Molecular Vision 2003; 9:624-634 <>
Received 28 August 2003 | Accepted 5 December 2003 | Published 8 December 2003

FGF-2 induced reorganization and disruption of actin cytoskeleton through PI 3-kinase, Rho, and Cdc42 in corneal endothelial cells

Hyung Taek Lee,1 EunDuck P. Kay1,2

1Doheny Eye Institute and 2Department of Ophthalmology, Keck School of Medicine, University of Southern California, Los Angeles, CA

Correspondence to: EunDuck P. Kay, Ph.D., Doheny Eye Institute, 1450 San Pablo Street, DVRC 203, Los Angeles, CA; Phone: (323) 442-6625; FAX: (323) 442-6688; email:


Purpose: Corneal endothelial cells (CECs) undergo endothelial to mesenchymal transformation (EMT) in response to FGF-2 stimulation. One phenotypic change that occurs during EMT is a change in cell shape from polygonal to elongated fibroblast-like cells. We investigated whether FGF-2 plays a role in this morphogenetic pathway by reorganizing actin cytoskeleton through the actions of phosphatidylinositol (PI) 3-kinase and the Ras related Rho family of small guanosine triphosphatases (GTPases).

Methods: Cell morphology was analyzed using phase contrast microscopy, and the organization of actin cytoskeleton and focal adhesions were analyzed by immunofluorescent staining. Expression of vinculin and β-actin was determined by immunoblot analysis. Pharmacologic inhibitors (LY294002, C3 exoenzyme, Y27632, or PD98059) or neutralizing antibody to FGF-2, respectively, were used to block PI 3-kinase, Rho, Rho associated kinase, extracellular signal regulated kinase, or FGF-2 pathways.

Results: CECs treated with FGF-2 became smaller and lost their characteristic polygonal cell morphology. Such cell shape change was completely blocked by treatment with LY294002. CECs in culture have abundant stress fibers that are oriented radially across the cell. However, FGF-2 caused a loss of these stress fibers and focal adhesions. The modulated cells contained a cortical actin ring while LY294002 completely abolished this action of FGF-2 on actin cytoskeleton. Treatment of cells with C3 exoenzyme or Y27632 in the presence of FGF-2 induced spindle shaped cells with prominent pseudopodia which were rapidly formed upon exposure to the inhibitor. The expression level of vinculin was found to be similar in all experimental conditions but vinculin was mostly translocated to the cytoplasm in response to FGF-2 stimulation. CECs plated on Matrigel matrix demonstrated findings similar to those from cells plated on the conventional culture dishes, except that Matrigel facilitated the formation of pseudopodia. We further investigated in vivo actin organization using organ cultures of corneal endothelium (CE) on Descemet's membrane. The contact inhibited endothelial monolayer demonstrated a circumferential actin ring, and no stress fibers were observed. When CE was treated with FGF-2, a half population of CE lost its characteristic contact inhibited cobblestone morphology. Actin cortex was greatly disrupted in these modulated cells. Both neutralizing antibody to FGF-2 and LY294002 completely impeded the modulating activity of FGF-2 on the endothelial monolayer. When CE was simultaneously treated with FGF-2 and Y27632, the circumferential actin cortex was greatly disrupted and the endothelial monolayer was transformed into multi-layers of fibroblastic cells containing pseudopodia. Both LY294002 and neutralizing antibody to FGF-2 antagonized the actions of FGF-2 and Y27632.

Conclusions: These data indicate that CECs in culture have constitutively active Rho activity as evidenced by stress fiber formation and that PI 3-kinase negatively regulates the formation of stress fibers and focal adhesions, perhaps antagonizing the Rho pathways. Formation of pseudopodia in response to FGF-2 and Y27632 may suggest that the Rho/ROCK pathway negatively regulates Cdc42.


Corneal fibrosis represents a significant pathophysiological problem that causes blindness by physically blocking light transmittance. One clinical example of corneal fibrosis observed in corneal endothelium is the development of a retrocorneal fibrous membrane (RCFM) in Descemet's membrane [1,2]. In RCFM, corneal endothelial cells (CECs) are converted to fibroblast-like cells. The contact inhibited phenotypes of normal CECs are lost, subsequently leading to multi-layers of the modulated cells [3,4]. These morphologically altered cells simultaneously resume their proliferation ability and deposit a fibrillar extracellular matrix (ECM) in the basement membrane environment. In our previous studies, we have reported that FGF-2 is the direct mediator for such endothelial to mesenchymal transformation [5-8]. Among the phenotypes altered during endothelial mesenchymal transformation, we have reported that FGF-2 directly regulates the cell cycle progression through the action of phosphatidylinositol (PI) 3-kinase, thus leading to a marked stimulation of cell proliferation [9,10]. We also reported that FGF-2 induces production of type I collagen and fibronectin, the two major components of fibrotic RCFM tissue [3,11]. However, it is not known how CECs undergo morphological alteration to acquire the fibroblastic cell morphology. In normal cornea, FGF-2 is a component of Descemet's membrane that may be necessary for wound repair [6,12,13]. In order to understand the whole spectrum of endothelial to mesenchymal transformation, we investigated the morphogenetic pathways mediated by FGF-2 in CECs.

