L-plastin, a malignant transformation-associated protein, is a member of a large family of actin filament cross-linkers. Here, we analysed how phosphorylation of L-plastin on Ser5 of the headpiece domain regulates its intracellular distribution and its interaction with F-actin in transfected cells and in in vitro assays. Phosphorylated wild-type L-plastin localised to the actin cytoskeleton in transfected Vero cells. Ser5Ala substitution reduced the capacity of L-plastin to localise with peripheral actin-rich membrane protrusions. Conversely, a Ser5Glu variant mimicking a constitutively phosphorylated state, accumulated in actin-rich regions and promoted the formation of F-actin microspikes in two cell lines. Similar to phosphorylated wild-type L-plastin, this variant remained associated with cellular F-actin in detergent-treated cells, whereas the Ser5Ala variant was almost completely extracted. When compared with non-phosphorylated protein, phosphorylated L-plastin and the Ser5Glu variant bound F-actin more efficiently in an in vitro assay. Importantly, expression of L-plastin elicited collagen invasion in HEK293T cells, in a manner dependent on Ser5 phosphorylation. Based on our findings, we propose that conversely to other calponin homology (CH)-domain family members, phosphorylation of L-plastin switches the protein from a low-activity to a high-activity state. Phosphorylated L-plastin might act as an integrator of signals controlling the assembly of the actin cytoskeleton and cell motility in a 3D-space.
Proteins capable of cross-linking actin filaments into networks or bundles play an important role in the assembly of the cellular actin cytoskeleton (Bartles, 2000). In addition, alteration of their functions contributes to pathologies such as cancer, where structural and functional modifications of the actin cytoskeleton accompany uncontrolled cell movement and signalling (Giganti and Friederich, 2003). Regulation of the activities of these proteins by second messengers and post-translational modifications such as phosphorylation allows them to act in a coordinated manner and to build-up specific actin structures in cells.
Initially detected in leucocytes (Matsushima et al., 1988) and cancer cells derived from solid tissues (Goldstein et al., 1985; Lapillonne et al., 2000; Lin et al., 1993a; Lin et al., 1993b; Park et al., 1994) L-plastin was identified as a closely related isoform of intestinal brush-border fimbrin (de Arruda et al., 1990) and is sometimes referred to as L-fimbrin. Several lines of evidence support an important role for L-plastin in the assembly of the actin cytoskeleton. L-plastin is a member of a large family of actin-cross-linking proteins, including α-actinin and filamin (Stossel et al., 2001). Sharing a common actin-binding domain that is made up by two calponin homology (CH)-domains, members of this family control the assembly of most of the actin networks in cells (Gimona et al., 2002). L-plastin and the other plastin isoforms exhibit a modular molecular organisation (de Arruda et al., 1990; Gimona et al., 2002) and similar in vitro activities. The N-terminal calmodulin-like headpiece domain comprises two helix-loop-helix EF-hand Ca2+-binding motifs, followed by two independent actin-binding domains (ABDs) within the same molecule. Cross-linking activity of the other family members requires dimerisation of two polypeptide chains comprising each an ABD. Conversely, the spatially close ABDs (120 Å) of plastins enable them to organise actin-filaments into tight bundles (Bretscher, 1981).
In cells, L-plastin localises to actin-rich membrane structures involved in locomotion, adhesion and immune defence, including cell-adhesions, immune complexes, filopodia and the phagocytic cup, supporting a role in the organisation of the actin cytoskeleton (Babb et al., 1997; Jones and Brown, 1996; Rosales et al., 1994; Samstag et al., 2003). In line with such a role, overexpression of L-plastin in transfected fibroblast-like CV1 cells induced the rearrangement of focal-adhesion-associated actin filaments (Arpin et al., 1994). L-plastin is also detected in solid tumours of epithelial and mesenchymal origin and was proposed to play a role in tumour cell invasion (Zheng et al., 1999).
Among the three plastin isoforms, L-plastin has the unique property of being phosphorylated in response to signals triggering the activation of the immune response, cell migration and proliferation. L-plastin is phosphorylated on residues Ser5 and Ser7 in hematopoietic cells in vivo, but most likely on Ser5 exclusively in non-hematopoietic cells (Lin et al., 1998; Messier et al., 1993; Shinomiya et al., 1995). Whereas in vivo and in vitro phosphorylation of Ser5 by cAMP-dependent protein kinase A (PKA) is well documented (Lin et al., 1998; Wang and Brown, 1999), there is still controversy over which kinase(s) phosphorylate(s) Ser7.
Importantly, despite the complexity of signalling cascades leading to phosphorylation of L-plastin in immune cells, phosphorylation of this protein occurs upstream of cytoskeleton rearrangements that underly processes such as chemotaxis (Paclet et al., 2004), T cell activation (Henning et al., 1994) or adhesion (Jones and Brown, 1996), suggesting a role for phosphorylated L-plastin in the assembly of actin structures. In support of such a possibility, phosphorylated L-plastin localises to the detergent-insoluble cytoskeletal fraction in adherent macrophages (Messier et al., 1993). However, although it was suggested that phosphorylation of L-plastin regulates its F-actin-binding properties and thus its biological function, experimental support for such a mechanism of regulation is still lacking.
Here, we combined cell biology and biochemical approaches to address the question how phosphorylation regulates the interaction of L-plastin with F-actin. We obtained the first direct evidence that Ser5 phosphorylation of L-plastin promotes its targeting to the actin cytoskeleton and increases its filament bundling activity in cells and in vitro. Importantly, ectopic expression of L-plastin endows epithelial cells with invasive properties in an in vitro collagen invasion assay and this effect depends on Ser5 phosphorylation, suggesting that Ser5 phosphorylated L-plastin contributes to the regulation of cell motility in a 3D-space.
