A gene duplication of a septin reveals a developmentally regulated filament length control mechanism

Septins are a family of filament-forming, GTP-binding proteins that function in many cell processes including cytokinesis (Hartwell et al., 1974), cell polarity (Gladfelter et al., 2005), and membrane organization (Luedeke et al., 2005; Yamada et al., 2016). Although septins are highly conserved from yeast through humans, the number of septin genes between organisms varies greatly, from 1 in Chlamydomonas to 13 in humans (Pan et al., 2007; Shuman and Momany, 2022). The variable number of septin genes within organisms is suggested to result from multiple gene duplications. Additionally, alternative splicing in mammalian cells has the potential to produce a wide array of variability in septin gene-products, both between different tissues and even within a single cell (Russell and Hall, 2011; Sellin et al., 2014). Despite this well-appreciated complexity in subunit composition, how the pool of septin proteins available in a cell contributes to septin properties and functions is poorly understood.

Septin proteins self-assemble into rod-shaped hetero-oligomeric complexes. These complexes are the soluble, minimal subunit of septins, measuring 32 nm in length (Bertin et al., 2008; Bridges et al., 2014). Octamers concentrate on either membrane surfaces or cytoskeletal networks (in animals) and elongate through annealing such that the terminal subunits of the octamers interact to create filaments (Field et al., 1996; Bertin et al., 2010; Bridges et al., 2014). Septin filaments are highly flexible, but their mechanical properties can be modulated through pairing and protein–protein interactions to form a variety of higher-order structures such as lattices, bundles, and rings (Bertin et al., 2010; Garcia et al., 2011; Ong et al., 2014). The process of assembly and polymerization has been studied in most detail in the budding yeast Saccharomyces cerevisiae, where five mitotic septins are expressed and arranged in the following order: X-Cdc12-Cdc3-Cdc10-Cdc10-Cdc3-Cdc12-X, where the terminal subunit X can be either Cdc11 or Shs1 (Garcia et al., 2011). Cdc11 and Shs1 impart different properties on octamers and filaments. These proteins are only 36% identical, making it difficult to dissect the molecular basis for the different properties. The interface between terminal subunits is critical to many parameters relevant to septin assembly and could influence properties such as flexibility, polymerization, and fragmentation rates, as well as bundling and crosslinking.

We set out to address how the sequence variations in the terminal subunit can impact septin filament properties by taking advantage of a duplication of the gene encoding CDC11 in the filamentous fungus Ashbya gossypii. This results in a situation with two versions of Cdc11, Cdc11a and Cdc11b, which show 91% sequence identity. Here, we report that Cdc11a and Cdc11b assemble septin structures in A. gossypii in a development-specific manner and that the deletion of either CDC11A or CDC11B results in distinct morphological phenotypes. This functional specialization arises from a small number of primary sequence changes. The position of Cdc11 at the termini of complexes, which mediate filament elongation, enables these modest residue changes to have substantial impacts on the biophysical features of septin filaments. Remarkably, different Cdc11 complexes produce septin filaments of distinct lengths, different membrane binding abilities, and sizes of higher-order assemblies in vivo. In particular, a single amino-acid substitution is sufficient to impart distinct biophysical properties to the different Cdc11 complexes. This work reveals how specialization after gene duplication can generate novel biophysical features that can control the size, shape, and function of cytoskeletal polymers.

The A. gossypii (“Ashbya”) genome encodes eight septin genes, all of which have homologs in S. cerevisiae (Gattiker et al., 2007). Interestingly, in Ashbya there are two copies of the CDC11 gene on two different chromosomes, with CDC11a (AER445C) on chromosome V and CDC11b (AFR436C) on chromosome VI. This is in striking contrast to S. cerevisiae, which has a single version of CDC11 and any other filamentous growing fungi studied to date. Notably, S. cerevisiae underwent a whole-genome duplication after divergence from a common ancestor and thus harbors many retained duplicated genes, but not CDC11 or other septins (Dietrich et al., 2004). The 411 amino acids of AgCdc11a share 91% identity (95% similarity) to the 408 amino acids of AgCdc11b. We predict that these two paralogs originated from an ancestor that had a tandem duplication of the CDC11 locus (Fig. S1 A). Thus, a duplication and genome rearrangement have preserved a lineage of closely related Ashbya species with an additional Cdc11 protein. Therefore, we set out to examine the cellular and biophysical significance of having a second Cdc11 protein of similar, but not identical sequence.

We first examined the expression, timing, and localization of Cdc11a- and Cdc11b-capped oligomers in cells. As the CDC11b locus could encode a pseudogene that is not expressed, we examined transcript levels of CDC11b over time. We found that both CDC11a and CDC11b transcripts are regulated throughout the life cycle of A. gossypii. Specifically, CDC11b transcripts are abundant inside spores, and following germination, the number of transcripts decreases until 18 h when the CDC11b transcript begins to increase in abundance. In contrast, we see a steady increase in CDC11a expression over time through 18 h (Fig. S2 B). Next, we characterized the localization of Cdc11 proteins during Ashbya development. During hyphal growth, septins can assemble into several major structures: (1) thin filaments which are dynamic, curving, and found all over the cell cortex; (2) interregion rings that consist of bundles of septin filaments aligned parallel to the long axis of the hyphal tube; (3) branch point structures where septins localize to sites of micron-scale curvature at the base of emergent hyphae that extend perpendicular to the main hyphal axis; (4) symmetrical tip-splitting structures where septins localize to the micron-scale curvature at the bases of bifurcated tips (Helfer and Gladfelter, 2006; DeMay et al., 2009; Kaufmann and Philippsen, 2009; Bridges et al., 2016; Fig. 1 A). To examine the localization of Cdc11a and Cdc11b-capped octamers at these various structures through time, we generated a strain that coexpressed Cdc11a-mCherry and Cdc11b-eGFP under the control of their native promoters and imaged using confocal microscopy. We observed robust Cdc11a localization at all septin structures from 12 to 24 h (Fig. 1, B–E). In contrast, Cdc11b localization is only visible at septin structures at 24 h. The ratio of Cdc11a- to Cdc11b-capped octamer fluorescence intensity at branch point and IR-ring structures shows a consistently higher level of Cdc11a relative to Cdc11b from 12 to 18 h, which is when all growth is through lateral branches and no tip-splitting has begun (Fig. 1, F and G). At ∼20 h, Ashbya ceases lateral branching and begins to undergo tip-splitting (Knechtle et al., 2003). Only at the 24-h timepoint the ratio of Cdc11a- to Cdc11b-capped octamers is reduced (Fig. 1, B–E).

In addition to the colocalization (overlap of fluorescent signals) of Cdc11a and Cdc11b within septin filaments at 24 h, we also observe continuous filaments that show a spatially resolvable signal for Cdc11a and Cdc11b (Fig. 1 B), suggesting that both types of octamers are capable of coassembling into a filament via end-on interactions. Interestingly, there is still remnant Cdc11a-capped octamer present at 24 h at all septin structures, including the tip-split saddle-points, suggesting that these two Cdc11-capped octamers colocalize in vivo, but the relative abundance of each protein in assemblies depends on the developmental stage of the cells (Fig. 1, F–H).

