Binding of the vesicle docking protein p115 to Golgi membranes is inhibited under mitotic conditions.

The vesicle docking protein p115 showed saturable, high affinity binding to interphase Golgi membranes. The affinity of binding was up to 20-fold lower using membranes preincubated with mitotic cytosol. In contrast, binding was not affected by mitotic pretreatment of p115. The reduction in p115 binding was mediated by phosphorylation, could be induced by a cyclin-dependent kinase, and was fully reversible. A shift of p115 from membranes to cytosol was also found after fractionating mitotic cells. The functional significance of the decreased binding was addressed by in vitro mitotic incubations which disassemble Golgi cisternae, predominantly producing transport vesicles. The addition of excess p115 decreased loss of membrane from cisternae, indicating that p115's action is limiting while transport vesicles accumulate. The cessation of intra-Golgi traffic in mitosis has been hypothesized to result from an inhibition of membrane fusion while budding of transport vesicles continues. This process also contributes to mitotic Golgi disassembly. Our results imply that there is a mitotic modification to Golgi membranes leading to a reduction in the affinity of the p115 receptor. Reduced p115 binding may play a part in the inhibition of membrane fusion by preventing prior vesicle docking.

The vesicle docking protein p115 showed saturable, high affinity binding to interphase Golgi membranes. The affinity of binding was up to 20-fold lower using membranes preincubated with mitotic cytosol. In contrast, binding was not affected by mitotic pretreatment of p115. The reduction in p115 binding was mediated by phosphorylation, could be induced by a cyclin-dependent kinase, and was fully reversible. A shift of p115 from membranes to cytosol was also found after fractionating mitotic cells. The functional significance of the decreased binding was addressed by in vitro mitotic incubations which disassemble Golgi cisternae, predominantly producing transport vesicles. The addition of excess p115 decreased loss of membrane from cisternae, indicating that p115's action is limiting while transport vesicles accumulate. The cessation of intra-Golgi traffic in mitosis has been hypothesized to result from an inhibition of membrane fusion while budding of transport vesicles continues. This process also contributes to mitotic Golgi disassembly. Our results imply that there is a mitotic modification to Golgi membranes leading to a reduction in the affinity of the p115 receptor. Reduced p115 binding may play a part in the inhibition of membrane fusion by preventing prior vesicle docking.
The exocytic pathway consists of a series of discrete organelles, traffic between which is mediated by vesicular carriers (1). The Golgi apparatus lies at the heart of the exocytic pathway, and consists of a stack of closely apposed cisternae, together with the cis-and trans-Golgi networks on either face of the stack (2). Molecules on the exocytic pathway are transported through the sequence of compartments that comprise the Golgi apparatus by vesicles which bud from one compartment and fuse with the next (3).
One approach to understanding Golgi structure and function has been to study the disassembly and reassembly of the Golgi apparatus which occur during division of animal cells (4). During mitosis, the single copy organelle is disassembled into hundreds of clusters, which become dispersed throughout the cytoplasm by metaphase (5,6), and shed vesicles into the cytoplasm (7). These fragments can then be partitioned stochastically between the two daughter cells (8). The majority of mitotic Golgi fragments, accounting for two-thirds of starting membranes, are uniform small vesicles (6,9,10), produced by a coat protomer (COP) 1 I-dependent mechanism (11,12), which during interphase is responsible for both anterograde (3) and retrograde (13) transport through the Golgi stack. There is also a COP I-independent pathway, accounting for the remaining third of membrane, which produces larger, more heterogeneous fragments (12). It has been hypothesized that the accumulation of Golgi fragments during mitosis is related to a mitotic inhibition of membrane fusion which also underlies the widespread mitotic cessation of intracellular transport (4,14).
The mitotic target(s) which mediate the inferred inhibition of vesicle fusion have not to date been identified, even though many of the components of the fusion machinery are now known. Vesicles specifically recognize their targets by an interaction between membrane proteins named v-and t-SNAREs (vesicle and target SNAP receptors), to produce a docked vesicle (15,16). Docking is followed by binding of ␣/␥SNAPs (soluble NSF attachment proteins) and NSF (N-ethylmaleimidesensitive factor), which bring about membrane fusion (17,18). The events which control vesicle docking might include the interactions of SNAREs with other proteins, for example, Rab proteins (19) and homologues of Sec1p (20) and of synaptophysin (21). In particular, the activity of v-SNAREs may be governed by Rabs (22,23), which have been implicated as regulators of vesicle fusion at many steps of intracellular traffic (24,25).
Proteins which have been studied in mitosis to date include a number of Rab proteins (26). For three Rabs found in the Golgi apparatus, only one (rab1p) is phosphorylated in mitosis, and as a result it becomes more membrane-bound. By comparison, neither change was shown by the other two (Rab2p and Rab6p) (26,27). A different pattern again is shown by Rab4p, which promotes recycling from early endosomes; it is phosphorylated in mitosis and as a result accumulates in the cytoplasm (26,28). Given these results, and since Rab proteins do not contain a common phosphorylation site, it seems unlikely that the widespread cessation of traffic is brought about by direct modification to Rabs.
