Stacking of Golgi cisternae in Schizosaccharomyces pombe requires intact microtubules

The central feature of the Golgi apparatus is the stack of closely-apposed and flattened cisternae present in almost all eukaryotic cells. The function of this stack is unknown but one obvious possibility is that it would facilitate transport of newly synthesised proteins (Palade, 1975). Vesicles bud from the dilated rims of one cisterna and fuse with the next cisterna in the stack towards the trans side (Rothman and Orci, 1992). The close proximity of cisternae in a stack would ensure rapid and efficient transport and even elimi-nate the need to target vesicles specifically to the next cis-terna. Budded vesicles would simply fuse with the nearest cisterna. The cup-shaped structure of the Golgi in many cells would even ensure fusion with the next cisterna rather than the one from which it had just budded. Unfortunately, all of the available evidence is against this interpretation. Rothman and colleagues have shown, both in vivo and in vitro, that budded vesicles will fuse with the correct cis-terna even when it is present in a different stack and do so with the same speed and efficiency (Rothman et al., 1984a,b). In addition, a number of fungi, most notably Saccharomyces cerevisiae, have Golgi that mostly comprise a single cisterna (Preuss et al., 1992). These nevertheless carry out a sequence of modifications to proteins undergo-ing transport, which are most easily explained as vesicles budding from one cisterna and fusing with a completely separate one (Graham and Emr, 1991). Taken together, these data argue that budded vesicles are highly diffusible and have targeting information that would preclude the need to arrange the cisternae in a stack.

Another possible function for the stack is to recover pro-teins that have been lost from the endoplasmic reticulum (ER) (Rothman, 1981). Stacked cisternae would allow repeated attempts to recover the lost proteins and return them to the ER. This model was originally proposed for ER membrane proteins but no supporting evidence was found at that time (Brands et al., 1985; Yamamoto et al., 1985). More recent work shows that ER membrane proteins are lost and can be recovered (Jackson et al., 1993) as can sol-uble ER proteins (Pelham, 1989). The latter are recovered by a receptor that recognises the C-terminal sequence-KDEL in mammals and-HDEL in S. cerevisiae. The absence of stacked cisternae in the latter argues against a role for stacking in the salvage operation.

The main experimental difficulty in determining the function of the Golgi stack has been the lack of an experimental system. Conditions that separate the cisternae invariably vesiculate them. These include treatment with ionophores such as A23187 (Blomfield et al., 1983) and physiological states such as mitosis, during which the entire Golgi apparatus is converted to thousands of small vesicles (Warren, 1993). Anti-microtubule agents have a different effect. Nocodazole breaks down the single-copy Golgi found in animal cells yielding multiple discrete stacks, dispersed throughout the cell (Thyberg and Moskalewski, 1985). Nocodazole has no effect on microtubules in Schizosaccharomyces pombe cells (Walker, 1982; Ayscough, unpublished observations) but a chemically related drug, thiabendazole (TBZ), was shown to inhibit cell division by blocking mitotic spindle formation (Walker, 1982). During studies with this drug on S. pombe we found that the Golgi could be reversibly unstacked. Here we have characterised this effect and shown that it depends on intact microtubules.

All chemicals used were from Sigma Chemical Co. Ltd or BDH, Poole, UK, except where otherwise stated. TBZ was made up at 10 mg/ml in DMSO and used on log phase cells in suspension at 100 μg/ml.

The strains used were 972 h (wild type) provided by P. Nurse, Oxford University, UK, and nda2-KM52 and nda3-KM311 pro-vided by M. Yanagida, Kyoto University, Japan. Cells were grown in rich medium, YES (Moreno et al., 1991), with shaking at 32°C.

Microtubules were visualised using the TAT1 monoclonal anti-bodies raised against α-tubulin of Trypanosoma brucei, a gift from K. Gull, Manchester University, UK. The antibodies to α-1,2 galactosyltransferase (GalT) were polyclonal raised in rabbit to a fusion protein of GalT and β-galactosidase (K. Ayscough and G. Warren, unpublished work). The serum was used at a dilution of 1:100.

Immunofluorescence was performed as described by Moreno et al. (1991) with methanol fixation except that all washes there-after used 100 mM Pipes-NaOH, pH6.9, 1 mM EGTA, 1 mM MgCl₂, 1.2 M sorbitol (PEMS buffer); cells were subsequently blocked in PEM containing 2% fish skin gelatin (PEMF) and anti-body incubations were at 4°C. The secondary antibody was rho-damine-conjugated goat anti-rabbit (TAGO Inc. Burlingham, CA, USA) used at 1:100. Cells were mounted in Citi-fluor antifade (Agar Aids) containing 1 μg/ml DAPI (4′6′-diamidino-2-phenylindoledihydrochloride) on coverslips coated with chrome alum gelatin. Cells were observed using a Zeiss Axiophot micro-scope through a ×100 objective and photographed using Ilford HP5 film.

