Distinct roles of Rho1, Cdc42, and Cyk3 in septum formation and abscission during yeast cytokinesis

Cytokinesis is the postmitotic event that physically separates the cytoplasms of the daughter cells. In both fungal and animal cells, cytokinesis is performed in two steps, cleavage-furrow formation and abscission. Formation and ingression of the cleavage furrow involve many cellular processes such as actomyosin ring (AMR) contraction, targeted membrane deposition, and ECM rearrangements (Balasubramanian et al., 2004; Barr and Gruneberg, 2007; Fededa and Gerlich, 2012), among which the AMR is the most extensively characterized. The majority of AMR components (F-actin, type-II myosin, and other associated proteins) are conserved in fungi, amoebae, and animals (Balasubramanian et al., 2004). Thus, the mechanisms of AMR function are likely to be conserved also in these species, although the degrees to which cleavage-furrow ingression depends on AMR contraction may vary (Balasubramanian et al., 2004; Uyeda and Nagasaki, 2004). Membrane trafficking also appears to be critical for furrow ingression (McKay and Burgess, 2011), and perturbation of various intracellular trafficking systems has a strong impact on this process and/or on subsequent abscission (Albertson et al., 2005). The involvement of ECM rearrangements in cytokinesis is less well understood, but many types of ECM components (proteins, proteoglycans, and polysaccharides) localize to the cleavage furrow, and some of them have been shown to be required for normal cytokinesis (Hwang et al., 2003; Mizuguchi et al., 2003; Ng et al., 2005; Xu and Vogel, 2011).

After cleavage-furrow ingression, the intercellular bridge connecting the two daughter cells is disassembled to irreversibly separate the daughter cells (Byers and Abramson, 1968; Schiel and Prekeris, 2010; Neto and Gould, 2011), a process termed abscission. In mammalian cells, abscission appears to involve anchoring of the plasma membrane to the midbody and subsequent resolution, in which the septins, secretory vesicles, and recycling endosomes are all implicated (Estey et al., 2010; Schiel and Prekeris, 2010; Neto and Gould, 2011). In Drosophila melanogaster S2 cells, F-actin appears to be essential for cleavage-furrow ingression but dispensable for abscission (Echard et al., 2004), suggesting that there are AMR-dependent and -independent events in animal cytokinesis. Among the key questions for a comprehensive understanding of cytokinesis are how these events are organized temporarily and spatially to accomplish successful cytokinesis and whether there is a central regulatory molecule or system that governs them.

The budding yeast Saccharomyces cerevisiae is an ideal model organism for study of the AMR-independent processes involved in cytokinesis because cells lacking the only type II myosin, and hence the AMR, are viable (Bi et al., 1998; Wloka and Bi, 2012). In wild-type cells, contraction of the AMR and cleavage-furrow ingression start concomitantly with synthesis of an extracellular polysaccharide, chitin, between the two layers of plasma membrane in the furrow to form the primary septum (PS; Shaw et al., 1991; Wloka and Bi, 2012). Chitin in the PS is synthesized mainly by the chitin synthase Chs2 (Shaw et al., 1991); the proteins Iqg1, Inn1, Hof1, and Cyk3 appear to be involved in PS formation by directly or indirectly contributing to Chs2 activation (Epp and Chant, 1997; Lippincott and Li, 1998; Korinek et al., 2000; Sanchez-Diaz et al., 2008; Jendretzki et al., 2009; Tully et al., 2009; Nishihama et al., 2009; Meitinger et al., 2010; Palani et al., 2012). Because Chs2 is delivered to the division site by exocytosis (Chuang and Schekman, 1996; VerPlank and Li, 2005), PS formation is a cytokinetic event achieved by the concerted actions of AMR contraction, membrane trafficking, and ECM rearrangement.

Because the transmembrane protein Chs2 and the AMR are both positioned at the leading edge of the cleavage furrow, where membrane fusion should take place during abscission, it is conceivable that furrow ingression and PS formation are inherently incapable of providing a mechanism for abscission because their machineries physically obstruct the process, necessitating another mechanism for abscission. An abscission defect in yeast has been observed as a consequence of perturbed spindle–midzone function (Bouck and Bloom, 2005), and a pathway named NoCut was found (Norden et al., 2006; Mendoza et al., 2009) to be responsible for this checkpoint function, which is thought to transduce a signal from the chromatin and spindle midzone to the cortical membrane. However, the precise mechanisms of abscission in yeast remain obscure.

As PS formation nears completion, the modes of ECM rearrangement and membrane trafficking change, and the secondary septa (SSs), consisting mainly of glucan, begin to be laid down on both mother and daughter sides of the PS (Cabib, 2004). The completed trilamellar septum is then partially degraded by a chitinase and glucanases that are secreted asymmetrically by the daughter cell (Kuranda and Robbins, 1991; Colman-Lerner et al., 2001; Cabib, 2004), thereby concluding cell separation.

In animal cells, the small GTPase RhoA appears to regulate multiple processes in cytokinesis (Piekny et al., 2005; Fededa and Gerlich, 2012). RhoA is bound to the membrane by CAAX prenylation at its C terminus and activated to its GTP-bound state and concentrated at the division site by a guanine nucleotide exchange factor (GEF). RhoA-GTP then binds to various downstream effectors such as the actin-nucleator formin and the Rho-associated protein kinase to regulate AMR assembly and initiation of contraction (Matsumura, 2005). In addition, RhoA may be involved in abscission, at least in some cell types. For example, in HeLa cells, RhoA has been reported to be activated in late telophase (Yoshizaki et al., 2004), localize to the midbody, and promote abscission (together with Citron kinase and anillin; Gai et al., 2011), and expression of a dominant-negative form of the RhoA GEF Ect2 caused a specific defect in abscission (Chalamalasetty et al., 2006).

During S. cerevisiae cytokinesis, the RhoA homologue Rho1 promotes AMR formation through the formin Bni1 (Tolliday et al., 2002). Rho1 activity during this process appears to be controlled mainly by the Polo kinase Cdc5 through two GEFs, Tus1 and Rom2 (Yoshida et al., 2006). Rom2 may also be regulated by phosphatidylinositol 4,5-bisphosphate (PIP₂) produced by the Mss4 phosphatidylinositol 5-kinase in the plasma membrane (Audhya and Emr, 2002). PIP₂ also contributes to the accumulation of Rho1 at the division site after mitotic exit through interaction with the polybasic sequence at the C terminus of Rho1; this pool of Rho1 can promote cytokinesis in myo1Δ cells via recruitment of the chitin synthase Chs3, without restoring actin ring assembly (Yoshida et al., 2009). In addition, Rho1 is known to use the glucan synthases Fks1/2, the protein kinase Pkc1, and the exocyst subunit Sec3 as effectors to regulate cell wall synthesis, the cell wall integrity MAPK pathway, and exocytosis, respectively, which contribute to a broad range of cellular events such as stress response, bud growth, wound healing, and cell cycle progression (Guo et al., 2001; Levin, 2005; Kono et al., 2008, 2012).

