The specific phosphatase inhibitor okadaic acid (OA) induced fragmentation of the Golgi apparatus in interphase HeLa cells. Immunoelectron microscopy for galactosyltranferase identified a major Golgi fragment composed of a cluster of vesicles and tubules that was morphologically indistinguishable from the ‘Golgi cluster’ previously described in mitotic cells. The presence of homogeneous immunofluorescence staining for galactosyltransferase in OA-treated cells also suggested that isolated Golgi vesicles, previously found in mitotic cells, existed along with the clusters. After removal of OA, both clusters and vesicles appeared to participate in a reassembly pathway that strongly resembled that occurring during telophase.
OA also induced inhibition of intracellular transport, another feature of mitotic cells. OA treatment prevented newly synthesised G protein of vesicular stomatitis virus (VSV) from acquiring resistance to endoglycosidase H and from arriving at the cell surface. In addition, fluid phase endocytosis of horseradish peroxidase (HRP) was reduced to less than 10% of control values. All these effects were dose-dependent and reversible.
OA should be a useful tool to study the Golgi division and membrane traffic.
As animal cells enter mitosis, fragmentation of the Golgi apparatus cisternal stack produces hundreds of ‘Golgi clusters’ and thousands of Golgi vesicles (Zeligs and Wollman, 1979; Lucocq et al. 1989). Both these structures appear to represent end products of a poorly characterised disassembly process, and possess important features that distinguish them from the interphase Golgi stack. Firstly they are both composed of numerous vesicles containing the Golgi proteins that are located within cisternae during interphase (Lucocq et al. 1989); and secondly, they lack cisternae and stacking observed in interphase Golgi apparatus. Together, these features allow us to propose mechanisms for mitotic Golgi apparatus disassembly.
One idea is that the vesicles are generated by the budding process that carries transported proteins across the stack and, due to a relaxation of a normal sorting mechanism, Golgi proteins are now included in the vesicles. If these vesicles can no longer fuse and the cisternae become unstacked then clusters of vesicles will form (Warren, 1985). This would also explain the general inhibition of transport observed in mitotic cells (Berlin and Oliver, 1980; Featherstone et al. 1985; Hesketh et al. 1984; Warren et al. 1983, 1984).
Another suggestion is that enzymes repeatedly cut and reseal Golgi membranes during interphase and mitosis. During interphase they would cut the inter-saccular regions (Rambourg and Clermont, 1990) generating discrete stacks, an effect mimicked by microtubule depolymerising agents (Thyberg and Moskelewski, 1985). During mitosis, unstacking would expose the cisternae themselves to these enzymes, generating vesicle clusters. Inhibition of intracellular transport would be the result of another mechanism.
Investigation of the disassembly mechanism has, up to now, been severely hampered by the difficulty of isolating enough prophase cells for biochemical and quantitative morphological analysis. It would therefore be of particular interest to identify new systems for studying disassembly. Here we report that, whilst screening possible modulators of the autophagic pathway, we found that the phosphatase inhibitor okadaic acid (Takai et al. 1987; Haystead et al. 1989) induced Golgi apparatus fragmentation and arrested intracellular transport in interphase HeLa cells.
MATERIALS AND METHODS
Chemicals and antibodies
Okadaic acid was obtained from either Gibco, (Basel, Switzerland), Calbiochem, (Luzern, Switzerland) or Moana Product, (Hawaii), and stored at -20 °C in 10% dimethylsulphoxide (DMSO) in water at a final concentration of at least 1 mM. Horseradish peroxidase (HRP), o-dianisidine and diaminobenzidine were obtained from Sigma Chemical Co. (Dorset, England). Affinity-purified antibodies to human galactosyltransferase were kindly supplied by Dr Eric Berger, Zurich. Monoclonal antibodies against VSV G protein have already been characterised (Lefran-cois and Lyles, 1982) and were provided by Dr Lefrancois. Endoglycosidase H of Streptomyces griseus was from Boehringer Mannheim, (UK). L-[35S]methionine (1000 Ci mmol−1) was from Amersham, (Bucks England).
