During wound healing, keratinocytes initiate migration from the wound edge by extending lamellipodia into a fibronectin-rich provisional matrix. While lamellipodia-like structures are also found in cultured keratinocytes exposed to epidermal growth factor (EGF), the signaling pathway that regulates the formation of these structures is not defined. In cultured human keratinocytes seeded on fibronectin, we found that protein-serine/threonine kinase inhibitors including staurosporine, induced concentration-dependent formation of extended lamellipodia (E-lams). The formation of E-lams was inhibited by the proteintyrosine kinase inhibitors herbimycin A and genistein and augmented by the protein-tyrosine phosphatase inhibitor sodium orthovanadate. Staurosporine treatment induced relocation of tyrosine phosphorylated phospholipase C-γ1 (PLC-γ1) to the tips of lamellipodia where actin assembly was initiated. Consistent with an involvement of PLC-γ1 in E-lam formation, intracellular free calcium (Ca2+) was elevated during the formation of E-lams and conversely, E-lam formation was blocked by intracellular Ca2+ chelation with BAPTA/AM, but not by extracellular reduction of Ca2+ by EGTA. Notably, glycogen synthase kinase-3α/β (GSK-3α/β) was activated by staurosporine as evidenced by reduced phosphorylation on Ser-21/9. Suppression of GSK-3 activity by LiCl2 or by a specific chemical inhibitor, SB-415286, blocked E-lam formation but without altering cell spreading. Furthermore, GSK-3 inhibitors blocked both staurosporine- and EGF-induced keratinocyte migration in scratch-wounded cultures. We propose that GSK-3 plays a crucial role in the formation of long lamellipodia in human keratinocytes and is potentially a central regulatory molecule in epithelial cell migration during wound healing.
After disruption of epithelia by injury, migration of epithelial cells is rapidly initiated to re-establish tissue integrity (Stenn and Malhorta, 1992; Woodley, 1996). Keratinocyte migration on the fibronectin-rich, provisional wound matrix is initiated by the formation of long lamellipodia that extend into the wound, a step that is considered crucial for epidermal wound closure (Odland and Ross, 1968; Larjava et al., 1996; Woodley, 1996). Similar cellular structures reflect invasive, migratory behavior of border cells during Drosophila melanogaster oogenesis (Fulga and Rørth, 2002). Lamellipodia formation and cell migration are dependent on the coordinated assembly of actin filaments and microtubules (Larjava et al., 1996; Mikhailov and Gundersen, 1998; Small et al., 1999). Actin assembly in turn relies on the activity of Rac1, a small GTPase of the Rho family (Nobes and Hall, 1995; Fenteany et al., 2000). Migration of keratinocytes on the provisional matrix also requires integrins (Grinnell, 1992; Hughes and Pfaff, 1998; Larjava et al., 2002), transmembrane receptors for matrix proteins whose binding activity is controlled in part by alterations of affinity. The pathways involved in the formation of extended lamellipodia in keratinocytes remain virtually unexplored.
Structures resembling the cytoplasmic extensions of keratinocytes in wounds have been observed previously in cultured mouse epidermal cells and megakaryoblastic leukemia cells following exposure to the protein kinase inhibitor, staurosporine (Sako et al., 1988; Yamazaki et al., 1999). Similarly, rat pheochromocytoma PC12 cells exhibit neurite outgrowth after stimulation with staurosporine (Rasouly et al., 1994; Rasouly et al., 1996). Although the signaling pathways underlying these phenomena are not defined, there may be a common mechanism that also controls the formation of cellular extensions in keratinocytes. Accordingly, staurosporine may provide a useful tool to study the cellular signaling events that regulate extended lamellipodia formation.
Staurosporine is a broad-spectrum protein-serine/threonine kinase inhibitor whose function is dependent on interactions with the ATP-binding site of its target kinases. As staurosporine can either stimulate or inhibit multiple protein kinases in a cell-type specific manner, multiple potential targets may be involved in the formation of cytoplasmic extensions. For example, molecular targets of staurosporine that can regulate cell shape include protein kinases C (PKC1), A, G and B/Akt as well as Ca2+/calmodulin-dependent kinase, myosin light chain kinase, p21-activated kinase 2, 3′-phosphoinositide-dependent protein kinase 1 and Rho-associated kinase (Tamaoki et al., 1986; Herbert et al., 1990; Feng et al., 1999; Masure et al., 1999; Zeng et al., 2000; Hill et al., 2001). Moreover, the effects of staurosporine appear to be strongly influenced by the cell type that is studied, the composition of the matrix and the concentration of staurosporine. Consequently, different studies employing staurosporine often present contradictory conclusions. For example, staurosporine can either inhibit or activate PKC in keratinocytes, depending on the nature of the experimental model (Sako et al., 1988; Dlugosz and Yuspa, 1991; Matsui et al., 1992; Matsui et al., 1993; Jones and Sharpe, 1994; Sudbeck et al., 1994; Stanwell et al., 1996).
