Lee et al. (2000) Nat. Struct. Biol. - Perelman School of Medicine at the ...

© 2000 Nature America Inc. • http://structbio.nature.com

articles

Redistribution and loss of side chain entropy upon formation of a calmodulin–peptide complex Andrew L. Lee, Sandra A. Kinnear and A. Joshua Wand


The response of the internal dynamics of calcium-saturated calmodulin to the formation of a complex with a peptide model of the calmodulin-binding domain of the smooth muscle myosin light chain kinase has been studied using NMR relaxation methods. The backbone of calmodulin is found to be unaffected by the binding of the domain, whereas the dynamics of side chains are significantly perturbed. The changes in dynamics are interpreted in terms of a heterogeneous partitioning between structure (enthalpy) and dynamics (entropy). These data provide a microscopic view of the residual entropy of a protein in two functional states and suggest extensive enthalpy/entropy exchange during the formation of a protein–protein interface.

Molecular recognition is fundamental to the mechanism for the transduction of cellular signals. Understanding how these signals propagate into their preprogrammed responses requires the full characterization of the thermodynamics underlying the formation of protein/protein interfaces, across which signal integrity is maintained. Protein–protein interactions involve a complex interplay between structure, dynamics and the fundamental thermodynamic parameters ∆H, ∆S, ∆Cp and ∆G1–5. A variety of experimental approaches can be employed to separate some of the various contributions to overall binding affinity. Enthalpy and heat capacity changes can be directly measured using calorimetric techniques. Alternatively they can be estimated from changes in solvent-accessible surface area using structural coordinates4. Regardless of the method used, an obvious and central component of the free energy of protein association, the change in conformational entropy of the protein (∆Sconf), is usually only inferred from the other quantities. As residual conformational entropy in proteins is quite substantial5–7 and apparently critical for function8–10, direct characterization of this elusive component of free energy is necessary for understanding the energetics of protein–protein interactions. In principle, characterization of internal protein dynamics should facilitate an independent and direct measurement of local and overall conformational entropy changes11–13. We have used NMR methods to examine the response of the picosecondnanosecond dynamics of the main chain and of methyl-bearing side chains of calmodulin to the formation of a complex with a peptide model of the calmodulin-binding domain of the smooth muscle myosin light chain kinase (smMLCKp). Calmodulin is the primary regulatory protein mediating target protein activities in response to fluctuations in intracellular calcium levels 14. Calmodulin has two globular domains, each containing a pair of helix-loop-helix motifs (EF-hands) that bind calcium ions15. The N- and C-terminal domains are connected by a central helix that is dynamically disordered at its center16. The binding of four Ca2+ ions by calmodulin results in the creation of a binding surface that forms an interface with the target protein15,17,18. In the calcium-saturated state (CaM), each domain contains a large

hydrophobic cavity lined with four methionine residues. In the four examples known in structural detail, binding of CaM to the calmodulin-binding domain of a target protein results in a significant reorganization of the central helix such that the two domains of CaM bury the vast majority of the target domain19–22. Although there is a major change in the relative orientations of the two domains, only small structural changes occur within the N- and C-terminal domains. Large hydrophobic side chains of the peptide insert into the hydrophobic cavities of CaM, and

Fig. 1 Illustration of various motions affecting the obtained generalized order parameters for methyl deuterons. Shown is a fragment of a leucine or valine amino acid side chain containing the isopropyl group. The trivial motion about the symmetry axis of the methyl group occurs on a subnanosecond time scale. Torsional oscillations about the Cβ–Cγ bond are often described using the Gaussian axial fluctuation model49 having a characteristic standard deviation of σ. In that model, the methyl group symmetry axis would follow the indicated line. More complicated motions, arising from additional remote torsional oscillations and bond librations, will result in deviations from the path defined by the indicated simple rotation and would result in a distribution of motions within the elliptical cone shown. The square of the generalized order parameter for the symmetry axis of the leucine δ-methyl groups (S2axis) reports on these kinds of motions occurring on time scales up to ∼10 ns. Similar motions are anticipated for other methyl-bearing amino acid residues.