It is now clear that the actin cytoskeleton with its polymerization dynamics is central to many aspects of cellular activities, such as cytokinesis, phagocytosis, cell migration, adhesion, polarity, and morphology [14-18]. Assembly and organization of the actin cytoskeleton is controlled by the Ras related Rho family of small guanosine triphosphatases (GTPases), Rho, Rac, and Cdc42 [19-22]. In particular, these GTPases have been shown to induce morphological changes associated with actin polymerization. In Swiss 3T3 fibroblasts, Rac1 induces membrane ruffling and lamellipodium formation, RhoA induces the formation of stress fibers, and Cdc42 induces the formation of microspikes and filopodia, all of which are dependent on filamentous actin (F-actin) organization [23-26]. These earlier studies also demonstrated that Rho GTPases are key signal transducers, mediating growth factor induced changes to the actin cytoskeleton [23,24]. As stated above, FGF-2 induced a change in the cell shape of CECs from a polygonal to a fibroblastic morphology. We hypothesized that these morphologic differences of CECs were related to changes in the cytoskeletal organization of cells as observed in many other cell systems. Our previous study reported that CECs in culture demonstrated a serum induced stress fiber formation [8], whereas a circumferential cortical actin mat was observed in corneal endothelium in vivo [27,28] and bovine CECs in culture when maintained in the absence of serum [29]. We also reported that FGF-2 altered the cell shape of CECs through PI 3-kinase and that these morphologically altered CECs demonstrated accumulation of cortical actin at the plasma membrane [8]. In the present study we tested whether FGF-2 antagonizes Rho activity, thus leading to the loss of stress fibers and to reorganization of actin cytoskeleton at the cortex. We further investigated whether there was the possibility of competition between Rho and Cdc42 in the FGF-2 mediated change in cell morphology, as observed in NIE-115 neuroblastoma cells, in which microinjection of C3 exoenzyme (Rho inhibitor) induced an increased formation of both filopodia and lamellipodia [30].

The cytoskeletal organization of cells grown in tissue culture is often very different from that of cells in living organisms. This casts some doubt on whether information that comes from studying actin dependent cellular processes in cells cultured under these conditions is physiologically relevant. We, therefore, compared the actin cytoskeleton of cells plated on tissue culture substratum with that of cells plated on Matrigel matrix, the structural components of which are similar to the basement membrane of CECs (type IV collagen and laminin). We also compared the actin cytoskeleton of the cultured CECs to that of organ cultured corneal endothelium, which maintains the in vivo phenotype.

In the present study, we show that FGF-2 alters cell morphology of CECs through the action of PI 3-kinase and that the cell shape change is closely related to the organization pattern of actin cytoskeleton. We further demonstrate that FGF-2 caused the loss of stress fibers and focal adhesions, but this action of FGF-2 is not sufficient to cause activation of Cdc42 pathways. Inhibitors of Rho or Rho associated kinase (ROCK) pathways are required to further promote the cell shape change to spindle shaped cells accompanied by the formation of prominent pseudopodia.



FGF-2 was purchased from Intergen (Purchase, NY) and neutralizing antibody to FGF-2 was purchased from Upstate Biotechnology Inc. (Lake Placid, NY). Monoclonal antibody against β-actin and LY294002 was purchased from Sigma-Aldrich (St. Louis, MO). Rhodamine-phalloidin was purchased from Molecular Probes (Eugene, OR). Monoclonal antibody against vinculin and fluorescein isothiocyanate (FITC) and rhodamine conjugated secondary antibodies were purchased from Chemicon (Temecula, CA). Y27632 and C3-exoenzyme were purchased from Calbiochem (San Diego, CA). Growth factor reduced Matrigel was purchased from BD Biosciences (San Diego, CA). Mounting solution and biotinylated secondary antibodies were purchased from Vector Laboratories (Burlingame, CA).

Cell cultures

Rabbit eyes were purchased from Pel Freez (Rogers, AR). Isolation and establishment of rabbit CECs were performed as previously described [31]. Briefly, the corneal endothelium-Descemet's membrane complex was treated with 0.2% collagenase and 0.05% hyaluronidase (Worthington Biochemical, Lakewood, NJ) for 90 min at 37 °C. Primary cultures were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 20% fetal calf serum (FCS) and 50 μg/ml of gentamicin in a 5% CO2 incubator. First passage CECs maintained in DMEM containing 10% FCS (DMEM-10) were used for all experiments. For subculture, confluent cultures were treated with 0.05% trypsin and 5 mM EDTA in phosphate buffered saline (PBS) for 5 min. When cells were treated with FGF-2, heparin (10 μg/ml) was added to the cultures since our previous study showed that CECs require supplemental heparin for FGF-2 activity to occur [5]. In some experiments, pharmacologic inhibitors were used in the presence of FGF-2 stimulation. These were LY294002 (PI 3-kinase, 20 μM), PD98059 (extracellular signal regulated kinase, ERK, 10 μM), C3 exoenzyme (Rho, 5 μg/ml), or Y27632 (Rho associated kinase, ROCK, 10 μM). Optimal concentration of each inhibitor was determined prior to using it in experiments. In most experiments, cells plated on the conventional culture plates and cells plated on Matrigel coated dishes were analyzed in parallel. For thin coating with Matrigel, diluted Matrigel solution (25 μl stock solution in 1 ml of serum free medium) was added to the petri dishes and incubated at 37 °C for 60 min. The unbound material was aspirated and the dish was gently rinsed with serum free medium. Phase contrast micrographic analysis was performed using a digital camera (Diagnostic Instruments Inc., Iowa City, IA). For organ cultures, the corneal endothelium-Descemet's membrane complex was peeled as one piece from young rabbit eyes. The isolated tissue was incubated for 4 days in one of the following conditions; DMEM-10, FGF-2, FGF-2 + LY294002, FGF-2 + neutralizing antibody to FGF-2, FGF-2 + Y27632, FGF-2 + Y27632 + LY294002, FGF-2 + Y27632 + neutralizing antibody to FGF-2, or FGF-2 and all three reagents.

