Glial Cell–Derived Neurotrophic Factor (GDNF)–Induced Migration and Signal Transduction in Corneal Epithelial Cells

Abstract

purpose. To identify signal-transduction pathways induced by glial cell-derived neurotrophic factor (GDNF) in corneal epithelial cells and to characterize its effect on cell migration.

methods. Expression of GDNF receptor (GFR) α-1 in human corneal epithelium was detected by RT-PCR and Western blot analysis. Expression and phosphorylation of Ret, activation of focal adhesion kinase (FAK) and mitogen-associated protein kinase (MAPK) signaling pathways, and phosphorylation of paxillin by GDNF were investigated by immunoprecipitation and Western blot analysis in primary human corneal epithelial cells and a corneal epithelial cell line. The tyrosine kinase inhibitor herbimycin A and Ras farnesyltransferase inhibitor manumycin were used to specifically inhibit GDNF-induced signaling pathways. In vitro wound-healing assays and modified Boyden chamber analysis were performed to investigate the effect of GDNF on epithelial cell migration.

results. Expression of GFRα-1 was detected in normal and transformed human corneal epithelium. GDNF induced tyrosine phosphorylation of Ret. Furthermore, tyrosine phosphorylation of FAK and phosphotyrosine kinase (Pyk) 2; serine phosphorylation of c-Raf, MEK1, and Elk 1; and tyrosine-threonine phosphorylation of Erk-1 and -2 were time-dependently activated in the presence of GDNF. Tyrosine phosphorylation of paxillin was also induced by GDNF. Migration of corneal epithelial cells was significantly stimulated by GDNF. Herbimycin A strongly inhibited the activation of Ret, FAK, c-Raf, and Erk-1 and -2; the phosphorylation of paxillin; and corneal epithelial cell migration. More specifically, the Ras inhibitor manumycin inhibited phosphorylation of c-Raf, MEK 1, Erk-1 and -2, and Elk 1, but not that of FAK.

conclusions. Corneal epithelial cells express receptors specific for GDNF that are used by GDNF to induce intracellular signaling. FAK and MAPK pathways seem to be activated by GDNF to modulate gene transcription and cell migration. FAK seems to be an upstream regulator of the MAPK cascade for GDNF signal transduction. As an inducer of FAK-dependent corneal epithelial migration, GDNF may play an important role in corneal regeneration and wound healing.

The underlying molecular mechanisms that regulate regeneration and repair processes in the corneal epithelium have yet to be elucidated. The discovery of various cytokine families and their effects on proliferation, migration, and differentiation of cultured cells suggests a functional role for polypeptide growth factors in the human cornea. ^{1 2 3 4} However, despite significant progress in the understanding of the cytokine network in the cornea, peptide growth factors have not been used successfully to treat patients with impaired regeneration of the corneal epithelium. Only recently, topical application of nerve growth factor (NGF) has resulted in healing of human corneal ulcers due to various diseases. ⁵ NGF and other neurotrophins, such as brain-derived neurotrophic factor, neurotrophin-3 and -4, and their corresponding receptors have been shown to be transcribed in corneal epithelium and to modulate proliferation and migration of corneal epithelial cells in vitro. ^{6 7} These observations suggest that members of the nerve growth factor family, which are primarily expressed in the central and peripheral nervous system, regulate homeostasis and repair in the human corneal epithelium. Therefore, it seems likely that other neurotrophic factors are also important in these processes. 

Glial cell-derived neurotrophic factor (GDNF) is a homomeric protein with seven conserved cysteine residues. ^{8 9} It is a distant member of the transforming growth factor (TGF )-β superfamily of growth factors and a member of the GDNF family (which also includes neurturin, persephin, and artemin). ¹⁰ GDNF binds to glial cell-derived neurotrophic factor receptor (GFR) α-1, a membrane-bound protein belonging to the GFRα family. ^{11 12} On ligand binding, GFRα-1 forms a heterotetrameric complex with the proto-oncogene product Ret. Phosphorylation of the tyrosine-kinase receptor Ret induces a signaling cascade, ultimately leading to gene transcription. ¹³  

GDNF is widely distributed in the central and peripheral nervous systems where it is primarily present in dopaminergic neurons and motoneurons. It is also expressed in the inner ear, olfactory epithelium, carotid body, kidney, and gastrointestinal tract. ¹⁴ In the eye, GDNF is primarily expressed in the retina, and several investigators have pointed out that GDNF could be used therapeutically to provide neuroprotection and to rescue photoreceptors in the context of retinal degeneration. ^{15 16 17} In contrast to the nervous system, the biological significance of GDNF in other cell types, such as epithelial cells, is unclear. We have recently reported that mRNA for GDNF is transcribed in corneal keratocytes but not in corneal epithelium, suggesting that this protein may be an important modulator of epithelial function, mediating signals originating in the corneal stroma. ⁷ This hypothesis seems to be supported by the observation that recombinant GDNF stimulates the proliferation of rabbit epithelial cells in vitro. ⁷  

To further test this hypothesis and to determine whether GDNF—similar to NGF—could serve as a pharmacologic treatment for nonhealing corneal ulcers in human patients, we have investigated its effect on migration and in vitro wound healing of human corneal epithelium. Until now the downstream intracellular kinase cascade that mediates signals induced by binding of GDNF to GFRα-Ret has not been fully described. We have therefore analyzed the GDNF-dependent activation of the mitogen-associated protein kinase (MAPK) and focal adhesion kinase (FAK) pathways leading to the phosphorylation of intermediate-type filaments, such as paxillin, and the induction of gene transcription. 

