Direct single-molecule observation of calcium-dependent misfolding in human neuronal calcium sensor-1

Significance

Protein misfolding can lead to neurodegeneration. Yet, the mechanistic details of this deleterious phenomenon are largely unknown, as experimental portrayals of the early and reversible molecular events leading to misfolded conformations have so far remained highly limited. Here we use single-molecule optical tweezers to monitor the structural rearrangements leading to misfolded conformations of human neuronal calcium sensor-1, a protein linked to serious neurological disorders. We identified two distinct and calcium-dependent misfolding trajectories originating from an on-pathway folding intermediate. Remarkably for a calcium sensor, pathologically high calcium concentrations impede correct folding by increasing the occupation probabilities of the misfolded states. The results open ostensible links between protein misfolding and calcium dysregulation that could be important in neurodegeneration and its potential inhibition.

Abstract

Neurodegenerative disorders are strongly linked to protein misfolding, and crucial to their explication is a detailed understanding of the underlying structural rearrangements and pathways that govern the formation of misfolded states. Here we use single-molecule optical tweezers to monitor misfolding reactions of the human neuronal calcium sensor-1, a multispecific EF-hand protein involved in neurotransmitter release and linked to severe neurological diseases. We directly observed two misfolding trajectories leading to distinct kinetically trapped misfolded conformations. Both trajectories originate from an on-pathway intermediate state and compete with native folding in a calcium-dependent manner. The relative probability of the different trajectories could be affected by modulating the relaxation rate of applied force, demonstrating an unprecedented real-time control over the free-energy landscape of a protein. Constant-force experiments in combination with hidden Markov analysis revealed the free-energy landscape of the misfolding transitions under both physiological and pathological calcium concentrations. Remarkably for a calcium sensor, we found that higher calcium concentrations increased the lifetimes of the misfolded conformations, slowing productive folding to the native state. We propose a rugged, multidimensional energy landscape for neuronal calcium sensor-1 and speculate on a direct link between protein misfolding and calcium dysregulation that could play a role in neurodegeneration.

Most proteins have evolved to fold rapidly into a specific and functional 3D structure immediately after translation from the ribosome. The folding process is, however, not adequately efficient to prevent the occurrence of misfolded states in vivo (1), especially in the case of larger multidomain proteins which comprise roughly 75% of the human proteome (2, 3). Normally, to tackle and destroy these unproductive structures, cells are equipped with competent clean-up machinery, such as chaperones, proteasomes, and unfoldases (4). If misfolding cannot be ameliorated, these nonnative states accumulate in the cell to form aggregates with potential pathophysiological consequences (5).

The emerging view that protein misfolding is a common phenomenon in living cells is still largely unsubstantiated, as detecting and characterizing misfolded states has been experimentally challenging (2, 6). The mechanistic details that have accumulated over the last decades on misfolding have mostly come from studies on the resulting oligomeric structures and amyloid formation (1), whereas our understanding of the structural rearrangements and pathways leading to precursory misfolded states is still highly incomplete. Importantly, the formation of prefibrillar monomeric and oligomeric misfolded states is, contrary to amyloids, reversible and thus these states provide a potential target for drug design.

Sparse populations and their associated weak signals limit the use of traditional bulk methods for monitoring the early events of misfolding, and relatively few systems have been studied in detail (7–13). Now, single-molecule force spectroscopy techniques, such as optical tweezers, enable detection of rare alternative folding pathways and short-lived misfolded states by direct mechanical manipulation (14–19). Although aggregation requires more than one molecule, nonnative structural rearrangements within a single molecule only report on monomeric misfolded states. Recent works have exploited these properties to study misfolding of well-known disease-related proteins, such as the prion protein, as well as proteins not generally associated with misfolding, such as the EF-hand calcium sensor calmodulin (CaM) (20–22).

