Biochemical Journal

Review article

Vesicle trafficking and membrane remodelling in cytokinesis

Hélia Neto, Louise L. Collins, Gwyn W. Gould

Abstract

All cells complete cell division by the process of cytokinesis. At the end of mitosis, eukaryotic cells accurately mark the site of division between the replicated genetic material and assemble a contractile ring comprised of myosin II, actin filaments and other proteins, which is attached to the plasma membrane. The myosin–actin interaction drives constriction of the contractile ring, forming a cleavage furrow (the so-called ‘purse-string’ model of cytokinesis). After furrowing is completed, the cells remain attached by a thin cytoplasmic bridge, filled with two anti-parallel arrays of microtubules with their plus-ends interdigitating in the midbody region. The cell then assembles the abscission machinery required for cleavage of the intercellular bridge, and so forms two genetically identical daughter cells. We now know much of the molecular detail of cytokinesis, including a list of potential genes/proteins involved, analysis of the function of some of these proteins, and the temporal order of their arrival at the cleavage site. Such studies reveal that membrane trafficking and/or remodelling appears to play crucial roles in both furrowing and abscission. In the present review, we assess studies of vesicular trafficking during cytokinesis, discuss the role of the lipid components of the plasma membrane and endosomes and their role in cytokinesis, and describe some novel molecules implicated in cytokinesis. The present review covers experiments performed mainly on tissue culture cells. We will end by considering how this mechanistic insight may be related to cytokinesis in other systems, and how other forms of cytokinesis may utilize similar aspects of the same machinery.

  • cholesterol
  • cytokinesis
  • cytoskeleton
  • membrane trafficking
  • phospholipid
  • soluble N-ethylmaleimide-sensitive fusion protein-attachment protein receptor (SNARE)

INTRODUCTION

Membrane traffic is required for cytokinesis

Mammalian cells exhibit pronounced shape changes during mitosis: cells become rounded in early mitosis, before compacting in metaphase and furrowing in telophase [1]. Such changes are underscored by alterations in membrane trafficking. Over the past decade, clear evidence for a role of membrane traffic in mammalian cytokinesis has been described, including the accumulation of membrane vesicles in the furrow and intercellular bridge, the localization of key trafficking proteins {e.g SNAREs [SNAP (soluble N-ethylmaleimide-sensitive fusion protein-attachment protein) receptors], dynamin and clathrin} to the cleavage furrow and intercellular bridge, and by several studies reporting a functional role for these proteins in abscission [25]. Plant cells cannot divide by an actomyosin-based ring because of the rigid cell wall. Instead, plant cells construct a new membrane at the division plane by directing post-Golgi traffic to this site where the vesicles fuse into a transient compartment, the cell plate, which expands towards the cell edge then fuses with the PM (plasma membrane). Similarly, mammalian cells deliver secretory vesicles to the site of abscission, but other membrane compartments also traffic into the intercellular bridge, in particular recycling endosomes (Figure 1). A role for recycling endosomes in cytokinesis was first suggested from observations that Rab11 is required for cellularization in Drosophila embryos [6,7]. Subsequent studies revealed that Rab11-containing vesicles accumulate in the midbody, and that depletion of Rab11 or expression of a dominant-negative mutant perturbs abscission in mammalian cells [8].

Figure 1 Membrane traffic in cytokinesis

Top panel: during metaphase/anaphase, endocytosis occurs mainly from the poles of the cell, where endosomal Rab proteins such as Rab5 are localized to the side of the daughter nucleus facing away from the furrow. Although endocytosis occurs normally, exocytosis (recycling) of membrane back to the cell surface is inhibited. Rab11-positive vesicles are located more proximal to the furrow, and traffic into the midbody in late telophase along curvi-linear paths. Secretory traffic (shown as light grey-shaded vesicles on the right-hand daughter cell for clarity) into the furrow is also observed. EMMAs are localized to the midbody in telophase. The midbody ring is shown in red. Lower panel: a cross-section through the midbody in late telophase. Rab11/FIP3-positive vesicles are recruited to the midbody: (i) by interaction with the centralspindlin component Cyk4 (after Ect2 is released); or (ii) via a Rab11- and Arf6-dependent interaction with the Exocyst (which in turn is recruited to the midbody ring (red) by centriolin (not shown for clarity). See the text for further details. (iii) Secretory and/or endosomal vesicles accumulate in the midbody prior to abscission and may represent a ‘platform’ on to which the abscission machinery is assembled (see the text).

Such observations support an important role for membrane traffic in cytokinesis, but it is unclear whether this reflects movement into the furrow/intercellular bridge from both daughter cells, an asymmetric delivery of membrane from only one of the two daughters, or a combination of both. Furthermore, other important questions surrounding aspects of these trafficking steps remain unanswered: (i) do the vesicles trafficked to the midbody coalesce into a cell-plate-like structure, or do they remain distinct?; (ii) what controls the accumulation of vesicles in the intercellular bridge?; (iii) how are the normal cellular trafficking pathways modulated during cytokinesis?; and (iv) what is the function of the vesicles which accumulate in the intercellular bridge? Further studies have begun to address these questions.

TRAFFIC FROM ONE CELL, OR TWO?

Using a GFP (green fluorescent protein) marker engineered to localize to secretory vesicles (lum-GFP), Gromley et al. [9] studied the dynamics of secretory vesicles during cytokinesis. Early in cytokinesis, lum-GFP vesicles do not accumulate in the intercellular bridge, but as the midbody thins later in cytokinesis, lum-GFP-containing vesicles accumulate in this region. Strikingly, in all cells analysed, Gromley et al. [9] reported that lum-GFP vesicles accumulated on only one side of the midbody, suggesting that they were delivered from only one of the daughter cells. As cytokinesis progressed, the GFP signal decreased, presumably reflecting the fusion of these vesicles with the furrow PM. Finally, in all cells examined, the intercellular bridge cleaved on the side that received the accumulation of vesicles, albeit some time after the disappearance of the GFP signal. These observations led to the suggestion that the compound fusion of these asymmetrically delivered vesicles with each other and the PM could mark the site of abscission (Figure 1) [9].

