A Specific Activation of the Mitogen-Activated Protein Kinase Kinase 1 (Mek1) Is Required for Golgi Fragmentation during Mitosis

The pericentriolarly organized Golgi apparatus of mammalian cells undergoes extensive fragmentation during mitosis. This has been known for almost three decades (Robbins and Gonatas 1964; Lucocq et al. 1987). This fragmentation is observed in mammalian cells and not in plants, yeast, and flies during mitosis (Hirose and Komamine 1989; Preuss et al. 1992; Stanley et al. 1997). However, the extent of fragmentation, the fate of the fragmented Golgi membranes, the significance of this fragmentation, and above all, the mechanism of the Golgi fragmentation are subject to debate.

To address these questions, we recently reconstituted Golgi fragmentation in permeabilized cells. In this system, normal rat kidney (NRK) cells are permeabilized, washed to remove cytosolic proteins, and then incubated with mitotic extracts prepared from NRK cells arrested in mitosis. The reaction mixture is incubated at 32°C for 60 min, and the organization of the Golgi membranes monitored by fluorescence microscopy using Golgi-specific antibodies. We have shown that mitotic cytosol is highly enriched in cdc2 kinase activity, but that the cdc2 kinase activity is not required for Golgi fragmentation and dispersal of the Golgi fragments from the pericentriolar region. On the other hand, inactivation of the mitogen-activated protein kinase kinase 1 (MEK1) by drugs, or depletion of MEK1 by affinity chromatography or fractionation of mitotic cytosol, results in a loss of Golgi fragmentation. Surprisingly, the cytosolic downstream targets of MEK1, i.e., extracellular signal–regulated kinase (ERK) 1 and ERK2 (MAPKs), are not required for the MEK1-dependent Golgi fragmentation (Acharya et al. 1998).

We now provide evidence that MEK1 undergoes specific activation by phosphorylation during mitosis. The mitosis-specific activation of MEK1 is required for the conversion of the Golgi into tubuloreticular elements, and this fragmentation step does not require cdc2 kinase.

All biochemical reagents were obtained from the sources described before unless otherwise mentioned (Acharya et al. 1998). Nickel agarose beads were purchased from Amersham Pharmacia Biotech. Rabbit anti–rat mannosidase II (MannII) and the anti–NH₂-terminal MEK1 antibody were kindly provided by Drs. Kelly Moremen (University of Georgia, Athens, GA) and James Ferrell (Stanford University, Stanford, CA), respectively. The anti–COOH-terminal MEK1 was purchased from Santa Cruz Biotechnology. The phospho-MEK antibody was purchased from New England Biolabs, Inc. Drs. Cesare Montecucco (University of Padua, Padua, Italy) and Nick Duesbery (Van Andel Research Institute, Grand Rapids, MI) kindly provided lethal factor. Recombinant His-MEK1 was a kind gift from Drs. Jocelyn Wright (SUGEN, South San Francisco, CA) and Natalie Ahn (University of Colorado at Boulder, Boulder, CO). Recombinant ΔN7-MEK1 was a gift from Dr. Nick Duesbery. All EM reagents were purchased from the sources described before (Takizawa et al. 1993). Trypsin was purchased from Calbiochem.

All tissue culture cells are grown in complete medium consisting of α-MEM (Gibco) with 10% FCS, 100 U/ml penicillin, and 100 μg/ml streptomycin at 37°C in a 5% CO₂ incubator.

Cytosol from NRK cells arrested in mitosis or interphase was prepared as described before (Acharya et al. 1998). Cell permeabilization and assay for Golgi fragmentation by mitotic extracts were carried out as described before (Acharya et al. 1998).

Mitotic extract was pretreated with lethal factor (30 ng/ml) for 10 min at 32°C. This mixture and an ATP regeneration system were then added to semi-intact cells and the incubation carried out for 1 h at 32°C. When PD and olomoucine were used, these compounds were added to cytosol pretreated with lethal factor at a final concentration of 75 and 200 μM, respectively. The mixture was incubated for 10 min at 32°C in the presence of an ATP regeneration system and then added to semi-intact cells for 60 min at 30°C to monitor effects on Golgi fragmentation using the permeabilized cell system.

