h ˆ hydrophobic amino acid) and adopt a poly-proline type II helix that packs into two pockets in the SH3 surface. However, the two classes bind in opposite orientations relative to the SH3 surface as determined by a salt bridge formed between an Arg residue in the ligand (at the N terminus of class I and the C terminus of class II) and an Asp residue in SH3.7 An additional speci®city pocket binds ¯anking sequences N and C-terminal to the core motif in type I and II ligands, respectively.8 The core interaction surface is predominantly hydrophobic, formed by conserved aromatic residues on the SH3 surface and by proline and hydrophobic residues of the ligand (Figure 1). The NMR structure of Src SH3 bound to RLP2, which lacks a ¯anking speci®city sequence, shows no intermolecular hydrogen bonds and is not signi®cantly perturbed from the free SH3 structure.7,9 However, measurement of trans-hydrogen bond scalar couplings reveal the propagation of minute ligand-induced changes in hydrogen bond length across the domain, demonstrating a subtle but global response of the protein to ligand binding.10 Recent thermodynamics studies of SH3 ligand binding have revealed that the binding energetics cannot be accounted for by structural considerations alone. As reported herein and elsewhere,11,12 the change in standard entropy for the binding reaction between SH3 and short peptide ligands measured by isothermal titration microcalorimetry (ITC) at or near 25  C is invariantly negative, a surprise considering the extensive hydrophobic interactions involved.5 Thus, the driving forces of SH3 ligand binding are still not fully understood. The role of protein dynamics in ligand binding has become an active area of research in protein NMR. Recently, several intriguing studies have correlated NMR-derived protein dynamics with function, particularly with protein-ligand interactions.13 ± 17 Approximate relationships between entropy and the generalized order parameter describing bond ¯uctuations18,19 have been derived,20,21 allowing estimation of the entropy change associated with changes in order parameters upon ligand binding. Such entropy

Functional Dynamics in the c-Src SH3 Domain

Figure 1. The SH3/RLP2 interface. The Figure was constructed using pdb ®le 1RLQ7 and Swiss PdbView (www.expasy.ch/spdbv/mainpage.html). SH3 residues Ê of RLP2 are shown in CPK, with non-polar within 5 A residues colored orange and polar residues colored by atom type (carbon, green; nitrogen, blue; oxygen, red). RLP2 (shown as stick model) is gray, with residue R1 highlighted in blue and selected residues labeled in italics. Hydrogen atoms are omitted for clarity. (a) End-on view of binding interface. The prism-like cross-section of the RLP2 poly-proline type II helix is highlighted by a broken triangle. (b) Top view of binding interface. The two Leu-Pro segments of the ligand pack into hydrophobic pockets 1 (between Y11 and Y57) and 2 (between Y57, Y13 and W39). RLP2 residue R1 salt bridges with SH3 residue D20.

changes can then be compared with the overall standard entropy change of the protein/ligand binding reaction measured by ITC to gain insight into the potential roles of backbone motions in determining ligand binding energy.

875

Functional Dynamics in the c-Src SH3 Domain

The primary goal of the work presented herein is to relate backbone dynamics of chicken c-Src SH3 to the binding function of this modular domain. Using the high af®nity class I ligand RLP2, we have characterized 15N backbone dynamics of both free and ligand-bound SH3 at three different time scales, and have measured the thermodynamic parameters of the binding reaction. Measurement of 15N relaxation parameters allows evaluation of motions occurring with time constants in both the ps-ns and ms-ms ranges, providing two temporal windows in which ligandinduced changes in dynamics may be detected. An additional time scale is provided by hydrogendeuterium (H/2H) exchange measurements, which allow characterization of changes in global or local unfolding events with exchange time constants on the order of minutes to days, possibly revealing cooperative units of unfolding. It will be shown that changes in motions on each of these time scales yield potentially signi®cant contributions to the ligand binding energetics, and that the dynamic response of the protein is not limited to the protein-ligand interface but extends to remote regions. The combination of NMR and ITC data allows ligand binding to be viewed from the perspective of the protein backbone to investigate potential sources of binding energy throughout the protein.

prism edges in contact with the SH3 binding surface, and hydrophobic packing dominates the protein-ligand interface (Figure 1). Accordingly, the observed difference in af®nity between two Src SH3 ligands (RLP2 and PLR1) was attributed to differences in hydrophobic contacts.7 At 25  C, a binding reaction so dominated by hydrophobic interactions should gain entropy upon ligand binding due to the release of ordered water molecules from the exposed hydrophobic surfaces of the free protein and ligand. Assuming that the free peptide randomly samples all accessible conformational space, conformational entropy will be lost upon adopting the bound conformation. However, one third of the residues comprising the RLP2 ligand are proline, allowing less conformational freedom for this peptide relative to the average unstructured peptide. For comparison, the thermodynamic parameters for binding of the Src SH2 domain to a phosphopeptide ligand (EPQpYEEIPIYL), a reaction with comparable hydrophobic surface area buried upon binding, shows a positive change in entropy as expected based on hydrophobicity arguments (C.W. & L.K.N., unpublished results). Hence, considering only the structure of the ligand binding interface of the SH3/RLP2 complex, the origins of binding energy are not immediately apparent. Measurement of

Results and Discussion Thermodynamics of SH3 ligand binding are not explained by the structure of the proteinligand interface The thermodynamic parameters for the SH3/ RLP2 ligand binding reaction in NMR buffer obtained from ITC measurements show that the interaction is driven by a favorable change in standard enthalpy (
Ho) that offsets an unfavorable reduction in standard entropy (
So) (Table 1). This is in agreement with ITC measurements made for other proline-rich ligands binding to this domain (C.W., J. L. Reinking and L. K. N., unpublished results) as well as to other SH3 domains.11,12 The origins of the favorable change in enthalpy and loss of entropy associated with the SH3/RLP2 binding reaction are not apparent from the structure of the protein-ligand interface. The prism-like structure of the ligand places two of the three

Table 1. ITC resultsa for c-Src SH3-RLP2 ligand binding Ka (104 Mÿ1) Kd (10ÿ6 M)
G(kcal/mol)
H(kcal/mol) T
S(kcal/mol)

5.3  0.7 19  2 ÿ6.4  0.1 ÿ10.2  0.6 ÿ3.8  0.6

a All data displayed are mean and standard deviations of three measurements.

15

N relaxation parameters

Backbone amide 15N and 1HN assignments were determined for all non-proline residues in both free and RLP2-bound SH3 and are deposited in the BioMagResBank (BMRB accession numbers 4888 and 4889, respectively). These assignments are in general agreement with those previously reported in different buffer conditions and for a slightly different construct.7 Relaxation parameters were determined for 51 and 53 residues within the canonical SH3 structure (residues 6 to 61) in free and RLP2bound SH3, respectively. Residues omitted from analysis due to spectral overlap include N34 (superimposed with Q65) in both free and RLP2bound SH3, and E18 and S44 in free SH3. The resulting T1, T1r, and NOE values are plotted as a function of residue in Figure 2. Relatively high NOE values indicate that the small SH3 fold is well ordered with the exception of a few residues in the RT and n-Src loops. Global tumbling of free and RLP2-bound SH3 is anisotropic Reduced data sets22 consisting of 31 and 33 residues for free and RLP2-bound SH3, respectively, were used to determine the global tumbling parameters for each form. As indicated by the resulting F statistics, the axially symmetric anisotropic model for global tumbling provides the optimal ®t for both free and RLP2-bound SH3 (Table 2). As expected, the effective correlation time for global tumbling increases upon ligand binding. The diffu-

876

Functional Dynamics in the c-Src SH3 Domain

Figure 2. Relaxation parameters for free (open triangles) and RLP2bound (®lled triangles) versus residue number. The overall increase in T1 and decrease in T1r is consistent with a small increase in correlation time for global tumbling upon ligand binding.

sion tensors for free and RLP2-bound SH3 are clearly distinguished as prolate (Dk/D? ˆ 1.22) and oblate (Dk/D? ˆ 0.85) ellipsoids, respectively (Table 2 and Figure 3). Distinction between prolate and oblate models has been previously demonstrated for another small globular protein.23