Immunofluorescent staining and confocal microscopy

CECs plated in 4 well chamber slides were treated with the respective experimental conditions. Cells were fixed with 4% paraformaldehyde in PBS for 5 min and permeabilized in PBS containing 0.1% Triton X-100, then blocked with 2% bovine serum albumin (BSA) in PBS for 60 min at room temperature. The subsequent incubation was carried out with 2% BSA in PBS, and all washes were carried out in PBS containing 0.1% Triton X-100 at room temperature. Cells were incubated with anti-vinculin antibody (1:400 dilution) for 1 h at 37 °C and then with FITC conjugated secondary antibody (1:200 dilution) for 1 h at 37 °C in the dark. For F-actin staining, cells were incubated with rhodamine-phalloidin (1:100 dilution) at 37 °C for 10 min and then rinsed. After extensive washing, the slides were mounted in a drop of Vectashield mounting medium (Vector Laboratories Inc.) to reduce photobleaching. Control experiments, performed in parallel with the omission of the primary antibodies, did not show the activity. Antibody labeling was examined using a Zeiss LSM-510 laser scanning confocal microscope. Optical slices (1.8 μM) were taken perpendicular to the cell monolayer (apical to basal orientation). A 488 nm argon laser was used in combination with a 499/505-530 nm excitation/emission filter set for fluorescein examination. For rhodamine, the 543 nm helium neon laser was used with a 543 nm excitation filter and a 560 nm emission filter. The captured images were then pseudocolored red for rhodamine and green for FITC. Image analysis was performed using the standard system operating software provided with the Zeiss LSM-510 series microscope. All illustrations were assembled and processed digitally using Adobe Photoshop 7 (Adobe, San Jose, CA).

Protein preparation and protein determination

Cells were washed with ice cold PBS and then lysed with cell lysis buffer (20 mM HEPES, pH 7.2, 10% glycerol, 10 mM Na3VO4, 50 mM NaF, 1 mM phenylmethylsulfonyl fluoride [PMSF], 0.1 mM dithiothreitol, 1 μg/ml leupeptin, 1 μg/ml pepstatin, and 1% Triton X-100) on ice for 30 min. The lysate was subjected to sonication on ice. The cell homogenates were then centrifuged at 15,000x g for 10 min. Protein concentration of the resultant supernatant was assessed with a Bradford reagent.

SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblot analysis

The conditions of electrophoresis were as described by Laemmli [32], using the discontinuous Tris-glycine buffer system. Protein (30 μg) was loaded on an 8% SDS-polyacrylamide gel and separated under the reduced condition. The proteins separated by SDS-PAGE were transferred to a nitrocellulose membrane (Bio-Rad Laboratories, Hercules, CA) at 0.25 ampere for 10 h in a semidry transfer system (transfer buffer; 25 mM Tris-HCl, pH 8.3, 190 mM glycine, 20% methanol). Immunoblot analysis was performed using a commercial ABC Vectastain kit (Vector Laboratories). All washes and incubations were carried out at room temperature in TTBS (0.9% NaCl, 100 mM Tris-HCl, pH 7.5, 0.1% Tween 20). Briefly, the nitrocellulose membrane was immediately placed in the blocking buffer (5% nonfat milk in TTBS) and kept for 1 h at room temperature. Incubations were carried out with primary antibodies (1:1000 dilution) for 2 h, with biotinylated secondary antibody (1:5000 dilution) for 1 h, and with ABC reagent for 30 min. The membrane was treated with the enhanced chemiluminescence (ECL) reagent (Amersham Pharmacia Biotech, Buckinghamshire, UK), and the ECL-treated membrane was exposed to ECL film.


Effect of FGF-2 on cell shape and actin cytoskeleton in cultured CECs

In our previous study, we found that FGF-2 caused CECs to lose their characteristic polygonal cell morphology [8]. We also reported that neutralizing antibody to PI 3-kinase and inhibitors of PI 3-kinase (wortmanin and LY294002) caused the modulated morphology induced by FGF-2 to revert to a normal polygonal cell shape. In the present study, we attempted to confirm that this morphogenetic activity of FGF-2 was mediated by PI 3-kinase, and we further determined whether ERK1/2 (a member of the mitogen-activated protein kinase, MAPK), the sustained activation of which is required in many signaling pathways, was also involved. The subconfluent cultures maintained in growth medium (DMEM-10) showed a spread and polygonal morphology. Neither 20 μM LY294002 nor 10 μM PD98059 altered the cell shape (Figure 1A). On the other hand, cells treated with FGF-2 lost the characteristic polygonal morphology and became smaller, whereas cells simultaneously treated with FGF-2 and LY294002 reverted in shape to the spread and polygonal phenotypes similar to those observed in cells in DMEM-10 (CECs maintained in DMEM-10 were thereafter designated as control cells). Unlike the PI 3-kinase inhibitor, PD98059 did not alter the cell shape caused by FGF-2 (Figure 1A).

