Conflict of interest. Peter Hall is Editor-in-Chief of the Journal of Pathology but recused himself (as is Journal policy) from all parts of the handling, peer review and acceptance procedures of this review, which was commissioned by the Guest Editors of the 2012 Annual Review issue.
With an origin in the Nobel Prize-winning work of Lee Hartwell 1 and the biology of cell cycle control and cytokinesis, the septins are now known to be a large family of evolutionarily-conserved genes encoding proteins with roles well beyond cytokinesis 2, 3. The origin of the term ‘septin’ lies in the laboratory jargon of the Pringle laboratory in the late 1980s and was first used in the published record in the context it is now generally employed by Sanders and Field in 1984 4. The term reflects the most obvious spatial localization of the protein products of the canonical Saccharomyces cerevisiae septin genes cdc3, cdc10, cdc11 and cdc12 at the point of septation between mother and daughter in the budding cycle 5. The history of the field has been comprehensively recorded by John Pringle 5 and it is, as with many areas of scientific endeavour, characterized by a slow and faltering initial development in the hands of a small number of enthusiasts followed by a burgeoning growth of interest as relevance in disparate areas emerges.
A plethora of reviews on the subject of septins have been published 2, 3, 6–8, 9–13 and the field now benefits from biannual workshops 14. Our focus here is to bring together recent perspectives on the genomics and transcriptomics of the septin gene family, the biochemistry of the filaments formed by septin proteins and the newly emerging themes of how septins participate in physiological and pathological processes in mammalian, and especially human, systems. The central thesis promoted is that the diversity of septin polypeptides provides a constellation of potential pools and filaments that provide dynamic and regulatable higher-order structures that participate in key cellular processes and in particular those that revolve around membrane shape and compartmentalization 9, 13. Such processes include cytokinesis, but also a diverse array of other membrane-associated events. Association with the actin and microtubule cytoskeleton is also a feature of the septins. Indeed, the evolutionary history of septins appears to be associated with acquisition of new roles for these dynamic, non-polar yet asymmetrical heteroligomeric complexes. During the preparation of this review, Spiliotis and Gladfelter 13 published an enormously valuable synthesis of the field, emphasizing the roles of septins in providing a set of tools by which the cell can orchestrate asymmetry.
Genomic and transcriptomic complexity
The origins of the septin family lie in deep evolutionary history 15. Careful analysis of the then available databases suggested that septins were restricted to fungi and animals (opisthokont eukaryotes) 16–18. More recently, and as more genomic data are available from a broader range of organisms, it has become clear that other eukaryotes, including algae and ciliates, contain septin-like sequences 15. Notwithstanding these important new insights, the septins fall into the broad family of P loop GTPases 19, 20, and phylogenetic studies have defined five groups of orthologous septin families across eukaryotes 16. Gain and loss of sequence elements characterize this phylogeny 16–18, but all septin genes encode polypeptides with a GTP-binding domain with conserved GxxxxGK(ST), DxxG and xKxD motifs. This is flanked by an N-terminal polybasic domain and a C-terminal septin element (sometimes called a septin unique domain, or septin unique element). There are variable-length N-terminal extensions and the C-terminus can (but does not have to) contain a coiled-coil region 3. In addition, some septins (eg SEPT6 in man) lack the ability to hydrolyse GTP, due to a variant in the switch I regions (N78*) preventing interaction with the γ-phosphate of GTP 21.
The number of septin genes in any given organism is rather variable 16–18. Saccharomyces cerevisiae has seven septin genes—the canonical cdc3,cdc10,cdc11 and cdc12 and three others (Shs1/Sep7,Spr3 and Spr28). Such an arrangement is also seen in Schizosaccharomyces pombe, while Drosophila melanogaster has five septin genes (Sep1,Sep2,Sep4,Sep5 and Pnut). In stark contrast, Caenorhabditis elegans has two septin genes (unc-59 and unc-61) and Chlamydomonas has only one 7, 18, 22, 23. In vertebrates extensive gene duplication events have given rise to a plethora of septin genes. In man there are 13 septin genes and at least 11 pseudogenes (see Table 121). Bioinformatic analysis of the mammalian septin genes and their products allows the identification of four distinct families (see Table 124), with clear parallels in yeast and other species 16–18. The SEPT2 group has four members (SEPT1, 2, 4 and 5) with two coiled-coils; the SEPT6 group has five members (SEPT6, 8, 10, 11 and 14) with one coiled-coil domain; and the SEPT3 group (SEPT3, 9 and 12) has no coiled-coil. A curious and ill-understood feature of these groups is that a single member of each has a particularly long N-terminal extension (SEPT4, SEPT8 and SEPT9) 3. Of note is that these long N-terminal regions are predicted to have minimal regular secondary structure 25, are proline-rich and contain putative PIN1-binding sites (SP and TP). Finally, the SEPT7 group stands out with a single member (SEPT7) and seemingly unusual structural features 25. Recent structural studies 21, 26 provide experimental support for this classification and the characterization of the crystal structure of a core septin filament 26 and reports of the basic rules of septin filament assembly (see below) are consonant with this broad classification.
