Análisis de los datos

The Golgi apparatus typically occupies a juxtanuclear, usually pericentriolar position in most mammalian cells. Which side of the nucleus Golgi apparatus resides can reflect which area of the cell surface exocytosis is directed towards. For example, the Golgi apparatus is orientated on the side of the nucleus proximal to the leading edge of a migrating cell (Nobes and Hall, 1999). Similarly, when a cytotoxic lymphocyte binds to an antigenic cell, there is a reorientation of the Golgi apparatus to the side of the nucleus proximal to the offending cell and a directed exocytosis of cytoplasmic granules toward it (Atkinson and Bleackley, 1995). This long range ordering of Golgi structure is dependent on interactions with the cytoskeleton. That Golgi apparatus positioning is dependent on microtubules is confirmed by observations that microinjection of anti­ tubulin antibodies (Wehland and Willingham, 1983) or depolymerization of microtubules with nocodazole (Sandoval et al., 1984) fragments the Golgi apparatus into a series of disseminated, discrete mini-stacks (Cole et al., 1996a; Shima et al., 1998; Storrie et al., 1998). These mini-stacks may represent the discrete stacked units of the Golgi ribbon and remain functional in that nocodazole treatment has no effect on the secretion or sialylation of VSVG (Featherstone et al., 1985; Rogalski et al., 1984). In contrast, disruption of microtubules in S. pombe, using thiabendazole or in a cold sensitive mutant cell line, resulted in unstacking of Golgi cistemae (Ayscough et al., 1993). Nocodazole induced dispersal of Golgi mini-stacks appears to be a kinesin driven process (Minin, 1997), and is currently a controversial subject. This is because some authors propose that nocodazole induced dispersal involves transit through the ER (Cole et al., 1996a; Storrie et al., 1998), while others suggest that this is not required (Shima et al., 1998). Contradictory results have been obtained whereby use of a dominant negative mutant Sarlp that blocks ER exit, either accumulates Golgi residents in the ER (Storrie et al., 1998) or does not (Shima et al., 1998) upon nocodazole treatment. The reasons for such discrepancies remain unclear, but may have to do with the different incubation times used.

Recovery from nocodazole also involves microtubule dependent saltatory movements which cluster Golgi mini-stacks in the pericentriolar region (Kreis et al., 1988; Ho et al..

1989). Such movements are consistent with a role for dynein in moving the Golgi apparatus to the pericentriolar region, and this has been confirmed in an assay where broken CHO cells ‘capture’ purified Golgi membranes and return them to the centrosomal region (Corthésy-Theulaz et al., 1992). This process can be accelerated by mapmodulin, a protein that may clear microtubule tracks of obstructing microtubule associated proteins (MAPs), so accelerating dynein mediated organelle movement (Ulitzur et al., 1997a, b). The precise mode of linkage of Golgi membranes to microtubules remains unclear, but may involve linkage via dynactin to dynein (Burkhardt et al., 1997; Harada et al., 1998; Ma et al., 1999) and an extensively coiled coil peripheral membrane Golgin GMAP-210 (Infante et al., 1999). In support of the latter, the C-terminal microtubule-binding domain of GMAP-210 when tagged with GFP and transfected into cells localized to the centrosome (Infante et al., 1999).

With the exception of the unstacking data from S. pombe (Ayscough et al., 1993), cytoskeletal proteins seem to be more responsible for the long range order of Golgi architecture, rather than the structure of individual stacks. Cytochalasin D treatment to disrupt actin based microfilaments or microinjection of antibodies specific for intermediate filaments has no effect on Golgi morphology (Ho et al., 1989). However, interactions with the actin cytoskeleton may affect the position of the Golgi in mammalian cells (di Campli et al., 1999), and may be more important for Golgi stack movements towards ER exit sites and sites of cell wall synthesis in plant cells (Nebenfuhr et al., 1999).

