Michael González-Durruthy*ab,
Adriano V. Werhlicd,
Luisa Cornetetcd,
Karina S. Machadocd,
Humberto González-Díazef,
Wilson Wasiliesky Jr.g,
Caroline Pires Ruash,
Marcos A. Geleskyh and
José M. Monserratab
aInstituto de Ciências Biológicas (ICB) – Universidade Federal do Rio Grande-FURG, Cx. P. 474, CEP 96200-970, Rio Grande, RS, Brazil. E-mail: gonzalezdurruthy.furg@gmail.com; Tel: +55 5332935196
bPrograma de Pós-Graduação em Ciências Fisiológicas-Fisiologia Animal Comparada-ICB-FURG, RS, Brazil
cCentro de Ciências Computacionais (C3) – Universidade Federal do Rio Grande-FURG, RS, Brazil
dPrograma de Pós-Graduação em Computação-C3-FURG, RS, Brazil
eDepartment of Organic Chemistry II, Faculty of Science and Technology, University of the Basque Country UPV/EHU, 48940, Leioa, Bizkaia, Spain
fIKERBASQUE, Basque Foundation for Science, 48011, Bilbao, Bizkaia, Spain
gInstituto de Oceanografia (IO), FURG, Brazil
hPrograma de Pós-graduação em Química Tecnológica e Ambiental, FURG, Brazil
First published on 14th June 2016
Interactions between the single walled carbon nanotube (SWCNT) family and a mitochondrial ADP/ATP carrier (ANT-1) were evaluated using constitutional (functional groups, number of carbon atoms, etc.) and electronic nanodescriptors defined by (n, m)-Hamada indexes (armchair, zig-zag and chiral). The Free Energy of Binding (FEB) was determined by molecular docking simulation and the results showed that FEB was statistically more negative (p < 0.05), following the order SWCNT-COOH > SWCNT-OH > SWCNT, suggesting that polar groups favor the anchorage to ANT-1. In this regard, it was showed that key ANT-1 amino acids (Arg 79, Asn 87, Lys 91, Arg 187, Arg 234 and Arg 279) responsible for ADP-transport were conserved in ANT-1 from different species examined to predict SWCNT interactions, including shrimp Litopenaeus vannamei and fish Danio rerio commonly employed in ecotoxicology. The SWCNT-ANT-1 inter-atomic distances for the key ANT-1 amino acids were similar to that with carboxyatractyloside, a classical inhibitor of ANT-1. Significant linear relationships between FEB and n-Hamada index were found for zig-zag SWCNT and SWCNT-COOH (R2 = 0.95 in both cases). A Perturbation Theory-Nano-Quantitative Structure-Binding Relationship (PT-NQSBR) model was fitted that was able to distinguish between strong (FEB < −14.7 kcal mol−1) and weak (FEB ≥ −14.7 kcal mol−1) SWCNT–ANT-1 interactions. A simple ANT-1-inhibition respiratory assay employing mitochondria suspension from L. vannamei, showed good accordance with the predicted model. These results indicate that this methodology can be employed in massive virtual screenings and used for making regulatory decisions in nanotoxicology.
Among the CNTs, single-walled carbon nanotubes (SWCNTs) have rapidly become one of the most widely studied nanomaterials, primarily on the basis that their unique physico-chemical properties increase the number of new applications in nanomedicine as active ingredients, supportive substrates and pharmaceutical excipients for the design of versatile drug delivery systems.1,2
Given these diverse and important applications, it is expected that the number and variety of manufactured CNTs will increase rapidly over the next years, imposing a need for new methods to quickly test the potential toxicity of these nanomaterials for their safe use in nanotechnologies.2
Implementation of in silico methods based in Docking Simulation (DS) appears to be an efficient alternative for the prediction of the potential toxicity and environmental impact of CNTs. DS methodology predicts the non-covalent binding of a receptor (usually an enzyme or protein) and a ligand (smaller molecules as SWCNT) at the atomic level.3 The algorithms defined in DS can test hundreds of thousands of ligand conformations and orientations to find the best receptor-ligand binding affinity by assigning and optimizing a score.3,4
In particular, Docking Simulation (DS) coupled to a Virtual Screening Framework (DS-VSF) represents a powerful new technique for the rational design of a SWCNT before its mass production because it allows the computational analysis of a large volume of hypothetical compounds and the selection of the compounds that might have a greater chance of interacting with certain receptors or targets.5–7 Currently there are no precedents of this methodology applied to the evaluation of potential toxicity of SWCNT.
