Spatiotemporal Regulation of RhoA during Cytokinesis

Abstract

The active form of the small GTPase RhoA is necessary and sufficient for formation of a cytokinetic furrow in animal cells. Despite the conceptual simplicity of the process, the molecular mechanisms that control it are intricate and involve redundancy at multiple levels. Here, we discuss our current knowledge of the mechanisms underlying spatiotemporal regulation of RhoA during cytokinesis by upstream activators. The direct upstream activator, the RhoGEF Ect2, requires activation due to autoinhibition. Ect2 is primarily activated by the centralspindlin complex, which contains numerous domains that regulate its subcellular localization, oligomeric state, and Ect2 activation. We review the functions of these domains and how centralspindlin is regulated to ensure correctly timed, equatorial RhoA activation. Highlighting recent evidence, we propose that although centralspindlin does not always prominently accumulate on the plasma membrane, it is the site where it promotes RhoA activation during cytokinesis.

Introduction

Cytokinesis, one of the most photogenic events in the life of a cell, requires precise positioning of the division machinery relative to the two segregated masses of DNA. Diverse strategies to accomplish this important cellular process have appeared during evolution. In plant cells, the division plane is determined by the converging plus ends of interpolar microtubules that direct the delivery and ultimate fusion of membranes containing cell wall materials to the cell center [1]. Many prokaryotes divide at the midcell. This site can be defined by the coordinated action of two inhibitory signals: an assembly inhibitor that oscillates between the two cell poles and a second inhibitory signal associated with the segregating masses of DNA [2]. The midcell is permissive for the assembly and treadmilling of prokaryotic tubulin, ftsZ, which directs the local synthesis of cell wall materials that mediate cell fission [3]. In animal cells, the position of the anaphase spindle directs the position of the cleavage furrow, as demonstrated by spindle manipulation experiments [4]. The spindle generates both positive and negative signals that create and pattern cortical contractility. At the peak of the positive signal and/or the minimum of the negative signal, the contractile ring, an actomyosin-based structure assembles, constricts, and eventually triggers abscission. Contractile ring assembly requires activation of the small GTPase RhoA and zones of active RhoA accumulate at sites of furrow formation [5–7]. Optogenetic induction of a membrane-associated zone of active RhoA reveals such zones are sufficient for furrow formation [8]. These findings suggest that a primary function of the spindle is to generate and pattern zones of RhoA activity.

RhoA directly activates formin-mediated f-actin assembly and indirectly promotes myosin II activation [9, 10]. The GTPase associates with the plasma membrane via C-terminal prenylation [11], a modification essential for its function. RhoA is activated by a conserved RhoGEF, Ect2 (for consistency, we will use the mammalian nomenclature throughout) (Figure 1 and 2) [12]. RhoA activation requires the interaction of RhoA•GDP with active Ect2 at the plasma membrane. This step is highly regulated, as Ect2 is autoinhibited (Figure 2) [13] and, in some cell types, not constitutively membrane bound [12]. Here we will review the current understanding of this process, with a focus on developments in the last ~5 years. The important steps that follow RhoA activation, contractile ring assembly, constriction and abscission have been recently reviewed [14, 15].

Figure 1. How is RhoA activated at the equatorial plasma membrane?

The centralspindlin complex (green) prominently accumulates at the spindle midzone and is responsible for recruiting and activating the RhoGEF Ect2 (orange) to the mid-plane during anaphase. Both Ect2 and active RhoA (blue) have been detected on the equatorial membrane, where they promote formation of an actomyosin-based contractile ring. There is also evidence for a pool of membrane-bound centralspindlin. Does this pool activate Ect2-RhoA at the plasma membrane, especially in cells where the spindle midzone does not lie adjacent to the membrane?

Figure 2. Domain organization of RhoA, Ect2 and centralspindlin.

Primary structure and predicted domains of the main proteins involved in RhoA activation during cytokinesis. The protein-protein interaction regions are indicated as are the main functional interactions. (N) N-terminal domain; (NL) neck linker region; (CC) coiled-coil domain; (PBC) poly-basic cluster.

Anaphase spindles are primarily composed of radial arrays of dynamic astral microtubules surrounding spindle poles and overlapping plus ends of microtubules at the middle of the spindle. Either one can spatially modulate furrow assembly, and a number of cell types enlist both strategies [16], presumably improving fidelity. Considerable variety is observed in the degree to which a given cell relies on one or the other. Anastral meiotic spindles rely exclusively on overlapping microtubules at the spindle midzone, representing one end of this spectrum. To a first approximation, overlapping microtubule plus ends are involved in promoting the local activation of RhoA at the equator and dynamic astral microtubules suppress it at poles. The centralspindlin complex, a major focus of this review, is a crucial component of the former. We begin by discussing the myriad domains of centralspindlin and how they contribute to RhoA activation. Subsequently we discuss how these reactions are spatially constrained to ensure RhoA activation at the equatorial plasma membrane - despite the lack of significant accumulation of centralspindlin at this site - and discuss our nascent understanding of the mechanism of inhibition of contractile activity by astral microtubules.

