Understanding Cytokinesis Failure

Abstract

Cytokinesis is the final step in cell division. The process begins during chromosome segregation, when the ingressing cleavage furrow begins to partition the cytoplasm between the nascent daughter cells. The process is not completed until much later, however, when the final cytoplasmic bridge connecting the two daughter cells is severed. Cytokinesis is a highly ordered process, requiring an intricate interplay between cytoskeletal, chromosomal, and cell cycle regulatory pathways. A surprisingly broad range of additional cellular processes are also important for cytokinesis, including protein and membrane trafficking, lipid metabolism, protein synthesis and signaling pathways. As a highly regulated, complex process, it is not surprising that cytokinesis can sometimes fail. Cytokinesis failure leads to both centrosome amplification and production of tetraploid cells, which may set the stage for the development of tumor cells. However, tetraploid cells are abundant components of some normal tissues including liver and heart, indicating that cytokinesis is physiologically regulated. In this chapter, we summarize our current understanding of the mechanisms of cytokinesis, emphasizing steps in the pathway that may be regulated or prone to failure. Our discussion emphasizes findings in vertebrate cells although we have attempted to highlight important contributions from other model systems.

Cytokinesis Occurs in Multiple Stages

The process of cytokinesis can be divided into four stages including specification of the cleavage plane, ingression of the cleavage furrow, formation of the midbody, and abscission (Fig. 1). Each stage is dependent on the proper execution of the prior stage, and thus interference with any stage may result in cytokinesis failure. The first stage of cytokinesis specifies the cleavage plane by recruiting a central regulator of cytokinesis, RhoA, to the site of cleavage. If this step is perturbed, cytokinesis will not initiate properly. In the second stage of cytokinesis, the cleavage furrow ingresses through formation of an actomyosin ring and myosin-dependent motor activity. Failure at this step may lead to a lack of furrow initiation or partial ingression of the furrow followed by regression. The third stage of cytokinesis is characterized by formation of the midbody and stabilization of the cytokinetic furrow. This stage requires proper function of proteins located in the central spindle, a microtubule-based structure that separates segregated chromosomes during anaphase, and on proteins that stabilize interactions between the actomyosin ring and the central spindle. A failure at this stage will lead to regression of the cleavage furrow. The final stage in cytokinesis, abscission, is the step in which the cytoplasmic contents are finally separated from one another. This event requires the presence of a functional midbody, but also additional proteins involved in vesicle trafficking and fusion. Failure at this stage may lead to regression of the cleavage furrow or to formation of a persistent connection between the two daughter cells. Cytokinesis is thus a series of linked processes and a problem at any step of this cascade may be sufficient to induce failure. Some proteins participate in multiple steps in cytokinesis, and thus perturbation of their abundance or activity may be especially prone to induce cytokinesis failure.

Figure 1. Multiple stages of cytokinesis.

Three populations of microtubules first specify the site of cleavage by activating RhoA in a narrow zone between segregating chromosomes (I). Formation and activation of the actomyosin ring next leads to furrow ingression (II). The constricting furrow compacts the central spindle microtubules leading to midbody formation (III). Abscission of the furrow occurs by physically separating the cytoplasm of the daughter cells (IV).

Stage I: Positioning the Division Plane and Initiating Cytokinesis

The Importance of Microtubules

Classic micromanipulation experiments determined that the mitotic spindle dictates the position of the cleavage furrow.^{1, 2} However, a bipolar spindle is not necessary for induction of a cleavage furrow,^{3, 4} suggesting that microtubules themselves play an essential role in initiating cleavage. Three separate populations of microtubules have been implicated in the regulation of cytokinesis (Fig. 1; reviewed by ref. 5). First, equatorial astral microtubules, which emanate from the spindle pole to the site of cleavage, may be stabilized in the equatorial cortical region³ and deliver positive signals that stimulate formation and contraction of the cleavage furrow.² In contrast, polar astral microtubules, which emanate from the spindle pole to sites away from the site of the furrow, may help position the cleavage furrow by inhibiting cortical contractility,^6–8 perhaps by spatially biasing the pattern of myosin recruitment.^{9, 10} Finally, central spindle microtubules, which form an overlapping network between the spindle poles following anaphase, send positive signals that become especially important during later steps of cytokinesis. The signals sent by these distinct microtubule populations are partially redundant, ensuring that selection of the division plane is robust.^{11, 12}

The RhoA Pathway Plays an Essential Role in Furrow Initiation

What are the positive signals delivered by microtubules that initiate furrowing at the correct place in the cell? A central event is the localized activation of the small GTPase RhoA at the site of the future furrow (Fig. 2; reviewed by Ref. 13). RhoA is essential for furrow formation in animal cells,^14–17 and activated RhoA localizes to a narrow zone within the furrow.^18–22 Localized activation of RhoA within this narrow zone is thought to be important for efficient furrowing, as perturbations that broaden the zone of RhoA activation often lead to a failure of the furrow to form or to ingress.¹⁹ A narrow zone of activation is established by tethering RhoA activators to the central spindle, delivering a strong yet spatially restricted signal for cytokinesis initiation.

Figure 2. Localization of cytokinesis components.

Interphase (a), anaphase (b), and late cytokinesis (c).

An essential activator of RhoA is the guanine nucleotide exchange factor ECT2,^{17, 19, 23–26} originally identified as a protooncogene.²⁷ ECT2 is sequestered in the nucleus during interphase (Fig. 2) and released following nuclear envelope breakdown in mitosis, but the protein remains inactive because it exists in an autoinhibited conformation.^{24, 28} In late anaphase, ECT2 localizes to the central spindle and associates with the centralspindlin complex, composed of the kinesin protein MKLP1 and the GTPase activating protein (GAP) MgcRacGAP.^{17, 19, 29, 30} MgcRacGAP binds to ECT2 and stabilizes it in an active conformation that permits it to interact with RhoA.¹⁹ Tethering of the centralspindlin complex to the central spindle is thought to restrict activated ECT2 within a narrow zone, resulting in a narrow zone of RhoA activation.^{19, 23, 25, 29} Depletion of MKLP1¹⁹ or disruption of the central spindle^{31, 32} leads to delocalization of ECT2 and MgcRacGAP from the central spindle, broadening the region of RhoA activation.¹⁹ Cells containing a broad zone of RhoA activation fail to form a furrow.¹⁹ Therefore, tethering of MgcRacGAP and ECT2 to the central spindle is not essential for RhoA activation, but is instead important for efficient furrowing by restricting the zone of RhoA to within a narrow zone at the equator of the cell.

These findings suggest that cytokinesis failure could result from failure to properly deliver RhoA activators to the cortex, causing insufficient activation of RhoA. Alternatively, cytokinesis could fail if RhoA is activated too broadly, in regions outside the cleavage furrow. Interestingly, ECT2 deregulation can lead to oncogenic transformation^{27, 28, 33} although it is not clear whether perturbation of cytokinesis is an important component of this phenomenon, as ECT2 may participate in other processes such as spindle assembly³⁴ and regulation of the Ras/MAP kinase pathway.³⁵ Like many genes involved in cell division, ECT2 expression is induced by growth factors³⁶ in a manner that depends on the Rb/E2F pathway.³⁷ ECT2 is overexpressed in some tumors^{38, 39} where it could broaden the region of RhoA activation, perturbing proper initiation of cytokinesis. Alternatively, elevated ECT2 could perturb late stages of cytokinesis, as RhoA may need to be inactivated for cytokinesis to be completed.²⁵. In fact, overexpression of some fragments of ECT2 has no effect on cytokinesis initiation, but specifically blocks later stages of cytokinesis.^{24, 25}

