Recent studies have demonstrated the existence of glycosyl-phosphatidylinositol (GPI)-anchored proteins in higher plants. In this study we tested whether GPI-addition signals from diverse evolutionary sources would function to link a GPI-anchor to a reporter protein in plant cells. Tobacco protoplasts were transiently transfected with a truncated form of the Clostridium thermocellum endoglucanase E reporter gene (celE') fused with a tobacco secretion signal (PR-1a) at the N-terminus and either a yeast (GAS1), mammalian (Thy-1) or putative plant (LeAGP-1) GPI-anchor addition signal at the C-terminus. The yeast and plant C-terminal signals were found to be capable of directing the addition of a GPI-anchor to the endoglucanase protein (EGE') as shown by the sensitivity of the lipid component of GPI to phosphatidylinositol-specific phospholipase C (PI-PLC) digestion. In contrast, the mammalian signal was poorly processed for anchor addition. When EGE' was fused to a truncated form of the LeAGP-1 signal (missing three amino acids predicted to be critical to signal cleavage and anchor addition), a GPI-anchor was not linked to the EGE' protein indicating the necessity for the missing amino acids. Our results show the conservation of the properties of GPI-signals in plant cells and that there may be some similar preferences in GPI-addition signal sequences for yeast and plant cells.