In eukaryotic cells, the completion of cytokinesis is dependent on membrane trafficking events to deliver membrane to the site of abscission. Golgi and recycling endosomal-derived proteins are required for the terminal stages of cytokinesis. Recently, protein subunits of the ESCRT (endosomal sorting complexes required for transport) that are normally involved in late endosome to lysosome trafficking have also been implicated in abscission. Here, we report that a subunit, CHMP3 (charged multivesicular body protein-3), of ESCRT-III localizes at the midbody. Deletion of the C-terminal autoinhibitory domain of CHMP3 inhibits cytokinesis. At the midbody, CHMP3 does not co-localize with Rab11, suggesting that it is not present on recycling endosomes. These results combined provide compelling evidence that proteins involved in late endosomal function are necessary for the end stages of cytokinesis.
Proteins that make up the ESCRT (endosomal sorting complexes required for transport) are involved in the sorting and trafficking of membrane proteins into MVBs (multivesicular bodies). Current models of ESCRT function place the ESCRT proteins into three complexes, ESCRT-I, -II and -III (reviewed in [1–4]). ESCRT-I is thought to be involved in the recognition of ubiquitinylated cargo membrane proteins that are to be sorted into MVBs. ESCRT-II and -III have been implicated in further protein sorting and invagination of the endosomal membrane away from the cytoplasm to form MVBs. Once assembled on membranes, removal of ESCRT proteins by an AAA (ATPase associated with various cellular activities) ATPase called Vps4 (vacuolar protein sorting 4) is required in order for ESCRT proteins to carry out multiple rounds of sorting . The latter stages of MVB formation are not currently well understood; however, it is known that the ultimate fate of cargo proteins that are sorted into MVBs is usually degradation in lysosomes. Perturbing the function of the ESCRT machinery in mammalian cells by protein knockdown or expression of dominant-negative proteins such as an ATPase defective Vps4 (Vps4E235Q) results in intracellular accumulation of cargo proteins that fail to be degraded in lysosomes [6–8].
In addition to their role in sorting of membrane proteins for destruction in lysosomes, ESCRT proteins have been implicated in membrane virus budding (reviewed in ), mRNA trafficking  and cytokinesis [11–13]. The involvement of ESCRT proteins in cytokinesis is particularly interesting as evidence is accumulating from model organisms that endocytosis and many proteins involved in endocytic pathways such as dynamin, clathrin and Rab11 are also essential for the successful completion of cytokinesis [14–18].
Cytokinesis is the separation of one cell into two daughter cells following mitosis. During the final stages of cytokinesis in animal cells, a midbody is formed, which is a thin membrane tether-like structure connecting the two daughter cells together. The end of cytokinesis is typified by the abscission of this structure at the midbody, which is an event heavily reliant on membrane dynamics. A model for abscission that seems to unify the seemingly different mechanisms of cytokinesis in animals and plants involves the trafficking of membrane vesicles to the midbody and phragmoplast respectively [14,19,20]. At the midbody, in animal cells, these vesicles have been suggested to fuse homotypically and heterotypically with the plasma membrane, causing abscission at this region and separation of the two daughter cells . The source of the membrane vesicles required for abscission is seemingly complex, as proteins usually associated with Golgi, early stages of endocytosis and endosomal recycling all appear to be important in the process . The recent discovery that ESCRT proteins, which are generally considered as functioning in late endosomal membrane trafficking, are required for cytokinesis adds additional complexity to the source of membranes present at the site of abscission. An ESCRT-I protein (TSG101), an ESCRT-related protein (Alix) and ESCRT-III proteins [CHMP (charged MVB protein) 2, CHMP4 and CHMP5] are all present at the midbody during the final stages of cytokinesis in mammalian cells [11,12]. Furthermore, interfering with the expression, by knockdown or overexpression, of TSG101, Alix and other ESCRT proteins results in impaired cytokinesis [11,12]. Thus it seems that a functional ESCRT machinery is required for the late stages of cytokinesis.
Following on from this work, we wanted to investigate whether the ESCRT-III protein CHMP3 is present at the midbody and is functionally required for cytokinesis in animal cells. To address this point we utilized a dominant-negative derivative of CHMP3, an ESCRT-III protein. We and others have shown that full-length CHMP3–FLAG and CHMP3–GFP (green fluorescent protein)-fusion proteins are cytosolic, do not noticeably affect endosome morphology and do not inhibit membrane trafficking to lysosomes or HIV particle production when expressed in cultured mammalian cells [22–24]. However, truncated CHMP3-fusion proteins, with the C-terminal autoinhibitory domain removed, become membrane-associated and act as dominant-negative proteins in that they dramatically alter endosome morphology, prevent trafficking to the lysosome and inhibit virus budding [22–25]. We now show that as well as perturbing endocytic trafficking to the lysosome a dominant-negative CHMP3 derivative (CHMP31–179–GFP) localizes to the midbody of dividing mammalian cells and dramatically inhibits cytokinesis.
