Report

Ligand-Dependent Equilibrium Fluctuations of Single Calmodulin Molecules

See allHide authors and affiliations

Science  30 Jan 2009:
Vol. 323, Issue 5914, pp. 633-637
DOI: 10.1126/science.1166191

Abstract

Single-molecule force spectroscopy allows superb mechanical control of protein conformation. We used a custom-built low-drift atomic force microscope to observe mechanically induced conformational equilibrium fluctuations of single molecules of the eukaryotic calcium-dependent signal transducer calmodulin (CaM). From this data, the ligand dependence of the full energy landscape can be reconstructed. We find that calcium ions affect the folding kinetics of the individual CaM domains, whereas target peptides stabilize the already folded structure. Single-molecule data of full length CaM reveal that a wasp venom peptide binds noncooperatively to CaM with 2:1 stoichiometry, whereas a target enzyme peptide binds cooperatively with 1:1 stoichiometry. If mechanical load is applied directly to the target peptide, real-time binding/unbinding transitions can be observed.

Single-molecule mechanical methods have made it possible to study and control biomolecular conformations with unprecedented precision (16). Whereas classical methods such as changes in temperature or chemical environment act globally and rather unspecifically on the energy landscape, forces can be applied locally to precisely manipulate selected structural elements of a protein and thus explore protein energy landscapes in a controlled manner (5). Due to limited resolution, single-molecule mechanical studies of protein folding and protein/ligand interactions have so far almost exclusively been carried out in nonequilibrium and have thus been restricted to the study of unfolding and unbinding reactions (79). Hence, half of the energy landscape (the part that controls re-folding or ligand re-binding) has been mostly inaccessible to these experiments. We have used a low-drift atomic force microscope (AFM) (10) to observe folding/unfolding fluctuations of single calmodulin proteins (CaM) under equilibrium conditions in the presence of Ca2+ and target peptides. Ligand-dependent folding/unfolding fluctuations, as well as ligand binding/unbinding fluctuations, can be induced mechanically and observed in real time.

CaM is the most prominent Ca2+ sensor in eukaryotic cells (11). When Ca2+ binds to CaM, the flexible Ca2+-free apo conformation converts into a rigid and stable holo conformation (12). Holo CaM binds to specific target sequences in downstream regulatory proteins, thus altering their signaling properties. To date, more than one hundred target proteins for CaM have been described (11). Little is known about the folding of CaM, but temperature-jump experiments have shown that folding of apo CaM occurs on the submillisecond time scale (13).

The experimental scheme, including a sketch of Ca2+-CaM, is shown in Fig. 1A. Following previously described methods, we have sandwiched a single CaM molecule between immunoglobulin domains from Dictyostelium discoideum filamin that serve as attachment points for the AFM tip and the surface. CaM consists of two structurally similar domains, a ∼75-residue N-terminal (DomN, red) and a ∼70-residue C-terminal (DomC, blue) globular domain, that each bind two Ca2+ ions. A typical force extension trace of CaM at a pulling velocity of 1 nm/s and 10 mM Ca2+ is shown in Fig. 1B. At pulling velocities generally used in force spectroscopy of ∼1 μm/s, unfolding events of CaM are buried in the thermal noise (14). Low instrumental drift allows us to achieve the slow pulling velocity (1 nm/s) necessary for the required force resolution of ∼2 pN (15). Two distinct peaks (marked in red and blue, respectively) correspond to the unfolding of the two globular CaM domains. To allow for a structural interpretation of the unfolding traces, we introduced a disulfide bond between residues 128 and 144 of DomC to shorten the extensible backbone of this domain (CaM 128×144). Ligand binding properties of the shortened version were indistinguishable from the wild type. The length gain upon unfolding of the domains can be measured using the wormlike chain model of polymer elasticity (16) (black solid lines in Fig. 1B, main graph). The average values are ΔLI = 25.7 ± 0.4 nm and ΔLII = 17.6 ± 0.7 nm. This result suggests that the red unfolding peak marks unfolding of DomN (expected value ΔL = 25.0 nm), whereas in the second peak DomC unfolds (expected value ΔL = 18.7 nm) (17). Closer inspection of the unfolding peaks reveals rapid transitions between the folded and unfolded states in both unfolding peaks. The insets above the peaks in Fig. 1B show time traces of the transition regions. Such transitions are the hallmark of thermodynamic equilibrium, and a kinetic analysis now allows extracting both equilibrium and nonequilibrium parameters and, hence, the energy landscape of CaM folding from a single molecule.

