Research ArticleMitosis

Mitotic phosphorylation by NEK6 and NEK7 reduces the microtubule affinity of EML4 to promote chromosome congression

See allHide authors and affiliations

Science Signaling  13 Aug 2019:
Vol. 12, Issue 594, eaaw2939
DOI: 10.1126/scisignal.aaw2939

NEKs and EML4 in cell division

Faithful cell division requires proper separation of duplicated chromosomes into daughter cells. When this process goes awry, the dividing cell or daughter cells carry too many or too few copies (known as aneuploidy), which can cause disease. Microtubule dynamics are critical to this process. The protein EML4, which is abnormally fused to other proteins in some cancers, binds to and stabilizes microtubules, and its presence in the cell is critical for mitotic progression. However, Adib et al. found that upon entry into mitosis, phosphorylation of EML4 by the kinases NEK6 and NEK7 induced its dissociation from microtubules to enable proper chromosome alignment at the spindle equator, ready for subsequent anaphase and cytokinesis. These findings reveal further insight into both the regulation of cell division and how NEKs promote tumor growth.


EML4 is a microtubule-associated protein that promotes microtubule stability. We investigated its regulation across the cell cycle and found that EML4 was distributed as punctate foci along the microtubule lattice in interphase but exhibited reduced association with spindle microtubules in mitosis. Microtubule sedimentation and cryo–electron microscopy with 3D reconstruction revealed that the basic N-terminal domain of EML4 mediated its binding to the acidic C-terminal tails of α- and β-tubulin on the microtubule surface. The mitotic kinases NEK6 and NEK7 phosphorylated the EML4 N-terminal domain at Ser144 and Ser146 in vitro, and depletion of these kinases in cells led to increased EML4 binding to microtubules in mitosis. An S144A-S146A double mutant not only bound inappropriately to mitotic microtubules but also increased their stability and interfered with chromosome congression. In addition, constitutive activation of NEK6 or NEK7 reduced the association of EML4 with interphase microtubules. Together, these data support a model in which NEK6- and NEK7-dependent phosphorylation promotes the dissociation of EML4 from microtubules in mitosis in a manner that is required for efficient chromosome congression.


Dynamic instability is an essential property of microtubules that allows them to play diverse roles in intracellular trafficking, organelle positioning, cell migration, and cell division (1). Furthermore, the capacity to switch between relatively stable and unstable states is key to the reorganization of the microtubule network that occurs between interphase and mitosis, which enables assembly of the mitotic spindle upon which chromosomes are segregated. Microtubule dynamics are dependent on the intracellular concentration of tubulin, the rate of guanosine triphosphate (GTP) hydrolysis by the tubulin heterodimers, and the activity of numerous microtubule-associated proteins (MAPs) that bind the tips and lateral surface of microtubules (25).

One family of MAPs that remains relatively poorly characterized is the EMAP-like, or EML, proteins. EMAP (for echinoderm MAP) was first identified in unfertilized sea urchin eggs as the major nontubulin component of the mitotic spindle (6, 7) and has since been determined to be highly conserved in various organisms, including flies, worms, and humans (810). Their association with microtubules has been shown in different systems, different family members are suggested to contribute to stabilization and/or destabilization of microtubules (1113), and an observed delay in mitotic chromosome congression upon depletion of EML3 or EML4 in human cells confirms their importance for cell division and supports roles for members of the EML family in spindle organization (14, 15). Nevertheless, their precise function in regulating microtubule dynamics is unclear.

Six human EMLs have been described (EML1 to EML6), with EML1 to EML4 sharing a similar organization of an N-terminal domain (NTD) of about 175 to 200 residues encompassing a coiled-coil motif and a C-terminal domain of about 650 residues consisting of WD (Trp-Asp) repeats (16). Crystallographic studies have revealed that the coiled-coil motif forms a homotrimer, referred to as the trimerization domain (TD), whereas the WD repeats assemble into two juxtaposed seven-bladed β-propellers, named the TAPE domain for tandem atypical β-propellers in EMLs (17, 18). A strongly conserved HELP (hydrophobic EMAP-like protein) motif located toward the start of the TAPE domain provides key residues at the interface between the two β-propellers necessary for proper folding. The TAPE domain in isolation does not localize to microtubules, although it does bind tightly to soluble tubulin. Interdigitation of the two β-propellers generates a curved sheet with a concave and convex surface; the concave surface shares considerable homology between the different EMLs, and point mutations in this region disrupt tubulin binding (17). Microtubule binding is conferred by the NTD and requires both the TD and a basic region that lies between the TD and start of the TAPE domain. EML5 and EML6 lack the NTD but have three contiguous copies of the TAPE domain. Hence, whereas EML1 to EML4 can assemble into trimeric complexes that together contain three TAPE domains, EML5 and EML6 contain three copies of the TAPE domain encoded within a single polypeptide. However, because they lack the NTD, it remains unclear whether EML5 and EML6 can bind microtubules.

EML proteins have attracted considerable interest from the cancer community since discovery of translocations involving the genes encoding EML1 and EML4. Although rare in frequency, an EML1-ABL1 fusion protein has been identified in T cell acute lymphoblastic leukemia (19). In contrast, the EML4-ALK fusion protein is present in a substantial proportion (~5%) of lung adenocarcinoma patients, as well as breast and colorectal tumors (2022). Both ABL1 (Abelson 1) and ALK (anaplastic lymphoma kinase) are tyrosine kinases, and fusion of the C-terminal catalytic domain of the kinases to the N-terminal region of the EMLs leads to constitutive kinase activation as a consequence of TD-mediated oligomerization (23). Different breakpoints in the EML4 gene lead to distinct EML4-ALK variants that are associated with variable disease progression and therapeutic response in different patients (24). EML1 is also implicated in the pathology of an inherited developmental brain disorder where point mutations in the EML1 TAPE domain that potentially destabilize the protein cause neuronal heterotopia in both rodents and humans (25).

The microtubule cytoskeleton undergoes notable reorganization upon entry into mitosis with a switch from long, relatively stable microtubules to short, unstable microtubules. This switch is largely driven through a change in the complement of MAPs associated with microtubules and major changes in microtubule nucleation capacity (26). Observations that both sea urchin EMAP and human EML4 undergo phosphorylation during mitotic progression suggest that regulation of EML proteins may contribute to these changes in the microtubule network (27, 28). EML2, EML3, and EML4 were identified in a large-scale proteomic analysis of human NIMA-related kinase 6 (NEK6) binding proteins (29). NEK6, and the closely related protein NEK7, are cell cycle–dependent serine/threonine kinases that are activated in mitosis, downstream of another member of this kinase family, NEK9 (3033). NEK9 is a dimeric protein that is phosphorylated in its noncatalytic C-terminal domain by the mitotic kinase, cyclin-dependent kinase 1 (CDK1). This creates a binding site for another mitotic kinase, polo-like kinase 1 (PLK1), which phosphorylates the catalytic domain of NEK9 activating it in mitosis. Activated NEK9 then undergoes autophosphorylation at a different site in its noncatalytic domain that enables direct binding of NEK6 or NEK7. NEK9 can subsequently activate NEK6 and NEK7 both through direct phosphorylation and by promoting their dimerization and autophosphorylation. The use of catalytically inactive mutants has shown that activation of NEK6, NEK7, and NEK9 in mitosis is essential for bipolar spindle assembly and efficient chromosome congression (14, 3436).

Here, using antibodies against the endogenous protein, we set out to explore how EML4 might be regulated through the cell cycle. We found not only that the microtubule affinity of EML4 was reduced in mitosis but also that this resulted from phosphorylation by the NEK6 and NEK7 kinases in the N-terminal microtubule-binding region. Our biochemical and structural data show that association of EML4 with microtubules occurs through electrostatic interactions between a basic region within EML4 and the acidic tubulin C-terminal tails. Together, our results indicate that EML4 is a microtubule stabilizing protein, and its displacement from spindle microtubules is essential for chromosome congression in mitosis.