Characterisation of an antibody against L-plastin phosphorylated on Ser5 (Ser5-P) and expression of wild-type L-plastin or phosphorylation variants in Vero cells.
Because pharmacological agents such as protein-kinase activators or inhibitors have broad effects on the organisation of the actin cytoskeleton (Howe, 2004), we decided to use a mutational approach to investigate how phosphorylation regulates the biological and biochemical properties of L-plastin (Fig. 1A). We replaced Ser5, its major phosphorylation site that is located in its regulatory headpiece domain (Lin et al., 1998; Shinomiya et al., 1995; Wang and Brown, 1999), with Ala (L-plastin Ser5Ala) or Glu (L-plastin Ser5Glu) to inactivate this site or to mimic constitutive phosphorylation, respectively. Although negatively charged residues and the phosphate group are chemically not identical, it is well documented that this approach can yield valuable information on the biological role of protein phosphorylation (Clarke et al., 2004; Gautreau et al., 2000; Huttelmaier et al., 1999).
To analyse phosphorylated wild-type L-plastin in cells, we raised an antibody against a synthetic phospho-peptide encompassing the phosphorylation site of L-plastin and harbouring a phosphate group on Ser5 (anti-Ser5-P antibody). To test the specificity of this antibody, we first took advantage of the observation that protein kinase A (PKA) exclusively phosphorylates wild-type L-plastin on Ser5 in vitro (Jones et al., 1998; Wang and Brown, 1999). Immunoblotting with anti-Ser5-P of in-vitro-phosphorylated recombinant L-plastin, revealed a time-dependent increase in signals (Fig. 1B, upper panel) whereas signals remained constant when the same samples were analysed with anti-L-plastin antibody (B, lower panel). Anti-Ser5-P did not react with Ser5Ala L-plastin incubated in the presence of PKA (B, upper panel, L-plastin S/A).
We used Jurkat T lymphoid cells to determine whether this antibody also allowed us to distinguish between the phosphorylated and non-phosphorylated forms of L-plastin in cell extracts. Whereas in non-stimulated Jurkat cells base-line phosphorylation of L-plastin is barely detectable, activation of PKA greatly increases phosphorylation on Ser5 (Jones et al., 1998; Wang and Brown, 1999). Jurkat cells expressed high amounts of endogenous L-plastin - as revealed with the anti-L-plastin antibody (Fig. 1C, lower panel). A faint signal was detected with anti-Ser5-P antibody in non-stimulated cells that greatly increased when cells were stimulated with the PKA activators forskolin or 8-Bromo-cAMP (Fig. 1C, upper panel). Conversely, no protein band was detected when cells were pre-incubated with a PKA inhibitor H-89 prior to stimulation (Fig. 1C, upper panel). In all cases, the total amount of L-plastin did not vary (Fig. 1C, lower panel). This result clearly showed that the anti-Ser5-P antibody specifically reacted in vivo with L-plastin phosphorylated at Ser5 but not with the non-phosphorylated form.
We first used fibroblast-like Vero cells, which exhibit a well-organised actin cytoskeleton, as a model system to study the role of L-plastin phosphorylation on Ser5 in transfection experiments. Notably, Vero cells do not produce L-plastin, as determined with an anti-L-plastin antibody (data not shown). Vero cells were transfected with wild-type L-plastin or L-plastin Ser5-substitution variants epitope-tagged with the 13 C-terminal residues of the Sendai virus L protein (sv-tag). Expression and phosphorylation of wild-type and mutants was then monitored by immunoblotting of the different cell lysates (Fig. 1D). Similar amounts of wild-type L-plastin or of its Ser5-substitution variants were produced in Vero cells (Fig. 1D, upper panel). However, probing the same cell lysates with anti-Ser5-P antibody, revealed a band in wild-type-transfected cells but not in those transfected with the Ser5-substitution variants (Fig. 1D, lower panel). This result further highlighted the specificity of the anti-Ser5-P antibody and showed that L-plastin was phosphorylated on Ser5 in Vero cells, as previously reported for related cell types (Lin et al., 1998).
L-plastin phosphorylated on Ser5 localises to the actin cytoskeleton in Vero cells
We first investigated the intracellular distribution of L-plastin phosphorylated on Ser5 (Ser5-P-plastin) in transfected Vero cells (Fig. 2). Immunofluorescence analysis with anti-Ser5-P antibody and Rhodamine-phalloidin as a probe for F-actin revealed that most of Ser5-P-plastin localised to F-actin-rich structures, including membrane ruffles and microspikes found at the cell periphery (Fig. 2A, upper panels). In addition, Ser5-P-plastin also weakly decorated stress fibres, frequently located in the centre of the cell. Co-staining of cells expressing L-plastin by using anti-Ser5-P antibody and an antibody against the sv-tag as probes for the total pool of L-plastin, revealed similar but not identical staining patterns (Fig. 2A, lower panels). In addition to the cytoskeleton-associated pool of L-plastin that was also detected by the anti-Ser5-P antibody, the anti-sv-tag antibody revealed a more intense cytoplasmic staining, suggesting that most of cytosplasmic L-plastin was not phosphorylated (Fig. 2A, left panel). The anti-Ser5-P antibody did not react with non-phosphorylable L-plastin containing the Ser5Ala mutation, confirming the specificity of the staining (Fig. 2B).
Intracellular localisation of L-plastin Ser5-phosphorylation variants.
