Following the Fate In Vivo of COPI Vesicles Generated In Vitro

Abstract

COPI vesicles are a class of transport carriers that function in the early secretory pathway. Their fate and function are still controversial. This includes their contribution to bidirectional transport within the Golgi apparatus and their role during cell division. Here we describe a method that should address several open questions about the fate and function of COPI vesicles in vivo. To this end, fluorescently labeled COPI vesicles were generated in vitro from isolated rat liver Golgi membranes, labeled with the fluorescent dyes Alexa-488 or Alexa-568. These vesicles appeared to be active and colocalized with endogenous Golgi membranes within 30 min after microinjection into mammalian cells. The COPI vesicle-derived labeled membrane proteins could be classified into two types that behaved like endogenous proteins after Brefeldin A treatment.

Transport of proteins and lipids between the different organelles of the secretory pathway is mediated by small transport vesicles. To date, at least three different types of vesicles have been well characterized, defined by their coat components (1). The protein coat of COPII vesicles, responsible for anterograde transport from the endoplasmic reticulum (ER) to the Golgi apparatus, comprises the heterodimeric complex Sec23/24, the heterotetrameric complex Sec13/31, and the small GTPase Sar1 (2–4). Transport between the trans-Golgi network, plasma membrane, endosomes and lysosomes is mediated by clathrin-coated vesicles (CCVs), which contain clathrin and are defined by members of the small GTPase family Arf (Arf1/Arf6) as well as different adaptor proteins (e.g., AP1–AP4 or GGA) (5,6). The coat of COPI vesicles is made of a heptameric protein complex termed coatomer and the small GTPase Arf1 (7–10). More recently, mammalian coatomer complex was found to represent a mixture of four complex isoforms, defined by the combination of two isoforms of γ-COP and ζ-COP, respectively (11).

COPI vesicles were first described as a class of non-CCVs that mediate transport between the Golgi stacks (12) and retrograde transport from the Golgi to the ER (13). The cycling of COPI vesicles between a donor and a target membrane involves recruitment of coat components, cargo uptake and coat polymerization, vesicle budding, and eventually uncoating, followed by fusion with the target membrane. COPI vesicle formation can be mimicked on liposomes with Arf1, coatomer, cytoplasmic tails of transmembrane proteins such as the p24 proteins and GTP as minimal components (14). As an initial step in COPI vesicle formation, Arf1-GDP is recruited by binding to the cytoplasmic tails of p24 proteins (15) or to the SNARE, membrin (16). After exchange of GDP to GTP, mediated by the exchange factor GBF1 (17), Arf1 dissociates from the p24 proteins and stably associates with the membrane via its N-terminal amphipathic helix and the attached myristoyl residue (18). Activated Arf1, together with the transmembrane proteins then recruits coatomer, by binding to several of its subunits (19,20). This causes polymerization of the complex and budding of a coated vesicle (21,22). Once formed, the vesicles must be uncoated before they can fuse with their target membrane. Uncoating is catalyzed by Arf-GTPase-activating proteins that stimulate hydrolysis of Arf-GTP, resulting in Arf-GDP that dissociates from the membrane and causes coatomer to be released (23–25).

Although studied for a number of years, many open questions exist concerning the fate and function of COPI vesicles. These include their contribution to anterograde transport of cargo molecules within the Golgi apparatus. Based on the assumption that only one type of coatomer exists, a model of cisternal progression/maturation has been developed, in which anterograde transport is mediated by an en bloc movement along the stack of newly formed cis-Golgi cisternae, which are eventually converted into trans-Golgi cisternae. COPI vesicles in this model are exclusively involved in the retrograde transport of ER- and Golgi-resident proteins, in order to maintain the identity of these compartments (26). Although this model explains transport of aggregates such as procollagen, too big for vesicular transport (27), it cannot explain the rapid secretion of several other proteins. Although some studies report an enrichment of retrograde cargo and/or a depletion of anterograde cargo (28), immunoelectron microscopy and biochemical studies revealed that anterograde cargo such as vesicular stomatitis virus G-protein and proinsulin are enriched in COPI vesicles (29,30), whereas Golgi-resident enzymes are depleted (31). In addition, the existence of different subpopulations of COPI vesicles, defined by their tethers (32), as well as a differential localization of the four more recently characterized coatomer isoforms (33), suggest different transport directions for COPI vesicles, including anterograde transport of secretory cargo and transmembrane proteins.

