Specificity and Symmetry in the Interaction of Calmodulin Domains with the Skeletal Muscle Myosin Light Chain Kinase Target Sequence*

The specificity of interaction of the isolated N- and C-terminal domains of calmodulin with peptide WFFp (Ac-KRRWKKNFIAVSAANRFK-amide) and variants of the target sequence of skeletal muscle myosin light chain kinase was investigated using CD and fluorescence. Titrations show that two molecules of either domain bind to 18-residue target peptides. For WFFp, the C-domain binds with 4-fold higher affinity to the native compared with the non-native site; the N-domain shows similar affinity for either site. The selectivity of the C-domain suggests that it promotes occupancy of the correct binding site for intact calmodulin on the target sequence. Far UV CD spectra show the extra helicity induced in forming the 2:1 C-domain-peptide or the 1:1:1 C-domain-N-domain-peptide complex is similar to that induced by calmodulin itself; binding of the C-domain to the Trp-4 site is essential for developing the full helicity. Calmodulin-MLCK-peptide complexes show an approximate two-fold rotational relationship between the two highly homologous domains, and the 2:1 C (or N)-domain-peptide complexes evidently have a similar rotational symmetry. This implies that a given domain can bind sequences with opposite peptide polarities, significantly increasing the possible range of conformations of calmodulin in its complexes, and extending the versatility and diversity of calmodulin-target interactions.

Calmodulin is a regulatory protein involved in a variety of Ca 2ϩ -dependent cellular signaling pathways. Its importance as a mediator of the second messenger Ca 2ϩ is reflected in its high conservation throughout evolution. This apparently contrasts with its unique ability to interact strongly with and to regulate selectively a variety of proteins (at least 30) without any obvious sequence homology in their calmodulin binding region (see Refs. 1-5 for reviews). Recently, structures of calmodulin and calmodulin-peptide complexes at atomic resolution have been determined (reviewed in Refs. 2, 6, and 7), showing two similar domains with two Ca 2ϩ binding sites each, which for calmodulin in solution are connected by a flexible linker (8 -15). The conformational change upon Ca 2ϩ binding to calmodulin exposes those residues that create the binding site for most target proteins (1,6,7,16).
Binding of the Ca 2ϩ saturated form of calmodulin to the target protein triggers its activation. Despite the wide target range, target affinities are strong (K d Ϸ 1 nM) (2)(3)(4). The interaction is apparently mediated by both hydrophobic and electrostatic forces (17). NMR (18) and x-ray structures (19) of two related Ca 2ϩ -calmodulin-target peptide complexes show that the ␣-helical MLCK target peptide lies in a hydrophobic channel composed by the two domains, with the predominant interactions being those between the N-and C-terminal domains and the C-and N-terminal portions of the target, respectively. Calmodulin can be cleaved by trypsin to generate two halfmolecules, i.e. the C-terminal and the N-terminal Ca 2ϩ binding domain (20,21). The equilibrium (22)(23)(24) and kinetic properties (25,26) of intact calmodulin in the Ca 2ϩ binding and dissociation reactions, as well as the secondary structure (24,27), are well represented by a summation of the properties of these fragments, suggesting that the two domains are effectively independent structures. The isolated domains are capable of activating target proteins, but the degree of activation varies with the target protein (26, 28 -35). In particular, skeletal muscle myosin light chain kinase (sk-MLCK) 1 is activated best by a 1:1 mixture of the domains (85% activation compared with calmodulin), but less by either the C-domain (65%) or the N-domain (20%) (31). This activation pattern is reproduced when calmodulin chimeras consisting of two linked N-domains or two C-domains are used (36). Although the C-domain activates target enzymes better than the N-domain in several cases (29 -31, 35), this is not always so (31,33). Therefore, differences in domain sequence and structure may contribute to the versatility of calmodulin's regulatory functions. Binding of the domains to the target enzyme is not necessarily sufficient for activation since isolated domains can inhibit calmodulin-induced activation of enzymes which are not activated by the domain itself (28,29,31,35).
Although the overall structure of the two domains show marked similarities (37), differences between them in sequence and Ca 2ϩ affinity have apparently been well conserved during evolution (38). This points toward possible functional differences of the two domains, which can be further amplified by variations in target sequences. Studies of enhancement of Ca 2ϩ affinities of calmodulin by target sequences suggest that cal-modulin which is half-saturated with Ca 2ϩ could bind to the target protein by only one domain without activation, allowing a rapid response in enzyme activity to an increase in Ca 2ϩ concentration (39).
In the present work, we address on a molecular level the specificity of the interaction of the individual calmodulin domains with target peptides derived from the target sequence of sk-MLCK. Previously, spectroscopic studies have been reported on the interaction of the domains with the short peptides melittin and mastoparan (40 -42). Here, the binding affinities, the molecular interactions in the complex, and the effects of domain binding on the conformation of the target peptides are investigated using a variety of spectroscopic techniques. The striking finding is that two molecules of either domain bind with good affinity to the 18-residue target peptides. The symmetry of the resulting complexes is considered by comparison with the calmodulin-MLCK peptide structure (2), and is discussed in relation to the known versatility of calmodulin in its specific interactions with a range of target proteins. The feasibility of calmodulin domains to bind with reversed polarity and alternative positions on a target sequence greatly extends the potential range of conformations of calmodulin in its bound form, and provides added diversity to the calcium sensitivity of calmodulin-dependent activation processes.

