T.Ng and M.Parsons contributed equally to this work
Protein kinase C (PKC) α has been implicated in β1 integrin-mediated cell migration. Stable expression of PKCα is shown here to enhance wound closure. This PKC-driven migratory response directly correlates with increased C-terminal threonine phosphorylation of ezrin/moesin/radixin (ERM) at the wound edge. Both the wound migratory response and ERM phosphorylation are dependent upon the catalytic function of PKC and are susceptible to inhibition by phosphatidylinositol 3-kinase blockade. Upon phorbol 12,13-dibutyrate stimulation, green fluorescent protein–PKCα and β1 integrins co-sediment with ERM proteins in low-density sucrose gradient fractions that are enriched in transferrin receptors. Using fluorescence lifetime imaging microscopy, PKCα is shown to form a molecular complex with ezrin, and the PKC-co-precipitated endogenous ERM is hyperphosphorylated at the C-terminal threonine residue, i.e. activated. Electron microscopy showed an enrichment of both proteins in plasma membrane protrusions. Finally, overexpression of the C-terminal threonine phosphorylation site mutant of ezrin has a dominant inhibitory effect on PKCα-induced cell migration. We provide the first evidence that PKCα or a PKCα-associated serine/threonine kinase can phosphorylate the ERM C-terminal threonine residue within a kinase–ezrin molecular complex in vivo.
Of the protein kinase C (PKC) family of protein kinases, several isoforms including PKCα (Platet et al., 1998), PKCδ (Kiley et al., 1999) and PKCθ (Tang et al., 1997) have been implicated in promoting a cell migratory phenotype. Recently, we and others have shown an enhancement of haptotactic cell motility/invasiveness following overexpression of PKCα (Ng et al., 1999a; Sun and Rotenberg, 1999). The precise mechanism(s) by which the catalytic activity of PKC influences cell motility has yet to be defined but is likely to involve components of the focal complexes, the small GTPases (Rho, Rac and Cdc42) and actin cytoskeleton-binding proteins (reviewed by Zigmond, 1996). In MCF-7 breast cancer epithelial cells, the signal machinery that drives efficient cell migration towards extracellular matrix β1 integrin substrates is dependent upon the PKCα-catalysed recycling of the activated [ligand-occupied, monoclonal antibody (mAb) 12G10-reactive] pool of β1 integrin receptors (Ng et al., 1999a). Overexpression of PKCα in MCF-10A human breast cells enhances cell motility, which is subject to control by the small GTPase Rac1 (Sun and Rotenberg, 1999). By using dominant inhibitor constructs, the activity states of RhoA, Rac1 and Rab5 have been shown to influence 12-O-tetradecanoylphorbol-13-acetate (TPA)-induced disassembly/reassembly of actin stress fibres and focal adhesions in MDCK cells (Imamura et al., 1998). In NIH 3T3 cells, growth factor-induced Rac1 activation is sensitive to PKC inhibition, also placing PKC upstream of Rac in this cell system (Buchanan et al., 2000).
The crystal structure of a dormant moesin N-terminal FERM (band four-point-one, ezrin, radixin, moesin homology domains)–C-terminal tail complex reveals the C-terminal segment to be bound as an extended peptide masking a large surface of the FERM domain (Pearson et al., 2000). The activation of ERM proteins is mediated by both C-terminal threonine phosphorylation (T567 in ezrin, T564 in radixin and T558 in moesin) (Matsui et al., 1999; Nakamura et al., 1999; Gautreau et al., 2000; Pearson et al., 2000) and exposure to polyphosphoinositides (reviewed by Mangeat et al., 1999). In vitro, the T558-phosphorylated form of moesin interacts with F-actin only in the presence of polyphosphoinositides such as phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2] (Nakamura et al., 1999).
Two protein kinases, namely PKCθ (Pietromonaco et al., 1998) and Rho-kinase/ROCK (Matsui et al., 1998; Tran Quang et al., 2000), were found to phosphorylate the C-terminal threonine residue of the ERM proteins in vitro. However, the authors who first showed that ROCK phosphorylates the site in vitro (Matsui et al., 1998) subsequently reported that the C-terminal threonine phosphorylation of ERM is in fact insensitive to ROCK inhibition in vivo (Matsui et al., 1999). To date, the protein serine/threonine kinase(s) that is responsible in vivo for the phosphorylation of the ERM C-terminal threonine residue has yet to be identified.
In the present study, we have found both green fluorescent protein (GFP)–PKCα and β1 integrins to co-localize with endogenous ezrin in vivo. Using fluorescence lifetime imaging microscopy (FLIM) and co-immunoprecipitation, GFP–PKCα was confirmed to form a molecular complex with vesicular stomatitis virus (VSV) G-tagged ezrin in MCF-7 cells. The C-terminal threonine phosphorylation of ERM is dependent on the catalytic domain function of PKC (not ROCK) in an in vivo model of cell migration. Finally, a phosphorylation site mutant of ezrin that is incapable of interacting with the cytoskeleton but retains the ability to bind PKC is shown here to exert a dominant inhibitory effect on PKC-driven cell migration.
PKC-enhanced cell migration in response to wounding
Previously we observed that transient expression of GFP–PKCα increased the haptotactic response of MCF-7 cells towards β1 (but not αvβ5) integrin substrates (Ng et al., 1999a). In this study, we confirmed that stable expression of a GFP–PKCα full-length construct also significantly enhanced the ‘persistence’ of 2C4 fibrosarcoma cell migration in response to wounding in comparison with cells stably expressing the corresponding GFP control vector [P <0.01 by analysis of variance (ANOVA), Figure 1A].
