Overview of the work of the BBSRC's Membrane Protein Structure initiative

In 2004, the Biotechnology and Biological Sciences Research Council (BBSRC), as part of their Structural Proteomics of Rational Targets (SPoRT) initiative, funded the MPSi (Membrane Protein Structure initiative) consortium. MPSi was formed from most of the main UK university-based groups involved in the determination of the structure and function of membrane proteins. It contains groups from the Universities of Glasgow, Leeds, Sheffield, Oxford, Manchester, Imperial College, Birkbeck College and Daresbury Laboratory. The main aim of this consortium was, using predominantly prokaryotic transporters and ion channels, to establish high-throughput methods designed to overcome the main bottlenecks in the processes involved in determining the structures of membrane proteins. At the same time there was the intention to embed the full range of skills required to breech these bottlenecks in each of the participating laboratories. The adoption of high throughput strategies requires automation and the MPSi grant funded state-of-the-art robotics to help achieve this.

A key reason why the structural biology of membrane proteins lags so far behind that of water-soluble proteins is that, at each stage of the process required to go from proteins to crystal structures, there are significant hurdles to be overcome, relative to the situation with water soluble proteins.

Very many membrane proteins are only present in their natural membranes at very low concentrations, too low to allow their isolation in sufficient quantities for structural studies. So, the first hurdle is to develop suitable, robust methods for overexpression. Within MPSi, initially, most of the expertise in overexpression, especially in the area of a wide range of prokaryotic transporters, resided in Leeds. Their recent work carried out as part of MPSi and their recommendations of how to optimize the chances of success in overexpressing in Escherichia coli are described here in the paper by Deacon et al.

Later on during the period of the MPSi grant the use of GFP labelled membrane proteins (Drew et al. [1]), which allows much easier detection of the over-expressed product, was introduced. This approach is reported in the paper by McLuskey et al. It is still not possible always to be able to overexpress any given membrane protein successfully, even if it is of prokaryotic origin. However, our experience in the endeavour has allowed us to produce some guidelines that offer quick and efficient methods with which to rapidly assess whether the protein of interest will over-express, or whether another orthologue should be tried. Once conditions have been optimized large scale expression is then required, usually achieved using fermentors. Now dealing with large volumes (10–40 litres) can be a bottleneck. The paper by Roach et al. describes how this can be overcome.

Once membranes have been obtained that contain sufficient quantities of the required membrane protein, it must then be solubilized, purified and checked for activity. This process can be very daunting. What detergent should be used? What purification method? How can the native structure be confirmed in an isotropic environment, rather that in the membrane? Even if the protein is pure and fully active is it mono-disperse? A mono-disperse sample is an important prerequisite for crystallization. When choosing overexpression strategies, it is wise to produce the over- expressed protein in a ‘tagged’ form, where the tag, e.g. His₆, can then be used to facilitate purification. Choices of solubilizing detergents must then be compatible with the use of the tag.

As described here by McLuskey et al., it is often convenient to start with DDM at about 1.5% w/v as the solubilizing detergent. Searching the literature suggests that this is a good choice. However, the sample size in the literature is very small and as a consequence these statistics are not very robust. If possible, it probably wise to try three or four different detergents that have significantly different characteristics and are preferably cheap. Once the membrane protein has been solubilized then, if it is tagged, affinity chromatography followed by gel filtration are useful ways to start the purification. If the overexpression is carried out in E. coli ArcB can be a really problematical contaminant (it even then crystallizes very well!). Indeed there have been several reports of ArcB being crystallized by mistake. ArcB can be readily removed if the His-tagged protein is removed from the Ni-column by proteolytic cleavage. ArcB then remains on the column Drew et al. (this issue).

If the purified protein has a convenient activity assay, this can now be carried out to check that it is native. However, if this is not possible, then indirect methods must be used. Here, the papers by O'Reilly et al., Cleverley et al. and Kean et al. describe indirect methods aimed at tackling this question.

If the protein is pure and active, the best thing to check before going to crystallization is whether it is monodisperse. This can be done most easily by either electron microscopy, or dynamic light-scattering or analytical size exclusion chromatography. All these methods can be used with relatively small amounts of protein.

When considering setting up crystallization trials, it is always wise to try both 2-D and 3-D methods (readers interested in 2-D crsytallization should refer to the excellent review by Kuhlbrandt [2]). It can be daunting for a new comer to protein crystallization to realize how many possible crystallization conditions can in principle be set up in an effort to obtain crystals suitable for structural analysis, and, therefore, how much precious protein will be required. In an effort to overcome these problems and to capitalize on the information already available about conditions in which membrane proteins have been successfully crystallized, Newstead et al. [3] have recently analyzed the available data and used it to produce two targeted sparse matrix screens, Memsys and Memgold, which can be used to initially screen for possible crystallization conditions. In the present issue, Newstead et al. describe how effective the screens have been with β-barrel membrane proteins. Although these screens represent the best place to start crystallization trials with membrane proteins, it must be remembered that this is based on rather limited data and moreover, the more they are used the more the data will be biased. Workers in this area are therefore encouraged, if these initial screens do not work, to be more catholic in the conditions they try. Robots that can dispense nanolitre amounts will make this no more expensive in terms of the quantities of precious protein that are required. The preferred application of these sparse matrix screens is to use a 10 mg/ml protein solution, in four different detergents such as DDM, βOG, C12E8 and LDAO, and set the trials at two different temperatures, such as 4°C and 18°C. This spread of conditions provides a reasonable start. Any ‘hits’ can then be optimized. It is worth saying again that failure to get ‘hits’ is not the end of the story. You can then adopt two possible ways forward. If you are committed to only work with one specific protein then other crystallization conditions can be tried. However if any orthologue will do, then others, ideally from another species, can be tried. If you are just in a high-throughput mode then this failure can be a criterion to reject that target.

Membrane proteins in their natural environment interact with lipids. These interactions can be both specific, such as in the case of purple bacterial reaction centres and cardiolipin (Jones et al. [4]), and non-specific. The paper by Chetwynd et al. describes a molecular dynamics approach to the study of non-specific, membrane protein/lipid interactions.

As part of MPSi a tailored laboratory information management system (LIMS) has been developed to allow the group to efficiently store and exchange information, as described here by Troshin et al. It was recognized at the outset of this project it was going to be just as important to record what did not work as to record what did. The LIMS, therefore, will be an important data source to interrogate when trying to decide how to proceed with new membrane proteins. Adopting strategies designed to capitalise on experience can help save a great deal of heartache and frustration.