The Johnson Research Foundation, Department of Biochemistry and Biophysics, School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA. Correspondence should be addressed to A.J.W. email: [email protected] 72

nature structural biology • volume 7 number 1 • january 2000


articles a


b

ing to a Gaussian torsion angle distribution of ∼60° (Fig. 1). The lower values are, on average, significantly lower than those found for other proteins, especially for leucine and isoleucine methyl groups31. Evidently, CaM is an uncommonly dynamic protein on a picosecondnanosecond time scale at the side chain level, but not at the backbone level as seen from 15N relaxation16 (see below). This novel behavior may in part result from CaM’s unusual hydrophobic core. Because the hydrophobic contacts are spread along the edges of the pocket formed between helices, there is apparently less restriction for hydrophobic side chain mobility than for interior side chains in canonical globular proteins. The effective internal correlation times (τe) ranged from 7 to 116 ps in CaM.

Reduced flexibility of CaM upon peptide binding When complexed with the smMLCKp peptide, 58 out of 79 methyl groups in CaM were sufficiently resolved for reliable analysis. As in free CaM and other proteins, heterogeneous methyl dynamics are Fig. 2 Methyl symmetry axis order parameters (S2axis) for methyl groups in CaM. a, The calcium-sat- observed (Fig. 2b). However, the order urated state; b, the calcium-saturated state complexed with smMLCK peptide. Methionine Cε parameters are, on average, higher than methyl groups are shown as red bars. Missing data results from spectral overlap of methyl resofor unbound CaM, reflecting an overall nances. loss of mobility of the side chains of CaM upon binding the smMLCKp domain. The overall values of S2axis in the complex are many other hydrophobic and ion pair interactions are formed more typical of proteins previously studied and are perhaps between CaM and the target domain. It is the thermodynamic indicative of the more globular nature of the CaM–smMLCKp nature of this interface that is of interest here. complex. Nevertheless, despite significant rigidification upon binding, the S2axis values of methyl-bearing amino acids of NMR characterization of methyl dynamics calmodulin indicate the presence of considerable residual Using 2H spin relaxation methods23, the degree of spatial restric- motion at the interface. The effective internal correlation times tion of a given methyl group (Fig. 1) was assessed via the square ranged from 5 to 88 ps in the CaM–smMLCKp complex. of the model-free generalized order parameter24 of its symmetry A difference plot (Fig. 3a) readily illustrates the changes in axis (S2axis). Values of the S2axis parameter can range from 0 to 1, S2axis values brought about by formation of the complex. Because corresponding to isotropic disorder and a fixed orientation in different sets of methyl resonances were overlapped in the free the molecular frame, respectively. For each generalized order and bound states, only 41 out of 79 methyl S 2axis values could be parameter, a corresponding effective internal correlation time compared. On average the S2axis parameters increase by 0.07 upon (τe) was obtained. We choose to interpret the latter parameter formation of the complex. The dynamic response to the binding only qualitatively, since its interpretation is not always straight- of smMLCKp is nevertheless a mixed one, varying from decreasforward in the context of methyl dynamics examined by deuteri- es in S2axis for a few valine and leucine methyls to large increases um relaxation25. Symmetry axis generalized order parameters in methionine methyls. The significant reorganization of side and internal correlation times for methyl groups in calcium-sat- chain dynamics that accompanies binding contrasts with the flat urated calmodulin and the 1:1 CaM–smMLCKp complex were backbone response as observed by 15N relaxation (Fig. 3b). This obtained. Calcium-saturated calmodulin binds to smMLCKp is particularly noteworthy in light of the relatively small structurwith high affinity (∼1 nM KD)26,27, and the complex is in the al response of the individual domains of CaM to the smMLCKp slow-exchange regime on the NMR time scale28–30. domain15,19–21. Clearly, it is the side chains that report on the protein’s unique dynamic behavior in ‘structured’ regions, whereas Ca2+-CaM methyl dynamics the backbone of CaM, in either functional state, displays the norIn the case of free CaM, 51 out of 79 methyl resonances were suf- mal order parameters (∼0.85) seen in so many other systems. ficiently resolved in the 13C-1H correlation spectrum to permit These observations underscore the effective independence of reliable analysis. The S2axis values indicate that the side chains of backbone and side chain dynamics that can occur in proteins. CaM exhibit a wide range of motional amplitudes (Fig. 2a). The most pronounced changes in methyl-bearing amino acid Values range from ∼1, corresponding to effectively no motion of side chain dynamics are found in methionine residues of CaM. the methyl group symmetry axis, to as low as ∼0.2, correspond- Such a general effect on calmodulin has been anticipated32–34, nature structural biology • volume 7 number 1 • january 2000

