Cell growth and division are tightly coordinated to maintain cell size constant during successive cell cycles. In Schizosaccharomyces pombe, the SAD kinase Cdr2 regulates the cell size at division and the positioning of the division plane. Cdr2 forms nodes on the medial cortex containing factors that constitute an inhibitory pathway for Wee1. This pathway is regulated by polar gradients of the DYRK kinase Pom1, and involves a direct inhibitor of Wee1, the SAD kinase Cdr1. Cdr2 also interacts with the anillin Mid1, which defines the division plane, and with additional components of the medial cortical nodes, including Blt1, which participate in the mitotic-promoting and cytokinetic functions of nodes. Here, we show that the interaction of Cdr2 with Wee1 and Mid1 requires the UBA domain of Cdr2, which is necessary for its kinase activity. In contrast, Cdr1 associates with the C-terminus of Cdr2, which is composed of basic and KA-1 lipid-binding domains. Mid1 also interacts with the C-terminus of Cdr2 and might bridge the N- and C-terminal domains, whereas Blt1 associates with the central spacer region. We propose that the association of Cdr2 effectors with different domains might constrain Cdr1 and Wee1 spatially to promote Wee1 inhibition upon Cdr2 kinase activation.
The cell cycle consists of alternative phases of growth and division that need to be properly balanced to maintain a constant cell size over generations. Given that cell size influences key aspects of cellular function, such as exchanges with the external environment, this rule applies from single-cell organisms to cells within complex tissues in multicellular organisms (Conlon and Raff, 2003; Lloyd, 2013; Mitchison, 2003).
The existence of cell size control mechanisms was established years ago in yeasts, single-cell eukaryotes with stereotyped round (budding yeast) or rod-shape morphology (fission yeast) (Turner et al., 2012). In budding yeast, cells born at different sizes adjust the length of G1 phase to compensate for cell size difference prior to ‘start’ and commitment to a new round of cell division (Johnston et al., 1977). In contrast, fission yeast cells adjust the time spent in G2, and divide only when a constant cell size has been reached. A G1/S cell size control also exists in this organism but can only be seen when the G2/M cell size control is abolished (Fantes and Nurse, 1977,, 1978; Fantes, 1977; Mitchison and Creanor, 1971; Nurse and Thuriaux, 1977; Sveiczer et al., 1996).
The delay in mitotic onset observed when S. pombe cells are too short to divide is imposed by the Wee1 kinase (Fantes and Nurse, 1978). Wee1 acts as a direct inhibitor of the cyclin-dependent kinase Cdc2 (the homologue of CDK1), preventing Cdc2 activation by phosphorylation of Tyr15 (Gould and Nurse, 1989; Russell and Nurse, 1987b). This inhibition is reversed by the Cdc25 phosphatase (Russell and Nurse, 1986). Mitotic entry takes place upon inversion of the balance of Wee1 and Cdc25 activity in a switch-like mechanism that involves positive- and negative-feedback loops exerted on Wee1 and Cdc25 by the polo-like kinase Plo1 (Hagan and Grallert, 2013; Navarro et al., 2012).
An important pathway regulating Wee1 activity and influencing mitotic commitment is the so-called cell geometrical network (CGN) (Martin, 2009; Martin and Berthelot-Grosjean, 2009; Moseley et al., 2009). This pathway is controlled by the SAD kinase Cdr2 (Breeding et al., 1998; Kanoh and Russell, 1998; Young and Fantes, 1987) and the antagonistic DYRK kinase Pom1 (Bahler and Pringle, 1998). Cdr2 forms nodes at the medial cell cortex (Morrell et al., 2004) where it recruits Wee1 and Cdr1 (also known as Nim1), the second fission yeast SAD kinase that has been shown to inhibit Wee1 activity in vitro (Coleman et al., 1993; Parker et al., 1993; Russell and Nurse, 1987a; Wu and Russell, 1993; Young and Fantes, 1987). Cdr2-deficient cells exhibit a delay in mitotic commitment and a long cell size at division, indicating that Cdr2 nodes represent active sites for Wee1 inhibition.
Cdr2 node distribution and activity are regulated by the DYRK kinase Pom1 (Almonacid et al., 2009; Martin and Berthelot-Grosjean, 2009; Moseley et al., 2009). Pom1 forms inhibitory diffusion gradients that emanate from the cell tips (Hachet et al., 2011; Saunders et al., 2012). These gradients restrict Cdr2 nodes to the cell middle and also inhibit their function towards Wee1. The molecular mechanisms underlying these inhibitions have recently been deciphered and are largely independent from one another. Inhibition of Cdr2 activity depends on the phosphorylation, by Pom1, of an autoinhibitory tail which can block the phosphorylation of Cdr2 T-loop on T166 by the stress kinase Ssp1, which is required for the activation of the kinase domain. Pom1 independently restricts Cdr2 node assembly to the medial cortex by reducing the membrane-binding affinity and clustering properties of Cdr2 in the cell tips where Pom1 concentration is high (Bhatia et al., 2014; Deng et al., 2014; Rincon et al., 2014).
Importantly, it has been proposed that growth might alleviate Pom1 inhibition over Cdr2 by moving apart the cell tips from which Pom1 gradients emanate (Bhatia et al., 2014; Deng et al., 2014; Martin and Berthelot-Grosjean, 2009; Moseley et al., 2009). In this way, the CGN might couple mitotic entry to cell size. Several models have been proposed to account for such a property (Bhatia et al., 2014; Deng et al., 2014; Martin and Berthelot-Grosjean, 2009; Moseley et al., 2009; Pan et al., 2014; Vilela et al., 2010). Nevertheless, given that pom1-deficient cells still exhibit cell size homeostatic properties (Wood and Nurse, 2013), whether the CGN constitutes an active cell size sensor that modulates Wee1 activity depending on cell size remains to be confirmed.
