Mitotic Golgi Partitioning Is Driven by the Membrane-Fissioning Protein CtBP3/BARS

See allHide authors and affiliations

Science  02 Jul 2004:
Vol. 305, Issue 5680, pp. 93-96
DOI: 10.1126/science.1097775

This article has a correction. Please see:


Organelle inheritance is an essential feature of all eukaryotic cells. As with other organelles, the Golgi complex partitions between daughter cells through the fission of its membranes into numerous tubulovesicular fragments. We found that the protein CtBP3/BARS (BARS) was responsible for driving the fission of Golgi membranes during mitosis in vivo. Moreover, by in vitro analysis, we identified two stages of this Golgi fragmentation process: disassembly of the Golgi stacks into a tubular network, and BARS-dependent fission of these tubules. Finally, this BARS-induced fission of Golgi membranes controlled the G2-to-prophase transition of the cell cycle, and hence cell division.

Fission and fusion of biological membranes are essential aspects of intracellular transport and organelle inheritance (13). During mitosis, the Golgi complex undergoes extensive fragmentation that begins in prophase and initially involves the severing of tubules connecting adjacent stacks (4). The isolated Golgi stacks are then shortened and transformed into tubular networks that fragment into clusters of vesicles and tubules (4). This fragmentation finally leads to Golgi membranes that are dispersed into isolated vesicles and tubulovesicular clusters (2) and, possibly depending on the cell type, may (5, 6) or may not (7, 8) be redistributed into the endoplasmic reticulum.

To examine whether CtBP3/BARS [C-terminus binding protein 3/brefeldin A (BFA) adenosine diphosphate–ribosylated substrate (9)] is involved in mitotic fragmentation of the Golgi, we used a well-established assay that reconstitutes this process in permeabilized normal rat kidney (NRK) cells incubated with mitotic cytosol (10). Mitotic cytosol induces a breakup of the Golgi ribbon into tubulovesicular clusters, which then disperse throughout the cell, similar to the Golgi clusters characteristic of prometaphase in intact cells (11). Immunodepletion of BARS inhibited Golgi fragmentation by more than 80%, and the addition of recombinant BARS to depleted extracts completely restored their fragmentation activity (Fig. 1, A to C) (fig. S1). The p50-2 antibody to BARS inhibited fragmentation by more than 75% (Fig. 1D); the same was true for the isolated Fab fragment (fig. S2).

Fig. 1.

BARS and mitotic fragmentation of the Golgi (see also figs. S1 and S2). (A) Immunoblot of mitotic extract (ME) and mock- and BARS-depleted cytosols. Upper panel: p50-2 antibody to BARS. Lower panel: antibody to α-tubulin. (B) Digitonin-permeabilized NRK cells incubated with interphase extract (IE), ME, or BARS-depleted ME in the absence (depleted) or presence (depleted + BARS) of 10 μg of recombinant BARS. Bright areas show immunofluorescence (antibody to giantin) of Golgi membrane organization. Scale bars, 10 μm. (C to F) Quantification of Golgi fragmentation index (14) in digitonin-permeabilized NRK cells, as in (B). (C) ME, mock-depleted ME (mock), or BARS-depleted ME without (-) or with (BARS) 10 μg of recombinant BARS. (D) ME alone or with 10 μg of generic IgGs (IgG) or 10 μg of p50-2 (anti-BARS). (E) ME alone or with 5 μg of GST, 5 μg of SBD, or 5 μg of SBD plus 10 μg of recombinant BARS. (F) As (E), but with 5 μg of NBD instead of SBD. In (C) to (F), data represent means ± SD from at least three independent experiments; 200 cells per experiment.

BARS is structurally similar to a family of D-hydroxyacid dehydrogenases and is composed of two separated domains: the NAD-binding domain (NBD; amino acids 113 to 308) and the “substrate”-binding domain (SBD; amino acids 1 to 112 and 309 to 430) (12). We reasoned that these domains could represent nonfunctional BARS “deletion mutants” that, when present in molar excess with respect to endogenous BARS, could act as dominant negative (DN) mutants. Indeed, the addition of SBD or NBD to the assay inhibited Golgi fragmentation by at least 55% and 75%, respectively (Fig. 1, E and F). This inhibition was due to interference with endogenous BARS, as it was completely overcome by addition of recombinant BARS (Fig. 1, E and F). BARS per se (i.e., without mitotic cytosol), or when added together with interphase cytosol, had no visible effects; these findings indicate that BARS is necessary but not sufficient for mitotic fragmentation of the Golgi (13).

