Introduction
HEF1 is a member of a group of scaffolding proteins that includes p130Cas and Efs/Sin1, 2. This group of Cas proteins localize to focal adhesions in interphase cells, and act as intermediates in a variety of integrin-dependent signalling processes, including the establishment of cell attachments, migration and cell survival signalling. In 1998, we proposed that HEF1 might have a previously unsuspected function in mitosis3, based on the observation that the HEF1 protein relocalized from focal adhesions to the mitotic spindle asters in M phase. Since that time, reports have appeared that suggest that other focal adhesion complex proteins, such as zyxin4, paxillin5, FAK and Pyk2 (ref. 6) associate with the mitotic spindle or other relevant structures, such as the microtubule-organizing centre (MTOC) or centrosome. Recent work has emphasized the dual activity of centrosomes in contributing to the control of cell polarization in interphase cell migration7, 8, but also in coordinating assembly of the mitotic spindle in M phase9. Centrosomally associated signalling proteins such as the Aurora-A (AurA) kinase also govern the timing of mitotic entry10, for instance by regulating the activation of cyclin B1 (ref. 11). In this study, we demonstrate a requirement for HEF1 in activation of AurA and Nek2 (ref. 12) — a second kinase important for centrosome cohesion — and we provide additional data indicating that HEF1 provides a novel bridge that coordinates cell attachment and cell division processes in mammalian cells.
Results
Cell-cycle-regulated HEF1 localization
HEF1 localizes to the centrosome of MCF-7 cells in a cell-cycle-regulated manner (Fig. 1a), with the centrosomal signal lowest in G1, and strongest in G2/M cells. This corresponds to relatively low levels of HEF1 detectable in G1 and to the fact that the bulk of HEF1 in interphase cells localizes to focal adhesions3 (Fig. 1b, c). As HEF1 levels increase during G2, a slower migrating, hyperphosphorylated species becomes more apparent, as we have previously reported3. At mitotic entry, a significant fraction of HEF1 moves to the spindle, and HEF1 is no longer detectable at the centrosome at cytokinesis. The endogenous HEF1 localization pattern described here disappeared following HEF1 depletion with siRNA, supporting signal specificity (see Supplementary Information, Fig. S1A, B). Furthermore, a HEF1-specific mouse monoclonal antibody, mAb-14A11, transiently overexpressing GFP-fused HEF1, generated the same pattern (see Supplementary Information, Fig. S1C–E). Finally, the HEF1 signal in the vicinity of the centrosome was coincident with the patterns seen for multiple centrosome-associated proteins, including 
-tubulin, c-Nap-1, pericentrin, ninein and Nek2 (see Supplementary Information, Fig. S1F). Using GFP- and FLAG epitope-fused HEF1 derivatives, we mapped the minimal sequence determinants that are necessary for localization of HEF1 to the centrosome as being HEF1 amino acids 1–405 (Fig. 1d, e). This sequence contains the SH3 domain and SH2-binding site-rich domains2, with sequences between 363–405 being an essential localization determinant. In addition to this fragment, carboxy-terminal derivatives of HEF1 (aa 351–653 and 654–834) also showed weaker association with the centrosome, suggesting that more than one interaction contributes to the localization, analogous to the discrete focal adhesion targeting elements within HEF1 (ref. 13).
Figure 1: HEF1 localization to the centrosome: cell-cycle and sequence dependence.
(a) Cell-cycle-synchronized populations were analysed for HEF1 localization to the centrosome and mitotic apparatus. HEF1 is indicated in green, and visualized with anti-HEF1-SB-R1 antibodies, as previously described3. -tubulin (red) is used to indicate centrosomes (arrows). DNA (blue) becomes condensed at mitotic entry. Enlarged views of boxed centrosomes are shown in the bottom right corners; arrows mark centrosomal location. Images are merged confocal sections. Scale bar of 8 
m applies to top row; 6 m scale bar applies to all other images. (b) FACS analysis demonstrating asynchronous (Asy.) MCF-7 cells, and MCF-7 cells synchronized in G1, S and G2/M phases, as used for immunofluorescence analysis. (c) Western analysis of HEF1 levels in the indicated phases of the cell cycle, with 
-actin as loading control. The HEF1 doublet represents two phosphorylation-induced isoforms of HEF1 with relative molecular masses of 105,000 (Mr 105K) and 115,000 (Mr 115K) (hyperphosphorylated)10. The broad band migrating at 
95K is a non-specific, cross-reacting species detected with the rabbit polyclonal antibody, described in ref. 10 (also see Supplementary Information, Fig. S1A). (d) Cells transfected with plasmids expressing GFP–HEF11–363 or GFP–HEF11–405, and stained with antibody to -tubulin (red). White lines in lower panels outline cell peripheries. Boxed centrosomes are enlarged in insets; arrows indicate other examples. (e) Fragments of HEF1 analysed as GFP- or FLAG epitope-tagged fusion proteins, amino acids and the degree of localization to the centrosome are shown. The degree of centrosomal localization was estimated by measuring the signal intensity at the centrosome in the same set of experiments (data not shown). HEF11–834 is full-length HEF1. Asterisk, protein poorly expressed.
Up- or downregulation of HEF1 expression causes centrosomal and spindle abnormalities
To determine the functional significance of HEF1 association with the centrosome and the mitotic spindle, we established three independent systems for manipulation of HEF1 in MCF-7 cells. First, we overexpressed the full-length HEF1 protein under control of a tetracycline-repressible promoter (Fig. 2a). Second, based on previous data indicating HEF1 cleavage and proteolysis at mitosis and apoptosis (discussed further below), we stabilized full-length endogenous HEF1 using peptide aptamers targeted to a previously mapped HEF1 cleavage site13, 14 (Fig. 2b). Third, we used an siRNA approach to deplete endogenous HEF1, using two independent siRNAs (siHEF1 and siHEF1a; Fig. 2c and see Supplementary Information, Fig. S2A). For each manipulation, we used a matching negative control set (corresponding to tetracycline-regulated GFP, for overexpression), non-specific peptides, scrambled siRNA, a GFP-targeted siRNA and, in some cases, a p130Cas-targeted siRNA for depletion (Fig. 2c and see Supplementary Information, Fig. S2A).
