Interface integrity in septin protofilaments is maintained by an arginine residue conserved from yeast to man

Abstract

The septins are conserved, filament-forming, guanine nucleotide binding cytoskeletal proteins. They assemble into palindromic protofilaments which polymerize further into higher-ordered structures that participate in essential intracellular processes such as cytokinesis or polarity establishment. Septins belong structurally to the P-Loop NTPases but, unlike their relatives Ras or Rho, do not mediate signals to effectors through GTP binding and hydrolysis. Biochemical approaches addressing how and why septins utilize nucleotides are hampered by the lack of nucleotide-free complexes. Using molecular dynamics simulations, we determined structural alterations and intersubunit binding free energies in human and yeast septin dimer structures and in their in silico generated apo forms. An interchain salt bridge network around the septin unique β-meander, conserved across all kingdoms of septin containing species, is destabilized upon nucleotide removal, concomitant with disruption of the entire G-interface. Within this network, we confirmed a conserved arginine residue, which coordinates the guanine base of the nucleotide, as the central interaction hub. The essential role of this arginine for interface integrity was experimentally confirmed to be conserved in septins from yeast to human.

• Septins are bound to nucleotides when they are incorporated in filaments. The role of the nucleotide in protofilament formation is unknown.

• The authors identified a species-overlapping conserved arginine residue that coordinates the nucleotide. Mutation of this amino acid results in failure of filament incorporation of human and yeast septins in vitro and in vivo.

• Properly assembled septin structures are indispensable for essential intracelllar processes. Understanding the mechanisms of protofilament formation is important to explain septin-related phenotypes and disease onset.

INTRODUCTION

Septins are a component of the cytoskeleton (Mostowy and Cossart, 2012). Individual septin subunits form hexameric or octameric linear protofilaments termed rods, serving as building blocks for higher-ordered structures such as filaments, gauzes, or rings (Bertin et al., 2008; Garcia et al., 2011; Schneider et al., 2013; Kaplan et al., 2015). These assemblies are required for essential intracellular processes such as cytokinesis, polarity establishment, or cellular adhesion (Woods and Gladfelter, 2021).

Septins belong to the GTPase superclass of the P-loop NTPases (Leipe et al., 2002). The structural topology of the septin GTPase (or short G-) domain resembles a modified Rossmann fold containing all features for GTP hydrolysis such as the P-loop and the switch 1 and switch 2 motifs (Vetter and Wittinghofer, 2001; Cavini et al., 2021; Grupp and Gronemeyer, 2022). However, some septins do not hydrolyze their bound nucleotide, classifying them as pseudo-GTPases (Stiegler and Boggon, 2020; Hussain et al., 2023). The G-domain is decorated with N- and C-terminal extensions of various length, ranging from a few (C-terminal extension of Cdc10 from yeast) to several hundred amino acids (N-terminal extension of human SEPT9). Another septin-specific structural element of the G-domain is the septin unique element (SUE), consisting of a three-stranded β-meander (SUE-βββ) (Sirajuddin et al., 2007; Cavini et al., 2021). Septin monomers assemble into apolar, palindromic protofilaments by alternating interactions via their N- and C-termini (NC-interface) and via adjacent G-domains (G-interface), respectively (Sirajuddin et al., 2007; Bertin et al., 2008; Mendonça et al., 2021; Grupp et al., 2024), with the SUE-βββ being a central element in G-interface formation.

Phylogenetic analyses grouped animal, fungi, and protist septins in five groups (Pan et al., 2007; Auxier et al., 2019; Shuman and Momany, 2022) and recently, two more groups were assigned for septins from algae and one additional for ciliates (Delic et al., 2024). In all eight groups conserved residues in the GTPase specific motifs and the SUE can be found whereas residues in the NC-interfaces are conserved in groups 1-6a but poorly conserved in the newly assigned groups 6b-8 (Delic et al., 2024). This finding led to the assumption that nucleotide binding and hydrolysis is an evolutionary old feature of the septins. What is the role of nucleotide interaction and hydrolysis in septins? Unlike their relatives, the small GTPases from the Ras family, septins do not mediate signals by binding to effector molecules in their active, GTP-bound state (Wittinghofer and Vetter, 2011). Experimental data suggest that nucleotide binding and hydrolysis is required for protofilament formation rather than for signal transduction (Sirajuddin et al., 2009; Zent et al., 2011; Weems et al., 2014; Zent and Wittinghofer, 2014; Weems and McMurray, 2017). However, except for the yeast septin Cdc11, structures of monomeric septins are not yet available (Brausemann et al., 2016) and all available dimeric or multimeric septin structures are nucleotide bound. These limitations hampered mechanistical studies aiming at solving the long-standing question how septins employ nucleotides in the process of protofilament formation although such a role was postulated already more than a decade ago (Sirajuddin et al., 2009).

Biochemical processes inaccessible to experimentation can sometimes be approached by in silico methodology. If sufficient structural information about the protein of interest is available, molecular dynamics (MD) simulations (Zheng et al., 2018) can be employed to study for example substrate translocations through channel proteins (Göddeke and Schäfer, 2020), conformational dynamics in small GTPases (Morris et al., 2016) or stability of protein complexes (Lemkul and Bevan, 2010). Previously, we employed MD simulations to determine the intersubunit affinity in human septin G-interface dimers, making use of available crystal structures (Grupp et al., 2023). In this experimental setup, the dimer structures and in silico generated apo structures were first relaxed in unbiased simulations. These equilibrated structures were subsequently subjected to steered simulations combined with umbrella sampling (Kästner, 2011) to extract eventually free binding energies.

By comparing the nucleotide-bound and apo structures after the unbiased simulations, we identified a conserved arginine residue in the SUE-βββ, Arg(βb), which undergoes a major conformational change (“flipping”) upon nucleotide removal in the investigated catalytically active subunits SEPT2 and SEPT7, accompanied by a reduction of the binding affinity between the respective subunits. The guanine base of the nucleotide is coordinated in a triangular network and Arg(βb) was previously suggested as a central interaction hub in this network in SEPT2 (Sirajuddin et al., 2009). We assumed that the nucleotide-induced reverse conformational transition involving Arg(βb) could explain the positive effect of nucleotide binding on septin assembly. If true, this conformational change should be conserved for all septins. Recently, structures of Saccharomyces cerevisiae septin G-interface dimers became available (Marques da Silva et al., 2023; Grupp et al., 2024), allowing to perform similar MD simulations with septins from yeast. We present here an improved structure of a yeast septin tetramer with an anisotropic resolution of 2.04 Å, providing an excellent experimental dataset for MD simulations. Based on these simulations, we provide herein evidence that the Arg(βb) plays indeed a species-overlapping pivotal role in G-interface formation. Our in vitro and in vivo experiments in human and yeast cells confirm the prediction that mutation of Arg(βb) results in G-interface disruption.

RESULTS

A refined structure of a yeast septin tetramer

We have previously presented the first structure of a septin heterotetramer comprised of the yeast-septin subunits Cdc10-Cdc3-Cdc12-Shs1 (PDB-ID 8PFH) (Grupp et al., 2024).

Because this complex contained GDP in all subunits, we attempted to crystallize the same complex in the transition state conformation. Inorganic MgF₃⁻ (forming spontaneously in a Mg²⁺ containing buffer from supplemented NaF) should occupy the γ-phosphate binding site and a GDP: MgF₃⁻ preparation of a small GTPase should thus mimic the transition state of the GTPase reaction (Jin et al., 2017). We crystallized the heterotetramer in the presence of Mg²⁺ and NaF. Diffracting crystals belonged to the space group P-1-2₁-1 and contained the tetramer Cdc10-Cdc3-Cdc12-Shs1 in the asymmetric unit (Figure 1A). As in our previous crystallization approach, a filament-like structure forming atypic Cdc10-Shs1 interfaces was present in the crystal. The resulting structure did not reveal the desired transition state conformation but contained again GDP in all subunits (Figure 1B). However, with anisotropic diffraction limits of 2.017, 3.018, and 2.645 Å as determined by STARANISO (Vonrhein et al., 2018) (Supplemental Table S1), the overall resolution was significantly improved compared with 8PFH. The structure was deposited in the PDB under identifier 9GD4.

