Cleavage furrow formation in animal cells results from a local increase in cortical contractility. During anaphase, the spindle contains, in addition to astral arrays of microtubules, a set of bundled microtubules known as the central spindle. Each of these populations of microtubules, the astral arrays and the central spindle bundles, is sufficient to direct cleavage furrow formation, yet in wild-type situations these sets of microtubules co-operate to induce furrow formation at the same site, between the segregating chromosomes. These pathways have distinct genetic requirements that reflect their differential control of cortical actomyosin. We review our current understanding of the molecular mechanisms of furrow formation, with particular emphasis on the central spindle-independent pathway.
Cytokinesis, the physical separation of nascent daughter cells, is one of life's most fundamental processes. The underlying molecular requirements appear to be well conserved among animal cells since all major model systems require a core set of molecules essential for cytokinesis (for reviews, see [1,2]). Chief among those molecules are actin and myosin, which assemble into a ring-like structure known as the contractile ring that forms at the presumptive site of furrow ingression. The progressive sliding of myosin minifilaments along actin filaments is believed to provide the contractile force that drives furrow ingression, although the exact mechanism is still subject to debate (see  for review). Spatial and temporal regulation of actin polymerization and myosin motor activity are therefore crucial for the proper execution of cytokinesis. The anaphase spindle plays an essential role in positioning the site of furrow ingression through the local regulation of the activity of the small GTPase RhoA and its ultimate effectors, F-actin and myosin II. Cytokinesis has to be versatile to allow for the division of cells of different size, shape or developmental context and many observations indicate that parallel pathways contribute to cytokinesis. In this review, we will focus on recent advances in understanding the mechanisms of these pathways.
RhoA: a central regulator of cytokinesis
RhoA regulates contractile ring formation by directly activating the actin assembly factor formin and by indirectly promoting phosphorylation of the regulatory light chain of myosin II (Figure 1). Activation of RhoA is therefore a central event in this process. The primary activator of RhoA during cytokinesis is the RhoGEF (guanine-nucleotide-exchange factor) ECT2 (for detailed review, see ). There appear to be at least two types of RhoGAP (GTPase-activating proteins) with distinct roles during cytokinesis: RGA-3/4 and CYK-4. RGA-3 and -4 are two Caenorhabditis elegans genes that encode highly related proteins due to their recent duplication. Loss of function analysis reveals that RGA-3/4 suppress RhoA activity during early embryogenesis [5,6]. However, clear structural or functional orthologues of RGA-3/4 have not yet been characterized in other species. In contrast, the RhoGAP CYK-4 is well known for its evolutionarily conserved roles in central spindle assembly and regulation of ECT2 .
Factors affecting furrow initiation
Central spindle dependent regulation of cytokinesis by local activation of RhoA
The central spindle has emerged as a critical regulator of cytokinesis. This structure is composed of an array of antiparallel bundles of microtubules situated between the two spindle poles during anaphase. Centralspindlin, a heterotetrameric complex consisting of parallel homodimers of both the kinesin-6 motor MKLP1 (mitotic kinesin-like protein) and the RhoGAP CYK-4 [7,8], is essential for central spindle assembly in all animal systems analysed so far [7,8]. Previous studies in human cells have shown that the aforementioned RhoGEF ECT2 localizes to the central spindle by binding to the CYK-4 subunit of centralspindlin [9–13]. This interaction appears to promote the activation of ECT2, which is relieved from an autoinhibitory state . Localized ECT2 activation triggers the accumulation of active RhoA in cortical regions surrounding the central spindle and promotes the recruitment of actin and myosin and the assembly of the contractile ring. The interaction between ECT2 and CYK-4 is promoted by the kinase Plk1 (polo-like kinase 1) which also localizes to the central spindle during anaphase [15–18]. Plk1 binding to the central spindle is mediated by the microtubule-associated protein PRC1 . Many of these protein–protein interactions are negatively regulated by the cyclin-dependent kinase CDK1, providing temporal control of contractile ring assembly and cytokinesis [19–22].