Using three-dimensional confocal time-lapse imaging, Goss and Toomre [10] reached somewhat different conclusions. They examined the distribution of a temperature-sensitive mutant of VSVG–YFP (vesicular somatitis virus glycoprotein–yellow fluorescent protein). At 39.5 °C, this protein accumulates in the ER (endoplasmic reticulum); this accumulation can then be released as a synchronous pulse by shifting cells to 32 °C. One crucial finding of this study was that 1 h after release of VSVG–YFP from the ER, 63% of cells undergoing cytokinesis exhibited bright punctae at the cleavage furrow. These structures were given the name of EMMAs (exocytic mid-zone membrane accumulations) (Figure 1), and do not co-localize with Golgi or trans-Golgi network markers, suggesting that EMMAs are constituted from post-Golgi cargo [10]. Interestingly, EMMAs exhibited a markedly heterogeneous distribution in mitotic cells, with one-third of cells examined exhibiting no EMMAs, one third with EMMAs only on one side of the midbody, and one-third on both sides of the midbody. Since this analysis was performed in fixed cell populations, Goss and Toomre [10] postulated that this distribution reflected the cells transitioning between these structures at different stages of cytokinesis. Real-time analysis revealed that both daughter cells traffic both VSVG–YFP-positive and YFP–glycosylphosphatidylinositol-anchored protein-positive vesicles into the furrow, and both daughter cells symmetrically accumulate these post-Golgi vesicles in the furrow (Figure 1). However, the authors observed a transition to asymmetrical accumulation, suggesting that cells can vary the trafficking and/or accumulation of these vesicles in different phases of cytokinesis [10]. Further support for asymmetric vesicle delivery was provided from studies of a large multi-functional protein, BRUCE (baculovirus inhibitor of apoptosis repeat-containing ubiquitin-conjugating enzyme; see below) [11]. Expression of a mutant form of BRUCE revealed that the midbodies of these cells accumulated ‘balloon-like’ accumulations of secretory vesicles in close proximity to the intercellular bridge, but only on one side. Similarly, BRUCE depletion resulted in the asymmetric accumulation of GFP–Rab11 on one side of the midbody [11].

IS THERE A CELL-PLATE-LIKE STRUCTURE IN MAMMALIAN CELLS?

A key question arising from the observation of EMMAs and the accumulation of secretory vesicles in the intercellular bridge is whether these structures are akin to a cell-plate; in other words do these accumulations of vesicles homotypically fuse with each other to generate a larger compartment? Using fluorescence recovery after photo-bleaching, Goss and Toomre [10] observed that VSVG–YFP vesicles remain as distinct entities after reaching the midbody. Similar conclusions were reached by Gromley et al. [9], arguing against homotypic fusion into a cell-plate-like organelle.

The SNAREs required for abscission {Sx2, VAMP8 (vesicle-associated membrane protein 8); [12]} have also been implicated in homotypic fusion [13]. Hence, the possibility remains that homotypic fusion may occur in EMMAs, but the nature of the VSVG–YFP cargo (an integral membrane protein) or a property of the vesicles may place spatial restrictions on the ability of this protein to diffuse into the limiting membrane of any fused vesicles. The presence of lum-GFP-containing vesicles in the midbody of heterogeneous size [11] could imply that some of these vesicles have fused with each other. High-resolution imaging will be required to definitively answer this issue.

Regardless of this point, there is agreement that there is fusion of both lum-GFP vesicles and VSVG–YFP vesicles with the PM in the furrow, and that this fusion precedes the completion of abscission. This may be important mechanistically, as the intercellular bridge is long lived, often persisting for several hours. Although it is generally viewed as being a stable structure, the bridge undergoes gradual thinning, with a notably thinned area often visible at one side of the midbody (reviewed in [14]). This thinning of the bridge may be an important pre-requisite to the abscission event, and the notion that the fusion of vesicles with the PM in the bridge may underpin this is gaining support [14]. Thus a present model suggests that the fusion of vesicles with the PM of the intercellular bridge (and perhaps their homotypic fusion) results in the gradual thinning of the midbody until the final abscission step, which is mediated by components of the ESCRT (endosomal sorting complex required for transport) complex (discussed further below; for a recent review, see [15]).

WHAT CONTROLS THE ACCUMULATION OF VESICLES IN THE INTERCELLULAR BRIDGE?

The observation that vesicles derived from recycling endosomes and post-Golgi secretory vesicles accumulate in the intercellular bridge begs the question of how this is controlled. A first clue regarding this came with the identification of the centrosomal protein centriolin, which was found to accumulate on the midbody ring in late telophase [16]. Centriolin was subsequently found to interact with proteins involved in vesicle docking, specifically components of the Exocyst complex [17], and the SNARE regulatory protein SNAPIN [9]. Depletion of centriolin impairs the localization of Exocyst components to the midbody, and depletion of Exocyst components results in an enhanced accumulation of lum-GFP-containing vesicles in the midbody [9]. Such data would apparently argue against a role for the Exocyst in tethering these vesicles, but rather implicates the Exocyst in their subsequent fusion.

The localization of the Exocyst complex to the midbody also requires Septin 9. Septins are a family of GTP-binding proteins that regulate aspects of membrane trafficking and have previously been implicated in cell division. Depletion of Septin 9 results in a 5-fold increase in the number of cells remaining joined by a midbody [18]. HeLa cells depleted of Septin 2, 11 or 17 reveal abnormal furrow ingression. By contrast, cells depleted of Septin 9 furrow normally but abscission fails in 70% of the cells, with midbodies remaining intact for many hours [19]. In an effort to determine the mechanism responsible for this, Estey et al. [19] examined the localization of known midbody proteins upon Septin 9 depletion. Although molecules such as Cep55 (centrosome protein 55), Plk1 (Polo-like kinase 1) and VAMP8 were correctly localized, the characteristic enrichment of the Exocyst component Sec8 in midbodies was impaired. This result is in agreement with studies in yeast, where septins act to localize Exocyst components [20]. Collectively, such data point to an important role for Septin 9 (possibly acting in concert with centriolin) in the localization of the Exocyst to the midbody.