In this study, several different protein-serine/threonine kinase inhibitors, including staurosporine, promoted dose-dependent spreading and extended lamellipodia (E-lam) formation in keratinocytes seeded on fibronectin. Our major finding is that staurosporine and epidermal growth factor can induce E-lam formation, and that ultimately, the activity of glycogen synthase kinase-3 (GSK-3) is required for epithelial cell migration and E-lam formation.
Materials and Methods
Staurosporine, herbimycin A, genistein and sodium orthovanadate were obtained from Sigma (St. Louis, MO, USA). LY-294002, H-7, K-252a, BAPTA-AM, U-73122 and thapsigargin were obtained from Biomol (Plymouth Meeting, PA, USA). Tyrphostin AG1478 was purchased from Calbiochem (La Jolla, CA, USA). SB-415286 was a generous gift from GlaxoSmithKline (Research Triangle Park, NC, USA).
Electron microscopy of 3-day-old wounds
Wounds were created in the palatal gingiva of human volunteers with a scalpel (2-3 mm deep, 2 mm wide and 8 mm long). The protocol was approved by the Clinical Research Ethics Board of the University of British Columbia. Tissue biopsies from 3-day-old wounds were fixed in 2% glutaraldehyde for 2 hours, impregnated with 1% osmium tetraoxide for 30 minutes, 2% uranylacetate for 30 minutes, alcohol-dehydrated and embedded in Epoxy-resin (Epon 812). For transmission electron microscopy, 500 nm thick sections were cut with a glass knife (Microtome MT 6000 Sorvall, Dupont, Wilmington, DE) and stained with 1% Toluidine Blue/1% borax for 5 minutes on a hot plate. Sections (80 nm thick) were made with a diamond knife and placed on nickel grids, stained with 5% uranylacetate for 20 minutes and with lead citrate for 5 minutes, and then viewed and photographed with a transmission electron microscope (Philips 300, Philips, Eindhoven, Holland).
The immortalized epidermal keratinocyte cell line HaCaT (a generous gift from Dr Norbert Fusenig, German Cancer Center, Heidelberg, Germany) was maintained in Dulbecco's modified Eagle's medium (DMEM; Gibco BRL Life Technologies, Rockville, Maryland, USA) supplemented with 23 mM sodium bicarbonate, 20 mM Hepes, antibiotics (50 μg/ml streptomycin sulfate, 100 U/ml penicillin) and 10% heat-inactivated fetal calf serum (FCS; Gibco BRL). The HaCaT cell line models many of the properties of normal epidermal keratinocytes, is not invasive and can differentiate under appropriate experimental conditions (Boukamp et al., 1988). Cells were always subcultured for 2 days before experiments to maintain consistency throughout the study. Primary mouse keratinocytes were isolated and maintained as described previously (Häkkinen et al., 2002).
To study whether growth factors could induce lamellipodia formation in HaCaT keratinocytes, cells were seeded on 6-well plates (30,000 per well) in complete growth medium. The next day, cells were changed to DMEM containing 1% FCS and epidermal growth factor (EGF; Invitrogen), transforming growth factor β1 (TGFβ1; Chemicon, Temencula, CA, USA), tumor necrosis factor α (TNFα; Chemicon) or keratinocyte growth factor (KGF; Upstate, Lake Placid, NY, USA). The morphology of the cells was recorded by digital photography after 48 hours in culture.
Cell spreading assays
Cell spreading experiments were done essentially as described before (Koivisto et al., 1999). Cell culture plates were coated with fibronectin (2 μg/cm2; from bovine plasma; Chemicon). Unoccupied sites were blocked with 1% heat-denatured bovine serum albumin. HaCaT cells (30,000/cm2) were seeded on plates in serum-free DMEM containing 50 μM cycloheximide to prevent de novo protein synthesis. Inhibitors were added either at the moment of seeding, or cells were allowed to attach and spread for 120 minutes prior to adding the agents, depending on the nature of the experiment to be performed. To quantify the effect of the drugs on cell spreading and E-lam formation, cells were fixed by carefully adding formaldehyde (8% formaldehyde and 10% sucrose in phosphate-buffered saline (PBS) containing Ca2+ and Mg2+). The wells were filled with distilled water to the top and covered with a glass plate for observation by phase-contrast microscopy. Four representative fields from each replicate well were studied. The percentages of spread cells and cells with E-lams of total cell number in each field were counted. Cell with lamellar cytoplasm were considered as spread. Alternatively, cells were collected and lysed at certain time points and analyzed for protein phosphorylation.