Association of EML4 with microtubules is reduced in mitosis

To examine whether the subcellular localization of EML4 is regulated in a cell cycle–dependent manner, two different EML4 antibodies were used. The first detects the N-terminal region between residues 150 and 200, whereas the second detects the C terminus between residues 951 and 981 (Fig. 1A). Antibody specificity was confirmed by detection of a band at the expected size for EML4 (120 kDa) by Western blot analysis and loss of this band upon depletion of EML4 with two distinct small interfering RNA (siRNA) oligonucleotides (Fig. 1B). Likewise, confocal microscopy with these two antibodies revealed cytoplasmic staining that colocalized with microtubules in human U2OS osteosarcoma cells and that was lost upon depletion of EML4 (Fig. 1, C and D). Comparison of EML4 localization in interphase and mitotic cells with these antibodies revealed that, although the protein was strongly detected on the microtubule network in interphase, it was difficult to detect on spindle microtubules in mitosis (Fig. 1, E and F). Quantification revealed an approximate twofold reduction in colocalization of EML4 with microtubules in metaphase compared to interphase (Fig. 1, G and H). A similar loss of association with microtubules was observed for recombinant EML4 upon transition from interphase to mitosis in U2OS cells stably expressing yellow fluorescent protein (YFP)–tagged EML4 (Fig. 1, I and J), supporting the hypothesis that affinity of EML4 for microtubules is reduced in mitosis.

Fig. 1 EML4 exhibits reduced affinity for microtubules in mitosis.

(A) Schematic representation of full-length (FL) human EML4 indicating the trimerization (TD; pink), basic region (brown), and TAPE (blue) domains. Epitopes recognized by the two commercial EML4 antibodies (red) map to residues 150 to 200 (N-terminal antibody) and 951 to 981 (C-terminal antibody). (B) Lysates prepared from U2OS cells that were transfected with either control siRNA (siMock) or one of two EML4-targeted siRNAs for 72 hours were analyzed by Western blot with the antibodies indicated. Blots are representative of three independent experiments. (C and D) Confocal microscopy of cells transfected as described in (B) and fluorescently stained for N-terminal (C) or C-terminal (D) EML4 (green) and α-tubulin (red). Scale bars, 5 μm. Images are representative of three independent experiments. (E and F) Confocal microscopy of U2OS cells captured in interphase or metaphase and fluorescently stained for N-terminal (E) or C-terminal (F) EML4 (green) and α-tubulin (red). “Zoom” panels are magnified views of the merged imaging. Scale bars, 5 μm. Images are representative of three independent experiments. (G and H) Colocalization, assessed as the mean Pearson’s correlation coefficient (R; y axis) between EML4 and microtubules for cells shown in (E) and (F). Data are means ± SD from five lines per cell in 10 cells per each of three independent experiments. I, interphase; M, metaphase. (I and J) Confocal microscopy of U2OS cells captured in interphase and metaphase that were transfected with YFP-EML4 24 hours and then stained for GFP (green), α-tubulin (red), and Hoechst 33258 (for DNA; blue). Scale bars, 5 μm. Colocalization between the GFP and microtubules for cells shown in (I) was calculated as in (G) and (H). Images are representative of and data are means ± SD from 30 cells pooled from three independent experiments. **P < 0.01 and ****P < 0.0001 by a one-tailed unpaired Student’s t test.

The EML4-NTD is phosphorylated in mitosis

To explore how this cell cycle–dependent change in microtubule affinity is regulated, we first examined whether the endogenous EML4 protein was modified upon mitotic entry. Although Western blot analysis revealed no change in abundance of EML4 between extracts prepared from interphase and mitotic HeLa and U2OS cells, there was a distinct reduction in electrophoretic mobility in mitosis compared to interphase in both cell types (Fig. 2A). We then tested which region of the EML4 protein was required for microtubule binding and, as previously shown for EML1 (18), confirmed that this property is conferred by the NTD (residues 1 to 207) and not the TAPE domain (residues 208 to 878) of EML4 (Fig. 2, B and C). The isolated EML4-NTD not only colocalized efficiently with microtubules but also led to formation of extensive microtubule bundles in the cytoplasm, a phenotype not seen upon expression of the full-length (FL) EML4 protein. Because microtubule association is dependent on the EML4-NTD, we asked whether a similar cell cycle–dependent change in gel mobility was detected upon expression of the isolated NTD. Whereas no change in migration was observed for YFP alone, the YFP-tagged EML4-NTD migrated as two distinct bands in lysates prepared from mitotic cells as compared to a single band in interphase cell lysates (Fig. 2D). This suggests that a substantial proportion (~25 to 30%) of the EML4-NTD is modified upon entry into mitosis. To confirm that this gel shift was the result of phosphorylation, we treated lysates taken from mitotic cells expressing the YFP-EML4-NTD protein with λ-protein phosphatase and observed a potentially dose-dependent loss of the slower migrating form (Fig. 2E), confirming that this is a phosphorylated version of the EML4-NTD protein. That only a fraction of the recombinant protein was phosphorylated is consistent with ectopic expression of the EML4-NTD saturating the cellular kinase(s) responsible for this modification.

Fig. 2 The EML4-NTD binds microtubules and is phosphorylated in mitosis.

(A) Western blotting for EML4, cyclin B1, phospho-histone H3 (pHH3), and α-tubulin in HeLa and U2OS cells that were either untreated (I) or treated with nocodazole (500 ng/ml) for 16 hours to arrest them in mitosis (M) before being lysed. Blots are representative of three independent experiments. (B) Confocal microscopy of U2OS cells transfected with YFP-tagged FL or isolated peptide (NTD, TAPE) EML4 constructs and then fixed and stained for GFP (green), α-tubulin (red), and Hoeschst 33258 (for DNA; blue). Scale bars, 5 μm. Images are representative of three independent experiments. (C) Colocalization, assessed as the mean Pearson’s correlation coefficient (R; y axis), between the EML4 construct and microtubules (GFP and α-tubulin signals) in cells represented in (B). Data are means ± SD from five lines per cell in 10 cells per each of three independent experiments. (D) Western blotting as described in (A) in U2OS cells transfected with YFP or YFP-EML4-NTD for 24 hours. Blots are representative of three independent experiments. (E) Western blotting for GFP in U2OS cells transfected with YFP-EML4-NTD for 24 hours from which lysates were treated with the indicated units of λ-PPase for 30 min. Blots are representative of three independent experiments. ****P < 0.0001 by a one-tailed unpaired Student’s t test.

Phosphorylation of EML4 by NEK6 and NEK7 regulates its microtubule affinity

To determine which kinase (or kinases) may be responsible for this phosphorylation, we took advantage of the previous observation of an interaction between EML4 and the mitotic NEK6 and NEK7 kinases (29). We incubated FL human EML4 protein purified from insect cells with recombinant NEK6 or NEK7 kinase and adenosine triphosphate (ATP). The EML4 proteins were then excised from an SDS–polyacrylamide gel and subjected to mass spectrometry. This revealed four sites that were phosphorylated by NEK6 (Ser144, Thr490, Thr609, and Ser981) and four sites phosphorylated by NEK7 (Ser134, Ser146, Thr609, and Ser981) (Fig. 3A and fig. S1). Analysis of published phosphoproteome data on EML4 through PhosphoSitePlus revealed clustering of phosphorylation within the NTD and at the extreme C terminus. Of the three sites within the NTD phosphorylated by NEK6 and NEK7 (Ser134, Ser144, and Ser146), the latter two were, by far, the most commonly reported serine/threonine phosphorylation sites indicated in the PhosphoSitePlus database based on high-throughput proteomic discovery mass spectrometry. Moreover, Ser144 and Ser146, but not Ser134, are conserved across vertebrate species and, in most cases, have a hydrophobic residue at position −3 (Fig. 3B), typical of NEK6 and NEK7 phosphorylation sites (37, 38). We therefore focused on the potential importance of these two sites and generated single S144A and S146A phospho-null mutants, as well as a combined S144/146A double mutant, in constructs expressing a YFP-tagged EML4-NTD. Western blot analysis of lysates from subsequently transfected U2OS cells revealed no substantial change in migration of the upper band with the single mutants in cells arrested in mitosis. However, the S144/146A double mutant exhibited a much reduced gel shift, confirming that phosphorylation of these two sites is likely to be responsible for the reduced gel mobility in mitotic cells (Fig. 3C). An S144/146A double mutant was therefore also generated in a construct expressing YFP-tagged FL EML4. Whereas the wild-type EML4 protein did not obviously localize to spindle microtubules, the S144/146A double mutant very clearly did (Fig. 3, D and E). Conversely, a double phosphomimetic EML4-S144/146D mutant associated less well with microtubules than the wild-type protein in interphase cells (fig. S2, A and B). Hence, we conclude that phosphorylation at Ser144 and Ser146 reduces the affinity of EML4 for microtubules.