Other proteins may regulate RhoA activity during cytokinesis.⁴⁰ These include additional Rho GEFs such as GEF-H1⁴¹ and MyoGEF,⁴² both of which are essential for cytokinesis in mammalian cells. Additional proteins may influence the location and timing of RhoA activation, including the armadillo protein p0071⁴³ and the Rho effector mDia1, which may sustain RhoA activation in a positive-feedback loop.⁴⁴ In contrast, the protein HEF1, which is upregulated in tumor cells, may impair the RhoA activation cycle.⁴⁵

GAP proteins are also important for controlling RhoA activation and inactivation. As stated earlier, RhoA may need to be inactivated during late cytokinesis to disassemble the cleavage furrow, and thus hyperactivation of RhoA could block cytokinesis completion. Two GAP proteins that may inactivate RhoA during cytokinesis are MgcRacGAP and p190 RhoGAP.⁴⁶ Although MgcRacGAP plays a critical role in activation of RhoA by recruiting and activating ECT2, phosphorylation of MgcRacGAP by Aurora kinases may stimulate its ability to serve as a RhoGAP,⁴⁷ contributing to RhoA inactivation. As its name suggests, MgcRacGAP may also inhibit the GTPase Rac. The activity of Rac is suppressed in the spindle midzone,²¹ and constitutively activated Rac induces a mutlinucleation.⁴⁸ Thus in addition to activating RhoA by recruiting ECT2, MgcRacGAP may inactivate Rac in the furrow to support cytokinesis.^48–51

Failure of Cytokinesis During Stage I

Together these studies emphasize the importance of microtubules in delivering signals that lead to localized activation of RhoA, and possibly suppression of Rac, in the furrow. Recent studies suggest that cytokinesis failure may occur in cells in which spindle elongation or spindle positioning is perturbed, disrupting delivery of activation signals to the cortex. The first example is binucleation of cells in the liver, which may be regulated physiologically.^{52, 53} In humans, the number of polyploid cells averages 30–40% in the adult liver.^{54, 55} Studies in rat hepatocytes indicate that tetraploid cells arise from cytokinesis failure in which diploid, mononucleated cells undergo mitosis but do not form a contractile ring.⁵² Cells do not undergo anaphase spindle elongation, perhaps because reorganization of the actin cytoskeleton is impaired.⁵³ Furthermore, astral microtubules fail to contact the equatorial cortex in cells that fail cytokinesis,⁵³ suggesting that the delivery of RhoA activators to the cortex is impaired. In rat liver, the number of binucleated cells increases following weaning, suggesting there may be important connections between liver physiology and cytokinesis regulation,⁵³ but how these pathways might impact microtubule organization remains unclear.

The second example is cytokinesis failure that occurs in cells that contain mutations the APC (Adenomatous Polyposis Coli) tumor suppressor. Some APC mutations may induce cytokinesis failure by interfering with microtubule-dependent anchoring of the mitotic spindle.⁵⁶ Although APC has important roles in formation of the mitotic spindle and the spindle checkpoint,^57–60 cells expressing APC mutants become polyploid over time,^{56, 59, 61} indicating that the protein is important for proper cytokinesis. Different APC alleles may have distinct effects on mitosis.⁵⁶ For example, in cells expressing a particular C-terminal truncation mutant of APC, microtubules make less contact with the cell cortex, spindles undergo considerable rotation during mitosis, and cells do not efficiently initiate cytokinesis.⁵⁶ The physiological relevance of these findings was confirmed by the finding that the Min allele of APC gives rise to similar mitotic defects and that the frequency of tetraploid cells is greatly increased in Min⁵⁶ and APC knockout mice.⁶⁰ Although it is likely that tetraploidy can arise through multiple mechanisms in tumors carrying different APC alleles, these findings suggest that failure to properly anchor the mitotic spindle can be an important source of tetraploidy.

Stage II. Ingression of the Cleavage Furrow

In the second stage of cytokinesis, activated RhoA leads to recruitment and activation of effector proteins that organize the furrow and stimulate ingression (Fig. 3). RhoA stimulates actin polymerization through activation of formins, and stimulates myosin activity by activating kinases such as Rho kinase (ROCK) and Citron kinase. Scaffolding proteins such as anillin and septins also play important roles in organizing the cleavage furrow and promoting cytokinesis. Here we discuss each of these processes and how they might be perturbed to result in cytokinesis failure.

Figure 3. Role of the RhoA pathway in furrow initiation.

Autoinhibition of ECT2 is suppressed by association of ECT2 with MgcRacGAP. ECT2 then activates RhoA by stimulating exchange of GDP for GTP. Active RhoA then activates formins to stimulate actin nucleation, and binds to ROCK and Citron kinases, stimulating phosphorylation and activation of myosin.

Stimulation of Actin Filament Assembly

Formins are proteins that nucleate formation of unbranched actin filaments in response to stimulation by RhoA (for review, see ref. 62). In the absence of active RhoA, most of the formins that regulate cytokinesis are autoinhibited.⁶³ RhoA binding relieves autoinhibition to promote actin polymerization.^{64, 65} The mammalian formin mDia1 is activated downstream of Rho signaling,⁶⁶ and cytokinesis is blocked if mDia1 is inhibited by antibody injection.⁶⁷ However, deletion of mDia1 does not perturb cytokinesis in mouse embryonic fibroblasts,⁶⁸ suggesting redundant pathways for actin nucleation. Cytokinesis may also depend on the use of preexisting actin filaments that are nucleated outside the furrow.^{69, 70} Formins may be important in later stages of cytokinesis, as they have been implicated in regulation of Src activity, which has been shown to be important for completion of cytokinesis.^{67, 71}

Localization and Activation of Myosin

Myosin II (hereafter simply referred to as myosin) is the principle motor protein required for cytokinesis (for review, see ref. 72). Myosin is recruited to the cleavage furrow at early stages of cytokinesis in a RhoA-dependent fashion. Myosin activity and localization are regulated by phosphorylation of its regulatory light chain (myosin light chain or MLC). Because myosin motor activity is directly required for furrow ingression,⁷³ perturbation of myosin localization or its activity could result in cytokinesis failure.

Phosphorylation of serine 19 of MLC simulates actin-activated ATPase activity of myosin,^{74, 75} whereas phosphorylation at threonine 18 promotes myosin assembly. Phosphorylation of myosin at these positions is important for proper localization of myosin to the furrow and for ingression.^76–79 In contrast, phosphorylation at serines 1, 2 and 9 of MLC inhibits myosin ATPase activity^{80, 81} During mitosis, MLC is phosphorylated at these positions by CDK1.^{76, 82, 83} At anaphase, inactivation of CDK1, controlled by the degradation of mitotic cyclins by the Anaphase-Promoting Complex/Cyclosome (APC/C), is important for MLC dephosphorylation and myosin activation during cytokinesis. Therefore, failure to degrade mitotic cyclins, or to fully inactivate CDK1, could therefore perturb myosin activation and disrupt cytokinesis.

Three kinases contribute to myosin activation by phosphorylating positions 18 and 19 of MLC (Fig. 3). Two of these kinases, ROCK and Citron kinase, are activated by RhoA. ROCK localizes to cleavage furrows,^{84, 85} and a small molecule inhibitor of ROCK slows cleavage.⁸⁴ Citron kinase also localizes to the cleavage furrow, and is required for cytokinesis in several systems.^{25, 50, 86–92} Citron kinase can phosphorylate MLC at both ser19 and thr18,⁹³ and its overexpression causes unregulated contraction of the cortex, supporting its role as a positive regulator of myosin activity.⁸⁷ Mouse knockout studies suggest that Citron kinase may play an especially important role in neurogeneic and spermatogenic cytokinesis.^94–97 It is likely that ROCK and Citron kinase play partially overlapping roles, explaining why each protein is not essential for cytokinesis in all systems. There is no evidence to suggest that ROCK or Citron kinase is overexpressed or mutated in human tumors, but Citron kinase interacts with the kinesin protein KIF14,⁸⁹ which is overexpressed in several tumor types.^98–101 Whether overexpression of KIF14 perturbs Citron kinase function in cytokinesis remains unknown. Knockdown of KIF14 induces cytokinesis failure,¹⁰² perhaps as a result of a failure to recruit Citron kinase to the cleavage furrow.