More probable mitotic targets a priori are the group of cytoplasmic proteins which act at all steps of fusion mediated by SNAREs. Apart from NSF and SNAPs, four other cytoplasmic factors have been identified from their activity in an in vitro intra-Golgi transport assay, including p115, which has been purified to homogeneity and cloned (29 -31), and Rab-guanine nucleotide dissociation inhibitor (Rab-GDI) (32), which can sequester many different Rabs in an inactive, cytoplasmic form (32,33). Rab-GDI is activated by phosphorylation, although whether this occurs in a cell cycle-dependent manner is not known (34). p115 has been implicated in three intracellular traffic steps: 1) intra-Golgi traffic (29); 2) endoplasmic reticulum to Golgi traffic, by inference from the action of p115's Saccharomyces cerevisiae homologue, Uso1p (35,36); and 3) transcytosis, from which p115 is also known as transcytosis-associated protein (37,38). The function of p115 has been shown to be in a docking step prior to membrane fusion (31,36). Since p115's action is widespread, we have investigated its role in mitosis. We have found that p115 binding to Golgi membranes is inhibited under mitotic conditions as a result of a modification to the membranes, not to p115. Furthermore, an early stage of mitotic Golgi disassembly in vitro was inhibited by increasing p115 receptor occupancy with excess, purified p115. Together these data suggest that the mitotic inhibition of p115 binding may contribute to the inhibition of membrane fusion by acting at the earlier step of vesicle docking.

EXPERIMENTAL PROCEDURES
Materials-All reagents were of analytical grade or higher and were purchased from Sigma or BDH (Dorset, UK) unless stated otherwise. Stock solutions of olomoucine (100 mM in Me 2 SO), staurosporine (10 mM in Me 2 SO, Calbiochem, Nottingham, UK), and microcystin-LR (1 mM in 10% methanol, Calbiochem) were stored at Ϫ20°C. Bacterially expressed bovine cyclin A tagged with His 10 and protein A sequences (39) was purified by nickel-nitrilotriacetic acid affinity chromatography. His-tagged NSF, ␣-SNAP, and ␥-SNAP were expressed in bacterial cells and similarly purified. Protein concentrations were determined by the method of Bradford (40).
Antibodies-The murine monoclonal antibody to p115, designated 8A6 or anti-p115-2 (29), was affinity-purified on protein G. Polyclonal rabbit antibody to mannosidase II was raised against the chymotryptic fragment of the purified protein (41).
Preparation and Treatment of Golgi Membranes-Stacked rat liver Golgi membranes were prepared at 2 mg/ml protein (42). The purification of ␤-1,4-galactosyltransferase over total protein was typically 100-fold.
To salt-wash membranes, 75 l of Golgi membranes were mixed with 1 ml of 500 mM potassium chloride in buffer A (10 mM magnesium acetate, 15 mM EGTA, 20 mM ␤-glycerophosphate, 50 mM potassium acetate, 200 mM sucrose, 2 mM ATP, 1 mM dithiothreitol (Boehringer Mannheim, East Sussex, UK), pH 7.5, at 4°C). After 30 min at 4°C, the membranes were layered over 0.5 ml of cushion of the same buffer, containing 300 mM sucrose, and centrifuged at 50,000 rpm for 10 min (TLA-100.3, Beckman). Membranes were resuspended in 75 l of buffer A and stored at Ϫ70°C. Alkali washing was carried out in a similar fashion using 0.1 M sodium carbonate (pH 11.5) and omitting a higher density cushion. For chymotrypsin treatment, salt-washed Golgi membranes (5 l, derived from 10 g of protein) were treated with chymotrypsin (3 g, type VII) in a final volume of 35 l at 4°C for 30 min. Phenylmethylsulfonyl fluoride was added to a final concentration of 400 g/ml for a further 10 min. For mock treatment, the inhibitor was added to the chymotrypsin first.
Preparation of Mitochondrial Membranes-2 g of rat liver homogenized in 10 ml of 0.5 M sucrose, 0.1 M potassium phosphate (pH 6.7) was centrifuged at 1000 ϫ g for 10 min. The supernatant was centrifuged for a further 10 min at 10,000 ϫ g. The resulting pellet was resuspended in 0.5 M sucrose, layered over steps of 0.86 M and 1.2 M sucrose, and centrifuged at 28,000 rpm for 1 h (SW40, Beckman). Pelleted membranes were more than 90% mitochondria, as assessed by electron microscopy.