In this study, we investigated the contribution of SS formation to abscission in yeast and identified genes involved in this process by a genetic screen. Further characterization of these genes has suggested a role for Rho1 in SS formation and a novel mechanism for its regulation by the catalytically inactive transglutaminase domain in Cyk3.

In mutants defective in anaphase-promoting complex function or degradation of Iqg1, such as cdh1Δ, the PS frequently fails to fully span the division plane, and the gap is filled with materials reminiscent of the SS (Tully et al., 2009). Similar discontinuities in the PS, typically ∼40 nm in diameter, were observed also in wild-type cells (Fig. 1 A), although in lower frequency (∼5% vs. the ∼27% observed in cdh1Δ cells). To ask whether such discontinuities were normal structures that had been missed in conventional EM thin sections, we performed serial section EM on wild-type cells. We found such discontinuities in four of the seven cells observed (Fig. 1 B). As the discontinuities are only ∼40 nm and the sections used were ∼80 nm thick, it is likely that some were missed and that most or all wild-type septa have a similar structure. Collectively, these results, those of Tully et al. (2009), and other evidence (see Introduction) suggest that PS formation alone is not sufficient to complete cytokinesis and that abscission is delayed in mutants defective in timely disassembly of the AMR.

Next, we asked whether the process that follows PS formation, SS formation, might contribute to abscission. We stained wild-type cells expressing Myo1-GFP with aniline blue, a dye that binds specifically to 1,3-β-glucan, the main constituent of the SS (Young and Jacobs, 1998). As shown in Fig. 1 C, the dye stained only general cell wall in the majority (∼90%, n = 104) of large-budded cells before (Fig. 1 C, image 1) and during (Fig. 1 C, image 2) AMR constriction. In contrast, in cells in which the AMR had contracted but not disassembled (n = 34), strong staining was usually obvious either at the periphery of the division plane (41%; Fig. 1 C, image 3) or as a continuous line across it (29%; Fig. 1 C, image 4). After AMR disassembly, cells always showed a continuous line of staining across the division plane (100%, n = 37; Fig. 1 C, image 5). Thus, SS formation appears to begin before the plasma membrane is resolved to separate the mother and daughter cytoplasms, suggesting that it might play a role in abscission.

SS formation and its regulation remain poorly understood, mostly because of a lack of mutants or conditions that exhibit specific defects in this process. We reasoned that a dosage-suppressor screen of mutants defective in PS formation might identify genes involved in SS formation. We performed such a screen using a cyk3Δ hof1Δ double mutant (see Materials and methods). This mutant is almost inviable (Fig. 2 A; Korinek et al., 2000), probably as a result of its severe inefficiency in PS formation (Fig. 3, A and E). In addition to the expected CYK3 and HOF1, we isolated nine suppressor genes from two independent genomic libraries (Fig. S1 A). Six of these genes (Fig. 2 A) encode proteins functionally related to two Rho family small GTPases, Rho1 and Cdc42: RHO1 itself and its homologue RHO2, proteins that function upstream and downstream of Rho1 (MSS4 and KDX1, respectively), and direct regulators of Cdc42 (RGD2 and a C-terminally truncated version of CDC24, designated cdc24*; see below). Interestingly, unlike CYK3, which induces Chs2-dependent PS formation (Nishihama et al., 2009), most of these genes were also able to suppress chs2Δ (Fig. 2 A), indicating that the Rho-related genes suppress through a pathway independent of PS formation. Characterization of the other three genes, PSP1, NAB6, and PAP1, will be described elsewhere.

Consistent with the independence of suppression from Chs2, little or no improvement in PS formation was observed during suppression by RHO1 or RHO2 (Fig. 3, B and E, right; and Fig. S2 A), RGD2 (Fig. 3, C and E), cdc24* (Fig. 3, D and E), or KDX1 (not depicted). Instead, the suppressed cells had very thick SS-like structures that spanned the bud necks with slightly better efficiency than was seen in control cells (Fig. 3 E, left), probably accounting for the higher growth rates of the suppressed cells. These results suggest that the Rho-related genes hyperactivate SS formation and that the SS can be an alternative to the PS as a driving force for successful (albeit abnormal) cytokinesis. Similar improvements in growth and hyperactivation of SS formation were observed using an iqg1Δ mutant (Fig. S1 A and Fig. S2, B–E), indicating that the suppression mechanisms do not require the AMR, which fails to form in iqg1Δ mutants (Epp and Chant, 1997; Lippincott and Li, 1998). Interestingly, many cyk3Δ hof1Δ cells suppressed by RHO1 or RHO2 had SS without detectable PS (Fig. 3, B and E; and Fig. S2 A), suggesting that these genes (but not cdc24* or RGD2) also have inhibitory effects on PS formation.

Each of the aforementioned suppressors either activates the Rho1 pathway or inactivates the Cdc42 pathway, with the apparent exception of cdc24*. We extended the test to other GEFs and GTPase-activating proteins (GAPs) that regulate these small GTPases. Activation of Rho1 by either overexpression of certain Rho1-GEF genes (TUS1 and ROM1) or deletion of certain Rho1-GAP genes (SAC7 or LRG1) suppressed cyk3Δ hof1Δ (Fig. S1 B and Fig. 2 B, left and middle images). In striking contrast to cdc24*, wild-type CDC24 failed to suppress cyk3Δ hof1Δ (Fig. S1 B). Further analyses (Fig. S3) suggested that the chimeric Cdc24* protein, consisting of portions of Cdc24 and of the tetA tetracycline resistance protein (from the vector backbone), functions as an inhibitor of Cdc42. Consistent with this hypothesis, overexpression of certain Cdc42 GAPs (RGA1 and BEM3) also suppressed the cyk3Δ hof1Δ mutant (Fig. S1 B). Interestingly, RGD1, which encodes a GAP for Rho3 and Rho4 (Prouzet-Mauléon et al., 2008), also suppressed the mutant, suggesting that these proteins may also be involved in cytokinesis. In most cases, suppression of an iqg1Δ mutant paralleled those with the cyk3Δ hof1Δ mutant (Fig. S1, A and B). Collectively, our genetic data support the model that either activation of Rho1 or inactivation of Cdc42 is sufficient to suppress effects in AMR and/or PS formation by inducing SS formation (Fig. 2 C).

