Structural and thermodynamic characterization of the recognition of the S100-binding peptides TRTK12 and p53 by calmodulin

Abstract

Calmodulin (CaM) is a multifunctional messenger protein that activates a wide variety of signaling pathways in eukaryotic cells in a calcium-dependent manner. CaM has been proposed to be functionally distinct from the S100 proteins, a related family of eukaryotic calcium-binding proteins. Previously, it was demonstrated that peptides derived from the actin-capping protein, TRTK12, and the tumor-suppressor protein, p53, interact with multiple members of the S100 proteins. To test the specificity of these peptides, they were screened using isothermal titration calorimetry against 16 members of the human S100 protein family, as well as CaM, which served as a negative control. Interestingly, both the TRTK12 and p53 peptides were found to interact with CaM. These interactions were further confirmed by both fluorescence and nuclear magnetic resonance spectroscopies. These peptides have distinct sequences from the known CaM target sequences. The TRTK12 peptide was found to independently interact with both CaM domains and bind with a stoichiometry of 2:1 and dissociations constants K_d,C-term = 2 ± 1 µM and K_d,N-term = 14 ± 1 µM. In contrast, the p53 peptide was found to interact only with the C-terminal domain of CaM, K_d,C-term =2 ± 1 µM, 25°C. Using NMR spectroscopy, the locations of the peptide binding sites were mapped onto the structure of CaM. The binding sites for both peptides were found to overlap with the binding interface for previously identified targets on both domains of CaM. This study demonstrates the plasticity of CaM in target binding and may suggest a possible overlap in target specificity between CaM and the S100 proteins.

Introduction

Calmodulin (CaM) is a small (17 kDa), acidic, calcium-binding protein that is ubiquitous in eukaryotes.¹ It has been shown to bind more than 300 targets in response to transient fluctuations of intracellular calcium concentrations and plays critical roles in a number of cellular processes, including metabolism, the immune response, muscle contraction, and apoptosis.¹ CaM is also thought to play a role in several neurodegenerative disorders, such as Parkinson's² and Alzheimer's disease.³ This small protein consists of two distinct N- and C-terminal domains each containing two EF-hand calcium-binding motifs. Calcium binding induces large conformational changes within each globular domain which expose hydrophobic patches that are generally required for protein binding and various CaM functions.⁴ In the majority of available structures, the alpha helical linker between the N- and C-terminal domains unwinds to allow both domains to wrap around a single, usually α-helical, target peptide sequence.^1,5 However, an increasing number of novel structures have been reported where CaM adopts a more open, flexible conformation and interacts with its targets with differing stoichiometries.^4,6–8

Interestingly, CaM is proposed to be functionally distinct from the related, eukaryotic calcium-binding S100 proteins, due to differing subcellular localization, expression, and target recognition.^9–11 In particular, the TRTKIDWNKILS peptide (TRTK12), derived from the C-terminal region (residues 265–276) of the actin-capping protein (CapZ) has been proposed to be a specific binding motif for the S100 proteins and structurally distinct from CaM-binding peptides.^12–14 TRTK12 was previously identified as a consensus binding sequence for S100B^12,13 and subsequently shown to effectively act as a reversible inhibitor for interactions of glial fibrillary acidic protein or the full-length CapZ protein with S100B.¹² In addition, the TRTK12 peptide has also been shown to bind S100A1, S100A2, and S100P in a calcium-dependent manner.¹⁵ Similarly, the p53 C-terminal peptide, derived from the negative regulatory domain (residues 367–388) of the p53 tumor-suppressor protein, has previously been shown to bind to some members of the S100 protein family, including S100B and S100A1, in a calcium-dependent manner.^16–18 The p53 protein is a tetrameric transcription activator that regulates cell cycle arrest and apoptosis in response to cellular stress.¹⁹ S100 protein binding has been shown to prevent kinase C-dependent phosphorylation of Ser376 and Ser378, which are important for p53 activation and are proposed to stabilize the biologically active, tetrameric form of the protein.^14,16,19

To probe the binding specificity of the TRTK12 and p53 peptides to the S100 protein family, the peptides were screened for binding using isothermal titration calorimetry (ITC) against 16 members of the human S100 protein family, as well as CaM. Considering previous reports of distinct functionality of the S100 proteins and CaM,^9–11,20 the latter was intended to serve as a negative control. Unexpectedly, novel interactions were detected for both the TRTK12 and p53 peptides with human CaM. To gain a better understanding of these interactions, the binding of TRTK12 and p53 peptides were further characterized using thermodynamic (ITC), kinetic (NMR), and structural (NMR) methods.

Results and Discussion

Interactions of the p53 peptide with the S100 proteins

The C-terminal p53 peptide was screened for binding against 16 representative members of the human S100 protein family using ITC. Interactions were detected with only four proteins, S100B, S100A1, S100A2, and S100A4, consistent with previous findings by us and others.^14,16,18 The thermodynamics of these interactions are summarized in Table I and in the Supporting Information Figure S1. For the S100B and S100A1 proteins, the titration data fit well to the simplest binding model with a 1:1 stoichiometry, where an individual p53 peptide binds each subunit of S100 dimer. In contrast, the p53 peptide appears to bind the S100A2 and S100A4 proteins with a stoichiometry of 1:2, where a single peptide binds the S100 dimer. Such variations in the binding stoichiometry for the S100 proteins and peptides derived from p53 have previously been observed.^16,21 The dissociation constants obtained in ITC experiments are between 2 and 200 µM and are modulated by ionic strength (Table I), likely as a result of electrostatic screening of the large number of charged residues in the p53 peptide. Under physiological salt concentrations (120 mM NaCl) the measured binding affinities are consistent with previously reported dissociation constants measured by other methods.¹⁶

Table I.