The cells were spun down, washed extensively in distilled water and then resuspended in freshly prepared 2% potassium permanganate for 45 minutes at room temperature. The cells were then washed again in water. Pellets were processed by dehydration in a graded series of ethanol followed by propylene oxide and embedded in Epon 812 resin (Taab Laboratories Equipment Ltd.). Sections were cut with a diamond knife and mounted on carbon/formvar-coated grids. Sections were stained for 10 minutes in a 2% (w/v) solution of uranyl acetate in 50% (v/v) ethanol and for 4 minutes with a solution of lead citrate according to Reynolds (1963). They were viewed in a Philips CM10 electron microscope.

Yeast cells were grown to mid-log phase in YES and then fixed with 1% monomeric glutaraldehyde in 0.1 M Sørensens phosphate buffer (pH 7.4) for 30 minutes at room temperature. After fixation the cells were treated with 0.5 M NH₄Cl in phosphate buffer for 2 hours. The cells were dehydrated in a graded series of methanols at progressively lower temperature and infiltrated with Lowicryl HM20 at −50°C. The resins were polymerised by UV light for 24 hours at −50°C. Ultrathin sections were mounted on carbon-coated grids and were labelled as follows. After 5 minutes on a drop of PBS and 1 hour preincubation on 10% normal goat serum, 10% ovalbumin plus 10% BSA in PBS, the grids were transferred onto drops of anti-GalT antiserum (1 in 10 dilution in PBS) overnight at 4°C. The sections were then washed three times over a 15 minute period with PBS, then incubated with immuno-gold conjugates (goat anti-rabbit IgG, 10 nm gold) for 2 hours. After three final rinses in PBS and three rinses in distilled water the sections were air-dried and then contrasted with uranyl acetate and lead citrate and examined with a Joel 1200 EXII electron microscope. As a negative control the primary antibody was omit-ted from the procedure to visualise any nonspecific binding of gold particles on cell organelles.

Cells were spheroplasted (Moreno et al., 1991) and acid phos-phatase activity measured (Schweingruber et al., 1986a). Spec-trophotometric assays were performed on the medium to measure secreted activity with p-nitrophenyl phosphate as the substrate for acid phosphatase.

Cells were grown to log phase, then spheroplasted and lysed (Moreno et al., 1991). Cell extracts were separated by PAGE and proteins detected by western blotting using the Amersham ECL protocol. A monoclonal antibody against S. pombe acid phos-phatase, 7B4, was used at 1 in 10 dilution, the secondary anti-body was horseradish peroxidase-conjugated goat antimouse (TAGO Immunochemicals) used at 1 in 75 dilution.

Log phase cells were treated for 4 hours with TBZ at 100 μg/ml, a concentration chosen in preliminary experiments as the level necessary for complete disassembly of the microtubular network. The cytoplasmic microtubules that extend from one end of the cell to the other (Fig. 1A) dis-appeared completely leaving only one or two dots (Fig. 1B).

When viewed in cells stained with the nuclear stain, DAPI, these dots were shown to be adjacent to the nucleus and most likely represented spindle pole bodies.

The Golgi apparatus comprises multiple, discrete stacks dispersed, apparently at random, throughout the cell cyto-plasm and can be visualised using an antibody to an α-1,2 galactosyltransferase (Chappell and Warren, 1989; Fig. 1C). TBZ caused an increase in cytoplasmic fluorescence and in the number of the discrete fluorescent spots that represent Golgi stacks (Fig. 1D). The number of Golgi stacks in a cell increases as cells grow (data not shown) so this observation could be explained by the average increase in the length of the cells in a TBZ treated population. TBZ pre-vents assembly of the mitotic spindle so cells grow but cannot divide.

TBZ had a dramatic effect on the Golgi apparatus when examined by electron microscopy. At low power most sections through untreated cells showed one or more transverse sections through Golgi stacks (Fig. 2A). After treatment with TBZ for 4 hours, few if any stacks remained (Fig. 2B). Fig. 2C-F shows high power images of the Golgi at increasing times after treatment. The clear impression is that the

Golgi unstacks yielding individual cisternae, which do not appear to break down further.