We also found that deletion of the paxillin homologue gene PXL1 weakly suppressed both cyk3Δ hof1Δ and hof1Δ (Fig. 2 B and Fig. S4 A). PXL1 was identified as an allele-specific dosage suppressor of cdc42 point mutations, and both genetic and biochemical evidence suggests that Pxl1 is a negative regulator of Rho1 (Gao et al., 2004; Mackin et al., 2004). Because the timing and pattern of Pxl1-GFP localization also coincide with that of SS formation (Fig. S4, B–D), it seems likely that Pxl1 connects Cdc42 and Rho1 during SS formation (Fig. 2 C), serving as a molecular brake for excessive SS formation.

The extensive formation of SS when Rho1 was overexpressed or activated in cytokinesis mutants suggested that it may have a role in SS formation in wild-type cells. To test this hypothesis, we induced the expression of a GDP-locked, dominant-negative mutant of Rho1 (T24N) in cells released from an anaphase block produced by a cdc15-2 mutation. Under these conditions, the AMR was already formed at the time of release (not depicted; Yoshida et al., 2006), and the initiation and rate of furrow ingression were not affected by the expression of Rho1^T24N (Fig. 4, A and D). However, the Rho1^T24N-expressing cells completely failed to separate even after 105 min (Fig. 4 A) or 12 h (not depicted). Some cells eventually formed tiny buds (Fig. 4, A and C), indicating that the cells did not simply die upon induction of Rho1^T24N and thus that the separation defect was triggered specifically by inhibition of a Rho1 pathway. EM analyses revealed that at 75 min, ∼87% of control cells had formed SS (Fig. 4 F, top), whereas only ∼30% of the Rho1^T24N-expressing cells had detectable SS even though normal-looking PSs were formed (Fig. 4 F, bottom), indicating that SS formation was blocked or delayed by the dominant-negative Rho1.

Importantly, although Rho1 activity in early stages of cytokinesis is required for AMR formation (and possibly also for its contractility), it appeared to be dispensable for cleavage-furrow ingression, as indicated by nearly normal constriction of Myo1-GFP rings in Rho1^T24N-expressing cells that had been synchronized with nocodazole at a stage before actin ring formation (Fig. 4 B and not depicted) as well as in asynchronous culture (Fig. 4 D and not depicted). These results are consistent with the dispensability of the AMR and its contractility for cleavage-furrow formation in yeast (see Introduction).

Examination of Rho1-GEF mutants revealed that tus1Δ cells also exhibited bud emergence before cell separation, whereas rom1Δ and rom2Δ cells did not (Fig. 4 E). Interestingly, of 30 tus1Δ cells examined by EM, 40% had apparently complete PSs without SS, and 50% had SS only on the daughter side of the PS (Fig. S5 A and not depicted). As Tus1 remains at the bud neck after the furrow and AMR have constricted (Yoshida et al., 2006), and overexpression of TUS1 (but not of ROM1 or ROM2) could suppress cyk3Δ hof1Δ, iqg1Δ, and chs2Δ mutations (Fig. S1 B), this GEF may be primarily responsible for Rho1 activation during SS formation. Conversely, hyperactivation of Rho1 either by expression of a constitutively active construct (Rho1^Q68L) or by deletion of the GAP gene LRG1 caused formation of abnormally thickened SS (Fig. S5 A). Surprisingly, deletion of another GAP gene, SAC7, did not have a strong impact on SS structure (Fig. S5 A), although this deletion was at least as effective as lrg1Δ in suppressing cyk3Δ hof1Δ (Fig. 2 B). Collectively, these results suggest that Rho1 has a role in SS formation as well as in actin ring formation during normal cytokinesis and that Tus1 and Lrg1 are its major positive and negative regulators for this role.

Both Rho1 (Yoshida et al., 2009) and its effector Fks1 (glucan synthase; Fig. 5) arrive at the division site before cleavage-furrow ingression and hence before the initiation of SS formation. These observations suggest that Rho1 activity might be specifically suppressed during PS formation to prevent premature activation of SS formation. Consistent with this hypothesis, expression of the constitutively active Rho1^Q68L caused a substantial increase in the number of cell clusters (Fig. S5 B), suggesting that Rho1^Q68L interferes with a step in cell division. To explore the hypothesis further, we conducted pull-down assays using cdc15-2–synchronized cells and a GST-fused Rho-binding domain (RBD) from Pkc1, which binds specifically to GTP-bound Rho1 (Kono et al., 2008). As reported previously (Yoshida et al., 2006), Rho1 activity was high in the arrested cells (Fig. 6, A and B, 0 min). It then decreased during the 30 min after release and generally remained low until 75 min, except for a small increase at 45 min that was observed in each of three independent experiments (Fig. 6, A and B). In parallel experiments, furrow ingression started at <40 min after release (Fig. 6 C) and was typically complete within 7–10 min in each cell (not depicted). Because SS formation begins before AMR disassembly (Fig. 1 C), the transient activation of Rho1 coincides with, and is probably responsible for, the initiation of SS formation. The later activation of Rho1 beginning at ∼75–90 min is presumably associated with bud growth in the new cell cycle (Kono et al., 2008).

Several observations suggest that Cyk3 might be involved in a Rho1 inactivation mechanism that functions during PS formation. First, in cdc15-2–synchronized cells, Cyk3-GFP localized to the bud neck just before furrow ingression (Fig. 6 C), which coincides with the timing of Rho1 inactivation. Second, in contrast to wild-type cells, cyk3Δ cells formed PS and SS simultaneously (Fig. 7 A, left and center), suggesting that the putative temporary suppression of SS formation by Rho1 inactivation does not function properly in this mutant. Third, overexpression of Cyk3 abolished the simultaneous formation of PS and SS observed in a mitotic exit network mutant that was forced to exit mitosis with the cyclin inhibitor Sic1 (Meitinger et al., 2010). Fourth, the severe clustering phenotype of cyk3Δ cells (Fig. 7 B, left and center) appears to involve activation of Rho1 because deletion of CYK3 did not exacerbate the similar phenotype produced by the constitutively active Rho1^Q68L (Fig. S5 B). Fifth, cyk3Δ showed a synergistic growth defect at 24°C with the Rho1-GAP mutation sac7Δ (Fig. 6 D) but not with another Rho1-GAP mutation, lrg1Δ (not depicted). Lrg1, like Cyk3, localizes to the neck (as well as to the bud tip; Watanabe et al., 2001), whereas Sac7 localizes throughout the cell cortex (Fig. 8 D, image 5), and a sac7Δ lrg1Δ mutant is inviable (Lorberg et al., 2001). Thus, collectively, the data suggest that Cyk3, like Lrg1 (Svarovsky and Palecek, 2005), is involved in controlling Rho1 activity at the neck, whereas Sac7 has a more general role, and that the excessive activation of Rho 1 when both systems are inactive is harmful or even lethal. Finally, the temporary inactivation of Rho1 completely disappeared in cyk3Δ cells (Fig. 6, A and B), although Myo1-GFP constriction initiated normally (Fig. 6 C).