A second function of Cdr2 nodes is to recruit the anillin Mid1 (Almonacid et al., 2009; Chang et al., 1996; Moseley et al., 2009; Sohrmann et al., 1996) as well as a series of non-essential contractile ring components, such as Blt1, Gef2, Klp8 and Nod1 (Jourdain et al., 2013; Moseley et al., 2009; Zhu et al., 2013), to form precursors for the contractile ring and pre-establish the division plane on the medial cortex in parallel to Mid1 export from the nucleus (Almonacid et al., 2009; Bahler et al., 1998a; Paoletti and Chang, 2000).
Medial nodes eventually mature at the onset of mitosis upon activation of Mid1 by Plo1 (Almonacid et al., 2011), which triggers the recruitment of essential ring components such as the IQGAP Rng2, myosin II, the formin Cdc12 and the F-Bar protein Cdc15 in an orderly manner (Almonacid et al., 2011; Laporte et al., 2011,, 2010; Padmanabhan et al., 2011; Pollard and Wu, 2010; Rincon and Paoletti, 2012; Wu et al., 2003) while Cdr2 gradually dissociates from the nodes (Akamatsu et al., 2014; Martin and Berthelot-Grosjean, 2009; Morrell et al., 2004). These events lead to contractile ring assembly by acto-myosin-dependent compaction of nodes (Laporte et al., 2012; Vavylonis et al., 2008).
Similar to Wee1 and Cdr1, the ring components that associate with Cdr2 during interphase depend on Cdr2 for their proper localization, whereas Cdr2 localization to the medial cortex is not compromised in their absence (Moseley et al., 2009). Cdr2 is therefore considered to be the key organizer of medial cortical nodes. Moreover, besides cytokinetic functions, these components also participate in the mitotic-promoting function of Cdr2 with Wee1 and Cdr1 because their deletion delays mitotic entry (Jourdain et al., 2013; Moseley et al., 2009; Zhu et al., 2013).
How Cdr2 scaffolds various node components and organizes them into functional pathways regulating mitotic entry and division plane position remains completely unknown. In this work, we have dissected the molecular mechanisms by which Cdr2 coordinates the recruitment of its major effectors – Wee1, Cdr1 and Mid1 – to regulate cell division spatially and temporally.
Cdr2 belongs to the family of AMP-activated protein kinase (AMPK)-like kinases, which is a subgroup of the calmodulin-dependent kinase superfamily (CaMK) (Marx et al., 2010). This family contains classical AMPKs (Ucp9, also known as Ssp2, in fission yeast) as well as a series of closely related kinases including the MARK kinases (Kin1 in fission yeast) and the SAD or BRSK kinases (Cdr2 and Cdr1 in fission yeast). AMPK-like kinases share a series of structural features including an N-terminal serine/threonine kinase domain associated with a ubiquitin-associated (UBA) domain or auto inhibitory domain (AID), a medial spacer and a C-terminal lipid-binding KA-1 domain (Marx et al., 2010; Moravcevic et al., 2010; Rincon et al., 2014).
The crystal structures of the kinase domain plus the UBA or AID domains of several MARK isoforms and of S. pombe AMPK Ucp9 have been solved (Chen et al., 2009; Marx et al., 2006; Panneerselvam et al., 2006). This revealed that UBA and AID domains consist of three α-helices, adopting a similar fold, that bind tightly to the N-lobe of the kinase domain for UBA domains of MARK kinases, or to the hinge between the N- and C-lobes opposite to the active site for the AID domain of Ucp9. In both cases, these interactions are largely mediated by the third helix of the UBA or AID domain. Interactions between the UBA or AID domain and helix C of the kinase domain were observed in both cases and might regulate the activity of the kinase domain in a positive or negative manner (Chen et al., 2009; Jaleel et al., 2006; Marx et al., 2010).
Sequence analysis identified a UBA- or AID-like domain in Cdr2 between amino acids 284 and 327, hereafter called the Cdr2 UBA domain (supplementary material Fig. S1A). This domain was modelled based on the crystal structure of the S. pombe Ucp9 kinase domain plus its AID (PDB 3H4J; Chen et al., 2009) (Fig. 1A; supplementary material Fig. S1B) or that of the Mark3 (the homologue of fission yeast Par1) kinase domain plus its UBA domain with a T211A, S215A double mutation (PDB 3FE3; Marx et al., 2010). Analysis of the models revealed that key hydrophobic residues linking the Ucp9 AID to the Ucp9 kinase domain were all conserved in Cdr2 (supplementary material Fig. S1B). In contrast, several residues linking Mark3 UBA to Mark3 kinase domain were not conserved in Cdr2 (unpublished results) suggesting that Cdr2 UBA might associate with Cdr2 kinase domain similarly to the Ucp9 AID.
The Cdr2 UBA domain is necessary for Cdr2 activity
Given that the UBA and AID domains can impact on kinase activity positively or negatively (Chen et al., 2009; Jaleel et al., 2006; Marx et al., 2010), we next determined the function of Cdr2 UBA. To do so, we created a Cdr2ΔUBA mutant lacking amino acids 281 to 330. This mutant was expressed from the cdr2 locus in replacement of wild-type Cdr2, in fusion or not with mEGFP at the C-terminus. The deletion of Cdr2 UBA did not affect Cdr2 localization (Fig. 1B) but increased its levels slightly as judged by fluorescence intensity measurements on the medial cortex (∼1.4 fold; supplementary material Fig. S1C,D). Importantly, it strongly increased cell length at division in both the presence (data not shown) and absence of the mEGFP tag (17.3 µm compared to 14.1 µm for wild-type cells; Fig. 1C). Cdr2ΔUBA cells were almost as long as cdr2Δ cells at division (17.3 µm as compared to 17.7 µm), indicating that the function of Cdr2 as a mitotic-promoting factor is severely compromised in the absence of the UBA domain.