Fig. 3 NEK6 and NEK7 regulate association of EML4 with microtubules.

(A) Schematic representation of FL EML4 with TD and TAPE domains for reference, indicating the phosphorylation sites detected by mass spectrometry after incubation in vitro with NEK6 (blue) or NEK7 (green). Sites phosphorylated by both kinases are indicated in red. (B) Sequence alignment from species across the five vertebrate classes of the EML4-NTD region spanning the three identified phosphorylation sites (in red). (C) Western blotting in U2OS cells transfected with wild-type (WT) or mutant YFP-EML4-NTD for 24 hours and then treated with nocodazole for 16 hours to arrest cells in mitosis. Blots are representative of three independent experiments. (D and E) Confocal microscopy (D) and analysis of Pearson’s correlation coefficient for colocalization (E) of U2OS cells transfected with wild-type or double-mutant YFP-tagged FL EML4 for 24 hours and then stained for GFP (green), α-tubulin (red), and Hoechst 33258 (for DNA; blue). Scale bar, 5 μm. Images are representative of three independent experiments. Data are means ± SD from five lines per cell in 10 cells per each of three independent experiments. (F) Western blotting for the indicated proteins in U2OS cells transfected with YFP-EML4-NTD and either control siRNA (siMock) or NEK6-, NEK7- or NEK9-targeted siRNA for 48 hours and then treated with nocodazole (500 ng/ml) for 16 hours to arrest cells in mitosis. Blots are representative of three independent experiments. (G and H) Confocal microscopy (G) and analysis of Pearson’s correlation coefficient for colocalization (H) of U2OS cells transfected as in (F) for 72 hours and then stained with EML4 (green), α-tubulin (red), and Hoechst 33258 (for DNA; blue). Scale bar, 5 μm. Images are representative of and data are means ± SD from five lines per cell in 10 cells per each of three independent experiments. ***P < 0.001 and ****P < 0.0001 by a one-tailed unpaired Student’s t test.

To determine whether NEK6 and NEK7 can regulate the association of EML4 with microtubules, these kinases were depleted from U2OS cells, and the migration of YFP-EML4-NTD was analyzed in mitotic lysates by Western blot. This revealed that depletion of NEK6, NEK7, or the upstream NEK9 kinase that is responsible for activating NEK6 and NEK7 or chemical inhibition of CDK1 that operates upstream of NEK9 all led to substantial loss of the slower migrating form of the EML4-NTD protein (Fig. 3F and fig. S2C). Furthermore, analysis of endogenous EML4 by immunofluorescence microscopy revealed that depletion of NEK6, NEK7, or NEK9 led to increased association of EML4 with spindle microtubules (Fig. 3, G and H). Depletion of NEK9 led to an increase above that of NEK6 or NEK7 depletion alone. Conversely, expression of constitutively active mutants of NEK6 (Y108A), NEK7 (Y97A), or NEK9 (ΔRCC1) in U2OS cells led to a significant reduction in association of endogenous EML4 with interphase microtubules (fig. S2, D and E). However, the reduction of EML4 association with interphase microtubules upon expression of constitutively active NEK9 was unaffected by addition of the CDK1 inhibitor, supporting the notion that NEK9 acts downstream of CDK1 (fig. S2F). Together, these data provide persuasive evidence that NEK6 and NEK7, acting downstream of NEK9 and CDK1, regulate microtubule association of EML4 through phosphorylation of residues in its NTD.

The EML NTD binds microtubules through interaction with tubulin C-terminal tails

In EML1, it was shown that the TD and a basic sequence that lies between the TD and TAPE domain confers microtubule binding to EMLs (18). The region between the TD and TAPE domain is also highly basic in the EML4 protein [residues 64 to 207; pI (isoelectric point) = 10.23], and so, we hypothesized that association may depend on electrostatic interaction between this basic region and the acidic surface of the microtubule created by the C-terminal tails of α- and β-tubulin. These C termini protrude on the exterior surface of the microtubule and are rich in glutamate (E) residues. A number of MAPs associate, at least in part, through interaction of basic domains with these so-called tubulin E-hooks, including MAP2 and tau (39). To seek evidence that the basic nature of the EML4-NTD mediates microtubule association, the tubulin C-terminal tails of microtubules were removed by limited proteolysis with subtilisin. SDS–polyacrylamide gel electrophoresis (PAGE) analysis confirmed that tubulin now migrated as a doublet after subtilisin treatment, indicative of partial cleavage of the C-terminal tails (Fig. 4A). When incubated with two different concentrations of purified EML4-NTD protein, there was a marked reduction in sedimentation of this protein with the subtilisin-treated microtubules as compared to untreated microtubules (Fig. 4, A to C). We planned to use total internal reflection fluorescence (TIRF) microscopy of fluorescently labeled proteins to confirm this dependence on the tubulin C-terminal tails. However, it proved impossible to generate a purified version of fluorescently tagged EML4-NTD protein. We therefore undertook this analysis with a purified YFP-EML1-NTD fragment (residues 1 to 174). Using a mixed population of untreated microtubules labeled with a 561-nm fluorophore and subtilisin-treated microtubules labeled with a 640-nm fluorophore, we found that the EML1-NTD protein bound strongly to untreated microtubules but not to microtubules treated with subtilisin (Fig. 4, D and E). Together, these data are consistent with the basic N-terminal region of EML proteins associating with the acidic tubulin C-terminal tails on the microtubule surface.

Fig. 4 Subtilisin treatment of microtubules leads to loss of association of EML4 and EML1 in vitro.

(A) Microtubules (MTs) were either untreated or incubated with subtilisin before repurification to remove the enzyme. These were then incubated with purified EML4-NTD before sedimentation to generate a supernatant (S) and pellet (P) fraction. Samples were then analyzed by SDS-PAGE and Coomassie Blue stain. A representative example of three independent experiments is shown. (B and C) The relative fraction of the EML4-NTD protein present in the pellet fraction represented in (A) is shown for two different concentrations of the EML4-NTD protein. Data are means ± SD from three independent experiments. *P < 0.05; n.s., nonsignificant; by a one-tailed unpaired Student’s t test. (D) Representative TIRF image of fluorescently labeled microtubules (left) treated with (magenta) or without (green) subtilisin before incubation with YFP-EML1-NTD (residues 1 to 174) (right). Images are representative of three independent experiments. (E) Box plots reveal YFP intensity (a.u., arbitrary units) associated with X-rhodamine (561) and Hilyte640-labeled microtubules that were either untreated or treated with subtilisin. Boxes represent quartiles; whiskers show 10/90% data ± SD (n > 200 microtubules, pooled from two independent experiments).