To examine whether Ser5 phosphorylation regulates the interaction of L-plastin with the cellular actin cytoskeleton, we compared the intracellular distribution of L-plastin phosphorylation variants in transfected Vero cells (Fig. 3). Wild-type L-plastin co-distributed with F-actin in peripheral microspikes and membrane protrusions and also decorated the proximal, membrane-anchored end of stress fibres (Fig. 3A, upper panels). By marked contrast, L-plastin Ser5Ala (Fig. 3A, middle panels) exhibited little co-distribution with these F-actin structures and a diffuse cytoplasmic staining. Notably, L-plastin Ser5Ala-Ser7Ala, a variant in which both potential phosphorylation sites were invalidated, yielded a similar phenotype (data not shown). Conversely, L-plastin Ser5Glu (Fig. 3A, lower panels) predominantly localised to peripheral membrane protrusions and the number of spike-like actin structures detectable in these cells was increased sixfold when compared with cells producing the Ser-Ala variant. In addition, similar to wild-type L-plastin, the Ser5Glu variant decorated a subset of stress fibers, mostly located in the centre of the cell.
The dot-like staining pattern observed with L-plastin Ser5Glu was reminiscent of that of cell-matrix adhesions. Co-staining of cells with vinculin as a marker for focal adhesions, showed that wild-type L-plastin and the Ser5Glu variant strongly colocalised with the distal, actin-associated portion of these structures. Of note, this staining pattern was most prominent in the cell centre (Fig. 3B, upper and lower rows). By contrast, Ser5Ala only weakly co-distributed with vinculin (B, middle row).
The prominent label in peripheral membrane extensions suggested that phosphorylated L-plastin and the Ser5Glu variant might target to sites of actin assembly (Pollard and Borisy, 2003). To confirm this, Vero cells expressing wild-type L-plastin or its phosphorylation variants were co-stained for the Arp2/3-complex-binding protein cortactin (Weed et al., 2000), used here as a marker for sites of actin polymerisation (Fig. 4). Cortactin concentrates in peripheral membrane extensions of migrating cells, such as lamellipodia (Bryce et al., 2005) and podosomes (Mizutani et al., 2002), and ventral membrane protrusions found in various cell types, including invasive cancer cells (Linder and Aepfelbacher, 2003). Cortactin co-distributed with Ser5Glu L-plastin (Fig. 4, upper panels) in peripheral plasma membrane extensions but not in dot-like structures. Triple staining of cells expressing wild-type L-plastin revealed that also Ser5-P-plastin co-distributed with cortactin in actin-rich plasma membrane extensions (Fig. 4, middle panels), but to a lesser extend when compared with the Ser5Glu variant. Conversely, Ser5Ala L-plastin (Fig. 4, lower panels) showed no significant co-distribution with cortactin. These results suggested that part of the phosphorylated pool of L-plastin localised to sites of Arp2/3-complex-mediated actin assembly. Although phosphorylated L-plastin was previously shown to co-distribute with vimentin in macrophages (Correia et al., 1999), we were unable to detect a co-distribution of both proteins in Vero cells (supplementary material Fig. S1).
To confirm that Ser5Glu stabilised peripheral actin-structures, we next analysed the effects of phosphorylation variants in LLCPK1 epithelial kidney cells that were previously used to assess the role of F-actin bundling proteins in the assembly of the cortical actin cytoskeleton (Arpin et al., 1994; Friederich et al., 1989; Loomis et al., 2003). In addition to stress fibres and micro-spikes present at their basal surface, these cells harbour numerous spike-like actin structures at their apical face, corresponding to the core bundles of microvilli. Elongation of these structures caused by forced expression of bundling proteins can be easily visualised by epifluorescence microscopy. A focal plane on the apical surface of these cells revealed that Ser5Glu mutants strongly colocalised with the F-actin bundles of surface microvilli that were more prominent than those detected in non-transfected neighbouring cells (Fig. 5, upper and lower panels). By marked contrast, the Ser5Ala variant exhibited a diffuse cytoplasmic distribution and its association with these surface projections was barely detectable (Fig. 5, middle panels). This result suggest that Ser5Glu L-plastin stabilises and/or bundles actin filaments of plasma membrane protrusions.
L-plastin phosphorylation on Ser5 increases its avidity for cellular F-actin.
Immunolocalisation studies with wild-type L-plastin and its phosphorylation variants suggested that the phosphorylated form bound more tightly to the actin cytoskeleton. To test this, we treated cells with Triton X-100, a non-ionic detergent, prior to fixation and processing for immunofluorescence staining. This approach allows to evaluate of the avidity of actin-binding proteins for different F-actin structures in the context of the cell (Arpin et al., 1994; Friederich et al., 1992). Whereas membrane and cytosolic proteins are extracted, proteins that are tightly associated with the actin cytoskeleton remain bound. Wild-type L-plastin remained associated with F-actin structures in detergent-extracted transfected Vero cells (Fig. 6A, upper panels), as previously reported (Arpin et al., 1994; Friederich et al., 1992). Like wild-type L-plastin, the Ser5Glu variant remained associated with actin cytoskeleton (Fig. 6A, lower panels). Conversely, L-plastin Ser5Ala was almost totally extracted. In a few cells, we detected residual labelling, restricted to the ends of stress fibres and peripheral microspikes (Fig. 6A, middle panels). In line with this qualitative result, the number of cells exhibiting detectable cytoskeleton-associated Ser5Ala staining decreased on average by 83% after detergent-extraction, wheras detectable signals decreased only by 50% for cells transfected with wild-type or Ser5Glu L-plastin (Fig. 6B). These results suggest that the Ser5Ala variant has a reduced avidity for cellular F-actin. The detergent-resistant pool of wild-type L-plastin was phosphorylated on Ser5 (Fig. 6C).
