A novel interplay between GEFs orchestrates Cdc42 activity during cell polarity and cytokinesis in fission yeast

Growth and division, fundamental processes in all cells, are essential for proper function and proliferation, and function through polarization. Cell polarization relies on the ability of the cytoskeleton to establish unique domains at the cell cortex to govern the local function and activity of specific proteins (Drubin and Nelson, 1996; Nance and Zallen, 2011). The Rho family of small GTPases serves as the primary regulators of cell polarity via actin regulation (Ridley, 2006). Although Rho GTPases have numerous functions, regulation of actin cytoskeleton organization is a prominent feature of these proteins. GTPases are active when GTP-bound and inactive once they hydrolyze GTP to GDP. Guanine nucleotide exchange factors (GEFs) activate GTPases by promoting the binding of GTP, while GTPase activating proteins (GAPs) inactivate GTPases by promoting GTP hydrolysis (Bos et al., 2007). A member of the Rho family of GTPases, Cdc42, is a master regulator of polarized cell growth and membrane trafficking in eukaryotes (Estravís et al., 2012, 2011; Etienne-Manneville, 2004; Harris and Tepass, 2010; Johnson, 1999). Cdc42 acts as a binary molecular switch and can respond to and initiate multiple signaling pathways. In most eukaryotes, Cdc42 is regulated by numerous GEFs and GAPs, complicating our understanding of GTPase regulation (Bos et al., 2007). In the fission yeast Schizosaccharomyces pombe, Cdc42 is activated by two GEFs, Gef1 and Scd1 (Chang et al., 1994; Coll et al., 2003). The presence of only two Cdc42 GEFs, and the well-documented process of cell polarization in these cells, make fission yeast an excellent model system to understand the mechanistic details of cell shape establishment. Here, we report that the two Cdc42 GEFs regulate each other during both cytokinesis and polarized growth. This finding provides new insights into the spatiotemporal regulation of Cdc42 during critical cellular events.

Fission yeast cells are rod-shaped and grow in a polarized manner from the two ends. Cells in early G2 phase are monopolar, and grow from the old end, which existed in the previous generation. Cells transition to a bipolar growth pattern in late G2 through the process of new end take-off (NETO), when growth initiates at the new end, which was formed during sister cell separation (Mitchison and Nurse, 1985). Due to this simple growth pattern, fission yeast is an excellent model to understand how a cell regulates polarized growth from multiple sites. In fission yeast, active Cdc42 displays anti-correlated oscillations between the two ends. The regulation that leads to this oscillatory pattern arises from both positive and time-delayed negative feedback, as well as from competition between the two ends, and controls cell shape and polarity (Das et al., 2012).

Cdc42 is activated at the cell ends to promote polarized growth and restricted from the cell sides to maintain cell shape (Das et al., 2012, 2015, 2009). Cdc42 is also involved in cytokinesis. During cytokinesis, Cdc42 activation promotes septum formation, and like in other systems, Cdc42 needs to be subsequently inactivated to promote cell separation (Atkins et al., 2013; Onishi et al., 2013; Wei et al., 2016). The regulatory mechanisms that allow for these spatiotemporal activation patterns are not well understood. To explain Cdc42 activation during polarized growth, it is important to first understand how Cdc42 regulators function. Gef1 and Scd1 are partially redundant but exhibit unique phenotypes when deleted (Chang et al., 1994; Coll et al., 2003), indicating that they may regulate Cdc42 in distinct, but overlapping, manners. Scd1 oscillates between the two cell ends, much like active Cdc42 (Das et al., 2012), and is essential for polarity establishment (Chang et al., 1994). Scd1 is also required for mating (Bendezú and Martin, 2013). In contrast, gef1 mutants are narrower and grow in a monopolar manner (Coll et al., 2003; Das et al., 2012). Given that Gef1 is sparsely localized to the cortex (Das et al., 2015; Tay et al., 2018) and not required for polarity establishment, it is unclear why Gef1 is required for bipolar growth.

Investigations into the behaviors of Gef1 and Scd1 during interphase are complicated since Gef1 is barely detectable at sites of polarized growth, overlapping with Scd1. However, these GEFs also localize to the site of cell division during cytokinesis (Wei et al., 2016). The temporal localization and function of the two GEFs are discernible during cytokinesis as they are recruited to the division site in succession to activate Cdc42. During cytokinesis, Gef1 localizes first to the actomyosin ring to activate Cdc42 and promote ring constriction (Wei et al., 2016). Next, Scd1 localizes to the ingressing membrane and regulates septum formation (Wei et al., 2016). The temporal difference between Gef1 and Scd1 localization at the division site enables investigation of the significance of the GEFs in Cdc42 regulation, which is unclear from studies solely of the growing ends.

Bipolar growth occurs when the new end is able to overcome the dominance of the old end. Here, we show that Gef1 enables the new end to overcome old end dominance and promote bipolar growth. We find that in gef1 mutants both the Cdc42 GEF Scd1 and its scaffold Scd2 are localized mainly to the old ends. Using cytokinesis as a paradigm we identify a novel crosstalk between Gef1 and Scd1. Our data indicate that Gef1 promotes the localization of Scd1 to the division site via Cdc42 activation and the scaffold Scd2. Gef1 activates Cdc42, which then promotes recruitment of Scd2, and consequently of Scd1. We extend these observations to the sites of polarized growth, where we show that Gef1 promotes bipolar Scd1 and Scd2 localization to initiate growth at a second site, while Scd1 prevents ectopic Gef1 localization to the division site and the cell cortex. By this manner of regulation, Cdc42 activation is promoted at the new end of the cell with no prior growth history, but is restricted from random sites. To our knowledge, such crosstalk has not been reported to function between GEFs of the same GTPase. The interplay between the Cdc42 GEFs operates in the same manner during both cytokinesis and polarized growth, suggesting that this is a conserved feature of Cdc42 regulation.