Materials and Methods

Ex Vivo Corneal Tissue

After informed consent was obtained from the patients in conformance with the tenets of the Declaration of Helsinki, ex vivo corneal epithelium was obtained from eyes that were enucleated for choroidal melanoma. After enucleation, the central and midperipheral corneal epithelia were removed by mechanical scraping, snap frozen, and stored in liquid nitrogen. 

Culture of Primary Corneal Epithelial Cells and an Epithelial Cell Line

Human corneal epithelial cells were cultured as outgrowth cultures, as described. ³ In brief, the Descemet membrane was removed from corneal buttons of transplant quality, and the corneas were dissected into small pieces. Explants were grown in SHEM (1:1 mixture of Dulbecco’s modified Eagle’s medium and Ham’s F10 with 10% FBS; Gibco, Grand Island, NY), supplemented with insulin (5μ g/ml), epidermal growth factor (EGF; 10 ng/ml), adenine (24μ g/ml), and hydrocortisone (5 ng/ml; all from Sigma, St. Louis, MO). The epithelial phenotype was confirmed by an antibody specific for keratin K12. Because these cultures yielded only a limited amount of cells, we used an SV40-transformed corneal epithelial cell line (kindly provided by K. Araki Sasaki, Kinki Central Hospital, Hyogo, Japan) for Western blot experiments. Similar to normal corneal epithelium, these cells exhibit clonal growth characteristics and display a corneal epithelial phenotype, including, for example, expression of keratin K12. ^{18 19} Cultures were performed in standard conditions (37°C, 95% humidified air and 5% CO₂). 

Isolation of Total RNA and mRNA Purification

Total RNA was isolated according to the guanidinium-thiocyanate-phenol-chloroform extraction method ²⁰ by use of an isolation system (RNAgents total RNA kit; Promega, Madison, WI). ⁷ For mRNA isolation, an extraction system (polyATtract system III; Promega) was used. ⁷ To minimize the risk of contamination by genomic DNA, mRNA was digested by DNase followed by phenol-chloroform-isoamylalcohol extraction and isopropanol precipitation. 

PCR Primer Design and Reverse Transcription–Polymerase Chain Reaction

PCR primers for detection of human GFRα-1 were designed to span intron 6 according to the GFRα-1 gene sequence (GenBank accession number: NM005264; GenBank is provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD, and is available at http://www.ncbi.nlm.nih.gov/) so that the amplification of potentially contaminating genomic DNA would produce PCR fragments that were substantially larger than the cDNA PCR products. The DNA sequences of forward and reverse primers are: AGACCATCGTGCCTGTGTGCT (forward), and GGGTCATGACTGTGCCAATAAG (reverse), with a resultant 216-bp product. First-strand cDNA was synthesized by incubating 0.1 μg mRNA with 0.5 μg oligo d(T)primer, 200 U Moloney murine leukemia virus (M-MLV) reverse transcriptase, desoxyribonucleotides (dATP, dCTP, dGTP, and dTTP in a concentration of 0.5 mM) and recombinant RNasin RNase inhibitor (25 U) in 25 μl for 1.5 hours at 42°C. PCR was performed using 0.5 μl single-strand cDNA with 3 U Thermus aquaticus DNA polymerase, desoxyribonucleotides (concentration of 0.2 mM), PCR buffer, and 25 pmol upstream and downstream primers in 50 μl (all reagents from Promega). A thermocycler (PTC-100; MJ Research, Watertown, MA) was used at 95°C for 3 minutes (predenaturation). Then, 35 cycles were performed including denaturation at 94°C for 1 minute, annealing at 55°C for 1 minute, and extension at 72°C for 1 minute. PCR products were size fractionated by 2% agarose gel electrophoresis. We used Phi X 174 DNA/HinfI fragments (Promega) as a molecular weight standard. PCR fragments were cloned into pCR2.1 vectors (Invitrogen, San Diego, CA) and sequences confirmed by standard methods. 