Ser5 phosphorylation increases the F-actin-binding activity of L-plastin in vitro.
Taken together, these results indicate that Ser5 phosphorylation of L-plastin positively regulates its association with F-actin. Therefore, we next compared F-actin-binding properties of non-phosphorylated and PKA-phosphorylated L-plastin in a co-sedimentation assay with F-actin in vitro (Fig. 7A). At a L-plastin to actin ratio of 1:2, almost twice as much phosphorylated L-plastin co-sedimented with F-actin compared with the non-phosphorylated form (Fig. 7B). Immunoblotting analysis with anti-Ser5-P confirmed that L-plastin in pellets was phosphorylated on Ser5 (Fig. 7C). As expected, Ser5-P signals increased as a function of L-plastin concentration, correlated with the increase of L-plastin in the pellets (Fig. 7C). Similar to non-phosphorylated wild-type L-plastin, also the Ser5Ala variant bound F-actin in vitro (Fig. 7B and supplementary material Fig. S2), suggesting that the Ser5 Ala substitution did not grossly affects F-actin binding.
In vitro phosphorylation of wild-type protein is prone to experimental variations and the extend to which the protein is phosphorylated is difficult to quantify. Therefore, we decided to use the Ser5Glu variant in further in vitro studies. In agreement with previous results (Namba et al., 1992), non-phosphorylated wild-type L-plastin bound F-actin in a dose-dependent manner (Fig. 7A), and the data fitted a hyperbolic function (Fig. 7B). Interestingly, the Ser5Glu variant exhibited a different binding curve and distinct saturation behaviour compared with non-phosphorylated wild type (Fig. 7B). Whereas similar amounts of wild-type L-plastin and the Ser5Glu variant bound at low molar ratios of plastin to actin, almost twice as much Ser5Glu variant bound at a high plastin to actin ratio (1:2), and F-actin binding saturation was not reached for this variant.
Bundling activity of L-plastin was assessed in a low-speed centrifugation assay allowing to sediment F-actin bundles but not single filaments (Glenney, Jr et al., 1981). Importantly, Ser5Glu L-plastin had markedly higher bundling activity (Fig. 8), even at low plastin to actin ratios, where no differences in F-actin binding was observed (Fig. 8B). At a molar ratio of 1:8, 47% of F-actin sedimented when non-phosphorylated L-plastin was added, whereas 84% of F-actin pelleted in the presence of the Ser5Glu variant (Fig. 8B). Formation of F-actin bundles was confirmed by transmission electron microscopy (Fig. 8C). Whereas only a few F-actin bundles were detected in the presence of non-phosphorylated L-plastin (Fig. 8C, left panel), F-actin preparations incubated with Ser5Glu variant exhibited a large number of tight F-actin bundles (Fig. 8C, right panel). Because Ser5 is spatially close to the EF-hand Ca2+-binding motif of the headpiece domain, we wanted to determine whether the presence of a negatively charged group at amino acid position 5 interfered with the inhibitory effect of Ca2+ (Namba et al., 1992). In pellets, an almost identical decrease in F-actin was observed at low-speed centrifugation when non-phosphorylated wild-type L-plastin or the Ser5Glu variant were incubated in the presence of increasing concentrations of free Ca2+ (Fig. 8D), suggesting that addition of a negative charge in amino acid position 5 does not influence Ca2+-dependency of F-actin bundling.
Since Ser5-P-plastin is targeted to peripheral membrane protrusions and focal adhesions known to be involved in cell migration, we tested whether L-plastin affects cell migration in a phosphorylation-dependent manner. Since L-plastin has been associated with cell invasion (Zheng et al., 1999), we expressed wild-type L-plastin or phosphorylation variants in human embryonic kidney HEK293T cells that do not produce endogenous L-plastin (data not shown) and investigated their invasive capacity in an in vitro collagen assay (Bracke et al., 2001; De Corte et al., 2002). This semi-quantitative assay measures to what extent cells can migrate through a 3D-collagen gel and mimic invasion of tumour cells in the stroma. Immunoblotting analysis of transfected HEK293T cells revealed that similar amounts of L-plastin variants were produced (Fig. 9A). Untransfected wild-type HEK293T cells are not invasive, in contrast to DHD-FIB rat colon myofibroblasts that were used as a positive control (Fig. 9B). We observed that expression of the wild-type L-plastin conferred invasive properties to HEK293T cells in this assay. Importantly, cells producing the Ser5Ala variant behaved like non-transfected control cells. Conversely, similar to wild-type L-plastin, the Ser5Glu variant endowed cells with invasive properties (Fig. 9B). These findings suggested that L-plastin phosphorylation at Ser5 is a prerequisite to elicit collagen invasion. Invasion of epithelial cells is associated in part with loss of cell-cell adhesion contacts. To investigate the effects of wild-type L-plastin and variants on cell-cell contact formation, we used a fast cell aggregation assay in which cells are cultured in suspension and their capacity to form aggregates is evaluated (Boterberg et al., 2001; De Corte et al., 2002). Non-invasive cells are expected to form larger aggregates than invasive cells. HEK293T cells transfected with wild-type L-plastin or the Ser5Glu variant showed drastic reduction in particle size (∼10 to 100 μm) at 30 minutes of culture (Fig. 9C) to a size comparable to mock-transfected cells at time zero (see arrow Fig. 9C). The particle diameter of HEK293T cells transfected with Ser5Ala or Ser5Ala-Ser7Ala at 30 minutes of culture was much larger (∼100 to 1500 μm) and comparable to that of mock-transfected (MT) cells at 30 minutes. Conversely to the collagen invasion assay, cells transfected with L-plastin variants behaved similar when tested in a migration assay on plastic dishes. HEK293T cells were grown to confluency and after 48 hours, a wound was made by scratching a line in the cell monolayer. When analysing the capacity of cells to migrate into the wounds, no significant difference could be measured for cells transfected with the various constructs (Fig. 9D).