Cryo–electron microscopy reveals binding of EML4 to C-terminal tubulin tails

To examine this interaction of the EML4-NTD with microtubules in more detail, we first performed structured illumination microscopy (SIM) of endogenous EML4 in U2OS cells. This revealed small, evenly sized puncta of EML4 distributed along the surface of the cytoplasmic microtubules (Fig. 5, A and B). There was no obvious concentration at microtubule ends as has been seen for plus-tip tracking proteins, such as the EB (end-binding) or ch-TOG (colonic, hepatic tumor overexpressed gene) proteins (5). We then used cryo–electron tomography to directly visualize purified EML4-NTD bound to microtubules in vitro. Low-resolution reconstructions showed density corresponding to EML4-NTD bound along the microtubule outer surface consistent with the SIM analysis of endogenous EML4 in cells (Fig. 5C). It also revealed that while there is occasional evidence of evenly spaced interactions with the microtubules every 4 nm (corresponding to tubulin monomers), binding of EML4-NTD is overall rather disordered. We processed images of EML4-NTD bound to microtubules using single-particle averaging algorithms and revealed additional density on both α- and β-tubulin due to bound EML4-NTD (Fig. 5, D and E, and fig. S3, A to D). Although the resolution of tubulin in our reconstruction is better than 4 Å (Table 1) and shows clear evidence of discrimination between α- and β-tubulin, density corresponding to EML4-NTD is present at substantially lower resolution. This EML4-NTD density lies above the C-terminal helices H11 and H12 of each tubulin monomer, next to where the C-terminal tails of each monomer emerge from the microtubule wall. This confirms the biochemical evidence for involvement of tubulin C-terminal tails in EML4-NTD binding. The low resolution of the EML4-NTD density and the limited extent of density that is visualized are likely due to the flexible nature of its interaction with the tubulin C-terminal tails, which themselves are unstructured and rarely visualized in cryo–electron microscopy reconstructions.

Fig. 5 Superresolution and cryo–electron microscopy reveal EML4-NTD binding microtubules through a flexible interaction with α- and β-tubulin C-terminal tails.

(A and B) U2OS cells were stained for EML4 (C-terminal antibody; magenta) and α-tubulin (green) and imaged by SIM. Boxed region in (A) is magnified in (B), as an example of clusters of EML4 foci that colocalize with microtubules. Scale bar, 5 μm. Images are representative of three independent experiments. (C) Section through a cryo-tomogram of an EML4-decorated microtubule. The microtubule lumen is indicated in light blue false color. Red arrows indicate clear microtubule-bound densities, with sizes consistent with EML4-NTD trimers. Red arrowhead notes where, in some regions, a periodicity of 4 nm for extra densities was observed, consistent with binding to both α- and β-tubulin. Scale bar, 30 nm. (D) C1 single-particle cryo–electron microscopy reconstruction of EML4-NTD–decorated 13 protofilament microtubules (resolution, 4.4 Å), low-pass–filtered to 20 Å. Density within 9 Å of the fitted α- and β-tubulin atomic models is indicated with light and dark gray, respectively, and defined, connected density > 9 Å away is indicated with red. (E) Symmetrized reconstruction (resolution, 3.6 Å), showing two tubulin dimers within a single protofilament. Local resolution–filtered α- and β-tubulin density is shown as transparent light and dark gray density, respectively. The fitted H12 atomic model is shown as ribbons, with the flexible C-terminal tails indicated by dashed lines. The reconstruction low-pass–filtered to 20 Å is also shown as mesh, with extra densities associated with α- and β-tubulin colored in red and orange, respectively.

Table 1 Cryo-EM data collection, refinement, and validation statistics.

Data collection and processing information for the EML4-NTD–bound 13-protofilament HeLa cell tubulin microtubule dataset and refinement statistics for the HeLa cell tubulin dimer model asymmetric unit built into the symmetrized electron density map.

View this table:

Depletion of EML4 reduces stability of interphase microtubules

As microtubules exhibit reduced stability in early mitosis, the loss of EML4 from microtubules at this time could be important for these altered dynamics. This would be consistent with a previous report that overexpression of EML4 stabilizes microtubules in Cos7 monkey cells (11). To test whether endogenous EML4 promotes microtubule stability in human cells, we first examined the consequences of depleting EML4 in U2OS cells on the sensitivity of microtubules to the depolymerizing agent, nocodazole. Incubation of mock-depleted cells with nocodazole did not affect microtubule organization; however, this dose (75 nM) led to loss of an intact microtubule network in EML4-depleted cells (Fig. 6A). Furthermore, EML4 depletion led to reduced sedimentation of microtubules in the presence of Taxol (10 μM), also suggestive of impaired stability (Fig. 6, B and C). Stable microtubules are subject to more posttranslational modifications, including acetylation and detyrosination. Depletion of EML4 led to significant reduction of both these markers as measured by Western blot or immunofluorescence microscopy (Fig. 6, D to F, and fig. S4). Conversely, time-lapse imaging of live cells incubated with the fluorescent SiR-tubulin probe demonstrated that mitotic spindle microtubules were more resistant to nocodazole-induced depolymerization upon transient expression of YFP-EML4-S144/146A as compared to wild-type YFP-EML4 protein (Fig. 6, G and H). Together, these data confirm that binding of EML4 promotes microtubule stabilization and provide a rationale for why it is removed from microtubules upon mitotic entry.

Fig. 6 Displacement of EML4 from microtubules reduces their stability in mitosis.

(A) U2OS cells were transfected with either control siRNA (siMock) or one of two siRNAs targeting EML4 (siEML4.1 or siEML4.2) for 72 hours and then either left untreated or treated with 75 nM nocodazole for 2 hours. Cells were fixed and stained with α-tubulin antibodies. Scale bars, 5 μm. Images are representative of three independent experiments. DMSO, dimethyl sulfoxide. (B) Microtubule sedimentation assay performed with lysates prepared from U2OS cells that were transfected as in (A) and Western blotting the pellet (P) and supernatant (S) fractions for α-tubulin. (C) Quantification of the % tubulin in the pellet fraction represented in (B). Data are means ± SD from three independent experiments. (D to F) Western blotting and analysis for acetylated (Ac.), detyrosinated (Detyro.), and total tubulin in lysates from U2OS cells transfected as in (A). Blots are representative of three independent experiments. Data are means ± SD from three independent experiments. (G) U2OS cells were transfected with YFP-EML4-WT or YFP-EML4-S144/146A and incubated with SiR-Tubulin to visualize microtubules, the intensity of which was measured every 60 s after addition of nocodazole (200 μg/ml). Stills from representative movies are shown at 0, 7, and 14 min. Scale bar, 5 μm. (H) The relative SiR-Tubulin intensity of cells described in (G) is shown. Data are means ± SD from three independent experiments with a minimum of 10 positions per condition. ***P < 0.001 and ****P < 0.0001 (C, E, and F) and the P value in (H) assessed by a one-tailed unpaired Student’s t test.

Phosphorylation of the EML4-NTD is required for efficient chromosome congression

To directly test the importance of EML4 displacement from microtubules by phosphorylation in mitosis, we undertook time-lapse confocal imaging of HeLa cells expressing either the wild-type or phospho-null EML4 mutant. These cells also stably express an enhanced green fluorescent protein (EGFP)–Lamin A construct that labels the nuclear envelope and an mCherry-histone H2B construct that labels the chromatin enabling visualization of nuclear envelope breakdown (NEBD) and chromosome dynamics. Whereas in untransfected cells, NEBD was quickly followed by chromosome congression and anaphase onset (as indicated by sister chromatid separation), these events were delayed in cells expressing the wild-type EML4 protein (Fig. 7, A to C). Again, this can be explained by overexpression of EML4 saturating the machinery required to regulate its localization. However, expression of the phospho-null S144/146A double mutant protein led to notable failure of both chromosome congression and anaphase onset with less than 30% cells successfully completing cell division after 4 hours in mitosis (Fig. 7, A to C). Fixed imaging of U2OS cells arrested in mitosis with MG132 to prevent anaphase onset revealed that 90% of untransfected cells and 75% of cells transfected with wild-type EML4 had condensed chromosomes that were fully aligned on the metaphase plate; in contrast, only 20% of cells transfected with the EML4-S144/146A mutant had fully congressed chromosomes (Fig. 7, D and E). Moreover, staining for the microtubule network revealed that the spindle microtubules in these cells were unusually long (Fig. 7D). Last, we tested whether the mitotic delay in cells expressing the phospho-null EML4 mutant was the result of activation of the spindle assembly checkpoint (SAC). Increased recruitment of the SAC component BUBR1 (budding uninhibited by benomyl-related protein 1) to chromatin-associated foci and loss of the mitotic delay upon treatment of cells with an inhibitor of the MPS1 kinase that is essential for SAC activity confirmed SAC activation in cells expressing the phospho-null EML4 mutant (Fig. 7, F to H). Thus, the kinase-dependent displacement of EML4 from microtubules in mitosis is necessary for assembly of a functional mitotic spindle capable of efficient chromosome congression and satisfaction of the SAC.