Monolayer and suspension HeLa cells were grown in minimal essential medium (MEM) supplemented with penicillin, streptomycin, 10% fetal calf serum (FCS), non-essential amino acids and glutamine in an atmosphere of 5% CO2/95% air. Spinner Hela cells were grown as a monolayer to 75% confluency in 150 cm2 plastic flasks and transferred to spinner flasks at 2 ×105 cells ml 1 twenty four hours before use.
VSV ts045 infection
Vesicular stomatitis virus (VSV) and the temperature-sensitive mutant ts045 of the virus were grown on BHK cells and VSV-infecting units determined (Featherstone et al. 1985). At 32 °C, a 0.5-1 ×108 suspension of HeLa cells was washed in infection medium (IM) (Featherstone et al. 1985) and infected for 2 h at 250 infecting units/cell VSVtsO45 in IM containing 1.8 mM CaCl2 (5×106cellsml−1). The cells (106 cells ml−1) were washed three times in growth medium, and returned to a spinner bottle for 4-5h at 40°C (2×10® cells ml−1). Infection with wtVSV was identical except that the virus incubation lasted for Ih and infection was continued for 4-5 h at 37 °C.
Labelling with [3SS]methionine
VSV-infected cells were washed twice with RPMI 1640 medium supplemented with 1% (v/v) dialysed FCS, 10 mM Hepes, pH 7.4, and 1μgml−1 methionine, and incubated for 60 min in this medium before resuspension at 5×106cellsml 1 in 0.5 ml of icecold RPMI 1640 medium containing 10/iCiml−1 [35S]methionine. The samples were warmed to 37 °C and after 10 min the cells were washed (3 times) in ice-cold calcium- and magnesium-free PBS (CMF-PBS) (containing 10 mM sodium phosphate, 150 mM NaCl, pH 7.4) and chased in 0.5 ml of growth medium containing 5mM methionine at 37 °C for 1 h. Subsequently the cells were washed (3 times) in 1ml of CMF-PBS and extracted with Triton X-114, treated with endo beta-AI-acetylglucosaminidase H (Featherstone et al. 1985) and G protein immunoprecipitated with a monoclonal antibody against VSV G protein (Lefrancois and Lyles, 1982) in a buffer containing 50 mM Tris, pH 8.8, 0.4 M NaCl, 1% deoxycholate, 1% NP40, 5 mM EDTA and protease inhibitors. Proteins were separated by SDS-polyacrylamide gel electrophoresis essentially according to the method of Laemmli, (1970). The fixed gels were fluorographed at —80 °C using Kodak X-OMAT X-ray film.
Fluorescence microscopy and fluorescence-activated cell sorting (FACS)
Cells were grown to confluency on polylysine-coated coverslips and fixed in methanol, acetone at -20 °C and air dried. The coverslips were incubated for 30 min with antibodies against galactosyltransferase, washed in CMF-PBS (3 times over 5 min each), and then placed for 30 min on FITC/goat anti-rabbit antibodies, washed again in CMF-PBS and distilled water and finally mounted in Moviol (Hoescht, Frankfurt, Germany). Observations and quantitation were performed either on a Zeiss Axiovert microscope, Biorad Laserscan MRC 600 confocal microscope or a Zeiss photomicroscope in.
Cells for FACS analysis were fixed in 3% formaldehyde in CMF-PBS. Following two washes with CMF-PBS, one 10 min wash in CMF-PBS containing 50 mM ammonium chloride, three 2 min washes in CMF-PBS and one 5 min wash in CMF-PBS containing 0.2% fish skin gelatin (CMF-PBS/gelatin), the cells were incubated for 20-30 min with culture supernatant from hybridoma cells secreting antibodies to VSV G protein. The cells were then washed with CMF-PBS/gelatin three times before incubating for 20-30 min with FITC-labelled goat anti-mouse second antibody. After further washes in CMF-PBS/gelatin and then in CMF-PBS, the cells were diluted to 1-2×10® cells ml−1 for quantitation of surface fluorescence by flow cytometry using a FACScan (Beckton and Dickinson).
Immuno- and conventional electron microscopy
HeLa cells were grown to confluency on 6 cm diameter dishes. Cells were fixed in 0.5% glutaraldehyde in 0.2 M Hepes, pH 7.2, for 30 min at room temperature. Ultrathin cryosections were prepared and incubated with galactosyltransferase antibodies followed by 6.2 nm protein A-gold essentially as described by Lucocq et al. (1989), except that the labelled sections were embedded and contrasted in methylcellulose/uranyl acetate (Griffiths et al. 1984). For conventional microscopy, glutaraldehyde-fixed cells were postfixed, dehydrated and embedded in epoxy resin as described previously (Lucocq et al. 1987). Ultrathin sections were contrasted with lead citrate.