The nature of the so-called mitotic haze, which is observed throughout the cytoplasm during division of mammalian cells, is another example of a controversial role for COPI vesicles. Several in vivo and in vitro studies have led to the assumption that this haze represents COPI vesicles that result from an extensive vesiculation of the Golgi at the beginning of mitosis (34,35). As the cell proceeds through mitosis, these vesicles do not fuse with the ER but remain distinct throughout the mitotic cytoplasm. Together with larger mitotic Golgi clusters that were also observed, this haze was suggested to be the Golgi partitioning unit in mammalian cells. Other studies, however, interpreted this mitotic haze as Golgi proteins that have been relocated into the ER at the onset of mitosis (36). These authors suggest a complete merger of Golgi membranes with the ER, which then acts as the partitioning unit. After cell division, new Golgi membranes are built up by COPII-dependent export of Golgi components from the ER, according to this view.

More precise in vivo investigations, which would help further to enlighten these issues are often restricted by the limited resolution of 200–300 nm in conventional widefield or even confocal microscopy. COPI vesicles with a size below 100 nm are too small to be resolved as single particles against a background of labeled Golgi or ER and therefore cannot be distinguished from these membranes.

In this study, we describe a method that allows identification of the fate and function of COPI vesicles in vivo. To this end, fluorescent COPI vesicles were generated, purified and characterized in vitro from isolated rat liver Golgi membranes, labeled with the fluorescent dyes Alexa-488 or Alexa-568, respectively. These vesicles were microinjected into the periphery of African green monkey fibroblasts (BSC-1) and, owing to their fluorescent label, their fate could be unambiguously followed and investigated in vivo. The microinjected vesicles colocalized with endogenous Golgi membranes. This method will help to address the questions of Golgi inheritance in mammalian cells and also help to uncover the functions of coatomer isoforms by using homogeneously coated isoformic COPI vesicles.

Results

COPI vesicles can be generated in vitro from isolated Golgi membranes by a reconstitution assay (12,32,37). To this end, Golgi membranes are incubated either with cytosol or, in a biochemically more defined system, with recombinant Arf1 and coatomer purified from rabbit liver cytosol (38). In order to perform the desired microinjection experiments, preparative amounts of coated vesicles of high purity and concentration are needed. In previous studies, the slowly hydrolyzable GTP analog GTPγS was often used to increase the yield of coated vesicles. These ‘GTPγS-vesicles’, however, are unable to uncoat, a process required for subsequent fusion with a target membrane, and are therefore not suitable for the desired microinjection experiments.

In order to obtain sufficient amounts of fusion-competent vesicles at high concentrations, we incubated large amounts of fluorescently labeled Golgi membranes with recombinant Arf1 and coatomer in the presence of GTP. Labeling was achieved by incubating the Golgi membranes with a succinimidyl ester of either Alexa Fluor 488 or Alexa Fluor 568, leading to covalent linkage of the dye molecules to several different Golgi proteins (see below). After the vesiculation reaction, the salt concentration was increased to release the vesicles, and donor membranes and larger Golgi fragments were pelleted by a low spin centrifugation. The resulting vesicle-containing supernatant was loaded on top of a sucrose gradient and centrifuged to equilibrium. After fractionation, the sucrose density of the fractions was determined and an aliquot of each fraction was analyzed by SDS-PAGE and Coomassie staining (Figure 1). Coated vesicles are typically found at concentrations between 38 and 45% sucrose (fractions 6–8), whereas non-coated vesicles peak at the density of Golgi remnants (30–35%, fractions 11–14), which are smaller Golgi fragments generated during vesiculation (12,37). Free coatomer, according to its molecular mass of 600 kDa, sediments under these conditions to fractions 9–11 and thus is separated from the vesicles.

Figure 1.

Purification of fluorescently labeled vesicles on a sucrose density gradient. After centrifugation for 2.5 h at 400000 ×g, fractions were collected and analyzed using a 12% SDS-PAGE gel followed by Coomassie staining. Vesicle-containing fractions between 38 and 45% sucrose (boxed in red) were pooled for further concentration.

In order to inject sufficient amounts, the vesicles had to be further concentrated. To this end, fractions 6–8 were pooled, diluted with buffer to a sucrose density of 12–15% and collected by centrifugation onto a small cushion consisting of three layers with 50, 45 and 38% sucrose, respectively. After this concentration step, a small ring of fluorescent vesicles was visible on top of the 45% sucrose layer. A major problem affecting the yield of COPI vesicles prepared in the presence of hydrolyzable GTP is GTP-hydrolysis by Arf1 and concomitant uncoating. Although partial uncoating cannot be excluded during vesicle preparation as described here, the procedure is rapid enough to generate a workable yield of COPI-coated vesicles. In order to avoid aggregation of uncoated and coated vesicles, the concentration step was performed in the presence of 1 mg BSA/mL and 150 mm KCl. The purity and concentration of the vesicles generated was analyzed by SDS-PAGE and Coomassie staining (Figure 2).

Figure 2.