Because L-plastin is a well-characterised, direct downstream target of PKA in hematopoietic cells (Wang and Brown, 1999), we determined whether PKA inhibitors affected L-plastin-dependent collagen invasion. Whereas PKA inhibition strongly reduced L-plastin-dependent cell invasion, PKC inhibitors did not affect this process (Fig. 9E). To confirm that PKA indeed phosphorylated wild-type L-plastin in HEK293T cells, L-plastin-transfected cells were treated with the cell-permeable PKA activator 8-Bromo-cAMP (Fig. 9F). When compared with untreated cells, 8-Bromo-cAMP stimulated L-plastin phosphorylation, as monitored by the threefold increase of the signal detected with anti-Ser5-P antibody (Fig. 9F). Incubation of cells in the presence of PKA inhibitor prior to 8-Bromo-cAMP addition inhibited this stimulation (Fig. 9F), without, however, inhibiting base-line phosphorylation of L-plastin.
Taken together, these results show that phosphorylated L-plastin affects cell migration in a 3D- but not in a 2D-space and that PKA-elicited signalling is required for L-plastin-mediated invasion.
L-plastin, a major F-actin-bundling protein of leucocytes and epithelial- or mesenchymal-derived cancer cells, is phosphorylated upon activation of pathways triggering adhesion and migration, concomitantly to remodelling the actin (Henning et al., 1994; Matsushima et al., 1988; Shibata et al., 1993; Wang and Brown, 1999). Here, we gained consistent evidence that phosphorylation of L-plastin positively regulates its F-actin-binding and cross-linking activities and that it might function as a regulatory switch in the assembly of actin-rich structures involved in cell migration and signalling. Importantly, phosphorylated L-plastin might contribute to the regulation of the migratory behaviour of cells in a 3D-space.
In line with previous data (Lin et al., 1998), we found that L-plastin is phosphorylated on Ser5 in fibroblast-like Vero cells and epithelial HEK293T cells. Conversely to Jurkat T lymphoid cells, Vero and HEK293T cells exhibited high-base-line phosphorylation of L-plastin on Ser5. Whereas phosphorylation of L-plastin is triggered by signals activating immune cells (Henning et al., 1994; Jones and Brown, 1996; Matsushima et al., 1988; Paclet et al., 2004), pathways controlling L-plastin phosphorylation may be constitutively activated in adherent fibroblasts and epithelial cells (Lin et al., 1998). Consistently, constitutive phosphorylation was observed in activated, adherent macrophages and leucocytes (Jones and Brown, 1996; Messier et al., 1993). Phosphorylated L-plastin preferentially localised to F-actin structures, including peripheral microspikes and membrane extensions, focal adhesions and, to a lesser extent, stress fibres. This result matches previous biochemical data, revealing phosphorylated L-plastin in the detergent-resistant fraction of adherent macrophages (Messier et al., 1993). Phosphorylation on Ser residues was proposed to be required for targeting L-plastin to the actin cytoskeleton (Messier et al., 1993). We found that also non-phosphorylated L-plastin weakly colocalised with actin-rich structures. However, by marked contrast, the Ser5Glu variant highly concentrated in actin-rich plasma membrane extensions, even when expressed at low levels, whereas the Ser5Ala variant exhibited a diffuse staining, reflecting a large pool of unbound protein. This difference in staining pattern was neither due to increased degradation of L-plastin containing the Ser5Ala mutation, as monitored by immunoblotting analysis, nor to impaired F-actin binding, as suggested by our in vitro data.
Taken together, our data suggest that phosphorylation of L-plastin increases its avidity for F-actin structures in cells - as also supported by our detergent-extraction experiments. However, we cannot exclude that targeting of phosphorylated L-plastin is mediated in part by other protein partners - as demonstrated for other CH-domain family members such as ABP 280 (Goldmann, 2001). Alternatively, phosphorylation-dependent binding of L-plastin to a protein ligand may increase its actin-binding or bundling activity, consequently influencing its intracellular location. In support of such a possibility, Iba1, a Ca2+-binding protein, increases the bundling activity of L-plastin in vitro (Kanazawa et al., 2002; Ohsawa et al., 2004). Although these mechanisms are not mutually exclusive, we favour a mechanism by which phosphorylation influences targeting of L-plastin by promoting its F-actin-binding property, as supported by our finding that Ser5 phosphorylation or Ser5Glu substitution increases binding to F-actin in vitro. This result also demonstrates that negatively charged residues can mimic phosphorylation, without excluding, however, that behaviours of the Ser5Glu variant and of phosphorylated wild-type L-plastin might be similar but not identical. The non-hydrolysable negatively charged glutamate residue might keep the Ser5Glu variant in a locked state on the filament, thereby creating the more pronounced phenotype we observed with this mutant in transfected cells. Conversely, switching the wild type from a phosphorylated and to a non-phosphorylated state might increase its off-rate.
