Cholesterol Loading Induces a Block in the Exit of VSVG from the TGN


Kirsten Sandvig,


Recent work from our laboratory demonstrated that increased cellular cholesterol content affects the structure of the Golgi apparatus. We have now investigated the functional consequences of the cholesterol-induced vesiculation of the Golgi apparatus and the role of actin for these changes. The results showed that cholesterol-induced vesiculation and dispersion of the Golgi apparatus is a reversible process and that reversal can be inhibited by cytochalasin D, an actin-disrupting reagent. Furthermore, electron microscopy revealed that jasplakinolide, which stabilizes actin filaments, prevented the dispersion, but not the vesiculation of the Golgi cisternae. Importantly, the different Golgi markers seemed to be separated even after vesiculation. To investigate whether transport through the different steps of the exocytic pathway was affected in cholesterol-treated cells, we visualized ER to plasma membrane transport by using ts045-VSVG-GFP. In COS-1 cells expressing ts045-VSVG-GFP increased cholesterol levels did not affect transport of VSVG into the vesiculated Golgi apparatus. However, increased levels of cholesterol resulted in retention of the nascent G protein in vesicles with the TGN-marker TGN46. Biotinylation of cell surface molecules to quantify arrival of VSVG at the plasma membrane confirmed that cholesterol treatment inhibited export of the VSVG protein. In conclusion, the data show that transport of VSVG into/through a vesiculated Golgi is feasible, but that cholesterol loading inhibits exit of VSVG from the vesicles containing TGN markers. Furthermore, the data illustrate the importance of actin filaments for Golgi structure.

Cellular cholesterol levels are precisely controlled by regulation of cholesterol biosynthesis, uptake and efflux (1,2). The Golgi apparatus maintains a concentration gradient of cholesterol, and cholesterol is at least partly transported through the Golgi and to the plasma membrane, where it serves a number of functions (3,4). Recent studies have revealed that cholesterol and lipid rafts play a fundamental role in formation of constitutive and regulated secretory vesicles from TGN, as well as for polarized secretion (5,6). Both too little and too much cholesterol can cause Golgi vesiculation (7,8), and in adrenocortical cells lacking the minus end directed motor protein KIFC3, cholesterol depletion leads to dispersal of Golgi cisternae, suggesting that cholesterol is important for regulation of microtubule motor proteins associated with the Golgi apparatus (9). Moreover, actin filaments can become associated with rafts (10), and actin filaments are essential for the morphology and positioning of the Golgi apparatus (11,12), as well as for protein transport from the Golgi to the plasma membrane and to the ER (13–16). A large number of studies have shown that actin and actin-associated proteins are associated with the Golgi apparatus. For instance, actin-based myosin motors, the actin-regulatory protein Cdc42, and different actin-binding proteins have been located in the Golgi membranes and/or Golgi-derived vesicles (17–22). However, in contrast to the established roles of microtubules and microtubule-based motors in Golgi function and location (9,23), the involvement of actin microfilaments in the regulation of the membrane dynamics and morphology of the Golgi apparatus is less well understood.

The extent to which increased levels of cholesterol affect anterograde transport through the Golgi apparatus and whether actin is involved in the cholesterol-induced changes of the Golgi were not known. To address these questions, we have in the present study used a complex of methyl-β-cyclodextrin and cholesterol (mβCD/chol) to insert cholesterol into the plasma membrane (8). Cytochalasin D (Cyt D) and jasplakinolide (Jas) were added to disrupt and stabilize the actin filaments, respectively. Jas was found to inhibit dispersal of the cholesterol-induced fragmented Golgi apparatus, and Cyt D but not Jas-inhibited reformation of the Golgi apparatus after removal of excess cholesterol. In order to study Golgi function after cholesterol treatment, we utilized a temperature-sensitive vesicular stomatitis virus glycoprotein tagged with green fluorescent protein (ts045-VSVG-GFP) as a marker protein for the exocytic pathway. This glycoprotein is retained in the ER at 39.5 °C, but upon temperature reduction to 31.5 °C, it moves out of the ER and into the Golgi complex before being transported to the plasma membrane (24). When the GFP is attached to the cytoplasmic tail of VSVG, its normal folding and transport properties are preserved (25–27). In the present article we demonstrate that VSVG can be transported through the cholesterol-induced Golgi vesicles until it arrives in a compartment colocalizing with TGN46. However, the increased level of cholesterol blocks transport to the cell surface.


