Consequences of Hydrophobic Nanotube Binding on the Functional

Jun 18, 2019 - Binding of two Ca2+ at each domain switches the closed conformation to .... The two-dimensional free-energy landscape along the reactio...
0 downloads 0 Views 3MB Size
Download PDF
This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes.

Article Cite This: ACS Omega 2019, 4, 10494−10501

http://pubs.acs.org/journal/acsodf

Consequences of Hydrophobic Nanotube Binding on the Functional Dynamics of Signaling Protein Calmodulin Wentao Zhu, Jianyang Kong, Jian Zhang, Jun Wang, Wenfei Li,* and Wei Wang* National Laboratory of Solid State Microstructure, and Collaborative Innovation Center of Advanced Microstructures and Department of Physics, Nanjing University, Nanjing 210093, China

Downloaded via 37.9.40.182 on August 29, 2019 at 16:45:32 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

S Supporting Information *

ABSTRACT: The wide applications of nanomaterials in industry and our daily life have raised growing concerns on their toxicity to human body. Increasing evidence links the cytotoxicity of nanoparticles to the disruption of cellular signaling pathways. Here, we report a computational study on the mechanisms of the cytotoxicity of carbon nanotubes (CNTs) by investigating the direct impacts of CNTs on the functional motions of calmodulin (CaM), which is one of the most important signaling proteins in a cell, and its signaling function relies on the Ca2+ binding-coupled conformational switching. Computational simulations with a coarse-grained model showed that binding of CNTs modifies the conformational equilibrium of CaM and induces the closed-to-open conformational transition, leading to the loss of its Ca2+-sensing ability. In addition, the binding of CNTs drastically increases the calcium affinity of CaM, which may disrupt the Ca2+ homeostasis in a cell. These results suggest that the binding of hydrophobic nanotubes not only inhibits the signaling function of CaM as a calcium sensor but also renders CaM to toxic species through sequestering Ca2+ from other competing calciumbinding proteins, suggesting a new physical mechanism of the cytotoxicity of nanoparticles.

■

INTRODUCTION

CaM is a calcium-binding signaling protein expressed in all eukaryotic cells. It has two symmetrically arranged globular domains connected by a flexible central linker21−23 (Figure 1). Each CaM domain is composed of two helix−loop−helix Ca2+-

Nanomaterials have been widely used in modern industry and biomedical fields because of their unique physical and chemical properties.1 However, the increasing environmental exposure to nanomaterials also raised public concerns regarding the possible toxicity to human body. Cumulative evidences showed that nanoscale particles may cause damages to the respiratory system, nervous system, and even cardiovascular system.2−7 Large-scale exposure may also produce physiological responses, such as inflammation, epithelioid granulomas, and fibrosis.6−9 Consequently, there is an urgent need to investigate the molecular mechanisms of the biological toxicity of nanoparticles, which is very important for designing biologically compatible nanomaterials to reduce their adverse biological effects. There are several possible ways that nanoparticles can lead to toxicity to a cell. Previous studies mainly focus on the effects of oxidative stress and the subsequent DNA damage induced by nanoparticles.3,5,6,9,10 In recent years, direct interactions between nanoparticles and biomolecules attract much attention.11−18 Both molecular dynamics (MD) simulations and spectroscopic experimental measurements suggested that the nonspecific hydrophobic interactions between nanoparticles and the hydrophobic residues of proteins tend to disrupt and block the active sites and competitively inhibit the binding of the target proteins.6,11,17−20 © 2019 American Chemical Society

Figure 1. Schematic diagram showing the steps of calcium signal transduction mediated by calmodulin (CaM). Upon Ca2+ binding, the CaM domains undergo closed-to-open conformational change, exposing their hydrophobic pocket for the binding of downstream target proteins. Here, only the helix responsible for the CaM binding is shown for the target protein. The binding of the hydrophobic nanotube can disrupt the signaling pathway, thereby leading to cytotoxicity. Received: April 28, 2019 Accepted: June 6, 2019 Published: June 18, 2019 10494

DOI: 10.1021/acsomega.9b01217 ACS Omega 2019, 4, 10494−10501

ACS Omega

Article

Figure 2. Impacts of CNT binding on the conformational motions of the CaM domain. (a,b) Representative trajectories showing the conformational change of CaMc without (a) and with (b) CNT binding. The simulations were conducted at a temperature of 300 K and [Ca2+] of 1.0 μM. (c,d) Two-dimensional free-energy landscapes along the reaction coordinates Qopen and Qclosed without (c) and with (d) CNT binding. (e,f) Two-dimensional free-energy landscapes along the reaction coordinates RMSDopen and HC (e) and along the reaction coordinates RMSDclosed and HC (f) with CNT binding. HC is the number of hydrophobic contacts between CaMc and CNT. (g) Cartoon representation of the representative conformations. State labels: UO, CNT-unbound open conformation; UC, CNT-unbound closed conformation; BO, CNT-bound open conformation; BC, CNT-bound closed conformation; BD, CNT-bound distorted conformation.