These results suggest that the central spindle acts as a self-assembling scaffold that concentrates key regulators of RhoA, thereby contributing to the spatial and temporal regulation of cytokinesis. The details of this mechanism has been recently reviewed  and this mechanism will not be discussed further, rather we will focus on alternative pathways for furrow formation.
Parallel pathways for furrow formation
ECT2-dependent activation of RhoA is essential for cytokinesis. However, formation of a central spindle is not an essential prerequisite for furrow initiation, although it is essential for proper completion of cytokinesis [23–25]. Evidence for a central spindle-independent pathway for furrow induction came from genetic and laser ablation experiments in the single cell C. elegans embryo. Neither depletion of the MKLP1 orthologue ZEN-4 (required for central spindle assembly) nor depletion of GPR-1/2 (required for establishment of cortical pulling forces and therefore proper spindle elongation) prevents cleavage furrow formation. However, if both central spindle formation and spindle elongation are perturbed, cytokinesis is blocked . An analysis of microtubule distribution in fixed specimens revealed that a local minimum of microtubule density forms at the site of furrow ingression. Formation of this local minimum requires separation of the opposing asters by cortical pulling forces or that antiparallel microtubules are bundled into the spindle midzone. Similarly, laser-severing experiments showed that two signals, one associated with the central spindle and one with spindle asters, can each induce furrows in one embryo . These observations suggest that a signal associated with the central spindle triggers furrowing in C. elegans, in parallel with a central spindle-independent mechanism regulated by astral microtubules.
Parallel pathways for furrowing have not only been demonstrated in C. elegans embryos but also in cultured human cells . However the two pathways have been analysed to different extents in the two systems and it is not clear if they are similar in all details. For example, the ECT2/CYK-4 interaction has not yet been confirmed in C. elegans embryos. This is interesting as depletion of CYK-4 in C. elegans embryos does not completely block cortical contractility, whereas depletion of CYK-4 has this effect in human cells [12,25]. This suggests that ECT2-mediated activation of RhoA may not be solely dependent on CYK-4 in C. elegans embryos. Conversely, the mechanistic details of the astral pathway have only begun to be characterized in human cells.
A local signal associated with astral microtubules that induces central spindle-independent furrow ingression?
There are two widely discussed classes of models for astral-mediated cytokinesis. Either astral microtubules can provide a local positive cue for furrow ingression at the cell equator, inducing the local recruitment of contractile components in a manner akin to the central spindle dependent signal. Alternatively, astral microtubules could inhibit cortical contractility in the polar cortex of the embryo, leading to a relative increase in equatorial contractility.
Classical experiments designed to distinguish between a positive equatorial stimulus and an inhibitory signal at poles overwhelmingly supported the former model . However, most of those experiments did not resolve the contributions from the central spindle (a functional equivalent of which could form wherever microtubules are arranged in an antiparallel configuration) and therefore could not unambiguously distinguish between these two possibilities. More recently, this issue has been reanalysed with modern tools and with recognition of the contribution of the central spindle. Some support for the existence of a local signal for astral-induced furrowing came from an analysis of GFP (green fluorescent protein)–RHO-1 and GFP–ABD (moesin's actin binding domain) to monitor localization of contractile ring components during anaphase . Both probes localize to the equatorial cortex shortly before onset of furrow ingression in wild-type embryos. In embryos defective in both proper spindle elongation and central spindle assembly, RhoA and actin detectably, but transiently, accumulate to the equatorial cortex. However, those embryos fail to form a cleavage furrow, suggesting that this localization is insufficient to trigger the formation of a functional contractile ring.