Multiple centralspindlin domains contribute to RhoA activation

To activate RhoA and induce a cleavage furrow, the RhoGEF Ect2 itself requires activation. This role is performed by centralspindlin, a protein complex conserved in all metazoans. A centralspindlin heterotetramer contains a dimer of a kinesin-6 motor protein MKLP1 (aka Kif23 in mammals; ZEN-4 in C. elegans; and Pavarotti in Drosophila) and a dimer of a RhoGAP protein, Cyk4 (aka MgcRacGAP in mammals; CYK-4 in C. elegans; and RacGAP50C/Tumbleweed in Drosophila). Each subunit of the complex contains multiple domains required for cytokinesis (Figure 2). While both centralspindlin and Ect2 are well conserved, as is the requirement that they interact, certain aspects of their localization patterns are not consistent across cell types. In most somatic cells during interphase, centralspindlin subunits and Ect2 are primarily nuclear localized [17, 18], but a fraction can also be detected in the cytoplasm and/or the membrane [19]. In mammalian cells, membrane recruitment of Ect2 is cell cycle regulated [20] and accompanied by prominent recruitment to the spindle midzone, whereas in C. elegans blastomeres it is constitutively membrane-bound and only weakly detected at the spindle midzone (K. Longhini and MG, unpublished results). Prominent centralspindlin localization to the spindle midzone is a conserved feature of anaphase cells, but its cortical recruitment is more variable. In Drosophila it is readily detected at the cortex [21, 22], in C. elegans it can be detected during mid-cytokinesis [23, 24], but it is rarely detected at the cortex in cultured mammalian cells, prior to midbody assembly when it is highly concentrated on the membrane [25].

MKLP1 has five functional domains. It has an N-terminal motor domain, an extended neck linker region, a coiled-coil dimerization motif, an oligomerization motif, and a C-terminal globular region. Cyk4 has a short N-terminal unstructured region, a coiled-coil dimerization motif, an extended region that interacts with Ect2, a lipid binding C1 domain, and a C-terminal RhoGAP domain. The N-terminal region of Cyk4 and the extended neck linker region of MKLP1 mediate complex assembly which also requires that each protein dimerize [26].

While this review focuses on the role of centralspindlin in RhoA activation for furrow induction, centralspindlin also is required early in cytokinesis to organize the central spindle and during the final stage of cell abscission [27]. In addition to its cytokinetic functions, distinct interphase functions for this complex are now being defined. For example, centralspindlin recruits Ect2, and thereby activates RhoA, to regulate epithelial junctional integrity [19]. The ability of MKLP1 to bundle microtubules is also required in post-mitotic germline cell development [28] and neuronal axon extension [29].

Structure and function of MKLP1

Kinesin motor domain

One of the major functions of MKLP1 is to bundle microtubules in the spindle midzone during anaphase via its motor domain [17]. Although MKLP1 can associate with microtubules as a dimer, these interactions are transient; processive motility requires MKLP1 oligomerization [30]. However, MKLP1 oligomers are insufficient to bundle midzone microtubules and concentrate centralspindlin at this site. Cyk4/MKLP1 complex formation is required for bundling in vivo [25] and establishes a strong preference for antiparallel microtubule bundling in vitro [31]. Cyk4 binding conformationally restricts the two motor domains of a MKLP1 dimer [31, 32].

MKLP1 oligomerization motif and its regulation

A small, functionally conserved, 16-residue domain in MKLP1 makes an important contribution to centralspindlin function during cytokinesis by inducing oligomerization [24, 30, 33]. Oligomerization potentiates the weak processivity of MKLP1 on microtubules and consequently the signature accumulation of centralspindlin on plus ends of antiparallel microtubules [30]. In human cells, the ability of centralspindlin to oligomerize is inhibited by 14–3-3 proteins that directly bind phosphorylated S710 on MKLP1 prior to anaphase [33]. 14–3-3 proteins are highly helical, dimeric, multi-functional proteins that bind to specific interactors via phosphoserine motifs [34, 35]. The 14–3-3 binding site on MKLP1 is conserved; it is generated through phosphorylation by Ndr kinases [36]. However, during anaphase, the Aurora B kinase subunit of the Chromosome Passenger Complex (CPC), phosphorylates MKLP1 at S708 [37–39], which abolishes 14–3-3 binding [33]. Thus, the CPC promotes central spindle assembly by inducing oligomerization of centralspindlin.