Some in vitro studies have been demonstrated that SWCNTs exert cytotoxicity after their accumulation in the mitochondria matrix and/or by affecting the function of mitochondrial proteins of the inner membrane.8 Previous in vitro research including drugs and environmental pollutants using sub-mitochondrial particles (respiratory chain complexes I, II, III, IV; ADP/ATP translocator, ATP synthase/ATPase) to predict the toxic impact of 92 different xenobiotics showed a strong correlations with the toxicity in human and indicated that the mitochondria are a relevant model for studying the relative toxicity of many xenobiotics.9 On the other hand, some authors using DS have found that carbon nanomaterials can block different protein channels for the transport of cations, anions and zwitterions, as the CNTs are rich in hydrophobic residues at the catalytic sites, similar to natural toxins and synthetic drugs specific for these channels.10–13.
A few theoretical works have predicted that the electronic properties of carbon nanotubes depend on both the chirality and diameter, which are both functions of n and m parameters, known as the Hamada indices and defined by wrapping specifies a translation vector of the graphene lattice (chiral vector; Ch). Each CNT topology is usually characterized by these two parameters, thus defining some peculiar symmetries such as armchair (n = m), zig-zag (n > 0; m = 0) and chiral (n > m > 0) forms.14–16 The indexes n and m determine the electronic properties of CNT, which can vary between being metallic and non-metallic. If (n − m) is a multiple of 3, then the CNT exhibits metallic behavior, otherwise the CNT exhibits semiconducting or non-metallic behavior. Furthermore, the presence of OH and COOH covalent functionalization may affect the electrochemical properties and reactivity of the CNT.14 However, the constitutional and/or electro-topological nanodescriptors of CNT (such as diameter, chirality, and functionalization) have not yet been considered from the point of view of quantitative structure–property/activity relationships (QSPR/QSAR) with respect to their interactions with mitochondrial proteins. This method may play an important role as a predictive tool for the risk assessment of nanomaterials, as indicated by the pioneering works on nano-QSPR studies of nanoparticles (NQSPR) published by Puzyn (2009).17 In general, the main assumption of QSPR/QSAR models is that similar molecules have similar properties. Consequently, smaller changes in the structure of the system should correlate linearly with smaller changes on the values of its properties. However, not all similar molecules have similar properties. The underlying problem is therefore how to define one smaller structural change on a molecular level. The problem is relevant because each type of property, e.g., partition coefficient, reactivity, or metabolism, is expected to depend on another difference. It means that it is necessary to quantify “smaller” variations (perturbations) in the molecular structural level that in turn imply a “smaller” linear change in the free energy of interaction of the nano-drug with the receptor (or CNT with mitochondrial proteins).18
Very recently, Gonzalez-Díaz et al. (2013) formulated a general purpose Perturbation Theory (PT) model for chemoinformatics problems with multiple-boundary experimental conditions.19 This new methodology is potentially useful to carry out Quantitative Structure-Binding Relationships (QSBR) in the context of the present work to predict the interaction of SWCNT with ADP/ATP mitochondrial carrier isoform 1 (ANT-1) in quantitative terms. However, to do so, the method must be adapted for QSBR studies in nanosciences. In this work, it was describe the re-formulation of this model to develop a new type of PT-Nano-QSBR model for nanoparticles (PT-NQSBR models) to be used for in silico studies of SWCNT–ANT-1 interactions.
ANT-1 catalyzes the electrogenic ADP3− and ATP4− exchange across the inner mitochondrial membrane. The transporter provides ATP4− efflux into the cytosol in exchange for the entry of ADP3− into the mitochondrial matrix during oxidative phosphorylation, which can be estimated in terms of efficiency as ADP/O ratio. Considering the relationship between the amount of ADP and total amount of oxygen consumed during state III of respiration ADP-dependent. By the other hand, the ADP transport can be specifically inhibited by carboxyatractyloside (CATR), which reduces the ADP affinity. Particularly in the cationic cluster consisting of key amino acids (Arg 79, Asn 87, Lys 91, Arg 187, Asp 231, Arg 234) involved in the ADP-transport through active site of ANT-1.20,21
Under several pathophysiological conditions such as cardiomyopathy, Alzheimer disease and lactic acidosis, ANT-1 is a component of the mitochondrial permeability transition pore (PTPM), a multiprotein complex that is directly implicated in apoptosis.21–23 The induction and/or inhibition of the PTPM-(ANT-1) by SWCNT could represent an attractive therapeutic strategy to induce cytotoxicity and/or cytoprotection based on the modulation of ANT-1 function. In these sense, some potential pre-clinic applications of SWCNT could be considered: (1) death of cancerous cells by conformational changes of ANT-1 associated with the inhibition of ADP transport and mitochondrial swelling, which act as an MPTP-(ANT-1) inducer;24,25 and (2) as an MPTP-(ANT-1) inhibitor, preventing mitochondrial calcium overload in the calcium binding domain of ANT-1 and the migration of ANT-1 during MPTP-(ANT-1) assembly in pathophysiological conditions, on the basis that ADP is an inhibitor of MPTP and has an important role in oxidative phosphorylation.22,26–28
In present study, it was analyzed how covalent functionalization (–OH and –COOH) and different structural geometries of SWCNT including armchair, chiral and zig-zag forms can act as specific inhibitors of ANT-1. Taking into account the information cited above, the main objective of this study was to evaluate the interactions between SWCNT and ANT-1 using DS-VSF and nano-QSBR-perturbation theory (PT) model to predict the structural attributes of SWCNT involved in the interactions with mitochondrial ADP transport by ANT-1.