MATERIALS AND METHODS
Proteins and Peptides-Drosophila melanogaster calmodulin expressed in Escherichia coli was purified as described previously (43). The purified protein ran as a single band on an SDS-polyacrylamide gel electrophoresis (15% gel; Laemmli system). The tryptic fragments of calmodulin were prepared as described in Ref. 31 with an additional gel filtration step (G75 column) included after the cleavage and before the anion exchange chromatography. For each fragment, no impurities could be detected on SDS-polyacrylamide gel electrophoresis. However, HPLC chromatography showed an impurity of unknown origin in the C-domain preparation corresponding to 2.4% of the protein mass. Mass spectroscopy showed that the predominant fragment for the N-domain preparation was residues 1-75 of calmodulin and, for the C-domain preparation, residues 78 -148.
Concentrations of proteins and peptides were determined spectrophotometrically using the following extinction coefficients: 5690 M Ϫ1 cm Ϫ1 at 280 nm for Trp-containing peptides (44); 585 M Ϫ1 cm Ϫ1 at 259 nm for FFFp (44) and 1578 M Ϫ1 cm Ϫ1 at 279 nm for Drosophila calmodulin in the presence of excess Ca 2ϩ (43). The same extinction coefficient was used for the C-domain of calmodulin assuming that the absorption of the single Tyr-138 in this domain is unaffected by the tryptic cleavage. For the N-domain, 975 M Ϫ1 cm Ϫ1 at 259 nm was used based on its five Phe residues (44). It was confirmed that the extinction coefficient of the N-domain changed by less than 10% upon denaturation with 6 M guanidine HCl.
Fluorescence Measurements-Uncorrected fluorescence emission spectra were recorded in UV-transmitting plastic cuvettes or in quartz cuvettes using a SPEX FluoroMax fluorimeter. Excitation was at 300 nm for measurements at Ն50 M peptide concentration, otherwise at 290 or 295 nm for the C-domain and at 280 or 300 nm for the N-domain (bandwidth 0.9 nm). The emission was scanned from 300 or 310 nm to 400 nm (bandwidth 4.3 nm). The temperature was 20°C and the buffer 25 mM Tris/HCl, 100 mM KCl, 1 mM CaCl 2 at pH 8.0. Data for the fluorescence titrations were obtained either by integrating the spectra in the region of the largest fluorescence change (300 -330 nm) or by measurements of 30 s in duration at a wavelength in the range 322-334 nm with bandwidth 17 nm. Unless otherwise stated, the dissociation constants are derived from at least three independent titrations. Circular Dichroism (CD) Measurements-CD spectra were recorded in fused silica cuvettes using a Jasco J-600 spectropolarimeter. The measurements were made at 20°C in 25 mM Tris/HCl, 100 mM KCl, 1 mM CaCl 2 at pH 8.0. Far UV-CD measurements (200 -280 nm) with calmodulin and all peptides as well as with WFFp and isolated domains were made using 1-mm cuvettes with peptide and protein concentrations in the range 7-25 M. With isolated domains and FFFp or FFWp the measurements were made in a 0.1-mm demountable cuvette at 10-fold higher concentrations. Spectra are presented as the molar CD absorption coefficient (⌬⑀ M ) using the molar concentration of the protein rather than the mean residue weight for the normalization. In the case of the domain mixture complexes, the concentration of the complex was used for the normalization to facilitate the comparison with the spectra of calmodulin (46). The difference between the CD absorption coefficient at 222 nm in the presence and the absence of the peptide was used to estimate the number of residues adopting a helical structure upon complex formation. The results of three experiments were averaged. The difference was expressed as molar in peptide (not protein) concentration and compared with the ⌬⑀ M value of fully helical peptides of different lengths, which were calculated according to Ref. 45.
Near UV-CD spectra (255-340 nm) were measured using 15-30 repetitive scans for peptide and protein concentrations in the range 80 -380 M with 10 mm cuvettes. The spectral range of 310 -340 nm was used to zero the curves. A base-line spectrum was recorded, smoothed over 10 nm, and subtracted, and the resulting spectra were slightly smoothed. The spectra are the average of at least two independent measurements. For near UV CD titrations, spectra of 10 -20 scans were recorded between 280 and 320 nm, and were zeroed using the data between 308 and 320 nm. A base-line spectrum was smoothed over 10 nm, subtracted, and the CD signal of the resulting spectrum integrated between 280 and 293 nm for titrations with the N-domain and between 292 and 299 nm for titrations with the C-domain. At least four independent titrations were evaluated. Under the conditions of the CD experiments with the N-domain and the short peptides WF10p and FW10p, incomplete complex formation is expected owing to the low affinity of the interaction. Using the dissociation constants of 60 and 75 M (see below), it is calculated that 53% (N-WF10p) and 48% (FW10p-N) of the peptides are bound under the conditions of the experiment, and these values were used to subtract the free peptide contribution to the spectra and to normalize them to 1 M complex concentration.