Figure 2: Overexpression, stabilization and depletion of HEF1.
(a) MCF-7 cells with tetracycline-repressed expression (MCF-7-tTA) of stably integrated HEF1 or GFP in the presence (+) or absence (-) of tetracycline, measured at 24 or 48 h after medium change. Western analysis with antibody to HEF1 demonstrates induction following tetracycline removal. -actin was used as a loading control. (b) Western blot analysis of MCF-7 cells infected by retroviruses expressing HA/thioredoxin-tagged peptide fusion proteins (HA–TRX). Levels of HEF1 are stabilized by specific HA–TRX peptides (P1-HEF1, P2-HEF1), but not by non-specific HA–TRX peptides (P1-NS), or by HA–TRX with no peptide inserted (NP). Antibody to HA shows comparable expression of HA–TRX fusions in all lanes. (c) Western blot analysis of MCF-7 cells treated with siRNAs to HEF1 (siHEF1 and siHEF1a) or a scrambled (Scr) or GFP (siGFP) oligonucleotide duplex shows efficient and specific HEF1 depletion at 48 and 72 h time points. The blot was stripped and re-probed with -actin as a loading control. All lanes shown were run on a single gel; white lines here and in the following figures indicate excision of intervening bands.
Both overexpression of exogenous HEF1 and peptide stabilization of endogenous HEF1 induced a high frequency of cells with spindle defects by 48 h following cell treatment (Fig. 3a, b and Supplementary Fig. S3A, B). Most notable was the increase in the number of cells with multipolar spindles, which represents 12% of the mitotic cell population with overexpression, and >16% with specific HEF1-targeted peptides, as compared with 2–3% for all negative controls. For every cell with multipolar spindles, each spindle originated from a -tubulin-, pericentrin- or GFP–centrin-positive structure (Fig. 3a; see Supplementary Information, Fig. S3A; and data not shown). In other cells, whereas no multipolar spindles were observed, the spindle was nevertheless defective. A common phenotype was the presence of 'bent' spindles (Fig. 3a), which would arise if the spindle poles had not fully moved to opposite sides of the cell. Overexpression of HEF1, or introduction of HEF1-stabilizing peptides, was also found to consistently induce supernumerary centrosomes, with >10% of all cells containing in excess of three centrosomes. (Fig. 3a, c and see Supplementary Information, Fig. S3A, B). Abnormally increased numbers of centrosomes can arise from either deregulation of the centrosomal duplication cycle during S phase, or as a result of failed cytokinesis15. We found that the increased numbers of centrosomes accumulated gradually over 24–48 h following HEF1 induction (Fig. 3c), and were due to secondary defects in cytokinesis, because centrosomes did not accumulate in HEF1-overexpressing cells held in hydroxyurea (Fig. 3d).
Figure 3: Overexpression and stabilization of HEF1 induce supernumerary centrosomes and multipolar mitotic spindles.
(a) MCF-7 cells with tetracycline-repressed HEF1 (tTA-HEF1) were un-induced (top row) or induced by tetracycline (Tet) removal (bottom 3 rows). Cells were treated to visualize DNA (blue) 
-tubulin (green) and -tubulin (red) for immunofluorescence; representative mitoses are shown. Scale bar, 4 m. (b) Quantification of multipolar spindles scored in cells with induced (-Tet) or un-induced (+Tet) HEF1 or GFP expression, or treatment with peptides stabilizing endogenous HEF1: specific for HEF1 (H), non-specific (NS), or peptide negative control (NP) HA–TRX fusions. The histogram indicates the percentage of cells with multipolar spindles under different conditions. Red bars indicate conditions with increased HEF1 levels. Three independent experiments were performed, resulting in the assessment of 150 mitoses in total for each condition. ***P < 0.001 versus negative control condition. (c) tTA-HEF1 cells were first synchronized using a double thymidine block, then released into medium with tetracycline (blue; HEF1 repressed) or without tetracycline (red; HEF1 induced), and the number of centrosomes scored at 0, 24, or 48 h. The percentage of cells with 
3 centrosomes was recorded. In three experiments 150 cells were counted. ***P < 0.001 difference, minus versus plus tetracycline for each time point; difference at 0 h is not significant. (d) tTA-HEF1 cells were presynchronized in thymidine, then placed in medium with hydroxyurea (HU), and with or without tetracycline, and centrosomes were scored up to 72 h. In three experiments 150 cells were counted. No significant differences were seen. Error bars indicate standard error of the mean (s.e.m.).