To examine the more detailed morphology of spreading keratinocytes, HaCaT cells were allowed to spread on fibronectin-coated coverslips in the presence or absence of staurosporine followed by fixation with glutaraldehyde (2.5% in 0.1 M PBS) and preparation for scanning electron microscopy (Aoki and Tavassoli, 1981; Firth et al., 1997).
Cells were allowed to spread on fibronectin and treated with staurosporine as described above and lysed at designated time points in Tris-buffered saline (pH 7.6) containing 1 mM phenylmethylsulfonyl fluoride, 10 μg/ml leupeptin, 5 mM EGTA, 1 mM sodium orthovanadate, 5 mM sodium fluoride, 100 mM n-octyl β-D-glucopyranoside and 0.5% Nonidet P-40. Cell lysates were separated by SDS-PAGE, transferred onto Hybond ECL membrane (Amersham, Little Chalfont, Buckinghamshire, UK) and immunoblotted with antibodies recognizing phospho-GSK-3α/β (Y279/Y216; BioSource International, Camarillo, CA, USA or S21/S9; Cell Signaling Technology, Beverly, MA, USA) or total GSK-3 (0011-A; Santa Cruz Biotechnology, Santa Cruz, CA, USA). Peroxidaseconjugated IgGs (Santa Cruz Biotechnology) were used as secondary antibodies. Detection was performed with Amersham's ECL kit. Cell lysates with or without staurosporine treatment were also probed for the quantitative expression of 33 known phosphoproteins using validated commercial antibodies in the Kinetworks™ Phospho-Site Screen (KPSS 1.0), immunoblotting analysis provided by Kinexus Bioinformatics Corporation (Vancouver, Canada). For this screen, cells were treated with staurosporine either for 10 minutes, which preceding E-lam formation, or for 60 minutes, by which time the E-lams have formed. Untreated control samples were collected at the same time points.
Cells were allowed to spread on fibronectin-coated SonicSeal Slide™ wells (Nunc, Naperville, IL, USA), treated with staurosporine as described above, formaldehyde-fixed and permeabilized with PBS containing 0.5% Triton X-100. Immunolocalization was performed as described previously (Larjava et al., 1990) using antibodies recognizing phospholipase Cγ-1 phosphorylated on Tyr-783 (Sigma) or GSK-3 (0011-A; Santa Cruz Biotechnology). Species-specific Alexa Fluor™ 546 IgG molecules (Molecular Probes, Eugene, OR, USA) were used as secondary antibodies. Alternatively, anti-mouse Vectastain® ABC and Vector® VIP peroxidase substrate kits (Vector laboratories Inc., Burlingame, CA, USA) were used for immunodetection of GSK-3. Rhodamine-labeled phalloidin (Sigma) was used to detect actin filaments in acetone-fixed samples.
Measurement of intracellular Ca2+
Cells were plated and allowed to spread on fibronectin-coated coverslips as described above and loaded with fura 2/AM (2 μM; Molecular Probes). Single cells were measured with an inverted microscope optically interfaced to a Photon Technology Incorporated (London, ON) ratio fluorimeter. The slit widths were set at 3 nm and excitation was set at alternating wavelengths (100 Hz) of 346 and 380 nm. Measurements were conducted at room temperature in serum-free PBS containing 20 mM Hepes buffer to maintain the pH at 7.4. A stable baseline Ca2+ ratio was obtained for ∼50 seconds after which 50 nM staurosporine was added. All cells were measured for 500 seconds.
For scratch wound migration assays, HaCaT cells were seeded on 24-well plates (1.7×105 cells per well) in DMEM containing 10% FCS. The cells were grown to confluence for 2 days after which the cultures were scratch-wounded with a pipette tip. Loose cells were removed by washing with PBS, and the cells were allowed to migrate for 24 hours in Minimum Essential Medium (EMEM; Bio-Whittaker, Walkersville, MD, USA) supplemented with 1% FCS in the presence of staurosporine (50 nM) or EGF (10 ng/ml) and inhibitors. The inhibition in cell migration produced by GSK-3 inhibitors was quantified by comparing the number of cells within the wound area of inhibitor-treated samples with the number of cells in control wounds. Eight representative fields from each wound were studied.
All experiments were repeated at least three times. Each of the spreading assays used two to four replicate wells, and four fields of cells for each well were counted (n=8-16). Statistical significance was set at P<0.05, and the difference between two groups was calculated using unpaired, two-tailed Student's t-test.