Summary of the Dissociation Constants and the Changes in the Heat Capacity upon Binding of the p53 Peptide to CaM, S100B, S100A1, S100A2, and S100A4 at Different Ionic Strengths

Protein	a (µM)	b (kJ mol−1 K−1)	c (kJ mol−1 K−1)
S100B	2 ± 1	−1.0 ± 0.1	−1.0 ± 0.1
S100B, 120 mM NaCl	20 ± 5	−1.2 ± 0.3	−1.0 ± 0.1
S100A1	20 ± 5	−1.1 ± 0.1	−1.1 ± 0.1
S100A1, 120 mM NaCl	>200d	n.d.	−1.1 ± 0.1
S100A2, 120 mM NaCl	50 ± 3	−0.8 ± 0.1	n.d.
S100A4, 120 mM NaCl	50 ± 4d	n.d.	n.d.
CaM	2 ± 1	−0.9 ± 0.1	−1.3 ± 0.2
CaM, 120 mM NaCl	20 ± 1	−0.8 ± 0.1	−1.3 ± 0.2

^aDissociation constants at 25°C obtained using ITC.

^bThe experimentally determined changes in the heat capacity upon binding (Fig. 2).

^cStructure-based calculations [Eq. (2)] of homology models based on one peptide binding to each subunit of the dimer of human s100 proteins.

^dDissociation constant determined at 35°C.

n.d. not determined.

In addition, Table I also summarizes the heat capacity changes, ΔC_p, on p53 binding the S100 proteins, obtained from the temperature dependencies of the enthalpies of binding (Supporting Information Fig. S1). The values of ΔC_p, appear to cluster around −1.0 kJ mol⁻¹ K⁻¹. Importantly, for the S100B-p53 interaction, the experimentally determined ΔC_p value does not appear to vary with increasing ionic strength (−1.0 ± 0.1 kJ mol⁻¹ K⁻¹ vs. −1.2 ± 0.3 kJ mol⁻¹ K⁻¹). The magnitude of these ΔC_p values_, as well as their similarity for all S100-p53 interactions, suggest a significant amount of hydrophobic surface area is buried in all S100-p53 complexes.^18,22

Thermodynamics of binding of the p53 peptide to CaM

Figure 1(A) shows representative ITC data in which the C-terminal p53 peptide was titrated into CaM in the presence of 5 mM calcium chloride. The titration data fit well to the simplest model, where the p53 peptide binds CaM with a 1:1 stoichiometry (Table I). The dissociation constant obtained in the absence of physiological ionic strength, K_d = 2 ± 1 µM, is an order of magnitude tighter than that obtained in its presence, K_d = 20 ± 2 µM, (Table I). Increasing ionic strength has previously been shown to reduce the affinity of several biologically-relevant targets for CaM^23,24 and is also observed, for example, in S100B-p53 interactions (Table I).

Figure 1.

Representative ITC experiments showing the binding of p53 and TRTK12 to human CaM in the presence of calcium. Upper plots in each panel represent the raw heat effects as a function of time and lower plots show the cumulative heat effects (▪) as a function of the molar ratio of peptide to protein, and the fits to the experimental data for binding of p53 to CaM (A), the binding of TRTK12 to CaM at 25°C (B), and the binding of TRTK12 to the CaM N-terminal fragment CaM^(1–80) at 25°C (C).

Figure 2(A) shows the temperature dependence of the enthalpy of binding, ΔH_cal, of p53 to CaM obtained in the absence and presence of physiological ionic strength. Although there is a change in the dissociation constant on p53 binding to CaM, the enthalpy of binding is independent of ionic strength. The slope of the temperature dependence is linear and corresponds to the heat capacity change on binding, ΔC_p. These values are summarized in Table II. The experimentally determined ΔC_p value does not appear to depend on ionic strength [Fig. 2(A)], as the heat capacity change in the absence of NaCl (−0.9 ± 0.1 kJ mol⁻¹ K⁻¹) is in excellent agreement with that obtained in the presence of 120 mM NaCl (−0.8 ± 0.1 kJ mol⁻¹ K⁻¹). The similarity between the enthalpies of binding and changes in heat capacity on binding in the absence and presence of physiological ionic strength, as well as the modest change in the dissociation constant, suggest that the interactions are not driven by nonspecific electrostatic interactions. Furthermore, the magnitude of the negative ΔC_p suggests the involvement of hydrophobic interactions in the formation of the p53 peptide and CaM complex.

Figure 2.

Temperature dependence of the enthalpies of binding, ΔH_cal, for the p53 and TRTK12 peptides to CaM. Panel A: Binding of the C-terminal p53 peptide to CaM in a buffer containing 5 mM calcium chloride and either 0 mM NaCl (red inverted triangles) or 120 mM NaCl (black inverted triangle). Panel B: Binding of the wild-type TRTK12 peptide to human CaM and the N-terminal fragment CaM^(1–80) in the presence of 5 mM calcium chloride. The individual enthalpies for peptide binding to the first (inverted black triangles) and second (black triangles) binding sites of full-length human CaM are shown, as well as the total enthalpy of binding (black circles). The enthalpies of TRTK12 binding to CaM^(1–80) are overlaid (inverted orange triangles). Solid lines represent linear fits of the ΔH_cal temperature dependencies for each peptide-CaM interaction. The slopes of these lines represent the changes in heat capacity, which are summarized in.

Table II.

Summary of the Dissociation Constants, the Changes in the Heat Capacity upon Binding, and the Kinetic Rate Constants of the p53 Peptide and the TRTK12 Wild-Type Peptide and Its Alanine Variants Binding to CaM at 25°C