An additional, vital role for centralspindlin oligomers emerged from analysis of PAR-5, a 14–3-3 protein in C. elegans [24]. Most species have several 14–3-3 isoforms, some of which are functionally redundant, making it challenging to characterize their cellular roles [35, 40]. PAR-5 is the only 14–3-3 isoform expressed in the early C. elegans embryo [41, 42]. Depleting this protein or, importantly, mutating its conserved binding site (S682) on ZEN-4/MKLP1 results in embryos with a global increase in active RhoA, and ectopic cleavage furrows associated with a striking localization of centralspindlin on the membrane. These results reveal a role for centralspindlin oligomers in RhoA activation at the plasma membrane. This function of centralspindlin is promoted by the CPC by antagonizing PAR-5 activity. As the CPC is dispensable for furrow formation in PAR-5-deficient embryos [24], centralspindlin oligomerization appears to be the primary function of the CPC in furrow formation (Figure 3). Thus, the property of centralspindlin to oligomerize promotes at least two functions of the complex, one based on microtubule binding and the other on membrane binding. Oligomerization likely increases the avidity of a weak microtubule binding motor domain in MKLP1 and a combination of weak membrane tethers in Cyk4.

Figure 3. Molecular pathway leading to RhoA activation during cytokinesis.

GTP-bound RhoA activates formin-mediated actin assembly, myosin motors and binds a scaffold protein anillin, to generate a functional contractile ring. The RhoA GEF Ect2 is spatiotemporally regulated by centralspindlin. Polo-like kinase 1 phosphorylates Cyk4 permitting its interaction with Ect2, once Cdk1 activity falls in anaphase. 14–3-3 proteins bind MKLP1 preventing centralspindlin oligomerization. Aurora B kinase phosphorylation of MKLP1 prevents 14–3-3 binding, thus activating a functional, oligomeric form of centralspindlin that can localize at the central spindle and the plasma membrane. PAR-5 is the only active 14–3-3 protein in the early C. elegans embryo. NOP-1 is a C. elegans-specific Ect2 activator.

Structure and function of Cyk4

N-terminal MKLP1 and Ect2 docking sites

The N-terminus of Cyk4 is intimately involved in assembly of the centralspindlin complex [25]. C-terminal to this region are important residues that mediate interaction with the RhoGEF Ect2 [43, 44]. Although their domain structures suggest that the RhoGEF Ect2 and the RhoGAP Cyk4 function antagonistically, there is strong evidence that they cooperate to promote RhoA activation. In most cell types, depletion of either Ect2 or Cyk4 results in a failure to form an ingressing cleavage furrow. Ect2 and Cyk4 are required for the accumulation of RhoA and its effectors at the equatorial plasma membrane [44–47]. Depletion of MKLP1 has a weaker affect on contractile ring assembly [44], perhaps due to variable extents of depletion.

Ect2 exists in an inactive conformation owing to an intramolecular interaction between N terminal BRCT domains and C terminal DH-PH domains [13]. How is this GEF activated to promote RhoA function? The N-terminal BRCT domains of Ect2 are phosphopeptide binding modules [48] that also bind phospho-Cyk4 [43, 44] relieving autoinhibition.

Chemical inhibitors and analog sensitive mutants of Plk1 reveal that it plays multiple roles in cytokinesis. Plk1 kinase activity is required for its spindle midzone localization, yet it antagonizes midzone recruitment of PRC1 and centralspindlin during metaphase [49]. Plk1 inhibition leads to loss of Ect2 from the midzone; RhoA and downstream effectors fail to accumulate [50, 51] and Ect2 does not detectably bind Cyk4 [50, 52]. In vitro, Cyk4 phosphorylation promotes, though is not strictly required for, binding with Ect2 [44, 52]. The N-terminus of Cyk4 has seven Plk1 phosphorylation sites, four of which are evolutionarily conserved (S149, S153, S164 and S170). Mutating these sites to alanine attenuates assembly of a stable complex between centralspindlin and Ect2, failure to activate RhoA, and loss of Ect2 from the midzone [52, 53]. In sum, these results point to a model where Cyk4 binding contributes to Ect2 activation.

