Evidence for Prebudding Arrest of ER Export in Animal Cell Mitosis and its Role in Generating Golgi Partitioning Intermediates


**Corresponding author: John Lucocq, j.m.lucocq@dundee.ac.uk


During mitosis the interconnected Golgi complex of animal cells breaks down to produce both finely dispersed elements and discrete vesiculotubular structures. The endoplasmic reticulum (ER) plays a controversial role in generating these partitioning intermediates and here we highlight the importance of mitotic ER export arrest in this process. We show that experimental inhibition of ER export (by microinjecting dominant negative Sar1 mutant proteins) is sufficient to induce and maintain transformation of Golgi cisternae to vesiculotubular remnants during interphase and telophase, respectively. We also show that buds on the ER, ER exit sites and COPII vesicles are markedly depleted in mitotic cells and COPII components Sec23p, Sec24p, Sec13p and Sec31p redistribute into the cytosol, indicating ER export is inhibited at an early stage. Finally, we find a markedly uneven distribution of Golgi residents over residual exit sites of metaphase cells, consistent with tubulovesicular Golgi remnants arising by fragmentation rather than redistribution via the ER. Together, these results suggest selective recycling of Golgi residents, combined with prebudding cessation of ER export, induces transformation of Golgi cisternae to vesiculotubular remnants in mitotic cells. The vesiculotubular Golgi remnants, containing populations of slow or nonrecycling Golgi components, arise by fragmentation of a depleted Golgi ribbon independently from the ER.

The Golgi apparatus of animal cells forms an interconnected array of closely stacked cisternae situated close to the cell nucleus, and an emerging challenge in cell biology is to understand how such a localised and interconnected structure can be partitioned equally between daughter cells at cell division. To this end, immunofluorescence studies have revealed the Golgi breaks down and disperses during mitosis to yield two types of ‘fragment’ that may be intermediates in the partitioning process (1–4). One type of ‘fragment’ appears as multiple discrete elements with vesiculotubular ultrastructure (5–7). The other ‘fragment’ appears as a finely dispersed organelle, the identity of which is the subject of a current controversy: a controversy which arises from two competing models of Golgi division (8).

In one model the Golgi remains independent of the endoplasmic reticulum (ER) and undergoes fragmentation because transport vesicles continue to bud without fusion (2). Presumably budded vesicles, containing Golgi residents, become dispersed throughout the cytoplasm and equal partitioning is ensured, either by diffusion of vesicles throughout the cytoplasm, or possibly by tethering of the vesicles to structures such as spindle microtubules (9). Support for this model comes from experiments showing Golgi residents remain independent of the ER (10,11) and from demonstration of Golgi residents in isolated vesicles of metaphase cells by immunoelectron microscopy (6). Vesiculation of purified Golgi cisternal stacks (essentially free of the ER) has also been observed in an in vitro system (7), in which disassembly occurs by two distinct mechanisms, one dependent, and the other independent, of COPI transport vesicle budding (7). These studies have identified potential mitotic mechanisms for arresting fusion of transport vesicles (12,13), although these have yet to be tested in vivo.

In the other model (14) arrested ER export (15) traps recycled Golgi components in the endoplasmic reticulum which is then divided equally at cytokinesis. This model gains support from in vivo studies showing endogenous Golgi residents (16), green fluorescent protein (GFP)–Golgi residents (17) and ER–Golgi intermediate compartment (IC) proteins (11) accumulate in the ER of mitotic cells. In its purest form the recycling model predicts all Golgi residents recycle to the ER, but emerging evidence indicates that the extent of recycling (18), and therefore partitioning, varies significantly for different Golgi components. Thus, while the rapidly recycling IC protein, ERGIC53, is completely within the ER of metaphase cells (11), slower recycling Golgi stack residents such as galactosyltransferase may remain in discrete non-ER structures (1,3,4,11). Such results indicate the recycling of Golgi residents during mitosis is a selective rather than an all or none process.

We have previously postulated a distinct version of the recycling hypothesis, in which recycling and trapping of different Golgi components is selective. In this scheme recycling of Golgi components drives transformation of cisternae into tubulovesicular remnants in which nonrecycling residents become progressively concentrated (11,19). This idea is supported by quantitative electron microscopy showing substantial, but not complete, loss of identifiable membrane in the mitotic Golgi (6), and studies demonstrating that experimental ER export arrest is accompanied by transformation of cisternae into vesiculotubular structures (20–23). Further support for selective recycling comes from recent studies in interphase cells showing that so-called Golgi matrix proteins fail to recycle to the ER (18). Such Golgi matrix proteins are good candidates for the Golgi components that would be ‘left behind’ in the vesiculotubular remnants.

In this paper we test a number of predictions of the selective recycling model. The first prediction is that ER export inhibition is sufficient to drive disassembly of the Golgi to form vesiculotubular remnants. Here we show that microinjecting dominant negative mutants of the Sar1 (which block assembly or processing of COPII vesicles budding from ER) induce formation of vesiculotubular aggregates similar to those seen in mitotic cells. Furthermore, in the context of mitosis, we find that microinjection of Sar1 mutants prior to telophase prevents substantial reassembly of cisternae in G1 cells.