Coimmunoprecipitation of GFRα-1 and Ret

To confirm the expression of GFRα-1 and to detect the interaction of GFRα-1 and Ret after stimulation with GDNF, coimmunoprecipitation of the GFRα-1 and Ret complex was performed. Total protein (100 μg) in 1-ml cell lysate was incubated with 40 μl protein G-agarose (Boehringer Mannheim, Mannheim, Germany) for 2 hours, followed by brief centrifugation at 12,000 rpm. The supernatant was incubated with 40 μl protein G-agarose and 10 μg polyclonal antibody against Ret (Santa Cruz Biotechnology, Santa Cruz, CA) overnight at 4°C with agitation. The protein G-agarose complex was collected by centrifugation and washed in lysis buffer. The bound protein pellet was eluted in SDS gel-loading buffer by boiling, and proteins were separated by 3% to 8% tris-acetate gel electrophoresis (NuPage; Novex, San Diego, CA) and blotting to nylon membranes. GFRα-1 protein was detected by incubating membranes with an antibody against GFRα-1 (Santa Cruz). 

Investigation of GDNF-Induced Signal-Transduction Cascade by Immunoprecipitation and Western Blot Analysis

To investigate which protein kinase cascades are activated by GDNF, corneal epithelial cells from the cell line (5 × 10⁵ cells/75 cm²) were cultured in SHEM with 10% FBS for 1 day, followed by incubation in serum-free SHEM without additives or with recombinant human GDNF (200 ng/ml) for 10 to 40 minutes. Cells were solubilized in buffer containing 50 mM tris₂Cl (pH 8.0), 150 mM NaCl, 0.02% sodium azide, 100 μg/ml phenylmethylsulfonyl fluoride, 1 mM Na₃VO₄, 1% Triton X-100, and a mixture of several protease inhibitors (complete TM; Boehringer Mannheim; 1 tablet per 30 ml buffer). Total protein per lane (80 μg) was fractionated by 10% SDS-3-(N-morpholino)propanesulfonic acid (MOPS) bis-tris gel (NuPage; Novex) or 3% to 8% tris-acetate gel (Novex) and blotted onto a nitrocellulose membrane. Phosphorylated proteins were detected with phosphospecific antibodies and visualized with the enhanced chemiluminescence (ECL) Western blot analysis system (Amersham Pharmacia Biotech; Uppsala, Sweden). Antibodies against phospho-Raf(ser259), phospho-MEK1/2 (ser217/221), phospho-p90 ribosomal S6 kinase (RSK; ser381) and phospho-Elk-1 (ser383) were obtained from New England Biolabs (Beverly, MA). Antibody against phospho-FAK (tyr397/tyr407/tyr576/tyr577/tyr861/tyr925), and phosphotyrosine kinase (Pyk) 2 (tyr402/tyr579/tyr/580/tyr881) were purchased from Biosource (Camarillo, CA). Antibodies against phospho-MAPK (Erk-1 and -2; thr202/tyr204) came from New England Biolabs and Promega and an antibody against unphosphorylated Erk-1 and -2 from SantaCruz Biotechnology. A monoclonal antibody against phosphotyrosine was purchased from Sigma. 

Immunoprecipitation was used to detect tyrosine phosphorylation of Ret and paxillin, as shown above for GFRα-1. In brief, 100 μg total protein was incubated with protein G-agarose and polyclonal antibody against Ret or paxillin. Phosphorylated Ret and phosphorylated paxillin were detected by Western blot analysis with an antibody against phosphorylated tyrosine. 

Treatment with Protein Kinase Inhibitor

To determine whether GDNF-specific intracellular signals are transduced through MAPK and FAK the protein-tyrosine kinase inhibitor herbimycin A was used at a concentration of 10 μM. After 2 hours of incubation with or without inhibitor, cultures were stimulated with GDNF (200 ng/ml) for 30 minutes, followed by protein isolation and Western blot analysis. Because herbimycin is a broad-spectrum inhibitor that targets various tyrosine kinases, we further substantiated the effect of GDNF on MAPK signaling by use of manumycin (Sigma) at 10μ M. This inhibitor specifically interferes with membrane translocalization of Ras, thus blocking the step between Ret and phosphorylation of the ERK-activating MAPK kinase kinase Raf. ^{21 22}  

In Vitro Wound-Healing Assay and Modified Boyden Chamber Analysis

Because we could not obtain sufficient quantities of GDNF to conduct reproducible in vivo experiments, we investigated by using an in vitro wound-healing model and a modified Boyden chamber assay. 

Corneal epithelial cells were seeded at 5 × 10⁵ onto 60-mm plastic dishes with a 2-mm grid (Sarstedt, Nümbrecht, Germany) and cultured until subconfluence. The cell layer was injured by inducing several parallel scratches with a cell scrubber (Falcon; Becton Dickinson, Heidelberg, Germany). The tip of the scrubber was cut to measure approximately 1 mm, and, consequently, the width of the scratch in the cell layer was also approximately 1 mm. Selected areas were marked, and consecutive images were taken at different time points at ×5 magnification under an inverted phase-contrast microscope equipped with a video camera. After injury, dishes were incubated in SHEM either containing no growth factors (control) or rhGDNF (250 ng/ml), NGF (250 ng/ml; Boehringer Mannheim), or EGF (10 ng/ml). For each condition, four representative areas were evaluated within three dishes. The mean diameter of the scratch in each representative area was set as 100% at the beginning of the experiment. Eighteen hours later, the mean diameter of the same scratch was recalculated and expressed as a percentage of the diameter at the beginning of the experiment. Data were analyzed with Student’s t-test. 