Although the structure of the Cyk4/Ect2 complex is not known, homology modeling of Ect2 BRCT domains allowed prediction of residues required for this interaction. Mutational analysis revealed that residues T153 and K195 are crucial for Cyk4 binding [52]. Recently, T153A and K195M (TK) mutations were generated in full-length Ect2 and expressed in cells. While one might expect that the TK mutation of Ect2 BRCT domains would phenocopy the non-phosphorylatable mutant of Cyk4, surprisingly, the TK mutant rescues cytokinesis defects in the absence of endogenous Ect2. This mutant is strongly impaired for its interaction with Cyk4 and accumulates extremely weakly at the spindle midzone and equatorial membrane. These results were interpreted to indicate that the Ect2/Cyk4 interaction is dispensable during wild-type cytokinesis [54]. However, an alternative interpretation bears consideration. As the BRCT domains are known to be involved in autoinhibition [13], T153 and K195 may participate in binding to the C-terminus of Ect2 to mediate Ect2 autoinhibition, perhaps via a C-terminal phosphorylation site or an acidic amino acid. If so, the TK mutant might be partially activated due to relief of autoinhibition. Importantly, Cyk4 is still required for cytokinesis in these mutants [54]. This suggests that a low level of interaction of the TK mutant with Cyk4, undetected via co-immunoprecipitation, might suffice for RhoA activation. Indeed, the same study shows that although cytokinesis requires membrane recruitment of Ect2, the required levels of Ect2 at the membrane can be very low, below the limit of detection.

Plk1 phosphorylation of Cyk4 is antagonized by PP2A activity. PP2A binds Cyk4 through these phosphorylations and an additional interaction with a nearby LxxIxE motif. Mutation of this motif results in hyperphosphorylation of Cyk4 and cytokinesis failure. However, cells are able to assemble intact central spindles and cleavage furrows that ingress extensively before regressing. Dephosphorylation of Cyk4 by PP2A is thus likely important for the late stages of cytokinesis and not for early RhoA activation [55–57].

Membrane binding via the C1 domain

Though conventionally and historically considered a microtubule bundling complex, centralspindlin has been reported to localize to the cleavage furrow in some cells [21, 58, 59]. However the mechanism by which it localizes, and the role of this pool of centralspindlin in cytokinesis has only been recently revealed. The Cyk4 component of centralspindlin contains a weak membrane-binding C1 domain. In human cells, this domain plays a role late in cytokinesis, linking the plasma membrane to the spindle microtubules as the midbody forms [60]. Because deletion of the HsCyk4 C1 domain does not impact cleavage furrow formation in these cells [60], it suggests that the membrane-binding property of the C1 domain is not required earlier for RhoA activation. However, the C1 domain is essential for centralspindlin-directed RhoA activation in C. elegans [24, 61]. This difference is likely due to redundant, microtubule-based mechanisms for generating membrane-associated centralspindlin [62], as disrupting spindle organization in human cells uncovers a role for the C1 domain in RhoA activation [24]. Ectopic localization of the Drosophila homolog of Cyk4 on the plasma membrane in S2 cells leads to hypercontractility [63], supporting the model that membrane-bound centralspindlin promotes RhoA activation.

RhoGAP domain

Although the GAP domain of Cyk4 is capable of serving as a canonical Rho family GAP in vitro, with a significant preference for Rac and Cdc42 over RhoA [27, 64, 65], genetic analyses indicate that the domain is primarily involved in RhoA activation during cytokinesis. Activation of RhoA is a curious function for a RhoGAP protein like Cyk4, which would ordinarily be predicted to turn a GTPase “off”. Another RhoGAP, MP-GAP (ARHGAP11A), has a conserved role in inactivating RhoA [66–68] (Figure 3).

The Cyk4 RhoGAP domain has been analyzed most extensively in the early C. elegans embryo. Two main models have been proposed for its function. One posits that Cyk4 GAP domain is primarily required to activate RhoA during cytokinesis. Another model proposes that its GAP activity is required during cytokinesis to turn off the GTPase Rac1. We recently elaborated on the merits of these models in a separate review [69]. A major concern with the latter model is that the experiments to support it have been performed in the presence of a parallel RhoA activation pathway peculiar to C. elegans (see NOP-1 section). In the absence of this second, non-essential, pathway, Cyk4 GAP mutants do not support furrow induction, thus favoring the model where this domain participates in RhoA activation. It is notable that, in this context, the active site of the GAP domain of Cyk4 is required for furrowing even when Rac1 is mutated [61].

The exact role performed by the Cyk4 GAP domain is likely to be context dependent. In Xenopus embryos, inactivating GAP function by abolishing the catalytic arginine finger increases the intensity and broadens the zone of RhoA activation during cytokinesis [70]. This is consistent with a conventional view of GAP function and favors a model where continuous GTPase flux is required for proper cytokinesis. However, deletion of the Cyk4 GAP domain in these embryos causes dramatic instability of the contractile ring [70], suggesting that the GAP domain performs an anchoring function in cytokinesis. Consistent with a role for the GAP domain in membrane recruitment, it is required for its recruitment to the bridges separating individual nuclei in the syncytial germline in C. elegans [71]. In other cell types, the Cyk4 catalytic arginine is either dispensable for furrow induction [72], required to limit cell-substrate adhesion via Rac effectors [65], or contrary to expectations for a conventional GAP, strictly required for furrow formation [73].