In a typical preparation, 20 μg of recombinant His-MEK1 (wild-type or ΔN7) was incubated in a total volume of 80 μl with interphase or mitotic extract at a final concentration of 4 mg/ml. The kinase buffer used was composed of 50 mM Hepes, pH 7.2, 10 mM MgCl₂, 1 mM DTT, 500 μM ATP, and 30 μCi [³²P]ATP. The reaction mixture was incubated for 30 min at 30°C. At the end of the incubation, His-MEK1 was adsorbed of nickel beads, resuspended in 50 mM Hepes, pH 7.2, 200 mM NaCl, 10 mM β-mercaptoethanol, 10 mM imidazole, 10% glycerol, and finally eluted with 200 mM imidazole.

His-MEK1 incubated with mitotic or interphase extract as described above was incubated at room temperature with trypsin (1:100, enzyme/substrate in wt/wt). At the indicated times, aliquots were removed and the reaction terminated by boiling the samples in Laemmli sample buffer. The samples were analyzed by SDS-PAGE followed by autoradiography and immunoblotting with the indicated antibodies as described in the figure legends.

The procedure was the same as described before (Acharya et al. 1998).

NRK cells were grown in special petri dishes (Matteck Corporation; Takizawa et al. 1993), permeabilized as described above, and processed for immunoelectron microscopy as described by Polishchuk et al. 1999. For the analysis of intact mitotic cells, NRK cells were incubated for 12 h with 2.5 μg/ml aphidicolin to arrest cells in S phase. The cells were washed by repeated changes of medium and placed in fresh medium. After 6–8 h, the mitotic cells were collected by gentle shake-off and processed for immunoelectron microscopy using anti-ManII antibody followed by HRP-conjugated secondary antibody. Ultrathin sections (80 nm) were observed at 80 KeV with either a JEOL 100C× or 2000F× electron microscope, and semithin sections (0.5 μM) at 400 KeV with a JEOL 4000E× intermediate voltage electron microscope. The mitotic stages were assigned as described by Lucocq et al. 1987.

It is known that activation of MEK1 is necessary for a number of mitosis-specific events in somatic cells (Acharya et al. 1998; Shapiro et al. 1998; Zecevic et al. 1998). To test whether the mitotic form of MEK1 is in any way different from its activated counterpart found in the nonmitotic stages, wild-type His-tagged MEK1 was expressed in Escherichia coli and the recombinant purified protein incubated with mitotic extracts or interphase extracts in the presence of [³²P]ATP at 30°C. MEK1 was isolated from the reaction mixtures by adsorption on Ni beads and analyzed by SDS-PAGE and autoradiography (Fig. 1 A). Quantitation of the autoradiogram revealed a fourfold increase in phosphorylation of MEK1 incubated with mitotic vs. interphase extract (Fig. 1 C).

An antibody that recognizes a generically activated form of MEK1 (phospho-MEK1 antibody) was used to test whether this antibody recognizes the mitotically phosphorylated form of MEK1 compared with MEK incubated with interphase cytosol. The phospho-MEK1 antibody recognizes an activated form of MEK1 phosphorylated at Ser217 and Ser221. This form of MEK1 activation occurs upon the incubation of cells with growth factors or cytokines, or during membrane depolymerization and calcium influx (Crews et al. 1992; Alessi et al. 1994; Robinson and Cobb 1997). The Ser217- and Ser221-phosphorylated MEK1 activates the downstream kinases called ERKs. An important question is whether the mitotically phosphorylated MEK1 is in some way different from the generically activated form of MEK1. To test this, His-MEK1 was incubated with mitotic cytosol or interphase cytosol and ATP. MEK1 was isolated on Ni beads, released from the beads by imidazole treatment, and analyzed by Western blotting with the phospho-MEK1 antibody. The phospho-MEK1 antibody is twofold less reactive with mitotically activated MEK1 compared with interphase MEK1 (Fig. 1B and Fig. D). Thus, even though MEK1 is hyperphosphorylated during mitosis, it is less reactive with the antibody that recognizes the generically phosphorylated (nonmitotic) form of MEK1.

