Protein structure

• 1. Introduction

  ›

  Protein structure is the three-dimensional arrangement of atoms in an amino acid-chain molecule, and understanding it sits at the heart of modern biology. A chain of amino acids, by itself, tells you relatively little. But fold that chain into a precise shape, and you unlock the difference between an enzyme that drives metabolism and a useless tangle of molecules. That gap between sequence and shape is what structural biology exists to close. How do proteins know which shape to take? What forces guide a flexible chain into something rigid enough to do work? And what happens when that process goes wrong? Those questions run through everything in this story.

• 2. Building Blocks and Bonds

  ›

  Amino acids link together through condensation reactions, and each reaction costs exactly one water molecule. The resulting bond between adjacent amino acids is called a peptide bond. By convention, any chain shorter than 30 amino acids is called a peptide rather than a protein. Longer chains carry a directional identity: one end is called the N-terminus, where a free amino group sits, and the other is the C-terminus. Counting always starts from the N-terminal end.

  Frederick Sanger made a discovery that changed how scientists think about proteins: he determined the sequence of amino acids in insulin. That work established that each protein carries a unique sequence, and that sequence encodes both structure and function. Insulin itself is composed of 51 amino acids arranged across two chains. One chain has 31 amino acids; the other has 20.

  The primary structure of a protein can be read directly from the gene that encodes it, but not entirely. Post-translational modifications such as phosphorylations and glycosylations are considered part of the primary structure, yet they cannot be read from the gene itself. The identity and order of amino acid residues form the foundation on which every higher level of structure is built.

• 3. The Four Levels

  ›

  In 1951, Linus Pauling described two main types of secondary structure: the alpha-helix and the beta-strand, or beta-sheet. These local sub-structures are defined by patterns of hydrogen bonds between groups along the polypeptide backbone. They are constrained to specific values of two angles, called psi and phi, on what is known as the Ramachandran plot. Both forms saturate the hydrogen bond donors and acceptors in the backbone, which stabilizes them.

  Above secondary structure sits tertiary structure, the full three-dimensional shape of a single polypeptide chain. Alpha-helices and beta-sheets fold into a compact globular form. Hydrophobic residues are buried away from water in the core, while hydrophilic residues face outward. Salt bridges, hydrogen bonds, and the tight packing of side chains lock the domain in place. Disulfide bonds also contribute, though they are extremely rare in cytosolic proteins because the interior of the cell is generally a reducing environment.

  Quaternary structure goes one level further, describing what happens when two or more polypeptide chains assemble into a single functional unit. A unit made of two chains is a dimer; three chains form a trimer; four form a tetramer; five form a pentamer. Hemoglobin is a well-known example of a heterotetramer, assembled from two alpha chains and two beta chains.

• 4. Folding and Its Forces

  ›

  Protein folding is the process by which a polypeptide chain collapses from a flexible, unfolded state into its stable, functional three-dimensional form. This native state is generally assumed to be determined by the amino acid sequence alone. That assumption is known as Anfinsen's dogma, a thermodynamic principle holding that the native structure is a unique, stable, and kinetically accessible minimum of free energy.

  The forces doing this work include hydrophobic interactions, hydrogen bonding, van der Waals forces, and Coulomb interactions. Hydrophobic collapse is a key step: the number of hydrophobic side-chains exposed to water is minimized, driving them into the protein's core. The result is a globular protein with a hydrophobic interior and a hydrophilic surface.

  Not all folding happens after the chain is fully synthesized. In some cases, folding begins co-translationally, while the ribosome is still building the chain. In other cases, specialized helper proteins called chaperones assist the process. The thermodynamic stability of a soluble globular protein is usually measured as the free energy difference between its folded and unfolded states, and that value typically does not exceed 50 kJ/mol.

• 5. Proteins in Motion

  ›

  Myosin, kinesin, and dynein are motor proteins, and each moves cargo through the cell in a different direction or by a different mechanism. Myosin is responsible for muscle contraction. Kinesin carries cargo away from the nucleus along structures called microtubules. Dynein moves cargo toward the nucleus and also drives the beating of motile cilia and flagella. These proteins are possible only because proteins are not rigid objects; they shift between conformational states as they work.

  A motile cilium has been described in published literature as a nanomachine composed of perhaps over 600 proteins in molecular complexes, many of which also function independently as nanomachines. Flexible linkers allow mobile protein domains to recruit binding partners and induce long-range changes in shape through a process called allostery.

  Some proteins take this flexibility to an extreme. Intrinsically disordered proteins exist and function in a state that lacks a stable tertiary structure. They resist description by any single fixed shape. To model these proteins, researchers use conformational ensembles, collections of multiple possible structures that together represent the protein's behavior. Ensemble files have been generated for proteins including Sic1/Cdc4, p15 PAF, MKK7, Beta-synuclein, and P27.

• 6. Determining Structure

  ›

  Around 90% of the protein structures in the Protein Data Bank have been determined by X-ray crystallography. The method works by measuring the three-dimensional density distribution of electrons in a crystallized protein sample, then using that data to infer the coordinates of every atom. Nuclear magnetic resonance techniques account for roughly 7% of known structures.

  For very large complexes, cryo-electron microscopy offers an alternative. Its resolution has historically been lower than X-ray crystallography or NMR, but it continues to improve, and it is particularly valuable for structures like virus coat proteins and amyloid fibers. Circular dichroism can determine the general secondary structure composition of a sample. Two-dimensional infrared spectroscopy has become a valuable method for studying flexible peptides that resist other approaches.

  Once a structure is experimentally determined, the work is not necessarily finished. Molecular dynamic simulations can probe the structure computationally, exploring how it behaves over time. A technique called fast parallel proteolysis, or FASTpp, can probe the structured fraction of a protein and its stability without requiring purification first. Computational prediction methods, including threading and homology modeling, can also build three-dimensional models for proteins whose structures have not been directly measured.

• 7. Classification and Prediction

  ›

  Despite roughly 100,000 different proteins being expressed in eukaryotic systems, the number of distinct structural domains, motifs, and folds is much smaller. The Structural Classification of Proteins database and the CATH database each provide their own independent classification scheme. When two proteins share significant structural similarity, that overlap is treated as evidence of a common evolutionary ancestor.

  Domains are independently stable structural units that often fold separately from the rest of the protein. Because of that independence, domains can be transferred between proteins through genetic engineering to create chimera proteins. A combination of a protein tyrosine phosphatase domain and a C2 domain that appears across different proteins has been called a superdomain, on the theory that it evolves as a single unit.

  Predictive machine learning approaches to structure work at four levels. The first level predicts secondary structure and solvent accessibility from the sequence alone. The second level maps distances and contact points along the chain. The third level estimates the full atomic coordinates of the protein. The fourth level predicts complexes of multiple proteins together. Progress across all four levels is assessed at the biannual Critical Assessment of Structure Prediction event, which tracks how well computational methods are advancing against known experimental benchmarks.