Peptide	a (µM)	a (µM)	b (µM)	c (kJ mol−1 K−1)	c (kJ mol−1 K−1)	d (kJ mol−1 K−1)
p53	2 ± 1	—	n.d.	−0.9 ± 0.1	—	−1.3 ± 0.2, n/a
p53, 120 mM NaCl	20 ± 1	—	n.d.	−0.8 ± 0.1	—	−1.3 ± 0.2 n/a
TRTK12	3 ± 1	14 ± 1 [25 ± 4]	7 ± 1	−0.6 ± 0.1	−0.6 ± 0.1 [−0.7 ± 0.2],	−0.9 ± 0.2, −0.6 ± 0.2
TRTKM1 (T1A)	4 ± 1	20 ± 1 [20 ± 1]	5 ± 1	−0.7 ± 0.1	−0.6 ± 0.1 [−0.8 ± 0.2]	−0.9 ± 0.2, −0.7 ± 0.2
TRTKM5 (I5A)	10 ± 1	100 ± 3	40 ± 1	−2.0 ± 0.3	−0.7 ± 0.1	−1.1 ± 0.2, −0.7 ± 0.2
TRTKM6 (D6A)	0.3 ± 0.1	6 ± 1	2 ± 1	−1.1 ± 0.1	−0.3 ± 0.1	−1.3 ± 0.2, −0.8 ± 0.2
TRTKM10 (I10A)	20 ± 2	100 ± 5	60 ± 4	−0.7 ± 0.1	−0.6 ± 0.1	−1.0 ± 0.2, −0.8 ± 0.2
TRTKM11 (L11A)e	80 ± 1	n.d.b.	70 ± 10	−1.0 ± 0.1	n.d.b.	−1.1 ± 0.2, n/a

^aDissociation constants obtained using ITC using two independent site binding model. Values in square brackets were obtained for titrations of the peptide into the N-terminal domain fragment of calmodulin CaM^(1–80) (see Materials and Methods).

^bValues were obtained using emission fluorescence assays and single site model and thus represent apparent binding constants.

^cThe experimentally determined changes in the heat capacity upon binding for each site (Figure 2). Values in brackets were obtained for titrations of the peptide into the N-terminal domain fragment of calmodulin CaM^(1–80) (see Materials and Methods).

^dStructure-based calculations (Equation (2)) of homology models based on one peptide (PDB: 1CFF) or two peptides (PDB: 2LLQ/2LLO) binding to human CaM. First value for C-terminal domain and second value for N-terminal domain.

^eITC Titration data fit best to a single-site binding model for this peptide.

n.d.b., no detectable binding; n/a, not applicable; n.d., not determined.

Thermodynamics of binding of the TRTK12 peptides with CaM

Figure 1(B) shows representative ITC data in which the wild-type TRTK12 peptide was titrated into CaM in the presence of 5 mM calcium chloride. The ITC data fit well to the two independent binding sites model. The 2:1 stoichiometry was confirmed by performing reverse titrations (data not shown). The dissociation constants, K_d,1 and K_d,2, obtained from the fit of the wild-type peptide binding to CaM were determined to be 3 ± 1 µM and 14 ± 1 µM at 25°C (Table II). There is approximately a fivefold difference in affinity between the two binding sites, suggesting the peptide preferentially binds to one of the domains of CaM.

In addition to the wild-type TRTK12, the thermodynamics of binding were measured for five alanine variants of this peptide (see Materials and Methods for sequences and rationale). There is reasonable agreement between the dissociation constants obtained from ITC for the individual sites and the apparent dissociation constants obtained using fluorescence spectroscopy (Table II). All peptides, except TRTKM11 (L11A), were found to bind with a stoichiometry similar to that of wild type, that is, two peptides per CaM. The TRTKM11 (L11A) peptide appears to bind CaM with a stoichiometry of 1:1, suggesting that L11 is a critical residue for binding and that the L11A substitution abolishes binding to one of the two available sites. Interestingly, the same TRTKM11 peptide was found to bind several S100 proteins, including S100B and S100P, with an atypical stoichiometry of one peptide per S100 dimer.¹⁵ For the majority of alanine variants, the dissociation constants for each of the two binding sites were found to differ from the wild-type peptides to different extents. For example, the dissociation constants for the first (K_d1 = 10 ± 1 µM) and second (K_d2 = 100 ± 2 µM) binding sites for the TRTKM5 variant are approximately threefold and sevenfold weaker, respectively, than those of wild type. The binding affinities for individual variants suggest that there may be important interactions associated with the hydrophobic side chains of I5, I10, and L11 of the TRTK12 peptide. For example, the binding affinities for the first (K_d1 = 20 ± 2 µM) and second (K_d2 = 100 ± 5 µM) binding sites of the TRTKM10 variant are approximately sevenfold weaker than those of wild type. Furthermore, the L11A substitution in the TRTKM11 peptide abolishes binding for one of the sites and significantly weakens binding for the other (K_d1 = 100 ± 2 µM). Electrostatic interactions also appear to modulate binding as the TRTKM6 (D6A) peptide was found to have a significantly tighter affinity than the wild-type peptide for both binding sites (Table II). This is consistent with the large number of negative charges on the surface of CaM, which may act to electrostatically repel the wild-type peptide.

Structural aspects of TRTK12 peptide binding to CaM

Previously, we have established that the TRTK12 and p53 peptides bind to CaM with a 2:1 and 1:1 stoichiometry, respectively. Here, we use NMR spectroscopy and computational modeling to map the location of the peptide binding sites on CaM. The interactions between TRTK12 and CaM were examined by monitoring binding-induced changes in the chemical shifts of ¹⁵N/¹H CaM backbone atoms between the peptide-free and peptide-bound states. The TRTK12 peptide appears to bind CaM with a stoichiometry of 2:1, and occupy distinct binding sites on both the N- and C-terminal domains [Fig. 3(B)]. The residues involved in binding the wild-type TRTK12 peptide in the N-terminal domain appear to include Glu14, Phe16, Phe19, Gly33, Val35, Met36, Glu47, Asp50, Ile52, Phe68, Thr70, Met71, Met72, Ala73, Arg74, and Asp80. Significant chemical-shift changes were also observed for residues in the first (Thr26, Ile27, and Thr29) and second (Ala57, Ile63, and Asp64) calcium-binding sites within the N-terminal domain. This is consistent with observations for other CaM-binding peptides, such as melittin,²⁶ and suggests the structural linkage between calcium binding and peptide binding to the N-terminal domain of CaM. The chemical-shift perturbations observed for residues from the C-terminal domain were similar to those involved in binding p53 and include Phe89, Val91, Leu105, Arg106, Asn111, Ala128, Phe141, Val142, Gln143, Met144, Met145, Thr146, Ala147, and Lys148. Significant binding-induced chemical-shift changes were also observed for residues in the third (Asp93 and Ala103) and fourth (Ile130 and Asp133) calcium-binding sites within the C-terminal domain.