We have reported that the GEF Scd1 localizes to the membrane adjacent to the actomyosin ring after Gef1 to activate Cdc42 along the membrane barrier (Wei et al., 2016). Since Scd1 arrives at the division site soon after Gef1, it is possible that Gef1 promotes Scd1 localization. Alternatively, if Gef1 and Scd1 act independently, then loss of gef1 would not impact Scd1 localization to the division site. To test this, we examined whether Scd1 localization to division site is Gef1-dependent. Both Gef1 and Scd1 are low-abundance proteins and are not suitable for live cell imaging over time. To overcome this limitation, we used the actomyosin ring as a temporal marker. The actomyosin ring undergoes visibly distinct phases during cytokinesis: assembly, maturation, constriction and disassembly. We determined the timing of protein localization to the division site by comparing it to the corresponding phase of the actomyosin ring. We have previously reported that ring constriction is delayed in gef1Δ mutants (Wei et al., 2016). To eliminate any bias in protein localization resulting from this delay, we only analyzed cells in which the rings had initiated constriction. In gef1Δ mutants, the number of constricting rings that recruited Scd1–3×GFP decreased to 15% from 96% in gef1+ cells (Fig. 1A,B, P<0.0001; Fig. S1A–C). Furthermore, the gef1Δ cells that managed to recruit Scd1–3×GFP did not do so as efficiently as gef1+ cells, given the 15% decrease in Scd1–3×GFP fluorescence intensity at the division site (Fig. 1A,C, P=0.0089; Fig. S1D–F). This indicates that Gef1 promotes Scd1 localization to the division site and that the two GEFs are not independent.

Next, we investigated how Gef1 promotes Scd1 recruitment to the division site. GEF recruitment to sites of Cdc42 activity occurs via positive feedback, as reported in budding yeast (Butty et al., 2002; Irazoqui et al., 2003; Kozubowski et al., 2008; Witte et al., 2017). In this model, activation of Cdc42 leads to further recruitment of the scaffold protein BEM1, which then recruits the GEF CDC24 to the site of activity, thus helping to break symmetry and promote polarized growth. A similar positive feedback may also exist in fission yeast (Das et al., 2012; Das and Verde, 2013). Indeed, a recent study suggests that the scaffold Scd2 establishes positive feedback (Lamas et al., 2019 preprint). We hypothesized that Gef1-activated Cdc42 acts as a seed for Scd1 recruitment to the division site. To test this, we asked whether constitutive activation of Cdc42 could rescue the Scd1 recruitment defect exhibited by gef1Δ cells. For this approach to work, the constitutively active Cdc42 must localize to the division site. Localization of active Cdc42 is visualized via the bio-probe CRIB–3×GFP, which specifically binds GTP-Cdc42. Since our previous work reported that Cdc42 activity is reduced at the division site in gef1Δ cells (Wei et al., 2016), we validated this approach by first testing whether constitutively active Cdc42 restores CRIB–3×GFP localization at the division site in gef1Δ cells. The constitutively active allele cdc42G12V and the bio-probe CRIB–3×GFP were expressed in gef1+ and gef1Δ cells. Mild expression of cdc42G12V was sufficient to restore CRIB–3×GFP intensity at the division site to physiological levels in gef1Δ cells (Fig. 1D,E, P<0.0001; Fig. S1G–I). Likewise, expression of cdc42G12V restored Scd1–tdTomato localization to the division site in cdc42G12V gef1Δ cells (Fig. 1F,G; Fig. S1J–L). This demonstrates that active Cdc42 can promote Scd1 recruitment. Next, we asked how active Cdc42 promotes Scd1 localization. We examined downstream targets of active Cdc42 for this purpose. The Cdc42 ternary complex consists of the GEF Scd1, the scaffold protein Scd2, and the downstream effector kinase Pak1 (also known as Shk1) (Endo et al., 2003). Observations in budding yeast suggest that the PAK kinase Cla4 mediates GEF recruitment (Kozubowski et al., 2008). Contrary to this hypothesis, we find that Scd1–3×GFP intensity increases in the nmt1 switch-off mutant allele of pak1, compared to pak1+ cells (Fig. S2H). These findings support similar observations reported in the hypomorphic temperature-sensitive pak1 allele, orb2-34 (Das et al., 2012), and indicate that pak1 does not facilitate Scd1 recruitment to the site of action.