The second prediction is that ER export inhibition occurs at an early step in vesicle budding. If ER export inhibition induces membrane loss from vesiculotubular structures, then export should be inhibited at an early stage; otherwise, budded vesicles would accumulate behind the block and the mitotic Golgi would increase in size. In animal cells ER export occurs at specialised ER exit sites that are focuses for budding of COPII vesicles (24–27). At these sites the ER membrane protein Sec12 promotes exchange of GDP for GTP on the small GTPase Sar1p, which is then recruited to the ER membrane and promotes recruitment of coat protein complexes Sec23/24 and Sec13/31 from the cytosol (28–30). We have previously documented that Sec13p is displaced from the ER exit sites in mitosis, consistent with an early budding arrest (11). In this report we show that ER buds and the exit sites of which they form a part, are depleted in mitotic cells. We also find that the coat components Sec23p, Sec24p, Sec31p, and Sec13p are displaced to the cytosol in mitotic cells, further suggesting an early block in COPII vesicle formation.

Finally we examine the idea that vesiculotubular remnants arise by fragmentation rather than recycling via the ER. In interphase cells ER exit sites are distributed in both the juxtanuclear Golgi region and in the cell periphery (26,27,31,32); and those exit sites in the juxtanuclear region are closely associated with the Golgi stack, whereas those in the periphery are not. Thus, if the vesiculotubular remnants were generated by fragmentation of the ER–Golgi region (33), then at least two types should be produced: one containing exit site and Golgi stack derivatives and the other composed only of exit sites.

Alternatively, if the vesiculotubular structures were generated by recycled Golgi components that emerge from the ER (14), the distribution of Golgi stack components would be rather even. Here we show that the distribution of Golgi stack proteins over vesiculotubular structures of mitotic cells is markedly heterogeneous, supporting the idea that vesiculotubular structures arise by fragmentation and not by recycling mechanisms.

Together, our results suggest a consensus model of mitotic Golgi partitioning in which prebudding ER export inhibition induces a selective decanting of recycled Golgi components into the ER. Recycling drives the morphological transformation of cisternae to vesiculotubular remnants that contain slow or nonrecycling components and arise by fragmentation of the Golgi ribbon.


Experimental ER export arrest is sufficient for disassembly of Golgi cisternae

To test the hypothesis that the inhibition of ER export is sufficient to induce the transformation of Golgi cisternae into vesiculotubular structures, we used GDP and GTP restricted Sar1 proteins that are dominant negative for ER–Golgi transport. The guanine nucleotide exchange factor Sec12p promotes GDP/GTP exchange on Sar1p, which is recruited to the ER membrane, and sequential recruitment of coat components Sec23/24 and Sec13/31 from the cytosol follows. The forming COPII structures concentrate cargo molecules in analogous fashion to receptor-mediated endocytosis of ligands at the cell surface (28,29,34), although some abundant soluble cargo molecules may be loaded into COPII vesicles nonselectively (35). GDPSar1p prevents COPII assembly and budding, whereas GTPSar1p allows COPII vesicle assembly/budding and induces accumulation of COPII coated vesicles by preventing uncoating.

GDP and GTP restricted Sar1 proteins were microinjected into nonsynchronised interphase cells and after 1.5–2 h processed for immunofluorescence. Under these conditions the Golgi markers galactosyltransferase (Figure 1) and GM130 (not shown) had acquired a diffuse distribution in most cells, although a minority of the cells contained weak punctate staining in the juxtanuclear region in addition to diffuse staining (upper arrow in Figure 1A). In buffer-injected cells the Golgi markers had a predominantly juxtanuclear distribution, and confirmation that ER export had been inhibited was obtained using the IC marker ERGIC53 which accumulated in the ER of the microinjected cells (not shown). To investigate the morphology of any residual elements, we examined serial ultrathin sections of microinjected cells using transmission electron microscopy (Figure 2). We found large vesiculotubular aggregates that were composed of lucent vesicles and smaller vesiculotubular elements. These aggregates lacked cisternal structures and were mainly located close to the nucleus. After microinjection of GTPSar1p, vesiculotubular structures contained budding profiles on ER elements, consistent with their origin from ER–Golgi structures. We conclude that inhibition of ER export was sufficient to induce Golgi cisternal stack disassembly and form vesiculotubular aggregates similar to those found in mitotic cells (see ref. 5).

Figure 1.

Figure 1.

Microinjection of GDPSar1p and GTPSar1p induces dispersion of the galactosyltransferase. Unsynchronised interphase cells were microinjected with Sar1 dominant negative mutant proteins and left for 1.5–2 h at 37 °C before fixation and processing for immunofluorescence microscopy. Bar 10 µm.

Figure 2.

Figure 2.

Microinjection of GTPSar1p induces disassembly of Golgi cisternae. Unsynchronised interphase cells were microinjected with Sar1 dominant negative mutant protein and left for 1.5–2 h at 37 °C before fixation and processing for electron microscopy. Bar 100 nm.