Keratinocytes form lamellipodia in mucosal wounds that can be simulated in vitro
Keratinocytes in 3-day-old human mucosal wounds formed long, finger-like projections, which flared at their termini, into the loose matrix of the wound bed (Fig. 1A). No hemidesmosomes or basement membrane were observed underneath the advancing cells (Fig. 1A). Primary mouse keratinocytes that were cultured in the presence of fibroblast-conditioned medium and EGF (Häkkinen et al., 2002) also spontaneously formed long, thin lamellipodia (Fig. 1B), similar to those seen in human wounds. We refer to these structures as `extended lamellipodia', or `E-lams'. Compared to classical lamellipodia, E-lams contain an additional thin cellular process resembling an arm. At the tip of the arm, a veil-like lamellipodium was formed (Fig. 1B). To test whether growth factors and cytokines commonly present in wound fluids (EGF, TGFβ1, TNFα, and KGF) could induce E-lam formation in cultured human keratinocytes, HaCaT cells were grown in the presence of these cytokines for 48 hours. Only EGF was capable of inducing E-lam formation in HaCaT cells (Fig. 1C). The observed E-lam formation in cultured HaCaT cells was in part dependent on EGF receptor (EGFR)-mediated signaling, because tyrphostin AG1478 (2 μM), a highly selective inhibitor of EGFR tyrosine kinase activity (Levizki and Gazit, 1995; Osherov and Levitzki, 1997), prevented EGF-induced E-lam formation (data not shown). As EGF-induced E-lam formation was a relatively slow process (no E-lams were observed before 24 hours), this phenomenon probably involves numerous pathways and alterations in patterns of gene expression and protein synthesis. Accordingly, we elected to search for model systems that produced more definitive and rapid formation of E-lams.
Staurosporine induces extended lamellipodia and cell spreading in human keratinocytes on fibronectin
To focus on a model that exhibited more rapid induction of E-lams and that would permit study of key signaling events in E-lam formation, we tested whether the protein-serine/threonine kinase inhibitor staurosporine could accelerate E-lam formation in keratinocytes. Staurosporine has been shown to promote lamellipodia extension in other cell types (Sako et al., 1988; Rasouly et al., 1994; Rasouly et al., 1996; Yamazaki et al., 1999). Accordingly, HaCaT cells were seeded on to fibronectin-coated substrata in the presence of staurosporine, a treatment that induced keratinocytes to form morphologically similar E-lams as seen in EGF-treated keratinocytes. The ability of staurosporine to induce formation of E-lams was not restricted to HaCaT keratinocytes as the same phenomenon was observed with primary mouse and human epidermal keratinocytes (data not shown). In contrast to EGF, staurosporine induced E-lams within 60 minutes in the absence of de novo protein synthesis (Fig. 2Ab). Staurosporine, therefore, possibly bypassed the requirement for EGFR signaling, as neither EGF (30 ng/ml) nor tyrphostin AG1478 (2 μM) exerted any effect on staurosporine-induced E-lam formation (data not shown). Staurosporine also induced a dose-dependent increase in spreading of keratinocytes on fibronectin. Maximal induction of spreading was achieved with 10 nM staurosporine. E-lam formation was, however, typically initiated at concentrations above 30 nM staurosporine (Fig. 2B). Staurosporine was able to induce E-lams to the same extent in spreading cells and in cells that had already spread (Fig. 2C). This characteristic allowed us to investigate signaling mechanisms of E-lam formation that are independent of cell spreading. Scanning electron microscopy was used to visualize staurosporine-induced morphological changes at higher resolution. In addition to E-lam formation, cells treated with 50 nM staurosporine exhibited fine filopodia that projected from the lamellipodia (Fig. 2A).
To confirm that E-lam formation was associated with the inhibition of protein-serine/threonine kinases, two other broad-spectrum protein-serine/threonine kinase inhibitors, H7 and K252a, were tested for their ability to induce E-lam formation in cells plated on fibronectin. Both inhibitors were able to induce cell spreading and E-lam formation, although they were less potent than staurosporine, indicating that E-lam formation was indeed related to inhibition of protein-serine/threonine kinases. For H7, maximal cell spreading was induced at 400 nM, whereas E-lams did not start to form until concentrations above 2 μM were used (data not shown). For K252a, both maximal cell spreading and the onset of E-lam formation were reached at 50 nM (data not shown). The maximal incidence of E-lam formation with H7 and K252a never exceeded 20% of the cells (data not shown). At 50 nM staurosporine, E-lams that were morphologically similar to those induced by EGF occurred in 15-25% of cells. Higher concentrations of staurosporine increased the proportion of cells exhibiting E-lams but their morphology became more irregular (Fig. 2B and data not shown). Accordingly, this 50 nM concentration was used in further studies that focused on the signal transduction mechanisms regulating lamellipodia formation in keratinocytes.