The scaffold protein Scd2 is a component of the Cdc42 ternary complex that binds active Cdc42 (Endo et al., 2003; Wheatley and Rittinger, 2005). We hypothesized that Gef1-dependent active Cdc42 recruits Scd1 to the division site through the scaffold Scd2. Indeed, we find that gef1Δ cells displayed a significant decrease in Scd2–GFP-containing assembled rings compared to gef1+ cells. In gef1Δ mutants, the number of rings that recruited Scd2–GFP prior to ring constriction decreased to 8% compared to 88% in gef1+ cells, indicating a delay in Scd2 recruitment (Fig. 2A,B, P>0.0001; Fig. S2A–C). Although gef1Δ cells were able to recruit Scd2 to the division site once ring constriction began, the fluorescence intensity of Scd2–GFP at the division site was reduced by 61% compared to gef1+ cells (Fig. 2A,C, P>0.0001; Fig. S2D, S2E–G). Gef1 thus promotes Scd2 localization to the division site. Since previous work indicates that Scd1 and Scd2 require each other for their localization, it is possible that a decrease in Scd2 at the division site observed in gef1 mutants is due to a decrease in Scd1 at this site (Kelly and Nurse, 2011). However, contrary to previous findings, we observed that Scd2–GFP localization at the division site is not impaired in scd1Δ cells (Fig. 2D) (Kelly and Nurse, 2011). In contrast, Scd1–3×GFP localization is completely abolished at the division site in scd2Δ cells (Fig. 2E). Scd2 localization is independent of Scd1 and depends on Gef1 instead.

Taken together, our data reveal that Gef1 promotes Scd2 localization to the division site, which then recruits Scd1. Based on these data, we hypothesized that Gef1, Scd2 and Scd1 sequentially localize to the division site. To test this, we examined the temporal localization of these proteins to the division site. Since these proteins do not lend themselves to extended time lapse imaging, we used the spindle pole bodies as an internal timer. The distance between spindle pole bodies is a well-established temporal marker to determine a cell's cytokinetic stage. The spindle pole body distance increases as mitosis progresses until the cell reaches anaphase B (Nabeshima et al., 1998), at which time the actomyosin ring starts to constrict (Wu et al., 2003). We report the spindle pole body distance at which Gef1–mNG (monomeric NeonGreen), Scd1–3×GFP and Scd2–GFP signals are visible at the division site before the actomyosin rings start constriction (Fig. 3A). Gef1–mNG and Scd2–GFP appear around the same time during mitosis with a mean spindle pole body distance of 5.9 µm, and 6.6 µm, respectively (Fig. 3B, not significant). Scd1–3×GFP arrives later with a longer mean spindle pole body distance of 7.6 µm (Fig. 3B, P=0.005). Furthermore, in cells expressing both Gef1–tdTomato and Scd2–GFP, 100% of assembled rings with Gef1 invariably also had Scd2 at the division site (Fig. 3C,D). In cells expressing Scd2–GFP and Scd1–tdTomato, only 59% of assembling rings localized both Scd2 and Scd1 (Fig. 3C,D). In cells expressing both Gef1–tdTomato and Scd1–3×GFP, 61% of assembling rings with Gef1 also had Scd1 (Fig. 3C,D). Taken together, this demonstrates that Gef1 recruits Scd2 to actomyosin rings prior to recruitment of Scd1.

To determine whether the Gef1-mediated regulation of Scd1 and Scd2 localization was conserved during cell polarity, we analyzed whether their localization at the cell poles was altered in gef1Δ mutants. Scd1 and Scd2, like active Cdc42, undergo oscillations between the two competing ends (Das et al., 2012); thus, a bipolar cell does not always display bipolar Scd1 or Scd2 localization. Fewer new ends in gef1Δ cells exhibited Scd1–3×GFP; bipolar Scd1–3×GFP was observed in 30% of interphase gef1+ cells, but only in 14% of gef1Δ cells (Fig. 4A,B; P=0.0004; Fig. S3A–C). Similarly, 70% of gef1+ cells displayed bipolar Scd2–GFP localization, but this was reduced to 30% in gef1Δ cells (Fig. 4A,B, P<0.0001; Fig. S3A–C). While this suggests that Gef1 promotes Scd1 and Scd2 localization to the new end to enable bipolar growth, this interpretation suffers from some complications. The old and new ends compete with each other for active Cdc42 (Das et al., 2012). It is possible that in gef1Δ mutants the new end fails to grow since the old end traps Scd1 and Scd2, resulting in monopolar distribution of these proteins. However, we did not find enhanced Scd1–3×GFP levels at the old end in gef1Δ cells compared to gef1+ cells (Fig. S3D). This suggests that Gef1 does not prevent accumulation of Scd1 and Scd2 at the old ends.

Next, we asked whether Gef1 recruits Scd1 to the new end via Cdc42 activation, just as it does at the division site. We first tested whether expression of constitutively active Cdc42 results in bipolar localization of active Cdc42 in gef1Δ cells, as indicated by CRIB–3×GFP localization. Low-level expression of cdc42G12V was sufficient to restore bipolar CRIB–3×GFP localization in gef1Δ cells, compared to the empty-vector-containing gef1Δ mutants (Fig. 5A,B, P<0.0001; Fig. S4A–C). We observed bipolar CRIB–3×GFP in 35% of gef1+ cells transformed with the empty vector pJK148 and in 36% of cells expressing cdc42G12V. In gef1Δ pJK148 mutants 8% of cells displayed bipolar CRIB–3×GFP. In contrast, in gef1Δ cdc42G12V cells, bipolar CRIB–3×GFP was restored to 34%. In the same cells, we observed bipolar Scd1–tdTomato in 21% of gef1+ pJK148 cells, and in 23% of cells expressing cdc42G12V. In gef1Δ pJK148 mutants, we observed bipolar Scd1–tdTomato in only 6% of cells. Furthermore, in gef1Δ cdc42G12V cells, bipolar Scd1–tdTomato was restored to 19% (Fig. 5A,C; Fig. S4D–F). Thus, expression of cdc42G12V was sufficient to restore bipolar Scd1–tdTomato localization to the cell ends in gef1Δ mutants, just as it was at the division site (Fig. 5A,C). This demonstrates that Gef1 promotes Scd1 localization to the new ends through Cdc42 activation.