Several lines of evidence indicate that Cyk3 might inactivate or block activation of Rho1 by physically interacting with its inactive form. First, a pull-down assay using bacterially expressed, GST-fused full-length Cyk3 (GST-Cyk3) or its C-terminal half (GST-Cyk3^480–885; Fig. 8 A) revealed that Rho1 interacted with the C-terminal half of Cyk3 (Fig. 8 B), in contrast to Hof1, which bound only to full-length GST-Cyk3 in this assay (Fig. 8 B). Second, a similar experiment using wild-type, constitutively active (Q68L), or dominant-negative (T24N) Rho1 revealed that the affinity of Rho1 for the C-terminal half of Cyk3 was significantly increased by the T24N mutation but not by the Q68L mutation (Fig. 8 C). Third, although we have not detected a Cyk3–Rho1 interaction in vivo by coimmunoprecipitation (presumably because of the membrane association of both proteins and/or the transient nature of the interaction), we have detected an interaction using bimolecular fluorescence complementation (BiFC). We constructed a Rho1 BiFC reporter by fusing the C-terminal portion (aa 155–238) of the Venus protein to the N terminus of Rho1 via a short linker (VC155-GGSGGS-Rho1; henceforth simply VC-Rho1) and expressed this construct from a low-copy plasmid in rho1Δ cells. This reporter successfully detected interactions of Rho1 with several GEFs and GAPs (Fig. 8 D, images 1–5). We then tagged the C terminus of Cyk3 with the N-terminal portion of Venus (aa 1–155) containing an amino acid substitution (I153L) to diminish potential self-assembly between the N- and C-terminal portions of Venus (Kodama and Hu, 2010). A CYK3-VN strain showed no detectable fluorescence (Fig. 8 D, image 6), but when VC-Rho1 was also expressed, clear BiFC signal appeared at the bud neck in a temporarily regulated manner (Fig. 8 E). In most cells with large Myo1-CFP rings, signal was not detected (Fig. 8 E, cells 1 and 2), but it did appear in some such cells (cells 3 and 4), and it remained while Myo1-CFP constricted to a dot (Fig. 8 E, cells 5 and 6). After AMR disassembly, BiFC signal remained at the bud neck (Fig. 8 E, cell 7), but this signal persistence may reflect only the tight binding between the N- and C-terminal Venus fragments (Kodama and Hu, 2010) rather than a continuing association of Cyk3 with Rho1 during SS formation.

The N-terminal portion of Cyk3 contains an SH3 domain and a proline-rich region, which bind to Inn1 and Hof1, which are involved in PS formation (Fig. 8 A; our unpublished data; Nishihama et al., 2009; Jendretzki et al., 2009; Meitinger et al., 2011; Labedzka et al., 2012). The C-terminal portion of Cyk3 contains a transglutaminase core (TGc) domain that resembles the protein and peptide cross-linking transglutaminases. Interestingly, several transglutaminases are implicated in the modification and activation of mammalian RhoA (see Discussion). However, the TGc domain in Cyk3 lacks the conserved cysteine that is essential for TGase activity (Makarova et al., 1999; Pollard et al., 2012). We constructed a strain in which the TGc domain (aa 519–580) is deleted (cyk3^TGcΔ). Cyk3^TGcΔ appears to retain most Cyk3 functionality because a cyk3^TGcΔ hof1Δ double mutant (in contrast to a cyk3Δ hof1Δ double mutant: Fig. 2 A) grew as well as a hof1Δ single mutant, Myo1-GFP constricted at a normal rate in cyk3^TGcΔ cells, and Cyk3^TGcΔ-GFP localized normally to the bud neck (not depicted). However, cyk3^TGcΔ sac7Δ cells grew as poorly as cyk3Δ sac7Δ cells (Fig. 6 D). In contrast, mutations in the SH3 domain (cyk3^W45A) or the proline-rich region (cyk3^P188,191A) did not cause such a synergistic growth defect with sac7Δ (Fig. 6 D). The cyk3^TGcΔ strain showed a partial deficiency in Rho1 inactivation as judged by the GST-RBD pull-down assay (Fig. 6, A and B), a moderate premature formation of SS formation (Fig. 7 A, right), and a moderate delay in cell separation (Fig. 7 B, right). Finally, no Cyk3–Rho1 interaction could be detected by the BiFC assay when the TGc domain was deleted (Fig. 8 F). Collectively, these results suggest that the TGc domain of Cyk3 is required for effective blockage of Rho1 activation, that this role of Cyk3 is independent of its binding to the PS formation proteins Inn1 and Hof1, and that the more severe phenotype of cyk3Δ cells (compared with cyk3^W45A, cyk3^P188,191A, and cyk3^TGcΔ cells) may be a synthetic effect of slowed PS formation together with a premature activation of SS formation.

In animal cells, it has long been recognized that abscission, the final resolution of the daughter cell membranes, is a step distinct from cleavage-furrow ingression (Schiel and Prekeris, 2010; Neto and Gould, 2011). However, in yeast, this idea has been controversial, in part because conventional EM images have typically shown a PS that appears to span the entire division plane (Weiss, 2012; Wloka and Bi, 2012). Nonetheless, studies using mutants and fluorescent membrane markers have suggested that a distinct abscission step exists and is specifically inhibited in response to spindle midzone and chromosome segregation defects by the NoCut checkpoint pathway (Norden et al., 2006; Mendoza et al., 2009). Some septin mutants also appear to be specifically defective in abscission (Dobbelaere and Barral, 2004; Wloka et al., 2011). Our serial section EM revealed a discontinuity in the PSs of most wild-type cells, supporting the hypothesis that a distinct abscission step is needed to complete cytokinesis. We suggest that the abscission machinery must be distinct from the AMR and PS formation machinery because these elements localize to the leading edge of the cleavage furrow, where they would presumably interfere with the membrane fusion needed for abscission. In support of this idea, the PS discontinuity is enlarged in mutants defective in timely disassembly of the AMR (Tully et al., 2009).