During telophase, the mitotic inhibition of ER export is reversed and export of newly synthesised proteins into the Golgi stack resumes. Previous work has shown that reassembly of the Golgi stack precedes ER–Golgi transport (36), suggesting stack reassembly occurs independently of ER export initiation. However, a study using microinjected mutant Sar1 proteins showed that maintenance of the ER export block during telophase was sufficient to prevent delivery of Golgi proteins into the juxtanuclear Golgi of telophase cells from dispersed locations (17), suggesting that Golgi stack reassembly might also be prevented. To test this idea directly, we microinjected Sar1 mutant proteins (GDPSar1 and GTPSar1) into prometaphase/metaphase and after 1.5–2 h observed the structure of the Golgi remnants in telophase/G1 cells by immunofluorescence and electron microscopy. In buffer-injected controls, immunofluorescence showed the Golgi markers ERGIC53, NAGT1 and TGN46 were concentrated in the juxtanuclear region of G1 cells (although ERGIC53 was also partially distributed in the ER as expected). In contrast, in Sar1 mutant-injected cells the staining for these Golgi markers was generally diffuse in the cytoplasm of the cells, and presumably located in the ER (Figure 3). In a minority of microinjected cells, punctate structures in the juxtanuclear region showed weak staining for Golgi markers in addition to the diffuse fluorescence. The structure of the ER, as displayed by calreticulin staining, was not substantially altered under these conditions (Figure 3).

Figure 3.

Figure 3.

Distribution of Golgi proteins in cells microinjected with Sar1 proteins and allowed to exit mitosis. Prometaphase or metaphase cells were microinjected with GDPSar1p or GTPSar1p as indicated and left for 1.5–2 h, fixed and stained for ERGIC53, TGN46, NAGT1 or calreticulin (CALR). Bar 10 µm.

In buffer-injected controls, transmission electron microscopy revealed that the Golgi cisternal stacks of telophase/G1 had reassembled normally (Figure 4a) and were located in the juxtanuclear region (not shown). In Sar1p mutant- injected cells, Golgi stacks were absent but vesiculotubular aggregates were observed in the cells (Figure 4b,c). Serial section analysis revealed that the absence of stacks was not an artifact of the sectioning procedure; and stereological analysis of single telophase/G1 pairs showed GDPSar1p and GTPSar1p microinjected cells contained less than 2% of the Golgi cisternal membrane surface found in buffer-injected cells. In microinjected cells the tubulovesicular aggregates were situated predominantly in the juxtanuclear region close to the midbody, i.e. in the same location at which reassembling Golgi stacks congregate in buffer-injected or noninjected cells, and in a similar distribution to the weakly stained NAGT1 and TGN 46 positive structures found in some cells by immunofluorescence. Stereology showed the vesiculotubular aggregates contained 46 and 58% of the Golgi volume of noninjected cells, for GTPSar1p and GDPSar1p, respectively.

Figure 4.

Figure 4.

Sar1 mutants prevent cisternal stack reassembly during telophase/G1 cells. Prometaphase or metaphase cells were microinjected with GDPSar1p or GTPSar1p and after 1.5–2 h were fixed, embedded and those cells that had progressed to telophase/G1 were sectioned for transmission electron microscopy. (a) Buffer injection (b) GDPSar1p (c) GTPSar1p. Bars 100 nm.

There were distinct differences between the morphology of the vesiculotubular aggregates generated by the different Sar1 mutants in the telophase/G1 cells. In GDPSar1p-injected cells, putative Golgi remnants were composed largely of lucent vacuoles of approximately 100 nm in diameter and the ER of these cells lacked detectable ER buds (Figure 4b). This was consistent with an early inhibition of COPII recruitment, combined with continued processing of the COPII vesicle pool. In contrast, GTPSar1p-injected cells contained vesiculotubular aggregates composed of numerous small 50–60-nm-coated vesicles characteristically associated with bud profiles on the ER and a small proportion of larger lucent vacuoles (Figure 4c). These observations are consistent with a continued COPII vesicle budding combined with arrested uncoating. In this case, smaller accumulations of 50–60 nm vesicles were also found in the periphery of microinjected cells and probably represent budding activity at peripheral ER exit sites.

ER buds are depleted in mitotic cells

To investigate the stage at which ER export is inhibited, we quantitated ER buds and exit sites and studied the distribution of COPII proteins in metaphase and telophase cells, using quantitative electron microscopy and immunofluorescence. Inhibition at a single early step would lead to a reduction in budding profiles and budded COPII vesicles, and also a reduction in membrane-associated COPII. Inhibition at a later step would lead to accumulation of early budding stages, combined with depletion of later stages of COPII vesicle budding and processing.

In initial experiments in HeLa cells, the numbers of ER buds in mitotic and telophase/G1 cells were counted in cells stained using prolonged osmium tetroxide treatment to reveal the endoplasmic reticulum and exit site ER buds (33). This analysis showed that the number of ER buds in metaphase cells was very greatly reduced compared to telophase/G1 cells (Figure 5A). As shown in Figure 5B, the density of buds over the ER was also less in metaphase cells, indicating that the small number of buds counted was not due to a reduction in the amount of ER in mitotic cells. In fact, the total volume of ER in metaphase cells was within 5% of that estimated for telophase/G1 cells (results not shown).

Figure 5.

Figure 5.

Numbers of ER buds are reduced in metaphase compared to telophase/G1.(A) ER buds were counted in complete serial section series from mitotic and telophase/G1 cells exposed to prolonged treatment with osmium tetroxide (see Materials and Methods). (B) The number of buds was related to the ER volume estimated stereologically (see Materials and Methods). n = 3 cells, bars standard error of the mean (SEM).

It is possible that in osmium-stained preparations very short buds may have been difficult to identify, and so we also examined more limited serial section samples of epoxy resin-embedded metaphase and telophase cells. Again, the estimates of ER bud number were substantially lower than in telophase/G1 cells, suggesting there is a marked inhibition of bud formation in mitosis (data not shown). Interestingly, by this method, short buds with a length substantially shorter than their width, were not observed (see below).