To determine whether the formation of E-lams also required changes in tyrosine phosphorylation in addition to the inhibition of protein-serine/threonine kinases, cells were allowed to spread on fibronectin and then treated with staurosporine (50 nM) in combination with the nonselective protein-tyrosine kinase inhibitors (herbimycin A; 10 μM, genistein; 200 μM) or a protein-tyrosine phosphatase inhibitor (sodium orthovanadate; 500 μM). Both herbimycin A (Fig. 3E,F) and genistein potently inhibited E-lam formation (Fig. 3G,H), indicating that E-lam formation was dependent on tyrosine phosphorylation that is mediated by protein-tyrosine kinases other than EGFR tyrosine kinase. As indicated above, E-lam formation was unresponsive to the inhibition of EGFR tyrosine kinase. In agreement with the importance of tyrosine phosphorylation in E-lam formation, sodium orthovanadate increased the proportion of cells forming E-lams from ∼20% to ∼60% (Fig. 3B,D). In control cells without staurosporine, sodium orthovanadate caused cell retraction (Fig. 3A,C).
Staurosporine induces GSK-3 activation
To examine changes in protein phosphorylation that are involved in E-lam formation, cells treated with 50 nM staurosporine were analyzed by Kinexus Kinetworks™ Phospho-Site Screen (data not shown). The results indicated that staurosporine may increase GSK-3 activation. Notably, GSK-3 is a major regulator of the cytoskeleton in other cell types (Wagner et al., 1996). The activity of GSK-3α/β is mainly regulated by inhibitory phosphorylation on serine residues (S21/S9) (Grimes and Jope, 2001). Accordingly, we found that staurosporine strongly induced dephosphorylation of GSK-3α/β on serine 21/9 (Fig. 4A,B). The reduction of GSK-3 serine phosphorylation was detectable within 1-5 minutes after incubation with staurosporine (Fig. 4A,B). Both α and β GSK-3 isoforms were dephosphorylated at serine residues by staurosporine treatment (Fig. 4A,B). Tyrosine phosphorylation of GSK-3α/β on residues 279/216 is a prerequisite for its activity (Hughes et al., 1993). Tyrosine phosphorylation of GSK-3 was not markedly altered by staurosporine (Fig. 4C). Thus, staurosporine induced changes in GSK-3 phosphorylation that were consistent with its activation.
Extended lamellipodia formation requires the activity of GSK-3
The direct involvement of GSK-3α/β in E-lam formation was tested by treating cells with staurosporine in the presence of the GSK-3 inhibitors LiCl2 and SB-415286 (Klein and Melton, 1996; Coghlan et al., 2000). Both of these GSK-3 inhibitors caused a specific and concentration-dependent reduction of the incidence of E-lams without a change in general cell morphology. This experiment indicated that GSK-3 activity is specifically involved with the extension of E-lams but does not affect lamellar spreading (Fig. 5A-C). In agreement with this function, immunostaining for GSK-3 was restricted to the extended part of the E-lam in staurosporine-treated cells and was not present in lamellae (Fig. 6B,C). In EGF-treated cells, the localization of GSK-3 was more diffuse, but it was still more abundant away from cellular perimeters (data not shown).
Intracellular free Ca2+ is essential for E-lam formation
GSK-3 is activated in response to an increase in intracellular Ca2+ (Hartigan and Johnson, 1999), and staurosporine can induce a long-lasting increase of intracellular free Ca2+ in primary human keratinocytes (Jones and Sharpe, 1994). We found that in HaCaT cells, 50 nM staurosporine induced a slow and steady increase in cytoplasmic free Ca2+ that was detectable within 30 seconds after adding the staurosporine and reached a plateau in 300 seconds (in 6 out of 15 cells measured; 40%). A typical response is shown in Fig. 7A. None of the cells treated with vehicle only showed any Ca2+ increase (data not shown). To test the importance of this Ca2+ influx for E-lam formation, we treated spread HaCaT cells with either BAPTA/AM (100 nM-10 μM) or EGTA (2 mM) to chelate intracellular or extracellular Ca2+, respectively, together with staurosporine. Chelation of extracellular Ca2+ by EGTA slightly increased cell spreading but did not affect E-lam formation (Fig. 7B). The effect of BAPTA/AM on cell morphology, however, was dependent on staurosporine (Fig. 7B): the spread morphology of control cells was unaffected by BAPTA/AM. The cells exposed to staurosporine, however, underwent concentration-dependent collapse in response to BAPTA/AM, and E-lam formation was completely abolished at 10 μM BAPTA/AM. Staurosporine thus sensitized the cells to chelation of intracellular Ca2+. The Ca2+ concentration-dependent disappearance of E-lams indicates that intracellular Ca2+ may play a crucial role in staurosporine-induced cell spreading and E-lam formation.