Cholesterol-induced dispersion of the Golgi apparatus is reversible and dependent on actin filaments

We have previously shown by immunofluorescence and electron microscopy that increased cellular cholesterol content causes Golgi fragmentation in HeLa cells (8). Such an effect of cholesterol seems to be general, since a dispersion of Golgi markers was observed also in A431, Hep-2, MDCK-1 (data not shown) and COS-1 cells that were used in this study, since they are easily transfected (see below). However, the time-course for the observed changes of the Golgi apparatus varied somewhat between different cell lines. Our previous results have shown that treatment of HeLa cells with 5 mm mβCD/chol for 30 min was sufficient to induce fragmentation and dispersion of the Golgi apparatus (8). In COS-1 cells, a similar change in Golgi morphology was obtained after 2 h incubation with the same concentration of mβCD/chol. Treatment of COS-1 cells with 5 mm mβCD/chol increased the cellular cholesterol content by 83%, 97% and 121% after 30 min, 60 min and 120 min, respectively. In HeLa cells we have earlier found that there is a similar (80%) increase in the cholesterol level after 30 min (8). However, since these numbers represent the total cellular content of cholesterol and not what is found in the Golgi apparatus, the possibility exists that there are cell-specific differences when it comes to the efficiency of cholesterol transport from the plasma membrane to the Golgi apparatus.

To investigate whether the cholesterol-induced dispersion of the Golgi apparatus is a reversible process, HeLa cells were incubated with 5 mm mβCD/chol to increase the cellular cholesterol content. After 60 min, the cells were washed, and then incubated for 12 h with normal growth medium. The Golgi apparatus was identified by immunostaining for GM130, a cis-Golgi marker (28). As shown in Figure 1, the dispersed Golgi apparatus seen after cholesterol treatment (Figure 1B) is reorganized to form a perinuclear organelle after 12 h incubation of the cells with normal growth medium (Figure 1E). The Golgi apparatus then looks similar to that of control cells (Figure 1A). Also, measurements of the cellular cholesterol levels revealed that recovery of the normal Golgi morphology after removal of extra cholesterol correlates with return of normal cholesterol contents (data not shown). Importantly, when the Golgi morphology had recovered, transport of VSVG to the cell surface occurred normally, indicating that not only the appearance of the Golgi apparatus, but also its function recovers upon removal of cholesterol (data not shown). To investigate whether this reassembly of the Golgi apparatus requires actin filaments, Cyt D, which disrupts the actin cytoskeleton and thereby blocks membrane transport along actin filaments (29), was added to the cells after removal of mβCD/chol. As shown in Figure 1(F), the reformation of a perinuclear Golgi apparatus was partially blocked by Cyt D treatment. When a similar experiment was performed with Jas, which stabilizes actin filaments (30), the reassembly of the Golgi apparatus occurred normally (data not shown). Similar data were obtained with COS-1 cells (data not shown). The results indicate that actin filaments are required for recovery of the Golgi structure.

Figure 1.

Figure 1.

    Cholesterol-induced dispersion of the Golgi apparatus is reversible and dependent on actin filaments. HeLa cells were incubated for 1 h without (A) or with (B) 5 mm mβCD/chol, or with 0.1 μm Cyt D for 1 h (C) or 12 h (D). The mβCD/chol complex was removed after 1 h and the incubation was con-tinued for 12 h in the presence of normal growth medium without (E) or with (F) 0.1 μm Cyt D. The cells were then fixed, per-meabilized and visualized with an antibody against the Golgi marker GM130, and rhodamine-conjugated goat anti-mouse IgG. Note that cholesterol induces vesiculation and dispersion of the Golgi marker GM130 (B) and that Cyt D in itself does not give a similar effect (C, D). However, Cyt D inhibits the reversal of the dis-persion induced by cholesterol (F). Bars, 20 μm.

    Stabilization of actin filaments with jasplakinolide prevents formation of a dispersed Golgi apparatus