FIGURE 1:

Crystal structure of the tetrameric septin complex. (A) The yeast septin complex consisting of Cdc10, Cdc3, Cdc12, and Shs1 in the asymmetric unit of the crystal. N- and C-termini of the subunits are indicated. (B) All subunits within the crystal contain GDP as shown by Polder-OMIT maps (contoured at 4.0 σ). (C) Nucleotide binding pocket of Cdc3 with a coordinated Mg²⁺ ion. A 2m-DFc electron density map (contoured at 1.5 σ (blue)) and a mFo-DFc electron density map (contoured at +3.0 σ (green) and −3.0 σ (red)) is shown. Both maps do not support to build GTP in this subunit. The model was aligned with the Cdc10:GDP-Cdc3:Mg²⁺/GTP dimer (black) (PDB-ID: 8FWP). (D) The conserved hydrogen bond network coordinating the guanine ring of the nucleotide in Shs1. Arg(βb) (here Arg288) forms a triangular bond network with the conserved residues Tyr(βb) (Tyr290), Asp(G4) (Asp220) and Glu(α4) (Glu188) from the neighboring subunit Cdc12. Nucleotides are displayed in ball and stick presentation throughout.

The overall fold of this and our previous structure is consequently very similar, with a Cα root-mean-square deviation (Cα-RMSD) of 0.975 Å. In the nucleotide binding pocket of Cdc3, a coordinated Mg²⁺ ion could be resolved in the same position as the published Cdc3-Cdc10 dimer (PDB-ID: 8FWP) and the Cdc3-Cdc10-Cdc10-Cdc3 tetramer (PDB-ID: 8SGD) (Marques da Silva et al., 2023) (Figure 1C).

MD simulations

We investigated human septin G-interfaces using MD simulations in previous work and determined the intersubunit affinities in dependence of the respective nucleotide (Grupp et al., 2023). We have identified a highly conserved Arg residue in the SUE, Arg(βb), as an essential element for G-interface integrity. Arg(βb) formed a triangular saltbridge network with the conserved residues Tyr(βb) and Asp(G4), which altogether coordinated the guanine ring of the nucleotide, and with a Glu(α4) residue from the neighboring subunit. The central role of Arg(βb) in coordination of the guanine base was already described (Sirajuddin et al., 2009). This network is also present in yeast septins (Figure 1D). In catalytically active human SEPT7 and SEPT2, Arg(βb) flipped by 90° upon nucleotide removal, thereby disrupting the entire bond network, suggesting that Arg(βb) plays a pivotal role in G-interface integrity (Grupp et al., 2023). Whether this feature is unique for human septins or a common mechanism across species could until now not be answered due to the lack of structural information of septin G-interfaces of other species.

The refined structure of a yeast septin presented here provided us with a crystal structure containing the yeast septin G-interface dimers with a favorable resolution to perform MD simulations, namely Cdc10:GDP-Cdc3:Mg²⁺/GDP and Cdc12:GDP-Shs1:GDP. We generated energy minimized coordinate files for these dimers in their nucleotide-bound conformation and obtained the apo state (Cdc10:apo-Cdc3:apo, Cdc12:apo-Shs1:apo) by in silico deletion of the nucleotide. Additionally, we created both dimer sets (with and without nucleotide) with Arg(βb) mutated to alanine. The overall setup of the in silico investigation followed the concept of our previous work (Grupp et al., 2023). Each coordinate file structure was temperature and pressure equilibrated with restraints applied, yielding input structures for unbiased simulations. Five independent unbiased simulations without restraints each with a duration of 700 ns were subsequently conducted.

Structural rearrangements within the Arg(βb) saltbridge network after the unbiased simulations were quantified by considering distances between the involved key residues. The pairwise distances between Arg(βb) and Asp(G4) as well as Arg(βb) and Glu(α4) were extracted from the unbiased simuations of wild-type (WT) apo and holo systems and probability densities were obtained by two-dimensional kernel estimation. From the obtained densities, relative free energy differences (ΔG) were calculated and plotted as free energy landscapes (Figure 2). The free energy landscape plot of the apo-bound state can be compared with the plot of the respective GDP-bound state to illustrate the structural changes in the Arg(βb) bond network over the timeframe of the simulation.

FIGURE 2:

Distances between key residues of the Arg(βb) hydrogen bond network, plotted as free energy landscapes. Pairwise distances from the unbiased simulations for each system were combined and probability densities were calculated by two-dimensional kernel estimation, from which relative free energy differences (ΔG) were calculated. The state with the highest calculated density across corresponding apo and holo systems served as a reference point. ΔG relative to the global energy minimum for both systems is shown as a color gradient. For each system, the average ΔG and its SD is given in the inserts. Reference lines in the plots indicate the distances present in the initial energy minimized coordinate structures, the intersection represents the observed distance pair. Each dot represents one distance pair. Free energy landscape plots are shown for distances between (A) Cdc10Arg(βb) - Cdc10Asp(G4) (x-axis) and Cdc10Arg(βb) - Cdc3Glu(α4) (y-axis), (B) Cdc3Arg(βb) - Cdc3Asp(G4) (x-axis) and Cdc3Arg(βb) - Cdc10Glu(α4), (C) Cdc12Arg(βb) - Cdc12Asp(G4) (x-axis) and Cdc12Arg(βb) - Shs1Glu(α4) (y-axis), and (D) Shs1Arg(βb) - Shs1Asp(G4) (x-axis) and Shs1Arg(βb) - Cdc12Glu(α4) (y-axis). For all panels, the GDP-bound state is shown on the left and the apo state is shown on the right.

Upon removal of the nucleotides from the Cdc10-Cdc3 dimer, the saltbridge networks became disturbed, as shown by a pairwise increased scattering of both distances between the respective residues. The average ΔG for the Cdc10Arg(βb) apo and Cdc3Arg(βb) apo system was increased compared with the respective GDP system, which confirms an overall destabilization of the interface in the apo state (Figure 2, A and B).

For Cdc10Arg(βb), the global energy minimum in the apo-bound state was shifted by ∼ 0.1 nm toward a smaller Arg(βb)-Asp(G4) spacing but 0.3 nm toward a higher Arg(βb)-Glu(α4) spacing, which suggests a rapid destabilization of the initially observed network over the time course of the simulation. An increased scattering along both axes could be seen, indicative of a weakening of both the Arg(βb)-Glu(α4) and Arg(βb)–Asp(G4) interactions (Figure 2A).

For Cdc3Arg(βb), the local energy minimum was not displaced in the apo state, suggesting a longer dwell time in the initial state of the simulation. As for Cdc10, both the Arg(βb)–Glu(α4) and Arg(βb)–Asp(G4) interactions weakened, as indicated by the increased number of low energy state observations with higher distances (Figure 2B).

Removal of the nucleotide from the Cdc12-Shs1 dimer resulted also in a global destabilization of the interface, indicated by a pairwise increased scattering of both distances and an increase of the average ΔG of both apo systems (Figure 2, C and D).

The Arg(βb)–Asp(G4) interaction, but not the Arg(βb)–Glu(α4) interaction was weakened for Cdc12Arg(βb) (Figure 2C). On the contrary, this weakening is more prominent for the Arg(βb)–Asp(G4) interaction than the Arg(βb)–Glu(α4) interaction, as determined by the appearance of a second energy minimum.

To complement the free energy landscape approach with a more global view on structural deformations of the whole SUE-βββ, we compared the RMSD distribution of the SUE-βββ, obtained from the pooled unbiased simulations of each septin subunit, with respect to the equilibrated input structure (Supplemental Figure S1). The 75% quantile and the maximum RMSD value was shifted for any introduced modification (removal of the nucleotide, Arg(βb) mutation or a combination of both) in any sepin subunit. This behavior indicated an increased probability of a structural deformation in the SUE-βββ. Compared with the other subunits, Shs1 displayed the most uniform violin plot shapes and smallest shift of the 75% quantile, which suggests that the SUE-βββ of this subunit is less prone to structural alterations.

Next, we proceeded to determine intersubunit affinities within the different systems. The output of the five unbiased simulations was subjected to backbone RMSD-based clustering (Daura et al., 1998) and the central structure in the largest cluster for each system was first examined with respect to structural alterations. Structural integrity of the septin dimers maintained after the unbiased simulations with the Cα-RMSD of the relaxed structures remaining below 0.6 nm compared with the energy minimized coordinate structures, with the largest deviations located as expected in the flexible switch loops and loops in the free NC-interfaces (Supplemental Figure S2).

We describe in the following in more detail the structural alterations of these structures compared with the respective coordinate file structures.