We next examined if and how the variable stoichiometries of Cdc11a/b influenced the features of higher-order septin assemblies branch points, interregion rings, and tip-splits. We find that at branch points, the average filament length decreases over time, with the shortest filament occurring at 24 h, coincident with Cdc11b localization to these structures (Fig. 1, C, F, and J). Interestingly, in contrast, when we examined septin filament length at interregion rings, we found that the average filament length increases over time, with the longest filaments occurring at 24 h (Fig. 1, D, G, and I). However, interregion rings are known to be regulated by the kinases Elm1 and Gin4 (DeMay et al., 2009), suggesting that posttranslational modification of septins could be another means to regulate filament length in these higher-order structures. Lastly, when we examined septin filament length at tip-splits, we observed that the filaments that follow the curve between the new tips were longer relative to either branch point or interregion ring filaments (Fig. 1, E, H, and K). However, it is important to note that differences in filament length observed branch points and tip-splits could also result from the varying degrees of membrane curvatures found at these sites, which have been shown to influence septin filament length (Shi et al., 2022). Moreover, due to the spatial resolution limit of our microscope, it is difficult to disentangle continuous filaments from smaller, juxtaposed filaments that run along the curved membrane. In all cases, however, measurements show a broad range of lengths presumably because the septin assemblies scored are from different aged cells and structures, and septin structures can change through time (Demay et al., 2009). Collectively, these data show that Cdc11b-capped complexes have the potential to promote and oppose filament length depending on the type of septin structure.

Next, we investigated if Cdc11a and/or Cdc11b could functionally compensate for one another. To do this, we generated deletion strains (where either CDC11A or CDC11B were deleted from the genome) and examined Ashbya cell morphologies using differential interference contrast (DIC) microscopy (Fig. 2). We find that cells lacking CDC11A exhibit multiple morphology defects at both 18 and 24 h, including hyper-branched hyphae (distance between branches = 21.73 ± 15.5 µm in wild-type cells, 13.03 ± 7.73 µm in cdc11aΔ cells, ****P < 0.0003 from unpaired t test) and long, persistent hyphae where no lateral branches emerge, as well as aberrant, asymmetric tip-splitting events (Fig. 2, D–F). Both lateral branch and tip splitting phenotypes suggest problems maintaining single axes of polarity in the absence of Cdc11A.

Cells lacking CDC11B exhibit no observable morphological defects at 18 h (Fig. 2 G); however, at 24 h, we observe aberrant tip-splitting events (Fig. 2 I) and larger hyphal diameters relative to wild-type (diameter = 4.45 ± 0.70 µm in wild-type cells, 5.75 ± 0.92 µm in cdc11bΔ cells, ****P < 0.0001 from unpaired t test). It is likely these two phenotypes are related as tip splitting is thought to emerge as a response to fast growth rates supported by a high flux of vesicles to tips (Knechtle et al., 2003). Thus, wide hyphae would result if such flux is not divided into the two split tips as normally seen. Notably, the tip-splitting defects observed in cells lacking CDC11A are different than those in cells lacking CDC11B. Tip-splitting events in hyphae cdc11aΔ are asymmetric and are wavy as opposed to straight hyphae (Fig. 2 F), whereas cdc11bΔ tip-splitting hyphae are either asymmetric or symmetric and observed to undergo tri- and quadfurcations (Fig. 2, H and I). This is in contrast to wild-type cells where tip-splitting events are always bifurcation events (Fig. 2, B and C). This data suggests that Cdc11a and Cdc11b have distinct, developmentally regulated functions despite their high degree of sequence identity.

To disentangle the degree to which specific functions were due to differences in expression timing or differences in the sequence of the two Cdc11 proteins, we generated strains that would either express two copies of CDC11A or CDC11B. This also ensures that the dose of septin protein is comparable with wild-type cells throughout development, unlike the null mutants. For these experiments, Cdc11a is expressed from both the CDC11A and CDC11B promoters (CDC11B-pr:CDC11A) and similarly for Cdc11b (CDC11A-pr:CDC11B). In cells with two copies of CDC11A, we observe normal growth relative to that of wild-type at 18 h; however, at 24 h we observe a higher frequency of tip-splitting events compared to wild-type cells (Fig. 2, J–L). Interestingly, these tip-splits (Fig. 2 L) appear to be morphologically indistinguishable from that of wild-type cells (Fig. 2 C), suggesting that CDC11A can functionally compensate for CDC11B with respect to hyphal morphology; however, it cannot properly compensate in regulating the number/frequency of tip-splitting events. Cells with two copies of CDC11B are still hyperbranched (Fig. 2 M) and retain aberrant tip-splitting phenotypes (Fig. 2, N and O) reminiscent of cdc11aΔ cells (Fig. 2 F). However, the tip-splitting phenotype is less severe than cdc11aΔ cells, suggesting that CDC11B can partially functionally compensate for CDC11A in tip splitting. Thus, Cdc11a contributes to the normal frequency and length of lateral branches in early growth and also the stable axis of polarity in tip-splitting at later stages. Loss of Cdc11b, consistent with it being expressed primarily late in development, leads to phenotypes indicating a key role in dictating the number and spacing of tip-splitting events. Collectively, these data suggest that despite their high sequence similarity, Cdc11a and Cdc11b proteins have discrete functions in septin higher-order assembly and morphogenesis throughout the growth cycle. We next look to understand the biophysical basis for these distinct cell functions.

Motivated by our measurements of filament length of septin structures in A. gossypii (Fig. 1), we set out to determine if Cdc11a or Cdc11b-capped octamers had distinctive biochemical properties such as length, membrane adsorption rate, and rigidity. To do this, we purified recombinant Ashbya septins expressed in E. coli. Septin complexes capped with either AgCdc11a or AgCdc11b form hetero-octamers and are capable of filament formation in solution (Fig. 3, A and B). Although septins can polymerize in solution in vitro, septin filament formation is dependent on septin-membrane interactions in fungal cells (Bridges et al., 2014; Bridges et al., 2016). We, therefore, reconstituted septin assembly with recombinant proteins and planar-supported lipid bilayers (SLBs; Fig. 3 C). The distribution of filament lengths for both Cdc11a- and Cdc11b-capped octamers at steady state could be approximated by a left truncated exponential function, consistent with isodesmic polymer growth models (Fig. 3 D; Romberg et al., 2001; Skillman et al., 2013; Woods et al., 2021). Interestingly, septin assembly on curved membranes has been shown to be a cooperative process, where membrane-bound septins drive the recruitment of septins from the solution onto the membrane (Shi et al., 2022). We speculate that the mechanism of septin polymerization (isodesmic vs. cooperative) is tuned by membrane geometry; however, future work is required to test this hypothesis. Fitting our data to a left-truncated exponential function (see Materials and methods for model details) allowed us to calculate an average filament length, despite the diffraction limit of the light microscope. We found that the average filament length for Cdc11a-capped octamers was 0.73 µm (∼23 octamers) and 0.52 µm (∼16 octamers) for Cdc11b-capped octamers (Fig. 3 D, ****P < 0.0001). Additionally, we observed no difference in the width of filaments containing Cdc11a- or Cdc11b-capped octamers based on fluorescence intensity (Fig. S3 A), suggesting that the presence of either Cdc11a or Cdc11b within septin octamers does not influence lateral pairing. Instead, this data suggests that the affinity for “end-on” Cdc11a-Cdc11a interactions is stronger than Cdc11b–Cdc11b interactions and is consistent with the shorter assemblies seen in a subset of Cdc11b-containing higher-order assemblies in cells.

