Controlling Cytokinesis through Promiscuous Phosphorylation outside BARs

Abstract

In this issue of Molecular Cell, Roberts-Galbraith and colleagues report that a key cytokinetic regulator in fission yeast, Cdc15, is phosphorylated on numerous sites that collectively, but not individually, control its oligomerization state and its associations with the plasma membrane and interacting proteins.

Cytokinesis in most eukaryotes is mediated by an actomyosin-based contractile ring that must be positioned at the appropriate place and time to ensure that the two daughter cells equally share the genome and other cellular components (Glotzer, 2005). Many organisms ensure this equal partitioning by coupling cytokinesis with cell-cycle progression. As protein phosphorylation controls many cell-cycle transitions, it is nearly self-evident that protein phosphorylation would also regulate cytokinesis. Many biochemical reactions are controlled by phosphorylation of specific sites that result in, for example, changes in enzyme activity or protein-protein interactions. In this issue of Molecular Cell, Roberts-Galbraith and colleagues report that an important regulator of cytokinesis, Cdc15, is phosphorylated on more than four dozen sites by multiple kinases, and these phosphorylations introduce a multitude of negative charges that have little individual effect but collectively profoundly affect the ability of Cdc15 to oligomerize and to associate with interacting proteins (Roberts-Galbraith et al., 2010).

Cytokinesis is actively studied in the widely used model systems, including budding yeast, fission yeast, dictyostelium, flies, worms, and mammalian cells. They each have strengths, weaknesses, and idiosyncrasies. In all of these organisms, cytokinesis involves an actomyosin ring, but only in some of them is it essential (Balasubramanian et al., 2004). Similarly, each of these organisms has one or more mechanisms to coordinate the position of the division plane with the position of the mitotic spindle, though these mechanisms differ among the species. Cytokinesis in animal cells and fungi involves conserved proteins such as actin, myosin, and the actin nucleating formin, whereas other factors appear specifically required in certain taxa. An example of this latter class is the Cdc15 protein from S. pombe (Fankhauser et al., 1995). Cdc15 has attracted particular interest, as it contains a PCH (Pombe Cdc15 homology) domain implicated in membrane deformation. This domain is present in numerous proteins throughout eukaryotes, and they are often followed by a coiled-coil dimerization domain that constitutes a domain variously known as F-BAR or EFC (Figure 1) (Aspenström, 2009).

Figure 1. Phosphorylation of Cdc15 Dimers Inhibits Oligomerization and Ring Assembly.

(A) Overall organization of Cdc15p, indicating approximate positions of the F-BAR and SH3 domains and the widely distributed phosphorylation site.

(B) Depiction of the characteristic curvature of an F-BAR domain dimer.

(C) End-to-end oligomerization of F-BAR dimers can generate rings or spirals. In vivo, the concave side of the oligomers is thought to contact membrane invaginations.

(D) Distributed phosphorylation of Cdc15 impairs oligomerization of the F-BAR domain.

An individual F-BAR domain contains three extended α helices that assume a curved, elongated shape, likened to a banana. Two F-BAR domains dimerize in an antiparallel configuration. The dimers then associate at their ends to form large rings or spirals that associate with membranes via their positively charged, concave surface and deform them to form membrane buds or tubules (Shimada et al., 2007). These deformed membranes can then be pinched off by dynamin-like proteins during processes such as endocytosis. Indeed, in S. pombe, Cdc15 concentrates near sites of endocytosis.

Cdc15 is a relatively large protein, containing an ~250 residue N-terminal F-BAR domain and an ~50 residue C-terminal SH3 domain (Figure 1). The extended central region of Cdc15 lacks obvious domains. Cell-cycle-regulated changes in the phosphorylation state of Cdc15 correlate with its recruitment to the contractile ring. To examine whether the phosphorylation state of Cdc15 alters its biochemical properties, Roberts-Galbraith et al. purified the hyperphosphorylated protein from yeast cells and examined it by electron microscopy (EM). Whereas hyperphosphorylated Cdc15 appeared as globular particles, dephosphorylation of the same preparation yielded striking, undulating filaments. Dephosphorylation-dependent large-scale structural changes were confirmed by analytical ultracentrifugation. Of interest, native Cdc15 filaments resembled those observed with the recombinant, isolated F-BAR domain of Cdc15, strongly suggesting that filamentation of native Cdc15 is directed by its F-BAR domain. Not only does dephosphorylation of Cdc15 promote its oligomerization, it also enhances its ability to interact with certain contractile ring components, including the formin Cdc12 and the type I myosin Myo1. Unfortunately, these experiments did not provide insight into the configuration of the remaining two-thirds of the protein.