Table 1. The current status of the human septin gene family with known pseudogenes
Septin protein class
Nonsense mediated decay
Entrez Gene ID
Alias and previous names
Septin 13 is now known to be an expressed pseudogene designated SEPT7P2
A recurring theme in the mammalian septins (and possibly throughout the vertebrate septins) is extensive alternative splicing to give rise to an array of potential septin polypeptides for any given septin gene 3, 22. While in some cases there is extensive characterization of these splicing events (SEPT927, SEPT828, SEPT529 and SEPT1130), the catalogue remains as yet incomplete. Moreover, the functional significance of these transcripts and isoforms is at present obscure 22. However, certain phenomena are consistently observed, with the known splicing events giving rise to N- or C-terminal variants with maintenance of the core septin GTP-binding domain.
Alternative splicing is usually thought of in terms of creating polypeptide diversity, but its regulatory roles are potentially even more important 22, 31. In addition, in some cases the splicing events lead to varying 5′ or 3′ non-coding regions that have potential regulatory functions. For example, the SEPT9_i4 isoform can be encoded by two distinct transcripts (SEPT9_v4 and SEPT9_v4*), which differ in their 5′ untranslated regions (UTR). In terms of their expression, the SEPT9_v4 transcript is widely expressed in a range of normal adult and fetal tissues with low or barely detectable levels of SEPT_v4* 32. The functional significance of two transcripts encoding one isoform lies in the regulatory properties of the different UTRs 33. It is worth noting that bioinformatic analysis of all mammalian septin genes would suggest that there is generally one isoform of each septin gene that can be encoded by multiple mRNAs, varying either in their 5′ or 3′ UTR sequences 3, 22.
The mRNAs encoding SEPT9_i4 (SEPT9_v4 and v4*) both have long 5′ UTRs—approximately 700 bp. The two transcripts can be distinguished by the exons that comprise the first half of their 5′ UTR. In SEPT9_v4 this is exon 1β and in SEPT9_v4* it is exons 1ζ and 2. The final 300 bases in each UTR is common and represented by the initial sequence of bases in exon 3, up to the start codon for translation at position 732 in SEPT9_v4 and position 719 in SEPT9_v4*. An internal ribosome entry site has been mapped to this common region, which is regulated by a short (eight amino acids) upstream open reading frame found only in exon 1β of the SEPT9_v4 transcript 33. We now know that these two mRNAs are translated with differing efficiencies, such that SEPT9_v4* might be considered the constitutive transcript. In conditions of cellular stress, such as hypoxia or treatment with etoposide, when cap-dependent translation is halted, the translation of the SEPT9_v4 mRNA is also arrested, presumably because translation of the short upstream peptide has been halted. However, translation of the SEPT9_v4* mRNA continues under such conditions 33. The cell thus ensures a mechanism of generating this isoform even under conditions of stress, when synthesis of all proteins other than those that are key for survival has been shut down.
The idea that septins might be regulated by p53 was initially highlighted by Kostic and Shaw 34, who identified SEPT4 as a putative p53-responsive gene. This prompted a bioinformatic analysis of the human SEPT9 promoters. The concept that translation of the SEPT9_v4* transcript might form part of a cellular stress response was then strengthened by the observation that there are multiple putative p53-binding sites upstream of the transcriptional start site of this mRNA, while such sites are infrequent upstream of the first exons of other SEPT9 transcripts. We then demonstrated that these are functional and that levels of SEPT9_v4* and SEPT9_i4 are elevated in response to DNA damage in a p53-dependent manner 33. In contrast, levels of mRNAs and proteins of the other SEPT9 isoforms are unaltered (McKee, unpublished data). The information regarding SEPT9 genomics and its various transcripts, plus patterns of expression of transcripts (see next section), is consistent with complex and highly regulated control of the expression of septin isoforms. While we now have some clues as to the mechanism, we do not know why so much effort is expended on the regulation of septin levels.
