Study of Carbon Nanotube–Bovine Serum Albumin Interaction Using the Tritium Radiotracer Technique and Supercomputer Simulation ^†

Abstract

Bovine serum albumin (BSA) was ³H-labeled via a tritium thermal activation method that allowed quantifying BSA adsorption on single-walled carbon nanotubes (SWCNTs) to be 740 mg/mg, which leads to the ζ-potential of the BSA–SWCNT complex changing from −10 to −16 mV. Supercomputer simulations were carried out with Gromacs and PM7 with MOPAC2016 with Berendsen, Nosè–Hoover and Parrinello–Rahman algorithms. The dominant interactions between BSA and SWCNTs are found to be hydrophobic, and hydrogen bonds are also present. The mean total energy of the Coulomb and Van der Waals interactions is −646 ± 8 kJ/mol, by gmx energy.

1. Introduction

Single-walled carbon nanotubes (SWCNTs) are referred to as promising material for biocompatible devices and novel medicines [1,2,3]. In these applications, the contact of SWCNT with blood should be expected. Serum albumin is the most abundant globular protein in mammalian blood [4], so it is of great importance to study interactions of SWCNTs with major blood components such as serum albumin, as their contact is possible with the incorporation of SWCNTs into common life. A radiotracer technique with tritium is a highly sensitive instrument that allows the quantification of biomolecules in their complexes with nanomaterials without significant changes in the biomolecule structure [5].

However, it is also known that carbon nanomaterials in biological media—naturally or in vitro (intentionally)—may become covered with proteins: the formed layer is usually called a “protein corona” [6]. The protein corona is able to change carbon nanomaterials’ cytotoxicity and biocompatibility [7,8]. Since serum albumin is the major mammalian blood component, studying serum albumin–SWCNT interactions may expand understanding of protein corona formation on SWCNTs.

Serum albumin interaction with SWCNTs in a biological media is supposed to be non-covalent if no special premodifications of SWCNTs are made, but the mechanisms of albumin–SWCNT interactions are quite complex and not clearly defined. Proteins may undergo conformational changes in complexes with CNT [9] that may lead to their denaturation. An investigation of conformational changes of serum albumin in complex with SWCNTs would expand the understanding of nanomaterial–organism interfaces. We have selected bovine serum albumin (BSA) as a model protein for our study because its structure is close to human serum albumin and it is being used for antifouling coatings for nanobiotechnology applications [10]. The method of obtaining of [³H]BSA has been developed earlier [11]. Studying BSA–SWCNT interactions may help in understanding adverse effects of SWCNTs and their complexes with BSA when applied in biological systems or for environmental issues.

A combination of spectroscopy and molecular modeling is known for BSA–SWCNT systems [12], resulting in energy values for SWCNT interactions with peculiar domains, but the length of the SWCNT was 1.0 nm, which is quite small even in comparison with BSA (~14 nm [13]). Spectroscopy results may be aberrated by many factors, which are difficult to mention. Also, an SWCNT was placed into the structure of BSA with the use of molecular docking, while the real SWCNT trajectory into the chosen part of the BSA molecule may be interfered. In work [12], hydrophobic interactions were found to be the major ones, but electrostatic forces were also present.

Molecular dynamics is a kind of in silico experiment [14] allowing the simulation and visualization of proteins with carbon nanomaterials interactions [15]. Since the protein is a big molecule, its conformational changes may be quite complex. An SWCNT is a nanomaterial consisting of a large number of atoms that altogether require a longer simulation time that demands significant computational power, so it is preferably carried out at a supercomputer facility [16], if possible.

The aim of our research is to investigate the type and contribution of interactions between BSA and SWCNT in aqueous suspension. We suppose both hydrophobic interactions and hydrogen bonds. Tritium-labeled BSA ([³H]BSA) obtained via a tritium thermal activation method was used for the quantification of BSA on SWCNT adsorption, since other analytical techniques are quite inefficient to use for this complex system due to SWCNT’s optical properties and the small quantities of substances under investigation. The simulated SWCNT has a length of about 15.5 nm, which is closer to the real system, because it is comparable with the BSA molecule in size. The trajectories for BSA were chosen according to several possible starting positions of BSA against SWCNT and include the whole pathway of BSA to SWCNT. Computer simulation was carried out in Gromacs 2020.3 software with use of HPC computing resources at Lomonosov Moscow State University [17].

2. Materials and Methods

2.1. Preparation of CNT/BSA Model Structure and Supercomputer Simulation

Bovine serum albumin (PDB:4F5S) as a model protein was obtained from the PDB database. A BSA molecule was prepared with the UCSF Chimera docking program [18]: H₂O molecules, ligand and chain B were removed; hydrogen atoms and charges (force field Amber ff14SB) were added to chain A according to the standard docking preparation procedure. In the model of fluorine-functionalized SWCNT with a length of about 15.5 nm, a pre-optimization of structure was carried out with ChemBio 3D/Draw Ultra 11.0.2 software (Perkin Elmer, Shelton, CT, USA). A pre-optimization of SWCNT was carried out in an MMFF94 force field (50,000 iterations, grad 0.01). Partial atom charges and bonds lengths of SWCNT were obtained after geometry optimization from quantum chemistry data with MOPAC2016 22.234W software (method PM7, EF optimizer, mulliken population analysis).

