To understand how a stable hourglass can form from a relatively dynamic septin ring (Figure 1A), one must first understand how septins are assembled into filaments, which are further organized into higher-order structures such as rings and hourglasses. Yeast septins were first seen at the bud neck as a membrane-associated structure made from filaments with a 10 nm width (Byers and Goetsch, 1976a). However, when expressed and purified in vitro from E. coli or budding yeast in the presence of high salt (>300 mM KCl or NaCl) only septin rods of the approximate length of an octamer were formed (Figure 1B, left) (Frazier et al., 1998; Bertin et al., 2008). When the ionic strength was lowered (50–75 mM salt) and with Cdc11 serving as the terminal septin subunit were filaments visualized (Frazier et al., 1998; Bertin et al., 2008), with the C-terminal extensions (CTEs) of Cdc3 and Cdc12 interacting on neighboring filaments to form the paired filaments (Figure 1B, right) (Bertin et al., 2008; Bertin et al., 2010). In contrast, the Shs1-capped septin rods associate laterally into curved bundles and rings in vitro under low salt condition (Garcia et al., 2011). While data from this study concerning the ability for Shs1-capped septin rods to assemble end-on-end into filaments is not as clear as that of Cdc11-capped septin rods, additional analysis by Förster Resonance Energy Transfer has more conclusively negated the possibility that two Shs1-capped rods can polymerize end-on-end (Booth et al., 2015). Given that the budding yeast cytoplasm has potassium and sodium concentrations of 200–300 and 20 mM, respectively (van Eunen et al., 2010; Cyert and Philpott, 2013), the ability of yeast septins to spontaneously generate filaments in the cytoplasm should be quite low.

Strikingly, in the presence of lipid monolayers containing 20% phosphatidylinositol-4,5-bisphosphate (PIP2), purified Cdc11-capped septin rods could form long paired septin filaments even at higher salt concentrations (Figure 1B, right) (Bertin et al., 2010; Bridges et al., 2014). This implies that septin octamers probably exist in the cytoplasm and are assembled into paired filaments once bound to the plasma membrane at the presumptive bud site or the bud neck. This indeed appears to be the case, as it was found by fluorescence correlation spectroscopy that septins in the cytoplasm exist as complexes and not as monomers or filaments (Bridges et al., 2014). The septin complexes appear to “sense” and favor the membrane curvature found at the bud neck, as the purified septin complexes preferentially engaged with lipid-coated beads of diameters producing a similar positive curvature as that found at the bud neck (Bridges et al., 2016; Cannon et al., 2019). The organization of the septin filaments into higher-order structures may also be governed by the changes in membrane curvature during bud growth and cytokinesis, as suggested by a recent study that examined and modeled the impact of varying degrees of positive and negative curvature on septin filament assembly and organization on lipid-covered surfaces in vitro (Beber et al., 2019).

Higher-order assemblies of septins are likely regulated by PTMs as well as SAPs (Gladfelter et al., 2001; Mcmurray and Thorner, 2009; Hernández-Rodríguez and Momany, 2012; Alonso et al., 2015; Perez et al., 2016). One such SAP, Bni5, localizes to the bud neck in a septin-dependent manner (Lee et al., 2002). Bni5 appears to interact with the CTEs of the terminal subunits Cdc11 and Shs1 in vitro and in vivo (Figure 1C) (Finnigan et al., 2015; Booth et al., 2016). Recombinant Bni5 can reduce the distance between individual filaments of a paired filament and can also bundle septin filaments, thereby contributing to a higher-order organization (Figures 1B,C) (Patasi et al., 2015; Booth et al., 2016). It is possible that a combination of phospholipid composition, PTMs, and SAPs is able to bypass the need for positive curvature during the nascent septin ring formation, as there is no discernable bud neck at this point and the membrane topology at the presumptive bud site does not match the curvature preference of the septins. Taken together, the in vitro analysis of the ability of septins to interact with membranes, recognize membrane curvature, and recruit SAPs for higher-order organization will undoubtedly help understand the mechanisms for the assembly and regulation of the plethora of septin structures in vivo.