In contrast to the results with overexpressed HEF1, depletion of HEF1 induced a high percentage of cells with mono-astral spindles, containing two -tubulin-positive structures (Fig. 4a; centre row), or with poorly formed spindles (Fig. 4a; bottom row). The centrosomes of cells with defective spindles showed weaker reactivity with antibody to -tubulin than did cells with undepleted HEF1 (Fig. 4a). Furthermore, the spindles were consistently less reactive with -tubulin antibody, particularly in cells with less well-formed spindles (see Supplementary Information, Figs S1B and S2B). Parallel staining that was done with antibody to HEF1 indicated that HEF1 itself was effectively depleted in individual cells with noticeable phenotypes (see Supplementary Information, Fig. S2B, bottom row); the most pronounced phenotypes were observed in cells with the greatest degree of HEF1 depletion. HEF1-depleted non-mitotic cells also manifested centrosomal abnormalities. Two distinct, widely separated, GFP–centrin-positive structures (Fig. 4b and see Supplementary Information, Fig. S2B) were observed in >70% of HEF1-depleted cells but in 25–30% of cells treated with a scrambled siRNA (Fig. 4c). Normally, two widely separated centrosomes are not observed until the G2/M transition15, 16. This increase in the frequency of split centrosomes was observed with both HEF1-directed siRNAs, but with neither of the two non-specific siRNAs; a weaker effect was seen with a p130Cas-directed siRNA (Fig. 4c). Fluorescence-activated cell sorting (FACS) analysis of HEF1-depleted cells versus scrambled siRNA-treated controls confirmed that their cell-cycle profile does not show G2 enrichment (Fig. 4d). Hence, the primary defect with HEF1 depletion is probably one of centrosome cohesion resulting in premature splitting, rather than a secondary consequence of altered cell-cycle compartmentalization.
Figure 4: Depletion of HEF1 induces centrosomal splitting and a mono-astral mitotic spindle.
(a) MCF-7 cells were transfected with either a control scrambled (Scr; top panels) siRNA, or an siRNA specific for HEF1 (siHEF1; middle and bottom panels) for 48 h, then processed for immunofluorescence using markers to -tubulin (red), -tubulin (green) and DNA (blue). Shown are representative mitoses. Insets show enlarged views. Scale bar, 5 m. (b) Immunofluorescence analysis of MCF-7 cells with stably expressed GFP–centrin2 (green) 48 h post-transfection of scrambled (Scr) or HEF1-directed (siHEF1) siRNAs was used to calculate the frequency of split (top panels) versus closely paired (bottom panels; Scr) centrioles (indicated by arrows, with boxed examples shown in insets). The centrioles were considered as split if the distance between them was 2 m or more12. Results shown are for depletion with siHEF1; comparable phenotypes are seen with siHEF1a. DNA is shown in blue. Scale bar, 5 m. (c) For quantification, >150 cells were counted from each of at least three separate experiments for HEF1-depleted (siHEF1) or control (Scr) siRNA-treated cells, and two experiments for sip130 and siGFP. All counting was done at a time point after 48 h of siRNA treatment. ***P < 0.001 for siHEF1 versus either Scr or siGFP; */**P < 0.01 versus Scr and P < 0.05 versus siGFP. Error bars indicate s.e.m. (d) FACS analysis of cells depleted with indicated siRNAs at 48 h post-transfection.
HEF1 interacts with AurA and is required for the activation of AurA kinase
The phenotypes described above for HEF1 depletion and overexpression are similar to those reported for inhibition or overexpression of other proteins known to regulate centrosomal maturation and cell-cycle progression, including the AurA kinase17, 18. Indeed, whereas in HEF1-depleted MCF-7 cells in the G2 phase of the cell cycle total levels of AurA at the centrosome were similar to those found in MCF-7 cells treated with a matched scrambled siRNA (Fig. 5a, b), levels of phospho-AurA (T288, indicative of kinase activation11) were, in contrast, greatly reduced or absent under conditions of HEF1 depletion (Fig. 5a, b). This implied that HEF1 has an important role in the activation of AurA. This role could be direct, with HEF1 being a physical component of an AurA activation complex, or it might be indirect, with HEF1 causing defects at an early stage of the centrosomal cycle that interfere with AurA activation. AurA and HEF1 co-precipitated from whole-cell lysates, with the greatest levels of association being in cells in G2 (Fig. 5c). This suggests that HEF1 control of AurA activation might be direct, through a physical interaction with AurA itself or with an AurA-containing complex.
Figure 5: HEF1 associates with AurA and controls AurA activation.
(a) MCF-7 cells were depleted with HEF1-directed siRNA (siHEF1) or a scrambled control (Scr). Mitotic cells stained with antibodies directed at AurA (green), phospho-AurA/T288 (P-AurA; red) and DNA (blue) are shown. Scale bar (5 m) applies to all panels. (b) MCF-7 cells were transfected with scrambled control (Scr) or HEF1-directed (siHEF1) siRNA for 48 h, then collected and part of the sample elutriated. Non-elutriated (Asyn.), G1/S, or G2/M enriched populations were used for western blot analysis using the antibodies indicated (left). White vertical lines indicate where intervening (irrelevant) lanes run on the gel have been excised from the image. (c) MCF-7 cells prepared as in b were used for co-immunoprecipitation. Top: immunoprecipitation with control IgG or Cas antibody (cross-reactive with p130Cas and HEF1; efficient direct precipitation of both proteins), and western blot using antibodies as indicated on the left. Bottom: immunoprecipitation with control IgG or AurA antibody, and western blot as indicated. For each experiment, co-immunoprecipitation is compared in cells depleted for HEF1, or treated with scrambled siRNA, to confirm the specific requirement of HEF1 in co-immunoprecipitation.
Using a series of GST-fused truncations of HEF1, we determined that GST–HEF11–363 was the minimal sequence required to efficiently pull down AurA in vitro (Fig. 6a). Our earlier results (Fig. 5a, b) predicted that HEF1 interaction with AurA might help to activate the kinase. Full-length HEF1 is not stable as a recombinant purified protein, prohibiting a direct in vitro test for this idea with the native protein (data not shown). However, as an alternative approach, we titrated the GST–HEF11–363 minimal AurA-interacting domain versus GST into an in vitro kinase reaction containing recombinant AurA purified from bacteria (Fig. 6b). Increasing levels of GST–HEF11–363, but not of GST, clearly induced the autophosphorylation of AurA and the ability of AurA to phosphorylate a histone H3 substrate, indicating that the association with HEF1 is sufficient to promote AurA catalytic activity11. Indeed, a higher level of activated AurA was observed in cells overexpressing HEF1, in contrast to the lower levels of AurA activation seen with HEF1 depletion (Fig. 6c). Furthermore, we found that both GST–HEF11–363 and HEF11–405 were phosphorylated by recombinant activated AurA in vitro (Fig. 6d).