In the second step, the SWCNTs-ligands (pristine-SWCNTs or SWCNT-H) structures were carefully modeled taking to account general CNT-nanodescriptors semi-empirical values for [n] and [m]-Hamada indexes calculated by H. Yorikawa and S. Muramatsu in 1995 (ref. 29) and others CNT-parameters like molecular weight, number of bonds, number of atoms, ratio, diameter, hexagons number/1D unit cell, metallic and/or semiconducting properties.29 For this instance it was used the software Nanotube Modeler (http://www.jcrystal.com/products/wincnt/) version 1.7.5 registered to one of the authors (J. M. Monserrat). Furthermore, some pristine-SWCNT structures were oxidized either with carboxyl (–COOH) or hydroxyl (–OH) moieties using an advanced semantic chemical editor Avogrado (Version 1.1.1 free software). All the SWCNT-ligands minimization was done using the MOPAC extension for geometry optimization based on the AM1-Hamiltonian method.
An ad hoc framework was developed to configure the virtual screening (VS) experiments to evaluate the various parameters. This framework has a web interface in which the user configures the experiment and obtains the respective Python script to automatically perform the VS steps. In the framework interface, the user provides information regarding the receptor file (ANT-1) and the folder in which all the nanotubes structures are stored. To evaluate the SWCNTs–ANT-1 in silico interactions Autodock Vina rigid docking it was implemented, open source software developed by Trott & Olson (2010) herein the receptor (ANT-1) and ligands (SWCNTs) were considered as a rigid molecules.30 Following this idea, conformational rigidification favors a significant gain of enthalpy of SWCNT–ANT-1 complexes associated to reduction of SWCNT-intramolecular deformation or vibrational decrease within ANT-1 active site.
In this context, the (SWCNTs–ANT-1) complexes free energy of binding (FEB) were calculated based on the score function which attempt to approximate the standard chemical potentials (ΔGbind). For this instance, the ΔG scoring function used combines the knowledge-based potential and empirical information obtained from experimental affinity measurements. Following this idea, the FEB of SWCNT–ANT-1 complex optimization it was performed with sophisticated gradient and efficient local optimization algorithm of energy based on quasi-Newton method like Broyden–Fletcher–Goldfarb–Shanno (BFGS). In this algorithm, a succession of steps consisting of a mutation and a local optimization are taken, with each step being accepted according to the Metropolis criterion.30 This theoretical procedure was performed to the receptor binding cavity using Cartesian coordinates for ANT-1 grid box size with the average dimensions of X = 30 Å, Y = 30 Å, Z = 30 Å and the ANT-1 receptor grid box center X = 18.8 Å, Y = 18 Å, Z = 32 Å to evaluate the SWCNT–ANT-1 interaction, considering the CATR-biophysical environment (ANT-1 active site) to evaluate the SWCNT-affinity. Several runs starting from random conformations were performed, and the number of iterations in a run was adapted according to the problem complexity. For this instance an exhaustiveness option of 8 (average accuracy) in each docking calculation was used.30 The docking output results or FEB values are similarly defined to ΔGbind values for all docked poses according to ΔG energy scoring function with the thermodynamic description represented below:
ΔGbind ≈ FEB = G(SWCNT/ANT-1 complex) − G(ANT-1 receptor) − G(SWCNT) |
FEB = ΔH + ΔG(solvation) − TΔS(bind) |
FEB = ΔE(MM) + ΔG(GB) + ΔG(SA) − TΔS(bind) |
The next step was the analysis of the FEB results inter-atomic distances between key amino acids of the receptor (ANT-1) and atoms at the best binding position for ligands (SWCNT).31–34 It was considered key amino acids those involved in the inhibition of ADP transport by CATR, in way to compare the SWCNT inhibitory potential with this inhibitor.