In the unmodified Cdc10-Cdc3 and Cdc12-Shs1 dimers, the bond network around Arg(βb) remained intact (Supplemental Figures S3A and S4A). Upon removal of the nucleotides from the Cdc10-Cdc3 dimer, flipping of the Arg(βb) could be seen in both subunits (Supplemental Figure S3B). Additionally, the bonds connecting the βb and βc strands in the Cdc10-SUE became distorted. As a result, Cdc10-SUE-βββ partially dissociated from the Cdc3-SUE, indicated by altered positioning of Trp(βb) relative to Cdc10 (Supplemental Figure S3B). Introduction of the Arg(βb)-Ala mutation in both subunits in the nucleotide-bound state left the Asp(G4) hydrogen bonding to the guanine ring intact. Otherwise, as in the apo form, the bonds connecting the βb and βc strands in the Cdc10-SUE became perturbed, resulting in partial dissociation of the Cdc10-SUE-βββ from the Cdc3-SUE (Supplemental Figure S3C).

Removal of the nucleotide from the Cdc12-Shs1 dimer resulted only in a partial disruption of the bond network, corresponding to the slight deformation of Arg(βb), but without flipping (Supplemental Figure S4B). Larger conformational displacements within the SUE occurred, as indicated by the partial dissociation of the Cdc12-SUE-βββ from Shs1. Furthermore, the two Tyr(βb) residues of both subunits tilted toward each other (Supplemental Figure S4B). Introduction of the Arg(βb)-Ala mutation lead additionally to disruption of the contact between the conserved Trp(βb) residues of both subunits (Supplemental Figure S4C). For both Cdc10-Cdc3 and Cdc12-Shs1 dimers, we also tested the Arg(βb)-Ala mutation in the apo form. No further structural alterations could be observed (Supplemental Figures S3D and S4D).

To calculate intersubunit affinities, protein-protein binding free energy profiles were determined by COM-pulling and umbrella sampling (Lemkul and Bevan, 2010; Patel and Ytreberg, 2018; Grupp et al., 2023), employing the central structure in the largest cluster from the unbiased simulations. Briefly, the filament axis of each structure was aligned with the x-axis, both subunits were separated along the x-axis in COM-pulling simulations and different conformations along the reaction coordinate were extracted. COM separation distances ranged from 3.5 nm (starting conformation) to ∼ 8.5 nm.

Binding free energies (∆G) for the different dimers were calculated by the weighted histogram analysis method (Hub et al., 2010). The corresponding free energy profiles are shown in Supplemental Figure S5, A and B. Binding free energy in the Cdc12-Shs1 G-interface was increased by ∼ 50 kJ mol⁻¹ upon nucleotide removal, introduction of the Arg(βb)-Ala mutation, or both (Table 1; Supplemental Figure S5C). In the Cdc10-Cdc3 dimer, ∆G increased by ∼ 30 kJ mol⁻¹ upon nucleotide removal and by ∼ 50 kJ mol⁻¹ upon combined introduction of the Arg(βb)-Ala mutation and nucleotide removal (Table 1; Supplemental Figure S5C). Introduction of the mutation alone did not result in an increase of ∆G. Consequently, these results suggest that nucleotide removal leads to a destabilization of the yeast septin interfaces, reflected in the increased ∆G, in agreement with our previous observations for human septins (Grupp et al., 2023).

TABLE 1:

Binding free energies for the Cdc10-Cdc3 and Cdc12-Shs1 systems.

System	ΔG [kJ*mol−1]	ΔΔG [kJ*mol−1]
Cdc10-Cdc3
Cdc10:GDP-Cdc3:GDP	−137 ± 11	—
Cdc10:apo-Cdc3:apo	−103 ± 14	33 ± 18
Cdc10R251A:GDP-Cdc3R360A:GDP	−135 ± 21	2 ± 24
Cdc10R251A:apo-Cdc3R360A:apo	−89 ± 17	48 ± 20
Cdc12-Shs1
Cdc12:GDP-Shs1:GDP	−189 ± 19	—
Cdc12:apo-Shs1:apo	−133 ± 9	57 ± 21
Cdc12R263A:GDP-Shs1R288A:GDP	−138 ± 35	51 ± 39
Cdc12R263A:apo-Shs1R288A:apo	−133 ± 21	56 ± 28

In the following, we present our efforts to investigate the role of Arg(βb) in vivo and in vitro both in human and yeast septins.

Arg(βb) is essential for filament incorporation of SEPT9- and SEPT9-mediated cell migration

We have recently created a SEPT9 knockout (ko) in human dermal fibroblasts (Hecht et al., 2024). This cell line provides an excellent tool to study the role of Arg(βb) (Arg516 in SEPT9) in the context of G-interface formation between SEPT7 and SEPT9 inside the living cell.

We generated an EGFP-SEPT9(R516A) construct and additionally EGFP-SEPT9(W520A). Trp520 in SEPT9 is an absolute conserved residue in the SUE that is essential for G-interface integrity. Mutating this Trp to Ala should disrupt the SEPT9-SEPT7 G-interface as shown for SEPT2 (Sirajuddin et al., 2007).

Both mutations and WT EGFP-SEPT9 were transfected into the ko cell line and stably expressing cell lines were selected. The intracellular localization of the EGFP-SEPT9 constructs was monitored using fluorescence microscopy with the endogenous septin cytoskeleton stained with an anti-SEPT8 antibody and actin stained with Phalloidin. EGFP-SEPT9 was incorporated readily into endogenous septin fibers, displaying the typical phenotype of WT septin filaments including partly colocalization with actin stress fibers as described by us and others (Dolat et al., 2014; Spiliotis et al., 2016; Hecht et al., 2024) (Figure 3A, upper panel). The Trp520 mutant did only weakly if at all align with the endogenous septin cytoskeleton (Supplemental Figure S6.) We found a similar result for cells expressing EGFP-SEPT9(R516A). Localization of this construct was mostly diffuse or in filament-like formations concentrated at the plasma membrane. Colocalization with the endogenous septin and actin cytoskeleton was only occasionally observed (Figure 3A, lower panel). Western blot analysis employing an anti-SEPT9 antibody of extracts from 1306 WT cells, SEPT9 ko cells and cells expressing EGFP-SEPT9 or EGFP-SEPT9(R516A) confirmed lower expression levels in the soluble extract and degradation products for EGFP-SEPT9(R516A) (Supplemental Figure S7). We conclude that the not-incorporated protein becomes either insoluble or unstable, leading to degradation inside the cell or during extraction. This is in line with the finding that unassembled protein complex subunits are instable and prone to degradation (Sousa et al., 2019).

FIGURE 3:

Arg(βb) is essential for filament incorporation of SEPT9 in vivo and SEPT9-mediated cell migration. (A) Human 1306 fibroblasts with SEPT9 ko, stably overexpressing EGFP-SEPT9 (upper panel) or EGFP-SEPT9(R516A) (lower panel). The endogenous actin was stained with Phalloidin and the endogenous septin cytoskeleton was visualized with an anti-SEPT8 antibody. Note that the majority of EGFP-SEPT9(R516A) is not incorporated into endogenous septin structures. In most observed cells, no colocalization at all could be detected. Shown are sum intensity projections. Scale bar 10 µM. (B) A total of 1306 fibroblasts with SEPT9 ko exhibit restricted filopodia as indicated by staining with an anti-VASP antibody (right). Stable overexpression of EGFP-SEPT9 in these cells restores filipodia length (middle) whereas EGFP-SEPT9(R516A) overexpressing cells maintain the ko phenotype (right). Filopodia are highlighted with white arrowheads. Shown are max intensity projections. Scale bar 10 µM. (C) Boyden chamber assay to monitor cellular migratory behavior. Overexpression of EGFP-SEPT9 in SEPT9 ko cells restores migration rate to WT level whereas EGFP-SEPT9(R516A) overexpressing cells maintain the migration rate of ko cells. Significance values were calculated by one-way ANOVA followed by Tukey post hoc test with N = 28 to N = 30 collected from three independent experiments. Data are depicted as scattered dot plot with means ± SD.

SEPT9 is functionally linked to focal adhesion maturation and cellular migration (Fuechtbauer et al., 2011; Dolat et al., 2014; Lam and Calvo, 2019). Accordingly, SEPT9 ko fibroblast cell line exhibited restricted, shortened filopodia (Hecht et al., 2024) (Figure 3B, left). Expression of EGFP-SEPT9 in SEPT9 ko cells restored filopodia length (Figure 3B, middle) whereas EGFP-SEPT9(R516A) overexpressing cells exhibited a similar, restricted filipodia shape as the ko cells (Figure 3B, right).