This, then, provides experimental data that p53 regulates septin expression and highlights the point that such regulation is more likely to be transcript/isoform-specific rather than gene-specific. A further link between septins and p53 comes from the observations of Kremer et al35, who provide evidence that an interaction between the canonical septin 2–6–7 complex and SOCS7 in the cytoplasm acts to retain NCK cytoplasmically until both NCK and SOCS accumulate in the nucleus in response to DNA damage, with activation of downstream members of the DNA-damage kinase cascade, including p53. The effect of DNA damage in rearranging septins with subsequent rapid nuclear accumulation of NCK and SOCS7 thus links septins to the DNA damage response. Recently, SEPT4 has been shown to be regulated by the notch signalling pathway 36. The relevance of this and other signalling systems to septin biology will be an interesting feature of future studies.
Septin expression patterns
To date we have only a patchy and certainly incomplete knowledge of the details of the expression of the array of septin transcripts and isoforms in human cells and tissues, and the regulatory events that control this 3. Much has been done in culture to define septin profiles, but the lack of robust tools has limited the field. A low-resolution analysis by expression microarray has been reported 37 and a number of other surveys of particular septins have been published 28, 29, 38, 39. While the former has the merit of a large sample size, the available probe set directed at septin genes was limited and failed to encompass the complexity of the then known transcripts. With regard to some septins, those data have been validated by RT–PCR and limited antibody-based methods 40, 41. In other cases we remain only partially informed of the catalogue of septin expression and perturbation in disease. As with the full definition of the details of septin genomics and transcriptomics, there is an urgent need for a comprehensive analysis of septin expression profiles 22. Nevertheless, it is clear that there are profound differences in the expression of septins across human tissues and in disease states. We would contend that defining the details of septin transcript expression, the subcellular distribution and stoichiometries of the various isoforms will be an essential step towards developing a full understanding of septin biology.
At a subcellular level, it was proposed that a small number of well-defined expression patterns exist for septins, irrespective of the organism studied 42. Indeed, in their comprehensive and insightful analysis, Lindsey and Momany 41 reviewed the then known literature and were prescient in their suggestion that the apparent differences between organisms would diminish as more was learned (Figure 1). This has been the case. Furthermore, these authors highlighted the diversity of localization patterns and any comprehensive understanding of the functions of septins will need to take the set of subcellular locations into account. That argument needs to be extended to add definition of the diversity of isoforms that are known to exist. Where are they in cells, and what pools and populations exist? Indeed, understanding the subtleties of septin isoform levels and locations will be a cornerstone of future progress.
Septins are frequently, but not universally, associated with cell membranes. The most obvious membrane association occurs during cell division and in particular in cytokinesis. Septins are also frequently, and possibly universally, associated with cellular projections such as primary cilia 43, 44, neuronal spikes and dendrite formation 45–47, sperm morphogenesis 48 and hyphal growth in fungi 13, 49 (Figure 1). They may associate with the base of such projections or be at the tip, or on occasions be found in a punctate distribution in the cytoplasm associated with the projection.
Association of septins with nuclear membranes or distribution throughout the nucleus has also been described in mammals, and association of septins with the nuclear envelope has been described in yeast 50. While in yeast it has been proposed that this may at least in part relate to spatial partitioning of nuclear materials, including ‘young’ and ‘aged’ nuclear pores during cell division, and hence contribute to the regulation of yeast ageing 50, the existence of significant septin pools in mammalian interphase nuclei is puzzling. Other cleavage furrow-associated proteins, such as anillin 51, tumbleweed and pavarotti 52, have been demonstrated to have nuclear localization and certainly in the case of tumbleweed and pavarotti this localization has functional significance in terms of modulating wnt signalling 52. In that context it is worth noting that SEPT9_i1 has a functional nuclear localization signal and potential nuclear roles in modulating HIF1a responses 53. Clearly, we currently have a remarkably poor understanding of the possible nuclear functions of septins.
Septins can be seen in association with both actin- and tubulin-based filaments. In the latter case, septins seem to be frequently associated with polyglutamated microtubules 54. Of note is the interaction of septins with microtubule-associated proteins, thus providing a mechanism for regulating microtubule dynamics 55, 56 and plus-end microtubule dynamics 57. These different spatial arrangements of septins are dynamic: while being more stable than both microtubules and actin filaments, they turn over much faster than intermediate filaments 13. However, a clear theme is the highly regulated assembly of septins into filamentous structures that facilitate the compartmentalization and shaping of pre-existing cellular materials and the provision of: (a) scaffolds on which to build other structures; and (b) boundaries that prevent diffusion of components between compartments 9, 11, 13, 58.
