In mammalian cells, the Golgi apparatus is disassembled at the onset of mitosis and reassembled at the end of mitosis. This disassembly–reassembly is generally believed to be essential for the equal partitioning of Golgi into two daughter cells. For Golgi disassembly, membrane fusion, which is mediated by NSF and p97, needs to be blocked. For the NSF pathway, the tethering of p115-GM130 is disrupted by the mitotic phosphorylation of GM130, resulting in the inhibition of NSF-mediated fusion. In contrast, the p97/p47 pathway does not require p115-GM130 tethering, and its mitotic inhibitory mechanism has been unclear. Now, we have found that p47, which mainly localizes to the nucleus during interphase, is phosphorylated on Serine-140 by Cdc2 at mitosis. The phosphorylated p47 does not bind to Golgi membranes. An in vitro assay shows that this phosphorylation is required for Golgi disassembly. Microinjection of p47(S140A), which is unable to be phosphorylated, allows the cell to keep Golgi stacks during mitosis and has no effect on the equal partitioning of Golgi into two daughter cells, suggesting that Golgi fragmentation-dispersion may not be obligatory for equal partitioning even in mammalian cells.
The Golgi apparatus occupies a central position in the classical secretory pathway, where it receives the entire output of de novo synthesized proteins from the ER, and functions to distill, posttransitionally process, and sort cargo to their ultimate destinations (Mellman and Simons, 1992). For its inheritance, there are several strategies: de novo formation, fission, and disassembly–reassembly. Glick and coworkers examined the Golgi during mitosis in the budding yeast Pichia pastoris and found de novo formation of Golgi in a daughter cell (Bevis et al., 2002). In the protozoa Toxoplasma gondii, Warren and coworkers reported binary fission of Golgi for its inheritance (Pelletier et al., 2002). In this case, the Golgi splits in half before mitosis and segregates into the two daughter cells. In contrast, animal cells utilize the strategy of disassembly–reassembly (Warren, 1993). During mitosis, the Golgi apparatus is fragmented into thousands of vesicles and short tubules that are dispersed throughout the cytoplasm. Some or all of them might be absorbed into ER, which is still a matter of contention (Zaal et al., 1999; Jokitalo et al., 2001). At telophase, a Golgi apparatus is rapidly reassembled from the fragments within each daughter cell (Lucocq et al., 1989).
This disassembly–reassembly process is regulated by phosphorylation (Shorter and Warren, 2002), but there are continuing debates as to which kinases are responsible. Suggested kinases include Cdc2 kinase (Lowe et al., 1998; Kano et al., 2000), MAPK kinase (Acharya et al., 1998; Kano et al., 2000), and Polo-like kinase 1 (Sutterlin et al., 2001). Since the Golgi apparatus has a complicated structure, its disassembly is thought to occur in several steps. These kinases might be involved in these disassembly steps. Golgi disassembly–reassembly requires the blocking of membrane fusion at early mitosis and its unblocking at late mitosis (Warren, 1993). Experiments using an in vitro assay showed that Golgi reassembly requires at least two distinct membrane fusion pathways: the NSF pathway and the p97/p47 pathway (Rabouille et al., 1995c). NSF binds to SNAREs via α-SNAP and activates it for membrane fusion (Sollner et al., 1993). NSF also needs the assistance of p115-GM130 tethering (Nakamura et al., 1997). At early mitosis, GM130 is phosphorylated on Serine (Ser)*-25 by Cdc2 kinase. This phosphorylation disrupts the tethering of p115-GM130, resulting in the mitotic inhibition of the NSF pathway (Lowe et al., 1998). Similar to NSF, p97 binds to SNAREs via the cofactor p47 (Kondo et al., 1997) and dissociates the SNARE complex with the assistance of the valosin-containing protein (p97)/p47 complex–interacting protein (VCIP135) during membrane fusion (Uchiyama et al., 2002). It has also been suggested that the interaction between p47 and monoubiquitin might be important (Meyer et al., 2002). In contrast to NSF, p97 does not require p115-GM130 tethering (Rabouille et al., 1995c). This leads to the question: how is the p97 pathway regulated mitotically during the Golgi disassembly–assembly?
In this paper, we first observed the main localization of p47 in the nucleus, which suggests that the major population of p47 does not function in the maintenance of the Golgi apparatus during interphase. Next, we found that mitotic phosphorylation of p47 inhibits its binding to Golgi membranes. This phosphorylation was shown to be required for Golgi disassembly in vitro. Finally, we succeeded in maintaining Golgi stacks in mitotic mammalian cells by the microinjection of p47(S140A), which is unable to be phosphorylated. Using this model cell, we tested whether the fragmentation-dispersion of Golgi during mitosis is essential for its equal partitioning.
p47 mainly localizes to the nucleus in an interphase cell
The subcellular distribution of p97 and p47 was determined in NRK cells by immunofluorescence microscopy using PFA fixation. Several groups have previously reported different staining patterns for p97: the nucleus-dominant pattern (Peters et al., 1990; Meyer et al., 2000) and the cytoplasm-dominant pattern (Madeo et al., 1998). Our results showed that p97 existed in both cytoplasm and nucleus (Fig. 1 A, left). Muller et al. (1999) reported that the amounts of p97 in the nucleus and cytoplasm varied widely in different tissues. Hence, these different staining patterns might depend on the cell lines used.