The cumulative speed profile of GFP–PKCα-2C4 cells in Figure 1A showed faster cell movement than that of the vector control cell line during the initial 4 h post-wounding but, when the whole time course was analysed statistically, the speed of the GFP–PKCα-expressing cells based on 5 min displacements was not different from the speed of the control cells by ANOVA. The total level of PKCα expressed in GFP–PKCα-2C4 cells (which is a mixed clone), after correction with the β-actin control, was ∼3 times that of the endogenous protein present in the vector control cells (Figure 1A). In GFP–PKCα-2C4 cell cultures wounded in the presence of the PKC inhibitor bisindolylmaleimide I (BIM, 10 μM) or the phosphatidylinositol 3-kinase (PI3K) inhibitor LY294002 (10 μM), there was a loss of the wound closure response (P <0.01 by ANOVA for all three inhibitors, showing a significant loss of persistence of the wound closure response compared with no treatment control). Persistence of motility measured for each cell was calculated as the ratio of the resultant displacement in 16 h over the sum of individual 5 min displacements. The ROCK inhibitor Y-27632 had no effect on the wound-induced migratory response (persistence or speed) compared with untreated cultures (P = 0.49 by ANOVA showing no significant difference in persistence of the wound response compared with no treatment control). An analysis of the speed data by ANOVA, however, indicated a significant reduction in speed with LY294002 compared with no treatment control.
Activation of PKC with phorbol 12,13-dibutyrate (PDBu) treatment (post-wounding) was shown to reverse the inhibition of cell migration by LY294002 [P = 0.096 by ANOVA, comparing the wound migratory response following pre-treatment with LY294002 (+PDBu) with control treatment, i.e. +PDBu post-wounding alone; Figure 1B].
PKC control of ERM C-terminal phosphorylation in response to wounding
A monoclonal antibody (297S) that recognizes the C-terminal threonine-phosphorylated (T567 in ezrin, T564 in radixin and T558 in moesin; Nakamura et al., 1999) form of ERM proteins (CPERM) was used to monitor their phosphorylation in relation to PKC activation. In MCF-10A monolayer culture, endogenous PKCα redistributed to a vesicular compartment at the wound edge (Figure 2A). This redistribution was coincident with an increase in PKCα phosphorylation at the Thr250 site (Ng et al., 1999b), indicative of PKC activation at the wound edge (T.Ng, unpublished observations). Endogenous ERM protein phosphorylation at the C-terminal threonine residue is low in confluent monolayer cultures (observed for MCF-7, MCF-10A epithelial cells and GFP–PKCα-2C4 fibrosarcoma cell cultures) (Figures 2 and 3). An increase in ERM C-terminal threonine phosphorylation in response to wounding was observed at the wound edge. Pre-treatment of MCF-7 cells at the wound edge with alkaline phosphatase, following paraformaldehyde fixation, significantly reduced the recognition of endogenous ERM proteins by mAb 297S, confirming the phosphospecificity of the antibody (data not shown). The anti-CPERM staining partially co-localized with both the GFP-tagged and endogenous PKCα at the wound edge. The site of PKC/CPERM co-localization at the wound edge varied according to cell type. Co-localization was seen predominantly in intracellular vesicular structures in mammary epithelial cells such as MCF-10A and at cell ruffles in the case of 2C4 fibrosarcoma cells (Figure 2).
ERM C-terminal threonine phosphorylation at the wound edge was sensitive to PKC inhibition by BIM and PI3K inhibition by LY294002, but was not affected by the ROCK inhibitor Y-27632 (Ishizaki et al., 2000) (Figure 2). In 2C4 fibrosarcoma cells stably transfected with GFP–PKCα, the abrogation of the ERM C-terminal threonine phosphorylation by these inhibitors was mirrored by a reduction of the wound closure response (Figure 1A).
GFP–PKCα complexes with VSVG–ezrin in vivo
Of the ERM proteins, ezrin has been shown to be an effector of growth factor-mediated cell migration (Crepaldi et al., 1997). To determine whether ezrin forms part of the PKC–integrin complex, we assessed whether PKCα and ezrin associate upon PKC activation. FLIM was undertaken to determine the extent of fluorescence resonance energy transfer (FRET) between GFP–PKCα (donor) and VSVG-tagged ezrin stained with a Cy3 (acceptor)-conjugated antibody to the VSVG epitope. The spatial separation (R0) between GFP and Cy3 at which the resonance energy transfer efficiency is 50% is 6 nm; however, detectable energy transfer can occur up to 9 nm (Ng et al., 1999b). FRET results in a shortening of the GFP (donor) fluorescence lifetime that is measured by two independent parameters, the phase shift (τp) and relative modulation depth (τm). Therefore, FRET should only be detected between two proteins that are closely associated or complexed in vivo.
When GFP–PKCα- and VSVG–ezrin-expressing cells were stained with VSVG-Cy3 following PDBu stimulation and fixation, it was evident that fluorescence lifetime, <τ>, measured as the average phase shift and relative modulation depth [(τp + τm)/2] for GFP was decreased at the plasma membrane and cell cortex beneath the membrane (Figure 3A). The presence of FRET (from GFP to Cy3) in these structures was confirmed by photobleaching the Cy3 acceptor, resulting in a lengthening of the donor <τ> after acceptor photobleaching (Figure 3C and data not shown). The FRET efficiency (Eff) pseudocolour plots, which only apply to the +anti-VSVG-Cy3 lifetime maps, give a qualitative indication of the pixel-to-pixel variation in the amount of PKC molecules that are complexed to ezrin. [Eff = 1 − τda/τd where τda is the lifetime map of the donor in the presence of acceptor and τd is the average lifetime of the donor in the absence of acceptor (average of the mean<τ> of at least five GFP–PKC alone control cells).] The efficiencies are approximately proportional to the fraction of the total steady-state fluorescence that originates from donor molecules in their bound state. Since the donor-tagged molecules transfer energy to the acceptor when they are in a complex, their lifetimes are reduced and hence the number of photons emitted would be reduced fractionally.
