The nature of the complex containing GRASP65, a membrane protein involved in establishing the stacked structure of the Golgi apparatus, and GM130, a putative Golgi matrix protein and vesicle docking receptor, was investigated. Gel filtration revealed that GRASP65 and GM130 interact in detergent extracts of Golgi membranes under both interphase and mitotic conditions, and that this complex can bind to the vesicle docking protein p115. Using in vitro translation and site‐directed mutagenesis in conjunction with immunoprecipitation, the binding site for GRASP65 on GM130 was mapped to the sequence xxNDxxxIMVI‐COOH at the C‐terminus of GM130, a region known to be required for its localization to the Golgi apparatus. The same approach was used to show that the binding site for GM130 on GRASP65 maps to amino acids 189–201, a region conserved in the mammalian and yeast proteins and reminiscent of PDZ domains. Using green fluorescent protein (GFP)‐tagged reporter constructs, it was shown that one essential function of the interaction between GRASP65 and GM130 is in the correct targeting of the two proteins to the Golgi apparatus.
The Golgi apparatus can be visualized under the light microscope as a reticulum adjacent to the nucleus in animal cells (Golgi, 1898; Burke et al., 1982). This reticulum consists of a number of flattened disc‐shaped cisternae aligned in parallel to form stacked structures linked at their edges by tubules (Rambourg and Clermont, 1990). Morphological studies have shown that filamentous material linking adjacent Golgi cisternae can be visualized under the electron microscope in a wide variety of plant and animal cells (Mollenhauer, 1965; Amos and Grimstone, 1968; Franke et al., 1972). Biochemical studies showed that these cross‐bridges were proteinaceous, and that proteolytic treatment can disrupt the stacked organization of isolated Golgi stacks (Cluett and Brown, 1992), presumably by digestion of these filamentous cross‐bridges. Subsequently, a putative intercisternal Golgi matrix was isolated from purified rat liver Golgi stacks, and was shown to be able to bind to enzymes of the medial/trans‐Golgi, but not to those of the cis‐Golgi and cis‐Golgi network (Slusarewicz et al., 1994). One component of this matrix, GM130, was identified, and turned out to be a member of a family of autoantigens possessing extensive coiled‐coil motifs (Nakamura et al., 1995). Many of these autoantigens are peripheral or integral membrane proteins associated with the Golgi apparatus (Kooy et al., 1992; Fritzler et al., 1993, 1995; Linstedt and Hauri, 1993; Seelig et al., 1994; Nakamura et al., 1995; Misumi et al., 1997), and thus are candidates for components of an intercisternal Golgi matrix (Barr and Warren, 1997). Further characterization of GM130 revealed that it acts as a receptor for p115 (Nakamura et al., 1997), a factor involved in the tethering of Golgi transport vesicles prior to their fusion (Waters et al., 1992; Sapperstein et al., 1995). GM130 is phosphorylated in mitosis and can no longer bind p115, an observation that has been linked to the block in vesicle traffic during cell division (Levine et al., 1996; Nakamura et al., 1997). How this role of GM130 in vesicle docking is related to its putative function in a structural matrix of the Golgi is unclear at present.
Recently, we identified a 65 kDa membrane‐associated protein, GRASP65, involved in the reassembly of Golgi stacks in a cell‐free system (Barr et al., 1997). Analysis of the sequence of GRASP65 revealed a central domain conserved between yeast and mammals, and a less well conserved C‐terminal domain containing many serine and proline residues possibly involved in cell‐cycle regulation (Barr et al., 1997). GRASP65 is thought to be associated with membranes by means of an N‐terminal myristic acid anchor, and is found in a complex with GM130 (Barr et al., 1997). It could thus provide a link between vesicle traffic and cisternal stacking. We therefore set out to define the interactions between GRASP65, GM130 and p115 in an attempt to understand better the roles of these proteins in cisternal stacking. This work uses a combination of biochemical studies on purified Golgi membranes, together with deletion and point mutagenesis of GRASP65 and GM130, to demonstrate and characterize the direct interaction between these two proteins and the role this might play in their targeting to Golgi membranes.
