Cellular functions of actin- and microtubule-associated septins

SUMMARY

Septins are an integral component of the cytoskeleton, assembling into higher-order oligomers and filamentous polymers that associate with actin filaments, microtubules and membranes. Here, we review septin interactions with actin and microtubules, and septin-mediated regulation of the organization and dynamics of these cytoskeletal networks, which is critical for cellular morphogenesis. We discuss how actomyosin-associated septins function in cytokinesis, cell migration and host defense against pathogens. We highlight newly emerged roles of septins at the interface of microtubules and membranes with molecular motors, which point to a ‘septin code’ for the regulation of membrane traffic. Additionally, we revisit the functions of microtubule-associated septins in mitosis and meiosis. In sum, septins comprise a unique module of cytoskeletal regulators that are spatially and functionally specialized and have properties of bona fide actin-binding and microtubule-associated proteins. With many questions still outstanding, the study of septins will continue to provide new insights into fundamental problems of cytoskeletal organization and function.

In Brief:

Elias Spiliotis and Konstantinos Nakos review the interactions of septins with actin and microtubules, and the septin-mediated regulation of the organization and dynamics of these cytoskeletal networks, which is critical for cellular morphogenesis.

Introduction

Alongside microtubules, actin and intermediate filaments, septins comprise a distinct filamentous network with important functions in cellular morphogenesis and physiology^1–4. Septins were originally discovered in the budding yeast Saccharomyces cerevisiae as a network of filaments that associates with the mother–bud neck domain of the plasma membrane and controls the localization of many proteins with roles in cell division and growth^2,5,6. While absent from plants, septin filaments are ubiquitous in fungi and animal cells, where they localize to the cytoplasm as well as on cell membranes^7–11. Intriguingly, septins associate with subsets of actin filaments and microtubules^12,13. Over the last twenty years, it has become evident that septins regulate the spatial organization and functions of the actin and microtubule networks in a diversity of cellular processes.

Septins are a family of GTP-binding proteins that assemble into oligomeric complexes and polymers that lack end-to-end polarity and are more stable than actin filaments and microtubules^2,3,14,15. Encoded by paralogous genes that expanded through evolution, septins are classified into four groups that vary mainly in GTPase activity and in the amino- and carboxy-terminal extensions of their core GTP-binding domain (Figure 1A)^16,17. Septins assemble in a combinatorial fashion by oligomerizing in tandem through homomeric and heteromeric interactions of their GTP-binding domains; this oligomerization is partly instructed by GTP binding and hydrolysis^18–23. Yeast septins assemble into an octamer, a palindromic dimer of a heterotetrameric complex (Cdc10–Cdc3–Cdc12–Cdc11 or Cdc10–Cdc3–Cdc12–Shs1), which is the repeating unit of septin polymers²⁴. Similarly, mammalian septins form heterooctamers with subunits from one of each of the four septin groups (Figure 1B)^25–28. Paralogs of the same septin group can replace one another, generating septin complexes of various composition and combination^17,28,29. In mammalian systems, smaller unit complexes (such as hexamers and tetramers) arise from non-canonical modes of assembly as a result of paralogs having different rates of GTP hydrolysis and/or expression levels^30–32. Septins associate preferentially with membrane domains of micron-scale curvature and subsets of actin filaments and microtubules (Figure 1C)^33–38. These interactions are mediated by septin paralogs with unique domains and biochemical properties. For example, paralog-specific amphipathic helices and repeat motifs mediate binding to curved membrane domains and microtubules, respectively^33,34. Thus, septins are a highly diverse and modular network of GTPases, the localization, interactions and functions of which are determined by their individual subunits.

Figure 1. Septin domains and assembly.

(A) The mammalian septin family consists of thirteen paralogs, which are classified into four groups: SEPT2 (SEPT1, SEPT2, SEPT4, SEPT5), SEPT3 (SEPT3, SEPT9, SEPT12), SEPT6 (SEPT6, SEPT8, SEPT10, SEPT11, SEPT14), and SEPT7. Septin paralogs have a highly conserved GTP-binding domain (G-domain) followed by a unique sequence termed the septin unique element (SUE). Septins of the SEPT6 group do not hydrolyze GTP, while some paralogs (e.g., SEPT9) have faster rates of GTP hydrolysis. The amino- and carboxy-terminal extensions (NTEs and CTEs) of the GTP-binding domains are highly variable. CTEs contain α-helical coiled-coil domains, and SEPT9 contains a unique and alternatively spliced NTE, which interacts directly with actin filaments and microtubules. This NTE sequence is divided into basic and acidic proline-rich domains and consists of a repeat motif (K/R-R/x-x-D/E) that interacts with the acidic carboxy-terminal tails of tubulin. Septins also contain polybasic domains and amphipathic helices, with which they bind to membrane bilayers of distinct phospholipid content and curvature. (B) Septins homodimerize and heterodimerize in tandem through two alternating interfaces of their GTP-binding domains, assembling into non-polar palindromic oligomers. The prevailing oligomeric unit is a dimer of tetramers consisting of a paralog from each of the four septin groups (SEPT2, SEPT3, SEPT6, SEPT7). Septins of the same group can replace one another, generating octamers of variable combinations. Depending on septin expression levels and rates of GTP hydrolysis, which influence the homodimeric and heterodimeric interactions of the G-domains, smaller unit oligomers can form (e.g., SEPT2–SEPT6–SEPT7 heterohexamers). Additionally, oligomeric complexes of atypical combinations that consist of paralogs of the same septin group have been reported. (C) Septin oligomers associate with actin, microtubules and cell membranes. Septins form higher-order filamentous networks on cell membranes and may also polymerize into filaments on the surface of actin and microtubules.

Here, we focus on actin- and microtubule-associated septins, and their functions in fungal and animal cells. We review how septins interact indirectly and directly with actin filaments. We highlight roles of septins in actomyosin assembly and contractility and describe how these roles are adapted in mechanisms of cytokinesis, cell migration and bacterial infection. Shifting to microtubule-associated septins, we review their functions in the nucleation, dynamics and post-translational modifications of microtubules and how these effects promote cellular morphogenesis. We discuss recent evidence on the regulation of kinesin- and dynein-driven transport by septins, postulating the existence of a ‘septin code’ for the spatial control of intracellular traffic. Lastly, we provide an overview of microtubule-associated septin functions in mitosis and meiosis and conclude with outstanding questions that point to future advances.

Septins associate with cortical and cytoplasmic actin filaments

From fungi to animals, septins are widely observed to colocalize with actin filaments^9,12,39. Septins are critical for the assembly and maintenance of functional actomyosin networks at distinct regions of the plasma membrane and cytoplasm^2,12,39. These networks are commonly associated with saddle-shaped curvatures of the plasma membrane, which outline the neck of membrane protrusions resembling the S. cerevisiae mother–bud neck, and perinuclear as well as lamellar actin stress fibers^10–12. Direct septin roles in the nucleation, dynamics, capping and/or branching of actin filaments have yet to be elucidated; however, septins can regulate actomyosin organization and contractility by interacting with, scaffolding and/or clustering actin-binding proteins and signaling effectors, which provide feedback regulation (Figure 2).

Figure 2. Septin interactions with regulators of actin growth, organization, contractility and disassembly.

Summary of septin interactions with animal (top) and fungal (bottom) proteins that function in actin polymerization (formins FHOD1 and Bnr1), Arp2/3-mediated nucleation (cortactin, Las17/WASP), actin organization and crosslinking (anillin, α-actinin-4, Tea1, Rvs167, coronin, Hof1), actin contractility (myosin II, myosin essential light chain (ELC) and regulatory light chain (RLC)), and actin disassembly (cofilin, MICAL-1). Septin interactions with the effector proteins of Cdc42 (Cdc42EP/BORG) and the kinases CIT-K and ROCK2, which regulate myosin and formin activities, are also shown. Mammalian and fungal (S. cerevisiae, M. oryzae) septins are shown in purple ovals. Solid lines denote direct interactions, and dotted lines represent functional interactions that might be due to indirect or direct binding. Arrows point to phosphorylation and dephosphorylation events.