Next, we asked whether bipolar localization of Scd1–tdTomato correlated with bipolar growth. To determine this, we stained cells with Calcofluor to mark the sites of polarized growth. Bipolar growth was exhibited in 40% of gef1+ pJK148 cells and 14% of gef1Δ pJK148 mutants (P<0.0001, Fig. S5). Expression of cdc42G12V in gef1+ and gef1Δ cells elevated bipolarity to 49% and 48%, respectively (P<0.0001, Fig. S5). To address whether cells grew at poles that contained high levels of Scd1–tdTomato, we examined the growth pattern of cells exhibiting bipolar or monopolar Scd1–tdTomato localization through time-lapse microscopy. Regardless of genotype, all cells with bipolar Scd1–tdTomato localization grew in a bipolar manner (Fig. S6). Approximately 20% of gef1+ pJK148, gef1+ cdc42G12V and gef1Δ cdc42G12V cells, and 9% of gef1Δ pJK148 mutants with monopolar Scd1–tdTomato grew in a bipolar manner (Fig. S6). That a fraction of cells grow in a bipolar manner even though Scd1–tdTomato appears monopolar could be due to oscillation of Scd1 between the two ends in bipolar cells (Das et al., 2012). Since Scd1–tdTomato fluorescence was only captured at time 0, it is likely that in these cells the signal was detected during the transition of Scd1–tdTomato from one end to the other. While Scd1 requires Scd2 for its localization, Scd2 localization is independent of Scd1 (Fig. S4G). Scd2 likely binds active Cdc42, so it is possible that Gef1-mediated Cdc42 activation leads to Scd2 recruitment. To test this, we monitored ectopic Gef1 localization. In cells treated with the actin-depolymerizing drug latrunculin A (LatA), Gef1–tdTomato ectopically localizes to the cell sides (Fig. 5D). In these cells, Scd2–GFP colocalizes with ectopic Gef1–tdTomato at the cell sides. Moreover, in gef1Δ cells Scd2–GFP fails to localize to the cell sides upon LatA treatment (Fig. 5E). In these cells, Scd2–GFP remains at the cell ends, but the levels are much reduced. This provides further evidence that Gef1 marks the site for Scd2 recruitment at the cell cortex.

Fission yeast transitions from monopolar to bipolar growth upon reaching a certain size (Mitchison and Nurse, 1985). It has been proposed that this size requirement is necessary to establish two stable regions of Cdc42 activity (Das et al., 2012). Because protein abundance scales with cell size, growth may allow for the accumulation of sufficient GEFs and other polarity factors to maintain two sites of Cdc42 activity (Das et al., 2012). These models suggest that in gef1Δ cells the total active Cdc42 levels are insufficient to allow bipolar growth, thus resulting in monopolarity. However, the dependence of bipolarity on cell size and protein abundance cannot be explained in G1-arrested cdc10-129 mutants (Marks et al., 1986). cdc10-129 cells shifted to 36°C remain monopolar even as they grow longer compared to cdc10+ cells (Fig. 6A).

To gain insight into the transition to bipolar growth we examined the growth patterns of gef1Δ mutants. The old ends initiate growth immediately after completion of division, and upon reaching a certain size the new end initiates growth, resulting in bipolarity (Fig. 6B) (Mitchison and Nurse, 1985). The two ends compete for active Cdc42; at first the old end, which is dominant, wins this competition (Das et al., 2012). A growth pattern in which one daughter cell is monopolar and the other daughter cell is prematurely bipolar is exhibited in 68% of monopolar gef1Δ mutant cells (Fig. 6C; Fig. S6A). In monopolar gef1Δ cells, growth occurs at the old end, which grew in the previous generation (Fig. 6C; Fig. S7A). The new end frequently fails to grow since it cannot overcome the old end's dominance. The daughter cell that inherits its parent cell's non-growing end typically displays precocious bipolar growth, indicating that these cells do not contain a dominant end. This suggests that a cell end is dominant if it grew in the previous generation. These results indicate that the new ends of gef1Δ cells are not well-equipped to overcome old end dominance. To further confirm this, we tracked the fate of cell ends with a known history of growth in gef1+ and gef1Δ cells. In gef1+ cells, 97% of daughter cells derived from a growing end display a normal growth pattern (Fig. 6D). In gef1Δ cells, 81% of daughter cells derived from a growing end failed to initiate growth at their new end and were thus monopolar (Fig. 6D). These data indicate that Gef1 enables the new end to overcome old end dominance to promote bipolar growth. Furthermore, our data indicate that Gef1 promotes bipolarity by promoting Scd1 localization to the new end. As a further test, we asked whether Scd1 localization would be enhanced at the new end in the precociously bipolar gef1S112A mutant, in which Gef1S112A constitutively localizes to the cortex. Indeed, 52% of gef1S112A cells show bipolar Scd1–3×GFP localization, while this is seen in only 33% of gef1+ cells (Fig. S7B,C, P<0.0001).