73



articles although the heterogeneous response was a unexpected. For example, residues Met 124 and Met 72 become much more rigid (∆S2axis > 0.5) upon peptide binding. Met 124 has already been identified as important for activation of smMLCK and is the methionine residue least permissive to mutation35,36. Met 72 has also been shown to be important for activation of smMLCK35. These intriguing correlations indicate the presence of functionally important interconversion of dynamics (local entropy) and structure (enthalpy) for Met 72 and Met 124. In contrast to the significant rigidification of these two residues (S2axis ∼0.8), other methionines remain relatively mobile in the complex (S2axis ∼0.4). The favorable enthalpic contribution inferred from the tight packing of Met 72 and Met 124 in van der Waals contacts (bonds) apparently compensates for the loss of local entropy upon forma- b tion of the complex. Similarly, the loss of hydration of exposed hydrophobic groups such as methionine is compensated by the gain in solvent entropy upon formation of the complex (that is, the classic hydrophobic effect). Thus, the CaM–smMLCKp complex illustrates enthalpy/entropy exchange on both sides of the thermodynamic equilibrium. This view also illustrates how flexibility in CaM may allow Fig. 3 Differences in order parameters (bound - free) in CaM. a, Side-chain methyl groups (∆S2axis); accommodation of significantly different b, backbone NH groups (∆S2). Secondary structure elements are indicated with solid lines. The dashed line represents the region of the ‘central helix’ of free CaM that is disordered in solution peptide sequences for high-affinity bind- and significantly reorganized upon complex formation with the smMLCKp calmodulin-binding ing. A deeper understanding of this will domain peptide. Methionine Cε methyl groups are shown as red bars. require further investigations of other peptide sequences and CaM mutants. Another group of methyl-bearing amino acid side chains with face. Notably, however, a few side chains (Val 55, Leu 116 and significantly perturbed dynamics are the residues found deep in Val 142) gain mobility upon binding of the smMLCKp domain the major hydrophobic pocket of each CaM domain: Ile 27, (Fig. 4). All of these residues are significantly solvent exposed. As Ile 63, Ile 100 and Val 136. These central residues structurally illustrated below, such an increase in dynamics corresponds to link the EF-hand pairs of each domain15. Binding of the an increase in conformational entropy and contributes to a stabismMLCKp domain to calmodulin significantly stabilizes this lization of the complex. The present observations are interesting in light of a recent interaction28. Although proper ∆S2axis values could be obtained for only Ile 27γ2, Ile 63γ2 and Ile 63δ, approximate ∆S2axis values (or study of the response of the methyl dynamics of the C-terminal lower limits) for the remaining methyls of this group of amino SH2 domain of phospholipase C-γ1 to the binding of a phosacids can be obtained. Previous experience31 suggests that pro-R photyrosine peptide substrate37. In that study, the ∆S2axis values and pro-S valine methyl S2axis values are generally similar and observed (≤0.10–0.15) were comparatively smaller than found that isoleucine δ methyl S2axis values are usually less than or equal here and were localized to the phosphotyrosine peptide-bindto those of the corresponding γ-methyls. Accordingly, the ing site. In contrast, significant ∆S2axis values occur throughout remaining methyls of Ile 27 and Ile 100 can be estimated to have the CaM structure. The different behaviors observed in these ∆S2axis values between 0.1 and 0.2, and S2axis values in the bound two protein–peptide systems may represent real differences in state in the 0.8–0.85 range. By this argument, Val 136 appears to binding strategies. One obvious difference between the systems have a ∆S2axis of ∼0. Thus, in general, the stabilization of the inter- is that the CaM–smMLCKp interface (∼2,000 Å2) is significantEF-hand interaction by the binding of the target peptide is ly larger than the interface formed between PLCC-γ1 SH2 and reflected in the behavior of associated side chains. When the its phosphopeptide substrate (∼600 Å2). In addition, the ligand smMLCKp domain binds, the side chains along the back walls of domain in the CaM–smMLCKp complex is essentially buried, both major hydrophobic pockets of CaM become about as rigid having a residual accessible surface area (ASA) of ∼720 Å2, as the backbone (Fig. 2b). This is illustrated in Fig. 4, in which whereas significantly more surface area of the smaller ligand S2axis values are color-coded and mapped onto the structure of peptide remains solvent accessible in the SH2 complex (ASA the complex. Finally, there appears to be no clear correlation ∼1,040 Å2). No significant correlation between S2axis values and solventbetween ∆S2axis values and proximity to the binding interface, although the largest positive ∆S2axis values do occur at the inter- accessible surface area was observed for free or bound CaM, 74



articles


a

b

are solvent exposed, the τe values are narrowly distributed around 8–10 ps. Burial in the complexed state apparently produces local environments that are sufficiently distinct to be reflected in an increased dispersion of τe values (Fig. 5).