Association of GRASP65 and GM130 under both interphase and mitotic conditions
We have shown previously that GRASP65 and GM130 can be co‐immunoprecipitated from a detergent extract of purified rat liver Golgi membranes, and that both GRASP65 and GM130 are present in high molecular weight fractions by gel filtration, indicating that they may be part of the same large complex (Nakamura et al., 1995; Barr and Warren, 1997). During mitosis, the interactions between Golgi cisternae are lost and the cisternae become fragmented into many small vesicles and tubules (Thyberg and Moskalewski, 1985; Lucocq et al., 1987). The putative GRASP65–GM130 complex is thought to be involved in these events and we believed that it might undergo some cell cycle‐regulated changes in size and composition, since it is known that both these proteins are phosphorylated during mitosis (Barr et al., 1997; Nakamura et al., 1997). To investigate this possibility further, detergent extracts of Golgi membranes pre‐treated with either interphase or mitotic cytosols were fractionated by gel filtration over a Superose 6 column, and the fractions analysed by SDS–PAGE and Western blotting with specific antibodies (Figure 1). Comparison of the distributions of GM130 and GRASP65 under interphase (Figure 1A) and mitotic (Figure 1B) conditions showed that the peak positions of both GM130 and GRASP65 were very similar under both conditions tested (Figure 1A and B, fraction 10), corresponding to an apparent molecular mass of ∼1.2 MDa. The only indication that a change in the composition or conformation of the GRASP65–GM130 complex had occurred was a shift in the shoulder of the peaks from the left under interphase conditions(Figure 1A, fraction 9) to the right under mitotic conditions (Figure 1B, fraction 11). As a control that should not undergo any cell cycle changes in molecular weight during gel filtration, the Golgi mannosidase II (Mann II) was investigated and found to gel filter at 200–250 kDa (Figure 1, fractions 14 and 15) under both interphase and mitotic conditions as expected for a dimer of a 120 kDa, mainly globular protein (Moremen et al., 1991; Nakamura et al., 1995). These data support the idea that GRASP65 and GM130 are present as part of a stable complex in Golgi membranes, under both interphase and mitotic conditions.
Direct interaction of GRASP65 with GM130 and binding of the GM130–GRASP65 complex to the vesicle docking protein p115
To provide evidence that GRASP65 binds directly to GM130, far Western analyses were performed. A detergent extract of Golgi membranes was fractionated by SDS–PAGE, and the proteins transferred to nitrocellulose membranes; these were probed with recombinant histidine‐tagged GRASP65 (Figure 2A, left panel), then stripped to remove the bound GRASP65 and reprobed with antibodies to GM130 (Figure 2A, right panel). Recombinant GRASP65 bound to a protein of 130 kDa, and this binding increased with increasing concentrations of probe. Western blotting with antibodies to GM130 revealed that the protein recognized by the recombinant GRASP65 had the same molecular weight as GM130. This provides evidence that GRASP65 directly interacts with a 130 kDa protein in Golgi membranes that is likely to be GM130.
GM130 is known to function as a receptor for the vesicle transport protein p115, thought to be involved in the tethering of membranes. It has been shown that this interaction is prevented in mitosis by phosphorylation of the N‐terminus of GM130 (Nakamura et al., 1997). The relationship between this function of GM130 in vesicle traffic and that of GRASP65 in establishing the stacked cisternal structure of the Golgi apparatus is unclear. One possible explanation is that two functionally discrete pools of GM130 exist in Golgi membranes, one complexed with GRASP65 and the other with p115. To investigate this, p115 binding assays were performed (Figure 2B), using a detergent extract of Golgi membranes incubated with control beads or beads on which purified rat p115 had been immobilized as described previously (Nakamura et al., 1997). Aliquots of the total extract, and the bound and unbound fractions from either control or p115 bead‐containing incubations were analysed by SDS–PAGE and Western blotting with antibodies to GM130 and GRASP65. Reactivity corresponding to GM130 and GRASP65 was found to bind to p115 (Figure 2B, p115 beads) but not to control beads (Figure 2B, Control), and the ratio between the GM130 and GRASP65 signals was found to be similar to that observed in the starting extract (Figure 2B, RLG extract). In previous studies using p115 beads, GRASP65 was not detected since it does not stain well with Coomassie Blue, and is in the same region of the gel as an abundant contaminating protein, serum albumin (see Figure 1 in Nakamura et al., 1997). It is unlikely, therefore, that two pools of GM130 exist distinguished by their ability to bind p115. This result supports the view that the complex of GRASP65–GM130–p115 functions in both vesicle docking and cisternal stacking.