Figure 3.

Changes in backbone ¹H and ¹⁵N chemical shifts for TRTK12 (A) and p53 (B) binding to Ca²⁺-bound CaM (PDB: 2L7L²⁵). Residues are color coded according to the degree of combined backbone ¹H and ¹⁵N chemical-shift perturbations (c.s.p.) on TRTK12 (C) and p53 (D) binding. Color assignment was done as follows: c.s.p. 0.3–0.4 (yellow), c.s.p. 0.4–0.5 (orange), and c.s.p. > 0.5 (red). Selected residues are labeled for orientation only.

The 2:1 binding stoichiometry of TRTK12 to CaM suggests that the binding mode is different from canonical one, in which both CaM domains to wrap around a single target peptide sequence. However, a similar 2:1 stoichiometry has been observed for the interactions of the NSCaTE and NtMKP1 peptides with CaM.^27,28 A sequence analysis of TRTK12 shows little similarity to the canonical CaM-binding epitopes. The peptide lacks the glycine and glutamine residues that are common to the IQ motif, as well as the hydrophobic residue spacing common to the 1-10, 1-14, and 1-16 motifs. Instead, the TRTK12 peptide aligns best with noncanonical binding sequences (Fig. 5), such as CaD-A and MLC, which have been shown to preferentially interact with the C-terminal domain of CaM.²⁹

Figure 5.

Alignment of the TRTK12 peptide with known CaM-binding peptides. The peptides are aligned according to the position of their charged residues (blue or red) and hydrophobic anchor (green), which is proposed to bind in a hydrophobic pocket of the C-terminal domain of CaM. The location of the N- (N) and C-termini (C) of each peptide is indicated.

To confirm the binding model of TRTK12 to CaM, the interactions of the wild-type peptide and its alanine variants with the isolated N-terminal domain fragment of CaM, CaM^(1–80), were also studied. Titrations of the TRTK12 peptides and CaM^(1–80) consistently resulted in stoichiometry of 1:1 [Fig. 1(C)] and K_d = 25 ± 4 µM, which compares well with the K_d,2=14 ± 1 µM for the second site of the full-length CaM (Table II). Furthermore, the enthalpies of binding for the TRTK12 peptides and CaM^(1–80) are in excellent agreement with those obtained for one of the two binding sites on full-length CaM [Fig. 2(B)]. Likewise, the ΔC_p value obtained for the binding of the TRTK12 peptide to CaM^(1–80), −0.7 ± 0.1 kJ mol⁻¹ K⁻¹, is similar to that obtained for TRTK12 binding to one of the individual sites on full-length CaM, −0.6 ± 0.1 kJ mol⁻¹ K⁻¹ (Table II). Together, these results confirm that the binding sites are delocalized and there is one for each CaM domain.

While the orientation of the peptide within the N-terminal domain is unknown, it is likely that tryptophan 7 (W7) of TRTK12 is buried in a hydrophobic pocket. This is supported by tryptophan fluorescence experiments performed by titrating CaM^(1–80) into a solution containing the wild-type TRTK12 peptide, where both an increase in intensity and a blue shift in λ_max were observed (Supporting Information Figure S2). In addition, significant chemical-shift perturbations were observed for a large number of hydrophobic and charged residues in the N-terminal domain that make up the canonical binding pocket of CaM. These residues include Glu14, Phe19, Val35, Met36, Glu47, Ile63, Phe68, Met71, Met72, and Asp80 [Fig. 3(B)]. Homology models based on the CaM-NSCaTE (PDB: 2LQC²⁷) and CaM-ER-α (PDB: 2LLO⁶) complexes are consistent with the burial of W7. For illustrative purposes, Figure 4 also shows a high-scoring complex of CaM-TRTK12 generated using HADDOCK (see Materials and Methods for details). The generated structures were furthered evaluated and scored based on their ability to recapitulate the experimental data (see Materials and Methods for selection criteria). As with the homology models, the HADDOCK-generated structures bury W7 in the canonical tryptophan-binding pocket. Furthermore, the highest-scoring structures place the peptide in the same conformation as the homology models: an antiparallel orientation with the peptide's C-terminus lying between helices III and IV of CaM. These models suggest D6 from TRTK12 is in a somewhat unfavorable environment, as it lies adjacent to similar charges in N-terminal domain of CaM, including Glu14. This is consistent with the enhanced binding affinities observed for the TRTKM6 (D6A) variant (Table II).

Figure 4.

Panel A: Homology model of the TRTK12 peptide (pink) in complex with the N-terminal domain of CaM (orange) generated using HADDOCK. This structure is superimposed with the complex of the NSCaTE peptide (cyan) binding to the N-terminal domain of CaM (PDB: 2LQC²⁷). Panel B: An alternate view of the structures shown in Panel A. The tryptophan side chain of both peptides is shown as sticks for comparison. Panel C: Homology model of the TRTK12 peptide (pink) in complex with the C-terminal domain of CaM (red) generated using HADDOCK. This structure is superimposed with the complex of the C20W peptide (cyan) binding to the C-terminal domain of CaM (PDB: 1CFF³⁰). Panel D: An alternate view of the structures shown in Panel C. The tryptophan side chain of both peptides is shown as sticks for comparison.

Homology models of TRTK12 in complex with the C-terminal domain of CaM also produced a consistent set of structures. Homology models based on the CaM-C20W (PDB: 1CFF³⁰) and CaM-ER-α (PDB: 2LLQ⁶) complexes suggest that the TRTK12 peptides lies in the C-terminal domain with the peptide's N-terminus pointing away from the center of the protein and the peptide's C-terminus lying between helices V and IV. This orientation is also consistent with the highest-scoring complexes generated using HADDOCK (see Materials and Methods). In these structures, W7 of TRTK12 is buried in the canonical tryptophan-binding pocket of CaM (Fig. 4), consistent with the significant chemical-shift perturbations observed for this cluster of nonpolar residues in the C-terminal domain. In addition, this arrangement allows for several favorable electrostatic interactions, placing the R2 and K4 residues from TRTK12 within range of several acidic residues from CaM, including Asp80, Glu84, and Glu127. Interestingly, this proposed conformation of the bound state aligns D6 of the peptide near several similar charges on CaM, including Glu123 and Glu127. As with the N-terminal domain, this orientation is consistent with the tighter dissociation constants observed for the TRTKM6 variant, which substitutes D6 with an alanine.

