The mechanisms of the coordinated assembly and disassembly of the septin/myosin ring is central for the understanding of polar growth and cytokinesis in yeast and other organisms. The septin- and myosin-binding protein Bni5p provides a dual function during the formation and disassembly of septin/myosin rings. Early in the cell cycle, Bni5p captures Myo1p at the incipient bud site and actively transforms it into higher-order structures. Additionally, Bni5p stabilizes the septin/myosin ring and is released from the septins shortly before the onset of cytokinesis. If this Bni5p dissociation from the septins is artificially prevented, ring disassembly is impaired and the untimely appearance of septin/myosin ring is induced. The prematurely formed septin/myosin rings delay the establishment of a new polarity axis and the progression into a new cell cycle. This observation suggests a negative feedback between septin/myosin ring formation and polarity establishment that might help to guarantee the singular assembly of this structure and the synchronization of its formation with the cell cycle.
Septin filaments provide a distinct compartment and scaffold for the contractile actin-myosin ring and other proteins involved in cell division (Dobbelaere and Barral, 2004; McMurray et al., 2011; McMurray and Thorner, 2009). The five mitotic septins of the budding yeast Cdc3p, Cdc10p, Cdc11p, Cdc12p and Shs1p form an octameric septin rod (Pringle, 2008). Septin rods assemble into non-polar filaments and other higher-order structures that appear as rings or collars at the bud neck of the cells (Bertin et al., 2008; Bertin et al., 2012; Byers and Goetsch, 1976; Haarer and Pringle, 1987). Septins mark the site of cell separation at the end of each cell cycle. Shortly before cytokinesis the septin structure at the bud neck splits into two rings. Contraction of the actin myosin ring, fusion of the membranes and synthesis of the extra-cellular septa occurs within this septin-defined compartment (Dobbelaere and Barral, 2004). After completion of cell separation, the septin rings are disassembled and reappear in G1 of the new cell cycle as a patch at the membrane where the new bud is formed (McMurray and Thorner, 2009). New bud formation requires the localization of a complex of polarity factors consisting of the GDP/GTP exchange factor Cdc24p, the scaffold protein Bem1p and the Cdc42p-stimulated kinase Cla4p. This complex locally activates the Rho-type GTPase Cdc42p (Bi and Park, 2012). Activated Cdc42p recruits through its effectors Gic1p/Gic2p the septins very probably as octameric rods to the incipient bud site (Gladfelter et al., 2002; Iwase et al., 2006). Further restructuring of the septins to ring- and collar-like structures is initiated and controlled by Cdc42p, multiple kinases, and by further proteins of still unknown functions (Bi and Park, 2012; Gladfelter et al., 2002; Iwase et al., 2006; Kadota et al., 2004). The establishment of polarity and the formation of the septin ring are tightly coupled and possibly interdependent processes in yeast. The observed robustness in forming only one bud and one septin structure during the cell cycle is attributed to positive- and negative-feedback mechanisms whose exact nature and modes of operation are actively investigated (Howell et al., 2012; Slaughter et al., 2009).
The type II myosin Myo1p and actin are successively recruited via either direct or indirect interactions with the septins to construct the contractile actin myosin ring at the bud neck (Lippincott and Li, 1998). By its binding to Myo1p and the septin subunit Cdc11p, Bni5p was proposed to anchor Myo1p early to the bud neck through a linear septin-Bni5p-Myo1 assembly (Lee et al., 2002; Fang et al., 2010). One region interacting with Bni5p was mapped between residues 991 and 1180 of Myo1p (Fang et al., 2010). Myo1p harbors a second targeting region C-terminal to residue 1180 that is bound by the yeast IQGAP protein Iqg1p (Fang et al., 2010). Iqg1p appears around the onset of anaphase at the site of cytokinesis. Together with the yeast formin Bni1p and the myosin light chain Mlc1p, Iqg1p is required for actin ring assembly and contraction (Tolliday et al., 2002; Vallen et al., 2000; Yoshida et al., 2009). In addition, Iqg1p keeps Myo1p at the site of cell separation after the splitting of the septin ring and the disappearance of Bni5p (Fang et al., 2010; Lister et al., 2006).
In higher eukaryotes, the capture of Myosin II at the site of future cytokinesis requires its assembly into thick bipolar filaments (Uehara et al., 2010). Here we provide evidence that Bni5p not only captures Myo1p, but also induces its assembly into higher-order structures at the bud neck of yeast cells. Our experiments with prematurely formed septin/myosin rings let us further propose the existence of a negative feedback between septin/myosin structure formation and polarity establishment.