In most of the other F-actin binding proteins, phosphorylation sites are located at an F-actin interface where the negatively charged phosphate group negatively interferes with actin binding (Azim et al., 1995; Harbeck et al., 2000; Izaguirre et al., 2001). By contrast, Ser5 phosphorylation of L-plastin occurs in the regulatory Ca2+-binding headpiece-domain and promotes F-actin-binding and -bundling. Although this domain is not required for filament cross-linking, it yet increases the stiffness of actin bundles (Klein et al., 2004). Based on our data and structural information (Klein et al., 2004; Volkmann et al., 2001), Ser5 phosphorylation might induce a conformational change in the headpiece domain that affects the interaction of the actin-binding domains with actin. An increase in bundling activity was observed at low L-plastin to actin ratios where amounts of the bound non-phosphorylated L-plastin or the Ser5Glu variant were similar, suggesting that the Ser5Glu variant more efficiently promoted cross-link formation. Notably, the binding curve of Ser5Glu to F-actin did not fit a first-order reaction indicating the existence of different L-plastin or F-actin binding states. Our results do not favour a mechanism where phosphorylation would influence the Ca2+-sensitivity of the F-actin bundling activity of L-plastin. Taken together, these results suggested that, although not an absolute requirement for L-plastin binding to F-actin, phosphorylation on Ser5 promotes binding and, thereby, targeting to actin-rich structures in cells.
Using a collagen invasion assay, we found that expression of L-plastin endows HEK293T cells with invasive properties. This is consistent with previous data showing that L-plastin repression by an RNA antisense approach decreased the invasive capacity of prostate carcinoma cells (Zheng et al., 1999). Cell invasion depended on Ser5 phosphorylation, suggesting that targeting of L-plastin to peripheral actin-rich membrane extensions and adhesions was required for this process. Cell migration in a 2D-space relies on coordinated cycles of cell extension, cell adhesion and actin-myosin contraction (Friedl, 2004). The fact that neither wild-type L-plastin nor L-plastin Ser5Glu increased the rate of cell migration in a wound healing assay, suggested that forced expression of these proteins does not grossly affect these processes. Because of the high transient-transfection rate we obtained in HEK293T cells (60-70%), it is unlikely that this result is caused by the low percentage of transfected cells present at the wound borders. In contrast to 2D cell movement, migration through a collagen matrix frequently involves matrix degradation by specific cell-surface proteases (Friedl, 2004). In macrophages, phosphorylated L-plastin localises to podosomes (Babb et al., 1997; Evans et al., 2003; Messier et al., 1993), plasma membrane protrusions involved in the regulation of adhesion, cell surface proteases (Buccione et al., 2004) and cancer cell invasion (Linder and Aepfelbacher, 2003). Phosphorylated L-plastin or the Ser5Glu variant did not colocalise with cortactin in podosome-like adhesion structures, yet both proteins co-distributed in membrane extensions and peripheral microspike adhesions where actin assembly occurs (Schirenbeck et al., 2005; Svitkina et al., 2003). Based on our in vitro F-actin-binding data and the observation that L-plastin Ser5Glu but not L-plastin Ser5Ala promoted formation of F-actin spikes, it is tempting to speculate that phosphorylated L-plastin might, by stabilising actin filaments, affect the assembly of membrane specialisations implicated in invasion. In line, the closely related T-plastin isoform promotes Arp2/3-mediated actin assembly and force generation in a biomimetic actin-based motility assay (Giganti et al., 2005). Alternatively, phosphorylated L-plastin might act as a scaffold for cytoskeleton-associated signalling complexes. In support of such a mechanism, L-plastin-deficient (L-plastin-/-) polynuclear monocytes (Chen et al., 2003) exhibited a marked reduction in phosphorylation of paxillin, an cortactin-interacting protein (Bowden et al., 1999) and Syk, a Src-kinase family member (Frame et al., 2002), signalling proteins involved in actin remodelling and cell invasion. These potential mechanisms might also account for the effect of L-plastin on cell cohesion, observed in the fast cell aggregation assay. Notably, filamin, a CH-domain actin-membrane linker protein, is required for the concentration and release of the protease epithin at cell-cell contacts (Kim et al., 2005).
PKA but not PKC inhibitors blocked L-plastin-induced collagen invasion. However, albeit PKA-mediated phosphorylation of L-plastin could be stimulated in HEK293T cells, we were unable to inhibit base-line phosphorylation. Several, non-exclusive explanations might account for these apparently discrepant results. Turnover of phosphate in L-plastin might be rather slow and longer incubation times with inhibitors, as in invasion assays, might be required to yield non-phosphorylated L-plastin. Alternatively, other, so far unknown kinases might phosphorylate L-plastin on Ser5 in HEK293T cells. Third, inhibition of PKA does affect invasion by targeting other phospho-proteins, as supported by previous studies (Howe, 2004). Further experiments are required to dissect signalling pathways linking L-plastin phosphorylation to invasion.
L-plastin is, so far, the only of the three plastin isoforms that is phosphorylated in cells (de Arruda et al., 1990; Lin et al., 1998). Unlike I-plastin and T-plastin, which are associated with relatively stable structures, such as the brush border microvilli or the stereocilia of the inner ear (Bretscher, 1981; Sobin and Flock, 1983), L-plastin might, during evolution, have acquired a specific regulatory system coupling the activity state of the protein to signalling cascades triggering events that require the rearrangement of the actin cytoskeleton (Henning et al., 1994; Jones and Brown, 1996; Matsushima et al., 1988; Shibata et al., 1993). Notably, plastins/fimbrins of various organisms harbour negatively charged residues in their N-terminal part that might have a function similar to phosphorylated Ser5 in L-plastin.
To conclude, phosphorylation of L-plastin increases its F-actin-binding affinity and is required for its efficient targeting to sites of actin assembly at the plasma membrane and for eliciting cell invasion. Our findings pave the way for further investigations of how phosphorylated L-plastin contributes to cell migration in a 3D-space during tumour cell invasion and embryonic development.