![]() | (1) |
![]() | (2) |
The first input term is the function f(FEB)ref = 〈FEB〉query which is the average value of FEB for all the SWCNT of the same class as query SWCNT. It means that 〈FEB〉query can be considered as the expected value of FEB for the interaction of a new SWCNT (query SWCNT) with the target protein ANT-1 (assuming a normal distribution). The second class of terms Vk are the values of the structural parameters of the query SWCNT. Last, the difference (ΔVk = queryVk − refVk) quantify the deviations or perturbations (changes, distortions, etc.) on the SWCNT-structural parameters (queryVk) of the new SWCNT compared with those of the original reference SWCNT (refVk). We can substitute each symbol Vk by the classic symbol of the respective property (SWCNT-nanodescriptors) and expand the input terms in order to understand better this mathematical formalism following the eqn (3) and (4):
![]() | (3) |
![]() | (4) |
Refers to Table 1 to see more details of the employed model. It was used the Linear Discriminant Analysis (LDA) forward-stepwise algorithms implemented in the software STATISTICA to fit the values of the parameters (a0, ak, bk, ck, dk and e0) and other parameters of the model. In the PT-NQSBR model, the output f(FEB)query is a function of the value of FEB for the new SWCNT-structure which contains the CNT-nanodescriptors (Hamada index n and m, diameter, molecular weight, number of atoms). Following this idea, it is important to note that the SWCNT-diameter as a relevant nanodescriptor on the prediction of the f(FEB)query it was considered in our chemoinformatics model through Hamada index n and m, because this structural parameter has a strong and direct proportionality relationship with the different geometry configurations of SWCNT evaluated as referred in the Section 2.2 according to eqn (5):
![]() | (5) |
Nano-QSBR model input variables | CNTs-nanodescriptors details (Vk) |
---|---|
〈FEB〉query | FEB-expected value for a CNT of the same type than the query |
mquery | m Hamada index values for CNTquery |
nquery | n Hamada index values for CNTquery |
ΔMw = Mwquery − Mwref | Difference in molecular weight (Mw) between the query and reference CNT |
ΔNa = Naquery − Naref | Difference in number of atoms (Na) between the query and reference CNT |
In Fig. 1, it is depicted the workflow for this theoretical process. These PT-NQSBR models should predict the probability of interaction of the CNT structures with a target protein (ANT-1 in this case).
Continuous-monitoring of oxygen consumption in mitochondrial suspensions was polarographically determined with a Clark-type electrode (Oxygraph System Hansatech Instrumens) in a 2 mL glass chamber equipped with a magnetic stirrer. Isolated mitochondria (1 mg protein per mL) from hepatopancreas were energized with 5 mM potassium succinate (plus 2.5 μM rotenone) in a standard incubation medium consisting of 125 mM sucrose, 65 mM KCl, 2 mM inorganic phosphate (K2HPO4) and 10 mM HEPES-KOH pH 7.4 at 20 °C in standard respiration medium. The experimental approach was calibrated using the oxygen content of air saturated medium.35,36
It were performed respiration protocols using three types of multi-walled carbon nanotubes MWCNT which were formed by 3 concentric tubes with distance of 0.34 nm between each wall and semiconducting behavior (conductivity = 100 S cm−1). The outer diameter of Dmax = 7.6 ± 1.5 Å for the three CNT-samples were similar to maximum diameter of zig-zag-SWCNTs (Dmax = 7.051 Å) used in the molecular docking experiments. The final concentration was 5 μg mL−1 in all cases.
Before the respiratory assays MWCNTs were dissolved in dimethyl sulfoxide (DMSO: 900 μL) and ultrapure Milli Q water (100 μL), to prepare individual stock suspensions at a concentration of 1 mg mL−1. In order to prevent CNTs agglomeration for the oxygen consumption assays, it was employed tip-sonication regime during 5–10 min to generate a non-agglomerated suspension or monodisperse state for these CNT-samples. The sonication power was 9.3 W, with an energy input of 16.7 kJ at 25 °C using a Ultronique/Eco-sonics Q-3.0/40A sonicator. After, samples were stirred for 10–15 min. The resulting diluted suspensions were cooled to room temperature and filtered through a 0.22 μm polycarbonate membrane (Millipore, USA), before exposure to mitochondria suspensions at a final concentration of 5 μg mL−1.