Septin filament formation and other protein interactions
Early studies of septins demonstrated that they could form filamentous structures (reviewed by McMurray and Thorner 59), both alone and in association with other cytoskeletal elements, such as actin and tubulin filaments. A great deal of important information accrued, using an array of elegant biochemical and biophysical approaches, but the crucial observations came from the structural studies of Wittinghofer's group 21, 26. These build on earlier studies that, for example, showed that equimolar complexes of three septins could be isolated from mammalian cells 3, 60 and structural studies of these septins (using bacterially expressed recombinant proteins) showed that the core GTP-binding domains were central to complex formation (reviewed in 59). A 5 nm diameter, 25 nm long repeat unit made up of SEPT7–SEPT6–SEPT2–SEPT2–SEPT6–SEPT7 could be defined with the N- and C-terminal-containing face (designated NC) or the nucleotide-binding face (designated G) being involved in the interactions, not the coiled-coil domains as had been anticipated by some. Of note is the use of GDP alone in the SEPT7 homotypic interactions, but GTP and GDP in the SEPT6–SEPT2 interaction [Figure 2]. The filaments made by septins are remarkably similar across phylogeny 9, 11, 59. In man and other organisms the repeat unit is non-polar but the C-terminal coiled-coils are orthogonal to the axis of oligomerization and point in one direction (ie the unit filament has a definite sidedness). The coiled-coil domains are presumptive docking sites for other protein interactions and also appear to facilitate the formation of parallel septin filament arrays 61.
Other septin complexes have been isolated from human cells, including SEPT4, SEPT8, SEPT7 and SEPT5, SEPT11 and SEPT7 62, 63. It is of note that SEPT7 appears to be a common element in most such septin complexes and it has both unusual genomic 22 and structural features 25. It is also of note that comprehensive two- and three-hybrid studies of human septin interactions show that these and a range of other filaments are possible 62, 63 and that the early prediction by Kinoshita et al.38 that complexes would contain one septin from each of the three basic groups (SEPT1, SEPT2, SEPT4 or SEPT5 with SEPT6, SEPT8, SEPT10, SEPT11 or SEPT14 with SEPT7) was indeed correct. What is less clear is the place of the other septins (SEPT3, SEPT9 and SEPT12), which all lack a coiled-coil region 3. Certainly SEPT9 is sometimes reported in sub-stoichiometric levels in human septin complexes and yeast two- and three-hybrid studies indicate that SEPT9 can certainly engage in a diverse set of interactions 62, 63. Similarly SEPT3 can also engage in a range of septin–septin interactions 62, 63).
Recent studies from Sellin et al64 begin to provide some insights into the process by which septin complexes are assembled, with their demonstration that SEPT9 is incorporated only into heteromers of six subunits indicating that it is introduced late in the filament assembly. Their data is therefore consistent with a model in which temporal assembly of filaments is determined by a septin homology subgroup. These authors also reported that it is the larger SEPT9 isoforms (SEPT9_i1–3) that are preferentially incorporated into higher-order structures, rather than the smaller ones (SEPT9_i4 and SEPT_i5) at sub-stoichiometric amounts. Again we are faced with the need to understand the roles of the diversity of septin isoforms in the assembly (and potentially disassembly) of the filaments. Certainly we know that a truncated isoform of SEPT9 (i4) can disrupt the filament-associated location of the larger SEPT9 isoforms 65 and might be a potential chain modulator or even disruptor. Could it be that spatial and quantitative regulation of septin isoforms and their pools could play key roles in septin function? Notwithstanding that idea, some functional redundancy in the formation of septin filaments is highlighted by the relative lack of phenotype in some mice lacking in specific septins 66.
Septins not only interact with other septins but they form complexes with an array of other proteins 3, 7, 62, 63. In yeast, such interactions with signalling molecules and enzymes is crucial for cell viability 67, 68. Moreover, septin interactions with a range of proteins are pivotal for yeast cell polarity determination 69. While certainly incomplete, we are now gaining some insights into the human septin interactome 3, 62, 63. Proteins with a diverse array of functions have been reported to interact with septins. These particularly include proteins with roles in microtubule function, cell cycle control and dynamics, cell motility, protein modifications including phosphorylation and sumoylation/ubiquitination, endocytosis and intracellular trafficking, apoptosis and signalling cascades (see Table 2 in 63). Despite this progress, we have little insight into the dynamics and control of such interactions. In addition, there is a pressing need to build an understanding of how septin–septin interactions might modulate interactions with non-septin proteins and how post-translational modifications of septins could add additional levels of control.