We investigated the precise role of BARS in the fragmentation process by electron microscopy (11, 14). In permeabilized NRK cells treated with interphase cytosol, the Golgi maintained a normal stacked organization (Fig. 2A), whereas with mitotic cytosol alone the Golgi was fragmented into small tubulovesicular elements (∼0.5 μm in diameter) that were dispersed throughout the cytoplasm (Fig. 2A) (fig. S3) (11, 14). In contrast, when NBD was included with mitotic cytosol, the Golgi was not fragmented but lost its typical stacked organization and was transformed into groups of large tubular-vesicular-saccular networks (∼1.2 μm in diameter) in the perinuclear area (Fig. 2A) (fig. S3). These networks appeared interconnected in serial thin sections, which suggests that they represent Golgi stacks that have undergone tubulation but have preserved their continuities (13). Similar results were observed when BARS was inhibited with p50-2 (13). Thus, mitotic cytosol contained two activities that could be separated experimentally by inhibiting BARS: the disassembly/tubulation of the Golgi stacks, and the fragmentation of this tubular network mediated by BARS.

Fig. 2.

BARS is necessary for fragmentation of Golgi membranes (see also fig. S3). (A) Digitonin-permeabilized NRK cells were treated and fixed for immunoelectron microscopy (antibody to mannosidase II; horseradish peroxidase staining). Thin section: a Golgi stack after IE treatment, a small tubulovesicular cluster after ME treatment, and a large tubulosaccular network after treatment with ME plus 5 μg of NBD (ME + NBD). Scale bar, 500 nm. Lower right panel: percentage quantifications of these Golgi membrane forms (stacked, dispersed, tubulated, undefined) (14). About 200 cells were analyzed under each condition in three independent experiments. (B) Negative staining (14) of isolated rat liver Golgi membranes after incubation for 20 min with IE, ME, or BARS-depleted ME (2 mg/ml each) in the absence (depleted) or presence (depleted + BARS) of recombinant BARS (5 μg/ml). Notations: compact tubular network (white asterisk), fenestrae (black arrowheads), vesicular profiles (white arrowheads), tubules (arrows), and fission sites (pointed arrowheads). Images are representative of three independent experiments. Scale bar, 500 nm. Lower left panel: whole length of the tubule starting in the framed area of the “depleted” sample. Blind quantification is shown in table S1.

We next made use of an assay designed for the analysis of the detailed three-dimensional ultrastructure of isolated Golgi membranes (15). Rat liver Golgi membranes were incubated for 20 min with interphase or mitotic cytosol, or with BARS-depleted mitotic cytosol in the absence or presence of recombinant BARS. Golgi membranes incubated with interphase cytosol appeared as groups of “normal” disk-like cisternae that were fenestrated at the rims and often associated with tubular-reticular networks (Fig. 2B). This incubation did not cause obvious changes in morphology [although very long incubation times did induce a partial fragmentation of the Golgi tubules (15)]. In contrast, incubation with mitotic cytosol for 20 min caused an extensive disruption of both the tubular and, more important, the cisternal Golgi membranes, which were transformed into clusters of vesicles of various diameters and included tubules exhibiting characteristic constrictions at specific sites [“fission intermediates” (15)] (Fig. 2B). Instead, BARS-depleted mitotic cytosol converted the Golgi membranes (including the cisternae) into a large network of smooth, and sometimes long, tubules that did not generate vesicles or present “fission intermediates” (Fig. 2B). Finally, when recombinant BARS was added to this depleted cytosol, the fragmentation of the tubular membranes was completely restored, with the formation of vesicles and “fission intermediates” (Fig. 2B) (table S1). These effects differed markedly from those previously described for BARS-enriched interphase cytosol, which fragmented the Golgi tubules but left the cisternae intact (15).

We then examined the role of BARS in living cells undergoing mitosis. NRK cells were blocked at the beginning of S phase by incubation with the DNA polymerase inhibitor aphidicolin (16). After aphidicolin washout, they were microinjected with either p50-2 or the DN BARS mutant and were then fixed at various times after injection (14). The number of control [immunoglobulin G (IgG)–injected] cells in mitosis increased with time, peaking 8 hours after aphidicolin washout. In these mitotic cells, the Golgi apparatus presented the typical fragmented and dispersed distribution (Fig. 3A). In contrast, cells injected with p50-2 (or the Fab fragment; fig. S2) presented two striking alterations: The number of cells undergoing mitosis was reduced by 75% (Fig. 3B), and in the few cells that did enter mitosis, the Golgi complex was not fragmented but instead appeared to be divided into two large aggregates located at the two poles of the mitotic spindle (Fig. 3C), reminiscent of similar findings (17). Injection of the DN BARS mutant SBD caused similar effects (Fig. 3D). These effects could be reversed by coinjecting BARS at the same concentrations, because the Golgi fragmented and the cells entered mitosis normally (Fig. 3D). Thus, BARS was required for mitotic fragmentation of the Golgi in vivo, and entry into mitosis was linked to this BARS-induced Golgi fragmentation.

Fig. 3.