In line with these findings, SEPT9 ko cells barely migrated in a transwell assay whereas overexpression of SEPT9 in WT cells increased the migration rate (Hecht et al., 2024) (Figure 3C). Stable overexpression of EGFP-SEPT9 in SEPT9 ko cells restored the migration rate to WT level whereas cells stably overexpressing EGFP-SEPT9(R516A) displayed low migration rates close to those of the ko cells (Figure 3C).

Arg(βb) is essential for yeast septin function in vivo

We investigated next the impact of Arg(βb) on subunit association in yeast septins. To test whether Arg(βb) is necessary for protofilament formation in vivo, we generated yeast strains each with a ko for one mitotic septin (Cdc10, Cdc3, Cdc12, Cdc11, Shs1). Each strain expressed its genomically deleted septin from an extrachromosomal rescue plasmid containing an uracil prototrophy. The cells were subsequently transformed with a second plasmid, containing either a WT copy of the deleted septin, a copy bearing the Arg(βb)-Ala mutation, or containing no additional coding sequence. Serial dilutions of each strain were spotted onto selective media containing 5-fluoorotic acid (FOA), selecting for cells having lost the plasmid. The plates were incubated at different temperatures and colony growth was documented. ko of each septin except of Shs1 was lethal, indicated by complete growth inhibition of the empty plasmid control on FOA medium (Figure 4). Arg(βb) to Ala mutation was lethal for cells expressing the corresponding cdc11 and cdc12 alleles at all tested temperatures. The same mutation resulted only in temperature sensitivity (restricted growth at 37°C) in cdc10 and cdc3 alleles. Strains containing the mutation in these subunits grew normal at 30°C (Figure 4).

FIGURE 4:

Growth assay with yeast expressing septins with Arg(βb) mutation. Yeast cells with a ko for each one septin (from top to bottom: Cdc10, Cdc3, Cdc12, Cdc11, Shs1) were transformed either with WT, Arg(βb)-Ala mutation or empty plasmid. Serial dilutions of each strain were spotted onto selective media containing 5-FOA, selecting for cells that have lost a rescue plasmid containing an Ura3 marker. Note that the Arg(βb) mutation results in temperature sensitivity for all tested strains and in lethality even at permissive temperature for Cdc12 and Cdc11.

To validate this finding, we constructed GFP fusion proteins with and without Arg(βb) to Ala mutation of each septin subunit and tested the intracellular localization in two different yeast strain backgrounds (Figure 5; Supplemental Figure S8). The budneck localization was quantified in relation to a Shs1-mCherry marker signal (Figure 5A). Consistent with the growth assay, Cdc12, Cdc11, and Shs1 containing the Arg(βb) mutation failed to localize at the budneck. The GFP signal was distributed unspecifically in the cytoplasm (Figure 5B). Budneck localization with significantly reduced fluorescence intensity was still detectable for the Arg(βb) mutants in Cdc10 and especially Cdc3 (Figure 5, A and B).

FIGURE 5:

Localization of septins containing Arg(βb) mutation inside the living yeast cell. Septins with and without Arg(βb)-Ala mutation were expressed as GFP fusion protein in yeast cells containing Shs1-mCherry as septin marker. (A) Quantification of the budneck signal (in relation to the marker signal) of the GFP fusion constructs. Only cells in G2/M phase with a clearly detectable Shs1-mCherry signal at the budneck were considered. Budneck localization is entirely abolished for Arg(βb) mutation in Cdc12, Cdc11, and Shs1 and reduced for Cdc10 and Cdc3. Significance values were calculated by an unpaired t test with Welch's correction with at least N = 39 cells. Data are depicted as scattered dot plot with means ± SD. (B) Representative microscopy images for each GFP-tagged septin subunit as WT or Arg(βb)-Ala mutation. In the mCherry channel, the Shs1-mCherry marker is shown. Note the diffuse localization of GFP-Cdc12, GFP-Cdc11, and GFP-Shs1 each bearing the Arg(βb)-Ala mutation. Shown are sum intensity projections. Scale bar 5 µM.

Arg(βb) is essential for G-interface formation in septin dimers in vitro

We asked next whether our findings could be reproduced in vitro using recombinantly expressed G-interface dimers. The central G-interface dimers of human hexameric or octameric rods, the SEPT7 homodimer and the SEPT7-SEPT9 heterodimer, respectively, were employed for this purpose. The truncated G-domains of SEPT7 and SEPT9, each with or without mutation of the Arg(βb) to Ala, were expressed from two compatible plasmids (Table 2) within the same cell. Because both G-domains have a comparable molecular weight, SEPT7 was fused to MBP, enabling its differentiation from SEPT9 on a SDS–PAGE gel. SEPT9 was fused to a His₆-tag for purification. For the SEPT7 homodimer, His₆-SEPT7 constructs with and without Arg(βb) to Ala mutation were expressed.

TABLE 2:

List of plasmids used in this study. pAc-MBP, pAc, and pES are an in-house constructed, pET15A based plasmids for the expression of MBP- or His₆-tagged proteins in E. coli. fl -full length, aa – amino acid residue.

	Construct	Vector backbone	Insert	Source
#1	EGFP-SEPT9	pEGFP-C2	fl	E. Spiliotis
#2	EGFP-SEPT9(R516A)	pEGFP-C2	fl	This study
#3	EGFP-SEPT9(W520A)	pEGFP-C2	fl	This study
#4	DsRed-SEPT9	pDsRed-Monomer C1	fl	M. Hecht
#5	EGFP-SEPT2	pEGFP-C2	fl	M. Hecht
#6	His6-SEPT9G	pES	aa 295-567	(Hecht et al., 2024)
#7	His6-SEPT9G(R516A)	pES	aa 295-567	This study
#8	MBP-SEPT7G	pAc-MBP	aa 48-318	This study
#9	MBP-SEPT7G(R265A)	pAc-MBP	aa 48-318	This study
#10	His6-MBP-SEPT7G	pAc	aa 48-318	This study
#11	His6-MBP-SEPT7G(R265A)	pAc	aa 48-318	This study
#12	His6-SEPT7G	pES	aa 48-318	This study
#13	His6-SEPT7G(R265A)	pES	aa 48-319	This study
#14	His6-MBP-Cdc10G	pAc	aa 30-end	This study
#15	His6-MBP-Cdc10G(R251A)	pAc	aa 30-end	This study
#16	His6-MBP-Cdc12G	pAc	aa 32-314	This study
#17	His6-MBP-Cdc12G(R263A)	pAc	aa 32-314	This study
#18	His6-Shs1G	pET-Duet1	aa 21-339	This study
#19	His6-Shs1G(R288A)	pET-Duet1	aa 21-339	This study
#20	His6-Cdc11G	pET-Duet1	aa 20-298	This study
#21	His6-Cdc11G(R247A)	pET-Duet1	aa 20-298	This study
#22	His6-Cdc3G	pET-Duet1	aa 117-410	This study
#23	His6-Cdc3G(R360A)	pET-Duet1	aa 117-410	This study
#24	PMET-Cdc10	pRS316	fl	(Grupp et al., 2024)
#25	PMET-Cdc3	pRS316	fl	This study
#26	PMET-Cdc12	pRS316	fl	This study
#27	PMET-Cdc11	pRS316	fl	This study
#28	PMET-Shs1	pRS316	fl	This study
#29	PMET-Cdc10	pRS313	fl	(Grupp et al., 2024)
#30	PMET-Cdc10(R251A)	pRS313	fl	This study
#31	PMET-Cdc3	pRS313	fl	This study
#32	PMET-Cdc3(R360A)	pRS313	fl	This study
#33	PMET-Cdc12	pRS313	fl	This study
#34	PMET-Cdc12(R263A)	pRS313	fl	This study
#35	PMET-Cdc11	pRS313	fl	This study
#36	PMET-Cdc11(R247A)	pRS313	fl	This study
#37	PMET-Shs1	pRS313	fl	This study
#38	PMET-Shs1(R288A)	pRS313	fl	This study
#39	PMET-Cdc10-GFP	pRS306	fl	This study
#40	PMET-Cdc10(R251A)-GFP	pRS306	fl	This study
#41	PMET-GFP-Cdc3	pRS306	fl	This study
#42	PMET-GFP-Cdc3(R360A)	pRS306	fl	This study
#43	PMET-Cdc12-GFP	pRS306	fl	This study
#44	PMET-Cdc12(R263A)-GFP	pRS306	fl	This study
#45	PMET-Cdc11-GFP	pRS306	fl	This study
#46	PMET-Cdc11(R247A)-GFP	pRS306	fl	This study
#47	PMET-Shs1-GFP	pRS306	fl	This study
#48	PMET-Shs1(R288A)-GFP	pRS306	fl	This study
#49	PMET-GFP	pRS306	fl	This study

The septin complexes were purified by IMAC via the His₆-tag and immediately afterward subjected to analytical size exclusion chromatography (SEC). The elution profile of the WT SEPT7-SEPT9 heterodimer contained besides the major dimer peak a second peak corresponding to monomeric His₆-SEPT9, likely resulting from unequal expression of the subunits from the respective plasmids (Figure 6A). Mutation of Arg(βb) to Ala in any of the two subunits shifted the peak entirely towards the retention volume of monomeric His₆-SEPT9. The SEPT7 homodimer eluted exclusively as dimer after IMAC purification and the retention volume was entirely shifted to the SEPT7 monomer upon mutation of Arg(βb) to Ala (Figure 6B).