The EF-hand superfamily of calcium sensors is responsible for translating changing levels of intracellular Ca²⁺ concentration into a biochemical signal through conformational changes that allow them to interact with an array of binding targets (23). The subfamily of neuronal calcium sensors (NCS) is mostly expressed in neurons and currently includes 15 members (24, 25). Neuronal calcium sensor-1 (NCS-1) is the most ancient member of this family (Fig. 1A), and it has been functionally associated with cognitive processes, such as learning and memory (26, 27), and with a number of cellular processes such as neurotransmitter release (28, 29), and regulation of ion channels, and G protein coupled receptors (GPCRs) (24, 30), including the dopamine receptor D2 (31). NCS-1 has also been linked to serious neurodegenerative disorders including schizophrenia, bipolar disorder (BD) (32), and autism (33, 34). However, the dysfunctions of NCS-1 are poorly characterized on the molecular level, and whether they involve altered functional profiles or loss of function due to formation of misfolded states is currently unknown.

Fig. 1.

Misfolding pathways of NCS-1. (A) The NMR structure of NCS-1 (PDB 2LCP), with the N domain (EF1/EF2) depicted in gray and the C domain (EF3/EF4) in blue. Black spheres represent Ca²⁺ ions. EF1 does not bind Ca²⁺ because of a conserved cysteine-proline mutation (35). (B) Sketch of the experimental setup. NCS-1 was tethered between functionalized beads via DNA handles and stretched and relaxed by moving the pipette relative to the optical trap (16, 62). (C) Native folding pathway of NCS-1. After being mechanically stretched and unfolded (red trace), NCS-1 refolds upon relaxation of the applied force into its native state via two intermediate states, I1 and I2. Dashed lines are worm-like-chain fits to the data. Color-coded arrows indicate the pulling/relaxing directions. (D) Misfolding pathways of NCS-1. During refolding, NCS-1 sometimes follows alternative pathways leading to misfolded states M1 (blue) or M2 (green), which are less compact than the native state (red). Dashed lines are worm-like-chain fits to the data. (E) Rescue of the native state of NCS-1. During relaxation (black), the molecule misfolded. During stretching (red), nonnative contacts of the misfolded conformation are progressively broken, until the molecule can find its native folding pathway (“rescue transition”). (F) Fraction of folding pathways leading to either M1 or M2 as a function of Ca²⁺ concentration and relaxation speed. Higher Ca²⁺ concentrations and relaxation speeds facilitate NCS-1 misfolding. Error bars indicate SEs of mean. At least five different molecules were used for each calcium concentration.

Because only a few systems have been studied experimentally, little is known about folding and/or misfolding mechanisms of members of the EF-hand superfamily (21, 35–37). The extensively studied CaM has been shown on the single-molecule level to frequently visit misfolded states that slow down the overall folding rate of the protein (21). The physiological consequences of CaM misfolding have not yet been explored. NCS-1 shares modest sequence homology with CaM, mostly within and around the calcium binding sites (24). Similar to CaM, NCS-1 contains four EF hands organized in two EF domains (Fig. 1A) yet it exhibits a larger number of interdomain contacts (38), a feature that has been suggested to increase the probability of misfolding in proteins (2). The formation of misfolded states along the folding pathway of NCS-1 may have important consequences with regards to its function as a calcium sensor and might also play a role in disease pathologies.

Using optical tweezers, we have recently characterized the native folding pathway of NCS-1 (39). Here we use a similar experimental approach to monitor individual NCS-1 molecules as they populate nonnative misfolded states in real time. We identified two misfolding trajectories leading to two distinct misfolded conformations, characterized by different extensions and different pathways on the energy landscape. Both misfolding pathways originated from a partially folded on-pathway intermediate state, and they competed with native folding. The occupancy probability of both misfolded states could be controlled by modulating either the relaxation rate of the applied force and/or the calcium concentration. Remarkably for a calcium sensor, higher calcium concentrations, even within physiologically relevant conditions, lead to an increased probability of NCS-1 misfolding.