Myosin light chain kinase (MLCK) is the third and final kinase that has been implicated in direct phosphorylation of myosin light chain. This kinase is activated by calcium/calmodulin, and some isoforms of MLCK and calmodulin localize to the cleavage furrow.^103–106 Inhibition of calmodulin or MLCK can disrupt cytokinesis in cultured cells,^106–108 but mice lacking MLCK develop normally, but die after birth, suggesting the kinase is not essential for cytokinesis in all tissues.¹⁰⁹ How myosin light chain kinase is regulated in the cleavage furrow is unclear, but hydrolysis of PIP2 may be important for IP3-induced calcium release, which could stimulate MLCK activity.¹¹⁰ The mild and varied phenotypes associated with MLCK inhibition again suggest functional redundancy in MLC phosphorylation during cytokinesis.

The overall degree of MLC phosphorylation is also affected by the activity level of myosin phosphatase. This enzyme is inhibited by several mechanisms during cytokinesis to favor MLC phosphorylation. Myosin phosphatase is a heterotrimeric enzyme consisting of a targeting subunit that bind myosin (MYPT1 or MBS), a catalytic subunit (the delta isoform of PP1c), and an additional small subunit. Both ROCK and Aurora B may phosphorylate MYPT1 in the furrow to inactivate the phosphatase.^111–113 In addition, a number of other kinases including Raf-1¹¹⁴ may negatively regulate myosin phosphatase (reviewed in ref. 72). Thus a number of signaling pathways could converge on myosin phosphatase to regulate cytokinesis.

Organization of Actin and Myosin in the Furrow

How actin and myosin are organized within the cleavage furrow, and how contraction occurs, is not well understood (for a more detailed discussion, see refs. 115 and 116). Phosphorylated myosin localizes to the furrow in early anaphase⁷⁹ and localization of myosin requires RhoA activation^{13, 117} but not myosin ATPase activity.^{73, 118, 119} Recruitment may require phosphorylation of MLC, as mutations or inactivation of kinases that perturb MLC phosphorylation also disrupt myosin localization. Other scaffolding components, such as anillin, may be important for maintaining myosin within the furrow, as discussed below.

Both actin and myosin are highly dynamic in furrows, and dynamic actin is important for cytokinesis.^{108, 120} For example, the actin disassembly factor cofilin is necessary for cytokinesis.¹²¹ Cofilin is negatively regulated by the kinase LIMK1, and thus upregulation of LIMK1, or loss of its negative regulator LATS1, is sufficient to enhance actin polymerization and induce cytokinesis failure.¹²² Interestingly, inhibition of myosin slows disassembly of actin filaments of the furrow,¹²³ suggesting that myosin motor activity may help drive actin disassembly.

Scaffolding Proteins in the Furrow

Anillin

Another conserved furrow component is anillin, which may act as a scaffold protein that binds F-actin, myosin, septins, and activated RhoA.^124–129 Although anillin localizes to the furrow at early stages of cytokinesis, it is not essential for ingression. Instead, it may stabilize the furrow and be important for later stages of cytokinesis including midbody formation and abscission.^{92, 127, 130–132} However, anillin becomes essential for ingression if the central spindle is disrupted, suggesting it may make early steps of cytokinesis more robust.¹²⁸ Anillin interacts with RhoA¹²⁸ and its localization to the furrow requires activation of RhoA.^{128, 129, 132, 133} MgcRacGAP may also be directly involved in targeting anillin to the furrow,¹³⁴ providing a link between the centralspindlin complex and anillin localization.

Anillin contains domains that permit it to interact with phosphorylated myosin,¹²⁷ actin filaments, and septins.^{124, 125, 135} These features make anillin ideally suited to crosslink the actomyosin and septin cytoskeletons within the contractile ring. Anillin may enhance the robustness of early stages of cytokinesis by promoting the anchoring of myosin in the vicinity of activated RhoA, favoring myosin phosphorylation. Anillin may be essential for asymmetric ingression of the cytokinetic furrow, which occurs when the furrow ingresses from only one side of the cell, rather than circumferentially.¹³⁶ Asymmetric ingression may be important in epithelia¹³⁷ and embryos^{136, 138} where it may serve a mechanical function, enhancing the robustness of cytokinesis. Anillin is also important for completion of cytokinesis, as anillin remains in the cytoplasmic bridge even after myosin and actin have dissociated.¹²⁷ Interestingly, whereas overexpression of anillin seems to have little phenotype in Drosophila S2 cells,¹²⁹ overexpression of anillin in mammalian cells is very toxic,¹³² suggesting anillin could have important functions independent of cytokinesis. Levels of anillin appear to be controlled by the ubiquitin-proteasome pathway, as anillin is targeted for ubiquitination and degradation during G1 by the APC/C.¹³²

Like the liver, the heart also contains a large number of tetraploid cells that arise through cytokinesis failure. Although it was originally proposed that cytokinesis failure might be a consequence of failure to disassemble myofibrils within cardiomyocytes, recent work suggests this is unlikely to be the case.¹³⁹ Instead, cells that fail cytokinesis show complete disassembly of the myofibril, but show abnormal localization of anillin, and failure of anillin to concentrate at the midbody.¹³⁹ However, these cells also show delays in furrow ingression, suggesting that earlier steps in cytokinesis may also be affected in these cells.

Septins

Septins represent a second class of scaffolding protein that may help to organize proteins within the cleavage furrow. Septins are GTP-binding proteins that can form filaments and localize to the cytokinetic ring.^140–144 Several human septins have been implicated in cytokinesis, including SEPT2 (Nedd5), SEPT9 (MSF), and SEPT12. SEPT12 localizes to the central spindle and midbody during anaphase and cytokinesis, respectively.¹⁴⁵ SEPT2 accumulates in the contractile ring and midbody,^146–148 and microinjection of antibodies¹⁴⁶ or antisense downregulation¹⁴⁷ of SEPT2 interferes with cytokinesis. Inactivation of SEPT9 by antibody microinjection or siRNA also induces cytokinetic defects.^{149, 150}

Septins may participate in several aspects of cytokinesis, including regulation of actin and microtubule dynamics. SEPT2 associates with actin, forming filaments in association with actin bundles and focal adhesions,¹⁴⁶ whereas SEPT9 associates with the microtubule network.^{150, 151} Septins may also play a direct role in cytokinesis by interacting with anilllin.^{124, 125, 135} Furthermore, SEPT2-containing filaments may provide a molecular platform for myosin and its kinases to ensure the full activation of myosin that is necessary for cytokinesis.¹⁵² Finally, septins may form a barrier that restricts the diffusion of membrane proteins in the furrow,^{153, 154} thus helping retain activated RhoA within the narrow zone required for efficient initiation of cytokinesis. In mammalian cells, the p85 subunit of PI3 kinase may regulate SEPT2 in cytokinesis,¹⁵⁵ linking cellular signaling pathways with steps in cytokinesis. Septins may be deregulated in tumors, either through gene fusions^156–158 or by overexpression.¹⁵⁹ The SEPT9 gene is amplified and overexpressed in mouse mammary tumors and human breast cancer cell lines,¹⁵⁸ and high SEPT9 expression in human breast cancer cells is associated with oncogenic phenotypes and cytokinesis defects.¹⁶⁰