The existence of several charge–charge interactions, as well as the burial of some polar residues from TRTK12, may explain the smaller magnitude of ΔC_p (−1.3 to −0.8 kJ mol⁻¹ K⁻¹ Table II) observed for binding to the individual CaM domains as opposed to the canonical binding of 1 peptide per CaM (−3.7 to −2.7 kJ mol⁻¹ K⁻¹ see Ref.8). Furthermore, the available structures of the TRTK12 peptide interacting with the closely-related S100 proteins suggests that only a portion of the peptide directly interacts with its binding partners, leaving the remaining sequence solvent exposed.^31,32 Nonetheless, homology models of the CaM-TRTK12 complex suggest that the total amount of surface area buried per domain, 1600 ± 100 Ų, is similar to what has been observed for several other CaM-binding targets, such as the NSCaTE²⁷ and MLC⁸ peptides. Several bulky, hydrophobic side chains of the TRTK12 peptide may anchor the peptide to CaM, contributing to the significant amount of buried surface area. This is consistent with the significantly weaker dissociation constants observed for the TRTKM5 (I5A), TRTKM10 (I10A), and TRTKM11 (I11A) variants (Table II). It is also consistent with the highest-scoring structures generated by HADDOCK, where I5, I10, and L11 are buried against hydrophobic residues in the binding pockets of both domains (Fig. 4). It is important to note that the conformation of the TRTK12 peptide in the bound form remains to be determined. Homology modeling suggests that the peptide in the bound form should adopt helical conformation, while the HADDOCK model, generated based on available constraints, suggests a more extended conformation.

Structural aspects of p53 peptide binding to CaM

The following CaM residues experienced significant changes in chemical shift on p53 peptide binding: Phe19, Asp50, Thr70, Ala73, Asp80, Ile85, Ala88, Val91, Phe92, Asp93, Asn97, Leu105, Val108, Met109, Thr110, Asn111, Lys115, Val121, Glu127, Ala128, Asp129, Phe141, Val142, Gln143, Met144, Met145, Thr146, Ala147, and Lys148. The vast majority of residues found to interact with p53 appear to structurally cluster to a region in the C-terminal domain of CaM. The only exception is residue F19 from the N-terminal domain, which appears to interact with p53 in the bound complex, possibly through an interaction with the tail portion of the peptide.

The analysis of ITC binding isotherms, together with the chemical-shift data obtained from NMR, strongly suggest that the p53 peptide binds CaM with a 1:1 stoichiometry. Furthermore, these interactions map to the C-terminal domain of CaM (Fig. 3). Previously, it has been shown that peptides bound to a single domain of CaM have a less negative ΔC_p than those bound to both domains via canonical binding.⁸ This is consistent with the relatively small absolute value obtained for the change in the heat capacity on p53 peptide binding (−0.9 ± 0.1 kJ mol⁻¹ K⁻¹) and the ΔC_p value for the noncanonical interactions of several peptides with CaM.⁸

Interestingly, many of the same CaM residues appear to be involved in binding both the p53 and TRTK12 peptides. This is consistent with the available structures of several targets in complex with the C-terminal domain of CaM alone, such as eNOS,³³ the ER-α,⁶ CaMKI²⁵ and suggests that the binding interface often consists of the same key residues. Within the bound complex, many peptides adopt an antiparallel orientation, with their N-termini pointing away from the center of CaM toward the protein's C-terminal end. However, significant heterogeneity has been observed for several peptides in complex with CaM and conformational changes in both CaM and its binding partners appear to be very sensitive to slight differences in the target sequences.^26,34,35 Furthermore, several peptides have been observed to bind CaM in both parallel and antiparallel orientations, particularly when they contain so called “palindromic sequences.”³⁶ This may be the case for the p53 peptide, which contains large hydrophobic anchors (L4 and F20) at both its N- and C-termini, but does not contain an intervening tryptophan. In addition, this peptide contains a large number of lysine, arginine, and histidine residues regularly spaced throughout its core. The inherent heterogeneity in the bound complexes suggests that identifying the residues of the binding interface of CaM may be more significant than the particular orientation of a single, bound conformation.

Line-shape analysis and the kinetics of peptide binding

For the TRTK12 peptide titration, residues that underwent significant changes in their backbone chemical shifts could be separated into two groups based on their individual titration profiles. One group includes residues Phe16, Phe19, Gly33, Ala57, Asp64, and Phe68 while the other includes residues Val91, Ala102, Leu105, Ala128, Ile130, Gln143, Ala147, and Lys148. Importantly, the residues from the first group map to a distinct location on the N-terminal domain while residues from the second group map onto the C-terminal domain of CaM (Fig. 3). One-dimensional (1D) line shapes extracted from the series of 2D HSQC spectra were analyzed using LineShapeKin (Material and Methods) to obtain the k_off rate constants for the TRTK12 peptides' dissociation from CaM. The quality of the fits from LineShapeKin (see Materials and Methods), as determined by both χ² analysis and visual inspection, suggests that the dissociation constants for the C-terminal and N-terminal domains are approximately 3 ± 1 and 12 ± 1 µM at 35°C, respectively (see Supporting Information Fig. S3). These are in excellent agreement with the dissociation constants obtained using ITC at 35°C: 5 ± 1 and 15 ± 1 µM, respectively. The kinetic parameters of binding obtained from the fits are as follows: k_off1 = 1400 ± 210 s⁻¹ and k_on1 = (4.5 ± 1.7) × 10⁸ M⁻¹ s⁻¹ for the C-terminal domain and k_off2 = 570 ± 120 s⁻¹ and k_on2 = (4.8 ± 1.1) × 10⁷ M⁻¹ s⁻¹ for the N-terminal domain.