Cdr2 also promotes medial division by recruiting Mid1 to the medial cortex during interphase (Almonacid et al., 2009; Moseley et al., 2009). To test whether this second function of Cdr2 was affected, a Cdr2ΔUBA Mid1helix* double mutant line was examined [Mid1helix* lacks a lipid-binding amphipathic helix and fully depends on Cdr2 to bind to the membrane and position the division plane (Celton-Morizur et al., 2004)]. Cdr2ΔUBA Mid1helix* double mutant cells displayed a very high percentage of misplaced septa, similar to what has been observed in the cdr2Δ Mid1helix* double mutant, indicating that Cdr2ΔUBA does not interact properly with Mid1 (Fig. 1D). We conclude that the Cdr2 UBA domain is necessary for the division-plane-positioning function of Cdr2.
With dual defects in cell size at division and division plane positioning, Cdr2ΔUBA mutant phenocopies the Cdr2E177A kinase dead mutant (Almonacid et al., 2009; Morrell et al., 2004; Moseley et al., 2009) (Fig. 1C,D), raising the possibility that Cdr2 UBA is necessary for Cdr2 kinase activity. The activation of kinases of the AMPK family requires the phosphorylation of conserved residues in the T-loop (Bayliss et al., 2012; Lizcano et al., 2004). Most kinases have a primary phosphorylation site on a conserved threonine, but additional phosphorylation on a conserved serine of the T-loop is often required for full activation. Both residues are conserved in Cdr2, T166, which has previously been shown to be a target of Ssp1 kinase and regulates Cdr2 activity (Deng et al., 2014), and S170 (supplementary material Fig. S1E). Mass spectrometry analysis of Cdr2 immunoprecipitated from cell extracts confirmed that both sites are phosphorylated in vivo (supplementary material Fig. S2). Single mutations of these residues to alanine to prevent phosphorylation gave rise to longer cell size at division compared to wild-type cells, with a stronger defect in the T166A mutant than in the S170A mutant (16.2 and 15.4 µm respectively as compared to 14.3 µm in wild-type cells, Fig. 1E). Cell size at division remained shorter than in the Cdr2E177A kinase dead mutant (16.8 µm) unless the T166A and S170A mutations were combined (17.3 µm, Fig. 1E). These results indicate that both sites contribute to Cdr2 activation in vivo.
To test whether Cdr2ΔUBA can be phosphorylated at the T-loop, we produced an antibody against Cdr2 phosphorylated at T166 (Cdr2-pT166), as reported previously (Deng et al., 2014). This antibody allowed us to recognize Cdr2-pT166 specifically in western blots after immunoprecipitation of Cdr2 with an anti-GFP monoclonal antibody (mAb). The Cdr2-pT166 signal was completely abolished upon mutation of T166 to alanine. Cdr2ΔUBA was not recognized at all by the anti-Cdr2-pT166 Ab, showing that the deletion of Cdr2 UBA blocks the phosphorylation at the T-loop, thereby preventing Cdr2 activation (Fig. 1F).
Finally, we mutagenized key hydrophobic amino acids involved in the interaction between the kinase domain and the UBA domain based on our model for the Cdr2 kinase and UBA domains, to simply disrupt their interaction (supplementary material Fig. S1B). We mutagenized either three hydrophobic amino acids in the Cdr2 kinase domain into polar serine residues (Cdr2KD3S mutant), or three hydrophobic amino acids in the Cdr2 UBA domain (Cdr2UBA3S mutant) or combined the six mutations (Cdr2KD3S-UBA3S mutant). All three mutants had a similar length at division than Cdr2ΔUBA (Fig. 2A) and showed synthetic division plane position defects with the Mid1helix* mutant although the defects in Cdr2KD3SUBA3S were less strong than in Cdr2ΔUBA (Fig. 2B). This suggests that mutations of hydrophobic residues on both interaction surfaces have compensatory effects. T166A phosphorylation was also strongly affected in the three mutants (Fig. 2C). We thus conclude that the Cdr2 UBA domain has a key role in promoting Cdr2 activation through phosphorylation at the Cdr2 T-loop and does not display auto-inhibitory properties like the Ucp9 AID (Chen et al., 2009).
Wee1 and Mid1 bind the Cdr2 kinase–UBA domain in a kinase-activity-dependent manner
The main effectors of Cdr2 for its mitotic-promoting and cytokinetic functions are Wee1 and Mid1, respectively. Both depend on Cdr2 for proper localization to the cortex during interphase. Their association with Cdr2 medial cortical nodes also depends on Cdr2 activity (Almonacid et al., 2009; Moseley et al., 2009). This suggests that the cell size and cytokinetic defects seen in Cdr2ΔUBA mutant might result from Wee1 and Mid1 delocalization from the cortex. Consistent with this hypothesis, GFP–Wee1 overexpressed from the nmt1 promoter (Moseley et al., 2009) and Mid1–mEGFP were delocalized from the cortex in the Cdr2ΔUBA mutant (Fig. 3A), which is similar to what happened in the Cdr2Δ591–640 mutant, which carries a truncation in the KA-1 domain and does not associate with the cortex (supplementary material Fig. S3A). In contrast, Cdr1–3GFP localization was not affected in the Cdr2ΔUBA mutant (Fig. 3A).
We then wondered how the activity of Cdr2 might control Wee1 and Mid1 association with Cdr2. We thus tested whether Mid1 and Wee1 could directly bind the Cdr2 kinase–UBA domain in two-hybrid assays. The Cdr2 kinase–UBA domain, used as a bait, interacted with Mid1 and Wee1, demonstrating that this region of Cdr2 contains direct binding sites for Wee1 and Mid1. Moreover, the interaction was disrupted by introduction of the E177A kinase dead mutation in Cdr2 kinase domain, showing that it requires Cdr2 activity (Fig. 3B).