Figure 6: Delineation of the HEF1-AurA interaction.
(a) In vitro translated 35S-labelled AurA was used for pulldowns with GST-fused fragments of HEF1 or GST. Autoradiography shows AurA (top panel), and Coomassie blue (CB) shows GST fusions (bottom panel). In lane 'R', 20% of a total AurA 35S-labelled reaction mixture is shown; for each pulldown, a complete reaction mixture is used. (b) Decreasing quantities (4, 2, or 1 g) of GST-fused HEF11–363 or GST were used in an in vitro kinase reaction with recombinant AurA. Reactions were visualized with antibody to AurA or phospho-AurA/T288, or with Coomassie blue to show GST fusions. Phosphorylated histone H3 was visualized by autoradiography. (c) AurA was immunoprecipitated from MCF-7 cells that were tetracycline repressible for vector (tTA) or HEF1 (tTA-HEF1) expression, + or - tet; and additionally, from cells treated with siRNA to HEF1 (siHEF1) or to scrambled control (Scr). The immunoprecipitated AurA was incubated with recombinant histone H3 with -32P-ATP in an in vitro kinase assay, resolved by SDS–PAGE, and visualized by antibody to AurA, Coomassie blue (CB) staining to detect H3, and autoradiography. (d) GST, GST-fusion proteins, or histone H1 or H3 (CB; left) were incubated with AurA and -ATP-32P in vitro. Right: autoradiograph of phosphorylated proteins. (e) GST–HEF11–363 wild type (wt), or with alanine or glutamic acid mutations (S296A (A), S296A/S298A (AA), S296E (E), S296E/S298E (EE)) were incubated with AurA and -ATP-32P in vitro. Top: CB shows GST–HEF1 fusions. Bottom: autoradiograph of phosphorylated GST-fusions. (f) The GST-fused derivatives of HEF1 described in e were mixed with recombinant AurA and histone H3 (H3) in the presence of -ATP-32P, incubated, and the level of histone-H3 phosphorylation was determined by autoradiography. Top, CB shows GST fusions; second row, input histone H3; third row, phospho-histone H3. After incubation, immunoprecipitated AurA (IP:AurA) was probed with antibody to AurA and GST. (g) GFP-fused full-length HEF1 (wt or with mutations described in e) were transfected into MCF-7 cells. After 24 h, sequential western blot analysis of immunoprecipitated AurA was done using anti-AurA and anti-GFP (top and middle panels). Bottom: anti-GFP shows expression of GFP fusions in total lysate; bands at Mr 70–80K are degradation products.
Although the AurA consensus site remains poorly defined, an RHQSer 296LSP motif closely resembles a site phosphorylated by the Aurora yeast orthologue Ipl1p (ref. 19). We used mass spectrometry analysis of in vitro phosphorylated GST–HEF11–363 to confirm in vitro phosphorylation of this site (data not shown), and mutated Ser 296 into alanine (unphosphorylatable) or glutamic acid (mimicking constitutive phosphorylated HEF1) alone or together with an adjacent serine, Ser 298. All Ser 296 mutants were no longer phosphorylated by AurA (Fig. 6e). However, whereas alanine mutants of GST–HEF11–363 maintained the ability to interact with AurA and activate the kinase, glutamic acid mutants of HEF1 lost both abilities (Fig. 6f). In parallel, we compared the relative interaction of HEF1 with AurA in the presence or absence of ATP in vitro (see Supplementary Information, Fig. S4A). We found that AurA co-immunoprecipitated much more efficiently with unphosphorylated HEF1, supporting the results obtained with the mutants.
In vivo, we then compared the ability of GFP-fused HEF1, HEF1S296E, HEF1S296E/S298E and HEF1S296A/S298A to immunoprecipitate AurA (Fig. 6g). Whereas HEF1S296A/S298A was similar to HEF1 in interacting with AurA, both phosphomimic variants were severely impaired for AurA interaction, similar to the in vitro results. Together, these data suggest a model in which an initial interaction of HEF1 with AurA prior to mitotic entry activates AurA, which then phosphorylates HEF1, promoting dissociation of the two proteins. Amino acids 1–405 are a minimum determinant of strong HEF1 association with the centrosome in vivo (Fig. 1d, e), with a critical localization determinant located in the serine-rich amino acids from 363–405. We transfected GFP-fused truncation derivatives of HEF1 into the MCF-7 cells, immunoprecipitated with antibody to GFP, and confirmed that GFP–HEF11–405 co-immunoprecipitated with AurA from whole-cell lysates, whereas GFP–HEF11–363 did not (see Supplementary Information, Fig. S4B). We next compared the activation of AurA that had been immunoprecipitated from cells expressing HEF1, HEF11–363 and HEF11–405 (see Supplementary Information, Fig. S4C). GFP–HEF11–405 was like GFP–HEF1 in that it promoted increased activity of AurA against a histone H3 substrate, whereas GFP–HEF11–363 did not. Together, these results imply that a primary role of aa 363–405 in promoting the HEF1–AurA interaction in vivo is through localizing HEF1 to the centrosome, where endogenous AurA is concentrated (Fig. 5a). We were unable to test whether GFP–HEF11–405 was similar to GFP–HEF1 in inducing multipolar spindles, because in a survey of hundreds of cells, no mitotic cells overexpressing GFP–HEF11–405 were ever observed, suggesting that this truncation may be disrupting HEF1-dependent processes prior to mitotic entry.