For the p53 titration, all CaM residues that experienced significant chemical-shift perturbations exhibited a monophasic dependence on the peptide concentration throughout the titration. This is consistent with the stoichiometry observed in the ITC titration (i.e., 1:1). Binding was observed to occur in the intermediate to fast exchange regime (k_off ≥ Δω), which allowed the position-averaged chemical shifts of the peptide free and peptide-bound forms to be easily tracked throughout the titration (see Supporting Information Fig. S3). The apparent dissociation constant calculated from line-shape analysis, using residues Asp93, Leu105, Ala128, Phe141, and Lys148, lies between 0.3 and 0.6 µM and is similar to 2 ± 1 µM obtained from ITC. The fitted data resulted in the same k_off value, 200 ± 40 s⁻¹, whether the K_d value was simultaneously fit with k_off or whether K_d was held constant using the value obtained from ITC. As the dissociation constants obtained from ITC have been proposed to be more accurate than those directly obtained from line-shape analysis,³⁷ the K_d from ITC was used to calculate the apparent k_on rate constant: (9 ± 5) ×10⁷ M⁻¹ s⁻¹.

The k_off rate constants determined using line-shape analysis for both TRTK12 and p53 are relatively fast compared to other CaM-binding peptides. For targets with nanomolar binding affinities to CaM, k_off is generally observed to lie between 0.1 and 10 s⁻¹ with associated k_on rate constants ≥1 × 10⁸ M⁻¹ s⁻¹.^29,38 There is limited kinetic data available for peptides that bind CaM with micromolar affinities, but k_off for at least one peptide has been observed to be greater than or equal to 400 s⁻¹.³⁹

The fast bimolecular rate constants observed in this study may suggest a mechanism of binding. It is well-established that the TRTK12 and p53 peptides lack significant secondary structure in solution and, therefore, represent intrinsically disordered peptides (IDP's).¹⁸ In addition, these peptides likely adopt a helical conformation in the bound state, given that this is the predominant structural motif for CaM-binding targets¹ and that these peptides contain significant helical content when in complex with the S100 proteins.^31,32,40 Previous work has demonstrated that the coupled binding and folding process for IDP's generally proceeds via one of two mechanisms: conformational selection (binding after folding) or “fly-casting,” (folding after binding).⁴¹ It has been suggested that these mechanisms can be discriminated in the case of IDP binding if the apparent k_on is sufficiently fast, ≥1 × 10⁷ M⁻¹ s⁻¹, due to the dependence of this rate constant on the folding equilibrium.⁴¹ The k_on (9 ± 5) ×10⁸ M⁻¹ s⁻¹ for p53 and (4.5 ± 1.7) ×10⁸ M⁻¹ s⁻¹ for TRTK12 binding to the C-terminus of CaM exceed this threshold by more than an order of magnitude, which suggests they may indeed bind via a “fly-casting” mechanism. The presence of certain kinetic intermediates is also considered to be direct evidence of a “folding after binding” mechanism.^41,42 Although not observed in this study, a variety of CaM-binding targets have been observed to associate in a biphasic manner.^38,42 This has been proposed to include a fast phase, where an initial encounter complex is formed, and a slower phase, where the target rearranges to exclude water, bury several ion pairs, and often forms an alpha helix.⁴² These data are consistent with previous suggestions that the “fly-casting” mechanism may be the predominant binding mode for IDP's and their binding partners.^43–45 The results presented here provide additional experimental evidence that a large number of unstructured ligands interact with their protein receptors through a “fly-casting” mechanism.^18,43,44

Target overlap between CaM and the S100 proteins

Although more than 300 target sequences are known to bind CaM, the vast majority can be classified as belonging to one of four binding motifs: 1-14, 1-10, 1-16, and the IQ motif.¹ The specific spacing of bulky hydrophobic groups, positively charged residues, and/or polar glutamine residues in these motifs is proposed to be crucial for the high affinity often observed for CaM binding partners. In addition, these binding epitopes are generally unstructured in solution, but adopt a helical structure in the bound complex.¹ Interestingly, the sequences of TRTK12 and p53 peptides do not fit into any of the aforementioned binding motifs despite their net positive charge, overall hydrophobicity, and lack of structure in solution. A search of the CaM Target Database produced no results, that is, no similar sequences were found in the database of known targets.¹ As such, both peptides appear to be novel in vitro binding partners for CaM. Furthermore, both peptides produced a maximum, normalized score of 0 for all residues using the putative CaM binding site analysis.¹

Canonical sequence recognition by CaM is characterized by a 1:1 stoichiometry, with both the N- and C-terminal domains wrapping around an α-helical peptide/protein.¹ However, several studies have also identified target peptide sequences that interact with CaM by only binding to the N- or C-terminal domain.^27,28,30 In addition, peptides have been observed to bind both domains individually, sometimes with micromolar affinities.^7,46,47 Importantly, the interactions between CaM and full-length proteins have been shown to depend on the presence of short, linear sequence motifs within its binding partners.¹ Therefore, using peptides that represent these sequences to study CaM binding is a well-established approach. The dissociation constants obtained for the TRTK12 (3 ± 1 µM and 14 ± 2 µM) and p53 (2 ± 1 µM) peptides are similar to the dissociation constants for other biologically relevant, CaM-binding partners. Furthermore, the dissociation constants appear to be independent of the binding mode, as they are similar for both canonical (e.g., α-synuclein 35 ± 10 µM⁴⁸) and noncanonical (e.g., caldesmon 1.4 ± 0.1 µM,^8,49 NSCaTE 0.1–2.0 µM⁷) target interactions. In addition, the binding affinity of the p53 peptide for CaM in the presence of physiological ionic strength is significantly tighter than those observed for several of the S100 proteins, which have been shown to interact with this sequence in vivo.^16,50,51 Similarly, the binding affinities of the TRTK12 peptides for CaM are similar to those observed for S100B (K_d= 2 ± 0.1 µM), which has been shown to bind both the TRTK12 peptide and the full-length CapZ protein.¹² Lastly, CaM has been shown to bind to proteins that perform similar functions as the p53 and CapZ proteins. These include NAP-22,⁵² p68,⁵³ and adducin.⁵⁴ Taken together, this suggests that the interactions of the p53 and TRTK12 peptides with CaM may be physiologically relevant. However, it should be noted that additional studies will be necessary to substantiate these interactions in vivo with the full-length p53 and CapZ proteins.