To evaluate the chemotactic effect of GDNF, a modified Boyden chamber assay was used. Cells (4 × 10⁵) were seeded onto tissue culture inserts containing a polyethylene terephthalate (PET) filter with 8-μm pore size (Falcon; Becton Dickinson). Within 4 hours after seeding, most cells had attached to the filter and formed a semiconfluent monolayer. The medium was then changed to SHEM without additives in the upper well and SHEM with GDNF (250 ng/ml) in the lower well. After 24 hours, cells were removed from the surface of the insert by gentle scrubbing. Cells on the bottom of the insert (which had migrated through the filter) were fixed with −20°C methanol and stained with crystal violet. The surface of the filters was screened for cells under the microscope (10 fields/filter). Data were analyzed with Student’s t-test. 

Results

Transcription of GFRα-1 in Human Corneal Epithelium

GFRα-1 binds GDNF and provides specificity for GDNF signaling. Figure 1 shows amplification of a cDNA fragment specific for GFRα-1 (216 bp) from ex vivo corneal epithelium, primary cultured epithelial cells (ex vitro), and a corneal epithelial cell line. PCR fragments were cloned and sequenced. The presence of GFRα-1 in corneal epithelium suggests a regulatory function of GDNF on corneal epithelial cells through receptor-mediated intracellular signals. 

GDNF Induction of Formation of GFRα-1–Ret Complex and Phosphorylation of Ret

After stimulation with GDNF, the receptor GFRα-1 recruited Ret to the cell membrane and activated its receptor tyrosine kinase domain to induce intracellular signaling proteins. To demonstrate protein message for GFRα-1 in corneal epithelial cells and to investigate whether the GFRα-1–Ret complex is formed after exposure of corneal epithelial cells to GDNF, we performed coimmunoprecipitation of GFRα-1 and Ret and observed a time-dependent increase of GFRα-1–Ret complex formation (Fig. 2A) . In comparison with serum-free medium, the level of GFRα-1 protein bound to Ret increased between 5 and 15 minutes after exposure to 250 ng/ml GDNF. To show GDNF-induced activation of Ret, we performed immunoprecipitation experiments with an antibody against total Ret protein and Western blot analysis with an antibody against phosphorylated tyrosine. Tyrosine phosphorylation of Ret was induced within 5 minutes after addition of GDNF in comparison with serum-free control cultures (Fig. 2B , co). This effect was specifically blocked by the tyrosine kinase inhibitor herbimycin A. 

Activation of Intracellular Signals by GDNF in Cultured Corneal Epithelial Cells

To determine which signal-transduction pathways are induced by GDNF, we sought to detect GDNF-dependent phosphorylation of FAK family members and MAPK signaling components. Western blot analysis showed that tyrosine phosphorylation of FAK slightly increased within 10 and more significantly within 20 and 40 minutes (Fig. 3) . Pyk2, another member of the FAK family, was also phosphorylated within 40 minutes (Fig. 3) . 

Within the MAPK signaling cascade cRaf belongs to the MAPK kinase kinase (MAPKKK) family and serves as initiator for the propagation of MAPK signals. ²³ MAPK kinase (MEK)-1 is a key protein that mediates signaling cascades from MAPKKK to the two components of MAPK, Erk-1 and -2. ^{24 25} Activation of Erk facilitates its translocation into the nucleus where it phosphorylates transcription activators such as Elk. ²⁶ All these signaling components were activated by GDNF (Fig. 3) . Serine phosphorylation of cRaf, MEK1, and Elk, as well as tyrosine-threonine phosphorylation of Erk-1 and -2 were induced within 10 minutes, and the level of phosphorylation gradually increased over the observation period. In contrast, phosphorylation of ribosomal kinase (RSK) was not detectable (Fig. 3) . These results indicate that Ret signaling in the corneal epithelium is mediated by the FAK(Pyk2)-MAPK-Elk pathways. 

Effect of Herbimycin A and Manumycin on GDNF-Dependent Phosphorylation of Intracellular Signaling Proteins

Treatment of cells with the inhibitor herbimycin A resulted in a reduction of GDNF-dependent phosphorylation of FAK, cRaf, and Erk. It is interesting to note that not only tyrosine phosphorylation of FAK, but also serine phosphorylation of cRaf and tyrosine-threonine phosphorylation of Erk-1 and -2, were inhibited by the protein-tyrosine kinase inhibitor herbimycin A (Fig. 4 , GDNF+herbi.). This implies that activation of the cRaf-Erk pathway may also be dependent on phosphorylation of FAK, which represents an upstream regulator. ²⁷  

Because herbimycin inhibits various kinases, we further substantiated the effect of GDNF on MAPK signaling by using manumycin, which specifically inhibits Ras activation. ^{21 22} Treatment of cells with manumycin significantly inhibited GDNF-dependent phosphorylation of cRaf, MEK, Elk, and Erk-1 and -2 (Fig. 5) , indicating that GDNF activates Erk-signaling through Ras. Of note, the level of phosphorylated FAK induced by GDNF did not decrease in the presence of manumycin. This suggests that activation of FAK is independent of Ras. 

