We begin our computational analysis of bipolar disorder with molecular modelling techniques, specifically targeting the cell membrane and its components.
The cell membrane is made up of two layers of lipids that surround the cell and proteins that provide protection and signaling for the cell. The membrane is selectively permeable, meaning only certain substances are allowed to come inside the cell while certain substances are banned from coming in. The soft assembly of lipid molecules arrange together in a double layer to isolate their water-hating (hydrophobic) part from their water-loving (hydrophilic) part. The hydrophilic part then faces the aqueous inside and outside of the cell. Meanwhile, transmembrane (TM) proteins can be channels or pores that allow materials to cross the membrane when they are required.
Using hydropathicity plots, we can display the hydrophobic and hydrophilic regions of a protein sequence, which we can then use to predict protein structure. Specifically, we can identify sequences of amino acids that form transmembrane helices – parts of the protein that span the lipid membrane.
This hydropathicity plot for ADCY2 (UniProt Q08462, 1091 aa) was calculated using Kyte & Doolittle’s (1982) hydrophobicity scale with a 19-residue window. A score greater than zero means the region is hydrophobic and positive peaks correspond to potential TM helices. The ExPASy ProtScale hydrophobicity profile below suggests that there are at least 12 membrane-spanning helices in ADCY2, with the major peaks indicating the approximate positions of these helices.
More recently, Kapcha & Rossky (2014) have developed a new hydrophobicity scale in an effort to better understand and rapidly characterise water-protein interactions. They introduce a simple binary but atomic-level method that allows for the classification of polar and non-polar moieties within single residues, including backbone atoms. The following surface plots shows the protein binding interface of ADCY2 based on Kyte & Doolittle’s residue-based method (left) and Kapcha & Rossky’s atomic-level method (right). The most hydrophobic residue ILE, with a hydrophobicity value of 1.00 will be fully saturated red. The most hydrophilic residue ARG, with a hydrophobicity value of 0.00 will be fully saturated blue. This scale uses 0.50 as the midpoint where the color is white. So, as the hydrophobicity value gets closer to 0.50 the saturation of the red or blue color decreases.
Surface hydrophobocity maps for ADCY2 showing good agreement between the residue-based and atomic-level methods
We have also used the MEMSAT-SVM server, which relies on neural networks and support vector machines, to predict the transmembrane helix domains of ADCY2, as shown below.
Three-dimensional analysis of protein structures continue to play a critical role in biological and medical discovery, providing fundamental insight into function that produces useful biochemistry, and dysfunction that leads to disease. Despite rigorous laboratory attempts to characterise the complete structural proteome, there remains a large number of proteins that cannot be resolved experimentally, either by X-ray crystallography or NMR. Fortunately, in silico structure prediction, namely homology or comparative modelling has allowed the biological disciplines to achieve reliable and accurate models based on template selection, sequence and spatial alignment, structural refinement and validation, and other computational techniques.
Among the several automated homology modelling servers available I-TASSER (Roy, 2010) was the preferred option due its consistent top ranking according to the Critical Assessment of Techniques for Protein Structure Prediction (CASP). Moreover, full-length structure models are constructed by reassembling fragments from threading templates using replica exchange Monte Carlo simulations. The top 5 structures predicted by I-TASSER for human ADCY2 are shown below.
C-score is a confidence score for estimating the quality of predicted models by I-TASSER. It is calculated based on the significance of threading template alignments and the convergence parameters of the structure assembly simulations. C-score is typically in the range of [-5,2], where a C-score of higher value signifies a model with a high confidence and vice-versa.
TM-score and RMSD are known standards for measuring structural similarity between two structures which are usually used to measure the accuracy of structure modeling when the native structure is known. In cases where the native structure is not known, it becomes necessary to predict the quality of the modeling prediction, i.e. what is the distance between the predicted model and the native structures? To answer this question, I-TASSER predicted the TM-score and RMSD of the predicted models relative the native structures based on the C-score.
In a benchmark test set of 500 non-homologous proteins, we found that C-score is highly correlated with TM-score and RMSD. Correlation coefficient of C-score of the first model with TM-score to the native structure is 0.91, while the coefficient of C-score with RMSD to the native structure is 0.75. We only report the quality of prediction (TM-score and RMSD) for the first model, because the correlation between C-score and TM-score is weak for the lower-rank models. However, we list the C-score of all models just for reference.
Molecular dynamics (MD) simulation of membrane proteins requires the setup of an accurate representation of lipid bilayers. The CHARMM-GUI Membrane Builder will be used to achieve this as it provides a web-based user interface designed to interactively build all-atom protein/membrane or membrane-only systems through an automated optimized process (Wu, 2014).
For our protein/membrane system we have chosen the DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine) lipid molecule.
The final assembly contained 364 lipids on the top layer and 348 lipids on the bottom layer, with a total of 337801 atoms, including the protein and the water molecules.
Protein embedded in lipid bilayer
All cells have a transmembrane electric potential across the plasma membrane. The potential has two basic functions, namely
- To provide a driving force for operating the numerous molecular machines embedded in the membrane
- To transmit signals in between “excitable” cells such as neurons and muscle cells
The potential is due to an ion concentration imbalance across the membrane established by the membrane protein Na+/K+-ATPase. These active transport proteins each continuously expel three sodium ions out of the cell and at the same time import two potassium ions. This charge separation thus gives rise to an electrical potential, which is about -60 mV. The excitability of cells then depends voltage-gated ion channels (VGIC), which are signal transducers that provide a regulated path for the movement of inorganic ions such as Na+, K+, Ca2+ and Cl– across the plasma membrane in response to various stimuli (Bjelkmar, 2011).
Of particular interest to our research are VGICs that produce neuronal action potentials. Signalling in the nervous system is accomplished by networks of neurons from one end of the cell through to the axon.
Essential to this signalling mechanism are:
- Voltage-gated Na+ channels
- Voltage-gated K+ channels
- Voltage-gated Ca2+ channels
These channels may be open or closed depending on whether the associated receptor has been activated by the binding of its specific ligand (such as a neurotransmitter) or by a change in the transmembrane electric potential (Delemotte, 2008).
A Simple Method for Displaying the Hydropathic Character of a Protein. Journal of Molecular Biology
A Simple Atomic-Level Hydrophobicity Scale Reveals Protein Interfacial Structure. Journal of Molecular Biology
I-TASSER: a unified platform for automated protein structure and function prediction. Nature Protocols
CHARMM-GUI Membrane Builder toward realistic biological membrane simulations. Journal of Computational Chemistry
Modeling of voltage-gated ion channels. PhD dissertation
Modeling membranes under a transmembrane potential. Journal of Physical Chemistry B