    To further investigate the role of actin filaments in cholesterol-induced dispersal of the Golgi apparatus, actin filaments were stabilized by adding Jas during the incubation with cholesterol. In COS-1 cells (Figure 2D) and HeLa cells (data not shown), the Golgi apparatus, localized by an antibody to GM130, then seemed to be preserved as a perinuclear complex. However, when Jas-treatment was performed after dispersion of the Golgi with cholesterol, Jas was unable to restore a perinuclear localization of GM130 (data not shown). Since the resolution of confocal microscopy did not allow us to conclude whether the perinuclear Golgi labeling seen in the presence of Jas and cholesterol actually represented intact Golgi cisternae or just a perinuclear localization of Golgi vesicles, we used EM to investigate this question. As shown in Figure 3, EM of COS-1 cells revealed that Jas did not inhibit the cholesterol-induced vesiculation of the Golgi cisternae, but it kept the vesicles formed after addition of cholesterol in a perinuclear location (Figure 3E). The size distribution of the vesicles formed after cholesterol treatment is shown in Figure 3(D), and the diagram illustrates that most of the vesicles are about 50–200 nm in diameter, probably reflecting a combination of small transport vesicles and larger vesicles deriving from vesiculation of the Golgi apparatus. These vesicles are larger than the typical Golgi transport vesicles (Figure 3B) that are reported to be about 70–75 nm (31). The size distribution of vesicles obtained after treatment with Jas and mβCD/chol (Figure 3F) was similar to that seen after mβCD/chol alone; the only difference was that the vesicle clusters concentrate in the perinuclear cytoplasm. Similar observations were made in cholesterol-loaded HeLa cells, i.e. addition of cholesterol induced formation of a combination of small transport vesicles and larger vesicles deriving from vesiculation of the Golgi stacks (data not shown). Thus, cholesterol treatment may act by a two-step process: it first leads to vesiculation and then to dispersal of the Golgi apparatus by a process inhibited by stabilization of actin filaments.

    Figure 2.

    Figure 2.

      Jasplakinolide (Jas) treatment blocks Golgi dispersion induced by mβCD/chol. COS-1 cells were incubated without (A) or with (B) 5 mm mβCD/chol, with 3 μm Jas (C), or 3 μm Jas together with 5 mm mβCD/chol (D) for 2 h. Sub-sequently, the cells were fixed, permeabilized and stained with anti-GM130 antibody and rhodamine-conjugated goat anti-mouse IgG. As shown, the cholesterol-induced dispersion seen in B is strongly inhibited in D where also Jas was added. Bars, 10 μm.

      Figure 3.

      Figure 3.

        Jasplakinolide(Jas) pre-vents dispersion but not cholesterol-induced vesiculation of the Golgi apparatus. COS-1 cells were either untreated (A and B), treated with 5 mm mβCD/chol (C and D), or 3 μm Jas together with 5 mm mβCD/chol for 2 h (E and F). The cells were then fixed and prepared for electron microscopy. Distributions of Golgi vesicle size are shown in B (control) D (treated with mβCD/chol) and F (treated with Jas and mβCD/chol) (n = 200 in each experiment). Note that the size of most of the Golgi vesicles formed after addition of cholesterol (C) is about 50–200 nm in diameter (D) which is different from the size of typical COPI vesicles. Bar, 1 μm.

        Golgi markers are in separate structures after cholesterol-induced vesiculation

        To investigate whether the increased cholesterol level affected the compartmentalization of the Golgi apparatus and induced a mixing of Golgi markers, we used confocal microscopy on cells double-labeled with antibodies against different markers. As shown in Figure 4, in control HeLa cells both GM130 and TGN46 were detected in the perinuclear region, where they showed an apparently high degree of colocalization, possibly due to the compact Golgi structure and not due to true colocalization in the same Golgi cisternae. However, after treatment with cholesterol, one can clearly see that the markers are found in different vesicles, indicating that there is still a compartmentalization of the Golgi apparatus. Similarly, in COS-1 cells the markers stay separated (Figure 5). In this case we used nocodazole treatment of the cholesterol-treated cells to obtain a better resolution (Figure 5). Also, other markers could be localized to different vesicles. AP-1 and giantin could be seen in different vesicles both in HeLa and COS-1 cells, and the same was the case for mannosidase II and AP-1 (HeLa cells) (data not shown). Furthermore, by using HeLa cells, we could demonstrate that even after cholesterol treatment, anti-ERGIC-53 (an antibody reacting only with human cells) (32) was only partially colocalized with giantin (data not shown), indicating that also ERGIC still exists as a separate compartment. Also, in COS-1 cells incubation at 15 °C, previously shown to block transport from ERGIC to Golgi (33), was found to arrest VSVG transport at the level of ERGIC after cholesterol treatment (data not shown). Thus, a number of the characteristics of the export machinery are intact in the cholesterol-treated cells. In addition, our immunofluorescence results showed that TGN46 did not colocalize with LBPA (a late endosome marker), Lamp1 (a lysosome marker) and EEA1 (a early endosome marker) in cholesterol-loaded cells (our unpublished data). A likely explanation for lack of colocalization of TGN46 with endosomal markers is that both endosome-to-Golgi and Golgi-to-endosome transport is inhibited by high cholesterol (see below).

        Figure 4.

        Figure 4.