We hypothesized that differences in affinity between neighboring Cdc11-capped complexes could influence the rate at which septins assemble onto the membrane, potentially by impacting the off-rate since filament length controls dissociation (Cannon et al., 2019). To test this hypothesis, we adsorbed either Cdc11a- or Cdc11b-capped septin octamers onto planar-supported lipid bilayers and monitored change in fluorescence intensity over time to measure rates of assembly. We observed a similar rate of increase in fluorescence intensity through time for both Cdc11a- and Cdc11b-capped octamers (Fig. 3 E). Surprisingly, however, when we examined the percent surface coverage for both types of filaments at steady-state, we found that Cdc11b-capped filaments occupy approximately twofold greater space on the membrane (Fig. 3 F). We speculate that the observed higher surface coverage by Cdc11b-capped filaments could result from a shorter average filament length, as shorter filaments might be able to more efficiently pack together on the membrane than longer filaments. Additionally, having a higher density of bound septins with shorter filaments would increase the number of available binding sites (free filament ends) for which septin octamers coming from the solution could interact, thereby potentially increasing the amount of septin bound to the membrane. Thus, Cdc11 identity did not impact the rate of assembly but can change the density of bound septins on a membrane.

The mechanical properties of cytoskeletal filaments are intimately tied to their functions. As Cdc11a or Cdc11b are the terminal subunits for filament formation, they are uniquely poised to influence the intrinsic flexibility of a septin filament. Therefore, we next characterized the persistence length (Graham et al., 2014), the average length over which filaments stay straight, for septin filaments composed of either Cdc11a- and Cdc11b-capped octamers. Interestingly, the persistence length of both classes of filaments is similar (1.38 and 1.08 µm, Cdc11a and Cdc11b complexes, respectively; Fig. 3 G), suggesting that despite apparently different affinities, the Cdc11a and Cdc11b generate filaments with similar flexibilities. When combined, these data show that the identity of septin protein in the terminal position of the oligomer can produce septins with different average lengths and packing densities on membranes, indicating that relatively small changes in the protein sequence can produce substantially different polymer behaviors.

Septins are the only identified sensor of positive micron-scale membrane curvature in eukaryotes. Previous work in our lab has utilized an in vitro reconstitution system to measure septin adsorption using positively curved SLBs and fluorescently tagged, recombinant protein through quantitative fluorescence microscopy (Bridges et al., 2016). Recombinant septins from budding yeast were shown to preferentially assemble onto spherical SLBs with a curvature, κ = 2 µm⁻¹ (corresponding to a diameter of 1 µm) when mixed with SLBs of various curvatures. We used this spherical SLB system to examine the curvature-sensing ability of Ashbya septins containing either Cdc11a or Cdc11b-capped septin octamers. When we measured adsorption for both octamer types at steady state, we found that both Cdc11a and Cdc11b have a preference for beads where κ = 2 µm⁻¹ (Fig. 4, A and B), consistent with the curvature preference for budding yeast septins (Bridges et al., 2016). Interestingly, we noticed that Cdc11b-capped octamers show a higher absorption value (∼0.8 and ∼1.0 for Cdc11a and Cdc11b, respectively, ****P < 0.001; Fig. 4 B), indicating that more of these septins are bound to the bead. This is consistent with our previous data where Cdc11b-capped octamers occupy a higher surface coverage on planar membranes relative to Cdc11a at a steady state (Fig. 3 F). Higher absorption values onto beads (diameter = 1 µm, κ = 2 µm⁻¹) at steady state could suggest that Cdc11b-capped complexes have a stronger affinity (K_d) and/or maximal binding capacity (B_max) for curved membranes than Cdc11a-capped octamers. To test if affinity could be the driving force behind the adsorption differences, we measured the accumulation of either Cdc11a- and Cdc11b-capped octamers onto 1-µm diameter beads (κ = 2 µm⁻¹) over time (Fig. S3 B). We found that there is no difference in the amount of time it takes either type of septin octamer to reach steady-state, suggesting that differences in affinity do not underlie the higher adsorption values for Cdc11b-capped octamers onto membranes. Unfortunately, we could not generate the saturation binding isotherms to calculate maximum binding values due to low protein yield. We speculate that Cdc11b-capped filaments are packed more efficiently on membranes than Cdc11a-capped filaments; however, future work using higher-resolution methods is required to test this hypothesis on curved membranes.

As Cdc11a and Cdc11b colocalize within the same structures in cells, we next investigated if recombinant septin complexes containing either Cdc11a or Cdc11b could colocalize within filaments (either through end-on or lateral interactions) in vitro in the absence of other cellular factors. First, we mixed equal amounts of Cdc11a- and Cdc11b-capped septin complexes onto planar SLBs and examined filaments using two-color TIRF microscopy. We observed filaments that contained both Cdc11a and Cdc11b septin complexes (Fig. 5, A and B), demonstrating that both classes of septin complexes are capable of colocalizing within filamentous structures. Whether this colocalization is indicative of Cdc11a- and Cdc11b-capped octamers assembling within the same single filament or two single filaments that consist of homotypic Cdc11a- and Cdc11b-capped octamer interactions that are laterally paired is difficult to determine with light microscopy. However, there are clear instances where the signal for both Cdc11a and Cdc11b is either colocalized within a filamentous structure (Fig. 5 A*) or appears to be speckled with alternating Cdc11a/Cdc11b signal (Fig. 5 A**). We suspect that Cdc11a and Cdc11b-capped octamers can form both end-on and laterally paired heterotypic filaments. Importantly, we did notice that there appeared to be a higher density of Cdc11b-capped complexes on the membrane (Fig. 5 A), despite adding equal concentrations of both septin octamers, which was consistent with our previous results (Fig. 3, C and F). Interestingly, when we measured the steady-state filament length distribution of coassembled polymers, we found that the average filament length was shorter than either Cdc11a or Cdc11b filaments alone (Fig. 5 C; and Fig. 3, C and D). This suggests that the affinity for heterotypic interactions between Cdc11a and Cdc11b is weaker than homotypic (Cdc11a-Cdc11a or Cdc11b-Cdc11b) interactions. Moreover, we observed an increase in filament persistence length (from ∼1 to ∼2 µm; Fig. 5 D), suggesting that the reduced affinity of Cdc11a for Cdc11b may manifest as a “mismatched” interaction that can impact the bending ability of septin filaments. Collectively, these data show that coassembly of these different septin complexes can lead to polymers of distinct length and flexibility compared with either type of homopolymer.

