As environmental conditions change, organisms respond by changing the expression patterns of the genes, and proteins that are encoded by their genomes, to cope with their altered circumstances. Changes in protein expression patterns in response to physiological phenomena can be tracked by two-dimensional (2-D) gel electrophoresis, where thousands of proteins from a tissue are separated on a single electrophoresis gel. However, most so-called`proteomics studies' to date have focused on model organisms who's genomes have been sequenced and gene products identified. With this information in hand, the masses of protein fragments can be determined and matched with theoretical values in order to identify individual proteins. Kevin Russeth and colleagues from the University of Minnesota in Duluth have questioned whether this `proteomic' approach can also be applied successfully to unsequenced non-model organisms with limited genomic information available. Specifically,they asked whether using combinations of protein search programs improves the identification of proteins in such organisms. By combining and comparing various software programs that are currently used in protein identification,they also tested the assumption that we all too often make: that we can trust the results of sophisticated software.
Russeth's team chose 7 proteins from a 2-D gel of tissues from the thirteen-lined ground squirrel (Spermophilus tridecemlineatus) that were expressed during hibernation, as the team were curious to identify protein's involved in the animals' physiological response to hibernation. After isolating the proteins, they were digested into peptide fragments with the protease trypsin and the masses of the resulting fragments measured with high-resolution mass spectrometry. Once the team knew the masses of the peptide fragments from each protein and their amino acid sequences (or content), they analysed the mass patterns with four different algorithms,Mascot, Pro ID, Sequest and Pro BLAST, to try to identify each protein.
Surprisingly, all four programs identified the same proteins from each mass spectrum, with a minimum of 2 and a maximum of 16 unique peptides required to confidently identify a protein. However, the number of peptides that were determined to belong to an identified protein varied from program to program. Thus all the programs obtained the same results, despite differences between the search algorithms, much to the authors' relief.
But how confident can we be in the protein identifications? While several programs achieving the same identification is promising, the authors were cautious and inspected the positively identified spectra manually and judged that two of the seven mass spectra were of insufficient quality to warrant a confident identification, despite the agreement among programs.
Equipped with some assurances that we can confidently identify proteins in organisms with limited genomic information, we might expect some juicy insights into the differences in proteomic expression levels between active and hibernating squirrels, and the authors deliver well on this. They found that a mitochondrial protein, succinyl coenzyme A transferase (SCOT), became significantly up-regulated in the heart of hibernating squirrels. High levels of SCOT are likely to facilitate the increased use of ketone bodies, produced in the liver and delivered by the blood to the heart, as a metabolic fuel while the mammal is dormant.
Thus, despite potential pitfalls that continue to challenge the fully-fledged application of proteomics in non-model organisms, Russeth and his team have shown that there is much promise. It seems that a combination of high quality mass spectrometry, sophisticated search algorithms and a bit of chutzpah are essential for the successful application of proteomics to study non-model organisms.