Data Analysis and Determination of Dissociation Constants-Titrations were performed by addition of the domain solution to the peptide solution, and recording changes in fluorescence or CD signals deriving from the Trp chromophore of the peptide. Titration curves of WF10p and FW10p peptide with either domain were fitted with a stoichiometry of 1:1, using standard fitting procedures (46). In the case of the long peptides WFFp and FFWp, the titration curves clearly indicated that saturation of the optical signal was achieved close to a stoichiometry of two molecules of domain per molecule of peptide. The simplest model for binding assumes that the optical signal monitors binding of a domain to the Trp-containing portion of the peptide; binding of a second molecule of the domain is revealed only indirectly via the competition of the domain between the two sites. The analysis is based on the known structure of the complex of calmodulin with the sk-MLCK target M13 peptide, in which the C-domain of calmodulin, interacting exclusively with the Trp residue of the peptide, binds predominantly with the (Trp containing) N-terminal portion of the peptide, and the N-domain binds predominantly with the C-terminal portion of the peptide. The two sites on the peptide are therefore designated according to their position in the peptide sequence as site N and site C; in the case of binding to peptide WFFp, binding at site N with K d N , produces an optical signal, and binding at site C with K d C is optically "silent," whereas for peptide FFWp, the optical properties of sites N and C are reversed.
The mechanism for the binding of two molecules of a given domain (D) per molecule of peptide with sites N and C (understood to be oriented N-peptide-C) is shown in Scheme 1.
According to the scheme, in the course of a titration of a fixed amount of peptide (P) with a single domain, D, the optical signal at any point is determined by ⌺X i ⅐S i , where X i is the mole fraction of various peptidecontaining species (and ⌺X i ϭ 1), together with S i , their intrinsic spectroscopic properties. On the above model, for WFFp peptide, and indices 1-4 referring to free peptide, D-peptide-, -peptide-D, and Dpeptide-D, respectively, then S 1 ϭ S 3 (known from free peptide) and S 2 ϭ S 4 , (known from, or fitted to, the plateau titration value). To evaluate the values of X i for a given concentration of peptide (fixed) and variable concentration of added domain [D], the analysis requires the values of K d C , K d N (and parameter f when f 1) with the solution of a cubic equation. The unknowns (K d C , K d N , and f) are then refined by least squares methods in fitting to the experimental titration data. Conversely, simulations were made, using chosen values of these parameters to examine their individual effects on model titration curves.
To simplify the analysis, the determination of the two binding constants K d N and K d C was done in two steps. First, in titrations of the fluorescence and the CD signals at high peptide concentration, the ratio K d C /K d N was determined. This value was then used as a fixed parameter in the fits to the fluorescence titrations at low concentration, which finally resulted in the individual K d values. The rationale behind this approach, demonstrated by simulation, is that titration curves at high concentration (i.e. Ͼ K d C and K d N ) depend strongly on the ratio K d C /K d N and less on their absolute values, whereas the opposite is true at lower concentrations.
The simplest analysis is where the two sites are non-interacting, i.e. there is no thermodynamic co-operativity in the binding of the second copy of the domain. For this case, the co-operativity factor in the scheme is given by f ϭ 1. Interaction can be included with f Ͼ 1 (negative co-operativity) or f Ͻ 1 (positive co-operativity). In titrations of a fulllength peptide with a single domain, it was found that the inclusion of the additional co-operativity parameter f gave no significant improvement to the numerical fitting, and thus fits were generally done with the simplest model with f ϭ 1. In calculations based on studies of the 1:1:1 complex of C-domain plus N-domain plus WFFp peptide, the inclusion of an f ϳ 0.5 was deduced, suggesting a small degree of positive co-operativity; in one out of four cases, evidence was found for a spectroscopic interaction. These analyses are discussed under "Results."

RESULTS
Fluorescence Spectra-Solutions of the free Trp-containing peptides have a fluorescence emission maximum at ϳ358 nm. The maximum of the enhanced fluorescence emission of all the peptide-domain complexes (Table I) lies between 335 and 338 nm, except for the N-WF10p complex (348 nm), suggesting that the N-domain is less effective than the C-domain in burying the Trp residue of this short peptide in a hydrophobic environment. As discussed in more detail below, two molecules of either domain were found to bind to one molecule of full-length pep-tide (WFFp or FFWp). Complexes are represented as e.g. X-WFFp-Y, indicating that domain X binds to the N-terminal portion of the peptide and domain Y to the C-terminal portion.
Near UV CD Spectra-Near UV CD spectra of the domains in the absence and the presence of a target peptide, as well as spectra of the free peptides are shown in Fig. 1. The free peptides WFFp (Fig. 1A), FFWp (Fig. 1B), and WF10p (data not shown) have similar near UV CD spectra; below 290 nm, the signal increases steadily and without fine structure to a ⌬⑀ M ϭ 0.4 M Ϫ1 cm Ϫ1 at 255 nm. Only the FW10p peptide (Fig. 1C) shows a signal above 290 nm.