Changes in ps-ns time scale motions occur both near and far from binding surface Employing the optimal global tumbling parameters for each form, 39 and 36 residues were adequately ®t by the two-parameter Lipari-Szabo (LSz) model (S2,te)18,19 in free and RLP2-bound

Table 2. Global tumbling analysis for free and RLP2-bound SH3 tac,eff (ns) Dxx (nsÿ1) Dyy (nsÿ1) Dzz (nsÿ1)



y ( )b

f ( )

b

c ( )b

w2/nc

F

Pd

Free SH3 Isotropic Axially symmetric Asymmetric

5.95 5.91 5.93

0.0280 0.0263 0.0258

0.0280 0.0263 0.0268

0.0280 0.032 0.032

24 24

179 175

20

4.89 1.19 1.15

32.2 1.44

4.65e-9 0.255

RLP2-bound SH3 Isotropic Axially symmetric Asymmetric

6.37 6.42 6.43

0.0262 0.0273 0.0282

0.0262 0.0273 0.0266

0.0262 0.0233 0.0230

65 62

31 36

155

8.41 2.34 2.27

28.6 1.46

8.35e-9 0.248

a tc,eff is calculated as (1/2)(Dxx ‡ Dyy ‡ Dzz)ÿ1, where the Dii`s are the magnitudes of the principal components of the diffusion tensor. b The angles y, f, and c are the Euler angles that de®ne the orientation of the diffusion tensor with respect to the moment of inertia frame of the X-ray crytallographic structure of SH3. c 31 and 33 residues were used to ®t the ®nal global tumbling parameters for free and RLP2-bound SH3, respectively. The number of degrees of freedom, n, is the number of residues minus the number of ®tted parameters (one, four, and six for isotropic, axially symmetric and asymmetric, respectively). d P represents the probability that the reduction in w2 with inclusion of additional parameter(s) is obtained purely by chance.

Functional Dynamics in the c-Src SH3 Domain

Figure 3. Distinction between prolate and oblate axially symmetric diffusion tensors. Values of w2g(Dk, D?, y, f) projected along Dk/D? for the axially symmetric anisotropic model, obtained using simulated annealing, are plotted for free (a) and RLP2-bound (b) SH3. The minima illustrate a clear distinction between prolate (Dk/D? > 1) and oblate (Dk/D? < 1) models for each case. (c) The corresponding orientation of the diffusion tensor relative to the SH3 fold for free (left) and RLP2bound (right) SH3. The unique axis of each diffusion tensor points out of the page.

SH3, respectively. For the remaining residues, two three-parameter models were considered: the addition of a chemical exchange term (S2,te,Rex),24,25 and an extended model which includes an intermediate timescale motion (S2f ,S2s ,te).26 In the free form, two residues required an intermediate timescale motion, and ten required a chemical exchange term, while in the RLP2bound form three residues required an intermediate timescale motion, nine required a chemical exchange term, and ®ve were not adequately ®t by any model (as de®ned in Materials and Methods). The residue-speci®c generalized order parameter (S2), which represents a measure of restriction of motion of the individual N-H bonds, plotted as a function of residue, reveals the general features of

877 ¯exibility across the backbone (Figure 4(a)). The most ¯exible residues (low S2) are located in the nSrc loop and the RT loop, while residues in the distal hairpin are relatively well ordered. In general, both free and RLP2-bound SH3 have relatively rigid backbones, each with mean S2 values of 0.86(0.07) (for residues 6 to 61), consistent with the range of S2 values generally obtained for N-H groups in folded proteins. The backbone N-H S2 values reported for SH3 domains of Hck,27 Btk,28 and the folded form of Drk,29 all obtained in the absence of ligand using the isotropic global tumbling assumption, closely parallel those of unliganded c-Src SH3 presented here. To our knowledge, the effect of ligand binding on backbone 15N dynamics in SH3 domains has not been previously examined. Ligand-induced changes in S2 are observed in regions both near and remote from the c-Src SH3 ligand binding interface. Increases in S2 values that exceed the associated uncertainty are observed at Y11, E36, W39, Q49, S55, N56 and V58 (Figure 4(a)). Of these residues, Y11, W39, S55 and Ê of the ligand (Figure 1). ResiN56 are within 5 A Ê further, and is adjacent to Y57 due V58 is only 1 A that directly contacts the bound RLP2 peptide. Hence, the majority of those residues displaying increased restriction in ps-ns motions upon ligand binding are at or near the protein-ligand interface. Ê ) from the However, E36 and Q49 are remote (>9 A bound ligand, indicating long-range effects. Decreases in S2 that exceed the associated uncertainty also occur in regions both near and far from the ligand binding surface. Increases in motion occur at R16, T17, and D20, which all reside in the RT loop, as well as at V32 and T35 (at the base of and in the n-Src loop, respectively). Of these resiÊ of the bound ligand. dues, only D20 is within 5 A Ê , while V32 Residues R16 and T17 are within 7 A Ê ) from the ligand and T35 are remote (>9 A (Figure 4(a) and (c)). In the context of structure, the changes in S2 values described above occur primarily in or near the three loops of the SH3 fold. In the n-Src loop, changes are observed for W39, which directly interacts with bound ligand, for T35 and E36, located directly across the loop from W39, and for V32, located at the base of the loop. It is possible that ligand-induced restriction of W39 may alter motions of residues located directly across this

loop, as well as of residues hydrogen bonded across the loop (such as V32). Similarly, correlated changes in motion are observed across the RT loop, where D20 and its cross-loop neighbors R16 and T17 display increases in motion upon ligand binding. In the c-Src SH3/RLP2 NMR structure,7 the side-chain carboxyl group of D20 in the RT loop participates in a salt bridge with the guanidino group of R1 in RLP2. In the X-ray structure of the repressed form of nearly intact Src (pdb accession code 2Src30), which lacks the D20-ligand salt bridge, several electrostatic interactions occur between D20 and both R16 and T17. Such inter-

878

Functional Dynamics in the c-Src SH3 Domain

Figure 4. Motions on the ps-ns and ms-ms time scales, and their locations relative to the ligand binding interface. (a) Generalized order parameters (S2) versus residue for free (open symbols) and RLP2bound (®lled symbols) SH3. Models that best ®t each residue are denoted (triangle, LSz; diamond, Rex; square, extended). The corresponding SH3 secondary structure is displayed across the top. (b) Rex versus residue for free (open bars) and RLP2-bound (®lled bars) SH3. (c) Residues displaying statistically signi®cant changes in S2 upon binding RLP2. Side-chains of Y11, Y57 and W39 are shown to highlight the SH3 binding surface (shown also in (d)). Color coding of the SH3 ribbon indicates lack of signi®cant change or not determined (gray), and signi®cant decreases (blue) or increases (red) in S2. (d) Residues displaying statistically signi®cant changes in Rex upon binding RLP2. The same color coding as in (c) is used for changes in Rex.

actions in free SH3 would be disrupted by formation of the D20-ligand salt bridge upon binding to RLP2, providing a plausible mechanism for increased motion of these residues upon ligand binding. Finally, residue Q49 is located in the distal loop, which is on the opposite side of the domain from the protein-ligand interface (Figure 1). Minute ligand-induced changes in hydrogen bond geometry have been characterized for this domain, and have demonstrated the propagation of strain to regions remote from the protein-ligand interface, including the base of the n-Src loop and the tip of the distal loop.10 The relatively small changes in ps-ns time scale motions observed here, both near and remote from the protein-ligand interface, similarly re¯ect the global response of the domain to ligand binding. The overall contribution to the change in free energy resulting from changes in the order parameters of N-H bonds is ÿ1.5(0.6) kcal/mol, estimated by two different methods.20,21 Although this value includes all residues ®t in both the free and bound forms, it compares well to the values obtained either when residues ®t with different models in the two forms are not included (ÿ1.1 kcal/mol) or when only residues ®t by the