Materials and Methods
Vero monkey kidney, epithelial HEK293T or LLCPK1 cells were grown in Dulbecco's modified Eagle's medium. Jurkat T lymphoid cells were maintained in RPMI 1640 medium (Bio Whittaker Europe, Verviers, Belgium). Media were supplemented with 10% fetal calf serum and cells were grown at 37°C in 5% CO2.
Antibodies and reagents
Polyclonal rabbit IgGs against L-plastin have been previously characterised (Lapillonne et al., 2000). Mouse monoclonal anti-vinculin antibody was a kind gift of M. Glukhova (Institut Curie, Paris, France); anti-cortactin antibody was purchased from Upstate (TE Huissen, The Netherlands) and anti-vimentin antibody from Santa Cruz Biotechnology (Tebu-Bio, Boechout, Belgium). Polyclonal anti-Ser5-P antibody against L-plastin phosphorylated at Ser5 was raised against a peptide encoding L-plastin residues 2-17 in which Ser5 was phosphorylated (ARGS(P)VSDEEMMELREA). Rabbit antiserum was purified by negative affinity on non-phosphorylated peptide followed by positive affinity on the phosphorylated peptide. The monoclonal VII-E-7 antibody against the 13 C-terminal residues of the Sendai virus L protein (sv-tag) was a kind gift of J. Neubert (Max-Planck-Institut für Biochemie, Martinsried, Germany). The rabbit polyclonal antibody against the same sequence was described previously (Arpin et al., 1994). The anti-rabbit IgG antibody coupled to horseradish peroxidase was purchased from Amersham Biosciences (Roosendaal, The Netherlands). Cy2-conjugated goat anti-rabbit IgG was purchased from Jackson Immuno-Research Laboratories (De Pinte, Belgium). Alexa Fluor 350- or 594-coupled phalloidin and secondary antibodies were purchased from Molecular Probes (Invitrogen, Merelbeke, Belgium). β-G-actin was purchased from Cytoskeleton (Boechout, Belgium). PKA catalytic subunit, PKA activator 8-Bromo-cAMP, PKA inhibitors H89 and KT5720 and PKC inhibitor GF109 and Gö6796 were purchased from Calbiochem (Leuven, Belgium). Forskolin was obtained from Sigma (Bornem, Belgium). Lipofectin reagent was purchased from Invitrogen.
Site-directed mutagenesis of cDNAs
To mutate Ser5 into Ala or a Glu by a PCR-based approach, Mut5A (5′-AAAAATGGCCAGAGGAGCAGTGTCC-3′), Mut5E (5′-AAAAATGGCCAGAGGAGAAGTGTCC-3′) sense primers corresponding to L-plastin nucleotides 4-21 were used. Underlined sequences indicate mutated amino acids. The common reverse primer (3′-CTTCCCCTCCTTCAGGTCCTCAGC-5′) was complementary to bases 802-780 of L-plastin cDNA and flanked at its 5′-end by an NcoI site. The PCR fragment containing the mutation was used to replace the corresponding fragment of the L-plastin cDNA inserted in the pCB6 expression vector, downstream of the cytomegalovirus promoter. The Ser5Glu and Ser5Ala cDNAs were also cloned into the BamHI-EcoRI sites of the pGEX-2T expression vector. Mutated DNAs were verified by sequencing.
Wild-type human L-plastin, as well as the Ser5Glu and Ser5Ala variants of L-plastin were produced in E. coli from the pGEX-2T expression vector and purified as described previously (Arpin et al., 1994). The concentration of thrombin-cleaved proteins was determined according to the Bradford method (BioRad, Nazareth, Belgium) and by SDS polyacrylamide gel electrophoresis (PAGE) using a BSA protein standard curve.
Transient expression of cDNAs in cells
Five or 10 μg of cDNA encoding wild-type human L-plastin or L-plastin variants was transfected respectively into HEK239T cells using calcium phosphate procedure (Chen and Okayama, 1988) or into 5×106 Vero cells by electroporation at 240 volt and 950 μF (Toneguzzo et al., 1986). LLCPK1 cells were transfected by Lipofectin.
Treatment of cells with pharmacological agents
Cells were washed with PBS and resuspended in HBSS++ (1× Hanks' buffered salt solution) containing 20 mM Hepes, 0.5 mM Mg2+ and 1.0 mM Ca2+. Jurkat T lymphoid cells or HEK293T cells were treated with PKA or PKC activators or inhibitors as indicated in the figure legends and described (Wang and Brown, 1999). After treatment, cells were lysed and processed for immunoblotting analysis.
Analysis of L-plastin or L-plastin variants by immunoblotting
Cells were lysed for 30 minutes in ice-cold RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.1% SDS, 1% Triton X-100, 1% NP40 and 1% Na-deoxycholate) containing a cocktail of protease and phosphatase inhibitors. Lysates were cleared by centrifugation in a microfuge at 16,000 g for 10 minutes at 4°C, and the protein concentration was determined using a modified Lowry method or Bradford assay (BioRad).
Total cell lysates (20 μg of protein) were separated by PAGE under reducing conditions and transferred onto nitrocellulose membrane (Schleicher & Schuell) using a semi-dry transblot apparatus. Primary antibodies indicated in figure legends were revealed by using secondary antibodies coupled to horseradish peroxidase and enhanced chemiluminescence (ECL). In some experiments, the membrane was stripped as previously described (Janji et al., 1999).