In contrast to p97, p47 mainly localized to the nucleus with a small amount present in the cytoplasm (Fig. 1 A, right). This staining pattern was the same in HeLa and CHO cell lines (unpublished data). Using antibodies to p47 NH2- or COOH-terminal fragment also revealed nuclear-staining patterns (unpublished data). Because p47 forms a trimer of ∼150 kD (Kondo et al., 1997), it is difficult for it to be transported into the nucleus by simple diffusion. A sequence search for peptide motifs predicted two candidate NLSs in p47: (119)PPRKKSPN(116) and (172)RRRH(175) (Hicks and Raikhel, 1995). Hence, we induced a point mutation in each candidate sequence (K112T, 112K→T; R173T, 173R→T) and expressed the mutants in NRK cells (Fig. 1 B). The expressed p47 was detected by using antibodies to an HA tag in order to distinguish it from endogenous p47. p47wt was expressed in the nucleus (Fig. 1 B, top left), displaying a distribution similar to endogenous p47. When p47 was mutated in either or both of the two NLS motifs, mutated p47 was expressed mainly in the cytoplasm (Fig. 1 B). Together, p47 mainly localizes to the nucleus of interphase cells, and the two NLS motifs were found to be important for this nuclear localization. Since p47 is an essential cofactor of the p97 membrane fusion pathway (Kondo et al., 1997), its nuclear localization suggests a low activity of p97/p47-mediated membrane fusion in the cytoplasm of interphase cells.
Although the major population of p47 was localized to the nuclear region, a small amount of p47 was still observed in the cytoplasm (Fig. 1 A, right). The next question was whether the tiny amount of p47 in cytoplasm functions in Golgi membrane fusion in interphase cells. To address this, we microinjected anti-p47 antibodies into the cytoplasm of interphase cells and investigated Golgi morphology. When prophase cells were injected with the same antibody (8 g/l) and incubated for 1.5 h, Golgi reassembly at the end of mitosis was completely inhibited. Only clusters of vesicles and tubules, not stacks, were observed in daughter cells (Uchiyama et al., 2002). Since the amount of cytoplasmic p47 in interphase cells is much less than that in mitotic cells, this concentration (8 g/l) of the antibody must be sufficient to block the function of cytoplasmic p47. Fig. 1 C shows Golgi staining with anti-GM130 antibodies in the injected cells. Here, the Golgi apparatus was still located in perinuclear regions without dispersion (Fig. 1 C, second and fourth panels). Next, we investigated Golgi ultrastructure by EM. Interphase cells were injected with this antibody at 8 or 16 g/l and incubated for 2 h. As shown in Fig. 1 D (the injected antibody, 16 g/l), the Golgi apparatus still showed normal stacking structures without fragmentation (right). Thus, we could not find any obvious evidence of a Golgi maintenance function mediated by p97/p47. We concluded that the fusion activity mediated by p97/p47 is less critical during interphase than at the end of mitosis.
Once the cell enters mitosis, the nuclear envelope is broken down and p47 comes out from the nucleus, which enables p47 to function in Golgi membrane fusion. However, Golgi disassembly at early mitosis requires the inhibition of membrane fusion. This leads to the question: how is the p97/p47 pathway blocked during early mitosis?
p47 is mitotically phosphorylated
To investigate whether there was mitotic phosphorylation of p47, NRK cells were synchronized and incubated in medium containing [32P]orthophosphate. Mitotic cells were collected by mechanical shake-off and used for the preparation of 32P-labeled mitotic cytosol. Hoechst DNA staining showed that >95% of cells were mitotic (unpublished data). Nonsynchronized cells were used for the preparation of interphase cytosol. p47 was isolated from the cytosol by denatured immunoprecipitation (Fig. 2 A). Western blotting showed that the band of mitotic p47 was shifted upwards (Fig. 2 A, top), and autoradiography revealed that p47 was strongly phosphorylated at mitosis (Fig. 2 A, bottom).