A two-step mechanism targets Myo1p to the bud neck
By comparing the localization of Myo1-GFP in cells expressing or lacking Bni5p we first confirmed the observation of Fang et al. that the early recruitment of Myo1p to the bud neck strictly depends on Bni5p (Fig. 1A) (Fang et al., 2010). We could additionally demonstrate that the presence of Bni5p at the bud neck is independent of Myo1p (Fig. 1A). Both observations are compatible with the role of Bni5p as a passive adaptor between septins and Myo1p. This linear targeting model would predict that Myo1p appears at the site of septin assembly only after the septin–Bni5p complex has formed (Fang et al., 2010). However, time-lapse analyses of cells co-expressing GFP- and Cherry-tagged fusion proteins demonstrated that the septins and Myo1p appeared simultaneously at the incipient bud while Bni5p could be detected there on average only 10 minutes later (Fig. 1B). In cells lacking Bni5p, Myo1p was still initially concentrated at the incipient bud site (Fig. 1C). It then passed the septin ring, and transiently accumulated in the newly formed bud (Fig. 1C). The quantification of the signal intensities of the fluorophore-labeled fusions revealed for the same proteins at the end of the cell cycle an inversed order of disappearance at the bud neck (Fang et al., 2010). Bni5p vanished ∼10 minutes before Myo1p started to disappear and 12 minutes before the splitting of the septin ring (Fig. 1D).
Quantification of Myo1-GFP and Bni5-GFP by measuring the GFP-fluorescence at the neck of small and medium sized buds demonstrated an approximately sixfold higher intensity of Myo1-GFP. This ratio is difficult to reconcile with the stoichiometric targeting of Myo1p by Bni5p (supplementary material Fig. S1).
Bni5p induces Myo1p higher-order structures
To better characterize the Bni5p-dependent targeting of Myo1p in the absence of septins, we fused Bni5-Cherry to the N-terminus of Sso1p (Bni5-Cherry-Sso1). Sso1p is a t-SNARE of the plasma membrane and redirected the Bni5-Cherry moiety of the fusion away from the bud neck to the plasma membrane of the cell (Fig. 2A). We expected that a stoichiometric recruitment of soluble Myo1p should lead to a matching distribution below the membrane. The co-expressed Myo1-GFP did instead form cable-like structures of Myo1-GFP that crossed the cytosol of the cells. These structures seemed to originate from the plasma membrane, and displayed a significant turnover (Fig. 2A,B). We never observed a similar high-order assembly of Myo1-GFP in cells lacking BNI5. A novel allele of BNI5 that missed the 107 C-terminal residues (bni51–341) still induced higher-order structures upon Myo1-GFP co-expression (Fig. 2C). Bni51–341-Cherry was predominantly found in the cytosol of yeast cells. However, the resultant higher-order structures of Myo1p were free of Bni51–341-Cherry-staining (Fig. 2C). We propose that Bni5p actively transforms Myo1p into cable-like structures without being permanently attached to them. To test whether this Myo1p-assembly is supported by actin, we stained actin filaments with Alexa-Fluor-546-phalloidin in cells expressing Bni5-Cherry-Sso1p and Myo1-GFP. We occasionally observed colocalization of Myo1p and actin but more often Myo1 higher-order structures were void of actin (Fig. 2D). In the absence of filamentous actin, Myo1 structures were still visible but appeared shorter and less frequently (Fig. 2E). The activities of the mislocalized Bni5p mutants were specifically directed towards Myo1p as Cdc11-GFP did not form cable-like structures and localized only occasionally to the membrane of cells co-expressing Bni5-Cherry-Sso1p (Fig. 2F).
Binding to Bni5p is not sufficient for formation of Myo1p higher-order structures
We examined N- and C-terminal truncations of Myo1p with a Split-Ubiquitin-based interaction assay to reevaluate the location of the binding site for Bni5p on Myo1p (Fig. 3A,B) (Hruby et al., 2011; Labedzka et al., 2012). Deleting 229 or 529 residues from the C-terminus of Myo1p (Myo11–1700, Myo11–1400) hardly affected the interaction with Bni5p whereas a Myo1-fragment covering only the N-terminal 1100 residues displayed only very weak binding (Fig. 3A). The ATPase domain of Myo1p (Myo11–805) did not interact with Bni5p whereas the C-terminal tail-domain (Myo1860–1929) bound as well as the full-length protein (Fig. 3A) (Fang et al., 2010). In vitro pull down experiments confirmed that a Myo1p fragment from residues 1044–1365 (Myo11044–1365) bound autonomously to Bni5p but not to the septin Cdc11p (Fig. 3B). The bacterially expressed GST fusion of Bni51–341 that induced cable-like structures of Myo1p in vivo also interacted with Myo11044–1365 in vitro (Fig. 3B).
Myo1p is recruited to the septins through its binding to Bni5p (Fig. 1; supplementary material Table S1) (Fang et al., 2010). The binding of Bni5p to the septin ring is probably mediated by Cdc11p although the involvement of other members of the septin rod is discussed (supplementary material Table S1) (Fang et al., 2010; Lee et al., 2002). To estimate the contribution of Cdc11p on the interaction of functional septin-rods with Bni5p we purified octameric septin rods (Cdc11-Cdc12-Cdc3-Cdc10-Cdc10-Cdc3-Cdc12-Cdc11) and hexameric septin rods, missing the peripheral Cdc11p subunit, from E.coli. The interaction of both preparations with Bni5p in a pull down assay revealed that Bni5p is only marginally bound by hexameric septin rods (Fig. 3C). Our conclusion that Cdc11p is the main binding partner for Bni5p is corroborated by the reported lack of Bni5p at the bud neck of cells deleted for Cdc11p (Lee et al., 2002; McMurray et al., 2011).