FIGURE 6:

Arg(βb) is essential for nucleotide-mediated G-interface formation in vitro. (A) His₆-SEPT9 and MBP-SEPT7 G-domains with and without Arg(βb) mutation were coexpressed in E. coli, purified by IMAC and subsequently subjected to analytical SEC. Arg(βb) mutation in any of the subunits leads to disruption of the interface, leaving only monomeric His₆-SEPT9 after the IMAC. Coomassie-stained SDS–PAGE sections of the indicated peaks from the SEC are shown besides the chromatogram. MBP-SEPT7: 75 kDa calculated MW, His₆-SEPT9: 36 kDa. (B) His₆-SEPT7 with and without Arg(βb) mutation was expressed in E. coli, purified by IMAC and subsequently subjected to analytical SEC. Arg(βb) mutation leads to disruption of the homotypic interface, leaving only monomeric His₆-SEPT7 after the IMAC. Commassie-stained SDS–PAGE sections of the indicated peaks from the SEC are shown. His₆-SEPT7: 33 kDa calculated MW. (C) Purified monomeric MBP-SEPT7 with and without Arg(βb) mutation was incubated with purified monomeric His₆-SEPT7, His₆-SEPT9 and His₆-SEPT9(R516A) in the presence of a GTP/GDP mixture. The Arg(βb) mutation suppressed G-interface formation entirely. Unspecific binding of human septin subunits to MBP-Cdc12 could not be detected. (D) Purified monomeric MBP-Cdc10 and MBP-Cdc12 with and without Arg(βb) mutation were incubated with the respective purified His₆-tagged G-interface partner with or without Arg(βb) mutation in the presence of a GTP/GDP mixture. The Arg(βb) mutation suppressed G-interface formation entirely, regardless of the tested combination. Unspecific binding of yeast septin subunits to MBP-SEPT7 could not be detected.

Does the Arg(βb) to Ala mutation interfere with the process of G-interface formation? To answer this question, we assembled the SEPT7-SEPT9 heterodimer and the SEPT7 homodimer from individual subunits in vitro and tested whether the Arg(βb) mutation inhibits complex formation. We generated His₆-MBP-SEPT7 with and without Arg(βb) mutation and His₆-SEPT9, His₆-SEPT9(R(βb)A) and His₆-SEPT7. All constructs were subsequently purified by IMAC and preparative SEC. The SEPT7 constructs tended to form dimers to a certain extent but addition of 1 mM MgCl₂ to the IMAC elution buffer resulted almost exclusively in monomers as judged by SEC (Supplemental Figure S9A).

The His₆-MBP-SEPT7 (and His₆-MBP-Cdc12) constructs were incubated with a 5-fold molar excess of the respective His₆-tagged G-interface partner in the presence of nucleotide.

Complexes were subsequently incubated with amylose resin, eluted with maltose and the composition of the complex was analyzed by SDS–PAGE. His₆-SEPT9 and His₆-SEPT7 were only able to bind to WT His₆-MBP-SEPT7 (Figure 6C). The Arg(βb) to Ala mutation suppressed G-interface formation entirely. No unspecific binding with the yeast septin His₆-MBP-Cdc12 could be detected.

To test the G-interface formation for yeast septins, we purified monomeric His₆-MBP-Cdc12 and His₆-MBP-Cdc10 with and without Arg(βb) to Ala mutation and the respective G-interface partners His₆-Cdc3, His₆-Cdc11 and His₆-Shs1. The monomeric state of the proteins was confirmed by SEC (Supplemental Figure S9B). While the G-domains of Cdc3, Cdc11, and Shs1 could be readily purified as monomers, Cdc10 and Cdc12 required the MBP-tag and an improved lysis buffer composition to ensure solubility (see Materials and Methods section).

As for the human G-interface dimers, the Arg(βb) to Ala mutation blocked formation of the Cdc10:3, Cdc12:11 and Cdc12:Shs1 G-interfaces entirely (Figure 6D). No unspecific binding to the human His₆-MBP-SEPT7 could be detected.

DISCUSSION

Small GTPases of the Ras family share common structural features and cycle between GTP- and GDP-bound conformations, each conformation exhibiting distinct affinities for effector molecules (Bourne et al., 1991). Although septins share the same structural features as small GTPases, their biochemistry seems to be different (Grupp and Gronemeyer, 2022) and the role of nucleotide utilization by septins is a subject of intense debate. The fact that all available dimeric or multimeric septin complexes contain nucleotide strongly suggests that septins utilize the nucleotide for protofilament formation (Grupp and Gronemeyer, 2022). We have previously employed MD simulations on available human septin structures to bypass the limitation that dimers in their apo conformation cannot be generated in vitro (Grupp et al., 2023). We found that an arginine residue in the SUE βββ-meander undergoes a conformational change upon nucleotide removal in the catalytically active SEPT2 and SEPT7 subunits. We could identify Arg(βb) related structural alterations also in yeast septins employing MD simulations.

When interpreting the output from MD simulations, one has to keep two intrinsic limitations in the context of septin dimers in mind: First, the simulations with septin dimers depict the scenario of disrupting an already formed dimer, but septin protofilaments are stable for hours in vitro once they are assembled (Renz et al., 2013). The simulations thus identify “weak spots” inside the dimer that also may determine dimer formation. However, these predictions have to be confirmed experimentally.

Second, the unbiased simulations cover a limited time frame (here 700 ns, requiring ∼ 2 wk of computing time per simulation) but folding of proteins or rearrangements within complexes may require microseconds (Chen et al., 2017), time scales exceeding reasonable computing time for unbiased MD. In our previous work, we argued that Arg(βb) flipping is a feature of catalytically active septins since we did not observe it in SEPT6. Contrary to this assumption, we could not show Arg(βb) flipping in the central clustered structure of the catalytically active Cdc12. Thus, we might have simply missed the flipping in Cdc12 as well as in the catalytically inactive Shs1 and SEPT6 if it occurs later on the reaction coordinate than what is captured by the time frame of the simulation.

To account for this presumable limitation, we have conducted free energy landscape and backbone RMSD analyses to detect and describe Arg(βb) related structural alterations in the SUE on a more global scale rather than to focus on the movement of one particular amino acid residue.

We performed in vitro and in vivo reconstitution experiments with human and yeast septins to confirm the evolutionary conserved significance of Arg(βb) for G-interface integrity.

SEPT9 is catalytically active and occupies the central position of the human octameric protofilament (Castro et al., 2020), making this subunit an excellent target to study Arg(βb)-mediated filament incorporation of a subunit inside the living cell. We made use of our recently presented SEPT9 ko cell line (Hecht et al., 2024) and performed complementation experiments using EGFP-SEPT9 with and without mutated Arg(βb). Although octamers are favored (Martins et al., 2022), the human septin cytoskeleton can consist of hexamers in vivo (Sellin et al., 2011; Soroor et al., 2020), with SEPT7 forming a central G-interface. Our SEPT9 ko cells show indeed a rudimentary septin cytoskeleton but display severe phenotypes such as defects in cell migration and focal adhesion and filopodia dysfunction (Hecht et al., 2024). SEPT9 with Arg(βb) mutated to alanine is not incorporated into septin filaments in vivo and our in vitro data support the conclusion that the SEPT7-SEPT9 G-interface is not properly formed. Cells expressing this mutant as only copy of SEPT9 show a ko phenotype regarding filopodia architecture and migratory behavior.

The outer subunits of the yeast septin protofilament Cdc11 and Cdc12 are essential for the cell (Hartwell, 1971) and the Arg(βb) mutation in these subunits is lethal. Shs1 is a nonessential septin that can be replaced by Cdc11 (Iwase et al., 2007; Garcia et al., 2011) and consequently neither the ko of this subunit nor its Arg(βb) mutation is lethal.