We tried to determine the responsible kinase. At first, we investigated the effect of several protein kinase inhibitors on the mitotic phosphorylation of p47 by incubating His-p47 with mitotic cytosol and [32P]ATP in the presence of the indicated protein kinase inhibitor (Fig. 2 B). Staurosporine (broad Ser/threonine (Thr) kinase inhibitor) and olomoucine (Cdc2 inhibitor) inhibited p47 phosphorylation (Fig. 2 B, fifth and sixth lanes). But, neither PD98059 (MEK inhibitor), SB203580 (p38 MAPK inhibitor), KT5720 (protein kinase A inhibitor), nor calphostin C (PKC inhibitor) had any effect on p47 phosphorylation. Hence, Cdc2 was thought to be a candidate kinase. This was supported by the previous report that p47 can be a substrate of Cdc2 in vitro (Mayr et al., 1999). To confirm this, we investigated whether Cdc2-depleted mitotic cytosol phosphorylated p47 or not (Fig. 2 C). p13-suc1 beads were used for the depletion of Cdc2 from cytosol, and Western blotting with anti-Cdc2 antibodies showed that >90% of Cdc2 was depleted (unpublished data). p47 phosphorylation was inhibited by Cdc2 depletion from mitotic cytosol (Fig. 2 C, lane 3), and this inhibition was rescued by supplementation with Cdc2 (Fig. 2 C, lane 4). In addition, the purified Cdc2 complex phosphorylated p47 in the absence of mitotic cytosol as shown in Fig. 2 D. Therefore, Cdc2 was confirmed as a responsible kinase.
The consensus motif for Cdc2 is Ser/Thr-Pro-X–Arg/Lys, with Pro at the +1 position being absolutely critical, and a basic residue at the +3 position preferred but not essential for kinase recognition (Holmes and Solomon, 1996). p47 has one Thr and three Ser residues with Pro at the +1 position: T57, S114, S140, and S272. His-p47 was mutated at each phosphorylation site and tested for phosphorylation. As presented in Fig. 2 E, all p47 mutants except p47(S140A) were phosphorylated by mitotic cytosol. Similar results were obtained by using purified Cdc2 instead of mitotic cytosol (unpublished data). Together, we concluded that p47 was mitotically phosphorylated on Ser-140 by Cdc2.
p47 forms a complex with p97 at mitosis
After determining the phosphorylation of p47 at mitosis, this led to the question of what effect the phosphorylation would have on the function of p47. p47 forms a tight complex with p97, which is essential for in vitro Golgi reassembly (Kondo et al., 1997). We investigated the effect of phosphorylation on the binding of p47 to p97. To test whether p97 binds phosphorylated p47, p97 and p47 were incubated in the presence of Cdc2 and [γ-32P]ATP (Fig. 3 A). p47 bound to p97 was coimmunoprecipitated with antibodies to the His tag on p97. Phosphorylated p47 was detected by autoradiography. Phosphorylated p47 was coimmunoprecipitated together with p97 (Fig. 3 A, lane 2), suggesting that phosphorylated p47 bound to p97. p97 is also reported to be phosphorylated at mitosis (Madeo et al., 1998), and therefore, we tested whether phosphorylated p97 bound to p47 (Fig. 3 B). Synchronized NRK cells were labeled with [32P]orthophosphate and used for the preparation of mitotic cytosol. p47 and its binding proteins were coimmunoprecipitated from the 32P-labeled mitotic cytosol using polyclonal anti-p47 antibodies and protein A beads. p47-binding proteins were eluted from the beads by 1 M KCl. p97 was isolated from the eluate by denatured immunoprecipitation using polyclonal anti-p97 antibodies, and phosphorylated p97 was detected by antoradiography. As shown in Fig. 3 B, lane 2, phosphorylated p97 bound to p47. We next compared p47 distributions in interphase and mitotic cell extracts. Extracts were size fractionated by gel filtration, and the localization of p97 and p47 were studied (Fig. 3 C). In both the interphase and mitotic extracts, p97 eluted at ∼670 kD, and p47 eluted at ∼670 and ∼200 kD. (Note that the bands of p47 were double in the mitotic extract.) Since p97-associated p47 is eluted at ∼670 kD (Kondo et al., 1997), these results indicate that p47 forms a complex with p97 in the mitotic extract as well as in the interphase extract. Thus, all the data from the binding experiments and gel filtration show that p47 phosphorylation has no effect on complex formation with p97.
Phosphorylation of Ser-140 in p47 inhibits its binding to Golgi membranes
It has been reported that p47 mediates the binding of the p97/p47 complex to Golgi membranes (Rabouille et al., 1998). Therefore, we investigated the effect of the phosphorylation on the binding of p47 to Golgi membranes. Purified rat liver Golgi membranes were incubated with mitotic HeLa cytosol to obtain mitotic Golgi fragments. Golgi membranes were incubated with interphase cytosol as a control. Fig. 4 A shows the amounts of p97, p47, and syntaxin5 in both membranes by immunoblotting. Mitotic Golgi fragments had very little p97 and p47 compared with those in the membranes incubated with interphase cytosol (Fig. 4 A, top and middle), although the amounts of the membrane protein syntaxin5 were almost the same in both membranes (Fig. 4 A, bottom).