HRP activity was revealed after pretreatment of cells with or without OA for lh followed by incubation in 5 mg ml−1 HRP for one hour. The cells were fixed as described above and the HRP reaction was performed as detailed by Smythe et al. (1989) before postfixation, dehydration and embedding in epoxy resin.
Fluid phase endocytosis
Cells were grown to confluency in 6 cm diameter plastic dishes and exposed to 1 mg ml−1 HRP in MEM with or without prior OA treatment. The cells were then cooled on ice, treated with 1 ml of trypsin-EDTA for 30 min, resuspended and washed three times in 3 ml of ice-cold Dulbecco’s PBS (DPBS) containing 10% fetal calf serum and six times in DPBS alone. Finally the cells were pelleted at 1000g and solubilized in 1ml of 0.1% (w/v) SDS in water. HRP was assayed within 1 h (Steinman and Cohn, 1972) and protein estimated according to the method of Bradford (1976). Measurements of HRP activity per mg protein were expressed as% of 37 °C uptake.
Golgi fragmentation induced by OA
Treatment with 1 UM OA for 1 h caused all interphase HeLa cells to round up. Immunofluorescence labelling for galactosyltransferase showed that the juxtanuclear Golgi apparatus (Fig. 1A) had become fragmented and dispersed (Fig. IB). This effect was dose-dependent with 0.5 /ZM causing partial Golgi fragmentation and 0.1 /ZM OA none at all (not shown). During the first hour of treatment, the juxtanuclear Golgi reticulum (Fig. 1A) appeared to undergo piecemeal fragmentation, with small clumps of fluorescence dispersed in cells around a residual juxtanuclear accumulation of fluorescence (not shown). At one hour the fluorescence appeared more homogeneous and was interspersed with small clumps of fluorescence usually less than one micrometre across (identical with Fig. IB). Prolonged incubation with OA (up to 3 h) did not change the fluorescence distribution although it then appeared more homogeneous. Overall, the fluorescence distribution appeared very similar to that observed in metaphase HeLa cells stained for galactosyltransferase (Fig. 1C).
Structures labelled for galactosyltransferase in OA-treated cells were revealed by gold labelling on ultrathin frozen sections (Fig. 2C). These showed a striking similarity to mitotic ‘Golgi clusters’ previously found in HeLa cells (compare OA-induced and mitotic clusters in Fig. 2A and B respectively). Both mitotic and OA cluster profiles had the following features in common: (i) they were composed of groups of vesicles and tubules (50 nm across) some of which labelled for galactosyltransferase, (ii) they lacked cisternal profiles and stacking of their components, (iii) they were of similar size, (iv) they contained occasional lucent vesicles (Fig. 2A,B), some of which were labelled for galactosyltransferase, (v) they were bounded by endoplasmic reticulum, sometimes on all sides, (vi) they appeared to be globular in shape, and (vii) they were present in multiple copies. No evidence for galactosyl transferase staining was found in endoplasmic reticulum or on the cell surface.
Golgi reassembly after OA treatment
Washing out the OA induced Golgi apparatus reassembly. Two hours after washout the Golgi remained fragmented and dispersed (Fig. 1B), but after 3-4 h some of the cells contained dispersed clumps of fluorescence (Fig. 1D). At this time the distribution of fluorescence was very similar to that in telophase HeLa cells (Fig. 1E). By 8h, the clumps of fluorescence were mainly located in the perinuclear region (Fig. 1F). Quantitation of the reassembly process is displayed in Fig. 3. This shows that after 2 h there was a rapid decline in the fraction of cells with small fragments of fluorescence, decreasing from 95% to 9% between 2 and 8 h. This was accompanied by the transient appearance of peripheral clumps of fluorescence which were most prominent at 3h but later declined in frequency. After 3h the proportion of cells containing perinuclear clumps increased rapidly to a maximum of 79% at 8h, at which time 16% of cells contained a single fluorescent perinuclear reticulum.