Next, we measured septin adsorption onto SLBs of various curvatures after mixing equal concentrations of Cdc11a and Cdc11b together (Fig. 5, E and F). Unsurprisingly, we saw a curvature preference for beads where κ = 2 µm⁻¹ (diameter = 1 µm) for both Cdc11a- and Cdc11b-capped octamers, consistent with our previous observations (Fig. 4, A and B). Moreover, Cdc11b complexes showed higher adsorption than Cdc11a-capped octamers (Fig. 5 F). Interestingly, we observed less total adsorption of protein onto all tested curvatures when we mixed Cdc11a and Cdc11b-capped octamers (Fig. 5 F) than we saw with either type of octamer alone (Fig. 4 B).

Much of the sequence variation between Cdc11a and Cdc11b occurs within the C-terminal extension (CTE) region of the protein (Fig. S4). Therefore, we generated Cdc11 chimeras by swapping the CTE’s of Cdc11a (residues 263–412) and Cdc11b (residues 263–414) to test if the CTE sequence was sufficient to impart the properties particular to either Cdc11a or Cdc11b. Using planar SLBs, we found that Cdc11a-Cdc11bCTE and Cdc11b-Cdc11aCTE-capped octamers had an average steady-state filament length of 0.94 µm (∼29 octamers) and 0.55 µm (∼17 octamers), respectively (Fig. 6, B and D). Interestingly, Cdc11a-Cdc11bCTE-capped octamers are longer than Cdc11a-capped octamers, whereas there is no difference in average filament length between Cdc11b-Cdc11aCTE- and Cdc11b-capped septin octamers. To rationalize why this might be the case, we measured the width of filaments for both Cdc11a-Cdc11bCTE and Cdc11b-Cdc11aCTE-capped octamers. We found that both CTE-chimera filaments are wider than Cdc11a- or Cdc11b-capped octamers (Fig. S3 A), although Cdc11a-Cdc11bCTE filaments were wider than Cdc11b-Cdc11aCTE filaments. We suspect that when fused to Cdc11a, the CTE of Cdc11b promotes tighter lateral associations that are less prone to fragmentation than any of the other complexes. Interestingly, we observed equal surface coverage for both types of polymers, despite their unequal average filament lengths (Fig. 6 C), suggesting that the density of septins on the membrane can be tuned by regulating the degree of lateral pairing between neighboring septin filaments. Notably, however, we observed that swapping the CTEs does not alter filament rigidity, as we calculated similar persistence length values for these filaments (Fig. 6, E and F).

Next, we examined if swapping the CTEs would result in any differences in curvature sensitivity. We found that both chimeras still showed a curvature preference for 1-µm beads (Fig. 6, G and H). However, we did notice an increase in total adsorption onto the beads for both chimeras relative to their wild-type Cdc11a- and Cdc11b-capped octamers (Fig. 6, F and G). Collectively, these data show that despite the CTE sequences harboring most of the sequence variation between the two Cdc11 paralogs, they are insufficient to recapitulate the biochemical/biophysical features of each Cdc11 polypeptide. We think the most likely explanation for this lack of transferability is due to potential unknown differences in Cdc11 protein structure introduced by the specific chimeric fusions or that the biophysical differences arise from changes in other parts of the protein sequence.

Given that the CTEs were not sufficient to interconvert the function of the different Cdc11 proteins, we examined the predicted structure of the globular portion of the protein to search for any differences between these two proteins. We threaded both Cdc11a and Cdc11b through Phyre 2 (Kelley et al., 2015b) and overlaid the predicted structures (Fig. 7 B). Despite most of the globular core being highly conserved among both gene products, we found a single sequence variation located just before a prominent difference in predicted structure (Fig. 7, A and B). Specifically, Cdc11a contains a threonine at position 62, whereas Cdc11b contains an alanine. Interestingly, the Cdc11a sequence flanking this substitution is predicted to be more helical, whereas the structure within this region of the Cdc11b polypeptide is predicted to be disordered. With these potential structural differences in mind, we purified Cdc11aT62A-capped complexes and examined the steady-state filament length distribution on planar SLBs. Remarkably, we observed that the average filament length of these complexes was dramatically reduced (L_avg = 0.44 ± 0.016 µm) when compared with wild-type Cdc11a-capped octamers (L_avg = 0.73 ± 0.04 µm), and much closer to the average filament length of wild-type Cdc11b-capped octamers (L_avg = 0.52 ± 0.01 µm; Fig. 7, C and D). Moreover, the surface coverage of Cdc11aT62A-capped octamers was similar to Cdc11b-capped octamers (Fig. 7 E). To test the effects of this mutant in Ashbya cells, we generated a strain where CDC11A was replaced at its endogenous locus and promoter by a gene encoding the Cdc11 aT62A mutation and examined cell morphology using DIC microscopy. Remarkably, cells grown for either 18 or 24 h exhibited hyperbranched and aberrant tip-split morphologies (Fig. 7 F) that phenocopy cells lacking Cdc11a (CDC11A-pr: CDC11B and cdc11aΔ; see Fig. 2). In sum, these data show that even a single amino acid difference can influence the biophysical properties of septin filaments, including their length and density on the membrane, which in turn can influence whole cell morphology.

In this study, we investigated the effects of having two similar, but distinct, terminal subunits (Cdc11a and Cdc11b) on septin filament properties using in vitro reconstitution, molecular genetics, and in vivo imaging. Our data demonstrate that small changes in individual septin subunits can lead to distinct roles in morphogenesis and higher-order assembly sizes in cells and alter septin filament lengths, which in turn can influence the kinetics of septin assembly onto membranes. Collectively, our data highlight how cells might utilize different pools of available septin polypeptides to alter their biochemical/biophysical properties of the septin cytoskeleton to best suit downstream functions.

All known fungi with budding yeast-like genomes, even those that underwent a genome duplication event, carry a single copy for each septin. Interestingly, Ashbya species are the exception by carrying two copies of CDC11 (Gattiker et al., 2007). Based on the current synteny of the loci, after the duplication event of the CDC11 locus in the Ashbya ancestor, a DNA double-strand break presumably occurred between the two CDC11 copies that were repaired through a complex gene fusion event (Fig. S1 A). This duplication was followed by a complex double-strand repair that presumably occurred relatively recently in the A. gossypii evolutionary history because the only other sequenced species closely related to A. gossypii with a second syntenic copy of CDC11 is Ashbya aceri (Fig. S1 B). The A. aceri Cdc11a protein is 100% identical to Cdc11a from A. gossypii, while the A. aceri Cdc11b shares 95% identity with A. gossypii Cdc11b (Fig. S1 C). What then makes Ashbya unique? Ashbya has evolved a lifestyle of a filamentous fungus with continuously growing and branching hyphae. The closely related Eremothecium cymbalariae, Eremothecium sinecaudum, and Eremothecium coryli carry one CDC11 gene, yet also form hyphae. However, none of these three species exhibit the two different branching patterns seen in Ashbya; lateral branching (low-speed growth, 0.2 µm/min for emergent germling hyphae) and symmetrical tip-splitting events (faster growth, 1.5–3.5 µm/min within 18–24 h; Köhli et al., 2008). It is important to note that lateral branches are the predominant mode of growth early in Ashbya development (Knechtle et al., 2003), whereas tip-splitting events begin in later stages of growth (∼18–20 h; Ayad-Durieux et al., 2000). The genome rearrangements leading to two CDC11 copies did not involve upstream sequences of CDC11A; however, the promotor region of CDC11B underwent a rearrangement potentially affecting its regulation. Indeed, we see that CDC11A is expressed at all developmental growth stages, whereas the expression of CDC11B is restricted to the fast tip-splitting phase (Fig. S2). This is consistent with the deletion and replacement data that suggest Cdc11a is important in lateral branching and tip-splitting. In contrast, Cdc11b plays a minor role in lateral branching and is more important later in development for tip-splitting events. Interestingly, cells with two copies of CDC11B still show asymmetric tip-splitting events, suggesting that CDC11A and CDC11B both contribute, yet have distinct functions at these sites.