To confirm this result in vivo in fission yeast, we produced a fusion between the Cdr2 kinase–UBA domain and the S. cerevisiae Kcc4 C-terminus, which contains the membrane-binding KA-1 domain, to target the Cdr2 kinase–UBA domain to the cortex. This construct was tagged with GFP in C-terminus and expressed from cdr2 locus in replacement of the wild-type Cdr2. This construct localized slightly to the nucleus and cytoplasm, and more strongly to the entire cell cortex (supplementary material Fig. S3B). The Cdr2 kinase–UBA–Kcc4C fusion protein was largely non-functional in terms of mitotic-promoting activity or cytokinesis regulation as judged by cell length at division (16.1 µm; supplementary material Fig. S3C) and the percentage of misplaced septa when combined with Mid1helix* (supplementary material Fig. S3D). Nevertheless, Wee1 and Mid1 co-immunoprecipitated with this construct as efficiently as with full-length Cdr2. Furthermore, the kinase dead mutation E177A completely abolished the interaction, like in the two hybrid assay (Fig. 3C,D).
Taken together, our results demonstrate that the main effectors of Cdr2, Wee1 and Mid1 bind to Cdr2 kinase–UBA domain in a kinase-activity-dependent manner. This could be because the Cdr2 kinase–UBA domain conformation is modified upon activation to create interaction sites for Wee1 and Mid1. Alternatively, Cdr2 might phosphorylate these proteins or their binding sites within Cdr2 kinase–UBA domain to promote their interaction.
Cdr2 activation by Ssp1 controls its association with Wee1 and Mid1
The above results suggest that, in wild-type cells, Cdr2 phosphorylation on T166 by Ssp1 kinase (Deng et al., 2014) might regulate Cdr2 function by controlling its interaction with Wee1 and Mid1. Accordingly, we found, using immunoprecipitation assays, that the interaction between Wee1 or Mid1 and Cdr2 was strongly reduced in presence of the T166A mutation that prevents Cdr2 activation by Ssp1 (Fig. 3E).
We also reciprocally checked whether Mid1 and Wee1 binding to the Cdr2 kinase–UBA domain could modulate the phosphorylation of T166 by Ssp1. To do so, we determined the level of Cdr2-pT166 in Mid1Δ400-450 and wee1Δ cells. We did not detect any significant alteration of the level of T166 phosphorylation in these conditions (supplementary material Fig. S3E–G). Similar results were obtained in cdr1Δ or blt1Δ mutants. This indicates that Cdr2 activation by Ssp1 is not modulated upon binding of other node components.
Cdr1 associates with the Cdr2 C-terminus
The role of Cdr2 as a mitotic-promoting factor depends largely on Cdr1, which belongs to the AMPK family of kinases, like Cdr2, and is recruited to medial cortical nodes by Cdr2 (Moseley et al., 2009). Cdr1 can directly phosphorylate and modulate Wee1 activity negatively in vitro (Coleman et al., 1993; Parker et al., 1993; Russell and Nurse, 1987a; Wu and Russell, 1993). The association of Cdr1 with medial nodes does not depend on Cdr2 kinase activity (Moseley et al., 2009) and, likewise, was not impaired in a Cdr2ΔUBA mutant (Fig. 3A). This suggests that Cdr1 association with Cdr2 might depend on a different domain to that used in its interaction with Mid1 and Wee1.
To gain insight into how Cdr1 interacts with Cdr2, we produced a deletion of the spacer region, including the Cdr2 UBA domain (Cdr2Δ281-590; Fig. 4A). This construct, expressed from the Cdr2 locus in replacement of wild-type Cdr2 was sufficient to promote Cdr1–3GFP association with the cortex (Fig. 4B). Next, a Cdr2 C-terminus construct (Cdr2-Cter; amino acids 591–747), which contained the Cdr2 basic domain and KA-1 domain involved in membrane binding and clustering (Rincon et al., 2014), was expressed under Cdr2 promoter from the leu1 locus in a strain deleted for cdr2. This construct was also sufficient to recruit Cdr1–3GFP to the cortex (Fig. 4B). The weak signals observed likely stem from weak expression levels and reduced clustering of this construct compared to full-length Cdr2, as judged in a strain expressing Cdr2-Cter tagged at the N-terminus with EGFP from the same locus (Fig. 4B). Co-immunoprecipitation assays between Cdr1 and Cdr2-Cter confirmed that Cdr1 can interact with Cdr2-Cter (Fig. 4C).
Mid1 has an additional C-terminal interaction site and Blt1 associates with Cdr2 spacer region
Given that Wee1 and Mid1 bind the Cdr2 kinase–UBA domain efficiently, and that Cdr1 associates with Cdr2-Cter, we reasoned that a minimal construct containing these domains might be sufficient for the Cdr2 mitotic-promoting and division-plane-positioning functions. The localization of a construct that contains these domains, Cdr2Δ330-590, expressed from the cdr2 locus resembled that of Cdr2Δ281-590 lacking the UBA domain (supplementary material Fig. S4B; Fig. 4B). Both constructs localized at the medial cortex, although with less efficiency than wild-type Cdr2. Nevertheless, Cdr2Δ330-590 had a long cell size at division (18.0 µm; supplementary material Fig. S4A) indicating that this construct is largely non-functional. This suggests to us that the spacer region between Cdr2 kinase–UBA domain and the Cdr2 C-terminus is functionally important.
To understand the role of Cdr2 spacer, we first performed a series of co-immunoprecipitation assays with constructs including the spacer in full or in part and Cdr2-Cter (Cdr2331-775–mEGFP, Cdr2481-775–mEGFP and Cdr2531-775–mEGFP constructs; Fig. 5A). Surprisingly, these constructs were all able to interact with Mid1, indicating that Mid1 has the ability to bind the Cdr2 C-terminal region in addition to Cdr2 UBA and therefore might bridge them together (Fig. 5B).