The very C-terminus of Cyk4 mediates a direct interaction with the microtubule bundling protein PRC1. While not strictly required for centralspindlin localization to the midzone, this interaction promotes stable accumulation of the complex and central spindle assembly, in the presence of spindle pulling forces [74].

Summary: Activation of Ect2 by Centralspindlin

Centralspindlin-mediated RhoA activation via Ect2 is complex and multifactorial. Genetic analysis indicates it requires physical interactions between all four proteins (Cyk4, MKLP1, Ect2, and RhoA) at the plasma membrane. Formation of this complex involves Plk1-mediated phosphorylation of Cyk4 and subsequent binding of this site by the auto-inhibitory N-terminus of Ect2 [52, 53]. Ect2 membrane recruitment is mediated by its PH domain and basic regions in the C-terminus [20]. Membrane recruitment of Cyk4 involves both its C1 and GAP domains and its ability to bind to MKLP1 [61]. The involvement of the GAP domain suggests that active RhoA could contribute to centralspindlin recruitment to the membrane and thus Ect2 activation, perhaps enabling positive feedback [61]. The Cyk4 GAP domain also appears to allosterically activate Ect2; this activation involves RhoA binding, potentially inducing a second means of positive feedback. These reactions also depend on centralspindlin oligomerization which requires antagonism of 14–3-3 by the CPC [24]. The oligomeric state of the complex may permit a division of labor - some GAP domains in the complex may directly engage Ect2, whereas others may contribute to membrane localization.

While individual interactions between domains have been established in vitro, activation of nucleotide exchange activity of Ect2 by centralspindlin has not yet been demonstrated. Because low levels of membrane-bound Ect2 are sufficient for furrow formation in vivo, and due to the relatively modest activity of Ect2 in vitro, it is likely that the stimulation by centralspindlin will be significant. A major challenge for the future is biochemical reconstitution of Ect2 activation. It is important to emphasize that not every interaction mentioned above is necessarily required during cytokinesis of every cell.

Centralspindlin-independent RhoA activation

Although in most cells, Ect2 activation during cytokinesis requires centralspindlin, in C. elegans embryos, centralspindlin is partially redundant with a protein called NOP-1. NOP-1 is a nematode-specific protein that functions as a global activator of Ect2 during polarization and early embryonic development [75]. C. elegans embryos in which centralspindlin function is disrupted can form cytokinetic furrows that ultimately regress. However simultaneous loss of NOP-1 and centralspindlin abolishes furrow induction and active RhoA does not accumulate on the membrane (Figure 3) [75]. The mechanism by which NOP-1 activates Ect2 is currently unknown; the protein contains low complexity regions but lacks other recognizable domains.

However, elimination of this pathway provides a straightforward means to study the role of centralspindlin in furrow formation in this system. Thus far, compelling evidence for parallel activators of Ect2 during cytokinesis in other organisms is lacking.

Spatial Control of RhoA activation

Spindle midzone

In light of our current understanding of the mechanism by which centralspindlin promotes RhoA activation (Figure 3) how is this activity spatially regulated by the spindle? Overlapping plus ends of microtubules in the spindle midzone are bundled by, among other factors, the coordinated activity of centralspindlin oligomers, the microtubule associated protein (MAP) PRC1, and Kif4, a kinesin motor protein that functions in association with PRC1 [76, 77]. The bundling reaction results in stabilization of these midzone microtubules and stable accumulation of centralspindlin, and in some cell types, centralspindlin recruits Ect2 to the midzone [44].

Centralspindlin accumulation at the spindle midzone positions it in the presumptive plane of cell division. However, prominent accumulation at this site is not essential for initiation of a furrow, as cells can furrow in a centralspindlin-dependent manner without significant centralspindlin concentrating on the spindle midzone, as in PRC1-depleted cells [23, 78, 79]. Furthermore, centralspindlin accumulation at the midzone is not always sufficient for furrowing as observed in C. elegans embryos expressing a Cyk4 ΔC1 variant [24] (Figure 4). While the spindle-bound pool of centralspindlin plays a crucial role in midbody stabilization during the terminal stages of cytokinesis [60], the membrane-bound pool of centralspindlin is likely the most relevant for RhoA activation and furrow initiation, as the plasma membrane is the site of RhoA activation. Microtubule bundles with associated centralspindlin may facilitate its association with Ect2 at the plasma membrane if they are positioned sufficiently close to the cell cortex, as is seen in human cells expressing the Cyk4 ΔC1 variant [60] and in Drosophila spermatocytes [62]. The stark difference in phenotype in human cells and C. elegans embryos caused by the same Cyk4 ΔC1 allele may be a function of the distance between the spindle and the plasma membrane (Figure 4) [24]. Membrane and microtubule-associated pools of centralspindlin may compete with each other, however this has not been experimentally shown. Once cytokinesis is underway, this is unlikely to be a particularly dynamic competition as microtubule-associated centralspindlin exchanges slowly [30].