Mitochondria (Mit) total oxygen consumption was calculated as the difference between oxygen concentration of respiration medium (Resp med) at time 0 and oxygen concentration at the end of the measurement (300 s). This allowed the estimation of % ADP-transport inhibition considering oxygen consumption of treatment (3) (see below) as 100%. For this instance, it were performed several combinations of treatments in order to evaluate the ANT-1 inhibition as following: (1) Resp med (blank control), (2) Resp med + Mit, (3) Resp med + Mit + ADP, (4) Resp med + Mit + ADP + CATR, (5) Resp med + Mit + ADP + MWCNT-(H), (6) Resp med + Mit + ADP + MWCNT-OH, and (7) Resp med + Mit + ADP + MWCNT-COOH. CATR refers for carboxyatractyloside, the specific inhibitor of ANT-1. A final concentration of 1 μM of the inhibitor was employed in the assays. For the whole assays, two different experiments with two different mitochondrial suspensions of L. vannamei were performed.
Taking into account that according to Pebay-Peyroula et al. (2003),37 the protein ANT-1 has dimensions of 20 and 40 Å for diameter and depth, respectively, it is expected that these are the maximum dimension limits that would allow interaction of a CNT with the protein. In fact, the diameters of the CNT assayed in the docking analysis ranged from 2.35 to 12.21 Å and in all cases the length was 10 Å. Note that Pebay-Peyroula et al. (2003) determined that the depth of the cavity for ADP binding is 10 Å. Thus, under the experimental conditions employed here, there were no steric constraints for the interaction of CNT with the catalytic site of ANT-1.37
The results showed interesting aspects regarding the potential for SWCNT to modulate the activity of ANT-1, indicating that the employed theoretical procedure can be considered for the rational design of carbon nanomaterials with higher affinity and specificity for ANT-1. The carboxylate (COO−)-moieties of SWCNT-COOH may be important for electrostatic interactions at the internal ANT-1 hydrophobic pocket, which is formed by five cationic arginine residues (Arg 79, Arg 187, Arg 231, Arg 234, and Arg 279). This observation highlights the importance of hydrophobic residues, particularly arginine rich regions, to form stable complex with SWCNT as has been suggested by Park et al. (2003) for studies of the molecular docking between SWCNT and K+-channels.12 Pebay-Peyroula et al. (2003) considered the COOH-moiety to be a toxicophore important for the ADP transport inhibition by carboxy-atractyloside (CATR).37 In contrast, the presence of the OH-moiety of CATR is considered to be a low affinity descriptor for the ANT-1 interaction.20,37
It seems reasonable to propose that the COOH-moiety of SWCNT-COOH can establish stable complexes of salt bridges with ANT-1, similar to the COO− group of CATR. In this way, deprotonated SWCNT-COO− could disrupt the association of the ADP3− anion with the positive amines of Arg or Lys residues present at the bottom of ANT-1 cavity. The same authors indicated that the carboxyl groups of CATR bind primarily to residues Arg 79, Asn 87, Lys 91, Arg 187, Arg 231 and Arg 234 at the active site of bovine ANT-1. These interactions showed an electrostatic attraction, which explains the high efficiency of this inhibitor to induce mitotoxic responses through the significant reduction of phosphorylation efficiency (ADP/O ratio) and ATP synthesis. According to the FEB values obtained for the CATR or SWCNT interactions with ANT-1, the following order of affinity can be postulated: ANT-1–CATR complex ∼ ANT-1–SWCNT pristine (armchair, zig-zag and chiral) ∼ ANT-1–SWCNT-OH (armchair and zig-zag) < ANT-1–SWCNT-OH (chiral) ∼ ANT-1–SWCNT-COOH (armchair, zig-zag and chiral) (Fig. 3). These in silico evidences suggest that SWCNT have great potential to exert drastic effects in the same biophysical environment as that affected by the specific inhibitor CATR, increasing the likelihood of inducing mitochondrial toxicity.
In this context, single walled carbon nanotubes can induce mitotoxicity through the modulation of the ADP/ATP transport in diseases such as cancer through the induction of the mitochondrial permeability transition pore (MPTP), mitochondrial dysfunction and apoptosis, in which ANT-1 is an important player on the cell bioenergetics triggering of the MPTP-(ANT-1) in pathophysiological circumstances.24,25,27,41
As mentioned in the Introduction section, ANT-1 plays an important role in maintaining the cellular redox potential and phosphorylation efficiency (ADP/O ratio), explaining its wide distribution in all eukaryotic species. In particular, the key amino acids (Arg 79, Asn 87, Lys 91, Arg 187, Asp 231, Arg 234) involved in the ADP transport by ANT-1 were shown to be fully conserved in all the species analyzed (Bos taurus, Mus musculus, Rattus norvegicus, Danio rerio, Lepeophtheirus salmonis, Litopenaeus vannamei, and Homo sapiens).28 In this way it was possible to extrapolate SWCNT affinity to ANT-1 from Bos taurus employed in this study to different animal species. To evaluate the coincidence of the key amino acids of ANT-1 involved in the interactions with the families of SWCNT (SWCNT, SWCNT-COOH, SWCNT-OH) in ANT-1 (Table 2), the ANT-1 sequence from bull Bos taurus (NP_777083.1) was aligned with the homologous sequences from other relevant species to extrapolate the interactions and/or potential toxicity of SWCNT based on comparison with the classical inhibitor of ANT-1 (carboxyatractiloside, CATR), proposed as control to compare the affinity (FEB). As shown in Fig. 4, these amino acids are fully conserved in all species analyzed.