Statistical analysis of cumulative results from all the cells at each time point confirmed that there was a time-dependent, PDBu-enhanced association between GFP–PKCα and VSVG–ezrin [see two-dimensional lifetime histograms and pixel counts versus Eff (%) profile in the lower panels of Figure 3A]. The maximal levels of acceptor fluorescence intensity (anti-VSVG ezrin-Cy3) were similar in both untreated and PDBu-stimulated cells, and there was no significant correlation between <τ> and the intensity of acceptor fluorescence at each pixel of the cell (data not shown). The increase in association by PKC activation observed in Figure 3A is therefore not an artefact due to protein overexpression. Phorbol esters bind to the C1 domains of conventional and novel PKCs, and thereafter induce a conformational change that results in a disruption of the intramolecular interaction between the pseudosubstrate site in the regulatory domain (RD) and the kinase domain. Figure 3B shows that the expression of the PKCα RD + V3 (hinge region) alone, without the catalytic domain, renders the PKC–ezrin association constitutive. This implies that the PDBu effect is mediated at least in part through a conformational change in the PKC regulatory domain, and that the association is not merely a kinase domain–substrate binding.
The ezrin-bound and uncomplexed subpopulations of GFP–PKCα are associated with distinct fluorescence lifetimes (τ1 and τ2, respectively). Quantitative maps of the populations of each form were derived using a global FLIM analysis as previously described (Verveer et al., 2000a,b), and two pseudocolour cell plots are presented in Figure 3B as ‘percentage population of PKC complexed to ezrin’. A cumulative analysis of all the cell data showed that the average populations of the ezrin-complexed form of PKCα RD + V3 were 0.48 ± 0.08 (mean ± SD, n = 7; τ1 = 0.56 ns and τ2 = 2.12 ns) and 0.51 ± 0.12 (n = 7; τ1 = 0.46 ns and τ2 = 2.06 ns), before and after 20 min of PDBu treatment, respectively. This is in perfect agreement with the equivalent cumulative pixel counts versus Eff (%) profiles in the lowest panel, which show that PKC–ezrin association was rendered constitutive by expressing the RD + V3 alone. There is no linear relationship between the two quantities (i.e. populations versus Eff); hence, the contrast in the Eff cell plots in Figure 3B appears to be slightly different from the population images. Importantly though, the biological conclusions drawn from the cumulative/statistical analysis of these two parameters are identical.
Figure 3C shows that the non-phosphorylatable T567A mutant of VSVG–ezrin, which associates poorly with the actin cytoskeleton (Gautreau et al., 2000; Tran Quang et al., 2000), still interacts with PKCα. Cell morphology and the degree of PKC/ezrin co-localization were similar between wild-type and T567A ezrin-co-transfected cells, in agreement with previous observations made in LLC-PK1 cells (Gautreau et al., 2000). Thus, phosphorylation at the C-terminal threonine residue is not a prerequisite for complex formation.
Co-sedimentation of GFP–PKCα and ezrin
Previously we found by cryoelectron microscopy that the 12G10-reactive β1 integrin localized to the filopodia and multivesicular bodies in breast carcinoma cells (Ng et al., 1999a). In pulse–chase experiments coupled with immunofluorescence staining, the vesicular structures containing activated β1 integrins were shown to recycle and co-localize with transferrin receptors. By sucrose gradient fractionation, GFP–PKCα and β1 integrin were found to co-sediment through both velocity (VG) and equilibrium (EG) sucrose gradients prior to and following PDBu stimulation (Figure 5A and B and data not shown). Consistent with the FLIM results, in unstimulated MCF-7 cells there was no significant co-sedimentation between GFP–PKCα and endogenous ezrin (Figure 4A). In PDBu-stimulated cells, however, ezrin became highly enriched in PKCα- and β1 integrin-containing EG fractions (Figure 4B). Among the post-PDBu EG fractions, endogenous transferrin receptors were only detected in fractions 1–4, which were enriched in GFP–PKCα, β1 integrin and endogenous ezrin (data not shown), indicating that these fractions contained recycling endosomes.
T567 hyperphosphorylation of the PKCα-associated ezrin subpopulation
To corroborate the FLIM and sucrose gradient co-sediment ation data, we determined whether a PKCα–ezrin complex could be demonstrated by co-immunoprecipitation. Immunoprecipitation of endogenous PKCα from PDBu-stimulated MDA-MB-231 (which, unlike MCF-7, contains a sufficient amount of endogenous PKCα detectable by immunoblotting) with the antibody MC5 and western analysis with an anti-ezrin rabbit serum reveals that endogenous ezrin can be co-immunoprecipitated with PKCα (Figure 5A). The pool of ezrin in the PKCα-bound fraction was composed predominantly of a slower electrophoretic mobility form (labelled ezrin*) when compared with the unbound ezrin. The type of large resolving acrylamide gels used in these precipitation assays was shown previously to distinguish the mature phosphorylated form of PKCα (80 kDa) from its dephosphorylated variant (with an apparent mol. wt of 76 kDa) (Cazaubon et al., 1994). Figure 5B shows a similar co-precipitation of GFP–PKCα with endogenous ezrin from PDBu-stimulated GFP–PKCα-2C4 cells, in two independent experiments (cultures 1 and 2); again the slow mobility form ezrin* was detected predominantly in the PKCα-bound fraction. The ratio of ezrin*:ezrin in the ‘bound’ fraction on the total ezrin blot in Figure 5B was 0.7 and 1.4 for cultures 1 and 2, respectively. An increase in the amount of the C-terminal threonine-phosphorylated form of ERM proteins (detected by an affinity-purified polyclonal rabbit anti-CPERM IgG) was found in the PKCα-bound pool, when compared with the unbound fraction. From densitometric analysis, the mean ratio of CPERM in PKC-bound versus unbound fraction ± SEM = 2.3 ± 0.3 (n = 7 independent cultures) and, when corrected for the total amount of ezrin present in each lysate, the mean ratio of CPERM in bound versus unbound ± SEM = 10.7 ± 3.1. The PKCα-associated ERM pool is therefore hyperphosphorylated at the C-terminal threonine residue. The phosphospecificity of this affinity-purified polyclonal IgG is demonstrated in Figure 5D.