Figure 4. Spatial control of RhoA activity by the mitotic spindle.

Furrow induction involves centralspindlin-dependent RhoA activation at the equatorial plasma membrane and astral inhibition at the poles (red). Centralspindlin (green) also strongly accumulates at midzone microtubules, but this location can be too distant from the membrane to directly induce RhoA activation. A. In small cells, in which the spindle extends towards the membrane (top), centralspindlin can concentrate near the plasma membrane; in such cells the Cyk4 C1 domain is not required to generate a furrow. In cell types where the central spindle is far away from the plasma membrane (bottom), the membrane-binding C1 domain of Cyk4 is critical in localizing centralspindlin and thereby activating RhoA at the equator to generate a furrow. Equatorial accumulation of membrane-bound centralspindlin is likely induced by a combination of spindle midzone accumulation, local CPC activity, and directed microtubule-based transport.

B. Astral microtubules negatively regulate RhoA activity at the poles. In cells where the centralspindlin complex is globally activated (in the absence of 14–3-3 proteins), asters can still restrict furrow formation to the equatorial region (top). In the absence of RhoA activators (such as a cyk-4;nop-1 double mutant C. elegans embryo), cytokinetic furrows do not form (bottom).

Chromatin modulates CPC accumulation

Centralspindlin binds to the ends of distinct sets of microtubules, depending on the organization of the spindle. In monopolar cells, centralspindlin accumulates robustly at a clustered subset of microtubule plus ends in the cell periphery [80]. However, these sites do not become populated when a bipolar spindle is present, presumably due to the preferential recruitment of centralspindlin to antiparallel bundles at the spindle midzone (Figure 5). Indeed, in assays with artificial asters, centralspindlin accumulates between adjacent asters [81].

Figure 5. Spatial control of RhoA activity by the CPC.

A. Schematic of a binucleate cell undergoing mitosis. During anaphase, the CPC localizes at (1) the spindle midzone near separating chromosomes and subsequently propagates to nearby overlapping microtubules. CPC can also accumulate on (2) microtubules from overlapping adjacent asters but less efficiently than at sister asters. It also can accumulate at (3) the plasma membrane. Known requirements for its association at each site are listed. The complex may not occupy all these sites in every cell type and the regulators of its localization are not fully conserved, and therefore require further investigation.

B. The CPC (gray) is known to accumulate at the equatorial region of dividing veterbrate cells, specifically at the central spindle and the plasma membrane during anaphase. This could result in zones of centralspindlin activity where centralspindlin (green) locally forms clusters and accumulates with high avidity, allowing Ect2 (orange) and RhoA (blue) activation at the equatorial membrane and subsequent recruitment of actomyosin to generate the contractile ring. Centralspindlin is inactive (pale yellow) in other regions of the cell because of 14–3-3/PAR-5 (red) inhibition.

Although centralspindlin binds to overlapping antiparallel microtubule ends, some overlapping plus ends recruit centralspindlin more efficiently than others. When cells contain multiple spindles, centralspindlin accumulation is favored at the overlapping plus ends between asters which had segregated chromosomes (“sister asters”) as compared to neighboring asters that did not participate in chromosome segregation [81]. The preference for sister asters is probably imposed by their strong accumulation of the CPC, apparently as a consequence of the dissociation of CPC from chromosomes and its subsequent binding on the nearest antiparallel microtubules (Figure 5) [82]. CPC accumulation creates gradients in Aurora B kinase activity [83] that appear to propagate CPC accumulation from an initial site at the center of an overlap zone to adjacent regions [81]. This bias in CPC accumulation is then amplified through a positive feedback loop in which Aurora B kinase activity promotes microtubule stabilization [84], and stabilized microtubules preferentially recruit CPC. This model is supported by experiments in which hyperactivation of the CPC or artificial stabilization of microtubules reduces the disparity in accumulation of CPC on sister asters relative to adjacent asters [81]. The CPC, in turn, regulates centralspindlin recruitment to microtubules, at least in part, by regulating centralspindlin oligomerization.

Chromosome Passenger Complex and MKLP2

The CPC also regulates centralspindlin recruitment to the membrane (Basant and Glotzer, unpublished). Suggestively, in human cells, CPC localizes to the equatorial cell cortex prior to furrow invagination [85]. The mechanism of CPC accumulation on the equatorial membrane is not well understood. INCENP, the scaffolding subunit of the CPC, contains a long, flexible, charged, single alpha helical (SAH) domain that has microtubule binding activity required for its localization to the spindle midzone, but it is dispensable for its accumulation at the cell cortex [86, 87]. Curiously, this helical region also associates with f-actin [84].