On fungal cell membranes, septins interact with actin indirectly via myosin II, which is linked to septins by the myosin-binding factor Bni5^40–43, and via actin-binding protein complexes that contain the formin Bnr1 as well as Bin/amphiphysin/Rvs (BAR) domain and ezrin/radixin/moesin (ERM) family proteins^44–46. In animal cells, by contrast, membrane and cytoplasmic septins interact both directly and indirectly with actin filaments^{32,37,47–51}. In vitro reconstitution assays demonstrated that insect septins (Drosophila Sep1–Sep2–Pnut) and mammalian septins (human SEPT2–SEPT6–SEPT7 and SEPT9) associate directly with polymerizing and pre-assembled actin filaments, crosslinking them into curved, circular and linear bundles^36,52,53. However, in vitro SEPT2–SEPT6–SEPT7 binding to actin filaments has also been shown to require the actin-binding protein anillin⁵⁰. It is unclear whether septins exist as oligomers or filamentous polymers on the surface of intracellular actin filaments, but the oligomeric and polymeric states of septins correlate with the in vitro formation of curved and linear bundles, respectively⁵². Hence, the actin-binding, -bending and -crosslinking properties of septins may depend on their oligomeric and polymeric states, which also have differential effects on microtubule dynamics^54,55.

In addition to binding linear actin, septins associate with Arp2/3-nucleated branched actin filaments. In the axons of sensory neurons, SEPT6 localizes to membrane actin patches, which consist of branched actin filaments, and promotes the recruitment of cortactin, an activator of Arp2/3-mediated actin polymerization³⁷. In vitro assays indicate that SEPT6 binding to actin filaments and branch points is enhanced in the presence of Arp2/3, while SEPT7 shows no such preference³⁷. In epithelial cells, SEPT6 overexpression increases the levels of cortactin in lamellipodia, which have also been reported to contain SEPT1, SEPT4 and SEPT5 in the absence of other septin paralogs^32,37. These findings suggest that septins may distinguish between actin networks of different organization in a paralog- and complex-specific manner.

The actin-binding domains of septins and the underlying mechanism(s) of their interaction with actin are not well understood. The actin-binding sequence of SEPT9 has been narrowed down to its amino-terminal basic domain, which does not appear to contain any known actin-binding motifs³⁶. Of note, SEPT9 has a uniquely long amino-terminal extension that is alternatively spliced, giving rise to a multitude of SEPT9 isoforms with different actin- and microtubule-binding properties^33,38,56. Electron microscopy studies have shown that the amino-terminal extension of SEPT9 isoform 1 (SEPT9_i1) binds actin using three different modes, two of which involve interactions with actin surface domains that are also bound by the ATP-bound myosin V subfragment 1 and the actin-severing protein cofilin (Figure 3A)³⁶. In contrast to these findings of direct septin–actin binding, septin association with actin is also mediated by anillin⁵⁰, non-muscle myosin II⁵⁷, and effectors of the small GTPases Cdc42 (Cdc42EP/Borg) and Rho (Rhotekin) (Figure 3B)^51,58–60. These factors recruit septins to actin or associate with septins and actin synergistically. Thus, septins can control the organization of actin filaments directly by physically crosslinking, bundling and/or bending them and indirectly through competitive or cooperative interactions with actin-binding proteins.

Figure 3. Septins interact with actomyosin and spatially regulate the organization of actomyosin networks.

(A) SEPT2 interacts directly with the coiled-coil domain of the heavy chain of non-muscle myosin II. SEPT9 binds actin filaments via its NTE basic domain, which can occupy three different regions of the actin surface domains (SD) 1, 2 and 4 (purple ovals and arrows). (B) Septin association with actin involves effectors of Rho (Rhotekin, anillin and septin-associated Rho guanine nucleotide exchange factor, SA-RhoGEF) and Cdc42 (Cdc42EP3). Anillin, which interacts with the myosin regulatory light chain, can recruit septins to actin filaments, while Rhotekin and Cdc42EP3 synergize with septins in binding actin. Actin-associated septins are posited to scaffold the kinases ROCK2 and CIT-K, which activate myosin II contractility by phosphorylating the myosin regulatory light chain. (C) In the budding yeast S. cerevisiae, septins localize at the mother–bud neck cortex. In the early stages of bud growth, septins form a network of filaments oriented along the axis of cell polarity. Septins scaffold and pattern the localization of the formin Bnr1 and the F-BAR protein Hof1 into evenly spaced pillars, which enable the formation of actin cables that are spaced apart and oriented toward the growing bud. In mitosis, septins initially provide a scaffold for the recruitment of myosin II and actomyosin ring (AMR) assembly. At the onset of cytokinesis, septins associate with the AMR indirectly through Hof1, which interacts with both septins and myosin II. Subsequently, septins split into two rings, flanking the AMR. (D) Infection of rice plants by the fungal pathogen M. oryzae requires the formation of the appressorium, a pressurized hyphal structure that penetrates into the cuticle. At the base of the appressorium, a septin ring organizes the formation of a toroidal actin network by corralling the localization of the I-BAR protein Rvs167 and the WASP homolog Las17, and scaffolding the ERM protein Tea1, which promotes actin–membrane association. (E) Septins are recruited to plasma membrane sites of S. Typhimurium attachment and provide a scaffold for the phosphorylation of the formin FHOD1 by the ROCK2 kinase.

Septins regulate actin organization on membranes by scaffolding and corralling actin-binding proteins

On the plasma membrane of yeast and animal cells, septins have evolutionarily conserved functions in scaffolding the organization of actomyosin filaments by controlling the localization and activity of actin nucleation-promoting factors, myosin II and actin-binding proteins (Figure 3C–E). In the early stages of bud growth of S. cerevisiae, septins associate with the mother–bud neck cortex, forming an hourglass-like structure consisting of filaments oriented along the mother–bud axis⁶. These septins (Cdc3–Cdc10–Cdc11 and Shs1) are organized as evenly spaced pillars that scaffold the localization of the formin Bnr1 and of Hof1, an F-BAR domain protein that binds and bundles actin (Figure 3C)⁴⁴. Septin-mediated patterning of Hof1 and Bnr1, the activation of which depends on septins and the septin-associated kinase Gin4, enables the capture and organization of actin filaments into evenly distributed cables that run along the axis of polarized growth^44,61.

In the pathogenic fungus Magnaporthe oryzae, which infects rice plants by breaking the cuticle with a pressurized protrusive structure termed the appressorium, septins (Sep3–Sep4–Sep5–Sep6) form a ring that organizes the assembly of a doughnut-shaped network of actin filaments (Figure 3E)⁴⁶. The septin ring corrals the I-BAR protein Rvs167 and the WASP homolog Las17 by restricting their diffusion and scaffolds the ERM family protein Tea1, which links actin filaments to the membrane⁴⁶. Septins also restrict the diffusion of membrane actin patches that contain coronin, an actin-remodeling protein that associates transiently with the septin-bound actin ring⁶². In coronin mutants, the actin ring is loosely bound to septins and fragmented, suggesting that septin-mediated confinement of coronin is critical for the assembly of a toroidal actin network⁶². Hence, the M. oryzae septin ring promotes actin assembly by confining factors with roles in actin polymerization and remodeling, whilst concomitantly scaffolding proteins that tether actin to the plasma membrane.