Purification of p115-Rat liver was homogenized in 0.5 M sucrose, 0.1 M potassium phosphate (pH 6.7), and human placenta was homogenized in 140 mM sucrose, 20 mM MES (pH 6.6), 70 mM potassium acetate, 1 mM EGTA, 0.5 mM magnesium acetate, and 1 mM dithiothreitol. After centrifugation to remove particulates, cytosols were snap frozen in aliquots. Cytosol was diluted to 10 mg of protein/ml, and a 40% ammonium sulfate cut was taken. Precipitated protein was resuspended in buffer Q0 (25 mM Tris/chloride, pH 7.3, 1 mM dithiothreitol), to achieve a conductivity equal to that of Q200 (ϭ Q0 with 200 mM potassium chloride), clarified by centrifugation at 4000 ϫ g for 10 min, loaded on to fast-flow Q Sepharose, and eluted with a gradient of Q200 -Q650. p115-containing fractions were diluted with Q0, loaded onto high performance Q Sepharose (Pharmacia, St. Albans, UK), and eluted with Q200 -Q570. The peak of p115 was then fractionated by velocity sedimentation on sucrose gradients (6 -28% w/w, in Q200), by centrifuga-tion for 1 h at 65,000 rpm in a VTi65.1 rotor (Beckman). p115, sedimenting at 5 S, was concentrated and further purified on monoQ (Pharmacia), and finally gel filtered on Superose 6 (Pharmacia). The p115 peak, which was more than 90% homogeneous by estimation from Coomassie staining after SDS-polyacrylamide gel electrophoresis (PAGE) and was purified approximately 4000-fold compared to cytosol, was snap frozen in aliquots. In some cases, partially purified p115 from the high performance Q Sepharose peak, containing 40 g/ml p115 at approximately 2% total protein, was used. Prior to use in binding assays, any aggregated p115 was removed by centrifugation at 100,000 rpm for 2 min (TLA-100, Beckman).
Preparation of HeLa Cytosols-Spinner HeLa cells were grown and mitotic and interphase cytosols were prepared as described by Stuart et al. (43). Cytosols were desalted into buffer A on P6-DG spin columns (Bio-Rad, Hemel Hempstead, UK), and then frozen in aliquots at Ϫ70°C. Prior to use, cytosol was supplemented with cytochalasin B (1 g/ml, final concentration) and an ATP-regenerating system from a 10 ϫ stock (final concentrations, 10 mM creatine phosphate, 20 g/ml creatine kinase (both Boehringer Mannheim) and 1 mM ATP), and diluted to achieve a concentration of 7.1 mg of protein/ml. To remove aggregated protein, cytosols were preincubated at 37°C for 10 min and precleared by centrifugation at 100,000 rpm for 5 min at 25°C (TLA-100).
Single-step p115 Binding Assay-70 l of either buffer A containing varying amounts of p115 or precleared cytosol desalted into buffer A was mixed with 5 l of salt-washed membranes (derived from 10 g of Golgi protein) and incubated at 4 or 37°C for up to 40 min. Incubations with cytosol contained 50 g of protein at a final concentration of 6.7 mg/ml. In some experiments, parallel incubations were performed without membranes. After incubation the mixture was layered over 75 l of buffer A containing 300 mM sucrose and centrifuged at 100,000 rpm for 10 min (TLA-100). Membrane pellets were resuspended directly into sample buffer (44).
Two-step p115 Binding Assay-For the first step of two-step assays, incubations of (non-salt-washed) membranes with cytosol, were scaled up, typically 8-fold to a total volume of 600 l. After 15 min the incubation was supplemented with microcystin-LR (final concentration, 10 M) and staurosporine (final concentration, 25 M) while still at 37°C. 200 l of 2 M potassium chloride were added to achieve a final concentration of 500 mM, the mixture was layered over an equal volume of buffer A containing 500 mM potassium chloride, 300 mM sucrose, 2 M microcystin-LR, and 5 M staurosporine, and membranes were pelleted by centrifugation at 75,000 for 15 min (TLA-100.2). For the second step, all solutions contained microcystin-LR (2 M) and staurosporine (5 M). Membranes were resuspended in buffer A and incubated with dilutions of p115 or with cytosol as for the single-step assay.
Histone Kinase Assay-Histone kinase activity was assayed after the incubations of cytosol plus membranes and the subsequent removal of membranes by centrifugation. The method of Felix et al. (45) was adapted by diluting the cytosols in buffer A, reducing the final ATP concentration to 0.6 mM, and incubating at 37°C for 10 min. These conditions matched those for p115 binding as closely as possible (see above). Results were expressed in terms of nanomoles of ATP/min/mg of cytosolic protein.
Western Blotting-Proteins were fractionated by SDS-PAGE and transferred to nitrocellulose (Hybond C, Amersham Life Science, Little Chalfont, UK). Blots were blocked in phosphate-buffered saline containing 5% milk and 0.05% Tween 20, and antibodies were diluted to 1 in 1000 in this solution. p115 was detected with the murine monoclonal designated as anti-p115.2 (8A6) (29). Similar results were obtained with another monoclonal, anti-p115.5 (data not shown). Horseradish peroxidase-conjugated goat anti-mouse/rabbit (Tago, Buckingham, UK) was used to detect primary antibodies. Bands were visualized by enhanced chemiluminescence (Amersham Life Science).