This information void is unfortunate, as mass spectrometry analyses of Cdc15 purified from yeast cells revealed up to 50(!) sites that were detectably phosphorylated, most of which lie within the central “unstructured” region (Figure 1). In order to get a handle on the significance of phosphorylation, the Gould lab generated a panel of Cdc15 variants. Due to the large number of sites and the immense number of possible combinations, they chose to group the phosphorylation sites into subgroups based on recognition sequence. One subgroup consisted of 13 RXXS sites. Phospho-RXXS sites can be bound by 14-3-3 proteins; indeed the 14-3-3 protein Rad24 is known to bind phospho-Cdc15. The other major subgroup contained 11 (S/T)P sites that are likely to be regulated by proline-directed kinases and phosphatases (e.g., cyclin-dependent kinases, MAP kinases, and the phosphatase Cdc14/Clp1). The third subgroup was filled by seven sites that didn’t match either consensus sequence. Cdc15 variants were constructed in which all sites in a subgroup were mutated to nonphos-phorylatable (Cdc15_SA) residues or phosphomimetic (Cdc15_SD) residues.

All of the mutants could rescue mutations in Cdc15 to viability, but morphological defects were readily apparent. Cytokinesis occurred with delayed kinetics with both types of mutants, though the basis for the delay may differ, as indicated by the localization of the mutant proteins. Wild-type Cdc15 localizes to small patches that likely reflect sites of endocytosis at cell tips during interphase and accumulates at the contractile ring during metaphase. Each of the Cdc15_SA variants exhibited increased localization near sites of endocytosis at cell tips during interphase and, of greater interest, precociously accumulated to the middle of the cell prior to mitotic entry. The extent of this phenotype positively correlated with the number of sites mutated. The Cdc15_SD mutations behaved similarly to wild-type, though a quantitative reduction in localization was observed, suggesting that other regulatory inputs affect Cdc15 localization.

These results are striking in that two nonoverlapping sets of phosphorylation sites have similar effects. In fact, the effects of these mutations appeared additive. The simplest explanation is that the behavior is dictated by the overall charge of the protein. This is not entirely without precedent, as one of the key interactions at the kinetochore is similarly regulated by its overall charge and also negatively regulated by phosphorylation (Guimaraes et al., 2008).

An important unresolved question is the nature of the interaction of the central unstructured region containing the majority of the phosphorylation sites and the filament forming F-BAR domain. As the phosphorylated protein behaves as a dimer, assembly of the filaments must be inhibited. In some F-BAR domains, filament assembly is mediated by charged residues at the interface between adjacent dimers (Shimada et al., 2007); perhaps these interact with the various phosphosites or with the 14-3-3 protein Rad24 that binds to some of them (Figure 1). Again, the recent literature provides a close analogy. A key regulator of cytokinesis in animal cells, the centralspindlin complex, also must oligomerize in order to function. In this case as well, 14-3-3 binding to phosphosites antagonizes oligomerization (Douglas et al., 2010).

Ectopic Cdc15 appears to be partially active. Cdc15_SA induced precocious recruitment of a subset of contractile ring components; however, functional contractile rings did not form. The Cdc15_SA foci resulted in ectopic endocytosis, perhaps indicating assembly of a hybrid site containing both endocytic site and contractile ring components.

The absence of intact contractile rings upon expression of Cdc15_SA is surprising because an activated version of the formin Cdc12 can induce fully functional rings during interphase (Yonetani and Chang, 2010). Although Cdc15_SA is able to recruit some Cdc12, it may not recruit Cdc12 to sufficient levels, or Cdc15_SA may be unable to activate it. Given that Cdc12 and Cdc15 are normally interdependent for their localization, it is likely, but not yet demonstrated, that overexpressed, activated Cdc12 recruits Cdc15, though this begs the question of whether this requires dephosphorylation of Cdc15.

Another important open question is whether an analogous regulatory mechanism pertains to a functional ortholog of Cdc15 in mammalian cells. An F-BAR domain containing protein, PSTIP, does concentrate at the cleavage furrow of dividing cells (Spencer et al., 1997), but there has been no loss-of-function studies demonstrating that it is required for cytokinesis. There are several structurally related proteins that may act redundantly. Nevertheless, given the strong evidence for membrane deformation and endocytosis at the cleavage furrow of dividing animal cells, this is probably fertile ground for future digging.