Neither Cdc11, nor Shs1, nor Cdc12 are incorporated into septin structures in vivo when bearing the Arg(βb) mutation, supported by the decreased intersubunit affinity in the Shs1-Cdc12 dimer upon nucleotide removal and/or introduction of the mutant.

For the inner subunits of the yeast septin protofilament Cdc3 and Cdc10, introduction of the Arg(βb) mutation leads only to temperature sensitivity although both subunits are also essential (Hartwell, 1971). Localization to septin structures at the budneck is weakened for both subunits with Arg(βb) mutation, but still detectable. In line with these findings, the intra-subunit affinity in the Cdc3-Cdc10 interface is only decreased for the apo form but not for the interface bearing the Arg(βb) mutation. The Cdc3 and Cdc10 subunits including the Arg(βb) mutation might have a remaining sufficient (but weakened) overall affinity to form a G-interface in vivo (maybe promoted by other intracellular factors or conditions), resulting in the partial complementation of the corresponding gene deletions.

However, this remaining affinity is obviously not sufficient to form the interface in vitro. In Shs1, Cdc11 and Cdc12 the destabilizing effects prevail and the interface is neither formed in vivo nor in vitro.

Taken together, we provided evidence that nucleotide coordination by Arg(βb) is a crucial factor for subunit incorporation into the septin cytoskeleton. Results found by others highlight the physiological relevance of this particular residue: Cla10-1 yeast (with CLA10 being allelic to CDC12) show cytokinesis defects in the absence of Cln1 and Cln2 cyclins (Cvrčková et al., 1995). The mutation in CDC12 was later identified as a conversion of Arg(βb) to Lysine (Weems et al., 2014). Several septin mutations can be found in human cancers (Angelis and Spiliotis, 2016). Mutations of Arg(βb) in human SEPT12, SEPT10, and SEPT8 is documented in colon cancer samples (Seshagiri et al., 2012; Mouradov et al., 2014; Robles et al., 2016). The “Somatic Mutations in Cancer Database” COSMIC (https://cancer.sanger.ac.uk/cosmic) lists further Arg(βb) mutations of SEPT14 in lung cancer, endometrioid, and kidney carcinoma and of SEPT3 in breast cancer. Asp(G4) is the only other residue of the triangular network that is mutated in SEPT2 and SEPT3 of cancer cells (Angelis and Spiliotis, 2016). These Arg(βb)-related physiological phenomena from different species encouraged us to evaluate the conservation degree of this residue across species and kingdoms in more detail. We performed a sequence analysis with 131 Opisthokonta and 123 Archaeplastida and Chromista septins employing the sequence collection from (Delic et al., 2024). All evaluated 254 septins contain Arg(βb) in their SUE-βββ and the other residues in the network around Arg(βb) show a high conservation score (Figure 7A). The conservation score was superimposed onto the structure of our yeast septin teramer (Figure 7B). Closer inspection of the nucleotide binding pocket in Cdc3 as example shows that all residues of the bond network coordinating the nucleotide with Arg(βb) in the center show the highest conservation score (Figure 7C). Because we have confirmed nucleotide coordination by Arg(βb) also experimentally as crucial factor for G-interface formation in septins from yeast and man, we conclude that Arg(βb) is the central hub for nucleotide coordination in all septins. We suggest that during evolution septins have repurposed the nucleotide from signal transduction to a tool to induce and/or report protofilament completion.

FIGURE 7:

Sequence analysis from 254 septins (131 Opisthokonta and 123 non-Ophisthokonta). (A) Excerpt of the sequence alignment, showing 41 out of 254 septin sequences. Conservation degree is visualized by a color coded conservation score calculated for all 254 septins using ConSurf (Ashkenazy et al., 2010). Cdc3 (bold letters) serves as reference sequence and structure. (B) Conservation score from A projected on the yeast septin tetramer structure (PDB-ID 9GD4). The color gradient for the conservation score obtained from ConSurf is shown. (C) Nucleotide binding pocket of Cdc3. All amino acid residues within 5 Å distance from the nucleotide with conservation score of 9 are indicated. These residues are also indicated in A on the top of the alignment.

MATERIALS AND METHODS

Protein crystallization and structure determination

Protein expression and purification for crystallization was performed as described previously (Grupp et al., 2024) with the exception that usage of Ethanol in the expression medium was abandoned and that 10 mM NaF was added to the protein preparation after the preparative size exclusion chromatography.

Septin complexes were crystallized by sitting-drop vapor diffusion. Diffracting crystals were obtained with 0.33 µl of protein solution (3 mg/ml) mixed with 0.21 µl of a reservoir condition containing 17.5% PEG 5000, 0.1 M (NH₄)₂SO₄ and 0.1 M Bis-Tris pH 6.5 and 0.06 µl of a seeding solution. Crystals appeared within one day upon incubation of the crystallization plates at 20°C. Crystals were cryoprotected in 10% 2,3-butandiol prior to flash-freezing in liquid nitrogen. The diffraction experiments were carried out at the ID30B beamline of the ESRF (doi.10.15151/ESRF-ES-928402160) at 100 K at a wavelength of 0.873129 Å.

Merged x-ray diffraction data were obtained from the ESRF autoprocessing pipeline using XDS (Kabsch, 2010) as part of autoPROC Staraniso (Vonrhein et al., 2018).

The structure was solved by molecular replacement in Phaser (McCoy et al., 2007) using the subunits from PDB-ID 8PFH as template. The final model was completed by iterative rounds of automated refinement and model building in the Phenix package (Adams et al., 2010), autoBUSTER (Bricogne et al., 2017) and Coot (Emsley et al., 2010). Initial automated refinement cycles in Phenix consisted of reciprocal and -real space positional refinement as well as grouped B-factor refinement and TLS refinement while applying secondary structure restraints. In later stages of model building, automated refinement was performed using autoBUSTER with individual B-factor and translation liberation screw (TLS) refinement. After refinement, the final model exhibited 97.85% Ramachandran favored residues, of these 2.15% in allowed regions, and a clash-score of 1.33 as determined by MolProbity as part of autoBUSTER.

The refinement statistics are shown as Supplemental Table S1.

MD simulations

Energy minimized coordinate files were generated as described in previous work (Grupp et al., 2023) with some modifications: The Cdc10:GDP-Cdc3:GDP:Mg²⁺ and Cdc12:GDP-Shs1:GDP dimers were extracted from PDB-ID 9GD4 and truncated to their G-domains (Cdc10_F31-I301, Cdc3_F118-A410, Cdc12_G33-E313, Shs1_I22-S338). Crystal waters within 0.6 nm of the truncated models were included for all further steps. Unresolved regions in the structures were filled using MODELLER (version 10.5) (Šali and Blundell, 1993) using subunits from PDB-ID 8SGD and 8PFH as templates. Assignment of protonation state, hydrogen bond optimization and structure relaxation were performed in Maestro (Schrödinger Release 2024-2, Maestro, Schrödinger, LLC, New York, 2021) as described previously (Grupp et al., 2023). Subsequently, the nucleotide coordinates were removed from the resulting PDB files, leading to structures of the dimers in the apo state. Both nucleotide-containing and apo coordinate files were generated with Arg(βb) to Ala mutations which were introduced with the “residue scanning” tool in Maestro.

Force field compatibility transformation as well as N- and C-terminal capping were performed essentially as described (Grupp et al., 2023).

All MD simulations were performed in GROMACS v. 2022.2 (Abraham et al., 2015) using the CHARMM36m force field (Huang et al., 2017) on the JUSTUS2 high performance cluster at Ulm University. Time steps for the integration were set to 2 fs and bonds to hydrogen atoms were constrained using the LINear Constraint Solver (LINCS) algorithm (Hess et al., 1997; Hess, 2008). No dispersion correction was applied. The particle mesh Ewald (PME) algorithm (Darden et al., 1998; Essmann et al., 1998) was employed for treatment of electrostatic interactions with a real-space cutoff of 1.2 nm. Short-range Lennard-Jones interactions were smoothly switched to zero between 1.0 and 1.2 nm. Periodic boundary conditions were applied in all directions.