Depletion of ER exit sites and redistribution of COPII to the cytosol in mitotic cells

In interphase cells, ER exit sites are characterised by clusters of ER buds distributed on one or more closely related ER cisternae (26). The mechanisms which lead to clustering of budding activity are unknown, but the degree of clustering may influence Golgi stack formation (25). We counted the number of exit site regions using serial sectioning of material exposed to prolonged osmium tetroxide treatment, defining an exit site as one or more (clustered) ER buds (see Materials and Methods). We found that the number of exit sites in metaphase cells was about half that of cells in telophase/G1 (Figure 6A). By combining the counts of exit sites and ER buds, the number of ER buds per exit site was calculated to be 1.68 (SEM 0.339 n = 3 cells) in metaphase and 3.28 (SEM 0.506 n = 3 cells) in telophase/G1. This ratio was also estimated on limited serial section series of conventionally processed material, yielding 1.182 buds per exit site for metaphase (n = 91 buds counted) and 1.95 buds per exit site for telophase/G1 (n = 156 buds counted). Thus by each method there was a 1.65–1.95-fold increase in the number of buds per exit site as cells enter telophase. Further analysis (Figure 7) revealed that a significant proportion of exit sites in telophase/G1 cells (23%) possessed 7–10 buds per exit site, while no exit sites with these numbers of buds were found in metaphase cells. Conversely, the fraction of exit sites with a single bud was 33%, compared to only 7.9% in telophase cells.

Figure 6.

Figure 6.

Numbers of exit sites is reduced in metaphase compared to telophase/G1.(A) The number of exit sites was counted in serial sections of complete metaphase or telophase/G1 cells as described in Materials and Methods. (B) Exit sites were counted using complete Z-series through cells stained using immunofluorescence with rabbit antibodies against Sec13 as described in Materials and Methods. Interphase cells were treated with 10 µg/mL nocodazole for 1 h to disperse exit sites and Golgi elements. In (A) n = 3 cells and in (B) n = 10 cells. Error bars are SEM.

Figure 7.

Figure 7.

Numbers of buds in serially sectioned exit sites. Limited serial section series over 1–2 µm of mitotic and telophase cells were examined and the number of buds counted in exit sites completely within the section series. 91 buds from metaphase and 156 buds from telophase/G1.

The presence of morphologically identifiable buds on the ER might not always reflect the presence of exit sites, especially when budding activity is low. To include exit sites that lacked buds but were associated with COPII-coated vesicles, we stained cells with a Sec13p antibody for immunofluorescence. Sec13p is a well-characterised subunit of the COPII coat and by immunoelectron microscopy is found at ER exit sites with little staining of other intracellular organelles (not shown, see ref. 25,37,38). This marker has recently been used to study exit sites in mitotic and interphase cells (4). The number of Sec13 labelled structures was counted in Z series in the confocal microscope and the number of resolvable structures in mitotic cells was compared with that obtained by counting ER budding regions by electron microscopy. The estimates obtained in each case were very similar (Figure 6A,B). We next attempted to compare the number of Sec13 structures in metaphase and telophase cells, but this was difficult because in many telophase cells the Sec13 structures had begun to cluster together, making identification of the structure ‘edges’ difficult. This clustering of exit sites most likely reflects Golgi reassembly in the juxtanuclear region, and is also observed in unsynchronised interphase cells (26). To overcome this problem, we treated cells with 10 µg/mL nocodazole for 1 h at 37 °C (see ref. 9 andMaterials and Methods), which disassembles the microtubules and disperses exit sites from the juxtanuclear region (not shown). The number of exit sites could now be counted more easily, especially in interphase cells, although it is important to note that the degree to which exit sites become dispersed did vary from cell to cell. The results (Figure 6B) show there were nearly twofold more resolvable exit sites in interphase cells compared to mitosis. The number of exit sites in interphase cells was similar to the number detected in telophase/G1 by electron microscopy.

The reduced number of exit sites in mitotic cells might have reflected a fusion of extant exit sites. If this had occurred, then there would have been a detectable increase in the size of exit sites in mitotic cells. To investigate this possibility, we measured the Sec13p positive structures in both mitotic and nocodazole-treated interphase cells (Figure 8). The structures were assigned to groups with median sizes of 0.25 µm, 0.5 µm and 0.75 µm. It is important to note that due to the resolution limits of light microscopy the confocal microscope may visualise groups of COPII buds and vesicles measuring between 50 nm and 250 nm diameter as 0.25 µm structures, a conclusion consistent with our observations of single ER buds lacking associated vesicles in both mitotic and interphase cells using electron microscopy. Despite its imprecision, this type of analysis nevertheless showed the exit sites of mitotic cells to be smaller than in nocodazole-treated interphase cells. Thus, taken together, these data indicate that the extent of exit site assembly is markedly restricted in mitotic cells compared to telophase/G1 or unsynchronised interphase cells.

Figure 8.

Figure 8.

Sec13p containing structures are smaller in metaphase than in interphase cells. Metaphase or interphase cells (treated with nocodazole, see Materials and Methods) were processed for confocal microscopy and fluorescent structures classified according to size.

