The EMLs are a conserved family of microtubule-associated proteins (MAPs). The founding member was discovered in sea urchins as a 77-kDa polypeptide that co-purified with microtubules. This protein, termed EMAP for echinoderm MAP, was the major non-tubulin component present in purified microtubule preparations made from unfertilized sea urchin eggs [J. Cell Sci. (1993) 104, 445–450; J. Cell Sci. (1987) 87(Pt 1), 71–84]. Orthologues of EMAP were subsequently identified in other echinoderms, such as starfish and sand dollar, and then in more distant eukaryotes, including flies, worms and vertebrates, where the name of ELP or EML (both for EMAP-like protein) has been adopted [BMC Dev. Biol. (2008) 8, 110; Dev. Genes Evol. (2000) 210, 2–10]. The common property of these proteins is their ability to decorate microtubules. However, whether they are associated with particular microtubule populations or exercise specific functions in different microtubule-dependent processes remains unknown. Furthermore, although there is limited evidence that they regulate microtubule dynamics, the biochemical mechanisms of their molecular activity have yet to be explored. Nevertheless, interest in these proteins has grown substantially because of the identification of EML mutations in neuronal disorders and oncogenic fusions in human cancers. Here, we summarize our current knowledge of the expression, localization and structure of what is proving to be an interesting and important class of MAPs. We also speculate about their function in microtubule regulation and highlight how the studies of EMLs in human diseases may open up novel avenues for patient therapy.
Expression, localization and structure of EMLs
Mammals express six EML proteins that have been reported under different names, but we refer to here as EML1 to EML6 (Figure 1). EML1, 2, 3 and 4 share a similar protein organization with an N-terminal region of approximately 175–200 residues that appears largely unstructured apart from a short coiled-coil, and a C-terminal structured domain of approximately 650 residues consisting of multiple WD (tryptophan–aspartate) repeats. EML5 and EML6 are distinct in that they lack the N-terminal region and have three contiguous repeats of the C-terminal WD repeat domain. EML1 and EML4 are highly expressed in early mouse embryos, but exhibit a lower and more restricted expression pattern in late embryos and adults with the most common site of expression being the nervous system, including the hippocampus, cortex, cerebellum, eyes and olfactory bulb [15,20]. EML5 is also primarily expressed in the nervous system of the adult mouse , whereas EML3 was detected not only in the brain but also in the liver and kidney [17,34]. Human studies suggest widespread expression of at least EML1 and EML2, although for all the EMLs there is clear evidence of differential splicing . Caenorhabditis elegans ELP-1 is also expressed in a variety of adult tissues, including the body wall, muscles and intestine .
EML protein organization.
Localization studies, mostly in cultured cells, together with in vitro biochemical studies provide overwhelming evidence that EMLs are microtubule-binding proteins. EMAP was originally purified using a biochemical approach from microtubules prepared from unfertilized sea urchin eggs. Antibodies raised against EMAP detect a similar sized protein by Western blot in microtubule preparations from a variety of echinoderm species, while it decorated microtubules in sea urchin embryos and adult coelomocytes by immunofluorescence microscopy . Expression of recombinant proteins confirmed the association of mammalian EML1, 2, 3 and 4 with the microtubule lattice [10,15,24,26,33]. However, the analysis of endogenous EML localization has, at least until recently, been hampered by the lack of good, commercially available antibodies. Peptide-specific antibodies raised to EML3 and EML4 have, however, confirmed localization of these proteins to microtubules in cultured cells [6,33]. Taken together, the expression and localization studies are indicative of a family of cytoplasmic microtubule-binding proteins with widespread expression, but which may have particular roles in neuronal tissues.