Analysis of the isolated vesicles. Samples from each step of the vesicle preparation were separated on a 12% SDS-PAGE gel and stained with Coomassie. A) 1% of the labeled Golgi membranes. B) 0.75% of the low spin supernatant after vesicle formation. C) 1.5% of the vesicle fractions pooled between 38 and 45% after isopycnic sucrose density gradient centrifugation (from Figure 1). D) 10% of the concentrated vesicle sample. The bands indicated were excised and analyzed by MALDI mass spectrometry (see table 1). Note that 1 mg BSA/mL was added to the solutions used for vesicle concentration (band 6).

Characterization of labeled vesicles

The various steps of the purification procedure were analyzed by SDS-PAGE and Coomassie staining (Figure 2). In order to characterize more precisely the purified vesicles, the proteins indicated with arrows were excised and analyzed by matrix-assisted laser desorption/ionization (MALDI) mass spectrometry (Table 1). Importantly, actin and myosin that were present in the vesicle-containing supernatant (lane B in Figure 2), loaded on top of the gradient, were clearly separated from the vesicles and stayed in remnant fractions (fractions 12–15 in Figure 1). The seven coatomer subunits and Arf1 constitute the vast majority of vesicle proteins (bands 3, 5, 7, 10 and 14). ζ-COP and Arf1 colocalize in this gel (band 14). Because predominantly Arf1 peptides were found in this band (not shown), it is highly likely that the purified vesicles contain multiple Arf1 molecules per coatomer. In addition to the coat components, the membrane proteins p23, p24 and p25 (19) were detected (band 11, 12 and 13). α-Mannosidase II (Mann II), a representative of a medial Golgi-resident enzyme (band 4) was also identified. The lectin ERGIC-53 (band 8) that predominantly localizes in the intermediate compartment (IC) (39) and cycles between the ER and IC were identified as another membrane protein. Beside these transmembrane proteins, the soluble luminal protein p54/NEFA was found. This potential calcium-binding protein that in vivo colocalizes with Mannosidase II was found before in COPI vesicles by electron microscopic studies (40). However, its function in the Golgi apparatus or in the lifecycle of a COPI vesicle is unknown.

Table 1.

Proteins found in a crude (lane B, Figure 2) and a purified (lane D, Figure 2) vesicle fraction.

	Crude fraction	Purified fraction
1	Myosin	—
2	Actin	—
3	ND	α-COP
4	ND	α-Mannosidase II
5	ND	β, β×, γ-COP
6	ND	BSA
7	ND	δ-COP
8	ND	ERGIC-53
9	ND	p54/NEFA
10	ND	ε-COP
11	ND	p25
12	ND	p24
13	ND	p23
14	ND	Arf1/ζ-COP

Fluorescent labeling of the vesicles was analyzed by SDS-PAGE and western blot using a specific antibody directed against Alexa-488. As depicted in Figure 3, the Alexa-488-NHS-ester reacts with a variety of Golgi proteins in contrast to the Biotin-NHS-ester that was described earlier to label predominantly golgin-84 (37). The protein pattern and the efficiency of the labeling reaction are highly reproducible as shown by comparing different vesicle samples.

The observed differences in staining pattern are partly because of the higher sensitivity of the western blot-based detection using an anti-Alexa antibody of high affinity when compared with Coomassie staining. In addition, Arf1 as well as the coatomer subunits, which comprise a large percentage of the proteins revealed by Coomassie staining, was added after the labeling reaction. Together, these explain the higher staining complexity observed by western blotting. Phosphatidylethanolamine lipids, which also contain a primary amino group, were not labeled, as determined by quantitative electrospray ionization tandem mass spectrometry (not shown).

Finally, the isolated COPI vesicles were investigated by electron microscopy using negative stain. A representative image shown in Figure 4 reveals a homogeneous population of coated vesicles with a size of ~90 nm.

Figure 4.

Electron microscopic analysis of the purified fluorescently labeled vesicles. An aliquot of the concentrated vesicle sample was diluted 1:10, directly applied on a grid and subsequently processed for negative staining with uranyl-acetate. Bar represents 1 μm.

Microinjection of fluorescently labeled vesicles

Owing to partial uncoating, the purified vesicles tended to form larger aggregates, which often led to clogging of the injection needle. In order to minimize aggregation, the vesicles were purified in the presence of 1 mg BSA/mL and 150 mm KCl, and the final vesicle sample was resuspended and diluted twofold with buffer containing 1 mg BSA/mL and 150 mm KCl, lowering the sucrose concentration of the injected solution to about 20%. This dilution reduced osmotic changes in the cytoplasm following injection. In order to remove larger aggregates, the solution was spun for 2 min at 12 000 ×g prior to loading the injection needle. The inner surface of the injection needle was coated with BSA in order to minimize absorption of vesicles to the surface. Care was taken to use needles with a wide diameter, because a too small aperture resulted in high shearing forces and thereby compromised the injected vesicles. The pretreated needles were tested for their suitability for injection prior to injection of vesicles.