A GFP fusion to a Myo1-fragment including the binding domain for Bni5p (Myo1991–1365-GFP) still localized to the bud neck (Fig. 3D) (Lister et al., 2006; Fang et al., 2010). The localization in small and medium budded cells was independent of the presence of native Myo1p but strictly dependent on Bni5p (Fig. 3D; middle). In contrast to full-length Myo1p, Myo1991–1365-GFP did not form cables but colocalized with Bni5-Cherry-Sso1p at the plasma membrane (Fig. 3E). Time-lapse analysis of these cells revealed that the binding occurs throughout the cell cycle. In large budded cells, a fraction of Myo1991–1365-GFP changed its location to the bud neck (Fig. 3E). This is very probably due to the presence of the Iqg1p binding sites at the C-terminus of Myo1991–1365 that might interact with the later appearing Iqg1p at the end of the cell cycle (Fang et al., 2010; supplementary material Table S1).
Bni5p might induce Myosin higher-order structures at the bud neck
Do the artificially induced cable-like structures of Myo1p reflect similar higher-order structures at the bud neck? Our observations are compatible with three basic models for the final organization of the septins, Bni5p and Myo1p at the bud neck: (I) Bni5p anchors soluble Myo1p to the septins (Fang et al., 2010). (II) The anchored Myo1p forms higher-order structures while attached to Bni5p. (III) Myo1p higher-order structures remain after their maturation without physical contact to the septin-bound Bni5p (Fig. 4A). The hierarchical structure of Model I offers testable predictions about the dynamics of its three components. The septin ring as the anchor to the bud should display the lowest mobility of the three components. The turnover of Bni5p at the neck is the sum of two separable events: The dynamics of the septins at the bud neck and the reversible binding of Bni5p to the septins. Myo1p-mobility at the bud neck is the sum of three independent events: Myo1p's binding to and dissociating from Bni5p, the dissociation of Bni5p from the septins and the mobility of the septins at the bud neck. Accordingly, Myo1p should have the highest turnover, followed by Bni5p, and the septins. Determining the rates of fluorescence recovery after photobleaching (FRAP) is a well-established method to quantify the dynamics of fluorescent proteins in vivo. We measured the FRAPs of Shs1-GFP, Bni5-GFP and Myo1-GFP and characterized each FRAP by its half-time of recovery (t0.5Shs1, t0.5 Myo1, t0.5Bni5, Fig. 4B,C; Table 1). The calculated t0.5Shs1 of 702 seconds can only be an approximation, as we never observed full recovery of the GFP signal after the bleaching (Fig. 4C,D; Table 1). In accordance with other published studies, we considered the septin structures as immobile during the duration of the FRAP measurements of Bni5- and Myo1-GFP (Dobbelaere et al., 2003). Myo1-GFP and Bni5-GFP displayed ∼10-fold faster recoveries than Shs1-GFP. The extent of recovery was higher than 80% of the fluorescence intensity before bleaching. By quantifying the fluorescence signals of the unbleached areas we could exclude that the recoveries are caused by the lateral movement of the GFP molecules from the unbleached structure (Fig. 4E). t0.5Bni5 described a maximum for medium-budded cells (34 seconds) with smaller values for small- (22 seconds) and large-budded cells (26 seconds) (Fig. 4D; Table 1). During bud growth, t0.5Myo1 steadily increased and always stayed above the corresponding t0.5Bni5 (36 seconds, 46 seconds, 61 seconds) (Fig. 4D; Table 1). The latter observation violates the central prediction of Model I. As the implied formation of higher-order structures prevents the fast dissociation of Myo1p from the bud neck, Model II provides a qualitative explanation for the higher t0.5Myo1. Two testable predictions can be deduced from Model II: (1) The formation of Myo1p higher-order structures should concomitantly decrease the mobility of Bni5p at the bud neck. (2) A Myo1p fragment that contains the Bni5p binding site but lacks the propensity to aggregate in cable-like structures should display a faster FRAP than the full-length Myo1p. We tested the first prediction by measuring the FRAP of Bni5-GFP in cells lacking Myo1p. In these cells, t0.5Bni5 did not increase during bud growth. The t0.5s of medium and large budded cells were nearly identical and stayed below the corresponding t0.5Bni5s of cells containing Myo1p (Fig. 4D; Table 1). The second prediction was tested by measuring the FRAP of Myo1991–1365-GFP. The t0.5 of the fragment was 3.1 seconds and 4.7 seconds in small and large budded wild-type cells and 4.1 seconds and 4.2 seconds in small and large budded Δmyo1 cells (Fig. 4D,F). Myo1991–1365-GFP was consistently more dynamic than Bni5-GFP and Myo1-GFP (Fig. 4D; Table 1). Our FRAP-measurements thus argue for Model II as a better presentation of the organization of the septins, Bni5p and full-length Myo1p at the neck of small and medium-sized buds. For their organization in large budded cells we favor a hybrid of Model II and III, as it can explain the further increase of t0.5Myo1 by the formation of a rigid multimeric Myo1p structure and the decrease of t0.5Bni5 by the progressive loss of the Myo1p-Bni5p contacts (Fig. 4A,D). The dynamic behavior of Myo1991–1365 is adequately represented by Model I.
This value is only an approximation owing to experimental constraints.
Deviations from the mean are calculated as s.e.m.