In Fig. 4 B, staurosporine and microcystin-LR were added to mitotic Golgi fragments to inactivate any remaining kinases and phosphatases. The resulting membranes were salt washed and incubated with p47 in the presence of staurosporine and microcystin-LR. The upward shifts of GM130 bands (Fig. 4 B, top, lanes 3 and 4) showed that the membranes remained “frozen” in a mitotic state. The frozen mitotic Golgi fragments could bind p47 (Fig. 4 B, middle, lane 4) as well as the control Golgi membranes (interphase Golgi membranes; Fig. 4 B, middle, lane 2). This indicates that the mitotic Golgi fragments still have the ability to bind p47. Accordingly, it is thought that the phosphorylation of p47 decreases its binding affinity to the membranes. To confirm this, we next performed several binding experiments using p47 and salt-washed Golgi membranes.
We first investigated whether phosphorylated p47 bound to Golgi membranes. p47 was phosphorylated by the incubation with Cdc2 and then its phosphorylation condition was frozen by the addition of staurosporine and microcystin-LR. The phosphorylated p47 was used for binding experiments with salt-washed Golgi membranes. As presented in Fig. 4 C, Cdc2-mediated p47 phosphorylation inhibited its binding to the membranes (Fig. 4 C, middle, lane 3). This inhibition was rescued by olomoucine, an inhibitor of Cdc2 (Fig. 4 C, middle, lane 4). These inhibition and rescue were also observed when a p97/p47 complex was added instead of p47 (unpublished data). Next, the mutated p47, p47(S140A), which cannot be phosphorylated in Ser-140, was used for the binding experiments (Fig. 4 D). p47(S140A) bound to Golgi membranes (Fig. 4 D, lane2), and Cdc2 had no effect on this binding (Fig. 4 D, lane 3). We finally tested another p47 mutant, p47(S140D), which mimics the phosphorylation of Ser-140, for the binding experiments (Fig. 4 E). Much less p47(S140D) bound to Golgi membranes (Fig. 4 E, top, lanes 4 and 5) when compared with p47wt (Fig. 4 E, top, lanes 2 and 3) and p47(S140A) (Fig. 4 E, top, lanes 6 and 7). In summary, our biochemical data showed that phosphorylation of Ser-140 in p47 inhibited its binding to Golgi membranes.
We also further investigated the localization of p47 in mitotic cells (Fig. 5). As shown in Fig. 1 A, when NRK cells were directly fixed using PFA, anti-p47 antibodies displayed very strong nuclear staining and weak cytoplasmic staining. However, when such cells were washed with 0.01% saponin-containing buffer to remove p47 that was not membrane-bound before fixation, perinuclear staining was prominent in addition to nuclear staining (Fig. 5 A, arrows). In interphase cells as shown in Fig. 5 A (the star-marked cells), the staining of p47 showed clear colocalization with the staining of GM130, a Golgi marker, at perinuclear regions. In early prometaphase cells (the top left cell in Fig. 5 A and its magnified images shown in Fig. 5 B), p47 staining did not colocalize with GM130 staining. These results strongly support the biochemical data that mitotically phosphorylated p47 does not bind to Golgi membranes. Fig. 5 C presents the localization of p47 in the cell at early telophase. Once the cell entered telophase, p47 staining again showed colocalization with GM130 (Fig. 5 C, arrowheads). It has been reported that GM130 is phosphorylated by Cdc2 at prophase and dephosphorylated at early telophase (Lowe et al., 2000). Similarly, it is likely that p47 was dephosphorylated at early telophase and rebound to Golgi membranes.
Phosphorylation of Ser-140 in p47 controls p97-mediated membrane fusion in vitro
We have established that p47 is mitotically phosphorylated at Ser-140 and that phosphorylated p47 is dissociated from Golgi membranes. Next, we wanted to ascertain whether this phosphorylation in p47 has any effect on Golgi disassembly and assembly in vitro.
If p47 is not phosphorylated at early mitosis, are Golgi membranes fragmented? We used mutated p47, p47(S140A), which cannot be phosphorylated on Ser-140. The early biochemical binding experiment had shown that p47(S140A) could bind to Golgi membranes in the presence of Cdc2 (Fig. 4 D). Purified Golgi membranes and mitotic cytosol were incubated with p97/p47wt or p97/p47(S140A) for the in vitro Golgi disassembly assay. Cisternal fragmentation was defined by the loss of cisternal membranes, with fragmentation induced by mitotic cytosol set to 100%. As shown in Fig. 6 A, the addition of p97/p47(S140A) inhibited cisternal fragmentation almost completely, whereas the addition of p97/p47wt had no effect. This indicates that the phosphorylation of Ser-140 in p47 is required for Golgi disassembly.
We also tested the effect of p47(S140D), which mimics Ser-140 phosphorylation, on Golgi reassembly. Mitotic Golgi fragments were incubated with p97/p47wt or p97/p47(S140D) for the in vitro Golgi reassembly assay. Fig. 6 B shows that p97/p47(S140D) had a very small effect on the cisternal regrowth compared with p97/p47wt, suggesting that the dephosphorylation of Ser-140 in p47 is necessary for Golgi reassembly.