The spectrum of the N-domain is composed of the Phe signals below 270 nm (Fig. 1B, bold line). The spectrum of the Cdomain (bold line in Fig. 1C) shows the additional contribution of the single Tyr residue above 270 nm. Spectra of the isolated domains of bovine testis calmodulin (24) and of the C-domain of Drosophila calmodulin (39) have been reported elsewhere. The spectrum of a 1:1 domain mixture (Fig. 1A) is closely similar to that of intact calmodulin (39,46) confirming earlier reports that the near UV CD spectrum of calmodulin can be represented by a summation of the CD spectra of the individual domains (24). Near UV CD Spectra of Complexes with Peptide WFFp-The near UV CD difference spectrum of the bound Trp in the C-WFFp-C complex (Fig. 1D) is stronger than that with the shorter WF10p peptide (Fig. 1D). The Trp signal is also affected by the end protection of the peptides, the unprotected versions (suffix "u") showing the weaker signal (39). At 295 nm, the signal decreases in the order:  Trp and surrounding residues. We conclude that there is some variation in the binding mode of the C-domain to the Trp residue in the different peptides. The Trp near UV CD difference spectrum of N-WFFp-N is characterized by a negative signal (Fig. 1D). The spectrum of the complex N-WF10p has the same shape, but a reduced intensity (not shown). Although binding to a different binding site cannot be ruled out, the preserved shape may suggest that the Trp is still bound to the same binding site. However, the fluorescence spectra indicate that the environment of the WF10p Trp residue is significantly less hydrophobic than that of the WFFp Trp residue. It is possible that the hydrophobic environment is only maintained as long as the Trp residue is restricted in mobility, but with increased mobility Trp gains access to less hydrophobic regions, which are probably close to the water surface.
Near UV CD Spectra of Complexes with Peptide FFWp-The spectrum of N-FFWp-N has the same form as that of the complex of calmodulin with FFWu (46) (and FFWp; data not shown), but with intensity reduced to 50%, suggesting that the Trp residue is bound similarly in both complexes but with increased mobility within the N-FFWp-N complex. For F10p-N, a significantly weaker signal than for N-FFWp-N is observed, which is not the case when the calmodulin complexes with FFWu and FW10u are compared (39).
C-FFWp-C shows the weakest CD spectrum of the complexes with the long peptides, and this is the only case where the spectrum with the corresponding short peptide FW10p does not show weaker intensity and is significantly different in shape. Therefore, the Trp binding modes are different for the long and the short peptide and are possibly a superposition of spectra of distinct complex conformations.
Near UV CD Spectra of Complexes with a 1:1 Domain Mixture-To find out whether the 1:1 mixture of C-and N-domain binds in a manner similar to that of intact calmodulin, near UV CD spectra were recorded of the mixture with and without peptide WFFp or FFWp. The Trp spectrum of the 1:1:1 complex with WFFp is very similar in shape to that of C-WFFp-C and of the calmodulin-WFFp complex (Fig. 1D). It is estimated that the Trp spectrum of N-WFFp-N contributes less than 15% to the 1:1:1 complex spectrum. Thus, in the 1:1:1 complex, the Trp residue of the WFFp peptide effectively binds only to the Cdomain, resulting in the C-WFFp-N complex, whose spectrum closely resembles that of intact calmodulin-WFFp.
In contrast, the spectrum of the 1:1:1 complex of C-and N-domain with FFWp is significantly different from those of N-FFWp-N (Fig. 1E) or the complexes of calmodulin with FFWu (46) and FFWp, but is very similar to that of FW10p-N and, despite a deviation at 295 nm, to the C-FFWp-C spectrum (Fig. 1F). This may suggest that the isolated C-domain can bind (in part) to the unusual Trp containing C-terminal site on the peptide.
Binding Stoichiometry and Affinities of Individual Calmodulin Domains for the Target Peptide-The fluorescence and CD spectra (Fig. 1, D-F) show that each domain binds to the single Trp in the N-terminal portion of WFFp and the C-terminal portion of FFWp and to the short peptides WF10p and FW10p. The fluorescence and CD signals from this Trp residue were used to determine the affinities of the domains for the peptides and the stoichiometry of the complexes.
For the short WF10p and the FW10p peptides, the fluorescence titrations with either domain were fitted well with 1:1 stoichiometry and the K d values obtained are listed in Table I. For the full-length peptides, titrations of the CD and fluorescence signals with a given domain show that the stoichiometry is clearly greater than 1 (see Fig. 2). These data were analyzed with the two binding site model described under "Materials and Methods." Fig. 2 shows typical examples of two CD titrations (C-domain and WFFp, Fig. 2A; N-domain and WFFp, Fig. 2B) and of two fluorescence titrations at low peptide concentration (C-domain and WFFp, Fig. 2B; N-domain and WFFp, Fig. 2D). The dissociation constants obtained are illustrated in Fig. 3 and listed in Table I together with the corresponding Gibbs free energies, calculated as RT log K d .