In gastrulating frog embryos, Sept7 promotes the enrichment of actin and phosphorylated myosin II in plasma membrane domains of anterior-to-posterior cell–cell contacts⁶³. This localization enables directional pulling forces that intercalate cells during convergent extension. At mammalian endothelial cell–cell adhesions, SEPT2 might similarly regulate actomyosin in subdomains of the plasma membrane, controlling the formation of the junction-associated lamellipodia that promote VE-cadherin contacts^64,65. Because actomyosin forces can promote membrane fusion by reducing energetic barriers and/or altering the conformations of fusogenic proteins^66,67, it is plausible that septins regulate vesicle fusion by modulating actomyosin at the plasma membrane–vesicle interface. Interestingly, SEPT7 forms a complex with myosin II and the SNARE protein SNAP-23 on the glucose transporter 4 (GLUT4) storage vesicles of renal podocytes⁶⁸. This interaction is proposed to inhibit myosin II and fusion of these vesicles with the plasma membrane prior to insulin stimulation, which downregulates SEPT7 expression⁶⁸.

In sum, septins can promote actin polymerization on membrane domains by scaffolding or corralling actin nucleation-promoting factors (such as formin, WASP and cortactin). Concomitantly, septins provide a template for the assembly of an actomyosin network by scaffolding myosin and actin-binding proteins that link actin filaments to the membrane. Thus, septins may directly amplify the density and linkages of actomyosin to the lipid bilayer, which has implications for the actin–spectrin skeleton, the rigidity and the protrusive activity of cell membranes. In light of new evidence indicating that membrane protrusions require thinner actin densities and looser actin–membrane linkages^69,70, septins could suppress protrusive activity at the neck regions of lamellipodia and filopodia, where they assemble due to preferential association with domains of saddle-shaped membrane curvature^34,71. Alternatively, septins may similarly modulate actin–membrane linkages on membrane domains enriched in phosphatidylinositol-4,5-bisphosphate (PIP₂), which promotes septin assembly into membrane-bound filaments⁷². As PIP₂ recruits and regulates actin-nucleating and -binding proteins⁷³, septins may function in coordination with PIP₂-binding factors to modulate the assembly and dynamics of membrane-bound actin.

Septin roles in actomyosin ring assembly and contraction during cytokinesis

Septins were discovered and named for their functions in yeast cytokinesis and the formation of the septum, a wall-like structure that develops concomitantly with the constriction of the cytokinetic actomyosin ring^5,74. In S. cerevisiae, septins scaffold the assembly of this ring^40,41,75,76. Septins are required for the initial recruitment of myosin II (Myo1), which is mediated by the Cdc11/Shs1-binding protein Bni5, and subsequently maintain the localization of the myosin essential light chain (Mlc1)^{40,41,75–78}. Septin-dependent assembly of Myo1 is critical for the capture and organization of actin filaments into the actomyosin ring^42,79. Septins (Cdc10) also recruit Hof1, which during cytokinesis is positioned between the actomyosin ring and the septin double rings that split from the septin hourglass structure of the mother–bud neck⁴⁵. Hof1 maintains the symmetry of Myo1 constriction and couples the actomyosin ring to the chitin synthase that generates the primary septum⁴⁵. Interestingly, septins (Cdc10) are not required for actomyosin ring constriction, which begins immediately after the septin hourglass splits into a double ring^77,80. It is hypothesized that clearance of the underlying septin filaments might be critical for the initiation of actomyosin ring closure⁸⁰.

In mammalian cytokinesis, septins have been proposed to function in actomyosin ring contractility. SEPT2 binds the coiled-coil domain of the myosin IIA heavy chain and this interaction is required for the localization of citron kinase (CIT-K) and Rho-activated kinase 2 (ROCK2) at the ingressing cleavage furrow⁵⁷. Through this scaffolding function, septins promote phosphorylation of the myosin regulatory light chain at residues Thr18 and Ser19, which activates myosin II contraction⁵⁷. Septins may function similarly in Drosophila melanogaster epithelia, the cell–cell junctions of which fail to disengage in Pnut mutants due to insufficient contractile force⁸¹. In cultured Drosophila cells, however, septins are thought to impact actomyosin ring closure indirectly⁵². Pnut and anillin form a network that is separate but adjacent to the actomyosin ring⁸² and facilitates closure of the ring by promoting the extrusion of membrane domains that disengage from actomyosin^82,83. In cellularizing Drosophila embryos, the Pnut and Sep1 septins are required for the circular organization of cortical actin, which becomes polygonal, discontinuous and less parallel to the membrane in pnut null mutants⁵². Although myosin II is not affected, these defects slow down membrane invagination and are attributed to the loss of the actin filament crosslinking and bending functions of Sep1–Sep2–Pnut complexes⁵².

In the single-cell Caenorhabditis elegans embryo, the septin UNC-59 and anillin polarize to the anterior cortex in a manner that depends on the cortical PAR proteins⁸⁴. Surprisingly, UNC-59 depletion increases the amount of actin filaments and the levels of myosin II in the cytokinetic contractile ring and rescues contractile ring closure in formin(ts) mutants⁸⁴. These findings suggest that septins inhibit actomyosin assembly. However, the observed effects might be due to septin-mediated sequestration of actomyosin throughout the anterior cortex, which might limit the availability of actomyosin for contractile ring assembly at the plane of cell division. Interestingly, during cleavage furrowing, which takes place unidirectionally, septins together with anillin and myosin II accumulate asymmetrically on the ingressing side of the ring⁸⁵. UNC-59 is required for the asymmetric concentration of both anillin and myosin II and the asymmetry of furrowing. Depletion of UNC-59 does not impact the completion of cytokinetic cleavage, but exacerbates defects caused by inactivation of myosin II⁸⁵. Thus, septins are posited to fine-tune the C. elegans actomyosin ring, promoting its mechanical robustness and asymmetric ingression.

On the whole, septin functions in cytokinesis vary from myosin II recruitment and activation to actomyosin ring assembly and closure. As septins are not universally essential for ring closure, they often have auxiliary roles and/or partial functional redundancies with other cytokinetic factors. This is likely a reflection of the diversity of mechanisms that different species and cell types employ to optimally execute cytokinesis. Septins are nevertheless a critical component of the cytokinetic actomyosin ring machinery, modifying actomyosin properties in both a direct and an indirect manner.

Actomyosin organization and functions in cell migration

Septins are essential for the migration of many cell types, including epithelia, fibroblasts, lymphocytes and neurons^53,86–88. Importantly, abnormalities in septin expression enhance the migratory and invasive properties of cancer cells^{9,38,56,60,89,90}. Depending on the cell type and mode of cell migration (that is, amoeboid or mesenchymal), septins regulate the organization and contractility of actomyosin on stress fibers and/or the plasma membrane (Figure 4). Septins physically reinforce the actomyosin network by crosslinking subsets of actin filaments, readying cells for the higher mechanical demands of migration that include elevated contractility, mechanotransduction and mechanical stress. Additionally, septins regulate the formation of membrane protrusions, which are critical for the directionality of cell migration, and scaffold Rho signaling effectors and kinases that control mechanotransduction. However, not all septins have pro-migratory roles, and septin paralogs and isoforms with inhibitory roles have been reported.

Figure 4. Septin localization and functions in the actin networks of migrating cells.