Stage III. Formation of the Midbody

The central spindle, also referred to as the spindle midzone, plays an important role in keeping separated chromosomes apart prior to cytokinesis completion, because when microtubules are depolymerized in late anaphase, the nuclei collapse back together.⁷³ Microtubules in the midzone may be locally nucleated, as the minus ends of the midzone microtubules are decorated with gamma-tubulin.^{161, 162} As cytokinesis progresses, the constricting furrow compacts the midzone microtubule array. The furrow ingresses until a cytoplasmic bridge is formed that is 1–1.5 microns in diameter. Several kinesin-like motor proteins and chromosomal passenger proteins move along the midzone spindle towards the plus ends and accumulate in the overlapping region, forming a phase-dense structure referred to as the Flemming body, stembody, telophase disc, or midbody (reviewed in ref. 163). Disassembly of the actomyosin ring may be an important step at this stage of cytokinesis, as loss of F-actin accompanies and may trigger midbody formation.¹⁶⁴ Once the cytoplasmic bridge matures and abscission begins, the bridge becomes insensitive to the actin inhibitor latrunculin,⁹² implying that the plasma membrane is linked to the midbody by a connection that does not involve dynamic f-actin. Scaffolding proteins such as anillin and septins may stabilize the bridge structure. In almost all systems, central spindle formation is essential for midbody formation, which in turn is necessary for abscission.^{16, 32, 165} In this section, we discuss the components that are required for formation of the central spindle and midbody.

PRC1 is a microtubule bundling protein that is critical for midzone formation in mammalian cells.^{32, 166} PRC1 accumulates on the central spindle in anaphase, and suppression of PRC1 expression causes failure of microtubule interdigitation.³² In the absence of PRC1, astral microtubules can guide the equatorial accumulation of anillin, actin, and chromosome passenger proteins, enabling cleavage furrow ingression, but abscission fails.¹⁶⁷ PRC1 has separate domains that independently target the protein to the midzone and bundle microtubules.³² PRC1 is targeted to the midzone by the kinesin protein KIF4, which transports PRC1 to the ends of microtubules. Absence of KIF4 leads to a failure to accumulate PRC1 in the central spindle, and abolishes central spindle formation.¹⁶⁸ PRC1 in turn recruits the centralspindlin complex, and additional mitotic kinesins including CENP-E, MCAK¹⁶⁹ and KIF14.⁸⁹ PRC1 also serves as an important docking site for the kinase Plk1 in the central spindle.¹⁷⁰ PRC1 expression may be perturbed in cancer cells or in response to checkpoint signaling pathways. PRC1 upregulation in tumors^{169, 171} may be a consequence of p53 inactivation, as induction of p53 can inhibit PRC1 expression and interfere with cytokinesis completion.^{172, 173}

Although the centralspindlin complex (MKLP1 and MgcRacGAP) is important for cytokinesis initiation as described earlier, centralspindlin is also necessary for spindle midzone and midbody formation, and ultimately for abscission.^{31, 165} Centralspindlin is recruited to the midzone by PRC1 (reviewed in ref. 174), and proper localization requires the presence of both members of the centralspindlin complex.¹⁶ A splice variant of MKLP1, called CHO1, includes an additional domain that can interact with F-actin,¹⁷⁵ suggesting that CHO1 could link the actin and microtubule cytoskeletons. Injection of antibodies that target this domain induce failure in late steps of cytokinesis,¹⁷⁵ suggesting CHO1 may stabilize interactions between midbody microtubules and the ingressing cleavage furrow in late steps of cytokinesis. Centralspindlin is also important for recruiting additional proteins to the midbody that are required for abscission, and both components of the complex appear to be regulated by phosphorylation. Each of these topics will be discussed in more detail below.

Proteomic approaches have identified a large number of proteins that concentrate at the midbody (Fig. 2),¹⁷⁶ and a functional role for some of these proteins in abscission has been supported by results of RNAi experiments.^{92, 162, 177} Several of these proteins localize to the Golgi apparatus during interphase, and are released from the Golgi during mitosis by phosphorylation.^178–180 Inhibition of Golgi disassembly during mitosis perturbs cytokinesis,¹⁸¹ perhaps by interfering with release of components that are essential for cytokinesis. Precisely how these proteins function in cytokinesis remains unclear, but one potential function is to recruit mitotic regulators such as Plk1 to the midbody.¹⁷⁹

Additional proteins that localize to the midbody and are required for cytokinesis include LAPSER1, which may recruit the microtubule severing protein katanin to the midbody,¹⁸² and annexin 11.¹⁸³ Annexins are Ca(2+)-binding, membrane-fusogenic proteins with diverse but poorly understood functions. Cells lacking annexin 11 fail to establish a functional midbody, and instead remain connected by intercellular bridges that contain bundled microtubules but exclude normal midbody components such as MKLP1 and Aurora B.¹⁸³ These data suggest that despite its potential role in membrane fusion, annexin 11 seems to be required at an earlier step for recruitment of MKLP1 and Aurora B to the midbody.

Stage IV. Abscission

Once the midbody is formed, it subsequently organizes the final event of cytokinesis, termed abscission. By the time of abscission, the cytoplasmic bridge has narrowed to 0.2 microns in diameter. At these late stages, microtubule bundles become compacted and begin to disappear.^{92, 184} In this process, the cytoplasmic bridge is reorganized to permit separation of the daughter cells. A wide variety of proteins involved in vesicle and protein trafficking, membrane fusion, and other processes are required for abscission, suggesting the final stage of cytokinesis is just as complex as earlier stages. Human cultured cells may remain connected by the cytoplasmic bridge for many hours before undergoing abscission.¹⁸⁵ In some systems, such as embryos, blastomeres often remain connected by intracellular bridges for many cell cycles. In spermatocytes, the cytoplasmic bridge is in fact stabilized,¹⁸⁶ and cytokinesis completion does not occur, enabling communication between the cytoplasm of adjacent cells. Thus in certain circumstances abscission may be the target of physiological regulation.

Membrane Trafficking and Cytokinesis

Membrane trafficking plays a critical role in the process of cytokinesis (Fig. 4). Three pathways have been implicated in the process of cytokinesis. First, the secretory pathway, including Golgi-derived components, may contribute new membranes and proteins to the ingressing furrow, and also participate in late steps of cytokinesis completion. Second, the endocytic pathway and recycling endosomes may remodel membranes in the cleavage furrow and also contribute vesicles that may participate in the final steps of cytokinesis. Finally, recent evidence suggests that components of the ESCRT machinery, best characterized for its role in multivesicular body formation, may also be essential for the final stages of cytokinesis. The relative contributions of each of these pathways in the process of cytokinesis is likely to be dependent on cell type.

Figure 4. Membrane trafficking in cytokinesis.

Secretory vesicles accumulate at the intercellular bridge in a centriolin-dependent manner by SNARE interaction and vesicle tethering by the exocyst complex. Vesicles originating from the recycling endosome and containing the complex Rab11/FIP3 move along microtubules. Interaction of FIP3 with both ARF6 and the exocyst permits vesicle targeting to the midbody. Vesicles then fuse with each other and the plasma membrane, physically separating the daughter cells.

