The ORFs of the isoforms i1 of SEPT2 and SEPT9 (or SEPT91-568) as well as SEPT7 and SEPT6 were cloned into two compatible bicistronic plasmids. SEPT6 was constructed with a N-terminal 6His-tag for IMAC purification while the other subunits were untagged.

Protein expression of hexameric and octameric rods was performed in the E. coli strain BL21DE3 in SB medium at 18°C. Protein purification from the crude extract was performed by immobilized metal affinity chromatography (IMAC) and subsequent size exclusion chromatography (SEC).

Integrity of the purified complex was determined by density gradient centrifugation and MS analysis. Western blot analysis was performed using an anti-SEPT9 antibody (mouse, clone 2C6, Sigma Aldrich). Detailed protocols on the cloning-expression and purification procedures as well as on the analytical assays can be found in the supporting methods.

Septin filament formation was performed as described elsewhere for yeast septins (Renz et al., 2013). Briefly, septin preparations were adjusted to 2 µM and cleared from aggregates by centrifugation. Filament formation was induced by dialysis into 50 mM Tris pH 8.0, followed by incubation at 4°C overnight and subsequent centrifugation at 100.000 xg. The pellet was resuspended in 50 mM Tris pH 8.0 and either subjected to SDS-PAGE or to negative staining for electron microscopy. Therefore, the protein solution was spotted onto a carbon coated TEM grid and negative stained with uranyl acetate using standard procedures. EM imaging was performed on a JEM 1400 TEM instrument (Jeol) at 120 kV acceleration voltage. The camera resolution was 2000 x 2000 px.

The nucleotide content of the rod preparations was determined by analytical anion exchange chromatography. Briefly, purified protein from the SEC was dialyzed into buffer without salt and subsequently heat denatured to release bound nucleotide. The denatured protein was pelleted and the supernatant was loaded onto a MonoQ HR 5/5 column (Amersham Pharmacia Biotech). The nucleotides were eluted by a linear salt gradient and detected at 254 nm. GTP and GDP were used to determine the elution profiles of the pure compounds.

Uptake of [γ³²P]-GTP to hexameric and octameric rods and isolated SEPT9 was performed as described elsewhere for yeast septins (Baur et al., 2018). Briefly, septin preparations were incubated with [γ³²P]-GTP in exchange buffer containing 5 mM EDTA. For hydrolysis, the exchange reaction was transferred into hydrolysis buffer containing MgCl₂. Radioactivity retained in the native septin proteins during uptake or during hydrolysis was detected with a filter binding assay.

Further details of the procedure and the subsequent data processing are outlined in the supporting methods.

Pyrene-labeled G-actin (shortly pyrene actin) was prepared from rabbit muscle acetone powder as described elsewhere (Cooper et al., 1983) with some modifications outlined in the supporting methods. Pyrene actin fluorescence increases upon actin polymerization and can be recorded in actin polymerization buffer at an excitation of 365 nm and emission of 407 nm. Details on the pyrene actin preparation, the assay protocol and data processing are outlined in the supporting methods.

The unprocessed raw data are shown in Supplementary Figure S4.

We expressed the subunits of the human octameric and hexameric septin rods in E.coli from two compatible bicistronic plasmids. SEPT6 and SEPT7 were expressed from one plasmid and SEPT2 (and SEPT9 for octameric rods) from the second plasmid as in a previous study (Sheffield et al., 2003). Only SEPT6 carried a hexa-histidine tag for purification by immobilized metal affinity chromatography (IMAC). The enriched fractions were subsequently subjected to size exclusion chromatography (SEC). Complexes eluted in a shoulder in the second half of the main peak (the first half representing the void peak of the column) and in a minor peak directly adjacent to the main peak (Supplementary Figure S1A).