The relative proportion of total ezrin/radixin/moesin bound to GFP–PKCα in unstimulated GFP–PKCα-2C4 cells was investigated by co-precipitation and is presented in Figure 5C. A comparison with Figure 5B and other similar experiments (n = 7 independent cultures) shows that in unstimulated cells, there is no C-terminal hyperphosphorylation of ERM in the PKC-associated fraction. The ezrin* band is not apparent under these conditions. A negligible amount of radixin was bound to PKCα. The stoichiometry of pull-down for moesin (PKCα-bound: unbound ∼0.2) was less than that for ezrin (PKCα- bound:unbound ∼0.7). Ezrin is therefore the predominant ERM protein that is associated with PKCα under these conditions.
Enrichment of GFP–PKCα and ezrin at plasma membrane protrusions
The precise localization of GFP–PKCα and ezrin was examined by immunoelectron microscopy (EM). Our initial efforts to localize PKCα by GFP immunogold labelling were unsuccessful using transiently transfected MCF-7 cells; the amount of endogenous ERM proteins in these cells was also below the detection limit using an affinity-purified rabbit anti-ezrin IgG (S.Gschmeissner and T.Ng, unpublished observations). A fibrosarcoma cell line stably expressing GFP–PKCα (GFP–PKCα-2C4) and transiently transfected with a VSVG-tagged ezrin was therefore chosen for EM studies. Electron micrographs in Figure 6A and B show the distribution of both VSVG-tagged ezrin [detected by an affinity-purified rabbit anti-ezrin IgG + 10 nm protein A–gold (PAG)] and GFP–PKCα (detected by a rabbit anti-GFP antiserum + 5 nm PAG) in the cytoplasm and plasma membrane of PDBu-stimulated GFP–PKCα-2C4 cells. Both proteins were found to be particularly enriched at plasma membrane protrusions. In contrast, there was no detectable labelling of other membranous structures including the mitochondria, Golgi and the nucleus. Control experiments where the second antibody (anti-GFP or anti-ezrin IgG) was omitted revealed no cross-reactivity of the 5 nm PAG with the first antibody (anti-ezrin or anti-GFP IgG, respectively) (data not presented). Figure 6C and D shows the immunolocalization of wild-type and T567A, VSVG-tagged ezrin in GFP–PKCα-2C4 cells, respectively (stained with an anti-VSVG mAb P5D4 + 10 nm PAG). The distribution is remarkably similar. Also the distinction between the membrane- and cytoskeleton-associated subpopulations of wild-type versus T567 ezrin would be difficult, even at the EM level of resolution.
PKCα phosphorylates ezrin in vitro
The activation-enhanced association/proximity between PKCα and ezrin (Figure 3) as well as the significant diminution of ERM C-terminal threonine phosphorylation by PKC inhibition (Figure 2) suggest that ERM proteins may be in vivo substrates of PKCα. The only PKC isoform that has been documented to phosphorylate the ERM proteins in vitro to date is PKCθ (Pietromonaco et al., 1998). Figure 7A shows that recombinant human PKCα can also phosphorylate full-length glutathione S-trans ferase (GST)-tagged ezrin in vitro in a Ca2+/TPA/phospholipid-dependent manner. After 30 min of in vitro phosphorylation at 37°C, the stoichiometry was 1.5 mol ATP incorporated per mol ezrin. The ratio of 32P incorporation in ezrin ± PKC activators was ∼3:1. Full-length ezrin was a better PKC substrate than the GST–C-terminal ezrin (amino acids 280–585) (<0.3 mol ATP incorporated per mol ezrin; data not shown), implying that components in the N-terminal part of ezrin might be required for kinase–substrate contact.
We next investigated the ability of immunoprecipitated PKCα to phosphorylate its associated endogenous ezrin at the C-terminal threonine residue ex vivo (procedure described in the legend of Figure 7B). In view of the findings in Figure 2, pre-treatment with LY294002 (10 μM for 6 h) was used to reduce the basal ERM C-terminal threonine phosphorylation in GFP–PKCα-2C4 cells, or more specifically in the PKCα-bound ERM proteins (−Mg/ATP, anti-CPERM ‘bound’ lane in Figure 7B). Notably, the PKC-associated ezrin* band was barely detectable in the total ezrin blot from cells pre-treated with LY294002 when compared with the untreated cultures shown in Figure 5B. After the ex vivo phosphorylation reaction (+Mg/ATP, anti-CPERM ‘bound’ lane in Figure 7B), there was an ∼3-fold increase in the amount of CPERM on beads, indicating that PKCα, or a PKCα-associated serine/threonine kinase, is capable of phosphorylating the ERM C-terminal threonine residue within a kinase–ezrin molecular complex.