So what do septins, the filaments that they form and the expanding catalogue of protein–protein interactions actually do? The bulk of data come from yeast and other fungi, although important observations have been made in other model systems, including Caenorrhabditis elegans and Drosophila. From these systems, as well as increasingly from study of mammalian septins, burgeoning evidence suggests that septin heteromeric complexes provide dynamic higher-order structures that can act as scaffolds of docking sites for other proteins at specific subcellular sites, as well as defining barriers to lateral mobility of membrane-associated proteins 9, 12, 13. Why are there so many septins (13 mammalian septin genes) and such a diversity of alternatively spliced transcripts and isoforms? The many parallels between septins across phylogeny argue for a set of basic properties and functions that have been added to over evolutionary time, new functions and roles appearing as nature employs the modular nature of septin biochemistry in new ways to facilitate new structures and functions. Gene duplication events perhaps providing the genetic substrate for diversity of role 16, 17, 18, 22.
Pathologists and other microscopists are all too well aware of the diversity of cell form and shape. Not only do the many cell types in Metazoa show morphological diversity, there is astonishing plasticity and the potential for dramatic changes over a range of time frames. The ability to shape and sculpt membranes using the contractile systems of the actin cytoskeleton, and in particular the cortical elements subjacent to the plasma membrane is poorly understood. Furthermore, it is increasingly clear that different regions of membranes, specific domains, can have remarkably different properties and attributes. These concepts are not simply a feature of the outer cell membrane but can also apply to other membranes systems and cell compartments. For example, a role in endoplasmic reticulum physiology has been shown in yeast 70. Many studies of septin expression have localized them to the cell cortex, being intimately associated with membranes. Indeed, the majority of septins have a polybasic region immediately N-terminal to the GTP-binding domain 3. This region has been shown to allow lipid binding and is important for membrane association 71, 72. What is less clear is why some septins do not have this polybasic domain 17.
The association of septins with particular regions of membrane appears to have important functional roles in differentiating and distinguishing specific domains and regions, one from another 73, 74. As with other aspects of septin biology, important clues to such roles came from studies in yeast 5, 7, 52, 75, 76. Again, however, the conservation of functionality means that septins retain and extend such roles in Metazoa. The existence of diffusion barriers in diverse cell types such as spermatozoa 48, epithelia 43, 44), neurons 45, 46, 47 and lymphocytes 77, and potentially all cell types, appears to depend on the association of septins with the cell cortex and in some way can prevent lateral movement of membrane-associated proteins 9, 73, 74. The roles of septins in defining asymmetry and cell polarity by defining diffusion barriers has recently been further highlighted in epithelia, where ciliogenesis and collective cell movement are regulated by septins 43, 44.
Those properties of septins in defining the mechanisms that control and affect the morphological and functional events underpinning cell polarization 13, 54 lead to an interesting possibility. Throughout biology, polarization of cells and cellular compartments is a constant feature 78. Both in non-cycling and cycling cells, the creation and maintenance of asymmetry and polarization are crucial. At the time of cell division there can be symmetrical or asymmetrical partitioning of cellular materials. In yeast the ability to distinguish mother and daughter depends on membrane and other compartmentalization events. Septins are known to have crucial roles in this by compartmentalizing and restricting the movement of diverse cellular factors and thus modulating their fate at the time of cell division 54. These functions seem to relate to two activities that may (or may not) be related; first, the scaffolding of non-septin moieties that are both enzymatic and regulatory in nature, and second by providing barriers that can prevent the movement of membrane-associated proteins 9, 69, 70, 71, 77.
While the majority of cell divisions in Metazoa are essentially symmetrical in terms of fate, crucially stem cell divisions are characterized by the potential for asymmetry of fate, with (on average) one daughter being committed to being a stem cell while the other has a non-stem cell fate. Asymmetry of cell division is recognized in mammalian cells and can contribute to stem cell function by contributing to polarizing the fate of cellular components, including DNA strands 79. It would be intriguing if the determinants of symmetry breaking in cell division were to be conserved across phylogeny from yeast to mammals. In that regard, we would then posit that septins might have a role in stem cell biology as determinants of asymmetrical divisions. Roles for p53 in regulating such events have been postulated 80, 81 and, as highlighted above, some septins (including SEPT4 and a specific SEPT9 transcript) are p53-regulated. Further support for a functional relationship between septins and stem cell properties comes from the observation that in mice Sept4 is required for stem cell apoptosis and by this means can contribute to tumour suppression 82. SEPT4_i1 has also been implicated in regulating the susceptibility of hepatocytes to apoptosis 83 and the properties of SEPT4 deserve careful consideration.