Reconstitution of the interaction between GRASP65 and GM130 in an in vitro assay
In an attempt to characterize the complex containing GRASP65 and GM130, it was decided to see if it could be reconstituted, at least in part, with only these two proteins, using coupled in vitro transcription–translation in conjunction with immunoprecipitation. In vitro translation reactions were carried out with plasmids encoding either GRASP65 (Figure 3, lane 1) or GM130 (Figure 3, lane 2) alone, or GRASP65 and GM130 together (Figure 3, lane 3). Aliquots of the total reactions were analysed by SDS–PAGE and autoradiography; the position of GM130 is marked by an open triangle and that of GRASP65 by an asterisk (Figure 3A). As expected, the production of GM130 and GRASP65 depended on the presence of their respective plasmids. To assay for an interaction between the two proteins, immunoprecipitations were performed from these in vitro translation reactions with antibodies to GM130 or GRASP65. Analysis of the bound material by SDS–PAGE and autoradiography revealed that antibodies to GM130 precipitated GM130 (Figure 3B, lane 2) but not GRASP65 (Figure 3B, lane 1), and antibodies to GRASP65 precipitated GRASP65 (Figure 3C, lane 1) but not GM130 (Figure 3C, lane 2). If GRASP65 was translated in vitro together with GM130, it was found to be precipitated by antibodies to GM130 (Figure 3B, lane 3) and, conversely, GM130 was found to be immunoprecipitated by antibodies to GRASP65 (Figure 3C, lane 3). Using this method, a complex between GRASP65 and GM130 can be reconstituted with only these two proteins, and thus their interaction is amenable to further analysis in vitro.
GRASP65 binds to the C‐terminus of GM130
To define the binding site for GRASP65 on GM130, a series of N‐ and C‐terminal deletion constructs in GM130 were constructed (Figure 4). These constructs were then used in the in vitro translation and immunoprecipitation assay together with a wild‐type GRASP65 construct. Analysis of the total in vitro translation reactions by SDS–PAGE and autoradiography revealed that all constructs were expressed equally, taking into account the different number of methionines present in the various deletion constructs (Figure 5A). Deletions from the N‐terminus of GM130 to position 441 and 690 (Figure 5B, lanes 9 and 10) had no effect on the interaction with GRASP65. The weaker signal for GRASP65 and GM130ΔN690 is due to a decreased efficiency in the immunoprecipitation rather than a lack of interaction between the two proteins, since the major epitope recognized by the polyclonal antiserum used for these experiments is deleted in this GM130 construct. Since the N‐terminal 75 amino acids contain the binding site for p115, this indicated that the interaction of GRASP65 with GM130 is distinct from that of p115. Deletions from the C‐terminus of GM130 to positions 887, 436 and 237, known to abolish the targeting of GM130 to the Golgi apparatus (Nakamura et al., 1997), also resulted in the loss of binding to GRASP65 (Figure 5B, lanes 2–4). The molecular weight of the GM130ΔC436 construct is similar to that of GRASP65, and it runs immediately below it on SDS–PAGE, giving the appearance of a very broad band in the total in vitro translation (Figure 5A, lane 3). Careful examination of the immunoprecipitation (Figure 5B, lane 3) reveals that the upper part of this broad band corresponding to GRASP65 is missing, indicating that there is no interaction of GRASP65 with this GM130 deletion construct. Introduction of a myc tag at the extreme C‐terminus (GMfmyc) also had the same effect, abolishing both binding to GRASP65 (Figure 5B, lane 8) and targeting to the Golgi apparatus (N.Nakamura unpublished data). Together, these data suggested that the binding site for GRASP65 lay in the last 100 amino acids of GM130 including the free C‐terminus, and it was decided to focus on this region more closely. Additional deletion constructs in GM130 were made in which the last 62, 26 or three amino acids were removed by the introduction of in‐frame stop codons. None of the constructs GM130ΔC924, ΔC960 or ΔC983 was able to interact with GRASP65 (Figure 5B, lanes 5–7), confirming that the binding site lay in this region. Alanine‐scanning mutagenesis of the last 11 amino acids of GM130 from D976 to I986 was then performed in order to define the GRASP65‐binding site. These GM130 constructs were then used in the in vitro translation assay; quantitations of such experiments are shown in Figure 5C. From this analysis, it can be seen that the four hydrophobic amino acids Ile–Met–Val–Ile lying at the C‐terminus, in conjunction with the Asn–Asp pair at position 978–979, form the site of GRASP65 interaction with GM130.
