The F-BAR Cdc15 promotes contractile ring formation through the direct recruitment of the formin Cdc12

Cytokinesis is the terminal stage in cell division that results in the physical separation of two daughter cells. In many eukaryotic cells, an actomyosin-based contractile ring (CR) forms between the two segregated genomes and eventually constricts, dividing the mother cell into two daughter cells. In Schizosaccharomyces pombe, a powerful model organism for cytokinesis studies, assembly of the CR relies on a single formin, Cdc12, which is essential for nucleation and elongation of F-actin during CR formation (Nurse et al., 1976; Chang et al., 1997; Kovar et al., 2003; Kovar and Pollard, 2004) and also contributes to CR maintenance by bundling F-actin (Bohnert et al., 2013). Cdc12 recruitment to the medial cortex during CR formation relies on two redundant genetic modules (Wachtler et al., 2006; Laporte et al., 2011). Specifically, mutation of either IQGAP Rng2 or Myosin II in combination with the F-BAR (FER/Cip4 homology Bin–Amphiphysin–Rvs) scaffold Cdc15 eliminates Cdc12 recruitment to the division site and CR formation (Laporte et al., 2011). However, the mechanistic contributions of these genetic pathways are unknown.

Cdc15 is the founding member of the Pombe Cdc15 Homology family of proteins, which generally dimerize and bind membranes through their conserved N-terminal F-BAR domains (Tsujita et al., 2006) and interact with proteins though C-terminal protein-binding domains (Roberts-Galbraith and Gould, 2010). Although Cdc15 is essential for cytokinesis, it is unclear whether Cdc15’s role in Cdc12 recruitment is important for CR assembly. Some studies reported that cdc15 temperature-sensitive mutants lack CRs or produce them in only a fraction of cells (Fankhauser et al., 1995; Chang et al., 1996; Carnahan and Gould, 2003). Others found that cells lacking cdc15 formed rings of F-actin at the cell middle during early stages of mitosis, but these rings were not maintained during anaphase, leading to the conclusion that CR formation per se is independent of Cdc15 (Balasubramanian et al., 1998; Arai and Mabuchi, 2002; Wachtler et al., 2006; Hachet and Simanis, 2008; Laporte et al., 2011; Arasada and Pollard, 2014). Given that Cdc15 defines one of two genetic pathways of formin recruitment (Laporte et al., 2011), it is important to clarify its role in CR assembly. To do so, we defined the Cdc15-binding motif within the Cdc12 N terminus and constructed Cdc12 mutants that cannot interact with Cdc15. Cells lacking the Cdc12–Cdc15 interaction assembled CRs but had reduced Cdc12 in the CR, a delay in the medial accumulation of F-actin and actin-binding proteins, delayed CR formation, and were unable to survive other perturbations to CR assembly. Thus, the Cdc12–Cdc15 interaction is an important contributor to Cdc12 localization and CR formation.

We previously detected an interaction between Cdc15 and the Cdc12 N terminus that depended on the phosphorylation state of Cdc15 (Carnahan and Gould, 2003; Roberts-Galbraith et al., 2010). The Cdc12–Cdc15 interaction was unprecedented because it involved the F-BAR domain of Cdc15 rather than its SH3 domain (Carnahan and Gould, 2003). Because the first 151 residues of Cdc12 localized GFP to the division site (Yonetani et al., 2008), we examined these amino acids for a candidate Cdc15 interaction motif. Sequence comparison of Cdc12 aa 1–151 with its orthologues in other Schizosaccharomyces species revealed one conserved motif (aa 24–36; Fig. 1 A). Deletion of this motif (Δ24–36) or mutation of a conserved proline within it (P31A) resulted in loss of interaction with the Cdc15 F-BAR domain in vitro (Fig. 1 B). A synthetic peptide containing the motif (aa 20–40) bound the Cdc15 F-BAR, whereas mutation of P31 to alanine within the peptide abolished the interaction (Fig. 1 C). Titration binding assays between the Cdc12 peptide and the Cdc15 F-BAR revealed a dissociation constant of 1.1 nM, indicating a strong affinity (Fig. 1 D). Because Cdc12 is a low abundance protein (Wu and Pollard, 2005), a strong interaction may be necessary to recruit or maintain it at the cell middle. In fact, other protein–protein interactions that promote the localization of other formins are also in the nanomolar affinity range (Brandt et al., 2007; Watanabe et al., 2010). As might be expected from this tight association, Cdc12 (aa 24–36) fused to GFP localized to the division site (Fig. S1), supporting the possibility that this motif participates in directing Cdc12 to the cell middle.