While the pathological literature tends to focus on mechanisms underlying disease processes or (more frequently) the diagnosis of disease, definition of prognosis and prediction of response to treatment, our perspective has been more in relation to how perturbation of septin expression and function in disease can illuminate septin biology 3, 6, 10, 36, 84, 85. That is, how can ‘nature's experiments’ provide us with insights into septin function? The catalogue of septin abnormalities in disease has been reviewed elsewhere 3, 84 and includes a diverse range of disease processes. As the tools for studying septins become more widely disseminated, this catalogue is likely to expand. Here we will highlight: (a) the observations made about septins in neoplasia; (b) the reports of how germline mutations in SEPT9 can contribute to hereditary neuralgic amyotrophy (HNA); and (c) the role of septins in the pathogenesis of intracellular infections, such as listeriosis.
Septins and neoplasia
The first clue to the possible role of septins in neoplasia came from the observation that SEPT9 could form an in-frame fusion with MLL in acute myeloid leukaemia 86. Since then, SEPT2, SEPT5, SEPT6, SEPT9 and SEPT11 have been shown to form similar in-frame fusions with MLL in haematological malignancies 3, 87. These result in chimeric fusion proteins where the N-terminus of MLL to the majority of the septin moiety. Such translocations are seen in de novo and treatment-related myeloid neoplasia, although septins are but one of the many fusion partners of the remarkably promiscuous MLL gene. Furthermore, there is considerable variation in the exact breakpoints involved in septin–MLL fusions 87. Despite this, several observations point to the non-random nature of the these genetic events, eg the involvement of multiple septins and the preference for 5′ events in the septin genes, the preferential association with AML, and the potential for the protein–protein interactions inherent in septin structure act to oligomerize the MLL moiety 87. The involvement of septins in non-haematological malignancy gene fusion events has not yet been reported but in this fast-developing field ‘absence of evidence is not evidence of absence’!
The second strand of evidence linking septins to neoplasia came from studies of allelic imbalance in ovarian and breast cancer 88, 89. Fine deletion mapping of a region of allelic imbalance on distal chromosome 17q suggested that the target gene was a new member of the septin family, subsequently renamed as SEPT9. This therefore implied that SEPT9 might act as a tumour suppressor gene; however, subsequent studies failed to identify coding region mutations 32. In independent studies, the orthologous murine locus was shown to be an integration site for the SL-3 retrovirus in T cell lymphomas 90. A second study in a murine model of breast cancer demonstrated high-level copy number increases in the majority of tumours analysed in the orthologous chromosomal region, and these tumours were shown to have elevated levels of Sept991. The picture became more complicated with emerging information regarding the series of transcripts encoded by SEPT9 and their individual analysis in normal and disease tissue. Putting together the exon information obtained from the two initial reports of allelic imbalance 88, 89, along with newly identified exons, it was shown that SEPT9 was represented by multiple splice variants with alternate 5′ and 3′ exons, but a common core of exons in which some transcripts were truncated versions of others 27. Many of these transcripts were similar in size. This therefore makes it difficult to interpret early studies of SEPT9 expression that relied on northern blotting or assessed general levels of SEPT9 transcripts. The availability of transcript-specific RT–PCR demonstrated that splice variants were expressed in a tissue-specific manner and that mammalian breast and ovarian neoplasia was accompanied by up-regulation of specific splice variants and down-regulation of another 32, 39, 40, 41, 92. The most striking changes involved three 5′ transcripts: SEPT9_v1,SEPT9_v4 and SEPT9_v4*.
The up-regulation of SEPT9_v1, a transcript encoding an isoform with the largest N-terminal extension has been linked to several tumour-associated phenotypes. There is now evidence that this isoform can localize to the nucleus where it stabilizes HIF1-α and leads to the activation of HIF-regulated target genes, thus linking SEPT9_i1 to angiogenesis 53, 93, 94. Over-expression of this isoform in immortalized human mammary epithelial cells lead to increased growth kinetics, cell motility and invasion, as well as cytokinesis defects and disruption of tubulin microfilaments 92. SEPT9_i1 over-expression may also promote genomic instability through both cytokinesis and mitotic spindle defects 95. The clinical relevance of this may manifest itself in resistance to microtubule-interacting agents 96. In addition to its interaction with HIF1-α, this isoform has also been linked to stabilization of JNK and the subsequent transcriptional activation of the JNK target genes. This, then, may be at least one mechanism by which SEPT9_i1 contributes to increased proliferation and indicates a role in cellular stress responses and survival 92–96.