Gel filtration of in vitro translated GM130 and GRASP65
To confirm that the complex formed by the in vitro translation of GM130 and GRASP65 was related to that present in Golgi membranes, it was fractionated by gel filtration over Superose 6 as described previously (Figure 1). For this purpose, three sets of assays were performed, with in vitro translation of: GM130 and GRASP65 individually, followed by their mixing on ice (Figure 6A); GM130 and GRASP65 together (Figure 6B); and GM130 lacking the last 100 amino acids (GM130ΔC887) and GRASP65 together (Figure 6C). When GM130 and GRASP65 were in vitro translated separately then mixed on ice, they gave two clearly resolved peaks by gel filtration, GM130 at fraction 10 (Figure 6A, ●) and GRASP65 at fraction 15 (Figure 6A, ○). From the column calibration, these elution positions would correspond to apparent mol. wts of 1.2 MDa for GM130 and 200 kDa for GRASP65. When GM130 and GRASP65 were in vitro translated together and fractionated by gel filtration, the peak of GM130 was shifted slightly to the left (Figure 6B, ●) and a second peak of GRASP65 was present at fraction 10 (Figure 6B, ○) consisting of ∼25% of the total GRASP65. Since GRASP65 is present in molar excess over GM130, the peak at fraction 15 (Figure 6B, ○) is still present. This is consistent with the formation of a stable interaction between GRASP65 and GM130 in the in vitro translation assay, as was also shown by immunoprecipitation with specific antibodies. Finally, when GM130ΔC887 and GRASP65 were in vitro translated together and fractionated by gel filtration, they gave two clearly resolved peaks by gel filtration, GM130 at fraction 10 (Figure 6C, ●) and GRASP65 at fraction 15 (Figure 6C, ○). This is consistent with the lack of an interaction between GRASP65 and GM130ΔC887 in the in vitro translation assay, as was also shown using immunoprecipitation with specific antibodies (Figure 5B). Gel filtration confirms the results obtained by immunoprecipitation implicating the C‐terminus of GM130 in the direct binding of GRASP65, and shows that these two proteins alone could form the 1.2 MDa complex present in detergent extracts of Golgi membranes.
GM130 binds to a conserved region of GRASP65
To define the binding site for GM130 on GRASP65, a series of N‐ and C‐terminal deletion constructs in GRASP65 were constructed (Figure 7A). These constructs were then used in the in vitro translation and immunoprecipitation assay together with a wild‐type GM130 construct. Analysis of the total in vitro translation reactions by SDS–PAGE and autoradiography revealed that all constructs were expressed equally, taking into account the different number of methionines present in the various deletion constructs (Figure 8A). Deletions of the C‐terminus of GRASP65 to position 237 and 314 (GRASP65ΔC237 and ΔC314) had little effect on the interaction with GM130 (Figure 8B, lanes 3 and 4). The decreased amount of the GRASP65ΔC314 immunoprecipitating with GM130 (Figure 8B, lane 4) may be due to incorrect folding of this protein, since larger deletions, for example to amino acid 237 (Figure 8B, lane 3), interact with GM130 to the same extent as the full‐length GRASP65 (compare lanes 1 and 3). Deletions of the C‐terminus of GRASP65 to position 152 (GRASP65ΔC152) and the N‐terminus to position 198 (GRASP65ΔN198) resulted in the loss of binding to GM130 (Figure 8B, lanes 2 and 5). These data suggested that the binding site for GM130 lay in the 90 amino acids of GRASP65 from amino acids 112 to 202, and it was decided to study this region, domain 2, more closely. A double deletion from the N‐ and C‐termini comprising only domain two of GRASP65 (GRASP65ΔN112ΔC202) was able to interact with GM130, indicating that this domain alone contains the binding site for GM130 (Figure 8B, lane 6). Interestingly, this region of GRASP65 is in the part of the protein most highly conserved between different organisms, including mammals and yeasts (Barr et al., 1997). In the rat GRASP65, the conserved sequence comprising Trp112–Pro202 (Figure 7A, domain 2), is repeated imperfectly from amino acids Phe16 to Ser108 (Figure 7A, domain 1). The deletion analysis indicates that domain 1, the imperfectly repeated sequence of domain 2, is unable to bind to GM130, since the GRASP65ΔC152 deletion possessing domain 1 but lacking domain 2 does not interact with GM130. Alignment of the conserved domain of rat GRASP65 from Trp112 to Pro202 containing the binding site for GM130 with this imperfectly repeated sequence from Phe16 to Ser108 reveals that they are