Another set of cultures was maintained in DMEM-10, with or without continuous FGF-2 stimulation, until they reached confluence. These cells were then treated with inhibitor for 24 h. Control cells demonstrated a monolayer of polygonal cells. Neither inhibitor altered the polygonal and contact inhibited monolayer of CECs (Figure 1B). On the other hand, cells treated with FGF-2 lost the characteristic phenotypes of polygonal monolayer, whereas LY294002 caused the complete reversion of the FGF-2 modulated phenotypes to the endothelial monolayer. Again, PD98059 did not block the action of FGF-2 on cell shape. These data suggest that FGF-2 modulates cell shape of CECs through the action of PI 3-kinase and that the MAPK pathway is not involved in the morphogenetic pathway of FGF-2.

We then double stained the cells for F-actin and vinculin to determine whether the FGF-2 mediated cell shape change was related to the actin cytoskeleton and focal adhesions. Cells plated and maintained in DMEM-10 for 1, 2, or 3 days showed stress fiber organization and focal adhesions (Figure 2). Interestingly, the rapidly growing CECs developed well defined stress fibers, oriented radially across the cell. Focal adhesions were formed simultaneously in a time dependent manner. On the other hand, CECs treated with FGF-2 demonstrated a gradual loss of the serum induced stress fibers. Actin cytoskeleton was reorganized into a cortical actin ring in a time dependent manner. Cells stimulated with FGF-2 for 24 h demonstrated a partial disruption of stress fibers. Accumulation of the cortical actin ring was largely apparent in the cells exposed to FGF-2 for 2 days. CECs stimulated with FGF-2 for 3 days were devoid of all visible stress fibers, but all cells acquired a cortical actin ring structure (Figure 2). Focal adhesion formation was simultaneously markedly decreased in a time dependent manner. After 3 days in FGF-2, vinculin staining was mostly observed in the cytoplasm. Loss of stress fibers and focal adhesions mediated by FGF-2 was further investigated to determine whether PI 3-kinase regulated these cellular activities. LY294002 did not alter the serum activated cytoskeletal arrangement in the control cells. On the other hand, the actin cortical ring structure mediated by FGF-2 was completely lost when cells were simultaneously treated with FGF-2 and LY294002. Stress fibers and focal adhesions were reassembled, suggesting that the action of FGF-2 on cytoskeletal architecture is mediated through the action of PI 3-kinase.

The data presented in Figure 2 demonstrate that the subconfluent CECs acquire serum mediated activation of the Rho/ROCK pathway and that FGF-2 antagonizes the action of Rho through PI 3-kinase, as evidenced by the loss of stress fibers. Therefore, we examined the effect of C3 exoenzyme and Y27632 in the cells treated with FGF-2. In the absence of FGF-2 stimulation, both inhibitors blocked the serum mediated stress fiber formation and reorganized actin cytoskeleton at the cortex (data not shown). Again, FGF-2 caused a cortical actin ring with cytoplasmic staining of vinculin (Figure 3A). On the other hand, simultaneous treatment of cells with FGF-2 and either C3 exoenzyme or Y27632 caused a spindle shaped cell with extending pseudopodia, which was positively stained with rhodamine-phalloidin (Figure 3A). Some cells with pseudopodia also exhibited laterally placed ruffles or ruffles localized to the distal ends of the pseudopodia. A similar phenomenon in which pseudopodia and lamellipodia were formed upon activation of Cdc42 was observed in NIH3T3 fibroblasts [33]. Induction of pseudopodia by blocking ROCK in the presence of FGF-2 stimulation was further determined in a time dependent manner (Figure 3B). Pseudopodia formation was very rapid. A 10 min treatment of cells with Y27632 was able to induce pseudopodia. After 30 min of exposure to the inhibitor, all cells developed pseudopodia, suggesting that inhibition of Rho pathways causes a rapid activation of Cdc42. Organization of pseudopodia was completely blocked by either neutralizing antibody to FGF-2 or LY2940002, suggesting that pseudopodia formation (or blocking of pseudopodia formation) is also a downstream event to PI 3-kinase (Figure 3C).

Since stress fiber formation is under the control of Rho, the expression level of Rho was measured in cells treated with FGF-2 with or without inhibitors (LY294002, C3 exoenzyme, or Y29632). Control cells showed a high level of Rho and FGF-2 did not alter the expression level of the serum activated Rho. None of the inhibitors caused differential expression of Rho at protein levels (data not shown). Numerous attempts to measure Rho activity were made using a Rho-GTP pull down assay, as described previously [34,35]. The Rho binding domain of Rhotekin was used to precipitate cellular Rho-GTP. A positive control using GTP-γS, the non-hydrolyzable analog of GTP, and a negative control using GDP were measured in the cell lysates of the subconfluent CECs. These were accordingly high in GTP-γS-loading conditions and low in GDP loading conditions in CEC extracts (data not shown). However, we could not measure the Rho-GTP activity in CECs maintained in DMEM-10 that contained well developed stress fibers and focal adhesions (data not shown). It should be noted that the terminology of Rho was rather inclusively used for all members of Rho since the antibody to Rho reacts with all three forms of Rho (RhoA, RhoB, and RhoC).