Structural studies on human EML1, 2 and 4 reveal interesting and unexpected features of this family of proteins [25,26]. The crystal structure of the C-terminal WD domain from human EML1, comprising residues 167–815, was determined to a resolution of 2.6 Å (Figure 2A). The WD repeats form 13 individual β-sheet structures that form the blades of two β-propeller domains that are closely and rigidly connected. The first β-propeller is formed from seven WD repeats. Unexpectedly, the second propeller is assembled from six WD repeats and an additional subdomain that is formed from separate regions of the primary sequence of EML1. This domain appears to be a ubiquitous and unique feature of EML family proteins, and was termed the tandem atypical propeller in EML (TAPE) domain. The tandem propeller arrangement creates a relatively planar structure with a concave and convex surface and, sequence comparison among the human EMLs, suggests stronger conservation of the concave surface. Crystal structures of the coiled-coil regions of human EML2 and EML4 were determined to resolutions of 2.1 and 2.9 Å, respectively (Figure 2B). In both structures, three molecules of the EML protein come together through a core of hydrophobic interactions stabilized by salt bridges to form a trimerization domain (Figure 2C). The conservation of primary sequences indicates that the coiled-coil domains of human EML1–4 are all trimeric, although co-precipitation experiments indicate that these proteins have the potential to assemble heterotrimers, as well as homotrimers .
Structured domains of the EML protein family.
Microtubule binding and regulation of EMLs
The mechanism of microtubule binding of EMLs is intriguing. Studies on sea urchin EMAP revealed that microtubule binding was conferred through a region towards the N-terminus of the protein . Indeed, it was suggested that a highly conserved ‘HELP’ (hydrophobic ELP) motif of approximately 40 residues within this N-terminus was responsible for microtubule binding [10,32]. However, chimeras that specifically fused the HELP motif to EGFP did not localize to microtubules , and structural studies have since revealed that the HELP motif is an integral part of the TAPE domain and contributes to protein folding rather than microtubule binding . An alternative feature of the N-terminus that could be responsible for microtubule association is its basic nature that is conserved across metazoans. Many MAPs interact with microtubules through electrostatic interactions between basic regions of the MAP and the negative surface of microtubules created by the exposed C-terminal tails of α-/β-tubulin that are rich in acidic residues, particularly glutamate . The one letter code of glutamate, E, has led to these tails being referred to as ‘E-hooks’ with various MAPs, including dynein and kinesin motors, binding via the E-hooks. Limited proteolysis with subtilisin can cleave the E-hooks from polymerized microtubules in vitro and this has been exploited to test whether the binding of MAPs is dependent on E-hooks. Subtilisin digestion of microtubules did not, however, prevent the association of sea urchin EMAP arguing against binding occurring via electrostatic forces . However, we have recently found that binding of the basic N-terminal fragment of human EML1 to microtubules is disrupted by subtilisin digestion, arguing in favour of an electrostatic mechanism of interaction for this protein (Montgomery et al., in preparation).
The N-terminal regions of EML1–4 have a coiled-coil that promotes trimerization . Deletion of the coiled-coil reduces microtubule binding of EML1, and this domain might contribute to microtubule association either by direct binding or via promoting oligomerization. EML5 and EML6 lack the coiled-coil, but do have three repeats of the TAPE domain encoded within a single polypeptide. Hence, they may adopt a similar overall tertiary structure that favours microtubule binding, although whether EML5 and EML6 associate with microtubules remains to be tested. Intriguingly, sea urchin EMAP does not have a coiled-coil and has only one copy of the TAPE domain, yet localizes to microtubules via its N-terminus. Moreover, Drosophila ELP-1, also called doublecortin (DCX)-EMAP, contains within its N-terminus a specific sequence that bears homology to a domain in DCX that directly contributes to microtubule binding . Together with the difference in subtilisin response of EMAP, this may suggest that EMAP, ELP-1 and the human EML proteins have different mechanisms of microtubule binding. On the other hand, it may simply reflect differences in the relative contributions of electrostatic and conformational factors within the N-terminal regions that contribute to microtubule binding. Furthermore, studies on sea urchin EMAP suggest a second, weaker microtubule-binding site in the C-terminal TAPE domain of the protein that could contribute to its overall microtubule affinity . The isolated TAPE domain of human EML1 did not localize to microtubules in cultured cells, but did associate tightly with soluble α/β-tubulin heterodimers via interactions with conserved residues on its concave surface [25,26]. Whether this reflects a distinct role in binding soluble tubulin, or rather a second microtubule-binding site that is only effective in the context of a trimer, will be important to examine.