Treatment with staurosporine induces relocalization of tyrosine phosphorylated phospholipase C-γ1 to E-lams and localized actin assembly
As phospholipase C-γ (PLC-γ) is a key mediator in the pathway that releases Ca2+ from intracellular stores (Berridge, 1993), we tested the involvement of PLC-γ in E-lam formation. Treatment of keratinocytes with the specific PLC-γ inhibitor U-71322 induced a concentration-dependent loss of cell spreading both in control and staurosporine-treated cells, as well as inhibition of E-lam formation (Fig. 8A). Thus PLC-γ activity was involved in the spreading behavior of keratinocytes, but a direct role for PLC-γ in E-lam formation could not be ascertained from inhibitor studies. As the activity of PLC-γ1 depends on phosphorylation on Tyr-783 (Kim et al., 1991), indirect evidence for the importance of PLC-γ1 in E-lam formation was obtained by immunocytochemistry. These experiments demonstrated localization of active, tyrosine phosphorylated PLC-γ1 at the tips of E-lams and co-localization with actin filaments (Fig. 8B). Consistent with the importance of PLC-γ in staurosporine-induced morphological change, disturbance of cellular Ca2+ gradients by thapsigargin (a sarcoplasmicendoplasmic reticulum ATPase (SERCA) inhibitor; ≥2 nM) blocked cell spreading (inhibition by 63%, n=8, P<0.001) and E-lam formation in staurosporine-treated cells. Thapsigargin also reduced the spreading of control cells but less potently (inhibition by 25%, n=8, P<0.001).
Agents that interfere with cellular Ca2+ signaling modulate GSK-3α/β serine phosphorylation
Because both the activity of GSK-3 and the intracellular Ca2+ signaling participated in E-lam formation, we tested whether these two pathways were connected. Notably, the activity of GSK-3 has been reported to be upregulated by increased intracellular Ca2+ (Hartigan and Johnson, 1999). Therefore, we tested whether disturbance of cellular Ca2+ metabolism by PLC-γ inhibitor U-71322 (4 μM), SERCA inhibitor thapsigargin (100 nM) or intracellular Ca2+ chelator BAPTAAM (10 μM) that all blocked E-lam formation affected staurosporine-induced serine dephosphorylation of GSK-3. None of the agents had any significant effect on tyrosine phosphorylation of GSK-3 (data not shown). Thapsigargin prevented staurosporine-induced GSK-3 serine dephosphorylation (Fig. 9A). U-71322 did not completely prevent but significantly delayed GSK-3 dephosphorylation: in cells treated with staurosporine only, the dephosphorylation of GSK-3 was evident after a 5-minute incubation (Fig. 9G), whereas in cells treated with a combination of U-71322 and staurosporine, the reduction in serine phosphorylation was only evident after 30 minutes incubation (Fig. 9C). In contrast, BAPTA-AM that buffers cytoplasmic Ca2+ greatly accelerated serine dephosphorylation of GSK-3 (Fig. 9E). Thus, agents that affect cellular Ca2+ signaling seem to modulate the degree of serine phosphorylation of GSK-3 in keratinocytes, but the GSK-3 dephosphorylation does not always lead to E-lam formation.
Staurosporine-induced cell migration is dependent on GSK-3 activity
To determine if E-lam formation induced by staurosporine is indeed associated with the migratory phenotype in keratinocytes, cells were allowed to migrate in the presence of staurosporine in scratch-wounded cultures. Staurosporine potently induced cell migration resulting in wound closure within 24 hours (Fig. 10B). Staurosporine also induced the migration of primary mouse keratinocytes, although these cells were more sensitive to staurosporine and a smaller concentration was required (data not shown). To investigate whether keratinocyte migration was dependent on GSK-3 activity, staurosporine- and EGF-treated HaCaT cells were allowed to migrate in scratch wound assays in the presence of GSK-3 inhibitors. Consistent with the importance of GSK-3 in E-Lam formation in keratinocytes, staurosporine-induced migration was completely blocked by 30 μM SB-415286 and by 30 mM LiCl2 (Fig. 10E,H). EGF-induced cell migration was also blocked when the wounded cell cultures were incubated with GSK-3 inhibitors before the addition of EGF to the cells (Fig. 10F,I), indicating that GSK-3 activity also contributes to EGF-induced cell migration. If the GSK-3 inhibitors were added to the cells simultaneously with EGF, the cell migration was still strongly reduced although not totally prevented: SB-415286 and LiCl2 inhibited the number of cell within wounds by about 35% and by 45%, respectively, compared to cells treated with EGF only (data not shown).