This result was not the consequence of using permanganate to fix the cells. Permanganate was used because it gives a high level of membrane contrast, however it does so by removing large parts of the cytoplasm. TBZ could simply have made the Golgi more sensitive to disruption by permanganate. Samples were therefore fixed with monomeric glutaraldehyde and embedded in Lowicryl resin. As can be seen in Fig. 3, the effect of TBZ was still the same. The loss of Golgi stacks (Fig. 3A) was accom-panied by an increase in structures (Fig. 3B-F) similar in appearance and distribution to those identified as single cisternae after permanganate staining. Their identity was con-firmed by using antibodies to GalT. This enzyme is present in most, if not all, cisternae in the stack (Fig. 3A) and in most of the single cisternae (Fig. 3B-F). Other organelles were also examined for an effect of TBZ. The mitochon-dria, vacuoles, nuclear envelope and peripheral ER were all unaffected.

The effect of TBZ on cisternal stacking was quantified in two ways. Cells were selected randomly and every cisterna in these cells was counted and scored according to the size of stack in which it was found. Only short lengths of mem-brane (< 1 μm) with dilated rims were judged to be single cisternae whilst stacked cisternae were easily recognisable. As shown in Fig. 4, TBZ had a dramatic effect on the number of cisternae in the stack, particularly on those stacks with three or more cisternae. In the absence of TBZ nearly 80% of identifiable cisternae were present in stacks, this dropped to less than 40% after 4 hours of treatment. There was a corresponding rise in single cisternae from about 20% to 70%. These changes occurred without any significant change in the absolute number of cisternae per cell section (from 4.6 to 4.0 cisternae/cell section) arguing strongly that TBZ causes cisternal unstacking. Furthermore, the average length of all cisternae changed from 62±9 μm before treat-ment to 58±7 μm afterwards arguing that single cisternae were the final products and further breakdown did not occur.

Identification of single cisternae relied largely on mor-phological criteria and the unequivocal identification of single cisternae in the cell sections was not always straight-forward. They could often be confused with internal ER, which is the fraction of the ER that does not underlie the plasma membrane and does not surround the nucleus (Pidoux and Armstrong, 1992). A second method, which relied on measurements of the surface density of Golgi and ER membranes was therefore used to confirm the results. The surface area of randomly chosen Golgi stacks, internal ER, nuclear envelope and peripheral ER was measured by point counting and expressed as a percentage of the area of the cell cytoplasm, calculated in the same way. Single Golgi cisternae were included in the measurements of internal ER. As shown in Fig. 5 there was a 12% increase in the sur-face density of internal ER, similar to the 18% fall in stacked Golgi membrane. The increase in internal ER was most likely due to an increase in single Golgi cisternae because the surface density of ER did not change. Both the nuclear envelope and the peripheral ER were little affected by treatment with TBZ (Fig. 5).

Disassembly of microtubules in the presence of TBZ occurred with a half-time of less than 10 minutes and was completely reversible after 4 hours of treatment (Fig. 6).

Reassembly was equally rapid with cytoplasmic micro-tubules reforming within 10 minutes. Cisternal unstacking took slightly longer than the disassembly of microtubules with a half-time of 20-25 minutes (Fig. 7A). The effect was reversible after 4 hours of treatment. When TBZ was washed out and cells resuspended in fresh medium the stacks reformed with a half time of between 30 and 60 minutes. (Fig. 7B).

The nda3 mutant was isolated as a cold-sensitive mutant that blocks cells in mitosis at the non-permissive tempera-ture (Umesono et al., 1983). Sequencing the gene showed it to be 75% homologous to that for chicken β-tubulin (Hiraoka et al., 1984). Mutations in this gene can confer resistance to TBZ and, under the conditions used in this study, intact microtubules were found in these cells at all stages of TBZ treatment when assessed by immunofluores-cence with the TAT1 antibody (data not shown).

When nda3 cells were grown at the permissive temper-ature of 32°C, TBZ had little effect on the surface density of Golgi stacks (Fig. 8A). The surface density fell from 17% to 14% whereas that of wild-type cells fell from 20% to 2%.

However, when the cells were shifted to the non-per-missive temperature of 20°C the mutant β-tubulin was unable to form a mitotic spindle so the cells became blocked in mitosis and cells contained no microtubules. After 10 hours incubation at 20°C, the surface density of Golgi stacks fell from 17% to 4% (Fig. 8B). In marked contrast, wild type cells were unaffected by this change in temper-ature (Fig. 8B).