Fig. 7 A phospho-null EML4 mutant interferes with chromosome congression and activates the SAC.

(A) HeLa:EGFP-Lamin A/mCherry-H2B cells were either untransfected (left column) or transfected with YFP-EML4 wild-type protein (WT; middle column) or YFP-EML4-S144/146A (right column) for 24 hours before time-lapse confocal imaging. Stills from representative movies are shown with time (in minutes) indicated from mitotic entry. Scale bar, 7.5 μm. Images are representative of three independent experiments. (B and C) Quantification of cells described in (A) indicating time from nuclear envelope breakdown (NEBD) to last chromosome congressed (B) and time from last chromosome congressed to anaphase onset (C). Data are means of cumulative frequencies ± SD; n = 20. (D) U2OS cells were either untransfected or transfected with YFP-EML4-FL, EML4-WT, or EML4-S144/146A for 24 hours before being treated with RO-3306 for 16 hours, followed by 4 hours with MG132. Cells were then fixed and stained for GFP (green), α-tubulin (red), and Hoechst 33258 (DNA; blue). Scale bar, 5 μm. Images are representative of three independent experiments. (E) Quantification of the percentage of cells described and represented in (D) with normal or uncongressed chromosomes in mitosis. Data are means ± SD from at least 30 cells pooled from three independent experiments. (F) U2OS cells were either untransfected or transfected with EML4-S144/146A for 24 hours before being fixed and stained for GFP (green), BUBR1 (red), and Hoechst 33258 (DNA; blue). Scale bar, 5 μm. Images are representative of three independent experiments. (G) The relative intensity of BUBR1 was quantified and plotted relative to that in control samples. Data are means ± SD from 30 cells pooled from three independent experiments. ****P < 0.0001. (H) Histogram shows the mitotic index upon MPS1 inhibition (MPS1i) in U2OS cells transfected with either EML4-WT or EML4-S144/146A constructs. Data are means ± SD from three independent experiments. *P < 0.05 and **P < 0.01 by a one-tailed unpaired Student’s t test.


Human EML4 is phosphorylated on serine/threonine residues in mitosis (27). Here, we found that EML4 undergoes phosphorylation at Ser144 and Ser146 in the N-terminal microtubule-binding region in mitosis and that this reduces its affinity for microtubules. This phosphorylation is catalyzed by the NEK6 and NEK7 kinases and perturbs binding of the EML4 protein with the surface of the microtubule lattice (Fig. 8A). Confocal and superresolution microscopy indicated the presence of EML4 foci that are distributed along the length of polymerized microtubules. Our data suggest that this interaction most likely occurs through electrostatic interaction of the basic NTD of EML4 with the acidic tubulin C-terminal tails (so-called E-hooks) that are exposed on the surface of the microtubule. This conclusion is based not simply on the fact that phosphorylation—the addition of negative charge—weakens the interaction but on the observation that limited proteolysis of the microtubules that cleaves off these C-terminal tails abrogated binding. Cryo–electron microscopy and image reconstruction provided definitive evidence that the EML4-NTD binds at the site on the microtubule where the C-terminal tails emerge on both α- and β-tubulin. Furthermore, spacing of the EML4-NTD foci suggests that the protein can bind to both tubulin monomers within an individual heterodimer. It will be intriguing to further explore the way in which the acidic C-terminal tails from tubulin interact with the trimerized EML4-NTD and how variations in tubulin isoforms and/or posttranslational modifications affect this interaction (Fig. 8B). What is also not yet clear is whether the FL EML4 protein binds in an identical manner and how the TAPE domain, which binds soluble tubulin heterodimers, is oriented with respect to the microtubule.

Fig. 8 Phospho-dependent regulation of EML4 microtubule binding affinity.

(A) Schematic model illustrating how the NEK9-NEK6-NEK7 kinases regulate affinity of EML4 for microtubules at the G2-M transition. Phosphorylation of two sites within the EML4-NTD as a result of activation of these kinases reduces electrostatic interaction of this basic region of EML4 for the acidic microtubule surface. (B) Cartoon showing electrostatic interaction of the basic NTD of the trimeric EML4 protein with the acidic tubulin C-terminal tails (green).

The mechanism by which EML4 stabilizes microtubules in cells, either directly or indirectly, also remains enigmatic. However, EML4 shares many features with the ch-TOG (XMAP215) family of MAPs in that both proteins have separable domains for binding the microtubule lattice and soluble tubulin. ch-TOG acts as a processive microtubule polymerase by binding to the microtubule with a basic region and then using its multiple TOG domains to add soluble tubulin to the growing microtubule plus-ends (40, 41). However, although EML4 has a basic NTD that binds the microtubule polymer and a TAPE domain that binds soluble tubulin, there was no detectable concentration of EML4 at plus ends of microtubules where it could promote growth through acting as a microtubule polymerase. Similar to tau, EML4 is abundant in the nervous system, suggesting that it may have a major function in stabilizing the long microtubules present in neurons (11). At the molecular level, proteins of the tau family bind to more than one tubulin dimer, thereby preventing catastrophe and slowing shrinkage and promoting rescue, while also being displaced from the microtubule lattice by phosphorylation (3, 4). The fact that EML4 exists as a trimer could allow it to bind to more than one tubulin dimer, thereby increasing its overall avidity for the microtubule surface and potentially directly stabilizing the polymer. However, in contrast to tau (42), EML4 does not interact precisely with the microtubule surface, but it rather loosely and flexibly associates with the tubulin C-terminal tails. Also, in contrast to tau and its relatives, EML2 and EMAP have been reported to destabilize microtubules by increasing the catastrophe rate and decreasing the rescue rate, respectively (12, 13). In this regard, it is interesting that the nonneuronal isoform of EML2 used in that study, as well as EMAP, lack the TD, raising the possibility that trimerization is necessary for EMLs to stabilize microtubules. Conversely, a longer isoform of EML2 that does have a TD is found in the brain and spinal cord (43), where it could also stabilize microtubules.

Another key finding of this study is the discovery of a new role for the mitotic NEK6 and NEK7 kinases. Depletion of NEK6 or NEK7, or the upstream kinase NEK9 that activates NEK6 and NEK7, leads to increased association of EML4 with spindle microtubules. Conversely, expression of constitutively active mutants of NEK6, NEK7, or NEK9 reduces binding of EML4 to interphase microtubules. This provides compelling evidence that these enzymes play a central role in regulating microtubule binding of EML4. Previous depletion and overexpression studies have shown that NEK7 increases the overall dynamicity of interphase microtubules supporting a role for NEK7 in regulating microtubule dynamics in interphase and mitosis (44). We found that NEK6 and NEK7 were capable of phosphorylating EML4 in vitro, including at Ser144 and Ser146, whereas mutation of these two sites or depletion of NEK6, NEK7, or NEK9 led to loss of a gel mobility shift of the EML4-NTD in mitosis. Together with the fact that EML4 is present in a complex with NEK6 and NEK7 in cells (29), this strongly suggests that EML4 is a direct physiological substrate for these kinases. However, we do not exclude the possibility that EML4 may also be subject to additional phospho-dependent regulation by other mitotic, as well as nonmitotic, kinases. The sea urchin EMAP protein was found to be efficiently phosphorylated by CDK1 (28). We also observed loss of the gel mobility shift of the EML4-NTD in mitosis upon addition of a CDK1 inhibitor, but this result can be explained by CDK1 acting upstream of NEK9. Analysis of the PhosphoSitePlus database suggests that whereas Ser144 and Ser146 are the two most commonly detected serines or threonines phosphorylated in EML4, there is even more frequent detection of phosphorylation at Tyr226. This residue sits at the junction of the NTD and TAPE domain, and it will be interesting in the future to test the role of this phosphorylation on EML4 function.