Centralspindlin has been reported to associate with the cortex at the site of furrow ingression, independent of its prominent association with the central spindle. The PRC1 orthologue, SPD-1, is required for formation of a well-organized central spindle, but dispensable for cytokinesis . Following on from this report, Verbrugghe and White  recently reported that SPD-1 and GPR-1/2 depleted embryos form cleavage furrows, although spindle elongation is equally compromised in ZEN-4 and GPR-1/2 depleted embryos, which do not furrow. The former embryos contained a cortical pool of centralspindlin and the latter did not. The authors suggested that these results were incompatible with the view that furrow induction is mediated by parallel pathways involving the central spindle and a local reduction in microtubule density. Rather, they proposed that a central spindle-independent, cortical pool of centralspindlin is responsible for inducing furrow formation. Implicit in this interpretation is that SPD-1 depletion ablates central spindle formation. This is not likely, as centralspindlin has been detected on the central spindle in fly and mammalian cells depleted of the SPD-1 orthologue PRC1 and even in C. elegans embryos [32–35]. Moreover, functional differences in furrow formation have been documented between spd-1 embryos and embryos in which the spindle midzone is ablated with a laser . Thus the most parsimonious explanation for these findings is that a poorly organized central spindle participates in furrow formation in SPD-1 depleted embryos and the furrow-localized pool of centralspindlin may be derived from microtubule bundles that are deformed by the ingressing furrow.
A screen for genes required for cytokinesis in spd-1 embryos highlighted the DEP (dishevelled, Egl-10 and pleckstrin homology)-domain-containing protein LET-99 . LET-99 is not generally required for cytokinesis , but LET-99 depletion in spd-1 mutants causes incomplete furrowing and LET-99 depletion in embryos with ablated central spindles abolishes furrowing altogether. Interestingly, LET-99 localizes to a cortical band overlapping with the presumptive site of furrow ingression and this localization is dependent on the position of the mitotic spindle during anaphase [36,38]. Furthermore, LET-99 localization is dependent on PAR-2 and GPR-1/2, which are both essential for cytokinesis in the absence of the central spindle. LET-99 regulates localization of GPR-1/2, leading the authors to hypothesize that G-protein signalling is required for central spindle-independent cytokinesis. However, LET-99 and GPR-1 are also required for the cortical pulling forces that mediate spindle elongation, so the role of LET-99 in central-spindle-dependent furrowing could be indirect . However, two alleles of LET-99 cause similar defects in spindle elongation, but only one fails to initiate a furrow when the midzone is laser ablated. It is not known how those two alleles affect the distribution of microtubules near the cortex and whether a difference in spindle elongation might be able to account for the differences in phenotype. Also, depletion of ZEN-4 in both LET-99 strains completely blocks cytokinesis (M. Werner and M. Glotzer, unpublished work). Further characterization of the let-99 alleles is required to determine the extent to which their spindle elongation defects contribute to their cytokinesis phenotypes. Depletion of LET-99 also affects the asymmetry of furrow ingression, reduces the residence time of cortical myosin foci, and delays the onset of furrow ingression [36,40], suggesting that LET-99 could affect the organization of the cortical actomyosin meshwork as well as modulating spindle elongation.
At this juncture, the evidence for a local stimulatory signal at the equatorial cortex from astral microtubules to induce cytokinesis independent of the central spindle is inconclusive, and further analysis, especially of the role of LET-99 in astral-mediated cytokinesis, is required.
Can negative regulation of cortical contractility by astral microtubules explain the formation of central-spindle-independent furrows?
Spindle repositioning experiments have led to numerous insights into cytokinesis. In the early C. elegans embryo, repositioning the spindle to the posterior leads to the formation of two distinct furrows: one in the anterior third of the embryo and the other in the posterior . The former shares strong similarities with the astral microtubule-dependent furrow and the latter with the central-spindle-dependent furrow. Interestingly, from the perspective of myosin dynamics, the central spindle-independent furrow and the astral-mediated anterior furrow observed in embryos with posterior spindles show striking similarities to the PCF (pseudocleavage furrow) that forms during polarity establishment in C. elegans embryos . Shortly after fertilization, the sperm centrosome migrates to the posterior pole and induces a local relaxation of the cortex, leading to an increase in anterior cortical contractility that ultimately becomes organized into a single furrow, a process known as pseudocleavage (for review see ). In all three of these cases, myosin assembles into large foci with similar size and spacing that appear to be interconnected in a meshwork-like structure. When the distribution of these foci is biased towards the anterior cortex, these foci move in a co-ordinated manner reflecting interlinked contractile elements. This contractile behaviour results in the formation of a deeply ingressing furrow in all three cases. The furrow forms at the boundary of the contractile anterior cortex rich in myosin foci and the less contractile posterior cortex largely devoid of myosin foci. In all three cases, furrow formation is sensitive to reduction of formin levels which disturbs the organization of the cortical actomyosin meshwork [40,43]. This indicates that each of these three types of furrows requires an organized cortical meshwork. Critically, the cortical distribution of these foci inversely correlates with levels of microtubule densities in wild-type embryos and embryos with posterior spindles.
