AmericanChemicalSociety.com

Transport of liquids through nanometer size structures is a ubiquitous process in modern biomedical engineering. The proposed applications of nanofluidics include lab-on-a-chip technology,  epigenetic information analysis, biosensors, protein crystallization,  drug delivery systems, and electronics. To transport biomolecules through nanoscale structures, hydraulic pressure gradients and/or electric fields are imposed across the device elements. Today, it is already possible to fabricate macroscopic length channels of a sub-ten nanometer cross-section, and assemble them into regular macroscopic arrays. However,  while reducing the size of a nanofluidic channel creates new opportunities for high-precision manipulation of biological macromolecules, it also makes the performance of the entire nanofluidic system more susceptible to undesirable interactions between the transported biomolecules and the walls of the channel.

Here we report the results of an ongoing theoretical and computational study that aims to develop a comprehensive model of protein adsorption to the inorganic surfaces in nanofluidic systems. In the simplest theoretical model of protein adsorption, a protein is assumed to bind to the surface with a certain probability so that the fractional coverage of the surface θ is described as a Langmuir isotherm θ = α C/(1 + α C), where C is the protein concentration and alpha is the Langmuir adsorption constant. The Langmuir constant incorporates all the effects of the surface properties of the channel and protein, and can be empirically associated with quantities such as the total surface area of the protein, and the charge, roughness and hydrophobicity of the channel surface. While being of great practical use for characterization of macroscopic surfaces, measuring such an empirical dependence is not always possible for submicron-size objects. Furthermore, carrying out such measurements for all possible combinations of the inorganic materials and all the biomacromolecules of a living cell formidable is practice. 

Ideally, one would want to predict the macroscopic behavior of the protein flow through a nanochannel taking into account the microscopic properties of the protein and the nanochannel material. Molecular dynamics (MD) is a computational method that can accurately predict the forces between biomolecules and inorganic surfaces in water if their atomic structures are known. Using this method, we have carried out the first (to our knowledge) simulations of pressure-driven flow through silica nanochannels that characterized, with atomic resolution, adsorption of a model protein to the nanochannel surfaces. Although the simulated adsorption of the proteins was found to be nonspecific, it had a dramatic effect on the rate of the protein transport. To determine the relative strength of the protein–silica interactions in different adsorbed states, we simulated flow-induced desorption of the proteins from the silica surface. Our analysis of the protein conformations in the adsorbed states did not reveal any simple dependence of the adsorption strength on the size and composition of the protein–silica contact, suggesting that the heterogeneity of the silica surface may be an important factor. Reducing or eliminating such nonspecific binding is one of the most challenging problems in the development of the next generation of nanofluidic devices for applications in biotechnology.

Since most proteins are made up of just twenty types of amino acids and a limited number of their modifications, it might be possible to reduce the gargantuan task of predicting the binding affinity of a protein in a random conformation on a heterogeneous inorganic surface to the characterization of the interactions of individual amino acids with the surface of interest.  By combining such information with the knowledge of the atomic structure of the protein, the binding affinity of the entire protein could be described in statistical terms, which would have numerous applications in bionanotechnology. To demonstrate that such an approach is, in principle, possible, we characterized the interactions of individual amino acids with a featureless surface (the so-called phantom surface) which models a frictionless hydrophobic material. To determine the binding free energy for each of the twenty amino acids, we have computed the potential of mean force of individual amino acids as a function of the distance between the amino acid and the phantom surface. The resulting potentials revealed little or no binding for charged and polar amino acids, whereas amino acids containing long nonpolar side chains or aromatic rings displayed strong affinity to the phantom surface.

Future studies will be directed toward expanding the approach outlined in this proof-of-principle demonstration to more realistic heterogeneous surfaces. While being computationally expensive, the massively parallel character of the calculations makes it possible to carry out such studies using currently available computational resources. The availability of atomic-scale models of inorganic surfaces in an aqueous environment would be critical to the success of such an approach, as heterogeneous nature  of such surfaces could only then be accurately taken into account.