Endoplasmic reticulum localization of Sec12p is achieved by two mechanisms: Rer1p-dependent retrieval that requires the transmembrane domain and Rer1p-independent retention that involves the cytoplasmic domain

Among the six chimeras that are composed of both Secl2p and Dap2p fragments, DSSm, DSDm, and SSDm did not produce any detectable halo in the wild-type cells (see Fig. 4 , spots 5, 6, and 8). This indicates that these proteins were retained in the cells before they reach the Kex2 compartment of the late Golgi. All these constructs contain the TMD of Secl2p (SRFFTNFILVLLSYILQFSL). It is striking that the DSDm chimera, whose only part from Secl2p is the TMD, is completely retained before the late Golgi. Complementary to this result, SDSm secreted c~-factor in the wild-type cells (see Fig. 4 , spot 3), again indicating that the TMD region of Secl2p is very important for its localization. The same interpretation can be made for SDDm (compare with SSDm). By replacing the Secl2p TMD by the Dap2p TMD, the retention of the molecule was seriously impaired. It should be noted, however, that the amount of the a-factor secreted by SDSm or SDDm is only 27-38 % (12-13 % if corrected by normalization of expression levels) as compared to DDDm, suggesting that another mechanism of retention operates for these proteins (see below). It seems that the lumenal domain of Secl2p has no effect on the localization as long as we could test by this halo assay. DDSm and SSDm were almost indistinguishable from DDDm and SSSm, respectively, in a-factor secretion.