The marked reduction of vinculin staining at focal adhesions caused by FGF-2 was also examined to determine whether it was due to the low expression level of vinculin. Cell lysates obtained from the subconfluent CECs treated with FGF-2 alone or with FGF-2 and one of the inhibitors (LY294002, C3 exoenzyme, or Y27632) were measured for vinculin expression (Figure 4). The steady state levels of vinculin were not altered regardless of the culture conditions. Neither FGF-2 nor the inhibitors modulated the expression levels of vinculin. This observation and the cytoplasmic staining profile of vinculin (Figure 3A) suggest that FGF-2 may facilitate translocation of vinculin from the focal adhesion complex to the cytoplasm.

Effect of FGF-2 on cell shape and actin cytoskeleton in CECs plated on Matrigel

The cytoskeletal organization of cells grown in tissue culture is often very different from that of cells in living organisms. Therefore, we compared the cellular behavior of CECs plated on tissue culture substratum with those of CECs plated on Matrigel matrix, the structural proteins of which are similar to the basement membrane of corneal endothelium. The subconfluent cells plated on Matrigel and maintained in DMEM-10 had a spread cell shape, but some cells demonstrated pseudopodia (Figure 5A), unlike the control cells that adhered to the culture dishes (Figure 1A). Cells treated with LY294002 caused the cell shape to revert completely to the spread and polygonal morphology without pseudopodia. CECs treated with FGF-2 became smaller spindle shaped cells with prominent pseudopodia. When inhibitors were present simultaneously with FGF-2, LY294002 caused the cell shape to spread morphology without pseudopodia, whereas PD98959 failed to do so (Figure 5A). The formation of pseudopodia in the growing CECs on Matrigel is likely due to some components present in Matrigel that antagonize Rho activity and subsequently activate Cdc42, although we used Matrigel containing reduced amounts of growth factors. However, this activity of Matrigel was no longer observed when cells reached confluency. Confluent CECs plated on Matrigel and maintained in DMEM-10 showed the characteristic monolayer of contact inhibited polygonal cells (Figure 5B), similar to those observed in control cultures (Figure 1B). Neither inhibitor altered the monolayers of polygonal cells. On the other hand, cells treated with FGF-2 became elongated and multi-layered (Figure 5B). Treatment with both FGF-2 and LY294002 caused a complete reversion of the FGF-2 mediated phenotypes to a contact inhibited monolayer, whereas PD98959 failed to block the action of FGF-2. These results, again, confirmed that morphological transformation induced by FGF-2 is mediated by a PI 3-kinase dependent pathway.

We also examined the actin cytoskeleton organization and focal adhesion complex in response to FGF-2 stimulation in CECs plated on Matrigel. The spindle shaped cells lost the FGF-2 mediated actin cortical ring. Furthermore, these cells have prominent pseudopodia that were positively stained with rhodamine-phalloidin (Figure 6). These cells also lost the focal adhesion complex and vinculin was localized in the cytoplasm instead of focal adhesions. Neither C3 exoenzyme nor Y27632 altered the phenotypes induced by FGF-2 and Matrigel (Figure 6). This observation is different from that of the control cells plated on tissue culture dishes in which pseudopodia formation requires both FGF-2 and inhibitors of Rho pathways (Figure 3A).

Effect of FGF-2 on cell shape and actin cytoskeleton in corneal endothelium ex vivo

Earlier studies demonstrated that F-actin was arranged into dense peripheral bands in individual CECs in vivo [28]. We therefore investigated the organization of actin cytoskeleton in organ cultured corneal endothelium on Descemet's membrane. We further determined whether FGF-2 was able to morphologically transform CECs on Descemet's membrane and to modulate actin cytoskeleton as observed in tissue culture conditions. Corneal endothelium on Descemet's membrane was peeled as one piece and maintained in FGF-2 for 4 days with one the following conditions; neutralizing antibody to FGF-2, LY294002, or Y27632, both Y27632 and LY294002, both neutralizing antibody to FGF-2 and Y27632, or all three components (Figure 7). Tissue maintained in DMEM-10 served as control. Corneal endothelium maintained in DMEM-10 demonstrated a characteristic cobblestone monolayer and cortical actin mat, whereas corneal endothelium treated with FGF-2 for 4 days showed a mixture of polygonal cells and cells with altered shape. The morphologically altered cells were observed to be multi-layered. In this mixture of cultures, the unmodulated cobblestone CECs demonstrated an actin cortical mat, whereas the modulated cells showed a disorganized actin cytoskeleton. When cells were treated simultaneously with FGF-2 and neutralizing antibody to FGF-2, the FGF-2 mediated phenotypes completely reverted to the cobblestone monolayer. LY294002 was also able to block the action of FGF-2 on cell shape and actin cytoskeleton. Of interest, the cell morphology of corneal endothelium treated with Y27632 in the presence of FGF-2 stimulation was markedly transformed to multi-layers of spindle shaped cells. In this culture actin cytoskeleton was completely disorganized and pseudopodia formation was greatly enhanced. However, this synergistic effect of Y27632 and FGF-2 was no longer observed when cells were simultaneously treated with one of the reagents that could block the action of FGF-2. These data suggested that the Rho and ROCK pathway was not activated in corneal endothelium in vivo and that FGF-2 was able to markedly modulate cell shape and actin cytoskeleton in corneal endothelium in vivo.