Quantitation of Western Blotting-Western blots for p115 were scanned by a video camera under the control of a Screen Machine card to produce digital images. The pixel density was calculated using a local background value adjacent to each band (NIH Image 1.51). The p115 bound to membranes (g) was calculated by comparison of each band to a standard curve, which was run in several lanes of each SDS-PAGE gel. In experiments with rat p115, standard curves were made up of untreated Golgi membranes, and these were calibrated separately to a Western blot of purified rat p115. In experiments with human p115, standard curves consisted of serial dilutions of purified human p115. To enhance the accuracy of quantitation, duplicate gels, blots and images were made from the same samples in some experiments, and each data point finally obtained was the average of duplicate values. In some experiments, the recovery of mannosidase II was measured by reprob-ing blots with rabbit anti-mannosidase II. In the event of a lack of uniformity of blotting, results were discarded.
Golgi Disassembly in Vitro-700 g of mitotic HeLa cytosol protein in 70 l were mixed with 10 l of either (i) buffer B (as buffer A, but replacing 50 mM potassium acetate with 50 mM potassium chloride and 50 mM Tris/chloride, pH 7.3) (11); (ii) concentrated p115; (iii) NSF, ␣ and ␥SNAPs; or (iv) all of p115, NSF, ␣ and ␥SNAPs. All of these mixtures were desalted into buffer B, and immediately precleared by centrifugation at 14,000 ϫ g for 3 min. The mixtures were then supplemented with ATP-regenerating system, cytochalasin B (1 g/ml final), and microcystin-LR (10 M final) prior to the addition of 10 g of Golgi membranes (final volume 100 l). The final concentrations of added proteins were p115, 50 g/ml; NSF, 25 g/ml; and both ␣/␥SNAPs, 5 g/ml. The 5 g of purified p115 added to the estimated 1.7 g of p115 present in the mitotic cytosol led to a 4-fold increase in the p115 concentration to 600 nM.
After incubation, membranes were recovered by centrifugation in a horizontal rotor at 14,000 rpm for 15 min, fixed, and processed for electron microscopy as described previously (11). Ultrathin sections (50 -70 nm) were viewed using a Philips CM10 electron microscope. Five fields per sample were photographed in a systematic, random fashion and printed at a final magnification of 29,000. The percentage of membranes present in cisternae was determined by the intersection method.

RESULTS
A Receptor for p115 on Golgi Membranes-Endogenous p115 is stripped from membranes by salt (29,46). To titrate the amount of salt required, highly purified rat liver Golgi membranes were incubated with buffer containing 0 to 2 M potassium chloride and pelleted through a sucrose cushion with the same amount of salt. The amount of p115 bound to the membranes was quantitated by SDS-PAGE and Western blotting. At most, 85% of endogenous p115 was stripped from the Golgi membranes, and this occurred at and above 500 mM potassium chloride, the level chosen to prepare salt-washed membranes in further experiments.
Salt-washed Golgi membranes (derived from 10 g of Golgi protein) were incubated at 4°C for 10 min with increasing amounts of p115 which had been partially purified from rat liver cytosol. The membranes were reisolated by centrifugation through a sucrose cushion, and fractionated by SDS-PAGE. Western blots for p115 ( Fig. 1) were scanned to produce digital images. The p115 bound to membranes was compared to a standard curve run in the same SDS-PAGE gels, and p115 in each lane was calculated in absolute terms (g). Parallel incubations without membranes (Fig. 1, lower series of bands) showed 3% of p115 in the pellet. Therefore, 3% of total p115 was assumed to have pelleted independent of membrane binding, and this amount was subtracted as background to produce the amount of membrane-bound p115.
At low p115 concentrations binding was linear, while at high p115 concentrations binding reached a plateau (Fig. 1). A similar binding curve was also seen using p115 purified to homogeneity from rat liver cytosol (data not shown). These data indicate that p115 associates with Golgi membranes as a ligand binding one or more receptors, and that there is no cytosolic cofactor for p115 which mediates its binding. An alternative model, that p115 is assembled into polymers in a manner nucleated or controlled by a Golgi protein, is not supported by our data, since this interaction would be expected to be cooperative, and Hill plots from several experiments produced Hill coefficients close to one, which indicates noncooperativity (data not shown).
Using a fixed, saturating amount of p115 (0.9 g), the amount of p115 bound in vitro was 70% of that seen on untreated Golgi membranes (Table I). Since 75% of membranes were recovered in the membrane pellet, as assessed by Western blotting for the Golgi-resident enzyme ␣-mannosidase II (47), binding of p115 normalized for the recovery of Golgi mem-branes was close to 100% of that seen in the starting material. Pretreatment of salt-washed Golgi membranes with chymotrypsin prevented more than 75% of subsequent p115 binding, while mock treatment had no effect (Table I). Heating the membranes to 95°C also abolished the binding of exogenous p115 (Table I). Washing the Golgi membranes with alkali (pH 11.5) reduced p115 binding by more than 60% (Table I). These results indicate that p115 binds a protein receptor on Golgi membranes. Either this receptor is an integral membrane protein partly denatured by alkali, or alternatively it is not an integral membrane protein and is partially extracted by alkali.
p115 was not detected on, and did not bind to, mitochondria or rough microsomes (Table I). Since these membranes would not be expected to bind p115, binding of p115 in vitro shows the same specificity as p115 binding in vivo.