Reassembly was also studied with electron microscopy. Initially, after washing out the OA, fragmentation continued with the cluster profiles reducing in size (0-2 h). However, by 4h Golgi stack profiles had appeared (Fig. 2D) that corresponded to the multiple clumps of fluorescence observed by fluorescence microscopy, and at 6h and 8h Golgi cisternal stacks, situated in the juxtanuclear region, had almost completely replaced the cluster profiles. Reassembly therefore occurred in two steps, initial rebuilding of multiple, peripheral Golgi stacks followed by their congregation and fusion in the juxtanuclear region of the cells.
Intracellular transport of newly synthesized VSV G protein
Inhibition of intracellular transport accompanies Golgi fragmentation in mitotic cells (for a review, see Warren, 1985). Newly synthesised G protein of vesicular stomatitis virus is no longer transported to the cell surface (Warren et al. 1983), fails to acquire endo H-resistant oligosaccharides and accumulates in the endoplasmic reticulum, indicating that transport into the Golgi apparatus is blocked (Featherstone et al. 1985). To study G protein transport after OA treatment, HeLa cells were infected with VSV, pulsed with [35S]methionine and chased. The G protein was extracted, digested with endo H, immunoprecipitated, then analysed using SDS-polyacrylamide gel electrophoresis (Fig. 4A). In the absence of OA, all of the G protein became resistant to endoglycosidase H after a 2h chase, marking its passage through medial Golgi cisterna(e). Pre-treatment with 1 μM OA prevented this occurring, an effect which over several hours was reversed by washing out the OA.
OA could have prevented construction of complex oligosaccharides in the Golgi apparatus without affecting intracellular transport. Therefore the arrival of the G protein at the cell surface was also assayed, by combining immunofluorescence localisation with flow cytometry. HeLa cells were infected with the temperature-sensitive 045 strain of VSV (tsO45-VSV) and incubated at the non-permissive temperature, which causes accumulation of the viral G protein in the endoplasmic reticulum. Shift to the permissive temperature resulted in the surface appearance of the G protein (Fig. 4B) that was prevented by pretreatment with OA and its continued presence after the shift. Removal of OA at the shift eventually resulted in surface G protein, even in the presence of cycloheximide.
Fluid phase endocytosis
Fluid phase endocytosis is also inhibited during mitosis in animal cells (Fawcett, 1965; Berlin and Oliver, 1980), an effect that was also mimicked by OA. Uptake of HRP at 37 °C was reduced to less than 10% in the hour following OA treatment whether or not OA was present (Fig. 5), and again, reversion occurred on washing out the OA, uptake returning to approx. 67% of normal 6 h later. Electron microscopy showed that staining for HRP present in numerous endosomal/lysosomal structures in non-treated cells was almost completely inhibited by OA treatment (not shown). Only rarely did OA-treated cells contain positive structures that were lightly stained.
OA-treated cells were examined for other features of mitosis. OA induced chromatin condensation in only 20% of the cells. On washing out the OA, the proportion of cells showing chromatin condensation fell from 22% at 2 h to 12.2% at 3 h, and 0% and 1% at 6 h and 8 h respectively, indicating reversion of this effect. Conventional electron microscopy showed that the nuclear envelope remained intact throughout OA treatment, and immunofluorescence localisation also showed no change in the location of lamin B, a component of the nuclear lamina which becomes dispersed in the cell during mitosis (Burke et al. 1983) (not shown).
In this report we describe the effects of OA on interphase HeLa cells. We found that OA induces fragmentation of the Golgi apparatus to produce multivesicular clusters that are indistinguishable from mitotic Golgi clusters, and we have also obtained indirect evidence that OA-treated cells contained isolated Golgi vesicles, which in mitotic cells represent a second type of Golgi fragment. As these disassembly products are so similar to the mitotic ones, OA treatment should now allow investigation of the disassembly process itself, overcoming the difficulty of isolating enough prophase cells.
The strong similarities between the effects of OA and mitosis on the Golgi apparatus were further emphasised when we studied the reassembly of the Golgi apparatus after washing out the OA. As we previously observed in telophase cells, reassembly occurs in two steps: first, multiple Golgi stacks reassemble in the peripheral cytoplasm, and these later congregate and fuse in the juxtanuclear region (Lucocq et al. 1989). The main difference was that the whole process took much longer than in telophase, taking hours rather than minutes, an observation that may be related to a slow reactivation of phosphatases inhibited by OA (see below).
