Retention mechanisms are usually reinforced by retrieval mechanisms, whieh ensure that if a resident does escape from its compartment of residence it is recognised and returned. Such retrieval requires retrograde vehicles, which are most likely COPI vesicles (Cosson and Letoumeur, 1994; Letoumeur et al., 1994; Cosson et al., 1996; Spang and Schekman, 1998; Girod et al., 1999). Retrieval has been most clearly demonstrated for ER residents, and was initially shown for the soluble ER residents BiP and PDI (Munro and Pelham, 1987; Pelham, 1988). Deletion of the four C-terminal amino acids (KDEL) from either protein induced their gradual secretion. Furthermore, grafting this C-terminal KDEL sequence onto a lysosomal resident, cathepsin, induced its accumulation in the ER. Cathepsin received early Golgi sugar modifications, suggesting that it left the ER only to be retrieved from an early Golgi compartment. Other C-terminal tetrapeptides induce ER retrieval most notably HNEL in mammals (Bu et al., 1995, 1997), HDEL in Saccharomyces cerevisiae (Pelham et al., 1988), and DDEL in Kluyveromyces lactis (Lewis et al., 1990). This retrieval is accomplished by interaction with a KDEL receptor (Semenza et al., 1990; Lewis and Pelham, 1992) which exists at steady state in the CGN and cw-Golgi (Griffiths et al., 1994). Interaction between receptor and ligand occurs specifically in the Golgi because of the lower pH of this compartment relative to the ER (Wilson et al., 1993). On binding KDEL ligand the KDEL receptor oligomerizes and is packaged into retrograde moving COPI vesicles (Majoul et al., 1998), and returned to the ER (Townsley et al., 1993). On reaching the ER, the more alkaline pH induces ligand release and return of monomeric, unoccupied receptor to the Golgi (Townsley et al., 1993). This retrieval pathway may operate as distal from the ER as the TGN or plasma membrane (Miesenbock and Rothman, 1995; Majoul et al., 1996).
ER residents with type I membrane topology are retrieved by virtue of a C-terminal KKXX or KXKXX (where X is any amino acid) motif in their cytoplasmic domain (Nilsson et al, 1989; Jackson et a l, 1990). In type II proteins the retrieval signal consists of two crucial arginine (RR) residues which must reside within the first five N- terminal amino acids (Schutze et al., 1994). Grafting these signals onto reporter molecules results in an ER localization, but the molecules obtain Golgi sugar modifications indicative of retrieval (Jackson et al., 1993; Schutze et al., 1994). KKXX containing molecules interact directly with coatomer, and this interaction is essential for retrieval, suggesting COPI vesicles are the retrograde vehicle (Cosson and Letoumeur, 1994; Letoumeur et al., 1994; Dominguez et al., 1998). Furthermore, this interaction between a KKXX bearing molecule and coatomer can actually promote COPI vesicle formation, at least from liposomes in a purified system (Bremser et al., 1999).
A retrieval mechanism has also recently been proposed to be important for maintaining the distribution o f Golgi enzymes within the Golgi stack (Glick et al., 1997), and such a mechanism may co-operate with the retention mechanisms outlined above. This model is contingent upon a cistemal maturation (Section 1.2.5) view o f Golgi dynamics and constructs a mathematical tautology to explain the asymmetric localization of Golgi enzymes within the stack. In this model czj-loealized Golgi enzymes are more likely to enter retrograde moving COPI vesicles than trans-\ocd\izQ& enzymes, and this enhanced retrieval maintains the observed asymmetric enzyme distribution despite the anterograde motion o f maturing cistemae which bear the secretory cargo (Glick et al.,
1997). Enzyme oligomers may still be incorporated into this model, in that they form and exclude enzymes from vesicles, until the cistemae reaches a certain age. In support of this the S. cerevisiae cw-Golgi mannoysltransferase O chip was found to obtain TGN modifications suggesting its czj-Golgi localization is maintained by retrieval (Harris and Waters, 1996) and similar movements have been reported for GPP 130 (Linstedt et al., 1997) and mammalian NAGTI although with slower kinetics (Hoe et al., 1995). Whether Golgi enzymes actually enter COPI vesicles is a highly contentious issue, and although present, they do not appear to be enriched in COPI vesicles isolated in the presence of the non-hydrolyzable GTP analogue GTPyS (Sonnichsen et al., 1996). However, a steady state population of vesicles can be isolated from tissue
Chapter 1_______________________________________________________ Introduction
culture cells which contain high amounts of Golgi glycosylation enzymes (Love et al, 1998), but the precise nature of these vesicles is unclear. Recently GTP hydrolysis by ADP-ribosylation factor (ART) 1 has been shown to affect the content of COPI vesicles both in vitro and in vivo (Nickel et al., 1998; Lanoix et al., 1999; Pepperkok et al., 2000), and in one such study Golgi enzymes were found to be efficiently packaged into putative COPI vesicles (Lanoix et al., 1999). Furthermore, NAGTI a medial-Golgi
enzyme was found to be incorporated into COPI vesicles in preference to GalT a
trans-GoXgi enzyme in accordance with the proposal of Glick and colleagues (Lanoix et al., 1999; Glick et al., 1997). The cytoplasmic tail of NAGTI has also been found to interact weakly with coatomer suggesting a mechanism of incorporation (Dominguez et al., 1998). However, whether the vesicles isolated by Nilsson and colleagues truly are COPI vesicles is not entirely clear. The ultimate test of this model will come by determining whether Golgi enzymes can be detected in COPI vesicles in vivo by quantitative immuno-EM. This approach has demonstrated the presence of both anterograde and retrograde moving cargo in distinct populations of COPI vesicles associated with the Golgi stack (Orci et al., 1997).
Retrieval of Golgi enzymes may also operate through non-COPI coated vehicles, as demonstrated in a recent studies that reveal a Rab6 dependent, COPI independent ER- Golgi recycling pathway for glycosylation enzymes, which is hijacked by the Shiga toxin (Girod et al., 1999; White et al., 1999). However, such recycling to the ER of Golgi enzymes may reflect the need to degrade or ‘fix up’ functionally expired transmembrane proteins.
The concepts of retention and retrieval are perhaps most clearly unified for the TGN resident TGN38. TGN38 is predominantly a TGN marker but cycles to the plasma membrane via an endosomal compartment (Luzio et al., 1990). Internalization and retrieval from the plasma membrane is induced by a -YQRL signal in the TGN38 cytoplasmic tail (Bos et al., 1993). Additionally, the TMD of TGN38 represents a retention signal, as its transplantation onto reporter molecules is sufficient to localize them to the TGN (Ponnambalam et al., 1994). Thus, TGN38 contains two non
overlapping localization signals that must co-operate in order to maintain TGN38 in the TON. The presence of multiple localization signals may be a general trait of resident proteins.