Results

Individual NCS-1 molecules were manipulated with polystyrene beads by means of DNA molecular handles covalently attached to cysteine residues engineered at positions 4 and 188, as previously reported (39) (Fig. 1B and Methods). To identify and characterize misfolding pathways of NCS-1, we first performed constant-velocity experiments, where the molecule is stretched and relaxed by moving the pipette at constant speed relative to the optical trap (16, 39, 40). Under these experimental conditions, we have previously shown that NCS-1 unfolds in an apparent three-state manner and folds back into its native state through a process that starts with the folding of the C domain, undergoing a major conformational change (U→I2) followed by a minor rearrangement (I2→I1), and ends with the folding of the N domain at lower forces (I1→N transition) (Fig. 1C). The I2→I1 transition is the rate-limiting step of the overall folding process and is mandatory for the subsequent folding of the N domain. Here we show that, in addition to the pathway leading to the native state, NCS-1 folding can follow alternative pathways leading to nonnative (misfolded) structures (Fig. 1D). In these cases, the protein folds into I2 but fails to transit into I1, thus never reaching the natively structured C domain. At lower forces, the molecule starts fluctuating between I2 and misfolded conformations, before being trapped in either of two main misfolded states, M1 and M2, differentiated by their extension (Fig. 1D, green and blue traces). A fit of the force vs. extension traces to the worm-like-chain (WLC) model of polymer elasticity (41) yielded a contour length change (ΔL_c) of 29 ± 2 nm (n = 35; errors are given as SD) for the I2−M1 transition, and ΔL_c = 18 ± 3 nm (n = 31) for the I2−M2 transition. These states are clearly different from the native state as the native I2−I1−N transition has an overall associated contour length change of 34 ± 2 nm (39). Furthermore, M1 and M2 displayed drastically reduced mechanical stability compared with the native state and unfolded at lower forces (Fig. S1). During stretching, kinetically trapped misfolded molecules were occasionally (<1%) observed to snap into their native state through a “rescue transition” that suddenly shortened the extension of the molecule (Fig. 1E, red trace). The rescue transition occurred exclusively from I2, not from M1 or M2, and usually nonnative interactions had to be mechanically broken in an unfolding step before the molecule could fold into N, supporting the off-pathway nature of the kinetically trapped conformations. Upon further increase of force, the molecule displayed the signature of high-force unfolding transitions of both the N and C domains, verifying a native fold.

To examine any possible effects of Ca²⁺ concentration and relaxation speed on the misfolding behavior of NCS-1, we performed experiments in 0.5 µM, 10 µM, 1 mM, and 10 mM CaCl₂, at various relaxation speeds. At all calcium concentrations, the misfolding probability of NCS-1 increased with increasing relaxation speed (Fig. 1F). Strikingly, the misfolding probability also increased with increased Ca²⁺ concentration. At a physiologically relevant calcium concentration (0.5 µM) and at slow relaxation speed (5 nm⋅s⁻¹), only 5% of the molecules misfolded. When the calcium concentration was raised to 10 mM, almost 50% of the molecules misfolded. These data are highly interesting, as various physiological or pathophysiological phenomena can cause a transient or sustained increase in neuronal calcium concentrations (42–44), as further addressed in Discussion. The two misfolded states, M1 and M2, were also diversely populated at different Ca²⁺ concentrations (Fig. S2). The M1 state was highly populated at 10 mM Ca²⁺ and less populated at 1 mM Ca²⁺, whereas it was absent at lower Ca²⁺ concentrations. The M2 state instead was equally populated under all four experimental conditions. Rescuing transitions were observed at all calcium concentrations, and their frequency increased with lower calcium concentrations (∼1% at 10 mM to ∼5% at 10 µM).

To characterize the kinetics and thermodynamics of the misfolded states, we performed constant-force measurements, both at high (10 mM) and low (10 µM) calcium concentrations. Under both conditions, the calcium concentration is above bulk dissociation constants (4−400 nM) (35), suggesting the calcium binding sites of NCS-1 to be saturated. In these experiments, a molecule was stretched and relaxed multiple times until it was observed to populate a pathway leading to a misfolded state (Fig. 2A). At this point, the relaxation of the molecule was stopped and the applied force was clamped via a force-feedback mechanism to a specific tension to observe NCS-1 fluctuate at equilibrium between I2 and misfolded conformations, before eventually folding into N (typically after 20–100 s) (Fig. 2B). In all cases (>300 events), the transition to N took place from I2, providing additional evidence for the off-pathway nature of the misfolded states. After the molecule had folded into the native state, it did not unfold again within the measuring time (>5 min). The probability density distribution of the extension signal revealed that in 10 mM Ca²⁺, NCS-1 primarily populates two misfolded states positioned at 3.2 ± 2.4 nm and 7.6 ± 1.1 nm from I2 (Fig. 2C). These extensions compare well to those of the misfolded states M2 and M1 observed in constant-velocity measurements (Fig. 1D), suggesting that the most probable nonnative conformations populated by NCS-1 at equilibrium are also the conformations in which it remains kinetically trapped at lower forces. The kinetic network of the on- and off-pathway states of NCS-1 is shown in Fig. 2D.