To determine the functional consequence of disrupting the Cdc12–Cdc15 interaction, we constructed cdc12 alleles at the endogenous locus in which the binding motif was mutated or deleted. Although cdc12-P31A and cdc12-Δ24–36 cells are viable, they displayed synthetic lethal genetic interactions with rng2-D5, myo2-E1, and mid1Δ (Fig. 2, A–C; and Fig. S2, A–C). Myo2, Rng2, and Mid1 contribute to Cdc12 recruitment through a common genetic pathway distinct from Cdc15 (Laporte et al., 2011), and therefore, synthetic lethality likely results from the combined disruption of both Cdc12 recruitment pathways. cdc12-P31A and cdc12-Δ24–36 did not show a strong genetic interaction with cdc15-140 (Fig. S2 D), as would be expected if these alleles disrupt a major Cdc15 pathway during cytokinesis.

To test whether Cdc12 binding to Cdc15 influences Cdc12 localization to the CR, we tagged wild-type and mutant alleles with a single copy of mNeonGreen (mNG), a brighter and more photostable variant of GFP (Shaner et al., 2013). We compared mutant and wild-type cells in the same field of view using TRITC-conjugated lectin cell wall staining to differentiate between strains (Fig. 3 A). In addition, we used the spindle pole body (SPB) marker Sid4-GFP to define the stages of mitosis based on the distance between spindle poles. At mitotic onset, SPBs separate to opposite sides of the nucleus as the spindle forms, and they maintain a constant distance from one another throughout metaphase and anaphase A of ∼2.5 µm (Hagan, 1998; Nabeshima et al., 1998). We refer to these stages combined as “early mitotic,” as only ∼1/9 of cells with constant spindle length are in anaphase A (Nabeshima et al., 1998). SPBs move apart again at the onset of anaphase B. CRs in early mitotic cells had 35–36% less Cdc12-P31A-mNG or Cdc12-Δ24–36-mNG than wild-type Cdc12-mNG, and the amount of mutant Cdc12 did not increase during anaphase B like wild-type Cdc12 (Fig. 3, A and B). However, there was no significant difference in total Cdc12 protein levels in mitotic cells among strains (Fig. 3 C). Thus, Cdc15 helps recruit/maintain Cdc12 at the medial cortex during mitosis. These results also indicate that significantly less CR-associated Cdc12 is sufficient for CR formation, a finding that is in accord with the survival of diploids with a single functional copy of cdc12⁺ (Chang et al., 1997).

Cdc15 also influences Cdc12 localization in abnormal cell cycle situations (Carnahan and Gould, 2003; Roberts-Galbraith et al., 2010). For example, a cdc15 phosphomutant (cdc15-S11A) precociously localizes to the medial cortex when cells are arrested in G2; concomitantly, Cdc12 forms one or two medial cortical spots (Roberts-Galbraith et al., 2010). To test whether this precocious Cdc12 recruitment depends on Cdc15 binding, we visualized cdc25-22 cdc15-S11A cells with Cdc12-mNG or Cdc12-P31A-mNG. In contrast to Cdc12-mNG, which localized as bright central dots, Cdc12-P31A-mNG was diffuse in the cytoplasm (Fig. 3, D and E). A Cdc12 mutant lacking the C-terminal oligomerization domain (C′; Cdc12-ΔC′-mNG; Bohnert et al., 2013) formed a central streak rather than focused spots under these conditions, in accord with its lack of self-association (Fig. 3, D and E). However, the double mutant (Cdc12-P31A-ΔC′-mNG) showed diffuse cytoplasmic localization (Fig. 3, D and E), revealing the importance of the Cdc15 association for precocious Cdc12 medial recruitment. Overexpression of cdc15 also results in the formation of large puncta of Cdc12 (Carnahan and Gould, 2003); this pattern was abolished by the P31A mutation in cdc12 (Fig. 3 F). Thus, in both normal and abnormal conditions, Cdc15 directs Cdc12’s localization via residues 24–36.