Mitotic Effects on p115 Binding-To investigate the binding of p115 to Golgi membranes under mitotic conditions, we used a system which mimics mitotic Golgi fragmentation in vitro, using rat liver Golgi membranes together with a high concentration of cytosol (final concentration, 6.7 mg of protein/ml) from mitotic HeLa cells (11). Salt-washed Golgi membranes were incubated with mitotic or interphase HeLa cytosols for up to 40 min at 4 and 37°C. The binding of p115 to Golgi membranes at 4°C was similar with mitotic and interphase cytosol ( Fig. 2A). Binding was rapid, since it was maximal after a very brief incubation followed by centrifugation for 10 min (t ϭ 0).
On warming the mitotic incubations to 37°C, 60% of the membrane-associated p115 was lost over the first 10 min, with a half-time of 3 min (Fig. 2B). Interphase incubations at 37°C showed little change in p115 binding during this time. Between 10 and 40 min there was a further slight reduction in p115 binding with both cytosols (Fig. 2B). Prior salt washing was not required for the loss of p115 from mitotic membranes, since the same time course and extent of p115 loss was seen with untreated Golgi membranes (data not shown). The change in p115 on mitotic membranes did not reflect a general loss of detectability, for example following proteolysis or a mitotic modification to p115, since the yields of p115 were unaffected by these incubations. Therefore, in mitotic incubations at 37°C, p115 binding was rapidly inhibited.
p115 Receptor as the Target of Mitotic Modification-The inhibition of p115 binding to Golgi membranes from mitotic cytosol was seen at 37°C but not at 4°C (Fig. 2). To test whether the membranes might be the target of a mitotic modification at 37°C, we performed a two-step assay. In the first step, membranes were incubated with mitotic cytosol for 15 min at 37°C. Mitotic membranes were then reisolated at 4°C in the presence of salt to strip membrane-bound p115. In the second step these membranes were incubated with increasing amounts of p115 at 4°C for 10 min to determine the binding affinity (Fig. 3). At the end of the first incubation, prior to the salt wash, two additions were made, microcystin-LR, and staurosporine. Microcystin-LR, which inhibits phosphatases of types 1, 2A, and X (48,49), was added to prevent the action of phosphatases during the reisolation and second incubation which might otherwise revert mitotic phosphorylations to membrane proteins. Staurosporine, a broad spectrum inhibitor of many kinases including cyclin-dependent kinases (50 -52), was added to inhibit mitotic kinases. These are carried over from the first to the second step, and might otherwise have modified the added p115. The overall intended effect was to "freeze" the phosphorylation state of the second step.
"Frozen" mitotic membranes showed reduced binding of partially purified rat p115 (Fig. 3A), while interphase membranes, which were made as a control, showed binding of p115 similar to that of plain, untreated membranes in parallel experiments (compare Figs. 1 and 3A). The amounts of membrane-bound p115 were converted to concentrations of p115 monomer (molecular weight, 108 kDa) and used to plot bound/free versus bound (Scatchard plot, inset). Interphase membranes had a high affinity binding site, with a dissociation constant (ϪK d ϭ slope of line) of 4 nM. Mitotically treated membranes showed a much lower binding affinity (K d ϭ 85 nM). Similar results were obtained using rat p115 purified to Ͼ90% homogeneity (data not shown). Since there was some variation between experiments, indicated by the width of the error bars in the binding curves, the K d values are only approximate. However, the differences between the interphase and mitotic sets of data indicate a reduction of 20-fold in the affinity of p115 binding to mitotically treated membranes, without a reduction in the number of binding sites.
The mitotic effect on Golgi membranes reducing p115 binding (Fig. 3A) was seen using rat p115, rather than the human source (HeLa cytosol) used in Fig. 2. To demonstrate that the same effect on membranes also applied for p115 binding across this species barrier, p115 was purified from human placental cytosol to Ͼ90% homogeneity. This bound untreated Golgi membranes in a saturable fashion similar to rat p115 (data not The characteristics of p115 binding to membranes Membranes were salt-washed/alkali-washed as indicated, incubated with rat p115 (0.9 g) for 10 min at 4°C, recovered by centrifugation, and analyzed for p115 binding by SDS-PAGE and Western blotting. The amounts of membrane used were: 10 g for starting and salt-washed Golgi membranes, 20 g for alkali-washed Golgi membranes, 100 g for mitochondria, and 50 g for rough microsomes. Values given are the amounts of membrane-bound p115/10 g of membranes compared to starting Golgi membranes (100%  p115 Binding to Mitotic Golgi Membranes shown). Prior treatment of membranes with mitotic HeLa cytosol produced a similar inhibition of binding of human p115 as seen with rat p115 (Fig. 3B), the K d being 95 nM compared to 3 nM for the interphase control.