Several lines of evidence suggest that SS formation is involved in abscission. First, 1,3-β-glucan, the major constituent of the SS, began to be deposited at the bud neck before the AMR was disassembled (and thus before abscission). Second, the glucan synthase Fks1 localized on the surface of the cleavage furrow, but in the area trailing the AMR, and so is positioned to function without physically interfering with membrane resolution. Finally, hyperactivation of SS formation restored abscission in mutants defective in PS formation. Mechanisms other than glucan synthesis are probably also involved in SS formation and abscission. In particular, the chitin synthase Chs3 is required for formation of remedial SS in a chs2 mutant (Shaw et al., 1991), as well as in myo1Δ cells overexpressing Rho1 (Yoshida et al., 2009), and so is likely to be involved in normal SS formation as well.

Our suppression data suggest that the small GTPases Rho1 and Cdc42 act antagonistically to regulate SS formation (Fig. 2 C and Fig. 9 C). Rho1 appears to promote SS formation, in that its overexpression or activation hyperactivated SS formation. Promotion of SS formation by Rho1 has also been observed in a myo1Δ mutant (Yoshida et al., 2009), and it has been hypothesized that Rho1 is involved in SS formation in wild-type cells (Weiss, 2012; Wloka and Bi, 2012). We have provided strong support for this hypothesis by showing that expression of a dominant-negative Rho1 in synchronized cells blocks cell division but not cleavage-furrow ingression. During SS formation, Rho1 effectors implicated in cell wall synthesis and stress response are likely to be activated, such as the glucan synthases Fks1 and Fks2 (Mazur and Baginsky, 1996; Qadota et al., 1996), the protein kinase Pkc1 (through activation of the MAPK pathway, in which Kdx1 [also known as Mlp1] is involved; Watanabe et al., 1997; Kim et al., 2008), and Chs3 (Valdivia and Schekman, 2003; Yoshida et al., 2009). In support of this hypothesis, deletions of these genes exhibit synthetic growth defects with myo1Δ (our unpublished data; Rodríguez-Quiñones et al., 2008; Yoshida et al., 2009).

In contrast, Cdc42 appears to inhibit SS formation (Fig. 2 C and Fig. 9 C), as its inactivation by overexpression of certain GAPs or the abnormal Cdc24* suppressed mutants defective in PS formation. The available evidence suggests that Cdc42 functions both by inactivation of Rho1 and by an independent mechanism. Deletion of PXL1, a factor downstream of Cdc42 that inactivates Rho1 (Gao et al., 2004; Mackin et al., 2004), suppressed cyk3Δ hof1Δ, but it did so more weakly than overexpression of either the Cdc42-GAP Rgd2 or Cdc24*. In addition, unlike Rho1, Cdc42 GAPs did not inhibit PS formation while activating SS formation. Although the exact mechanisms remain to be determined, Atkins et al. (in this issue) have shown that the p21-activated kinase Ste20 functions downstream of Cdc42 in cytokinesis, and we have independently confirmed that ste20Δ bypasses the defects in cyk3Δ hof1Δ by hyperactivation of SS formation (unpublished data). The relevant substrate of Ste20 remains to be identified.

The most downstream known components of the NoCut checkpoint pathway are the homologous Boi1 and Boi2 proteins in the cell cortex (Mendoza et al., 2009), which interact both genetically and physically with Cdc42, Rho3, Rho4, and the scaffold protein Bem1 (Bender et al., 1996; Matsui et al., 1996). Thus, it will be interesting to test whether the NoCut pathway acts through these GTPases. Similarly, the ER surveillance pathway delays cytokinesis upon ER stress, and the Pkc1-MAPK cascade downstream of Rho1 becomes essential for cell survival (Babour et al., 2010). An attractive model would be that the ER surveillance pathway delays cytokinesis at abscission, giving the cell time to adjust to the stress, and then hyperactivates SS formation through Rho1 activation to resume cell division.

As Rho1 is required for AMR formation in anaphase (Fig. 9 A; Yoshida et al., 2006), our results suggest that Rho1 activity is required at both the beginning and the end of cytokinesis. Because Rho1 localizes to the division site throughout cytokinesis (Yoshida et al., 2009; our unpublished data), it seemed likely that Rho1 simply remained active during the entire process. However, we found that in cdc15-2–synchronized cells, in which AMR formation and cleavage-furrow ingression are temporarily separated, global Rho1 activity decreased substantially before furrow ingression, suggesting that the activity Rho1 might decrease even more dramatically at the division site. Moreover, Rho1^T24N did not block AMR constriction or PS formation, suggesting that Rho1 activity is dispensable for these processes. Indeed, overexpression of Rho1 significantly inhibited PS formation in cells (cyk3Δ hof1Δ) in which this process is already inefficient, suggesting that active Rho1 may actually inhibit PS formation. The dispensability of Rho1 for PS formation seems surprising given that proper localization of Chs2 requires the exocyst subunit Sec3 (VerPlank and Li, 2005), which is regulated by Rho1 as an effector (Guo et al., 2001). However, another Rho protein, such as Rho3 or Cdc42 (Wu and Brennwald, 2010), may be primarily responsible for Chs2 delivery during PS formation.

It remains possible that a small quantity of active Rho1 localizes to the leading edge of the cleavage furrow and promotes AMR contractility as seen in many other organisms (Fig. 9 B; Balasubramanian et al., 2004; Piekny et al., 2005; Bement et al., 2006). However, because of the force provided by PS formation, the AMR in S. cerevisiae can constrict at a nearly normal rate even when no force is provided by actomyosin contractility (Lord et al., 2005; Fang et al., 2010), making it difficult to assess intrinsic AMR contractility in this system.

Consistent with our observations in yeast, it appears that in at least some types of mammalian cells, a deficiency in RhoA activity at the division site does not block furrow ingression (O’Connell et al., 1999; Yoshizaki et al., 2004). Thus, once the AMR is formed and its contraction is initiated, RhoA may not be required for further cleavage-furrow ingression, and its role in abscission (Chalamalasetty et al., 2006; Gai et al., 2011) may be more important.