To date, most published studies have reported no significant differences in the association of EMLs with microtubules in interphase and mitotic cells. Indeed, sea urchin EMAPs, as well as mammalian EML1–4, have all been reported to localize to spindle microtubules in mitosis, as well as the microtubule cytoskeleton in interphase [10,20,24,30,33]. However, we recently found that there may be significant differences in the relative affinity of EMLs for microtubules in interphase and mitosis, with EML3 and EML4 exhibiting reduced localization to microtubules in mitosis as compared to interphase (Montgomery et al., in preparation). This is potentially important as there is a dramatic difference in the dynamic properties of microtubules between these two phases of the cell cycle, and the altered localization of EMLs could reflect a functional contribution to these changes. Moreover, a question is raised as to how the affinity of EMLs for microtubules is regulated through the cell cycle. A strong candidate is phosphorylation with hyperphosphorylation of EML4 reported in mitotic cells, and evidence that sea urchin EMAP is not only phosphorylated in mitosis, but can also interact with the mitotic kinase, CDK1 [5,24]. EMLs have also been identified in interactome studies with the mitotic NEK6 kinase . Our preliminary studies support a functional relationship with NEK6 phosphorylation reducing the affinity of EML3 for microtubules in mitotic cells (Montgomery et al., in preparation). Phosphoproteome data reveal a concentration of phosphorylation sites within the N-terminal basic regions of human EMLs, consistent with the hypothesis that phosphorylation could directly regulate microtubule affinity by altering electrostatic interactions.
EML functions in differentiated and proliferating cells
Biochemical studies on the sea urchin EMAP first revealed the potential of these proteins to alter microtubule dynamics. Purified EMAP caused an increase in microtubule dynamics with the suppression of rescue events in vitro that is consistent with overall destabilization . The limited studies that have since been undertaken on the human EMLs suggest that they might represent an unusual class of MAPs in which some members promote microtubule stabilization, while others promote destabilization. This notion is based on the fact that overexpression and depletion studies argue that EML4 is a microtubule-stabilizing protein [15,24], whereas EML2 can act as a microtubule destabilizer reducing growth rates and promoting catastrophe .
The key question with regard to the biochemical activity of EMLs is how EMLs influence the dynamic properties of microtubules at the molecular level. Structure–function studies with both sea urchin EMAPs and mammalian EMLs have revealed the presence of two distinct domains, one that binds to polymerized microtubules and another that binds to soluble α-/β-tubulin heterodimers [25,26]. In this respect, they resemble the chTOG/XMAP215 family of microtubule polymerases that associate with microtubules via a disordered basic region, while binding to tubulin heterodimers via multiple TOG domains . The combination of these two properties within a single protein, together with their preference for plus ends, allows them to stimulate microtubule growth and prevent catastrophes by processively tracking plus ends and catalyzing incorporation of tubulin dimers. The presence of multiple TOG domains increases the local concentration of tubulin heterodimers also promoting growth and it is intriguing that the human EMLs, through trimerization or having multiple WD domains in a single polypeptide, could have the same effect. Hence, it is attractive to speculate that this could explain how EML4 acts to stabilize microtubules. However, there is no evidence to date that EML4 concentrates at plus ends. Equally, it would not explain how EML2 promotes microtubule destabilization unless this reflects a subtle change in equilibrium between a conformation that promotes assembly and one that promotes disassembly. It is also undoubtedly true that in the ‘busy’ confines of the microtubule cytoskeleton, some of the consequences of EML proteins in regulating microtubule dynamics will arise from competition with other microtubule stabilization and destabilization factors.