Phosphorylation of Ser-140 in p47 is required for Golgi fragmentation-dispersion during mitosis in living cells
The in vitro Golgi disassembly assay showed that the addition of p97/p47(S140A) to mitotic cytosol inhibited Golgi fragmentation (Fig. 6 A). We next investigated the effect of p47(S140A) in living cells. Either p97/p47wt or p97/p47(S140A) was microinjected into prophase (or early prometaphase) cells and fixed at several intervals. Fig. 7 A shows Golgi staining with anti-GM130 antibodies after injection. In the cells injected with p97/p47wt, the Golgi membranes dispersed all over the cell at prometaphase and metaphase, although there remained some small blobs, and they were reorganized to form Golgi clusters at telophase (Fig. 7 A, top). The cells injected with p97/p47(S140A) showed entirely different Golgi staining: the dispersion of Golgi membranes was mostly suppressed, and Golgi clusters showing bright staining were observed through all mitotic phases (Fig. 7 A, bottom). The Golgi clusters induced by p97/p47(S140A) were doubly stained with antibodies to GM130 and β1,4-galactosyltransferase (GalT), markers of cis- and trans-Golgi, respectively (Rabouille et al., 1995a). As shown in Fig. 7 B, some parts of the structures were stained by only one antibody, although others were stained by both. This suggests that these Golgi clusters were polar structures.
The Golgi apparatus generally vesiculates and disperses all over the cytoplasm from prometaphase to early telophase. We thus quantified the total cellular fluorescence of GM130-containing structures in the injected cells. As presented in Fig. 7 C, in the cells injected with p97/p47wt, the amount of GM130 in identifiable fragments fell to ∼24% at metaphase. In the p97/p47(S140A)-injected cells, the reduction was very small; it fell to only ∼80% at metaphase. Zaal et al. (1999) suggested that the absorption of Golgi vesicles to ER might be related to this reduction, although this is a matter of controversy (Jokitalo et al., 2001). Accordingly, our results suggest that p47(S140A) inhibited the vesiculation-dispersion of Golgi (and its absorption to ER, if this absorption exists).
The double staining images presented in Fig. 7 B suggest that the Golgi clusters induced by p97/p47(S140A) may be organized polar structures. Therefore, we investigated the ultrastructure of injected cells by EM. Fig. 8 A shows representative images of the injected cells. In the cells injected with p97/p47wt, Golgi stacks were almost lost and fragmented at late prometaphase, similar to uninjected cells (Fig. 8, left). In contrast, the cells injected with p97/p47(S140A) had highly organized Golgi stacks at the same mitotic phase: 80.1% of cisternal membranes were found in stacks (Fig. 8, right). The connections between stacks were lost, and the Golgi ribbon was broken into discrete stacks. Note that the nuclear envelope had been lost and chromatin was well condensed in this cell (Fig. 8, right, inset). Fig. 8 B presents the quantitative results. The injection of p97/p47(S140A) significantly increased the percentage of Golgi membranes in cisternal structures threefold and significantly decreased the percentage of Golgi membranes in vesicles by a third compared with the injection of p97/p47wt. The amount of tubules/tubular networks did not change. Neither Golgi stacks nor cisternae are generally observed in mitotic cells, whereas the injection of p97/p47wt also allowed a few cisternae to remain at early mitosis (Fig. 8 B). Some of the injected p47wt might not be phosphorylated, which enabled occasional cisternae to remain.
The fragmentation-dispersion of Golgi is generally believed to be necessary for its equal partitioning into two daughter animal cells (Warren, 1993). Now, we have succeeded in preventing Golgi fragmentation-dispersion at mitosis. This suggests that the equal partitioning of Golgi should not be achieved in the cells injected with p97/p47(S140A). Therefore, we tested the partitioning of Golgi in the injected cells. Because it is reported that GM130 remains in Golgi fragments at mitosis without being absorbed to ER (Seemann et al., 2002), the staining of GM130 was used to trace Golgi membranes. To estimate the accuracy of the partitioning, we calculated median deviations from equality. Uninjected cells were used to calculate the control value. Table I presents the results. There were no significant differences among the three groups. The cells injected with p97/p47(S140A) showed 3.60%, which was not bigger than those in the uninjected cells and p97/p47wt-injected cells. This indicates that the p97/p47(S140A)-injected cells achieved equal partitioning of Golgi at the same accuracy as the control cells.