Cyk3 was originally identified as a dosage suppressor of iqg1Δ (Korinek et al., 2000), and its overexpression has been found to promote PS formation in both wild-type and mutant (hof1Δ, myo1Δ, inn1Δ, and a hypomorphic chs2 mutation) strains (our unpublished data; Nishihama et al., 2009; Oh et al., 2012; Wloka and Bi, 2012). Subsequent studies have shown that binding of the SH3 domain and PXXP motifs of Cyk3 to Inn1 and Hof1, respectively, is important for PS formation (unpublished data; Nishihama et al., 2009; Meitinger et al., 2011). In addition, it has been reported that overexpression of Cyk3 suppressed SS formation in a mitotic exit network mutant forced to exit from mitosis (Meitinger et al., 2010), although the mechanism was not defined. In this study, we found a previously unknown link between Cyk3 and Rho1 that involves the TGc domain of Cyk3. Both cyk3Δ and cyk3^TGcΔ mutants showed simultaneous formation of PS and SS, a synthetic growth defect with deletion of the Rho1 GAP gene SAC7, and a complete loss of the temporary inactivation of Rho1 during furrow ingression. Pull-down experiments showed that the C-terminal region of Cyk3 bound preferentially to inactive Rho1 and that this interaction was lost in the cyk3^TGcΔ mutant, and BiFC assays reproduced these findings in vivo.

Transglutaminases from bacterial toxins and human cells have been shown to directly deamidate or transglutaminate Gln-63 of human RhoA (Horiguchi et al., 1997; Schmidt et al., 1997; Singh et al., 2001), causing a loss of GTPase activity and thus constitutive activity of the protein. Interestingly, the TGc domain in Cyk3 lacks a Cys (Pro in Cyk3) that is one of the three residues critical for transglutaminase activity (Makarova et al., 1999; Pollard et al., 2012) and so should be enzymatically inactive. Therefore, Cyk3 may function by binding to Rho1 without modifying or activating it. Because the switch II region of RhoA is necessary and sufficient for its recognition by the bacterial toxin transglutaminases CNF1 and DNT (Lerm et al., 1999; Jank et al., 2006) and is also the central recognition site for Rho-GEFs (Hakoshima et al., 2003), the binding of Cyk3 to Rho1-GDP could interfere with its normal activation by Rho-GEFs (Fig. 9 B).

Because Rho1 appears to have an inhibitory effect on PS formation and thus furrow ingression, it is possible that the promotion of PS formation by overexpression of Cyk3 may be through an inhibition of Rho1. However, we consider this to be unlikely because the phenotype of cyk3Δ is more severe than cyk3^TGcΔ, and unlike cyk3^TGcΔ, mutations in the N-terminal region (cyk3^W45A and cyk3^P188,191A) did not show genetic interactions with sac7Δ. Both of these results suggest that the N-terminal and C-terminal regions of Cyk3 have genetically separable roles. Therefore, Cyk3 appears to be a bifunctional protein, with its N-terminal region involved in promoting PS formation through interactions with Inn1 and Hof1, and its C-terminal region involved in suppression of SS formation through interaction with Rho1.

It is presently unclear how Rho1 is reactivated for SS formation (Fig. 9 C). Although the BiFC signal between Cyk3 and Rho1 persisted after AMR disassembly, this signal may well be an artifact caused by the inherent irreversibility of binding of the BiFC probes that hinders its use for analysis of dissociation between two molecules (Kodama and Hu, 2010). Thus, it seems more likely that Cyk3 dissociates from Rho1 to allow its reactivation and that the dissociation is regulated by some mechanism such as phosphorylation.

There appears to exist a subclass of eukaryotic transglutaminase-like proteins that lack one or more of the three residues required for catalytic activity (Makarova et al., 1999), and some of them have been implicated in cytokinesis. For example, a Schizosaccharomyces pombe cyk3Δ mutant shows reduced efficiency of PS formation and premature SS formation in certain mutant backgrounds (Pollard et al., 2012), suggesting that the S. pombe and S. cerevisiae proteins have similar roles despite the apparent differences in cell wall compositions and mechanisms of cytokinesis between these yeasts. Moreover, overexpression of S. pombe Cyk3 induces cell wall damage and morphology defects (Pollard et al., 2012), which could be a consequence of Rho1 inhibition. Drosophila hillarin is concentrated at the cleavage furrow during cell division, and in a septin mutant background, a mutation in hillarin causes increased polyploidy (Ji et al., 2005). Similarly, Caenorhabditis elegans Ltd-1 is expressed during seam cell division and localizes in a pattern reminiscent of actin (Vargas et al., 2002). It will be interesting to determine whether these other transglutaminase-like proteins also function, at least in part, by regulating Rho proteins.

The yeast strains and plasmids used are listed in Table 1 and Table 2. Standard culture media (including the buffered rich medium YM-P) and genetic techniques were used (Lillie and Pringle, 1980; Guthrie and Fink, 1991). The PCR method (Longtine et al., 1998) was used for gene deletion and tagging except where noted. 1 mg/ml 5-fluoroorotic acid (5-FOA; Research Products International) was added where noted. Cell cycle synchronization using cdc15-2 or nocodazole (Toronto Research Chemicals) was performed essentially as described by Futcher (1999). In brief, cdc15-2 cells were cultured at 24°C in either YM-P medium or synthetic complete (SC) medium with appropriate supplements and carbon sources to early log phase (OD₆₀₀ ≈ 0.2) and then arrested by gradually increasing the temperature to 37°C. After 3 h, the cells were released by rapidly cooling the medium to 24°C on ice. For nocodazole synchronization, cells were cultured in YPD medium at 24°C to early log phase, and nocodazole was added at 15 µg/ml. After 2 h, the cells were collected and washed by filtration and then resuspended in fresh medium. Aniline blue was obtained from EMD Millipore, and a 0.1% Ponceau S solution was obtained from Sigma-Aldrich.

To construct pGEX2T-CYK3 and pGEX2T-CYK3^480–885, the appropriate regions were PCR amplified from yeast genomic DNA with BamHI and HindIII sequences on the 5′ and 3′ ends, respectively, and inserted between the corresponding sites in pGEX-2T. YEp13-CDC24 was constructed by PCR amplifying a DNA fragment containing nucleotides 2,296–2,859 relative to the start site of CDC24 and 1–40 relative to the SphI of YEp13 (incorporated in the primer) using pRS315-CDC24 (provided by E. Bi, University of Pennsylvania, Philadelphia, PA) as a template. This fragment was cloned by gap repair into SphI-digested YEp13-cdc24* (Table 2) by cotransforming into YEF473A.

pRS315GW-HOF1, pRS315GW-MYO1, pRS425GW-CYK3, pRS425GW-TUS1, pRS425GW-ROM1, pRS425GW-ROM2, pRS425GW-RGD1, pRS425GW-SAC7, and pRS425GW-LRG1 were constructed as follows. Genomic DNA fragments containing each gene were amplified by PCR (positions relative to the start sites: HOF1, −1,000 to 2,510; MYO1, −782 to 6,150; CYK3, −900 to 3,248; TUS1, −500 to 4,424; ROM1, −1,000 to 3,968; ROM2, −500 to 4,571; RGD1, −1,000 to 2,501; SAC7, −1,000 to 2,465; and LRG1, −1,000 to 3,554), cloned into the pCR8/GW/TOPO vector (Invitrogen), and transferred into pRS315-attR or pRS425-attR (this study) by Gateway recombination (Invitrogen). YEp181-RGA2 was constructed by ligating a SacI–KpnI fragment from pDLB1981 (Gladfelter et al., 2002) into the corresponding sites of YEplac181 (Gietz and Sugino, 1988).