After microinjection of fluorescently labeled vesicles the cells were incubated for 30 min at 37°C, in order to allow diffusion to the target membrane, and subsequently labeled for the endogenous Golgi marker Giantin using immunofluorescence microscopy. In Figure 5, representative images of microinjected African green monkey kidney fibroblasts (BSC-1) are depicted. For colocalization studies with beta-1,4-N-acetylgalactosaminyltransferase-2 (GalNAcT2-YFP), Alexa-568-labeled vesicles were injected into BSC-1 cells expressing the yellow fluorescent protein (YFP)-tagged transferase (YT2 cells) (34). Both Alexa-488 and Alexa-568 signals colocalize, 30 min after injection, with endogenous Golgi markers such as the tether protein Giantin, or the Golgi-resident enzyme GalNAcT2-YFP. The Alexa signals represent exogenous COPI vesicle proteins (see Figure 3). In addition, it was possible to label these injected vesicles using an antibody that specifically recognizes rat Mann II (53FC3), but not the monkey homolog. The observed colocalization with endogenous Golgi markers of exogenous Alexa-labeled proteins shows that intact and active vesicles were microinjected vesicles and were able to diffuse freely in the cytosol and eventually localize to endogenous Golgi membranes within 30 min. Punctate structures outside and inside the cells (the latter not colocalizing with Giantin) were most probably inactive vesicle clusters.

Figure 5.

Microinjection of fluorescently labeled vesicles into mammalian cells. A) Epifluorescence image of BSC-1 cells stained for endogenous Giantin (in red) and vesicle-derived rat Mann II (in blue), fixed and labeled 30 min after microinjection of green vesicles labeled with Alexa-488 (injected cell is marked with an asterisk). Bar, 20 μm. B) Confocal image of YT-2 cells, stably expressing GalNAcT2-YFP, fixed and labeled for Giantin (in blue) 30 min after microinjection of red vesicles labeled with Alexa-568. Bars, 20 μm.

Figure 3.

Analysis of the labeled proteins. Aliquots of four different vesicle samples were analyzed by SDS-PAGE and western blot. The labeling intensity was compared by using an anti-Alexa-488 and an anti- α-Mann-II antibody. Vesicles in lane 2 were labeled with Alexa-568 instead of Alexa-488, showing the specificity of the anti-Alexa-488 antibody (control).

In vivo dynamics of the injected vesicles

The observed Alexa signals colocalizing with endogenous Golgi markers are because of labeled exogenous proteins derived from the microinjected vesicles. This Golgi localization indicates that active vesicles were injected, which are able to diffuse freely in the cytosol, dock to endogenous Golgi membranes, and potentially fuse with it. However, the vesicles might have also fused with ER membranes followed by a COP II-dependent transport of the labeled proteins from the ER to the Golgi. In order to elucidate whether the injected vesicles directly move and dock with Golgi membranes, we microinjected the dominant negative mutant Sar1dn (Sar1H79G) prior to vesicle injection (41). This results in a complete block of anterograde COPII-dependent transport from the ER to the Golgi. As shown in Figure 6B, under conditions where ER to Golgi transport is blocked, fluorescently labeled vesicle proteins colocalized with the endogenous Golgi markers Giantin and GalNAcT-2. This further suggests that the injected vesicles move directly to Golgi membranes, independently of the ER. Recently different subpopulations of COPI vesicles, defined by their tethers, were described, of which only some contain α-mannosidase II (32). In Figure 6C, colocalization of exogenous α-mannosidase II is shown with endogenous Golgi membranes in the presence of Sar1dn (Figure 6C), emphasizing that α-mannosidase II containing vesicles move, dock and probably fuse directly with the Golgi apparatus.

Figure 6.

Analysis of the target compartment. A) Microinjection of red vesicles in untreated YT2-cells (control). B) Injection into YT2-cells of Sar1dn, with subsequent injection of Alexa 568-labeled COPI vesicles after 60 min. C) Injection into BSC-1-cells of Sar1dn with subsequent injection of Alexa-488-labeled COPI vesicles. The proteins visualized in fixed and labeled cells are indicated below the individual images. All images were taken with a confocal microscope. Bars, 10 μm.