The regeneration of corneal endothelium after in vivo injury appears to have two distinct pathways: The regenerative pathway, by which endothelial cells do not replicate but are replaced by migration and spreading of existing endothelial cells; and the nonregenerative pathway (or fibrosis), by which endothelial cells not only resume cell proliferation but alter collagen phenotypes, which in turn leads to the production of an abnormal ECM. Production of an abnormal ECM causes corneal fibrosis in Descemet's membrane. One such clinical example is the formation of a RCFM, the physical presence of which causes loss of vision. In order to elucidate the mechanism of RCFM formation in vivo, we established an in vitro model of endothelial mesenchymal transformation in which we carefully investigated the three altered phenotypes observed in RCFM; cell proliferation, cell shape changes, and production of type I collagen. We have shown that FGF-2, a component of Descemet's membrane, exerts a key role in such endothelial to mesenchymal transformation: FGF-2 directly regulates cell cycle progression through PI 3-kinase and PLC-γ1, thus leading to a marked stimulation of cell proliferation [8-10]. Synthesis and secretion of type I collagen are greatly induced by treating cells with FGF-2, with or without another protein factor released from polymorphonuclear leukocytes [5,11]. In contrast to these well characterized phenotypes observed in our in vitro model, it is not known how FGF-2 exerts its activity to promote morphogenesis of the polygonal monolayer of CECs to multi-layers of elongated fibroblastic cells. To understand the whole spectrum of RCFM formation in vivo and endothelial mesenchymal transformation in vitro, it is essential to elucidate the morphogenetic pathways mediated by FGF-2 in CECs.

It is now clear that cytoskeletal elements with their polymerization dynamics are central to many cellular activities, including morphogenesis and wound healing [36-38]. Assembly and organization of the actin cytoskeleton is regulated by Rho small GTPases, Rho, Rac, and Cdc42 [19-22]. The conclusion that Rho, Rac, and Cdc42 regulate three separate signal transduction pathways (stress fiber, lamellipodia, and filopodia) linking plasma membrane receptor to the assembly of distinct F-actin structures has been confirmed in a wide variety of cell systems [19-26,39]. In the present study, we investigated the morphological and cytoskeletal changes of CECs mediated by FGF-2 and its signal transduction pathways. When CECs in culture are maintained in serum containing medium, they demonstrate well defined stress fibers and focal adhesions, suggesting that Rho and Rho kinase pathways are activated in CECs under conventional tissue culture conditions. The activation of Rho and ROCK pathways is likely mediated by lysophosphatidic acid, an abundant component present in serum [40,41]. Interestingly, FGF-2 completely blocked this serum mediated Rho activation, as evidenced by the loss of stress fibers and focal adhesions. FGF-2 simultaneously altered cell shape from polygonal to fibroblast-like cells. In these modulated cells, F-actin was extensively reorganized into a cortical actin mat. This action of FGF-2 is completely blocked by LY294002, the addition of which resumes organization of stress fibers and focal adhesions. These data suggest that FGF-2 may antagonize Rho pathways through the action of PI 3-kinase. It is also likely that FGF-2, via PI 3-kinase, activates the Rac pathway that subsequently downregulates Rho activity. Recent studies demonstrate that PI 3-kinase is able to activate Rac via the activation of guanine nucleotide exchange factors and that the Rho activity can be negatively regulated by Rac [42,43]. Another interesting finding is that simultaneous treatment of CECs with FGF-2 and inhibitors of Rho or ROCK leads to pseudopodia formation of the FGF-2 modulated cells, presumably through Cdc42 mediated actions, suggestive of crosstalk between Rho and Cdc42. It has been reported that cells transformed by Cdc42 grow in an adhesion independent manner, exhibit altered morphologies, and display increased motility and invasiveness, all of which reflect alterations in the actin cytoskeleton [44,45]. Thus, it is likely that Cdc42 is involved in endothelial mesenchymal transformation of CECs.

The cytoskeleton organization of cells grown in tissue culture may be different from that of cells in living organisms because artificial substrates such as plastics are likely to force cells to adjust to artificially flat and rigid surfaces. By contrast, the authentic substrate for most cells in living organisms is the ECM, which is three dimensional, complex, and dynamic in its molecular composition. Therefore, we used growth factor reduced Matrigel matrix, the structural proteins of which are similar to those in Descemet's membrane (type IV collagen and laminin) [4]. Rapidly growing and confluent CECs that adhered to Matrigel were examined under conditions identical to those used for the control cells plated on tissue culture dishes. In contrast to the subconfluent control cells, rapidly growing CECs on Matrigel demonstrate pseudopodia formation. These findings suggest that some growth factors present in Matrigel, albeit at low concentrations, may activate Cdc42 pathways. This unexpected effect of Matrigel on CECs raises a question about the potential application of this soluble basement membrane extract of Engelbreth-Holm-Swarm tumor, at least for growing and subconflucent CECs. However, this yet to be characterized effect of Matrigel is no longer observed in confluent CECs, which demonstrate a monolayer of polygonal cells, identical to that of control cells. This observation, thus, supports the application of this widely used matrix for the study of adhesion and differentiation of many normal and transformed cell types [46-48]. Our data together suggest that FGF-2 causes cell shape change and induces multi-layers of the modulated cells through PI 3-kinase, regardless of the substratum used.