SDS-PAGE and Western blot analysis with an anti-SEPT9 antibody documented the presence of shorter fragments of SEPT9 (Figure 1A). SEPT9 has two flexible regions, a long extension at the N-terminus and a shorter one at the C-terminus. We suspected one or both of these extensions to cause the shorter fragments. The long SEPT9 isoforms used by Iv et al. (2021) in their rod preparations contained both the N-terminus or larger parts thereof, but their SEPT9i3 terminates at residue Glycine 568. This residue represents the terminal residue of the α6 helix of the G domain. We decided to test first if a version of SEPT9 truncated at Glycine 568 (SEPT9_G568) still produces degradation products in our purification pipeline. Octameric rods expressing SEPT9_G568 showed less degradation products than the native octamer and the purified complex could be readily separated from other subcomplexes and contaminants by the preparative SEC (Figure 1A). The IMAC enriched septin preparation was analyzed on a glycerol gradient centrifugation to confirm the integrity of the complex. Septin subunits, SEPT9 degradation products and contaminants migrated in the fractions corresponding to a lower molecular weight whereas the stoichiometric combination of all four subunits migrated just before the 670 kD marker protein became abundant, indicative of an octameric septin complex (Supplementary Figure S1B).

We analyzed the fractions from the major and minor peak from the SEC of SEPT9_FL and SEPT9_G568 complexes by mass spectrometry (MS). All four subunits are present in nearly stoichiometric abundance (Figure 1B). We conclude that our septin preparation in the main peak is indeed an octameric complex containing all four subunits. We performed the MS analyses with several purified complexes. The peptide intensity scores of the individual subunits varied slightly between the preparations, but we observed in all MS runs that SEPT9 was less abundant in the complex of the minor SEC peak. As this subcomplex poorly formed filaments (see below), all subsequent experiments were conducted with a septin preparation from the main peak.

We subsequently determined the nucleotide content of our rod preparations. Bound nucleotide was released from the protein by heat denaturation and afterwards analyzed by analytical anion exchange chromatography (Figure 1C). We determined the peak integrals and calculated the ratio of GDP to GTP. Both SEPT9_FL and SEPT9_G568 complexes contained between 3.5 and 5 times more GDP than GTP (varying between several purifications performed). Overall, we could not detect any difference in SEPT9 and SEPT9_G568 containing octamers regarding the nucleotide content after purification.

Hexameric SEPT2/6/7 complexes were purified using the same purification pipeline and yielded a highly purified after SEC (Supplementary Figure 1C).

We asked next whether septin filament formation is altered by the absence of the SEPT9 C-terminal extension. Septin preparations were cleared from aggregates by centrifugation, dialyzed into low salt buffer and subjected subsequently to ultracentrifugation. Surprisingly, we could detect filaments already at 300 mM NaCl for SEPT9_FL octamers (Supplementary Figure S2A) whereas filaments of SEPT9_G568 containing octamers and hexamers required 50 mM NaCl to form (Supplementary Figures S2B, C). Filament formation was almost abolished at 100 mM NaCl. Negative stain EM of SEPT9_FL octamers revealed the presence of long filament bundles consisting of parallel aligned septin filaments, consistent with the findings of Iv et al. (2021) (Figure 2A). SEPT9_G568 containing filaments formed also bundles, but the filaments forming the bundle appeared rather curved and less parallel aligned (Figure 2B). Complexes isolated from the minor peaks of the SEC only poorly formed filaments (data not shown) which we were not able to visualize by EM.

The septin subunit SEPT9 was reported to bundle actin and microtubules (Bai et al., 2013; Dolat et al., 2014; Mavrakis et al., 2014; Smith et al., 2015).