A second independent approach was also used to address the question of vesicle docking versus fusion, fluorescence recovery after photobleaching (FRAP). If the vesicles fused with the Golgi, FRAP of fluorescently labeled vesicles should follow kinetics similar to green fluorescent protein (GFP2)-tagged Golgi-resident proteins. GalT-GFP2 expressing BSC1 cells (42) were first microinjected with Alexa-568-labeled COPI vesicles. After incubation for 30 min at 37°C to allow fusion of the vesicles with the Golgi, half of the Alexa-568 and GalT-GFP2-positive Golgi structures were photobleached and the recovery of fluorescence in both channels was followed. Analysis of the mobile fractions and of the half time of maximum recovery showed that recoveries of vesicle fluorescence and of the Golgi localized enzyme GalT-GFP2 were comparable, suggesting that the vesicles had fused with the Golgi (Figure 7).

Figure 7.

FRAP of GalT-GFP2-expressing cells microinjected with fluorescently labeled vesicle. A) Alexa-568 labeled vesicles and GalT-GFP2 were subjected to FRAP and monitored over time. B) Golgi region fluorescence recovery in both channels plotted over time. Error bars represent SEM. C) Analysis of the mobile fraction and of the half time of maximum recovery after fitting of the experimental data to a single exponential curve. Data are expressed as mean values ± SEM. Bar, 10 μm.

The behavior and fate of the Alexa-labeled vesicle proteins were further investigated and compared to those of endogenous proteins. To this end, cells were treated with Brefeldin A (BFA) for an additional 30 min after injection of the fluorescently labeled vesicles. Localization of fluorescently labeled vesicular proteins was subsequently compared by immunofluorescence staining for the early Golgi marker GM130 and late Golgi marker golgin-97 (Figure 8A). Golgi proteins can be classified into two types with respect to their fate after BFA treatment (Figure 8B). One type, such as the Golgi matrix protein GM130, stays in smaller Golgi matrix structures (43), whereas the other type does not. An example of the second type is GalNAc-T2, which is redistributed to the ER (34). Some of the Alexa-labeled proteins still colocalized with GM130 in Golgi remnant structures, whereas others appeared more diffuse. It was, however, not possible to determine whether these had distributed to the ER because the dilution of the signal was too great.

Figure 8.

Dynamics of endogenous and vesicle-derived labeled proteins. Epifluorescence images of BSC-1 cells before A) and after B) BFA treatment. Cells were microinjected with green vesicles, incubated for 30 min and subsequently treated for 30 min with 5 μg/mL BFA B). After fixation and permeabilization, the early Golgi marker GM130 (in red) and the late Golgi marker Golgin 97 (in blue) were stained using specific antibodies. Fluorescence not colocalizing with endogenous Golgi markers is indicated by arrowheads and probably represents inactive vesicle clusters. Alexa-labeled proteins originating from the injected vesicles that colocalize with GM130 after BFA treatment are indicated with arrows. Injected cells are marked with an asterisk. Bar, 10 μm.

Discussion

Here we have developed a method to follow the fate and behavior of COPI vesicles in vivo to help overcome the low resolution of conventional fluorescence microscopy. To this end, we made use of techniques to generate and purify COPI vesicles using an in vitro system, so that fluorescently labeled COPI vesicles could be injected into the cytoplasm of mammalian cells.

Generation of fluorescently labeled vesicles

One of the biggest challenges in developing a method to microinject vesicles into mammalian cells was the generation of fluorescently labeled active vesicles in preparative amounts. GTPγS, which was used in the past to obtain high amounts of COPI vesicles, could not be used, because this would have led to inactive vesicles unable to lose their coat and therefore incompetent to fuse with the target membrane. Furthermore, vesicles generated in the presence of non-hydrolyzable GTP are known to be compromised in the uptake of cargo (44–46). In order to generate the required amounts of vesicles, we used Golgi, together with purified coatomer and Arf1, and modified the purification of COPI vesicles in order to minimize uncoating. The desired fluorescent labeling was achieved by incubation of the Golgi with succinimidyl-esters of the Alexa-488/568 dyes prior to the vesicle reaction. In a similar study, biotinylated vesicles were generated which were, however, not suitable for experiments in living cells (37). In contrast to the biotinylation reaction, which was reported to mark predominantly Golgin-84, the incubation with NHS-esters of Alexa-488/568 led to labeling of several different Golgi proteins. No co-labeling of phosphoglycerolipids containing a primary amino group was observed, excluding a fast redistribution of the dye because of lipid movement and/or exchange. Therefore, the Alexa signals observed in vivo resulted from labeled Golgi proteins that were incorporated into the vesicles. The labeling reaction did not inhibit the functioning of COPI vesicle tethering and perhaps fusion machineries because the injected vesicles were found in the Golgi area.