          GM130 and TGN46 are in separate structures after cholesterol-induced vesiculation in HeLa cells. Control cells and cells treated with 5 mm mβCD/chol for 1 h were fixed, permeabilized and double-stained with antibodies against GM130 and TGN46. GM130 was visual-ized by rhodamine-conjugated donkey anti-mouse IgG and TGN46 was detected using a FITC-conjugated donkey anti-goat IgG. Please note that after cholesterol loading, GM130 and TGN46 do not colocalize to any significant degree. Bars, 10 μm.

          Figure 5.

          Figure 5.

            GM130 and TGN46 are in separate structures after cholesterol-induced vesiculation in COS-1 cells. Cells were treated for 2 h with either 5 mm mβCD/chol, or 5 mm mβCD/chol together with 2.5 μg/ml nocod-azole. The cells were then fixed, permeabilized and double-stained with antibodies against GM130 and TGN46, as described in Figure 4. In COS-1 cells, addition of nocodazole is required to visua-lize that GM130 and TGN46 do not colocalize to any significant extent. Bars, 20 μm.

            VSVG is transported into but not out of the TGN vesicles formed in the presence of cholesterol

            We have previously shown that ricin transport from endosomes to the Golgi apparatus is inhibited by increased levels of cholesterol in HeLa cells (8). To investigate whether transport through the exocytic pathway also was inhibited in the cholesterol-loaded cells, we investigated the transport of ts045-VSVG-GFP from the ER to the cell surface. COS-1 cells were used for this study because they are easily transfected with ts045-VSVG-GFPct plasmid DNA (25), as compared with HeLa cells. COS-1 cells transfected with VSVG-GFP were incubated at 39.5 °C for 20 h, and then the temperature was shifted to 31.5 °C in the presence of cycloheximide. As shown in Figure 6 (control), VSVG-GFP fluorescence was localized to the ER (a reticular staining) at the restrictive temperature in most cells. Upon lowering the temperature to 31.5 °C for 30 min, the G protein was redistributed to a Golgi location in the transfected cells as judged by immunolabeling with anti-GM130 antibody. As expected, 3 h after the temperature shift, all the G protein had been delivered to the cell surface. This result is consistent with the temperature-sensitive folding phenotype of this glycoprotein, and is similar to the results obtained by others (13,25–27). When VSVG transport was studied in COS-1 cells treated with cholesterol, the increased cholesterol level did not alter the intracellular distribution of ER-associated G protein at 39.5 °C (Figure 6, mβCD/chol). However, within 30 min of a temperature shift from 39.5 °C to 31.5 °C, the G protein was accumulated in a dispersed Golgi-like structure at the perinuclear region. Labeling with antibody against GM130 showed a striking colocalization of the accumulated G protein with this marker. Even after 3 h, the G protein was still accumulated in this structure, but there was still some colocalization between G protein and GM130. This indicates that a transport step in the exocytic pathway is impaired by increased cholesterol level.

            Figure 6.

            Figure 6.

              Increased cholesterol level blocks the transport of VSVG protein at the Golgi apparatus. COS-1 cells expressing ts045-VSVG-GFP were incubated at 39.5 °C for 20 h. The cells were then incubated in the presence or absence of 5 mm mβCD/chol for 2 h at the same temperature, and then shifted to 31.5 °C for 30 min or 180 min. Subsequently, the cells were fixed, permeabilized, and stained with antibody against GM130, followed by rhodamine-conjugated goat anti-mouse IgG. Note that even after 180 min (the lower panels) the G protein is not transported to the cell surface in cholesterol-loaded cells. Bars, 20 μm.

              To investigate the location at which increased levels of cholesterol block the transport of G protein, the cells were incubated at 20 °C for 2 h after the 39.5 °C incubation. At this temperature secretory proteins are normally accumulated in the TGN (34,35). As expected, at steady-state the G protein displayed a perinuclear Golgi staining, whereas in cholesterol-loaded cells the G protein was localized in a dispersed Golgi structure (Figure 7). In both cases, the G protein colocalized with a TGN marker, TGN46 (36), indicating that the transport of G protein into TGN46-containing structures occurs normally. Then, the temperature was raised to 31.5 °C for 1 h to allow the G protein to exit the TGN and reach the cell surface. As shown in Figure 7, in control cells the G protein was transported out of TGN and had reached the plasma membrane. In contrast, in cells with increased cholesterol levels most of the G protein was arrested in vesicular structures, largely colocalizing with TGN46, indicating that increased levels of cholesterol block the transport of G protein at the level of the TGN.

              Figure 7.

              Figure 7.