Cells treated with LatA show ectopic Cdc42 activation at the cell sides (Mutavchiev et al., 2016). As reported earlier (Kelly and Nurse, 2011), we find that LatA treatment leads to a severe loss in Scd1–mNG levels at the cell ends and at the site of cell division (Fig. S8G). We asked whether ectopic Gef1 localization in LatA-treated cells results from loss of Scd1 from the cell cortex. To test this, we analyzed Gef1–mNG localization in scd1Δ mutants. In scd1+ cells, Gef1–mNG is mostly cytoplasmic and displays sparse but polarized localization at cell ends (Fig. 7Ai). In scd1Δ mutants, Gef1–mNG shows depolarized cortical localization with random patches at the cortex like that in scd1+ cells treated with LatA (Fig. 7Aii,v). Gef1–mNG localization in scd1Δ cells treated with LatA is similar to that in untreated scd1Δ cells (Fig. 7Aii,vi). This suggests that Scd1 is required to prevent ectopic Gef1 localization and restricts it to the cell ends.

Next, we asked whether ectopic Gef1 in scd1Δ cells and in LatA-treated cells leads to ectopic Cdc42 activation. We find that active Cdc42 appears depolarized in scd1Δ mutants during interphase. While CRIB–3×GFP remains restricted to the ends in scd1+ cells, in scd1Δ mutants it appears localized to random patches at the cortex (Fig. 7Bi,ii). LatA treatment leads to ectopic CRIB localization (Mutavchiev et al., 2016) and we observed that ectopic CRIB-3×GFP is abolished in gef1Δ cells (Fig. 7Bvii,ix). In gef1Δ cells treated with LatA, CRIB–3×GFP remains at the cell ends, but at lower levels compared to untreated cells. This indicates that Scd1 is functionally epistatic to actin in the removal of Gef1 from the cell cortex. Taken together, these data demonstrate that Scd1 at the cell ends prevents ectopic Gef1 localization and Cdc42 activation. This observation is supported by previous work, in which deletion of gef1 restores polarized growth in cells expressing only minimal levels of scd1 (Tay et al., 2018).

Because we find that Scd1 regulates Gef1 localization at sites of polarized growth, we asked whether a similar relationship also occurs at the division site. To test this, we analyzed Gef1 localization in scd1Δ mutants. We observed persistent Gef1 localization in scd1 mutants after ring constriction. In scd1Δ mutants, after completion of ring constriction and disassembly, Gef1 remains at the membrane that was adjacent to the ring (Fig. 7C). Among scd1Δ cells that had completed constriction, 70% show persistent Gef1–mNG at the newly formed membrane barrier, as confirmed by the absence of Rlc1–tdTomato (Fig. 7C,D; Fig. S8A–C). Similar Gef1–mNG localization was observed in only 20% of scd1+ cells (Fig. 7C,D, P<0.0001; Fig. S8A–C).

We find that Gef1–mNG removal from the division site is also dependent on actin. We analyzed Gef1 localization in cells with an actin cytoskeleton disrupted via LatA treatment. In LatA-treated cells that were fully septated following completion of ring constriction, we observed persistent Gef1 localization at the division site. Gef1–mNG persists on both sides of the septum barrier in 40% of cells treated with LatA, but not in mock DMSO-treated cells (Fig. 7E,F; Fig. S8D–F). This demonstrates that the actin cytoskeleton promotes Scd1 localization and that Scd1 promotes Gef1 removal from the division site. To confirm that actin promotes Gef1 removal via Scd1, we looked for a functional epistatic relationship between Scd1 and actin. We treated scd1+ and scd1Δ cells expressing Gef1–mNG with LatA or DMSO. In cells treated with DMSO, Gef1–mNG persists in 20% of septated scd1+ cells and in 63% of septated scd1Δ cells. In cells treated with LatA, Gef1–mNG persists in 40% of septated scd1+ cells and in 61% of septated scd1Δ cells (Fig. 7F; Fig. S8D–F). The extent of Gef1 persistence in scd1Δ cells does not increase with the addition of LatA. This indicates that Scd1 is functionally epistatic to actin in the process of Gef1 removal (Fig. 7F; Fig. S8D–F). Taken together, these data suggest that Scd1 removes Gef1 from the division site after ring disassembly.

To determine the significance of Scd1-mediated removal of Gef1 from the division site, we looked at the constitutively localized gef1S112A mutant. Gef1S112A–3×YFP but not Gef1–3×YFP persists at the division site after ring constriction, indicated by the absence of Cdc15–tdTomato-marked actomyosin ring (Fig. S8H). Live cell imaging revealed that cell separation is delayed in gef1S112A mutants. Cell separation occurred 28 min after ring constriction in gef1+ cells, but at 34 min in gef1S112A mutants (Fig. S8I, P=0.009). We have previously shown that prolonged Cdc42 activation at the division site impairs cell separation, as seen in other model systems (Atkins et al., 2013; Onishi et al., 2013; Wei et al., 2016). This would necessitate the removal of the GEFs from the division site. Our data indicate that Scd1-mediated Gef1 removal promotes cell separation.