The purity and composition of the isolated vesicles was controlled biochemically and by electron microscopy. SDS-PAGE and Coomassie staining revealed 14 prominent proteins that comprised the majority of the vesicular protein content (Table 1). Among these, we identified by MALDI mass spectrometry the coat components Arf1 and coatomer, with all of its seven subunits. Furthermore, the membrane proteins p23, p24, p25, ERGIC 53 and α-mannosidase II were detected. p54/NEFA was found as a major soluble protein, described as a potential calcium-binding, Golgi-resident protein that colocalizes with α-mannosidase II in early and medial cisternae (40). During mitosis or after BFA treatment p54/NEFA localizes in Golgi remnant structures, similar to matrix proteins such as GM130 or the members of the p24-protein family. Its relative abundance within the purified vesicles and the described membrane association are interesting features of this soluble Golgi-resident protein of yet unknown function. As shown in Figure 2D, a limited number of Coomassie bands of similar intensity were visualized after separation of highly purified vesicles. This is surprising given the enormous number of different proteins attributed to COPI vesicles in a recent proteomic study (47). In this study, for example, actin and myosin would be described as COPI vesicle proteins. This is in contrast to our results depicted in Table 1, clearly demonstrating separation of these proteins from COPI vesicles in the last purification step by isopycnic gradient centrifugation. These discrepancies observed in protein composition might be because of technical and methodological differences in vesicle generation and isolation.

The fate in vivo of microinjected vesicles

Because of their size, microinjection of fluorescently labeled vesicles is technically more challenging than microinjection of macromolecules. Although Arf1 has a relatively low intrinsic activity, partial uncoating during preparation and handling of the vesicles cannot be completely prevented. The resulting mixture of coated and uncoated vesicles showed a tendency to form bigger aggregates that resulted in clogging of the injection needle. A more homogeneous suspension of vesicles was obtained in the presence of 150 mm salt and 1 mg BSA/mL. High pressures, such as those needed to flush clogged needles, compromised the vesicle suspension, probably because of high shearing or expansion forces. Therefore, we used needles with an aperture wide enough to allow low injection pressures. Furthermore, coating of the inner surface of the needle with a BSA solution turned out to enhance the yield of active vesicles within the injected cells.

These measures allowed us to microinject active, fluorescently labeled COPI vesicles into the cytoplasm of African green monkey kidney fibroblasts, which appeared especially suitable because of their large size. Although injected into the periphery of these cells, fluorescence was observed in perinuclear Golgi regions within 30 min after microinjection, showing that these vesicles are able to travel over long distances. Based on the western blot and mass spectrometric analysis, the fluorescence represents proteins of ectopic rat liver Golgi and IC membranes that were imported by the injected COPI vesicles. The specific identification of ectopic rat α-mannosidase II within the Golgi region further corroborates that the observed fluorescence represents rat liver Golgi proteins imported by the injected COPI vesicles. Because α-mannosidase II is a Golgi-resident protein, it is unlikely that COPI vesicles containing α-mannosidase II fuse with the ER and, therefore, points towards an intra-Golgi transport function of these vesicles. This assumption was further underlined by microinjection into Sar1dn-treated cells. COPII-dependent ER to Golgi transport is inhibited in these cells, indicating that the observed Golgi localization of rat α-mannosidase II is likely to result from a direct movement and docking of the injected vesicles to endogenous Golgi membranes. Because α-mannosidase II localizes mainly to the medial Golgi (48), one might assume that it gets incorporated into vesicles that mediate transport from the cis face in anterograde direction. Recent studies, however, report that α-mannosidase II, in contrast to members of the p24 protein family is taken up into vesicles that are defined by the tether pair golgin-84 and CASP, which are involved in retrograde transport (32).

Because of their hydrophilic character, the Alexa dyes are presumably not able to penetrate the Golgi membrane, and therefore their contact is restricted to proteins on the cytosolic face of the Golgi. These labeled proteins can be classified into two types with respect to their fate after BFA treatment. One set relocates to a diffuse structure, perhaps ER membranes, analogous to Golgi-resident enzymes such as α-mannosidase II or Galactosyltransferase. The second set, however, remains in Golgi remnant structures positive for Golgi matrix proteins such as GM130, similar, for example, to the members of the p24 protein family (49,50). The finding that the ectopic vesicle-derived proteins behave similarly to endogenous proteins after BFA treatment suggests that the vesicles dock and probably fuse with endogenous Golgi membranes. This was corroborated by FRAP of Alexa-labeled vesicles and GalT-GFP expressed in the microinjected cells, both of which followed similar kinetics. This strongly suggests that fluorescently labeled proteins of the vesicles freely diffuse in the Golgi, rather than simply being docked to the target membrane.