Induction by GDNF of Phosphorylation of Paxillin in Cultured Corneal Epithelial Cells

The focal adhesion protein paxillin plays an important role in cell motility and migration. Tyrosine phosphorylation of paxillin is essential for formation of paxillin-containing focal adhesions and paxillin-mediated actin reorganization. To evaluate the level of phosphorylated paxillin in corneal epithelial cells after stimulation with GDNF, immunoprecipitation with an antibody against paxillin and Western blot analysis with an antibody against phosphorylated tyrosine were performed. Tyrosine phosphorylation of paxillin was moderately induced by GDNF within 10 minutes and more significantly after 20 minutes. Herbimycin A decreased levels of phosphorylation below control levels (Fig. 6) . 

GDNF-Induced Migration of Corneal Epithelial Cells

After a “wound” was made in a subconfluent monolayer of primary human corneal epithelial cells in control medium (SHEM without additives; Fig. 7 , Co), less than 10% of the gap was filled within 18 hours (Fig. 7A) . GDNF, NGF, and EGF significantly increased in vitro wound healing, evident in the fact that 16.5% ± 6% of the gap was filled in the presence of 250 ng/ml GDNF, which represents a significant increase over the control (P < 0.01; Fig. 7A ). Similarly, NGF (250 ng/ml) or EGF (10 ng/ml) resulted in approximately 20% closure of the gap (19.6% ± 3.1% and 19.0% ± 8.6%, respectively). These data indicate that NGF and EGF significantly promote wound closure (P < 0.0001 and P < 0.0001, respectively). 

These results were confirmed by using corneal epithelial cells from the SV40-transformed cell line (Fig. 7B) . In control medium (Fig. 7B , Co) 20% ± 4.5% of the gap was filled after 18 hours but 30% ± 6.9% in the presence of GDNF (250 ng/ml; P < 0.001). Similarly, NGF (250 ng/ml) or EGF (10 ng/ml) resulted in 27% ± 7.1% and 44.6% ± 9% wound closure (P < 0.0001 and P < 0.0001, respectively; Fig. 7B ). 

To further confirm the effect of GDNF on cell migration, a modified Boyden chamber analysis was performed. In control medium (SHEM without growth factors) only a few cells (19 ± 6.8) migrated through a filter of 8-μm pore size (Figs. 8A 8D) . However, when 250 ng/ml GDNF was added to the lower well, the number of cells that migrated through the pores of the filter increased more than sixfold (117 ± 37.6; P < 0.0001; Figs. 8B 8D ). 

Inhibition of GDNF-Induced Wound Healing and Migration by Herbimycin A

To further test the significance of the FAK-MAPK signaling pathways during wound healing, we investigated the effect of herbimycin A on GDNF-mediated cell migration in an in vitro wound-healing model. Herbimycin A blocked cell migration in the presence of GDNF. In a representative culture containing 250 ng/ml GDNF, 37.9% ± 8.1% of the wound gap was filled with cells after 18 hours. In the presence of GDNF and 10 μM herbimycin A, only 19.2% ± 5.5% of the gap was closed (P < 0.001). Similarly, migration through filters was inhibited by herbimycin A (Figs. 8C 8D) . Here the number of cells migrating through the filter in the presence of both GDNF and herbimycin (Figs. 8C 8D ; GDNF+INH) was significantly smaller than in the presence of GDNF alone (P < 000.1; Figs. 8B 8D , GDNF). 

Discussion

The response of corneal epithelial cells to a full-thickness wound initially consists of the transition from a quiescent state to a migratory state. After reduction of desmosomal adhesion complexes and changes in cell shape, the wound margin retracts, and adjacent cells slide onto the denuded basement membrane until the wound is closed. ^{28 29} Therefore, cell migration seems to be one of the most important factors in corneal wound healing. Many cytokines that stimulate corneal wound healing in laboratory animals also induce migration of corneal epithelial cells in vitro. This effect has recently been described for NGF, which is the only cytokine successfully used in patients to enhance corneal epithelial healing. ³⁰ In the current study, we report that a second neurotrophic factor, GDNF, promotes in vitro wound healing and migration of human corneal epithelial cells in tissue culture. These findings suggest that GDNF could also be used to enhance corneal wound healing in laboratory animals and humans. To further test this hypothesis we treated one patient with a progressive neurotrophic corneal ulcer unresponsive to conventional and medical therapy with recombinant human GDNF. In this patient, complete epithelial healing was achieved during 3 weeks of topical application (data not shown). Further clinical studies are needed to confirm the efficacy of such treatment for impaired wound healing. 