The observations of reduced size and number of exit sites suggested that there were less COPII components associated with exit sites in mitotic cells, and this was supported by previously published electron microscopy data showing an increased proportion of Sec13p labelling over the cytosol (11) and by a recent GFP study (4). However, it is not known whether other COPII components also show this modified distribution in mitosis. Using a sedimentation assay combined with immunoblotting, we first confirmed that Sec13p was distributed more into a cytosolic fraction (not shown), and then using immunofluorescence and confocal microscopy we assessed the distribution of Sec23p, Sec24p, Sec13p and Sec31p as (Sec31p in Figure 9; Sec23p, Sec24p, Sec13p not shown). In single confocal slices through interphase cells, punctate staining for each of these components was found in perinuclear region and periphery of interphase cells, but diffuse staining attributable to the cytosol was low, which is consistent with the previous studies of these components (4,37–39). In contrast, the staining in mitotic cells was distributed in some punctate structures and there was now a significant increase in the diffuse cytosolic staining for all of these proteins (see below).

Figure 9.

Figure 9.

Immunofluorescence localisation of Sec31p in interphase (A) and mitotic cells (B). Bar 10 µm.

ER export buds of metaphase cells are shorter than in interphase

Our data show that ER export buds are depleted in mitotic cells. However, some buds do persist in significant numbers during the marked inhibition of ER export. These buds could represent structures that have passed through an early stage because the inhibition is leaky, or even structures frozen at all stages of budding. In either case, if the kinetics of each stage of the bud elongation and vesicle scission process were unchanged, then the buds would have identical morphological characteristics to those of telophase and interphase cells. Alternatively, however, if inhibition also acted at a specific stage during the budding process we might detect a skewed distribution of bud lengths. To distinguish these possibilities, we studied the ER export buds in epoxy resin sections of metaphase and telophase/G1 cells and then measured them. The qualitative results are represented in Figure 10. In metaphase cells ER buds were often found as single profiles associated with relatively few vesiculotubular profiles. In general, the buds appeared to have a ‘stumpy’ appearance. Occasionally, buds belonging to distinct profiles of ER were found grouped together (Figure 10C). In telophase/G1 cells, ER exit sites not only contained more ER buds but these appeared individually longer and more varied in length and morphology; they were similar in morphology to those found in unsynchronised interphase HeLa cells. Often these structures appeared bent and/or interconnected to other membrane profiles in the vicinity. The buds were characteristically associated with large accumulations of vesiculotubular profiles and reassembling Golgi cisternal stacks.

Figure 10.

Figure 10.

Morphology of ER exit sites in mitotic and telophase/G1 cells. The buds of mitotic cells (A, B and C) are shorter and more homogeneous in length (arrows) compared to telophase/G1 buds (arrows in D, E and F), of which some show pinching off morphology (D). Bar 200 nm.

Measurement of ER buds in metaphase cells (Figure 11) showed them to be quite homogeneous in length. Most of the buds had a length to width ratio of about unity, and few were shorter than this. By comparison, the ER buds of telophase/G1 cells were longer, and showed a much larger variation in length-to-width ratio. These observations suggest COPII buds spend relatively less budding time in the elongate form during metaphase compared to telophase/G1. This could arise if transport was slowed during the early phases of budding in mitotic cells, in which case pinching off (scission) steps and pools of free COPII should be depleted. In fact, close examination showed very few of the buds in metaphase cells had a pinching off appearance (see Figure 10), while this was a significant feature of telophase ER exit sites (Figure 10D). Furthermore, a number of observations indicate COPII vesicles are depleted in mitotic cells, including: (i) a reduction in the number and size of Sec13p-containing structures; (ii) the reduction in the amount of membrane-associated Sec13p; and (iii) the reduction in the number of juxta-ER vesicles in mitotic cells by electron microscopy (not shown).

Figure 11.

Figure 11.

Length of ER bud profiles in metaphase and telophase/G1 cells. (A) Average lengths. (B) Length/Width ratios. (n = 24 for each value, bars SEM)

Organisation of ER exit sites

The exit sites of interphase cells can be divided into two principal classes: (a) peripheral elements that are not associated with demonstrable Golgi cisternae and generate ER–Golgi intermediate compartment vesicles that are transported to the central Golgi region of the cells; and (b) more central exit sites which are closely associated with Golgi stacks (26,27,32). If vesiculotubular remnants arise by fragmentation (33), then this inhomogeneous distribution might be preserved, whereas if they arise by recycling of Golgi components via the ER (14) the Golgi residents might be more evenly distributed over the vesiculotubular structures. Previous work has shown that over half the tubulovesicular clusters of mitotic cells contain Golgi residents, but their distribution in relation to exit sites has not been assessed in detail (6), and a recent immunofluorescence study was unable to clarify this relationship (4). We therefore colocalised Sec13p and GalNac T2 as representatives of exit sites and GalNacT2, respectively. In interphase cells we found that labelling for GalNacT2 was present mainly over the juxtanuclear Golgi cisternal stacks closely associated with small clusters of vesicles staining for Sec13p (not shown). Clusters of vesicles situated in the periphery of the cells were also labelled for Sec13p, but were virtually free of GalNacT2 labelling (not shown).

In metaphase cells a significant proportion of the vesiculotubular structures were found to be double-labelled for Sec13 and GalNacT2 proteins (72% n = 32). The labelling for these markers was most often located in distinct nonoverlapping regions, characteristically at opposite poles of the aggregate (Figure 12A). Quantitation of the labelling distribution (Figure 13) revealed approximately 80% of the labelling was in clearly segregated regions, staining for Sec13p or GalNacT2, respectively. Interestingly, labelled vesiculotubular aggregates could be classified into two types: one large and centrally located with Sec13p and GalNacT2 labelling (Figures 12A and 14); the other smaller and peripherally located with Sec13p labelling and little GalNacT2 (Figures 12B and 14). Thus these results show that vesiculotubular structures have a markedly heterogeneous distribution of GalNacT2 labelling.