Perspective

A method was established that allows more precise investigations of the function and behavior of COPI vesicles in vivo. By fluorescent labeling their fate can be unambiguously visualized in living cells. Once the four isoforms of coatomer are available as recombinant proteins, fluorescently labeled COPI vesicles uniformly coated with an individual isoform of coatomer could be used in microinjection experiments to study the function of isoformic coats. Moreover, Golgi membranes isolated from cell lines expressing fluorescently tagged proteins could be used to follow the direction taken by COPI vesicles containing them. Thus, tracking of individual COPI subpopulations as well as measurement of diffusion coefficients should provide further insights and help to resolve at least some of the controversies concerning the role of COPI vesicles in various transport routes. In addition, this system will allow a closer look at the nature and function of the ‘mitotic haze’, that has been observed during cell division.

Materials and Methods

Reagents were used in the following working concentrations (stock solutions): Alexa Fluor 488 carboxylic acid, Alexa Fluor 568 carboxylic acid, succinimidyl ester (A-20000, Invitrogen), 0.4 μg/μL (10 mg/mL in dimethyl sulphoxide), and BFA (Sigma-Aldrich), 5 μg/mL (1 mg/mL in ethanol).

Antibodies and cell lines

The following antibodies were used in this study: anti-Alexa-Fluor-488 (A-11094, Invitrogen), anti-GM130 (BD-Biosciences), anti-Golgin 97 (Clontech), anti-Giantin (51,52) and anti-rat Man II (53). BSC-1 cells (ATCC, CCL-26) were routinely cultured in DMEM (Sigma-Aldrich) supplemented with 10% fetal calf serum (PAA Laboratories), 100 units penicillin/mL, 100 μg streptomycin/mL, 2 mml-glutamine (Invitrogen), at 5% CO2 and 37°C. YT2 cells, stably expressing GalNAc-T2-YFP (34), were routinely cultured as BSC-1 cells in the presence of G418-sulfate (800 μg/mL; Invitrogen).

Purification of rat liver Golgi membranes, rabbit liver coatomer, recombinant myristoylated Arf1 and Sar1dn

Rat liver Golgi membranes were isolated as described in (54). Coatomer was purified from rabbit liver cytosol (38), N-myristoylated human Arf1 and Sar1dn (Sar1H79G) were expressed and purified form E. coli as described in (55,56).

Labeling of rat liver Golgi membranes

Rat liver Golgi membranes were labeled with Alexa dye in analogy to the biotinylation reaction described in (37). Briefly: 1.5 mg of Golgi membranes (300 μL with 5 mg/mL) were mixed with 60 μL HEPES pH 7.4 and 15 μL of the dye stock solution. This mixture was incubated for 1 h on ice. Residual-free dye was quenched with 25 μL of 1 M NH4Cl for 10 min on ice. In order to recover the membranes, the salt concentration was raised to 500 mm by adding 75 μL of 3 M KCl and the suspension diluted to 9% sucrose with 750 μL dilution buffer (25 mm HEPES pH 7.0, 2.5 mm MgAc, 500 mm KCl). Twice 600 μL of this suspension was layered on top of 150 μL 12% sucrose, 20 μl 17%, 20 μl 20%, 10 μl 23%, 5 μl 25% and 5 μL 30% sucrose (each in 150 mm KCl, 25 mm HEPES, pH 7.0, 2.5 mm MgAc) and spun in a TLS 55 rotor at 12 000 ×g for 10 min using TLA 45 tube. In order to remove any free dye, the supernatant was carefully sucked off, replaced by 12% sucrose and removed again. The remaining labeled membranes were resuspended in 12% sucrose to a total volume of 200 μL.

Generation and purification of labeled COPI vesicles

Fluorescently labeled vesicles were generated and purified in analogy to (37). The labeled Golgi membranes were incubated with 600 μg coatomer and 60 μg Arf1 in the presence of 1.2 mm GTP (from a 40 mm stock solution, Roche Applied Science) in a total volume of 500 μL in 25 mm HEPES, pH 7.0, 2.5 mm MgAc, 100 mm KCl (AB) supplemented with 9% sucrose and an ATP regenerating system (57) for 10 min at 37°C. The reaction was stopped by adding 500 μL ice cold AB that was supplemented with 9% sucrose and 400 mm KCl (final concentration of 250 mm KCl), in order to release the vesicles of the membranes. The incubation mixture was spun at 14 000 g for 10 min at 4°C and the vesicle-containing supernatant layered on top of a step gradient composed of 1.5 mL 30%, 1.5 mL 35%, 0.25 mL 37.5%, 0.25 mL 50% and 2 mL 55% (wt/wt) sucrose in gradient buffer (25 mm HEPES, pH 7.0, 2.5 mm MgAc, 150 mm KCl). The tube was filled to capacity with gradient buffer containing 9% sucrose and subsequently sealed. Membranes were centrifuged to equilibrium at 4°C for 2.5 h at 400 000 ×g in a vertical rotor (VTi 65.1 Beckman instruments, 65 000 ×g, accel/deccel set to slow). The bottom of the tube was punctured and the first 1.5 mL collected in 0.5 mL fractions, followed by fractions of 120–150 μL. Fractions corresponding to 38–44% were pooled and diluted to 12% sucrose using assay buffer containing 1.5 mg/mL BSA. This suspension was spun in a SW55 rotor (Beckman instruments) at 35 000 ×g for 1 h (accel/deccel set to slow) on top of a cushion of 5 μL 50%, 5 μL 45% and 5 μL 38% sucrose (each in AB containing 150 mm KCl and 1 mg BSA /mL) using a TLA 45 tube. The visible band of fluorescently labeled vesicles was carefully taken up in a volume of maximum 10 μL using a long loading tip, and subsequently diluted twofold using assay buffer containing 1 mg BSA/mL and 150 mm KCl. Aliquots of 5 μL were snap frozen in liquid nitrogen and stored at −80°C.