While studies in yeast continue to pioneer insights into the regulation of cell polarity, one aspect of S. pombe polarity remains elusive: how cells initiate growth at a second site during NETO (new end take-off). We find that the Cdc42 GEF Gef1 enables the new end to overcome old end growth dominance to initiate bipolar growth. Our data indicate that Gef1 promotes the localization of the other GEF Scd1 to the new end. Uncovering the mechanism through which this occurs is not trivial, since the function of Gef1 at sites of polarized growth has proven elusive due its sparse and transient localization at these regions. Therefore, we examined the relationship between these two GEFs at the division site, where the localization of both these GEFs is easily monitored. We have previously shown that Gef1 and Scd1 localize sequentially to the division site to activate Cdc42 during cytokinesis (Wei et al., 2016). Here, we take advantage of the temporal difference between Gef1 and Scd1 localization at the division site to determine the significance of these two GEFs in Cdc42 regulation. We uncover a novel interplay between the Cdc42 GEFs that functions in both cytokinesis and polarized cell growth (Fig. 8A). Given the conserved nature of Cdc42 and its regulators, we posit that this interplay between the GEFs is a common feature of Cdc42 regulation.

Analysis of Gef1-dependent Scd1 recruitment at the new end is complicated by the dynamic nature of their localization patterns, and the presence of two competing cell ends. Gef1 and Scd1 localize to the division site in a sequential manner. Moreover, during cytokinesis, the division site exclusively activates Cdc42 and thus does not compete with any other site in the cell. Therefore, we investigated Gef1 and Scd1 recruitment at the division site. We find that Scd1 localization is reduced at the division site in gef1Δ cells. Our observation that expression of constitutively active cdc42G12V can restore Scd1 localization in gef1Δ cells suggests that active Cdc42 promotes Scd2 recruitment, either through direct interaction or through an alternate mechanism. A recent study indicates that Scd2 optogenetically responds to active Cdc42, and is required for Scd1-mediated positive feedback (Lamas et al., 2019, preprint). Furthermore, Scd2 interacts with GTP-Cdc42 in vitro (Endo et al., 2003; Wheatley and Rittinger, 2005). Similarly, the budding yeast scaffold BEM1 interacts with GTP-Cdc42 in vivo (Yamaguchi et al., 2007). Taken together, this suggests that Scd2 likely binds active Cdc42. Thus, we posit that Gef1 localizes to the division site first, which enables recruitment of Scd1 (Fig. 8B).

After division, the old end always initiates growth first. The two cell ends compete for active Cdc42, and initially the new end is incapable of overcoming dominance at the old end (Das et al., 2012). Gef1 promotes bipolar growth and cells lacking gef1 are mostly monopolar (Coll et al., 2003; Das et al., 2012). In gef1Δ cells, new ends frequently fail to overcome old end dominance, resulting in monopolar growth. Bipolar growth in gef1Δ mutants is typically observed in cells that do not contain a dominant old end. We find that Cdc42 regulation at the cell ends resembles that at the division site. While Scd1 does not require Gef1 for its localization to the old end, our data suggest that Gef1 facilitates bipolar growth by promoting Scd2 localization at the new end, which is needed for Scd1 localization at this site. Polarized growth requires Cdc42 activation, mediated through positive feedback (Das and Verde, 2013; Wu and Lew, 2013). Scd1 is the positive-feedback-mediating GEF. Scd1 and Scd2 together promote a positive feedback pathway for Cdc42 activation. For such a pathway to function, Cdc42 first needs to be activated for the recruitment of Scd2, and consequently Scd1. Our data indicate that Gef1-mediated Cdc42 activation provides the seed for Scd1–Scd2 recruitment. This enables Scd2 recruitment to nascent sites that have no prior history of Cdc42 activation and enables Scd1 recruitment to those sites. Thus, Gef1 activates Cdc42 to establish a Scd1-dependent positive feedback pathway at the new end to overcome old end dominance and establish bipolar growth (Fig. 8C). A recent report indicates that Gef1 localization to the division site and to the sites of polarized cell growth depends on the F-BAR protein Cdc15 (Hercyk and Das, 2019). This provides further evidence that while the division site and cell ends are functionally distinct, some aspects of Cdc42 regulation are conserved at these sites.

Scd1 is the primary GEF that promotes polarized growth (Chang et al., 1994). We find that cells lacking scd1 are depolarized due to ectopic Cdc42 activation, as a result of mislocalized Gef1. In the presence of Scd1, Gef1 shows sparse cortical localization and is restricted to the cell ends. Cell ends that do not localize Scd1 either due to LatA treatment or scd1 knockdown display ectopic and enhanced cortical localization of Gef1. Thus, we posit that Scd1 prevents ectopic Gef1 localization. A recent report shows that ectopic Cdc42 activation in LatA-treated cells depends on the stress-activated MAP kinase Sty1 (Mutavchiev et al., 2016). Cells treated with LatA did not display ectopic Cdc42 activation in the absence of sty1. It is possible that in the absence of actin, reduced Scd1 at the cell cortex elicits a stress response, leading to Sty1 activation that results in the mislocalization of Gef1. Further analysis is necessary to test this hypothesis.