Our findings are consistent with compiled evidence concerning the function of GDNF in other organs. It is interesting that differential expression of GDNF is not limited to the cornea. For instance, mRNA specific for GDNF that is not present in corneal epithelial cells is also undetectable in the ureteral epithelium. ^{31 32} However, the underlying nephrogenic mesenchyme expresses message specific for GDNF that is also present in corneal stroma. ^{31 32} These observations suggest that GDNF may function as a paracrine morphogen secreted by cells of mesenchymal or neuroectodermal origin to modulate epithelial cells. The functional significance of GDNF is implied by the observation that a targeted disruption of the GDNF gene leads to severe kidney malformations. This indicates that this protein is important for coordinated development during ureteral bud outgrowth and branching. ^{33 34} In concordance with our findings in the cornea, recent in vivo studies using renal Madin-Darby canine kidney (MDCK) cells have shown that recombinant GDNF induces cell motility, loss of cell adhesion, and increased migration toward a localized source of GDNF. ³⁵ In summary, these observations suggest that GDNF can act as a paracrine modulator of epithelial cell function, such as cell migration during development and postnatal life. 

In a variety of cells, intracellular signals initiated by GDNF are mediated primarily by Ret, which is a receptor tyrosine kinase. ^{11 12} This makes GDNF different from other members of the TGF-β superfamily, a group of proteins that was previously thought to exclusively signal through serine-threonine kinases. ³⁶ Unlike most other tyrosine kinases, Ret cannot bind the ligand on its own but needs the glycosyl-phosphatidylinositol–linked coreceptor, GFRα. Translocation of Ret to its anchored coreceptor seems to be important for sufficient downstream signaling. ³⁷ In this study we have provided evidence that GFRα is expressed in corneal epithelial cells and that phosphorylation of Ret is induced by exposure of corneal epithelial cells to GDNF. Signals deriving from receptor tyrosine kinases are often transduced by the Ras-MAPK cascade, which plays a pivotal role in mediating growth factor-dependent cell growth and differentiation. ^{23 24 25 26} In corneal epithelial cells NGF, substance P, platelet activating factor, keratocyte growth factor (KGF), and hepatocyte growth factor (HGF) have recently been shown to activate MAPK. ^{7 38 39 40}  

GDNF-induced signaling has been investigated in a variety of neuronal cells. In a motoneuron hybrid cell line, GDNF induced activation of Ras, which is one of the initial steps in the MAPK signaling pathway. ⁴¹ Furthermore phosphorylation of Ret induces activation of c-Jun NH2 terminal kinases (JNK) such as JNK 1, which belong to the MAPK signaling system. ⁴² Activation of JNK1 by GDNF-Ret has been shown to be due to a pathway that is different from that leading to activation of another component of MAPK—that is, Erk 2 by GDNF. Finally, phosphorylation of Ret by GDNF induces activation of Elk, which induces a transcriptional response downstream of the MAPK signaling cascade. ⁴³  

Our results now show that GDNF-induced Ret phosphorylation causes activation of Erk-signaling in corneal epithelial cells in a time-dependent fashion. The finding that manumycin blocks this activation suggests that Ras serves as upstream regulator of Ret-mediated signaling in corneal epithelial cells. Similar to cells of neuronal origin, GDNF induces phosphorylation of MAPK, resulting in activation of Elk, which then evokes transcriptional responses. ²⁶ In contrast, phosphorylation of RSK, another transcription activator that is also controlled by MAPK is not augmented by GDNF. It is notable that RSK can be activated by NGF through MAPK. ⁴⁴ This observation indicates that various neurotrophic factors (such as NGF and GDNF) can activate different transcriptional regulators through the same signal transduction pathway. 

FAK is a nonreceptor tyrosine kinase that localizes to sites of integrin receptor clustering. FAK becomes phosphorylated at several sites when integrin interacts with matrix proteins. ⁴⁵ In this study we have shown that tyrosine phosphorylation of FAK increases in the presence of GDNF and that this effect is inhibited by herbimycin. The observation that Ret-mediated phosphorylation of FAK in corneal epithelial cells was not inhibited by manumycin suggests that FAK phosphorylation is independent of Ras. These results are supported by the observation that GDNF induces phosphorylation of FAK in neuroblastoma cells. ⁴⁶ Although the mechanism is not clear, FAK phosphorylation seems to be mediated by a Rho-dependent pathway downstream of phosphatidylinositol-3′kinase. ⁴⁶ Alterations of tyrosine phosphorylation status of FAK can be induced by growth factors, such as EGF and platelet-derived growth factor (PDGF) in several cell types. ⁴⁷  