In order to address further the identity of the CPERM that is bound to PKCα, the PKCα immunoprecipitation experiment was repeated with GFP–PKCα-2C4 cells that had been pre-treated with LY294002 (10 μM for 6 h) to reduce the basal ERM C-terminal threonine phosphorylation before lysis. Under conditions where ERM C-terminal threonine phosphorylation is reduced, the CPERM in the PKC-bound fraction had the same apparent molecular weight as that in the unbound fraction. Using a large resolving acrylamide gel, the major band in the anti-CPERM blot (indicated by a filled arrowhead in Figure 7C) co-migrated with the major band in the anti-ezrin blot (developed with an affinity-purified anti-ezrin IgG that has been cross-adsorbed for moesin and radixin; Algrain et al., 1993). Very little radixin that corresponds to the major radixin band in the unbound fraction was found in the PKCα immunoprecipitate. The major moesin band had a similar apparent molecualr weight but the stoichiometry (PKCα-bound:unbound ∼0.1) of pull-down was less than that for ezrin (PKCα-bound:unbound ∼0.2). Ezrin is therefore also the predominant ERM protein that remains associated with PKCα under conditions where ERM C-terminal threonine phosphorylation is reduced.
Ezrin T567A mutant inhibits PKCα-driven migration in wound closure
To investigate the role of ERM proteins in PKCα-dependent migration, GFP–PKCα-2C4 cells transiently co-expressing ezrin (wild-type/T567A variant) were tracked during overnight wound migration assays. The T567A mutant of VSVG–ezrin associates poorly with the actin cytoskeleton (Gautreau et al., 2000; Tran Quang et al., 2000) but retains the capacity to interact with PKCα in vivo (Figure 3C). Figure 8 demonstrates that the injection of a wild-type ezrin plasmid into GFP–PKCα-2C4 cells did not affect the PKCα-stimulated wound closure response. However, introduction of the T567A phosphorylation site mutant had a dominant inhibitory effect on PKCα-induced cell migration (P = 0.02 by ANOVA for the difference between wild-type- and T567A ezrin-injected cells, showing a significant loss of persistence of the wound closure response due to T567A ezrin expression). There was no significant cell toxicity associated with T567A ezrin expression, as assessed by the propidium iodide exclusion assay (data not shown).
The results here demonstrate that on activation, PKCα co-localizes and interacts with ezrin at the plasma membrane and cell cortex beneath the membrane, particularly at the membranous protrusions of the cell. Associated with this interaction, ERM proteins become phosphorylated on the C-terminal threonine residue, and agents that block this phosphorylation also abrogate PKCα-dependent migration. In addition, PKC-associated endogenous ERM proteins are shown to be hyperphosphorylated preferentially at the C-terminal threonine residue, i.e. activated. Among the endogenous ERM proteins, ezrin is the predominant ERM protein that is associated with PKCα. Recombinant human PKCα is capable of phosphorylating full-length GST-tagged ezrin in vitro in a Ca2+/TPA/phospholipid-dependent manner. Full-length ezrin is a better PKC substrate than the GST–C-terminal ezrin, implying that components in the N-terminal part of ezrin might be required for stabilizing kinase–substrate contact. The expression of the regulatory domain of PKCα alone is sufficient to support this interaction. We have also provided the first evidence that PKCα, or a PKCα-associated serine/threonine kinase, is capable of phosphorylating the ERM C-terminal threonine residue within a kinase–ezrin molecular complex in cells.
The pool of ERM proteins bound to PKCα contains a higher proportion of the C-terminal threonine-phosphorylated form than the ‘unbound’ pool. The increase in C-terminal threonine phosphorylation in the bound fraction correlates with the finding of a subset of ezrin that is associated with a slower electrophoretic mobility (ezrin*). The increase in the apparent molecular weight of ezrin* is likely to be related to its phosphorylation by PKC (or a PKCα-associated serine/threonine kinase) within an intermolecular complex, as well as any additional biochemical modification that occurs as a consequence of the phosphorylation. This is because in unstimulated cells where there is no C-terminal hyperphosphorylation of ERM in the PKC-associated fraction (Figure 5C), or under conditions where ERM C-terminal threonine phosphorylation is reduced (Figure 7C), the CPERM in the PKC-bound fraction has the same apparent molecular weight as that in the unbound fraction. The ezrin* band is also not apparent under these conditions.
Recent work (Gautreau et al., 2000) clearly demonstrates that the C-terminal threonine-phosphorylated form is enriched in the membrane and exists as a monomer. One possible consequence of the C-terminal threonine phosphorylation within the PKC–ezrin molecular complex is therefore to promote a transition of ERM from the oligomeric to monomeric state (active form). The activated form of ERM then acts as a downstream effector of PKCα involved in the control of polarized cell movement. The observed dominant inhibitory effect of expressing the T567A mutant of ezrin on PKC-enhanced wound closure can be explained in a number of ways. First, the T567A variant has an almost identical cytosol/membrane distribution as the wild-type protein but associates poorly with the cytoskeleton (Gautreau et al., 2000). This is not surprising since the T567A variant can oligomerize with endogenous ERM proteins and localize to the same compartments in cells. The localization of the T567A variant is indistinguishable from that of the wild-type protein, even at the EM level (Figure 6). Through oligomerization with endogenous ERM proteins, this mutant may exert a dominant inhibitory effect by sequestering the active (C-terminally phosphorylated) form of ERM. Alternatively, since T567A ezrin will still interact with PKCα in intact cells, it may sequester the kinase away from foci of actin polymerization.