The The independent site model used for these fits assumes that only two distinct spectroscopic species have to be accounted for. However, it seems possible from the differences in the K d C /K d N ratios that the spectroscopic properties of the Trp site may be affected to a small degree by binding of a domain to the silent binding site, i.e. S 2 S 4 (Scheme 1). Simulations show the following results. (i) For the CD data, S 4 ϭ 1.3 ϫ S 2 was needed to give K d C /K d N ϭ 8.0 Ϯ 0.5, giving slightly better fits compared with the assumed S 2 ϭ S 4 . With these parameters, K d N ϭ 10 Ϯ 4 nM and K d C ϭ 80 Ϯ 5 nM were obtained with the low concentration fluorescence titrations. (ii) Alternatively, for the high concentration fluorescence data, S 4 ϭ 0.8 ϫ S 2 was needed to give K d C /K d N ϭ 2.5 Ϯ 0.5, giving a fit only slightly worse than that with S 2 ϭ S 4 . With these parameters, K d N ϭ 25 Ϯ 11 nM and K d C ϭ 63 Ϯ 3 nM were obtained with the low concentration fluorescence data. There is therefore only marginal evidence of spectroscopic interaction between the two bound domains, and this makes relatively small differences in the derived K d N and K d C values.
Far UV CD Spectra of Complexes of Calmodulin or Calmodulin Domains with Target Peptides-Secondary structure changes upon WFFp, FFFp, or FFWp binding to the domains have been monitored using far UV CD spectroscopy. The domain and the complex spectra were normalized using the protein concentration, except in the case of the domain mixture, where the concentration of one domain was used to facilitate the comparison with the spectra of calmodulin (46). The spectra of the free domains are dominated by the contributions from ␣-helical structures with the characteristic minima near 207 and 222 nm. The spectrum of the 1:1 domain mixture agreed within experimental error with the corresponding spectrum of calmodulin confirming earlier results for bovine testis and brain calmodulin (24,27) that the domain structure is preserved in the tryptic fragments. Upon addition of the (unstructured) peptide to the domains, there is an increase in ␣-helicity consistent with the peptide adopting an ␣-helical structure in the complex. The conformation of the calmodulin domains is largely unaffected by binding of the target sequence (18), although slight structure perturbations have been reported (47). Far UV CD was measured at different domain:peptide ratios and the change in the CD spectrum was normalized to peptide concentration. The corresponding ⌬⌬⑀ values at 222 nm are collected in Table II together with the change in the number of helical residues. Calmodulin, the C-domain, and the domain mixture induce the same high degree of helicity in each peptide, whereas the N-domain is as effective only with FFFp. The induced helicity tends to be highest for peptide WFFp, decreasing with the number of replacements for FFFp and FFWp. The exception again is the N-FFFp-N complex, which shows the same induced helicity as N-WFFp-N. Table II shows that the increased helicity saturates at a 2:1 domain:peptide ratio for peptides WFFp, FFFp, or FFWp. A 1:1 domain:peptide ratio induces more than 50% of the maximum helicity, suggesting  that binding of the first molecule of either domain to the long peptides induces more helicity than binding of the second.

The Number of Domain Binding Sites on the 18-residue Peptides WFFp and FFWp-
In the complex of calmodulin with the target sequence of sk-MLCK, the C-domain of calmodulin binds mainly to residues in the N-terminal portion of the target peptide and the N-domain mainly to those in the C-terminal portion (18). This work clearly shows that two molecules of either the C-domain or the N-domain bind to the 18-residue sk-MLCK target peptide WFFp and to the variants FFFp and FFWp. In agreement with this, either domain will bind with 1:1 stoichiometry to the short peptides WF10p or FW10p, which represent the N-and the C-terminal portion of the WFF and FFW target sequences, respectively.
The Hydrophobic Pockets of the Domains-The maximum of the enhanced fluorescence at 335-338 nm (Table I) shows that either domain binds the bulky Trp residue of the peptides, which becomes buried in a hydrophobic pocket. Generally, the near UV CD Trp spectrum is unique for each 2:1 or 1:1 domainpeptide complex, and the high intensities (⌬⑀ Ϯ 1-2 M Ϫ1 cm Ϫ1 ) are consistent with strong immobilization of the chromophore. Both static and dynamic properties of the environment provided by the peptide and binding pockets of the domains can contribute to the CD signal, and are therefore characteristic for each complex. We conclude that the binding mode of the Trp residue depends on its position in the peptide sequence, the length of the peptide, and the domain to which it is bound. Near UV CD spectra of complexes of the short peptides WF10p and FW10p differ from those the 18-residue peptides WFFp and FFWp. Therefore, the exact binding mode of the Trp residue is not solely determined by local interactions between Trp and the protein, but is also sensitive to longer range interactions.
The Trp CD spectrum of C-WFFp-N has the same shape and sign, and 90% of the intensity of the Trp spectrum of the calmodulin-WFFp complex (Fig. 1D). This implies close similarity of not only the interactions local to Trp, but also those occurring at longer range. However, in the case of the permuted target sequence FFWp, the interactions that determine the Trp binding mode are different in the complexes with calmodulin, two N-domains, or the N-and C-domain mixture.