His-tagged MEK1 was incubated with either interphase or the mitotic extract in the presence of [³²P]ATP for 30 min at 30°C. The tagged protein was isolated on Ni beads and subjected to limited tryptic digestion. At various times the reaction was analyzed by SDS-PAGE and autoradiography and Western blotting with either anti–COOH-terminal or NH₂-terminal–specific MEK1 antibody or the phospho-MEK1 antibody. Interestingly, trypsin digestion of mitotically activated MEK, and not interphase MEK1, results in the production of a prominent phosphorylated polypeptide of ∼20 kD (Fig. 2 A). This polypeptide is recognized by the COOH-terminal–specific antibody, but not by the phospho-MEK1–specific antibody or the NH₂-terminal MEK1 antibody (Fig. 2 B). The mitotic form of MEK1 is less susceptible to digestion with trypsin compared with its interphase counterpart (Fig. 2 B). Thus, the overall susceptibility and pattern of trypsin digestion is different in MEK1 incubated with interphase vs. the mitotic cytosol. The 20-kD polypeptide of MEK1 is not generated when it is incubated with interphase extracts. This would explain the increase in phosphorylation of MEK1 during mitosis, without any increase in reactivity to the phospho-MEK1 antibody. We propose that MEK1, upon phosphorylation by the mitotic extracts, undergoes a change in conformation and this new conformation is differentially sensitive to proteolysis, rendering it active for the Golgi fragmentation process.

Bacterial anthrax toxin contains a protein called lethal factor, which cleaves the first eight amino acids from the NH₂ terminus of MEK1 (Duesbery et al. 1998; Vitale et al. 1998). This cleaved form of MEK1 is incapable of binding its downstream target, the MAPKs ERK1 and ERK2. We asked whether the mitotic-specific phosphorylation is possible without the first eight amino acids of MEK1, and whether cleavage with the lethal factor affects MEK1 activity in the Golgi fragmentation process. To address the first question, we used a recombinant form of MEK1 lacking the first seven amino acids. The purified recombinant His-tagged ΔN7-MEK1 was incubated with the mitotic or the interphase cytosol and [³²P]ATP for 30 min at 32°C. The ΔN7-MEK1 was isolated on Ni beads and then analyzed by SDS-PAGE followed by autoradiography and Western blotting with the anti–phospho-MEK antibody. Incubation of ΔN7-MEK1 with mitotic extracts results in a fourfold increase in ³²P incorporation, and the mutant protein is about fourfold less reactive with the phospho-MEK1 antibody compared with its counterpart incubated with the interphase cytosol (Fig. 3). Therefore, the wild-type MEK1 and the ΔN7-MEK1 have the same properties with respect to phosphorylation and reaction with the phospho-MEK1 antibody upon incubation with mitotic extract.

Does the treatment of mitotic cytosol with lethal factor affect its Golgi fragmentation activity? To test this, mitotic cytosol was incubated with purified recombinant lethal factor at 32°C for 10 min and the reaction was then divided into three aliquots and used for the following analyses:

(a) The reaction mixture was analyzed by SDS-PAGE followed by Western blotting with the COOH-terminal–specific MEK1 antibody and the NH₂-terminal–specific antibody (Fig. 4). Treatment of the cytosol with lethal factor cleaves MEK1. Western blotting with the COOH-terminal–specific antibody reveals a shift in the molecular weight of MEK1 (Fig. 4, top panel). As expected, the NH₂-terminal–specific antibody fails to react with MEK1 in cytosol treated with lethal factor (Fig. 4, lower panel), indicating that this treatment is effective in cleaving MEK1.