Factors that regulate CPC recruitment may regulate local RhoA activation (Figure 5). In mammalian cells, one such regulator is the kinesin-6 motor protein MKLP2. CPC and MKLP2 interact and prominently co-accumulate on the spindle midzone, as well as at the equatorial cortex [88]. MKLP2 is structurally related to MKLP1, however, MKLP2 does not interact with Cyk4. Nevertheless, these kinesins are regulated by similar mechanisms. Microtubule binding of both motors is inhibited by direct Cdk1 phosphorylation and activated by dephosphorylation upon mitotic exit; they also both oligomerize [30, 89, 90]. During anaphase, MKLP2 promotes dissociation of the CPC from chromosomes. This suggests a hierarchy of localization dependence: MKLP2 > CPC > centralspindlin.

Given this hierarchy, what are the determinants of MKLP2 localization? MKLP2 contains a myosin binding domain and its localization depends on myosin accumulation. Myosin has been proposed to promote MKLP2 localization [91]. However, given that myosin II accumulation is RhoA dependent and RhoA activation requires centralspindlin, and CPC and MKLP2 promote membrane localization of centralspindlin, this model appears circular, in that it suggests that a downstream effector of MKLP2 promotes its recruitment. Perhaps low levels of active RhoA can be generated at the equator prior to MKLP2/CPC accumulation.

Indeed, although MKLP2 localization is important, in cultured human cells with bipolar spindles, furrow formation is not strictly dependent on either MKLP2 or CPC, though cytokinesis fails to complete in their absence. The simplest interpretation is that a weak, albeit functional, MKLP2/CPC-independent mechanism promotes and regulates cortical accumulation of centralspindlin that creates an initial concentration of equatorial myosin that is then amplified by MKLP2 and CPC-induced recruitment of centralspindlin.

An additional mystery surrounds the role of MKLP2. While the function of Aurora B and centralspindlin are largely similar in C. elegans and vertebrate cells, C. elegans lacks an MKLP2 ortholog, raising the question of how CPC dissociates from chromosomes during anaphase in such cells. One possibility is that MKLP1 fulfills the function of both MKLP1 and MKLP2 in these organisms; however, Aurora B does not remain chromosome bound in C. elegans embryos depleted of MKLP1 [92]. Alternatively, another factor may contribute to CPC localization. A third possibility is that CPC may not have as high an affinity for chromatin in these organisms as it does in vertebrate cells, in which case specific factors would not be required for its dissociation during anaphase. Curiously, although the MKLP2 ortholog in Drosophila regulates CPC localization, it is not essential for cytokinesis in most cells [93].

Astral microtubules

Experiments in a variety of systems indicate the existence of a second pathway for spatial control of RhoA activation that works in parallel to the local accumulation of centralspindlin at the cell equator (Figure 4). This mechanism might be responsible for the initial, CPC-independent, equatorial activation of RhoA discussed above. Although the underlying molecular mechanism is not known, this pathway appears to involve inhibition of cortical contractility by dynamic astral microtubules, a model proposed many decades ago [94, 95].

Many insights into astral inhibition have come from studies of the early C. elegans embryo. When microtubules are attenuated, spindle assembly occurs in the embryo posterior, aligned with the short axis of the elliptical cell. Under these conditions, one furrow will form in the embryo anterior, distal to the astral microtubules and second furrow will form between the two asters [96]. Likewise, furrow formation still occurs when the spindle midzone in C. elegans embryos is disrupted genetically or by laser ablation and the position of the resulting furrow can be predicted from the position of the spindle asters in anaphase [97]. When the midzone is disrupted by inactivation of centralspindlin, these furrows are dependent on the presence of the aforementioned RhoA activator NOP-1, which given the lack of conservation of NOP-1, raises the possibility that these observations might not represent features of astral inhibition in all cell types. However, the well-conserved centralspindlin-directed furrowing pathway, when hyper-activated in C. elegans, is also restricted by astral inhibition. Depletion of 14–3-3/PAR-5 or mutation of MKLP1 such that it is not subject to regulation by 14–3-3/PAR-5 and the CPC, results in embryos that still furrow at the equator, albeit with less precision than in wild-type. This is surprising, as RhoA is initially globally activated under these conditions. When such embryos are treated with nocodazole to misposition the spindle to the posterior, ectopic oligomeric centralspindlin induces furrows in the anterior of the embryo, away from the asters, both in the presence or absence of NOP-1 [24]. These results suggest that astral microtubules can regulate contractility that results from either NOP-1- or centralspindlin-directed RhoA activation in the embryo. The simplest explanation of these results is that dynamic microtubules do not regulate NOP-1 or centralspindlin, but rather a common downstream factor.