In the present study, we demonstrated that GSK-3 regulates the formation of long lamellipodia in human keratinocytes and their migration into in vitro wounds. This is a novel function for GSK-3 and makes it potentially a key molecule for tissue repair. GSK-3 is known to regulate many signaling as well as structural molecules providing multiple targets for GSK-3 action (Grimes and Jope, 2001). There are two highly homologous isoforms of mammalian GSK-3, GSK-3α and GSK-3β that are encoded by separate genes (Woodgett, 1990). The two GSK-3 isoforms show different tissue expression patterns and their expression is differentially regulated, suggesting that they may be functionally distinct (Lau et al., 1999), although in vitro they exhibit similar biochemical and substrate specificities (Woodgett, 1991). The only known difference between GSK-3α and GSK-3β to date is the specific ability of GSK-3β to activate the transcription factor NF-κB (Hoeflich et al., 2000). GSK-3 has an important role in determination of cell fate during development (Kim and Kimmel, 2000). GSK-3 complexes also serve tumor-suppressive functions in cells (Kim and Kimmel, 2000). An excessive and dysregulated activity of GSK-3 is, however, linked to many pathological conditions (Cohen and Frame, 2001). The majority of studies about the function of GSK-3 are attributed to GSK-3β because of its high expression in brain tissues and its role in Alzheimer's disease (Lau et al., 1999; Grimes and Jope, 2001). Accordingly, most of the data concerning the function of GSK-3 comes from neuronal cells, and there is no information about its function in keratinocytes during wound repair. We show that both GSK-3 isoforms are expressed by cultured HaCaT keratinocytes and that GSK-3 catalytic activity is involved in the regulation of E-lam formation.
GSK-3α/β catalytic activity is positively regulated by tyrosine phosphorylation (Y279/Y216) and is inhibited by phosphorylation on Ser-21/9 (Hughes et al., 1993; Grimes and Jope, 2001). GSK-3 is a serine/threonine kinase, and staurosporine can inhibit GSK-3 activity in vitro (Leclerc et al., 2001). However, this inhibition has not been shown to occur in living cells. In contrast, Bhat et al., (Bhat et al., 2000) reported an increase in GSK-3 tyrosine phosphorylation and activity in neuronal cells treated with staurosporine. In the present study, staurosporine did not, however, markedly regulate tyrosine phosphorylation of GSK-3. The protein-tyrosine kinase inhibitor genistein can induce GSK-3 dephosphorylation and inactivation, and this inhibition can be reversed by the proteintyrosine phosphatase inhibitor sodium orthovanadate (Yu and Yang, 1994; Yu et al., 1997). In agreement with these previous results, we found that treatment of HaCaT cells with genistein and herbimycin A inhibited E-lam formation whereas sodium orthovanadate enhanced E-lam formation. In addition to GSK-3, tyrosine phosphorylation regulates the activities of PLC-γ1 and Rac1 (Kim et al., 1991; Sastry et al., 2002), molecules that are required for E-lam formation. In the present study, we also demonstrated that treatment of keratinocytes with staurosporine potently reduced serine phosphorylation of GSK-3, an alteration that has been linked to its inactivation. Staurosporine has been reported to inhibit cAMP-dependent protein kinase, PKB/Akt and several isoforms of PKC, all of which can potentially phoshorylate GSK-3 on Ser-21/9 (Goode et al., 1992; Cross et al., 1995; Fang et al., 2000; Li et al., 2000). These findings suggest one potential mechanism for GSK-3 activation by staurosporine. Based on the relative degree of serine dephosphorylation in HaCaT cells, GSK-3α shows greater relative activity than GSK-3β. We have not, however, been able to discriminate the functions of GSK-3α and GSK-3β as both LiCl2 and the specific inhibitor, SB-415286, inhibit both isoforms to the same extent (Klein and Melton, 1996; Coghlan et al., 2000).
Staurosporine is commonly used to induce apoptosis; the putative functions of GSK-3 are often discussed in this context. In concentrations significantly higher than that used in this study, staurosporine (at 100-500 nM) induces nuclear translocation of GSK-3β in neuronal cells, a phenomenon linked to the onset of apoptosis (Bhat et al., 2000; Bijur and Jope, 2001). The narrow window of staurosporine concentrations that induce E-lam formation indicates a high level of regulation of cell signaling by staurosporine. At optimal concentrations of staurosporine, GSK-3 localizes to the nascent E-lam whereas at higher concentrations GSK-3 may associate with the nucleus. Conceivably, one of the factors that distinguishes normally regulated wound healing from malignant epithelial invasion is the susceptibility of the migrating cells to undergo programmed cell death. The mechanisms that regulate cell migration and apoptosis may be linked, the end result being determined by the degree of activation and/or cellular location of regulatory molecules such as GSK-3.
A pool of GSK-3 is active in resting cells, and it can be further activated by transient Ca2+ influxes (Hughes et al., 1993; Hartigan and Johnson, 1999). Our findings suggest that an increase in intracellular Ca2+ concentration was involved in staurosporine-induced changes in cell morphology and probably participated in GSK-3 activation. Staurosporine induced an early but relatively slow increase in the total level of intracellular free Ca2+ in HaCaT cells. Similar results demonstrating a slow Ca2+ increase were obtained previously with primary human keratinocytes exposed to staurosporine (Jones and Sharpe, 1994). The increase in cytoplasmic Ca2+ influx is likely caused by the release of Ca2+ from intracellular storage compartments (Jones and Sharpe, 1994) and is consistent with our finding that chelation of extracellular Ca2+ with EGTA did not inhibit E-lam formation whereas the intracellular Ca2+ chelator BAPTA-AM did. BAPTA-AM induced a selective rounding of the staurosporine-treated keratinocytes and an inhibition of E-lam formation but did not affect the untreated control keratinocytes.