Depletion of NEK6 or NEK7 disrupts spindle formation in mitosis, and both the kinesin Eg5 and the chaperone HSP72 have been reported as mitotic substrates of NEK6 (35, 45). Our work here indicates that EML4 should be added to that growing list of substrates for these kinases and that, like CDK1, PLK1, and Aurora-A, the mitotic NEKs phosphorylate multiple substrates to drive robust spindle assembly. Removal of EML4 from microtubules appears to be crucial for the decrease in stability that accompanies assembly of a highly dynamic mitotic spindle. Live-cell imaging demonstrated not only that microtubules were more stable but also that the ability of condensed chromosomes to congress to the metaphase plate was severely hampered in cells expressing a phospho-null EML4 mutant. This was accompanied by a delay in anaphase onset that was mediated by persistent activity of the SAC. EML4 is nevertheless required for mitotic progression, arguing that, despite the reduced microtubule binding affinity, it still plays an important role at this stage of the cell cycle. EML4 was found to bind the nuclear distribution C (NUDC) protein via its TAPE domain and target NUDC to the spindle promoting kinetochore capture by microtubules (15). Whether this requires the microtubule binding activity of EML4 is not known. Of course, although microtubule stability initially drops upon mitotic entry during the process of search and capture, specific populations of microtubules, notably K-fibers, become highly stabilized upon biorientation of chromosomes (46). Residual staining of spindle microtubules was detected, and it is plausible that EML4 could yet contribute to K-fiber stability if a localized fraction of EML4 was dephosphorylated.

Many proteins specifically interact with microtubules during mitosis. These fulfill a variety of important functions that include the regulation of microtubule stability, the trafficking of checkpoint complexes, and the motility of the condensed chromosomes (47). Collectively, these proteins enable the assembly of a dynamic and finely tuned mitotic spindle. In contrast, EML4 preferentially interacts with microtubules during interphase, and the regulated disruption of this interaction through phosphorylation is what is necessary for proper mitotic spindle assembly and function. This raises the question of how many other proteins might dissociate from microtubules during mitosis. It seems likely that the electrostatic interactions that other MAPs depend on might also be disrupted through phosphorylation, as protein phosphorylation peaks during this phase of the cell cycle. This mode of regulation could therefore be relevant to other proteins, including potentially other members of the EML family, that interact with the negatively charged tubulin C-terminal tails.


Plasmid construction, mutagenesis, and recombinant protein expression

Generation of YFP-EML4-FL and YFP-EML1-NTD constructs was previously described (18), whereas YFP-EML4-NTD and YFP-EML4-TAPE were generated by polymerase chain reaction–based amplification of the NTD and TAPE domain fragments from the YFP-EML4-FL plasmid and subcloning into pLEICS-12 (Protein Expression Laboratory, University of Leicester). Generation of Flag-NEK6-Y108A, Flag-NEK7-Y97A, and Flag-NEK9-ΔRCC1 constructs were as described (31, 32, 34). Mutations were introduced into the YFP-EML4-FL and NTD constructs using the GeneTailor Site-Directed Mutagenesis Kit (Invitrogen), and all constructs were confirmed by Sanger sequencing at the University of Leicester. The YFP-EML1-NTD and YFP-EML4-NTD proteins were expressed and purified as described (18), whereas Flag-Strep-EML4-FL protein used for phosphomapping was expressed in insect cells and purified as described (17).

Cell culture, transfection, and drug treatment

U2OS, HeLa, and human embryonic kidney 293 cells were grown in Dulbecco’s modified Eagle’s medium (DMEM; Invitrogen) supplemented with 10% heat-inactivated fetal bovine serum (FBS), penicillin (100 IU/ml), and streptomycin (100 μg/ml) at 37°C in a 5% CO2 atmosphere. HeLa Kyoto H2B-mCherry/EGFP-Lamin A cells were maintained in DMEM containing 10% FBS, penicillin (100 IU/ml), streptomycin (100 μg/ml), G418 (500 μg/ml), and puromycin (0.5 μg/ml). Transient transfections were performed with Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions. Cells were synchronized in M phase either by incubation for 16 hours with nocodazole (500 ng/ml) or by incubation with 10 μM RO-3306 (Enzo Life Sciences) for 16 hours, followed by transfer into fresh medium with 20 μM MG132 (EMD Millipore) for 2 hours. M-phase arrested cells were collected by mitotic shake-off after 16 hours of treatment. SAC inactivation was achieved through incubation with the MPS1 inhibitor, AZ3146 (Sigma), for 4 hours.

Western blotting

Cells were lysed in ice-cold radioimmunoprecipitation assay (RIPA) or NEK extraction buffer (NEB) lysis buffer [50 mM tris-HCl (pH 8), 150 mM NaCl, 1% (v/v) Nonidet P-40, 0.1% (w/v) SDS, 0.5% (w/v) sodium deoxycholate, 5 mM NaF, 5 mM β-glycerophosphate, ribonuclease (30 μg/ml), deoxyribonuclease I (30 μg/ml), 1× protease inhibitor cocktail, and 1 mM phenylmethylsulfonyl fluoride (PMSF)] and subjected to SDS-PAGE and analysis by Western blotting. Primary antibodies were rabbit NEK6 (1 μg/ml) (32), goat NEK7 (1:250; Aviva Systems Biology), rabbit NEK9 (0.4 μg/ml; Atlas Antibodies), mouse α-tubulin (0.3 μg/ml; Sigma), mouse acetylated tubulin (1:2000; Sigma), rabbit glyceraldehyde-3-phosphate dehydrogenase (1:500; Cell Signaling Technology), rabbit EML4 (N terminus, A301-908A; C terminus, A301-909A; both 1:500; Bethyl Laboratories), rabbit green fluorescent protein (GFP; 0.5 μg/ml; Abcam), mouse pHistone H3 (1:1000; Abcam), and mouse cyclin B1 (0.5 μg/ml; Santa Cruz Biotechnology). Secondary antibodies were horseradish peroxidase–labeled secondary antibodies (1:1000; Amersham).

RNA interference

Cells at 30 to 40% confluency were cultured in Opti-MEM Reduced Serum Medium with 10% heat-inactivated FBS and transfected with 50 nM ON-TARGETplus siRNA duplexes using Oligofectamine (Invitrogen) according to the manufacturer’s instructions. siRNA duplexes were as previously described for NEK6 and NEK7 (34), NEK9 (AM51334-1113, Ambion), and EML4 (HSS120688 and HSS178451, Dharmacon). Cells were fixed or lysed for analysis after 72 hours of transfection.

In vitro kinase assay and mass spectrometry

Kinase assays were carried out using 0.1 μg of purified NEK6 or NEK7 kinase (Millipore). Proteins were incubated with 5 μg of substrate and 1 μCi of [γ-32P]-ATP in 40 μl of kinase buffer [50 mM Hepes-KOH (pH 7.4), 5 mM MnCl2, 5 mM β-glycerophosphate, 5 mM NaF, 4 μM ATP, and 1 mM dithiothreitol (DTT)] at 30°C for 30 min. Reactions were stopped with 50 μl of protein sample buffer and analyzed by SDS-PAGE and autoradiography. Phosphomapping was performed using an LTQ Orbitrap Velos ETD (Thermo Fisher Scientific) as previously described (35).