It is interesting that the TRTK12 and p53 peptides bind not only to CaM but also to several members of the S100 protein family.^12,15,16,18 Early work^9,55 on these proteins sought to differentiate them from CaM and related calcium-binding proteins. In particular, they were shown to exhibit different subcellular localization and oligomerization, unique responses to differentiation, distinct binding clefts, and to interact with a variety of distinct intracellular and extracellular targets.^9–11,20,56 However, the functional relationship between CaM and the S100 proteins remains unclear. It is now thought that the S100 proteins arose through a gene duplication of C-terminal domain of CaM, with subsequent perturbation of the calcium binding affinity in the remaining N-terminal EF-hand.⁵⁷ Phylogenetic analysis of the human S100 proteins shows they consist of four subgroups: A2/A3/A4/A5/A6, A13/A14/A16, A7/A8/A9/A12/G, and A1/A10/A11/B/P/Z.⁵⁷ Pairwise sequence alignments of the S100 proteins and CaM show that the latter subgroup, which consists of A1/A10/A11/B/P/Z, shares the largest degree of sequence similarity with CaM (Fig. 6). This is also evident from comparisons of the binding site residues of these proteins (Supporting Information Table S1), which show that the A1/A10/A11/B/P/Z and A2/A3/A4/A5/A6 subgroups contains a similar percent of hydrophobic, polar, and acidic surface area as CaM, relative to other S100 proteins.²⁰

Figure 6.

Percent identity matrix of 16 representative human S100 proteins and the C-terminal domain of CaM generated using ClustalOmega.⁵⁸ CaM shares the largest percent identity with members of the ancestral S100 proteins: S100A1, S100A11, S100B, S100P, and S100Z.

These sequence homologies may explain the similarity in plasticity of binding that has been observed for these proteins. In particular, CaM has been shown to interact with its binding partners with different stoichiometries, multiple target orientations, large variations in affinity, and using one or both domains.^1,4,7,8,59 Similar binding plasticity has been observed for the S100 proteins.^10,18,60,61 For example, a recently published structure indicates that S100A4 is able to interact with nonmuscle myosin IIA tail in an asymmetric manner, where a single peptide simultaneously interacts with both canonical binding sites and stretches across the S100A4 dimer interface.⁶⁰ Likewise, S100A1, S100A2, and S100A6 have been reported to interact with peptides derived from the MDM2 proteins through a sequential binding model with strong, negative cooperativity.⁶¹ Atypical binding has also been observed for the closely-related S100B¹⁸ and S100A10/A11 proteins.¹⁰ In addition, several studies have noted overlapping target specificity between CaM and the S100 proteins, particularly those derived from the S100B branch of the phylogenetic tree. The overlapping targets include melittin,^8,62 caldesmon,^8,63 tau,^64,65 TRPM3,⁶⁶ IQGAP1,^67,68 bHLH transcription factors,^69,70 p37,^71,72 NDR,^18,73 neuromodulin,^14,74 the ryanodine receptor,⁷⁵ twitchin kinase,^76,77 tropomyosin,⁷⁸ and now the TRTK12 and p53 peptides. Taken together, these results raise the question of whether or not a functional overlap may exist between CaM and the S100 proteins. Whether these interactions occur in vivo remains to be tested.

Materials and Methods

Protein purification

Full-length human CaM and representative members of the human S100 protein family, including S100A1, S100A2, S100A3, S100A4, S100A5, S100A6, S100A7, S100A8, S100A9, S100A10, S100A11, S100A12, S100A13, S100P, S100Z, and S100B, were overexpressed in Escherichia coli and purified as previously described.^8,10,18 The molar extinction coefficients at 280 nm (ε₂₈₀) were calculated as previously described.¹⁵ The N-terminal domain of human calmodulin, CaM^(1–80), was transformed into BL21 (DE3) cells using the PT7-7 bacterial vector and purified as previously described.⁷⁹ The sequence of CaM^(1–80) contains no tryptophans, tyrosines, or cysteine residues, so the concentration was measured using the extinction coefficient for the peptide backbone determined at 205 nm:⁸⁰

Peptide purification

All peptides were synthesized at the Penn State College of Medicine Macromolecular Core Facility using standard Fmoc chemistry with their N- and C-termini acetylated and amidated, respectively. The C-terminal p53 peptide [p53^(367–388); amino acid sequence, Ac-YSHLKSKKGQSTSRHKKLMFKTE-Am], and the TRTK12 wild-type peptide [Cap-Z^(265–276); amino acid sequence, Ac-TRTKIDWNKILS-Am], derived from the actin-binding protein, CapZ, have previously been used in structural studies.^{31,32,81–83} For clarification, the tryptophan located at the seventh position in the TRTK12 peptide sequence [Cap-Z⁽²⁷¹⁾] will be referred to as “W7.” The alanine variants of this peptide sequence, including TRTK12M1 (Ac-ARTKIDWNKILS-Am), TRTK12M5 (Ac-TRTKADWNKILS-Am), TRTK12M6 (Ac-TRTKIAWNKILS-Am), TRTK12M10 (Ac-TRTKIDWNKALS-Am), and TRTK12M11 (Ac-TRTKIDWNKIAS-Am) were previously studied by this lab.¹⁵ These variants were designed based on the hydrophobic and electrostatic interactions reported in the literature between the wild-type TRTK12 peptide and the S100B and S100A1 proteins.^31,32,81,82 The peptide concentrations were determined using molar extinction coefficients (ε₂₈₀) of 5500 M⁻¹cm⁻¹ for the TRTK12 wild-type and alanine variants and 1490 M⁻¹ cm⁻¹ for the p53 peptide.⁸⁴

Isothermal titration calorimetry

ITC measurements were performed using a VP-ITC instrument (MicroCal, Northhampton, MA) as previously described.^10,18,62,85 The resulting titration curves were analyzed using the Origin scripts for ITC data analysis provided by MicroCal. The data were analyzed using a variety of binding models, in the order of increasing complexity, as described previously.¹⁵ The simplest model that adequately fit the data was accepted.

