A PROTEOGENOMICS FRAMEWORK FOR DISCOVERY OF TRANSLATED MICROPEPTIDES IN MOUSE BRAIN AND OTHER SYSTEMS

In this thesis, we developed methods for micropeptide discovery and applied these methods to several biological systems to find evidence of novel translated micropeptides. Micropeptides are small proteins translated from open reading frames shorter than 100 codons, and have historically been understudied in genome annotations. This framework allows discovery of these peptides, by an approach called proteogenomics, which combines mass spectrometry-based proteomics with deep sequencing-based transcriptomics in order to study the genome. We developed a bioinformatics framework for micropeptide discovery experiments, and also developed and validated an LC-MS/MS platform for micropeptide discovery. Finally, to show these methods are broadly applicable, we used these methods to analyze the liver proteome, the HIV-1 virion proteome, and the mouse cerebellum proteome, finding evidence of novel micropeptides in each case. Micropeptides are thought to be low-abundance, and therefore difficult to detect by biochemical methods. A lack of sensitivity in proteomics is often due to suppression effects of high abundance molecules which co-elute with low abundance molecules. We developed a two dimensional chromatography platform, in which a strong cation exchange column is coupled in series with two hydrophobicity columns all coupled to a tandem mass spectrometer, allowing online fractionation of peptides. Applying this system to micropeptidomics, we have shown that sensitivity for micropeptides greatly increases as well. This system is applied in to discover novel micropeptides in human liver. The framework was applied to several biological systems with varying degrees of genome complexity: the HepG2 liver cell line, to study the human liver proteome, purified HIV-1 virions, to study the packaged virion proteome, and the mouse brain. The three experiments represent a large range of genomic complexity: HIV-1 has a very small genome, the human genome is much larger, and the mammalian CNS genome is believed to be highly complex due to alternative splicing. The human liver HepG2 cell line was subjected to proteogenomics analysis. A transcriptome assembly was translated in silico and used to search for peptide products from all possible open reading frames shorter than 100 codons. Two novel peptides were found with high confidence. Next, HIV-1 virions were subjected to proteogenomics. The HIV-1 genome is very small, so the whole genome was translated. In order to capture strain-specific differences, three biological replicates were genomically sequenced, and the resulting HIV-1 genome assemblies translated in silico. One novel peptide was observed in each replicate with high confidence. Finally, the mouse brain was subjected to proteogenomics in order to find novel secreted peptides. The soluble fraction of mouse brain was used to generate proteomics data. As the mammalian CNS proteome is believed to be very complex due to alternative splicing we utilized ribosome profiling data, which allows reading frame and translation start sites to be determined. A translatome was assembled from the frame determination and used to generate the database for proteomics analysis. Several micropeptides were again found, and in particular one highly expressed peptide was found to be enriched in behavior regulating regions of the brain. Thus, applying the framework to several different biological systems shows that it is robust against various genomes so that micropeptides can be discovered in various experimental contexts. The results also suggest the ubiquity of micropeptide expression in all organisms