To investigate the influence of septin rods on the formation of the actin cytoskeleton in vitro, we performed actin polymerization assays with 2 µM pyrene labeled G-actin and recorded the increase of fluorescence over time. G actin without additives or with purified GST (serving as control to exclude unwanted side effects by crowding) displayed a typical sigmoidal polymerization curve (Supplementary Figure S3A) (Cooper et al., 1983) which can be described mathematically by a growth curve including lag phase, exponential growth and plateau phase. From this curve, parameters such as the rate constant, the lag time or the doubling time can be calculated. All parameters calculated from the polymerization curves are summarized in Supplementary Table S1. Actin polymerization did not occur in G-buffer without salts (Supplementary Figure S3B). The presence of hexameric septins increased the lag time in a concentration dependent manner, leading to a shift of the inflection point relative to the inflection point of the actin curve (Figure 3A). For octamers, the shift was more pronounced at lower concentration but did not change significantly at higher septin concentrations (Figure 3B). More important, the actin monomer turnover in the elongation phase (represented by τ, the inverse value of the rate constant) was slightly decreased about 1.4 fold for hexameric rods and 0.5 µM octamer relative to actin alone and more pronounced decreased by 2.1 fold for 1 µM octamer. Increased τ values are reflected by a shallower slope of the curve.

A spin down assay with septin octamers previously dialyzed in G buffer showed that septin filaments did not measurably bind G-actin (Supplementary Figure 3C).

We have previously measured the uptake of [γ³²P] labelled GTP in the presence of EDTA by the hexameric- and octameric septin rods from yeast (Baur et al., 2018). All following experiments were conducted with SEPT9_G568 containing octamers. [γ³²P]-GTP uptake followed an exponential one phase association kinetics from which the rate constant and the reaction half time t_1/2 can be calculated. The reaction half time t_1/2 was 10 min for hexameric rods and 16 min for octameric rods, respectively (Figure 4A). All t_1/2 values of this and the subsequently performed experiments are summarized in Figure 4C and the full evaluation of all performed measurements are shown in the supporting information (Supplementary Tables S2–S4).

Unlike for yeast septin rods, adding MgCl₂ to the exchange reaction after 20 min did not change the overall shape of the uptake kinetics but prolonged t_1/2 slightly (and barely significantly) to 14.5 min for hexamers and 21 min for octamers (Supplementary Figure S5A and Figure 4C).

To evaluate whether the bound nucleotide remains in the septin rod or exchanges with free GTP, we pre-loaded septin preparations with non-radioactive GDP or GTP and removed excess nucleotide by micro-dialysis. Subsequent nucleotide exchange against [γ³²P]-GTP showed a shift in reaction half times to 42 min for both GDP and GTP loaded hexameric rods and to 33 min for GDP loaded octameric rods and 38 min for GTP loaded octameric rods, respectively (Supplementary Figures S5B, C and Figure 4C).

We also subjected the octameric rods to exchange of previously bound [γ³²P]-GTP against non-radioactive nucleotide. After loading, excess nucleotide was removed via a desalting column and the uptake of non-radioactive GTP was monitored via the decrease of protein-bound radioactivity. The calculated reaction half time of 46.5 min was slightly (and not significantly) slower than the t_1/2 for the inverse exchange reaction (GTP against [γ³²P]-GTP) (Figure 4B).

We conclude that human septin octamers exchange bound nucleotides quite readily.

Previous studies showed that isolated septin subunits may have other GTPase properties than septin rods (Sheffield et al., 2003). We tested the isolated SEPT9_FL subunit and the SEPT9 G-domain (SEPT9_Q295-E567) for nucleotide uptake and found that both did not accept [γ³²P]-GTP (Figure 4D).

To find out whether human septins follow the same GTP hydrolysis mechanism like other small GTPases, we measured the released or protein-bound radioactivity by [γ³²P]-GTP preloaded septins under conditions used for other small GTPases (i.e. magnesium and an excess of non-radioactive GTP to avoid unwanted re-association of [γ³²P]-GTP molecules) (Tucker et al., 1986; Schweins et al., 1995). In contrast to common small GTPases of the Ras family, neither hexameric nor octameric rods showed a release of hydrolyzed [γ³²P] under these conditions in the applied time frame, regardless of the applied reaction temperature (25°C or 37°C, respectively) (Figure 4E and Supplementary Figure S6).