Lipari-Szabo model in both forms are used (ÿ1.2 kcal/mol). These values are of the same order of magnitude and sign as T
S (ÿ3.8  0.6) measured by ITC for this binding reaction. Thus, although the changes in order parameters are relatively small, the net restriction of backbone N-H bonds may account for part of the unfavorable entropy change that occurs upon ligand binding. While our measured order parameters report on the changes in ¯exibility of the backbone, additional changes in entropy will certainly arise from ligand-induced changes in side-chain motions. For example, four aromatic side-chains in the SH3 binding surface (Y11, Y13, W39 and Y57) pack closely with the bound RLP2 ligand.7 As an upper bound for the change in entropy associated with ligand-induced restriction of these sidechains, we assume that each adopts a single conformation in the bound form, and that each sidechain samples the major residue-speci®c rotamer populations31 in the free form. For this limiting case, the loss in entropy due to conformational restriction of these four side-chains is ÿ3.1 kcal/ mol (ÿ0.7 kcal/mol for each Tyr and ÿ1.0 kcal/ mol for Trp).32 The relaxation parameters obtained for the W39 indole NH group were not adequately

879

Functional Dynamics in the c-Src SH3 Domain

®t by the Lipari-Szabo formalism in either form (w2 values were 8.19 and 8.86 for free and bound forms, respectively, data not shown). However, the best-®t S2 values (0.66 and 0.83 for free and bound, respectively) do indicate a ligand-induced restriction of ps-ns time scale motions of this side-chain that translate into a change in entropy of ÿ0.3 kcal/mol, considerably less than the upper bound predicted above but a signi®cant contribution. Additional contributions of side-chains to the net change in entropy in this system are being investigated via methyl group relaxation studies and will be reported elsewhere (unpublished results). Changes in m s to ms time scale motions occur primarily near the ligand binding surface Interpretation of the relaxation data for a number of residues in liganded and unliganded SH3 required inclusion of a chemical exchange term (Rex), indicating the occurrence of motions on the ms-ms time scale. Several proteins have been found to display enrichment of ms-ms time scale motions in binding surfaces, suggesting an important role for such motions in molecular recognition.33 ± 37 For measurements reported herein, an Rex value of 1 sÿ1 corresponds to an approximate error in T1r of 0.014 seconds, threefold higher than the average experimental uncertainty of the T1r measurement (0.004 second). Hence, for a residue with Rex > 1 sÿ1, it is more than 99.7 % certain that requirement of an Rex term is not due to pure chance. There are four such residues in free SH3 (T19, I53, S55 and V58) and three in RLP2-SH3 (E18, T19, and V32) (Figure 4(b)). With the exception of V32, all are located in two regions of the SH3 fold that contribute to the binding surface, the 310 helix and the RT loop. The most dramatic change in Rex induced by ligand binding occurs at three residues located in or near the 310 helix. These residues (I53, S55 and V58) have large Rex terms that are eliminated upon ligand binding (Figure 4). This region is important in formation of binding pockets 1 and 2 (Figure 1), providing the scaffold that supports the position of Y57 (Figure 4(d)). Packing interactions between Y57 and the two Leu-Pro segments of RLP2 may inhibit motions on the ms-ms time scale in this region, potentially contributing to the reduction of system entropy upon ligand binding. Chemical exchange is also observed for residues in a localized region of the RT loop in both free and RLP2-bound SH3. Although not all greater than 1 sÿ1, residues 18-20 near the tip of the RT loop require an Rex term in RLP2-bound SH3, while residues 19 and 20 show such a requirement in the free form (spectral overlap prevents assessment of E18 in this form). This indicates the occurrence of a localized dynamic process in this region on the ms-ms time scale in both forms, which may play a role in recognition. The speci®city pocket of the SH3 binding surface (pocket 3) consists of a

valley between the RT and nSrc loops into which a core-¯anking sequence may bind. The RLP2 ligand employed herein lacks such a ¯anking sequence. Hence, in both free and RLP2-bound forms, pocket 3 remains unbound. It is interesting to note that RT loop residue T19 has been identi®ed as a strong speci®city modulator in c-Src SH3,8 and displays high Rex values in both forms. In the RLP2-bound form, the observed chemical exchange in this region may also re¯ect transient formation of the R1-D20 salt bridge. In the Grb2 N-terminal SH3 domain in complex with a peptide derived from SOS, peptide residue R8 (equivalent to R1 in RLP2) displayed ms time scale motions.38 Such motion of R1 in RLP2 could account for the chemical exchange observed here for residues in the vicinity of D20 in the RLP2-SH3 complex. In addition to residues in the RT loop, residue G51, in b4 and located adjacent to D20 in the tertiary structure, also displays a small Rex term in the RLP2-bound form, further supporting an R1-associated chemical exchange process. The remaining residue displaying a large Rex term, V32, is located at the base of the n-Src loop remote from the RLP2 ligand binding interface. This residue requires an Rex term only in the RLP2bound form, and also exhibits a signi®cant decrease in S2 upon ligand binding (Figure 4). Anticorrelation of Rex and S2 has been previously noted, where the artifactual reduction of S2 values for residues displaying large Rex values was estimated to be (0.018  0.004)Rex.39 The reduction in S2 for V32 signi®cantly exceeds this approximate artifactual reduction, indicating that ligandinduced increases in motion occur on both the psns and ms-ms time scales at this remote location. For the two additional residues displaying anticorrelated Rex and S2 values, S55 and V58, the changes in S2 also exceed the estimated artifactual effect. Hydrogen-deuterium exchange rates are significantly altered by ligand binding It has been previously noted that ligand binding increases the number of slowly exchanging amide protons in the chicken c-Src SH3 domain,7,8 indicating a slowing of the cooperative unfolding of this domain. To further probe the energetic implications of this important difference, we have measured the hydrogen-deuterium (H/2H) exchange rates (kex) for free and RLP2-bound SH3. These rates were converted into protection factors (P) using site-speci®c intrinsic exchange rates (kint).40 The kint values for SH3 residues calculated for the experimental conditions employed range from 0.99 sÿ1 to 90 sÿ1, while the closing rate (kc) for the global folding reaction of the c-Src SH3 domain is 56.7 sÿ1 in a similar aqueous solution.41 The necessary condition (kint5kc) for EX2 exchange, which allows the residue-speci®c apparent Gibbs free energy for the opening reaction (
Gop,i) to be obtained,42 does not hold for all residues. Speci®cally, ®ve residues (G26, H43, S44,

880 G51 and N56) have kint values that exceed 10 sÿ1 and are excluded from consideration in terms of
Gop,i values. All of the measured exchange rates are presented in terms of the more general protection factors (Figure 5). Differences between P values for residues with similar kint values re¯ect differences in the site-speci®c opening and/or closing rates, and hence indicate local unfolding rather than uniform global unfolding. Conversely, similar P values for structurally clustered residues for which the EX2 mechanism is valid can reveal cooperative folding units. In agreement with P values reported for free cSrc SH3(41), the most protected residues in both free and RLP2-bound SH3 include those located at the base of the RT loop and in b-strands 3 and 4 which form a b-hairpin (Figure 5(a)). Protection of all NH groups for which exchange rates were determined increased upon ligand binding,

Functional Dynamics in the c-Src SH3 Domain

although not by a uniform amount, with the largest changes occurring at residues Y11, D20, A42, N56 and Y57 (Figure 5). Of these ®ve residues, four are located in the ligand binding surface and make direct contact with bound ligand (Figure 1). However, A42 is well removed from the binding interface, located near the center of strand b3. Several additional residues remote from the ligand binding interface show signi®cantly enhanced protection, such as residues in the diverging turn, b2, b3 and b4. These results indicate that ligandinduced increases in levels of protection are not simply due to steric blocking resulting from the presence of bound ligand. Hydrogen deuterium exchange rates reveal cooperative units of unfolding A single hydrogen bond within a b-sheet participates in two potentially cooperative types of

Figure 5. The hydrogen bond topology of c-Src SH3 and hydrogen/deuterium exchange protection factors. Hydrogen bonds are shown as arrows from the amide proton to the carbonyl oxygen. Protection factors for free (ln(Pf)) and RLP2-bound (ln(Pb)) SH3, and the difference (
ln(P) ˆ ln(Pb) ÿ ln(Pf)) are plotted. Open bars indicate residues for which the EX2 assumption is questionable. Errors in ln(P) were on average 0.5 kcal/mol, with the exception of Y11 and A42 in the SH3/RLP2 complex, which had larger errors (6.7 and 3.4 kcal/mol, respectively) due to the slow decay of peak intensity for these residues. (a) The ®ve-stranded b-sheet of the SH3 fold, showing inter-strand (b3-b4, b2-b3, b1-b2, b5-b1) and cross-strand (broad shaded lines 1-9) hydrogen bond networks, and the corresponding protection factors. Protection factors are plotted for each inter-strand hydrogen bond network versus cross-strand network number. (b) Hydrogen bonds stabilizing the RT loop and diverging turn, and the corresponding protection factors. Protection factors are plotted versus residue for those N-Hs hydrogen bonded (top) and those not (bottom).