binding motifs (EF-hands). In the absence of Ca2+, the two helices of the EF-hand motif are nearly parallel to each other and the whole domain adopts a closed conformation with the hydrophobic residues deeply buried. Binding of two Ca2+ at each domain switches the closed conformation to open conformation, in which the helices of the EF-hand motif become perpendicular to each other and the hydrophobic patches are largely exposed to the solvent, facilitating the binding of downstream target proteins which also have conserved hydrophobic residues.21,24−26 Such Ca2+ bindinginduced conformational switching of CaM is the key step of Ca2+ signal transduction and has been utilized in regulating a number of biological processes, including muscle contraction, metabolism, apoptosis, immune response, and so on.27,28 Because of its critical role in regulating cell processes and wellorganized hydrophobic patches, CaM has been considered to be one of the main targets that the exogenous hydrophobic nanoparticles bind with.17 Previous all-atom MD simulations showed that a carbon nanotube (CNT) can plug into the hydrophobic patch of the CaM domains with high affinity and occupy the binding site of the target proteins.17 These previous computational and experimental works inspired us to further investigate the effects of CNT binding on the functional dynamics of CaM, including Ca2+ binding, allosteric motions, and target protein recognitions, by which we can get deeper understanding of the molecular mechanism of the nanotoxicity of hydrophobic nanoparticles. Because of the tremendous degrees of freedom involved, the timescale of conformational changes and calcium binding of

CaM is far beyond the accessible scope of conventional allatom MD simulations. To overcome such timescale difficulty, in this work, we constructed a coarse-grained (CG) model to integratedly describe the Ca2+ binding, conformational motions, target recognition, and CNT binding by MD simulations within the framework of the dynamic energy landscape theory.29 Recently, such a CG model has been successfully applied in studying the Ca2+ binding-coupled folding, allosteric motions, and target recognition of CaM by several laboratories.30−33 Our simulations revealed that the CNT tends to bind at the same binding site of the target protein and therefore can inhibit the target recognition by competitive interactions. In addition, we showed that the CNT binding modifies the conformational equilibrium of CaM, which is highly biased to the active open conformation, leading to the loss of the responding ability of Ca2+. More interestingly, the conformational switch induced by the CNT binding increases the Ca2+ binding probability by nearly six folds, which may disrupt the calcium homeostasis in a cell. These results suggest that the hydrophobic nanotube binding not only inhibits the signaling function of CaM as a calcium sensor but also renders CaM to toxic species through sequestering the calcium ions from other competing calcium-binding proteins because of the increased Ca2+ affinity.

■

RESULTS Impacts of CNT Binding on the Conformational Equilibrium of the CaM Domain. An appropriate conformational equilibrium of CaM is the key to its biological 10495

DOI: 10.1021/acsomega.9b01217 ACS Omega 2019, 4, 10494−10501

ACS Omega

Article

stage of CNT binding. Then, driven by hydrophobic interactions, the CNT occupies the hydrophobic patch of CaMc, leading to the formation of more hydrophobic contacts between the CNT and CaMc. Interestingly, CaMc with the CNT bound at its hydrophobic patch hops between the open conformation (which favors intraprotein interactions) and the distorted conformation (which favors CNT−CaMc interactions), demonstrating the competition between intraprotein interactions and CNT−CaMc interactions. Impacts of CNT Binding on the Ca2+ Affinity of the CaM Domain. Undoubtedly, the alteration of the conformational equilibrium of CaMc by CNT binding will affect its Ca2+-binding ability. Figure 3a,b shows two representative

function. Experimental measurements showed that CaM is dominated by closed conformation in the absence of Ca 2+ . 33−36 Binding of Ca2+ shifts the conformational equilibrium, leading to the dominance of active open conformation. Here, we investigated how the CNT binding affects the functional dynamics of the CaM domains. For this purpose, we conducted MD simulations for a system containing the isolated C-terminal domain of CaM (CaMc) and the CNT(5,5) with the Ca2+ concentration of 1.0 μM. The simulation was performed with the interaction strength between the CNT and CaMc optimized according to atomic MD simulations (ϵHP = 0.4 kcal/mol). As a control, we conducted additional simulations in which the attractive component of the interactions between CaMc and CNT was removed, so that the binding of CNT to CaMc is eliminated. We used the reaction coordinates RMSDopen, RMSDclosed, Qopen, and Qclosed to characterize the conformations of CaMc. RMSDopen (RMSDclosed) is the root-mean-square deviation from the open (closed) conformation. Qopen (Qclosed) is the fraction of the formed native contacts in a given structure, with the open (closed) conformation used as the reference native structure, and varies between 0 and 1. The higher value of Qopen (Qclosed) represents that the conformation of a given structure is more open (closed)-like. We can see that without CNT binding, CaMc hops between the closed (with larger RMSDopen values) and open (with smaller RMSDopen values) states frequently (Figure 2a), with the latter being more dominantly populated. In addition, CaM can become unfolded occasionally. The two-dimensional free-energy landscape along the reaction coordinates Qopen and Qclosed shows three basins, corresponding to the open (UO, see figure caption), closed (UC), and unfolded states, with the inactive closed state having the lowest free energy (Figure 2c). Such results are consistent with our previous work for the same protein.30 In comparison, for the system with CNT−CaMc interactions, the CNT tends to bind with CaMc. Once bound, the conformational fluctuation of CaMc is much suppressed, and the active open conformation becomes dominant (see Figure 2b,d). A more detailed analysis showed that upon the binding of the CNT, the closed conformation is rarely visited. Instead, CaMc hops between the active open conformation (BO) and a distorted conformation (BD). In this distorted conformation, helices are closely wrapped around the CNT, with the helical content remaining intact (Figure 2g). These results suggest that the binding of the CNT shifts the conformational equilibrium and biases the conformational distribution of the CaM domains to the open state. Apparently, the suppression of the functionally relevant conformational motions arising from the binding of the CNT can lead to cell dysfunction, contributing to the cytotoxicity of CNT. To more clearly demonstrate the conformational dynamics of CaM and the binding mechanism between the CNT and CaMc, we plotted two-dimensional free-energy landscapes along the reaction coordinates RMSDopen and HC (Figure 2e) and along the reaction coordinates RMSDclosed and HC (Figure 2f). Here, the reaction coordinate HC is the number of hydrophobic contacts between CaMc and the CNT, which monitors the extent of wrapping of the CaMc helices around the CNT. One can see that at the physiologically relevant calcium concentration (1.0 μM), the CNT binds first to the closed conformation of CaM (BC) with a small number of contacts between the CNT and CaMc. This conformation corresponds to a transient state and occurs only at the early