The mechanism responsible for the tethering of recycling endosomal vesicles in the midbody is incompletely understood, but some studies are beginning to provide us with clues (Figure 1). The recycling endosome-associated protein FIP3 was originally identified as a Rab11 and Arf6 GTPase-binding protein that is required for abscission [21]. Both Rab11 and Arf6 are known to be involved in endosomal trafficking into the midbody [22], and it has been shown that FIP3-associated endosomes associate with centrosomes until late telophase, whereupon they rapidly move into the midbody and accumulate there late in telophase [8,21]. Rab11 and Arf6 are known to interact with components of the Exocyst complex [23], and depletion of Exo70 was found to decrease the localization of FIP3-containing endosomes into the midbody [24]. This is consistent with a model in which the tethering of Rab11/Arf6/FIP3-positive vesicles in the intercellular bridge involves the Exocyst (Figure 1). However, previous work has revealed that FIP3 can also bind the centralspindlin component Cyk-4 and offered a different view of how these vesicles could be tethered [25]. The centralspindlin complex [comprised of Cyk-4/MgcRacGAP (GTPase-activating protein) and MLKP1] regulates the formation of the central spindle and controls actomyosin ring constriction [26]. Cyk-4 recruits the Rho GEF (guanine-nucleotide-exchange factor) ECT2 to the centralspindlin complex; ECT2 activates RhoA and so regulates contractile ring function [26]. It has now been shown that in late telophase, FIP3 interacts directly with Cyk-4 and it does so at the same site on Cyk-4 that binds ECT2 [25]. This led Prekeris and co-workers to propose a model whereby in early telophase (during furrow ingression), centralspindlin-localized Cyk-4 interacts with ECT2 and so controls RhoA activity during contractile ring constriction [25]. Later in cytokinesis, there is a delocalization of ECT2 from centralspindlin (by an as-yet uncharacterized mechanism), resulting in Cyk-4 now interacting with FIP3 localized on recycling endosomal vesicles trafficking into the furrow (Figure 1). Consistent with this model, Prekeris and co-workers showed that ECT2 and FIP3 are sequentially recruited to the midbody as cells progress through cytokinesis [25]. These data make a compelling case for the tethering of recycling endosomal vesicles in the midbody in late telophase involving the centralspindlin complex.

So how can one integrate these observations into a model? A clue may come from the observation that FIP3 vesicle dynamics in telophase is complex [8,25]. Early in anaphase, FIP3 is associated with vesicles around the centrosome. As anaphase proceeds, traffic of FIP3-positive vesicles to and from the cleavage furrow is observed, with the majority of FIP3-containing organelles re-localizing to the furrow side of the nucleus. In late telophase, FIP3-containing vesicles accumulate at the midbody, and are notably more ‘static’ [25], suggesting that different tethering/targeting machinery operates at different stages of the cell cycle. Alternatively, multiple mechanisms of tethering may operate simultaneously, acting as a ‘belt-and-braces’ to insure that appropriate accumulation takes place. Consistent with this, Prekeris and co-workers noted that ECT2 knockdown alone is insufficient to fully drive FIP3 accumulation in the midbody in anaphase [25], and is also supported by data showing that the simultaneous depletion of Rab11 and Arf6 results in a more pronounced cytokinesis defect than either alone [27].

Further studies have identified other vesicle-tethering complexes, including the GARP (Golgi-associated retrograde protein) complex, and TRAPP I (transport protein particle I) and TRAPP II complexes [28], which may also function in cytokinesis. Mutation of the Drosophila homologue of the TRAPP II-specific gene TRS120 [Bru (Brunelleschi)] results in the failure of furrow ingression in male meiotic cells. The localization of Rab11 to the cleavage furrow requires Bru, and Bru and Rab11 genes exhibit synthetic lethality [29]. Bru also exhibits a genetic interaction with a PtdIns 4-kinase [fwd (Four Wheel Drive); see below], prompting the suggesting that this tethering complex acts in concert with phosphoinositides to regulate membrane delivery to the cleavage furrow [29]. These data clearly implicate membrane-tethering complexes, in addition to the Exocyst, as key players in cytokinesis, and further studies regarding their specific function are clearly warranted.

The identification of a new player in abscission dynamics may also have an impact on vesicle tethering in the midbody. BRUCE is a large (528 kDa) protein which is necessary for abscission [11]. This protein localizes to the spindle midzone and midbody in cytokinesis, where it adopts a ring-like structure adjacent to the mitotic kinase Aurora A. BRUCE interacts with Rab8 and Rab11 and also with components of the Exocyst complex [11]. Expression of a mutant form of BRUCE (or its depletion) modulated membrane traffic into the midbody (see above), which led Pohl and Jentsch [11] to argue that BRUCE targets vesicles to the midbody via an N-terminal domain to bind the Exocyst and endosomal components, and a C-terminal-targeting domain for midbody localization. Although it remains unclear whether BRUCE is a bona fide tethering factor, it is clear that this protein plays a vital role in the accumulation of vesicles in the midbody.

HOW ARE NORMAL CELLULAR TRAFFICKING PATHWAYS MODULATED DURING CYTOKINESIS?

During mitosis, cells exhibit dramatic shape changes [1]. Detailed quantitative analysis of cell volume changes in the cell cycle reveal that as a cell transitions from the characteristic spread morphology in interphase through prophase to metaphase, its volume decreases by approximately 30% (‘pre-mitotic condensation’) [1]. Furrowing during cytokinesis results in the volume of the two daughter cells becoming equivalent to that of the mother cell before pre-mitotic condensation.

These changes in cell volume/surface area are underpinned by dramatic, cell-cycle-dependent changes in membrane traffic to and from the PM. Clathrin-dependent endocytosis continues throughout the cell cycle, but in anaphase there is a dramatic decrease in the recycling of internalized material back to the PM, resulting in the accumulation of internalized material in endosomal compartments (Figure 1) [30]. In anaphase, this endocytosis occurs predominantly from the poles of the cells. As mitosis progresses, recycling of the internalized material back to the PM resumes, with traffic now directed towards the furrow [30]. This is consistent with the dynamics of FIP3 distribution (recycling endosomes) described above: in early telophase there is an accumulation of FIP3-positive structures on the furrow side of the nucleus, which then traffic into the intercellular bridge [25]. Similar changes in transferrin receptor trafficking were also reported [30,31]. These latter studies also revealed that the recovery in cell surface area during late stages of mitosis does not require a functional Golgi complex, suggesting that de novo synthesis of lipids and proteins is not required, but rather is underpinned by the redistribution of existing stores, consistent with the suggestion that the delivery of recycling endosomal vesicles plays a key role in this step. [30] Additionally, these changes in endocytic recycling are essential for cytokinesis, as blockade of clathrin-dependent endocytosis resulted in significant cytokinesis defects [31]. Studies in Drosophila spermatocytes revealed similar changes in membrane dynamics, but further showed that there appear to be two distinct phases of membrane addition to the growing furrow [32,33]. The first is a slow increase in surface area during anaphase, followed by a dramatic increase during cytokinesis. Interestingly, this latter increase requires the GTPase Arf6, which was observed to localize to Rab-11 and Rab-4 positive structures in the midzone region, further supporting a role for recycling endosomes in cytokinesis [32].