Energetics of binding The results presented here help to illustrate the breadth and complexity of the fast side chain dynamics of proteins and of protein–protein interfaces. It appears that CaM possesses greater than average side chain mobility, which is considerably reduced upon binding to the smMLCKp domain. This is indicative of a reduction in conformational entropy of calmodulin upon formation of the complex (∆Sconf < 0)12,13. Using a simple independent harmonic oscillator as a model for the underlying motion12 and assuming that nonmethyl-bearing residues respond as methyl-bearc d ing residues, one may estimate from an average per-residue ∆S2axis of 0.07 a corresponding T∆Sconf for CaM on the order of -35 kcal mol-1 at 35 °C. Though limited by the implicit assumption that all motions are uncorrelated and reflected in the measured ∆S2axis values, the result of this simple treatment agrees remarkably well with the calorimetrically determined T∆Sconf of -50 to -100 kcal mol-1 estimated by Wintrode and Privalov1. As pointed out by these authors1 and others2,3, identification of the change in internal entropy of proteins upon complex formation is fraught with difficulty and is sensitive to errors in estimates of Fig. 4 Color-coded representation of the side chain dynamics of calmodulin. S2axis values rotational-translational entropy and internal conare color-mapped onto individual methyl groups in the three-dimensional structures of formational entropy of disordered polypeptide free CaM (PDB code, 3cln) and of CaM complexed to smMLCKp (PDB code, 1cdl). S2axis values range from 0 (white) to 0.5 (cyan) to 1 (blue) on a continuous scale as shown in the chains such as the uncomplexed smMLCKp color legend at the top of the figure. Methyl groups not characterized because of spec- domain. Therefore, the qualitative agreement here tral overlap are shown in yellow. Methyl groups that undergo significant changes in S2axis is encouraging, as is the agreement between NMR values upon peptide binding are annotated, as are all methionine methyl groups. and calorimetrically derived conformational a, N-terminal domain of CaM. b, C-terminal domain of CaM. c, N-terminal domain of CaM–smMLCKp complex. d, C-terminal domain of CaM smMLCKp complex. In panels (c) entropy changes observed for the basic leucine zipand (d) the smMLCKp peptide is shown in green. The figure was prepared with MolMol50. per of GCN4 upon binding DNA41. We note that the T∆Sconf value of -35 kcal mol-1 for CaM upon binding peptide is significantly larger than the free although one might intuitively expect these parameters to be energy change of complex formation (-12 kcal mol-1). correlated. A lack of significant correlation has also been seen Consequently, changes or reorganizations of internal dynamic in previous studies31,37,38. There is no hint of the heterogeneous motions appear to represent a viable mechanism for the moduladynamical response in the B-factors obtained during crystallo- tion of binding affinities. graphic analysis. Finally, no significant correlations were Beyond providing a route to estimating the overall entropic found between S2axis (or τe) values and methyl group packing cost to CaM as a result of binding the smMLCKp domain, the efficiencies39. deuterium relaxation data perhaps more obviously yield siteAlthough it is difficult to interpret τe values, it is worth noting specific estimates of residual conformational entropy. that there is a slight overall decrease in τe values in CaM methyl Notwithstanding the aforementioned caveats with respect to groups upon binding smMLCKp (data not shown). In the con- correlated motions, the pervasive redistribution of motion text of 2H relaxation, obtained τe values are dominated by methyl (residual entropy) appears to reflect the complexity underlying rotation rates25 so that binding of the smMLCKp domain might the thermodynamic role of CaM dynamics in the binding of appear to facilitate methyl rotation. Alternatively, a small smMLCKp. Obviously, the spatial distribution of entropy would decrease in effective correlation times could reflect a more sub- not be apparent from an overall ∆S measurement. Such site-spestantial increase in side chain vibrational-librational frequencies cific quantification of both enthalpic and entropic contributions concomitant with a ‘stiffening’ of the dynamics upon substrate to protein stability and binding affinity should prove useful to binding40. A more reliable characterization of site-specific fre- issues underlying protein and drug design. Knowledge of such quency changes in methyl dynamics would entail analysis of patterns will facilitate dissection of thermodynamic parameters field-dependent 13C relaxation rates25. Also potentially revealing into sequence- and context-dependent contributions. is the response of the various methionine τe values to the binding An idea that has been widely expressed in the literature is that of the smMLCKp domain. In free CaM, in which all methionines CaM’s broad target specificity originates from the ‘methionine nature structural biology • volume 7 number 1 • january 2000

