Image courtesy of Andrew Emili,
University of Toronto, Toronto
Generating high-quality proteomics data for complex organisms can be a dirty job. Compared to the stability of the genome, the proteome can change from organ to organ and cell to cell. Furthermore, the localization of a protein within the cell is crucial to its function. A new study in Cell uses multidimensional protein identification technology (MudPIT) and comparative proteomics to analyze organ- and organelle-specific protein expression. The resulting dataset, the Mouse Proteome Project, is freely available on the web as a community resource (http://tap.med.utoronto.ca/~mts/).
MudPIT, also called shotgun proteomics, involves extensive protein separation by high-performance liquid chromatography (HPLC) followed by tandem mass spectrometry. The authors separated mouse brain, heart, kidney, liver, lung and embryonic placenta into cytosol, mitochondria, nuclear and membrane fractions by ultracentrifugation. Protein lysates from each fraction were trypsinized and loaded on a dual-phase HPLC column. The first phase, a strong cation exchange column, shifted proteins into the second phase, a reversed-phase chromatography column, when washed with buffers of increasing salt concentrations. Increasing concentrations of acetonitrile displaced proteins from the second phase into the mass spectrometer. The authors analyzed mass spectra using the SEQUEST database, which incorporates both sequence and structural criteria to identify proteins. Each organelle fraction underwent at least seven independent MudPIT analyses.
In all, 203 MudPIT experiments yielded ~8 million mass spectra. The authors identified 4768 proteins, with on average 2000 proteins per tissue and 1000 proteins per organelle. Most of the identified proteins demonstrated organelle (~75%) and tissue (~50%) specific expression. In the brain, 1366 proteins localized to the cytosol, 1075 to mitochondria, 907 to nuclei, and 1040 to membrane fractions.
The authors included inverted peptide sequences in their identification database to help estimate their error detection rate. These 'decoy' proteins comprised only 0.3% of the total proteins identified, suggesting a low rate of false positives. Proteins with transmembrane helices, usually underrepresented in proteomics assays, comprised 14% of the total proteins. Although the authors used different mouse strains and sexes, the MudPIT data largely agreed with a prior organ-specific microarray study. Only 25% of the molecules common to both studies showed differences in expression pattern. The MudPIT data also generally agreed with prior work on protein expression in individual organelles.
This study determined subcellular localizations for 1503 proteins for the first time, although 1494 of the detected proteins could not be assigned with confidence to a particular organelle. Studies on a less global scale will help address whether the unassigned proteins are 'shuttled' between organelles. Regardless, this study is the first of its kind to tackle the complexities of the mammalian proteome in a genome-like fashion and will undoubtedly influence our ability to examine proteins on the organismal level.