Perhaps more intriguing is the altered ratio in tumours of two splice variants which encode the same isoform SEPT9_i4. As described above, in neoplasia there is a shift from one transcript (SEPT9_v4), which encodes this isoform in normal adult and fetal tissue, to SEPT9_v4*, where levels are normally low or undetectable. The down-regulation of SEPT9_v4 mRNA is apparently by methylation of an upstream CpG island since 5-azacytosine treatment of tumour cell lines can reactivate expression of the mRNA 32. In fact, this CpG island lies upstream of the first exons of both SEPT9_v2 and _v4 and up-regulation of both transcripts was observed by 5-azacytosine, with no detectable changes in levels of any other splice variants. This CpG island was also identified in a general screen for methylated genomic fragments in plasma of patients with colorectal cancer and has clinical utility as an early marker of disease 97, 98. Methylation of SEPT9 has also been reported in head and neck cancer 99. More recently there has been evidence that SEPT9_v3 might also be regulated epigenetically in breast cancer progression 100.
The significance of the switch between the two transcripts encoding SEPT9_i4 is unclear. This isoform is expressed predominantly in the cytoplasm 101, partially colocalizes with actin and tubulin filaments and was found at the base of actin-containing surface projections. In over-expression studies it led to enhanced motility and loss of polarity and seems to behave quite differently to SEPT9_i1 65. As noted earlier, isoform 4 expression could also disrupt larger SEPT9-containing structures. The switch within neoplasia to a transcript which encodes isoform 4 with altered regulatory control (ie SEPT9_v4 to SEPT9_v4*) might suggest that this truncated isoform has a different role to the larger isoforms in a septin-containing scaffold—perhaps one that acts to regulate the scaffold. The observation that all septin family members have one isoform encoded by multiple transcripts emphasizes the importance of such an isoform and perhaps gives us clues to mechanisms by which the cell can regulate overall septin scaffold function.
Finally, there are also altered ratios of 3′ splice variants in ovarian neoplasia, again with increased prevalence of variants with low levels of expression in normal tissues (‘b’ and ‘c’; Cunnea and Russell, unpublished data). As stated above, the ‘a’ variant has a long 3′ UTR, whereas ‘b’ and ‘c’ have short 3′ UTRs. We do not know the significance of this, but a plethora of functions have been attributed to 3′UTRs, including regulation of translation, miRNA binding or intracellular localization; regulation of expression would be in keeping with our other observations of septins 102. Also worth noting is a potential CAAX box present in the COOH terminus of the ‘a’ variant, which is absent in the ‘b’ and ‘c’ forms (Cunnea and Russell, unpublished observations). This would imply the existence of a mechanism by which some septin isoforms might be subject to post-translational regulation (prenylation) prior to their localization at cell or nuclear membranes, where they fulfil a particular role 103.
Septins and hereditary neuralgic amyotrophy
Hereditary neuralgic amyotrophy (HNA) is a rare autosomal dominant relapsing–remitting disorder characterized by episodes of pain and weakness in the distribution of the brachial plexus. While a sporadic condition (Parsonage–Turner syndrome) with similar clinical features has been reported 104, both this and the autosomal dominant HNA can be precipitated by infections, immunization, childbirth and exercise. The locus involved in HNA had been mapped to a 1.8 mb region of 17q25 105, 106. Further analysis identified mutations (131 G → C, 262 C → T, 278 C → T) in the SEPT9 gene among 10 families with HNA 107. Other families with 262 C → T and 278 C → T have been reported 108, 109 and non-recurrent duplications in SEPT9 have also been linked to HNA 110.
