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

Apart from the =Rertp dependency, it is also puzzling that even the LeuX19 TMD can localize the protein to the ER to some extent. Although weak, this artificial TMD has an ability to retain the chimeric protein in the ER, which is not present in the Dap2p TMD. How could this observation be explained? Let us presume here that this is due to the retention in the ER because it is Rerlp-independent. Possibly, the introduction of this artificial stretch of leucines interferes with the assembly of the DSDm protein into the transport vesicles and thus retards the ER-to-Golgi traffic. The quality control system of the ER might be somehow involved in this process. Alternatively, the partial ER retention may take place without any particular signal, and the fast exit of Dap2p from the ER may depend on a selective mechanism that positively transports the protein to the Golgi apparatus. This is a revival of the selectivity model of Lodish (1988) and may be consistent with the arguments that cargo molecules are selectively concentrated in transport vesicles (Mizuno and Singer, 1993; Balch et al., 1994) . Of course, we cannot exclude the possibility that the effect of the LeuX19 TMD in ER localization is due to another retrieval mechanism that does not require Rerlp. It is known that many Golgi membrane proteins are localized by virtue of the TMD signals. However, the structural requirements of such signals are still unclear (Swift and Machamer, 1991; Bretscher and Munro, 1993; Nilsson et al., 1993; Weisz et al., 1993) . If the TMD alone can distinguish ER and Golgi proteins, what is the difference between them? Munro (1995) has recently demonstrated that the length of the TMD is critical for the Golgi retention in the case of sialyltransferase and suggested the importance of the lipid-based sorting. In his experiment, the LeuX17 TMD functions as a Golgi localization signal. We have performed a similar experiment to test whether such a length effect of the TMD could be involved in the case of Sec12p as well. The result indicates that all of the leucineonly TMDs we have examined show a partial retention ability that is independent of Rerlp. There was no special preference to the length of leucine stretch (M. Sato, unpublished data). The different requirement of similar hydrophobic TMD between the ER and the Golgi localization remains to be clarified experimentally.