881

Functional Dynamics in the c-Src SH3 Domain

hydrogen bond groupings (Figure 5(a)). The most easily visualized is the group of hydrogen bonds connecting two adjacent b-strands, referred to herein as an inter-strand network. Protection from H/2H exchange of amide protons within such a group may be correlated by separation, or ``unzipping'', of the two b-strands. A second type of group runs orthogonal to the strands of a b-sheet, with hydrogen bonds connecting peptide planes on different strands across the sheet. This type of group is referred to herein as a cross-strand network. Protection within this type of group may be correlated by changes in the position or orientation of any of the linked peptide planes within a given cross-strand network. The distribution of protection factors over the hydrogen bond topology of the ®ve-stranded b-sheet shows that although the structured regions of SH3 display the highest levels of protection, these levels are not uniform and do not respond uniformly to ligand binding (Figure 5). As was previously noted for free SH341, amide protons within the b3-b4 and b5-b1 inter-strand networks display the highest and lowest levels of protection, respectively, in both free and RLP2-bound forms (Figure 5(a)). Protection levels vary signi®cantly across both inter-strand and cross-strand networks, indicating that H/2H exchange in much of the SH3 fold occurs through local unfolding events. However, the presence of a core of hydrogen bonds across the ®ve-stranded b-sheet with similarly high ln(P) values (9.1(0.5) kcal/mol in free and 14.8(0.5) kcal/mol in bound form, black bars in Figure 5(a)) indicates that the most protected residues are likely to exchange primarily through cooperative global unfolding. The difference between these values (5.7(0.7) kcal/mol) is in good agreement with the thermodynamically measured binding free energy (Table 1), indicating that H/2H exchange of these core residues appears to re¯ect the difference in thermodynamic stability between the free SH3 domain and the complex. This further supports that exposure of these residues to solvent occurs through global unfolding. Hence, exchange is facilitated by a combination of both local and global unfolding events in both free and RLP2-bound SH3. In addition to the ®ve-stranded b-sheet, hydrogen bonds that stabilize the SH3 fold are found in the RT loop and the diverging turn (Figure 5(b)). Residues 11-13 and 21-28 have similar protection factors in free SH3, an unexpected result since the NH groups of several of these residues (D12, S22 and R28) do not appear to participate in hydrogen bonds. Upon ligand binding, a higher increase in stabilization is observed in this region for amide protons involved in structural interactions (Figure 5(b)). These include hydrogen bonds across the RT loop (Y13, F23 and K25), across the diverging turn (K24 and E27), and between the RT loop and b4 (L21), and residues Y11 and D20 that directly interact via their side-chains with the bound ligand. Residues Y13, F23, K24, K25 and

E27 all exhibit nearly identical protection in free SH3, and increase by nearly identical amounts upon ligand binding (Figure 5(b)). Hence, this local cluster of hydrogen bonds appears to behave as a cooperative unfolding unit that is stabilized by ligand binding. The similarity between these protection factors and the core residues of the b-sheet suggests that unfolding of this region is coupled to global unfolding.

Correlation between protection factors and hydrogen bonds Of the 24 amide protons in either free or RLP2-bound SH3 for which hydrogen bonds to carbonyl or carboxyl groups were observed via trans-hydrogen bond scalar couplings (h3JNC),10 all are signi®cantly protected from H/2H exchange with the exception of V32 and T35 (in the nSrc loop), L45 (at the base of the b2-b3 connection) and S61 (at the base of the b1-b5 connection). It is particularly notable that the amide group of K25, which hydrogen bonds to the side-chain carboxylate of D12 at the base of the RT loop, is signi®cantly protected in both free and RLP2-bound SH3. This backbone-to-sidechain hydrogen bond on the solvent-exposed side of the RT loop is highly conserved in SH3 domains and makes a signi®cant energetic contribution to the ensemble of folding transition states for both Src and a-spectrin SH3 domains.43 The other backbone-to-sidechain hydrogen bond observed via scalar coupling, NHT35-OdN33 in the nSrc loop, was not signi®cantly protected from exchange in either free or RLP2-bound SH3. Comparison of the changes in protection induced by RLP2 binding with the corresponding measured changes in hydrogen bond length reveals rather surprising results. Most hydrogen bonds observed via h3JNC lengthened upon ligand binding, contrary to the large increase in protection factors observed here for these residues. Three exceptions are L21, I31 and L41, that all display a decrease in hydrogen bond length and an increase in protection upon ligand binding. While the h3JNC0 coupling constant re¯ects the distance between the coupled nuclei and hence, is a direct measure of hydrogen bond length, P depends on the change in free energy of the system for the opening reaction. This change in free energy upon opening a given hydrogen bond depends not only on the strength of the hydrogen bond, but rather on the net difference between all energy terms associated with the closed and open states, which includes all electrostatic and van der Waals interactions as well as entropy involving both the protein and the solvent. The lack of correlation between h3JNC0 and P indicates that P should not be directly interpreted in terms of hydrogen bond strength, and that the h3 JNC0 coupling constant should not be directly interpreted in terms of protein stability.

882 Redistribution of the native state ensemble and implications for ligand binding energy As has been observed for other proteins,44,45 ligand binding to the Src SH3 domain signi®cantly decreases the H/2H exchange rate of amide protons not only at the protein-ligand interface, but also in remote regions of the protein. What implications might this have for ligand binding energy? Considerable groundwork has been laid in support of a paradigm in which the native state of a protein consists of an equilibrium ensemble of microstates with different ligand binding af®nities.46 The microstates differ from one another in the number of interactions broken, such as hydrogen bonds or van der Waals interactions, and sampling of different microstates facilitates H/2H exchange. In this paradigm, ligand binding alters the most probable distribution of microstates, since microstates with high ligand af®nities are favored. Equations expressing the apparent change in enthalpy and entropy of the system upon ligand binding due to alteration of the native state population distribution have been derived.47 Calculation of these perturbations requires knowledge of the population distribution over microstates as well as the ligand af®nities of the microstates (gi). As discussed in detail elsewhere, for cases where gi decreases with increasing number of interactions broken, the population of the higher energy (i.e. open) microstates would be reduced, and a decrease in protein mobility (as detected by methods such as H/2H exchange or ¯uorescence quenching) would be expected.47 Our observation of increased protection from H/2H exchange indicates that ligand binding does reduce the population of open microstates, narrowing the population distribution and resulting in a favorable change in enthalpy and an unfavorable change in entropy. This correlates with the observed thermodynamic parameters for the RLP2/SH3 binding reaction. Hence, from the equilibrium ensemble perspective, the entire domain may contribute to ligand binding energy. Experimental support for the equilibrium native state ensemble paradigm is provided by several recent studies.13,48 ± 50 For example, direct evidence of a population-shift mechanism was recently reported where the unphosphorylated inactive state of the signaling protein NtrC was shown to be composed of an ensemble of microstates that includes the active state, and phosphorylation of NtrC shifts the equilibrium to the active form.50 In a comparison of linear and cyclic forms of the Nterminal SH3 domain of c-Crk, the head-to-tail cyclized form bound a proline-rich ligand with sevenfold increased af®nity compared with the corresponding linear form.49 This corresponds to a change in
Gbinding of 1.1 kcal/mol, and is consistent with a ``pre-ordering'' of the cyclized form into a more narrow set of microstates with higher ligand af®nity. We note that covalent linkage of the N and C-terminal ends of the SH3 fold would

Functional Dynamics in the c-Src SH3 Domain

inhibit unzipping of the b1-b5 inter-strand network of hydrogen bonds, the least protected of the structural interactions observed herein for the Src SH3 domain. Hence, the cyclized form may represent the subset of microstates in which the b1- b5 interstrand interactions are predominantly formed. Our H/2H exchange results clearly show a signi®cant increase in protection in regions both near and remote from the binding interface. This adds to the growing evidence that supports the view of proteins as dynamic ensembles of similar microstates in rapid equilibrium, with the population distribution over microstates altered by ligand binding. In light of the magnitude of the change in af®nity between linear and cyclized c-Crk N-SH3, we further suggest that such dynamics can contribute signi®cantly to the energetics of SH3 ligand binding.