YEp13-cdc24ΔPB1 was constructed by using a mutagenesis kit (QuikChange; Agilent Technologies) to introduce a stop codon into YEp13-cdc24* immediately before the BamHI site at the junction between the CDC24 and tetA coding sequences. YEp13-cdc24*-DH8 was constructed in two steps. First, a fragment containing nucleotides 764–1,659 relative to the start site of CDC24 and two mutations (N452G and E453G) within the DH (Dbl homology) domain was PCR amplified using plasmid pRS414-3×mycCdc24-DH8 (provided by R. Arkowitz, Université de Nice, Nice, France) as a template. The PCR product was then cloned by gap repair into BbvCI-cut YEp13-cdc24*. To construct pRS425-CDC24-GFP, an XhoI fragment from pRS315-CDC24-GFP (provided by E. Bi) was integrated by gap repair into NotI-digested pRS425 (Christianson et al., 1992). YEp13-cdc24*-GFP was constructed in three steps. First, a DNA fragment containing base pairs 2,261–2,281 of CDC24, the GFP sequence, and base pairs 1–40 relative to the BamHI site of YEp13 (incorporated in the primer) was PCR amplified using pRS315-CDC24-GFP as a template. Second, pRS315-CDC24-GFP was digested by NheI and BglII to generate a DNA fragment containing base pairs −146 to −1 relative to the BamHI site of YEp13 and the CDC24 gene. Finally, these two fragments were integrated into BamHI-digested YEp13 by gap repair. YEp13-cdc24ΔPB1-GFP was constructed in the same way except that a primer generating a stop codon between the GFP sequence and the BamHI site was used in the PCR step.

YCp-P_GAL-RHO1^T24N was constructed as follows. First, site-directed mutagenesis was performed in pRS316-RHO1 (this study) using the QuikChange kit (primer 5′-CTGTGGTAAGAACTGTTTATTAATC-3′ and its complement). The resulting pRS316-RHO1^T24N did not yield any yeast transformants, presumably as a result of dominant lethality of this allele. Next, the RHO1^T24N ORF was PCR amplified with DNA fragments homologous to base pairs −40 to −1 and 1–40 relative to the SalI site of YCpIF2 (Foreman and Davis, 1994) at the 5′ and 3′ ends, respectively. Finally, this PCR product was integrated into SalI-digested YCpIF2 by gap repair. YIp128-P_GAL-RHO1^Q68L was constructed as follows. First, the NdeI–SacI fragment containing RHO1^Q68L from pRS304-P_cdc42-12myc-RHO1^Q68L was inserted into the corresponding sites of pRS304-P_GAL1 (both plasmids provided by D. Lew, Duke University, Durham, NC). From the resulting plasmid, a DNA fragment containing P_GAL-RHO1^Q68L was released by digestion with EcoRI and SacI and ligated into the corresponding sites of YIp128 (Gietz and Sugino, 1988).

To construct pRS306-3HA-RHO1, plasmid pRS316-RHO1 was digested at the NotI site in the vector, blunt ended, and self-ligated to eliminate the site. A new NotI site was then introduced immediately downstream of the RHO1 start codon by site-directed mutagenesis using the QuikChange kit. A DNA fragment encoding 3HA with NotI site ends was then PCR amplified using YEpHA-BUD8F (Schenkman et al., 2002) as a template and inserted at the new NotI site. Finally, the insert of this plasmid was transferred into pRS306 using the HindIII and SacI sites in both plasmids. YCp-P_GAL-3HA-RHO1, YCp-P_GAL-3HA-RHO1^T24N, and YCp-P_GAL-3HA-RHO1^Q68L were constructed using the same approach as for YCp-P_GAL-RHO1^T24N (see preceding paragraph), using pRS316-RHO1, pRS316-RHO1^T24N, and pRS304-P_cdc42-12myc-RHO1^Q68L (provided by D. Lew) as PCR templates. The PCR products were integrated into SalI–PstI-digested YCpIF2-3HA-BUD8 (this study).

Plasmid pFA6a-VN155^I153L-TRP1 was constructed by PCR amplifying a fragment encoding aa 1–155 of Venus using pFA6a-VN173-TRP1 (Sung and Huh, 2007) as a template and incorporating an I to L mutation at aa 153; the PCR used an F2 primer (Longtine et al., 1998) with a BamHI site upstream of the coding sequence and a reverse primer (5′-ATTTAGAAGTGGCGCGCCCTAGGCGGTGAGATAGACGTTGTGGCT-3′) that included an AscI site (underlined). The PCR product was digested with BamHI and AscI and then ligated into the corresponding sites of pFA6a-VC155-TRP1. Plasmid pRS315-VC-GGS₂-RHO1 was constructed by PCR amplifying a sequence encoding the C-terminal portion of Venus using pFA6a-VC155-TRP1 as a template and primers (5′-GAAAGATGTGCGGCCGCCGTCCGGCGTGCAAAATCCCG-3′ and 5′-GGGGATCCCTTGTACAGCTCGTCCATGCCGAGA-3′) that included NotI and BamHI sites (underlined). The PCR product was digested with NotI and BamHI and then ligated into the corresponding sites of pRS315-GFP-GGS₂-RHO1 (this study).

To construct plasmid pRS315-CYK3^TGcΔ, two DNA fragments containing genomic regions around the CYK3 ORF were PCR amplified using plasmid pRS315-CYK3 (this study) as a template and primers 5′-ATTAACCCTCACTAAAGGGA-3′ and 5′-CATTTCCTAAGTATAGTATTTTCTCTAAATTTTGG-3′ (fragment 1) and 5′-AATACTATACTTAGGAAATGTAACCAACCCAATTC-3′ and 5′-TAATACGACTCACTATAGGG-3′ (fragment 2). Fragment 1 contained the promoter and base pairs 1–1,555 of the CYK3 ORF, and fragment 2 contained base pairs from 1,740 of the ORF through the stop codon and terminator. These two fragments have 20 bp of homologous sequence at their 1,555 and 1,740 ends and were integrated into ApaI–SacI-digested pRS315 (Sikorski and Hieter, 1989) by gap repair.