Figure 3. Impacts of CNT binding on the calcium-binding affinity of the isolated CaMc domain. Representative trajectories showing the binding energies of the two Ca2+-binding sites of CaMc without (a) and with (b) CNT binding. (c) Population of the Ca2+-bound state of CaMc as a function of [Ca2+] without (red) and with (blue) CNT binding. The [Ca2+] at which the population of the Ca2+-bound state equals to 0.5 can be used to characterize the calcium-binding affinity. Here, the population of the calcium-bound states was calculated without distinguishing the two sites.

trajectories monitoring the binding energies of the two Ca2+binding sites with and without CNT binding at the Ca2+ concentration of 1.0 μM. One can see that the calcium ions can bind and unbind reversibly. However, upon CNT binding, the dissociation of Ca2+ becomes much slower, leading to increased Ca 2+ dwelling time and affinity. To more quantitatively characterize such an effect, we conducted additional simulations with different Ca2+ concentrations. As shown in Figure 3c, at a wide range of Ca2+ concentrations, the binding of the CNT can significantly enhance the population and the dwelling time of the Ca2+-bound state, leading to an increased binding affinity. Particularly, at physiologically relevant Ca2+ concentrations (∼1.0 μM), the population of the Ca2+-bound state can be enhanced by approximately six folds. Such an effect of CNT binding on the Ca2+ affinity is easy to understand, as the CNT binding tends to stabilize the open conformation of CaMc, in which the residues of the binding site are more compatible with the Ca2+ coordination geometry, and therefore has higher Ca2+ affinity, compared to the closed conformation.22 As the Ca2+ concentrations in living cells are strictly regulated, the coexistence of the normal CaM and CNT-bound CaM may introduce a crosstalk between the two species. Because of the much higher Ca2+ affinity, the population of the CNT-bound CaM may destroy the Ca2+ homeostasis in a cell by making Ca2+ compete with the normal CaM and other Ca2+-binding proteins. As a result, the normal CaM may lose the ability to transduct the Ca2+ signal, leading to cell dysfunction. These results suggest that the hydrophobic 10496

DOI: 10.1021/acsomega.9b01217 ACS Omega 2019, 4, 10494−10501

ACS Omega

Article

nanotube may impose cell toxicity by two possible mechanisms at the molecular level, including (i) modifying the conformational equilibrium of the CaM domains, therefore disrupting its Ca2+-sensing ability and (ii) enhancing the calcium affinity of the CNT-bound CaM and disrupting the calcium homeostasis in the cell. Effects of the CaMc−CNT Interaction Strength. In the above discussions, the interaction strength between CaMc and CNT was optimized according to the all-atom force field, as described in the Method section, which represents the situation of the intact CNT. However, in practice, such interaction strengths can be modified by different oxidation extents or constituting elements of the hydrophobic nanotube.12 Therefore, it is interesting to investigate how the abovediscussed effect of nanotube binding on the CaM conformation and Ca2+ affinity can be modulated by interaction strengths. For this purpose, we conduct simulations at a wide range of interaction strengths (ϵHP ranges from 0.0 to 1.0 kcal/ mol) and Ca2+ concentrations. A detailed analysis of MD trajectories shows that the effects of the interaction strength can be complex and tightly coupled with the calcium concentration. With the increase of the interaction strength, the mostly populated conformation changes from the closed conformation to open conformation and distorted conformation. To more quantitatively characterize the effect of CNT binding on the conformational distribution, we calculated the population for each of the above states with different interaction strengths and Ca2+ concentrations. As expected, the CNT is more likely to be bound with CaMc with a higher CaMc−CNT interaction strength (Figure S1). With the increase of ϵHP from 0 to 0.3 kcal/mol, the population of the CNT−CaMc complex increases from 0 to 1.0. The population of the CNT−CaMc complex is less sensitive to [Ca2+]. Figure 4a−c shows the results of the populations of the open state (a), closed state (b), and distorted state (c), represented by a color map. According to the populations of these states, we can further construct the phase diagram projected onto the two collective variables represented by Ca2+ concentration and interaction strength (Figure 4d). One can see that when the interaction strength is weak and Ca2+ concentration is low, the closed state is dominant, whereas at intermediate interaction strengths (e.g., ϵHP ranges from 0.2 to 0.4 kcal/mol, which includes the interaction strength for the standard hydrophobic CNT), the open state is dominant at the whole range of Ca2+ concentrations. With a further increase of the interaction strength, the distorted state is more populated. We plotted the populations of the calcium-bound conformations on the same space (Figure 5a). When the CaMc− CNT interaction is weak, the calcium-bound state can be significantly populated only at high calcium concentrations, whereas at the range of ϵHP = 0.2−0.4 kcal/mol, the calciumbound state can be significantly populated even at low calcium concentrations, which is consistent with the results shown in Figure 3 that CNT binding enhances the calcium affinity. When the interaction strength is stronger, which corresponds to the distorted conformation, the population of the calciumbound state is reduced again. The calcium binding is probable only at very high calcium concentrations in this case. Such a reduced calcium affinity suggests that the calcium-binding pockets have been distorted because of the strong CNT− CaMc interactions.