The increase of intracellular Ca2+ by staurosporine was probably mediated by PLC-γ1, a PLC-γ isoform that is expressed by HaCaT cells (Haase et al., 1997). Notably, PLC-γ1 is activated by tyrosine phosphorylation (Kim et al., 1991). In the present study, we demonstrated that tyrosine phosphorylated PLC-γ1 colocalized with actin filaments at the tips of lamellipodia, in agreement with previously published data (Yu et al., 1998). Conceivably, PLC-γ1 activity may also be spatially restricted in this keratinocyte model system. The importance of PLC-γ1 activity in E-lam formation was also suggested by the blockade of E-lam extension following a disturbance of cellular Ca2+ gradients by thapsigargin. The treatment of cells with PLC-γ1 inhibitor or thapsigargin also delayed or prevented the staurosporine-induced dephosphorylation of GSK-3. It appears, therefore, that the localized and controlled release of Ca2+ by the activity of PLC-γ1 is essential for formation of long lamellipodia and may also contribute to dephosphorylation of GSK-3 by an unknown mechanism in keratinocytes. Surprisingly, treatment of HaCaT cells with intracellular Ca2+ chelator BAPTA-AM, which inhibited E-lam formation, accelerated rather than inhibited GSK-3 serine dephosphorylation. The effects of BAPTA-AM on cellular Ca2+ are, however, complex, and by buffering intracellular Ca2+, BAPTA-AM may actually increase the amount of Ca2+ in cells in a high-calcium medium such as DMEM (Hofer et al., 1998). Ca2+ is a ubiquitous intracellular messenger that regulates a number of cellular processes, and the amplitude, speed and spatiotemporal patterning of Ca2+ signals affect the way cells interpret them (Berridge et al., 2000). Therefore, these results suggest that intracellular Ca2+ also regulates other aspects in E-lam formation than just GSK-3 activity, as witnessed by the coincident enhanced serine dephosphorylation of GSK-3 and inhibition of E-lam formation by BAPTA-AM. In addition keeping in mind the broad effect of staurosporine on cellular signaling events, while GSK-3 activity is required for E-lam formation, GSK-3 activation is unlikely the sole cause of E-lam formation, but it also requires other coordinated changes in cellular signaling.
As demonstrated in the present and previous studies, long cellular extensions are found in keratinocytes migrating into the wound provisional matrix (Odland and Ross, 1968; Larjava et al., 1996; Woodley, 1996). Long lamellipodia seen in vitro may represent simple forms of lamellipodia seen inside three-dimensional matrix in vivo. EGF is a natural wound cytokine that may regulate the formation of long lamellipodia in vivo. Stimulation of cells with EGF activates multiple downstream signaling pathways in a complex and cell-type-specific manner that is not currently completely understood (Bogdan and Klämbt, 2001; Yarden and Sliwkowski, 2001; Danielsen and Maihle, 2002). It is commonly believed that EGFR activation leads to the activation of PKB/Akt followed by inactivation rather than activation of GSK-3 (Burgering and Coffer, 1995). EGFR activation, however, also leads to activation of Fyn, a member of Src family tyrosine kinases (Mariotti et al., 2001). Interestingly, Fyn has been shown to be involved in GSK-3 activation (Lesort et al., 1999). GSK-3β is rapidly and transiently activated followed by its inhibition by extracellular stimuli including insulin (Lesort et al., 1999). It is speculated that even this rapid and transient GSK-3β activation plays a significant role in modification of cytoskeletal structures in neuronal development by modulating cellular plasticity and by inducing neuronal outgrowth (Lesort et al., 1999). In fibroblasts, GSK-3α appears to be transiently deactivated downstream of EGFR activation (Shaw and Cohen, 1999), whereas in human epidermoid carcinoma A431 cells, EGF caused cellular translocation of active GSK-3α without affecting its overall activity (Yang et al., 1989). Induction of long lamellipodia (E-lams) by EGF in keratinocytes in vitro appears to be a slow process probably involving several second messenger signals that may eventually lead to GSK-3 activation. Similarly, keratinocyte migration induced by EGF may involve cyclic transient activation and inactivation of GSK-3 as well as modulation of its cellular localization.
Based on the results and the data discussed above, we propose that E-lam formation involves the activation and membrane translocation of PLC-γ1. PLC-γ1-induced elevation of intracellular Ca2+ may contribute to the activation of GSK-3, which mediates E-lam formation by modulating the organization of the cytoskeleton. We suggest, therefore, that GSK-3 is potentially a key molecule in re-epithelization during wound repair.
This study was supported by grants from The Canadian Institutes of Health Research and The Finnish Cultural Foundation.