The Role of the Secretory Pathway

In large embryos, such as Xenopus and Sea Urchins, furrow ingression is coupled to insertion of new membrane via microtubule-dependent exocytosis.^187–190 In smaller cells, such as mammalian tissue culture cells, it is less clear whether new membrane insertion is required. Brefeldin A (BFA), which disrupts ER-Golgi-dependent trafficking, blocks cytokinesis completion in some studies^{191, 192} but not others.^193–196 Further evidence for a role of Golgi-derived vesicles in cytokinesis completion has emerged from studies of the protein centriolin, which may help recruit secretory vesicles to the site of abscission at the midbody.¹⁹¹ Centriolin was initially identified as protein that localizes to the maternal centriole during interphase and accumulates on mature centrioles during metaphase.¹⁹⁷ Knockdown of centriolin in mammalian cells causes cytokinesis failure, with the two cells remaining connected by a cytoplasmic bridge.¹⁹⁷ Centriolin localizes to a ring-like structure within the midbody,¹⁹¹ that also contains gamma-tubulin, GAP-CenA, and the centralspindlin complex. Recruitment of centriolin to the midbody ring is dependent on the centralspindlin complex, explaining why centralspindlin may be essential for cytokinesis completion.¹⁹¹

Centriolin may facilitate cytokinesis completion by recruiting to the midbody proteins involved in vesicle tethering and fusion. Centriolin interacts with components of the exocyst,¹⁹¹ a protein complex that tethers secretory vesicles to the plasma membrane (reviewed in ref. 198). Several components of the exocyst localize to the midbody ring in a centriolin-dependent manner, and depletion of exocyst components by siRNA interferes with cytokinesis completion.¹⁹¹ Centriolin also interacts with snapin, a snare-associated protein, and centriolin is required for the recruitment of snapin and SNARE proteins to the midbody.¹⁹¹

Secretory vesicles, derived from the Golgi apparatus, accumulate at the intercellular bridge during late steps in cytokinesis.¹⁹¹ At the time of abscission, the vesicles disappear, suggesting they undergo homotypic fusion with each other, and also heterotypic fusion with the plasma membrane, releasing their contents.¹⁹¹ Interestingly, vesicles seem to accumulate only on one side of the midbody, suggesting that delivery is asymmetric.¹⁹¹ This finding is consistent with the fact that following abscission, the midbody remains attached to one of the two daughter cells where it may play additional roles in signaling or marking the age of the cell.¹⁹¹ Why abscission typically occurs on only one side of the midbody remains unclear. It has been suggested that abscission may be triggered by arrival of the maternal centriole from one daughter cell,¹⁸⁴ but this event has not been observed consistently.¹⁹¹ Asymmetric abscission may be important to enable the midbody to remain attached to a daughter cell, or it may provide an opportunity to regulate the timing of abscission. However, asymmetric abscission may be inherently more prone to failure than if abscission were to occur on both sides of the midbody. The mechanism and significance of asymmetric abscission is an interesting topic for future investigation.

The Role of Endocytosis and the Recycling Endosome Pathway

Several lines of evidence suggest that the endocytic and the recycling endosome pathways play critical roles in cytokinesis completion. Endocytosis within the furrow may be important for remodeling the plasma membrane during ingression. In addition, endocytosis from other regions of the cell may serve as a source of vesicles destined for delivery to the cleavage furrow either directly or through the recycling endosome. For example, some endocytic vesicles internalized from the polar region are subsequently trafficked to the midbody area during later stages of cytokinesis.¹⁹⁹ Inhibition of proteins essential for endocytosis, including clathrin, dynamin, and alpha-adaptin, perturb cytokinesis in several systems,^200–204 and inhibitors of clathrin-dependent endocytosis block cytokinesis completion in mammalian cells.^{195, 199} In addition, there may be direct interactions between the endocytic machinery and proteins required for cytokinesis such as anillin.²⁰⁵

Small GTPases that regulate membrane trafficking have been directly implicated in cytokinesis completion. Arf GTPases initiate the budding of coated carrier vesicles by recruiting coat protein complexes onto donor membranes, whereas Rab GTPases regulate the targeting and docking/fusion of vesicles with acceptor membranes.²⁰⁶ Two different GTPases, Arf6 and Rab11, have been implicated in regulation of cytokinesis. Rab11 localizes preferentially to the recycling endosome (RE), and is required for proper RE organization and the recycling of vesicles to the plasma membrane. Both Arf6 and Rab11 concentrate near the cleavage furrow and are required for late steps of cytokinesis in mammalian cells.^{199, 207–210} Both GTPases interact with a common set of effector proteins that assist in delivery of endosomal vesicles to the cleavage furrow, termed FIP3 (Arfophilin-1) and FIP4 (Arfophilin-2).^{207–209, 211–214} FIP3-containing endosomes accumulate near the cleavage furrow and are required for successful completion of cytokinesis.²⁰⁹ Recruitment of FIP3 to the midbody requires ARF6, and recruitment of ARF6 to the midbody requires FIP3.²¹⁵ Other studies show that Arf6 interacts with MKLP1, suggesting the centralspindlin complex is important for targeting Arf6 to the cleavage furrow.^{207, 216} The exocyst has also been implicated in targeting of vesicles derived from the recycling endosome. For example, Exo70, a component of the exocyst complex, colocalizes with Arf6 in Rab11-positive endosomes.²⁰⁸ Exo70 interacts with FIP3 and FIP4 biochemically, and depletion of Exo70 impairs FIP3 and Rab11 localization to the furrow and midbody.²⁰⁸ Together these studies suggest the following model of delivery of endosomal vesicles to the midbody (Fig. 4). Rab11 first recruits FIP3 to endosomes. FIP3 in turn associates with ARF6, and together this complex localizes to the midbody via interactions with the exocyst and MKLP1.

Membrane Fusion During Abscission

Following vesicle targeting to the site of abscission, membrane fusion is necessary to complete cytokinesis. SNARE proteins are critical components required for membrane fusion (reviewed in ref. 217). Several SNARE proteins or associated components have been implicated in cytokinesis completion in different organisms.^{92, 218–220} In mammalian cells, two SNARE proteins, syntaxin 2 and endobrevin/VAMP-8, localize to the midbody during cytokinesis.^{191, 221, 222} Expression of dominant negative mutants or depletion of SNAREs impairs abscission, but has no effect on ingression of the cleavage furrow, suggesting that SNARE-mediated fusion is required only in the latest steps of cytokinesis.^{191, 221} Septin proteins may assist in membrane fusion by restricting the diffusion of membrane-associated components such as the exocyst to the region of abscision.¹⁵³ Furthermore, septins may assist in abscission by directly recruiting the exocyst²²³ and SNARE proteins.²²⁴

Role of the ESCRT Machinery

Recently, protein subunits of the Endosomal Sorting Complex Required for Transport (ESCRT) that are normally involved in late endosome to lysosome trafficking have also been implicated in abscission. These proteins are best known for their roles in multivesicular body formation (reviewed in ref. 225), where they are important for membrane invagination. ESCRT complexes also play important roles in the topologically equivalent process of viral budding. Because abscission likely requires changes in membrane organization, a role for the ESCRT complex in cytokinesis is very intriguing. However, the precise mechanism of membrane invagination mediated by the ESCRT complex remains unknown, and it is unclear whether the ESCRT pathway functions independently in abscission or whether it assists in secretory- or endosomal vesicle-mediated cytokinesis completion.

Components of the ESCRT machinery localize to the midbody, and inhibition of some ESCRT complexes blocks late steps in cytokinesis. For example, CHMP3, a subunit of the ESCRT-III complex, localizes to the midbody, and deletion of a C-terminal autoinhibitory domain of CHMP3 inhibits cytokinesis.²²⁶ Other subunits of the ESCRT machinery implicated in abscission include tumor-susceptibility gene 101 (Tsg101), a subunit of the ESCRT-I complex, and Alix, an ESCRT-associated protein.^{222, 227} Alix may interact with actin and microtubules,^{228, 229} establishing a link between the ESCRT machinery and cytoskeletal components that are present at the midbody. Alix and Tsg101 are recruited to the midbody by interaction with centrosome protein 55 (Cep55), a centrosome and midbody protein essential for abscission.^{222, 230, 231} Interestingly, Tsg101 has been implicated in cancer, and may have additional functional roles in cell cycle and transcriptional regulation (reviewed in Ref 225).