We interpret these observations as follows: through a mechanism that is not yet understood, astral microtubules inhibit cortical contractility. This inhibitory mechanism, coupled with the cell cycle-regulated activation of RhoA, causes the anaphase accumulation of myosin foci in the embryo cortex. These foci accumulate at cortical sites that are relatively free from microtubules. In wild-type embryos or embryos with disrupted central spindles, this corresponds to the anterior pole and the equatorial region. In embryos with a posterior spindle, this corresponds to the anterior half of the embryo. These foci form an organized meshwork that contracts to form a single furrow at the boundary of contractile and less contractile cortex. This behaviour is independent of the central spindle and is similar to that observed during PCF formation. Remarkably, the anterior bias of myosin localization during polarization that causes PCF formation is initially induced by locally inhibiting cortical contractility by a signal associated with either the centrosome  or, as recent data suggests, its associated microtubules . We propose that the PCF and the various central spindle-independent furrows have a similar mechanistic basis.
This model predicts that regulators of cortical myosin dynamics required for PCF formation should also affect aster-mediated cytokinesis (Figure 2). Indeed, depletion of ECT-2 or RhoA blocks both PCF and cytokinesis [25,26]. Conversely, simultaneous depletion of RGA-3/4 leads to cortical hypercontractility, an increase of cortical myosin levels and mispositioning of the PCF [5,6]. In parallel, the actual cytokinetic furrow is over-pronounced and sometimes misplaced, suggesting that a hyperactive cortical actomyosin meshwork can affect proper progression of cytokinesis.
Summary of the morphological and genetic similarities between the pseudocleavage furrow and the furrow that forms during anaphase in embryos lacking the central spindle
Another gene involved in cortical myosin dynamics in the early embryo is ani-1, one of three C. elegans homologues of human anillin. Anillin-like proteins are present in yeast and metazoans. Mid1p, a fission yeast anillin, is the first component of the contractile ring to localize to the equatorial cortex . The C. elegans protein, ANI-1, localizes to cortical myosin foci during polarization as well as to the ingressing furrow during cytokinesis . When ANI-1 is depleted by RNAi (RNA interference), formation of those foci is prevented and PCF is blocked. Furthermore ANI-1 is required for asymmetric ingression of the cleavage furrow, which enhances the robustness of cytokinesis . The similarities between PCF and astral-mediated furrowing would suggest an anillin requirement for the latter process. Indeed, we observed that cleavage furrow ingression is severely compromised when ANI-1 is depleted in the absence of the central spindle, and the formation of cortical myosin foci upon anaphase onset is perturbed in the absence of ANI-1 (Figure 3).
Anillin is required for robust focal accumulation of cortical myosin during pseudocleavage and anaphase
Taken together these data suggest that in C. elegans, two fundamentally different mechanisms for regulating cortical myosin distribution act redundantly to induce furrowing. While the central spindle-induced furrow is most likely the consequence of a local accumulation/activation of myosin in the surrounding cortical region, astral-induced furrowing is driven by the organized contraction of an organized cortical actomyosin meshwork whose asymmetric distribution is essential for furrow formation. This asymmetry is achieved by the relative inhibition of myosin recruitment at cortical regions of high microtubule densities. Although the astral pathway is sufficient to induce furrowing, it is not sufficient for cytokinesis, as these furrows cannot progress to completion in the absence of a central spindle. The astral pathway therefore is a secondary pathway that can accelerate furrowing, perhaps being of critical importance to large cells where the central spindle may be far from the cell cortex.