Septins associate with the actin networks of the plasma membrane, stress fibers and lamellipodia. (Top left) In migrating melanoma cells, SEPT9 and Cdc42EP5 are required for the organization of cortical actin. (Top right) In renal epithelia undergoing epithelial–to-mesenchymal transition (EMT), septins are enriched on contractile transverse arc stress fibers and at the interface of these fibers with the distal ends of the radial (dorsal) stress fibers that emanate from focal adhesions. Septins crosslink the actin filaments of these networks and are required for the maintenance of the transverse arc network and focal adhesion maturation. (Bottom right) In squamous cancer cells, SEPT1, SEPT4 and SEPT5 are distinctly enriched in lamellipodia, and in vitro SEPT6 associates with the branch points of Arp2/3-nucleated branched networks. (Bottom left) In various cells, septins associate with dorsal stress fibers that stretch in between the ventral nuclear and plasma membranes. Septin association with the stress fibers involves CDC42EP3 and the myosin II heavy chain. Septins may regulate mechanotransduction by scaffolding the phosphorylation of the myosin regulatory light chain by ROCK2 and controlling the nucleocytoplasmic shuttling of actin regulators such as Nck1, which might occur in a mechanosensitive manner. Interestingly, septins recruit Nck1 to stress fibers via SOCS7 and thereby prevent Nck1 from shuttling to the nucleus. SEPT9 isoform 2 (SEPT9_i2), which inhibits cell migration, suppresses the formation of a subset of stress fibers that localizes underneath the nucleus through an unknown mechanism.

In kidney cells undergoing the epithelial-to-mesenchymal transition, which reprograms gene expression and retools the cytoskeleton for a migratory phenotype, SEPT2–SEPT6–SEPT7–SEPT9 localizes on the actin transverse arc, a contractile network of curved stress fibers, and on the interface of this network with the radial (dorsal) stress fibers that grow from focal adhesions⁵³. In SEPT2-depleted cells, the transverse arc network is severely disrupted or absent, and nascent focal adhesions have shorter lifetimes, failing to stabilize and mature⁵³. Consistent with a function for SEPT9 isoform 1 (SEPT9_i1) in the crosslinking and bundling of actin filaments, which was demonstrated in vitro, this phenotype is rescued by the actin-bundling protein α-actinin-1⁵³. In vitro assays reveal that SEPT9_i1 might also protect actin filaments against depolymerization by cofilin or the contractile forces of myosin II³⁶. Interestingly, SEPT9_i1 expression is upregulated during partial epithelial-to-mesenchymal transition⁵³. Therefore, SEPT9_i1 may reinforce the transverse arc network and its linkages to radial (dorsal) stress fibers for the enhanced mechanotransduction and focal adhesion turnover of migrating cells. SEPT9 is also involved in the activation of focal adhesion kinase–Src–paxillin and RhoA–ROCK signaling, but it is unclear whether actin-bound SEPT9 provides a scaffold for the phosphorylation of these proteins⁸⁹.

The pro-migratory functions of SEPT9, which have been reported for isoform 1, can vary from other isoforms. SEPT9_i2 inhibits cell migration and the formation of a subset of actin stress fibers that localize on the ventral side of the nucleus³⁸. New work shows that these stress fibers assemble through re-organization of the cortical actin by myosin II, which might be regulated by SEPT9 isoforms⁹¹. SEPT9_i3 has also been found to interact with and inhibit a septin-associated Rho guanine nucleotide exchange factor (SA-RhoGEF)⁵⁹. In endothelial cells, this inhibition may underlie the SEPT9-dependent reduction of stress fibers, which occurs in response to low-stiffness extracellular matrix⁹². Alternatively, SEPT9 isoforms might promote disassembly of actin by recruiting proteins of the MICAL (molecules interacting with CasL) family⁹³, which disassemble actin filaments by oxidizing actin residues.

In cancer-associated fibroblasts, septins promote the establishment of an extensive actin stress fiber network, which facilitates the remodeling of the extracellular matrix into a pro-tumorigenic matrix through enhanced contractility and mechanotransduction⁵¹. The Cdc42 effector protein Cdc42EP3 (BORG2), which is upregulated in these fibroblasts, interacts and colocalizes with SEPT2 and SEPT7 in a filamentous crossbridge-like network between the actin filaments of stress fibers⁵¹. Cdc42EP3 possesses an actin-binding domain and synergizes with septins in generating and maintaining the stress fibers of cancer-associated fibroblasts⁵¹. In addition to Cdc42EP3, septin association with stress fibers involves SEPT2 binding to myosin IIA (see above), the activation of which in turn depends on septin-scaffolded kinases (for example, ROCK2)⁵⁷. Thus, it is plausible that septins are involved in the contractile and mechanotransducing properties of actin stress fibers. As the expression of septins such as SEPT6 and SEPT9 has been reported to change in response to extracellular matrix stiffness^92,94, septins might be part of a feedback loop that promotes mechanotransduction by increasing septin presence on stress fibers. Interestingly, stress-fiber-associated septins maintain the cytoplasmic localization of Nck1, a signal-transducing adaptor that shuttles between the nucleus and cytoplasm⁴⁹. In principle, septins could affect gene transcription by releasing nucleocytoplasmic proteins from stress fibers in a mechanoresponsive manner. Alternatively, septins might promote mechanotransduction from the extracellular matrix into the nucleus by stabilizing the actomyosin network. Consistent with this possibility, Cdc42EP3 and SEPT2 are required for the nuclear translocation of the oncogenic transcriptional activator YAP1 (Yes-associated protein 1) in response to stiffer matrices, which upregulates Cdc42EP3 and SEPT2 levels on actin filaments⁵¹. Thus, septins could enhance the mechanical tension exerted by actin filaments on focal adhesions, activating the focal adhesion kinase signaling cascade that triggers YAP1 translocation to the nucleus; the latter might be further amplified by opening of the nuclear pores due to forces transduced by actomyosin to the nuclear envelope⁹⁵.

In mouse melanoma cells with an amoeboid mode of migration, Sept9 is the only septin paralog whose upregulation correlates with the invasive metastatic phenotype⁶⁰. The rounded amoebalike morphology of these cells is sustained by contractile actomyosin filaments that associate with the plasma membrane⁶⁰. The cortical actomyosin network depends on Sept9 and Cdc42EP5 (BORG3), which localizes preferentially to the cell cortex with actin and phosphorylated (pSer19) myosin regulatory light chain⁶⁰. Cdc42EP5 does not bind actin on its own, but it enhances the actin-binding and -crosslinking properties of Sept9, promoting Sept9 association with cortical actin⁶⁰. Surprisingly, Sept7 does not impact melanoma cell migration⁶⁰. In amoeboid T cells, Sept7 regulates migration by modulating the rigidity of the plasma membrane without affecting cortical actomyosin⁹⁶. Moreover, Sept7 inhibits the migration of glioma cells through a mechanism that involves suppression of cofilin activity⁹⁷. This paralog-specific difference between Sept9 and Sept7 is also seen with other septins. In squamous cell carcinoma cells, SEPT1, SEPT4 and SEPT5 localize to lamellipodia, whereas SEPT9, SEPT11 and SEPT14 associate mainly with microtubules³². Notably, SEPT1 is required for cell spreading and accumulates in the lamellipodia of migrating cells after the establishment of front–rear polarity³². Taken together, these findings indicate that septin paralogs are functionally specialized for different actomyosin networks and mechanisms of cell migration.