Fig. 2.

Misfolding of NCS-1 at equilibrium. (A) Force vs. extension cycle showing a mechanically denatured NCS-1 molecule (red trace) that, during relaxation of the applied force, follows a misfolding pathway (black trace). Relaxation was stopped at 6.2 pN, and (B) the force was clamped via a feedback mechanism to observe NCS-1 fluctuate at equilibrium between I2 and misfolded states M1 and M2, until it eventually folded into its native state. (Inset) A magnified view of the data in the blue box. (C) Probability density distribution of the extension values of B, resolving misfolded states M2 and M1. The distribution was fit by the sum of four Gaussian functions (blue line) to show that M2 and M1 are located at 3.2 ± 2.4 nm and 7.6 ± 1.1 nm, respectively, from I2. (D) Diagram of the different molecular states populated during the folding of NCS-1. Fully folded EF hands are shown as gray or blue ellipses, for the N or C domain, respectively, and Ca²⁺ ions are shown as green circles.

To characterize the energies and transition kinetics of M1 and M2, extension-time traces were recorded at different forces, in the range between 5.4 pN and 7.0 pN. A population shift from I2 to the misfolded states was observed when the applied force was reduced (Fig. 3A). The force-dependent rates of the I2−M2 and I2−M1 transitions were subsequently determined by analyzing each extension trace with a hidden Markov model (HMM) algorithm (45–47). Experimental data are well modeled by a four-state Markov system where M1 and M2 are only connected to I2, while the transition from I2 to N (in our force range, the transient occupation of state I1 is too short-lived to be detected) is only one way. The position of the transition states along the reaction coordinate and the zero force rates were then estimated by fitting the data to a linearized form of the Bell model (39, 48) (Fig. 3 B and C). From the position of the transition states, we deduced a difference in extension of 4.9 ± 1.3 nm between I2 and M2, and of 8.4 ± 1.3 nm between I2 and M1 (Table 1). These distances compare well with the observed jumps of the extension in the recorded traces (Fig. 2B and Fig. S3), showing the overall consistency of our HMM reconstruction. The dwell time distributions of I2, M1, and M2 determined from the HMM analysis could be well fit to single exponentials, which is compatible with a two-state behavior of the transitions among these states (Fig. S3). From the rates, the salient features of the energy landscapes for the I2−M1 and I2−M2 transitions were reconstructed (Fig. 3D and Table 1). The height of the free-energy barriers depends on a preexponential factor, which we cannot measure directly with our instrument. We adopted the value 1.2 × 10⁴ Hz that has been recently measured with a similar setup (49).

Fig. 3.

Energetics and kinetics of NCS-1 misfolding in 10 mM Ca²⁺. (A) Extension vs. time traces of NCS-1 fluctuations at indicated forces. As the force is lowered, the M1 and M2 states are increasingly populated. (Inset) A magnified view of the misfolding transitions in the blue box, and the corresponding fit based on the HMM (red line). B and C show unfolding (red) and refolding (blue) rate constants at different forces for the I2−M1 and I2−M2 transitions, respectively. Solid lines are fits with a linearized form of the Bell model (40, 48). Error bars represent SDs. (D) Sketch of the free-energy landscapes for the I2−M2 (green) and I2−M1 (blue) transitions at 5.5 pN and 10 mM Ca²⁺, reconstructed using the HMM analysis. The I2 state was considered the reference state at zero energy.

Table 1.