1.4.3 Stacking.

Since Golgi membranes can be isolated from tissues as stacks (Morré and Mollenhauer, 1964; Morré et al., 1970; Fleischer and Fleischer, 1970; Leelavathi et al., 1970; Hino et al., 1978; Hui et al., 1998) it would appear likely that the stack is a stable structure. In vivo the Golgi apparatus is embedded in an electron dense ‘zone of exclusion’, which represents a fibrous matrix and is so-called because it excludes structures the size of ribosomes and above (Morré and Ovtracht, 1977; Mollenhauer and Morré, 1978). Close inspection of cistemae by EM reveals the presence of fibrillar structures running

Chapter 1_______________________________________________________ Introduction

parallel to cistemae (Mollenhauer, 1965; Amos and Grimstone, 1968; Staehelin et al., 1990) and electron dense proteinaceous material that bridges the intercistemal space (Franke et al., 1972; Cluett and Brown, 1992). Such intercistemal crossbridges are c. 8.5nm in width and l l nm in height and recur at lOnm intervals along individual cistemae (Cluett and Brown, 1992). These structures can be degraded by proteolysis which also unstacks Golgi cistemae (Mollenhauer et al., 1973; Cluett and Brown,

1992). Such structures constitute a cisternal exoskeleton or matrix that probably acts to maintain the Golgi stack. Such an exoskeleton may also co-operate with a cistemal endoskeleton to maintain Golgi architecture. Consistent with this, electron dense intracistemal crossbridges can also be visualized by EM, that span the cross-sectional width of cistemal lumen and are sometimes continuous with intercistemal crossbridges (Franke et al., 1972). The fact that single cistemae liberated by proteolysis maintain their disk shaped morphology indicates that this Golgi endoskeleton or lumenal matrix maintains the cistemal shape. It may be that the large oligomeric structures formed by Golgi enzymes, such as Mannll and NAGTI, function to maintain the characteristic disk shape of cistemae (Nilsson et al., 1994, 1996). Consistent with this possibility is the observation that the induction of the glycosylation enzymes in G. lamblia during trophozoite encystation correlates with the formation of prominent, flattened cistemal membranes (Lujan et al., 1995). Furthermore, the absence of disk shaped cistemae in P. falciparum may be due to the lack of glycosylation enzymes in this organism (Haidar,

1998).

Attempts have been made to isolate these matrix like structures by detergent extraction of purified Golgi membranes in a manner similar to that used in the identification of the nuclear lamina (Dwyer and Blobel, 1976). Extraction of purified Golgi stacks with detergent at low salt reveals fibrous structures highly reminiscent of stacked cistemae (Slusarewicz, 1994; Slusarewicz et al., 1994; Fath et al., 1997). Furthermore, this Golgi matrix was found to bind to the medial-Go\g\ enzymes NAGTI and Mannll with high affinity (Slusarewicz et al., 1994). This in itself suggests a stacking mechanism for the Golgi apparatus since these enzymes have been localized to both medial- and trans-

cistemae within the Golgi stack (Nilsson et al., 1994; Rabouille et al., 1995a). Were the

cytoplasmic tails of the NAGTI and Mannll in adjacent cistemae to interact with the intercistemal matrix, this would anchor cistemae together. However, the identity of matrix proteins that interact with only the cytoplasmic tails of enzymes has so far proven elusive. Furthermore, peptides representing the cytoplasmic tails of Mannll and NAGTI were unable to compete for binding of the enzymes to the matrix. This coupled to the fact the lumenal domains of the enzymes could also bind to the matrix suggest that a fraction of the matrix may be composed of lumenal proteins (Slusarewicz et al., 1994; Slusarewicz, 1994). However, consistent with the importance of Golgi enzymes in maintaining stacked structure is the observation that mutation of the TMD of NAGTI to a series of leucine residues resulted in unstacking of cistemae, and loss of cistemal shape (Nilsson et al., 1996). The conformational change this mutation induced in NAGTI may have disrupted inter- and intracistemal interactions essential for maintenance of Golgi morphology (Rabouille and Nilsson, 1995).

Fractionation of the Golgi matrix by SDS-PAGE revealed the presence of 12-15 major proteins and numerous minor proteins (Slusarewicz et al., 1994; Slusarewicz, 1994; Hui, 1997). Amongst these proteins were GM130 (Nakamura et al., 1995), GRASP65 (Francis Barr, personal communication), giantin and p i 15 (Nobuhiro Nakamura, personal communication) the COPI vesicle tethering machinery (Sonnichsen et al., 1998; Section 1.3.3). This hinted at another mechanism by which the Golgi stack may be maintained or established, if p i 15 were to simultaneously bind GM130 and giantin in adjacent cistemae (See Chapter 4). Furthermore, GRASP65 has been implicated in stacking Golgi cistemae (Barr et al., 1997, 1998; Shorter and Warren, 1999; Shorter et al., 1999; See Section 1.5.7; Chapter 4 and 5). Other components of the Golgi matrix were identified as actin, cytokeratins 8 and 18 (Hui, 1997), suggesting the Golgi matrix may be linked to elements of the cytoskeleton. A number o f other bands were subjected to peptide-mass fingerprinting analysis, but were not found in the database (Pappin et al., 1993; Hui, 1997).