HEF1 negatively regulates Nek2 and contributes to accumulation of pericentriolar material (PCM)
In normal cells, after telophase, centrosomes mature through the cell cycle, accumulating PCM that includes signalling proteins that govern centrosomal duplication and other functions such as microtubule nucleation15. Cohesion of the centrosomes is maintained by c-Nap-1: levels of c-Nap-1 are reduced 10-fold at the G2/M boundary, with phosphorylation by the Nek2 kinase, and potentially AurA, Cdk1 and Plk1 promoting its removal from the PCM and centrosomal disjunction, allowing formation of a bipolar mitotic spindle12, 20, 21. Moreover, overexpression of Nek2 induces precocious centrosomal disjunction in interphase12, whereas AurA depletion does not (see Supplementary Information, Fig. S3C–E). We examined the status of Nek2 and c-Nap-1 (Fig. 7a), and the centrosomal maturation marker ninein (see Supplementary Information, Fig. S4D) at the centrosome in MCF-7/GFP–centrin cells depleted of HEF1. HEF1 depletion reduced the signal intensity of all of these proteins at the centrosome, suggesting a contribution of HEF1 to the stable assembly of proteins with the PCM. We next asked whether HEF1 regulates Nek2 activation. Nek2 immunoprecipitated from HEF1-depleted cells was much more active in phosphorylating an MBP substrate than Nek2 from control-depleted cells (Fig. 7b). Moreover, overexpression of GFP–HEF1 decreased, whereas overexpression of GFP–HEF11–405 increased activation of Nek2 relative to background levels in cells expressing GFP. As in the other assays, GFP–HEF11–363 exhibited no activity. Finally, we found that endogenous HEF1 and Nek2 co-immunoprecipitated efficiently and specifically from MCF-7 cells (Fig. 7c). Together, these data imply that HEF1 contributes to Nek2 inhibition during the normal cell cycle.
Figure 7: HEF1 depletion affects Nek2 activation and association of proteins with the PCM.
(a) MCF-7 cells with integrated GFP–centrin were treated with scrambled control siRNA (Scr), or siRNA to HEF1 (siHEF1), and stained for immunofluorescence with antibodies to HEF1, c-Nap-1 or Nek2, as indicated. Scale bars, 10 m. (b) MCF-7 cells were either treated with siRNAs (left panel), or infected with retrovirus expressing full-length HEF1 or HEF1 truncations, as for Supplementary Information, Fig. S4C. Nek2 kinase was immunoprecipitated and used for in vitro kinase assays with myelin basic protein (MBP) as a substrate. A comparable level of HEF1 or AurA protein depletion (see Supplementary Information, Fig. S3C), and similar levels of overexpression of HEF1 and truncations (data not shown) was confirmed with antibodies to GFP and/or HEF1and AurA, in whole-cell lysates. (c) Antibody to Nek2 or to Cdc2 was used for immunoprecipitation from asynchronous MCF-7 cells. Immunoprecipitates were probed with antibodies to HEF1, Nek2, or Cdc2, as indicated.
Separation of HEF1 actions at the centrosome and in attachment
The main functions previously ascribed to HEF1 and the Cas family of proteins have been involved in regulation of cell attachment and motility1, 2. We had previously shown that overexpression of HEF1 induced cell spreading13, 22, whereas overexpression of the HEF1 C terminus had a dominant-negative function on cell attachment, causing cell rounding13. To evaluate whether HEF1 control of cell attachment, and regulation of centrosomal splitting, were linked or separable, GFP-fused HEF1 and HEF1 truncations were transfected into MCF-7 and HeLa cells, and the ability of different HEF1 domains to act as dominant negatives by inducing centrosomal splitting versus inhibiting cell attachment was scored (Fig. 8a–c). On the basis of this analysis, full-length HEF1 weakly induced centrosomal splitting but induced spreading. However, the HEF11–405 domain proved to have a potent phenotype, and the HEF1351–653 fragment a weaker phenotype, in inducing centrosomal splitting (Fig. 8a), whereas neither affected the degree of cell spreading (Fig. 8b). Conversely, HEF1654–834 induced significant cell rounding (Fig. 8b), as reported before, but did not affect centrosomal splitting (Fig. 8a). These results indicate that separable HEF1 domains were required for centrosomal and attachment activities.
Figure 8: HEF1 activities at the centrosome and in cell spreading.
For histograms, ***P < 0.001; **P < 0.01; and *P < 0.05, in reference to the GFP control. Error bars indicate s.e.m. (a) Centrosomal splitting in GFP-positive cells transfected with HEF1 truncations. Top, MCF-7 cells; bottom, HeLa cells. Bars represent the percentage of split centrosomes. Three (MCF-7) and two (Hela) independent experiments were performed (n = 100 cells per experiment). FL, full-length HEF1. (b) Cell spreading in GFP-positive cells transfected with GFP-fused HEF1 truncations. Top, MCF-7 cells; bottom, HeLa cells. One hundred cells were counted in each of three experiments. Cell spreading analysis was performed as described previously13, with calculation of area based on pixels within a traced cell perimeter. Immunofluorescence with -paxillin was used to confirm increased size and formation of focal adhesions in cells plated on laminin and fibronectin. (c) Western analysis of expression of different HEF1 truncations expressed as GFP-fused proteins in MCF-7 cells. Blot was probed with antibody to GFP. Lanes are GFP fusions to: 1, 654–834; 2, 351–653; 3, 1–363; 4, FL; 5, 1–405. (d) Mitotic spindle defects induced by HEF1 overexpression in cells plated on different matrices. Tet-repressible MCF-7-derived cells were grown in the presence (+) or absence (-) of tet 48 h after plating on glass coverslips (-), or coverslips coated with poly-L-lysine (PLL), fibronectin (FN), or laminin (LAM). Bars represent the percentage of multipolar or malformed spindles. One hundred mitotic cells were counted in each of three independent experiments. (e, f) Centrosomal splitting induced by depletion of HEF1 (siHEF1), or overexpression of dominant-negative HEF1 (GFP-fused HEF11–405), in cells plated on different matrices. e, MCF-7 cells with integrated GFP–centrin2 were transfected with siRNA containing scrambled sequence (Scr) or targeted to HEF1 (siHEF1). f, MCF-7 cells were transfected with GFP-fused HEF11–363 or HEF11–405, and centrosomes were visualized by antibody to -tubulin. Following treatment, cells were plated on glass coverslips (-), or coverslips coated with poly-L-lysine (PLL), fibronectin (FN), or laminin (LAM). Bars represent the percentage of split centrosomes. One hundred cells were counted in each of three experiments. For d–f, the differences induced by matrix were not statistically significant, except *P < 0.05 for PLL in d.
