Cytokinesis must be temporally and spatially coordinated with chromosome segregation in order to avoid genome polyploidization, centrosome amplification, and chromosome breakage 
. In stem cell niches, the plane of division must also be correctly oriented relative to other asymmetrically distributed determinants of cell fate 
. Conversely, defects in cytokinesis increase the likelihood of carcinogenesis 
, vividly highlighting the need to develop a detailed understanding of this process. While individual components of the cytokinetic apparatus are increasingly well annotated, our knowledge of how these components interact with one another at the systems level remains fragmentary 
. In this study, we show that the self-regulated recruitment of Plk1 to the spindle midzone and its phosphorylation of the HsCYK-4 subunit of centralspindlin encode positive cues for cleavage furrow formation, thus providing new mechanistic insight into how mammalian cell division is regulated.
First, by manipulating Plk1 geometry in anaphase cells, we define a clear requirement for Plk1 activity at the midzone prior to recruitment of Ect2, cortical up-regulation of the RhoA GTPase, and contractile ring assembly. Although Plk1's PBD-dependent localization to the spindle midzone has long been appreciated 
, conducting a direct test of whether (and how) this spatial regulation contributes to the initial events in cell division has been virtually impossible, as cells expressing only PBD-deficient alleles of Plk1 are incapable of satisfying the spindle checkpoint and entering anaphase 
. As an indirect alternative, RNAi-mediated depletion of individual Plk1 docking partners such as PRC1 
was used to block Plk1 enrichment at the midzone. Because such cells continue to form furrows, it was concluded that the midzone-associated pool of Plk1 plays no significant role upstream of the Ect2-RhoA network and contractile ring assembly 
. Based on the results obtained here, we suggest instead that such depletions did not abolish the full spectrum of PBD-dependent positive-feedback loops at the spindle midzone. In support of this view, Plk1 eventually accumulates on subcortical interzonal MTs in PRC1-depleted cells 
, presumably via redundant interactions with other MT-binding proteins such as MKLP2 
and HsCYK-4 (). It is also possible that the lack of a bona fide spindle midzone in PRC1-depleted cells actually enhances furrowing by increasing the number of interzonal MTs that contact the equatorial cortex 
In contrast, by using chemical genetics to inactivate or bypass all PBD-dependent positive-feedback loops in anaphase cells, we are now able to show that Plk1 self-organization at the midzone is essential for cytokinesis onset in human cells. This regulatory mechanism is conceptually similar to other positive-feedback loops that govern Cdk1 activation at the G1/S and G2/M transitions 
and cyclin B and securin proteolysis at anaphase onset 
. However, it is unique in that it works by concentrating Plk1 near its midzone-specific substrates, rather than by up-regulating Plk1's activity throughout the cytoplasm. Nevertheless, like other positive-feedback loops, Plk1 self-organization makes cytokinesis more switchlike—that is, more rapid and synchronous relative to other late mitotic events—than would otherwise be the case. Indeed, in cells expressing the active but unanchored kinase fragment (Plk1cat
), furrows were not only rarer but also shallower and more temporally heterogeneous than those induced by midzone-localized versions of Plk1 ( and Figure S6
). We suggest that Plk1-mediated cytokinesis synchrony could play an important role in ensuring that cells initiate and complete division before their license to do so expires, typically around 30 to 60 min after anaphase onset 
Like Plk1, Aurora B also relocalizes onto the spindle midzone at anaphase onset, resulting in an axial phosphorylation gradient that can be detected using a fluorescence resonant energy transfer (FRET)-based biosensor 
. Although a Plk1-generated gradient has yet to be detected by this method, our results show that Plk1's association with the midzone predicts whether or not an equatorial cleavage furrow will form ( and ). If an anaphase Plk1 gradient does exist, its detection may require the development of newer biosensors that can engage in PBD-mediated positive feedback and thus mimic the behavior of Plk1's anaphase-specific substrates.
We have also demonstrated that HsCYK-4 is an in vivo substrate of Plk1 at the spindle midzone, and that this phosphorylation allows HsCYK-4 to be recognized by the tandem BRCT repeats at the N terminus of Ect2. These results explain why Ect2 becomes mislocalized in cells treated with Plk1 inhibitors 
, and together with the contractile ring assembly defect in cells expressing non-phosphorylatable HsCYK-4 (), they suggest that HsCYK-4 is Plk1's major rate-limiting substrate upstream of RhoA. Curiously, this pronounced phenotype contrasts with the behavior of cells expressing dominant-negative Ect2 fragments, as the latter activate RhoA and ingress their furrows normally 
. By their very nature, dominant-negative experiments are difficult to judge for penetrance and specificity, as it can always be argued that the endogenous protein's function has been inadequately antagonized, or that another factor (in this case, a hypothetical negative regulator of furrowing) was also titrated. However, a more interesting notion is that HsCYK-4 mutations prevent not only Ect2 recruitment to the midzone but also relief of its auto-inhibition () 
, whereas dominant-negative Ect2 fragments allow the RhoGEF to cycle on and off HsCYK-4 in an activated state. Further experiments will be required to test this idea and clarify whether additional modes of Ect2 regulation operate in parallel with the mechanisms described here.
A model for Plk1-mediated generation of the midzone-derived signal for cell division.
In summary, our findings indicate that Plk1 generates the midzone-encoded stimulus for cell division by priming two structurally distinct phosphopeptide-binding modules within Ect2 and Plk1 itself, thereby ensuring the robustness and fidelity of this process (). Curiously, although the proximal and distal elements of this cytokinesis network (Plk1 and RhoA) are universally conserved 
, the intermediate components that couple this network to MTs (centralspindlin and Ect2) are either missing or lack cognate phosphopeptide-binding domains in unicellular organisms such as budding and fission yeast, which establish their division sites without reference to spindle geometry 
. It is tempting to speculate that the MT dependence of cytokinesis in metazoans is specifically related to the evolutionary history of these regulators of cell division.
Finally, we note that although genetic and pharmacologic probes are increasingly available for many cell cycle regulators, including Plk1, these tools typically involve substantial tradeoffs between temporal and spatial resolution. For instance, RNAi-based knockdown/addback methods allow researchers to deplete or relocalize a given enzyme, but require days to take effect. Conversely, pharmacologic agents have much faster kinetics of onset, but their high diffusibility limits their usefulness in selectively activating or inhibiting an enzyme at a defined subcellular location. Furthermore, distinguishing between the on-target and off-target effects of such compounds can be quite difficult. As illustrated here, chemical genetics can overcome these limitations and provide novel insights that are inaccessible via other routes. Given existing methods for gene replacement, similar systems can now be envisioned for most or all of the 600 kinases in the mouse and human genomes, providing powerful tools for both fundamental physiologic studies and preclinical evaluation of these enzymes as therapeutic targets.