                The transport of VSVG protein is blocked at TGN in cholesterol-loaded cells. COS-1 cells expressing ts045-VSVG-GFP were incubated with or without 5 mm mβCD/chol for 2 h at 39.5 °C, and then shifted to 20 °C for 2 h. The cells were either fixed or warmed up to 31.5 °C for 1 h. Following incubation, the cells were fixed, permeabilized and labeled with antibody to the trans-Golgi network marker, TGN46 and visualized using a rhodamine-conjugated donkey anti-goat IgG. Note that in cholesterol-loaded cells the G protein is colocalized with TGN46, indicating that the transport of G protein into the TGN46-containing structures occurs normally. Bars, 10 μm (for 20 °C) and 20 μm (for 31.5 °C).

                Quantitative measurements demonstrate a block in transport of the VSVG protein to cell surface in cholesterol-loaded cells

                In contrast to morphological studies, which are carried out on a limited number of cells, biotinylation offers a means to quantitatively detect the VSVG protein at the cell surface in a large number of cells. Thus, NHS-LC-Biotin was used to analyze ER to plasma membrane transport of the VSVG protein. COS-1 cells were transfected with VSVG and incubated overnight at 39.5 °C, and then incubated at the same temperature in the presence or absence of mβCD/chol for 2 h. Subsequently the cells were either placed on ice or chased at 31.5 °C with or without mβCD/chol for 30 min or 90 min, and then cell surface proteins were biotinylated before lysis. Biotinylated proteins were retrieved with streptavidin-agarose, and then the proteins were analyzed by SDS-PAGE and transferred to nitrocellulose membranes, before immunoblotting was carried out with an antibody against VSVG and an HRP-conjugated secondary antibody. The amount of biotin-VSVG was visualized by ECL. Quantitation of the data showed that the transport of the VSVG protein to the cell surface was completely blocked by increased cholesterol levels (Figure 8). Experiments performed with confocal microscopy illustrated that VSVG in control cells was able to reach the cell surface after 90 min, although the transport to the cell surface was less complete than after 3 h (data not shown). Thus, both morphological studies and biochemical measurements demonstrate that an increased cellular cholesterol level can block the transport of the VSVG protein to the cell surface.

                Figure 8.

                Figure 8.

                  The transport of G protein from ER to cell surface quantitatively measured by biotinylation. COS-1 cells were transfected with ts045-VSVG-GFP as described in Materials and Methods. After incubation either in the presence (open circle) or absence (filled circle) of 5 mm mβCD/chol for 2 h at 39.5 °C, cells were either immediately placed on ice or chased at 31.5 °C with (open circle) or without (filled circle) mβCD/chol for the indicated times. Biotinylated cell surface proteins were precipitated with streptavidin-agarose, separated by 10% SDS-PAGE and transferred to nitrocellulose membranes. The membranes were incubated with anti-VSVG antibody, visualized using HRP-conjugated secondary antibody and ECL. The resulting bands were quantitated by Image Quant 5.0. The error bars show deviations among three independent experiments. Note that the transport of the VSVG protein to the cell surface is completely blocked by an increased cellular cholesterol level.


                  The main finding in the present article is that although increased cholesterol leads to a vesiculation of the Golgi apparatus, it does not abolish the compartmentalization of the Golgi, and the VSVG protein can still be transported into but not out of the vesiculated Golgi apparatus. We have previously found that transport from endosomes to the Golgi apparatus is reduced under the same conditions (8). Thus, increased cholesterol imposes a bidirectional block in transport at the interphase of the TGN. Furthermore, the studies indicate that actin is involved both in the dispersal of vesicles seen after addition of cholesterol and in the reformation of the Golgi apparatus obtained upon removal of excess cholesterol.

                  Multiple members of the dynein, kinesin and myosin families have been found in connection with the Golgi apparatus, and also the large GTPase dynamin has been implicated in the formation of vesicles in the Golgi apparatus (37). However, since inhibition of cytoplasmic dynein 1 (CD1) has been reported to induce the same phenotype as disruption of microtubuli, i.e. formation of several small Golgi structures close to the exit sites of the ER, the cholesterol-induced fragmentation and dispersion is probably not caused by a separate effect on the dynein/microtubuli system. Furthermore, disruption of the actin cytoskeleton may induce a collapse of the Golgi close to the centrosome (11), also a picture that is different from what we observe here. Since anterograde vesicles are enriched in lipid raft constituents, including cholesterol, and retrograde vesicles are normally depleted of these lipids (38), an increased cholesterol level might somehow overwhelm the sorting machinery and lead to vesiculation. A changed trafficking might also be due to direct interaction of motors with lipid rafts and, as a consequence, their activation (37). For instance, KIFC3 and KIF1A might be sorted to and/or activated by lipid rafts. A direct effect of cholesterol on membrane curvature (5,39,40), as well as an effect of cholesterol on membrane thickness (3,41) and thereby on proteins that stay associated with the Golgi membranes, might also play a role in the vesiculation. The vesicles that arise are typically larger than COP-coated transport vesicles, indicating that they do not represent unfused transport vesicles but rather fragments of cisternae.