75



articles puddles’32, with the inherent malleability of methionines allowing CaM to adapt its interface locally to complement a range of target sequences. It is now apparent that the free state is generally flexible and unusually so, whereas the bound state has distinctive regions of flexibility (Fig. 3a). Some methionines have unpredictably high order parameters (Met 72, Met 124) whereas others retain much of the flexibility characteristic of the free state. It seems likely that the locations of flexibility and rigidity at the interface will change as a function of target peptide sequence. Although significant changes in S2axis occur throughout CaM upon complex formation, the larger increases tend to cluster at the binding interface (Fig. 4). Clustered responses may indicate localized ‘hot spots’ of binding free energy42. Clusters of increased rigidity (decrease in entropy) seem to indicate localization of significant binding enthalpy11–13,42. Indeed, it has been illustrated that for complexes involving multiple weak interactions (∆E ∼kT), extensive enthalpy/entropy compensation is anticipated5. As described, the side chains forming the back wall of the hydrophobic pockets in CaM become collectively rigid upon peptide binding. In the context of Weber’s original analysis5, the results here are quite different than those obtained for the association of subunits of naturally oligomeric proteins. In that case, the association is generally an entropydriven process, with extensive gains in entropy being obtained from the conversion of protein–water interactions to protein–protein interactions, as well as the more usually recognized water–water interactions. The underlying enthalpy/entropy exchange results from bonds with energies on the order of kT and a gain in protein entropy is derived when protein–water bonds are nominally stronger than protein–protein bonds. The reverse is observed here and indicates that the interfacial protein–protein bonds in the CaM–smMLCKp complex are generally stronger than their protein–water counterparts in the free components. This may be pointing to a feature that distinguishes protein complexes where the components are not designed to stably exist in isolation from those that are so designed. It is concluded that functionally relevant partitioning between structure (enthalpy) and dynamics (entropy) occurs primarily on the side chains. These data represent the first demonstration of microscopic enthalpy/entropy exchange in the formation of a protein–protein interface. Because peptide binding appears to be entropically unfavorable, these results are consistent with the formation of the complex being an enthalpically driven process1. Methods Samples. Chicken calmodulin was prepared in a similar manner to that described43. Escherichia coli BL21(DE3) cells were grown on minimal media containing 50% H2O, 50% D2O, and using 15NH4Cl and D-glucose (U-13C6-99%) as the sole nitrogen and carbon sources, respectively. The final NMR sample conditions for Ca2+CaM were 1.0 mM protein, 100 mM KCl, 5.5 mM total CaCl2 concentration, 10 mM imidazole-d4, pH 6.5, and 0.02% sodium azide, 90% H2O/10% D2O. The chicken smooth muscle myosin light chain kinase peptide (smMLCKp, GSARRKWQKTGHAVRAIGRLS) was subcloned behind a T7 promoter as a GST fusion using the expression vector pGEX-4T2 (Pharmacia), which was subsequently transformed into E. coli BL21(DE3) cells. Cells were grown on LB at 37 °C, and overexpression was induced at an OD600 reading of 0.6 with IPTG at a final concentration of 1 mM. After 3 h of induction, cells were harvested and bound in batch to Glutathione Sepharose4B resin. The resin was washed with PBS buffer before thrombin cleavage on

76

a

b

Fig. 5 Effective correlation times (τe) for methionine methyl groups in a, CaM and b, the CaM–smMLCKp complex.