Upon bud emergence or shortly after, the septins transition from a dynamic nascent ring at the presumptive bud site to a stable septin hourglass at the bud neck (Caviston et al., 2003; Dobbelaere et al., 2003). This stabilization could be caused by filament pairing as observed by thin-section EM and EM tomography (Byers and Goetsch, 1976a; Bertin et al., 2012; Bertin and Nogales, 2012) and by polarized fluorescence microscopy (Vrabioiu and Mitchison, 2006; Demay et al., 2011). These studies are not in agreement as to the orientation of the paired filaments in relation to the mother-bud axis, but are all in agreement that the hourglass seems to be made up of mainly paired filaments. This was further confirmed by our group in cell cortices visualized by platinum-replica electron microscopy (PREM) (Ong et al., 2014). The PREM data favors the orientation of radial paired filaments (about 300–400 nm in length) in an “early hourglass” being parallel to the mother-bud axis as postulated by the groups using polarized fluorescence microscopy (Vrabioiu and Mitchison, 2006; Demay et al., 2011), because the presumed same paired filaments in a “late or transitional hourglass” lie perpendicular to the visualized AMR that runs circumferentially at the bud neck (Ong et al., 2014). Despite the consensus on the paired filaments making up the septin hourglass, whether and how filament pairing is regulated in vivo has remained elusive.

We recently provided the first evidence of such regulation using genetic perturbations and PREM analyses of the septin hourglass in vivo (Marquardt et al., 2020). The protein kinase Elm1 localizes to the bud neck from bud emergence to the onset of cytokinesis in a septin dependent manner (Bouquin et al., 2000; Kang et al., 2016; Marquardt et al., 2020). This localization pattern matches perfectly with the timing of septin hourglass assembly and maintenance at the bud neck. Elm1 had already been implicated in maintaining the stability of septin structures at the bud neck (Bouquin et al., 2000; Asano et al., 2006), but the underlying mechanism had remained unknown at the time. We discovered that the septin hourglass was selectively destabilized with mostly the daughter half of the hourglass mislocalized to the growing bud cortex in elm1Δ mutants in both a kinase-dependent and -independent manner (Figure 1D, left and center panels) and that these phenotypes are at least partially mediated through Bni5 (Marquardt et al., 2020). This destabilization appears to be largely due to an increase in single septin filaments comprising the septin hourglass-like structure when visualized by PREM (Figure 1D, right panels) (Marquardt et al., 2020). This was confirmed by artificially pairing septin filaments using the GFP-nanobody/binding protein (GBP) system (Rothbauer et al., 2006; Kubala et al., 2010) and witnessing a significant rescue of the septin phenotype at the hourglass stage in elm1Δ cells (Figure 1E) (Marquardt et al., 2020). The same experiment also provided evidence that septin filament pairing and unpairing are likely regulated throughout the cell cycle, because aberrant septin structures at the old division site remained even in otherwise wild-type (WT) cells when the septins were forced to stay paired (Figure 1E, arrows) (Marquardt et al., 2020). Thus, to understand the structural transitions or architectural remodeling, one must understand the mechanisms of septin filament pairing and unpairing during the cell cycle.

Two major events are associated with the septin hourglass-to-double ring (HDR) transition: septin filament disassembly and reorganization. FRAP analysis indicates that septins become more dynamic during the HDR transition (Caviston et al., 2003; Dobbelaere et al., 2003). This dynamic change is accompanied by the net loss of septins at the division site by 20–30%, implicating septin filament disassembly in the process (Caviston et al., 2003; Dobbelaere et al., 2003; Demay et al., 2011; Wloka et al., 2011). Analysis by polarized fluorescence microscopy suggests that paired septin filaments reorganize from a radial arrangement in the hourglass to a circumferential arrangement in the double ring, a 90° change in filament orientation (Vrabioiu and Mitchison, 2006; Demay et al., 2011). Despite these insights, the detailed architectures of the septin hourglass and double ring at the filament level and the molecular mechanisms controlling the HDR transition remained unknown.

By combining cell synchronization with PREM, we have determined the architecture of different septin assemblies at the division site during the cell cycle (Ong et al., 2014; Chen et al., 2020; Marquardt et al., 2020). The “early hourglass” from pre-anaphase cells contains exclusively paired filaments arranged radially and parallel to the mother-bud axis (Ong et al., 2014). This structure is converted into a “transitional hourglass” before cytokinesis, in which the paired filaments are connected perpendicularly by periodic circumferential single septin filaments on the membrane-proximal side (Ong et al., 2014). Strikingly, these intersecting septin filaments, which define a septin gauze, are located exclusively at the outer zones of the transitional hourglass whereas myosin-II filaments occupy the middle zone but exclusively on the cytoplasmic side of the hourglass (Figure 2B, left) (Ong et al., 2014; Chen et al., 2020). At the onset of cytokinesis and under the control of the mitotic exit network (Cid et al., 2001a; Lippincott et al., 2001; Ong et al., 2014; Tamborrini et al., 2018; Chen et al., 2020), this “zonal” transitional hourglass is remodeled into a double ring that consists of exclusively circumferential paired and single septin filaments (Figure 2B, right) (Ong et al., 2014; Chen et al., 2020). These distinct architectures revealed by PREM analysis provide a foundational blueprint for mechanistic analysis of septin high-order assembly and remodeling.