Septins and actomyosin in bacterial infection

Septins have important functions in microbial pathogenesis and host immunity against bacteria, which hijack the actin cytoskeleton for cell entry and proliferation⁹⁸. Septins are an integral component of actomyosin in all stages of bacterial infection, from cell-surface attachment and internalization to intracellular motility and entrapment for autophagic destruction. Internalization of Listeria monocytogenes, which docks on the cell surface by interacting with the hepatocyte growth factor receptor Met, is characterized by the recruitment of septins and the formation of a septin–actin ring⁹⁹. In SEPT2-depleted cells, Met-triggered activation of phosphatidylinositol 3-kinase and Listeria internalization are markedly diminished⁹⁹. In contrast, SEPT11 knockdown enhances internalization without impacting Met signaling¹⁰⁰. The differential roles of SEPT2 and SEPT11 suggest their involvement in distinct pathways that respectively promote and inhibit the actin polymerization that enables bacterial engulfment through phagocytosis. In contrast to the effects on L. monocytogenes, knockdown of SEPT2, SEPT7 or SEPT9 decreases the internalization of Salmonella enterica serovar Typhimurium¹⁰¹. In SEPT7-depleted cells, ROCK2 fails to accumulate at actin-rich sites of bacterial contact with membrane ruffles¹⁰¹. Myosin II recruitment and phosphorylation remain unaltered, but phosphorylation of the actin-nucleating and -bundling formin FHOD1 on Thr1141 is abrogated¹⁰¹. Thus, septins promote the activation of the formin FHOD1 by acting as a scaffold to facilitate its phosphorylation by ROCK2 (Figure 3E). Septins may similarly function in the reorganization of actin during the invasion of the diarrheal pathogens Shigella flexneri and enteropathic Escherichia coli (EPEC)¹⁰². Interestingly, EPEC phosphorylates a conserved serine in the amino terminus of SEPT9 isoforms 1–3, which might be critical for the formation of the actin pedestals that strengthen EPEC adhesion to the cell membrane¹⁰².

Following internalization, septins associate with the actin surrounding bacteria-containing vacuoles and with actin comet tails, which promote intracellular movement of the bacteria. On the membrane vacuoles that contain Chlamydia trachomatis, septins (SEPT2, SEPT9, SEPT11) colocalize with filamentous actin¹⁰³. SEPT2 and SEPT9 depletion reduces actin assembly on C. trachomatis inclusions and suppresses extrusion from infected cells without impacting myosin IIB¹⁰³. In contrast to these findings, which indicate that septins contribute to the progression and spread of bacterial infection, septins have been shown to restrict the movement and spread of S. flexneri and vaccinia virus^104,105. Septins (SEPT2, SEPT9, SEPT11) form rings around the actin tails of intracellular S. flexneri and L. monocytogenes bacteria¹⁰⁴. Although septin-ring assembly depends on actin, neither SEPT2 nor SEPT9 depletion affects the velocity of bacterial movement¹⁰⁴. Strikingly, however, septins (SEPT2, SEPT9, SEPT11) form a cage-like structure around tail-less bacteria¹⁰⁴. These cages consist of circumferential bundles of septin filaments that are spaced between actin filaments and partially overlap with phosphorylated myosin light chain¹⁰⁴. Actin polymerization, myosin II heavy chain and actomyosin contractility are all required for the assembly of septin cages, which suppress the formation of actin tails¹⁰⁴. Septins restrict the intercellular spread of Shigella and promote its degradation through autophagy and fusion with lysosomes^104,106.

Overall, bacterial pathogens exploit septins for reorganizing the actin cytoskeleton at plasma membrane sites of attachment and internalization, and on the endocytic membranes of their vacuolar inclusions. Mechanistically, bacteria hijack septin functions in scaffolding actomyosin and signaling kinases that regulate actin assembly. However, host defense mechanisms also utilize septin–actomyosin interactions for restricting the spread and inducing the autophagic destruction of bacteria that escape from endocytic vacuoles.

Septins have properties of bona fide microtubule-associated proteins

Septins associate with microtubules in a diversity of cell types (for example, epithelial cells, neurons and platelets), colocalizing with subsets of microtubules, such as perinuclear¹⁰⁷, Golginucleated^108,109, subcortical^110,111 and axonemal¹¹² microtubules. This specificity arises from the α- and β-tubulin isotypes, microtubule-associated proteins (MAPs) and post-translational modifications of microtubules that favor septin binding^{33,109,113–115}. Conversely, it is also determined by the microtubule-binding properties of different septin paralogs and their membrane or cytoplasmic regions of accumulation^12,33.

Mechanistic insight into how septins associate with microtubules has emerged from studies of mammalian SEPT9_i1, which contains an amino-terminal repeat motif (K/R-R/x-x-D/E) that interacts preferentially with the carboxy-terminal tails of βII-tubulin (Figure 5A)³³. In addition, septin–microtubule binding involves the carboxy-terminal tyrosine of α-tubulin and polyglutamylation of the carboxy-terminal tail^109,115. SEPT9_i1 and SEPT2–SEPT6–SEPT7 complexes associate preferentially with microtubule lattices bound to the GTP analog GMPCPP^54,55. This is consistent with increased septin–microtubule binding in cells treated with taxol^107,114,116, which stabilizes microtubules by inducing a GMPCPP-like conformational state¹¹⁷. Because GMPCPP and taxol alter the longitudinal intra-protofilament interactions of α- or β-tubulin monomers without impacting lateral inter-protofilament contacts¹¹⁷, septin complexes may bind along the length of individual protofilaments rather than wrapping around protofilaments. Interestingly, SEPT2–SEPT6–SEPT7 has a higher affinity for GDP-bound microtubules than SEPT9_i1, which uniquely binds and recruits unpolymerized tubulin to microtubules^54,55. Although it is unknown whether the SEPT2–SEPT6–SEPT7 complex and other septins associate with microtubules via K/R-R/x-x-D/E repeat motifs, septin paralogs and complexes appear to have different microtubule-binding properties³³. Additionally, septin–microtubule interactions are influenced by MAPs (such as MAP4), which interact and compete with septins for microtubule binding^109,112,113.

Figure 5. Functions of microtubule-associated septins.

(A) Crystal structures of SEPT9 (PDB: 4YQF) and αβ-tubulin (PDB: 1TUB) depict their intrinsically disordered amino-terminal extension and carboxy-terminal tails (hand-drawn lines), respectively, which mediate septin–microtubule binding. Purple boxes on the amino-terminal extension of SEPT9 represent the microtubule-binding repeat motifs K/R-R/x-x-D/E. The carboxy-terminal tyrosine and polyglutamylated side chains of αβ-tubulin are denoted with the letters Y and E, respectively. (B) SEPT7 and SEPT1 promote the nucleation of microtubules from the centrosome and Golgi membranes, respectively. SEPT7 is required for the centrosomal localization of p150^Glued, and SEPT1 forms a complex with GM130, CEP170 and the γ-tubulin ring complex (γ-TURC). (C) Septin oligomers (SEPT9, SEPT2–SEPT6–SEPT7) promote persistent microtubule growth and crosslink microtubules into bundles; SEPT9 recruits αβ-tubulin dimers to the microtubule lattice. At higher concentrations, microtubule-associated septins pause and stunt plus-end growth. Septins also promote the capture and zippering of microtubule plus ends with the lattices of septin-coated microtubule bundles. (D) Septins function in microtubule acetylation and polyglutamylation. SEPT7 interacts with histone deacetylase (HDAC6) and is posited to scaffold the deacetylation of α-tubulin in the cytosol. Microtubule-associated septins have been proposed to scaffold tubulin tyrosine ligase-like (TTLL) and cytosolic carboxypeptidase 1 (CCP1) enzymes that respectively elongate and trim the polyglutamylated side chains of the carboxy-terminal tails of polymerized tubulin. (E) Microtubule-associated septins differentially regulate the motility of kinesin motors and their cargo. Microtubule-associated SEPT9 inhibits kinesin-1/KIF5 and dynein and enhances the motility of kinesin-3/KIF1A.