Changes in cell shape during the cell cycle are co-ordinated by changes to the actin cytoskeleton and the microtubule network. Similar networks participate in the regulation of integrin-dependent cell adhesion [34]. When viewed in the context of a study showing that adherent cells held in suspension fail cytokinesis [35], such observations prompt the question of whether integrin recycling plays a role in cytokinesis. Pellinen et al. [36] showed that the Rab21-dependent trafficking of integrins to and from the cleavage furrow is required for the completion of cytokinesis, where the integrins anchor the ingressing furrow to the matrix. Importantly this group also identified a chromosomal deletion and loss of Rab21 expression in human cancer which resulted in the appearance of multi-nucleated cells, strongly supporting the idea that Rab21-dependent integrin traffic into the furrow is a crucial facet of cytokinesis [36]. It is well established that integrins act to regulate other facets of the mitotic machinery, including spindle positioning, orientation and bipolarity. For example, integrins regulate the production and localization of PtdIns(3,4,5)P3 (see below), which acts as a cortical signal to direct spindle orientation (for a review, see [37]). It is therefore likely that endosomal recycling of integrins play roles in the cell cycle beyond cytokinesis. What is clear, however, is that endosomal recycling to direct traffic into the furrow and midbody (and perhaps other regions of the cell) is carefully co-ordinated during the cell cycle.

So how is cell-cycle-dependent modulation of endocytic traffic achieved? At present, very little is known regarding how cells can turn endocytic recycling on and off. In mitosis, phosphorylation of Golgi structural proteins by Cdk1 (cyclin-dependent kinase 1), Plks and mitogen-activated protein kinases are known to underlie structural changes in the Golgi ribbon [38,39]. Similarly, the Arf exchange factor GBF1 (Golgi-specific brefeldin A-resistant GEF 1) is phosphorylated by Cdk1-cyclin B in mitosis, resulting in its dissociation from Golgi membranes [40]. Such observations suggest that phosphorylation/dephosphorylation of key target proteins in mitosis probably underlies these changes in membrane trafficking characteristics.

Previous studies offer a potential clue to this mechanism. Traffic of endosomal vesicles to the intercellular bridge is a major mechanism contributing to the completion of cytokinesis, and Arf6 is key to that process. Arf6 localizes to both the PM and endosomal vesicles and regulates both actin remodelling and the targeting of vesicles to the intercellular bridge [22,41]. Of relevance to this latter function, Arf6 interacts with the JNK (c-Jun N-terminal kinase)-interacting proteins JIP3 and JIP4 [42]. JIP3 and JIP4 bind the microtubule-associated proteins kinesin-1 and dynactin/dynein, suggesting that Arf6 may control trafficking via modulation of these motor protein interactions [42]. Montagnac et al. [42] showed that endosomal traffic into and out of the intercellular bridge occurs constitutively, with traffic occurring in both directions without apparent vesicle fusion. Based upon elegant biochemical analysis, Montagnac et al. [42] proposed that FIP3/Rab11/Arf6-positive vesicles move within the intercellular bridge towards the plus ends of microtubules, via an interaction between GDP-Arf6 and kinesin-1. Upon activation of Arf6 at the midbody, GTP-Arf6 interacts with the JIP4–dynein–dynanctin complex and moves towards the microtubule minus ends and so carries vesicles out of the bridge [42]. Arf6/JIP4 would then act as a ‘motor switch’ controlling a continuous vesicle polarized trafficking to and from the midbody. Presumably, the availability of docking machinery within the midbody (e.g. Cyk-4 and Exocyst, see above) would then dictate whether a vesicle moves back into the cell body, or is ‘captured’ [42]. A recent study using cryo-electron tomography revealed that, just before abscission, the two daughter cells are connected by a thin cytoplasmic bridge, dense with tightly packed microtubules. The midbody contains both a core of continuous microtubules between the two daughter cells, and anti-parallel polar microtubules whose plus ends overlap at the midbody ring [43]. This dense, overlapping region possibly acts as a barrier to vesicle trafficking between the daughter cells, and may facilitate the switching of motors on vesicles delivered into and out of the bridge.

WHAT IS THE FUNCTION OF THE VESICLES WHICH ACCUMULATE IN THE INTERCELLULAR BRIDGE?

A prevailing view of the midbody is that it provides a platform for the assembly of the machinery required for abscission. Proteomic analysis of midbodies identified over a hundred proteins, many of which were subsequently shown to be essential for abscission [2]. Hence one view of membrane traffic in this context may be to ensure delivery of key components to the midbody in the correct temporal sequence. An alternative role for the accumulated vesicles in the midbody prior to abscission is to deliver membrane area, such that upon fusion with each other and the PM there is a thinning of the intercellular bridge (see above).

A further model proposes that the assembly of a pre-abscission ‘platform’ of densely packed endosomal/secretory vesicles within the midbody acts both as a diffusion barrier prior to abscission and facilitates the formation of the abscission machinery [44]. A clue as to the mechanism in which this may operate comes from studies of protein recruitment to the PM. The presence of phosphoserine (10–20% of PM lipids) and phosphoinositides endows the PM with a net negative charge important for the targeting of proteins containing polycationic motifs [45]. Studies from Grinstein and co-workers revealed that receptor activation alters the PM inner surface potential during phagocytosis. This results in the release of molecules such as K-Ras, Rac1 and c-Src from the membrane [46], and thus modulation of surface charge may re-direct proteins normally resident on PMs on to endosomal compartments [47]. We have speculated that similar alterations in surface charge (arising from alterations in lipid composition; see below) on endosomes localized in distinct cellular regions could facilitate distinct ‘endosome functionality’, serving to recruit the abscission machinery on to accumulated vesicles in the midbody (Figure 1) [44]. This notion of the co-ordinated recruitment of factors that regulate abscission is similar to that proposed for the recruitment of coat components to specific membrane domains during transmembrane protein sorting. For example, the adaptor protein AP-1A coat assembly triggers the concomitant recruitment of many other proteins, including Rac1, the Scar/WAVE (Wiskott–Aldrich syndrome protein verprolin homologous) complex, Rab11 and Rab14 in a fashion dependent upon the lipid composition of the membrane [48]. Baust et al. [48] concluded that “the combinatorial use of low-affinity binding sites present on the same membrane domain accounts not only for a selective coat assembly but also for the co-ordinated assembly of selected machineries required for actin polymerization and subsequent membrane fusion”. The identification of a range of protein domains that interact with anionic phospholipids {such as the FYVE, PH (pleckstrin homology) and PX (Phox homology) domains that interact with different phosphoinositides [49]} or domains which recognise more general physical properties of membranes such as charge or shape (e.g. annexins, BAR and C2 [50]) offers support for the concept of coincidence detection playing an important role in the assembly of an abscission machine on the surface of vesicles in the intercellular bridge. Finally, a recent report has identified a new membrane-targeting domain, kinase-associated-1 domain, that appears to modulate the association of key signalling molecules with the PM [51]. We hypothesize that similar machinery operates on the surface of vesicles accumulated in the midbody to regulate cytokinesis, and in the following section, we offer some ideas of how the membrane itself may help co-ordinate aspects of trafficking in cytokinesis.