Figure 4. Conformational distributions and phase diagram of the isolated CaMc domain. (a−c) Color maps showing the populations of CaMc with open (a), closed (b), and distorted states (c) with different [Ca2+] and interaction strengths ϵHP between CaMc and CNT. (d) Phase diagram of the CaMc conformation. The boundaries correspond to the location of the parameter values which give the equal populations of the neighboring states.

Effects of the CNT Binding on the Full-Length CaM. As CaM has two nearly symmetrically arranged domains, it is interesting to investigate the effect of CNT binding on the conformational distributions and calcium affinity of the two component domains of the intact CaM. We conducted additional simulations for the full-length CaM. Figure 5b,c shows the Ca2+-binding affinities of the CaMn and CaMc domains of the full-length CaM at a wide range of [Ca2+] and ϵHP values. The obtained two-dimensional free-energy landscapes were similar to that of the isolated CaMc (see Figure 5a) except for two details. First, at the same [Ca2+], the Ca2+binding probability of the CaMn domain was obviously lower than that of the CaMc domain, which is in accordance with the fact that CaMc has a higher Ca2+ affinity than CaMn as observed in previous experimental and computational works.30,37 Second, compared to the isolated CaMc domain, CaMc of the full-length CaM has a much higher Ca2+ binding probability event at larger ϵHP values. A more detailed analysis showed that in the full-length simulations, because of the competition between the two CaM domains for the hydrophobic CNT atoms, the interactions between the individual domain and the CNT are much reduced (Figure S1). Such an elimination of interactions makes the population of the distorted conformation less probable, leading to higher stability of the open state and increased calcium-binding affinity. We also studied the specific sequence of conformational changes of the two CaM domains starting from the closed structure in the presence of CNT. The results showed that the CaMn and CaMc domains change their conformations from the closed to open states with similar rates at a wide range of Ca2+ concentrations. Consequently, two pathways, in which either the CaMn or the CaMc domain changes the conformation first, are possible and have comparable probabilities. 10497

DOI: 10.1021/acsomega.9b01217 ACS Omega 2019, 4, 10494−10501

ACS Omega

Article

Figure 5. Impacts of the interaction strength between the CaM domains and the CNT on calcium binding. (a) Population of the Ca2+-bound state for the isolated CaMc with different [Ca2+] and interaction strengths ϵHP. (b,c) Same as (a) but for CaMn (b) and CaMc (c) of the full-length CaM. (d,e) Cartoon representation of the native structures of the full-length CaM at apo (d) and holo (e) states. (f) Cartoon representation of a representative conformation of the complex formed by the full-length CaM and the CNT. (g) Cartoon representation of the complex formed by the full-length CaM and the binding peptide of its target protein CaMKI.

The above simulation results demonstrate that the conformational distributions of the CNT-bound CaMc are determined by the interplay of intradomain interactions, Ca2+ binding energies, and the CNT binding energies. From the closed state to the open state and distorted state, the overall intradomain interactions are eliminated gradually, which is accompanied by an increase of the CNT−CaMc hydrophobic interactions (Figure S1b). At the same time, the calcium binding tends to stabilize the open conformation. These three factors together determine the responding ability of CaMc to Ca2+ concentrations. Notably, at ϵHP = 0.2−0.4, the CNT binding has the most significant effect on the Ca2+ binding and may suggest high toxicity, which covers the realistic interaction strength (0.4 kcal/mol) for the standard hydrophobic CNT without oxidation, as estimated by matching the data of CG simulations with those obtained in previous atomic MD simulations (Figure S2).38 Previous circular dichroism experiments also supported that CNTs tend to lock the conformation of CaM rather than destroy the overall structure.17 Interestingly, the CNT occupies the same binding site as that of the target protein CaMKI for the full-length CaM as shown in Figure 5f,g, which also suggests that the CNT can also introduce toxicity to CaM by competing with the target protein for binding with CaM, which is consistent with previous computational studies with atomic MD simulations.17 In addition, the CNT binding leads to an interdomain rearrangement compared to the conformations of the isolated CaM (Figure 5a,b), demonstrating the CNT binding-induced conformational change of CaM. Size Effect of the Hydrophobic Nanotube. CNTs with different sizes can have different impacts on the CaM structures as demonstrated in a recent work.17 Here, we conducted additional MD simulations, investigating the conformational equilibrium and Ca2+ affinity of CaM in the presence of varisized CNTs (from CNT(3,3) to CNT(9,9)). All simulations were conducted at [Ca2+] = 1.0 μM and ϵHP = 0.4 kcal/mol. As shown in Figure 6a,b, increasing CNT sizes tends to decrease both the Ca2+ affinity and the population of the open conformation, particularly for CNT sizes larger than