Regulation of Cytokinesis

Thus far we have outlined the core pathways and components essential for each stage of cytokinesis. In the remaining part of the chapter, we discuss how these components are regulated to ensure that cytokinesis occurs at the proper place and time. Many regulatory pathways impinge upon the cytokinesis machinery, suggesting that cytokinesis may be responsive to a variety of different cues within the cell. The complexity of cytokinesis regulation suggests that cytokinesis failure could result from alterations in the activity of these regulatory pathways.

Regulation of Cytokinesis by Protein Kinases

Cytokinesis is regulated by mitotic protein kinases, including cyclin-dependent kinases (CDKs), Polo kinase (Plk1), and the Aurora B kinase complex (Fig. 5). Mitotic CDK activity prevents cytokinesis onset until anaphase by phosphorylating cytokinesis components in a manner that inhibits their activity. For this reason, CDK1 must be inactivated for cytokinesis to proceed.²³² In fact, inhibition of CDK1 with a small molecule is sufficient to induce the initial events of cytokinesis,^233–235 suggesting that CDK1 inactivation is the trigger for cytokinesis initiation.

Figure 5. Regulation of cytokinesis by mitotic kinases.

A major function of CDK1 is to prevent precocious cytokinesis before proper chromosome segregation. CDK1 thus negatively regulates some of the main players of cytokinesis. At the same time, CDK1 plays a positive role in cytokinesis by releasing cytokinesis proteins from the Golgi apparatus, and by facilitating binding of Plk1 to its substrates. Plk1 and Aurora B phosphorylate substrates that are important for both early and late steps of cytokinesis. Solid arrows indicate phosphorylation; dashed arrows indicate changed protein localization; dotted arrows indicate protein interactions. IF, intermediate filaments.

In contrast, Polo kinase and Aurora B kinase positively regulate the events of cytokinesis, and must remain active for a period of time following CDK inactivation to promote cytokinesis. This period of the cell cycle, which lasts for about an hour in HeLa cells, has been referred to as “C-phase”.^{3, 236} C-phase is initiated by inactivation of CDK1, mediated by cyclin destruction catalyzed by the APC/C. At later times following anaphase, the APC/C also ubiquitinates other proteins that are essential for cytokinesis, including anillin,¹³² Polo kinase,²³⁷ and Aurora B.²³⁸ Thus the APC/C may also be responsible for terminating C-phase, an idea that is consistent with the finding that treatment of cells with proteasome inhibitors doubles the duration of C-phase.⁷³

Regulation of Cytokinesis by CDK Activity

Because CDK1 is a central negative regulator of cytokinesis, it is possible that failure to fully inactivate CDK1, perhaps as a consequence of failure to fully degrade mitotic cyclins, could inhibit cytokinesis at some step. One setting in which this might occur is in cells that become arrested in mitosis due to persistent activation of the spindle checkpoint, which normally inhibits the APC/C until chromosomes become properly aligned and attached at the metaphase plate.²³⁹ Prolonged activation of the checkpoint may result in abnormal mitotic exit, resulting in incomplete activation of APC/C, or improper timing of degradation of different substrates, leading to cytokinesis failure. There is also evidence that APC/C activity may be spatially regulated within the cell, with the subpopulation of APC/C that is associated with the spindle poles remaining inhibited until later stages of mitosis.²⁴⁰ It is therefore possible that perturbation of spindle organization could interfere with the timing of degradation of mitotic regulators, thus perturbing cytokinesis.

CDK1 activity restrains multiple steps in cytokinesis. Cytokinesis initiation is inhibited because the RhoA pathway is kept inactive. This is a consequence of phosphorylation of ECT2 by CDK1 at a site that blocks its association with MgcRacGAP.^{19, 26} In addition, myosin light chain is phosphorylated by CDK1 at sites that inhibit myosin activation,⁸² and high CDK1 activity also inhibits cortical recruitment of myosin.²⁴¹ Central spindle formation is also inhibited by CDK1-dependent phosphorylation. Phosphorylation of PRC1 by CDK1 inhibits its ability to bundle microtubules^{32, 166, 242} and its ability to interact with Plk1.¹⁷⁰ CDK1 also phosphorylates MKLP1, inhibiting its motor activity by reducing its affinity for microtubules.²⁴³ Thus CDK1-dependent phosphorylation acts at many steps to block cytokinesis.

Though CDK1 restrains cytokinesis onset, CDK1-dependent phosphorylation is also essential for cytokinesis because it primes Plk1-dependent phosphorylation that occurs during early stages of cytokinesis. CDK1 activity is also important during mitosis to promote dissociation of cytokinesis proteins from cellular organelles that would otherwise sequester the protein. For example, the protein Nir2 is required for cytokinesis, and must dissociate from the Golgi apparatus in order to participate in cytokinesis; dissociation is mediated by CDK1-dependent phosphorylation.¹⁷⁹ CDK1-dependent phosphorylation is also important for dissociation of Cep55 from the centrosome, and its subsequent phosphorylation by Plk1.²³¹

Regulation by Polo Kinase

Polo kinase is an essential positive regulator of cytokinesis in multiple organisms. In mammalian cells, Plk1 localizes to the midzone during anaphase and to the midbody during telophase and cytokinesis,²⁴⁴ and plays an essential role in the initiation of cytokinesis.^245–248 Plk1 activity is required for recruitment of itself and ECT2 to the central spindle, and inhibition of Plk1 with small molecule inhibitors abolishes RhoA GTPase localization to the equatorial cortex, suppressing cleavage furrow formation.^246–248 Plk1 also appears to be important for the interaction between ECT2 and MgcRacGAP.²⁴⁷ Other evidence suggests that Plk1 may bind to ECT2 in a CDK1-dependent manner.²⁴⁹ Another study using a distinct Plk1 inhibitor demonstrated that when Plk1 is inhibited, it spreads over the arms of chromosomes, resembling the localization of its binding partner PICH.²⁴⁵ Therefore, Plk1 activity is required for its own proper localization during cytokinesis, and also for recruitment and activation of RhoA.

Plk1 activity may be required for later steps in cytokinesis as well, as Plk1 is targeted to the central spindle by the motor protein MKLP2, and phosphorylation of MKLP2 by Plk1 is required for cytokinesis.²⁵⁰ Phosphorylation of MKLP2 by Plk1 may be necessary for the spatial restriction of Plk1 to the central spindle during anaphase and telophase, although interaction with PRC1 also appears to be important for docking of Plk1 to the central spindle.¹⁷⁰ Plk1 may also interact with and phosphorylate MKLP1 during cytokinesis,^{244, 251} although others suggest that Aurora B may be the relevant kinase.²⁵² Cep55 also appears to be a Plk1 substrate whose phosphorylation is primed by CDK1 but the consequences of this phosphorylation remains unknown.²³¹ A recent proteomic screen identified a large number of proteins that bind to the Polo-box domain of Plk1, including the Rho kinase ROCK2,²⁵³ where Plk1 and RhoA may function together to enhance ROCK2 activity. Another substrate of Plk1 that may be involved in regulation of mitosis and cytokinesis is NudC,^{254, 255} a dynein/dynactin associated protein that is essential for midzone formation and cytokinesis completion in C. elegans and mammalian cells.²⁵⁴