similar throughout their entire length (Figure 7B). Since the deletion mutagenesis of GRASP65 indicates that the binding site for GM130 lies after amino acid 152, the differences in the alignment from this point on are likely to explain why the conserved domain but not its repeated form constitute a binding site for GM130. The least conserved part of the alignment between the conserved domain and its repeat coming after amino acid 152 of GRASP65 lies between amino acids 194 and 202, with only one of 12 amino acids being identical (Figure 7B). This region is, however, conserved from rat GRASP65 domain 2 to the yeast homologues (Figure 7C). It was decided, therefore, to focus on the region lying between Cys191 and Pro202 of GRASP65, together with some additional conserved flanking residues. Alanine‐scanning mutagenesis of the 22 amino acids of GRASP65 from Pro179 to Pro202, excluding the alanine residues at 181 and 182, was then performed in order to define the GM130‐binding site. These GRASP65 constructs were then used in the in vitro translation assay; quantitations of such experiments are shown in Figure 8C. From this analysis, it can be seen that the key residues for the interaction of GRASP65 with GM130 are from Gly194 to Ile201 and that these lie in the least conserved region of the repeated domain as predicted. To provide further evidence that the effects of the mutations in GRASP65 define a specific binding site for GM130, and do not grossly alter the structure of the protein, gel filtration analysis of all the constructs used was performed. In vitro translated full‐length, N‐ and C‐terminal deletions and the alanine‐scanning point mutant forms of GRASP65 were analysed by gel filtration over Superose 6 in 200 mM KCl, and the apparent molecular masses were calculated from their elution positions relative to standard marker proteins. The results of such an analysis are summarized in Table I, and indicate that wild‐type GRASP65 and the GRASP65Δ314 deletion behave as dimers or trimers under these conditions. All of the alanine‐scanning mutants were found to be indistinguishable from the wild‐type GRASP65, indicating that these proteins are also forming dimers or trimers. The GRASP65Δ237 deletion lacking the predicted elongated tail, domain 3, forms dimers, whereas the GRASP65Δ152 deletion lacking domains 2 and 3 only forms monomers. While it is possible that di‐ or trimerization of GRASP65 is important for its binding to GM130, the observations presented here support the conclusion that the deletion and alanine‐scanning point mutations in GRASP65 have defined a binding site for GM130 on this protein.
Targeting of the GRASP65–GM130 complex to the Golgi apparatus
Having defined the sites at which GM130 and GRASP65 interact, we then investigated the role of this interaction in the targeting of the two proteins to the Golgi apparatus. Transient transfection studies were carried out using either the tail region, residues 816–986, of GM130 fused to the C‐terminus of green fluorescent protein (GFP), or the first 202 amino acids of GRASP65 fused to the N‐terminus of GFP. The same mutations that disrupted the interactions between the two proteins were then introduced into these constructs, and the effects these had on the targeting of each protein were examined.
When fused to GFP, the wild‐type GM130 tail is able to target correctly to the Golgi apparatus (Figure 9A), whereas deletion of the last three amino acids of the tail abolishes this (Figure 9B). Alanine‐scanning point mutations through the tail of GM130 from amino acids 978 to 986 reveal that the ability to bind GRASP65 (Figure 5C) correlates with the targeting of the corresponding GFP reporter construct to the Golgi apparatus (Figure 9C–K). Point mutations that lowered the binding of GM130 to GRASP65 (GM130 V985A and I986A) resulted in an intermediate phenotype, with a small amount of Golgi staining and a large soluble pool of the GFP reporter construct (Figure 9J and K). When fused to GFP, the wild‐type N‐terminal 202 amino acids of GRASP65 are able to target correctly to the Golgi apparatus (Figure 9L), whereas mutation of the glycine at position two of GRASP65 to abolish the myristoylation site prevents this (Figure 9M). Point mutations at amino acids 196 and 199, abolishing the ability of GRASP65 to bind to GM130 (Figure 8C), prevent its localization to the Golgi apparatus (Figure 9N and O). This indicates that the N‐myristoylation site and the domain defined as the binding site for GM130 are important for targeting GRASP65 to the Golgi apparatus.