Wild-type Cdc12 and Cdc15 localize to the medial cortex independently of F-actin (Wu et al., 2003, 2006). In contrast, N-terminal truncations of Cdc12 (Δ503-cdc12 and Δ841-cdc12) require F-actin for medial recruitment (Coffman et al., 2013). To determine whether abrogation of Cdc15 binding led to F-actin dependence, we treated wild-type and the cdc12 mutant cells with a high dose of Latrunculin-A (Lat-A) to deplete F-actin. Both Cdc12-mNG and Cdc12-Δ24–36-mNG localized in medial cortical dots in Lat-A–treated cells, but just as in CRs (Fig. 3, A and B), Cdc12-Δ24–36-mNG was less abundant (Fig. S3 A). Thus, lack of Cdc15 binding does not render Cdc12 localization dependent on F-actin.

Our results indicate that cells lacking the Cdc12–Cdc15 interaction have 33% less Cdc12 at the division site. Assuming that both wild-type and mutant Cdc12 nucleate and elongate F-actin at the same rate, cells lacking the Cdc12–Cdc15 interaction could then take 33% longer to form the F-actin in the CR. Thus, if Cdc12 activity is not otherwise regulated, mutant cells lacking the Cdc12–Cdc15 interaction may still have time to produce sufficient amounts of F-actin for the CR. On the other hand, because F-actin in the CR is rapidly and constantly turned over (Pelham and Chang, 2002), less CR-associated Cdc12 might lead to less F-actin. Therefore, we compared the amount of F-actin in the CR between early mitotic and anaphase B wild-type and mutant cells. There was ∼26% less F-actin, visualized with LifeAct-mCherry, in the CR of cdc12-P31A cells compared with wild-type cells during early mitosis (Fig. 4, A and B). However, there was no difference in the amount of F-actin in the CR between the two strains during anaphase B (Fig. 4, A and B). Thus, the mutant Cdc12 produces the same amount of F-actin in the CR as wild type given time. We did not observe a difference in longitudinal F-actin cable number (unpublished data), in accord with Cdc12 specifically producing F-actin for the CR (Chang et al., 1997). Formins compete with the Arp2/3 complex for G-actin, and mutations disrupting formin function have increased F-actin patch density (Burke et al., 2014). As expected, we found an ∼24% increase in fluorescence intensity of actin patches at cell tips in early mitosis but not anaphase B in the mutants compared with wild-type (Fig. 4, A and C) as expected given that less F-actin was incorporated into the CR in the mutant cells.

Although there was less F-actin in the CRs of cdc12 mutants lacking the Cdc12–Cdc15 interaction in early mitosis, we expected that proteins targeted there independently of F-actin, such as Cdc15 (Wu et al., 2006) and Myosin II (Naqvi et al., 1999), would not be affected in the mutant cells. Indeed, mCherry-Cdc15 and the myosin light chain Rlc1-GFP did not have altered protein levels in CRs of mutant cells compared with wild type in any mitotic stage (Fig. 4, D and E; and Fig. S3, B and C). In contrast, we anticipated that proteins localizing to the CR in an F-actin–dependent manner would exhibit decreased abundance in the CR of cdc12 mutants. Indeed, there was an ∼18–19% decrease in Ain1-GFP (an α-actinin protein; Wu et al., 2001) and ∼39% decrease in GFP-Adf1 (a cofilin protein; Nakano et al., 2001) abundance in the CR in the cdc12 mutants compared with wild type during early mitosis (Fig. 4, D and F; and Fig. S3, B–D). Like F-actin, Ain1-GFP recovered to wild-type levels during anaphase B, although GFP-Adf1 did not. We do not have an explanation for this difference.

Because these results indicated that loss of the Cdc15–Cdc12 interaction adversely affects CR assembly, we used time-lapse microscopy to examine cytokinetic progression. Using Rlc1-GFP and Sid4-GFP as markers for the CR and mitotic progression, respectively, we found that CR formation in cdc12-P31A and cdc12-Δ24–36 mutants was 26% and 23% slower, respectively, from the time of initial SPB separation compared with wild type (Fig. 5, A and B), in remarkable agreement with the reduction in F-actin. In contrast to some other mutations affecting CR formation (Coffman et al., 2009; Roberts-Galbraith et al., 2010; Tebbs and Pollard, 2013; Wang et al., 2014), we did not observe a compensatory decrease in the amount of time for CR maturation (the time between CR formation and the beginning of CR constriction).