In MDCK cells, PKCα (not PKCδ or ζ) is the only PKC isotype that changes its cellular localization concomitantly with the development of Ca2+-dependent desmosomes in response to wounding (Wallis et al., 2000). Treatment with an antisense oligonucleotide to PKCα abolishes the reversal of desmosomes to Ca2+ dependence at the wound edge, illustrating the important role that this specific PKC isozyme plays in the wounding response. The molecular mechanisms underlying the wound closure migratory response are, hitherto, poorly understood. After an overnight incubation, an intermediate zone of subconfluent cells is usually observed between the wound edge and the unperturbed part of the confluent monolayer culture. Cells in this subconfluent zone are motile and have less extensive intercellular contact than cells from the confluent region. PKC-induced wound closure (such as that shown in Figure 1A) is characterized by an increase in ‘persistence’ in the direction of cell movement, particularly in this intermediate zone where there is less contact inhibition of cell migration. Based on our previous work using the Transwell chamber migration sytem where the filter was coated with either bovine serum albumin (BSA; random motility) or specific integrin substrates (Ng et al., 1999a), the expression of PKCα in MCF-7 cells was observed significantly to enhance random motility as well as haptotaxis towards β1 integrin substrates on filters. Here we conclude that stable expression of a GFP–PKCα full-length construct significantly enhances the ‘persistence’ of 2C4 fibrosarcoma cell migration in response to wounding. GFP–PKCα-2C4 cells tend to undergo a more accelerated cell movement (speed) during the initial 4 h post-wounding in comparison with the vector control. Speed is a measure of random motility (Surtees, 1964) and is mathematically an entirely independent entity from persistence/directionality. These conclusions do not just apply to confluent cells induced to move in response to wounding, since the expression of PKCα in subconfluent MCF-7 cells also significantly enhances the directional cell movement in a chemotaxis assay (M.Parsons et al., in preparation).
ERM C-terminal phosphorylation at the wound edge is sensitive to PI3K inhibition with LY294002, but not the ROCK inhibitor Y-27632. The PKC-driven wound closure response is significantly impaired by these same inhibitors that abolish the ERM C-terminal threonine phosphorylation; this response is also insensitive to Y-27632. PI3K has long been implicated in the control of cell migration/haptotaxis (Wang et al., 1996; Derman et al., 1997; Shaw et al., 1997; Meng and Lowell, 1998; Ng et al., 1999a; Vanhaesebroeck et al., 1999). At least in NIH 3T3 fibroblasts, the di-C16-phosphatidylinositol-3,4,5-trisphosphate (PtdIns-P3)-induced cell motility involves PKC and can be inhibited by calphostin C (Derman et al., 1997). Also, in lymphocytes, PI3K blockade has an inhibitory effect on moesin redistribution to the uropod during migration (Vicente-Manzanares et al., 1999). Here we demonstrate that in a GFP–PKCα stably transfected fibrosarcoma cell line (GFP–PKCα-2C4), ERM C-terminal threonine phosphorylation, which is sensitive to inhibition of the catalytic activity of PKC in vivo, is also subject to PI3K control. PI3K may, by enhancing phospholipase C-γ activity in vivo (Bae et al., 1998), increase the availability of diacylglycerol (DAG), which in turn activates PKC and ERM phosphorylation. This would be consistent with the inhibitory effect that calphostin has (via competition for the DAG-binding site on PKC; Bruns et al., 1991) on PtdIns-P3-induced cell motility (Derman et al., 1997). Furthermore, we show that activation of PKC with PDBu treatment can reverse the inhibition of the wound migratory response by LY294002, placing allosteric activation of conventional or novel PKC isozymes downstream of PI3K (Figure 1B). These data further define the functional relationship between PI3K and PKC as downstream effectors of the integrin signalling axes (Berrier et al., 2000). Finally, the p85 regulatory subunit of PI3K has been shown to bind to the N-terminal part of ezrin; a point mutant of ezrin that does not interact with PI3K has an inhibitory effect on the activation of Akt, a downstream effector of the PI3K pathway (Gautreau et al., 1999). This complex array of feedback mechanisms between PKC, PI3K and ezrin may constitute an important mechanism for signal amplification and cell migratory function.
The PKC isozymes constitute a major family of protein kinases that influence cell motility through a variety of mechanisms that are involved in the regulation of cytoskeletal turnover (Arber et al., 1998) and possibly microtubule function (Steiner et al., 1990). Some of the downstream components may, like ezrin, form stable complexes with PKC in response to appropriate physiological stimuli, while others, such as cofilin, may be regulated indirectly through other signalling proteins (Arber et al., 1998). Further studies are needed to integrate our current knowledge of these individual signal components in a physiological context with the aim of intervening in these PKC-dependent processes in a pathological milieu such as tumour cell metastasis.
Materials and methods
Cell culture and plasmid constructs
Human breast carcinoma cells (MCF-7 cells) were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal calf serum at 37°C, in a 10% CO2 atmosphere. In addition, the MCF-7 cultures were grown in media supplemented with insulin (10 μg/ml). MCF-10A cells were established originally from mammary tissue of a patient with fibrocystic breast disease and display characteristics of luminal epithelial cells. Transient transfection was performed using Fugene (Boehringer Mannheim). 2C4 fibrosarcoma cells (derivative of HT1080 cells) were transfected with either GFP–PKCα plasmid (Ng et al., 1999a,b) or the control vector. At 48 h after transfection was initiated, cells were replated in complete medium at a 1:5 ratio. Selection for neomycin resistance was started the following day by the addition of the antibiotic G418 (Sigma) to 400 μg/ml. GFP-expressing cells were enriched by fluorescence activated cell sorting. After 10–14 days, G418-resistant cells were isolated as a mixed clone (see Figure 1A for the level of exogenous PKCα) and subsequently maintained in complete growth medium containing 40 μg/ml G418. Stable transfectants were cultured for up to 20 passages. The GFP–PKCα RD + V3 (Ng et al., 1999a) was made by subcloning the corresponding PKCα sequence encoding amino acids 1–337 into the pEGFP-N1 vector (Clontech), permitting fusion of the coding sequence of eGFP. The VSVG-tagged ezrin (wild-type and T567A variant) constructs were published previously (Gautreau et al., 2000).