In a further test, we determined the consequences of varying the degree of cellular attachment induced by extrinsic stimuli on HEF1-associated centrosomal phenotypes. Cells with induced overexpressed full-length HEF1 (Fig. 8d), depleted HEF1 (Fig. 8e), or overexpressed dominant-negative HEF11–405 (Fig. 8f) were plated on either normal tissue substrates, or on poly-L-lysine, fibronectin, or laminin to increase spreading. Cells plated on the last three substrates were significantly more spread, and were marked by more pronounced paxillin staining at focal adhesion structures (data not shown). However, in scoring the number of supernumerary centrosomes induced by overexpressed HEF1 (Fig. 8d), or the number of split centrosomes induced by removal or dominant-negative blockade of HEF1 function (Fig. 8e, f), the greater attachment status did not affect the observed phenotypes. Together with the earlier results, these findings demonstrate that HEF1-induced cell spreading and enhancement of focal adhesions do not cause the centrosome abnormalities that we have described here; rather, a distinct function of HEF1 is involved.
Discussion
The results of this study for the first time establish the focal adhesion protein HEF1 as a regulator of AurA and Nek2 activation, and of centrosome cohesion and amplification. Proteins that were initially defined as components of the cell attachment machinery, including APC23 and Ajuba11, have recently been found to also function in cell-cycle controls, and elegant genetic studies in lower eukaryotic models for development such as Caenorhabditis elegans (reviewed in refs 24, 25) have begun to allow a model to be elucidated, in which dynamic interconnections between the centrosome and structures at the cell cortex control the plane of mitotic spindle orientation, and cleavage furrow formation. It is likely that this mechanism will prove to be important for higher eukaryotes as well, given the need of many cells to limit cell division to specific planes (for example, to maintain barrier function). An economical view of cellular function would suggest that the re-use of proteins that govern cell attachment and cytoskeletal dynamics in interphase cells might not only be efficient, but might also provide a means of synchronizing changes in cell contacts during the mitotic process. It is possible that the pools of HEF1 used for centrosome, mitotic spindle and focal adhesions are completely distinct. However, it may be that migration of proteins such as HEF1 between these structures provides polarity and attachment cues that influence the entry to, and exit from, mitosis. Given the particular abundance of HEF1 in polarized epithelial and lymphoid cell populations, our work would define it as an excellent candidate for such a role.
Our data suggest a model in which in normal cells, HEF1 initially interacts with AurA in G2 prior to AurA activation, with the centrosome being one important point of interaction. In this model, as mitosis approaches, focal adhesion disassembly releases more HEF1, and the increasing interaction of HEF1 with AurA promotes AurA activation. In turn, phosphorylation of HEF1 by the activated AurA reduces the affinity of interaction between the two proteins, perhaps contributing to the relocation of HEF1 away from the centrosome, or perhaps contributing to the preferred interaction of AurA with other partner proteins in the context of the centrosome. AurA activity is known to be regulated by several other protein partners, including TPX2, Ajuba and PP2. In cells depleted of HEF1, AurA does not become activated, suggesting that the association with HEF1 is functionally important. In cells that have HEF1 overexpressed, and are able to associate with the centrosome, the stoichiometry of HEF1 is significantly increased, allowing the protein to continue to interact with AurA in spite of phosphorylation by AurA, and thus promoting elevated AurA activity. For both AurA and HEF1, the centrosomal amplification and multipolar spindles seen with overexpression of the proteins is a secondary consequence of cytokinetic failure. The exact mechanism is yet to be defined, but may involve regulation of a common effector by these proteins.
HEF1-depleted cells have abnormally split centrosomes, which accumulate reduced levels of -tubulin in G2 (refs 15, 16, 26), have abnormally reduced accumulation of ninein, c-Nap-1 and other proteins, and are deficient in organizing microtubules at mitosis. These defects are likely to be independent of HEF1–AurA signalling, because AurA depletion does not result in centrosomal splitting (see Supplementary Information, Fig. S3C–E), and may be direct (at the centrosome) or indirect. Our data suggest that HEF1 may normally act to restrain the activity of Nek2 (refs 12, 20), because HEF1 co-immunoprecipitates with Nek2. Nek2 is hyperactivated in cells with depleted HEF1 or dominant-negative HEF11–405 (Fig. 7b, c), with increased Nek2 activation previously reported as being sufficient to induce splitting. HEF1 may have additional activities required for centrosomal cohesion, as hyperactivation of Nek2 is not sufficient to completely remove c-Nap-1 from centrosomes in interphase cells12, whereas accumulation of ninein is an early step in the maturation of daughter centrosomes to mother centrosomes, and has not been described as being influenced by Nek2. Such a role for HEF1 is separable from any secondary effect due to defects in cell attachment, as different domains of HEF1 caused splitting versus cell rounding. Intriguingly, recent protein interaction studies of the ancestral HEF1/p130Cas homologue in Drosophila27 have suggested that this protein associates with a component of the -tubulin ring complex (-TuRC), which is important for microtubule nucleation28. It is also interesting that we have previously found that the cleavage of HEF1 at amino acid 363 by caspases produces a p55 species3, 13. Although we originally found this p55 species in both mitotic and apoptotic cells, our ongoing work has suggested that the initial suggestion of HEF1 cleavage at mitosis may have arisen at least in part from contamination of drug-synchronized mitotic populations with apoptotic cells (data not shown). However, the fact that HEF11–353 does not associate with centrosomes or with AurA in cells, whereas the slightly larger HEF11–405 does, implies that cleavage of HEF1 at this site may have a functional significance in disrupting centrosomal function during cell death.