DNA manipulations and constructs
A cDNA fragment encoding amino acid residues 1–179 of rat CHMP3 was cloned into the BglII/HindIII sites of pEGFP-N1 (Clontech) as a BglII/HindIII fragment to create a vector for the expression of the fusion protein CHMP31–179–GFP. Site-directed mutagenesis was performed using the QuikChange® method (Stratagene). The mutant protein CHMP31–179–GFPM1 had amino acids Arg24, Lys25 and Arg28 of CHMP3 changed to serine, alanine and asparagine residues respectively. All other constructs have been described previously [24,26].
Mouse monoclonal anti-β-tubulin antibodies were purchased from Sigma, anti-ubiquitin (FK2) from Biomol and antiRab11a from BD Biosciences. The rabbit anti-EEA1 (early endosome antigen 1) was a gift from Dr Michael Clague (University of Liverpool, Liverpool, U.K.), and the rabbit anti-CI-M6PR [cation-independent M6RP (mannose 6-phosphate receptor)] was a gift from Dr Paul Luzio (University of Cambridge, Cambridge, U.K.). Species-specific fluorophore (Alexa Fluor® 546)-conjugated anti-IgG secondary antibodies were all purchased from Molecular Probes.
Cell culture and transfections
Cos-7 and HeLa cells were maintained at 37 °C and 5% CO2 in DMEM (Dulbecco's modified Eagle's medium) supplemented with 10% (v/v) fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin. Cells were plated on to 13 mm coverslips in 24-well plates (Nunc) and grown until approx. 60% confluent when they were transfected with Trans-IT (Mirus) according to the manufacturer's instructions.
Immunofluorescence and multinucleation counts
At 24 h post-transfection, cells were fixed with 4% (w/v) paraformaldehyde for 20 min and permeabilized using methanol at −20 °C for 5 min and then blocked with 10% (v/v) NCS (newborn calf serum). Primary and secondary antibodies were diluted in 2% NCS-PBS (2% NCS in PBS) and cells were incubated with primary antibodies for ∼2 h at 18 °C and ∼1 h for secondary antibodies. Cells were washed five times for 5 min with 2% NCS-PBS following all antibody incubations. Stained cells were then mounted in Mowiol (Calbiochem, San Diego, CA, U.S.A.) and examined on a Zeiss LSM510 laser-scanning confocal microscope and appropriate images taken. For cell multinucleation counts, transfected cells were counted on a coverslip and scored for either a single nucleus or multiple (two or more) nuclei. Cells with continuous plasma membrane and connected by tethered tubulin ‘bridges’ between them were defined as multinucleated, provided both ‘cells’ contained nuclei.
EGF (epidermal growth factor) degradation assay
HeLa cells were seeded on to 13 mm coverslips 24 h prior to transfection and grown to 60–80% confluency. Cells were transfected using TransIT reagent as described by the manufacturer, with CHMP31–179–GFP or CHMP31–179–GFPM1, and allowed to express the constructs for 24 h. The medium was replaced, following a wash in warm PBS, with DMEM containing 1% (w/v) BSA (no fetal calf serum). Cells were serum-starved in this medium for 16 h and then incubated with 500 ng/ml Alexa Fluor® 555-conjugated EGF (Invitrogen) for 2 or 60 min. Following EGF stimulation for given times, cells were washed twice with cold PBS and then fixed in paraformaldehyde, stained with DAPI (4′,6-diamidino-2-phenylindole) for 30 min and mounted on to coverslips using Mowiol. Coverslips were examined on a Zeiss LSM510Meta laser-scanning confocal microscope.
The dominant-negative protein CHMP31–179 is present at the midbody during cytokinesis
Recent work has shown that TSG101, an ESCRT-I component, localizes to Flemming bodies during the late stages of cytokinesis and that its knockdown by siRNA (small interfering RNA) inhibits cytokinesis at abscission . In the same study, it was shown that ESCRT-III components may also have a role to play in abscission. In another study, the ESCRT-III proteins CHMP2, 4 and 5 have been localized to the midbody of dividing cells . In order to investigate this further and determine whether other ESCRT-III components are also present at the midbody during the final stages of cytokinesis, we made use of a dominant-negative CHMP3 construct (CHMP31–179–GFP). Transient transfection of this dominant-negative truncated form of CHMP3 resulted in a swollen vacuolar phenotype typified by CHMP31–179–GFP bound to large vacuolar structures in Cos-7 cells (Figure 1) and HeLa cells (results not shown). These structures were endosomal in origin as they contained both early and late endosomal markers (EEA1 and M6PR) (Figures 1a–1f). Ubiquitinylated proteins also accumulated on the CHMP31–179–GFP-containing endosomes (Figures 1g–1i). It is possible that the accumulated ubiquitinylated proteins are cargoes destined for lysosomal degradation; however, we cannot rule out the possibility that they are ubiquitinylated cytosoplasmic proteins recruited to membranes or even ESCRT components. In an EGF degradation assay, fluorescent EGF accumulated intracellularly after 60 min incubation with EGF in CHMP31–179–GFP-expressing cells but disappeared almost completely from neighbouring untransfected cells (Figures 2d–2f). These results indicate that the CHMP31–179–GFP-fusion protein is dominant negative, as its expression blocks trafficking of EGF to the lysosome and prevents its degradation.