Therefore, the interaction between GRASP65 and GM130 is required not only for them to form a stable complex, but also for their correct localization to the Golgi apparatus.
A protein complex involved in Golgi transport, vesicle docking and the stacking of Golgi cisternae
We have shown that the 1.2 MDa complex present in Golgi membranes containing GRASP65 and GM130 is able to interact with the vesicle transport factor p115 under interphase conditions. Using far Western blotting and an in vitro translation system, we were able to demonstrate a direct interaction between GM130 and GRASP65, and to map the binding sites on both proteins. For GM130, the region interacting with GRASP65 proved to be at the extreme C‐terminus and to correlate with the region previously shown to be involved in the targeting of GM130 to the Golgi apparatus. This indicates that one function of GRASP65 is in the Golgi targeting of GM130. Previously, using a functional assay for the identification of proteins involved in the stacking of Golgi cisternae, we found GRASP65 and, GM130. The fact that GM130 and GRASP65 directly interact and were identified by this approach implies that they may both play a role in the stacking of Golgi cisternae. The known function of p115 as a factor tethering Golgi transport vesicles to their target membranes and the role of GM130 as one of its receptors suggests that they may also function in the process by which cisternae form stacks. In addition to tethering vesicles during docking events, p115 might also, in conjunction with GRASP65, link cisternae together during the process of stacking, allowing the establishment of more stable cisternal cross‐bridges, analogous to the SNARE pair formation during vesicle docking.
The interaction of GRASP65 with GM130 is reminiscent of the binding of PDZ domains to their ligands
The binding site for GM130 on GRASP65 was mapped to a region between amino acids 194 and 201 in the most conserved part of GRASP65 which extends from amino acids 112 to 202, and is imperfectly repeated between amino acids 16 and 108. Within the region identified as the binding site for GM130 is the sequence Gly–Tyr–Gly–Tyr (GYGY) that is reminiscent of the sequence Gly–Leu–Gly–Phe (GLGF), a conserved motif of the PDZ domain (Ponting et al., 1997). It should be noted that this motif is not absolutely conserved, and exists in many different variants (Ponting et al., 1997). The binding of GRASP65 to the last four hydrophobic amino acids of GM130 is reminiscent of the interaction of PDZ domain‐containing proteins with the extreme C‐terminal four amino acids of their target proteins (Saras and Heldin, 1996). Specific PDZ domains have been shown to bind specific peptide sequences in vitro, which explains their specificities in vivo (Songyang et al., 1997). Typically, they act as adaptors in the assembly of multifunctional protein complexes involved in signalling events at the surfaces of cellular membranes (Cowburn, 1996; Ponting et al., 1997), for example the binding of PSD‐95, a guanylate kinase, to the N‐terminus of the NMDA receptor 2B (Kornau et al., 1995). In the case of GRASP65, this domain is involved in the targeting to the Golgi membrane of GM130, a molecule that would otherwise be cytosolic. It should be noted that although the PDZ domain has been found only in multicellular animals, there is evidence for the presence of PDZ‐like domains capable of binding C‐terminal tetrapeptide sequences in yeasts, plants and bacteria (Ponting, 1997).
The most conserved region of GRASP65 forms part of a domain required for the binding of GM130, and localization to the Golgi apparatus. GRASP65, although lacking a transmembrane domain, is N‐myristoylated (Barr et al., 1997). This modification is necessary for its interaction with membranes (Figure 9M), but not its specific localization to the Golgi apparatus (Johnson et al., 1994; Bhatnagar and Gordon, 1997). One possibility is that the GRASP65–GM130 complex targets specifically to the Golgi apparatus by interaction of the GRASP65 N‐terminal domains 1 and 2 with Golgi‐specific proteins possessing transmembrane domains. Alternatively, it is possible that GM130 can interact with a Golgi‐specific transmembrane protein only when bound to GRASP65, and that this explains how the GRASP65–GM130 complex targets only to Golgi membranes. We currently are investigating these possibilities.