Recent studies suggest that FAK not only interacts with integrins but also responds to growth factor receptors. ^{47 48} Furthermore, FAK links growth factor receptor– and integrin-signaling pathways. ⁴⁸ FAK associates with activated receptor tyrosine kinases such as EGF-R or PDGF-R. ⁴⁸ The site of this interaction the N-terminal domain (band 4.1), whereas interaction with integrins takes place in the C-terminal domain. ⁴⁸ Interaction of EGF-R with FAK leads to phosphorylation at the Y397 site, and this event is necessary for EGF-induced cellular motility. ⁴⁸ Therefore, for growth factor–activated cellular motility, FAK phosphorylation may be initiated by several receptor tyrosine kinases. ⁴⁸ In this respect, FAK seems to be an important link that bridges growth factor– and integrin-mediated intracellular signals. 

Because the most important function of FAK is related to cell migration, ⁴⁹ our data suggest that FAK could be a regulatory element of GDNF-induced migration in corneal epithelial cells. FAK has been shown to bind and phosphorylate a variety of adapter and signaling molecules such as paxillin, which is associated with protein tyrosine phosphatase and can regulate integrin-mediated phosphorylation. ^{50 51 52} Ret-dependent phosphorylation of paxillin occurs during cell migration and represents an event that is crucial for a coordinated rearrangement of cytoskeleton proteins such as actin filaments. ⁴⁶ Our results show that paxillin is phosphorylated in response to GDNF, suggesting that Ret-GFRα–mediated phosphorylation of FAK induces phosphorylation of paxillin in corneal epithelial cells. Support for this novel function of the Ret-GFRα signaling pathway comes from the notion that Ret regulates several cellular events, such as cell motility and cell migration, which are mediated by FAK. ⁴⁶  

Besides receptor tyrosine kinases, protein kinase C (PKC) and FAK can regulate the MAP kinase pathway, and these kinases have been reported to modulate several steps of the MAPK cascade. ^{27 53} In particular, FAK induces activation of the Ras-MAPK pathway by forming a complex with the GRB2 adapter protein in NIH 3T3 fibroblasts. As a mediator between growth factor receptor tyrosine kinase and the Ras-MAPK pathway, GRB2 associates with the Ras guanosine diphosphate–guanosine triphosphate (GDP-GTP) exchange protein Sos to regulate Ras activation. ⁵⁴ The observation that PTEN, which dephosphorylates FAK has an inhibitory effect on Erk activation could help in the understanding of the potential role of FAK as regulator of multiple signal-transduction pathways in the cornea. ⁵⁵ Furthermore, the use of specific FAK inhibitors that are not commercially available at this time could be used to further substantiate the role of FAK in corneal epithelial migration. 

In summary, our findings suggest that GDNF-dependent activation of GFRα-1 and Ret in corneal epithelial cells induces phosphorylation of Ras and FAK(Pyk2) followed by Ras-dependent activation of the MAPK (Erk) pathway to initiate gene transcription and FAK-dependent phosphorylation of paxillin to initiate cell migration (Fig. 9) . In this system, FAK could function as a mediator for multiple signals derived from GDNF receptors and integrins. The effect of GDNF on cell migration and proliferation is mediated by a complex intracellular signal network. Further investigation of signaling pathways in corneal epithelial cells should augment current knowledge concerning the basic mechanisms of cell migration and wound healing. 

 Supported by Deutsche Forschungsgemeinschaft Grant Kr 993/12-1 and Gertrud Kusen Stiftung, Hamburg, Germany.

 Submitted for publication January 8, 2001; revised April 25, 2001; accepted May 31, 2001.

 Commercial relationships policy: F.

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked“ advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

GFRα-1 in ex vivo and cultured corneal epithelium. Transcription of GFRα-1 (216-bp fragment) was detected by RT-PCR in ex vivo corneal epithelium (ex vivo), primary cultured corneal epithelial cells (ex vitro), and a corneal epithelial cell line (cell line). A representative experiment is shown.

Figure 2.

 

Time-dependent formation of the Ret-GFRα-1 complex and tyrosine phosphorylation of Ret by GDNF (Western blot). (A) For immunoprecipitation, cell lysates were incubated with agarose beads and an antibody against Ret (IP: Ret Ab). Immunoprecipitated proteins were subjected to Western blot analysis and detected by an antibody against GFRα-1 (IB: GFRa-1Ab). In serum-free medium (co) very small quantities of activated GFRα-1 protein (∼60 kDa), which can form a complex with Ret, were detected. After addition of GDNF (200 ng/ml), a gradual, time-dependent increase was observed over 15 minutes. GFRα-1 was not detected in cell lysate incubated with beads alone. Equal loading of total Ret protein is shown in the bottom panel. (B) Cell lysates were incubated with agarose beads and antibody against Ret for immunoprecipitation (IP: Ret Ab). Immunoprecipitated proteins were subjected to Western blot analysis and detected by an antibody against phosphorylated tyrosine (IB: P-Tyr Ab). In serum-free medium (co) the level of phosphorylated Ret (∼170 kDa) was relatively low. After addition of GDNF (200 ng/ml) the level increased time dependently over 15 minutes. Tyrosine phosphorylation of Ret in cells pretreated with herbimycin A and stimulated with GDNF (for 10 minutes) remains at a lower level than in cells without incubation of herbimycin A. Equal loading of total Ret protein is shown in the bottom panel. A representative experiment is shown.