Normally, Gef1 localization to the division site is lost after ring constriction (Wei et al., 2016). Similar to our observation at the cell cortex, we find that Gef1 persists at the division site in cells lacking Scd1 or in cells treated with LatA. Here, we show that Scd1 promotes the clearance of Gef1 from the division site after ring disassembly (Fig. 8A). Failure to inactivate Cdc42 results in cytokinesis failure in budding yeast and HeLa cells, and prevents cellularization in Drosophila embryos (Atkins et al., 2013; Crawford et al., 1998; Dutartre et al., 1996; Onishi et al., 2013). Our data suggest that Scd1 ensures Gef1 removal from the division site in the final stages of cytokinesis, preventing inappropriate Cdc42 activation. Taken together, our data demonstrate an elegant regulatory pattern in which Gef1-mediated Scd1 recruitment to the division site promotes septum formation, and Scd1-mediated Gef1 removal promotes cell separation.

Polarized cell growth requires symmetry breaking, and several models have indicated a need for Cdc42 positive feedback loops in this process (Bendezú et al., 2015; Irazoqui et al., 2003; Kozubowski et al., 2008; Slaughter et al., 2009a,b; Wedlich-Soldner et al., 2004). Elegant experiments in budding yeast demonstrate that local activation of Cdc42 establishes positive feedback through the recruitment of additional GEFs to amplify the conversion of GDP-Cdc42 to GTP-Cdc42 (Butty et al., 2002; Kozubowski et al., 2008). A caveat of positive feedback is that the site that first activates Cdc42 can act as a sink that traps the GEFs, thereby preventing Cdc42 activation at other sites. Such a trap can be undone via negative feedback regulation of Cdc42 that results in an oscillatory pattern at the cell ends (Das et al., 2012; Howell et al., 2012). Negative feedback in S. pombe and S. cerevisiae acts through Pak1 kinase activity that antagonizes either the Cdc42 scaffold or the GEF (Das et al., 2012; Gulli et al., 2000; Kuo et al., 2014; Rapali et al., 2017). Indeed, pak1 mutants fail to activate bipolar growth (Das et al., 2012; Verde et al., 1998). Although fission yeast contains another Pak homolog, Shk2, only Pak1 (Shk1) is essential. Furthermore, Shk2 functions primarily during G0 and meiosis, and its loss does not result in the same polarity defects as loss of Pak1 (Vještica et al., 2018; Yang et al., 1998). Importantly, Shk2 binds neither Scd1 nor Scd2 (Sells et al., 1998; Yang et al., 1998). Therefore, it is unlikely that Shk2 mediates Scd1 localization and further investigations will test this. Our data show that besides the kinase Pak1, Gef1 also contributes to initiation of bipolar growth.

The Cdc42 oscillatory pattern can be explained by the presence of positive feedback, time-delayed negative feedback, and competition between the two ends for active Cdc42. We posit that Scd1 activates Cdc42 through positive feedback at the dominant old end. Dominance at the old end ensures that Scd1 localization is mainly restricted to this end at the expense of the new end. A previous model suggests that as the cell reaches a certain size, the GEFs reach a threshold level that allows the new end to overcome old end dominance to initiate growth and promote bipolarity (Das et al., 2012). Threshold GEF levels alone cannot explain our findings since gef1S112A cells display bipolar growth at a smaller cell size (Das et al., 2015), while G1-arrested cdc10-129 mutants grow to longer cell lengths but remain monopolar. Our data suggest that the regulatory crosstalk between the Cdc42 GEFs may provide an advantage to the cell and enable the new end to overcome old end dominance. Gef1 triggers a positive feedback at the division site via a feed-forward pathway (Fig. 8A,C). Given that Gef1 promotes Scd1-mediated polarized growth at the new end, it is conceivable that Gef1 itself is tightly regulated to prevent random Cdc42 activation. Indeed, Gef1 shows sparse localization to the cell ends and is mainly cytoplasmic (Das et al., 2015). The NDR kinase Orb6 prevents ectopic Gef1 localization via the 14-3-3 protein Rad24-mediated sequestration to the cytoplasm (Das et al., 2015, 2009). Here, we show that while Gef1 promotes Scd1 recruitment to a nascent site, Scd1 itself restricts Gef1 localization to the cell ends to precisely activate Cdc42 (Fig. 8C). Taken together, our findings describe an elegant system in which the two Cdc42 GEFs regulate each other to ensure proper cell polarization.

In budding yeast, CDC24 is required for polarization during bud emergence and is essential for viability (Sloat et al., 1981; Sloat and Pringle, 1978), unlike Scd1 in fission yeast. Budding yeast also has a second GEF, BUD3, which establishes a proper bud site (Kang et al., 2014). During G1 in budding yeast, bud emergence occurs via biphasic CDC42 activation by the two GEFs: BUD3 helps select the bud site (Kang et al., 2014), and CDC24 allows polarization (Sloat et al., 1981; Sloat and Pringle, 1978). This is analogous to new end growth in fission yeast, which requires Gef1-dependent recruitment of Scd1 for robust Cdc42 activation. It would be interesting to test whether crosstalk also exists between BUD3 and CDC24 in budding yeast.