NMR spectroscopy

NMR spectra were collected at 25 or 35°C on a Bruker AVANCE II 600 or 800 MHz spectrometer equipped with a triple-resonance cryoprobe with z-axis gradients and processed using the Bruker software suite TopSpin (version 2.1). Backbone resonance assignments were confirmed for human CaM in the peptide-free state using ¹H-¹⁵N HSQC and ¹⁵N-separated 3D TOCSY experiments and previously published assignments as input.^5,25 Reassignment of the amide groups in the peptide-bound state were determined by monitoring changes in chemical shifts throughout the peptide titrations in a series of 2D ¹H-¹⁵N HSQC spectra. The experimental buffer consisted of 10 mM Tris, 5 mM CaCl₂, 15 mM NaCl, and 0.34 mM NaN₃ and 10% D₂O (pH 7.5). Chemical-shift mapping experiments were performed by titrating 0–1 mM peptide into 0.1–0.4 mM calcium-bound CaM and monitoring changes in backbone ¹⁵N and ¹H. Titrations were continued until there were no observable changes in the chemical shifts. The combined perturbation of the proton and nitrogen chemical shifts, ΔCS, was quantified according to the following weighted metric:^18,86

(1)

where δ_N and δ_H are the changes in chemical shift (parts per million) in the nitrogen and proton dimensions, respectively. Spectra were analyzed using Sparky.⁸⁷ Chemical-shift perturbations were considered to be significant if they were greater than a weighted average of 0.3.

LinShapeKin analysis

Line-shape analysis was performed using the BiophysicsLab Matlab package,⁸⁸ as previously described.¹⁸ The 1D line shapes in the ¹⁵N and ¹H dimensions from 2D HSQCs were extracted in Sparky^87,89 and the intensities were normalized to account for exchange broadening in the other dimension.⁸⁸ Titration data for the p53 peptide were fit to a model in which there is one binding site per CaM. Titration data for the TRTK12 peptide were analyzed independently for each of the two binding sites of CaM (see Supporting Information Methods for details). The LineShapeKin software simultaneously solves for both the apparent binding constant, K_d, and the off rate, k_off, analytically, which allows the on rate, k_on, to be easily calculated. As previously described,¹⁸ amide peaks that underwent significant chemical-shift changes during the titration and met the necessary criteria were selected for analysis. Structurally-clustered, residue-specific data were assumed to be reporting on the same binding event. This assumption was validated by comparing the fits of individual residues with each other and the global fit for structurally-clustered residues, which were in good agreement. All reported values are the result of global fits, which include both ¹⁵N and ¹H dimensions of all relevant amide groups across all ligand concentrations.

Structure-based calculations of ΔC_p

Using the available 3D structures of CaM in the calcium bound state and/or in complex with a peptide target, ΔASA_tot was calculated as previously described.²² Homology models of the peptide-bound state were generated using Modeller⁹⁰ and PDB entries 1CDL,⁹¹ 1CFF,³⁰ 2LLO,⁶ 2LLQ,⁶ and 2LQC.²⁷ Changes in the accessible surface area on binding, ΔASA_total, were calculated as previously described.^10,18,62 Changes in ΔASA were further subdivided into four categories (aliphatic surface area, aromatic surface area, peptide backbone surface area, and polar surface area) and converted into ΔC_p using the following empirical relationship:^{8,10,18,22,92}

(2)

where ΔASA_alp, ΔASA_arm, and ΔASA_pol are the changes in ASA for aliphatic, aromatic, and polar amino acids, respectively, and ΔASA_bb is the change in ASA for the polypeptide backbone.

HADDOCK docking

To further explore the interaction between the p53 and TRTK12 peptides with CaM, a series of docked complexes were generated using HADDOCK.^93,94 The p53 (PDB: 1DT7) and TRTK12 (PDB: 1MWN) structures were extracted from existing complexes and docked with calcium-bound, full-length CaM (PDB: 2M55 or 1CFF), or the individual calcium-bound N-terminal (PDB: 2M55, residues 1–78, or PDB: 1CFF, residues 1–78) and C-terminal (PDB: 2LQP) domains. Ambiguous interactions restraints were defined for CaM by the residues that underwent significant backbone chemical-shift perturbations, as monitored by ¹⁵N/¹H heteronuclear HSQCs, during the peptide titrations. The restrains for the p53 peptide were defined by the putative hydrophobic anchors based on sequence alignments (Supporting Information Fig. S4). For the TRTK12 peptide, they were defined by the residues that appeared to significantly weaken the dissociation constant on alanine mutagenesis (i.e., I5, I10, and L11). The best 200 structures from each run were clustered and scored according to the HADDOCK algorithm using the default options.^93,94 All 200 structures were further evaluated based on their ability to recapitulate the experimentally-obtained ΔC_p values according to Eq. (2). This is, to the best of our knowledge, the first case when structural thermodynamic information has been included into refinement. The 10 best structures were chosen and further analyzed based on their ability to recapitulate the observed chemical-shift perturbations, the position of the putative hydrophobic anchors, and, for the CaM-TRTK12 complexes, the relative exposure of the peptide's tryptophan (W7). A representative structure for the binding of each TRTK12 peptide to individual domains of CaM was selected for illustrative purposes.

Glossary

CaM

human calmodulin

EF-hand

calcium-binding domain consisting of a helix-loop-helix structure

Fmoc

fluorenylmethyloxycarbonyl chloride

ITC

isothermal titration calorimetry

TCEP

tris(2-carboxyethyl)phosphine; HSQC, Heteronuclear Single Quantum Coherence, MLC, C-terminal fragment of melittin.

Supporting Information

Additional Supporting Information may be found in the online version of this article.

Supplementary Information