(b) MEK1 was immunoprecipitated from the lethal factor–treated or -untreated cytosol using the COOH-terminal–specific anti-MEK1 antibody. The immunoprecipitated proteins were incubated with purified recombinant ERK2 for 30 min in the presence of ATP at 30°C. The reaction was then centrifuged to remove MEK1, and the supernatant containing ERK2 was incubated with myelin basic protein (MBP), [³²P]ATP at 30°C for 30 min. The reaction was then analyzed by SDS-PAGE and autoradiography and the phosphorylation of MBP taken as a measure of ERK activation. Fig. 5 A shows that cleavage with lethal factor inactivates the ability of MEK1 to activate ERK.

(c) Lethal factor–treated mitotic cytosol was incubated with permeabilized NRK cells that had been prewashed with 1 M KCl. After 60 min of incubation at 32°C, the cells were fixed for immunofluorescence using anti-ManII antibody. 200 cells on two different coverslips were counted, and this quantitation revealed that lethal factor has no effect on the ability of mitotic extracts to fragment Golgi membranes (Fig. 5 B).

We have shown previously that cdc2 kinase is not required for conversion of the pericentriolar Golgi membranes into smaller fragments (punctate structures at the light microscopy level; Acharya et al. 1998). To confirm that this is also the case with lethal factor–treated mitotic extracts, they were incubated in the presence of the generic inhibitor of cdc2 kinase activity, olomoucine. This mixture was then applied to permeabilized cells (washed with 1 M KCl) and an ATP regenerating system at 32°C for 60 min. The samples were visualized by fluorescence microscopy. Results show that under conditions where cdc2 kinase is inactive (the cdc2 kinase activity measured in the histone H1 phosphorylation was completely inhibited under these conditions [data not shown]), lethal factor–treated mitotic cytosol can convert the pericentriolarly organized stacks of Golgi cisternae into smaller structures (Fig. 6a and Fig. b).

We have shown previously that an inhibitor of MEK1 activation called PD 98059 (Alessi et al. 1995; Dudley et al. 1995) inhibits Golgi fragmentation (Acharya et al. 1998). Does this compound affect the Golgi fragmentation activity of lethal factor–treated mitotic cytosol? To test this, mitotic cytosol was incubated with lethal factor for 10 min at 32°C and then incubated with PD. The reaction was then supplied to permeabilized NRK cells, prewashed with 1 M KCl and an ATP regenerating system for 60 min at 32°C. The analysis of the cells at the light microscopy level using the anti-ManII antibody reveals a complete inhibition of the Golgi fragmentation (Fig. 6a and Fig. b). Thus, a PD-sensitive activation of MEK1 is necessary for Golgi fragmentation. These results confirm and further support our proposal that the fragmentation of the pericentriolar Golgi stacks during mitosis is by a PD-sensitive, lethal factor–insensitive, activation of MEK1, and does not require active cdc2 kinase.

We have shown before that mitotic cytosol causes extensive fragmentation of the pericentriolar Golgi membranes (Acharya et al. 1998). However, the analysis was performed only at the light microscopy level and the organization or the morphology of the fragmented Golgi membranes was not reported because of technical limitations, which we have now resolved. NRK cells were permeabilized, washed with 1 M KCl, and then incubated with mitotic cytosol and an ATP regenerating system. A light microscopic analysis reveals the conversion of the pericentriolar Golgi apparatus (Fig. 7 A) into small punctate structures distributed throughout the cells. A similar preparation was processed for immunoelectron microscopy with an antibody to the cis/medial specific Golgi protein ManII. Thin sections of the cells revealed the presence of ManII in small structures composed of small cisternae and tubuloreticular elements (Fig. 7B and C–E). Fig. 7 F shows the control Golgi membranes by immunofluorescence and immunoelectron microscopy with anti-ManII antibody in intact nonmitotic NRK cells. A calculation of the size of the Golgi fragments stained with ManII antibody reveals that the stacks of Golgi cisternae of 1.5 μm average size are converted into tubuloreticular fragments of ∼0.60 μm in size by the action of mitotic cytosol. Incubation of the permeabilized cells with cytosol made from interphase cells and an ATP regeneration system at 32°C for 60 min does not affect the pericentriolar organization of the Golgi membranes (data not shown).