Structural Change of the Peptide-The far UV CD data show that calmodulin, the C-domain, and the domain mixture induce a high degree of helicity in all three peptides: WFFp, FFFp, or FFWp. Compared with the C-domain, the N-domain is equally effective with FFFp, and significantly less so with WFFp and FFWp. The helicity induced by the 1:1 mixture of isolated domains is for all peptides close to that induced by intact calmodulin. The C-domain can replace the N-domain from its site in the calmodulin complexes without affecting the peptide helicity. On the other hand, replacing the C-domain from its site in the calmodulin complexes with the N-domain reduces the helicity of Trp-containing peptides (with an estimated loss of five helical residues). Thus, the C-domain is generally more effective than the N-domain in maintaining the helicity of the full peptide, as e.g. in the C-WFFp-C complex. For most complexes, the induced helicity clusters between 120 and 150 M Ϫ1 cm Ϫ1 corresponding to 13-15 helical residues. For the peptide to become fully helical, the presence of a Trp residue at position 4 of the target peptide sequence as well as its interaction with the C-domain is advantageous. Trp-4 does not have this function of inducing maximum helicity in the interaction with the N-domain. Thus, the binding pocket of the C-domain and the WFFp peptide sequence appear mutually optimized to obtain a highly helical target peptide.
The complex with the lowest induced helicity is N-FFWp-N, evidently an effect of the replacement Phe-17 3 Trp. However, the 1:1:1 complex of C-domain ϩ N-domain ϩ FFWp and the calmodulin-FFWp complex are closely similar. Thus, there appears to be a somewhat disordered C-terminal peptide portion in N-FFWp-N. Interestingly, the pattern for the natural target sequence WFFp matches the activation pattern of sk-MLCK. Separated N-domains hardly activate the enzyme, whereas the C-domain activates it to 65% and a 1:1 mixture of the domains to 85% (31). This suggests that the ability to induce helicity in the target sequence may be connected with the ability to activate the enzyme. However, this is not the only prerequisite for activation since the helicity criterion alone does not explain the different activation profiles of calmodulin, the C-domain, and a domain mixture.
The Role of the Individual Domains in the Target Recognition Process-Both C-domain and N-domain form complexes with either WFFp or FFWp with a stoichiometry of 2 mol of domain/ mol of peptide. The N-and the C-domain bind to both sites on the peptide with surprisingly similar affinities (see Table I), with the highest affinity interactions being those between the C-domain and the Trp-containing portion of the WFFp or FFWp target sequences. Similar conclusions were reached from studies of sk-MLCK activation (31); two binding sites for calmodulin domains were found on the enzyme with dissociation constants for the C-domain of K d N ϭ 300 nM and K d C ϭ 20 M and for the N-domain of K d N ϭ 12 M and K d C ϭ 3-5 M (the values are presented according to the nomenclature in this paper from the tentative assignment in the original paper). The K d values obtained here with a target peptide tend to be 1-2 orders of magnitude smaller than those derived from the enzyme activation studies. This may suggest that binding of the domains to the whole enzyme may require energetically costly disruption of interactions between the calmodulin binding site and other parts of the enzyme (1, 2).
The subtle affinity differences of the two domains for the binding sites on the peptide support the view that the key recognition process between calmodulin and the native target peptide is due to the interaction of the C-domain with the N-terminal portion of the target sequence (31,39,48): its preference for the "correct" binding site on the peptide and its higher affinity for this site as compared with the N-domain ensure the correct orientation of the complex. Consistent with this, a 1:1 mixture of the separated N-and C-domains appears to bind to the native target sequence WFFp in the same orientation as intact calmodulin (see "Results"). The resulting C-WFFp-N complex is virtually indistinguishable spectroscopically from the calmodulin-WFFp complex, in terms of the interactions leading to the exact mode of Trp binding and helicity induced in the peptide. This indicates that it is an intrinsic property of the C-domain and N-domain when present together to form a complex of almost identical structure as with calmodulin.
Evidence has recently been presented for a potential functional difference between the two domains. At intermediate Ca 2ϩ concentrations (Ϸ1 M) in the presence of the target peptide, the C-domain binds Ca 2ϩ preferentially and is therefore able to bind to the peptide, while the N-domain remains in the apo state (39). The intrinsic preference of the C-domain for the correct binding site on the peptide promotes the efficient formation of the appropriate intermediate complex, en route to the full complex required for activating the enzyme in response to an elevated Ca 2ϩ signal. The mechanism outlined above, which resembles that proposed for the Ca 2ϩ dependent activation of troponin C with tropomyosin (49), suggests that the conservation of sequence differences in the domains of the calmodulin molecule is important in its function.
The Role of the Trp Residue in the Target Sequence on the Recognition Process-The experiments with the FFWp and FFFp variant peptides show the importance of the interaction between Trp-4 of the peptide and the C-domain. Replacing Trp-4 by Phe appears to decrease the helicity in the N-terminal part of the bound peptide, and diminishes the selectivity of that peptide portion for the C-domain. With FFWp, the C-terminal binding site shows the higher affinity for both C-and N-domain. Binding of the C-domain with the C-terminal site of FFWp implies a type of interaction that is not observed in the calmodulin-FFWp complex, which apparently has the same peptide orientation as WFFp (48). This suggests a possible functional role of linking the two domains together in one molecule, which may ensure a defined orientation of the interaction.