Conclusions Characterization of the structural and dynamic response of a protein to ligand binding provides insights into the biophysical basis for fundamental biological processes such as allosterism and the regulation of protein-protein interactions involved in signal transduction. Previously, through measurement of trans-hydrogen bond scalar couplings, structural paths were characterized through which strain is cooperatively propagated from the binding site to remote locations within the c-Src SH3 domain.10 Herein we have shown that changes in dynamics on three different time scales also occur in backbone regions remote from the protein-ligand interaction surface. Hence, the effects of the localized binding interaction are clearly communicated both structurally and dynamically across the domain. The biological importance of communication of the c-Src SH3 ligand binding interaction to remote regions is potentially twofold. First, all sources of changes in free energy are critical in the delicate energy balance that determines the equilibrium populations of protein-ligand or protein-protein interactions. Different binding partners are recognized by the c-Src SH3 domain during progression through the cell cycle. Through participation of the whole domain in modulating binding energy, more than a single binding surface is involved which may enable ®ner tuning of switching between these interactions. We have shown here that on the ps-ns time scale, assuming that all observed motions are uncoupled from both global and other local motions, the observed changes in motion upon ligand binding would result in an increase in free energy of approximately 1.5 kcal/ mol, a signi®cant contribution. Additional motions on the ms-ms time scale are suppressed by ligand binding, which may also contribute to an unfavorable change in entropy of the system. The large increase in stability observed in the H/2H exchange measurements not only provides further

883

Functional Dynamics in the c-Src SH3 Domain

evidence for sources of entropy loss upon ligand binding, but also supports the notion that stabilization of pre-existing conformations within a native state ensemble is a fundamental paradigm for ligand binding.46,50 Second, propagation of the ligand binding interaction across the SH3 domain may have additional implications for mediating the assembly of multiprotein compexes in the activated state of Src. Recent solution NMR studies suggest that ligandinduced structural changes occur in the highly conserved linker (LINK32) between the SH3 and SH2 domains. In the context of the contiguous SH3LINK32-SH2 domains, LINK32 is ordered.51 Binding of RLP2 induces changes outside of the SH3 domain of not only chemical shifts52 but also of the LINK32 structure and of the relative orientations of the SH3 and SH2 domains (J. Reinking et al., unpublished results). Since numerous Src binding partners interact via both its SH3 and SH2 domains, the relative orientations of these binding modules could in¯uence the speci®city and af®nity of such interactions. Hence, the propagation of ligand binding beyond the SH3-ligand interface has potential consequences in target selection through altering both free energy and geometry in the intact Src tyrosine kinase.

For ITC studies, puri®ed SH3 was dialyzed into NMR buffer (50 mM sodium phosphate (pH 6.5), 200 mM sodium sulfate, 1 mM DTT), to which was added 10 mM DTT, 10 mg/ml pepstatin, 50 mM EDTA, 10 mg/ml benzamidine and 50 mM chloramphenicol. The SH3 protein was concentrated to approximately 0.1 mM using a 3 kDa molecular weight cut-off Centriprep concentrator (Amicon, Inc.). The same procedures were used to produce 15N- and 13 C/15N-labeled SH3 for NMR measurements, with the exceptions that isotopically enriched M9 medium was used, IPTG induction was initiated at A595 ˆ 0.4-0.6, and the SH3 protein solution was further concentrated to 1 mM. The ligand RLP2 (with amino acid sequence RALPPLPRY) was synthesized and puri®ed by the Cornell BioResources Center using solid-phase peptide synthesis and preparative HPLC. The purity of peptides exceeded 98 % as judged by analytical HPLC, quantitative amino acid analysis and capillary electrophoresis. The appropriate molecular mass was con®rmed by mass spectroscopy. The extinction coef®cient for RLP2 (911.2 Mÿ1 cmÿ1) was empirically determined by measuring the UV absorbance of a solution for which the concentration was determined by amino acid analysis in Cornell's BioResources Center. Samples of the SH3/RLP2 complex contained four molar equivalents of RLP2 ligand.

Materials and Methods

Isothermal titration calorimetry (ITC) experiments were carried out on a MCS (MicroCal Inc., North Hampton, MA) microcalorimetor. Lyophilized ligand was weighed and dissolved in the desired amount and type of solution (the same solvent as was used for the protein) to obtain the desired concentration. Ligand concentration was determined by UV absorption using the empirically determined extinction coef®cient for RLP2 (911.2 Mÿ1 cmÿ1). In each experiment, 3 ml of 2.34 mM RLP2 was injected at ®ve minute intervals into the 1.35 ml microcalorimetry cell containing 0.06 mM of protein at 25  C, for a total of 30 injections. The heat of dilution for RLP2 was measured and was used to correct binding isotherms prior to data analysis with the ORIGIN software (MicroCal Inc., North Hampton, MA). This software uses the standard Marquardt non-linear least squares ®tting method and equations detailed elsewhere53 to relate the heat evolved and the ligand:protein molar ratio to the thermodynamic parameters for binding. Since the stoichiometry of ligand binding for SH3 is known to be 1:1, the precise protein concentration was determined by ®tting the parameter n (stoichiometry of binding) to be 1 using the known ligand concentration.54 The standard enthalpy change upon ligand binding,
Ho, and the equilibrium constant of protein-ligand binding, Ka were derived directly from curve ®tting while the standard Gibbs free energy change,
Go, and standard entropy change,
So, of the binding reaction were calculated by:

Sample preparation The vector pGEX-SH3 (a generous gift from David Shalloway), which encodes a GST-fusion protein containing chicken c-Src residues 77-146 with non-c-Src residues GS at the N terminus and GIHRQ at the C terminus, was transformed into Escherichia coli BL21(DE3) cells. Natural abundance fusion protein was expressed, cleaved and SH3 was puri®ed as follows. One liter of cell culture in LB medium was grown in a 3L Fernbach ¯ask at 37  C with vigorous shaking to an A600 of 0.8, and protein expression was induced with addition of 0.595 g IPTG. The growth was continued for four hours before cells were harvested by centrifugation with two washes in R buffer (25 mM Hepes, 150 mM NaCl, 2.5 mM EDTA, pH 7.5). All subsequent procedures were carried out at 4  C. The cell pellet was resuspended in 30 ml R buffer containing 0.2 mM DTT, and 10 mg/ml aprotinin, 0.5 mM PMSF, 10 mg/ml pepstatin. The cells were lysed by two passes through a French pressure cell (900 PSIG), the cell lysate was cleared by centrifugation at 15,000 g for 30 minutes, and was then passed through a glutathione-Sepharose column (Pharmacia). The column was washed with 500 ml R buffer, and cleavage of the fusion protein was initiated by addition of ®ve units thrombin (Boerhinger-Mannheim) to the column. The cleavage reaction was allowed to proceed for eight to ten hours with constant mixing. The cut SH3 molecules were eluted with 20 ml R buffer. Thrombin was removed by passage through a 0.5 ml benzamidine-agarose column (Sigma). The protein was exchanged into the desired buffer and concentrated to approximately 0.1 mM using a Centriprep concentrator (3 kDa molecular weight cut-off, Amicon, Inc.). SH3 purity was determined to be 595 % by analytical HPLC and SDS-PAGE.

Microcalorimetry

ÿRT ln Ka ˆ
Go ˆ
Ho ÿ T
So where T is the absolute temperature and R is the universal gas constant, 1.987 cal/(mol K). The dissociation constant of the binding reaction, Kd ˆ 1/Ka, is used herein as a measure of ligand af®nity. Three independent measurements of the reaction were made, and the reported values represent the mean and standard deviation of the measurements.