Although the molecular details remain elusive, there is sufficient evidence that EMLs contribute to microtubule organization to expect them to have roles in both differentiated and proliferating cells (Figure 3). First of all, the pronounced expression of EMLs in the nervous system suggests a functional significance in microtubule-dependent processes in neuronal cells. They may be important for maintaining the particular architecture of these cells or they may have specific roles in mechanical or sensory signal transduction. In C. elegans, expression of ELP-1 in mechanoreceptor and ciliated neurons suggests a role in mechanotransmission and this is backed up by observing reduced touch sensitivity in cells lacking ELP-1 . Meanwhile, in muscle cells, ELP-1 may promote force generation by enabling attachment of adhesion complexes on the cell surface to the underlying microtubule network . In flies, the ELP-1 (DCX-EMAP) protein localizes to ciliated neurons in the auditory organ of the fly and insertional mutants have mechanosensation defects, including deafness and unco-ordinated movement . Hence, in both these organisms, there is evidence for mechanosensory roles, including in ciliated cells, although the mechanisms remain far from clear. Most differentiated epithelial cells in vertebrates possess a primary cilium that contributes to the detection of various external stimuli, including chemicals, movement and light . However, there is no evidence to date implicating the mammalian EMLs in cilia; for example, they are not present in ciliary proteomes or transcriptomes regulated by ciliary-specific transcription regulators, such as RFX or FoxJ1 . There is also no evidence for localization of EMLs to ciliary, or centriolar, microtubules, both of which exhibit increased stability and post-translational modifications such as acetylation and glutamylation . This lack of localization to ciliary or centriolar microtubules might suggest that the mammalian EMLs are not required for the assembly of highly stabilized microtubule structures nor for intraflagellar trafficking. On the other hand, they may have specific mechanosensory roles in restricted tissue types that have yet to be identified.
Potential functions of EML proteins in cell cycle progression.
There is good evidence that EMLs also play a role in proliferating cells and particularly during cell division. During mitosis, the microtubule cytoskeleton undergoes a dramatic reorganization as microtubules become short and highly dynamic in contrast with the long, relatively stable microtubules characteristic of interphase cells . This change is regulated at least in part via differential binding of MAPs. EML3 co-localizes with the mitotic spindle and mid-body microtubules in HeLa cells and is important for metaphase chromosome alignment . As well as being hyperphosphorylated in mitosis, EML4 is localized to the mitotic spindle and its depletion inhibits cell proliferation [5,24]. EML4 is required for the organization of the mitotic spindle and specifically for the proper attachment of spindle microtubules to kinetochores in metaphase. This function seems to involve the recruitment of the nuclear distribution gene C protein, NUDC, to the mitotic spindle . The study of EML1 function using HeCo mice, a spontaneous model of a neurodevelopment disorder, suggests that EML1 may contribute to spindle orientation . Unlike in wild-type brains, cells at the ventricular lining of mutant animals exhibited a reduced frequency of vertically oriented spindles. The authors proposed that the consequence of this was the inappropriate release of cells from the ventricular lining that retained progenitor markers; the presence of these ectopic cells in the white matter is the cause of neuronal dysfunction. The question, however, of how EMLs contribute to spindle orientation, as well as spindle dynamics and chromosome capture, remains to be addressed.
While the interaction of EMLs with mitotic kinases may reflect regulation of EML function by phosphorylation, it is possible that EML proteins may act as scaffolds to localize mitotic kinases to microtubules. Indeed, besides the direct regulation of microtubule dynamics by the EMLs themselves, it is worth more generally considering the potential role of EMLs in recruiting other regulators of microtubule dynamics to the microtubule cytoskeleton in both interphase and mitosis.