Transient expression of CHMP31–179–GFP results in localization to early and late endosomal compartments as well as impaired trafficking
Dominant-negative CHMP3 perturbs endo-lysosomal trafficking of EGF
In cells fixed during the late stages of cytokinesis, the midbody was observed to contain CHMP31–179–GFP protein apparently present on membrane vesicles (Figure 3a–3c). The CHMP31–179–GFP specifically localizes to the central region of the midbody, where there is lack of β-tubulin staining. This distribution was observed in both Cos-7 cells (Figures 3a–3c) and HeLa cell lines (Figures 3d–3f), indicating that this phenomenon is not cell-type-specific. CHMP31–179–GFP was often seen distributed along the microtubules in the midbody channel (results not shown), suggesting that vesicles maybe being transported along microtubules towards the midbody. Interestingly, although Rab11, an endosomal protein required for abscission , was present in the midbodies of dividing cells, it did not seem to co-localize with CHMP31–179–GFP (Figures 3g–3i).
CHMP31–179–GFP localizes to the midbody in late stages of cytokinesis
Dominant-negative CHMP3 perturbs cytokinesis
HeLa cells were transfected with constructs for the expression of CHMP31–179–GFP, GFP–Vps4E235Q and GFP–Vps4WT (wild-type GFP–Vps4) or a GFP control. They were then fixed and immunostained for β-tubulin and treated with DAPI to stain nuclei. Transfected cells were quantified under the fluorescence microscope for the percentage of multinucleate cells. Multinucleate cells were defined as cells connected by a continuous plasma membrane to another cell, thus containing two or more nuclei (Figure 4b). Of the GFP control transfected cells, 12% were multinucleate (Figure 4a). The percentage of multinucleate cells was dramatically increased (to 48%) in CHMP31–179–GFP-expressing cells. This indicates that CHMP31–179–GFP acts as a dominant-negative protein in cytokinesis in addition to lysosomal trafficking. In agreement with a previous study , the positive control, GFP–Vps4E235Q blocked cytokinesis, while GFP–Vps4WT had very little effect. GFP–Vps4E235Q appears to block cytokinesis at a late stage, similar to CHMP31–179–GFP, as it is also enriched on vesicles at the midbody of dividing cells (results not shown). The ATPase activity of Vps4, which is required to disassemble ESCRT-III from membranes, seems therefore to be required for abscission. GFP–Vps4WT, which does not block cytokinesis, is not enriched at the midbody.
Dominant-negative ESCRT component mutants disrupt proper cytokinesis
In a recent study, Muziol et al.  showed that the dominant-negative effect of a truncated CHMP3 construct on viral budding could be abrogated by mutating three basic amino acids (Arg24, Lys25 and Arg28). We mutated these same amino acids to create CHMP31–179–GFPM1 and expressed this protein in mammalian cells. Unexpectedly, CHMP31–179–GFPM1 associated with endosomal membranes (Figures 5a and 5b) caused the intracellular accumulation of ubiquitinylated proteins (Figure 5c), internalized EGF (Figures 2g–2h) and blocked cytokinesis (Figures 4d and 5d). Thus mutations that abrogate the dominant-negative effect of truncated CHMP3 on viral budding do not block the dominant-negative effect of truncated CHMP3 on trafficking to the lysosome or cytokinesis.
Expression of mutant CHMP31–179–GFPM1 results in early and late endosomal localization as well as disruption of endosomal trafficking and localization to the midbody in cytokinesis
Vesicular membrane traffic is important for the successful completion of cytokinesis in animal cells and in plants [14,19,20]. In plants, membrane vesicles provide material for cell plate formation, and, in animals, membrane trafficking is required for midbody channel closure. The source of membranes that are targeted to the midbody channel is seemingly complex, with Golgi-associated proteins involved in exocytosis and proteins that control endocytic recycling being required for late stages of cytokinesis. Recently, three studies, one in plants and the others in mammalian cells have implied that proteins best characterized as being involved in MVB biogenesis are also required for cytokinesis [11–13].