Septins function in the nucleation, organization, dynamics and post-translational modifications of microtubules (Figure 5B–E). SEPT7 and SEPT1 are uniquely involved in the nucleation of microtubules from the centrosome and Golgi complex, respectively (Figure 5B)^108,118. SEPT7 is required for the centrosomal localization of the dynactin subunit p150^Glued, and SEPT7 depletion results in the nucleation of a more dense and disorganized microtubule network¹¹⁸. On Golgi membranes, SEPT1 interacts with the centrosomal protein CEP170, promoting the nucleation of acentrosomal microtubules¹⁰⁸. In vitro reconstitution assays have shown that septins can directly crosslink microtubules into elongated bundles and promote the growth of microtubules by suppressing catastrophe^33,54,55. Interestingly, these effects depend on septin concentration and state of assembly (Figure 5C)^54,55. Loss of SEPT9 dimerization diminishes its ability to recruit unpolymerized tubulin and suppress microtubule catastrophe⁵⁵. Moreover, a biphasic effect is observed with SEPT9 and SEPT2–SEPT6–SEPT7 complexes, with oligomers promoting microtubule polymerization and higher-order polymers pausing or stunting growth^54,55. It is possible that oligomeric septins promote growth by physically stabilizing the microtubule lattice and suppressing catastrophe, and higher-order septin complexes amplify this effect to such a level that microtubule plus-end growth is completely constrained. Although some evidence indicates that septins (SEPT5 and SEPT2–SEPT6–SEPT7) can interact directly with the microtubule plus-end-binding and -tracking protein EB1, septins (SEPT2, SEPT9 and SEPT2–SEPT6–SEPT7) have not been observed to track on growing plus ends in vitro or in living cells^54,55,107. Surprisingly, in vitro binding of SEPT2–SEPT6–SEPT7 to the microtubule lattice induces EB1 dissociation from plus ends^54,119. Therefore, SEPT2–SEPT6–SEPT7 complexes might induce long-range conformational changes that affect microtubule plus-end dynamics via propagation along the microtubule lattice¹²⁰.

In addition to modulating microtubule dynamics, septins also impact the acetylation and polyglutamylation of microtubules. SEPT7 functions as a cytoplasmic scaffold of histone deacetylase 6 (HDAC6), which deacetylates unpolymerized αβ-tubulin dimers, regulating the levels of microtubule acetylation¹²¹. In cancer cells that are resistant to the microtubule-stabilizing drug taxol, SEPT9 and SEPT2 interact with the polyglutamylating enzymes of the tubulin tyrosine ligase-like (TTLL) family TTLL11 and TTLL1, respectively, and SEPT2 associates with cytosolic carboxypeptidase 1 (CCP1), which trims long glutamate chains¹¹⁵. Thus, microtubule-associated septins are posited to scaffold enzymes that regulate the polyglutamylation of the carboxy-terminal tails of tubulin.

Collectively, the evidence thus far indicates that septins associate with microtubules selectively, having the capacity to distinguish between microtubules composed of different tubulin isotypes and bearing distinct post-translational modifications. Possessing properties of bona fide MAPs, septins promote microtubule nucleation, growth, elongation and bundling. Moreover, septins complete for binding with other MAPs and also scaffold enzymes that maintain distinct microtubule post-translational modifications.

Roles of microtubule-associated septins in cellular morphogenesis

Septins are essential for the morphogenesis of a diversity of cell types including epithelial cells and neurons, which depend largely on microtubules for their polarized shapes and compartments (i.e., apical and basolateral regions of the plasma membrane, and axons and dendrites)^122,123. Septins play key roles in the spatial regulation of microtubule organization and microtubule-dependent membrane traffic, which underlies the development of cell polarity. For example, septins guide microtubule–microtubule interactions and microtubule motor–cargo movement during the development of apical and dendritic compartments, respectively^107,109,124. While this regulation takes place in the cytoplasm, septins also function at the interface between microtubules and the plasma membrane or nuclear membrane, impacting the morphogenesis of platelets and spermatozoa^111,125.

In polarizing columnar epithelia, where the microtubule network reorganizes along the axis of apicobasal polarity, SEPT2-containing complexes associate with perinuclear microtubule bundles and polyglutamylated microtubules that underlie sites of vesicle export from the trans-Golgi network (TGN)^107,109. On perinuclear microtubules, septins capture and promote the persistent growth of incoming microtubule plus ends, steering microtubule growth toward the cell periphery and apex¹⁰⁷. Through this navigating mechanism, septins contribute to the establishment of an apical microtubule network¹⁰⁷. At TGN exit sites, microtubule-associated septins promote vesicle egress by inhibiting the binding of MAP4, which impedes kinesin motility¹⁰⁹. SEPT2-depleted epithelia fail to form a columnar morphology because post-Golgi traffic stalls, abrogating membrane growth and polarization into apical and basolateral domains¹⁰⁹. Competition between septins (SEPT2–SEPT6–SEPT7) and MAP4 for microtubule binding might also play a role in the elongation of the primary cilia of epithelia; SEPT2–SEPT7–SEPT9 and MAP4 associate with axonemal microtubules, promoting and suppressing the elongation of cilia, respectively¹¹². Thus, formation of the apical, basolateral and ciliary compartments of epithelia are critically dependent on microtubule-associated septins.

Similar to their function in epithelia, septins play important roles in the organization and function of microtubules during neuronal morphogenesis. In developing (E15) mouse embryos, Sept7 deletion diminishes the length and complexity of dendritic arbors as well as the branching of the axonal terminals of pyramidal neurons¹²¹. This phenotype correlates with enhanced microtubule stability due to loss of tubulin deacetylation by HDAC6 and Sept7¹²¹. In sensory neurons, which have no dendrites, SEPT7 localizes at the base of axonal filopodia, promoting microtubule entry and thereby the maturation of nascent filopodia into collateral axon branches³⁷. SEPT7 may similarly function in the capture and guidance of microtubule plus ends at the base of dendritic spines in hippocampal neurons and at cortical sites of neurite reformation in neural crest cells following mitosis^126–128. Consistent with this possibility, SEPT2–SEPT6–SEPT7 filaments are more potent than actin in capturing microtubule plus ends in vitro¹²⁹, and membrane septins (SEPT2–SEPT6–SEPT7–SEPT9) capture microtubules in cells infected with a Clostridium difficile toxin, which depolymerizes actin and induces microtubule-driven membrane protrusions¹¹⁹. Notably, septins are also required for the attachment of mitotic microtubules to the plasma membrane of the mother–bud neck cortex of S. cerevisiae^130,131. Therefore, septins might have an evolutionarily conserved function in microtubule capture.

Neurite differentiation into an axon and dendrites involves the microtubule-associated SEPT9_i1 isoform, which appears to localize and function independently of SEPT7¹²⁴. It is unknown whether SEPT9 assembles into homomeric or heteromeric complexes free of SEPT7, but SEPT14 and SEPT4 are also reported to function independently of SEPT7 in neuronal migration during corticogenesis⁸⁶. Asymmetric neurite growth is enhanced by SEPT9_i1 and requires its amino-terminal microtubule-binding and -bundling domain, mutations in which cause a rare shoulder-arm neuropathy termed hereditary neuralgic amyotrophy^33,132. Studies in rat hippocampal neurons show that, after the formation of an axon, SEPT9_i1 becomes enriched in developing dendrites¹²⁴. On dendritic microtubules, SEPT9_i1 reinforces the polarity of membrane traffic by hindering the movement of vesicles with axonal proteins and promoting the anterograde transport of vesicles with dendritic cargo¹²⁴. Thus, SEPT9_i1 promotes the differentiation of dendritic membranes, which become enriched with postsynaptic proteins.

Microtubule-associated septins are also implicated in the morphogenesis of spermatozoa and platelets. SEPT12 is a testes-specific, microtubule-binding septin that when mutated causes abnormal sperm morphology and sterility¹³³. In spermatozoa from Sept12^+/+/+/− mouse chimeras, perinuclear microtubules are disorganized, the nucleus-containing head is abnormally shaped and flagellar tails are bent or abnormally short¹²⁵. These defects might arise from a possible role for SEPT12 in the assembly and function of the manchette, a transient skirt-like microtubule-based structure that morphs the nucleus into an oval and supplies the axoneme with proteins for intraflagellar transport¹³⁴. Interestingly, SEPT12 interacts with the SUN-domain-containing nuclear membrane protein SPAG4 and may therefore link the spermatid nucleus to microtubules¹³⁵. SEPT14 also interacts with α-actinin-4 at the nuclear envelope and the manchette and might link both microtubules and actin to the nuclear membrane¹³⁶. Septins have been similarly implicated in linking microtubules to the plasma membrane. In platelets, septins localize to a cortical microtubule ring¹¹¹, and in neurulating frog embryos SEPT7 organizes microtubules into cortical bundles that traverse the long cell axis, promoting cell elongation during wound healing¹³⁷.