Second, we produced serial deletions within the spacer of Cdr2. Two of these internal deletion mutants Cdr2Δ381-430 and Cdr2Δ481-530 showed a long cell length at division (15.7 and 17.2 µm respectively, Fig. 5C), whereas the rest of deletion mutants divided at a normal length, except Cdr2Δ531-590 whose long cell size at division (16.1 µm) might result from non-specific unfolding of the C-terminal anchoring domain directly adjacent to it (supplementary material Fig. S4D). The Cdr2Δ381-430 and Cdr2Δ481-530 mutants also had a high percentage of misplaced septa when combined with Mid1helix* (Fig. 5D) unlike other deletion mutants (supplementary material Fig. S4E). Interestingly, both Cdr2Δ381-430 and Cdr2Δ481-530 localized normally at the medial cortex, indicating that the observed defects did not arise from localization problems (supplementary material Fig. S4C).
We also found that the Cdr2Δ381-430 and Cdr2Δ481-530 mutants were properly phosphorylated on T166 (Fig. 5E) in contrast to the Cdr2ΔUBA mutant. This result suggested to us that the defects in these two mutants are not a consequence of Cdr2 kinase inactivation and might rather result from deficient interaction with another component of medial cortical nodes involved in both the mitotic-promoting and cytokinetic-related functions of Cdr2.
Blt1 was a good candidate because its absence perturbs cell size at division (Moseley et al., 2009) and leads to strong division plane defects in combination with Mid1helix* (Guzman-Vendrell et al., 2013).
We tested this hypothesis by checking Blt1–mEGFP localization in Cdr2Δ381-430 and Cdr2Δ481-530 mutants and observed a partial dissociation of Blt1 from the cortex in Cdr2Δ481-530 (Fig. 5F). Given that Blt1–mEGFP has a C-terminal lipid-binding domain that can mediate its retention to the cortex independently of Cdr2 (Guzman-Vendrell et al., 2013), we next made a Cdr2Δ481-530 Blt1Δ5 double mutant (Blt1Δ5 is deficient for lipid binding; Guzman-Vendrell et al., 2013) and observed a very strong detachment from the cortex (Fig. 5F). We conclude that a region that spans amino acids 481–530 of the Cdr2 spacer is necessary for the association of Blt1 with Cdr2. Given that Blt1 also interacts with Mid1 through Gef2 (Guzman-Vendrell et al., 2013), this interaction might provide a means to strengthen the interaction of Mid1 with Cdr2 further in medial cortical nodes, in an indirect manner.
Fission yeast medial cortical nodes regulate two important parameters of cell division, its timing and its position. Because both parameters influence cell size, medial cortical nodes are important players in cell size homeostasis.
Medial cortical nodes are hetero-oligomeric structures. They are primarily organized by the SAD kinase Cdr2, which possesses the necessary membrane-binding and clustering properties for self-assembly into nodes (Rincon et al., 2014). Membrane binding relies on Cdr2 C-terminus that contains basic and KA-1 domains, which can establish electrostatic interactions with acidic phospho-lipids of the plasma membrane. We have shown that the Cdr2 KA-1 domain is also involved in Cdr2 clustering. Cdr2 clustering is independently mediated by Mid1, which is itself endowed with oligomerization properties (Celton-Morizur et al., 2004; Saha and Pollard, 2012).
Medial cortical nodes have been shown to contain six additional components so far, the two effectors that regulate mitotic entry, Cdr1 and Wee1, as well as a series of four non-essential components of the contractile ring, Blt1, Gef2, Klp8 and Nod1 (Jourdain et al., 2013; Moseley et al., 2009; Zhu et al., 2013). These four components remain associated with the contractile ring throughout its assembly and constriction phases and have been shown to regulate its function (Goss et al., 2014; Guzman-Vendrell et al., 2013; Jourdain et al., 2013; Ye et al., 2012; Zhu et al., 2013).
Blt1 contains a membrane-binding motif in its C-terminus and can associate with the cell cortex independently of Cdr2 (Guzman-Vendrell et al., 2013; Moseley et al., 2009). It has been recently proposed that Blt1 nodes containing Gef2 and Klp8 might be released from contractile ring remnants at the end of cytokinesis, diffuse from the division site, towards the medial cortex, and fuse with Cdr2 nodes containing Mid1 and Cdr2 (Akamatsu et al., 2014).
Nevertheless, there is no evidence so far for long range movement of Blt1 nodes (Akamatsu et al., 2014) and Blt1 nodes remain concentrated at the cell tips in the absence of Cdr2 (Moseley et al., 2009). This suggests a possible alternative model where Blt1 recruitment to Cdr2 nodes might drive the disassembly of old Blt1 nodes inherited from the previous division cycle, by competition.
Given that Cdr2 is necessary for the recruitment of all node components identified, whereas none of these components are necessary for Cdr2 node assembly, it has been postulated that the primary function of Cdr2 is to ensure the scaffolding of these components and organization into a functional structure able to inhibit Wee1 and pre-establish the division plane. However, how Cdr2 coordinates the binding of these components remained completely unknown. Our work provides the first insights into the molecular organization of interphase node components by Cdr2 and the functional relevance of such an organization.
The Cdr2 UBA domain is required for Ssp1-dependent activatory phosphorylation at the T-loop
We first found that the Cdr2 UBA domain plays a crucial role in controlling Cdr2 kinase activation and binding of Wee1 and Mid1. These two processes are intimately linked to one another.
Cdr2 UBA modelling based on the crystal structure of Ucp9 kinase plus the AID domain revealed that the Cdr2 UBA domain might be directly attached to Cdr2 kinase domain through a hydrophobic interface and, like other UBA and AID domains, might contact the helix C of the kinase domain (Marx et al., 2010). This contact might provide a way for Cdr2 UBA to influence the Cdr2 kinase domain conformation and promote Cdr2 activation by phosphorylation at the T-loop on T166 by Ssp1 (Deng et al., 2014). We have found that S170 is also phosphorylated in vivo and controls Cdr2 activity too. Nevertheless the identity of the kinase performing the phosphorylation of S170 and tools to test the relationship between Cdr2 UBA and S170 are not yet available.