Molecular dynamics was simulated by Gromacs software with use of HPC computing resources at Lomonosov Moscow State University [17]. NVT and NPT simulations were applied for achieving equilibrium of the system under investigation containing 171,288 solvent molecules (water, TIP3P model). The basic parameters of molecular dynamics were 5360 nm³ (volume of cubic box area for simulation); simulation time of one trajectory ≤ 200 ns; iteration step of 2 fs; number of steps from 100 to 200 million; OPLS-AA force field; temperature 300 K; pressure 1 bar. Berendsen and Nosè–Hoover thermostats, as well as Berendsen and Parrinello–Rahman barostats algorithms were carried out, respectively. The simulated trajectories were chosen wisely and are displayed in Figure 1.

2.2. BSA on SWCNT Adsorption

We used bovine serum albumin (BSA) (Fraction V modified, M_W = 66.5 kDa, Biowest, Lakewood Ranch, FL, United States); fluorine-functionalized single-walled CNTs (SWCNTs) (powder, purity >99 wt %, SA = 570 m²/g (BET), fluorine content 7 ± 2 mol. %, Cheap Tubes, Cambridgeport, VT, USA); graphene oxide (GO) (Cheap Tubes, Cambridgeport, VT, USA); deionized water (Millipore Milli-Q membrane purification system); liquid scintillator (OptiPhase HiSafe3, PerkinElmer, Shelton, CT, USA).

Tritium-labelled [³H]BSA was obtained with a tritium thermal activation method according to [5]. The radioactivity of [³H]BSA was measured with a liquid scintillation counter (RackBeta 1215, LKB Wallac, Turku, Finland). The experimental stages of isotope exchange between tritium atoms and BSA, the adsorption experiment, and the ζ-potential study were described previously [11,19].

3. Results and Discussion

The total energies of the BSA–SWCNT complex in aqueous media for different trajectories (Figure 1) were calculated with means of gmx energy (Figure 2). The trajectories in Figure 2 (I–VI) correspond to the trajectories in Figure 1. The lowest mean total energy is observed for the IV trajectory and consists of both the Coloumb and Van der Vaals interactions and is equal to −646 ± 8 kJ/mol. It indicates that the IV trajectory is the most possible one. We would give the higher priority to the IV–VI trajectories as they were simulated with a Nosè–Hoover thermostat, which is suitable for canonical ensembles [20].

For molecular dynamics simulations, we have chosen the IV trajectory as the most possible one. For the IV trajectory, we observed two bonding sites that were formed consequently at 70 and 100 ns (Figure 3a,b) that correlate with the energy shifts in Figure 2 and indicate bonds formed by 70 and 100 ns, correspondingly. Figure 3a illustrates the first binding site formed by 70 ns, and Figure 3b illustrates the second binding site formed by 100 ns with the first binding site being preserved. Hydrophobic interactions are formed between Pro (67,96), Lys (204,4), His (67,246), Ile202, and Leu (66,103) of BSA and sp²-hybridized carbon atoms of SWCNTs. A hydrophobic bond between His (67,246) and SWCNT is the result of π–π interactions between aromatic cycle of His and conjugated system of the SWCNT surface. Hydrogen bonds are formed between Lys4, Glu100, Gln203 and carbon atoms of SWCNT due to the polarization of the SWCNT surface in the close presence of charged functional groups of BSA amino acids residues [21].

The adsorption isotherm of BSA on SWCNTs is linear (Figure 4a) and may be approximated with an equation analogous to the Henry equation (1) with R² = 0.998 and K_H = 862 ± 11 mL/mg. The adsorption of BSA on mostly hydrophobic SWCNTs (BSA–CNT) is higher than the adsorption of BSA on negatively charged graphene oxide (BSA–GO), indicating the major role of hydrophobic interactions between BSA and SWCNTs. The maximum adsorption of BSA on SWCNTs is 740 mg/mg in the studied BSA concentration range. We observe a ζ-potential decrease from −10 to −16 mV as the BSA adsorption increases on CNT in aqueous media (Figure 4b), while the pristine SWCNTs ζ-potential is close to neutral, indicating their hydrophobic character.

$$\mathrm{\Gamma }=K_{H}C$$

(1)

where

• Γ—BSA on SWCNTs adsorption, mg/mg;
• K_H—adsorption–desorption equilibrium constant, mL/mg;
• C—BSA concentration in equilibrium solution, mg/mL [22].

4. Conclusions

The dominant interactions between BSA and SWCNTs are hydrophobic with two sites of BSA bonded with sp²-hybridized carbon atoms of SWCNTs. Hydrogen bonds between BSA and SWCNT are also present due to the polarization of SWCNTs. The mean total energy of the Coulomb and Van der Waals interactions is −646 ± 8 kJ/mol, as determined by gmx energy that was obtained for wisely chosen trajectories of BSA with starting positions away from SWCNTs. The maximum absorption of BSA on the SWCNTs was quantified with the use of a radiotracer technique with [³H]BSA to be about 740 mg/mg, which is consistent in all the simulations and leads to the ζ-potential of the BSA–SWCNT complex changing from −10 to −16 mV as the adsorption increases. The combination of methods applied may help with the understanding of protein interactions with carbon nanomaterials that may lead to their sophisticated surface modification for drug delivery or creating biocompatible materials. Also, the obtained results may contribute to an understanding of protein corona on carbon nanomaterials formation in biological media.