884

Functional Dynamics in the c-Src SH3 Domain

Determination of relaxation parameters 

All NMR experiments were carried out at 25 C on a 600 MHz INOVA spectrometer equipped with a 13 C/15N/1H probe and Z-axis pulsed ®eld gradient. Backbone 15N and 1HN assignments were obtained using 15 N separated NOESY55,56 and TOCSY55 spectra, and were con®rmed with an HNCA spectrum.57,58 T1 and NOE measurements were performed using pulse sequences described by Farrow et al.59 The T2 sequence in the same paper was adapted according to Yamazaki et al.60 to measure T1r. All spectra were processed identically with nmrPipe using a cosine bell window function in both dimensions. For T1, the following time points were taken in the exact order: 0.011, 1.1, 0.088, 0.66, 0.165, 0.495, 0.33, 0.011, 0.088 seconds. Similarly, T1r time points were 0.008, 0.096, 0.016, 0.064, 0.024, 0.048, 0.032, 0.008, 0.016 seconds. Each relaxation parameter was measured at least twice (three measurements were obtained for unliganded SH3 T1r and NOE). Data sets were analyzed separately to evaluate the reproducibility of the resulting relaxation parameters. Peak volumes in the resulting 2D spectra were determined using non-linear least squares ®tting of lineshapes as implemented in the program nlinLS (Frank Delaglio, NIH/NIDDK). T1 and T1r0 values were determined by ®tting each peak volume to an exponential decay function. Optimal values of T1, 1r0 were determined by conjugate gradient minimization and the corresponding errors were generated by Monte-Carlo simulation.61 T1r values were obtained from the measured T1r0 values by correcting for off resonance effects.62 The reported errors for T1 and T1r are the larger of the standard deviation over repeated measurements or the square root of the sum of the squares of the Monte Carlo generated errors for repeated data sets. The steady-state NOE values were determined by the ratio of peaks with and without proton saturation and errors were estimated as the standard deviation over repeated measurements. Global tumbling analysis The relaxation of relatively rigid NH bond vectors is dominated by the global motion, and the T1/T1r ratio of such groups may be used to determine parameters for global tumbling that best ®t the data. A reduced data set from which mobile residues have been excluded is created for determination of the global tumbling parameters. A two-step ®lter, described in detail elsewhere22 is applied to the full data set to remove those residues whose relaxation is strongly affected by internal motions on the ps-ns and ms-ms time scales. Hence, the relaxation of the remaining residues in the reduced data set is dominated by the global tumbling. Anisotropic tumbling of the SH3 domain in both free and bound states was considered. The moment of inertia tensor for the SH3 domain structure was ®rst calculated using a high resolution X-ray crystal structure (PDB code 2src) and standard procedures. The orientation of this tensor relative to the molecular frame was used as a starting point in ®tting global tumbling parameters to the reduced data set. The global tumbling parameters were determined by minimizing w2g, the error-weighted difference between the experimental and calculated T1/ T1r ratios.63 The error in experimental T1/T1r was obtained by propagation of errors.61 The theoretical T1/ T1r ratio was calculated from standard analytical equations,64 using the form of the spectral density function for anisotropic global tumbling65 ± 67 and NH bond vector

orientations extracted from the 2src structure. The value of w2g was minimized for each of three global tumbling models (isotropic, axially symmetric anisotropic, and fully anisotropic). Coarse minima were ®rst identi®ed using a simulated annealing approach to search parameter space. Identi®ed minima were then re®ned using conjugate gradient minimization to optimize the principal axes of the diffusion tensor and a grid search to optimize the angular parameters.22 The appropriate model for global tumbling was determined using the statistical F-test.68 The probability that an improvement in the ®t due to increased model complexity is obtained purely by chance is given by a probability function P(F; n-m, N-n), and the percent con®dence that the improvement is signi®cant is given by (1-P)%.69 Analysis of internal motions Three models for motion consisting of the following parameters were considered: (1) the simple Lipari-Szabo (S2 and te),18,19, (2) the simple Lipari-Szabo with chemical exchange (S2, te, and Rex),24,25 and (3) the extended model (S2, S2f , and te, where S2 ˆ S2s S2f ).26 The simple Lipari-Szabo model was used if the resulting residual of the ®t, or w2 value, was smaller than 6.8, corresponding to the 99 % con®dence limit.70 The residual for a given residue is given by: w2 ˆ …T1c ÿ T1e †2 =sT1 ‡ …T1rc ÿ T1re †2 =sT1r ‡ …NOEc ÿ NOEe †2 =sNOE where subscripts c and e represent calculated and experimentally determined relaxation parameters, respectively, and sT1, sT1r, and sNOE are the estimates of the standard deviation of the experimentally determined relaxation parameters for the given residue. If the residual of the simple model exceeds 6.8, either the chemical exchange or extended model was considered. For either of these more complex models, a ®t was considered adequate if the corresponding residual was less than 1  10ÿ5. A residue was considered not ®t by any of the three models if none of the above criteria were met. In addition, any residues used in the ®nal determination of the global tumbling that require a threeparameter model of internal motion are considered underdetermined. For these residues, chemical exchange and anisotropic tumbling cannot be distinguished, and they are treated as not ®t by any model. The error estimates of the extracted parameters of motion were determined by Monte Carlo analysis as described elsewhere.71 The entire analysis, starting from values of T1, T2 and NOE and ending with internal motion parameters, was performed using a suite of programs collectively referred to as NORMAdyn (NMR; Optimized Relaxation Modeling with Anisotropy, for dynamics analysis) (N.H.P. and L.K.N., Cornell University). Hydrogen-deuterium exchange measurements Six milligrams of 15N labeled SH3 were dissolved into 600 ml NMR buffer (pH 6.5), and the solution was separated into two equal volumes. Four molar equivalents of RLP2 were added to one sample, and both samples were then lyophilized. Each sample was dissolved in 300 ml 2 H2O (99.996 %, Cambridge Isotope Labs) and 9.4 minute 15 N-1H HSQC spectra were acquired at 21 time points from 8.3 minutes to 196.3 minutes (value represents the time between 2H2O addition and the mid-point of the

Functional Dynamics in the c-Src SH3 Domain experiment duration). For RLP2-bound SH3, 16 additional 0.95 hour 15N-1H HSQC spectra were obtained from 3.93 to 18.18 hours. An 15N-1H HSQC spectrum for both free SH3 and RLP2-SH3 was also obtained at 19 hours, and an additional one for RLP2SH3 was taken at 124 hours. Peak volumes obtained at the different time points were ®t to an exponential decay function to yield the exchange time constant and its inverse, the observed rate of H/2H exchange (kex), for each peak. The theoretical interpretation of H/2H exchange rates has been described in detail elsewhere.42 Errors in measured kex values were estimated by jackknife simulations.72 Errors in kint were taken to be 34 % based on a comparison between theoretical and experimental rates examined by Koide et al.73 Errors in protection factors were propagated accordingly using standard methods.72

Acknowledgments We thank Frank Delaglio and Dan Garrett (NIH/ NIDDK) for use of their software tools, and Lewis Kay (U. Toronto) for use of his pulse sequence library. This research was funded by the National Science Foundation (grant no. MCB-9808727) and the National Cancer Institute (grant no. 1 R55 CA77478-01A1). N.H.P. and C.W. were supported by National Physical Sciences Consortium and Olin Foundation fellowships, respectively.

References 1. Cohen, G. B., Ren, R. B. & Baltimore, D. (1995). Modular binding domains in signal-transduction proteins. Cell, 80, 237-248. 2. Pawson, T. (1995). Protein modules and signalling networks. Nature, 373, 573-580. 3. Xu, W., Harrison, S. C. & Eck, M. J. (1997). Threedimensional structure of the tyrosine kinase c-Src. Nature, 385, 595-602. 4. Bjorge, J. D., Jakymiw, A. & Fujita, D. J. (2000). Selected glimpses into the activation and function of Src kinase. Oncogene, 19, 5620-5635. 5. Kuriyan, J. & Cowburn, D. (1997). Modular peptide recognition domains in eukaryotic signaling. Annu. Rev. Biophys. Biomol. Struct. 26, 259-288. 6. Dalgarno, D. C., Bot®eld, M. C. & Rickles, R. J. (1997). SH3 domains and drug design: ligands, structure, and biological function. Biopolymers, 43, 383-400. 7. Feng, S., Chen, J. K., Yu, H., Simon, J. A. & Schreiber, S. L. (1994). Two binding orientations for peptides to the Src SH3 domain: development of a general model for SH3-ligand interactions. Science, 266, 1241-1247. 8. Feng, S., Kasahara, C., Rickles, R. J. & Schreiber, S. L. (1995). Speci®c interactions outside the proline-rich core of two classes of Src homology 3 ligands. Proc. Natl Acad. Sci. USA, 92, 12408-12415. 9. Yu, H., Rosen, M. K., Shin, T. B., Seidel-Dugan, C., Brugge, J. S. & Schreiber, S. L. (1992). Solution structure of the SH3 domain of Src and identi®cation of its ligand-binding site. Science, 258, 1665-1668. 10. Cordier, F., Wang, C., Grzesiek, S. & Nicholson, L. K. (2000). Ligand-induced strain in the c-Src SH3 domain detected by NMR. J. Mol. Biol. 304, 497-505.