Defining HEF1 as a component of the AurA activation machinery is an important finding, providing evidence of a new channel for cross-signalling between cell adhesion and mitosis. Furthermore, AurA and Nek2 overexpression and hyperactivation have been observed in many tumours, and are associated with genomic instability29, 30, 31. It was shown for Cas proteins32 that upregulation of p130Cas and/or HEF1 correlates with poor prognosis in breast cancer. It has additionally been shown by us and others that upregulation of Cas proteins influences the transcription of a number of gene pathways associated with cancer development, enhancing activation of the MAPK pathway and promoting matrix metalloproteinase production (refs 22, 33 and others). It is entirely possible that altered Cas protein levels also induce transcriptional changes that influence genomic stability. Our data imply that beyond their well-defined functions in regulating susceptibility to apoptosis and cell migration, HEF1 and potentially other members of the Cas family may make additional contributions to the processes of cell transformation through the regulation of mitosis.
Methods
Plasmids and constructs.
FLAG-, GFP- and GST-fused HEF1 and derivatives were expressed from the vectors pCatch-FLAG, pEGFP-C4 (ref. 13) and pGST, respectively. AurA expressed from pCMV-SPORT6-C6 (OpenBiosystems, Huntsville, AL) was used for in vitro translation. AurA in pFAST-HT was used for production in baculovirus, and purified by Ni-Sepharose 6FF (Amersham, Piscataway, NJ). HEF1-specific and non-specific peptides14 were expressed from the retroviral vector pUP. Tet-repressible HEF1 in MCF-7 was made using HEF1 in the expression vector pUST-4. Lentiviral constructs were obtained by cloning GFP–HEF1, -HEF11–363, or -HEF11–405 into pLV-CMV-H4-puro-vector. HEF1 mutants S296A, S296E, S296A/S298A, and S296E/S298E were made using a QuikChange XL Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA). Primer sequences are available on request.
Cell culture.
MCF-7 (breast adenocarcinoma) and HeLa (cervical carcinoma) cell lines were grown in DMEM plus 10% fetal calf serum. The MCF-7-GFP-centrin2 and HeLa-GFP-centrin2 stable cell lines were obtained by transfection of parental cells with pEGFP-centrin2 plasmid34, and G418 selection. Tetracycline-regulated MCF-7-tTa-HEF1 stable cell lines were obtained by first infecting the parental MCF-7-tTa cell line (BD Biosciences, San Jose, CA) with the pUST-4-HEF1 retroviral vector, then selecting with G418/puromycin to produce a mass culture. For analysis of centrosome number in cells that had not undergone mitosis, MCF-tTa-HEF1 cells were plated for 24–72 h +/- tet, with 1 mM hydroxyurea (Sigma, St Louis, MO). Alternatively, after using a double thymidine block in the presence of tetracycline, cells were released and grown for 24 and 48 h in fresh media +/- tet. For growth of cells on poly-L-lysine, laminin, or fibronectin, the procedure described in ref. 13 were used to prepare coverslips. Cells for analysis were plated on these versus uncoated (normal) coverslips, grown for 48 h, then centrosomal composition and spreading were scored. Examination of centrosomes in non-adherent cells plated on poly-HEMA13 was impossible because onset of apoptosis within 24 h precluded reliable analysis (data not shown). For lentiviral infection, pLV constructs were transfected into the packaging cell line 293-T. After 24 h, media was collected, filtered through a 0.45-m PVDF filter (Millipore, Billerica, MA), and applied to MCF-7 cells with polybrene for 2 days, with fresh viral supernatant added every 12 h. After 48 h, cells were lysed, analysed by western blot analysis, and used for immunoprecipitation kinase reaction.
Protein expression, western blotting and immunoprecipitation.
Recombinant proteins were expressed in BL21 (DE3) bacteria, induced with IPTG, and purified using the MicroSpin GST Purification module (Amersham, Piscataway, NJ). Purified recombinant AurA was purchased from Upstate (Charlottesville, VA). For western blotting and immunoprecipitation, mammalian cells were disrupted by M-PER lysis buffer (Pierce, Rockford, IL) or NET2 buffer plus protease inhibitor cocktail, and whole-cell lysates used either directly for SDS-polyacrylamide gel electrophoresis (SDS–PAGE), or for immunoprecipitation. Immunoprecipitation samples were incubated overnight with antibody at 4 °C, subsequently incubated for 2 h with protein A/G-sepharose (Sigma), washed and resolved by SDS–PAGE. Western blotting was done using standard procedures and developed by chemoluminescence using the West-Pico system (Pierce). Transiently transfected or infected cells were analysed for protein expression at 24–96 h h post-transfection with Lipofectamine 2000 for plasmids, or Oligofectamine for siRNA (both from Invitrogen, Carlsbad, CA). Antibodies used included: rabbit polyclonal antibody to HEF1 (ref. 3) at a 1:100 dilution, mouse monoclonal antibody anti-HEF1 14A11 made for this study (1:500), anti--tubulin mouse monoclonal antibody (Sigma; 1:10,000), mouse monoclonal antibody anti--tubulin (GTU-88; Sigma; 1:5,000), anti-p130Cas (BD Biosciences; 1:2,000), anti-AurA (BD Biosciences; 1:1,000) for western, anti-AurA rabbit polyclonal (Abcam, Cambridge, MA; ab1287) for immunoprecipitation, anti-Phospho-AurA/T288 (Cell Signaling, Beverly, MA; 1:1,000), anti-HA-antibody (16B12; BabCo, Berkeley, CA; 1:5,000), mouse monoclonal antibody anti--actin (AC15; Sigma; 1:10,000), mouse monoclonal antibody anti-cyclin B (GNS-1; BD Biosciences; 1;1000). Rabbit anti-GFP (Abcam ab290) was used for immunoprecipitation, and mouse anti-GFP (JL-8; BD Biosciences; 1:2,000) was used for western blotting. Monoclonal antibody anti-GST (Cell Signaling; 1:2,000) and rabbit anti-Nek2 (Abcam; 1:200) were used for immunoprecipitation. Mouse anti-Nek2 (BD Biosciences; 1:500) and mouse monoclonal antibody anti-cdc2 antibody (Oncogene, Cambridge, MA; 1:1,000) were used for western blot analysis. Secondary anti-mouse and anti-rabbit HPR conjugated antibodies (Amersham) were used at a dilution of 1:10,000 or 1:20,000.