THE ROLE OF LIPIDS IN REGULATING CYTOKINESIS

In most membranes, phospholipids are distributed asymmetrically between the two leaflets of the bilayer, forming distinct PM sub-domains enriched in specific species of phospholipids. For instance, in eukaryotic cells, phosphatidylserine and PE (phosphatidylethanolamine) are present predominately in the inner leaflet of the PM, whereas phosphatidylcholine, glycolipids and sphingomyelin reside in the outer leaflet [52]. It is becoming increasingly clear that localized domains of different lipid species can result in the congruent localization of (for example) signalling domains. With this in mind, it is worth re-visiting the role of lipid species on cytokinesis.

Phospholipids

It has been long-established that, during cytokinesis, PE accumulates in the outer leaflet of the PM in the cleavage furrow [53,54]. The movement of the charged PE head-group through the hydrophobic core of the bilayer is energetically expensive, and a family of type 4 P-type ATPases (‘flippases’) has been implicated in mediating the transport of phospholipids between inner and outer leaflets of the PM [55]. Perhaps the localized activity of flippases may mediate the transient enrichment of PE at the cell surface in the furrow. Consistent with this, a phospholipid scramblase has been reported to localize symmetrically over the PM of metaphase cells, but to become asymmetrically localized to the furrow in telophase [56].

This exposure of PE at the surface is important for cytokinesis, since inhibition of the retro-translocation of PE to the inner leaflet by a PE-binding peptide blocks abscission (Figure 2) [53,54]. HeLa cells treated with a PE-binding peptide initiate furrow formation normally, and furrow with normal kinetics, but fail to complete abscission and remain connected for many hours after initiation of cytokinesis [54,57]. Examination of a mutant CHO (Chinese-hamster ovary) cell line defective in PE biosynthesis grown in PE-depleted medium revealed that the contractile ring formed and contracted normally, but did not disassemble, resulting in cell division arresting in late telophase. This suggests that PE transposition is intimately coupled to the disassembly of the contractile ring (Figure 2) [53,58]. Consistent with this, Emoto et al. [53,59] discovered that the immobilization of PE in the exofacial leaflet inhibited the disassembly of contractile ring components such as myosin II and radixin. In bacteria, cell division involves the formation of a septal ring (the FtsZ ring). This ring invaginates the cell wall and results in cell cleavage. In a PE-deficient bacterial strain, FtsZ proteins and other components of the cell division apparatus are correctly recruited to the division site, but fail to constrict [60]. Such observations offer the tantalizing view that PE is required for the cytoskeletal rearrangements required for cytokinesis in prokaryotes as well as eukaryotes.

Figure 2 Lipid distribution in the furrow

The asymmetric accumulation of cholesterol/sphingolipids into raft domains may influence the ability of exocytic vesicles to fuse with the PM. (i) One possible mechanism for this is that the localization of t-SNAREs to these raft domains may result in their preferentially adopting a ‘closed’ conformation (in which the SNARE domain required for fusion is inaccessible to the incoming v-SNARE). The syntaxin (t-SNARE) is shown in blue, with the N-terminal Habc domain folded back on to the SNARE domain. Note that PtdIns(4,5)P2 may also inhibit syntaxin protein function (see the text). (ii) We speculate that within the furrow or midbody, the absence of raft domains in certain areas may facilitate the t-SNARE adopting an ‘open’ conformation (where the Habc domain is moved away from the SNARE domain) and thus may represent areas of the furrow/midbody in which fusion may be enhanced. (iii) PE has been shown to be localized to the exofacial leaflet of the furrow. This PE ‘flipping’ is required for the down-regulation of RhoA activity in the furrow, and may represent a crucial control point for the disassembly of the actomyosin ring (shown in red). (iv) Cholesterol and/or PtdIns(4,5)P2 domains may also act as foci for the assembly of signalling domains, possibly including Ras and phospholipase C-γ. See the text for details.

So how might coupling between contractile ring dynamics and PE levels be achieved? PE adopts an inverted-cone (hexagonal II-structures) in model membranes. Such structures have been shown to modulate the behaviour of integral membrane proteins [61,62] and the activity of signalling molecules [56]. Perhaps the role of PE in cytokinesis is to form a unique PM domain within the furrow to control key machinery for cytokinesis. A central component of the contractile ring assembly pathway is the RhoA GTPase [26]. Similar to most small GTPases, RhoA is regulated by GEFs and GAPs. In the case of RhoA, the critical GEF appears to be the centralspindlin component ECT2, and the GAP is Cyk-4 (see above). RhoA-GTP activates pathways which culminate in actin polymerization and myosin II activation, so driving contractile ring constriction, and active RhoA has been localized to the ingressing furrow membrane [26,6365]. In late cytokinesis, down-regulation of RhoA activity in the midbody is needed for contractile ring disassembly. In cells treated with the PE-binding peptide, RhoA remained localized in the furrow membrane late in telophase and accumulated in the intercellular bridge, suggesting that the PE-binding peptide prevents RhoA inactivation [57,59]. Such observations offer the intriguing hypothesis that PE microdomains may modulate the interaction of RhoA and the contractile ring, and so affect contractile ring dynamics.

Cholesterol, PM rafts and cytokinesis

Sterols, such as cholesterol, have been implicated in cytokinesis in many organisms, including yeast, zebrafish and mammalian cells [6669]. The localization of sterols is cell-cycle-dependent: cholesterol accumulates at the site of cytokinesis in mammalian cells and Schizosaccharomyces pombe [66,67], and depletion of cholesterol perturbs cytokinesis in many systems [6669]. The role of cholesterol enrichment at the equatorial ring is unresolved. The sterol-rich membrane domain of the midline has been proposed to be involved in actomyosin ring positioning and maintenance, as prolonged treatment with the cholesterol-sequestering agent filipin disrupts the localization of the actomyosin ring marker Cdc4–GFP [66]. This was mirrored by overexpression of C-4 sterol methyl oxidase, an enzyme in the sterol biosynthetic pathway [66]. Additionally, Bgs4–GFP, an integral membrane protein which co-localizes with the sterol ring, is mis-localized on overtreatment of filipin [66]. Placement of the division machinery and sterol enrichment at the site of division may be linked, since a mutant of S. pombe with a mis-positioned actomyosin ring and septa has a similarly misplaced sterol ring [66]. Further mutants of S. pombe blocked in G2/M-phase show that the actomyosin ring forms and nuclear division occurs before sterol ring formation in a medial band, but this band persists as the ring constricts and the septum forms [66]. Membrane traffic may be involved in formation of a cholesterol-rich band at the equator since, in S. pombe, a functional secretory pathway is required as brefeldin A-treated cells show limited filipin localization to the midline [66]. This notion is supported by studies of Supervillin, a protein that binds tightly to cholesterol-rich lipid rafts. Supervillin interacts with F-actin (filamentous actin) and myosin-II in the furrow, and with the kinesins such as KIF14 in the midbody, and acts with Arf6-associated vesicle recycling at the cell periphery, thus co-ordinating trafficking and signalling with motor functions in these raft domains [70,71].