Fig. 1.

(A) Sketch of the experimental setup, showing CaM attached to AFM cantilever tip and surface by means of filamin domains that serve as handles (not to scale). (B) Sample trace of CaM 128×144 at 10 mM Ca2+. The folding/unfolding transitions of DomN are shown in red, and those of DomC are shown in blue. Wormlike chain fit curves (black traces) allow determining the contour length increase upon unfolding ΔL and thus identifying the individual domains. In the insets above the peaks, time traces of the transition region are depicted. (C) Time traces of the transition region of isolated DomC at 10 mM Ca2+. (D) Potential energy landscape for the folding of DomC at 10 mM Ca2+. N, native state; TS, transition state; U, unfolded state.

In agreement with experiments using single-domain CaM constructs, our results suggest that the two domains fold and unfold independently from each other (18). Therefore, we investigated single-domain constructs of both DomN and DomC. A sample time trace of the unfolding fluctuations of isolated DomC is shown in Fig. 1C. Due to the large spring constant of the AFM probe (6 pN/nm), the fluctuations do not occur at constant load. Unfolding occurs at high loads (12 pN), whereas refolding occurs from the relaxed state (6 pN). The slope of the broken lines marking the folded and unfolded levels, respectively, reflects the slow force ramp imposed during the experiment. To analyze the fluctuation kinetics, we define the characteristic transition rate as the averaged folding and unfolding rate in the whole transition region (DomC: 11.7/s, number of molecules analyzed n = 23). Using a Monte-Carlo simulation, we adjusted the unfolding and folding rates at zero force so that the characteristic transition rate matched our experiment [for details, see the supporting online material (SOM) text]. Independent measurements of the speed dependence of the nonequilibrium unfolding forces at higher pulling velocities yielded a distance of the transition state from the folded state of ΔxN-TS = 2 nm, which is a typical value for an all–α helical protein (SOM text and fig. S1A). Combining these results, we can now draw an energy landscape for the folding of CaM at 10 mM Ca2+ (Fig. 1D). A zero-force folding rate of ∼2 × 105/s under these conditions makes CaM one of the fastest folding proteins known to date (19).

To investigate the influence of ligand binding on the folding dynamics of DomC in greater detail, we performed concentration scans of Ca2+ and mastoparan (Mas), a wasp venom peptide known to bind to CaM (20). Figure 2A shows equilibrium time traces of the transition region at varying Mas concentrations with Ca2+ kept fixed at 10 mM. The most obvious influence of Mas binding is the decreased speed of the transition kinetics with increasing Mas concentration. We plotted the characteristic transition rate as a function of Mas concentration in Fig. 2B. The effect of Mas on the midpoint unfolding force (the average value of the unfolding forces in the transition region) is shown in the inset. As expected, higher concentrations of Mas increase the stability of DomC and, hence, its midpoint unfolding force. The effect of Mas on both force and kinetics can be readily explained by assuming that Mas binding exclusively slows the unfolding rate ku, as depicted in Fig. 2C for two sample concentrations. In the narrow force range where we observe folding/unfolding transitions, both unfolding rates (green) and folding rates (purple) will depend in good approximation exponentially on the applied force (2123). The slopes of the green and purple lines are determined by the distance of the transition state from the native and unfolded states, respectively (21, 22). At 1 μM Mas, the midpoint force of the transition region, where unfolding and folding occur at the same rates, is defined by the intersection of the thin green line and the purple line. Because Mas binding acts on the unfolding rate, increasing the concentration to 100 μM will shift the green line down (thick green line). The new intersection point now lies at an increased force as well as a reduced characteristic transition rate. A detailed modeling of ku and kf in dependence of ligand concentration and force in a Monte Carlo simulation could reproduce the full effect of Mas concentration on midpoint unfolding forces and transition rates (solid lines in Fig. 2B; for details, see SOM text). This simulation includes equilibrium constants for Mas binding to the native, unfolded, and transition state, respectively.

Fig. 2.