                  We here demonstrate that stabilization of actin with jasplakinolide prevents the spreading of the cholesterol-induced vesicles, but neither jasplakinolide nor cytochalasin D inhibits the formation of these vesicles, supporting the idea that actin plays a role in the localization of the vesicles but not in their formation. However, actin is required for the reversal of the process. If cytochalasin D is added upon removal of cholesterol, the Golgi morphology is not restored. This is in contrast to what happens when the Golgi apparatus is restored upon removal of brefeldin A. In that case the Golgi apparatus is reformed even in the presence of latrunculin B (15). Little is known about the orientation of the actin filaments associated with the Golgi apparatus. If the Golgi actin filaments have the same direction all the way through the Golgi and outwards, and if the orientation is as suggested by Warner and his colleagues in their study of myosin VI (42), where they suggested that myosin VI moves the vesicles away from the Golgi in a minus-end directed way, then the restoration of the Golgi apparatus is dependent on a plus-end directed actin-binding motor.

                  The changes in the Golgi apparatus were visualized by studying diverse Golgi-resident proteins, such as GM130, mannosidase II, giantin and TGN46. All of these markers could be seen in dispersed vesicular structures, indicating that all the Golgi subcompartments are affected by the cholesterol treatment. Interestingly, there is no noticeable mixing of Golgi markers upon addition of cholesterol. TGN and cis-Golgi markers stay apart.

                  Remarkably, trafficking within the vesiculated Golgi may still occur, since the Golgi markers are separated and the VSVG protein becomes localized with TGN46. Whether this is due to a maturation of Golgi vesicles, as suggested for Golgi cisternae, is not known. If maturation occurs, there is probably retrograde transport of Golgi markers in the vesiculated Golgi apparatus. However, trafficking out of the TGN is inhibited. At present, one can only speculate on why that is the case. A tubular TGN might be required for budding of vesicles destined for the cell surface. Similarly, in mitosis the Golgi apparatus is vesiculated and the transport to the cell surface is blocked (43). The apparent compartmentalization of the disrupted Golgi apparatus might facilitate the restoration of the Golgi apparatus that is observed upon removal of the extra load of cholesterol.

                  It has been reported that Golgi-to-plasma membrane traffic of VSVG-GFP is mediated by large pleiomorphic tubular structures (13,27,44) and also by small vesicles (27), and that the transport is microtubule dependent (13,27), and dependent on kinesin family members (45). Although we could not detect any changes of microtubuli by confocal microscopy after immunolabeling in cholesterol-treated cells (our unpublished data), we cannot exclude that subtle, important changes have occurred. Also, when actin was visualized by labeling with rhodamine-labeled phalloidin, no changes could be seen (our unpublished data). However, we cannot exclude that there are some local changes at the level of the TGN. When Jas was added to COS-1 cells, VSVG transport to the cell surface was somewhat inhibited (about 50%, data not shown), and we have therefore not included studies of transport of VSVG in cells treated with both Jas and cholesterol. There is controversial evidence whether elements of the actin cytoskeleton are involved in the secretory pathway. Some investigators found no evidence for actin filaments involvement in anterograde transport of VSVG (11,27), whereas others showed that in cytochalasin B-treated cells the export of VSVG protein out of the Golgi was delayed, and Golgi-derived tubules were significantly longer than in control cells (13), implicating that actin filaments facilitate protein transport from the Golgi complex. Additional evidence for implication of actin in the transport of Golgi-derived vesicles to the plasma membrane comes from studies of myosin VI (42). Those data imply a function for myosin VI in maintaining Golgi complex morphology and in vesicle transport from TGN.

                  An increased level of cholesterol induces fragmentation and dispersion of the Golgi apparatus in different cell types at a cholesterol level that is about double the normal cellular level. This is actually a lower level than found in Niemann-Pick C disease (46), where cholesterol levels can be increased by a factor of 5. However, the survival of those cells must imply that they manage to keep the cholesterol level in the Golgi apparatus sufficiently low to prevent the changes described here, or that their Golgi structures have become resistant to cholesterol-induced fragmentation.