Human hexameric septin rods are available for in vitro experimentation and even crystallization for several years (Sheffield et al., 2003; Sirajuddin et al., 2007). However, purified octameric rods containing SEPT9 were only very recently obtained (Iv et al., 2021). We introduce here an additional two step purification protocol for human octameric rods harboring a hexa-histidine tag at only one of its subunits. Gycerol density gradient centrifugation of the complex after the first purification step separated contaminants and septin containing subcomplexes from the octamer, which migrated at the expected density in the gradient. After preparative SEC, we obtained the purified octamer including all four subunits in nearly stoichiometric abundance as judged by MS analysis.

The central subunit SEPT9 is prone to C-terminal degradation during expression and purification of the rod. Removal of the terminal 18 amino acids stabilized the protein and led to less degradation products. To our surprise we found that the SEPT9_FL containing rods formed filaments already at 300 mM NaCl whereas filament formation by hexamers and SEPT9_G568 containing octamers was abolished already at 100 mM NaCl. The first study of filament formation of SEPT2/6/7 hexamers raised the expectation of another—at that time largely unknown - septin subunit (SEPT9) with a stabilizing effect on the filaments (Sheffield et al., 2003). Based on our herein presented experimental data we propose that the stabilizing effect of the SEPT9 subunit on filament formation is mediated by the α6 helix and its C-terminal extension. The α6 helix of the SEPT9 G domain is an essential component of the central SEPT9-SEPT9 NC interface, contributing to an interaction network with the α2 helix of the neighboring subunit (Castro et al., 2020). The C-terminal extensions would point out from the interface, maybe interacting with each other they further stabile the interface. We and others (Iv et al., 2021) postulate that the assembly of higher ordered septin structures is driven in vitro by the N- and C-terminal extensions.

Several studies have already demonstrated the capability of SEPT9, septin hexamers and octamers to bind and bundle filamentous actin (Mavrakis et al., 2014; Smith et al., 2015; Iv et al., 2021). All these studies monitored the effects of septins on filamentous actin. Actin bundling was detected subsequently by electron- or fluorescence microscopy. The pyrene actin assay introduced in this study looks at actin polymerization in real time, however, it measures the net polymer weight and does not provide information on the longitudinal distribution of the filaments (Cooper et al., 1983). It was recently shown by in vitro microscopy that recombinant septin rods cross-link actin filaments into straight and curved bundles (Iv et al., 2021). Our assay revealed that the presence of septin rods decreases the rate of actin polymerization. As we could not detect any interaction between septin rods and G-actin we propose that the septins exert their influence on polymerization through interaction with the forming actin filament and subsequently bundle filaments by lateral joining. Why this effect is more pronounced for octamers requires further investigation. Maybe the interaction of the octamer with the actin filament is favored over hexamer binding.

Whether SEPT9 is a bundling factor on its own is currently under debate (Mavrakis et al., 2014; Smith et al., 2015; Iv et al., 2021). As the influence on actin monomer turnover is more pronounced for octamers, we support a the idea of a direct contribution of SEPT9 on actin polymerization.

Septins were first labeled as GTP binding proteins in Drosophila (Field et al., 1996) and afterwards nucleotide binding was confirmed also for human and yeast septin complexes (Sheffield et al., 2003; Vrabioiu et al., 2004) and subsequently evaluated in further studies (Farkasovsky et al., 2005; Baur et al., 2018). However, the functional role of nucleotide binding- and hydrolysis of the septins in cellular processes and in septin (proto-)filament assembly is still under debate and not well understood.