We wondered whether binding to Cdc15 cooperated with previously described regulatory inputs into Cdc12 localization and function. We therefore carefully assessed CR dynamics in cdc12 mutant cells lacking its oligomerization domain (cdc12-ΔC′) and a phosphomutant whose multimerization cannot be regulated by the septation initiation network (SIN; cdc12-4A; Bohnert et al., 2013). The latter mutant renders CR assembly completely dependent on Mid1 (Bohnert et al., 2013). Although no changes in CR dynamics were observed in cdc12-ΔC′ cells compared with wild type, CR constriction took significantly longer in cdc12-4A cells (Fig. S3 E). Next, we assayed CR dynamics in double mutants with cdc12-P31A. The phenotypes reflected the defects of the individual single mutants, and therefore, the mutations are not additive (Fig. S3 E). Thus, Cdc15 and the SIN affect different aspects of Cdc12 function: Cdc15 binding to Cdc12 promotes its medial recruitment, whereas SIN-dependent phosphorylation of Cdc12 controls its oligomerization state that is important later in cytokinesis.

Although little Cdc12 is required for CR formation, its medial recruitment depends on multiple inputs. Here, we have established that Cdc15 is one of these inputs, contributing to Cdc12 recruitment during mitosis via direct binding. Our findings provide a rationale for why Cdc15 localizes early in the process of CR assembly (Fankhauser et al., 1995) despite its other described roles later in cell division (Wachtler et al., 2006; Vjestica et al., 2008; Roberts-Galbraith et al., 2009, 2010; Arasada and Pollard, 2014). The existence of multiple inputs for formin recruitment occurs in other biological contexts although the details differ (Chesarone et al., 2010). Of the formins with described recruitment pathways, many rely on a combination of intrinsic properties (membrane binding and/or dimerization/multimerization; Bohnert et al., 2013; Rousso et al., 2013) and extrinsic pathways (protein binding partners; Seth et al., 2006; Copeland et al., 2007; Chesarone et al., 2010; Liu et al., 2012). Homo sapiens mDia2, for example, requires interactions with both RhoA and Anillin to direct its localization and control its function during cytokinesis (Watanabe et al., 2010). Saccharomyces cerevisiae Bni1 localization to the bud neck requires dimerization of its N terminus in combination with two protein-binding domains, one for Rho1 and another for Spa2 (Fujiwara et al., 1998; Liu et al., 2012). Thus, it is likely that still more formins will be found to use a combination of sequence motifs to direct their intracellular localization, and some may be similar to the Cdc12 mechanism.

In the absence of Cdc15 binding, Cdc12 is still recruited for CR formation, likely through other N-terminal interactions with Rng2 and/or Myo2 based on genetic evidence (Laporte et al., 2011) and recent synthetic targeting experiments (Johnson et al., 2014; Tao et al., 2014). To elaborate, precocious targeting of Myo2 or Rng2 to the medial cortex is sufficient for Cdc12 recruitment and vice versa (Tao et al., 2014). Future studies will be aimed at determining the relative importance of various regulatory inputs into Cdc12 recruitment and function. We note that although Cdc15 can promote Cdc12 localization to medial cortical spots during interphase, this is insufficient for Cdc12 activation and the formation of CR F-actin (Fig. 3 D; Roberts-Galbraith et al., 2010). Thus, other interactions and/or protein modifications must promote Cdc12 F-actin nucleation and elongation functions.

Formin Cdc12 recruitment to the cell middle via direct binding to the Cdc15 F-BAR domain provides one of the first examples of a bifunctional F-BAR domain that interacts with a protein partner as well as the membrane. Since our original observation (Carnahan and Gould, 2003), only one other example has emerged (Kostan et al., 2014). Pacsin2, in addition to binding membranes, can interact with F-actin via its F-BAR domain (Qualmann et al., 1999; Kessels et al., 2006; Kostan et al., 2014). However, these two partners of the Pascin2 F-BAR compete to bind the same surface of the domain (Kostan et al., 2014). Although we do not yet know whether Cdc12 competes with membrane for binding to the Cdc15 F-BAR, it is likely that the F-BAR domain binds both simultaneously because both Cdc15–Cdc12 and Cdc15–membrane interactions are restricted to the same narrow time window when Cdc15 is dephosphorylated during mitosis (Wachtler et al., 2006; Roberts-Galbraith et al., 2010). Simultaneous binding would position the N terminus of the formin very close to the plasma membrane to establish a tight bridge between the membrane and CR during cell division. Further study of other F-BAR domains may reveal additional examples of bifunctional binding properties.