An alternative view of the role of cholesterol in membrane function has received impetus from the development of the lipid raft hypothesis, which posits that the preferential association of sterols (such as cholesterol) and sphingolipids endows PMs with the ability to form laterally segregated structures (rafts) that can act as platforms for the selective association of specific molecular machines (e.g. cell signalling platforms) (Figure 2) [72]. Although there remains some debate regarding the size and dynamics of these structures, the prevailing view of lipid rafts as organizing platforms is generally accepted [72]. Several observations have suggested that PM raft-like domains are formed in the furrows of dividing cells, raising the intriguing possibility that raft domains play a fundamental role in cytokinesis, perhaps acting as foci for the assembly of signalling systems. Using sea urchin eggs, Burgess and co-workers demonstrated that the ganglioside GM1, a marker for cholesterol-rich membrane domains, exhibits a striking re-distribution during mitosis [67]. As eggs elongate in anaphase, GM1 moves towards the site of the future furrow. As the furrow ingresses, GM1 concentrates in the furrow by moving away from the poles (compare with the movement of endosomes discussed above); a similar re-distribution of cholesterol into the furrow was also observed. By contrast, analysis of the fluidity of the PM revealed that the furrow membrane was more liquid ordered than the rest of the PM. The raft domains in the furrow were enriched in signalling molecules such as Src and phospholipase C-γ in the raft domains, supporting the notion that these domains act as foci for the formation of signalling units which control furrowing (Figure 2) [67]. Although we agree with this thesis, the possibility exists that a further role for cholesterol-rich domains may be to modulate SNARE protein function in the furrow.

SNAREs, CYTOKINESIS AND LIPID RAFTS

As discussed above, membrane traffic plays a central role in cytokinesis, with the cell exhibiting a carefully orchestrated set of trafficking events during mitosis, the fusion of vesicles at the tips of the growing furrow during cellularization in Drosophila providing a particularly striking example [73]. Membrane trafficking events in all eukaryotic cells, from yeast to humans, are controlled by the formation of specific SNARE complexes [74]. Members of the t-SNARE (target SNARE) family of proteins mark specific organelles. Formation of a SNARE complex between t-SNAREs and their cognate v-SNARE (vesicle SNARE) on the appropriate donor membrane is sufficient to drive bilayer fusion [75,76], thus controlling SNARE complex formation enables the cell to regulate membrane traffic in space and time. Studies of the distribution of t-SNAREs in other systems offer an interesting perspective on the potential role that PM domains may play in cytokinesis.

Exocytic t-SNAREs (comprising of one member of the syntaxin family, and one member of the SNAP25/SNAP23 family [74]) are at least partially clustered within cholesterol/sphingolipid-rich raft domains which are thought to mark the site of exocytosis, and disruption of these rafts by cholesterol depletion modulates exocytosis, suggesting that lipid rafts may act to restrict or localize t-SNAREs into functional domains and so may contribute to the spatial control of membrane traffic (Figure 2) [7780]. By generating a range of mutants of the t-SNARE proteins SNAP23 or SNAP25 with different affinities for lipid rafts, Salaün et al. [79,80] observed that mutants of SNAP25 with an increased affinity for lipid rafts displayed a reduced ability to support exocytosis, and that SNAP23 mutants with decreased affinities for lipid rafts supported increased exocytosis. Such studies argue that the localization of t-SNAREs to rafts may decrease SNARE-dependent fusion. These studies have been further supported by the demonstration that two different t-SNARE conformations exist that are separated both spatially and functionally into distinct clusters in which one of either conformation predominates [81]. Intriguingly, cholesterol depletion results in an increase in the proportion of t-SNAREs adopting a conformation which is functionally competent for fusion [81], prompting the suggestion that cells could regulate the position of exocytosis sites by controlling the lipid microenvironment within which the t-SNAREs reside (Figure 2). This is further supported by a recent study of lysosomal SNARE proteins in lysosomal storage diseases which are characterized by the abnormal accumulation of cholesterol in endosomal/lysosomal compartments [82]. Under these conditions, there is a pronounced reduction in the ability of lysosomes to fuse with either endocytic or autophagocytic compartments. Fraldi et al. [82] showed that the SNARE proteins involved in these fusion steps are abnormally sequestered into cholesterol-rich regions of lysosomal membranes, accounting for their diminished fusion capacity. With such studies in mind, the apparent accumulation of cholesterol and raft domains within the furrow of cells during cytokinesis could perhaps reflect a mechanism by which exocytic fusion is confined, perhaps to the growing tip of the furrow, or regions adjacent to the furrow region, so driving invagination (Figure 2). Clearly such a model is highly speculative, but super-resolution imaging approaches offer the ability to test these hypotheses directly.

Both genetic and biochemical studies have suggested that lipid species may themselves directly modulate SNARE-dependent fusion (reviewed in [83]). It is now generally accepted that SNARE-mediated fusion occurs through a hemi-fused state, in which the outer leaflets fuse first followed by the inner leaflets [84,85]. This imposes geometrical constraints on the two leaflets, since the formation of the hemi-fused state will require the outer leaflet to bend with high negative curvature, and progression to the fused state requires the inner membrane to adopt a positive curvature. Such models suggest that the presence of lipid species with geometries that favour these highly curved states could influence fusion directly [8587]. PtdIns(4,5)P2 is localized to the inner leaflet of the PM and is an inverted cone-shaped lipid [88]. As such, it would be expected that this species would antagonize the high negative curvature which is required for the formation of the hemi-fused intermediate (Figure 2). Consistent with this, James et al. [89] showed that incorporation of PtdIns(4,5)P2 into proteoliposomes containing SNARE proteins significantly slowed the rate of SNARE-dependent fusion. Similarly, incorporation of phosphatidic acid (which favours the formation of negative membrane curvature) enhances SNARE-dependent fusion in vitro [90,91]. Such studies support the hypothesis that the asymmetric distribution of phospholipids and membrane microdomains in the furrow provides an organizing platform for the control of furrowing/abscission (Figure 2). Consistent with this notion, the rate of fusion of secretory vesicles with the PM in the furrow has been observed to be significantly higher than that in the cell body [10].