Figure 6. Size dependence of the impacts of CNT binding on the calcium affinity and conformational equilibrium of CaM. (a) Population of the Ca2+-bound state as a function of CNT sizes for the CaMn (red) and CaMc (blue) domains of the full-length CaM. (b) Population of the open conformation as a function of CNT sizes for the CaMn (red) and CaMc (blue) domains of the full-length CaM. For comparison, the data of the simulations without CNT binding are also shown (dash lines). (c,d) Correlations between the populations of the Ca2+-bound state and the open state for the CaMn (c) and CaMc (d) domains of the full-length CaM. The Pearson correlation coefficients are labeled in the panels.

CNT(6,6). Consequently, CNTs with sizes smaller than CNT(6,6) have the most significant impacts on calcium binding. As the calcium-binding sites are far apart from the hydrophobic patch for CNT binding, the decrease of the calcium affinity does not arise from direct interactions between 10498

DOI: 10.1021/acsomega.9b01217 ACS Omega 2019, 4, 10494−10501

ACS Omega

Article

theory.29 Previous experimental measurements and computational studies showed that the conformational motions of the two individual domains involve two-state transitions.33,35 The above double-basin potentials were used separately for the two individual domains and their interface of CaM. In addition, the interdomain rearrangement is mainly controlled by the interactions arising from the CNT binding, which does not rely on the double-basin potential as will be described in the subsequent discussions. The atomic interaction-based CG model (AICG2+), which has been successfully used in simulating the large-amplitude conformational motions and folding of allosteric proteins and large protein complexes in previous works,30,41−43 was used in constructing structurebased potentials. The energy function of the AICG2+ force field is given by

the calcium-binding sites and CNTs but from the destabilization of the open conformation, which has higher calcium affinity. Such an interplay between the destabilization of the open conformation and the decreased calcium affinity as the CNT size increases can also be demonstrated by the strong correlation between the populations of the calcium-binding state and the open conformation, as shown in Figure 6c,d for CaMn and CaMc of the full-length CaM. A more detailed analysis showed when the CNT size is larger than CNT(7,7), the bound CNT is not compatible with the hydrophobic pocket of CaM, leading to the distortion of the overall CaM structure. Therefore, the impact of CNTs with larger sizes is similar to that of graphene sheets, which can distort the structures of peptides and proteins like protein G upon binding.18,39

■

flp nat nat V (R |R 0) = Vbond + Vloc + Vloc + Vnloc + Vexv

CONCLUSIONS In this work, to study the molecular mechanisms of the cytotoxicity of hydrophobic nanotubes, we constructed a CG model that can integratedly describe the conformational motions, calcium binding, and CNT interactions of the signaling protein CaM with a much longer timescale than that accessible by conventional all-atom MD simulations. Such a CG model enables the direct simulations of the impacts of CNT binding on the functionally relevant signaling dynamics of CaM. Our results suggest that there are three possible mechanisms by which the CNT binding can lead to the dysfunction of the CaM-based signal transduction, namely (1) inhibiting the target recognition by competitive interactions, which is consistent with previous observations;17 (2) modifying the conformational equilibrium of CaM; and (3) increasing the calcium affinity and therefore disrupting the calcium homeostasis. The mechanism (3) is particularly interesting and implies that the CNT binding not only disrupts the signaling function of CaM as a calcium sensor but also renders CaM to toxic species through sequestering the calcium ions from other competing calcium-binding proteins because of the increased calcium affinity. Our results provide new insights into the molecular mechanism of the cytotoxicity of hydrophobic nanoparticles.

The energy function includes the bond potential (Vbond), flp flexible local potential (Vloc ), structure-based local and nat nonlocal terms (Vnat and V loc nloc), and the excluded volume term. Details of the above terms can be found in ref 30. In eq 1, ΔV controls the conformational equilibrium of the open and closed states. Here, the values given in ref 30 were used. For the CaMc domain, the experimental data showed that at the apo condition, the population of the open conformation is around 5−10%.35 In ref 30, we showed that a ΔV value of −3.8 kcal/mol can well reproduce the experimental data. For CaMn, the populations of the conformations at the apo condition were not very well defined experimentally. A ΔV value of 1.0 kcal/mol was used, which leads to the same conformational distribution as that of CaMc. The parameter Δ controls the barrier height of the opening/ closing transitions, and its value was chosen such that one can observe frequent conformational transitions within a reasonable simulation time. Ca2+ was considered implicitly with the energetic effect of ligand binding being modeled as attractive interactions between the residues of the binding site. The binding energy of one calcium-binding pocket is given by44