In general, septins can regulate microtubule organization directly and indirectly as microtubule-associated and membrane-associated proteins, respectively. Guidance and capture of dynamic microtubule plus ends by microtubule- and membrane-associated septins might constitute a chief mechanism for orienting and organizing the microtubule network along the emerging axes of cell asymmetry.

Septins regulate the traffic of microtubule motors and their cargo

Cellular morphogenesis and physiology require not only a properly organized cytoskeleton, but also spatiotemporally accurate transport and positioning of membrane vesicles and organelles. Recent developments have shown that microtubule post-translational modifications and MAPs can selectively regulate the motility of distinct microtubule motors and their cargos^129,138,139. It is now posited that these microtubule properties constitute a ‘tubulin code’, which functions as a traffic code that guides intracellular transport by determining which motor–cargo complexes move on which microtubule tracks and in what direction. In this new era, microtubule-associated septins emerged among the first MAPs that differentially regulate the motility of select kinesin motors and thereby determine the intracellular destination of their cargo. Interestingly, septins also modulate microtubule motor interactions with membrane cargo, providing a two-pronged regulation at the interface of motors with microtubules and membrane cargo^124,140,141.

Using single-molecule in vitro motility assays, our group sought to test whether the microtubule-associated SEPT9_i1 isoform can directly impact the motility of kinesin and dynein motors¹²⁴. We found that SEPT9_i1 impedes the motility of kinesin-1/KIF5 and the dynein–dynactin–bicaudal D motor complex, promotes the motility of kinesin-3/KIF1A and does not impact kinesin-2/KIF17¹²⁴. Time-lapse imaging assays in living hippocampal neurons showed that SEPT9_i1 overexpression and depletion impact kinesin motility in a manner that is consistent with the in vitro effects¹²⁴. On dendritic microtubules, which are enriched with SEPT9_i1, this differential regulation impedes axonal cargo of kinesin-1/KIF5 from moving anterogradely and promotes kinesin-3/KIF1A motor–cargo movement into distal dendrites¹²⁴. SEPT9_i1 enhances the motility of kinesin-3/KIF1A in a manner that involves the lysine-rich loop 12 of KIF1A, which is more positively charged than other kinesins — a feature that enables KIF1A to undergo long processive runs without dissociating from microtubules¹²⁴. SEPT9_i1 might provide additional attachment sites for KIF1A through its amino-terminal acidic domains and/or induce conformational changes at the microtubule–kinesin interface. The latter may enhance microtubule docking of the lagging unbound motor domain of the KIF1A dimer, a rate limiting-step in KIF1A motility¹⁴², and/or weaken the intramolecular interactions between the microtubule-bound motor domain and the neck linker, which can also accelerate docking of the lagging motor domain¹⁴³. Of note, the neck linker of KIF1A is uniquely shorter than other kinesin motors^143,144.

Differential regulation of kinesin motor motility by SEPT9 poses the question of whether this is a unique property of SEPT9 or whether other septin paralogs and complexes function similarly. SEPT9 is the sole septin paralog with a long amino-terminal microtubule-binding extension and therefore other septins may not have the same regulatory effects. Furthermore, it is unclear whether SEPT9_i1 retains its differential regulation in heteromeric complexes, some of which might be permissive and other restrictive. Ongoing studies aim at testing how different septins, including SEPT9 isoforms with shorter amino termini, and heteromeric septin complexes impact kinesin motility. While it’s still early days, the possibility of selective regulation of kinesin motor–cargo transport by different septin paralogs and complexes will be an exciting development. Taken together with the differential localization of septins to distinct subpopulations of microtubules, septins could provide a ‘highway code’-like function for orienting membrane traffic on the microtubule network. This septin code may be part of the broader tubulin and MAP codes, which are currently hypothesized to guide intracellular transport.

Interestingly, new work shows that septins can regulate the attachment and activity of kinesin motors not only on microtubules, but also on membrane cargo¹⁴¹. Coupling of SEPT9_i1 to membrane organelles results in the recruitment of dynein–dynactin motors, which trigger retrograde transport to the perinuclear cytoplasm¹⁴¹. SEPT9_i1 interacts with the dynein intermediate chain (DIC) in its multimeric GDP-bound state, which indicates a scaffolding role in the recruitment of dynein–dynactin motors¹⁴¹. Indeed, SEPT9 promotes the retrograde transport of lysosomes in response to arsenite-induced oxidative stress, which increases the lysosomal membrane levels of SEPT9¹⁴¹. SEPT9_i1 expression is similarly upregulated in hepatocytes infected with the hepatitis C virus, with lipid droplets accumulating at perinuclear regions in a SEPT9_i1-dependent manner¹⁴⁵. Phosphatidylinositol 5-phosphate promotes SEPT9_i1 binding to lipid droplets and movement toward the cell center¹⁴⁵. Although SEPT9_i1 appears to associate with DIC independently of the SEPT2–SEPT6–SEPT7 complex, SEPT9_i1 may associate with this complex, which in turn can provide an additional link to dynactin through the interaction of SEPT7 with p150^Glued118.

In addition to binding dynein, SEPT9_i1 interacts with the cargo-binding tail of kinesin-2/KIF17, competing with the cargo adaptor/scaffold mLin-10/Mint1, which links KIF17 to the NMDA receptor subunit 2B (NR2B)¹⁴⁰. Taken together, these findings raise the possibility that SEPT9_i1 functions in the regulation of bidirectional transport by interacting concomitantly with motors of opposing directionality. This may enable switching between retrograde and anterograde movement in response to signaling cues, and hitchhiking of motors to intracellular locales that they cannot reach on their own (for example, transport of the dynein minus-end-directed motor by kinesins to microtubule plus ends). Hence, membrane cargo-bound septins can serve as cargo adaptors that interface with the protein sorting and vesicle transport machinery. Consistent with this function, septins (SEPT2, SEPT7, SEPT9) were recently found in a complex with the sorting nexin SNX21 and the dynein adaptor huntingtin¹⁴⁶. Furthermore, SEPT9_i1 interacts directly with tumor susceptibility gene 101 protein, a subunit of the endosomal sorting complex required for transport (ESCRT)¹⁴⁷. Therefore, it is possible that more septin paralogs and complexes function as adaptors of microtubule motors in a cargo-specific manner.

Taken together with their microtubule-associated properties, septins have emerged with hitherto unknown roles in regulating kinesin and dynein interactions with microtubules and membrane cargo. As a multitude of septin paralogs and complexes as well as kinesin motors exist, current findings point to a broader regulatory code by which septins select the type of motors that associate with microtubules and cargo.

Septin functions on mitotic and meiotic microtubules

Historically, the first evidence of septin association with microtubule-dependent functions and transport came from dividing cells^{131,148–150}. Expanding on the early findings, a growing number of studies have shown that microtubule-associated septins have pivotal roles in spindle assembly, chromosome alignment and segregation, and cytokinetic abscission.