Antibodies and direct conjugation to fluorophores, and Fab fragment generation
Anti-VSVG mAb was purchased from Boehringer Mannheim. MC5 is a murine mAb that recognizes the V3 region of PKCα (Young et al., 1988). 8E3 is a pan-β1 integrin murine mAb (a kind gift of M.Humphries). The rat mAb (297S) that recognizes the C-terminal threonine-phosphorylated (T567 in ezrin, T564 in radixin and T558 in moesin; Nakamura et al., 1999) form of ERM was a kind gift of S.Tsukita (Kyoto, Japan). A similar polyclonal rabbit anti-CPERM antibody was obtained by immunizing with a keyhole limpet haemocyanin (KLH)-coupled CRDKYKT*LRQIR peptide (the asterisk indicates the phosphorylated residue) and purifying the serum obtained through a GST–ezrin C(280–585) column to eliminate the non-phosphospecific antibodies. Phosphospecific antibodies were then affinity purified on a column coupled to the immunizing peptide. Goat polyclonal IgG against the C-terminus of ezrin was purchased from Santa Cruz and used in immunocytochemical staining. For western blotting, both a crude rabbit serum against ezrin and its affinity-purified IgG that has been cross-adsorbed for moesin and radixin (Algrain et al., 1993) were used. Also, crude sera raised against radixin and moesin (kind gifts of Professor Paul Mangeat, Universite Montpellier II) were used for blotting. Fab fragments were generated and isolated from mouse IgG using the ImmunoPure Fab preparation kit (Pierce) according to the manufacturer's protocol. Direct conjugation of IgG/IgG Fab fragments to the fluorophores Cy3 and Cy5 (Amersham Life Science) was performed at pH 8.5 (IgG) or pH 9.0 (IgG Fab) as described previously (Bastiaens and Jovin, 1998).
Wound migration assays
Cells were plated at confluence onto 22 mm coverslips and allowed to adhere. In the inhibitor studies, compounds were added to the media and cells were pre-incubated for 1 h prior to wounding. A single wound was made using a plastic pipette tip through the confluent monolayer and media were changed. For the ezrin studies, a wild-type or T567A VSVG-tagged ezrin construct was microinjected into the nuclei of confluent cells using an Eppendorf system, along with a Cy3-labelled anti-transferrin IgG for the subsequent identification of injected cells. Monolayers were then wounded in such a way that most of the injected cells would remain at the wound edge. Cell migration at the wound edge was recorded overnight at a 5 min lapse interval using a 10× objective, and the images were acquired using Kinetic Imaging (Merseyside, UK) AQM software.
Cell tracking and statistical analysis of migration
Cell trajectories were acquired by manually tracking all the cells over the sequence of time-lapse digital images (Motion Analysis, Kinetic Imaging, Merseyside, UK). The data were used to calculate speed as distance in μm per 5 min and converted to μm per h. Another important aspect of the migration response is the persistence of motility that we measured for each cell as the ratio of the resultant displacement in 16 h over the sum of individual 5 min displacements (Figures 1A and B, and 8). Comparisons between different groups of experiments were made by applying ANOVA to the persistence data. We used ANOVA with a hierarchical and unbalanced model that allows more general testing between treatment groups, and takes into account possible variations between experiments and among different cells in our recordings.
Immunocytochemical staining and confocal microscopy
Immunocytochemical staining was performed as described elsewhere (Kiley and Parker, 1997) except for the following modifications. Cells were permeabilized with 0.2% (v/v) Triton X-100/phosphate-buffered saline (PBS) following fixation in 4% (w/v) paraformaldehyde, with the exception of the slides processed for the GFP–PKCα/VSVG–ezrin association study by FLIM, which were fixed in methanol for 4 min at −20°C, due to the Triton X-100 extractability of some of the ezrin constructs. Cells processed for the determination of ERM C-terminal threonine phosphorylation were first fixed in 10% ice-cold trichloroacetic acid (TCA) as described (Matsui et al., 1998). Primary antibodies were diluted 1:200 in PBS containing 1% BSA, except for the fluorophore-conjugated antibodies that were used at 1:20–1:50. The Cy5-labelled anti-rat conjugate was obtained from Jackson ImmunoResearch Laboratories. Confocal images were acquired on a confocal laser scanning microscope (model LSM 510, Carl Zeiss Inc.) equipped with both 40×/1.3Plan-Neofluar and 63×/1.4Plan-Apochromat oil immersion objectives. Each image represents a two-dimensional projection of 2–3 slices in the Z-series, taken across the mid-depth of the cell at 0.2 μm intervals.
