Structure determination
Transmembrane proteins, reveal your structures - or we may use knowledge-based force
Although transmembrane proteins are considered to be the most attractive targets for drug discovery and may represent up to 30% of all human genes, there is frustratingly little structural information available for these molecules. With steadily increasing demand for detailed structures of transmembrane proteins, several academic groups and companies have recently achieved interesting progress towards new methods for predicting structures, potentially enabling three-dimensional models to be created for many of these proteins in the foreseeable future.
Transmembrane (TM) proteins of eukaryotic origin provide several significant challenges for structural biologists. First, they are often difficult to express or purify at sufficiently high levels as compared to many of their bacterial homologues. Second, post-translational modifications such as glycosylation events are complicating factors during crystallisation. Finally, despite the remarkable progress made in recent years in determining high-resolution structures for bacterial TM proteins, the fact that over half of eukaryotic TM proteins have no bacterial homologues is another important factor in the persisting lack of experimentally derived structures for these molecules.
One of the leading groups in working towards solving TM protein structures is the Bioinformatics Unit of University College London. Led by David Jones, this team of investigators is developing strategies for applying various structural prediction methods in tandem, thereby creating an effective platform for creating reliable three-dimensional models for large numbers of TM proteins. In a recent review in the Philosophical Proceedings of the Royal Society, Prof. Jones outlines his group's approach to this problem and describes the status and usefulness of the many structural prediction techniques currently available.
One of the first key steps in determining the structure of integral membrane proteins is the elucidation of membrane helix locations and topology. This can be done through the use of fusion proteins, proteolytic digestion, or antibody binding. However, given the relatively limited amount of information that topology provides, as well as the appreciable challenges in obtaining this data experimentally, a growing number of bioinformatic tools have been developed over the last several years for topology prediction and these are now widely used and accepted. More recently, consensus methods have emerged that enable users to efficiently combine the results from the major topology prediction methods. The first of these to be developed was the approach of Nilsson and colleagues, which finds a consensus between TMHMM, HMMTOP, MEMSAT, PHD, and TOPPRED, and which is able to correctly predict partial consensus topologies for over 90% of transmembrane proteins analysed.
With regard to the prediction of tertiary structure, however, the picture quickly becomes somewhat more complex. As is the case with globular proteins, the most widely applied method to date has been comparative modelling. An important limitation of this approach is still the lack of suitable templates (for example, the Protein Data Bank currently contains 95 sequence-unique transmembrane proteins versus approximately 20,000 globular proteins).
Nevertheless, several groups are now developing template-independent comparative modelling methods, beginning with the prediction of relative helix orientation, which is considered to be the most readily predicted structural feature after topology. One of several promising knowledge-based methods is FILM, which makes use of an intriguingly simple model of the physiochemical constraints provided by the lipid bilayer. FILM was developed by Jones' group at UCL and assembles super-secondary structural fragments, from a library of structurally known proteins, to predict the helix topology and conformation of small transmembrane proteins with reasonable accuracy.
By improving upon and integrating these various structure prediction tools, it may soon become routine to accurately predict the structures of eukaryotic transmembrane proteins, thereby resolving an important bottleneck in drug discovery.
Click here for useful links