How do septins contribute to tip-splitting? This is a fascinating morphological transition that is in major contrast to budding yeast where there is strict singularity in polarization sites, with competition leading to a single, winning site (Howell et al., 2009). We speculate that septin filament length/density at tip-splits must be precisely regulated to control the localization and restrict the diffusion of polarity proteins including Cdc42, Bni1, and Spa2 to discrete sites to promote symmetry breaking (Seiler and Plamann, 2003; Schmitz et al., 2006; Kelley et al., 2015a). It is conceivable that the septin assembly that is at the saddle point between tips helps insulate the two tips. Somehow the use of two different septins with different lengths and membrane binding supports faster growth rates and larger hyphal diameters that are coincident with the transition of lateral growth to tip-splits (Knechtle et al., 2003). The morphological defects observed during tip-splitting in either deletion strain could be attributed to the role of septins in cell wall deposition (DeMarini et al., 1997) and lipid metabolism (Mela and Momany, 2021) through the recruitment of chitin synthases or by the septins themselves providing a mechanical force to counterbalance turgor pressure and cortical tension (Gilden and Krummel, 2010). It is possible that when the normal mixture of a/b complexes is missing, more than two polarity axes emerge in a small area.

Our in vitro data show that when Cdc11a- and Cdc11b-capped complexes were seeded onto planar SLBs, Cdc11a-capped complexes formed filaments ∼1.5× longer than Cdc11b-capped complexes. These data suggest that the affinity between Cdc11a and Cdc11a subunits is higher than that between Cdc11b and Cdc11b subunits. However, we were surprised when we saw that Cdc11b-capped complexes showed a higher density of membrane-bound septin than Cdc11a-capped complexes. Our previous data show that single septin octamers are unable to stay bound (high off rate) to the membrane, suggesting that septins must form filaments (to effectively lower the off rate) to remain associated with the membrane (Bridges et al., 2016; Cannon et al., 2019). Thus, we thought that septin complexes that could form longer filaments would result in a higher density of septins on the membrane. In contrast, we see that Cdc11b-capped complexes, despite forming shorter filaments, show a higher degree of membrane binding than Cdc11a-capped complexes. We speculate that the interaction strength of Cdc11b–Cdc11b interactions is strong enough to form filaments that are capable of staying bound to the membrane (an effective off rate that is low). Furthermore, because the affinity of Cdc11b–Cdc11b interactions is weaker than Cdc11a–Cdc11a interactions, this could allow a higher number of shorter, but stably bound filaments to form on the membrane, effectively increasing the number of binding sites (free filament ends) for which single octamers, either coming from the bulk solution or diffusing on the membrane, could interact. This, in turn, would increase the density of septins on the membrane (Fig. 8). Thus, the shorter filaments formed by Cdc11b have essentially a comparable and very low off rate as the filaments formed from Cdc11a, but by being shorter, there are more binding sites at the ends to enhance recruitment of new subunits, yielding a higher density of septins bound to the membrane. However, this might not be due to filament length alone as Cdc11a–Cdc11bCTE-capped complexes show similar membrane binding densities despite having different filament lengths. Interestingly, CTE regions of septins have been shown to influence septin filament pairing (Bertin et al., 2010). When we measured the width of Cdc11a-Cdc11bCTE- and Cdc11b-Cdc11aCTE-capped complexes, we observed an increase in the width of these filaments with respect to their wild-type counterparts. These data suggest that the density of septins bound to the membrane may be further influenced by filament pairing. We speculate that as single septin filaments polymerize on the membrane, they can serve as a template for single septin octamers coming from the bulk solution to polymerize. This would potentially lower the high off rate of single septin octamers on the membrane by offering an additional point of contact with a neighboring filament. However, it is unclear why these chimeric CTE complexes show an increase in filament width with respect to the filaments formed from wild-type Cdc11a- or Cdc11b-capped octamers. It is possible that the nature of the fusion has oriented the CTEs in such a way that it induces new or stronger associations leading to longer filaments. Moreover, it was surprising to us that despite wider filaments, we saw no changes in persistence length when we compared wild-type filaments with CTE-swapped filaments. We suspect that we are not able to spatially resolve small fluctuations in filament bending due to the diffraction limit. Future work will have to use higher-resolution methods to examine the molecular basis for the biophysical properties of these mutants.

Interestingly, we found that Cdc11a- and Cdc11b-capped octamers are capable of copolymerizing into filaments (both end-on and lateral interactions) that show different biochemical/biophysical properties than filaments arising from either type of octamer alone. We saw that copolymerized filaments showed an increase in filament rigidity relative to filaments capped with either complex alone, highlighting the importance of the terminal subunit in controlling filament flexibility. It should be noted, however, that the persistence length values reported here are lower than those previously reported for septins. However, we are confident in our values as we have measured a higher number of filaments than in our previous reports (Bridges et al., 2014; Khan et al., 2018) and used both the end-to-end and cosine correlation analysis (data not shown) methods to calculate persistence length (Graham et al., 2014), which yielded similar values.

When Cdc11a- and Cdc11b-capped complexes are mixed at a 1:1 ratio, the average filament length at steady-state and total adsorption onto membrane curvature is decreased relative to Cdc11a- or Cdc11b-capped complexes. Consistent with this, the average filament length at branch points decreases as Cdc11b expression increases. In contrast, as Cdc11b expression increases, we observe longer filaments at IR rings. However, IR rings are known to be regulated by the kinase Elm1 (DeMay et al., 2009), suggesting that septin filament length can be further regulated by posttranslational modifications. As a major function of septins is to serve as a scaffold, we suspect that by controlling the local concentration of septins (through regulating filament formation) at the membrane, the cell can subsequently tune the concentrations of septin-interacting proteins that promote processes such as polarized growth, cell wall deposition, and mitotic progression. Future work should determine the precise stoichiometry between Cdc11a- and Cdc11b-capped octamers and how these relative levels regulate the total septin concentration on the membrane.

The septin superfamily is thought to have emerged from ancient gene duplications that gave rise to the different classes of septins shared across many eukaryotes (Pan et al., 2007; Auxier et al., 2019). This recent duplication of a terminal septin subunit reveals how small changes in the primary sequence can lead to substantial biophysical differences for the polymer providing a glimpse at how functional diversification may have occurred in this gene family.