Under certain conditions the rate constant of a FRAP can be equated with the dissociation rate constant of the corresponding interaction (Bulinski et al., 2001; Sprague et al., 2004). A strong discrepancy between both values would indicate that basic assumptions of our proposed models do not apply. We coupled purified SNAP-tagged Cdc11p or Bni5p onto SPR sensor chips and determined the in vitro dissociation rate constants of the 6-His-Bni5p–Cdc11-SNAP complex (koffBS = 0.055 seconds−1) and of the complex between 6-His-Myo11044–1365 and Bni5-SNAP (koffMB = 0.064 seconds−1) (supplementary material Table S2; Fig. S3). From koffBS we calculated a half-time for the in vivo FRAPs of Bni5p t*0.5Bni5 of 12.6 seconds that is below the observed values. t*0.5Bni5 agreed best with the t0.5Bni5 of cells lacking Myo1p (17.8 seconds, Fig. 4D; Table 1). The dissociation rate constant for the interaction between Myo11044–1365 and the septins (koffMS) was calculated according to Model I as the sum of koffBS and koffMB (supplementary material Table S2). koffMS then yielded a t*0.5Myo1 of 5.8 seconds. This value is slightly larger than the measured t0.5Myo1(991–1365) of ∼4 seconds but still in agreement with Model I.
The differences between predicted and observed values might reflect the conceptual difficulty to quantitatively compare in vivo with in vitro data. Cellular Bni5p is highly phosphorylated (Bodenmiller et al., 2008). Phosphorylation might lead to a higher stability of the Bni5p–Cdc11p complex that is not recapitulated by the dissociation rates of the proteins expressed in E.coli.
Bni5p stabilizes septin rings
Membrane localization of Cdc11-GFP in Bni5-Cherry-Sso1p-expressing cells could only be observed during the interval between septin disassembly and new patch formation, exactly when native Bni5p is not found in complex with the septins (Fig. 1B,D; Fig. 2F). The membrane-anchored Bni5p thus seemed to have escaped a septin ring-dependent modification and/or degradation that normally inhibits its binding to the septins during that time (Fig. 1B) (Lee et al., 2002) (Nam et al., 2007a; Nam et al., 2007b).
We measured a moderate affinity of 0.17 µM (±0.012; n = 3) between Bni5p and the SPR sensor chip-immobilized Cdc11p (supplementary material Fig. S3A). To find out whether this binding confers distinct features on the septin structures, we addressed the influence of Bni5p on the stability of the septin rings by two experiments: Firstly, we determined the ratio of bud neck-localized to cytosolic Shs1-GFP in wild-type and Δbni5-cells (Fig. 5A). Secondly, we compared the extent of FRAP of Shs1-GFP between both cell types (Fig. 5B). The higher fraction of cytosolic Shs1-GFP as well as the higher value of fluorescence recovery in Δbni5-cells demonstrated a positive contribution of Bni5p onto the stability of the septin structures (Lee et al., 2002). In contrast, the deletion of Myo1p increased the pool of soluble septins only slightly and left the extent of the FRAP of Shs1-GFP unaffected (Fig. 5A,B).
Prematurely formed septin rings attenuate polarity establishment
Bni5p dissociates from the septins and Myo1p at the end of the cell cycle before the septins split into two rings and arrives later than the septins and Myo1p at the incipient bud at the beginning of the next cell cycle (Fig. 1). To address whether the strict timing of interaction between Bni5p, Myo1p and the septins is indeed important for the cell cycle dependent changes of the septin/myosin structures, we attached Bni5p covalently to the septin-structures by replacing the native SHS1 by a SHS1-BNI5-Cherry fusion. SHS1 is a non-essential septin and shows a close spatial proximity to Bni5p in our Split-Ub interaction assay (supplementary material Table S1) (Carroll et al., 1998; Garcia et al., 2011; Mino et al., 1998). The transformed cells, additionally expressing Myo1-GFP, showed a superficially normal looking septin- and myosin ring (Fig. 6A). Shs1-Bni5-Cherry-harboring septin rings split shortly after the onset of Myo1p disassembly with the disassembly proceeding slightly slower than in wild-type cells (Fig. 6B, left panel). The released Myo1-GFP returned to the two remaining septin rings harboring the covalently attached Bni5p after contraction (Fig. 6A, left panels). Myo1-GFP-containing double rings were never observed in wild-type cells. The disassembly of the septin structure during cytokinesis started with similar kinetics in Shs1-Bni5-Cherry-expressing cells and in wild-type cells (Fig. 6B, right panel). However, at the end of cytokinesis Shs1-Bni5-Cherry-expressing cells retained a threefold higher concentration at the bud neck (Fig. 6B, right panel). Unexpectedly, 30% of the cells expressing Shs1-Bni5-Cherry were temporarily arrested in G1 and displayed Cherry-stained rings with a diameter similar to the septin ring (Fig. 6C,D; supplementary material Movie 1). Often these rings changed their locations frequently within the cell as if they were only loosely attached to the membrane (Fig. 6C; supplementary material Movie 1). Through colocalization of the correspondingly GFP-tagged proteins we could demonstrate that besides Shs1-Bni5-Cherry both Myo1p, and Cdc11p were also components of these rings (Fig. 6A,D). New bud formation was visible in wild-type cells ∼50 minutes after splitting of the septin rings (Fig. 6E). This period was much