Figure 3.

 

Time-dependent phosphorylation of intracellular signals by GDNF (Western blot analysis). Tyrosine phosphorylation of FAK (∼130 kDa) and Pyk2 (∼130 kDa); serine phosphorylation of cRaf (∼80 kDa), MEK1 (45 kDa), and Elk (∼60 kDa); and tyrosine-threonine phosphorylation of Erk-1 (44 kDa) and -2 (42 kDa) gradually increased within 40 minutes after exposure to GDNF. Phosphorylation of p90RSK (RSK; 90 kDa) was not induced in response to GDNF stimulation. To show that equal amounts of total protein were loaded in each lane (80 μg) a blot with an antibody against total Erk (phosphorylated and unphosphorylated Erk-1 and -2) is presented. A representative experiment is shown.

Figure 4.

 

Inhibition of GDNF-dependent phosphorylation of FAK, cRaf, and Erk by herbimycin A (Western blot analysis). GDNF-dependent tyrosine phosphorylation of FAK, serine phosphorylation of cRaf, and tyrosine-threonine phosphorylation of Erk-1 and -2 were significantly decreased after preincubation with herbimycin A for 2 hours and incubation with GDNF for 30 minutes. Total Erk (phosphorylated and unphosphorylated Erk-1 and -2) remained unchanged. A representative experiment is shown.

Figure 5.

 

Effect of manumycin on GDNF-dependent phosphorylation of FAK, cRaf, MEK, Erk, and Elk (Western blot analysis). GDNF-dependent serine phosphorylation of cRaf, MEK1, and Elk and tyrosine-threonine phosphorylation of Erk-1 and -2 were significantly decreased after preincubation with manumycin for 2 hours and incubation with GDNF for 30 minutes. GDNF-dependent tyrosine phosphorylation of FAK was not inhibited by manumycin. Total Erk (phosphorylated and unphosphorylated Erk-1 and -2) remained unchanged. A representative experiment is shown.

Figure 6.

 

Time-dependent tyrosine phosphorylation of paxillin by GDNF (Western blot analysis). Cell lysates were incubated with agarose beads and antibody against paxillin for immunoprecipitation. Immunoprecipitated proteins were subjected to Western blot analysis and detected by antibody against phosphorylated tyrosine. The level of tyrosine-phosphorylated paxillin (∼70 kDa) was low in serum-free medium (co). Addition of GDNF (200 ng/ml) caused a time-dependent increase over 20 minutes. Tyrosine phosphorylation of paxillin in herbimycin A–pretreated cells stimulated with GDNF (for 10 minutes) remained at a lower level than in cells without incubation of herbimycin A. The level of total paxillin remained unchanged. A representative experiment is shown.

Figure 7.

 

Effect of GDNF on in vitro closure of “wounds” in semiconfluent monolayers. Closure of scratch wounds of 1-mm diameter was significantly (*P < 0.01) enhanced by 250 ng/ml GDNF, 250 ng/ml NGF, or 10 ng/ml EGF in comparison with control cultures (co) in both primary corneal epithelial cells (A) and a corneal epithelial cell line (B). Data are expressed as a percentage ± SD of the initial wound gap in representative cultures at 18 hours. A representative experiment is shown (n = 12).

Figure 8.

 

Effect of GDNF on corneal epithelial cell migration in a modified Boyden chamber system. (A) In control medium, only a few (19 ± 6.8) corneal epithelial cells migrated from the upper chamber through the filter. (B) Addition of 250 ng/ml GDNF into the lower chamber resulted in a sixfold increase of cell migration through the filter (117 ± 37.6; P < 0.0001). (C) The presence of herbimycin A resulted in a significant inhibition of GDNF-induced cell migration. (D) Graphic representation of response ± SD in each condition. A representative experiment is shown (n = 10).

Figure 9.

 

Proposed activation of signal-transduction pathways by GDNF in corneal epithelial cells. After binding to GDNF, GFRα-1 dimerizes and forms a complex with Ret that is phosphorylated. This induces phosphorylation of FAK and Pyk2 and, consequently, FAK-dependent phosphorylation of paxillin, leading to actin reorganization. GDNF also induces activation of the Ras-MAPK pathway for Elk-dependent, p90Rsk-independent gene transcription in corneal epithelial cells. Both phosphorylated Pyk2 and FAK act as upstream regulators of Ras and the Ras-MAPK pathway.

The authors thank Biopharm GmbH, Heidelberg, Germany, for kindly providing GDNF.