This experiment of nature, and the first Mendelian disorder to map to a septin gene, provides a new insight into septin biology but also poses yet more questions. How can this germline abnormality manifest in such a highly localized and episodic manner? The mutations map to exon 3 and can lead to the production of a mutant SEPT9 protein (R88W) but, as has been highlighted above, there is extensive splicing of the SEPT9 gene with varying N-terminal polypeptides and 5′ UTRs. Nagata's group have proposed that the R88W mutation has effects on perturbed septin complex formation and altered rho signalling 111. Such changes would be presumed to be constitutive in nature and it is hard to link that to the episodic and stress-related clinical features. An alternative perspective might come from observations on the regulatory roles of the SEPT9_v4 and SEPT9_v4* 5′ UTRs 33. While some HNA mutations map to the protein-coding region of the larger isoforms, for the smaller isoforms they are non-coding and at least two are present in the common region of the 5′ UTRs of the SEPT9_v4 and SEPT9_v4* transcripts. Specifically, they map to the beginning of exon 3 and to the highly structured internal ribosome entry site. Experimentally, the introduction of the 262 C → T mutation into the 5′ UTR of both transcripts abrogates the usual stress-related translational repression of the SEPT9_v4 transcript, leaving the SEPT9_v4* transcript unaffected 33. Provocatively, the stoichiometry of this isoform may play a part in determining the specifics of septin complex formation (see below) and has been shown to delocalize septins from filaments 65. Could this then alter, for example, microtubule dynamics 55, 56, 57, which are key participants in the function of the long neurons of the brachial plexus? While we cannot discount the role of altered protein structure, the possibility that HNA is a disorder of septin regulation is attractive.
Septins, phagocytosis and intracellular bacterial infections
As highlighted above, septins have roles in membrane dynamics. The ability of cells to take up materials by phagocytosis and macro-pinocytosis depends on such membrane events and linkages to the cell cortex and actin cytoskeleton. The details of such linkages and their regulation remain poorly understood. Grinstein and Trimble hypothesized that septins may have a role in the mechanisms of FcgR-mediated phagosome formation 112. With a combination of localization studies, knock experiments, expression of exogenous transgenes and the use of pharmacological inhibitors, they conclusively demonstrated that this is indeed the case and that septins, and in particular SEPT2 and SEPT11, are required for phagosome formation 112. In parallel studies, Cossart's group have provided compelling evidence for the role of septins in the entry of intracellular organisms such as Listeria into non-phagocytic cells. Their early studies demonstrated the association of SEPT9 isoforms with phagosomes containing Listeria surface protein-coated beads 113. More recently, a role for SEPT2 and SEPT11 have been demonstrated in this process 114–118, indicating a clear commonality between the biology of phagocytic and non-phagocytic cells.
We now know that septin scaffolds play a role in multiple and diverse cellular processes, and we are beginning to develop insights into the structure–function relationships that underpin this plethora of roles. A unifying theme would seem to be a dynamic septin scaffold with a hexameric structure composed of three septin family members (eg 2–6–7) linked via C-terminal coiled-coils to form parallel arrays. Additional septin moieties (such as SEPT9) can modify such arrays. The existence of multiple septin isoforms, varying in their precise amino- and carboxy-termini, would then be a mechanism by which the cell confers variable properties on a given scaffold. This would be consonant with the complex regulation of each gene and the independent regulation of its isoforms. Furthermore, the observations that pathological processes may be associated with a breakdown of this tight regulation, where imbalance of individual transcripts predominates, rather than coding sequence mutation, would support this perspective.
Of course, many of the reports of disease associations may simply be that, associations, and may be the consequences of pathological processes, as opposed to having pathogenetic significance per se. Even such epiphenomena require definition and explanation and might yet provide insights into septin biology. Moreover, there remains the possibility that there are other contexts where septins have significant mechanistic relevance to disease processes. One is struck, for example, by the plethora of reports linking septins to neuropsychiatric disorders, and this may reflect the biology of septins in spine and neurite formation and neuronal plasticity and architecture. Given the burgeoning knowledge of how septins contribute to the definition of regions of membrane space and zones of cytoskeletal architecture, and can hence define cellular spatial asymmetry 13, it would not be surprising if a wealth of information about septins is yet to be uncovered in the study of pathological states. Central to that would be the ability to quantitate and image specific septin isoforms and complexes, and the tools to do this are becoming available. Pathology will be a fertile field for septin research in the coming years.
We thank current and past members of our laboratories and the many members of the septin community for their extraordinary collegiality and spirit of interaction. Septin research by the authors has been funded by the Department of Employment and Learning (Northern Ireland), WellBeing of Women, Northern Ireland Leukaemia Research Fund, Action Cancer and the Pathological Society. PAH acknowledges the generous support of the King Faisal Specialist Hospital and Research Centre. Finally, despite being a young field, there is already a huge literature on septins, and we apologize to authors whose contributions have not been cited because of space limitations.
Both authors participated fully in the writing of this review.