Fluorescence microscopy

For immunofluorescence microscopy, cells were grown on acid-etched glass coverslips, fixed with ice-cold methanol or methanol:acetone (1:1), and processed as previously described (34). Briefly, media were aspirated, and cells were fixed with ice-cold methanol at −20°C for 30 min. Cells were blocked in 1× phosphate-buffered saline (PBS) supplemented with 3% bovine serum albumin (BSA) and 0.2% Triton X-100 before incubation with antibodies in 1× PBS supplemented with 3% BSA. Primary antibodies used were mouse α-tubulin (0.3 μg/ml; Sigma-Aldrich), rabbit Flag (1:1000; Sigma-Aldrich), rabbit GFP (1 μg/ml; Abcam), rabbit EML4 (N terminus, A301-908A; C terminus, A301-909A; both 1:500; Bethyl Laboratories), mouse acetylated tubulin (1:2000; Sigma), goat Flag (1:1000; Abcam), and rat α-tubulin–conjugated 647 (1:200; Abcam). Secondary antibodies used were Alexa Fluor 488, 594, and 647 donkey anti-rabbit, donkey anti-mouse, and donkey anti-goat immunoglobulins (1 μg/ml; Invitrogen). DNA was stained with Hoechst 33258 (0.8 μg/ml). Fixed and time-lapse imaging were performed on a Leica TCS SP5 confocal laser scanning microscope fitted on a Leica DMI 6000B inverted microscope using a HCX Plan Apo 63× oil objective (numerical aperture, 1.4). For fixed images, Z stacks comprising 30 to 50 0.3-μm sections were acquired using Leica Application Suite Advanced Fluorescence (LAS-AF) software (Leica), and deconvolution of three-dimensional (3D) image stacks was performed using Huygens Essential software (Scientific Volume Imaging). For time-lapse fluorescence imaging, cells were cultured in glass-bottomed culture dishes (MatTek Corporation) and maintained on the stage at 37°C in an atmosphere supplemented with 5% CO2 using an environmental chamber (Life Imaging Services). Z stacks comprising of 0.5-μm sections were acquired every 10 min for a minimum of 16 hours. Stacks were processed into maximum intensity projections using LAS-AF software (Leica), and movies were prepared using ImageJ. To quantify colocalization, ImageJ software was used to draw five lines across the cytoplasmic regions of a cell on a single z section of an image. A total of 10 cells from three independent experiments were used to calculate the mean Pearson’s correlation coefficient (R value).

For SIM, cells were grown on acid-etched high-performance coverslips, fixed in ice-cold methanol, and stained for immunofluorescence microscopy with EML4 and α-tubulin antibodies as described above. However, after incubation with the secondary antibody, cells were subjected to a second, postfixation step in which cells were incubated with 4% paraformaldehyde for 10 min, before coverslips were mounted onto microscope slides with Citifluor mounting medium (Electron Microscopy Sciences). Imaging was performed using a Zeiss PS1 super resolution microscope, and images were processed using Zen software (Zeiss).

For TIRF imaging, microtubules were prepared from a mixture of 2 μM biotin-labeled, 2 μM Hilyte647-labeled, and 20 μM unlabeled tubulin and 1 mM GMPCPP {guanosine-5-[(alpha, beta)-methyleno]triphosphate} in MRB80 [80 mM K-Pipes, 4 mM MgCl2, and 1 mM EGTA (pH 6.9)]. Microtubules were then treated with subtilisin (25 μg/ml) or left untreated and incubated for 20 min at 37°C. To terminate digestion, 2 mM PMSF was added, and the reaction mix was centrifuged for 5 min at 150,000g. The supernatant was removed, and microtubules were resuspended in MRB80 buffer. Microtubules were attached to the surface of a flow cell using PLL–PEG (poly-L-lysine-polyethylene glycol)–biotin and streptavidin, and binding of 100 nM recombinant EML1 (1 to 174) was assessed using TIRF microscopy as described (18).

In vitro microtubule binding and sedimentation assays

Microtubules were assembled from 40 μM tubulin in the presence of 5 mM GTP in MRB80 for 1 hour at 37°C and diluted 1:3 in MRB80 and 2 μM Taxol. Three-fifth of the sample was treated with subtilisin (25 μg/ml; P5380, Sigma), and the remainder was left untreated and incubated at 37°C for 10 min. To terminate digestion, 5 mM PMSF and complete protease inhibitors (Roche) were added, and the reaction mix was loaded onto a 30% sucrose cushion and centrifuged in a TLA55 rotor at 100,000g for 45 min at 30°C. The microtubule pellets were resuspended in BRB25 [25 mM Pipes (pH 6.8), 1 mM MgCl2, 1 mM EGTA, and 2 mM DTT] supplemented with complete protease inhibitors and 2 μM Taxol and incubated with EML4(1-207)-Avi on ice for 15 min before pelleting through a 30% sucrose cushion in a TLA55 rotor at 100,000g for 45 min at 4°C. The pellet was taken up in equal volume to supernatant, and samples were analyzed on 12% SDS-PAGE gels, stained with Instant Blue (Expedeon), imaged on a G:BOX (Syngene), and quantified using ImageJ.

For microtubule sedimentation assays, cells were lysed at room temperature in RIPA lysis buffer, including 5 mM NaF and 5 mM β-glycerophosphate. Samples were loaded on the top layer of 30% sucrose in tubulin stabilization buffer [TSB; 1 mM EGTA, 5 mM MgCl2, and 80 mM Pipes (pH 7.0)] before centrifugation at 100,000g for 40 min at 21°C. Supernatants were collected, and pellets were washed with TSB for 10 min at 21°C. Pellets were diluted into a volume equal to that of the supernatants. Samples were subjected to SDS-PAGE and analysis by Western blot.

Live-cell microtubule stability assays

For measurement of microtubule stability in live cells, U2OS cells were transfected with YFP-EML4-WT or S144/146A for 24 hours before being grown in microwell 8-well chamber slides (Ibidi). Cells were incubated with 25 nM SiR-Tubulin (Cytoskeleton Inc.) for 4 hours, and MG132 was added to the media 30 min before imaging. Fields of view containing transfected cells with bipolar mitotic spindles were selected, and media were supplemented with nocodazole (200 μg/ml) along with MG132 and SiR-Tubulin, and image capture immediately commenced. Z stacks comprising sections of 29 μm by 0.5 μm were captured every 60 s for 15 min using a VisiTech VT-Infinity3 confocal microscope fitted with a Hamamatsu C11449-22CU Flash 4.0 V2 sCMOS camera and Nikon Plan Apo 100× objective (numerical aperture, 1.47). Images were cropped to single cells and deconvolved before analysis for SiR-Tubulin intensity in MATLAB.