There are now a number of other proteins that are candidates for components of the Golgi matrix. One group of proteins, the Golgins, represent a large family of long, rod­

Chapter 1 Introduction

like coiled coil proteins which often contain one or more flexible hinges (Chan and Fritzler, 1998). The Golgins are either peripherally or integrally associated with the Golgi apparatus with the C-terminus anchored at the Golgi membrane and N-terminus projecting out into the cytoplasm to capture interacting molecules. They were initially identified using sera from patients with auto-immune diseases such as Sjogren’s syndrome and rheumatoid arthritis (Chan and Fritzler, 1998; Table 1.4). A number of Golgins still remain to be identified: Golgins 35kDa to 260kDa (14 proteins), of which only Golgin-245 is sequenced so far (Kooy, 1994).

Table 1.4 The Golgins.

Protein MW Interactions Features Reference

Golgin-67 67kDa ? Cdc2 and Src kinase

phosphorylation motifs

Jakymiw et al., 2 0 0 0

Golgin-84 84kDa 7 _{Transport vesicle}

associated.

Bascom et al.,

1999

Golgin-95 95kDa ? A shorter form of

GM130.

Fritzler et al., 1993

Golgin-97 97kDa Rab6 60nm in length Barr, 1999

Golgin-160 (Grp Ip)

160kDa Sec34p,

Sec35p

COPII vesicle tether? Microtubule binding? Kim et al., 1999 Misumi et al., 1997 GM130 130kDa p l l 5 GRASP65

Part of COPI vesicle tether. 8 Onm in length Nakamura et al., 1995 Barr et al., 1997 Sonnichsen et al., 1998

GMAP-210 210kDa Minus ends

of

microtubules

May localize Golgi to centrosome.

Rios et al., 1994 Infante et al., 1999

Golgin-245 245kDa Rab6 Associated with non-

clathrin coated vesicles. Granin signature. 15 Onm in length Barr, 1999 Fritzler et al., 1995 Erlich et al., 1996 Gleeson et al., 1996

Giantin 372kDa p l l 5 Part of COPI vesicle

tether. 25 Onm in length. Linstedt and Hauri, 1993 Sonnichsen et al., 1998 85

The best characterized Golgins are GM130 and giantin (Nakamura et a l, 1995; Linstedt and Hauri, 1993), and function with p i 15 as components of a heterotemary COPI vesicle tether (Section 1.3.2; Sonnichsen et al., 1998; Shorter and Warren, 1999; Lesa et al., 2000; Dirac-Svejstrup et al., 2000). By analogy the other Golgins have been proposed to be involved in the tethering of vesicles to cistemae or cistemae to each other, however, this has not been so clearly demonstrated as for GM130 and giantin. The yeast homologue of Golgin-160 Grp Ip interacts with Sec34p, a COPII vesicle tethering protein, and has been proposed to be important for the maintenance of Golgi structure (Kim et al., 1999). Golgins -97 and -245 are targeted to the Golgi apparatus by a highly conserved C-terminal GRIP domain of approximately 50 amino acids (Munro and Nichols, 1999; Kjer-Nielsen et al., 1999; Barr, 1999) and this seems to involve an interaction with Rab6 (Barr, 1999). Rab6 has been implicated in intra-Golgi transport (Martinez et al., 1994, 1997; Mayer et al., 1996b), endosome to Golgi transport (Tsukada et al., 1999) as well as ER-Golgi retrograde transport (Echard et al., 1998; Girod et al., 1999; White et al., 1999) suggesting a possible role for these Golgins in a vesicle tethering reaction. The Golgins may also serve to anchor the Golgi matrix to the surrounding cytoskeleton. Both GMAP-210 and Golgin-160 have been implicated in microtubule binding (Infante et al., 1999; Misumi et al., 1997), and so may contribute to the position of the Golgi apparatus within the cell. Furthermore, isolated Golgi matrices are able to move along microtubule tracks in the presence of the molecular motor dynein (Fath et al., 1997).