PHOSPHOINOSITIDES: PtdIns(4,5)P2

Genetic studies implicate phosphoinositide metabolism as a crucial facet of cytokinesis. For example, a mutation in the Drosophila gene fwd, which encodes PtdIns 4-kinase, resulted in defective cytokinesis during male meiosis [33,92]. In S. pombe, both PtdIns4P 5-kinase and its product PtdIns(4,5)P2 are concentrated in the medial ring during cytokinesis, and both PtdIns 4-kinase and PtdIns4P 5-kinase are required for completion of cytokinesis [88]. Studies using reporter constructs comprised of different PH domains to inform on the localization of different phosphoinositides have established that there is an accumulation of PtdIns(4,5)P2 in the cleavage furrow of mammalian cells (Figure 3) [88]. In a contemporaneous study, PtdIns(4,5)P2 was also observed in the cleavage furrow of Drosophila spermatocytes [93]. Both these studies established that PtdIns(4,5)P2 in the cleavage furrow is important for the completion of cytokinesis. Although expression of the PH domain reporters at low levels (for image analysis) did not impair cytokinesis, expression at higher levels resulted in cytokinesis failure. A range of approaches led both groups to concur that interference with PtdIns(4,5)P2 production in the furrow interfered with the adhesion of the PM to the contractile ring, and that a given level of PtdIns(4,5)P2 production is required in the furrow to sustain ingression and maintain an active link with the underlying actin cytoskeleton (Figure 3) [88,93]. This suggestion is supported by the observation that (at least in vitro) PtdIns(4,5)P2 binds to septins and ERM (ezrin/radixin/moesin)-family proteins that are known to link the actin cortex to the PM (reviewed in [94]).

Figure 3 Phosphoinositides in cytokinesis

Top panel: PtdIns(4,5)P2 accumulates in the furrow and may control actin dynamics and the constriction of the actomyosin ring. PtdIns4P is localized to Rab11-positive vesicles trafficking to the furrow. PtdIns3P localizes to vesicles observed in the furrow/midbody in late telophase. Lower panel: Vps34/Beculin generates PtdIns3P on vesicles in the midbody. PtdIns3P then recruits FYVE-CENT via the FYVE domain, which trafficks into the midbody as a ternary complex with KIF13A (a kinesin superfamily member which uses microtubules to deliver the complex) and TTC19. TTC19 interacts with CHMP4B, a component of the ESCRT complex proposed to mediate the final abscission step. See the text for details.

PHOSPHOINOSITIDES: PtdIns4P

Previous studies have added a further level of understanding to the role of fwd and PtdIns4P dynamics in cytokinesis, and begun to integrate these pathways with those of the endosomal trafficking system. Fwd is required for the synthesis of PtdIns4P on Golgi membranes and for the formation of PtdIns4P-containing vesicles that move to the midzone during cytokinesis in fly spermatocytes [95]; importantly, these vesicles were shown to be Rab11-positive (Figure 3). PtdIns4P has been suggested to play an important role in the recruitment of both Rab11 and actin-regulatory proteins (Rac1, Scar/WAVE) during AP-1-dependent protein sorting at the Golgi [92]. Since (i) membrane vesicular transport is known to deliver actin to the midzone in Drosophila embryos [7,73], and (ii) the Rab11-effector Nuf promotes the polymerization of actin in the furrow [96], Brill and co-workers suggested that PtdIns4P-containing organelles may serve to concentrate or recruit factors that maintain F-actin in the contractile ring [92]. In support of this hypothesis, they note that mutations in fwd, nuf and rab11 are all associated with a failure to maintain actin organization during cytokinesis [92]. How these studies of fwd in spermatocyte cytokinesis relates to cytokinesis in other cell types is not clear, as fwd is dispensable for normal development and female fertility. It is likely that (an)other gene product(s) is/are capable of compensating for fwd mutation outside of the male germline. PtdIns4P localizes to the cell plate during plant cytokinesis [97], further emphasizing commonality between plant and mammalian cytokinetic mechanisms.

PHOSPHOINOSITIDES: PtdIns3P

A role of PtdIns3P in cytokinesis was first suggested from studies in Ustilago maydis and Arabidopsis. In U. maydis, there is an asymmetrical accumulation of vesicles enriched in PtdIns3P on the daughter side of the primary septum [98], and in plants, PtdIns3P-enriched vesicles accumulate around the cell plate [97]. Until recently, the role of PtdIns3P in mammalian cell cytokinesis was largely unexplored. Using a GFP-tagged tandem FYVE-domain reporter to localize to PtdIns3P, Stenmark and co-workers observed PtdIns3P-positive vesicles in the intercellular bridge which co-stained with endocytosed transferrin and thus represent recycling endosomal-derived vesicles (Figure 3) [99]. Knockdown of VPS34 (the main phosphoinositide 3-kinase) or its regulator, Beclin-1, delayed abscission and perturbed cytokinesis. Both VPS34 and Beclin-1 localize to the intercellular bridge, suggesting that local production of PtdIns3P is intimately involved in abscission [100]. These studies complement work in plant cells, where disruption of AtVps34 results in a severe growth defect [101], and in Trypanosoma brucei, where RNAi (RNA interference) depletion of TbVps34 results in a cytokinetic block [102], supporting the notion that evolutionarily conserved accumulation of PtdIns3P at the site of cytokinesis is important for abscission.

To delve into the mechanism underpinning this, Stenmark and co-workers performed an siRNA (small interfering RNA) screen in which the effect of depletion of approximately 80 proteins containing either FYVE or PX domains on cytokinesis was examined [99]. Depletion of one such protein, termed FYVE-CENT (FYVE domain-containing centrosomal protein), was found to result in severe cytokinesis defects, similar to those observed upon knockdown of VPS34 or Beclin-1. FYVE-CENT is centrosomally localized in interphase, but localizes to the midbody during cytokinesis. The midbody association requires an intact FYVE domain (i.e. is PtdIns3P-dependent), whereas the centrosomal localization does not [99]. Further insight into the biology of FYVE-CENT came from an examination of its binding partners: FYVE-CENT interacts with KIF13A (a kinesin super-family member) and TTC19 (a protein which contains four tetratricopeptide repeats, identified as protein–protein interaction domains in cell-cycle proteins). Similar to FYVE-CENT, these proteins localize to the centrosome in interphase, but move into the midbody during late telophase and are required for completion of abscission. The proteins exist in a ternary complex, and working models suggest that the kinesin superfamily member KIF13A transports the FYVE-CENT–TTC19 complex into the midbody along microtubules from the centrosomal location (Figure 3) [99].