The availability of yeast and human septin octamers enabled us to measure nucleotide association kinetics in these protein complexes and to compare the properties of both species. GTP-loading in human septin rods remains on a constant level after reaching a plateau whereas in the yeast septin rods the radioactive signal decreases about 50% after reaching a maximum (Baur et al., 2018). This indicates the release of the γ-phosphate or the entire nucleotide in at least some subunits of the protofilaments. Mg²⁺ suppresses this release but shows no significant effect on the GTP association rate in the yeast septin rods. In contrast, Mg²⁺ slightly slows down the GTP binding in human septin rods. In the absence of Mg²⁺, hexameric and octameric rods from both species are able to exchange previously bound GTP and GDP. The exchange rates for GDP and GTP of the human septin rods do not differ significantly. In contrast, the S. cerevisiae septin rods exhibit a much faster GDP-GTP than GTP-GTP exchange. These discrepancies indicate that, although displaying a high structural similarity, the septins from different species follow each an individual biochemistry. Even within one species the different subunits exhibit different properties. Comparing the GTP and GDP loaded structures of SEPT2 (PDB IDs 3FTQ and 2QNR, respectively) reveals a conformational change in the Switch1 region with the invariant Thr78 and Mg²⁺ being in close contact with the γ-phosphate exclusively in the GTP conformation. In the GDP conformation Switch1 is either unstructured or oriented away from the nucleotide (Supplementary Figure S7A). This conformational change cannot be seen in the structures of SEPT9 (PDB IDs 5CYP and 5CYO) where the corresponding Thr64 in Switch1 is associated with Mg²⁺ towards the nucleotide in both nucleotide states (Supplementary Figure S7B). This inflexibility of Switch1 might explain the inability of the isolated SEPT9 to accept [γ³²P]-GTP and why we do not detect significant differences in nucleotide uptake between hexamers and octamers. However, it should be kept in mind that the mentioned structures do not contain a native G interface and thus they can provide only indications explaining our experimental findings.

GTP hydrolysis is routinely detected in small GTPases of the Ras family. Experimentally, the small GTPase is loaded with GTP in the presence of EDTA and then subjected to a Mg²⁺ containing buffer without EDTA. The Mg²⁺ ion is an indispensable component for the GTPase reaction. The γ-phosphate is coordinated by invariant threonine and glycine residues in the Switch1 and Switch2 regions. The conformational change following the GTPase reaction was termed “loaded-spring mechanism” which allows the release of the γ-phosphate after hydrolysis and the relaxation of the switch regions into the GDP conformation (Vetter and Wittinghofer 2001). We were not able to detect the release of [γ³²P]-GTP or hydrolyzed [γ³²P] from any rod species examined. This indicates that bound nucleotides and presumable hydrolysis products remain in the binding pocket and are not released into the surrounding medium, at least not in the timeframe of the experiment (120–180 min) and at least not under the conditions applied. This is more in line with our findings for yeast septins (Baur et al., 2018) rather than with the properties of small GTPases of the Ras family.

GTPase hydrolysis products were detected in yeast and mammalian septin preparations after denaturation of the protein (Sheffield et al., 2003; Versele and Thorner 2004; Sirajuddin et al., 2009; Castro et al., 2020). In contrast, our assay was conducted under native conditions. Consequently, hydrolysis occurs either at very slow pace, undetectable within the chosen timeframe, or the hydrolysis products remain in the active site after the reaction took place. This raises the question which role nucleotides and their hydrolysis play. This question was addressed in yeast septins and it was proposed that nucleotide hydrolysis plays a role during formation of the rod (Weems and McMurray 2017). This would indeed imply that one GTP hydrolysis event is enough and that cycling between the GDP and GTP bound state would not be necessary. However, more studies both in vivo and in vitro are necessary to convincingly address this issue for human septins.

Taken together, we propose that GTP binding and hydrolysis in septins follow a different mechanism than in other small GTPases and that septin rods from budding yeast differ mechanistically in their guanine nucleotide interactions from their human counterparts. More structural information from yeast septins is needed to understand these differences.