■

Vbind(R ) =

METHOD The calcium exchange and conformational motions of CaM involve the timescale of milliseconds, which is far beyond the accessible timescale of conventional atomistic MD simulations. To overcome the timescale difficulty, we used a CG multibasin model30 to simulate the conformational motions of CaM. In this model, the residues of CaM are represented by spherical beads located at the Cα positions. The energy function of the multibasin model is given by40 VMB(R ) =

i (r IJ − r0IJ )2 yz zz zz 2 2 σ k {

∑ ϵlig expjjjjj− IJ

(3)

where ϵlig was set as 2.2 kcal/mol, with which CaM dominantly stays at an open conformation after Ca2+ binding, consistent with the experimental observations.22,23,45 σ represents the width of the Gaussian potential for the ligand-mediated contacts and was set as 0.15 Å. r0ij is the distance between ligand residues in the native conformation. The transitions between the Ca2+-bound and unbound states were realized by standard Metropolis Monte Carlo simulations. Specifically, Ca2+ binding is assumed to be diffusion-limited and occurs with the rate kon[Ca2+], where kon is the second-order rate constant and [Ca2+] is the calcium concentration. The Ca2+ dissociation rate is given by k0off exp(−Vbind/kBT). The values of the parameters kon and koff were taken from ref 30, where Vbind is the binding energy that depends on the conformations of the binding site and kBT is the thermal energy. The Ca2+ binding/ unbinding leads to switching of the overall energy function, and the conformational motions and CNT binding of CaM are dictated by the dynamic energy landscape. On the other hand, the CNT binding modifies the relative populations of the open and closed conformations of the two individual domains and

V (R |R 0closed) + V (R |R 0open) + ΔV 2

closed open jij V (R |R 0 ) − V (R |R 0 ) − ΔV zyz jj zz + Δ2 j z 2 k { 2

−

(2)

(1)

Here, R collectively represents the coordinates of the residues in a protein and R0 is the coordinate of the reference structure. closed V(R|Ropen ) are the structure-based potentials 0 ) and V(R|R0 centric to the open and closed structures, respectively, and are constructed within the framework of the energy landscape 10499

DOI: 10.1021/acsomega.9b01217 ACS Omega 2019, 4, 10494−10501

ACS Omega

Article

therefore the Ca2+-binding energies, which in turn modulate the rate of Ca2+ dissociation. Such an implicit Ca2+ binding model has been successfully used in previous works to reproduce and explain a number of experimental works on the Ca2+ binding-coupled folding and the conformational motions of CaM.30,46 CNTs were represented by all-atom structures, with the bonds, angles, and dihedral angles being restrained to the standard values by harmonic potentials. The atoms of CNTs interact with the hydrophobic residues of CaM by the following Lennard−Jones potential ÄÅ ÉÑ Ñ 0 y10 Ñ ÅÅÅ ji r 0 zy12 i ÑÑ r j z ij Å ij VHP = ∑ ϵHPÅÅÅÅ5jjjj zzzz − 6jjjj zzzz ÑÑÑÑ j rij z ÑÑ ÅÅ j rij z i,j k { ÑÖÑ (4) ÅÇÅ k {

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the National Natural Science Foundation of China (11574132, 11774157, 11774158) and the HPC center of Nanjing University.

■

where r0ij = (σi + σj)/2, with σi and σj being the radii of the residue i of CaM and the carbon atom j of CNT, respectively. The radius of the hydrophobic residues is taken from CafeMol.47 The radius of the carbon atoms is taken from the force field AMBER99SB.48 A single-wall CNT was used in the simulation. The interactions between the carbon atoms of the CNT and all nonhydrophobic residues are given by the excluded volume term. The interaction strength ϵHP between CNT atoms and CaM residues is determined by matching the potentials of mean force between a short peptide Gly−Phe− Gly and a graphene sheet calculated by previous all-atom MD simulations (Figure S2).38 Such CaM−CNT interactions dictate the interdomain rearrangements of CaM and do not rely on the double-basin potentials discussed above. It is worth noting that π−π interactions have crucial contribution to the interactions between CNTs and proteins as discussed in previous works.17,19,38 In the above CG model, all hydrophobic residues have the same interaction strength with CNT atoms. Therefore, the current CNT−CaM interaction potential can only distinguish the hydrophobic and hydrophilic residues, which limits its application to discuss in more detail the sequence dependence of CNT−CaM interactions. In our simulations, the CNT and CaM (or its isolated domains) are put into a cubic box with a size of 130 Å × 130 Å × 130 Å. We note that the resultant CNT−CaM concentration with the above box size is much higher than the typical experimental values, which may lead to fast binding between the CNT and CaM. All simulations were conducted by the software CafeMol with Langevin dynamics at a temperature of 300 K.47

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsomega.9b01217. Simulation details and additional data (PDF)