For both rat and human p115, mitotic pretreatment did not reduce the number of binding sites for p115 on Golgi membranes (intercept on y axis in the Scatchard plots, Fig. 3). Rather, there was a slight (1.5-fold) increase in the number of binding sites.
p115 as the Mitotic Target-To determine whether p115 itself could be a mitotic target, interphase and mitotic HeLa cytosols were preincubated for 15 min at 37°C. To preserve any modifications to p115, cytosols were supplemented with microcystin-LR and staurosporine to freeze its phosphorylation state. Varying amounts of these frozen cytosols were then incubated with salt-washed Golgi membranes at 4°C for 10 min, to determine the binding affinity by quantitative Western blotting. p115 in frozen interphase and mitotic cytosols produced similar binding curves and Scatchard plots as each other (Fig. 4), the K d values for p115 binding being 13 nM and 9 nM, respectively. Therefore, the binding of p115 to Golgi membranes was not reduced by a mitotic preincubation of p115.
Reversibility of the Mitotic Inhibition of p115 Binding-Intracellular transport rapidly resumes during cytokinesis at the same time as the reassembly of Golgi stacks (53). If the mitotic effect on p115 binding is relevant in vivo it might be expected to be reversible. This was investigated in a two-step, crossover assay. Golgi membranes were first incubated with HeLa cytosols at 37°C for 10 min, after which there was approximately twice as much p115 on the interphase membranes compared to mitotic membranes (Fig. 5). Treated membranes were then salt-washed to remove bound p115, and incubated for a second time for 10 min. When the second incubation was with buffer at 4°C, little p115 was detected on the membranes, showing the effect of the salt wash. When the second incubation was with interphase cytosol at 37°C, the same high level of bound p115 was obtained no matter whether the membranes had previously been interphase (solid bars) or mitotic (open bars). When the second incubation was with mitotic cytosol at 37°C, a low level of p115 binding was obtained for both sets of membranes. Therefore the mitotic inhibition of p115 binding was reversible.
The Role of Phosphorylation in Mitotic Inhibition of p115 Binding-The level of p115 binding from HeLa cytosols was inversely correlated with the histone kinase activity, which is a reflection of the activity of the mitotic kinase, p34 cdc2 . Thus, interphase cytosol produced high p115 binding and had low histone kinase activity, while mitotic cytosol showed an inhibition of p115 binding and had high histone kinase activity (Fig. 6). The mitotic pattern was mimicked in interphase cytosol by addition of cyclin A (Fig. 6), which shows that a cyclindependent kinase can support the mitotic modification to the receptor for p115.
As expected, olomoucine, a specific inhibitor of cyclin-dependent kinases including complexes containing cyclin A (54), inhibited the effects of cyclin A on interphase cytosol (Fig. 6). However, olomoucine did not affect the ability of mitotic cytosol, which contains cyclin B not cyclin A (55), to inhibit p115 binding, although the histone kinase activity of this cytosol was FIG. 3. Two-step assay to determine whether the p115 receptor is the mitotic target. In the first step, Golgi membranes were incubated for 15 min at 37°C with either interphase (triangles) or mitotic (circles) HeLa cytosol. The phosphorylation states were then "frozen" by addition of microcystin-LR and staurosporine. Potassium chloride (500 mM) was added to remove any membrane-bound p115, and the membranes were reisolated by centrifugation. In the second step, increasing amounts of A, rat and B, human p115 were incubated with the pretreated, frozen, salt-washed Golgi membranes for 10 min at 4°C in the continuing presence of microcystin-LR and staurosporine. Membranes were pelleted again, and, following SDS-PAGE, p115 was detected by Western blotting to produce binding curves from the average of three experiments (bars ϭ S.E.), the bands from one of which are shown at the top. The insets show Scatchard plots of the ratio of bound to free p115 (x axis) versus membrane-bound p115 (y axis, nM). partially inhibited (Fig. 6). In contrast, staurosporine, a broad spectrum kinase inhibitor, completely inhibited histone kinase activity and prevented the p115 receptor modification (Fig. 6). Staurosporine did not affect the binding of p115 from interphase cytosol (data not shown). The difference between the effects of olomoucine and staurosporine is discussed below.
Distribution of p115 between Membranes and Cytosol in Interphase and Mitosis-The concentration of p115 in HeLa cytosols was lower in interphase cytosol than mitotic cytosol (74 versus 160 nM final concentration in these experiments), the mitotic to interphase ratio being 2.1. This increase of p115 concentration in mitotic cytosol was not reflected in the total cellular levels (mitotic to interphase ratio ϭ 1.0). These findings imply that p115 shifts from membranes to cytosol during mitosis in vivo.
Golgi Disassembly in Vitro in the Presence of Excess p115-Stacked Golgi cisternae are rapidly disassembled in vitro by incubation with mitotic cytosol under conditions similar to those which inhibit binding of p115 (11). In the initial stages, cisternae are converted into small, homogeneous vesicles in a COP I-dependent fashion. A separate COP I-independent process is also active throughout disassembly (12, 56), but it is relatively less active in the initial stage of disassembly. We have examined the effect of excess p115 over the first 15 min of in vitro mitotic disassembly and quantified loss of membrane from cisternae morphologically. This time point was chosen to maximize the role of the COP I-dependent pathway.
