An Insoluble Problem?
The challenges of crystallizing membrane proteins—and how they’re being overcome
Membrane proteins represent only a handful of the total number of protein structures defined to date. Yet these proteins, which represent nearly 40 percent of all known proteins, including receptors, channels, and signaling molecules, are essential for cell communication and their malfunctions are implicated in many diseases. Structure-based design is one powerful way of developing drugs tuned to the precise actions and minimal side effects required for effective treatments. X-ray crystallography—still the only general method for solving the atomic structures of proteins of any size—has been hampered by the extreme difficulty of preparing and crystallizing pure membrane proteins.
The problem is a practical one: hydrophilic proteins, such as those in the cytoplasm, can form crystals in solution relatively easily, but membrane proteins also have hydrophobic parts that buoy the protein in the lipid layer. To maintain their shape, these lipid-loving domains must be surrounded by components that resemble the natural membrane—a requirement that makes it difficult to grow well-diffracting crystals. However, an array of technical advances over the last 2 years has advanced our ability to determine these structures.
Advances are the result of developments at multiple steps in the crystallization process. One example comes from Raymond Stevens and colleagues at the Scripps Research Institute who discovered that lipids were essential for determining the structure of a G protein-coupled receptor (GPCR) that responds to adrenaline. When Stevens tried to crystallize his GPCR, he found that cholesterol molecules were necessary for crystal formation, and from the structure, showed that cholesterol also acted as glue between the dimeric receptor molecules.1 This gave a structural explanation for the observation that cholesterol in the membrane is essential for the dimerization, and hence the signaling function, of this receptor.
The choice of detergent used to isolate membrane proteins can also have a profound effect on their ability to crystallize. Classical detergents, such as beta-octyl glucoside, create relatively large lipid globules, called micelles, which contain a single layer of phospholipids with their single-chain tails facing inward. Large micelles increase the ratio of lipid to protein, making it difficult to pack the protein to a density sufficient for crystal formation. Stevens collaborated with chemists to design new detergents, such as the cholate- based amphiphiles, that create smaller micelles, allowing the protein molecules to pack more closely and form better crystals.2
A third factor that can improve crystallization is the origin of the protein. The volume and quality of proteins produced depends on the organism, cell type, promoter, and vector used to generate them. However, there is no way of knowing in advance which species or expression method is going to yield a well-behaved protein. Because many membrane proteins are expressed by multiple species, homologous genes should be tested and screened to identify proteins that are most amenable to crystallization.3 Researchers can now use high-throughput methods to screen large numbers of expression and purification conditions, which helps to speed the process.4
A remarkable example of such optimization comes from Rod MacKinnon’s lab at The Rockefeller University. Interested in the mechanism of a CLC chloride transporter protein, they tried to improve the data they obtained from their crystal by replacing a naturally occurring amino acid with a methionine at each of 30 sites in the protein. They then tagged the methionines with the heavy element selenium, which helped them confirm the atomic structure of the protein with better precision.
Chopping off the termini of proteins is a trick long-used to make soluble cytosolic proteins more amenable to crystallization. Similarly, removing the hydrophilic and flexible ends of a membrane protein can improve crystal formation. By excising flexible regions of human aquaporin 4, a water channel implicated in the ALS-like autoimmune disease neuromyelitis optica, we obtained a high-resolution structure.5 Another approach is to insert mutations that alter the protein sequence in a manner that stabilizes one particular conformation, and screen to find those that might rigidify otherwise flexible regions.
More drastic engineering—wholesale gene redesign—can make it possible to solve structures of proteins that were not amenable to study by other means. Some amino acids have multiple codons—the triplets of nucleotides that make up the genetic code—and different species prefer to use different equivalent codons to make the same protein, an effect known as codon bias. Because codon bias affects translation efficiency, taking a gene from one organism and expressing it in a different one can improve protein production. We recently engineered a Plasmodium falciparum aquaporin gene to express in the plasma membrane fraction of Escherichia coli. But we altered the Plasmodium gene sequence to use the codons favored by E. coli, a process known as codon optimization. Although the DNA sequence changes, the protein sequence remains the same.6
The dramatic technical progress of the last few years has lent new energy to the prospects for determining structures of membrane proteins and how they function. We can now begin to approach the mechanisms of transmembrane processes implicated in a variety of illnesses, such as cancer, diabetes, schizophrenia, and depression, where signaling defects in membrane proteins have previously been difficult to study at the level of atomic structures.
Robert Michael Stroud is at the University of California, San Francisco.
|This article is adapted from a review in F1000 Biology Reports.
DOI: 10.3410/B3-8 (open access at http://bit.ly/NewStruc).
For citation purposes, please refer to that version.