In the present study, we show that a dominant-negative ESCRT-III protein, CHMP31–179–GFP localizes to the midbody in dividing cells. As TSG101, an ESCRT-I component, localizes to the Flemming body , a phase dense structure involved in abscission, and the ESCRT-III proteins CHMP2, CHMP4 and CHMP5 localize to the midbody of dividing cells , our results strengthen the hypothesis that a complete ESCRT machinery may be present at the midbody. We further show that expression of CHMP31–179–GFP inhibits cytokinesis, as seen by a large increase in the percentage of multinucleate cells compared with controls. This indicates that the ESCRT-III machinery is not simply passively present at the site of abscission but is also functionally required.
So what is the role of the ESCRT machinery in abscission? It has been suggested that the ESCRT machinery may be involved in cytokinesis at the step of membrane fission, as it is likely that ESCRT protein function in cytokinesis, virus budding and MVB formation is mechanistically conserved . To support this, ESCRT proteins, in particular those in ESCRT-III, are required for a late stage, possibly fission, in virus budding [29–31]. This process is topologically similar to the separation (budding) of two cells and the scission of inwardly budded vesicles from the limiting membrane of late endosomes to form MVBs. However, it has recently been shown that CHMP3 is not absolutely required for intraluminal vesicle formation at late endosomes . Therefore it has been proposed that CHMP3 is important for the fusion of multivesicular endosomes with lysosomes  and not fission of inwardly budded endosomal vesicles. This raises the intriguing possibility that CHMP3 may be involved not in membrane fission but in vesicle fusion events that take place at the site of abscission [14,21,32]. A further possibility is that the ESCRT machinery is involved in the function of recycling endosomes that are required for cytokinesis [27,33,34]. Several studies have shown that perturbation of ESCRT function results in a defect in endosomal recycling [7,8]. However, although we observe some co-localization of CHMP31–179–GFP with Rab11 on endosomes in the cytoplasm, there is little overlap of CHMP31–179–GFP with Rab11 at the midbody (Figures 3g–3i). This suggests that there are at least two distinct populations of endosomal membranes, late and recycling endosome-derived membranes, in addition to Golgi-derived membranes, present at the site of abscission. The complexity of components present at the midbody that are necessary for abscission highlights the sophistication of the mechanisms required for the final stage of cytokinesis. Much work is still required to understand these final stages.
In order to investigate whether membrane association of CHMP3 is required for localization to the midbody and inhibition of cytokinesis, we assessed the effect of the mutant CHMP31–179–GFPM1 on cytokinesis. CHMP3 with the same three positively charged amino acid residues mutated had previously been shown to lose both its ability to associate with membranes and its dominant-negative effect on viral budding . CHMP31–179–GFPM1 localized at the midbody (Figure 5d) and inhibited cytokinesis, although to a slightly lesser extent than the non-mutated protein (Figure 4). On further analysis, it was observed that unexpectedly CHMP31–179–GFPM1 clearly associated with endosomal membranes and also perturbed trafficking to the lysosome, as indicated by the accumulation of ubiquitinylated cargo proteins and internalized EGF in cells expressing CHMP31–179–GFPM1 (Figures 2g–2i and 5). Thus it was not possible to determine whether membrane association is required for localization to the midbody and inhibition of cytokinesis. In their study, Muziol et al.  observed plasma membrane association of truncated CHMP3–GFP constructs, whereas we (Figure 1 and results not shown) and others  did not, except at very high expression levels (results not shown). We have previously shown that CHMP3 binds to the endosomal lipid phosphatidylinositol 3,5-bisphosphate in vitro and have argued, as have others, that overexpressed, truncated CHMP3 may associate with membranes via endosomal-specific lipids rather than protein–protein interactions [23,24]. It would be interesting to determine whether endosome-specific lipids are required for abscission as establishment of specialized lipid composition seems to be important in cytokinesis . It is difficult to explain the discrepancies in membrane association of the CHMP3 mutant constructs between our study and that of Muziol et al. , but it is possible that cytokinesis and trafficking to the lysosome are more sensitive to ESCRT-III perturbation than viral budding.
In summary, we have shown that a dominant-negative ESCRT-III protein, CHMP31–179–GFP, localizes to the midbody and inhibits cytokinesis at a late stage. It is likely that the ESCRT machinery is involved in abscission, but further studies will be required to resolve the detailed mechanism of the role of CHMP3 in cytokinesis.
This work was supported by The Wellcome Trust (project grant 070085 to P. W.) and the BBSRC (Biotechnology and Biological Sciences Research Council) (Ph.D. studentship to J. D. D.). We thank Dr David Tosh for a critical reading of this paper prior to submission.
ATPase associated with various cellular activities
charged MVB protein
Dulbecco's modified Eagle's medium
early endosome autoantigen 1
epidermal growth factor
endosomal sorting complexes required for transport
green fluorescent protein
mannose 6-phosphate receptor
newborn calf serum
vacuolar protein sorting 4