So what may be the function of this delivery of FYVE-CENT–TTC19 to the midbody? A tantalizing clue was provided by the same study which identified TTC19 as a binding partner for CHMP4B [99]. The ESCRT machinery has been implicated as playing a key role in abscission [15,103]. In particular, ESCRT-III complex proteins have been proposed to mediate abscission via interactions with centrosomal proteins and membrane/membrane vesicles in the midbody. CHMP4B is a member of the ESCRT-III complex [104] that localizes to the midbody in cytokinesis, suggesting perhaps that a crucial role for PtdIns3P is to contribute to the proper assembly of ESCRT-III machinery in the midbody in late telophase via delivery of TTC19 (Figure 3) [99]. This exciting model awaits further experimental verification, but certainly supports the view that membrane remodelling represents a crucial facet of cytokinesis; in this case by defining the assembly of a molecular machine (the ESCRT-III complex) in the correct spatial and temporal co-ordinates.

THE ESCRT COMPLEX AND CYTOKINESIS

A central facet of the completion of abscission is the ability of the dividing daughter cells to cleave the final narrow ‘neck’ of membrane in the intercellular bridge. Studies have focussed on the ESCRT complex in this event, largely because of the remarkable in vitro studies from the Hurley laboratory (see e.g. [105]) which have revealed that ESCRT proteins have the ability to cleave such membrane necks in vitro, and from studies in which various components of the ESCRT complex were localized to the midbody and/or shown to be required for cytokinesis (for reviews, see [15,103,105]). As both the structure and function of ESCRT complexes, and their role in cytokinesis, have previously been reviewed, we will not delve into this area here. However, it is clear that cytokinesis requires ESCRT-III components, but not ESCRT-0 or ESCRT-II. One model proposes that the centrosomal protein Cep55 is recruited to the membrane neck, followed by ESCRT-I and ALIX (vps31 in yeast), which then recruits ESCRT-III to drive the final abscission [103,106,107]. In this context, ESCRT-III is thought to recruit the microtubule-severing enzyme spastin, thus facilitating cleavage of the central spindle at the proposed site of abscission [108,109]. Remarkably, some of this machinery (notably ESCRT-III and vps4-like proteins) function in cell division in archaea, strongly suggesting that this vital facet of the abscission machinery is evolutionaily conserved [110,111].

Several crucial questions remain to be addressed. These include determining the temporal order of arrival of the ESCRT machinery at the site of abscission, and defining how the lipid environment of the membrane neck is modulated to control the interaction with the ESCRT machinery (or vice versa). Super-resolution imaging will no doubt soon provide answers to some of these issues.

SUMMARY AND PERSPECTIVES

How membrane traffic and/or membrane remodelling are integrated and controlled during cytokinesis represents an important challenge for cell biologists. As discussed above, changes in the dynamics and nature of membrane trafficking pathways and the composition of the PM and the accumulated vesicles in the midbody all offer potentially important mechanistic control points: our challenge now is to develop a clearer picture of how these systems operate collectively. The application of ultra-fast high-definition imaging approaches, coupled to genetic manipulation, offers one route to address some of these issues.

The work described above was largely performed on model cell lines grown in culture, reflecting the ease of genetic/pharmacological manipulation and imaging that such systems offer. Cultured cells use a contractile ring to divide (the ‘purse-string’ mode of cytokinesis, sometimes called cytokinesis A). However, it is clear that under certain circumstances and in some cell types, cytokinesis proceeds apparently normally without a contractile ring. For example, a mutant form of Dictyostelium discoideum lacking myosin-II uses traction forces to drive the separation of the two daughter cells when grown on adherent surfaces [112,113]. Similarly, some mammalian cells exhibit a form of cytokinesis which is adhesion-dependent but contractile ring-independent (often referred to as ‘cytokinesis B’) [114]. It is presently unclear whether these two forms of cytokinesis represent back-up mechanisms/redundancies, or whether these different forms of cytokinesis are utilized differentially, e.g. in tissues or when a cell is surrounded by a matrix. The studies described above revealing a role for integrin trafficking in cytokinesis could be interpreted to imply that adhesion forces may operate in ‘cytokinesis A’ as well as in ‘cytokinesis B’, consistent with the evolutionary conservation of key facets of cytokinesis (e.g. Rab11-dependent endosomal vesicle trafficking in mammals, flies and worms; ESCRT function from archaea to mammals, etc.). A further challenge for the field will be to determine whether the molecular machinery identified in tissue culture cells (e.g. endosomal trafficking into the intercellular bridge, the accumulation of cholesterol in the furrow, etc.) also operate during cytokinesis in tissues, or in other more specialized forms of cytokinesis such as the division of larger cells (e.g. in early embryonic division). The next few years will undoubtedly provide us with new insight into these questions.

FUNDING

Work in the Gould laboratory is supported by Cancer Research U.K., the Association for International Cancer Research, Diabetes U.K. and The Wellcome Trust.

Acknowledgments

The authors thank Nia J. Bryant for helpful comments on the manuscript and Figures. We apologize to those authors whose work was not cited for space constraints.

Abbreviations: Bru, Brunelleschi; BRUCE, baculovirus inhibitor of apoptosis repeat-containing ubiquitin-conjugating enzyme; Cdk1, cyclin-dependent kinase 1; Cep55, centrosome protein 55; EMMA, exocytic mid-zone membrane accumulation; ER, endoplasmic reticulum; ESCRT, endosomal sorting complex required for transport; F-actin, filamentous actin; fwd, Four Wheel Drive; FYVE-CENT, FYVE domain-containing centrosomal protein; GAP, GTPase-activating protein; GEF, guanine-nucleotide-exchange factor; GFP, green fluorescent protein; JIP, JNK (c-Jun N-terminal kinase)-interacting protein; PE, phosphatidylethanolamine; PH, pleckstrin homology; Plk, Polo-like kinase; PM, plasma membrane; PX, Phox homology; SNAP, soluble N-ethylmaleimide-sensitive fusion protein-attachment protein; SNARE, SNAP receptor; TRAPP, transport protein particle; t-SNARE, target SNARE; VAMP8, vesicle-associated membrane protein 8; v-SNARE, vesicle SNARE; VSVG, vesicular somatitis virus glycoprotein; WAVE, Wiskott–Aldrich syndrome protein verprolin homologous; YFP, yellow fluorescent protein

References

View Abstract