more variable in the fraction of SHS1-BNI5-Cherry-expressing cells that displayed the prematurely assembled septin rings [Fig. 6E (Shs1-Bni5 wandering)]. Here, the time between septin splitting and bud appearance could extend to more than 160 minutes whereas the larger fraction of cells not harboring the premature septin rings behaved like wild-type cells [Fig. 6E (Shs1-Bni5 stable)]. We further investigated the timing of cell polarity establishment by monitoring the localization of Bem1p (Fig. 6F). In wild-type cells, Bem1p assembles together with Cdc42p and its GEF Cdc24p at the incipient bud site shortly before the septins are recruited (Bender and Pringle, 1991; Chenevert et al., 1992; Howell and Lew, 2012; Kozubowski et al., 2008; Slaughter et al., 2009; Wedlich-Soldner et al., 2004). In SHS1-BNI5-Cherry-expressing cells harboring premature septin rings, Bem1-GFP was distributed in clusters dispersed at the plasma membrane (Fig. 6F). These clusters were also detectable in wild-type cells but persisted much longer in cells displaying the prematurely formed septin rings (34.9±1.8 minutes versus 100±16.6 minutes) (Howell et al., 2012). Once Bem1p was concentrated as a patch, septin recruitment and budding was initiated without further delay (Fig. 6F; supplementary material Fig. S4A; Movie 2). The same behavior was displayed by Spa2p-GFP, a member of the polarisome and a further early marker of polarity establishment (Sheu et al., 1998). Again, in un-budded cells displaying wandering septin rings Spa2-GFP stained the plasma membrane before being concentrated at the incipient bud site. Septin recruitment and bud outgrowth followed subsequently (supplementary material Fig. S4B; Movie 3).
The actin and microtubule cytoskeletons obtain much of their structural and dynamic diversities through interactions with proteins that influence their stability, rate of subunit gain and loss, degree of crosslinking, and filament lengths and organization. Not much is known about the roles of septin-binding proteins in yeast although polymerization of the septins into ring-like structures, the splitting of these rings shortly before cytokinesis and the formation of a septin-based diffusion barrier might necessitate similar activities (Qiu et al., 2008; Eluère et al., 2012).
Bni5p was originally described as Cdc11p binding protein that suppresses the temperature sensitive effects of certain septin mutants (Lee et al., 2002). The protein was more recently shown to be responsible for recruiting Myo1p to the septin ring during the early phases of the cell cycle (Fang et al., 2010). Fang and colleagues proposed a linear targeting model where Bni5p would anchor soluble Myo1p to the septins through the simultaneous binding of Myop1 and the septin subunit Cdc11p (Fang et al., 2010).
We wish to extend this model by two novel findings. Contrary to the expectation of a linear targeting model, Bni5p arrives later at the bud neck than Myo1p. Supported by our observation that in Bni5p-deficient cells Myo1p is still transiently associated with the incipient bud site, we propose a two-step mechanism for the localization of Myo1p (Fig. 1): a Bni5p-independent concentration of Myo1p at the incipient bud site is followed by its Bni5p-dependent capture at the newly formed bud neck.
During our attempts to reconstitute the interaction between Bni5p and Myo1p at a heterologous site in the cell, we discovered higher-order structures of Myo1p in cells expressing two mutants of Bni5p that fail to properly localize at the bud neck (Fig. 2). As these structures are not seen in cells lacking Bni5p, we propose that Bni5p might actively induce the formation of these Myo1p higher-order structures. To approach the question whether similar structures of Myo1p are also induced at the bud neck of wild-type cells, we compared the t0.5 of the FRAPs of the full-length Myo1p with the t0.5 of a short, soluble Myo1991–1365 fragment containing the binding site(s) to Bni5p. In contrast to Myo1991–1365, the t0.5s of Myo1p increased during bud growth and were clearly above the t0.5s of Myo1991–1365 and a theoretically predicted t*0.5 that was calculated from the dissociation rate constants under the assumptions of the linear targeting model (Fig. 4D; Table 1). The quantitative differences between Myo1p and Myo1991–1365 in the rate of dissociation from the bud neck correlate with their different propensities to be transformed by mislocalized Bni5p into cytosolic Myo1-structures. We thus propose the formation of higher-order structures as the additional force that retains Myo1p more strongly than its fragment Myo1991–1365 at the septin ring. It is not clear how these proposed higher-order structures at the bud neck compare with the artificially induced cable-like structures in the cytosol or the thick filaments of myosin observed in other eukaryotes.
At this point alternate mechanisms for retaining Myo1p at the bud neck cannot be strictly excluded. For example, Bni5p might stimulate a conformation of the septins with a certain affinity to Myo1p. Alternatively, a Bni5p-dependent recruitment of a further Myo1p-binding partner to the septins might strengthen the interaction between septins and Myo1p. In the absence of evidence for any other mechanism, we favor the herein proposed minimal model of Myo1p anchoring and subsequent higher-order structure formation.