Incubation with mitotic cytosol at 4°C produced the same proportion of membrane in cisternae as seen in the starting FIG. 4. Two-step assay to determine whether p115 is the mitotic target. In the first step, interphase and mitotic cytosols were incubated for 15 min at 37°C. The phosphorylation states were then "frozen" by addition of microcystin-LR and staurosporine, and aggregates were removed by centrifugation. In the second step, dilutions of frozen cytosols were incubated with salt-washed Golgi membranes for 10 min at 4°C in the continuing presence of microcystin and staurosporine. Membranes were recovered by centrifugation, and p115 in the pellets was detected by Western blotting (bands at the top), and quantitated to construct binding curves and Scatchard plots (inset). The data are taken from one of two similar experiments.

FIG. 5. Reversibility of the mitotic inhibition of p115 binding.
In the first incubation of a two-step, crossover assay, Golgi membranes were incubated with either interphase (I, solid bars) or mitotic (M, open bars) HeLa cytosol at 37°C for 10 min. An aliquot of membranes was pelleted to measure bound p115 at this stage. Remaining membranes were salt-washed with 500 mM potassium chloride, to remove any membrane-bound p115, and recovered by centrifugation. In the second incubation, the pretreated membranes were incubated for 10 min with buffer at 4°C, interphase cytosol at 37°C, or mitotic cytosol at 37°C. Membrane-bound p115 was quantitated by SDS-PAGE and Western blotting.
FIG. 6. The role of phosphorylation in the inhibition of p115 binding. Interphase and mitotic cytosols were preincubated either alone or with various additions, as shown, for 10 min at 37°C, and precleared by centrifugation. Additions were at the following concentrations: staurosporine, 25 M; cyclin A, 2 M; olomoucine, 1 mM. The cytosols were then incubated with salt-washed Golgi membranes for 10 min at 37°C, followed by centrifugation. Solid bars, the pellets were fractionated by SDS-PAGE and Western blotted to determine membrane-bound p115. Hatched bars, histone kinase activity in the supernatant at the end of the incubation. material (45%). Incubation at 37°C produced considerable cisternal disassembly, only 22% of membrane remaining in cisternae, a loss of 23% (Table II). Addition of purified p115 to this incubation to increase the final p115 concentration 4-fold reduced the extent of cisternal disassembly. 40% of membrane remained in cisternae, a loss of only 5% (Table II). In contrast, excess NSF and SNAPs only marginally inhibited cisternal disassembly, and did not synergize with the effect of p115 (Table II). The activity of these purified components was demonstrated elsewhere in two separate assays, 20 S particle formation (data not shown) and Golgi reassembly (57). Overall, these results show that, during mitotic Golgi disassembly in vitro, there is a stage when the docking action of p115 is rate-limiting. DISCUSSION The binding of the vesicle docking protein p115 to Golgi membranes in vitro was studied in incubations with cytosol from mitotic HeLa cells arrested in prometaphase by growth for one cell-cycle in the microtubule depolymerizing drug nocodazole. In all experiments, cytosol from unsynchronized cultures (interphase) was used as a control. Our initial observation was that p115 binding decreased during mitotic incubations at 37°C. Loss of p115 from mitotic Golgi membranes in vitro has also been inferred elsewhere (57). p115 purified from both rat liver and human placenta bound at 4°C to membranes preincubated with interphase cytosol with a K d of 3-4 nM. However, incubation of the membranes with mitotic cytosol reduced the affinity of p115 binding, both human and rat, with an increase in the K d of 20-fold to 80 -95 nM, with a small increase in the number of binding sites. By comparison, p115 in interphase and mitotic cytosols bound to Golgi membranes with very similar K d values to each other (13 and 9 nM, respectively), showing no mitotic inhibition acting via p115 itself. Together these results strongly suggest that the receptor for p115, or some cofactor of the receptor, is the target of a mitotic modification that decreases the receptor's affinity for p115. The inhibition of binding was reversible, consistent with the rapid reassembly of Golgi stacks in telophase.
Fractionation of whole cells showed a 2-fold increase in p115 concentration in cytosols prepared from mitotic cells compared to interphase cells, while the total cellular levels were the same. Therefore, there is evidence from a second source that binding of p115 to Golgi membranes is inhibited under mitotic conditions. Together our data suggest that the same inhibition is seen during mitosis in vivo, but do not provide a direct demonstration of this. This raises the question of the applicability of our in vitro findings to the in vivo situation. In particular, we observed decreased binding of p115 using cytosol at 6.7 mg/ml of protein, while the protein concentration of cytoplasm is an order of magnitude greater than this. Therefore, the total cellular concentration of p115 is higher than the range of concentrations at which the observed mitotic inhibition of binding acts (Fig. 3). However, there are several reasons why p115 binding may be inhibited in mitosis in vivo despite this difference. First, as mentioned above, there is evidence from two separate sources that p115 binding is inhibited under mitotic conditions. Second, total cellular p115 does not equate free cytoplasmic p115, the concentration of which cannot be estimated. A considerable fraction may be bound to membranes and the cytoskeleton by mechanisms additional to the Golgi receptor described here. Finally, although the conditions of the in vitro binding assay approximate physiological conditions, the differences might radically alter the K d values. For example, the difference between the affinity values for binding to interphase membranes by p115 in HeLa cytosols (9 -13 nM) and purified p115 (3-4 nM) might reflect the different experimental design, especially the presence of 6.7 mg/ml cytosolic protein.