the resin, and the cleaved mixture was immediately passed over a p-aminobenzamidine column. The smMLCK peptide was collected in the flowthrough and HPLC purified over a C18 reversed phase resin. The complex between Ca2+-CaM and smMLCKp was formed under dilute conditions by titration of smMLCKp into CaM (50% 2H,13C,15N) until a 1:1 complex was observed in the NMR spectrum. This complex was then lyophilized and resuspended in 90% H2O/10% D2O to reach the final NMR sample condition of 1.2 mM protein, 100 mM KCl, 6.0 mM total CaCl2 concentration, 10 mM imidazole-d4, pH 6.5, and 0.02% sodium azide. NMR spectroscopy. All NMR experiments were carried out at 35 °C on Varian Unity Inova spectrometers equipped with 1H/15N/13C probes and z-axis pulsed-field gradients. Backbone and side chain assignments for the CaM–smMLCK complex were extended from previous assignments30 using standard triple-resonance techniques. Methionine methyl resonances were assigned using HMBC and 3D LRCC 1H/13C correlation experiments44. The 2H relaxation experiments of Kay and coworkers were employed to measure the relaxation rates of the multispin coherences IzCz, IzCzDz and IzCzDy generated within each methyl group23. For Ca2+-CaM, each of the three experiments was conducted at field strengths of 11.7 and 14.1 T. Relaxation decays for these coherences were sampled at identical time points at the two field strengths. The IzCz relaxation time points were 12.2*, 22.4, 33.6, 44.8*, 56.0, 67.2, 78.4, 89.7* and 101.2 ms; I zCzDz time points were 4.3*, 11.1, 19.8, 30.2*, 42.0, 55.0, 69.1, 84.3* and 100.6 ms; IzCzDy time points were 1.0, 3.1*, 5.7, 8.8*, 12.4, 16.3, 20.5, 25.1* and 30.0 ms. For the complex of Ca2+-CaM and smMLCKp, 2H relaxation data was collected at 14.1 and 17.6 T. At 14.1 T, the IzCz relaxation time points were 12.3*, 14.2, 16.7, 19.7*, 23.0, 26.8, 30.8, 35.2*, 39.8, 44.7, 49.8, 55.3 and 61.2 ms; IzCzDz time points were 1.3, 3.6, 7.3*, 12.4, 14.6*, 19.1*, 27.2, 31.6*, 36.9, 48.0 and 60.1 ms; IzCzDy time points were 1.0, 2.2*, 4.0, 6.2, 8.9, 12.1*, 15.8 and 20.0 ms.



articles Relaxation decays were sampled similarly at 17.6 T. Asterisks indicate duplicate points. All 2H relaxation data were fitted to a single exponential as described25. Relaxation analysis. 2H relaxation data was analyzed using the Lipari–Szabo model-free formalism, the details of which have been given elsewhere25. Overall rotational correlation times were determined for each domain of Ca2+-CaM and for CaM/smMLCKp based on 15N T1, T2 and NOE measurements at 14.1 T. In agreement with a previous study16, the N- and C-terminal domains of Ca2+-CaM tumble independently and have isotropic correlation times of 7.5 and 6.5 ns, respectively. Rotational tumbling anisotropy45–47 was determined to be small (D||/D⊥ = 1.27 for N-terminal domain; D||/D⊥ = 1.45 for Cterminal domain) using the local Di approach. The CaM–smMLCKp complex was determined to have an isotropic rotational correlation time of 8.3 ns. There is a possibility that the rotational correlation times are slightly overestimated, since T2 data were employed in the 1.


2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.

17. 18. 19. 20. 21. 22. 23. 24. 25. 26.

27.