Plasmid AGB141 (CDC11A-mCherry-NAT) was generated by yeast homologous recombination in S. cerevisiae. An mCherry-NAT fragment was amplified from AGB048 using AGO190/AGO299 and integrated into AGB125 by selection on YPD containing 50 µg/ml clonNAT (EMD Millipore). Plasmid AGB1320 (CDC11B-GFP-G418) was generated by yeast homologous recombination. A GFP-G418 fragment was amplified from AGB005 using AGO2823/AGO2824 and integrated into AGB124 by selection on YPD containing 200 µg/ml G418 (Calbiochem). Plasmid AGB214 (CDC11A-GFP-G418) was generated via yeast homologous recombination. The GFP-G418 cassette was PCR amplified with AGO540/AGO524 from AGB005 and integrated into AGB141 by selection on G418. Plasmids AGB1369 (cdc11A::G418) and AGB1370 (cdc11B::G418) were both generated via yeast homologous recombination. The G418 resistance cassette was PCR amplified with end homology using either AGO3326/AGO3327 (CDC11A) or AGO3324/AGO3325 (CDC11B) from AGB005. The resulting fragment was integrated into EcoNI digested AGB125 (for CDC11A) or SnaBI digested AGB124 (for CDC11B) and selected on G418. Plasmid AGB1374 (cdc11a::CDC11B-GFP-NEO) was generated via Gibson Assembly (NEB HiFi DNA Assembly Master Mix) in a stepwise fashion. First, a 3′ fragment of CDC11A was PCR amplified from AGB125 using AGO3336/AGO3337 and cloned into BamHI/XbaI digested pUC19. Second, a fragment containing CDC11B-GFP-G418 was PCR amplified from AGB1320 using AGO3338/AGO3339 and cloned into BamHI digested plasmid generated from the first step. Finally, a 5′ fragment of CDC11A was PCR amplified from AGB124 using AGO3340/AGO3341 and cloned into ApaI digested plasmid resulting from the second step. All steps were selected on ampicillin. AGB1352 (cdc11b::CDC11A-GFP-G418) was generated by Gibson Assembly. PCR fragments were amplified using AGO3285/3286 (5′ upstream CDC11B), AGO3287/AGO3288 (CDC11A-GFP-G418) and AGO3289/AGO3290 (3′ upstream CDC11B) from AGB124 (for 5′ and 3′ CDC11B) or AGB214 (for CDC11A-GFP-G418). The three fragments and BamHI digested pUC19 were assembled and selected on ampicillin. AGB1502 (cdc11a-T62A-mCherry-NAT) was made by restriction-mediated cloning. A PCR fragment containing the T62A mutation was amplified with AGO1942/1771 from AGB1481. The resulting fragment was digested with BclI/SgrI and ligated into BclI/SgrI-digested AGB141 and selected on ampicillin.

Septin expression plasmids were generated as follows: AGB113 is the AgCDC10 ORF TA cloned into pCR2.1 (Invitrogen). AGB120 is the AgCDC12 ORF TA cloned into pCR2.1. AGB121 is the AgCDC12 ORF from NotI/EcoRI digested AGB120 cloned into NotI/EcoRI digested pET-Duet (Novagen). AGB130 (pET-Duet1 AgCDC12 AgCDC10) was made by ligating an AgCDC10 NdeI/Kpn1 fragment from AGB113 into NdeI/KpnI digested AGB121. AGB827 (pET-Duet ScCDC12 AgCDC10) was made by ligating BsrGI/BstEII fragment from AGB130 into BsrGI/BstEII digested AGB401, creating a hybrid S. cerevisiae/A. gossypii expression vector. AGB116 (pACYC CDC3) was made by ligating a BamHI/Not1 restriction fragment containing CDC3 from AGB127 into BamHI/NotI digested pACYC-Duet1 (Novagen). AGB119 was made by PCR amplification of CDC11A from AGB125 using primers AGO128/129. The resulting fragment was digested with NdeI/XhoI and ligated into NdeI/XhoI digested AGB116. Plasmid AGB1281 was made by PCR amplification of CDC11B from AGB214 using primers AGO2823/2824. The resulting fragment was digested with NdeI/XhoI and ligated into NdeI/XhoI AGB119. C terminal extension (CTE) swaps were achieved via fusion PCR followed by restriction enzyme-dependent cloning. For the AGB1505 (Cdc11a-Cdc11b CTE) construct, the CDC11A N terminus was amplified from AGB119 using AGO3069/3103, and the CDC11B CTE was amplified from AGB1281 using AGO3102/2826. Fusion PCR was performed using AGO3069/2826, and the resulting fragment was digested with NdeI/XhoI and ligated into NdeI/XhoI AGB119. For the AGB1506 (Cdc11b-Cdc11a CTE) construct, CDC11B N-terminus was amplified from AGB1281 with AGO2825/3103, and the CDC11A CTE was amplified using AGO3070/3102. Fusion PCR was performed using AGO2825/3070, and the resulting fragment was digested with NdeI/XhoI and ligated into NdeI/XhoI AGB1281. PCR and fusion reactions were done using Phusion polymerase (NEB). AGB1481 (CDC3 Cdc11a T62A) was generated via site-directed mutagenesis (Q5 Site Directed Mutagenesis Kit, NEB) using AGO3135/3136 in AGB119.

All cloning transformations were done in DH5alpha chemically competent cells (NEB). All plasmids were confirmed by restriction digest and underwent Sanger sequencing to confirm the fidelity of the target gene sequence. For a list of strains and plasmids used in this study, see Tables 1 and 2. For a full list of oligonucleotides used in this study, please see Table 3.

A. gossypii strains used in this study can be found in Table 1. Description of the plasmids and oligos used to generate the different strains can be found in Table 2 and Table 3, respectively. To create AG230 (CDC11A-mCherry-NAT), a 4,949-bp DNA fragment was obtained by digesting AGB141 with HindIII/NheI, electroporated into A. gossypii strain AG416 (ΔlΔt; Altmann-Jöhl and Philippsen, 1996), and grown on agar media containing 100 μg/ml clonNAT (EMD Millipore). Transformants were verified by PCR and spores isolated to obtain homokaryons. Strain AG 973 expressing both CDC11A-mCherry-NAT and CDC11B-GFP-NEO was constructed by transforming AG230 with a 7,678-bp fragment from PciI/NotI/PspXI-digested AGB1320 and selecting in a media containing 100 μg/ml G418 (Calbiochem). Spores were isolated to obtain homokaryons (AG974.5, AG976.2). It was noted that not all homokaryotic spores within an isolate showed GFP fluorescence. Deletion strains of CDC11A and CDC11B were made by transforming AG416 (WT) with either a 4,480-bp fragment or a 4,000-bp fragment from Xba1/AgeI/NcoI-digested AGB1369 and AGB1370, respectively, and grown on G418 selection (AG969, AG970). These CDC11 knockout strains were verified by PCR; however, no homokaryotic spores were able to be recovered. To make a strain in which the CDC11A ORF was swapped for the CDC11B ORF, AGB1374 was digested with ApaI/SbfI, transformed by electroporation into WT Ashbya, and grown on G418 selection (AG971). To make a strain in which the CDC11B ORF was swapped for the CDC11A ORF, a PCR fragment was amplified using template AGB1352 and AGO3285/AGO3290, transformed by electroporation into WT Ashbya, and grown on G418 selection (AG972). Isolates of both CDC11 swapped strains were verified by PCR. Recovery of homokaryotes was not attempted. To create a strain that expresses CDC11A-mCherry with the T62A mutation, AGB1502 was digested with HindIII/XmnI, transformed via electroporation into AG416, and grown on a selective media containing clonNAT. Recovery of homokaryotes was not attempted.