Kinetics and thermodynamic parameters calculated from the force dependence of transition rates at 10 mM and 10 µM Ca²⁺ concentration

	10 mM Ca2+		10 µM Ca2+
Transition	x‡, nm	ΔG‡, kBT	x‡, nm	ΔG‡, kBT
M2→I2	3.1 ± 0.6	6.1 ± 1.2	1.2 ± 0.4	6.8 ± 0.7
I2→M2	1.8 ± 1.2	7.4 ± 2.3	3.0 ± 0.8	9.5 ± 1.4
xtotal, nm	4.9 ± 1.3		4.2 ± 1
M1→I2	1.9 ± 0.6	7.1 ± 1.2	—	—
I2→M1	6.6 ± 1.2	8.0 ± 2.3	—	—
xtotal, nm	8.5 ± 1.3		—

Here x^‡ is the distance to the transition state for the unfolding and refolding reaction, and x_total is the distance between two states. ΔG^‡ is the activation energy barrier for the different transitions at 5.5 pN calculated from the rates, using a preexponential factor of 1.2 × 10⁴ Hz. Errors are estimated from fit parameters’ uncertainties; k_BT = 2.5 kJ mol⁻¹.

In constant-force experiments and low calcium concentration (10 µM Ca²⁺), in keeping with what was observed in constant-velocity experiments (Fig. S2), NCS-1 populated only the misfolded state M2 (Fig. 4A), although very rare excursions to a state resembling M1 has been observed. Here the best reconstruction was obtained by positing a three-state Markov model with forbidden transitions between M2 and N. Our HMM analysis (Fig. 4B) yielded distances to the transition state for the I2−M2 transition that sum to 4.2 ± 0.8 nm (Table 1), in good agreement with the observed jumps of the extension in the recorded traces (Fig. 4C and Figs. S4 and S5). From the rate constants analysis, the energy landscape governing the I2−M2 transition was reconstructed (Fig. 4D), showing that at low calcium, the I2 to M2 transition has an activation energy quite similar to that observed at 10 mM Ca²⁺. These data are consistent with the weak Ca²⁺ dependence of the occupancy probability of M2 observed in constant-velocity experiments (Fig. S2).

Fig. 4.

Energetics and kinetics of NCS-1 misfolding in 10 µM Ca²⁺. (A) Extension vs. time traces of NCS-1 fluctuations at three different forces. An increase in force leads to a decrease in the occupation probability of M2. The expected extension of the M1 state is also indicated. (Inset) A magnified view of the misfolding transitions in the blue box and the corresponding fit based on the HMM (red line). (B) Unfolding (red) and refolding (blue) rate constants at different forces for the I2−M2 transition as determined from the HMM analysis, and fits with the Bell model (solid lines). Error bars represent SDs. (C) Probability density distribution of the extension values of the 3.7 pN trace of A. The distribution was fit to a double-Gaussian function showing that M2 is located at 4.0 ± 1.7 nm from I2. (D) Sketch of the free-energy landscape for I2−M2 transition at 5.5 pN and 10 µM Ca²⁺, reconstructed using the HMM analysis. The I2 state was considered the reference state at zero energy.

The impact of varying the calcium concentration on the extent of misfolding suggests that one or more of the three calcium-binding sites plays a key role in the formation of misfolded states. We therefore separately introduced classical mutations that prevent ion binding into each of the three calcium-binding EF-hand loops (39). Because folding reaches I2 before any misfolding occurs and because previous manipulation of the EF3 binding site disables folding, with no apparent native transitions and with similar behavior to the apo form of NCS-1 (39), this rules out EF3 as the origin for generation of misfolded states. In contrast, variants carrying disabled EF2 (NCS-1^EF2) or EF4 (NCS-1^EF4) binding sites still retain some native and nonnative folding transitions (39). Importantly, NCS-1^EF2 never populates M1 or M2, as the C domain always folds properly into its native structure, even at high Ca²⁺ concentrations and relaxation speeds (Fig. S6). This behavior is strikingly different from that of the wild-type NCS-1 that, under the same experimental conditions, fails to natively fold its C domain, and thus misfolds, in ∼80% of the cases (Fig. 1E). In contrast to NCS-1^EF2, NCS-1^EF4 never transits from I2 to I1, because this transition involves binding of Ca²⁺ to EF4 (39), and thus its C domain never reaches its native conformation. Nonetheless and surprisingly, with NCS-1^EF4, we could not detect any transitions from I2 to conformations resembling M1 or M2 (Fig. S6). Instead, at low forces, this variant folded into a structure with an extension significantly larger than those of the misfolded states populated by the wild-type NCS-1. These results collectively allowed us to conclude that both the EF2 and EF4 calcium binding sites are critical in the formation of misfolded states, as highlighted in Discussion.