Each dimeric complex was placed in an octahedral box with at least 1.5 nm distance between the protein and the surrounding box in all dimensions. The box was explicitly solvated using the CHARMM-modified TIP3P water model (Durell et al., 1994; Jorgensen et al., 1998; Neria et al., 1998). The system was further supplemented with 300 mM KCl, including neutralizing counterions. System construction was followed by steepest descent potential energy minimization of the system with a maximum force tolerance of 500 kJ mol⁻¹ nm⁻¹. Subsequently, equilibration to a target temperature of 300 K was performed over 50 ps under an NVT ensemble using the velocity rescaling thermostat (Bussi et al., 2007). The time constant for temperature regulation was set to 0.1 ps. Initial velocities were assigned using random seeding according to the Maxwell–Boltzmann distribution. Equilibration was continued for 500 ps under an NPT ensemble to reach the target pressure (1.0 bar). Pressure was regulated isotropically with the stochastic cell rescaling barostat (Bernetti and Bussi, 2020) with a time constant of 2 ps for pressure coupling. During temperature and pressure equilibration coordinates of heavy atoms in proteins and ligands were restrained with a force constant of 1000 kJ mol⁻¹ nm⁻² in all dimensions. Production simulations were conducted under the same conditions as the NPT equilibration but without restraints and with pressure regulated by the Parrinello-Rahman barostat (Parrinello and Rahman, 1998). We performed unbiased MD simulations for 700 ns, during which we expected to be able to cover the Arg(βb) side chain motions and molecular rearrangements. Equilibration and production simulations were performed five times with different random seeds during velocity generation, resulting in sets of five independent production simulations per system.

Structural rearrangements within the Arg(βb) bond network after unbiased simulations were analyzed by considering distances between key residues of this network, plotted as free energy landscapes. The residue identities were as follows: Asp(G4): Cdc10_D182, Cdc3_D289, Cdc12_D182, Shs1_D220; Glu(α4): Cdc10_E188, Cdc3_D295, Cdc12_D188, Shs1_D226; Arg(βb): Cdc10_R251, Cdc3_R360, Cdc12_R263, Shs1_R288. Free energies were obtained after extracting the pairwise distances between Arg(βb) and Asp(G4) as well as Arg(βb) and Glu(α4) from the five independent 700 ns simulations for each system. The reference atoms were Cζ in Arg(βb), Cγ in Asp(G4) and Cδ in Glu(α4). Distances from the five simulations were combined, followed by two-dimensional kernel density estimation in OriginPro v. 2023 using the bivariate kernel density estimator bandwidth method and the exact estimation formula. From the obtained densities, relative free energy differences (ΔG) were calculated between each observed state and the state exhibiting the highest probability density across corresponding apo and holo systems.

For quantification of global structural deformation in the SUE-βββ, the backbone RMSD for the corresponding amino acids (Cdc10_S237-E261, Cdc3_S345-D370, Cdc12_S248-E273, Shs1_S277-D298) was calculated after least square fitting of the investigated subunit on the equilibrated input structure. This was performed for each performed simulation independently. For the generation of distribution functions, the RMSD values obtained in the five independent 700 ns simulations for each system were combined separately for each SUE-βββ.

Input structures for COM-pulling simulations for both dimers were generated as described previously (Grupp et al., 2023) by pooling the last 100 ns of the five independent unbiased simulations and backbone RMSD-based clustering (Daura et al., 1998) after least-squares fitting. Clustering was performed with a cutoff of 0.2 nm except for the complex Cdc10_R251A:GDP-Cdc3_R360A:GDP:Mg. for which the cutoff was set to 0.25 nm as the lower cutoff yielded two dominant clusters of equivalent size. For each system, the central structure in the largest cluster was subsequently used for further simulations. Surrounding waters and closely associated ions within 0.6 nm around the protein complexes were also retained for the following steps. Resulting models were placed in a rectangular box with x/y/z dimensions of 20/10/10 nm aligning the filament axis with the x-axis. Subsequent box solvation, energy minimization and equilibration were performed as described for the unbiased simulations. After equilibration, COM-pulling simulations were carried out in the absence of position restraints. Analysis of different combinations of pull rate and spring constant showed a similar behavior as described for human septin dimers (Grupp et al., 2023) resulting in the use of the same spring constant (500 kJ mol⁻¹ nm⁻²) and a pulling rate (0.0075 nm ps⁻¹). Using these parameters low structural deformation and reproducible maximum forces were obtained. COM-pulling simulations, umbrella sampling and calculation of free energy profiles were again performed as described in our previous work but at a temperature of 300 K and sampling COM distances until ∼ 8.5 nm (Grupp et al., 2023).

Plasmids and strains

All plasmids used and/or created in this study are listed in Table 2. All generated plasmids were verified by sequencing.

Plasmids for cell culture were created by modifying the EGFP-SEPT9 expression plasmid #1, kindly provided by Elias Spiliotis (University of Virigina, School of Medicine, VA, USA). Point mutations were introduced via suitable primers, making use of an internal BamH1 restriction site located close to the respective codons, yielding plasmids #2 and #3, respectively.

For coexpression studies of human septins in Escherichia coli, the G-domains of SEPT7 and SEPT9 were cloned into the in-house generated pET15b derivatives pES and pAc-MBP, allowing for simultaneous expression in E. coli due to compatible antibiotic resistance cassettes (Amp and Kan, respectively) and origins of replication (pBR322 and p15A, respectively). Inserts with point mutation were generated by SOE-PCR (SEPT7) using available plasmids as template (Fischer et al., 2022) or amplifying from #2 (SEPT9).

For expression of monomeric human and yeast septins for pulldown experiments, the respective G-domains were PCR amplified from available plasmid templates and cloned as His₆- or His₆-MBP tagged constructs into the in-house generated pET15b derivatives pES and pAc or into the first MCS of pET-Duet1 (Novagen). Inserts with point mutation were generated by SOE-PCR.

For growth assays in yeast, the ORF of Cdc10, Cdc3, Cdc12, Cdc11, and Shs1 with and without R(βb)(A) mutation were cloned into the centromeric plasmid pRS313 (Sikorski and Hieter, 1989) enabling expression under the control of a P_MET17 promoter yielding plasmids #29-#38. Inserts with point mutation were generated by SOE-PCR.

The ORF of Cdc10, Cdc3, Cdc12, Cdc11, and Shs1 were cloned into the centromeric plasmid pRS316 (Sikorski and Hieter, 1989) bearing an Ura3 prototrophy and a P_MET17 promoter, yielding plasmids #24-#28. These rescue plasmids were transformed into the yeast strain JD47 (Dohmen et al., 1995) by standard methods.

The WT copy of the respective septin in the resulting strains was replaced by a NAT cassette as described elsewhere (Dünkler et al., 2021) and plasmids #29-#38 were subsequently transformed into the respective ko strains bearing the rescue plasmids.

The strains were grown in YPD medium overnight and serial dilutions were subsequently spotted on suitable selection media containing 5-FOA, selecting for cells without the plasmid, and on respective control media without FOA.

For microscopy, the ORF of Cdc10, Cdc3, Cdc12, Cdc11, and Shs1 (with and without R(βb)(A) mutation) were cloned das GFP fusion into the plasmid pRS306 (Sikorski and Hieter, 1989) under the control of a P_MET17 promoter, yielding plasmids #39-#48. The plasmids were linearized with StuI and subsequently transformed into the haploid yeast strains JD47 and ULM53 (Grinhagens et al., 2020) bearing a Shs1-mCherry fusion (Schneider et al., 2013), resulting in homologous recombination into the native Ura3 locus.

Protein expression and purification

The following human septin dimers, truncated to their G-domains, were employed for coexpression, with the respective compatible plasmids in the same cell: His₆-SEPT9G (#6) + MBP-SEPT7G (#8), His₆-SEPT9G(R516A) (#7) + MBP-SEPT7G (#8), His₆-SEPT9G (#6) + MBP-SEPT7G(R265A) (#9), His₆-SEPT7G (#10, SEPT7 homodimer), His₆-SEPT7G (R265A) (#11).

The respective septin G-domain heterodimers and homodimers were expressed in E. coli BL21DE3 LOBSTR in super broth medium at 18°C for ∼ 20 h after induction with 1 mM IPTG.

Cells were resuspended in IMAC A buffer (25 mM Tris pH8.0, 300 mM NaCl, 5 mM MgCl₂, 15% (vol/vol) glycerol), supplemented with 0.1% Tween20 and e-complete protease inhibitor cocktail (Roche), and lysed by lysozyme treatment and sonication. The resulting extract was submitted onto a 5 ml HisTrap Excel column (Cytiva) which was subsequently washed with 3% IMAC buffer B (25 mM Tris pH8.0, 300 mM NaCl, 200 mM imidazole, 5 mM MgCl₂, 15% (vol/vol) glycerol) until a stable A₂₈₀ baseline was obtained. The His₆-tagged protein complexes were then eluted with 100% IMAC buffer B.