Cdr2 activity is also modulated by Pom1 in a negative manner. This regulation involves the phosphorylation of Cdr2 C-terminal tail which inhibits Ssp1-dependent phosphorylation at the T-loop (Bhatia et al., 2014; Deng et al., 2014). It will now be interesting to test whether this regulation involves the Cdr2 UBA domain or not.
The Cdr2 UBA domain is also necessary for Wee1 and Mid1 binding to Cdr2. Given that Cdr2 activation is necessary for their binding (Almonacid et al., 2009; Moseley et al., 2009), deciphering whether Cdr2 UBA represents the actual binding site for these two components or if the binding site is located on Cdr2 kinase domain or spans on both kinase domain and UBA is difficult to test. Two non-exclusive models could explain how Mid1 and Wee1 binding are linked to Cdr2 kinase activation: activation could modify the conformation of Cdr2 kinase domain or UBA to create interactions sites that are not available when Cdr2 is inactive. Alternatively, Cdr2 might phosphorylate these two proteins to promote their interaction.
Interestingly, Cdr1 also contains a UBA domain that serves as a binding site for Nif1 and Skb1. However, these interactions do not require Cdr1 kinase activity or the presence of the kinase domain (Deng and Moseley, 2013; Wu and Russell, 1997). Another sharp difference is that Nif1 and Skb1 binding leads to kinase inhibition, indicating that the Cdr1 UBA domain can be classified as an AID. In contrast, Mid1 and Wee1 binding to Cdr2 does not influence Cdr2 kinase activity. Thus, our data confirms that UBA and AID domains of AMPKs can serve as modulators of kinase activity and potential binding sites for partner proteins, albeit with large functional differences between family members. The Cdr2 UBA function might be most similar to the reported functions of the UBA domains of MARK, SIK and BRSK or SAD kinases in mammals (Jaleel et al., 2006).
Cdr2 C-terminus contains Cdr1 interaction site and a secondary Mid1-binding site whereas Blt1 associates with Cdr2 spacer
To understand the relationship between Cdr2 partners within nodes, we next defined how Cdr1 interacts with Cdr2 and found that Cdr1 interacts with the Cdr2 C-terminus, which is composed of the lipid-binding basic and KA-1 domain, and is separated by a long spacer from the kinase–UBA domain where Wee1 binds. This was surprising because Cdr1 acts as a direct inhibitor of Wee1 (Coleman et al., 1993; Parker et al., 1993; Wu and Russell, 1993), and raised the question of how Cdr1 might phosphorylate Wee1 efficiently with two distant sites of interaction. Further analysis revealed two mechanisms which could potentially bring Cdr1 and Wee1 in close contact.
First, we identified a second interaction site for Mid1 in C-terminus of Cdr2. Unlike the N-terminal binding site, this site is constitutive as it does not require the presence or activity of the Cdr2 kinase domain. This interaction seems to be independent of Blt1 (M.G.-V. and A.P., unpublished results) which can bridge Cdr2 with Mid1 through Gef2 (Guzman-Vendrell et al., 2013). The existence of this secondary interaction site might explain why Mid1 association with Cdr2 is not fully abolished upon deletion of the Mid1 domain necessary for Mid1 interaction with Cdr2 (amino acids 400–450) or inactivation of Cdr2 (Almonacid et al., 2009; Moseley et al., 2009). With these two interaction sites, Mid1 might bridge the Cdr2 N-terminus and C-terminus. Nevertheless, a Mid1 mutant deficient for Cdr2 binding (Mid1Δ400-450) divides at the same cell length as the wild-type strain (our unpublished results), indicating that Cdr2 N-terminus and C-terminus bridging by Mid1 is either not essential for proper inhibition of Wee1, or is redundant with a parallel molecular mechanism.
Second, we found an interaction site for Blt1 in the middle of the Cdr2 spacer. As mentioned above, Blt1 interacts with Mid1 indirectly through Gef2 (Guzman-Vendrell et al., 2013). Its association with Cdr2 nodes is strongly affected in the absence of Mid1 or upon inactivation of Cdr2 (Moseley et al., 2009), whereas Blt1 or Gef2 reciprocally stabilize the association of Mid1 with Cdr2 nodes (Guzman-Vendrell et al., 2013; Ye et al., 2012).
We finally propose that whereas Cdr1 associates with the Cdr2 C-terminus regardless of Cdr2 status, Cdr2 activation might trigger a series of events that stabilize node components in a conformation favouring Cdr1-dependent inhibition of Wee1, that is, binding of Wee1 and Mid1 to the Cdr2 UBA domain first, which might in turn favour Blt1 association with the Cdr2 spacer and stabilize Mid1 further through Gef2 and through binding to Cdr2 C-terminus (see Fig. 6).
It is worth noting that the interactions between Blt1 and the Cdr2 linker, Mid1 and the Cdr2 C-terminus, or Cdr1 and the Cdr2 C-terminus were shown by immunoprecipitation and localization studies and might therefore either be direct, or involve known or unknown node components. We speculate that these multiple interactions might ‘freeze’ Cdr2 in a conformation where Wee1 and Cdr1 are in close vicinity, allowing efficient Wee1 inhibition. Such a mechanism could account for the long cell size at division reported for Blt1- or Gef2-deficient cells (Moseley et al., 2009; Ye et al., 2012).
Given that there is a large pool of Wee1 at spindle pole bodies compared to the small pool that can be detected at medial cortical nodes (Moseley et al., 2009), an interesting question that remains to be solved is how Wee1 inhibition at medial nodes can promote mitotic entry. One can imagine that there is a fast exchange between these two pools so that local inhibition of Wee1 at medial cortical nodes can influence Wee1 activity globally and alleviate Cdk1 inhibition.