In the yeast homologues of rat GRASP65, the sequence implicated in the binding of GM130 is highly conserved, indicating that these proteins may be binding to the C‐terminus of a GM130 homologue in these organisms. Analysis of the Saccharomyces cerevisiae genome reveals no protein that possesses the C‐terminal sequence in GM130 that is recognized by GRASP65. This implies that the changes (Figure 7C) from rat (GIGYGYLHRI) to Schizosaccharomyces pombe (GVGHGVLHRL) and S.cerevisiae (NVGYGFLHRI) GRASP65 have altered its specificity, such that it recognizes a different tetrapeptide signal, something that has also been observed for classical PDZ domains (Ponting et al., 1997).
Stoichiometry of GRASP65 and GM130
Fractionation of the complex containing GM130 and GRASP65 in detergent extracts of Golgi membranes by gel filtration gave an apparent molecular mass of 1.2 MDa, similar to that obtained for the complex reconstituted by in vitro translation of the two proteins. Since GM130, based on its sequence, is thought to form a long rod‐like structure similar to the myosin heavy chain (Nakamura et al., 1995), these values are unlikely to represent the true molecular weight of the complex. It is known that elongated rod‐like proteins have anomalous behaviour when analysed by gel filtration, compared with the globular protein standards used to calibrate the column, and hence give much higher apparent molecular weights (Ackers, 1970). Consistent with this interpretation is the observation that in vitro translated GM130 behaves almost exactly the same as GM130 from Golgi membranes on gel filtration, irrespective of the presence of GRASP65. Thus, from this analysis, it is not possible to estimate the stoichiometry of GRASP65 and GM130 or make any statements as to the presence of other proteins in the complex isolated from Golgi membranes. Under the conditions used, no p115 or giantin were found co‐fractionating with this complex, although it should be noted that the extraction conditions were chosen to disrupt low‐affinity interactions in order to simplify the analysis of the GRASP65–GM130 complex. Consideration of the interaction of this complex with p115 does give some clues as to the stoichiometry of GM130 and GRASP65. It is known that p115 forms a parallel homodimer comprising a C‐terminal coiled‐coil rod and a globular N‐terminal domain (Sapperstein et al., 1995). The N‐terminus of GM130 binds to this p115 homodimer (Nakamura et al., 1997), and is therefore likely to be a dimer itself, something also suggested by its sequence similarity with other proteins either known or thought to form coiled‐coil homodimers (Nakamura et al., 1995). It is probable that GRASP65 is also a dimer, an interpretation supported by the behaviour of in vitro translated GRASP65 as a protein of 200 kDa, corresponding to a dimer or trimer based on its mol. wt of 65 kDa by SDS–PAGE. This view is supported by observations that PDZ domains can be important for dimerization, and the fact that the form of GRASP65 lacking its PDZ‐like domain (Table I, GRASP65Δ152) behaves as a monomer.
Therefore, the minimal complex of GRASP65 and GM130 is likely to consist of a dimer of GRASP65 binding to the C‐terminus of a coiled‐coil dimer of GM130. This would result in the orientation of the GM130 rod such that the N‐terminal p115‐binding site projected away from the membrane, possibly enhancing its function in the capture and tethering of vesicles in docking and perhaps cisternal membranes during Golgi stack formation.
Materials and methods
Antibodies were as follows: NN5‐1 rabbit polyclonal (Nakamura et al., 1995) and mouse monoclonal 2C10 (Nakamura et al., 1997) raised against the 97 kDa fragment of GM130; mouse monoclonal 53FC3 to Mann II (Burke et al., 1982); and rabbit polyclonals FBA30 and 31 to GRASP65.
SDS–PAGE, Western and far Western blotting
Protein samples were solubilized in SDS–PAGE sample buffer, boiled for 3 min and analysed on 10 and 12% SDS–polyacrylamide gels (Blobel and Dobberstein, 1976). Western blotting was performed using a semi‐dry blotter onto Hybond‐C (Amersham Life Science, UK). Blocking and antibody incubations were performed in phosphate‐buffered saline (PBS) plus 10% (w/v) low fat skim milk powder. All secondary antibodies or streptavidin were horseradish peroxidase (HRP) conjugates (Tago, Buckingham, UK), detected using ECL (Amersham Life Science, UK). Far Westerns were performed in PBS plus 4% (w/v) low fat skim milk powder and 0.2% (w/v) Triton X‐100. Blocking incubations were performed for 12–18 h at 4°C, and binding was carried out for 8 h at 4°C. Antibody incubations were for 20 min at room temperature, and all washes were three times for 5 min at room temperature. Detection was using ECL plus (Amersham Life Science, UK).