Sample preparation for cryo–electron microscopy

Lyophilized tubulin purified from HeLa cells was purchased from Cytoskeleton Inc. (Denver, CO, USA) and reconstituted to 2.5 mg/ml in BRB80 (80 mM Pipes, 1.5 mM MgCl2, 1 mM EGTA, and 1 mM DTT) containing 1 mM GMPCPP at 4°C. After 10 min of incubation at 4°C, the tubulin was transferred to a water bath and polymerized at 37°C for 45 min. To increase GMPCPP occupancy, microtubules were double-cycled by pelleting at 13,000 rpm on a desktop centrifuge at room temperature, removing the supernatant, and resuspending the microtubule pellet to ~2.5 mg/ml at 4°C in BRB80 and 1 mM GMPCPP. Microtubules were then repolymerized by incubation at 37°C for 45 min. Stabilized microtubules were left at room temperature for 3 hours and then diluted in room temperature BRB25 and 1 mM GMPCPP to 0.5 mg/ml before use (25 mM Pipes, 1.5 mM MgCl2, 1 mM EGTA, and 1 mM DTT). Four microliters of microtubules was preincubated on glow-discharged C-flat Holey Carbon electron microscopy grids (Protochips, Morrisville, NC) at room temperature for 1 min, excess buffer was manually blotted away, and then, 4 μl of EML4-NTD (2.2 mg/ml) in BRB25 and 30 mM NaCl and 1 mM GMPCPP was added for 45 s. Excess buffer was again manually blotted away, followed by another 4-μl application of EML4-NTD. In an additional step for cryo–electron tomography only, excess buffer was again blotted away after 45 s of incubation, and 4 μl of 10-nm nanogold fiducial–BSA solution (Sigma), concentrated twofold in BRB25 and 30 mM NaCl and 1 mM GMPCPP, was added to the grid. Grids were then placed in a Vitrobot Mark IV (FEI Co., Hillsboro, OR) at room temperature and 80% humidity, incubated for a further 45 s, and then blotted and vitrified in liquid ethane.

Single-particle cryo–electron microscopy data collection and processing

Low-dose movies were collected automatically using EPU software on a K2 summit direct electron detector (Gatan) installed on a FEI Titan Krios (Astbury BioStructure Laboratory, University of Leeds) operating at 300 kV with a Quantum post-column energy filter (Gatan), operated in zero-loss imaging mode with a 20-eV energy-selecting slit. A defocus range of 0.5 to 2.5 μm and a calibrated final sampling of 1.37 Å/pixel was used with the K2 operating in counting mode at 6 e/pixel per second. The total exposure was 48 e2 over 30 frames (1.6 e2 per frame). Movie frames were aligned using MotionCor2 (48) with a patch size of 5 to generate full-dose and dose-weighted sums. Full-dose sums were used for contrast transfer function (CTF) determination in gCTF (49), and then, dose-weighted sums were used in particle picking, processing, and generation of the final reconstructions. Particle processing was performed with helical methods in RELION v2.1.0 (50), using a custom pipeline and scripts to account for microtubule pseudo-helical symmetry. Briefly, microtubules were boxed manually in RELION with a box separation distance of 82 Å (roughly the microtubule dimer repeat distance). The protofilament number of all segments within each microtubule was assigned according to the modal class of those segments after one iteration of 3D classification to 12-Å low-pass–filtered references of undecorated 11- to 16-protofilament microtubules built from multiple asymmetric units of the GMPCPP microtubule atomic model [Protein Data Bank (PDB) ID: 6DPU (51)]. Once the protofilament number of each microtubule had been established, the main 13-protofilament microtubule class was selected for further processing. Rough alignment parameters of each microtubule to its corresponding 7-Å low-pass–filtered 13-protofilament reference were assigned. On the basis of φ angles determined for each segment, median φ angles were assigned to all segments in a given microtubule. Assigned φ angles for each microtubule were checked by 3D classification against 7-Å low-pass–filtered undecorated 13-protofilament microtubule references rotated and translated to represent all possible seam positions and αβ-tubulin registers. For all references used in this step, pixel values were doubled for atoms within the S9-S10 and H1-S2 loops, being the main distinctive features between α- and β-tubulin. Rough final φ angles were assigned according to the modal 3D class of all segments within each microtubule. Finer local refinement was then performed, with or without applied helical symmetry. Final displayed reconstructions were sharpened to local resolutions as determined in RELION, unless stated. Four times binned data were used for the protofilament number assignment 3D classification, unbinned data for the final local refinements, and two times binned data for all other steps. The atomic models of six dimers of undecorated GMPCPP MTs (PDB: 6DPU) with incorporated Taxol [taken from PDB: 5SYF (52)] were fitted into symmetrized density and used as starting models for iterative rounds of model building in Coot (53) and Phenix (54). Extra density contributed by EML4-NTD had low resolution, presumably due to the flexible nature of the MT-EML4-NTD interaction; therefore, EML4-NTD was not modeled into density.

Cryo–electron tomography data collection and processing

Single-axis cryo–electron tomography of EML4-NTD–decorated microtubules was performed using a Tecnai G2 Polara at 300 kV with a Quantum post-column energy filter (Gatan) operated in zero-loss imaging mode with a 20-eV energy-selecting slit. Data at 5- to 6-μm defocus were collected on a K2 Summit direct electron detector operating in counting mode at 9 e/pixel per second (measured without sample obstructing the beam) with a final sampling of 5.39 Å/pixel. Tilt series of a total dose 114 e2 from −60 to +60° tilt were collected in 3° increments using the Hagen dose-symmetric tilt scheme (55). For each tilt, movies of 9-s exposure at four subframes per second were aligned using MotionCor2 (48). Dose weighting of tilt series was performed using custom scripts calling functions in SumMovie (56). Fiducial-based alignment of tilt series was performed in the Etomo graphical user interface to IMOD (v.4.9.0). CTF determination on each aligned tilt series without dose weighting was performed with CTFFIND4 (57) and 3D CTF correction, and tomogram reconstruction was performed by weighted back-projection of dose-compensated and aligned tilt series with novaCTF (58). Final tomograms displayed were binned four times using IMOD, and a B-factor of 35,000 was applied to amplify low-frequency information.

Statistical analysis

All quantitative data represent means and SD of at least three independent experiments. Statistical analyses were performed using a one-tailed unpaired Student’s t test assuming unequal variance (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; n.s., nonsignificant). For live-cell microtubule stability assays, the area under the curve for each dataset was calculated, and statistical analysis on these values was performed using a one-tailed unpaired Student’s t test assuming unequal variance.


Fig. S1. Mass spectrometry profiles of EML4 phosphorylation sites.

Fig. S2. Activated NEK6, NEK7, and NEK9 reduce the association of EML4 with interphase microtubules.

Fig. S3. Cryo–electron microscopy diagnostics and resolution estimation.

Fig. S4. EML4 depletion leads to reduced microtubule acetylation.


Acknowledgments: We are very grateful to P. Meraldi and his colleagues (University of Geneva) for introducing us to the kinetic SiR-Tubulin microtubule depolymerization assay. We acknowledge support from the University of Leicester Core Biotechnology Services (CBS) for DNA sequencing, construct preparation, and confocal imaging and R. Markus at the School of Life Sciences Imaging Facility at the University of Nottingham for superresolution microscopy. Single-particle cryo–electron microscopy data were collected with R. Thompson at the Astbury BioStructure Laboratory, University of Leeds. Funding: This work was funded by grants from Worldwide Cancer Research (16-0119) and the Wellcome Trust (204801/Z/16/Z) to A.M.F.; the Medical Research Council, U.K. (MR/R000352/1) to C.A.M.; the Cancer Research UK (C24461/A23302) to R.B.; and the Wellcome Trust (200870/Z/16/Z) and Lister Institute of Preventive Medicine to A.S. R.A. was supported by a PhD studentship from the Biotechnology and Biological Sciences Research Council, U.K. The FEI Titan Krios microscopes were funded by the University of Leeds (UoL ABSL award) and Wellcome Trust (108466/Z/15/Z). Cryo–electron tomography data were collected on a Polara microscope funded by grants from the Wellcome Trust (079605/Z/06/Z) and the BBSRC (BB/L014211/1). Author contributions: R.A., J.M.M., J.A., L.O., A.S., R.B., C.A.M., and A.M.F. designed the research; R.A., J.M.M., L.O., and D.R. performed the research; R.A., J.M.M., L.O., and M.W.R. contributed new reagents; R.A., J.M.M., J.A., L.O., K.R.S., D.R., A.S., R.B., and A.M.F. analyzed the data; R.A., J.M.M., J.A., L.O., A.S., R.B., C.A.M., and A.M.F .wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Cryo–electron microscopy structures have been deposited as EMD-0331, PDB ID: 6I2I. All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article