In summary, our work provides new molecular understanding of Cdr2 function as a major scaffolding factor for medial cortical node components and sheds light on the organization of these components in functional pathways. Given that the structure of AMPKs is largely conserved, this work might also help us to understand regulatory mechanisms affecting other AMPKs, in particular on the closest homologue of Cdr2 in mammals, the Brsk2 kinase (also known as SadA), which plays important functions in neuronal differentiation in the brain (Kishi et al., 2005; Lilley et al., 2014,, 2013) and insulin secretion in the pancreas (Nie et al., 2013a,,b,, 2012).
MATERIALS AND METHODS
Strains and plasmids
All S. pombe strains used were isogenic to 972 and are listed in supplementary material Table S1. Standard S. pombe molecular genetics techniques and media were used (Moreno et al., 1991). Strains were selected from genetic crosses by random spore analysis. All Cdr2 mutant alleles were integrated at their respective endogenous locus, except those cloned in pJK148-derived plasmids described previously (Keeney and Boeke, 1994) that were integrated at the leu1 locus. Transformations were performed using the lithium-acetate–DMSO method as described previously (Bahler et al., 1998b). All plasmids used are listed in supplementary material Table S2.
All Mid1 constructs were based on pJK148 (Keeney and Boeke, 1994). pAP159, pAP221 and pMA32 have been described previously (Almonacid et al., 2009; Celton-Morizur et al., 2004; Paoletti and Chang, 2000) respectively. To produce pMG50, the NotI-SacI fragment of pAP221 containing 4GFP–tnmt1, was subcloned into pAP159.
Yeast two-hybrid constructs
The yeast two-hybrid constructs are based on the bait plasmid pGBT9 and the prey plasmid pGAD424. Inserts were amplified by PCR from genomic DNA from a wild-type strain for pMG67, pMG75 and pMG77, and from genomic DNA of strain AP3346 for pMG98. For pMG67 (Cdr21-330) and pMG98 (Cdr2 1–330E177A) the inserts were cloned in XmaI–SalI-digested pGBT9. For pMG75, the Mid1 PCR insert was cloned in XmaI–SalI-digested pGAD424.
Cdr2 integration plasmids
Most mutants were cloned in pSR20, a pFA6a-mEGFP-KanMX6-derived plasmid carrying a cdr2 promoter between the SalI and BamHI sites, EGFP for N-terminal tagging between BamHI and PacI sites, followed by the original pFA6a mEGFP for C-terminal tagging and keeping the structure of the original plasmid until the end of KanMX6 where the cdr2 terminator is inserted between the SacI and SpeI sites. Most of Cdr2 constructs C-terminally tagged with mEGFP were cloned between BamHI and PacI in replacement of GFP, or between BamHI and AscI for the untagged mutants, before integration at the cdr2 locus in a cdr2Δ::NatMX6 strain AP2804, in replacement of the NatMX6 cassette by homologous recombination (Guzman-Vendrell et al., 2013). Plasmids were digested with NotI before transformation.
All the inserts of the Cdr2 mutants carrying internal deletions (pMG21, pMG23, pMG25, pMG27, pMG31 to pMG37, pMG41, pMG64, pMG65 and pSR55) were made by double PCR as described previously (Almonacid et al., 2009). The full-length inserts (pMG30, pMG40 and pSR34) were produced by simple PCR [pSR34 has been described previously (Rincon et al., 2014)]. The PCR products where then cloned into pSR20 as explained above. In the same way, the plasmids containing C-terminal fragments (pSR45, pSR46 and pSR156) were obtained by amplification of cdr2 from the codon 531, 591 or 331, respectively, to the last codon with oligonucleotides containing a BamHI site plus an ATG codon and a PacI site, and cloning into pSR20. pSR25 (described in Rincon et al., 2014) was produced by amplification of cdr2 from codon 591 to 747 with oligonucleotides containing the PacI and AscI sites plus the stop codon TAA, and cloning in frame after an EGFP tag in the pSR20 plasmid.
Plasmids carrying point mutations (pMG54, pMG108, pMG110 to pMG112, pMG141 to pMG146) were produced by site-directed mutagenesis of pSR3, a pBluescript plasmid carrying the Cdr2 open reading frame, or by double PCR, and were cloned into pSR20.
The inserts corresponding to Cdr21-330–kcc4C plasmids and its tagged and mutated versions (pMG92, pMG93, pMG131 and pMG133) were obtained by double PCR on pSR135, pMG54 or pMG123 in order to link the Cdr2 kinase–UBA domain (amino acids 1–330) mutated or not to Kcc4C (amino acids 917–1037) through a (GS)×3 linker, followed by GFP. As opposed to the rest of pSR20-based plasmids, the chimeric plasmids containing the KA-1 of Kcc4 were digested with SpeI instead of NotI to linearize before transformation.
pJK148 Cdr2591-747–12myc plasmid
pSR160 (described in Rincon et al., 2014) was produced by amplification of cdr2 from codon 591 to 747 with oligonucleotides containing XhoI (plus an ATG codon) and NotI sites, and cloning into a pJK148 plasmid containing the cdr2 promoter, a C-terminal Myc12 tag derived from pINV-Myc (Iacovoni et al., 1999) and the nmt1 terminator. The plasmid was digested with NruI before transformation.
3HA–Wee1 tagging plasmid
To produce the modified pFA6a plasmid to N-terminally tag Wee1 without overexpression (pMG95) we exchanged the nmt promoter of a pFA6a-KanMX6-P3nmt1-3HA plasmid for the wee1 promoter. The insert was created by PCR on genomic DNA, digested with BglII-PvuI, and cloned into the BglII-PacI-digested backbone. The plasmid was then used as an ordinary pFA6a plasmid and the PCR product was transformed into AP3200 (NatMX6-P81nmt1-GFP-wee1 h+), to prevent homologous recombination in the cloned pwee1 promoter.