siRNA.
RNA oligonucleotides duplexes (sequences on request) were synthesized targeted to HEF1, AurA and to p130Cas, as well as negative controls including scrambled and GFP-directed sequences (Dharmacon, Lafayette, CO; Ambion, Austin, TX). After transfection of siRNAs, degree of depletion of target proteins was determined by western blot.
Cell synchronization.
Cells were incubated for 16–18 h with 2 mM thymidine, washed twice in PBS, then either assayed directly (for observation at the G1/S boundary), or returned to fresh medium and allowed to grow for 9–12 h to observe synchronized progression to mitosis. For synchronization at the G2/M boundary, cells were incubated in 1 M nocodazole for 14 h, collected by shake-off, washed in PBS, then either re-plated in fresh medium on glass coverslips, cultured at 37 °C for up to 90 min, then fixed for immunofluorescence analysis; or lysed for western and immunoprecipitation analysis. As an alternative drug-free synchronization approach, an elutriating centrifuge (Beckman J, Fullerton. CA) was used to enrich G1 or mitotic cell fractions (details on request). For all the synchronization procedures, the predicted cell-cycle compartmentalization was confirmed using FACS analysis.
Immunofluorescence.
For immunofluorescence, cells growing on coverslips were fixed with 4% paraformaldehyde, permeabilized, blocked and incubated with antibodies using standard protocols. Alternatively, to maximize clear signals at centrosomes, cells were fixed in cold methanol (-20 °C) for 10 min, blocked and incubated with antibody (see figure legends). Primary antibodies included mouse monoclonal antibody anti-AurA (BD Biosciences; 1:300), rabbit polyclonal anti-phospho-AurA/T288, (Cell Signaling; 1:200), rabbit polyclonal anti-HEF1 1:100, mouse monoclonal antibody anti-HEF1 (14A11) 1:100, rat monoclonal antibody anti--tubulin (Abcam; 1:200), rabbit polyclonal anti--tubulin (Abcam; 1:200), mouse monoclonal antibody anti-pericentrin (Covance, Berkeley, CA; 1:250), mouse monoclonal antibody anti-phosphohistone 3 (Upstate), rabbit anti-ninein antibody (1:200), mouse monoclonal antibody anti-c-Nap-1 (BD Biosciences; 1:100), and mouse monoclonal antibody anti-Nek2 (BD Biosciences; 1:100). Secondary antibodies anti-mouse Alexa-488, anti-rabbit Alexa488, anti-mouse Alexa-568, anti-mouse Alexa-633, anti-rabbit Alexa 633, and TOTO-3 dye to stain DNA, were from Invitrogen. Confocal microscopy was performed using a Radiance 2000 laser scanning device coupled to a Nikon Eclipse E800 upright microscope (Carl Zeiss, Thornwood, NY). Statistical analysis of data by one-way ANOVA was performed using GraphPad Instat 3.0 (San Diego, CA).
Kinase assays.
For phosphorylation of HEF1 by AurA, an in vitro kinase assay was performed using bacterially expressed GST-fused HEF1 derivatives. Histone H3 (Upstate) and H1 (Upstate) were used as positive and negative controls for recombinant AurA (Upstate) phosphorylation, using standard methods except as noted in the Results. In parallel, aliquots without -32P(ATP) were processed for SDS–PAGE/Coomassie staining (Invitrogen). GST-pulldown assays used wild-type AurA translated (pCMV-SPORT6-C6) using TnT coupled reticulocyte lysate system (Promega, Madison, WI) mixed with titrated quantities of GST-fused HEF1 derivatives. To analyse HEF1 activation of recombinant, baculovirus-produced AurA, GST–HEF11–363 or GST was titrated into a mixture containing recombinant AurA, immunoprecipitated with anti-AurA, and used for a kinase reaction with -32P(ATP) and histone H3 substrate. Aliquots of the reaction mixture were used for SDS–PAGE and western analysis to confirm levels of AurA; phospho-histone H3 was visualized by autoradiography or by phospho-specific antibody. AurA kinase used in Fig. 6c was precipitated from MCF-7-tTA-neo or HEF1 cell lines as well as from MCF-7 cells treated with control or specific oligonucleotide duplex against HEF1 (siHEF1), using anti-AurA antibody (ab1287). The Nek2 kinase assay was performed in standard kinase buffer with addition of an Mg/ATP cocktail (Upstate), with MBP as the substrate.
BIND identifiers.
Three BIND identifiers (www.bind.ca) are associated with this manuscript: 334317, 334318 and 334319.