Microinjection of labeled COPI vesicles and Sar1dn

For microinjection, cells were cultured on 12 mm glass coverslips over night. Preparation of the injection needle: 5 μL of a 25% (wt/wt) sucrose solution in assay buffer containing 1.5 mg BSA/mL was loaded in a femtotip II microinjection needle (Eppendorf). After leaving this needle for 5 min at room temperature, the sucrose solution was removed again and the needle screwed to the Microinjector 5242 equipped with the micromanipulator 5171 (Eppendorf). BSC-1 cells were injected in order to determine the injection pressure and subsequently any remaining sucrose solution was blown out. In order to inject vesicles without harming them, pressures below 50 mbar were applied.

Frozen vesicle aliquots were thawed rapidly, spun for 2 min at 12 000 ×g in a TLS55-Rotor (Beckman instruments) and subsequently loaded into the prepared needle. About 100 cells were injected within 15 min and subsequently incubated for 30 min at 37°C. Injection time was up to 1 second and needles with a wide aperture used, in order to inject sufficient amounts of vesicles.

Microinjection of Sar1dn was performed as described in (56), using dextran cascade blue (Invitrogen, Karlsruhe) as injection marker. Sar1dn microinjected cells were incubated for 1 h at 37°C prior to injection of vesicles. The effect of Sar1dn injection was monitored by immunofluorescence microscopy of the cycling protein p23, which instead of the Golgi matrix protein GM130 relocates to the ER upon injection of Sar1dn.

BFA treatment

After microinjection of the vesicles cells were incubated for 30 min at 37°C in normal DMEM and subsequently BFA in a final concentration of 5 μg/mL added. After incubation for additional 30 min, the cells were processed for immunofluorescence and microscopy.

FRAP experiments

For FRAP experiments, GalT-GFP2 expressing BSC1 cells (42) were microinjected as described with Alexa-568-labeled COPI vesicles. After incubation for 30 min at 37°C to allow fusion of the vesicles to the Golgi apparatus, FRAP experiments were performed with a Perkin Elmer Improvision Ultraview Vox spinning disk confocal microscope. After 10 pre-bleach acquisitions, half of the Alexa-568 and GalT-GFP2-positive Golgi structures were photobleached using both 488 and 561 laser lines at maximum power, and then the recovery of fluorescence in both channels was followed at 37°C for 320 seconds with 10-second intervals. After normalizing the fluorescence for the total intensity of the Golgi in each cell, the data from 7 to 10 experiments were averaged. The analysis of the mobile fraction and of the half time of maximum recovery of each experiment was carried out after best fitting of the experimental data obtained to a single exponential curve using IgorPro software (WaveMetrics Inc.).

Immunofluorescence and light microscopy

After treatment, cells on coverslips were washed with PBS, fixed in 3% paraformaldehyde for 20 min and subsequently permeabilized using 0.5% Triton X-100 for 5 min. After blocking with 5 % BSA in PBS for 15 min, the samples were incubated with the respective primary antibodies diluted in PBS containing 5 % BSA for 30 min. After washing twice with PBS for 10 min, the samples were incubated with secondary antibodies conjugated to Alexa fluorophores (Molecular Probes) for another 30 min, washed twice with PBS for 10 min, and subsequently mounted in Mowiol 488 (Calbiochem, San Diego, Calif.). All steps were carried out at room temperature.

Confocal images were taken with an inverted laser scanning confocal microscope (LSM 510, Carl Zeiss), using a 63 × 1.4 NA objective lens, 4× zoom and a pinhole size equivalent to one Airy disk diameter. Eppifluorescence images were taken with a Zeiss Axioplan 2 upright microscope equipped with an Orca II cooled charged-coupled device camera (1280 × 1024 pixels, Hamamatsu) controlled with the Openlab 3.0.8 software package (Improvision) and a Plan-Apochromat 100 × 1.4 NA objective.