The similarities between Golgins and SNAREs are also intriguing. Both are coiled-coil proteins with the same membrane orientation and predicted rod-like structure. Golgins are generally much longer than SNAREs so they could act at a greater distance from the membrane. Coupled with their flexibility they may permit the vesicle to sample a target membrane for a cognate SNARE, or simply increase the efficiency of transport by preventing vesicles from diffusing away. Similarly in Golgi stacking Golgins may act to bring nascent cistemae together in the initial stages of stacking, and then hand over to the SNAREs to complete the reaction. The Gradgrindian view o f science: facts come first, ideas later, is very rarely true. This is clearly demonstrated by one of the

Chapter 1_______________________________________________________ Introduction

extensions of the SNARE hypothesis, that states that the stacking of Golgi cistemae may be mediated by cognate v-/t-SNARE pairs (Rothman and Warren, 1994). Cistemal stacking would then reflect a specialized form of SNARE mediated docking (Section 1.3.3), and would differ in that the docking state must be frozen, and not be allowed to proceed on to the fusion step. This may be mediated by a hypothetical fusion clamp, or there may be SNARE isoforms that are only able to dock and not fuse membranes. Such a SNARE mediated stacking mechanism would be consistent with the ordered stacking (i.e. medial cistemae stack/dock with cis cistemae, and trans cistemae stack/dock with medial cistemae), close apposition and constant spacing of stacked cistemae (Rothman and Warren, 1994).

Another group of proteins, more commonly thought of as components of the actin cytoskeleton, may contribute to Golgi architecture, and these are the Golgi localized isoforms of ankyrin and spectrin (Beck et al., 1994, 1997; Devarajan et al., 1996; Stankewich et al., 1998; Beck and Nelson, 1998; De Matteis and Morrow, 1998). Given the importance of spectrin and ankyrin in maintaining erythrocyte plasma membrane stmcture (Marchesi and Steers, 1968; Morrow et al., 1997), it would seem likely that spectrin could play an analogous role in maintaining Golgi architecture. Spectrin forms a long, thin flexible rod of c. lOOnm in length and exists as a heterodimer of homologous but distinct subunits, termed a and P, arranged in a head to tail orientation. Spectrin heterodimers can self-associate end to end to form linear oligomers that act as a minimal stmctural unit of a membrane skeleton (Shotton et al., 1979). pill spectrin is a ubiquitously expressed spectrin isoform and localizes with the Golgi and unidentified cytoplasmic vesicles (Stankewich et al., 1998). As yet no Golgi associated a-spectrin has been detected and it may be that pIII spectrin forms homodimers (De Matteis and Morrow, 1998). Two isoforms of Golgi ankyrin have been identified a 119kDa form and a 195kDa form, and these may serve to link the spectrin lattice to the Golgi membrane (Beck and Nelson, 1998). Golgi localized actin or centractin have been hypothesized to crosslink linear spectrin oligomers to generate a two dimensional lattice on the cytoplasmic face of the Golgi membrane (Beck and Nelson, 1998). These Golgi spectrin and ankyrin isoforms form part of the detergent insoluble Golgi matrix

(Beck et al., 1997), but it is as yet unclear whether the putative spectrin skeleton is found in the intercistemal space or restricted to the cis- or rm^j-aspects of the Golgi stack (De Matteis and Morrow, 1998).

Assembly of the spectrin/ankryin Golgi matrix is regulated by the small GTPase A RFl, and is dependent on ARFl stimulated phosphatidylinositol bisphosphate (PIP2) synthesis, which binds a PH domain in Golgi spectrin (Godi et al., 1998). ARPl acts to a recruit phosphatidylinositol-4-OH kinase-|3 (PI-4-0H kinase) to the Golgi membrane and in so doing stimulate PIP2 synthesis (Godi et al., 1999). Disruption of the spectrin/ankyrin Golgi matrix by transfecting cells with truncated forms of Golgi spectrin inhibits ER-Golgi transport of VSVG and NaVK"^-ATPase, but not that of E- cadherin (Devarajan et al., 1997). However, disruption of the matrix did not seem to perturb Golgi morphology (Devarajan et al., 1997).