Figure 12.

Figure 12.

Sec13p and GalNacT2 in distinct subregions of tubulovesicular structures in mitotic HeLa cells.(A) Large tubulovesicular structure from the central region of the cell close to the mitotic spindle containing labelling for Sec13p (small gold) and GalNacT2 (large gold). The labelling is segregated in distinct nonoverlapping regions. (B) A small vesiculotubular structure from the cell periphery labels only for Sec13p. Bar 50 nm.

Figure 13.

Figure 13.

Polarisation of Sec13/GalNacT2 labelling in vesiculotubular structures of metaphase cells. Metaphase cells were fixed cryosectioned and double labelled with different sizes of gold particles as described in Materials and Methods. Labelling areas were defined as the polygonal area enclosed by gold particles. Unmixed labelling was defined as no overlap between the areas enclosed by gold particles of each size (n = 32; bars, sem).

Figure 14.

Figure 14.

Relationship between size, location and composition of vesiculotubular structures in mitotic cells. Cryosections were labelled for Sec13p and GalNacT2. In sectioned metaphase cells the area of labelled vesiculotubular structure profiles and their distance from the plasma membrane was estimated as detailed in Materials and Methods. (A) Larger vesiculotubular structures are rich in GalNacT2 labelling while smaller ones are poorly labelled for this antigen. (B) Larger vesiculotubular structures are more distant from the plasma membrane but smaller ones predominate close to the plasma membrane.


In this paper we have examined some key predictions of the recycling model of mitotic Golgi partitioning, in which ER export inhibition traps Golgi residents in a dispersed ER prior to cell division (14). We have previously proposed that the recycling removes subsets of itinerant Golgi-located proteins selectively, a process which induces cisternae to shrink and take on a vesiculotubular form (19). Here, we tested this idea by microinjecting mutants of Sar1, which block transport between the ER and the Golgi stack. When Sar1p mutants were microinjected into interphase HeLa cells, the itinerant ERGIC53 and Golgi residents acquired a dispersed distribution, presumably in ER structures. Crucially, electron microscopy revealed the existence of large vesiculotubular remnants that lacked both cisternae and were depleted of Golgi markers, and displayed similar morphology to mitotic Golgi remnants (see ref. 5). We conclude that blockade of ER export is sufficient to drive disassembly of the Golgi cisternae to form aggregates of vesiculotubular structures, and other published results support this (18). This conclusion was further supported by our experiments in which Sar1p mutants were present during the reinitiation of ER export during telophase. In this case the Sar1 mutant proteins maintained ER export inhibition into telophase/G1 (as evidenced by studies of ERGIC53 distribution) and prevented the reappearance of Golgi cisternae, suggesting ER export supplies components required to rebuild Golgi cisternae. To determine whether these components have recycled from the Golgi or are essential factors newly synthesised during mitosis will require further work. Interestingly, despite the fact that vesiculotubular Golgi remnants mostly lacked the Golgi residents we have studied, they underwent translocation and clustering in the juxtanuclear region close to the midbody. It is therefore likely these remnants still contain the Golgi components needed for interaction with, and movement on, microtubules during telophase (40). Certainly, recent experiments support the view that certain Golgi matrix components can be separated from functional stack components by Golgi–ER recycling pathways (18), although it is worth pointing out that in our hands one of these proteins (GM130) became dispersed from the juxtanuclear region after microinjection of Sar1 mutants. It will therefore be of interest to characterise the remnants further.

A second prediction of the recycling hypothesis is that the ER export block would be positioned at an early stage. Our results show, for the first time, a marked reduction in the number of budding profiles on the ER of mitotic cells and a reduction in the number and size of exit sites. These observations therefore indicate a significant limitation on the assembly of budding profiles and the assembly of exit sites in metaphase cells. They are consistent with previous studies that demonstrated dramatic slowing of export from the endoplasmic reticulum (11,15) and show that mitosis acts at a stage which precedes the formation of buds on the ER membrane. Bud formation is known to correlate with the extent of COPII coat assembly and in vitro studies have revealed that Sar1/Sec23/24 can bind cargo in the absence of significant budding activity (29). Subsequent recruitment of Sec13/31 is needed for completion of the coat and for bud formation and possibly vesicle scission. Therefore the inhibition of COPII bud formation we have documented here is likely to act prior to addition of Sec13/31 complex to the coat. This conclusion is consistent with our data showing a redistribution of Sec31p and Sec13p from membrane to the cytosol by immunoelectron microscopy (this study; 11), sedimentation analysis and the immunofluorescence (this study and 4). Additional evidence for an early acting limitation in the assembly of prebudding COPII complexes is also indicated by our finding that Sec23p and Sec24p redistribute into the cytosol of mitotic cells.

The mechanism of this early block in COPII mitotic assembly is at present unclear, but our results do now suggest a number of possibilities. One is that mitosis produces post-translational modifications that abrogate the ability of this sec23/24 complex to bind cargo and facilitate coat assembly. Another possibility is that mitosis regulates the guanine nucleotide exchange factor Sec12p thereby restricting Sar1p recruitment and coat formation. Sec12p has yet to be characterised in the animal cells. Finally, a regulation based on Sec16p should also be considered because it could produce inhibition at the multiple levels suggested by our morphological budding data. This large hydrophilic multidomain protein interacts genetically with all components of the coat, except Sar1p. It may function as a foundation for construction of the COPII coat from soluble protein complexes by stabilising the interaction between the coat components (41). A modulation in Sec16p function could conceivably destabilise COPII coats and drive their dissociation. Further work is now required to identify this important mechanism.