■

REFERENCES

(1) Dresselhaus, M. S.; Dresselhaus, G.; Eklund, P. C. Science of Fullerenes and Carbon Nanotubes: Their Properties and Applications; Academic Press, 1996. (2) Muller, J.; Huaux, F.; Moreau, N.; Misson, P.; Heilier, J.-F.; Delos, M.; Arras, M.; Fonseca, A.; Nagy, J. B.; Lison, D. Respiratory toxicity of multi-wall carbon nanotubes. Toxicol. Appl. Pharmacol. 2005, 207, 221−231. (3) Belyanskaya, L.; Weigel, S.; Hirsch, C.; Tobler, U.; Krug, H. F.; Wick, P. Effects of carbon nanotubes on primary neurons and glial cells. NeuroToxicology 2009, 30, 702−711. (4) Salvador-Morales, C.; Townsend, P.; Flahaut, E.; Vénien-Bryan, C.; Vlandas, A.; Green, M. L. H.; Sim, R. B. Binding of pulmonary surfactant proteins to carbon nanotubes; potential for damage to lung immune defense mechanisms. Carbon 2007, 45, 607−617. (5) Firme, C. P.; Bandaru, P. R. Toxicity issues in the application of carbon nanotubes to biological systems. Nanomed. Nanotechnol. Biol. Med. 2010, 6, 245−256. (6) Zhao, X.; Liu, R. Recent progress and perspectives on the toxicity of carbon nanotubes at organism, organ, cell, and biomacromolecule levels. Environ. Int. 2012, 40, 244−255. (7) Smart, S. K.; Cassady, A. I.; Lu, G. Q.; Martin, D. J. The biocompatibility of carbon nanotubes. Carbon 2006, 44, 1034−1047. (8) Lam, C.-W.; James, J. T.; McCluskey, R.; Arepalli, S.; Hunter, R. L. A review of carbon nanotube toxicity and assessment of potential occupational and environmental health risks. Crit. Rev. Toxicol. 2006, 36, 189−217. (9) Liu, Y.; Zhao, Y.; Sun, B.; Chen, C. Understanding the toxicity of carbon nanotubes. Acc. Chem. Res. 2012, 46, 702−713. (10) Kagan, V. E.; Tyurina, Y. Y.; Tyurin, V. A.; Konduru, N. V.; Potapovich, A. I.; Osipov, A. N.; Kisin, E. R.; Schwegler-Berry, D.; Mercer, R.; Castranova, V.; et al. Direct and indirect effects of single walled carbon nanotubes on RAW 264.7 macrophages: role of iron. Toxicol. Lett. 2006, 165, 88−100. (11) Zhou, R.; Gao, H. Cytotoxicity of graphene: recent advances and future perspective. Wiley Interdiscip. Rev.: Nanomed. Nanobiotechnol. 2014, 6, 452−474. (12) Ge, C.; Du, J.; Zhao, L.; Wang, L.; Liu, Y.; Li, D.; Yang, Y.; Zhou, R.; Zhao, Y.; Chai, Z.; et al. Binding of blood proteins to carbon nanotubes reduces cytotoxicity. Proc. Natl. Acad. Sci. U.S.A. 2011, 108, 16968−16973. (13) Saptarshi, S. R.; Duschl, A.; Lopata, A. L. Interaction of nanoparticles with proteins: relation to bio-reactivity of the nanoparticle. J. Nanobiotechnol. 2013, 11, 26. (14) Parveen, R.; Shamsi, T. N.; Fatima, S. Nanoparticles-protein interaction: role in protein aggregation and clinical implications. Int. J. Biol. Macromol. 2017, 94, 386−395. (15) Kopp, M.; Kollenda, S.; Epple, M. Nanoparticle−protein interactions: therapeutic approaches and supramolecular chemistry. Acc. Chem. Res. 2017, 50, 1383−1390. (16) Miceli, E.; Kar, M.; Calderón, M. Interactions of organic nanoparticles with proteins in physiological conditions. J. Mater. Chem. B 2017, 5, 4393−4405. (17) Gao, J.; Wang, L.; Kang, S.-g.; Zhao, L.; Ji, M.; Chen, C.; Zhao, Y.; Zhou, R.; Li, J. Size-dependent impact of CNTs on dynamic properties of calmodulin. Nanoscale 2014, 6, 12828−12837. (18) Zuo, G.; Zhou, X.; Huang, Q.; Fang, H.; Zhou, R. Adsorption of villin headpiece onto graphene, carbon nanotube, and C60: effect of contacting surface curvatures on binding affinity. J. Phys. Chem. C 2011, 115, 23323−23328.

AUTHOR INFORMATION

Corresponding Authors

*E-mail: wfl[email protected] (W.L.). *E-mail: [email protected] (W.W.). ORCID

Jian Zhang: 0000-0003-2338-0395 Jun Wang: 0000-0003-2954-2607 Wenfei Li: 0000-0003-2679-4075 Wei Wang: 0000-0001-5441-0302 10500

DOI: 10.1021/acsomega.9b01217 ACS Omega 2019, 4, 10494−10501

ACS Omega

Article

tions of allosteric proteins. Proc. Natl. Acad. Sci. U.S.A. 2011, 108, 3504−3509. (42) Li, W.; Terakawa, T.; Wang, W.; Takada, S. Energy landscape and multiroute folding of topologically complex proteins adenylate kinase and 2ouf-knot. Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 17789− 17794. (43) Lequieu, J.; Schwartz, D. C.; de Pablo, J. J. In silico evidence for sequence-dependent nucleosome sliding. Proc. Natl. Acad. Sci. U.S.A. 2017, 114, E9197−E9205. (44) Okazaki, K.-I.; Takada, S. Dynamic energy landscape view of coupled binding and protein conformational change: induced-fit versus population-shift mechanisms. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 11182−11187. (45) Stigler, J.; Rief, M. Calcium-dependent folding of single calmodulin molecules. Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 17814− 17819. (46) Nandigrami, P.; Portman, J. J. Coarse-grained molecular simulations of allosteric cooperativity. J. Chem. Phys. 2016, 144, 105101. (47) Kenzaki, H.; Koga, N.; Hori, N.; Kanada, R.; Li, W.; Okazaki, K.-I.; Yao, X.-Q.; Takada, S. CafeMol: a coarse-grained biomolecular simulator for simulating proteins at work. J. Chem. Theory Comput. 2011, 7, 1979−1989. (48) Hornak, V.; Abel, R.; Okur, A.; Strockbine, B.; Roitberg, A.; Simmerling, C. Comparison of multiple Amber force fields and development of improved protein backbone parameters. Proteins 2006, 65, 712−725.