Wintrode, P.L. & Privalov, P.L. Energetics of target peptide recognition by calmodulin: a calorimetric study. J. Mol. Biol. 266, 1050–1062 (1997). Murphy, K.P., Xie, D., Garcia, K.C., Amzel, L.M. & Freire, E. Structural energetics of peptide recognition: angiotensin II/antibody binding. Proteins 15, 113–120 (1993). Lee, K.H., Xie, D., Freire, E. & Amzel, L.M. Estimation of changes in side chain configurational entropy in binding and folding: general methods and application to helix formation. Proteins 20, 68–84 (1994). Makhatadze, G.I. & Privalov, P.L. Energetics of protein structure. Adv. Prot. Chem. 47, 307–425 (1995). Weber, G. Thermodynamics of the association and the pressure dependence of oligomeric proteins. J. Phys. Chem. 97, 7108–7115 (1993). Karplus, M. & Kushick, J.N. Method for estimating the configurational entropy of macromolecules. Macromolecules 14, 325–332 (1981). Doig, A.J. & Sternberg, J.E. Side-chain conformational entropy in protein folding. Protein Sci. 4, 2247–2251 (1995). Rasmussen, B.F., Stock, A.M., Ringe, D. & Petsko, G.A. Crystalline ribonuclease A loses function below the dynamical transition at 220 K. Nature 357, 423–424 (1992). Feher, V.A. & Cavanagh, J. Millisecond-timescale motions contribute to the function of the bacterial response regulator protein SpoOF. Nature 400, 289–293 (1999). Stock, A. Relating dynamics to function. Nature 400, 221–222 (1999). Akke, M., Brüschweiler, R. & Palmer III, A.G. NMR order parameters and free energy: an analytical approach and its application to cooperative Ca2+ binding by calbindin D9k. J. Am. Chem. Soc. 115, 9832–9833 (1993). Li, Z., Raychaudhuri, S. & Wand, A.J. Insights into the local entropy of proteins provided by NMR relaxation. Protein Sci. 5, 2647–2650 (1996). Yang, D. & Kay, L.E. Contributions to conformational entropy arising from bond vector fluctuations measured from NMR-derived order parameters: application to protein folding. J. Mol. Biol. 263, 369–382 (1996). Vogel, H.J. Calmodulin: a versatile calcium mediator protein. Biochem. Cell Biol. 72, 357–375 (1994). Babu, Y.S. et al. Three-dimensional structure of calmodulin. Nature 315, 37–40 (1985). Barbato, G., Ikura, M., Kay, L.E., Pastor, R.W. & Bax, A. Backbone dynamics of calmodulin studied by 15N relaxation using inverse detected two-dimensional NMR spectroscopy: the central helix is flexible. Biochemistry 31, 5269–5278 (1992). Zhang, M., Tanaka, T. & Ikura, M. Calcium-induced conformational transition revealed by the solution structure of apocalmodulin. Nature Struct. Biol. 2, 758–767 (1995). Kuboniwa, H. et al. Solution structure of calcium-free calmodulin. Nature Struct. Biol. 2, 768–776 (1995). Ikura, M. et al. Solution structure of a calmodulin–target peptide complex by multidimensional NMR. Science 256, 632–638 (1992). Meador, W.E., Means, A.R. & Quiocho, F.A. Target enzyme recognition by calmodulin: 2.4 Å structure of a calmodulin–peptide complex. Science 257, 1251–1256 (1992). Meador, W.E., Means, A.R. & Quiocho, F.A. Modulation of calmodulin plasticity in molecular recognition on the basis of x-ray structures. Science 262, 1718–1721 (1993). Osawa, M. et al. A novel target recognition revealed by calmodulin in complex with Ca2+-calmodulin-dependent kinase kinase. Nature Struct. Biol. 6, 819–824 (1999). Muhandiram, D.R., Yamazaki, T., Sykes, B.D. & Kay, L.E. Measurement of 2H T1 and T1ρ relaxation times in uniformly 13C-labeled and fractionally 2H-labeled proteins in solution. J. Am. Chem. Soc. 117, 11536–11544 (1995). Lipari, G. & Szabo, A. 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 (1982). Lee, A.L., Flynn, P.F. & Wand, A.J. Comparison of 2H and 13C NMR relaxation techniques for the study of protein methyl group dynamics in solution. J. Am. Chem. Soc. 121, 2891–2902 (1999). Lukas, T.J., Burgess, W.H., Prendergast, F.G., Lau, W. & Watterson, D.M. Calmodulin binding domains: characterization of a phosphorylation and calmodulin binding site from myosin light chain kinase. Biochemistry 25, 1458–1464 (1987). Kemp, B.E., Pearson, R.B., Guerriero, V., Bagchi, I.C. & Means, A.R. The calmodulin


analysis, as is necessary for τm determination when τm exceeds 6–7 ns48. Overestimation of τm typically leads to underestimation of 2Hderived S2axis parameters29. Such systematic errors are not expected to affect the present results, however, since differences in order parameters are of primary interest here and are effectively insensitive to this issue. In the case of Ca2+-CaM, it was found that the determination of τm was sensitive to protein concentration, as reported previously16, due to aggregation. For this reason, protein concentrations of