In animal cells, the Golgi apparatus is disassembled at the onset of mitosis and reassembled at the end of mitosis. Golgi disassembly requires the inhibition of membrane fusion, which is necessary for reassembly (Warren, 1993). Two ATPases are required for this membrane fusion: NSF and p97 (Rabouille et al., 1995b). The mitotic inhibition of the NSF pathway for Golgi disassembly is achieved by blocking the tethering of p115-GM130 (Nakamura et al., 1997), and its inhibition is rescued by dephosphorylation of GM130 at telophase (Lowe et al., 2000). In this paper, we have clarified the mechanism of the cell cycle–dependent regulation of the other membrane fusion pathway, the p97/p47 pathway. p47 is mainly localized to the nucleus during interphase and moves to the cytoplasm when cells enter mitosis. p47 is phosphorylated by Cdc2 at early mitosis, and this phosphorylation prevents the binding of the p97/p47 complex to Golgi membranes, resulting in the inhibition of p97-mediated membrane fusion. Thus, the fusion machinery itself is mitotically blocked in the case of the p97 pathway, in contrast to the NSF pathway; the function of NSF/SNAPs, the SNARE-priming activity, is intact through mitosis (Muller et al., 2002).
Phosphorylation of Ser-140 in p47 inhibited its binding to Golgi membranes (Fig. 4). The Ser-140 is in a highly unstructured region (Yuan. X., and S. Matthews, personal communications) which could become structured upon binding to SNAREs. The phosphorylation could alter the nature of the unstructured loop region, affecting the conformational changes. Also, the additional negative charges on phosphorylated Ser-140 could alter the local conformation of the structure, masking the interaction site with SNAREs. Our data show that p47 binds to Golgi membranes at interphase, comes off at prometaphase, and rebinds to the membranes at telophase (Fig. 5). These results indicate that p47 is phosphorylated at early mitosis and dephosphorylated at late mitosis, although we have no data on the responsible phosphatase at present. This is consistent with the results of the in vitro function assays which showed that Golgi disassembly requires the phosphorylation of p47 and that Golgi reassembly requires the dephosphorylation of p47 (Fig. 6). Therefore, we propose that the phosphorylation-dephosphorylation cycle of p47 at mitosis as well as its nuclear localization at interphase play key roles in Golgi disassembly–reassembly during the cell cycle. Fig. 9 shows a schematic model. In an interphase cell, p47 is mainly in the nucleus. Once the cell enters mitosis, its nuclear envelope is broken down, and p47 enters the cytoplasm. At the same time, p47 is phosphorylated by mitotically activated Cdc2 kinase. Since the phosphorylated p47 cannot bind to Golgi membranes, p97/p47-mediated membrane fusion is inhibited, resulting in the fragmentation of Golgi membranes. At late mitosis, p47 is dephosphorylated and binds to Golgi membranes. This allows p97/p47-mediated membrane fusion to reassemble the Golgi apparatus. When nuclear envelopes are formed in daughter cells, p47 moves to the nucleus, and the membrane fusion mediated by p97/p47 is suppressed. Hetzer et al. (2001) has reported that p97/p47 is also required for nuclear envelope formation. This implies that, when the nuclear envelope as well as the Golgi apparatus are reassembled and p97/p47-mediated fusion is not required anymore, membrane fusion will be suppressed by the nuclear envelope that p97/p47-mediated fusion assembled. This feedback control system seems to be very tight. Moreover, Ser-140 in p47 is conserved in Drosophila, Xenopus, mouse, and human but not in yeast. In yeast, the nuclear envelope is not broken down at mitosis, and hence, the phosphorylation-dephosphorylation of the yeast p47 homologue shp1p may not be necessary.
Microinjection of p47(S140A), which cannot be phosphorylated, inhibits mitotic Golgi disassembly in living cells (Figs. 7 and 8). This is the first report showing that Golgi stacks can be successfully maintained during mitosis in mammalian cells. In p97/p47(S140A)-injected cells, the Golgi ribbon was broken into discrete stacks, but no further disassembly occurred (Fig. 8). This supports the general view that the disassembly occurs in two phases: first the ribbon is broken into stacks, and then each stack is converted into clusters of vesicles and tubules. p47(S140A) seems to have its effect on the second phase of disassembly. The function of p97/p47 is thought to be mediated by its membrane fusion activities, and there is no previous evidence concerning its function on Golgi stacking (Rabouille et al., 1995c; Kondo et al., 1997; Shorter and Warren, 1999). We observed no obvious effects of p97/p47(S140A) on the stacking in the in vitro assay (unpublished data). Therefore, it was a surprise that Golgi stacks remained in mitotic cells after injection of p97/p47(S140A). Nevertheless, this result is in good agreement with our previous observation in vivo that the injection of anti-p47 antibodies inhibited stacking formation as well as cisternal growth at the end of mitosis (Uchiyama et al., 2002). The discrepancy between in vitro and in vivo might be caused by lack of cytoskeletons in the in vitro system: cytoskeletal elements such as actin filaments were removed or depolymerized in the in vitro assay. The fragmentation of cisternae might be necessary for the effective destruction of Golgi stacks, because big cisternae are thought to prevent the access of mitotic modulators, such as kinases, to the stacking machinery between them. Fig. 8 B also shows the decrease of the amount of Golgi vesicles by the injection of p97/p47(S140A). In a previous paper, we showed that the in vivo inhibition of the p97/p47 pathway caused an accumulation of tubules, as opposed to vesicles, suggesting a function for p97/p47 in homotypic membrane fusion (Uchiyama et al., 2002). Accordingly, the vesicles might not be consumed by p97/p47(S140A)-mediated membrane fusion. The existence of big smooth cisternae at mitosis might result in the decrease of the regions for vesicle budding, and thus, it is most likely that budding of vesicles from Golgi was suppressed in p97/p47(S140A)-injected cells.
