Analyses of protein evolution, function, and architecture

Sammanfattning: Proteins can evolve over time in many different ways. An ancestral protein sequence inherited in different species will gradually undergo changes in primary sequence and sometimes in domain architecture. Some of these changes will affect its function, and evolutionary analyses can be used to predict function shift. A common paradigm is that orthologs, i.e. genes in different species that derive from the same gene in the last common ancestor, are functional counterparts. Orthology is a special case of the more general concept homology, which means any form of shared ancestry. This thesis investigates the functional conservation of orthologs compared to non-orthologs, and further explores gene and protein domain architectural changes during evolution. A set of 17 proteins were selected between human and the nematode C. elegans such that they were predicted to be orthologous, membrane-spanning, and did not have a known function. By experimental studies in the nematode, functional clues were obtained for 12 of them that thus have high relevance for the human orthologs. Several of the genes were expressed in the nervous system. One of them was a presenilin-like protein, which was subjected to further bioinformatic analysis, including prediction of its transmembrane topology. Mutations in presenilin are known to cause Alzheimer's disease, the main type of dementia in humans. Resolving the molecular structure of presenilin has not been possible yet because it is a transmembrane protein. Instead, many attempts to elucidate the transmembrane topology biochemically have been made, but the results were often contradictory. We therefore approached the problem by reconciling the output from several transmembrane topology predictors and previously published experimental studies. This allowed us to propose a novel nine-transmembrane topology with the C-terminus located in the extracytosolic space, which has subsequently been verified by several other researchers. To study the evolution of protein domain architecture we developed a new algorithm based on the maximum parsimony criterion to infer ancestral architectures. We analyzed 96 species across all kingdoms to find cases where a domain architecture had been created multiple times independently. In contrast to previous studies we found that such events are relatively frequent, up to 12.4%. Among the architectures displaying reinvention we could find no strong functional bias, implying that it is a widespread phenomenon. In this thesis, the focus is on evolutionary analysis and applying it when investigating various aspects of protein function and architecture. Incorporating new discriminating features is important to further enhance the accuracy of phylogenetic inference. To this end, we investigated conservation of intron positions among orthologs versus nonorthologs that are equally similar in sequence. We found that ortholog-ortholog gene pairs on average have a significantly higher degree of intron position conservation compared to ortholog-closest non-orthologs. This implies that shared intron positions could be used as an additional discriminating feature in evolutionary analysis.