Finally, we examined the actin cytoskeleton of corneal endothelium adherent to Descemet's membrane. Similar to the earlier report [28], corneal endothelium on its own basement membrane maintains the monolayer of cobblestone cells, and these cells contain a cortical F-actin ring with no visible stress fibers, suggesting that Rho and Rho kinase pathways are not activated in corneal endothelium under physiologic conditions. However, FGF-2 is able to disrupt the cortical actin mat, change cell shape, and decrease contact inhibition through the action of PI 3-kinase. These modulating activities of FGF-2 observed in this ex vivo-like organ culture further support our working hypothesis that FGF-2 is the direct mediator of endothelial mesenchymal transformation. Interestingly, both FGF-2 stimulation and blocking Rho pathways are required to further promote the mesenchymal transformation of corneal endothelium. It is likely that FGF-2 antagonizes the Rho pathways and that these pathways, in turn, release the negative regulation on Cdc42. However, in growing CECs, FGF-2 activity is sufficient to cause loss of stress fibers (inhibition of Rho) but not to activate the Cdc42 pathways (pseudopodia formation). Thus inhibitors to Rho or ROCK are required to further promote cell shape change accompanied by extensive remodeling of the actin cytoskeleton. The spindle shaped cells with prominent pseudopodia are likely the activated mesenchymal cells that actively participate in wound healing, as observed in an in vivo wound healing model [49]. What causes the additional inhibition on Rho pathways in CECs during the wound repair process is not known. However, our data suggest that differential cues (i.e., growth factors, cytokines, and their respective signal transduction) may be responsible for each stage of morphogenesis and remodeling of actin cytoskeleton. These cues generated during the wound healing process may act according to the stage of repair process, thus leading to a regenerative or nonregenerative pathway.


This study was supported by National Eye Institute Grants EY06431 and EY03040, and a grant from Research to Prevent Blindness.


1. Brown SI, Kitano S. Pathogenesis of the retrocorneal membrane. Arch Ophthalmol 1966; 75:518-25.

2. Michels RG, Kenyon KR, Maumence AE. Retrocorneal fibrous membrane. Invest Ophthalmol 1972; 11:822-31.

3. Kay ED, Cheung CC, Jester JV, Nimni ME, Smith RE. Type I collagen and fibronectin synthesis by retrocorneal fibrous membrane. Invest Ophthalmol Vis Sci 1982; 22:200-12.

4. Leung EW, Rife L, Smith RE, Kay EP. Extracellular matrix components in retrocorneal fibrous membrane in comparison to corneal endothelium and Descemet's membrane. Mol Vis 2000; 6:15-23 <>.

5. Kay EP, Gu X, Ninomiya Y, Smith RE. Corneal endothelial modulation: a factor released by leukocytes induces basic fibroblast growth factor that modulates cell shape and collagen. Invest Ophthalmol Vis Sci 1993; 34:663-72.

6. Kay EP, Gu X, Smith RE. Corneal endothelial modulation: bFGF as direct mediator and corneal endothelium modulation factor as inducer. Invest Ophthalmol Vis Sci 1994; 35:2427-35.

7. Gu X, Seong GJ, Lee YG, Kay EP. Fibroblast growth factor 2 uses distinct signaling pathways for cell proliferation and cell shape changes in corneal endothelial cells. Invest Ophthalmol Vis Sci 1996; 37:2326-34.

8. Kay EP, Park SY, Ko MK, Lee SC. Fibroblast growth factor 2 uses PLC-gamma1 for cell proliferation and PI3-kinase for alteration of cell shape and cell proliferation in corneal endothelial cells. Mol Vis 1998; 4:22 <>.

9. Lee HT, Kay EP. Regulatory role of PI 3-kinase on expression of Cdk4 and p27, nuclear localization of Cdk4, and phosphorylation of p27 in corneal endothelial cells. Invest Ophthalmol Vis Sci 2003; 44:1521-8.

10. Lee HT, Kay EP. Regulatory role of cAMP on expression of Cdk4 and p27(Kip1) by inhibiting phosphatidylinositol 3-kinase in corneal endothelial cells. Invest Ophthalmol Vis Sci 2003; 44:3816-25.

11. Kay EP, Gu X, Choi SH, Ninomiya Y. Posttranslational regulation of type I collagen in corneal endothelial cells. Invest Ophthalmol Vis Sci 1996; 37:11-9.

12. Vlodavsky I, Folkman J, Sullivan R, Fridman R, Ishai-Michaeli R, Sasse J, Klagsbrun M. Endothelial cell-derived basic fibroblast growth factor: synthesis and deposition into subendothelial extracellular matrix. Proc Natl Acad Sci U S A 1987; 84:2292-6.

13. Landshman N, Belkin M, Ben-Hanan I, Ben-Chaim O, Assia E, Savion N. Regeneration of cat corneal endothelium induced in vivo by fibroblast growth factor. Exp Eye Res 1987; 45:805-11.

14. Drechsel DN, Hyman AA, Hall A, Glotzer M. A requirement for Rho and Cdc42 during cytokinesis in Xenopus embryos. Curr Biol 1997; 7:12-23.

15. Caron E, Hall A. Identification of two distinct mechanisms of phagocytosis controlled by different Rho GTPases. Science 1998; 282:1717-21.

16. Nobes CD, Hall A. Rho GTPases control polarity, protrusion, and adhesion during cell movement. J Cell Biol 1999; 144:1235-44.