Bni5p dissociates from the bud neck shortly before the onset of cytokinesis without disrupting Myo1 localization even in cells lacking Iqg1p (Fang et al., 2010) (Fig. 1D). These observations might imply that the interaction between Bni5p and Myo1p does not contribute to Myo1p-localization at later phases of the cell cycle. This conclusion is supported by our findings that Bni5p is not detected in the artificially induced cable-like structures of Myo1p, and that the stabilizing influence of Myo1p on the dissociation of Bni5p from the bud neck is lost in large budded cells (Fig. 2A,C; Fig. 4D; Table 1).
Taken together, our initially proposed three alternative models for the targeting of Myo1p are not exclusive but might correspond to stages in the binding and subsequent maturation of Myo1p (Fig. 4A). After the initial binding of soluble Myo1p to the Cdc11p-attached Bni5p (Fig. 4A, I), Myo1p is converted into higher-order structures (Fig. 4A, II), to be finally released as an even more interconnected multimer (Fig. 4A, III). The herein proposed targeting pathway differs from those found in other eukaryotic cells by directing myosin assembly very early to the predetermined site of the future cell separation. However, the movement of myosin to the assembly site, the role of the septins in targeting and anchorage as well as the existence of higher-order structures of myosin are conserved features among yeast and other eukaryotic cells (Goldbach et al., 2010; Joo et al., 2007; Kosako et al., 2000; Matsumura, 2005; Vale et al., 2009; Yumura, 2001; Yumura et al., 2008).
Cdc11p caps the septin rods at both its ends (Bertin et al., 2008). The simultaneous binding of Bni5p to Cdc11p and Myo1p will align the Myo1p molecules that emanate from the different poles of the septin rod in anti-parallel orientation. Rods are supposed to be connected through a Cdc11p-Cdc11p contact within a septin filament (Bertin et al., 2008). This contact would add further Myo1p molecules that could then align in the septin filament with the Myo1p molecules of the previous rod in parallel orientation.
We investigated the reasons for the strict timing of the interaction between Bni5p and the septins during the cell cycle by fixing Bni5p covalently to one of the septin subunits. The most striking effect of this manipulation was the appearance of septin rings before the establishment of cell polarity. Our experiments cannot discriminate whether these septin rings constitute the remnants of the previous cell cycle or are de novo structures formed immediately after cytokinesis. Independent of the mechanisms of their formation, these septin rings delayed the formation of a new polarity axis (Fig. 6). We implied a negative feedback between matured septin structure and the proteins of cell polarity establishment. A possible mechanism could involve the septin-ring dependent sequestration of a critical factor for polarity establishment. This feedback would ensure a hierarchical and singular formation of one bud. Accordingly, the early dissociation of Bni5p from the bud neck is required for an efficient disassembly of the septin/myosin structure at the end of cytokinesis whereas the reasons for its delayed appearance during bud formation might be twofold: to prevent premature septin/myosin ring assembly before polarity establishment and to avoid formation of Myo1p higher-order structure before targeting. Once assembled, septin/myosin higher-order structures repress the search for a new site of polarity establishment.
Our observation that inappropriately formed or mutated septin rings can delay progression into a new cell cycle is supported by an earlier study where certain non-phosphorylatable alleles of CDC3 were shown to lead to two simultaneously occurring septin rings (Tang and Reed, 2002). Cells with two septin rings show a considerable delay in cell cycle progression. As we never observed two septin rings in the Shs1-Bni5-Cherry-expressing cells, the molecular mechanisms leading to the cell cycle delay in the two different yeast mutants might thus be related but still different (Tang and Reed, 2002). Based only on the study of mutants it is furthermore difficult to evaluate the significance of the postulated feedback between septin structure and polarity establishment during the unperturbed cell cycle. The recently observed oscillation of Bem1p during polarity establishment was explained by a negative feedback loop of unknown mechanism (Howell et al., 2012). The damping of the amplitude correlated with the appearance of the septins at the site of polarity, suggesting that septins are involved but not primarily responsible for establishing this feedback loop (Howell et al., 2012).
Cells keep their polarity axis throughout the cell cycle. Mutations that allow cells to reorient their polarity axis without grossly affecting the structure of the septins demonstrate that a septin ring is not the only and probably not the major factor for keeping the polarity axis during the later stages of the cell cycle. For example, certain mitotic arrest mutations of LTE1 induce the growth of buds that is orthogonal to the original axis and that will eventually lead to a characteristic hammerhead morphology (Geymonat et al., 2010). These and other experiments seem to indicate the presence of multiple pathways that are either used simultaneously or at different periods of the cell cycle to guarantee the observed robustness in determining the morphology of the yeast.
Materials and Methods
An overnight culture of the respective yeast strain was diluted 1∶10 in 3 ml of fresh medium and grown for 3–4 hours at 30°C. The cells were spun down and resuspended in 15–50 µl fresh medium. 3 µl of the resuspension were spotted on a microscope slide, covered with a glass coverslip and immediately observed under the microscope. For live cell imaging, the resuspension was spread over a 1.7% agarose pad. All live cell imaging experiments were performed at 30°C in humidified atmosphere for up to 6 hours.