Mitotic extracts treated with the lethal factor were incubated with permeabilized NRK cells and an ATP regenerating system at 32°C for 60 min. The samples were fixed and thin sections analyzed by immunoelectron microscopy using anti-ManII antibody. The results of the analysis are shown in Fig. 8 C. Golgi stacks are converted into tubuloreticular membranes and small cisternae in cells incubated with mitotic extract treated with lethal factor. For comparison, we have included the intact mitotic cells (prometaphase/metaphase; Fig. 8 A) and permeabilized cells treated with mitotic extracts (without the lethal factor treatment; Fig. 8 B). The results show that in all cases the Golgi membranes stained with ManII antibody are composed of tubuloreticular membranes. The size of these tubuloreticular elements on average is ∼0.60 μm. The Golgi fragments in mitotic cells, permeabilized cells incubated with mitotic extracts that are treated with or without the lethal factor, appear morphologically similar and are of similar size and distribution.

We have shown that MEK1 is activated specifically during mitosis, and that this activation does not require the presence of the first eight amino acids of MEK. A comparison of the ManII-stained structures in intact mitotic cells in vivo and in our permeabilized cell system shows Golgi membranes in the form of small tubules and reticular membranes. These structures are highly similar in morphology, size (0.60 μm), and distribution (small elements dispersed in the cell), strengthening the proposal that our assay reconstitutes the first step in Golgi fragmentation during mitosis. This process was MEK1-dependent and cdc2 kinase–independent.

Our results show that MEK1 is hyperphosphorylated in mitosis. Interestingly, this hyperphosphorylation results in a decreased affinity towards the generic phospho-MEK1 antibody. The mitosis-specific activation of MEK1 is inhibited by pretreatment with PD but not with the lethal factor of anthrax toxin. Thus, the fragmentation of the pericentriolar Golgi stacks into tubuloreticular elements is due to a PD-sensitive and lethal factor–insensitive activity of MEK1, which occurs during mitosis. We believe that this mitosis-specific activation by phosphorylation is responsible for a change in the substrate specificity of MEK1. The 20-kD phosphopeptide of MEK1 is generated only as a result of mitosis-specific phosphorylation. It is also interesting to note that Golgi fragmentation by MEK1 does not require the only known substrates, ERK1 and ERK2. Thus, specific activation by phosphorylation changes the substrate specificity of MEK1 so that it can cause Golgi fragmentation without using the ERKs. How a change in the substrate specificity is achieved based on the phosphorylation state of MEK1 is currently not known.

Warren and colleagues have argued that MEK1 is not required for entry into mitosis and Golgi fragmentation (Lowe et al. 1998). This interpretation is based on experiments in which they loaded cells with the lethal factor and found that this resulted in inactivation of ERK1 and ERK2. The cells entered mitosis and the Golgi membranes were found as clusters of small vesicles (Lowe et al. 1998). However, we can explain their findings based on our results that cleavage with lethal factor prevents activation of ERK1 and ERK2, but this form of MEK is still activated by its upstream kinase and can convert Golgi membranes into tubuloreticular elements (Fig. 5, Fig. 7, and Fig. 8).

It has also been argued that MEK activity is lower in mitosis and that this is consistent with the inhibition of MEK1 activity by cdc2 kinase. Lowe et al. 1998 suggested that the low MEK1 activity is consistent with MEK1 not being a key mitotic regulator of Golgi fragmentation. We argue that MEK1 activity is low as measured by the activation of ERK1 and ERK2 phosphorylation. However, its activity towards other potential substrates, such as those involved in Golgi fragmentation, spindle dynamics, and checkpoint control may be regulated by an alternative mechanism (Wang et al. 1997; Acharya et al. 1998; Shapiro et al. 1998; Zecevic et al. 1998).

The fact that cdc2 kinase inhibits MEK1 activation is based on in vitro analysis. Whether this also happens in vivo is not known (Rossomondo et al. 1994). Although cdc2 kinase is required for entry into mitosis and controls the fate of many cellular process, this does not mean that all the mitotic-specific events are regulated by cdc2 kinase or that all phosphorylations by cdc2 kinase (in vitro and in vivo) have a physiological significance (Moreno and Nurse 1990). For example, in the case of lamin disassembly, a recent report shows that this involves sequential phosphorylation first by PKC, which is then followed by cdc2 kinase (Collas 1999).