S. pombe strains (Table S1) were grown in yeast extract (YE). cdc12⁺, cdc12-P31A, and cdc12-Δ24–36 were tagged endogenously at the 3′ end with mNG:kan^R or mNG:hyg^R using pFA6 cassettes as previously described (Wach et al., 1994; Bähler et al., 1998). The pFA6 plasmid contains a replication origin and amp^R for selection in bacteria. mNG, a recently reported GFP derived from the lancelet Branchiostoma lanceolatum, was chosen for imaging experiments because of its superior brightness of ∼93 (product of extinction coefficient and quantum yield), compared with ∼34 for EGFP (Shaner et al., 2013; Allele Biotechnology). mNG has an excitation maximum of 506 nm and emission maximum of 517 nm and is currently the brightest known monomeric fluorescent protein in the green–yellow spectrum. A lithium acetate method (Keeney and Boeke, 1994) was used in S. pombe tagging transformations, and integration of tags was verified using whole-cell PCR and/or microscopy. Introduction of tagged loci into other genetic backgrounds was accomplished using standard S. pombe mating, sporulation, and tetrad dissection techniques. Fusion proteins were expressed from their native promoters at their normal chromosomal locus unless otherwise indicated. Lat-A was used at a final concentration of 100 µM from a 1-mM stock solution in DMSO.

The mNG:kan^R and mNG:hyg^R pFA6 cassettes were constructed by PCR amplifying the mNG sequence with the PacI restriction site in the forward primer and the Asc1 restriction site in the reverse primer. The sequence was subcloned into the pFA6 vector with the respective resistance cassettes.

To make the endogenous cdc12-P31A and cdc12-Δ24–36 alleles, a pSK vector (pBluescript backbone) was constructed that contained in the following order 5′ cdc12 flank including its promoter, full-length cdc12⁺, kan^R cassette, and 3′ cdc12 flank (pKG 5431; Bohnert et al., 2013). The cdc12-P31A mutation and cdc12-Δ24–36 deletion were created with mutagenesis of pKG 5431 by PCR and confirmed with sequencing. The mutant constructs were then released from the vector by digestion with Xba1 and Sac1 and transformed into wild-type S. pombe cells using a lithium acetate method (Keeney and Boeke, 1994). G418-resistant cells were selected, and the cdc12 locus was sequenced to identify transformants containing the desired and correct mutations.

For growth assays, cells were grown to log phase at 25°C in YE, 10 million cells were resuspended in 1 ml of water, and 10-fold serial dilutions were made. 2.5 µl of each dilution was spotted on YE plates, and the plates were incubated at the indicated temperatures.

The GFP-Cdc12 (aa 24–36) fragment was placed under the control of the nmt81 promoter in pREP81 plasmid vector. This construct and a construct with GFP alone were expressed in a wild-type strain. Cells were grown in media containing thiamine and then washed into media without thiamine to induce protein production and grown for 20 h at 32°C before imaging.

Cdc12 (aa 1–764) was cloned into pMAL-2c for expression as a maltose-binding protein (MBP) fusion (Carnahan and Gould, 2003). pMal-2c contains the P_tac promoter and amp^R gene. Cdc15 F-BAR (aa 19–312) was cloned into pET15b for expression as a 6×His fusion. Proteins were induced in Escherichia coli Rosetta2(DE3)pLysS cells with 0.4 mM IPTG overnight at 18°C. Protein was purified on amylose beads (New England Biolabs, Inc.) or cOmplete His-Tag resin (Roche) according to the manufacturer’s protocols. Cdc15 (aa 19–312)’s 6×His tag was removed with thrombin protease, and the protein was further purified on a HiTrap Q SP anion exchange column (GE Healthcare) and concentrated.