Finally, once the molecule had populated a pathway leading to misfolded states, we analyzed the overall folding rate of NCS-1 as a function of force. At both high (Fig. 5A) and low (Fig. 5B) calcium concentrations, the probability of observing native folding decreased at lower forces, displaying an essentially linear and positive correlation with force. These results seem puzzling because we would expect lower forces to favor the native state. In fact, the rate for the transition I2−I1 as estimated by HMM analysis increases as the applied tension is reduced (Fig. 5 C and D), but such an effect is outweighed by the increased lifetime of the misfolded states, as shown in Figs. 3 B and C and 4B.

Fig. 5.

Occupation of misfolded states slows the overall folding rate of NCS-1. In A and B, τ is the total time elapsed from the beginning of the force clamp until the protein eventually folds into N (this takes into account also the time spent visiting misfolded states). In C and D, we plot the force dependence of the rate constants for the I2−I1 transition in 10 mM and 10 μM CaCl₂, respectively. The solid line represents fits to the data according to the Bell model. The extrapolated zero-force folding rate at 10 mM CaCl₂ is 40 ± 4 s⁻¹, in remarkable agreement with the value of 38 ± 5 s⁻¹ that we found performing experiments at higher force (around 10 pN) (39). The folding rate at 10 µM is lowered by a factor of ∼40 (1.1 ± 1.6 s⁻¹), whereas the distance to the transition state is similar at both calcium concentrations (x_f = 3.3 ± 1.0 nm at 10 mM CaCl₂; x_f = 2.9 ± 0.4 nm at 10 µM CaCl₂).

Discussion

The application of single-molecule techniques has lately proven to be a powerful approach to detect and characterize protein misfolded states (3, 20, 22, 50), and to investigate their link to neurodegenerative diseases (1, 51). Here we have used optical tweezers to study the molecular rearrangements leading to misfolded conformations of the NCS-1 protein. We have shown that the misfolding behavior of NCS-1 can be controlled by modulating the relaxation rate of the applied force, a phenomenon previously observed with RNA hairpins (52) but not, to our knowledge, with proteins (16, 21, 40, 53). Precision control over protein folding energy landscapes may find practical applications in the study of molecules that have established links to misfolding diseases. The fate of a protein during folding strictly depends on the topography of its energy landscape. Our results suggest a complicated and rugged energy landscape for NCS-1, with several local maxima and minima that can act as kinetic traps. During folding, nonnative contacts can become accessible before native contacts form (54). If the force is decreased faster than the rate at which nonnative contacts can break, the molecule becomes kinetically trapped in misfolded structures. Mechanical force can then be used to pull out misfolded species from their energy minima and rescue native folding.

Our results add a new layer of complexity to the folding process of NCS-1, revealing a rugged multidimensional folding energy landscape governed by a kinetic partitioning mechanism. The process invariably starts with the folding of the polypeptide chain into the intermediate state I2. From here, it can proceed toward the native state by transiting into intermediate state I1 or, in a calcium-dependent manner, take alternative pathways leading to misfolded states M1 and M2. Two distinct misfolded states have previously been observed in the structurally related two-domain EF-hand protein CaM (21), yet those states appear to be different from the ones observed for NCS-1. First, the modulation of misfolding by calcium has, to the best of our knowledge, not yet been reported for CaM, or for any other protein. Second, for both misfolded states of NCS-1, the C domain is only partially folded, whereas both CaM misfolded states consist of EF hands that are either unfolded or fully folded, although incorrectly paired. The difference in the misfolding mechanisms may be a reflection of the different structural architectures of the two proteins. CaM has a rather symmetrical structure, with a long α-helix separating the two almost identical N and C domains, each binding two Ca²⁺ ions (55). In contrast, NCS-1 is characterized by a certain asymmetry, as the N domain binds only one Ca²⁺ ion, and by significantly more interdomain contacts (38). Despite these differences, the occupation of misfolded states has been shown to slow down the overall folding rates of both NCS-1 and CaM, with a potential impact on their functions (21). Our constant-velocity and constant-force data suggest partial or near-complete folding of the N domain in misfolded states M2 and M1, respectively, and indicate that both the EF2 and EF4 Ca²⁺-binding sites play important roles in misfolding. The ability of EF2 and EF4 to bind calcium appears to be a prerequisite for wild-type NCS-1 to populate M1 and M2. If either of the two EF hands is disabled, NCS-1 behaves differently. Subsequent to calcium binding to EF3, the EF2 and EF4 sites may compete both for calcium and for successive interaction with a folded EF3. This slows productive folding because native folding requires EF3 to interact with a folded EF4 and hence any nonnative EF2−EF3 contacts need consequently to be broken. We have previously shown how the crosstalk between different EF hands is important for the folding process of NCS-1 (39). Here we report data that highlight the importance of EF-hand crosstalk for the misfolding of NCS-1.