Despite the low number of buds on the ER of metaphase cells, some buds were nevertheless present and these were shorter than those found in telophase cells. One explanation for these results might be a low level of continued budding activity in mitotic cells (15). A slowing, rather than complete inhibition, of forward traffic could still generate the observed membrane depletion in post-ER compartments if it were combined with efficient retrograde traffic. One way to test for such a ‘leaky’ transport arrest would be to microinject the dominant negative mutant GTPSar1 and look for accumulation of COPII vesicles in mitotic cells. However, low levels of ongoing export do not explain the difference in bud length in between metaphase and telophase cells. One possibility is that elongated buds of telophase could reflect a relative retardation of later elongation/scission steps, perhaps because these steps are overloaded with cargo accumulated in the ER of mitotic cells (see ref. 42). On the other hand, shortened buds of mitosis might result from inhibited budding both prior to, and also during, early bud formation or COPII vesicle production. Multiple levels of inhibition have been suggested in arrested coated pit invagination at the plasma membrane in mitotic cells (43,44). It will therefore be of interest to determine the kinetics of each budding stage in mitotic cells.

Finally, we have tried to distinguish pathways that generate multiple copies of mitotic vesiculotubular Golgi remnants. In one model, Golgi components recycle via the ER and emerge through dispersed ER exit sites of mitotic cells (14), a model with parallels in nocodazole-induced dispersion of Golgi elements in interphase cells (45). In this case, Golgi components would become homogenised in the ER and emerge to become evenly distributed over the ER exit sites of mitotic cells. In another model, recycling is selective and the vesiculotubular remnants remain independent of the ER and arise by fragmentation of existing Golgi structures. Here, the expectation is that the distribution of Golgi components over exit sites is likely to be just as markedly heterogeneous as found in interphase cells. By localising exit site (Sec13p) and a slow recycling Golgi stack markers (GalNacT2, ref. 11) by immunoelectron microscopy, we detected a very large variation in the composition of vesiculotubular structures. Those vesiculotubular structures in the periphery of mitotic cells were small, poor in GalNacT2 and relatively rich in Sec13p, while more centrally located vesiculotubular structures were large and rich in GalNacT2. These distributions have striking parallels with those found in interphase cells and support a fragmentation mechanism rather than a recycling model for generating the vesiculotubular structures of mitotic cells. They further suggest that positional specialisation of exit sites may be maintained from interphase into mitosis. It is worth noting that our observation of an early prebudding arrest of ER export in mitosis is inconsistent with dispersal by recycling.

Taken together, our results suggest a model for Golgi partitioning involving both fragmentation and recycling mechanisms. ER export arrest allows decanting and partitioning of more rapidly recycling Golgi residents into the ER and induces transformation of cisternal stack into vesiculotubular aggregates containing nonrecycling residents. Multiple vesiculotubular aggregates (and possibly dispersed vesicles) may be generated by fragmentation of residual Golgi elements.

Materials and Methods

Electron microscopy

Mitotic Hela cells were isolated and processed for prolonged osmication or conventional epoxy resin embedding and serial sectioning at 100–200 nm, essentially as previously described (5,6,33). Fixation for epoxy resin sectioning was for 30 min in 0.5% glutaraldehyde in either 0.2 mm cacodylate buffer pH 7.4 or a mixture of 5% w/v sucrose in 0.1 mm cacodylate buffer pH 7.4 at 37 °C. Some preparations were fixed by adding an equal volume of double-strength fixative in buffer carrier to the cell suspension. Prolonged osmication produces electron-dense reaction product predominantly within ER cisternae (33) present in the periphery of mitotic cells, and cisternae at this location also labelled for ER markers such as ERGIC53 and calreticulin using immunoelectron microscopy on ultrathin cryosections and immunofluorescence (unpublished results). We defined ER buds as any protrusions with a diameter of 40–60 nm which were visualised on micrographs at a final magnification of at least 20 000×. To estimate the density of buds on the ER, the number of buds counted was related to the ER volume estimated using point counting with a square lattice grid of 1 cm spacing on micrographs taken at 2000–3000×. This analysis was carried out on all the serial sections through each cell. ER buds were assigned to ER exit sites if they were within 0.6 µm of the nearest bud (26). For electron microscopic study after microinjection cells were seeded onto CELLocate coverslips (Eppendorf UK, Cambridge, UK) and following fixation in 4% paraformaldehyde in PBS, the location of the injected cells on the locator grid was recorded using a CCD camera (Digital Pixel, Falmer, Brighton, UK) on a Zeiss Axioplan (Carl Zeiss Ltd, Welwyn Garden City, Herts, UK) microscope. The cells were then postfixed with 1% glutaraldehyde followed by 1% OsO4/1.5% potassium ferrocyanide and dehydrated and embedded in Durcupan resin. The CELLocate grid transferred to the resin, and combined phase and fluorescence CCD images were used to localise the cell pairs prior to trimming and sectioning. The region of injected cells was serially sectioned and sections collected on slot grids were contrasted with uranyl acetate and lead citrate. Stereology on systematic series of sections was performed essentially as described in Lucocq et al. 1989 (6) with the spacing between sections adjusted to produce at least 8 sections per telophase/G1 pair of cells.