(A) Time traces of isolated DomC at 10 mM Ca2+ and varying Mas concentrations. (B) Characteristic transition rate and unfolding force at different Mas concentrations (blue symbols). The ligand effect can be fully reproduced by Monte Carlo simulations (black symbols and black interpolation curve). (C) (Left) Force dependence of folding rate (purple) and unfolding rate (green). An increase in Mas from 1 μM (thin green line) to 100 μM (thick green line) moves the intersection point (transition region) to higher forces and lower rates. (Right) Potential energy landscape at low (thin green line) and high (thick green line) Mas. (D) Time traces of isolated DomC at 10 μM Mas and varying Ca2+ concentrations. (E) Transition rate and unfolding force at different Ca2+ concentrations (blue symbols). Results of a Monte Carlo simulation are shown in black. (F) (Left) Increase in Ca2+ from 1 mM Ca2+ (thin purple line) to 100 mM Ca2+ (thick purple line) moves the intersection point to higher forces and higher rates. (Right) Potential energy landscape for folding at low (thin purple line) and high (thick purple line) Ca2+.

Similarly, we investigated the effect of Ca2+ on midpoint unfolding forces and kinetics (Fig. 2D). From the data traces taken at three sample concentrations, it can be readily seen that unlike Mas, Ca2+ increases the characteristic transition rate of DomC. Forces and transition rates are quantified in Fig. 2E. Despite the increase in transition rates, the midpoint unfolding force (see inset) also increases (24). Increased transition rates, together with increased stability, can only be explained by Ca2+ raising predominantly the folding rate (see the scheme in Fig. 2F). Increasing the Ca2+ concentration from 1 mM to 100 mM will shift up the folding rate (purple lines) but leave the unfolding rate unaffected. The new intersection point lies at both higher forces and higher rates. Again, a Monte Carlo simulation incorporating the Ca2+ dependence of ku and kf can fully explain the data (solid lines in Fig. 2E; for details, see SOM text).

Thus, the two CaM ligands, Ca2+ and Mas, act on folding in different ways (see energy landscapes in Fig. 2, C and F). Mas stabilizes the already folded form but does not interact with the transition state structure or the unfolded protein because it leaves the folding rate unaffected. Hence, the transition state has only negligible affinity to Mas (Fig. 2C) (8). In contrast, at the concentrations measured (100 μM to 100 mM), Ca2+ predominantly acts on the folding rate. Hence, Ca2+ binding stabilizes both the transition state and the folded state, indicating a dissociation constant of Ca2+ binding to the transition state of ∼1 mM (see energy landscape in Fig. 2F and SOM text). At very low Ca2+, CaM folds in to the flexible and highly dynamic apo structure that undergoes a large conformational change upon addition of Ca2+ (12). Higher Ca2+ opens a new folding pathway directly to the holo form via a transition state that has already bound Ca2+ ions. At high Ca2+, this pathway prevails over the apo folding route (see scheme in fig. S3 and SOM text). Along this holo folding route, the Ca2+ binding sites might act as nucleation seeds for folding because Ca2+ binding appears to occur early on the folding pathway.

Our mechanical folding/unfolding assay now offers the possibility to study CaM-target peptide interaction modes of full-length CaM directly on the single-molecule level. Figure 3A (upper trace) shows a mechanical unfolding curve of CaM 128×144 in the presence of 100 μM Mas. In comparison to peptide-free CaM, a mechanical stabilization of both DomN and DomC becomes obvious, which manifests itself in higher forces as well as slower transition rates (Table 1). In the full-length construct, the stabilizations of the individual domains obviously occur independently from each other because they behave identical to single-domain constructs at the same Mas concentration (Fig. 3A, lower traces, and Table 1). This result is surprising, given that Mas binding to CaM is discussed in the literature almost exclusively in the context of a binding stoichiometry of 1:1 (25). Cooperative binding of both domains simultaneously to a single Mas peptide should occur with a higher energy than the sum of the interaction energies of the individual domains. Our result indicates that full-length CaM binds Mas noncooperatively in a 1:2 stoichiometry where each peptide interacts with only one CaM domain. The 1:2 stoichiometry can be confirmed by analyzing the single-molecule thermodynamics of ligand binding to CaM (SOM text). The Mas dependence of forces and transition rates of the first unfolding peak (DomN) of CaM 128×144 is shown in Fig. 3B. A 1:1 stoichiometry is not consistent with our data (dashed black line obtained from the Monte Carlo simulation), whereas a 1:2 stoichiometry fits our data well (solid black line) (26).

Fig. 3.