885 11. Renzoni, D. A., Pugh, D. J., Siligardi, G., Das, P., Morton, C. J., Rossi, C. et al.(1996). Structural and thermodynamic characterization of the interaction of the SH3 domain from Fyn with the proline-rich binding site on the p85 subunit of PI3-kinase. Biochemistry, 35, 15646-15653. 12. Lemmon, M. A. & Ladbury, J. E. (1994). Thermodynamic studies of tyrosyl-phosphopeptide binding to the SH2 domain of P56(Lck). Biochemistry, 33, 5070-5076. 13. Bracken, C., Carr, P. A., Cavanagh, J. & Palmer, A. G., III (1999). Temperature dependence of intramolecular dynamics of the basic leucine zipper of GCN4: implications for the entropy of association with DNA. J. Mol. Biol. 285, 2133-2146. 14. Kay, L. E., Muhandiram, D. R., Wolf, G., Shoelson, S. E. & Forman-Kay, J. D. (1998). Correlation between binding and dynamics at SH2 domain interfaces. Nature Struct. Biol. 5, 156-162. 15. Lee, A. L., Kinnear, S. A. & Wand, A. J. (2000). Redistribution and loss of side-chain entropy upon formation of a calmodulin-peptide complex. Nature Struct. Biol. 7, 72-77. 16. Zidek, L., Novotny, M. V. & Stone, M. J. (1999). Increased protein backbone conformational entropy upon hydrophobic ligand binding. Nature Struct. Biol. 6, 1118-1121. 17. Lu, J., Lin, C. L., Tang, C., Ponder, J. W., Kao, J. L., Cistola, D. P. & Li, E. (2000). Binding of retinol induces changes in rat cellular retinol-binding protein II conformation and backbone dynamics. J. Mol. Biol. 300, 619-632. 18. Lipari, G. & Szabo, A. (1982). Model-free approach to the interpretation of nuclear magnetic resonance relaxation in macromolecules. 2. Analysis of experimental results. J. Am. Chem. Soc. 104, 4559-4570. 19. Lipari, G. & Szabo, A. (1982). Model-free approach to the interpretation of nuclear magnetic resonance relaxation in macromolecules. 1. Theory and range of validity. J. Am. Chem. Soc. 104, 4546-4559. 20. Akke, M., Bruschweiler, R. & Palmer, A. G. (1993). NMR order parameters and free-energy-an analytical approach and its application to cooperative Ca2‡ binding by calbindin- D(9k). J. Am. Chem. Soc. 115, 9832-9833. 21. Yang, D. W. & Kay, L. E. (1996). Contributions to conformational entropy arising from bond vector ¯uctuations measured from NMR-derived order parameters: Application to protein folding. J. Mol. Biol. 263, 369-382. 22. Pawley, N. H., Wang, C., Koide, S. & Nicholson, L. K. (2001). An improved method for distinguishing between anisotropic tumbling and chemical exchange in analysis of 15N relaxation parameters. J. Biomol. NMR, 20, 149-165. 23. Gagne, S. M., Tsuda, S., Spyracopoulos, L., Kay, L. E. & Sykes, B. D. (1998). Backbone and methyl dynamics of the regulatory domain of troponin C: anisotropic rotational diffusion and contribution of conformational entropy to calcium af®nity. J. Mol. Biol. 278, 667-686. 24. Szyperski, T., Luginbuhl, P., Otting, G., Guntert, P. & Wuthrich, K. (1993). Protein dynamics studied by rotating frame 15N spin relaxation times. J. Biomol. NMR, 3, 151-164. 25. Farrar, T. C. & Becker, E. D. (1971). Pulse and Fourier Transform NMR, Academic Press, New York. 26. Clore, G. M., Szabo, A., Bax, A., Kay, L. E., Driscoll, P. C. & Gronenborn, A. M. (1990). Deviations from

886

27.

28.

29.

30.

31. 32.

33.

34.

35.

36.

37. 38.

39.

40.

the simple two-parameter model-free approach to the interpretation of nitrogen-15 nuclear magnetic relaxation of proteins. J. Am. Chem. Soc. 112, 49894991. Horita, D. A., Zhang, W., Smithgall, T. E., Gmeiner, W. H. & Byrd, R. A. (2000). Dynamics of the HckSH3 domain: comparison of experiment with multiple molecular dynamics simulations. Protein Sci. 9, 95-103. Hansson, H., Mattsson, P. T., Allard, P., Haapaniemi, P., Vihinen, M., Smith, C. I. & Hard, T. (1998). Solution structure of the SH3 domain from Bruton's tyrosine kinase. Biochemistry, 37, 2912-2924. Farrow, N. A., Zhang, O., Forman-Kay, J. D. & Kay, L. E. (1997). Characterization of the backbone dynamics of folded and denatured states of an SH3 domain. Biochemistry, 36, 2390-2402. Xu, W., Doshi, A., Lei, M., Eck, M. J. & Harrison, S. C. (1999). Crystal structures of c-Src reveal features of its autoinhibitory mechanism. Mol. Cell, 3, 629-638. Lovell, S. C., Word, J. M., Richardson, J. S. & Richardson, D. C. (2000). The penultimate rotamer library. Proteins: Struct. Funct. Genet. 40, 389-408. D'Aquino, J. A., Freire, E. & Amzel, L. M. (2000). Binding of small organic molecules to macromolecular targets: evaluation of conformational entropy changes. Proteins Struct. Funct. Genet. Suppl(4), 93107. Crump, M. P., Spyracopoulos, L., Lavigne, P., Kim, K.-S., Clark-Lewis, I. & Sykes, B. D. (1999). Backbone dynamics of the human CC chemokine eotaxin: fast motions, slow motions, and implications for receptor binding. Protein Sci. 8, 20412054. Fehrer, V. A. & Cavanagh, J. (1999). Millisecondtime-scale motions contribute to the function of the bacterial response regulator protein SpoOF. Nature, 400, 289-293. Gao, G., Semenchenko, V., Arumugam, S. & Van Doren, S. R. (2000). Tissue inhibitor of metalloproteinases-1 undergoes microsecond to millisecond motions at sites of matrix metalloproteinase-induced ®t. J. Mol. Biol. 301, 537-552. Wyss, D. F., Dayie, K. T. & Wagner, G. (1997). The counterreceptor binding site of human CD2 exhibits an extended surface patch with mutliple conformations ¯uctuating with millisecond to microsecond motions. Protein Sci. 6, 534-542. Ye, J., Mayer, K. L. & Stone, M. J. (1999). Backbone dynamics of the human CC-chemokine eotaxin. J. Biomol. NMR, 15, 115-124. Wittekind, M., Mapelli, C., Lee, V., Goldfarb, V., Friedrichs, M. S., Meyers, C. A. & Mueller, L. (1997). Solution structure of the Grb2 N-terminal SH3 domain complexed with a ten-residue peptide derived from SOS: direct re®nement against NOEs, J- couplings and 1H and 13C chemical shifts. J. Mol. Biol. 267, 933-952. Mandel, A. M., Akke, M. & Palmer, A. G. I. (1996). Dynamics of ribonuclease H: temperature dependence of motions on multiple time scales. Biochemistry, 33, 16009-16023. Bai, Y., Milne, J. S., Mayne, L. & Englander, S. W. (1993). Primary structure effects on peptide group hydrogen-exchange. Proteins: Struct. Funct. Genet. 17, 75-86.