We observed NCS-1 misfolding even at low calcium levels, suggesting a physiological relevance. Intracellular Ca²⁺ levels are tightly regulated in neurons, maintained at 40–100 nM at resting levels and rising to about 1–10 µM under excited conditions (42). However, even under normal conditions, Ca²⁺ concentrations can transiently (seconds to minutes) reach hundreds of micromolars, especially in the vicinity of Ca²⁺ channels (42, 43). When calcium homeostasis fails, the resulting and sustained high concentrations of cytosolic Ca²⁺ can have severe consequences for neurons, including deleterious effects on synaptic function and eventually cell death (44). A hypothesis gaining increasing support in recent years postulates that aging is a result of Ca²⁺ dysregulation (56, 57). Many factors in aging neurons, such as increased release of Ca²⁺ from intracellular stores and increased Ca²⁺ influx through L-type voltage-gated calcium channels, lead to sustained elevation of free Ca²⁺ concentration (56, 58). Moreover, calcium dysregulation has been implicated in the development of major neural disorders such as Alzheimer's disease, BD, and schizophrenia, and drugs that reduce Ca²⁺ signaling activity have in some cases proved successful in alleviating symptoms (59, 60). Interestingly, NCS-1 is highly up-regulated in schizophrenic and BD patients, and substantial evidence suggests that resting and activated levels of Ca²⁺ are elevated in BD (61). Whether NCS-1 is up-regulated to counteract loss of function as a consequence of Ca²⁺-induced increased misfolding is unknown but might be a highly interesting avenue of research.

Despite both Ca²⁺ homeostasis and protein misfolding being linked to neurodegeneration (5, 59), a direct causation between the two former has been missing. Here, using single-molecule optical tweezers, a putative link has been revealed between Ca²⁺ dysregulation, misfolding, and an NCS protein involved in neurodegenerative disorders. Thus, Ca²⁺ may not only shape the function of the NCS family but also cause its dysfunction in vivo.

Methods

The engineering, expression, and purification of the double-cysteine variants of unmyristoylated human NCS-1, the preparation of protein−DNA chimeric constructs, and the coupling of protein−DNA chimeras to polystyrene beads were performed exactly as described (39, 62). All experiments were performed using a custom-built optical tweezers instrument with a dual-beam laser trap of 840-nm wavelength. Measurements were conducted at ambient temperatures in 10 mM Tris, 250 mM NaCl at pH 7.0 with varying concentrations of CaCl₂ in the range 0.5 µM to 10 mM (SI Methods).

Changes in contour length associated with (un)folding transitions were estimated by fitting constant-velocity traces to the WLC model, as previously described (16, 39). The thermodynamics and kinetics of the transitions between molecular states I2, M1, and M2 in constant-force measurements were characterized using a HMM algorithm, as previously described (39). The force dependence of the transition rates was analyzed with the phenomenological Bell model (63) to estimate the position of the barriers and the barriers’ heights, using a preexponential factor of 1.2 × 10⁻⁴ Hz (39, 49). A total of 84 and 62 extension vs. time traces, from roughly 20 and 10 molecules, were used in our energy landscape reconstruction at 10 mM and 10 μM Ca²⁺, respectively (SI Methods).