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

A hint on this possibility comes from the result on Sed4p. Sed4p shares high homology with Secl2p in the cytoplasmic domain and the TMD (Hardwick et al., 1992) . In fact, both of these domains are capable of localizing the Dap2p chimeras in the ER as those of Secl2p do. If the ER retention by the cytoplasmic domain of Secl2p is fulfilled by interaction with other molecules, is it also the case with the cytoplasmic domain of Sed4p? If so, are there common partners with which Secl2p and Sed4p can interact? Sed4p has been shown not to possess the guaninenucleotide exchanger activity toward Sarlp despite the high similarity to Secl2p (Barlowe and Schekman, 1993) . However, Gimeno et al. (1995) reported that synthetic lethal interaction exists between one mutant allele of SAR1 and the disruptant of SED4 and suggested the possibility that their products may function in a multisubunit complex in the wild-type cells. Sed4p has also been shown to interact with Secl6p (Gimeno et al., 1995) . Since SAR1 has a strong suppressor activity on sec12 and sec16 ts mutants (Nakano and Muramatsu, 1989) , there may be a link between Secl2p and Secl6p. Sarlp and Secl6p could be somehow involved in the ER retention of Secl2p and Sed4p. How Sarlp and Sec16p are associated with the ER membrane is totally unknown. To solve this kind of chicken-and-egg problem, further efforts to dissect the structural requirements of the cytoplasmic domains of Secl2p (and Sed4p) will be necessary.