Recombinant proteins conjugated to amylose beads or synthetic biotinylated peptides (5 µg; GenScript) conjugated to streptavidin beads were incubated with recombinant Cdc15 F-BAR for 1 h at 4°C in binding buffer (50 mM Tris-HCl, pH 7.0, 250 mM NaCl, 2 mM EDTA, and 0.1% NP-40). For Fig. 2 A, 0.3 µM MBP-Cdc12N was incubated with 30 nM Cdc15 F-BAR. For Fig. 2 C, 30 nM peptide was incubated with 30 nM Cdc15 F-BAR. For the titration binding assay, 30 nM Cdc12 (aa 20–40) was added for all concentrations of Cdc15 F-BAR (1, 5, 10, and 50 nM). After extensive washing in binding buffer, samples were resolved by SDS-PAGE for Coomassie blue staining or Western blot analysis. For detection of biotinylated peptides, a streptavidin infrared dye was used (LI-COR Biosciences). The Cdc15 F-BAR was detected using a rabbit polyclonal antibody raised against GST-Cdc15 (aa 1–405; Roberts-Galbraith et al., 2009).

Live-cell images of S. pombe cells were acquired using a Personal DeltaVision (Applied Precision) that includes a microscope (IX71; Olympus), 60× NA 1.42 Plan Apochromat and 100× NA 1.40 U Plan S Apochromat objectives, fixed and live-cell filter wheels, a camera (CoolSNAP HQ2; Photometrics), and softWoRx imaging software (Applied Precision). Images were acquired at 25–29°C, and cells were imaged in YE media. Images in figures were maximum intensity projections of z sections spaced at 0.2–0.5 µm. Images used for quantification were not deconvolved. Other images not used for fluorescence quantification were deconvolved with 10 iterations.

Time-lapse imaging was performed on cells in log phase using a microfluidics perfusion system (CellASIC ONIX; EMD Millipore). Cells were loaded into Y04C plates for 5 s at 8 psi, and YE liquid media were flowed into the chamber at 5 psi throughout imaging.

Intensity measurements were made with ImageJ software (National Institutes of Health). For all intensity measurements, the background was subtracted by creating a region of interest (ROI) in the same image where there were no cells (Waters, 2009). The area of the background was divided by the raw intensity of the background, which was multiplied by the area of the ROI. This number was subtracted from the raw integrated intensity of that ROI (Waters, 2009). For CR intensity quantification, an ROI was drawn around the CR and measured for raw integrated density, and for whole cell intensity quantification, an ROI was drawn around the entire cell.

To compare populations of cells for all genotypes, cells were imaged on the same day with the same microscope parameters. In addition, to compare two populations of cells within the same field of view, one population was incubated with fluorescently conjugated lectin (Sigma-Aldrich), which labels cell walls. Specifically, 1 µl of a 5-mg/ml stock of TRITC-lectin in water was added to 1 ml of cells for a final concentration of 5 µg/ml. Cells were then incubated for 10 min at room temperature, washed three times, and resuspended in media. The lectin-labeled cell population and unlabeled cell population were mixed 1:1 immediately before imaging. The reciprocal labeling of populations was also performed to account for any bleed through. The fluorescence intensity of the cell tips was quantified by using the exact same size circle, 3.5 µm in diameter, placed at the tip of each cell end and then subtracting background (Burke et al., 2014).

Fig. S1 shows that Cdc12 (aa 24–36) fused to GFP localizes to the cell division site. Fig. S2 shows that the cdc12-Δ24–36 allele is synthetically lethal with rng2-D5, myo2-E1, and mid1Δ, whereas cdc12-P31A and cdc12-Δ24–36 do not have a significant genetic interaction with cdc15-140. Fig. S3 shows that Cdc12 lacking the Cdc12–Cdc15 interaction localizes independently of F-actin to the CR. Table S1 lists the genotypes of all strains used in this study.

We thank Christine Jones, MariaSanta Mangione, Rodrigo Guillen, Quanwen Jin, and especially Dr. Janel Beckley for helpful discussions and critical reading and editing of this manuscript.

We are grateful for the following support: National Institutes of Health grants GM101035 (to K.L. Gould) and T32-CA119925 (for K.A. Bohnert) and American Heart Association grant 14PRE19740000 (to A.H. Willet).

The authors declare no competing financial interests.