FRET determination by FLIM measurements
A detailed description of the FLIM apparatus used for FRET determination in this work can be found elsewhere (Squire and Bastiaens, 1999). The lifetime instrument performs phase- and modulation-based imaging fluorimetry by microscopy. All images were taken using a Zeiss Plan-Apochromat 100×/1.4NA phase 3 oil objective and the homodyne phase-sensitive images recorded at a modulation frequency of 80.224 MHz. The donor (GFP–PKCα) was excited using the 488 nm line of an argon–krypton laser and the resultant fluorescence separated using a combination of dichroic beamsplitter (Q 505 LP; Chroma Technology Corp.) and narrow band emitter filter (BP514/10; Lys and Optik). Acceptor images (Cy3-conjugated IgG or Fabs) were recorded using a 100 W mercury arc lamp (Zeiss Attoarc) as a source of sample illumination combined with a high Q Cy3 filter set (exciter: HQ 535/50, dichroic: Q 565 LP, emitter: HQ 610/75 LP; Chroma Technology Corp.). Global analysis was performed to calculate the precise fraction of donor molecules that have a reduced lifetime due to FRET (i.e. the fraction of GFP–PKCα RD + V3 bound to ezrin) as described (Verveer et al., 2000a,b).
Cellular fractionation by sucrose gradient centrifugation
GFP–PKCα-transfected MCF-7 cells were scraped from plates, washed and homogenized using a cell cracker in 0.25 M sucrose, 10 mM HEPES–KOH pH 7.2 containing 1 mM EDTA and 1 mM magnesium acetate. The post-nuclear supernatant was loaded onto a velocity gradient as described (Dittie et al., 1997). PKCα and β1 integrin were enriched in VG fractions 2 and 3 (T.Ng and P.J.Parker, unpublished data). EG fractions were subsequently prepared from these VG fractions as described (Dittie et al., 1997). Proteins were precipitated from these fractions in 10% TCA overnight, washed with ethanol/acetone (50/50), then denatured in Laemmli's sample buffer, separated on an 8% poly acrylamide gel under reducing conditions and transferred electrophoretically to a polyvinylidene difluoride (PVDF) membrane. Incubation with primary antibodies was performed overnight at 4°C. Detection was with enhanced chemiluminescence (ECL; Amersham) according to recommended procedures.
Immunoprecipitation and western blotting
GFP-tagged or endogenous PKCα was immunoprecipitated from whole-cell lysates (each derived from an 80% confluent cell monolayer in a 15 cm round tissue culture dish, pre-treated with 1 μM PDBu for 20 min) using 10–20 μg of protein G-coupled MC5 as described previously (Kanner et al., 1989), except that 1% (w/v) n-octyl β-D-glucopyranoside was used instead of NP-40. The samples were separated on large 8% resolving polyacrylamide gels. Blots were then probed with a rabbit IgG that recognizes anti-CPERM (in 0.1% TBS-Tween containing 3% BSA) followed by an anti-rabbit-horseradish peroxidase (HRP) conjugate (Amersham). The same membrane was stripped and reprobed (according to procedures recommended by Amersham) with an affinity-purified rabbit anti-ezrin IgG, then subsequently with an anti-PKCα C-terminal polyclonal serum. Incubation with primary antibodies was performed overnight at 4°C. Detection was with ECL according to recommended procedures.
Immunogold electron microscopy
Cells were scraped carefully and fixed in 4% paraformaldehyde for 1 h, before being processed for routine sectioning on a Leica ultra-cryotome for immunolabelling. Single and double labelling was carried out as described previously (Slot and Geuze, 1985). In the double labelling, the first antibody used was an affinity-purified rabbit polyclonal anti-ezrin IgG (cross-adsorbed against moesin and radixin) recognized by 10 nm PAG; this was followed, after an additional glutaraldehyde fixation step (1% monomeric glutaraldehyde for 5 min to cross-link the primary antibody), by a rabbit anti-GFP antiserum (a kind gift of D.Shima, ICRF) detected with 5 nm PAG. In some studies, the second antibody was omitted to control for cross-reactivity of the 5 nm PAG with the first antibody. After antibody labelling, sections were examined using a Jeol 1010 microscope.
In vitro kinase assays with recombinant or PKCα-co-precipitated endogenous ezrin
The potential phosphorylation of GST-tagged full-length ezrin by PKCα was examined in an in vitro32P incorporation assay where 5 μl (0.27 μg) of human recombinant PKCα (PanVera Corporation) were put in a 35 μl reaction buffer containing 5 μl (1.7 μg) of GST–ezrin, 5 μl (0.8 mM) of ATP containing 1 μCi of [32P]ATP (255 c.p.m./pmol), 5 μl each of MgCl2 (0.1 M) and CaCl2 (6 mM), 10 μl of a pure lipid sonicated suspension of 50 μg of phosphatidylserine, 5 μg of phosphatidylinositol 4,5-bisphosphate and 0.05 μg of TPA in 0.2 M HEPES pH 7.5/2 mM EGTA. The control reaction was carried out in the absence of lipids/TPA. The in vitro phosphorylation reactions were incubated at 30°C and stopped after various time points by adding 40 μl of modified Laemmli's sample buffer (containing final concentrations of 2% SDS, 4 M urea and 10 mM EDTA). The samples were then separated on an 8% polyacrylamide gel and the phosphorylated proteins were detected by autoradiography. The in vitro phosphorylation assay was also performed with protein G–agarose bead-bound, PKCα-co-precipitated endogenous ERM (see the legend of Figure 7B for experimental details).
We wish to thank Dr Gipi Schiavo and Professor Ian Hart (ICRF) for their valuable comments. 2C4 fibrosarcoma cells stably expressing GFP–PKCα (GFP–PKCα-2C4) or GFP control vector were kindly provided by B.Lillemeier. We also thank Professor Clare Isacke (Imperial College, London) and Professor Paul Mangeat (Universite Montpellier II, Montpellier) for kindly providing the GST-tagged ezrin plasmid and crude sera against radixin and moesin, respectively. We thank Christine T.Quang for many helpful discussions. This study was supported by the Imperial Cancer Research Fund (UK) and the UK Medical Research Council (in the form of a Clinician Scientist Grant awarded to T.N.).