We utilized Treehouse to generate a subtree of the Ashbya lineage (Steenwyk and Rokas, 2019). This tree was based on a previously described Saccharomycotina yeast phylogeny (Shen et al., 2018). National Center for Biotechnology Information Blast was used to identify homologs of CDC11 in the different species.

Transcription data were obtained from the Ashbya Genome Database. In short, highly purified spores from the ∆l∆t laboratory strain were used, and all the data were obtained in duplicate. For the time course experiment, spores were inoculated into AFM broth and samples were acquired at each time point. The center represents a region enriched for sporulation, while the rim represents a region of high-speed hyphal growth via tip-splitting.

Bl21 (DE3) E. coli cells were transformed using a duet construct expression system (AGB827 and AGB1281 or 1481 depending on desired purification) using ampicillin and chloramphenicol selection (Bridges et al., 2014). Selected transformants were then grown overnight in 25 ml of Luria broth (LB) with ampicillin and chloramphenicol selection at 37°C while shaking. 1/60 of LB liquid cultures were added to 1 liter of terrific broth with ampicillin and chloramphenicol selection and were grown to an O.D.600 nm between 0.6 and 0.8. Upon reaching the appropriate O.D.600 nm, 1 mM of isopropyl-B-D-1-thiogalactopyranoside was added to the cultures to begin induction. To achieve a stoichiometric septin complex of Ashbya septins, we use Cdc12 with a primary sequence derived from S. cerevisiae, while all other subunits were Ashbya-derived sequences. Induced cultures were grown for 18 h at 18°C while shaking before harvesting by centrifugation at 10,000 elative centrigual force for 15 min. Pellets were resuspended in lysis buffer (1 M KCl, 50 mM Hepes pH 7.4, 40 mM imidazole, 1 mM MgCl2, 10% glycerol, 0.1% Tween-20, 1 mg/ml lysozyme, and a 1× protease inhibitor tablet (Roche) at 4°C for 30 min to generate cell lysates. Cell lysates were then sonicated on ice for 10 s every 2 min. Lysates were clarified by spinning at 20,000 RPM for 30 min using an SS-34 rotor at 4°C. The supernatant was passed through a 0.44-um filter and then incubated with 2 ml of equilibrated cobalt resin at 4° for 1 h. The lysate-resin mixture was then added to a gravity flow column. Following the initial flow-through of unbound lysate, the resin was washed 4× (50 ml each wash) using wash buffer (1 M KCl, 50 mM Hepes pH 7.4, 40 mM imidazole, 5% glycerol). Bound protein was then eluted using elution buffer (500 mM imidazole, 300 mM KCl, 50 mM Hepes pH 7.4, 5% glycerol) and then dialyzed into septin storage buffer (300 mM KCl, 50 mM Hepes pH 7.4, 1 mM BME, and 5% glycerol) for 24 h using two steps. Protein purity was determined via SDS-PAGE and protein concentration was determined using a Bradford assay.

Proteins were diluted to 50 nM (in buffer containing 10 mM MOPS, 2 mM MgCl₂, 0.1 mM EGTA, pH 7.0) with 50 or 300 mM NaCl as indicated in the text. 3 h after dilution, samples were applied to UV-treated, carbon-coated EM grids and stained using 1% uranyl acetate. Micrographs were recorded on a JEOL 1200EX microscope using and AMT XR-60 CCD camera at a nominal magnification of 40,000×. Rod-shaped particles were picked manually for each dataset (n = 2,300 for CDC11a, n = 1,777 for CDC11b. Reference-free image alignments and classifications were conducted using SPIDER software. Each aligned dataset was classified into 100 classes using K-means classification and 30 example class averages were selected to produce each montage shown in Fig. S1. The whole micrographs shown in Fig. S1 are bandpass filtered between 2 and 40 pixels for clarity (FIJI FFT bandpass filter).

75 mole percent of dioleoylphosphatidylcholine (DOPC) and 25 mole percent phosphatidylinositol (soy) were mixed in chloroform in a glass cuvette. For bead assays, 0.05 mole percent of rhodamine-phosphotitdyl-ethanolamine (Rh-PE) was added to the above mixture. Lipid films were made by evaporating the chloroform using argon gas followed by an overnight incubation under negative vacuum pressure. The following day, the lipid films were hydrated using an aqueous supported lipid bilayer (SLB) buffer (150 mM KCl, 20 mM Hepes pH 7.4, and 1 mM MgCl₂) to a final concentration of 5 mM. Hydrated lipid films were vortexed for 10 s and allowed to sit at 37°C for 5 min. This process of vortexing and incubation at 37°C was repeated five more times (or until the lipid film was hydrated). The hydrated film was subject to water bath sonication at 2-min intervals until the opaque solution became transparent (vindication of small unilamellar vesicle formation).

Planar-supported lipid bilayers were prepared by plasma (oxygen) treatment of glass coverslips (PE25-JW; Plasma Etch) on high power for 15 min. Reaction chambers were prepared on the plasma-treated glass using a previously established method (Bridges and Gladfelter, 2016). 1 mM of small unilamellar vesicles was added to the reaction chamber (using SLB buffer) followed by the addition of 1 mM CaCl₂, followed by incubation at 37°C for 20 min to promote bilayer formation. After bilayer formation, excess small unilamellar vesicles were washed vigorously with 150 µl of SLB buffer six times. It is essential not to touch the cover glass during this step as bilayer continuity can be affected. Prior to adding septins to the membrane, the bilayers are washed again with 150 µl of reaction buffer to lower the salt to promote septin binding (50 mM Hepes pH 7.4, 0.13 mg/ml BSA, 1 mM β-mercaptoethanol). This step is repeated five more times. In the last wash, 125 µl of the reaction buffer was removed. 25 µl of septins at the desired concentration were added to the bilayer and imaged by total internal reflection fluorescence microscopy using a Nikon TiE TIRF system equipped with a solid-state laser system (15 mW, Nikon LUn-4), a Nikon Ti-82 stage, a 100× Plan Apo 1.49 NA oil lens, and Prime 95B CMOS camera (Photometrics).