Assembly of bipolar mitotic and meiotic spindles requires properly duplicated centrosomes, which nucleate microtubules and are located at the opposing poles of the spindle. Recent work shows that SEPT7 localizes to the centrosome in the absence of SEPT2 and SEPT6 and interacts with p150^Glued, which localizes to the subdistal appendages of centrioles¹¹⁸. Indicative of a function in centriole duplication, SEPT7 knockdown causes multipolar spindle formation and consequently impairs chromosome segregation¹¹⁸. SEPT7 also interacts with centromere-associated protein E (CENP-E)/KIF10, a kinesin-7 motor that mediates microtubule attachment to misaligned chromosomes, promoting chromosome alignment and bi-orientation¹⁵¹. At the metaphase plate, a network of SEPT2-containing fibers that juxtapose and partially colocalize with kinetochore microtubules maintain the kinetochore localization of CENP-E/KIF10¹⁵⁰. In SEPT2- and SEPT7-depleted cells, CENP-E/KIF10 is mislocalized to spindle poles and chromosomes fail to align properly at the metaphase plate^150,151. Notably, the spindle checkpoint is compromised, leading to multinucleation and polyploidy, hallmarks of genomic instability in cancer¹⁵⁰. Thus, the microtubule-associated roles of septins in mitosis are important for genomic stability, which might be undermined by abnormal septin expression.

Following cleavage furrow ingression in late cytokinesis, septins localize to microtubules of the intercellular bridge^147,152. It is unclear how septins function on these microtubules, but recent evidence suggests roles in microtubule severing and inheritance of cell-fate regulators¹⁵². Sept7^−/− fibroblasts fail to undergo abscission, which requires microtubule severing prior to membrane scission by the ESCRT-III complex¹⁵². This phenotype correlates with the hyperacetylation of microtubules and can be rescued by expression of stathmin, a tubulin-sequestering factor that promotes microtubule depolymerization¹⁵². Therefore, SEPT7 could indirectly impact microtubule severing through tubulin deacetylation. On intercellular bridge microtubules, SEPT7 was recently reported to colocalize and interact with the kinesin-6 motor KIF20A (also known as MKLP2), which mediates the transport of the chromosomal passenger complex to the central spindle for initiating cleavage furrow ingression¹⁵³. In neural progenitor cells, SEPT7 is required for localization of KIF20A to the intercellular bridge¹⁵³. Intriguingly, abscission is not affected following loss of SEPT7 in these progenitor cells, but they exit the cell cycle and differentiate into neurons precociously¹⁵³. Given that KIF20A interacts with the regulator of G-protein signaling RGS3, knockout of which phenocopies the effects of SEPT7 deletion, SEPT7 might control the localization and inheritance of cell-fate regulators during abscission^153,154. Stem-cell-fate decisions are influenced by the asymmetric inheritance of the midbody, which is released or inherited by a daughter cell through a double or single scission event, respectively¹⁵⁵. Therefore, microtubule-associated septins may function in stem-cell differentiation by regulating midbody inheritance.

On the cortical membrane of dividing cells, septins can regulate the position of the mitotic spindle through interactions with microtubule plus ends. In budding yeast, the cortical septin ring is required for the capture and pulling of astral microtubules, the plus ends of which associate with septins through Kar9, the yeast homolog of the adenomatous polyposis coli (APC) protein¹³¹. Temperature-sensitive septin mutant strains (cdc12–1) exhibit defects in nuclear and spindle positioning, phenocopying cells that lack dynein or Kar9¹³¹. This finding suggests that septins could impact spindle positioning and the axis of cell division in stem cells. Interestingly, septin knockdown alters the division axis in Drosophila neuroblasts, but it is unknown whether this is due to septin functions in spindle positioning¹⁵⁶.

In meiotic mouse oocytes, SEPT7 localizes to spindle microtubules and plays a role in spindle assembly, chromosome alignment and extrusion of the polar body¹⁵⁷. In S. cerevisiae cells that undergo meiosis and sporulation in response to nutritional deprivation, septins relocalize from the mother–bud cortex to microtubules¹⁵⁸. Cdc10, Cdc11 and the sporulation-specific septin Spr28 associate with the spindle microtubules of sporulating cells during meiosis I and II, promoting the formation of the prospore membrane and the cell walls of the emerging spores¹⁵⁸. Thus, septins appear to have conserved their association with meiotic microtubules throughout evolution.

Future questions and challenges

Since the early 2000s, a plethora of studies has cemented septins as an integral component of the cytoskeleton. However, many of the fundamentals of the cytoskeletal biology of septins elude our understanding. The organization and assembly state of septins on actin filaments is unclear. Knowledge of the septin domains, paralogs and complexes that bind actin is lacking. And whether septins regulate the nucleation and polymerization of linear or branched actin filaments has not been explored. Probing these questions mechanistically using structure–function approaches and in vitro biochemical assays will bring much-needed clarity and shed light on the differential roles of septins in actomyosin assembly and contractility. Phenotypic differences resulting from the depletion of septin paralogs and isoforms that are presumed to function in the same complex (for example, SEPT7 and SEPT9) are confounding and suggest that septins may function outside their canonical heteromeric complexes. Systematic work with septin paralogs, isoforms and complexes beyond the prototypical hexamers (SEPT2–SEPT6–SEPT7) and octamers (SEPT2–SEPT6–SEPT7–SEPT9) is necessary to fully understand the functional specificities of various septins. As septin expression is altered in a variety of disease states, likely leading to disruptive effects on stoichiometric assembly, it is critical that we gain a comprehensive understanding of the actin-binding properties and localizations of individual septin paralogs and isoforms. In the long term, this will enable us to determine which septin complexes associate with which actin networks. Taken together with studies that investigate how actin–septin binding is regulated and how in turn septins regulate actin-binding protein localization, future findings will provide new insights into how different actin networks are coordinated and spatially regulated, a key challenge in the actin field.

From a disease standpoint, we may also get closer to understanding how abnormal septin expression contributes to pathogenesis by selectively enhancing the actin networks of structures associated with malignant cell behaviors, such as invadopodia and lamellipodia in cancers. In the context of pathogenesis, it is also critical to study the role of septins in mechanotransduction, investigating whether septins associate with actin in a mechanosensitive manner and whether septins impact force transduction along actomyosin and into the nucleus.

On the microtubule front, ultrastructural studies are necessary to elucidate the precise organization and interactions of septins with microtubules at an atomic-scale resolution. Progress in this area will lead to a mechanistic understanding of how septins regulate microtubule dynamics and the motility of kinesin and dynein motors. In vitro reconstitution assays have enabled the elucidation of the direct functions of microtubule-associated septins. Using this approach with less-studied septin paralogs and complexes (such as SEPT1, SEPT14 and SEPT5–SEPT7–SEPT11) will enhance our knowledge of septin-specific functions and test the hypothesis of a septin code that specifies motor–cargo movement on microtubules. This hypothesis may also apply to the actin network, given that new evidence shows that septins interact with processive myosins^159,160 and thereby might regulate the transport of myosin motors and their cargo.

Recent studies have led to a number of new questions regarding how septins function at the interface of cell membranes with actin, microtubules and motors. Outstanding questions include: whether or how membrane septins nucleate, capture and organize actin filaments and microtubules; whether septins regulate and coordinate motors of opposing directionality on membrane organelles and vesicles; and how septin localizations and functions shift between membranes, microtubules and actin filaments. A relocalization between actin and microtubule polymers might be key for adjusting septin functions to meet the needs of cellular processes that might depend more heavily on actin than microtubules, and vice versa. Thus, it is critical to determine whether septins shift between actin and microtubules as a result of signaling cascades, mutually exclusive interactions and/or the relative abundance of actin and microtubule polymers. Furthermore, it is worth exploring whether septins provide functional crosstalk between actin and microtubules, either by shuttling between the two networks or by physically coordinating their dynamics and localization. Taken together with studies of septin mutations and abnormalities, which are linked to the pathogenesis of a variety of diseases, future investigations into these questions will provide novel insights into our knowledge of the septin cytoskeleton in health and disease.