Similar experiments were carried out on the nda2 mutant, which carries a mutation in the α-tubulin gene (Toda et al., 1984). The strain used, nda2-KM52, is not resistant to TBZ and at the permissive temperature behaved in the same way as wild-type cells. At the non-permissive temperature the cells lacked microtubules and the surface density of Golgi stacks fell dramatically (data not shown).

Cell growth can be used to measure Golgi function since growth will be prevented if plasma membrane and cell wall components are not secreted by the cell. Log phase cells were incubated in the presence or absence of TBZ and their growth followed by measurements of turbidity at 595 nm. As shown in Fig. 9A the growth rate was barely affected for at least 2 hours, much longer than the time taken to break down the microtubules and unstack the Golgi appa-ratus. At later times, cells became blocked in mitosis, a time during which the rate of secretion is reduced (Mitchison, 1991).

The secretion of acid phosphatase was measured directly. Spheroplasts were incubated for 3 hours at 32°C in the absence or presence of TBZ and the amount of acid phosphatase secreted into the medium was measured. As shown in Fig. 9B the secretion of acid phosphatase was relatively unaffected by TBZ. Only after 2 hours of total treatment did significant differences become apparent.

Acid phosphatase acquires a large and variable number of mannose and galactose residues as it passes through the Golgi stack. Cell extracts were fractionated by SDS-PAGE, blotted and probed with the monoclonal antibody, 7B4 (Fig. 9C, lane 1). Two forms of the acid phosphatase were present. (Schweingruber et al., 1986b). The ER form had a molecular mass of 72 kDa and was core glycosylated as shown by its sensitivity to endoglycosidase H (Endo H) (Fig. 9C, lanes 1 and 2). The Golgi forms were a ‘smear’ ranging in molecular mass from 95-130 kDa and were also sensitive to endo H. Treatment with TBZ had no effect on the ratio of these two forms (Fig. 9C, c.f. lanes 1 and 3) or on their sensitivity to Endo H (Fig. 9C, c.f. lanes 2 and 4) suggesting that the rate of transport through the Golgi was not affected by cisternal unstacking.

There is a strict correlation between the presence of micro-tubules in S. pombe and the presence of Golgi stacks. Microtubules were broken down either using TBZ or by shifting the nda3 mutant to the non-permissive temperature. Golgi stacks disappeared under both conditions. Further-more, the nda3 mutant at the permissive temperature was resistant to TBZ and under these conditions the Golgi stacks were unaffected. This shows that the effect of TBZ on Golgi stacks is a consequence of its action on microtubules and cannot be attributed to a side effect of the drug. Kinetically, the breakdown of microtubules was closely followed by the loss of Golgi stacks and in all cases the effects were reversible. A requirement for protein synthesis during reassembly could not be tested because the structure of the Golgi apparatus is also sensitive to inhibitors of protein syn-thesis (Ayscough and Warren, 1993). An inhibition of pro-tein synthesis could not, however, explain the results pre-sented here. Inhibitors of protein synthesis cause vesiculation of the Golgi apparatus rather than unstacking, and major changes in protein synthesis were not observed after addition of TBZ.

The effect of microtubule breakdown was to convert stacks of cisternae to single cisternae. Two counting meth-ods were used. In the first, random cell sections were scored for the presence of single Golgi cisternae or multiple cis-ternae in a stack. In untreated cells more than 80% of the cisternae were present in stacked structures. After treatment with TBZ, the total number of cisternae was very similar but only 28% of the cisternae were in stacked structures. The other 72% were present as single cisternae.

In the second method, the surface density of Golgi stacks and ER membranes were measured. The criteria for stacked cisternae were more rigorous than those for the first method requiring that a scored cisterna had another immediately next to it. This explains why treatment with TBZ, or a shift to the non-permissive temperature in nda3 mutants, caused 90% of the stacks to disappear compared to 65% for the first method. In addition, this method allowed direct com-parison between membranes so it was not necessary to iden-tify single Golgi cisternae. They were included in the mea-surements for internal ER, the membrane with which they were most often confused. The results showed that an 18% fall in the surface density of Golgi stacks after treatment for 4 hours with TBZ was accompanied by a 12% increase in the surface density of internal ER. The surface density of peripheral ER and nuclear envelope did not change sig-nificantly during treatment suggesting that most if not all of the increase in internal ER could be attributed to single Golgi cisternae. Since the increase accounted for most but not all of the fall in Golgi stacks it is possible that some of the Golgi cisternae were broken down further, yielding small vesicles that would not have been counted using these procedures. However, partial vesiculation should have caused a change in the average length of Golgi cisternae but the observed change from 62±9 μm before treatment to 58±7 μm afterwards was not statistically significant.

