Protein tertiary structure

In biochemistry and molecular biology, the tertiary structure of a protein or any other macromolecule is its three-dimensional structure, as defined by the atomic coordinates.

Relationship to primary structure
Tertiary structure is considered to be largely determined by the protein's primary structure - the sequence of amino acids of which it is composed. Efforts to predict tertiary structure from the primary structure are known generally as protein structure prediction. However, the environment in which a protein is synthesized and allowed to fold are significant determinants of its final shape and are usually not directly taken into account by current prediction methods. Most such methods do rely on comparisons between the sequence to be predicted and sequences of known structure in the Protein Data Bank and thus account for environment indirectly, assuming the target and template sequences share similar cellular contexts. Stanford University's Folding@Home project is a distributed computing research effort which uses its approximately 5 petaFLOPS (~10 x86 petaFLOPS) of computing power to attempt to model the tertiary (and quaternary) structures of proteins, as well as other aspects of how and why proteins fold into the inordinately complex and varied shapes they take. No currently existing algorithm is yet able to consistently predict a proteins' tertiary or quaternary structure given only its primary structure; learning how to accurately predict the tertiary and quaternary structure of any protein given only its amino acid sequence and the pertinent cellular conditions would be a monumental achievement. Although this ambitious goal has yet to be achieved, researchers have discovered how to combine several of the best of Folding@Home's algorithms to accurately predict the folded structure of some proteins under certain conditions. The calculations performed by the algorithms are constantly evolving, increasing in complexity and nuance, and involve enormous numbers of variables. These techniques are superficially comparable to weather models that show hurricane storm tracks; each of several algorithms independently models a complex system (the weather, in this case) somewhat differently from each of its sister weather algorithms, and the average of all the algorithms' output is taken to be the most likely "storm track". The shape of proteins can be elucidated through a somewhat similar process.

Researchers are also interested in proteins that can fold into more than one stable configuration; protein aggregation diseases such as Alzheimer's Disease and Huntington's Disease as well as prion diseases such as Mad Cow disease can be better understood by constructing (and deconstructing) disease models; the most common way of doing this is by developing a way of inducing the desired disease state in test animals (administering MPTP to give the animals Parkinson's disease, or knocking out a gene essential for the prevention of certain tumors from the animals' genomes ). Folding@Home allows for the modelling of disease states that are not as easily induced, without the need for test animals. Perhaps more importantly, fully human proteins encoded by fully human genes can be used without any of the ethical problems that arise in studying living human beings. Due to its enormous flexibility, which has only briefly been discussed here, coupled with its ability to improve over time, Folding@Home and projects like it are quickly becoming indispensable tools among researchers from a broad variety of disciplines. The possibilities in medicine, biology, pathology, nuclear physics, and other scientific disciplines should a reliable way to accurately model the final tertiary or quaternary structure of human proteins are almost limitless. Proteins, due to the precise conformations they fold into, are nature's original nanomachines; developing an inexpensive and practical way to design and target proteins would completely revolutionize medicine and would have incredibly far-reaching implications. The significance of such a discovery cannot be overstated. To date, over 78 scientific papers have been published on discoveries that relied on Folding@Home.

Determinants of tertiary structure
In globular proteins, tertiary interactions are frequently stabilized by the sequestration of hydrophobic amino acid residues in the protein core, from which water is excluded, and by the consequent enrichment of charged or hydrophilic residues on the protein's water-exposed surface. In secreted proteins that do not spend time in the cytoplasm, disulfide bonds between cysteine residues help to maintain the protein's tertiary structure. A variety of common and stable tertiary structures appear in a large number of proteins that are unrelated in both function and evolution - for example, many proteins are shaped like a TIM barrel, named for the enzyme triosephosphateisomerase. Another common structure is a highly stable dimeric coiled coil structure composed of 2-7 alpha helices. Proteins are classified by the folds they represent in databases like SCOP and CATH.

Stability of native states
The most typical conformation of a protein in its cellular environment is generally referred to as the native state or native conformation. It is commonly assumed that this most-populated state is also the most thermodynamically stable conformation attainable for a given primary structure; this is a reasonable first approximation but the claim assumes that the reaction is not under kinetic control - that is, that the time required for the protein to attain its native conformation before being translated is small.

In the cell, a variety of protein chaperones assist a newly synthesized polypeptide in attaining its native conformation. Some such proteins are highly specific in their function, such as protein disulfide isomerase; others are very general and can be of assistance to most globular proteins - the prokaryotic GroEL/GroES system and the homologous eukaryotic Heat shock proteins Hsp60/Hsp10 system fall into this category.

Some proteins explicitly take advantage of the fact that they can become kinetically trapped in a relatively high-energy conformation due to folding kinetics. Influenza hemagglutinin, for example, is synthesized as a single polypeptide chain that acts as a kinetic trap. The "mature" activated protein is proteolytically cleaved to form two polypeptide chains that are trapped in a high-energy conformation. Upon encountering a drop in pH, the protein undergoes an energetically favorable conformational rearrangement that enables it to penetrate a host cell membrane.

Many serpins (serine protease inhibitors) are metastable, and undergo a conformational change when a loop of the protein is cut by a protease.

Experimental determination
The majority of protein structures known to date have been solved with the experimental technique of X-ray crystallography, which typically provides data of high resolution but provides no time-dependent information on the protein's conformational flexibility. A second common way of solving protein structures uses NMR, which provides somewhat lower-resolution data in general and is limited to relatively small proteins, but can provide time-dependent information about the motion of a protein in solution. Dual polarisation interferometry is a time resolved analytical method for determining the overall conformation and conformational changes in surface captured proteins providing complementary information to these high resolution methods. More is known about the tertiary structural features of soluble globular proteins than about membrane proteins because the latter class is extremely difficult to study using these methods.

Interactions stabilizing tertiary structure

 * Disulfide bonds
 * Hydrophobic interactions
 * Hydrogen bonds
 * Ionic bonds

History
Since the tertiary structure of proteins is an important problem in biochemistry, and since structure determination is relatively difficult, protein structure prediction has been a long-standing problem. The first predicted structure of globular proteins was the cyclol model of Dorothy Wrinch, but this was quickly discounted as being inconsistent with experimental data. Modern methods are sometimes able to predict the tertiary structure de novo to within 5 Å for small proteins (<120 residues) and under favorable conditions, e.g., confident secondary structure predictions.