EMLs in human disease
Genetic defects involving EMLs have been associated with neuronal disorders and cancer. However, the first suggestion of a disease link came when the EML1 gene was mapped to a locus on chromosome 14 that was responsible for Usher syndrome type 1, in which patients suffer from deafness and blindness . These symptoms are typical of syndromic ciliopathies, inherited disorders that result from defects in primary cilia . Primary cilia, as indicated above, are microtubule-based organelles raising the possibility that EML1 may represent a ciliopathy disease gene. However, causative mutations in EML1 responsible for Usher syndrome type 1 have yet to be identified and there is no evidence for localization of EML1 to either primary or motile cilia. Mutations in EML1 on the other hand have more recently been shown to lead to a developmental brain disorder in both humans and mice . Specifically, point mutations in EML1 that most likely disrupt folding of the TAPE domain cause a null phenotype that manifests as neuronal heterotopia. This condition involves the presence of misplaced neural progenitor cells in the white matter of the brain that potentially arise as a result of defects in spindle orientation in the neocortex. Epilepsy is one symptom of neuronal heterotopia, and as it happens, unusually high expression of EML5 is detected in the anterior temporal neocortex of patients with intractable epilepsy . In this latter case, however, it remains to be seen whether EML5 expression contributes to the disease phenotype.
Particularly exciting is the finding that EMLs are present in oncogenic fusion proteins in human cancer. EML1–ABL1 fusions were first identified in T-cell acute lymphoblastic leukaemia, whereas EML4–ALK fusions were subsequently identified in lung, breast and colorectal cancers [8,22,27,28]. The fusions have the catalytic domain of the ABL or ALK tyrosine kinases at the C-terminal end of the protein joined to variable amounts of the N-terminal region of the EML protein (Figure 1). All fusions have at least the coiled-coil motif of the EML and there is good reason to believe that this promotes constitutive activation of the tyrosine kinase through oligomerization and trans-autophosphorylation. However, in the case of non-small-cell lung cancer patients, multiple different EML4–ALK variants have been identified in which the breakpoint can lead to the presence of substantially different amounts of the EML4 protein. At first sight, this may not seem particularly important. However, there is emerging evidence that the position of the breakpoint can be crucial in terms of both progression of disease and response to particular treatments .
Despite identification of the EMLs as abundant and conserved MAPs three decades ago, our journey to understand their mechanisms of action and biological roles has only just begun. Yet what we have learnt to date reinforces the fact that these proteins are likely to have essential roles in cell and tissue organization. Clearly, there is a need for detailed biochemical studies in purified systems to get to grips with the regulation of MT dynamics by EML proteins. Equally, it is imperative to know how they interact and co-operate with other MAPs, and how they respond to the complex tubulin ‘code’ created by tubulin isotype expression and post-translational modification . In terms of studying their role in interphase and mitotic events, interpreting the data will be complicated by the fact that the human proteins can exhibit hetero-oligomerization , and there is the potential for redundancy and adaptation by alternative splicing within the family. The discovery of EML1 mutations in a neurodevelopmental disease makes one suspect that EML mutations are going to underlie other diseases, including not only neuronal conditions but also perhaps ciliopathies. In this regard, studies of knockout animals should be highly informative in relation to the roles in developmental and tissue-specific processes. Finally, there is the exciting opportunity to use our growing knowledge of the biology of EMLs to consider novel therapeutic approaches for cancers driven by EML fusion proteins. The appreciation, through structural analysis of the TAPE domain that oncogenic fusions that disrupt ordered domains are heavily dependent on chaperones for stability , has justified testing chaperone inhibitors not only against EML4–ALK-positive tumours but also other cancers driven by inherently unstable fusion proteins. Moreover, the demonstration that the fusion breakpoint can influence stability and therapeutic response strengthens the argument for stratified approaches to cancer treatment. Furthermore, the involvement not only of EMLs but also of other MAPs in oncogenic fusions, such as FGFR3-TACC3 , will lead to new rationales for more targeted use of the well-proven microtubule poisons in future cancer treatment.
DCX, doublecortin; ELP, EMAP-like protein; EMAP, echinoderm MAP; EML, EMAP-like protein; HELP, hydrophobic ELP; MAPs, microtubule-associated proteins; TAPE, tandem atypical propeller in EML.
We acknowledge the support for this research in our laboratories from Worldwide Cancer Research, Cancer Research UK, The Wellcome Trust and the Biotechnology and Biological Sciences Research Council.
We particularly thank Dr Kathy Suprenant for her pioneering work in this field.
The Authors declare that there are no competing interests associated with the manuscript.