Normal prometaphase NRK cells had neither Golgi stacks nor cisternae, and vesiculated Golgi membranes were dispersed all over the cytoplasm. In contrast, the p97/p47(S140A)-injected prometaphase cell had Golgi cisternae and stacks (Fig. 8) and seemed to have very little absorption of Golgi vesicles into ER (Fig. 7 C). The injection of p97/p47(S140A) had no effect on cell cycle progression (e.g., chromatin segregation and cytokinesis were pursued normally; see Table S1). Hence, we used these p97/p47(S140A)-injected cells as model cells and tested whether Golgi fragmentation-dispersion is required for equal partitioning in mammalian cells. As a result, we found that the p97/p47(S140A)-injected cells achieved equal Golgi partitioning. Therefore, we propose that Golgi fragmentation-dispersion is not obligatory for its equal partitioning even in mammalian cells. Generally, organelles adopt one of two partitioning strategies: stochastic or ordered. Our result strongly suggests that the partitioning of the Golgi is controlled in a highly ordered way. A separate stack can be a minimum unit for partitioning even in mammalian cells. This is similar to the cases of algae and protozoa (Munro, 2002). Stochastic and ordered ways of partitioning need not be mutually exclusive and our results do not belittle the significance of mitotic Golgi fragmentation-dispersion. This might assist the ordered mechanism to increase the accuracy of partitioning. Sutterlin et al. (2002) also reported that Golgi fragmentation-dispersion is required for entry into mitosis, showing that this process is essential for mammalian cells. Taking our results into account, the fragmentation-dispersion of Golgi may have an important, presently unknown, significance besides partitioning.
Materials And Methods
Proteins, constructs, antibodies, and reagents
His-tagged p97 and p47 were prepared as described previously (Kondo et al., 1997; Uchiyama et al., 2002). Endogenous p97 was purified from rat liver cytosol (Rabouille et al., 1995c; Kondo et al., 1997). Recombinant p47 without a tag was expressed in Escherichia coli and purified biochemically.
For the expression of HA-tagged p47 in cultured cells, p47 cDNA was subcloned to pCG-N-BL vector (a gift from Dr. Taketani, Kyoto Institute of Technology, Kyoto, Japan). Point mutations were directly introduced into the p47 cDNA in pQE30 or pCG-N-BL by PCR reactions, using the Quick-change mutagenesis kit (Stratagene). All clones were verified by DNA sequencing.
Polyclonal antibodies to p47 and syntaxin5 were prepared as described (Hui et al., 1997; Kondo et al., 1997). A rabbit anti-p97 polyclonal antibody was raised using His-p97(1–198) as an antigen. Monoclonal anti-GalT antibodies and polyclonal anti-GM130 antibodies were gifts from Dr. Suganuma (Miyazaki Medical College, Miyazaki, Japan) and Dr. Nakamura (Kanazawa University, Kanazawa, Japan), respectively. mAb to p97, His-tag, and GM130 were purchased from Progen, QIAGEN, and BD Transduction, respectively.
The following reagents were purchased from Calbiochem; staurosporine, olomoucine, PD98059, SB203580, KT5720, calphostin C, microcystin-LR. Cdc2 kinase (p34cdc2/cyclin B) was from New England Biolabs, Inc.
In vivo metabolic 32P labeling
For enrichment of mitotic NRK cells, aphidicolin (2.5 μg/ml) was added to the medium for 14 h. The cells were washed with fresh medium, released from the S phase block for 2 h, and labeled with [32P]orthophosphate (200 μCi/ml) for another 4 h at 37°C. Mitotic cells were flushed from the dish, washed with PBS, and extracted with buffer (20 mM Hepes, 0.1 M KCl, 5 mM EDTA, 2 mM EGTA, 40 mM β-glycerophosphate, 10 mM NaF, 1 mM sodium orthovanadate, 1 mM DTT, protease inhibitor cocktail, 0.5% Triton X-100, 10% glycerol, pH 7.4). The lysate was cleared by centrifugation and used for the native or denatured immunoprecipitation with anti-p47 antibodies.