Functional Dynamics in the c-Src SH3 Domain 41. Grantcharova, V. P. & Baker, D. (1997). Folding dynamics of the src SH3 domain. Biochemistry, 36, 15685-15692. 42. Bai, Y., Sosnick, T. R., Mayne, L. & Englander, S. W. (1995). Protein folding intermediates: native-state hydrogen exchange. Science, 269, 192-197. 43. Martinez, J. C. & Serrano, L. (1999). The folding transition state between SH3 domains is conformationally restricted and evolutionarily conserved. Nature Struct. Biol. 6, 1010-1016. 44. Williams, D. C., Jr, Benjamin, D. C., Poljak, R. J. & Rule, G. S. (1996). Global changes in amide hydrogen exchange rates for a protein antigen in complex with three different antibodies. J. Mol. Biol. 257, 866876. 45. Engen, J. R., Smithgall, T. E., Gmeiner, W. H. & Smith, D. L. (1997). Identi®cation and localization of slow, natural, cooperative unfolding in the hematopoietic cell kinase SH3 domain by amide hydrogen exchange and mass spectrometry. Biochemistry, 36, 14384-14391. 46. Freire, E. (1999). The propagation of binding interactions to remote sites in proteins: analysis of the binding of the monoclonal antibody D1. 3 to lysozyme. Proc. Natl Acad. Sci. USA, 96, 10118-10122. 47. Eftink, M. R., Anusiem, A. C. & Biltonen, R. L. (1983). Enthalpy-entropy compensation and heat capacity changes for protein-ligand interactions: general thermodynamic models and data for the binding of nucleotides to ribonuclease A. Biochemistry, 22, 3884-3896. 48. Sadqi, M., Casares, S., Abril, M. A., Lopez-Mayorga, O., Conejero-Lara, F. & Freire, E. (1999). The native state conformational ensemble of the SH3 domain from alpha-spectrin. Biochemistry, 38, 8899-8906. 49. Camarero, J. A. & Muir, T. W. (1999). Biosynthesis of a head-to-tail cyclized protein with improved biological activity. J. Am. Chem. Soc. 121, 5597-5598. 50. Volkman, B. F., Lipson, D., Wemmer, D. E. & Kern, D. (2001). Two-state allosteric behavior in a singledomain signaling protein. Science, 291, 2429-2433. 51. Tessari, M., Gentile, L. N., Taylor, S. J., Shalloway, D. I., Nicholson, L. K. & Vuister, G. W. (1997). Heteronuclear NMR studies of the combined Src homology domains 2 and 3 of pp60 c-Src: effects of phosphopeptide binding. Biochemistry, 36, 1456114571. 52. Reinking, J. L., Gentile, L. N., Tessari, M., Vuister, G. & Nicholson, L. K. (1998). Mechanisms of regulation in the regulatory apparatus of pp60 c-Src. In Structure, Motion, Interaction and Expression of Biological Macromolecules. Proceedings of the Tenth Conversation in Biomolecular Stereodynamics, State University of New York, Albany, NY 1997 (Sarma, R. H. & Sarma, M. H., ed.), Adenine Press, Schenectady, NY, pp. 53-61. 53. Wiseman, T., Williston, S., Brandts, J. F. & Lin, L. N. (1989). Rapid measurement of binding constants and heats of binding using a new titration calorimeter. Anal. Biochem. 179, 131-137. 54. Sigurskjold, B. W., Altman, E. & Bundle, D. R. (1991). Sensitive titration microcalorimetric study of the binding of Salmonella O-antigenic oligosaccharides by a monoclonal-antibody. Eur. J. Biochemistry, 197, 239-246. 55. Marion, D., Driscoll, P. C., Kay, L. E., Wing®eld, P. T., Bax, A., Gronenborn, A. M. & Clore, G. M. (1989). Overcoming the overlap problem in the assignment of 1H NMR spectra of larger proteins by

887

Functional Dynamics in the c-Src SH3 Domain

56.

57. 58.

59.

60.

61.

62.

63.

use of three-dimensional heteronuclear 1H-15N Hartmann-Hahn-multiple quantum coherence and nuclear Overhauser-multiple quantum coherence spectroscopy: application to interleukin 1 beta. Biochemistry, 28, 6150-6156. Fesik, S. W. & Zuiderweg, E. R. P. (1988). Heteronuclear 3-dimensional nmr-spectroscopy - a strategy for the simpli®cation of homonuclear two-dimensional nmr-spectra. J. Magn. Reson. 78, 588-593. Grzesiek, S. & Bax, A. (1992). Improved 3d triple-resonance nmr techniques applied to a 31-Kda protein. J. Magn. Reson. 96, 432-440. Ikura, M., Kay, L. E. & Bax, A. (1990). A novelapproach for sequential assignment of H-1, C-13, and N-15 spectra of larger proteins-heteronuclear triple-resonance 3-dimensional nmr-spectroscopyapplication to calmodulin. Biochemistry, 29, 46594667. Farrow, N. A., Muhandiram, R., Singer, A. U., Pascal, S. M., Kay, C. M., Gish, G. et al. (1994). Backbone dynamics of a free and a phosphopeptide-complexed Src homology-2 domain studied by N-15 nmr relaxation. Biochemistry, 33, 5984-6003. Yamazaki, T., Muhandiram, R. & Kay, L. E. (1994). NMR experiments for the measurement of carbon relaxation properties in highly enriched, uniformly 13C, 15N-labeled proteins: application to 13C alpha carbons. J. Am. Chem. Soc. 116, 8266-8278. Nicholson, L. K., Kay, L. E., Baldisseri, D. M., Arango, J., Young, P. E., Bax, A. & Torchia, D. A. (1992). Dynamics of methyl groups in proteins as studied by proton-detected 13C NMR spectroscopy. Application to the leucine residues of staphylococcal nuclease. Biochemistry, 31, 5253-5263. Akke, M. & Palmer, A. G. (1996). Monitoring macromolecular motions on microsecond to millisecond time scales by R(1)rho-R(1) constant relaxation time NMR spectroscopy. J. Am. Chem. Soc. 118, 911-912. Palmer, A. G., Wright, P. E. & Rance, M. (1991). Measurement of relaxation-time constants for

64.

65.

66.

67. 68. 69. 70.

71.

72. 73.

methyl-groups by proton-detected heteronuclear nmr-spectroscopy. Chem. Phys. Letters, 185, 41-46. Kay, L. E., Torchia, D. A. & Bax, A. (1989). Backbone dynamics of proteins as studied by 15N inverse detected heteronuclear NMR spectroscopy: application to staphylococcal nuclease. Biochemistry, 28, 8972-8979. Tjandra, N., Wing®eld, P., Stahl, S. & Bax, A. (1996). Anisotropic rotational diffusion of perdeuterated HIV protease from N-15 NMR relaxation measurements at two magnetic ®eld strengths. J. Biomol. NMR, 8, 273-284. Lee, L. K., Rance, M., Chazin, W. J. & Palmer, A. G. (1997). Rotational diffusion anisotropy of proteins from simultaneous analysis of N-15 and C-13(alpha) nuclear spin relaxation. J. Biomol. NMR, 9, 287-298. Woessner, D. E. (1962). Nuclear spin relaxation in ellipsoids undergoing rotational brownian motion. J. Chem. Phys. 37, 647-654. Bevington, P. R. & Robinson, D. K. (1992). Data Reduction and Error Analysis for the Physical Sciences, 2nd edit., McGraw-Hill, Boston, MA. Devore, J. (1982). Probability and Statistics for Engineering and the Sciences, Brooks/Cole, Monterey, CA. Press, W. H., Teukolsky, S. A., Vetterling, W. T. & Flannery, B. P. (1996). Numerical Recipes in C: The Art of Scienti®c Computing, 2nd edit., Press Syndicate of the University of Cambridge, New York, NY. Nicholson, L. K., Yamazaki, T., Torchia, D. A., Grzesiek, S., Bax, A., Stahl, S. J. et al. (1995). Flexibility and function in Hiv-1 protease. Nature Struct. Biol. 2, 274-280. Mosteller, F. & Tukery, J. W. (1977). Data Analysis and Regression. A Second Course in Statistics, Addison-Wesley, Reading, MA. Koide, S., Jahnke, W. & Wright, P. E. (1995). Measurement of intrinsic exchange rates of amide protons in a 15N- labeled peptide. J. Biomol. NMR, 6, 306-312.

Edited by P. E. Wright (Received 21 May 2001; received in revised form 10 September 2001; accepted 11 September 2001)