Amino acids | CATR distance | a-SWCNT (9, 9) distance | c-SWCNT (5, 4) distance | z-SWCNT (9, 0) distance |
---|---|---|---|---|
Arg 79 | 2.65 | 3.81 | 9.90 | 2.30 |
Asn 87 | 3.12 | 8.50 | 2.90 | 3.80 |
Lys 91 | 2.73 | 3.60 | 2.44 | 2.51 |
Arg 187 | 3.0 | 5.02 | 8.77 | 3.06 |
Asp 231 | 3.0 | 15.40 | 20.10 | 10.60 |
Arg 234 | 3.0 | 5.30 | 10.70 | 7.80 |
![]() | ||
Fig. 4 Alignment of the ANT-1 sequences from bull Bos taurus (NP_777083.1), human Homo sapiens (NP_001142.2), mouse Mus musculus (NP_031476.3), rat Rattus norvegicus (NP_445967.1), fish Danio rerio (NP_999867.1), copepod Lepeophtheirus salmonis (ACO12396.1), and shrimp Litopenaeus vannamei (AEZ68611.1). The sequence of B. taurus is shown in green and the key amino acids known to interact with the ANT-1 inhibitor carboxyatractiloside are highlighted in yellow (also see Table 1). |
Furthermore, the inter-atomic distance values between SWCNT and the key amino acids (Arg 79, Asn 87, Lys 91, Arg 187, Asp 231, Arg 234) for binding to ANT-1 are, in most of the cases tested, very similar to the critical values for the interactions of the same amino acids with the classical inhibitor of ANT-1 (carboxyatractiloside, CATR) crystallographycally determined by Pebay-Peyroula et al., 2003.37 This was true particularly for Arg 79, Asn 87, Lys 91, Arg 187, which are known to be directly involved in the inhibition of ADP-transport (Table 2).37 In ESI S4† it can be found the 3D structural alignment of ANT-1 from different species (Fig. S4†) as well as the root-mean-square deviation of atomic positions (RMSD) (Table S4a†) and the FEB values after performing rigid docking simulations (Table S4b†) that showed no significant differences (p > 0.05) between species.
The analysis presented in this study should provide relevant information about biochemical models used for the evaluation of the interactions of CNT with ANT-1. In particular, this methodology may be used to understand the inhibitory mechanism and to infer a “binding site common substrate” with a location for interaction similar to that of the carbon nanotubes. This would permit extrapolation of the interactions and/or potential toxicity induced by the family of SWCNTs on ANT-1 independently of the phylogenetic position of evaluated species.
Consistent with our goal, the results obtained for the electro-topological properties (diameter-chirality, and functionalization) and the ANT-1 affinity (FEB) relationship indicated that the chiral index n is a relevant electro-topological descriptor to predict the interaction of some carbon nanotubes (zig-zag SWCNT and zig-zag SWCNT-COOH) with ANT-1. In this sense the correlation between electro-topological nano-descriptors and affinity (FEB) by ANT-1 was determined. When the influence of the various geometric configurations of SWCNT on the affinity to ANT-1 were analyzed, the chiral index n for the zig-zag carbon nanotubes (pristine and carboxylated) was shown to have an excellent linear correlation (R2 = 0.95) with FEB (p < 0.05; Fig. 5a and b). In contrast to these results, a low R2 (0.65) was observed for the linear relationship between FEB and n in zig-zag SWCNT-OH (Fig. 5c). The two dimensional (2D)-contour plot analysis to the FEB values for SWCNT, SWCNT-COOH and SWCNT-OH are depicted in the Fig. 5e and f to show that from both Hamada index, only n was relevant to describe SWCNT–ANT-1 interaction.