The GM130 constructs in pBluescriptII have been described previously (Nakamura et al., 1997). GRASP65 constructs were made using the Quickchange method (Stratagene, UK) for point mutations, PCR mutagenesis to introduce stop codons for C‐terminal deletions, or convenient restriction sites for N‐terminal deletions in the pcDNA3.1 mammalian expression vectors (Invitrogen). GFP‐tagged constructs were constructed in the pEGFP mammalian expression vectors (Clontech Laboratories Inc.). Details of the primers used are available from the corresponding author.
In vitro transcription–translation
Constructs for in vitro transcription–translation were in pcDNA3.1+ for GRASP65 and pBluescriptII for GM130. Reactions of 50 μl were performed with the T7 polymerase for GRASP65 and the T3 polymerase for GM130, according to the manufacturer's instructions (Promega) using 0.5 μg of plasmid DNA and methionine minus amino acid mix plus 4 μl of [35S]l‐methionine (typically 1400 Ci/mmol and 11 mCi/ml, ICN Pharmaceuticals Inc.). For in vitro translation reactions involving both GRASP65 and GM130 constructs, 50 μl reactions were set up for each construct, these were incubated separately for 60 min at 30°C, then mixed and incubated for a further 2 h at 30°C.
Immunoprecipitations were performed in IP buffer [20 mM HEPES–KOH pH 7.3, 200 mM KCl, 0.5% (w/v) Triton X‐100] using 4 μl of the appropriate antiserum and either 10 μl of packed protein A− or protein G–Sepharose (Pharmacia) for rabbit polyclonal and mouse monoclonal antibodies, respectively. After binding for 60 min at 4°C, the beads were washed four times with 500 μl of IP buffer, eluted with 30 μl of SDS–PAGE sample buffer, and the eluate analysed as appropriate.
p115 binding assay
Golgi membranes (10 μg) were extracted in 200 μl of IP buffer for 15 min on ice then centrifuged at 14 000 r.p.m. for 2 min in an Eppendorf microfuge to remove any insoluble material. Binding assays were performed in a total volume of 200 μl using 10 μl of packed p115 beads and 200 μl of Golgi membrane extract. After rotating for 60 min at 4°C to allow any interactions to form, the beads were washed four times with 500 μl of IP buffer, eluted with 30 μl of SDS–PAGE sample buffer, and the eluate analysed by SDS–PAGE and Western blotting.
Standard disassembly assays were carried out using interphase or mitotic HeLa cytosols at a final concentration of 10 mg/ml and 50 μg of purified rat liver Golgi membranes, as described previously (Misteli and Warren, 1994). After incubations with cytosol, Golgi membranes were recovered by centrifugation through a 1 ml cushion of 0.5 M sucrose in 100 mM potassium phosphate, pH 6.7, at 55 000 r.p.m. for 20 min at 4°C in the TLS55 rotor.
Samples to be analysed by gel filtration were adjusted to 500 μl and injected on to a Superose 6 HR10/30 column (Pharmacia) pre‐equilibrated in 20 mM HEPES–KOH pH 7.3, 200 mM KCl, 0.25% (w/v) Triton X‐100. Fractions of 1.0 ml were collected at a flow rate of 0.3 ml/min. Aliquots of each fraction were analysed by SDS–PAGE and either Western blotting or autoradiography as appropriate.
Transient transfections and immunofluorescence
HeLa cells were plated at 30% density on glass coverslips in 2 cm wells and left to attach for 12 h. They were then transfected using lipofectin (Gibco‐BRL, UK) according to the manufacturer's instructions. After 28 h, the cells were fixed in 3% paraformaldehyde and processed for immunofluorescence with appropriate antibodies.
F.A.B. would like to thank Dr Sylvie Urbé and Dr Martin Lowe for reading the manuscript, Dr Martin Lowe for purified p115 immobilized on beads, Michael D.Baron for his excellent implementation of Boxshade for the Macintosh, and Iain Goldsmith and the ICRF oligonucleotide synthesis service for providing oligonucleotides. N.N. was supported by a Wellcome Trust Postdoctoral Fellowship. F.A.B. was the recipient of a Beit Memorial Fellowship and an ICRF postdoctoral fellowship during the course of this work.
- Copyright © 1998 European Molecular Biology Organization