The Rho family of GTPases includes Rho, Rac and Cdc42. In certain mammalian cells, Cdc42 and Rac1 appear to activate cell growth in a biphasic manner (de Beco et al., 2018; Yang et al., 2016). For example, during motility, the GTPases, Rho, Rac and Cdc42, regulate the actin cytoskeleton (Heasman and Ridley, 2008; Machacek et al., 2009). During cell migration, these GTPases form bands or ‘zones’ in the leading and trailing regions of the cell (Ridley, 2015). Their spatial separation is mediated by the organization of their GEFs and GAPs, and by regulatory signaling between these GTPases (Guilluy et al., 2011). Cdc42 and Rho are mutually antagonistic, explaining how such zones of GTPase activity can be established and maintained (Guilluy et al., 2011; Kutys and Yamada, 2014; Warner and Longmore, 2009). Similarly, Cdc42 can refine Rac activity (Guilluy et al., 2011). Recent experiments demonstrate that during cell migration, reorganization of the actin cytoskeleton occurs in a biphasic manner, in which Cdc42 activation at new sites sets the direction, while robust Rac activation determines the speed (de Beco et al., 2018; Yang et al., 2016). Unlike most eukaryotes, the genome of S. pombe does not contain a Rac GTPase. However, the GEFs differentially activate Cdc42, such that Gef1 sets the direction of growth by establishing a new site of polarization, while Scd1 promotes efficient growth through robust Cdc42 activation. In conclusion, we propose that the crosstalk between the Cdc42 GEFs themselves is an intrinsic property of small GTPases and is necessary for fine-tuning their activity.

The S. pombe strains used in this study are listed in Table S1. All strains are isogenic to the original strain PN567. Cells were cultured in yeast extract (YE) medium and grown exponentially at 25°C, unless specified otherwise. Standard techniques were used for genetic manipulation and analysis (Moreno et al., 1991). Cells were grown exponentially for at least three rounds of eight generations each before imaging.

Cells were imaged at room temperature (23–25°C) with an Olympus IX83 microscope equipped with a VTHawk two-dimensional array laser scanning confocal microscopy system (Visitech International, Sunderland, UK), electron-multiplying charge-coupled device digital camera (Hamamatsu, Hamamatsu City, Japan), and 100×/numerical aperture 1.49 UAPO lens (Olympus, Tokyo, Japan). Images were acquired with MetaMorph (Molecular Devices, Sunnyvale, CA) and analyzed using ImageJ (National Institutes of Health, Bethesda, MD).

The growth pattern of gef1+ and gef1Δ cells was observed by live imaging of cells through multiple generations. Cells were placed in 3.5-mm glass-bottom culture dishes (MatTek, Ashland, MA) and overlaid with YE medium plus 1% agar, and 100 μM ascorbic acid to minimize photo-toxicity to the cell. A bright-field image was acquired every minute for 12 h. Birth scars were used to distinguish between, and to measure, old end and new end growth.

The forward primer 5′-GGATCCGTGTTTACCAAAGTTATGTAAGAC-3′ with a 5′ BamHI site and the reverse primer 5′-CCCGGGAACCCTCGCAGCTAAAGA-3′ with a 5′ XmaI site were used to amplify a 3 kb DNA fragment containing gef1, the 5′ UTR, and the endogenous promoter. The fragment was then digested with SalI and XmaI and ligated into the SalI-XmaI site of pKS392 pFA6-tdTomato-kanMX and pKG6507 pFA6-mNeonGreen-kanMX. Constructs were linearized by digestion with XbaI and transformed into the gef1 locus in wild-type cells.

pjk148-nmt41x-leu1⁺ or pjk148-nmt41x:cdc42G12V-leu1⁺ were linearized with NdeI and integrated into the leu1-32 locus in gef1+ and gef1Δ cells expressing either CRIB–3×GFP and Scd1–tdTomato, or Scd1–tdTomato and Rlc1–GFP. The empty vector pjk148-nmt41x-leu1⁺ was used as control. Cells were grown in EMM with 0.05 µM thiamine to promote minimal expression of cdc42G12V.

To visualize areas of cell growth, cells were stained in YE liquid with 50 μg/ml Calcofluor White M2R (Sigma-Aldrich, St Louis, MO) at room temperature.

Cells in YE were incubated at room temperature with 10 μM or 100 μM latrunculin A (Millipore-Sigma) dissolved in dimethyl sulfoxide (DMSO) for 40 min prior to imaging. Control cells were treated with 1% DMSO and incubated for 40 min.

Mutants expressing fluorescent proteins were grown to OD 0.5 and imaged on slides. Cells in slides were imaged for no more than 3 min to prevent any stress response as previously described (Das et al., 2015). Depending on the mutant and the fluorophore, 16–28 z-planes were collected at a z-interval of 0.4 µm for either or both the 488 nm and 561 nm channels. The respective controls were grown and imaged in an identical manner. ImageJ was used to generate sum projections from the z-series, and to measure the fluorescence intensity of a selected region (actomyosin ring, or growth cap at cell tip). The background fluorescence in a cell-free region of the image was subtracted to generate the normalized intensity. Mean normalized intensity was calculated for each image from all (n>5) measurable cells within each field. A Student's two-tailed t-test, assuming unequal variance, was used to determine significance through comparison of each strain's mean normalized intensities.

GraphPad Prism was used to determine significance. One-way ANOVA, followed by a Tukey's multiple comparisons post-hoc test, was used to determine individual P-values when comparing three or more samples. When comparing two samples, a Student's t-test (two-tailed, unequal variance) was used to determine significance.

We thank Joshua Bembenek and Tessa Burch-Smith for critical review of our manuscript; Kathy Gould (Vanderbilt University, Nashville, USA) for supplying plasmids; and Mohan Balasubramanian (University of Warwick, Coventry, UK) and Sophie Martin (University of Lausanne, Switzerland) for providing strains.