In vitro phosphorylation
p47 or its mutant was incubated in buffer A (50 mM Tris, 50 mM KOAc, 10 mM MgOAc, 20 mM β-glycerophosphate, 0.2 mM DTT, 40 μM ATP, 30 μCi/μl [γ-32P]ATP, pH 7.4) with mitotic cytosol or purified kinase for 30 min at 30°C followed by denatured immunoprecipitation. The reactions were terminated by adding an equal volume of the buffer (100 mM Tris, 2 mM EDTA, 2% SDS, pH 7.4) and boiled for 4 min. After adding 20 vol of buffer (50 mM Tris, 0.15 M KCl, 0.5% Tween-20, 1 mM EDTA, 1 mM EGTA, 10 mM NaF, 40 mM β-glycerophosphate, 1 mM sodium orthovanadate, 1 mM DTT, pH 7.4), p47 or its mutant was immunoprecipitated with antibodies to a His tag and protein G beads followed by SDS-PAGE and autoradiography.
For the binding experiments of p47 to salt-washed Golgi membranes, 1 mM ATP was added to buffer A instead of 40 μM ATP and 30 μCi/μl [γ-32P]ATP. After incubation, the reactions were supplemented with microcystin-LR (10 μM) and staurosporine (25 μM) followed by the binding experiments using Golgi membranes.
Depletion of Cdc2 kinase was performed using p13-suc1 beads (Upstate Biotechnology) as described (Hirota et al., 2000).
Binding experiments and gel filtration
p47 or its mutant was incubated with salt-washed Golgi membranes in buffer (0.1 M KCl, 20 mM Tris, 1 mM MgCl2, 1 mM ATP, 0.4 g/l BSA, 0.2 M sucrose, pH 7.4) for 1 h on ice, and then the membranes were recovered by centrifugation.
Interphase and mitotic cytosol were prepared from HeLa cells (Rabouille et al., 1995b). The homogenizing buffer was supplemented with microcystin-LR (10 μM) and staurosporine (25 μM) in order to maintain their phosphorylation conditions. Cytosolic proteins were fractionated by a Superose 6 in the presence of microcystin-LR (2 μM) and staurosporine (5 μM). Molecular weight markers were thyroglobulin (1,340 and 670 k), apoferritin (440 k), and β-amylase (200 k).
In vitro Golgi disassembly and reassembly assays
NRK cells grown on coverslips were fixed with 3% PFA/PBS for 20 min at room temperature and stained with antibodies after permeabilization, unless stated otherwise. 0.1% Triton X-100 and 0.1% saponin were used for the permeabilization of interphase and mitotic cells, respectively.
For the staining of mitotic cells with antibodies to p47 and GM130, cells were incubated in buffer (80 mM PIPES, 1 mM MgCl2, 5 mM EGTA, 0.01% saponin, pH 7) for 7 min at room temperature before the fixation with 3% PFA (Zerial et al., 1992). Saponin (0.01%) was used instead of Triton X-100 for the following permeabilization and labeling with antibodies.
The total cellular fluorescence of GM130-containing structures was calculated using a series of sections (1 μm in thickness) throughout a whole cell: 9–11 sections for an interphase cell and 17–19 sections for a mitotic cell. For the calculation, well-isolated cells were chosen. Background was subtracted and a threshold introduced to exclude diffuse cytoplasmic staining. All pixels above this threshold value were counted as GM130-containing structures in all sections and summed up. The resulting value was normalized to a uninjected interphase cells on the same coverslip.
Microinjection into living cells
The microinjection was performed as described previously (Uchiyama et al., 2002). Fluorescence of coinjected Cy3-BSA (∼1 μg/μl) was used as an injection marker to identify injected cells. For the injection into mitotic cells, the synchronized cells were used as described above. Prophase (or early prometaphase) cells were microinjected with p97/p47 samples 4 h after the release from aphidicolin block: a phase–contrast microscope was used for the identification of their phases. The injected cells were incubated for another 0.5–1.5 h and fixed.
For EM study, cells were grown on a coverslip on which a square area of ∼1 mm × ∼1 mm was outlined by a diamond pen (Uchiyama et al., 2002). The cells within the area were injected. In case of the injection into mitotic cells, all injections were done within 5–10 min. Uninjected cells within the area were completely removed by an injection needle using fluorescence of coinjected Cy3-BSA as a marker. After fixation and embedding into Epon, the cells within the area were ultrathin sectioned. Membrane profiles in the Golgi area were divided into three categories: cisternae, tubules, and vesicles. The relative proportion of each category of membranes was counted using an intersection method.
Online supplemental material
Table S1, showing the cell cycle progression after microinjection, is available.
We would like to thank M. Lowe for his kind and helpful advice, X. Yuan and S. Matthews for stimulating discussions; C. Rabouille for critically reading the manuscript; and K. Harasaki for his kind assistance in the preparation of the manuscript. Special thanks to P. Luzio and K. Maruyama for their kind support and encouragement.
E. Jokitalo is supported by Biocentrum Helsinki. This work was supported by a Wellcome Trust grant to H. Kondo.
The online version of this article includes supplemental material.
Abbreviations used in this paper: GalT, β1,4-galactosyltransferase; Ser, serine; Thr, threonine.