A cyk3Δ hof1Δ strain carrying a low-copy, URA3-marked HOF1 plasmid (MOY66) was transformed with a LEU2-marked high-copy (2µ) library in plasmid YEp13 (DeMarini et al., 1997), grown on a SC-Ura-Leu plate to form colonies, and replica plated onto SC-Leu + 5-FOA to select for ability to lose the HOF1 plasmid. From 35 transformants identified initially, plasmid recovery and retransformation yielded 22 plasmids that reproducibly rescued strain MOY66. Subcloning and/or deletion analyses identified the suppressor genes in 16 of these plasmids as (numbers of isolates in parentheses) CYK3 (3), HOF1 (2), RHO1 (2), KDX1 (2), RGD2 (1), cdc24* (1), PAP1 (2), NAB6 (2), and the 5-FOA resistance gene YJL055W (1; Ko et al., 2008). The remaining six plasmids contained distinct genome fragments in which the suppressor genes have not been determined. A parallel screen used a yeast genomic tiling library (Thermo Fisher Scientific) that was pooled from 384-well plates to five sublibraries and transformed separately; it yielded 26 plasmids that contained HOF1 (five plasmids), RHO1 (five plasmids), RHO2 (two plasmids), MSS4 (three plasmids), and RGD2 (11 plasmids), plus other plasmids that were isolated once each and not analyzed further.

Differential interference contrast and fluorescence microscopy were performed using a microscope (Eclipse 600-FN; Nikon), an Apochromat 100×/1.40 NA oil immersion objective, a cooled charge-coupled device camera (ORCA-2; Hamamatsu Photonics), and MetaMorph version 7.0 software (Molecular Devices). Most time-lapse microscopy was performed using the same equipment, by sandwiching cells suspended in SC medium between a coverslip and a slide and sealing with VALAP, as described previously (Nishihama et al., 2009). In one case (Fig. 5), the cells were spotted on a thin layer of YPD medium + 1% low-melting agarose (SeaPlaque; FMC Corporation), which was then placed on a coverslip. EM was performed essentially as described by Pollard et al. (2012): cells were collected by filtration, fixed with glutaraldehyde and potassium permanganate, embedded in LR white resin (Fluka; Sigma-Aldrich), and stained with uranyl acetate and lead citrate. Images were obtained using a microscope (JEM-1230; JEOL) equipped with a cooled charge-coupled device camera (967 or Orius SC1000A; Gatan). Images were postprocessed using the MetaMorph, Photoshop (Adobe), and/or ImageJ (National Institutes of Health) software.

Expression and purification of GST-tagged proteins from Escherichia coli BL21 (DE3 and pLysS) were performed as described by Frangioni and Neel (1993). In brief, E. coli strains transformed with the expression vectors were cultured in LB medium at 37°C to OD₆₀₀ ≈ 0.3, and IPTG was added at 0.1 mM. The cells were collected by centrifugation after 3 h of incubation at 37°C, resuspended in STE (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, and 1 mM EDTA) + 0.1 mg/ml lysozyme, and incubated for 15 min on ice. The cells were lysed by addition of 1.5% (final concentration) sarkosyl and brief sonication on ice, and the lysate was cleared by centrifugation at 12,000 g for 5 min at 4°C. Supernatants were supplemented with 2% (final concentration) Triton X-100, incubated with glutathione–Sepharose beads for 1 h at 4°C and then washed three times with Rho pull-down (RPD) or RIPA buffer (see below) depending on the downstream application. The beads were then resuspended in the same buffer supplemented with 10% glycerol and stored at −80°C. The pull-down assays using the RBD from Pkc1 was performed as described by Kono et al. (2008) with modifications. In brief, protein extracts were prepared from frozen yeast cells using glass beads in RPD buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 12 mM MgCl₂, 1 mM DTT, and 1% NP-40) supplemented with a complete protease inhibitor cocktail (Roche), and centrifuged at 12,000 g for 10 min. 3 mg of each cleared extract was incubated with 10 µl GST-RBD beads for 1 h at 4°C, washed three times with RPD buffer, and eluted by boiling in SDS sample buffer for 3 min. Samples were analyzed by SDS-PAGE (14%) and Western blotting using a peroxidase-conjugated mouse anti-HA antibody (3F10; Roche). 15 µg of the cleared extract was loaded in the input lanes. The pull-down assay using GST-Cyk3 and GST-Cyk3^(480–885) was performed similarly but using RIPA buffer (6 mM Na₂HPO₄, 4 mM NaH₂PO₄, 150 mM NaCl, 2 mM EDTA, 1 mM DTT, 50 mM NaF, 0.1 mM NA₃VO₄, 1% Triton X-100, 1% sodium deoxycholate, and 0.1% SDS) for protein extraction and incubation with the beads. Samples were analyzed by SDS-PAGE (14% for 3HA-Rho1 and 10% for Hof1–tandem affinity purification (TAP) and Western blotting using the anti-HA antibody or (for Hof1-TAP) peroxidase antiperoxidase soluble complex (Sigma-Aldrich).

Fig. S1 shows an expanded version of the data shown in Fig. 2 A, and also includes the iqg1Δ suppression results and the suppression by some GEFs and GAPs. Fig. S2 shows EM images of cyk3Δ hof1Δ and iqg1Δ cells expressing some of the suppressors. Fig. S3 shows a detailed characterization of cdc24*, which lead to the conclusion that it inhibits the Cdc42 pathway. Fig. S4 shows data implicating Pxl1 in regulation of SS formation. Fig. S5 shows supporting evidence of cytokinesis defects caused by ectopically activated Rho1.

We thank M. Wang of our laboratory for strains and permission to cite her unpublished data; K. Nakashima, R. Schekman, M. Hall, D. Lew, W.-K. Huh, E. Bi, Y. Ohya, K. Kono, S. Yoshida, and D. Pellman for strains and plasmids; J. Mulholland and J. Perrino (Stanford University Cell Sciences Imaging Facility) for assistance with EM; and B. Atkins, S. Yoshida, and D. Pellman for sharing unpublished information.

M. Onishi was supported in part by a postdoctoral fellowship from the Uehara Memorial Foundation. This work was supported by National Institutes of Health grant GM31006 (to J.R. Pringle).