BARS is required for Golgi fragmentation and mitotic progression (fig. S2). S phase–arrested NRK cells after aphidicolin washout (14) are shown. (A and C) After 1 hour, cells were injected with generic IgG (5 mg/ml) (A) or p50-2 (5 mg/ml) (C); after 7 hours, they were labeled for immunofluorescence of Golgi morphology (antibody to giantin; upper panels), injected cells (goat antibody to rabbit IgG; not shown), and cell cycle phase (Hoechst 33258; lower panels): interphase (left), metaphase (middle), and anaphase (right). Scale bars, 10 μm. (B and D) Quantification of relative mitotic index. (B) Cells were injected as in (A) and (C). After 3.5 hours, they were treated with buffer (nontreated), nocodazole (500 ng/ml), or BFA (200 ng/ml) for an additional 3.5 hours and processed as in (A) and (C). (D) Cells were injected with GST (1 mg/ml) or with SBD (1 mg/ml) without or with recombinant BARS (2 mg/ml) and were processed as in (A) and (C). In (B) and (D), data represent means ± SD from three independent experiments; 200 cells per experiment. Relative mitotic index: percentage ratio of mitotic injected cells to mitotic noninjected cells from the same cover slip.

To verify the link between entry into mitosis and Golgi fragmentation, we tried to bypass the block caused by p50-2 by forcing the dispersion of the Golgi membranes with either the microtubule disrupting drug nocodazole, which disperses the Golgi ribbon into ministacks (18), or BFA, which induces disassembly of the Golgi complex and causes it to fuse with the endoplasmic reticulum (19). Indeed, in cells injected with p50-2, both nocodazole and BFA completely restored mitotic progression (Fig. 3B), indicating that the prevention of entry into mitosis caused by p50-2 or by SBD was due to the block of Golgi dispersal. Blocking Golgi fragmentation with antibodies to the Golgi-stacking protein GRASP-65 also prevents entry into mitosis (16).

To further characterize the role of BARS in mitosis in vivo, we treated NRK cells with antisense BARS oligonucleotides (14). These reduced the levels of endogenous BARS by more than 80% (Fig. 4A). Treated cells grew more slowly, were larger and flatter but viable, and had a well-spread and highly tubulated perinuclear Golgi (Fig. 4C). When examined for alterations in the cell cycle, most of these cells (>70%) were arrested in G2, as shown by the immunofluorescence pattern of antibodies to phosphorylated histones H1 and H3 (Fig. 4B). When these cells were injected with recombinant BARS, the Golgi morphology was rapidly restored and the cells resumed mitotic progression (Fig. 4C). Furthermore, when they were treated with nocodazole or BFA, their Golgi became dispersed and they entered mitosis 5 to 6 hours after drug addition (Fig. 4D), albeit less efficiently than with the BARS injection.

Fig. 4.

BARS depletion by antisense oligonucleotides results in G2 arrest. (A) Immunoblot for endogenous BARS and tubulin in total cell lysates from untreated NRK cells (control) or cells treated with scrambled or antisense oligonucleotides. Upper panel, p50-2; lower panel, antibody to α-tubulin. (B) Labeling of cells with antibodies to phosphorylated histones H1 and H3. Scrambled oligonucleotide treatment (5 days); most cells show diffuse labeling typical of G1. Antisense oligonucleotide treatment; >70% of cells show punctuate nuclei typical of G2. (C) Golgi morphology (antibody to giantin) after scrambled oligonucleotides or antisense oligonucleotides without (antisense) or with injection of recombinant BARS (2 mg/ml) for 6 hours (antisense + BARS). Lower right panel: quantification of mitotic index relative to untreated control cells (C). Scale bars in (B) and (C), 10 μm. (D) Quantification of mitotic index relative to “scrambled”/“nontreated” cells. Cells as in (B); scrambled and antisense oligonucleotide treatments in absence and presence of nocodazole (500 ng/ml) or BFA (5 μg/ml) (final 6 hours of oligonucleotide treatment); 1000 cells counted per experimental condition.

In conclusion, we find two activities in mitotic cytosol: one resulting in the transformation of the Golgi stacks into tubular networks, the other resulting in the BARS-dependent fission/fragmentation of these networks. This agrees with, and partly explains, previous morphological observations in vivo (4). We propose that the initial Golgi tubuloreticular transformation is due to the loss of structural restrictions imposed by the “Golgi matrix” proteins, such as GRASP-65 and GM130, which are phosphorylated during mitosis (2023). The function(s) of this tubular disassembly may be to facilitate the effect of BARS by providing a more suitable geometry for fission than that of a Golgi stack, and/or to expose BARS receptors or activators in the Golgi membranes. The mechanisms of the BARS-dependent fission, and whether BARS is activated specifically during mitosis, remain to be elucidated. We find that BARS is active also in the fission of transport carriers during interphase traffic (24), but this activity may be enhanced by mitotic phosphorylation (25). Also of note, BARS has a slow acyltransferase (ACT) activity that has been associated with fission (15); however, a BARS point mutant (G172) devoid of ACT activity still induced fission, albeit with a potency lower than that of native BARS by a factor of ∼20 (fig. S1), which suggests that the ACT activity may have a facilitating rather than a necessary role.

Finally, the observation that BARS-induced Golgi fission is required for the G2-prophase transition supports the generality of Golgi fragmentation as a regulator of the cell cycle, possibly through the centrosome (16). This suggests that drugs acting on BARS (26) should have the potential to control the G2-prophase transition stage of this cycle.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article