                  Altogether, the results presented in this study underscore the importance of cellular control of the amount of cholesterol in the Golgi apparatus. Furthermore, these data suggest that actin filaments can be important for restoring the Golgi apparatus after disruption, and actin is clearly involved in localization of the vesiculated Golgi that is a result of increased cholesterol. Whether actin has a similar role also under other conditions where the Golgi apparatus undergoes structural changes remains to be clarified.

                  Materials and Methods


                  Streptavidin-agarose was from Amersham (Buckinghamshire, UK). Jasplakinolide was from Molecular Probes (Eugene, OR, USA). NHS-LC-Biotin, bicinchoninic acid proteins assay kit and Western blotting detection reagents were from Pierce (Rockford, IL, USA). FuGENE 6 was from Roche Molecular Biochemicals (Indianapolis, IN, USA). Glutaraldehyde was from Serva (Heidelberg, Germany). Nitrocellulose membrane was from Schleicher and Schuell Inc. (Keene, NJ, USA). Mowiol was from Calbiochem (San Diego, CA, USA). Methyl-β-cyclodextrin (average degree of substitution, 10.5–14.7 methyl groups per molecule), cholesterol, cholesterol assay kit, cycloheximide, cytochalasin D and cacodylic acid were obtained from Sigma Chemical Co. (St. Louis, MO, USA).


                  The following antibodies were used in immunofluorescence and immunoblotting studies: mouse monoclonal antibody against GM130 was obtained from BD Transduction Laboratory (San Diego, CA, USA); the polyclonal antibody against TGN46 was purchased from Serotec (Oxford, UK); the monoclonal antibody against the VSVG glycoprotein was obtained from Roche Molecular Biochemicals; human monoclonal G1/93 antibody against ERGIC-53 was a generous gift from Dr Hans-Peter Hauri (Biozentrum of the University of Basel, Switzerland). Secondary FITC- or rhodamine-conjugated antibodies and horseradish peroxidase-conjugated IgG were from Jackson Immunoresearch (West Grove, PA, USA).

                  Cell culture

                  All cell culture materials were purchased from Life Technologies (Paisley, UK). HeLa and COS-1 cells were grown in DMEM supplemented with 2 mm L-glutamine, 100 i.u./ml penicillin, 100 μg/ml streptomycin, and 10% (v/v) fetal calf serum. For HeLa cells, 0.2 mg/ml geneticin and 0.5 μg/ml puromycin were added. 20 °C incubation was performed on a water bath in Hepes medium without NaHCO3 (pH 7.4).

                  Preparation of methyl-β-cyclodextrin saturated with cholesterol

                  Methyl-β-cyclodextrin saturated with cholesterol (mβCD/chol) was prepared as previously described (47). Briefly, 1 g of mβCD was dissolved in 20 ml H2O, and then 30 mg of cholesterol was added. The mixed solution was incubated at 37 °C overnight with rotation. The solution was filtrated, freeze-dried and stored at room temperature.

                  Cholesterol and protein determination

                  COS-1 cells were grown to ∼90% confluence on 5-cm dishes. Cells were washed twice with PBS and lysed for 5 min at 37 °C with a lysis buffer containing 0.1% (w/v) SDS, 1 mm EDTA and 0.1 m Tris-HCl, pH 7.4. Subsequently, the lysates were homogenated by using a 19-gauge needle attached to a 1-ml syringe. The cholesterol and protein contents of homogenates were determined using a cholesterol assay kit and a bicinchoninic acid protein assay kit, respectively, according to the manufacturer's instructions.

                  VSVG-GFP transport assay

                  COS-1 cells (∼ 50% confluent) grown on 10-mm coverslips in a 35-mm dish were transiently transfected using FuGENE 6 with ts045-VSVG-GFPct plasmid DNA (25), a kind gift from Dr Suzie Scales (Stanford University Medical Center, Stanford, California). Transfection was carried out according to the manufacturer's instructions using 1 μg of plasmid DNA per transfection. After transfection, the cells were incubated at 39.5 °C for 20 h. Before a temperature shift from 39.5 °C to either 20 °C, or 31.5 °C, 100 μg/ml cycloheximide was added to each dish. The cells were fixed at the indicated time points, and then processed for confocal fluorescence microscopy.