Transfected cells were fixed with 3% paraformaldehyde and processed for immunofluorecence labelling as described previously (Friederich et al., 1999). Labelled cells were analysed by epifluorescence microscopy (Leica DMRX microscope, HCX PL APO ×63 or ×100) or a Zeiss laser scanning confocal microscope (LSM-510 Meta). Images were acquired with a linear CCD camera (micromax, Princeton Instruments) and analysed with Metaview or Metamorph software (Universal Imaging Cooperation)
Detergent extraction of cells
Transfected Vero cells were plated onto glass coverslips. After 48 hours, coverslips were cut in halves. One half was directly processed for indirect immunofluorescence, while the other was detergent-extracted with 0.5% Triton X-100 for 16 seconds at 20°C, as previously described (Arpin et al., 1994; Friederich et al., 1992) and then processed for immunofluorescence labelling. For quantification, number of cells with detectable sv-tag labelling per 100 cells were determined for each half coverslip, yielding quantitative information on the transfection efficiency and on the resistance of transfected proteins to detergent extraction.
Phosphorylation of L-plastin in vitro
L-plastin protein was phosphorylated using PKA catalytic subunit (specific activity ⩾750 units/μg of protein) at 30°C in a reaction mixture containing 20 mM Tris-HCl pH 7.4, 20 mM MgCl2, 10 mM dithiothreitol and 100 μM ATP. To monitor phosphorylation, aliquots were removed, the reaction was stopped by addition of boiling SDS-sample buffer and proteins were analysed by immunoblotting following the ECL detection method (Amersham Bioscience). To analyse F-actin binding capacity of L-plastin, aliquots from the reaction were added to G-actin and processed as described below.
Actin binding and bundling assays
G-actin was polymerised overnight at 4°C or for 4 hours at room temperature in the presence of L-plastin variants in polymerisation buffer (100 mM KCl, 1 mM MgCl2, 1 mM ATP, 0.5 mM EGTA, 50 mM sodium phosphate buffer, pH 7.0) as indicated in figure legends. Sedimentation of actin filaments and L-plastin was achieved by high-speed centrifugation at 200,000 g for 30 minutes. To sediment actin bundles, samples were centrifuged for 15 minutes at 12,000 g (Glenney, Jr et al., 1981). Proteins in pellets and supernatants were separated by SDS-PAGE. Coomassie-stained protein bands were scanned and densities were quantified using BioCapt and Bio-PROFIL Bio 1D software (Windows applications) or LabImage software 2.7.2 (Kapelan Bio-Imaging Solution).
Transmission electron microscopy
Samples prepared for the above-described bundling assay were analysed by transmission electron microscopy. Aliquots were removed from the mixture before centrifugation and immediately negatively stained with 1% uranyl acetate. Electron micrographs were obtained using a Philips CM 120 microscope at a magnification of 28,000× or 45,000×.
Collagen invasion assays
Gels were prepared in a six-well plate from a collagen-type-I solution (Upstate Biotechnology, Lake Placid, NY). Cells (1×105) were incubated on top of the gels for 24 hours at 37°C. HEK293T cells inside the gel were scored with a phase contrast microscope controlled by a computer program (Bracke et al., 2001; De Corte et al., 2002). Invasive and superficial cells were counted in 12 fields of 0.157 mm2. The invasion index is the percentage of cells invading the gel over the total number of cells counted. DHD-FIB rat colon myofibroblasts were used as a positive control. Experiments were performed in triplicate. Mean values and standard deviations were calculated.
Fast aggregation assay
Cell-cell adhesion was numerically evaluated in an aggregation assay as described earlier (Boterberg et al., 2001). Briefly, single-cell suspensions of HEK293T cells were prepared according to an N-cadherin-saving procedure and allowed to aggregate on a Gyrotory shaker at 80 rpm for 30 minutes in aggregation buffer (1.25 mM Ca2+, 0.1 mg DNase /ml, 10 mM Hepes and 0.1% BSA) and equilibrated at physiological pH and osmolarity. Cell aggregation was measured with an LS particle size analyzer (LS 200, Coulter Electronics) after 0 and 30 minutes of aggregation. The relative volume in function of the particle size was used as an index of aggregation. Semi-quantitative evaluation of Fast Aggregation (FA) was determined as follow. Less than 20 μm, no aggregates I; between 20 and 100 μm, no aggregates II; between 100 and 1000 μm, aggregates III; more than 1000 μm, aggregates IV.
Wound healing assay
HEK293T cells (8×105) were seeded into six-well cell culture plates and 18 hours after seeding, cells were transfected with cDNA constructs. After 48 hours, a wound was made by scratching a line in a confluent monolayer. Cell debris was removed by washing the cells with serum-free medium. Migration of cells into the wound was then observed at different time points. Cells were followed for 24 hours.
We wish to thank M. Goethals and L. Cabanié for their excellent technical help. This work was supported by the Centre National de Recherche Scientifique, France, the `Fondation Luxembourgeoise Contre le Cancer', the Fondation `Aide aux enfants atteints d'un cancer, Luxembourg'. A.G. is supported by fellowships from the MCESR, Lions Vaincre le Cancer, Luxembourg and by a grant of the EMBO Advanced Light Microscopy Facility. J.G. and V.D.C. greatly appreciate support from Marc Mareel and Joël Vandekerckhove, and acknowledge support of the Interuniversity Attraction Poles (IUAP/P5), Fortis Bank Verzekeringen, and the Belgian Federation against Cancer. V.D.C. is a Postdoctoral Fellow of the Fund for Scientific Research-Flanders, Belgium (F.W.O.).
Supplementary material available online at http://jcs.biologists.org/cgi/content/full/119/9/1947/DC1
↵* These authors contributed equally to this work
- Accepted January 4, 2006.
- © The Company of Biologists Limited 2006