In addition to septin filament disassembly (Figure 2B, center) (Caviston et al., 2003; Dobbelaere et al., 2003; Demay et al., 2011; Wloka et al., 2011), careful analysis of septin behavior during the HDR transition by FRAP, photo-activation, photoconversion, and super-resolution microscopy suggests that there must be “spatial cues” present at the ends of the septin hourglass that controls septin filament reassembly and reorganization into a double ring (Figure 2B) (Ong et al., 2014). What are the spatial cues? How do they control the HDR transition? These questions remained unanswered.

Our recent work suggests that the RhoGEF Bud3 and the anillin-like protein Bud4 function as the spatial cues to drive the HDR transition (Chen et al., 2020). Bud3 and Bud4 were initially identified as factors essential for axial budding in S. cerevisiae (Chant et al., 1995; Sanders and Herskowitz, 1996). Bud3 also acts as a GEF for Cdc42 in early G1 phase to spatially link successive polarization or budding events in haploid cells (Kang et al., 2014). Bud3 and Bud4 interact with each other (Kang et al., 2012; Kang et al., 2013; Wu et al., 2015) and associate with the septin hourglass after S/G2 and with the septin double ring after the HDR transition in a nearly identical pattern (Figure 2C, left) (Chant et al., 1995; Sanders and Herskowitz, 1996; Wloka et al., 2011; Kang et al., 2013; Chen et al., 2020). PREM coupled with immunogold labeling indicates that Bud3 and Bud4 specifically associate with the septin gauze at the outer zones of the transitional hourglass (Figure 2C, right) (Chen et al., 2020). Both proteins have been implicated in septin organization, likely at the stage of double ring formation (Lord et al., 2000; Gladfelter et al., 2005; Guo et al., 2011; Wloka et al., 2011; Eluere et al., 2012; Kang et al., 2013; Wu et al., 2015; Mcquilken et al., 2017). Taken together, these observations suggest that Bud3 and Bud4 likely define the spatial cues that act in the right place at the right time to drive the HDR transition, but how?

Live-cell imaging and PREM analyses demonstrate that Bud3 and Bud4 play distinct and essential roles in controlling the septin HDR transition (Chen et al., 2020). As indicated above, the transitional hourglass in WT cells possesses a zonal architecture where the myosin-II filaments are sandwiched by Bud3 and Bud4-decorated septin gauze (Figure 2C; Figure 2D, top row, left). During the HDR transition, the septins in the middle of the hourglass are preferentially lost whereas those at the ends of the hourglass are preferentially stabilized (Figure 2D, top row, right), leading to the double ring formation. In the absence of Bud3, however, the circumferential single filaments at the ends of the transitional hourglass are completely lost, with many of the structures displaying elongated paired filaments at one side of the hourglass (Figure 2D, middle row, left) (Chen et al., 2020). This filament-level observation may explain the phenotype of bud3Δ cells detected by live-cell imaging that, during the HDR transition, the septin hourglass was first thinned at its edges, followed by the disintegration or fragmentation of the remaining structure in the middle (Figure 2D, middle row, right) (Chen et al., 2020). These observations suggest that Bud3 is specifically required for the stabilization of the transitional hourglass at its edges, and, furthermore, this selective stabilization mechanism must act before the cell cycle-triggered disassembly of the septin filaments in the middle of the hourglass (located between the PM and the myosin-II filaments). In the absence of Bud4, the entire transitional hourglass became destabilized, often in an asymmetric manner where one side of the structure was affected more than the other (Figure 2D, bottom row, left) (Chen et al., 2020). This is consistent with live-cell imaging analysis showing that the mother side of the septin double ring is preferentially lost during the HDR transition (Wloka et al., 2011; Chen et al., 2020). In the absence of both Bud3 and Bud4, the transitional hourglass was severely compromised and, consequently, was hardly detectable by PREM and the double ring was essentially abolished (Chen et al., 2020). Taken together, these data suggest that Bud3 and Bud4 function as the spatial cues by exclusively localizing to the outer zones of the transitional hourglass and instructing the double ring formation at the ends of the hourglass. Bud3 is essential for the circumferential single filament assembly whereas Bud4 is likely required for the stability of both paired and single filaments in the transitional hourglass, especially at the mother side of the bud neck.