Cells were examined with an Axio Observer Z.1 Spinning Disc confocal microscope (Zeiss, Göttingen, Germany). Images were acquired under the control of the AxioVision 4.8.2 software (Zeiss, Göttingen, Germany) as Z-series with five to twelve layers. The distance between two Z-layers was set to 260 nm–400 nm. For image acquisition, a Plan-APOCHROMAT 63×/1.4 Oil DIC ∞/0.17 objective was used with the following camera settings: 2×2 binning, gain = 1. For excitation of GFP or Cherry, a 488 nm diode laser at 20% laser power or a 561 nm diode laser at 30% laser power was used, respectively.
Fluorescence recovery after photobleaching (FRAP)
FRAP experiments were performed on a DeltaVision microscope system (GE Healthcare, Freiburg, Germany) equipped with a Quantitative Laser Module (GE Healthcare, Freiburg, Germany), a 100×UPlanSApo 100×1.4 Oil ∞/0.17/FN26.5 objective (Olympus, Hamburg, Germany) and a CoolSNAP HQ2 camera (Photometrics, Tucson, USA). For image acquisition, a camera gain of 1 and 1×1 binning was used. A region of interest (ROI) of the specimen was completely bleached using one 488 nm laser spot at 20% laser power. The temperature was kept at 30°C during the experiments. The microscope was controlled by the softWoRx 5.0 software (GE Healthcare, Freiburg, Germany).
Quantitative analysis of microscopy data
For quantitatively comparing the fluorescence intensities of Myo1-GFP and Bni5-GFP at the bud neck, we introduced a centromeric plasmid expressing Myo1-GFP from its native promoter in a strain lacking the genomic copy of MYO1. We introduced into a second strain expressing Bni5-GFP from the chromosomal locus a centromeric plasmid expressing Sso1-Cherry from the PMET17-promoter. Immediately before the measurements, both logarithmically grown strains were mixed in equal amounts and the mixture prepared for fluorescence micrcoscopy. The fluorescence intensities of the signals at the bud neck were quantified as above. The identities of the cells were subsequently determined by the presence or absence of Cherry fluorescence.
Analysis of FRAP experiments
Each FRAP curve shown represents the average of the overlay of at least four independent measurements. Only cells exhibiting similar pre-bleach intensities and similar bleaching levels were analyzed.
Data evaluation and statistical analyses were performed using Prism. t-tests were used to compare the fluorescence intensities obtained from static images. The datasets obtained from time-lapse analyses and FRAP experiments were compared with each other by two-way ANOVA tests to evaluate the significance of their differences.
Predicting the rate constants for the off-rates of Bni5p and Myo1p from the bud neck
We initially assumed that Bni5p is bound to the bud neck through its interaction with the septins and that Myo1p is anchored to the bud neck in its soluble form through Bni5p. The predicted rate constant of dissociation of Myo1p from the bud neck is thus the sum of the experimentally determined koff values of the Myo1p–Bni5p and the Bni5p–septin complexes. From the koffs the predicted half-time for the FRAP of Myo1p at the bud neck (t*0.5Myo1), was calculated by using equation (4). The combined standard deviation of t*0.5Myo1 was calculated by adding the individual variances of both observations.
Expression of GST-tagged proteins and in vitro binding assays
GST-tagged proteins were expressed and immobilized on glutathione-coupled sepharose beads. In vitro binding assays were performed as described (Labedzka et al., 2012).
Expression and purification of Cdc11p, Bni5p and the Myo1 fragment
CDC11-SNAP was cloned in frame to the 6×His-tag into the multiple cloning site of pET-Duet1 (Novagen, Merck, Darmstadt). The SNAP tag represents an engineered version of the human O6-Alkylguanine-DNA-Alkyltransferase (Gronemeyer et al., 2006).
MYO11044–1365 and BNI5 were PCR amplified from yeast genomic DNA and cloned in frame to the 6×His-tag into a pET15b-derived expression plasmid containing or lacking the SNAP tag to yield 6×His-MYO11044–1365, 6×His-BNI5 and 6×His-BNI5-SNAP.
All proteins were expressed in E.coli strain BL21 (DE3) at 18°C in super broth medium over night. Cells were resuspended in IMAC Buffer A [50 mM KH2PO4 pH 8.0, 20 mM imidazole, 300 mM NaCl (containing an additional 1 mM DTT and 0.5% TritonX100 for the purification of Cdc11-SNAP)] supplemented with ‘complete protease inhibitor cocktail’ (Roche, Penzberg, Germany) and lysed by lysozyme treatment and sonication. Purification was achieved by immobilized metal affinity chromatography (IMAC) on a 5 ml HisTrap column (GE Healthcare, Freiburg, Germany) by applying a linear gradient of IMAC buffer B (50 mM KH2PO4 pH 7.4, 200 mM imidazole, 300 mM NaCl), followed by size exclusion chromatography against HBSEP buffer (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, 0.05% Tween 20, pH 7.4) using a Superdex 200 16/60 column (GE Healthcare, Freiburg, Germany). Protein-containing fractions were concentrated using an Amicon Ultra device (Millipore, Schwalbach, Germany) and used immediately for further experiments.