The only data that do not, at first sight, fit this interpre-tation are the immunofluorescence images in Fig. 1D. GalT is present in most cisternae in the Golgi stack (T. G. Chap-pell et al., unpublished), so unstacking should have caused a 2-3 fold increase in the number of labelled structures in each cell. An increase was observed but this was in line with the increase in cell length caused by the inhibition of cell division. The simplest explanation is to suggest that unstacked cisternae do not migrate far from the original stack and so cannot be resolved as separate structures by immunofluorescence microscopy. Evidence in support of this view comes from images such as those seen in Fig. 2E,F where unstacked cisternae remain clustered together. Such clusters might help explain the lack of effect of TBZ on transport through the Golgi stack. Overall, cell growth was unaffected for several hours as was the rate of transport of acid phosphatase both through the Golgi and into the surrounding medium. This suggests that transport vesicles can reach their target cisternae as quickly even if they are not ordered in a stack. An alternative possibility is that the remaining Golgi stacks can cope with all of the newly synthesised proteins leaving the ER. Since our most rigorous quantitative estimates put the figure as low as 10% this would mean that, under normal conditions, the Golgi is only working at 10% of its capacity. Though this seems unlikely it remains a formal possibility that cannot yet be dismissed.

How does the breakdown of microtubules cause unstack-ing? One obvious possibility is that the microtubules are themselves stacking proteins. Microtubules have a width of 250 Å, much wider than the intercisternal width in animal cells of ∼100 Å (Cluett and Brown, 1992). The inter-cisternal width in S. pombe is bigger, 133 ű52 Å in per-manganate-fixed samples, 218 ű37 Å in glutaraldehyde/formaldehyde-fixed samples, but still not sufficient to accommodate microtubules. In addition, there was no co-localisation of the Golgi with microtubules by immunoflu-orescence microscopy (Fig. 1, and data not shown). It is of course possible that the microtubules are linked to stacking proteins but they would only have access to those at the cisternal edges so it is difficult to see how they could control stacking. In addition, none of the known microtubule-binding proteins have been located to the inter-cisternal space.

Another possibility is suggested by the work on the bud-ding yeast, S. cerevisiae. The Golgi exists mostly as single cisternae (Preuss et al., 1992) yet certain secretory (sec) mutants accumulate what appear to be Golgi stacks (Novick et al., 1981; Svoboda and Necas, 1987). In other words, a stacking mechanism exists but the Golgi spends most of its time as single cisternae. Perhaps these cisternae are moved around the cell on microtubules. Microtubules bind to Golgi membranes (Corthésy-Theulaz et al., 1992; Ho et al., 1989) and one of their best characterised functions is the movement of membranes from one part of a cell to another (Schroer and Sheetz, 1991). Since a stacking mechanism exists this suggests that the cisternae need to come together occasionally, perhaps for the purpose of exchanging mate-rial by vesicle-mediated transport. Movement along micro-tubules would ensure that cisternae have the opportunity for stacking and once they had exchanged material the cister-nae could separate and move back onto microtubules. Pro-viding the time spent together was much less than the time spent travelling the Golgi would appear morphologically as single cisternae. In S. pombe the cisternae would spend longer together and less time travelling on microtubules so that most of the Golgi would appear to be stacked. How-ever, in the absence of microtubules, when the cisternae separate, they would be unable to go anywhere. This would explain unstacking and why the single cisternae appear clustered together (see above and Fig. 2E,F).

What could be the reason for such a mechanism? Vesi-cles that bud from the dilated rims of Golgi cisternae con-tain proteins being transported but not resident enzymes. These must be retained. In animal cells the membrane-span-ning region of Golgi enzymes is required for retention (Machamer, 1991) though there is a role for the cytoplas-mic tail (Nilsson et al., 1991). In budding yeast, the Golgi enzymes, Kex1p (Cooper and Bussey, 1992) and Kex2p (Wilcox et al., 1992), have been shown to require their cyto-plasmic tails for retention in the Golgi. Perhaps these tails become anchored to the tails of proteins in an adjacent cis-terna. Stacking would therefore provide a retention mech-anism ensuring that only unattached proteins are transported (Warren, 1989). What this does not explain is why the stack is a transient rather than a stable structure as it appears to be in animal cells. Whatever the reason, this system pro-vides, for the first time, a means of testing functions that might depend on the stacking of Golgi cisternae.

We thank P. Nurse and M. Yanagida for yeast strains, K. Gull and M. E. Schweingruber for antibodies, Tom Misteli, Deborah Mackay and James Close for critical reading of the manuscript and Marc Pypaert for advice on stereology.