Throughout this study we used the following definitions for mitotic stages in the transmission electron microscope: Metaphase – equatorial band of chromosomes; telophase/G1 – reformed nuclear envelope, elongated nuclear profile and deep cleavage furrow (definitive evidence for the presence of a midbody and distinction between telophase and G1 was not always achievable).

To analyse the composition of the exit site regions, cryosections of metaphase cells were double-labelled using antibodies against Sec13 and mouse monoclonal antibodies against GalNacT2, essentially as described in Prescott et al. (46), and clusters of vesicles photographed at 15 000× and the number of gold particles of each size counted. The size of vesiculotubular aggregates, labelled for Sec13p and or GalNacT2, was estimated using point counting with a square lattice grid with line spacing of 1.775 mm. On low magnification micrographs (2000–3000), the shortest distance from the centre of the vesiculotubular aggregate to the plasma membrane was measured.

Purification of GDPSar1 mutant proteins

The Sar1 T39N (GDP-bound) and H79G (GTP-bound) mutants have been expressed as histidine-tagged proteins in E. coli and described previously (24,47). Bacterial cultures were grown to mid-log phase and induced with 0.75 mm IPTG for 3 h. Cells were pelleted and lysed by sonication in buffer A (50 mm Tris pH 8.0, 100 mm NaCl, 5 mm MgCl2, 1 mm EGTA, 1 mm PMSF, 10 mm b-mercaptoethanol, 0.1 mg/mL lysozyme). Cell debris was removed by centrifugation at 100 000g for 30 min and supernatant was adjusted to 300 mm NaCl and 10 mm MgCl2. 1 mL of Ni-NTA resin (Qiagen) was added to each litre equivalent of bacterial culture and incubated on a rotating wheel for 2 h. Resin containing bound histidine-tagged Sar1 was washed extensively with buffer B (50 mm MES pH 6.0, 300 mm NaCl, 5 mm MgCl2, 1 mm EGTA, 1 mm PMSF, 10 mm b-mercaptoethanol). Proteins were eluted in buffer B containing 250 mm imidazole. Eluted protein was dialysed extensively against buffer C (25 mm HEPES pH 7.2, 125 mm potassium acetate, 5 mm MgCl2, 1 mm EGTA, 1 mm glutathione), snap frozen in liquid nitrogen in small aliquots at a protein concentration of 1.5–2 mg/mL and stored at − 20 °C.

Sedimentation assay

Interphase and mitotic cells were homogenised and postnuclear supernatants prepared as described previously (11). Postnuclear supernatants were centrifuged at 100 000g for 1 h and supernatants were lysed in 1% tritonX100 and immunoblotted for Sec13p as previously described (11).


HeLa cells (or HeLa cells stably expressing myc tagged N-acetylglucosaminyltransferase I (NAGT1)) (48) were seeded onto coverslips 24 h before microinjection and injected with either GTP-or GDP-restricted Sar1 mutant proteins or buffer only each containing 1 mg/mL FITC conjugated dextran as a cell marker. To ensure these proteins were present during telophase cells in early mitosis (pro-metaphase to metaphase as judged by phase imaging) were injected and left at 37 °C for 1.5–2 h after microinjection to allow cells to pass through telophase/G1. The cells were then fixed with 4% paraformaldehyde in PBS and processed as detailed above. Interphase cells were microinjected, incubated and processed by the same protocol.

Fluorescence microscopy

Sec13 (38), Sec23A (49), Sec24C and Sec31 (50) proteins were localised by confocal immunofluorescence in monolayer HeLa cells, fixed either in methanol at − 20 °C or formaldehyde, using rabbit antibodies to human proteins and Texas red conjugated antibodies, essentially as described in Farmaki et al. (11). For microinjection, cell pairs were found using the injected FITC-dextran fluorescence and imaged using a Zeiss LSM-410 confocal microscope. Antibodies to ER–Golgi intermediate compartment (ERGIC53; kindly provided by Hans-Peter Hauri, Basel), ER marker (calreticulin; Sressgen, UK), galactosyltransferase, GM130 (a gift from Martin Lowe, Manchester, UK) or myc on myc-tagged NAGT1(9E10) were localised using secondary antibodies conjugated to Texas Red and Cy5 (Jackson Immunochemicals, Luton, UK) so that both compartments could be assessed in the same injected cells. Final images were colour separated in Adobe Photoshop. To disperse exit sites, HeLa cells monolayers were treated with 10 µg/mL nocodazole for 1–2 h and fixed for immunofluorescence microscopy. Sec13p positive structures (primary magnification 1000×) were measured on slices taken at 2-µm intervals through mitotic cells on untreated monolayers or through interphase cells on nocodazole-treated cultures.


We thank W.E. Balch for the Sar1 plasmids, T. Nilsson and G. Warren for the myc-NAGTI expressing cells, A. Lamond for use of confocal microscope facilities, Henrik Clausen (Copenhagen, Denmark) and Tatsuo Suganuma (Miyazaki, Japan) for antibodies against GalNacT2 and galactosyltransferase, respectively. JML was supported by a Research Leave Fellowship from the Wellcome Trust (059767/Z/99/Z) and by Tenovus Scotland. We are grateful to Liz Smythe for critical comments on the manuscript.