Glycosylphosphatidylinositol-modified (GPI) proteins share structural features that allow their identification using a genomic approach. From the known S. cerevisiae and C. albicans GPI proteins, the following consensus sequence for the GPI attachment site and its downstream region was derived: [NSGDAC]-[GASVIETKDLF]-[GASV]-X(4,19)-[FILMVAGPSTCYWN](10)>, where > indicates the C-terminal end of the protein. This consensus sequence, which recognized known GPI proteins from various fungi, was used to screen the genomes of the yeasts S. cerevisiae, C. albicans, Sz. pombe and the filamentous fungus N. crassa for putative GPI proteins. The subsets of proteins so obtained were further screened for the presence of an N-terminal signal sequence for the secretion and absence of internal transmembrane domains. In this way, we identified 66 putative GPI proteins in S. cerevisiae. Some of these are known GPI proteins that were not identified by earlier genomic analyses, indicating that this selection procedure renders a more complete image of the S. cerevisiae GPI proteome. Using the same approach, 104 putative GPI proteins were identified in the human pathogen C. albicans. Among these were the proteins Gas/Phr, Ecm33, Crh and Plb, all members of GPI protein families that are also present in S. cerevisiae. In addition, several proteins and protein families with no significant homology to S. cerevisiae proteins were identified, including the cell wall-associated Als, Csa1/Rbt5, Hwp1/Rbt1 and Hyr1 protein families. In Sz. pombe, which has a low level of (galacto)mannan in the cell wall compared to C. albicans and S. cerevisiae, only 33 GPI candidates were identified and in N. crassa 97. BLAST searches revealed that about half of the putative GPI proteins that were identified in Sz. pombe and N. crassa are homologous to known or putative GPI proteins from other fungi. We conclude that our algorithm is selective and can also be used for GPI protein identification in other fungi.
Copyright 2003 John Wiley & Sons, Ltd.