The product peak was pooled and directly submitted to analytical size exclusion chromatography on a Superose6 column (Cytiva) with 25 mM Tris pH8.0, 300 mM NaCl, 5 mM MgCl₂ as mobile phase. The dimerization state of the eluting proteins was judged according to a gel filtration calibration standard (Bio-Rad).

Septin monomers for pulldown were expressed as described above.

#6, #7, #10, #11, #12, #13, #18-#23 expressing cells were resuspended in IMAC mono-A buffer (50 mM Tris pH8.0, 500 mM NaCl, 5 mM MgCl₂), supplemented with 0.1% Tween20 and e-complete protease inhibitor cocktail (Roche), and lysed by lysozyme treatment and sonication. The resulting extract was submitted onto a 5 ml HisTrap Excel column (Cytiva) which was subsequently washed with IMAC mono-A including 3% IMAC buffer mono-B (50 mM Tris pH8.0, 500 mM NaCl, 200 mM imidazole, 5 mM MgCl₂) until a stable A₂₈₀ baseline was obtained. The proteins were eluted with 100% IMAC mono-B. For all SEPT7 constructs (#10, #11, #12, #13), IMAC mono-B was supplemented with 1M MgCl₂ to dissociate SEPT7 homodimers into monomers.

Eluates of #12 and #13 tended to precipitate under these conditions and were centrifuged for 5 min at 16,000 × g immediately before SEC.

Cells expressing His₆-MBP-Cdc10 (#14 or #15) and His₆-MBP-Cdc12 (#16 or #17) were lysed in IMAC buffer lowsol-A (50 mM Tris-HCl pH 8.0, 500 mM NaCl, 5 mM MgCl₂, 50 mM L-Arg, 50 mM L-Glu, 15% Glycerol, 0.1% Tween20) supplemented with e-complete protease inhibitor cocktail (Roche) and lysed by lysozyme treatment and sonication. The resulting extract was submitted onto a 5 ml HisTrap Excel column (Cytiva) which was subsequently washed with IMAC lowsol-A including 3% IMAC lowsol-B (50 mM Tris-HCl pH 8.0, 500 mM NaCl, 5 mM MgCl₂, 50 mM L-Arg, 50 mM L-Glu, 15% Glycerol, 200 mM imidazole 0.1% Tween20) until a stable A₂₈₀ baseline was obtained. The proteins were then eluted with 100% IMAC lowsol-B.

The product peaks of all septin monomers were pooled and subjected to preparative SEC on a Superdex200 column (Cytiva) with 25 mM Tris pH8.0, 300 mM NaCl as mobile phase.

All purified proteins were assayed by SDS–PAGE on a 4–12% BisTris gradient gel (Life Technologies) with subsequent Coomassie brilliant blue staining.

Pulldown experiments

To enable complex formation, 2 µM MBP-tagged protein was incubated with 10 µM corresponding His₆-tagged protein in pulldown buffer (25 mM Tris pH8.0, 300 mM NaCl, 5 mM MgCl₂, 100 µM GTP, 100 µM GDP) in a total volume of 500 µl and incubated o/n at 4°C.

The complexes were then captured onto 30 µl amylose resin (NEB Biolabs) previously equilibrated with pulldown buffer, washed, and subsequently eluted with 80 µl MBP-elution buffer (25 mM Tris pH8.0, 300 mM NaCl, 5 mM MgCl₂, 10 mM maltose).

The eluates were subjected to SDS–PAGE on a 4-12% BisTris gradient gel (Life Technologies) and subsequently stained with Coomassie brilliant blue.

Mammalian cell culture and microscopy

A total of 1306 fibroblasts (immortalized human dermal fibroblasts, female Puerto Rican donor, kind gift of Sebastian Iben, Department of Dermatology, Ulm University Hospital, Germany) and a SEPT9 ko cell line in the same cellular background, generated previously in our laboratory (clone 2c3, Hecht et al., 2024), were used throughout this study. Cells were maintained at 37°C and 5% CO₂ in a humidified incubator in DMEM (Life Technologies) supplemented with 10% FBS (Capricorn). Plasmid transfection was performed using Lipofectamine 3000 (Invitrogen) according to the manufacturer's instructions. Stable cell lines were selected by treatment 750 µg/ml geneticin G418 (Formedium) 48 h posttransfection. The antibiotic concentration was lowered to 250 µg/ml after ∼ 2 wk and the cells were further maintained under these conditions. Immunfluorescence, transwell migration assays and microscopy were performed essentially as described in previous work (Hecht et al., 2024). The following antibodies were used in this study: The endogenous actin cytoskeleton was stained using AlexaFluor555-Phalloidin (Invitrogen, #A34055). The endogenous septin cytoskeleton was stained using an anti-SEPT8 antibody (produced in rabbit, kind gift of Michael Krauss, Molecular Pharmacology and Cell Biology Group, Leibnitz Research Center for Molecular Pharmacology, Berlin, Germany) combined with an AlexaFluor647 conjugated anti-rabbit secondary antibody produced in goat (Invitrogen, #21244). A total of 1306 fibroblasts contain SEPT8 as member from the 1b group instead of SEPT6 (Hecht et al., 2019). The anti-SEPT8 antibody stained septin cytoskeleton colocalized with DsRed-SEPT9 or GFP-SEPT2 decorated septin filaments in 1306 fibroblasts (Supplemental Figure S10), proving specificity of the antibody in the employed cell line. Filopodia were stained with an anti-VASP antibody produced in rabbit (Invitrogen, #MA5-35728) combined with an AlexaFluor647 conjugated anti-rabbit secondary antibody produced in goat (Invitrogen, #21244).

For Western blot analysis, confluent cells from one T25 cell culture flask were collected and lysed in 100 µl lysis buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 0.2 mM EDTA, 0.1% Triton X100) supplemented with protease inhibitor cocktail (Roche). Cell debris was removed by centrifugation for 20 min at 16,000 × g at 4°C and the protein concentration in the soluble extract was determined by Bradford assay. Equal amounts of protein for each tested cell lysate were loaded onto a 4–12% BisTris gradient gel (Life Technologies) which was subsequently blotted onto nitrocellulose. SEPT9 was detected using an anti-SEPT9 antibody (Sigma, #WH0010801M1) and an anti-GAPDH-HRP conjugate (Santa Cruz Biotechnology; #sc47724) was employed as loading control.

Microscopy of yeast strains

Strains were grown overnight in suitable selection medium containing 70 µM Methionine, diluted subsequently in 3 ml fresh medium and grown for 1.5–2 h at 30°C to midlog phase. Directly before the microscopy analysis 1.5 ml of the cultures were spun down and resuspended in 100 µl sterile H₂O. A total of 3.5 µl of the suspension were spotted on a glass slide and immobilized with a coverslip. Microscopy was performed with a 63x magnification objective with a series of 10 equally spaced z-slices, each separated by 0.26 µm. The same microscope setup as for the immunfluorescence was used (Hecht et al., 2024). Image analysis was conducted in Fiji. Sum projections of the acquired Z-planes from both channels were generated for subsequent analysis. To quantify budneck signals, an elliptical region of interest (ROI) measuring exactly 8 pixels in length and 6 pixels in width was positioned over the budneck, aligning the longer axis with the Shs1-mCherry signal and the shorter axis with the mother-daughter axis. Cytoplasmic signals in the mother and daughter cells were measured by placing an elliptical ROI over the remaining cell volume of the mother and daughter cells, respectively. Care was taken that these ROIs didn't incorporate budneck associated fluorescence signal in either channel. Only cells in G2/M phase displaying clearly discernible Shs1-mCherry signal at the budneck were considered. Finally, the average intensity within each ROI was calculated for both channels. The ability of septin variants to be incorporated in septin filaments at the budneck was quantified using the following equation:

Using this equation, budneck signals in both channels were corrected by subtracting the average cytoplasmic signals from the mother and daughter cells. GFP accumulation was then normalized relative to Shs1-mCherry accumulation.

Data availability

Sequences of the employed oligonucleotide primers for cloning are available from the authors upon request. The refined structure of the yeast septin tetramer was deposited in the PDB under identifier 9GD4. Raw data underlying the figures including the full alignment containing all 254 sequences (Figure 7), sequences of recombinant DNA and the MD data that support the findings of this study are openly available in Zenodo at doi.10.5281/zenodo.14622194

Abbreviations used:

COM

center of mass

ko

knock out

MBP

maltose binding protein

MD

molecular dynamics

ORF

open reading frame

RMSD

root mean square deviation

SEC

size exclusion chromatography

SUE

septin unique element

WT

wild type.