We found that a cyclin-dependent kinase present in interphase cytosol can support the mitotic modification to the receptor for p115. The likely candidate is p34 cdc2 , since this is the kinase in interphase cytosol which has been previously shown to be activated by cyclin A in vitro for both the inhibition of intra-Golgi transport (58) and the disassembly of Golgi stacks (11). These studies both showed that a second, cyclin-independent kinase could mediate similar effects on the Golgi apparatus (11,58). Similar second kinases which can inhibit membrane fusion have also been found in other steps of intracellular traffic (59,60). The difference between the effects of olomoucine and staurosporine added to mitotic cytosol in this study point to the same explanation. Olomoucine, which is specific for cyclindependent kinases, did not affect mitotic inhibition of p115 binding. Staurosporine, a kinase inhibitor with a broad spectrum including cyclin-dependent kinases (50 -52), completely prevented the mitotic effect. The second kinase may either act downstream of p34 cdc2 , or it may be mitotically activated in parallel to p34 cdc2 .
The involvement of a kinase was corroborated by the finding that microcystin-LR, a phosphatase inhibitor, preserved the mitotic modification to p115 binding during two-step assays. In similar experiments performed without microcystin-LR, the mitotic inhibition of p115 binding was not seen (data not shown). This suggests that the reversibility of the mitotic effect is brought about by a phosphatase present on the membranes. Thus, the mitotic modification may be a phosphorylation, or, if not, it is controlled by a pathway with a regulatory phosphorylation upstream.
The functional significance of the mitotic inhibition of p115 binding was investigated by adding excess p115 to incubations which mimic mitotic Golgi disassembly (12, 56). A 15 min time point was chosen to maximize the role of the COP I-dependent pathway, since the COP-dependent pathway acts rapidly, while the COP I-independent pathway is relatively inactive during the initial stages of disassembly (12). Normal mitotic incubation reduced the proportion of membrane in cisternae from 45 Golgi membranes were incubated with mitotic HeLa cytosol at either 4 or 37°C for 15 min. Incubations were supplemented with p115, or a mixture of NSF, ␣ and ␥SNAPs, or all of these. Membranes were recovered by centrifugation, fixed, and processed for electron microscopy. The proportion of membrane in cisternae was quantified as the percentage of total membrane by the intersection method, and the values shown are the mean Ϯ S.D. The right-hand column shows the loss of membrane from cisternae, given an initial value of 45% of total membrane. p115 Binding to Mitotic Golgi Membranes to 22% of total membranes. In the presence of 4-fold excess of purified p115, mitotic incubation only reduced the proportion of membrane in cisternae to 40%, a 78% reduction in the extent of cisternal disassembly. The ability of p115 to inhibit Golgi disassembly was specific, in that excess amounts of other components of the fusion machinery (NSF, ␣SNAP, and ␥SNAP) had only a marginal effect on the loss of cisternal membranes. We interpret this to mean that during mitotic disassembly, the reduced p115 binding made its action rate-limiting, which led to reduced vesicle docking and fusion. Since budding continued (11,12), the Golgi apparatus disassembled. Addition of excess p115 inhibited the mitotic effect by raising the levels of receptor occupancy back toward interphase levels, restoring the docking of transport vesicles. Excess p115 did not completely abolish disassembly, which may reflect either that not enough p115 was added, or the presence of other mitotic targets in the fusion machinery, or the action of the COP I-independent pathway.
Clues to the mode of action of p115 have been provided in recent studies. Rotary shadowing electron microscopy reveals a myosin shaped dimer, comprising closely aligned coiled-coil rodlike domains 45 nm in length and independent globular domains at one end (30). Sequence comparisons between p115 and its yeast homologue Uso1p show three highly conserved regions, two in the globular domain (N terminus) and one at the extreme distal end of the rod (C terminus) (30,31). Uso1p's action lies upstream of SNAREs and promotes their activity in vesicle docking (36). This concurs with the findings that p115 promotes the interaction of transcytotic vesicles with their target membrane prior to membrane fusion, and possibly prior to the formation of tight v/t-SNARE interactions (31). One model is that p115 mediates the initial link between transport vesicles and Golgi membranes, enhancing the efficiency with which the vesicle samples the target for the cognate tSNARE. In the case of incorrect pairing of vesicle and target, dissociation of p115 would allow the vesicle to diffuse away. We have shown that the binding of p115 is inhibited in mitosis by a modification to its receptor or a cofactor. Reduced p115 binding may lead to reduced efficiency of vesicle docking, and this may be an important factor in the mitotic inhibition of membrane fusion which underlies the cessation of membrane traffic. In the face of continued budding of transport vesicles, this would also facilitate disassembly of the Golgi apparatus. Our priority now is to identify the receptor for p115 and any other interacting molecules.