Supported lipid bilayers on silica microspheres were generated as previously reported by (Bridges et al., 2016). 50 nM of small unilamellar vesicles (SUVs; 75% DOPC, 25% Soy PI, 0.05% Rh-PE) were added to silica microspheres of various membrane curvatures at a total surface area of 440 mM². Note: The total surface area of each bead size is equal. SUVS were incubated with the microspheres for 1 h at room temperature on a rotator to induce bilayer formation. After bilayer formation, the microspheres were spun down each bead size at the minimal sedimentation velocity for each bead (it is important not to exceed this number as bilayer continuity can be affected). 50 µl of the supernatant was removed and discarded. 200 µl of prereaction buffer (33.3 mM KCl and 50 mM Hepes pH 7.4) was used to resuspend/wash the beads. The microspheres were spun down again (at the appropriate sedimentation velocity) and 200 µl of supernatant was removed. 200 µl of fresh prereaction buffer was added to resuspend/wash the microspheres. This process was repeated three more times. Washed microspheres were then mixed at a 1:1 ratio. 29 µl of this mixture was added to 721 µl of reaction buffer (33.3 ml KCl, 50 mM Hepes pH 7.4, 0.1% methyl cellulose, 0.13 mg/ml BSA, 1 mM β-mercaptoethanol). 75 µl of this mixture was added to a reaction chamber glued to a polyethylene glycol-coated coverslip (Bridges et al., 2016; Cannon et al., 2019). 25 µl of septins at a desired concentration were added to the reaction chamber and allowed to reach steady state (1 h) and were imaged using spinning disc confocal microscopy.

Images of fluorescent-tagged septins adsorbed onto curved supported bilayers on microspheres were acquired using a spinning disc (Yokogawa W1) confocal microscope (Nikon Ti-82 stage) using a 100× Plan Apo 1.49 NA oil lens and a Prime 95B CMOS camera (Photometrics). Images were analyzed using Imaris 8.1.2 software (Bitplane AG) as previously described (Bridges et al., 2016; Cannon et al., 2019).

10 locations on a given lipid bilayer were blindly selected to form 10 separate images. Each image was background subtracted. The total number of pixels with a septin signal was divided by the total number of pixels in the image to get the fraction of septins bound to the surface. This number was multiplied by 100 to get a percent surface coverage.

Filament lengths were measured by uploading raw images to FIJI and using the Ridge detection (Steger, 1998) plugin. After image segmentation and processing, septin length distributions are extracted from each field of view. We are unable to resolve the frequencies of the smallest septin filaments due to the diffraction limit of light (∼200 nm) and the small size of septin octamers (∼32 nm) and are thus left with an incomplete length distribution. Therefore, an arithmetic average of the observed lengths will not be an accurate estimate of the true mean length of the population, and we must use a model to extrapolate to the true length distribution. Observation, physical models of septin polymerization, and robust model fits suggest that this distribution is a left-truncated exponential. Since we have incomplete data, we utilize a convenient property of exponential distributions to obtain model fits and estimate the population mean length.

An exponential length distribution has the following PDF: f(x) = λexp(−λx), where 1/λ is the mean length. If we let X be an exponentially distributed random variable, i.e. X∼Exp(X), then X is memoryless, which means that: P(X>x+a|X>a) = P(X>x), for some cutoff value a. In practice, this implies that if the true septin length population is exponentially distributed with some mean length 1/λ, then our left-truncated data will be described by the same λ. To fit, we simply choose a cutoff value a below which data is ignored, and this value is subtracted from the observed lengths and the fit parameter λ is obtained To ensure robustness of fits, fit quality for a small range of cutoff values is assessed, and the smallest cutoff value that corresponds to a stable value of the fit parameter, λ, is the one used for the final model fit. Cutoff values are close to 200 nm. The true length distribution is related to the left-truncated data by the scaling factor exp(λa).

Persistence length measurements were calculated from raw TIRF microscopy images of septin filaments seeded onto planar-supported lipid bilayers using a previously published MATLAB GUI method (Graham et al., 2014).

A. gossypii spores were inoculated into a full medium at 30°C for either 12, 16, 18, or 24 h before harvesting mycelial cells. Harvested cells were washed 3× using low fluorescence media. Cells were subsequently mounted onto low fluorescence medium-based agar pads (2% agar). Images were acquired using a spinning disc (Yokogawa) confocal microscopy (Nikon Ti-82 stage), a 100× 1.49 NA oil lens, and a 95B Prime sCMOS camera (Photometrics). A. gossypii deletion and replacement strain images were acquired using a wide-field microscope equipped for differential interference contrast imaging. (Nikon Ti-82 stage, a 40× air Plan Apo 0.95 NA objective using a 95B Prime sCMOS camera (Photometrics).

Filament lengths were measured by merging both Cdc11a and Cdc11b channels. The channels were then displayed in the same lookup table (monochromatic) to ensure that the entirety of the filament was measured using ImageJ. Measurements of filament lengths were made manually. For ratiometric analysis, images were background subtracted (in both channels individually). Summed fluorescence intensity was exacted at each of these sites. The summed fluorescence intensity sum of the Cdc11a channel was then divided by the summed fluorescence intensity of the Cdc11b channel. Measurements from all structures were plotted using PlotsofData (Postma and Goedhart, 2019). Hyphal diameters and the distance between branches were measured from DIC images of various strains using ImageJ.

In all cases, unless otherwise stated, unpaired, two-sided t tests were used to determine statistically significant differences in quantified data. Unless otherwise stated, the data distribution was assumed to be normal but was not formally tested. Calculations of standard deviation and error were performed in Microsoft Excel and P values were determined using Prism (Graphpad).

Fig. S1 shows the schematics of the molecular origin of CDC11a and CDC11b as well as their sequence similarity. Fig. S2 shows CDC11a and CDC11b transcription over the course of A. gossypii’s lifecycle. Fig. S3 highlights additional biophysical and biochemical properties of septin octamers capped with either Cdc11a or Cdc11b. Fig. S4 shows a sequence comparison between Cdc11a and Cdc11b.

We would like to thank members of the Gladfelter lab for lively and stimulating discussions around septins. Kevin would like to thank Joanne Ekena for having perseverance, patience, grace, and the heart of a champion. We appreciate significant insights from Peter Philippsen on the genome rearrangements that likely gave rise to the current genomic context of the duplicated septins as well as his extensive insights and thoughts about the morphology of Ashbya cells. Thank you to the National Heart, Lung, and Blood Institute Electron Microscopy Core Facility.

This work was supported by National Science Foundation MCB-1615138 and MCB-2016022, and a Howard Hughes Medical Institute Faculty Scholars award to A.S. Gladfelter. K.S. Cannon was supported in part by a grant from the National Institute of General Medical Sciences under award T32 GM 119999. J.M. Vargas-Muniz was supported by National Institutes of Health Training Grant 5T32AI052080-14.

Author contributions: Conceptualization: A.S. Gladfelter, K.S. Cannon, and J.M. Vargas Muniz. Methodology: A.S. Gladfelter, K.S. Cannon, J.M. Vargas Muniz, N. Billington, I. Seim, J. Ekena, and J.R. Sellers, Investigation: A.S. Gladfelter, K.S. Cannon, J.M. Vargas Muniz, N. Billington, I. Seim, J. Ekena, and J.R. Sellers, Visualization: K.S. Cannon, J.M. Vargas Muniz, N. Billington, and I. Seim. Funding Acquisition: A.S. Gladfelter and J.R. Sellers. Project Administration: A.S. Gladfelter. Supervision: A.S. Gladfelter. Writing: Original draft: K.S. Cannon. Writing: Review and editing: A.S. Gladfelter, K.S. Cannon, J.M. Vargas Muniz, N. Billington, J. Ekena, and J.R. Sellers.