Cooperativity and the Energetics of the Domain-Peptide Interactions-The formation of the 1:1:1 mixed complex C-WFFp-N raises the question of co-operativity in the interactions of calmodulin domains with the target sequence. The sum of the total Gibbs free energy change (RT log K d ) for formation of C-WFFp and WFFp-N as calculated from the values in Table  I is Ϫ82 kJ/mol and is larger than the corresponding value, Ϫ78.9 kJ/mol, obtained for formation of WFFp-C and N-WFFp. This shows that the C-WFFp-N complex is energetically preferred, as deduced from the near UV CD spectra (see "Results"). Simulations using the K d values of Table I show that, under the conditions of the CD experiment with the 1:1 domain mixture, up to 34% of the complexes would involve "non-native" binding of the peptide Trp to the N-domain, exceeding the upper limit (Ͻ15%) estimated from the near UV CD data. This suggests a somewhat higher degree of specificity of the individual domains in the 1:1 C:N-domain mixture for their "native" binding sites on WFFp than is predicted from the interaction of two identical domains with the peptide. Introducing in the simulation a positive cooperativity factor of f ϭ 0.5 in forming the C-WFFp-N complex decreases the proportion of non-native complex to 11%, i.e. consistent with the near UV CD data. Thus, this factor of 0.5, corresponding to a change of less than 5% of the free energy change in the interaction, is sufficient to reconcile the homodomain titrations with the mixed domain near UV CD results. Thus, although the independent binding site model is generally adequate for the homodomain case, there is some evidence for a small degree of cooperativity between the complex of WFFp with the mixed N-and C-domains, and it is likely that this would be further enhanced in the interaction of calmodulin itself with the target sequence.
The estimated Gibbs free energy changes for formation of C-WFFp-N (Ϫ82 kJ/mol), C-WFFp-C (Ϫ84.6 kJ/mol), and N-WFFp-N (Ϫ76.2 kJ/mol) are significantly larger than that for WFFp binding to intact calmodulin (Ϫ66.5 kJ/mol). 3 These values may be compared with data of Persechini et al. (31) for the activation of the sk-MLCK enzyme by calmodulin fragments. They derived a value of the ratio of K C ⅐K N /K Cam ϭ 0.9 mM and equate this with an effective N-domain concentration for binding in the calmodulin-enzyme complex where the Cdomain is already bound. (K C and K N are the dissociation constants of the isolated C-domain and N-domain, respectively, from their native sites on the target sequence; K Cam is the dissociation constant for calmodulin.) The values obtained from the present studies of domain binding to the sk-MLCK target peptide correspond to values of 0.9 -1.8 mM, the higher value allowing for the inclusion of a small degree of cooperativity in domain binding. Given the completely different types of exper-iment, these values may be taken as showing an unexpectedly good degree of agreement.
The Gibbs free energy change for domain binding to the short peptides WF10p and FW10p accounts for 60 -80% of that of the domain binding to the corresponding end on the long peptides WFFp and FFWp (see Table I). This smaller free energy change for the short peptides is expected since, in the structure of calmodulin-M13 (18), the domains interact with more than the WF10p or the FW10p sequence of the target peptide; the number of C-domain residues in close contact with WF10p is estimated as 70% of those with WFFp, and the corresponding number for the N-domain and the FW10p portion is 80%.
Structural Model for the Domain-Peptide Complexes-The complexes of calmodulin with target peptide M13 (sk-MLCK; Ref. 18) and the RS20 peptide (sm-MLCK; Ref. 19) show an approximately two-fold rotational relationship between the Nand the C-domain of calmodulin as indicated in Fig. 4A. For the interaction of the two copies of one individual domain with the full target sequence, two distinct structural models may in principle be considered. They are presented schematically in Fig. 4 (B and C). It is assumed that the one copy of C-domain that binds to its native binding site on WFFp (colored red in Fig. 4, B and C) does so in its normal orientation as in calmodulin-WFFp. This is supported by the CD results, which give evidence for closely related binding modes for Trp in C-WFFp-C, C-WFFp-N, and the respective complexes with calmodulin (see above). The copy of the domain that binds to the non-native site (shown in green) may in principle do so in an orientation related to the native site either by rotational symmetry (axis out of plane perpendicular to the peptide, Fig. 4B) or by helical symmetry (translation along and rotation around the peptide axis, Fig. 4C). These distinct models have significant implications for the principles of calmodulin-target-sequence interactions. It is therefore important to analyze the similarities of the two domains together with the contacts which each makes with the target sequence.
Examination of the NMR structure of the calmodulin-M13 complex shows that of the 47 contact residues of the N-domain, 37 are conserved identically in the C-domain, and, of the 44 contact residues of the C-domain, 37 are conserved identically in the N-domain. Thus, the high degree of homology between the two domains and especially the 80% conservation of the residues in close contact with the peptide suggest that either isolated domain would bind to its non-native site on the peptide in the same orientation as does the alternate domain in the peptide complex with intact calmodulin, as illustrated for the C-domain in Fig. 4B. A detailed examination of the individual contact amino acids and their orientation relative to peptide residues confirms that the non-conserved residues are not expected to impair this type of interaction, either sterically or electrostatically.