Plk1 is overexpressed in a broad range of human tumors (for review see Ref. 256). Overexpression of Plk1 in HeLa cells leads to an increase of cells with large, often fragmented nuclei or multiple nuclei²⁵⁷ as well as centrosome amplification,²⁵⁸ suggesting that increased expression of Plk1 observed in some tumors may have an effect on cytokinesis completion as well as chromosome segregation. This finding has been corroborated in human primary cells.²⁵⁹

Regulation by Aurora B and the Chromosome Passenger Complex

The chromosome passenger complex (CPC) consists of the proteins Aurora B, INCENP, survivin, and borealin. The complex plays many important roles throughout mitosis, and has been implicated in the regulation of cytokinesis (see ref. 260 for review). At the metaphaseanaphase transition, the CPC relocalizes from centromeres to the spindle midzone and the equatorial cortex,^261–263 and ultimately concentrates near the midbody, adjacent to the centriolin ring.¹⁹¹ MKLP2, a kinesin-6 family motor protein, is required for relocalization of Aurora B, and also Plk1, to the central spindle in human cells.^{250, 252, 264, 265}

Aurora B activity is necessary for several events in cytokinesis (Fig. 5). First, Aurora B is required for proper localization and function MKLP1. Treatment of human cells with a small molecule inhibitor of Aurora B in early mitosis inhibits localization of MKLP1 (and its binding partner MgcRacGAP) to the central spindle.²⁶⁶ However, addition of an Aurora inhibitor at later stages of mitosis inhibits phosphorylation of MKLP1 without disrupting its localization,^{252, 267} yet perturbs cytokinesis completion, indicating that MKLP1 must remain phosphorylated to permit abscission. How phosphorylation regulates MKLP1 is not completely clear, as MKLP1 is phosphorylated at multiple sites that may have distinct effects.^{252, 267} Phosphorylation of MKLP1 could be important for stabilizing interactions between the cortex and midbody, or be important for recruiting proteins such as centriolin that are necessary for abscission. In addition, phosphorylation of MKLP1 by Aurora B is important to prevent the protein from being sequestered back in the nucleus as cells enter interphase.²⁵² It is interesting to note that many components required for cytokinesis are located in the nucleus or associated with Golgi apparatus during interphase (Fig. 2), and thus localization of cytokinesis components to the midbody could require sustained phosphorylation that prevents the proteins from being resequestered by these structures as cells exit mitosis.

Another important substrate of Aurora B is MgcRacGAP, whose phosphorylation appears important for completion of cytokinesis.^{47, 268, 269} Phosphorylation of MgcRacGAP has been proposed to stimulate its activity as a GAP for RhoA, which could be important for terminating RhoA activity in late stages of cytokinesis.⁴⁷ Another study indicates that phosphorylation of MgcRacGAP by Aurora B at a different site might activate the protein by stimulating release of the GAP domain from an inhibitory interaction with PRC1.²⁶⁸

There are several other important Aurora substrates include vimentin, an abundant intermediate filament protein. Intermediate filaments must be disassembled during mitosis to allow cell division, and mitotic phosphorylation is important for filament dissociation, as expression of nonphosphorylatable mutants of vimentin leads to cells that show a persistent filamentous bridge.^{270, 271} Following mitotic exit and during later stages of cytokinesis, ROCK²⁷² and Aurora B^{112, 270, 271, 273} maintain vimentin phosphorylation after CDK1 is inactivated. Aurora B may also promote cytokinesis by inhibiting myosin light chain phosphatase.¹¹² Aurora B also phosphorylates CENPA, which appears to play an important role in cytokinesis.²⁷⁴ Cells expressing mutants of CENPA that cannot be phosphorylated result in mislocalization of the passenger complex, and cause a delay in the final stages of cytokinesis.

Although Aurora B is a critical positive regulator of cytokinesis in vertebrate cells, this role does not seem conserved in yeast as the budding yeast ortholog Ipl1 and the fission yeast ortholog Ark1 are not essential for cytokinesis. However, in budding yeast, Ipl1 may negatively regulate late steps of cytokinesis in cells with spindle defects,^{275, 276} perhaps by regulating the localization of anillin-like proteins. This pathway may prevent abscission until segregating chromosomes have cleared the midzone. Whether a similar pathway operates in mammalian cells is not yet clear.

We are just beginning to learn about the mechanisms that regulate Aurora B activation. New work suggests that the TD60 protein may play an important role in activating Aurora B at centromeres.²⁷⁷ In multiple organisms, components of the mitotic exit network, including the phosphatase CDC14, play important roles in regulating cytokinesis,^278–280 in part through regulation of targeting of the Aurora B complex.²⁸¹ Aurora B is also regulated by a Cul3-containing ubiquitin ligase, which is important for removing Aurora B from mitotic chromosomes and allowing its accumulation on the central spindle during anaphase.²⁸²

In vertebrate cells, Aurora B is expressed in a cell-cycle dependent manner, peaking in the G2/M phase of the cell cycle,^{283, 284} and is highly expressed in a number of cancers.^{263, 284–291} However, in these tumors, expression of other proliferative markers, such as Ki-67, MCM2, geminin, and Aurora A is also increased,^{291, 292} suggesting that Aurora B upregulation may be part of a broader upregulation of mitotic components in tumor cells. Because Aurora B is a positive regulator of cytokinesis, it is unclear whether its overexpression would perturb cytokinesis. Elevated levels of Aurora B may promote cytokinesis completion in cells that would otherwise undergo cytokinesis failure due to other abnormalities in the mitotic machinery. However, it is possible that perturbing Aurora B expression, or other components of the CPC, could alter the stoichiometry of the complex, perturbing cytokinesis.²⁹³

Regulation of Cytokinesis by Tyrosine Kinases

Tyrosine kinase signaling pathways may also regulate cytokinesis completion. Small molecule inhibitors of Src, including PP2 and SU6656, inhibit abscission in HeLa cells.²⁹⁴ Src activity appears to be required in early mitosis, followed by delivery of tyrosine-phosphorylated proteins to the midbody via Rab11-driven vesicle transport.²⁹⁴ Src co-localizes with the diaphanous-related formins mDia1 and mDia2 in endosomes and midbodies of dividing cells, and inhibition of Src blocks cytokinesis.⁶⁷ Other tyrosine kinases, such as Fyn and its associated proteins are required for cytokinesis in lymphocytes,²⁹⁵ through mechanisms that remain obscure.

Regulation of Cytokinesis by Lipids

Several studies indicate that phosphoinositide-containing lipids may be important for cytokinesis, with phosphatidylinositol-4,5-bisphosphate (PtdIns(4,5)P2) playing a central role. In mammalian cells, PtdIns(4,5)P2 accumulates at the cleavage furrow, and overexpression of proteins that bind to PtdIns(4,5)P2 perturbs cytokinesis completion but not ingression of the cleavage furrow.²⁹⁶ Overexpression of dominant negative kinases required for PtdIns(4,5)P2 generation also inhibits cytokinesis completion in mammalian cells.²⁹⁶

Hydrolysis of PtdIns(4,5)P2 by PLC yields inositol trisphosphate (IP₃), which stimulates calcium release from internal stores. Inhibitors of PLC can interfere with cytokinesis,²⁹⁷ which can in some cases be rescued by addition of calcium.²⁹⁸ Alternatively, PtdIns(4,5)P2 may play a direct role in recruiting membrane proteins required for stability of the furrow. Several proteins required for cytokinesis, such as septin, profilin, and anillin can bind to PtdIns(4,5)P2, and overexpression of a protein containing a PtdIns(4,5)P2-binding domain blocks cytokinesis completion by interfering with adhesion of the plasma membrane to the contractile ring at the furrow.²⁹⁶

Other studies have shown that the membrane lipid phosphatidylethanolamine (PE) is exposed on the cell surface of the cleavage furrow during late cytokinesis.²⁹⁹ Addition of a cyclic peptide that binds tightly to PE inhibits cytokinesis completion,²⁹⁹ perhaps by interfering with contractile ring disassembly.³⁰⁰ Mutant cell lines that fail to synthesize adequate PE also show defects in cytokinesis completion that can be rescued by PE addition.³⁰⁰ Proper PE organization may be essential for RhoA inactivation at late stages of cytokinesis, which may in turn be necessary for actin disassembly.³⁰¹

Coupling of Cytokinesis to Other Cellular Pathways

The complexity of cytokinesis regulation provides opportunities for linking cytokinesis to other cellular pathways. Emerging evidence suggests interesting new connections between cytokinesis and the pathways involved in regulation of protein synthesis, DNA replication, and DNA damage.

