Organelle identity is maintained in the secretory pathway despite the extensive bidirectional anterograde and retrograde membrane flux. This requires that proteins involved in post-translational modifications such as molecular chaperones and glycosylation enzymes must not be transported to cellular destinations other than their resident compartment, and on arrival at their resident compartment must be retained there. Two general principles have been elucidated as to how this is achieved: retention and retrieval (see Section 1.2.4), and both mechanisms tend to co-operate to maintain the localization of any resident protein. A retention mechanism acts to actively maintain proteins in their target compartment, possibly by excluding them from transport vesicles (Nilsson and Warren, 1994). A corollary of this is that the retention signal must act in, but not before or after the correct compartment. It must therefore allow anterograde transport along the secretory pathway until arrival at the destination compartment.
The localization o f lumenal Golgi enzymes is achieved in part by retention mechanisms, which operate through combinatorial protein-protein and protein-lipid interactions. Two models have been proposed to explain how resident proteins in the Golgi stack are retained. The ‘bilayer thickness’ or lipid sorting model (Bretscher and Munro, 1993; Munro, 1991, 1995, 1998) proposes that the length of the transmembrane domain (TMD) is the pivotal factor in sorting of Golgi resident proteins. This proposal is based upon the observation that a cholesterol gradient exists across the Golgi stack, whereby the cholesterol concentration increases on moving from
cis- to ^ra«5-compartments (Orci et al., 1981), and as a consequence so too does the bilayer thickness. Coupled to this is the observation that endogenous Golgi enzymes have shorter TMDs (c. 17 amino acids) than plasma membrane (which has an even higher cholesterol content) proteins (c. 21-22 amino acids). Early Golgi residents are therefore excluded from upstream compartments because their TMD is too short to allow thermodynamically stable incorporation into thicker bilayers. This proposal was supported by the finding that the TMD of SialylT can be replaced by 17 leucine residues, indicating that the primary sequence per se is not essential for retention, and
Chapter 1_______________________________________________________ Introduction
that increasing the length of the SialylT TMD to 23 leucines resulted in a plasma membrane loealization (Munro, 1991, 1995; Dahdal and Colley, 1993). The TMD has also been shown to be sufficient to loealize GalT and NAGTI to the trans- and medial-
Golgi and grafting of these TMDs on to reporter molecules induces their localization to the appropriate Golgi compartment (Nilsson et al., 1991; Teasdale et al., 1992; Burke et al., 1992; Tang et al., 1992). However, it is not clear in these cases whether these effects are due to the biophysieal features of the primary sequence or the length of the TMD. Furthermore, hydrophilic residues flanking the TMD may also be required for Golgi localization (Munro, 1991). The bilayer thickness model is challenged by observations that the TMDs of GalNAc-Tl and T3 are 25 and 18 amino acids respectively (Bennett et al., 1996), suggesting a plasma membrane localization for GalNAc-Tl and di trans-Qo\g\ localization for GalNAc-T3. This is totally inconsistent with the available localization data, which reveals that GalNAc-Tl is found throughout the Golgi stack, and GalNAc-T3 in the medial Golgi cistemae (Rottger et al., 1998). The length of the TMD of GalT is also insufficient for Golgi retention (Masibay et al.,
1993).
The alternative model proposes that lumenal Golgi proteins are retained by the formation of oligomeric complexes that is induced by the prevailing microenvironmental conditions of the appropriate compartment and prevents access to transport vesicles (Swift and Machamer, 1991; Weisz et al., 1993). A related concept entitled ‘kin recognition’ assumes that resident Golgi enzymes are incorporated into large pre existing hetero-oligomers upon ingress to their specific compartment (Nilsson et al., 1993b). Distinct Golgi enzymes which reside in the same compartment are termed ‘kin’ and are thought to interact via their lumenal stalk domains, as has been demonstrated for Mannll and NAGTI (Nilsson et al., 1993a, 1994, 1996). Consistent with this is that retention of NAGTI in the ER also results in the accumulation of Mannll in the ER and vice versa (Nilsson et al., 1994). Detergent extraction of purified rat liver Golgi membranes reveals large oligomeric complexes eontaining Mannll and NAGTI that bind an intercistemal matrix (Slusarewicz et al., 1994; Hui, 1997). This ‘kin recognition’
may also reflect the formation of highly specific multienzyme complexes, which have also been found in yeast (Jungmann and Munro, 1998).
The existence of these large oligomeric complexes is difficult to equate with the high diffusional mobility of Golgi enzymes tagged with GFP in the Golgi membranes of living cells, as revealed by fluorescence recovery after photobleaching (FRAP) experiments (Cole et al., 1996b). Although these observations cannot exclude the possibility that oligomers of a few hundred molecules exist, they are inconsistent with the presence of larger oligomers that are anchored to an intercistemal matrix (Munro, 1998; Slusarewicz et al., 1994). However, other studies suggest that this high diffusional mobility may be due to the transient expression system used. When similar experiments are performed in cell lines stably expressing GFP tagged Golgi enzymes at endogenous levels, no such high diffusional mobility is observed in Golgi membranes, but only upon redistributing these enzymes to the ER by BFA treatment (George Banting and Dave Shima, personal communication). It is also unclear whether the GFP tag affects oligomer formation. However, the kin recognition model does not seem to apply to GalT, SialylT (Munro, 1995), GalNAc-Tl, -T2 nor -T3 (Rottger et al., 1998). It therefore seems likely that both features of the TMD and kin recognition are likely to play a role in retention.
Peripheral Golgi membrane proteins must also be targeted to the correct cistemae, and this may be achieved in part by binding specifically to the cytoplasmic tails of specific lumenal residents, which would then act as membrane receptors. N-terminal fatty acylation has been determined to be important for a number of proteins (e.g. endothelial nitric oxide synthase [Liu et al., 1997], SCGIO [DiPaulo et al., 1997], glutamate decarboxylase [Solimena et al., 1994], GRASP65 [Barr et al., 1997] and GRASP55 [Shorter et al., 1999]), but how this imparts specificity to targeting remains unclear, and may require additional signals (McCabe and Berthiaume, 1999). Other proteins such as GM130, Golgin-245 and Golgin-97 may ‘piggy back’ target to their compartment via such acylated proteins (Barr et al., 1997; Barr, 1999). Certain pleckstrin homology (PH) domains may also impart Golgi targeting specificity (Levine
Chapter 1_______________________________________________________ Introduction
and Munro, 1998). The distribution of peripheral membrane proteins could also be biased by how easily they are incorporated into COPI vesicles (Linstedt, 1999), as has been proposed for Golgi enzymes (Glick et al., 1997).