                  Biotinylation of cell surface VSVG protein

                  Cells were cultured on a dish of 10-cm diameter and transfected with ts045-VSVG-GFPct as described above. After 20 h incubation at 39.5 °C, the temperature was shifted to 31.5 °C for the indicated times, and then surface proteins were biotinylated essentially as described by de Hoop and Cid-Arregui (48). Briefly, cells were washed three times with ice-cold PBS containing 0.1 mm CaCl2 and 0.1 mm MgCl2, pH 7.4 (PBS+), and then surface proteins were biotinylated by incubation with 1 mg/ml NHS-LC-biotin in PBS+ for 30 min on ice. The reaction was quenched by washing the dishes twice with ice-cold GB solution [100 mm glycine and 0.3% (w/v) bovine serum albumin in PBS+], and subsequently washed with PBS+ twice. After biotinylation, the cells were lysed for 10 min on ice with a lysis buffer containing 2% (v/v) NP-40, 0.2% (w/v) SDS and protease inhibitors in PBS. The lysates were centrifuged at 14 000 r.p.m. for 15 min at 4 °C to remove cellular debris and insoluble material, and the resulting supernatant was incubated with streptavidin-agarose overnight at 4 °C by rotation. The beads were washed twice with 2% (v/v) NP-40 and 0.2% (w/v) SDS in PBS, twice with 1% (v/v) NP-40 in PBS and twice with 0.1% (v/v) NP-40 in PBS. Subsequently, the beads were incubated for 5 min at room temperature with 25 mm DTT in 50 mm Tris-HCl, pH 7.5, and then washed once with 50 mm Tris-HCl, pH 7.5. The proteins were eluted from the beads with 30 μl of SDS-loading buffer and boiled for 5 min; then the proteins were separated by 10% SDS-PAGE and electrophoretically transferred onto nitrocellulose membranes using a standard protocol for the Bio-Rad minigel and blotting apparatus. The membranes were incubated with blocking solution containing 5% (w/v) dry milk and 0.1% (v/v) Tween-20 in PBS, pH 7.5, and subsequently probed with mouse monoclonal anti-VSVG protein (dilution 1 : 500). The signal was visualized using HRP-conjugated secondary antibody and enhanced chemiluminescence according to manufacturer's protocol (Pierce). The bands were quantitated by Image Quant (Amersham Biosciences Inc., Piscataway, NJ, USA).

                  Confocal fluorescence microscopy

                  Cells grown to ∼ 50% confluence on 10-mm glass coverslips, were fixed for 15 min with 3% (w/v) paraformaldehyde in PBS, washed with PBS, and then permeabilized for 5 min with either 0.3% (for HeLa) or 0.1% (for COS-1) Triton X-100. Subsequently, the cells were washed with PBS and incubated for 30 min with 5% FCS in PBS (blocking buffer). Then primary antibodies diluted in blocking buffer were added to the cells. After 30 min, the coverslips were washed, incubated with rhodamine-labeled goat anti-mouse or donkey anti-mouse, or FITC-labeled donkey anti-goat secondary antibodies for 30 min, washed again, and mounted on slides with Mowiol. The specimens were viewed on a Leica (Wetzlar, Germany) confocal microscope using a 63× or 100× oil immersion objective. Montages of images were prepared with the use of Photoshop 7.0 (Adobe, Mountain View, CA, USA).

                  Electron microscopy

                  COS-1 cells were grown to ∼ 70% confluence on a 25-cm2 flask. After washing with Hepes medium, the cells were incubated either with or without 5 mm mβCD/chol, 3 μm Jasplakinolide (Jas), or 3 μm Jas together with 5 mm mβCD/chol for 2 h at 37 °C. Cells were then washed with PBS, fixed with 2% glutaraldehyde in 0.1 m cacodylate buffer, pH 7.2, for 60 min at room temperature and postfixed with 1% OsO4 followed by 1% uranyl acetate. The samples were then dehydrated through a series of graded ethanol concentrations and embedded in Epon. Sections were contrasted with lead citrate, followed by uranyl acetate if necessary, and viewed on a Philips CM 100 electron microscopy (Phillips, Eindhoven, the Netherlands). In connection with Figure 3, it should be noted that our definition of a Golgi ‘cistern’ is that the longest ‘diameter’ is at least 2 times the shortest ‘diameter’; if not, it is a ‘vesicle’. Quantitation of the distribution of Golgi vesicle size was performed on 200 cells randomly selected from 5 different cell profiles per experiment.


                  We thank Britt Solvår Morken for her excellent technical assistance. We are grateful to Drs Suzie Scales and Hans-Peter Hauri for their generous gifts of ts045-VSVG-GFPct plasmid DNA construct and antibody against ERGIC-53, respectively. This work was supported by The Norwegian Research Council, The Norwegian Cancer Society, The Danish Cancer Society, The Danish Medical Research Council, the Novo-Nordisk Foundation, Blix legacy, Torsted's legacy, the Jahre foundation, and Jeanette and Søren Bothner's legacy.