On the other hand, the chiral index m does not seem to be a relevant descriptor of SWCNT–ANT-1 interactions, given the lack of significant correlation (p > 0.05) with the FEB values for the three geometric configurations of SWCNT studied (zig-zag, armchair, chiral) and for all functional groups evaluated (hydrogen, carboxyl and hydroxyl groups) (see Table S1 of ESI S1†). Chirality has been widely used to describe the metallic and/or semiconducting properties, specifically for pristine SWCNT with zig-zag geometry, according to Pumera (2010).42 Also it is known that n index is proportional to the carbon nanotube diameter.15 As shown in Fig. 6, a higher diameter could induce partial or total distortions in the key amino acids within the active site of ANT-1 resulting in decreased FEB values and increased inhibition of the ADP transport by ANT-1 (see ESI S3† to see the results of the docking experiments used to construct Fig. 6). In addition, we performed control simulations with some examples of zig-zag SWCNT (3.0; 6.0; 9.0) tested, using flexible docking considering the cationic cluster formed by the arginine residues (Arg 79, Arg 187, Arg 231, Arg 234, and Arg 279) of the ANT-1 active site as flexible residues, and the FEB values obtained were very similar (not significant differences; p > 0.05) to FEB values from rigid docking simulation when these computational procedures were compared (see Table S2a in ESI 2†). Furthermore, it was verified that a high increase of the exhaustiveness of simulation from 8 to 100 keeping the same FEB results. The modification of the mentioned docking parameters (receptor flexibility and exhaustiveness) only increase the simulation time (see Table S2b in ESI 2†).
It has been reported that the zig-zag nanotubes are good conductors (similar to metals) or are semiconducting when n is a multiple of 3.15 Some theoretical studies have shown the existence of symmetry in the distribution of electric charges in the zig-zag SWCNT. In this case, the CNT maintain almost constant charges in the walls and great charge variation at the tips, a phenomenon called “edge effects”, that only appear in semiconducting zig-zag topologies of SWCNTs.43 This electronic feature may play an important role in the interactions of zig-zag SWCNTs with several channel proteins, including ANT-1. In this sense the “edge effects” has not been reported for metallic-armchair SWCNT, wherein the cycloparaphenylene aromatics system of the extreme carbon nanotubes are closed and without tips charge variation. In the case of zig-zag SWCNT-COOH, the high correlation with n could also include the presence of COOH groups that lead to more negative FEB values as shown in Fig. 5. The low correlation (R2 = 0.65) between n and FEB for zig-zag SWCNT-OH indicates that the presence of the OH group is not a good nanodescriptor and suggests that OH-functionalization could modify their electronic properties, reducing the influence of the chiral index n in the interaction energy (FEB) with ANT-1.
Experimentally it has been found that the covalent sidewall functionalization generates sp3 carbon sites in the CNT, which disrupt the band-to-band transitions of π electrons and cause loss of the novel properties of CNT including their high conductivity and remarkable mechanical properties. Other factors associated with loss of chirality is the presence of defects produced by functionalization of the sidewall including vacancies or pentagon–heptagon pairs (Stone–Wales defects) associated with full or partial covalent functionalization of SWCNT with OH and COOH (Charlier et al., 2002).14 Recently the use of computational tools focused to theoretical quantitative analysis have been extended to address pharmacological and/or toxicological properties of multiple xenobiotics in biological complex systems, including the interactions at the nanoscale range. In the last years, several articles have been published in the emerging field of Nano-QSAR studies of nanoparticles (NQSAR).17,18,44,45 In this sense, new chemoinformatics ideas based in PT-QSPR models are very useful for the study of complex molecular systems with simultaneous variation of multiple experimental boundary conditions. In present study, the LDA method was used to seek a PT-QSPR approach (PT-NQSBR) for the prediction of SWCNT–ANT-1 binding interactions. The PT-NQSBR equation infers the binding of a query SWNCT using the expected values free energy of binding for this type of SWCNT and structural parameters of SWCNT of reference as input. The output of the model is the scoring function f(FEB)query of the value of the FEB for the mentioned query SWCNT or new SWCNT. The scoring function f(FEB)query increases for higher values of probability of binding with FEB <−14.7 kcal mol−1. The cutoff value (−14.7 kcal mol−1) represents the FEB value for the interaction of CATR with ANT-1 (calculated in this work, see Fig. 3). The best equation found is indicated below, where Mw stands for molecular weight, Na for number of atoms and n and m are Hamada index.
![]() | (6) |
This equation is able discriminate the SWCNTs that bind strongly to ANT-1 (FEB < −14.7 kcal mol−1) from those with weak binding (FEB ≥ −14.7 kcal mol−1). The equation showed very high values of accuracy, specificity, and sensitivity in the range of 86.5–99.5% in both the training and external validation series (Table 3). The first input variable of reference is the expected value of f(FEB)ref = 〈FEB〉new (average value of FEB) for all SWCNT of the same type as the query SWCNT. The values of 〈FEB〉 new for different types (I, II, III, and IV) of SWCNT are shown in Table 4. The other input variables are the structural variables (mquery, nquery, Mwquery, and Naquery) for the query SWCNT and the structural variables for the reference SWCNT (Mwref, and Naref).