Pulldown with recombinant septin rods
Recombinant octameric septin rods were expressed in E.coli and purified with minor modifications as described (Bertin et al., 2008). Briefly, we followed the established protocol (Bertin et al., 2008) but reduced it to a two-step procedure consisting of an IMAC and an ion exchange separation (Renz et al.; manuscript submitted). A hexameric version of the rod lacking the terminal Cdc11p subunit was expressed and purified using the same protocol with an additional size exclusion chromatography step (Bertin et al., 2012). Both rods eluted from an analytical size exclusion column in a single peak containing all expressed subunits indicating the intactness of the structures and a 1∶1∶1∶1 stoichiometry in case of the octameric and a 1∶1∶1 stoichiometry in case of the hexameric septin rod. Octameric and hexameric septin rods and purified SNAP protein as control (200 µl each of a 100 nM solution) were covalently bound to 60 µl SNAP capture pulldown resin (NEB, Heidelberg, Germany). The beads were subsequently blocked with 1 mg/ml BSA, incubated with 200 nM of BG-TMR-labeled 6His-Bni5-SNAP or 6His-SNAP in a total volume of 200 µl. All steps were performed under high salt conditions (300 mM NaCl). After washing, bound proteins were eluted from the beads with 2×Lämmli buffer and directly subjected to SDS-PAGE. TMR fluorescence was directly detected ‘in gel’ (data not shown). Subsequently, the proteins were transferred onto nitrocellulose membrane and subjected to western blot analysis.
SPR analysis and data evaluation
The interactions of Bni5p with Cdc11p and of Myo11044–1365 with Bni5p were quantified by SPR using a Biacore X100 system (GE Healthcare, Freiburg, Germany). Purified Cdc11-SNAP or Bni5-SNAP was covalently labeled with BG-Biotin by SNAP tag chemistry and excess substrate was removed using a NAP5 desalting column (GE Healthcare, Freiburg, Germany). For the capture of biotinylated proteins, the surface of a CM5 Chip (GE Healthcare, Freiburg, Germany) was coated with an anti-biotin antibody (US Biologicals, Hamburg, Germany) with NHS ester chemistry in 10 mM sodium actetate pH 5.0. HBSEP was used as running buffer for all measurements. Purified analyte protein (6×His-Bni5p or 6×His-Myo11044–1365) was prepared in suitable concentrations in the same buffer. Regeneration of the sensor chip was performed with a 15-second injection pulse of 12 mM NaOH. The contact time of the analyte with the ligand was set to 60 or 180 seconds, respectively.
For both interactions analyzed, koff(fast) describes a fast dissociation and koff(slow) an ∼10-fold slower dissociation reaction. We considered only the fast reaction as relevant for determining the koff of the interaction and the slow dissociation as a consequence of unspecific adherence of 6×His-Myo11044–1365 and 6×His-Bni5p to the surface of the chip. Only values from curves with a goodness of fit (R2) grater than 0.990 and a predominant contribution of the fast reaction (%fast >80%) were considered.
Construction of fusion genes and other molecular manipulations
GFP-, Cherry- and Cub-fusion genes were constructed by homologous recombination in the yeast genome as described (Hruby et al., 2011; Dünkler et al., 2012). The BNI5 ORF was cloned in frame behind the last 402 bp of the ORF of SHS1 in the SHS1-Cherry-pRS304 vector to construct SHS1-BNI5-Cherry-pRS304. Linearization at the single NdeI site in the ORF of SHS1 and transformation into yeast replaced after homologous recombination the SHS1 gene by the full-length SHS1-BNI5-Cherry fusion gene. The ORF of BNI5 was inserted in frame between the PMet17 promoter sequence and the Cherry sequence in the PMet17 Cherry-SSO1-pRS313 vector to yield PMet17 BNI5-Cherry-SSO1-pRS313. After transformation of the centromeric vector, the expression of the gene fusion could be regulated by the amount of methionine in the medium. Gene deletions were performed by one-step PCR-based methods as described (Janke et al., 2004). A list of plasmids and yeast strains used in this study can be found in supplementary material Tables S3, S4.
Growth conditions, yeast strains and genetic methods
Culture media and yeast genetic methods were performed following standard protocols (Guthrie and Fink, 1991). Media for the Split-Ubiquitin interaction assay contained 1 mg/ml 5-fluoro-orotic acid (5-FOA, Fermentas, Heidelberg, Germany). Latrunculin A (Life Technologies, Darmstadt, Germany) was added to the yeast culture in a final concentration of 0.1 mM 15 minutes before the analysis. For staining with Alexa-Fluor 546 Phalloidin (Life Technologies, Darmstadt, Germany), cells were fixed with formaldehyde (3.7% final concentration) and incubated with Alexa-Fluor 546 Phalloidin (final concentration 66 nM) for 30 minutes at 4°C. All yeast strains used were derivatives of JD51 (Dohmen at al., 1995).
Split-Ub interaction analysis
We thank Dr K. Labedzka for the initial analysis of Myo1p; Dr R. Li for the Myo1-GFP expressing plasmid; and Dr J. Müller for improving the manuscript.
C.S., with contributions from J.G., C.R. and T.G., performed and analysed the experiments. N.J., C.S., and T.G. designed the research. N.J. wrote the manuscript with contributions of all other authors.
This work was supported by a German Research Foundation (DFG) Research Grant [grant number JO 187/5-1]. C.S. was supported by the International Graduate School in Molecular Medicine Ulm, DFG/GSC270.