Identifying Thermostabilizing Mutations in Membrane Proteins by Bioinformatics

Main Text

Membrane protein structures are being determined at an ever-accelerating rate, reflecting parallel increases in both our understanding of how to generate well-diffracting crystals and the amount of resources worldwide dedicated to membrane protein structural biology. Stability of detergent-solubilized membrane proteins has long been recognized as a prime factor in dictating the probability of crystallizing any given membrane protein (1) and has underpinned many of the advances made recently. For example, this was the basis for the development of a facile screening system for the rapid identification of stable membrane proteins by fusing targets to GFP and screening the unpurified detergent-solubilized membrane proteins by fluorescence-detection size exclusion chromatography (FSEC) (2). Using this or similar strategies, a wealth of structures have been determined that has provided the foundation for a molecular understanding of transporters and ion channels.

Screening for stable membrane proteins is a successful strategy, but what happens if a stable homolog is not found? What strategies are available if the structure of a specific transporter is required? Recently, protein engineering has been used to solve these difficulties through strategies that are designed to thermostabilize the target membrane protein through the introduction of point mutations that improve stability (3). This technology has been most successfully applied to G protein-coupled receptors, which are far less stable than most self-respecting bacterial transporters. The first demonstration that a previously intractable target could be thermostabilized and its structure determined was the β₁-adrenergic receptor (4), which is the target for drugs such as β-blockers, used in the treatment of heart problems. The use of thermostabilizing mutations has now facilitated the structure determination of at least nine different eukaryotic membrane proteins that were previously unattainable through standard techniques.

So, if thermostabilization of a membrane protein can ensure structure determination of any membrane protein, why is this strategy not used more widely? The difficulty lies in determining which 4–6 mutations will improve thermostability. As of this writing, the process involves making 300–400 mutants and testing their thermostability using high-affinity ligands and/or FSEC. However, this methodology is very expensive, time-consuming, and laborious, especially when used to thermostabilize a large membrane protein such as a mammalian serotonin transporter that contains 630 amino-acid residues (5). Thus there is a significant advantage in finding a cheap, rapid methodology for identifying thermostabilizing mutations in silico. However, this has proven extremely challenging even if a structure is available (6), which of course will not be the case if the aim of the thermostabilization procedure is to determine the structure! This is where the work of Sauer et al. (7), published in this issue of the Biophysical Journal, enters new territory for membrane proteins. They rationalized that, within any large family of membrane proteins found in organisms with different optimum growth temperatures, there will be specific mutations that underlie thermostabilization—which will therefore be more conserved in proteins from organisms growing at high temperatures. As proof of principle, Sauer et al. (7) worked on the tetracycline antiporter from Bacillus subtilis, BsTetL. Two different methodologies were tested for their ability to identify potentially thermostabilizing mutations, referred to as the global analysis method and the pairwise method. In the global analysis strategy, a total of 2343 homologs were identified, and after removing very close homologs, 1513 sequences were analyzed that included 140 sequences from thermophiles. Out of this analysis, 10 mutations that correlated well with thermostability were identified. The pairwise method used all 2343 homologs in the initial dataset (154 thermophiles) and identified 15 potentially thermostabilizing mutations. Interestingly, the two methods identified nonoverlapping sets of amino-acid residues that were found throughout the protein (Fig. 1). All these substitutions were made in BsTetL, and the thermostability of the detergent-solubilized mutants was assessed using FSEC. Out of 22 mutants tested for thermostability (three of the potentially thermostabilizing amino-acid residues were already in BsTetL), seven proved to increase the thermostability of BsTetL. This is a remarkable hit rate, considering there was no experimental data to guide the selection, and it represents an extremely cost-effective strategy to start thermostabilizing a membrane protein.

Figure 1.

Position of thermostabilizing mutations in a model of BsTetL.

The noteworthy success of Sauer et al. (7) in finding thermostabilizing mutations for BsTetL is an excellent start to building a thermostable protein. However, much more work may be required to improve the thermostability of BsTetL to ensure a crystal structure is obtained, because crystallization may require the use of harsher detergents than the dodecylmaltoside that was used in this study. This will require combinations of mutations. Sauer et al. (7) have already experienced loss of expression and incompatibility of mutations where two or more mutations are combined, which is commonly observed during the thermostabilization of other membrane proteins (8). There is also a question regarding whether the transporter should be stabilized as a functional transporter or whether it should be stabilized in a single conformation, as has proven so successful for G-protein-coupled receptors. Is crystallization prevented by conformational dynamics, poor thermostability, or all of these factors? Sauer et al. (7) emphasizes that many of the single mutations identified in BsTetL do not affect transport, but from our experience with G-protein-coupled receptors, I would maintain that reducing conformational dynamics during crystallization is advantageous, providing that the protein is folded correctly as defined by substrate/inhibitor binding. It was interesting to find that the thermostabilized serotonin transporter was biased toward a single conformation that can bind cocaine analogs and serotonin with high affinity, but it cannot transport serotonin (5). The converse is probably not true, because the identification of transport-negative mutants in the tetracycline antiporter TetA(B) that still bind tetracycline with high affinity did not identify thermostabilizing mutations (9). Thus considerable further development of the transporter toolbox is necessary, both from computational and experimental perspectives, to generate a series of complimentary approaches that will allow us to determine the structure of any transporter.

Editor: Hagan Bayley.