(A) (Upper trace) Sample trace of CaM 128 × 144 at 10 mM Ca2+ and 100 μM Mas. DomN (red) and DomC (blue) are stabilized approximately equally by Mas. (Lower traces) Isolated DomN (red) and DomC (blue) at 100 μM Mas. (B) Unfolding forces and transition rates of DomN of CaM 128×144 at different Mas concentrations (red). Solid lines, concentration dependence in the case of a 1:2 stoichiometry; broken lines, 1:1 stoichiometry (Monte Carlo simulations). (C) (Upper trace) Sample trace of CaM 128×144 at 10 mM Ca2+ and 10 μM MLCK. DomN (red) is stabilized considerably more strongly than DomC (blue). (Lower traces) Isolated DomN (red) and DomC (blue) at 10 μM MLCK. (D) Unfolding forces and transition rates of DomN of CaM 128×144 at different MLCK concentrations (red). Solid lines, concentration dependence for a 1:2 stoichiometry; broken lines, 1:1 stoichiometry. (E) Sample trace of CaM with MLCK fused to the C terminus at 10 mM Ca2+. The protein construct is shown in the scheme. (Inset) Time traces of first unfolding peak. An intermediate state, corresponding to CaM with unbound peptide (black level), appears.

Table 1.

Midpoint unfolding forces and characteristic transition rates (10 mM Ca2+).

View this table:

Because binding of the venom peptide Mas to CaM probably has a very different physiological role than binding to target enzymes, we also investigated binding of CaM to the interaction site peptide of its target enzyme skeletal muscle myosin light chain kinase (MLCK) (27). As expected for a strongly and cooperatively binding peptide (28), the stabilization of full-length CaM upon MLCK peptide binding is dramatic at concentrations of 10 μM (Fig. 3C, upper trace, and Table 1). The unfolding peak of DomN is increased to ∼19 pN, whereas DomC is stabilized to an extent similar to Mas. For comparison, the unfolding traces of the isolated domains are shown in the lower traces of Fig. 3C. In contrast to Mas, isolated DomN seems to exhibit almost no affinity for the MLCK peptide. In agreement with the literature, we find a binding stoichiometry of 1:1 for this peptide (Fig. 3D).

The large interaction energy of CaM and MLCK peptide, together with a diffusion-limited on-rate (29), suggests that the exchange of the bound peptide from the complex may be slow. Mechanical force can be used as control parameter to affect the peptide off-rates directly and bring them into an observable regime while still leaving the structure of CaM intact. To this end, we fused the MLCK peptide directly to the C terminus of CaM (30), enabling us to apply force to the peptide directly (see sketch in Fig. 3E). Again, the observed traces exhibit near equilibrium fluctuations. In the first unfolding peak, rapid transitions between three levels can be observed: peptide bound (green), peptide unbound (black), and DomN unfolded (red) (Fig. 3E). For clarity, the inset shows time traces of those fluctuations taken from three independent molecules. This data directly reveals the thermodynamic hierarchy of the folding/peptide-binding process. The black level (CaM folded and peptide unbound) seems to be obligatory both on the folding/binding pathway and on the unfolding/unbinding pathway. Even though the stabilization due to peptide binding (∼20 kBT as estimated from the area enclosed by the green and black levels) is higher than the folding free energy of the individual domains (∼15 kBT), binding/unbinding and folding/unfolding are still separate processes.

At forces of ∼20 pN, the peptide-bound state still has a long lifetime on a time scale of seconds (green levels). Using a transition state position of 1 nm for peptide unbinding, we estimated a zero-force lifetime for the bound peptide of 102 to 103 s. Such a long lifetime may be essential for proper functioning of the CaM signaling pathway, specifically because typical free cellular Ca2+-CaM concentrations are lower than the overall concentration of target sites (31).

We demonstrated that single-molecule force spectroscopy by AFM allows online observation of the interaction dynamics of a signaling protein with its target. How CaM binding kinetics and mechanics are affected by full-length target proteins and how their activity is modulated will be an important question for future experiments. We anticipate that direct single-molecule measurements of equilibrium fluctuations, as presented in this study, will provide an important tool to measure protein-target interaction dynamics in real time.

Supporting Online Material

www.sciencemag.org/cgi/content/full/323/5914/633/DC1

Materials and Methods

SOM Text

Figs. S1 to S4

References

References and Notes

View Abstract

Navigate This Article