Training | Statistical parameter | Observed values | ||
---|---|---|---|---|
(FEB ≥ −14.7)pred | (FEB < −14.7)pred | |||
(FEB ≥ −14.7)obs | Specificity | 86.59% | 4560 | 706 |
(FEB < −14.7)obs | Sensitivity | 99.64% | 30 | 8373 |
Total | Accuracy | 94.62% | ||
![]() |
||||
Validation | ||||
(FEB ≥ −14.7)obs | Specificity | 86.55% | 1518 | 236 |
(FEB < −14.7)obs | Sensitivity | 99.50% | 14 | 2788 |
Total | Accuracy | 94.51% |
Class | 〈FEB〉 | 〈Mw〉 | Type | Funct. | Conductivity |
---|---|---|---|---|---|
I | −12.3 | 1127.9 | Pristine | H | Metallic-SWCNT |
Semi-metallic-SWCNT | |||||
II | −12.7 | 1096.1 | Pristine | H | Semi-conducting-SWCNT |
III | −24.6 | 4778.5 | Oxidized | COOH | No conducting properties |
IV | −19.9 | 3507.1 | Oxidized | OH | No conducting properties |
The model showed a remarkable efficiency (see Table 3) for the correct classification of different forms of SWCNT having strong or weak binding to ANT-1, showing its potential application in the prediction of nanomaterial–protein interactions. The values of accuracy, specificity, and sensitivity obtained with this LDA model are similar to those obtained with other PT-NQSAR models reported by other authors for the toxicity and ecotoxicity of nanoparticles. For instance, Luan et al. (2014) published a PT-NQSAR model that described the cytotoxicity of nanoparticles in multiple experimental conditions.46 Kleandrova et al. (2014) extended the use of PT-NQSPR to studies of the ecotoxicity and cytotoxicity of uncoated and coated nanoparticles under different experimental conditions47,48 Please note that, as depicted in Fig. 1, the estimated PT-NQSBR model can be employed as a prospective tool or pre-screening filter to SWCNT structure-assigned with predicted low or high affinity to mitochondrial channels like ANT-1 prior to docking experiments.
In order to validate the in silico evidences, experiments were performed with mitochondria isolated from hepatopancreas of shrimp Litopenaeus vannamei. The profile of oxygen consumption increment after ADP addition (state 3 of respiration-ADP dependent, compare black with red trace). A mild inhibition was registered with CATR (20%, blue light trace in Fig. 7). Maximum inhibition of state 3 of respiration-ADP dependent (26%) was registered with MWCNT-COOH (light green, Fig. 7), a result that fits with the more negative FEB values obtained for this kind of nanotube in the docking experiments (Fig. 3). Taking to account that the oxidative phosphorylation by ATP-synthase is depending of ADP/ATP equilibrium concentrations and the important role of ANT-1 in the ATP-transport to cytosol in physiological normoxic conditions, the in vitro results suggest potential toxicity for shrimp mitochondria. The results of this case study suggest that the severity of inhibition of the mitochondrial ADP-transport, should depend primarily of the CNT-electronic nanodescriptors as type of functionalization (H, OH, COOH) considering the following order according to the severity of state 3 inhibition: MWCNT-COOH (26%) > MWCNT-(H) (20%) > MWCNT-OH (8%).
Recent experimental evidences using oxidized-CNT porin have show the high potential of carboxylated-SWCNT as synthetic analogues of biological membrane channels with high efficiency and selectivity for transporting ions and molecules.53,54 In this regard carboxylated-CNTs can spontaneously insert into cellular and mitochondrial membrane lipid bilayers to form channels that exhibit a unitary conductance of 70–100 picosiemens under physiological conditions and at the same time the negative charge of (COO−)-moiety of carboxylated-CNT could create electrostatic barrier for the anions passage like ADP3− through the positive amines of Arg or Lys residues present at the bottom of ANT-1 cavity as mentioned above.37,53
According to in silico and experimental physiological results, the CNT-functionalization type (COOH > H > OH) can be considered a relevant CNT-nanodescriptor to explains the ANT-1 biochemical interactions in this context. By the other hand, ANT-1 may be used as a good model for theoretical binding studies because of their special electrostatics properties due to the accumulation of positively charged residues near the binding site that are similar to other cellular and sub-cellular molecular carriers and in this way address structure-relationship studies for new carbon nanomaterials.49–53
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra08883j |
This journal is © The Royal Society of Chemistry 2016 |