Astral microtubules also regulate contractility in other contexts. For example, dynamic microtubules locally inhibit RhoA activation in both cultured mammalian cells and excitable cortices observed in activated echinoderm and Xenopus blastomeres [98, 99]. Microtubule disruption results in hypercontractility in C. elegans embryos [100].

The molecular basis for astral inhibition is not fully understood. Optogenetic experiments in cultured human cells indicate that dynamic astral microtubules do not impact the contractility resulting from artificial accumulation of a minimal RhoGEF domain [8]. If the astral pathway is active in these cells, then it must act upstream of the activators of RhoA, as it does not inhibit ectopically activated RhoA or the downstream effectors. In C. elegans embryos, Aurora A and its activator TPXL-1 are required for microtubules to inhibit accumulation of f-actin and the RhoA effector anillin from the anterior cortex [101]. Interestingly, Aurora A is largely dispensable for clearance of these downstream components from the posterior cortex, suggesting the existence of a separate regulatory mechanism.

A subset of astral microtubules may also play a positive role in delivering activators of contractility. For example, equatorially-directed astral microtubules can contribute to equatorial CPC localization [102] and centralspindlin localization (Figure 4, 5) [45]. In Xenopus, an SxIP motif in centralspindlin enables it to bind microtubule plus ends via EB1, but this motif is not well-conserved [103]. In echinoderm zygotes, but not in subsequent divisions, equatorially-directed microtubules appear to be preferentially stabilized [104, 105]. It is not clear whether this is a common role that microtubules play, nor is it known what distinguishes this subset of microtubules from the larger, radial array of microtubules that surrounds the spindle poles.

Concluding remarks

Activation of RhoA with precise spatiotemporal control during cytokinesis involves intricate, often redundant, mechanisms to regulate the centralspindlin-Ect2 complex. The importance of this complex in furrow induction and several aspects of regulation were well-established. Formerly, much attention focused on the role of this complex at the spindle midzone, particularly since the centralspindlin/Ect2 complex accumulates prominently at this site. However, protein abundance does not necessarily correlate with function. Additionally, given that RhoGEFs are potently activated by tethering to the plasma membrane [8, 106] and the demonstration that centralspindlin-mediated RhoA activation during cytokinesis in C. elegans embryos requires membrane targeting domains in centralspindlin and correlates with the degree of membrane recruitment [24, 61], we propose that the plasma membrane is the primary site where centralspindlin/Ect2 complex induces RhoA activation during cytokinesis. Recent observations support the conjecture that biologically relevant levels of this complex do not require high levels of protein accumulation [54].

In no way does this proposal imply that association of centralspindlin with the plasma membrane is independent of microtubules. In fact, we suggest that microtubules both promote and inhibit the membrane association of centralspindlin and hence its ability to complex with Ect2 (Figure 5). First, by promoting equatorial accumulation of CPC, microtubules indirectly induce direct binding of centralspindlin to the plasma membrane. Second, through the ability of antiparallel microtubule bundles to recruit centralspindlin directly and indirectly via CPC, such bundles at the periphery of the spindle midzone may bring centralspindlin/Ect2 in the immediate vicinity of the plasma membrane. Third, plus-end directed trafficking by centralspindlin could promote its delivery to the plasma membrane [30, 45]. In addition to these mechanisms that promote membrane associated centralspindlin/Ect2, the spindle midzone may functionally sequester centralspindlin/Ect2. Finally, through mechanisms that are yet to be fully defined, astral microtubules are proposed to inhibit the accumulation of centralspindlin and/or Ect2 at polar regions of the cell. We must also consider that there are examples of polarity-derived cues directing the site of cleavage furrow formation, independent of the spindle [22]. Presumably such cells have the capacity to locally induce RhoA activation, much like the pseudocleavage furrow of the early C. elegans embryo. As emphasized throughout this review, different cells are known to utilize different combinations of these mechanisms to promote equatorial activation of RhoA. Indeed, it is likely that the aforementioned redundancy has impeded our ability to understand the mechanism of cytokinetic furrowing.

The coming years promise to reveal important features of the above mechanisms. We need to better understand how asters clear RhoA activators from the poles, how CPC localization on the membrane is controlled, and the molecular details by which centralspindlin enhances Ect2 activity towards RhoA and whether active RhoA has a direct role in this process thereby generating positive feedback. Finally, while these core mechanisms are largely conserved among metazoans, a number of additional context-dependent factors such as polarity proteins, inter-cellular tension [107, 108], cell adhesion [8, 109], cell size, and relative abundance of cytokinetic proteins can impact the position of the division plane; these mechanisms also remain to be fully understood.