Warren and colleagues reported that 50 μM completely inactivated MEK1 in vivo in cells grown in the presence of serum without affecting the ability of cells to enter mitosis or undergo Golgi fragmentation (Lowe et al. 1998). We found that 80 μM PD is required for MEK1 inactivation and inhibits the entry of NRK cells into mitosis (data not shown). We carried out this experiment as follows: NRK cells were treated with aphidicolin (2.5 μg/ml) in complete medium (containing 10% FBS) for 12 h at 37°C. Cells were washed extensively to remove aphidicolin and incubated with fresh medium containing serum. The cells were then washed and incubated with complete medium containing either PD (80 μM) or DMSO (the solvent for PD) at 37°C. After 5 h, the cells were analyzed by fluorescence microscopy with the DNA-specific fluorescent dye HOECHST 33258 to quantitate the mitotic index. In DMSO-treated cells, 30% of the cells were in mitosis, whereas only 5% of the cells were in mitosis in incubations with PD. Our observations are supported by those of Wright et al. 1999, who have recently shown that when PD is applied to cells grown in defined conditions (and not in complete medium containing serum), MEK1 is inhibited and the cells are arrested in G2 (Wright et al. 1999).

Our data show that MEK1 is involved in the breakdown of the pericentriolarly organized Golgi membranes into numerous smaller tubuloreticular elements (Acharya et al. 1998; this work). Similar fragments have also been found in cells undergoing mitosis naturally and in cells induced to enter mitosis by okadaic acid treatment (Lucocq 1992; Lucocq et al. 1995). But do these fragments undergo further processing into clusters of vesicles or fuse with the ER (Thyberg and Moskalewski 1992; Zaal et al. 1999)? If the Golgi membranes are converted into numerous smaller fragments (resembling tubuloreticular elements), this should be sufficient for their partitioning into the daughter cells. But there is a large volume of data showing that Golgi membranes are converted into clusters of vesicles in mitotic cells (Lucocq et al. 1987; Warren 1993). Jesch and Linstedt 1998 also find that Golgi membranes are converted into vesicles of 60 nm diameter and can be separated from the ER in HeLa cells treated with nocadazole for 24 h. It is possible, however, that these cells are apoptotic not mitotic, and Sesso et al. 1999 have shown that in apoptotic cells, Golgi membranes are found distributed throughout the cell in the form of clusters of 40–60-nm vesicles. Is Golgi vesiculation a property of cells undergoing mitosis, apoptosis, or both?

In summary there is a general agreement in the field that the pericentriolarly organized Golgi apparatus is converted into smaller elements (tubuloreticular elements and smaller stacks) during mitosis. It remains to be determined whether these fragments undergo further processing to form vesicles or fuse with the ER. A comprehensive understanding of Golgi partitioning will require a resolution of this issue, as the process by which Golgi membranes are rebuilt into stacks (postmitosis) will differ depending on the final domicile of the Golgi during mitosis.

Members of the Malhotra lab are acknowledged for useful discussions and limiting the use of verbiage in the text. We thank Drs. Usha Acharya (National Cancer Institute) and Arrate Mallabiabarrena (UPF, Barcelona, Spain) for some of the early observations on MEK1 activation during mitosis. We thank Dr. Christine Sutterlin of the Malhotra lab for providing large quantities of recombinant MEK. We thank Drs. Nick Duesbury and Cesare Montecucco for lethal factor and Drs. Natalie Ahn and Jim Ferrel for various MEK1-specific reagents.

Some of the work reported was conducted at the National Center for Microscopy and Imaging Research, a National Institutes of Health (NIH) Research Resource supported under NIH grant RR04050 to M.H. Ellisman. The work in the Malhotra lab is supported by the NIH grants to V. Malhotra. V. Malhotra is an established investigator of the American Heart Association. A. Colanzi is supported by a fellowship from the Consiglio Nazionale Delle Ricerche, Italy.