In the calmodulin-MLCK peptide complex, the domain orientation relative to the peptide polarity is different for the Nand the C-domain (Fig. 4A) as a direct consequence of the approximate two-fold rotational symmetry of the domains. Progressing from the N terminus of the peptide to the C terminus, the peptide first contacts EF-hand IV of the C-domain and then EF-hand III, whereas with the N-domain it interacts with EF-hand I before EF-hand II. The ability of the isolated domains to form the C-WFFp-C and N-WFFp-N complexes, for which we infer a rotational symmetry as in the calmodulin complex, strongly suggests that either domain can accommodate peptide sequences with opposite polarities. Although this principle has been suggested by molecular modelling studies (17, 52), it has not yet been observed in the calmodulin-target peptide complexes for which detailed structures are available (2), although binding of melittin to calmodulin is known to occur with reversed polarity. 4 An analogous rotational relationship occurs in the complex of the regulatory light chain with myosin, where the EF-hands occur in the sequence III, IV, II, I (52,53). The principle of binding of peptides with either polarity is also consistent with the lesser importance of interactions with the peptide backbone as compared with those with the peptide side chains (17), the binding of calmodulin to a peptide composed of D-amino acids (51), and the calmodulin-mediated activation of sm-MLCK containing a reversed calmodulin binding sequence (50). Fig. 4 (B and C) shows that these considerations lead to some predictions about the relationship between target peptide sequence and the symmetry of the interactions. If a target sequence contains a palindromic feature (e.g. X-Y-Z . . . ZЈ-YЈ-XЈ, where the prime indicates at least a conservative homology between X and XЈ etc., if not a direct identity X ϭ XЈ), then the calmodulin domains would be most likely to adopt a two-fold rotational relationship. In fact, there is something of a palindromic relationship between residues 2 and 8 of M13 (R-R-W-K-K-N-F) and residues 13-19 (A-A-N-R-F-K-K), where bold type indicates the two key hydrophobic residues Trp-4 and Phe-17. By contrast, if a target sequence shows a repeating feature within a linear structure (e.g. X-Y-Z . . . XЈ-YЈ-ZЈ), then the domains would be most likely to adopt a helical (translational ϩ rotational) relationship. This does not appear to have been seen for calmodulin; however, it is intriguing that there appears to be a double (or triple) repeated feature in the sequence 96 -148 of skeletal and cardiac TnI apparently respon-sible for TnC binding (54), which might suggest that the two domains of TnC can bind with identical polarity relative to the TnI target sequence.
In conclusion, this work shows that the conformation and relative orientation of peptide polarity in the complexes formed between calmodulin and target sequences is likely to be characterized by specific features of the target sequence, and subtle affinity differences of the domains for the two binding sites. In the case of the WFFp sequence from sk-MLCK, a key role is played by the interaction of the C-domain with Trp-4 of the peptide. Recognition of the correct binding sites on the native target peptide is an intrinsic property of the isolated domains in a 1:1 mixture. However, the recognition process is clearly sensitive to the target sequence; two conservative point mutations in WFFp to generate the FFWp sequence suffice to change the conformation of the complex significantly as compared with that with calmodulin, and disturb the fine affinity balance of the domains for the two binding sites on the peptide. The observation of 2:1 complexes between calmodulin C-domain and WFFp target sequence (and between N-domain and target sequence) and detailed consideration of the known calmodulinpeptide complex structures indicate that a given domain need not bind to a target peptide with an exclusive polarity. The relative orientation of domain and peptide will be strongly determined by the peptide sequence. The combination of reverse polarity and possible interchangeability of position of the N-and C-domains implies that the potential versatility of calmodulin interactions, and their calcium dependence, can be significantly more diverse than has hitherto been inferred from the known high resolution molecular structures of calmodulintarget peptide complexes.  (18) (data are taken from Protein Data Bank file 1CDL.pdb using Rasmol). Red, target peptide, with polarity indicated by position of Trp-4 and the N and C termini. I, II, III, and IV refer to Ca 2ϩ ions (green) in the four EF-hands (turquoise). Helices (eight) are shown in purple, and odd-numbered ones H1, H3, H5, H7 are identified for clarity; loops joining two EF-hands within a domain are yellow, and the interdomain loop is orange. The position of the approximately two-fold rotational symmetry axis perpendicular to the paper and relating corresponding structural features of the N-domain (top right) and C-domain (bottom left) is indicated with ϩ. B and C, illustrative schematic models for the 2:1 C-domain-peptide complex. The domains are shown with two representative binding pockets. The domains binding to the native and non-native sites on the peptide are shown as red and green, respectively. B, orientation of the domains with rotational symmetry, as observed (approximately) for the C-and the N-domain in the complexes with intact calmodulin (see A). C, the red domain binding to its native site on the peptide does so in the same orientation as in the complex with calmodulin, whereas the green domain at the non-native site is reversed relative to B, and the relative orientation of the domains shows translation ϩ rotation (i.e. helical) symmetry.