(19) Zuo, G.; Huang, Q.; Wei, G.; Zhou, R.; Fang, H. Plugging into proteins: poisoning protein function by a hydrophobic nanoparticle. ACS Nano 2010, 4, 7508−7514. (20) Zuo, G.; Kang, S.-G.; Xiu, P.; Zhao, Y.; Zhou, R. Interactions Between Proteins and Carbon-Based Nanoparticles: Exploring the Origin of Nanotoxicity at the Molecular Level. Small 2013, 9, 1546− 1556. (21) Yamniuk, A. P.; Vogel, H. J. Calmodulin’s flexibility allows for promiscuity in its interactions with target proteins and peptides. Mol. Biotechnol. 2004, 27, 33−58. (22) Chattopadhyaya, R.; Meador, W. E.; Means, A. R.; Quiocho, F. A. Calmodulin structure refined at 1.7 Å resolution. J. Mol. Biol. 1992, 228, 1177−1192. (23) Zhang, M.; Tanaka, T.; Ikura, M. Calcium-induced conformational transition revealed by the solution structure of apo calmodulin. Nat. Struct. Mol. Biol. 1995, 2, 758. (24) Bayley, P. M.; Findlay, W. A.; Martin, S. R. Target recognition by calmodulin: dissecting the kinetics and affinity of interaction using short peptide sequences. Protein Sci. 1996, 5, 1215−1228. (25) Bähler, M.; Rhoads, A. Calmodulin signaling via the IQ motif. FEBS Lett. 2002, 513, 107−113. (26) Stefan, M. I.; Edelstein, S. J.; Le Novere, N. An allosteric model of calmodulin explains differential activation of PP2B and CaMKII. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 10768−10773. (27) O’Day, D. H. CaMBOT: profiling and characterizing calmodulin-binding proteins. Cell. Signalling 2003, 15, 347−354. (28) Hoeflich, K. P.; Ikura, M. Calmodulin in Action. Cell 2002, 108, 739−742. (29) Onuchic, J. N.; Luthey-Schulten, Z.; Wolynes, P. G. Theory of protein folding: the energy landscape perspective. Annu. Rev. Phys. Chem. 1997, 48, 545−600. (30) Li, W.; Wang, W.; Takada, S. Energy landscape views for interplays among folding, binding, and allostery of calmodulin domains. Proc. Natl. Acad. Sci. U.S.A. 2014, 111, 10550−10555. (31) Liu, F.; Chu, X.; Lu, H. P.; Wang, J. Molecular mechanism of multispecific recognition of Calmodulin through conformational changes. Proc. Natl. Acad. Sci. U.S.A. 2017, 114, E3927−E3934. (32) Nandigrami, P.; Portman, J. J. Comparing allosteric transitions in the domains of calmodulin through coarse-grained simulations. J. Chem. Phys. 2016, 144, 105102. (33) Chen, Y.-G.; Hummer, G. Slow conformational dynamics and unfolding of the calmodulin C-terminal domain. J. Am. Chem. Soc. 2007, 129, 2414−2415. (34) Anthis, N. J.; Doucleff, M.; Clore, G. M. Transient, sparsely populated compact states of apo and calcium-loaded calmodulin probed by paramagnetic relaxation enhancement: interplay of conformational selection and induced fit. J. Am. Chem. Soc. 2011, 133, 18966−18974. (35) Malmendal, A.; Evenäs, J.; Forsén, S.; Akke, M. Structural dynamics in the C-terminal domain of calmodulin at low calcium levels. J. Mol. Biol. 1999, 293, 883−899. (36) Kuboniwa, H.; Tjandra, N.; Grzesiek, S.; Ren, H.; Klee, C. B.; Bax, A. Solution structure of calcium-free calmodulin. Nat. Struct. Mol. Biol. 1995, 2, 768. (37) Linse, S.; Helmersson, A.; Forsén, S. Calcium binding to calmodulin and its globular domains. J. Biol. Chem. 1991, 266, 8050− 8054. (38) Dasetty, S. Understanding molecular interactions between proteins and carbon nanomaterials. Ph.D. Thesis, Clemson University, 2015. (39) Sun, T.; Chan, K. K.; Su, H.; Zhang, D. Dissimilar stability of proteins in graphene bilayer: a molecular dynamics study. Mol. Phys. 2013, 111, 545−552. (40) Okazaki, K.-I.; Koga, N.; Takada, S.; Onuchic, J. N.; Wolynes, P. G. Multiple-basin energy landscapes for large-amplitude conformational motions of proteins: Structure-based molecular dynamics simulations. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 11844−11849. (41) Li, W.; Wolynes, P. G.; Takada, S. Frustration, specific sequence dependence, and nonlinearity in large-amplitude fluctua10501

DOI: 10.1021/acsomega.9b01217 ACS Omega 2019, 4, 10494−10501