Human anillin is required for central- spindle-independent cytokinesis by restricting contractile ring components to the equatorial cortex
Two distinct mechanisms also contribute to furrowing in human cells. As in C. elegans embryos, the central spindle is dispensable for furrow initiation in human cells . However, furrow ingression in the absence of MKLP1 correlates with formation of an equatorial zone of active RhoA, although this zone broadens significantly when MKLP1 is depleted. This suggests that central spindle-independent mechanisms refine the localization of active RhoA, and therefore contractile ring components, to the equatorial cortex. One such mechanism involves anillin. Anillin localization overlaps with that of contractile ring components at the cell equator during anaphase, yet depletion of anillin in human cells by RNAi does not prevent furrow ingression [13,49–51]. However, as furrow ingression occurs, anillin-depleted cells undergo extensive oscillations and cytokinesis ultimately fails. Structure–function analysis indicates that anillin's actin- and myosin-binding domains are collectively required for anillin function . The C-terminus of anillin has sequence homology to the RhoA-binding protein Rhotekin. This homology spans a conserved domain, the AHD (anillin homology domain) and a PH (pleckstrin homology) domain. Moreover, RhoA can bind to anillin via the AHD in vitro and in vivo. As furrow ingression proceeds in anillin-depleted cells, both actin and myosin become mislocalized from equatorial to polar regions of the cell, most likely causing the cortical oscillations. These data suggest that anillin acts as an assembly scaffold for contractile ring components and RhoA, retaining them at the equatorial cortex as cytokinesis progresses. Interestingly, the ability of cells lacking a central spindle to form a cleavage furrow is dependent on anillin, since depletion of both anillin and MKLP1 leads to a complete block of cytokinesis and both actin and myosin are mislocalized all around the cell . Thus, partially redundant pathways also exist in human cells to localize and maintain contractile ring components at the equatorial cortex, assuring proper initiation and completion of cytokinesis.
Interestingly, when human cells containing monopolar spindles are forced into anaphase, a cleavage furrow forms on one side of the monopolar spindle . How do these spindles induce a furrow in the absence of a bipolar central spindle? One possibility is that the monopolar spindle is asymmetrically positioned and furrow induction is mediated by the astral inhibition mechanism. Alternatively, or in addition, centralspindlin or other regulators of central-spindle-dependent cytokinesis may localize to the tips of microtubules. Indeed, INCENP (inner centromere protein) concentrates on the microtubules that extend beyond the chromosomes, near the site of furrow ingression . Some of these divisions complete furrowing, suggesting that centralspindlin probably also accumulates on these microtubules. Since monopolar spindles are often radially symmetric and symmetrically positioned during mitosis, yet the cells always become asymmetric during anaphase, it is important to determine the origin of the asymmetry in these cells. Perhaps a slight imbalance in cortical contractility induces a reorganization of the monopolar spindle so that it is no longer radially symmetrical, which could in turn amplify the differences in cortical contractility. This would be reminiscent of the interplay between central spindle and cortex previously described in Drosophila cells , and may reflect the recently reported interaction between anillin and the Cyk-4 orthologue .
The debate over the positioning of the cell division plane during cytokinesis has long been dominated by two models, astral stimulation and astral relaxation, that aimed to explain how the mitotic spindle specifies the position of the furrow. Recent findings have provided experimental support for variations of both models. The central spindle pathway for furrow formation can explain many of the observations that lent support to the astral stimulation model. These mechanisms are, in fact, related. In the original astral stimulation model, a stimulatory signal concentrates at the equator as a consequence of high microtubule density on the cortex due to the combined influence of the two asters, whereas in the central spindle model the stimulatory signal concentrates due to a high density of antiparallel microtubules in the central spindle. These mechanisms do not explain all the available data and there is direct evidence for a second pathway in which microtubules have an inhibitory effect on cortical myosin accumulation and cortical contractility. It must be emphasized that this second pathway contributes to furrowing and not cytokinesis, since components of the central spindle are critical to allow for cell separation (abscission). A major challenge for the future will be to establish the molecular mechanism by which astral microtubules modulate cortical contractility. Furthermore, it remains to be determined if most animal cells utilize redundant mechanisms for furrow formation.
Mechanics and Control of Cytokinesis: Biochemical Society Focused Meeting held at Royal College of Surgeons, Edinburgh, U.K., 9–12 January 2008. Organized and Edited by Gwyn Gould (Glasgow, U.K.) and Iain Hagan (Manchester, U.K.).