There are other agents that cause Golgi apparatus disassembly. Nocodazole, for example (Thyberg and Moskalewski, 1985), produces Golgi apparatus fragments, but these are distinct from mitotic Golgi fragments because they still contain cisternal stacks and are not exclusively composed of numerous vesicles observed in Golgi clusters (HeLa cells, unpublished observations). Another substance that perturbs the Golgi apparatus structure is Brefeldin A (Misumi et al. 1986). Again, despite the fact that it causes a number of effects such as fusion of Golgi apparatus cisternae (Orci et al. 1991) and redistribution of their enzymes to the endoplasmic reticulum (Lippincott-Schwartz et al. 1989), it also fails to induce formation of any structures resembling mitotic Golgi fragments. OA therefore, appears to be the only agent available that closely reproduces parts of the Golgi division process in interphase cells.
Another major feature of mitosis is arrested intracellular transport, and this was also observed in OA-treated interphase cells. We found that both transport of newly synthesized VSV G protein into the Golgi apparatus (Featherstone et al. 1985) and fluid phase endocytosis were inhibited, suggesting that, as in mitosis, inhibition of transport is a generalised phenomenon (Berlin and Oliver, 1980; Featherstone et al. 1985; Hesketh et al. 1984; Warren et al. 1983). Again reversion of these effects was slow, occurring over a similar period to the Golgi reassembly, a feature that may facilitate detailed studies of these processes.
But how could okadaic acid cause these effects? It seems unlikely that OA has non-specific toxic effects on the cells, because neither changes in ATP and ADP levels (Haystead et al. 1989) nor morphological evidence of cell demise (this study) are evident after OA treatment. Most likely, the effects of OA are explained by a reversible inhibition of protein phosphatases such as PPI and PP2A. The concentration used here (1 UM) is certainly sufficient to inhibit both phosphatases completely without directly affecting other known phosphatases or kinases (Haystead et al. 1989), and its effects are slowly reversible even in the presence of cycloheximide. The slow reversion is probably related to the extremely tight binding of OA that is indicated by the significant inhibition observed when the concentration of OA is close to that of these two phosphatases (Haystead et al. 1989).
Protein phosphatases are however unlikely to be directly involved in causing the effects of OA on Golgi apparatus and intracellular transport. Rather, protein kinases, generally implicated in generating mitotic effects such as organelle disassembly, are likely to be responsible (e.g. Peters et al. 1990); OA has in fact been shown to activate the cdc2 kinase (Nurse, 1990) and induce a mitotic-like state (Goris et al. 1989; Felix et al. 1990; Picard et al. 1989; Yamashita et al. 1990). However, in HeLa cells, we found no increase in histone Hl (cdc2) kinase activity (25+5.7 S.E.M. pmoles Hl phosphorylated min−1 mg−1 cell protein) over interphase levels (23±4.5 S.E.M. pmoles min−1 mg−1) (6 measurements over 4h). Moreover we did not observe some other features of mitosis, such as nuclear envelope and lamina disassembly or extensive chromatin condensation, that are thought to be dependent on kinase activities (Baitinger et al. 1990; Labbe et al. 1988 and Peters et al. 1990). OA therefore seems to act independently of cdc2-kinase in interphase HeLa cells and selectively induces only certain mitotic features. The simplest explanation is that Golgi fragmentation is due to kinases that are activated by cdc2-kinase during mitosis. In interphase their activity is kept at low levels by the action of phosphatases but is revealed by the action of phosphatase inhibitors such as OA, leading to Golgi fragmentation and inhibition of transport. Only further work will establish the exact site at which OA acts.
In summary, OA mimics mitotic features of the Golgi apparatus and membrane traffic pathways and should therefore be a useful tool to study processes such as Golgi disassembly.
We thank Corinne Hug for technical assistance, Mr E. Souter for performing the Hl kinase assays, Dr E. Berger for the antigalactosyltransferase antibody, Dr Lefrancois for the monoclonal anti-G protein antibody and Professor P. Cohen for gifts of okadaic acid. J.L. was supported by grant No.31-27636.89 from the Swiss National Science Foundation.