The Design and Structure Prediction of Protein Oligomers

Abstract

The minimum free energy state of a protein (the native state) is encoded by its amino-acid sequence. Due to the many torsional degrees of freedom (DOF) available to a polypeptide chain, a vast number of conformations is possible. Therefore, to predict the native state of a protein directly from sequence, a computer algorithm must evaluate a large number of possible conformations using accurate scoring functions. Due to their larger size and the presence of extra rigid-body DOF, protein oligomers present additional challenges to structure prediction algorithms. These problems are somewhat alleviated if the simulated system possesses some form of symmetry, which limits the sampling of its configuration space and makes the folding simulation computationally tractable. First we focused on developing a method for the structure prediction of a special class of symmetrical protein oligomers: the coiled coils. The major challenge faced was to come up with a scheme that could discriminate between the native and the multiple alternate oligomeric states. We tested several approaches to predict both native subunit orientation and oligomeric state, and found that the free energy of helix folding may be an important factor in determining oligomer stability. Our most successful prediction approach was able to correctly predict native oligomeric states (and topologies) for 23 out of 33 coiled coils in a benchmark set. The accuracy of our prediction method was further evaluated by examining whether the obtained structural models could be used to determine the structures of crystallized coiled coils using molecular replacement (MR) phasing techniques. To that end, we implemented an automated structure-solving pipeline (CCsolve) that combines MR, model building, and refinement. We found that our de novo coiled-coil models were sufficiently accurate to enable effective structure determination for nearly all of the 24 test cases in our benchmark set. Somewhat inverse to the structure prediction problem is the problem of protein design. Here the goal is to find an amino-acid sequence that is most compatible with a desired backbone conformation. First we focused on the design of higher-order coiled coils and how their innate structural and energetic similarities could be exploited to create conformational switches (i.e. proteins that interconvert between two or more distinct structural states as a response to an external stimulus). To that end we designed a de novo peptide sequence (termed pHios) that switches between a symmetrical pentameric and a new type of hexameric assembly as a function of pH. We then focused on designing a protein-based nanocage that could be disassembled and reassembled in order to enable the encapsulation of various cargo molecules. Such a system would be useful as a drug delivery device, as well as a nanobioreactor for the study of enzyme catalysis in confined space. We used the icosahedral Hepatitis B virus capsid as a scaffold, and introduced an affinity motif that enabled the encapsulation of specifically tagged cargo proteins. We subsequently showed that our designed encapsulation system was able to load significant quantities of guest molecules, and thereby demonstrated its potential in the abovementioned applications.

Abstract (Swedish)

Popular Abstract in English

Proteins are a class of biological macromolecules that perform countless fundamental and diverse functions in living organisms. Chemically speaking, they are polymers built out of monomeric units called amino acids. All biological proteins are constructed out of 20 different amino-acid types. To form protein molecules, amino acids are linked together in a linear chain like pearls on a string. The physical and chemical characteristics of proteins are determined by the sequence of amino acids. Some proteins play structural roles, serving as building blocks for many cellular architectures, while others act as regulators, signal receptors, or enzymes that catalyze various chemical transformations necessary for life.

To carry out its biological function, a protein must “fold” into a specific three-dimensional structure. This structure is determined solely by the protein’s amino-acid sequence, which means that it could in theory be readily predicted. Such an endeavor would be extremely beneficial for many biological, medical, as well as industrial applications. This is because the information about protein structure is essential in research areas that aim to develop various protein-targeting drugs, as well as in research areas that aim to elucidate protein interaction networks (an important piece of information when treating various diseases for example). As experimental determination of protein structures is extremely laborious, time consuming, and costly, structure prediction offers an attractive alternative for such research endeavors. Furthermore, accurate structure prediction methods are a prerequisite for efficient protein design, the ultimate goal of which is to create new protein structures with predefined functions. As such, the field of protein design holds incredible potential for revolutionizing medicine, biology, as well as nanotechnology. However, problems faced by both protein structure prediction and protein design remain largely unsolved. In this work, we tackled some of the issues that plague these fields of research, with a specific focus on protein oligomers, which are structures assembled from multiple protein molecules that come together and interact closely.

Protein oligomers are mostly homomeric, meaning that they are composed of multiple copies of the same protein, and almost exclusively possess some form of symmetry. Here we focused on two classes of protein oligomers: coiled coils and icosahedral viral capsids.

Coiled coils are an important class of protein oligomers that perform a number of essential biological functions, including the regulation of gene expression, cell division, viral infection etc. In paper I, we developed a method that can predict the structures and oligomeric states of homomeric coiled coils with approximately 70% accuracy.

In addition to being able to provide structural models of coiled coils, this method can further aid the experimental determination of coiled-coil structures. Without accurate structural models of proteins whose structure we wish to determine experimentally, the structure solving process can be extremely laborious and time consuming. The availability of accurate structural models makes this endeavor much simpler. This is what we did in paper II: used the predicted structures of coiled coils to actually obtain the true, experimentally determined structures.

In paper III, we further designed a coiled-coil oligomerization switch that forms one type of assembly in solutions of low acidity, and a completely different type of assembly in solutions of high acidity. Such a system could be beneficial in synthetic biology as a biosensor for example.

In paper IV, we focused on developing a container capable of encapsulating specific protein cargo. Such systems are of great benefit in medicine for instance, as they can be used to deliver encapsulated drugs and/or imaging probes to specific organs, tissues, and cells.

• author: Lizatovic, Robert
• supervisor: Ingemar André
• opponent: Dr Grigoryan, Gevorg, Dartmouth College, Hanover, USA
• organization: Biochemistry and Structural Biology
• publishing date: 2015
• type: Thesis
• publication status: published
• subject: Bioinformatics (Computational Biology); Structural Biology
• keywords: virus capsids, coiled coils, molecular replacement, symmetry, Rosetta, protein design, structure prediction
• pages: 146 pages
• defense location: Hall B, Kemicentrum, Getingevägen 60, Lund
• defense date: 2015-10-09 10:00:00
• ISBN: 978-91-7422-410-8
• language: English
• LU publication?: yes
• id: d9b8e762-8336-4d1c-9e40-4db87ad45375 (old id 7862636)
• date added to LUP: 2016-04-04 10:26:35
• date last changed: 2018-11-21 20:58:46