Cytokinesis and Protein Synthesis

Recent work suggests that proper regulation of protein synthesis may be essential for cytokinesis to proceed with high efficiency. Interestingly, of 214 genes identified in a genome-wide RNAi screen in Drosophila S2 cells, 22% were ribosomal proteins, and another 5% were involved in translation.¹⁷⁷ Recent work suggests that the protein 14-3-3σ may play an important role in regulating protein synthesis during mitosis.³⁰² In normal cells, cap-dependent translation is suppressed during mitosis, whereas cap-independent translation is increased. Cells lacking 14-3-3σ do not make this switch, perturbing the pattern of proteins that are synthesized.³⁰² Downregulation of 14-3-3σ perturbs localization of Plk1 to the midbody, and leads to cytokinesis failure.³⁰² These effects may be a consequence of failure to properly synthesize proteins containing an internal ribosomal entry site during mitosis, such as Cdk11.³⁰² Interestingly, 14-3-3σ expression is often reduced in tumor cells by targeted degradation or promoter hypermethylation. Loss of 14-3-3σ may in turn result in defective cytokinesis as a consequence of alterations in protein synthesis.

Cytokinesis and DNA Replication

Interestingly, one of the components of the Origin Recognition Complex (ORC), which is required for initiation of DNA replication, may also play a role in cytokinesis in metazoans. In vertebrate cells, Orc6 localizes to kinetochores and to a reticular-like structure around the cell periphery, and ultimately to the cleavage furrow and midbody.³⁰³ Elimination of Orc6 induces mutlipolar spindles and formation of multinucleated cells in both human cells³⁰³ and Drosophila,³⁰⁴ suggesting this function is conserved. In Drosophila, Orc6 interacts with a septin protein that may be important for cytokinesis. Domains of Orc6 required for DNA replication and cytokinesis appear separable, suggesting that Orc6 has evolved a domain that participates specifically in cytokinesis.³⁰⁴ How Orc6 might couple the processes of DNA replication and cytokinesis completion remains unclear.

Cytokinesis and DNA Damage

Several lines of evidence suggest that cytokinesis may be regulated in response to DNA damage. Components required for DNA repair, such as BRCA2, may be directly involved in cytokinesis. Other evidence suggests that DNA damage pathways may regulate the expression of cytokinesis proteins, or regulate their activity by post-translational modification. Coupling of DNA damage pathways to cytokinesis regulation could be important for preventing the cleavage furrow from cutting damaged DNA that cannot be accurately segregated during mitosis. The existence of such pathways may explain why spontaneous chromosome missegregation is tightly coupled to cytokinesis failure in human cells.¹⁸⁵ In this model, DNA damage, or perhaps incompletely replicated DNA, may trigger pathways that prevent segregation of unreplicated or damaged sister chromatids, and at the same time activate pathways that block cytokinesis completion.

BRCA2 is an example of a protein that may play direct roles in both DNA repair and cytokinesis. BRCA2 is required for recombination-based repair of DNA double-strand breaks.³⁰⁵ However, BRCA2-deficient cells also show centrosome amplification that may be a consequence of defective cytokinesis.³⁰⁶ BRCA2 localizes to the midbody, and inactivation of BRCA2 in murine embryonic fibroblasts and HeLa cells interferes with cytokinesis.³⁰⁷ BRCA2 may be regulated during mitosis, as it is a Plk1 substrate whose phosphorylation is inhibited in the presence of DNA damage.³⁰⁸ Interestingly, downregulation of a BRCA2-interacting protein (BCCIP) also leads to defective cytokinesis.³⁰⁹ Other proteins involved in DNA damage responses may influence cytokinesis regulation by interacting with cytokinesis components. For example, the DNA damage checkpoint kinase Rad53 has been shown to associate with septins in budding yeast.³¹⁰ In mammalian cells, Ku70, a DNA-binding protein required for DNA damage repair, forms a complex with ARF6 during mitosis,³¹¹ suggesting a possible link between the DNA damage pathway and completion of cytokinesis.

Transcriptional controls may provide another mechanism for inhibiting cytokinesis in response to DNA damage. The expression of several cytokinesis proteins, including Plk1, ECT2, anillin, and survivin, is repressed when DNA is damaged, in a manner that depends on an intact Rb pathway.³¹² Other studies suggest that expression of cytokinesis proteins may be inhibited by activation of the p53 pathway.³¹³ For example, it has been shown that ECT2 expression is repressed by p53 via protein methyltransferases, suggesting that cytokinesis could be more likely to fail under conditions of p53 activation.³¹⁴

Post-translational modifications may also regulate cytokinesis in response to DNA damage. For example, Aurora B becomes highly poly-ADP-ribosylated when DNA is damaged, a modification that inhibits its kinase activity.³¹⁵ Poly(ADP-ribosyl)ation is an immediate cellular response to DNA strand breaks that is catalyzed by NAD⁺-dependent enzymes, poly(ADP-ribose) polymerases (PARPs).³¹⁶ This effect is mediated by direct interaction between the BRCT domain of PARP1 and Aurora B.³¹⁵ Because Aurora B activity is essential for chromosome segregation and cytokinesis, induction of DNA damage could lead to errors in chromosome segregation and failure of cytokinesis.

Conclusions

Cytokinesis is a surprisingly complex process that requires the interplay of many components and regulatory pathways. Cytokinesis failure can arise through defects in any of the four stages in cytokinesis, and as a consequence of inactivation or hyperactivation of any of a large number of different components (summarized in Fig. 6). Although many cytokinesis proteins have been identified, we are just beginning to understand how these proteins interact with one another, and how they are regulated. Understanding the causes of cytokinesis failure is important, as it may set the stage for genetically unstable tetraploid cells that give rise to tumors.³¹⁷ However, cytokinesis failure also seems to occur physiologically in some tissues, even in those that are not tumor prone such as the heart. Understanding how cytokinesis is regulated physiologically in response to different signals, or under conditions of cell stress or damage, remains an important area for future research. Although cytokinesis failure may be accompany certain pathological states such as cancer, it is likely that pharmacologically-induced cytokinesis failure may be an important issue to consider as new medicines are developed. Inhibitors of Rho kinase are being developed for cardiovascular medicine,³¹⁸ and inhibitors of Aurora kinase are under development as anticancer agents.³¹⁹ Because these compounds are likely to induce cytokinesis failure in normal tissues, it will be important to determine how sensitive various tissues are to cytokinesis failure, and the consequences of production of tetraploid cells in different tissue types.

Figure 6. Summary of different phenotypes resulting from cytokinesis failure.

Inhibition (downward arrow) or excessive activation (upward arrow) of different cytokinesis components can give rise to distinct phenotypes, including precocious ingression before the chromosomes have been separated, regression of the furrow giving rise to binucleated cells, or stabilization of the cytoplasmic bridge where daughter cells remain connected. The list is not comprehensive; see text for additional examples.