Cells were grown exponentially at 25°C in YE5S medium until the exponential phase. For the epi-fluorescence images shown in Fig. 1B, Fig. 3A, Fig. 5F, supplementary material Fig. S4A–C, microscopy was performed on a DMRXA2 upright Microscope (Leica Microsystems) equipped with a 100×1.4 N.A. Plan Apochromat objective and a Coolsnap HQ CCD camera (Roper) (2 s exposure, binning 1, gain 1). Images were analysed with Metamorph (Molecular Devices). In Fig. 4B we used a Nikon Eclipse TE2000-U microscope equipped with a 100×1.45 N.A., oil immersion objective, a PIFOC Objective stepper, a Yokogawa CSU22 confocal unit and a Roper HQ2 CCD camera (300 ms exposure; 80% of GFP laser power; binning 1, gain 3).
To determine the percentage of displaced septa, cells were stained for septa with Fluorescent Brightener 28 (Sigma). The percentage of displaced septa was determined by visual inspection. Displaced septa include septa that are not perpendicular to the cell long axis, not dividing the cell equally and misshaped septa. Cell length measurements were made with Metamorph software on DIC images of septating cells taken on the DMRXA2 microscope described above for strains with identical auxotrophies.
Generation of an anti-Cdr2-pT166 antibody
The anti-Cdr2-pT166 antibody was generated in rabbit against the phosphopeptide IQQPGKLLQ(pT)SC (ProteoGenix SAS). Specific immunoglobulins (Igs) were purified in two steps: the serum was first submitted to a column of non-phosphorylated IQQPGKLLQTSC peptide covalently linked to agarose beads using the sulfolink immobilization kit for peptides (Thermoscientific); second the flow through of the column was submitted to a column of phosphorylated peptide IQQPGKLLQ(pT)SC prepared in a similar manner. Specific Igs bound to the second column were eluted with 200 mM glycine pH 2.2, neutralized by addition of 1 M Tris, dialyzed against PBS and concentrated fivefold in centricon devices. Purified Igs were supplemented with 3% bovine serum albumin (BSA), 0.1% sodium azide and 50% glycerol and kept at −20°C.
Total protein extracts, co-immunoprecipitation assays and western blotting
For co-immunoprecipitations, 200 ml of cells were grown to an optical density of 1 at 595 nm at 30°C in YE5S medium concentrated two times compared to regular YE5S medium (YE5S2×). The cells were first washed with 1 ml of Stop buffer (150 mM NaCl, 50 mM NaF, 10 mM NaEDTA and 1 mM NaN3), then resuspended in 600 µl 1D buffer [50 mM HEPES pH 7.5, 100 mM NaCl, 1 mM EDTA, 1% NP-40, 20 mM β-glycerophosphate, 50 mM NaF, 0.1 mM Na3VO4, 1 mM phenylmethylsulfonyl fluoride complemented with complete EDTA-free antiprotease tablets (Roche)] together with 600 µl of glass beads and broken using a FastPrep FP120A instrument (Qbiogene; two cycles of 40 s at maximum speed). Lysates were then spun at 10,000 g for 10 min at 4°C, and supernatants were recovered. Soluble extracts were incubated with anti-mouse-IgG magnetic beads (M-280 Dynal; Invitrogen) coupled to 6 µg of anti-GFP monoclonal antibody (mAb; Roche) or anti-Myc mAb 9E10 (Roche) for 2 h at 4°C; then, the beads were washed five times with 1D buffer and the beads were resuspended in SDS-PAGE sample buffer. Immunoprecipitation samples and soluble extracts were submitted to SDS-PAGE and transferred onto nitrocellulose membranes.
Western blot assays were performed with anti-GFP mAb (1:500; Roche), and anti-Mid1 affinity-purified Ab (1:200) or affinity-purified anti-Cdr2-pT166 antibody (1:100). Secondary antibodies were coupled to peroxidase (Jackson ImmunoResearch) or to alkaline phosphatase (Promega). Signal quantification was performed in Metamorph. The signals of co-immunoprecipitated proteins were normalized relative to the protein concentration in the input and the amount of primary precipitated protein.
Analysis of phosphorylation sites on Cdr2 by mass spectrometry
For mass spectrometry analysis, Cdr2–mEGFP was purified from cells expressing Cdr2–mEGFP using anti-GFP mAb (Roche) as described previously (Almonacid et al., 2011). After immunoprecipitation, proteins were migrated on SDS-PAGE gels. Excised gel slices were processed as described previously (Rincon et al., 2014) for mass spectrometry analysis and identification of phospho-peptides.
Cdr2 kinase–UBA domain 3D homology modelling
3D homology modelling was performed as described previously (Sali and Blundell, 1993), using the Modeler 9.0 in the DS Modeling 1.7 software package (Accelrys, San Diego, CA). The model for the Cdr2 kinase–UBA domain was generated by using the coordinates of Ucp9 kinase plus its AID domain (PDB code 3H4J). The structural quality of the models was assessed according to the Modeler probability density functions as well as Profiles-3D analysis (DS Modeling 1.7). Out of the ten models generated, the one with the lowest energy was selected. 3D molecular representation was obtained by using PyMol™ (DeLano Scientific LLC).
We thank M. Riou and P. Paoletti for invaluable help in producing the Cdr2 kinase–UBA domain 3D model. We thank James Moseley for strains and plasmids. Microscopy was performed in the PICT-IBISA, Institut Curie, Paris, member of France-BioImaging national research infrastructure.
M.G.-V. and S.R. performed experiments, F.D. and D.L. performed the mass spectrometry analysis, M.G.-V. and A.P. conceived experiments and wrote the article.
This work was supported by Ligue contre le cancer; Comité de Paris; Fondation ARC pour la Recherche sur le Cancer; and Agence Nationale de la Recherche to A.P. A.P. is a member of LabEx CelTisPhyBio, part of IdEx PSL*. M.G.-V. received PhD fellowships from Université Paris-Sud and Fondation ARC pour la Recherche sur le Cancer. S.A.R. received post-doctoral fellowships from Fundacion Ramon Areces and Marie Curie FP7 program.
The authors declare no competing or financial interests.