17. Allen WE, Zicha D, Ridley AJ, Jones GE. A role for Cdc42 in macrophage chemotaxis. J Cell Biol 1998; 141:1147-57.

18. Settleman J. Rho GTPases in development. Prog Mol Subcell Biol 1999; 22:201-29.

19. Van Aelst L, D'Souza-Schorey C. Rho GTPases and signaling networks. Genes Dev 1997; 11:2295-322.

20. Hall A. Rho GTPases and the actin cytoskeleton. Science 1998; 279:509-14.

21. Ridley AJ. Rho family proteins: coordinating cell responses. Trends Cell Biol 2001; 11:471-7.

22. Etienne-Manneville S, Hall A. Rho GTPases in cell biology. Nature 2002; 420:629-35.

23. Ridley AJ, Paterson HF, Johnston CL, Diekmann D, Hall A. The small GTP-binding protein rac regulates growth factor-induced membrane ruffling. Cell 1992; 70:401-10.

24. Ridley AJ, Hall A. The small GTP-binding protein rho regulates the assembly of focal adhesions and actin stress fibers in response to growth factors. Cell 1992; 70:389-99.

25. Nobes CD, Hall A. Rho, rac, and cdc42 GTPases regulate the assembly of multimolecular focal complexes associated with actin stress fibers, lamellipodia, and filopodia. Cell 1995; 81:53-62.

26. Johnson DI, Pringle JR. Molecular characterization of CDC42, a Saccharomyces cerevisiae gene involved in the development of cell polarity. J Cell Biol 1990; 111:143-52.

27. Petroll WM, Jester JV, Barry-Lane P, Cavanagh HD. Assessment of f-actin organization and apical-basal polarity during in vivo cat endothelial wound healing. Invest Ophthalmol Vis Sci 1995; 36:2492-502.

28. Petroll WM, Hsu JK, Bean J, Cavanagh HD, Jester JV. The spatial organization of apical junctional complex-associated proteins in feline and human corneal endothelium. Curr Eye Res 1999; 18:10-9.

29. Chifflet S, Hernandez JA, Grasso S, Cirillo A. Nonspecific depolarization of the plasma membrane potential induces cytoskeletal modifications of bovine corneal endothelial cells in culture. Exp Cell Res 2003; 282:1-13.

30. Kozma R, Sarner S, Ahmed S, Lim L. Rho family GTPases and neuronal growth cone remodelling: relationship between increased complexity induced by Cdc42Hs, Rac1, and acetylcholine and collapse induced by RhoA and lysophosphatidic acid. Mol Cell Biol 1997; 17:1201-11.

31. Kay EP, Smith RE, Nimni ME. Basement membrane collagen synthesis by rabbit corneal endothelial cells in culture. Evidence for an alpha chain derived from a larger biosynthetic precursor. J Biol Chem 1982; 257:7116-21.

32. Laemmli UK. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 1970; 227:680-5.

33. Hirsch DS, Pirone DM, Burbelo PD. A new family of Cdc42 effector proteins, CEPs, function in fibroblast and epithelial cell shape changes. J Biol Chem 2001; 276:875-83.

34. Maddala R, Reddy VN, Epstein DL, Rao V. Growth factor induced activation of Rho and Rac GTPases and actin cytoskeletal reorganization in human lens epithelial cells. Mol Vis 2003; 9:329-36 <>.

35. Pawlak G, Helfman DM. MEK mediates v-Src-induced disruption of the actin cytoskeleton via inactivation of the Rho-ROCK-LIM kinase pathway. J Biol Chem 2002; 277:26927-33.

36. Connolly JO, Simpson N, Hewlett L, Hall A. Rac regulates endothelial morphogenesis and capillary assembly. Mol Biol Cell 2002; 13:2474-85.

37. Yuan XB, Jin M, Xu X, Song YQ, Wu CP, Poo MM, Duan S. Signalling and crosstalk of Rho GTPases in mediating axon guidance. Nat Cell Biol 2003; 5:38-45.

38. Sachdev P, Zeng L, Wang LH. Distinct role of phosphatidylinositol 3-kinase and Rho family GTPases in Vav3-induced cell transformation, cell motility, and morphological changes. J Biol Chem 2002; 277:17638-48.

39. Doughman RL, Firestone AJ, Wojtasiak ML, Bunce MW, Anderson RA. Membrane ruffling requires coordination between type Ialpha phosphatidylinositol phosphate kinase and Rac signaling. J Biol Chem 2003; 278:23036-45.

40. Brahmbhatt AA, Klemke RL. ERK and RhoA differentially regulate pseudopodia growth and retraction during chemotaxis. J Biol Chem 2003; 278:13016-25.

41. Imamura F, Horai T, Mukai M, Shinkai K, Sawada M, Akedo H. Induction of in vitro tumor cell invasion of cellular monolayers by lysophosphatidic acid or phospholipase D. Biochem Biophys Res Commun 1993; 193:497-503.

42. Welch HC, Coadwell WJ, Stephens LR, Hawkins PT. Phosphoinositide 3-kinase-dependent activation of Rac. FEBS Lett 2003; 546:93-7.

43. Nimnual AS, Taylor LJ, Bar-Sagi D. Redox-dependent downregulation of Rho by Rac. Nat Cell Biol 2003; 5:236-41.

44. Lin R, Bagrodia S, Cerione R, Manor D. A novel Cdc42Hs mutant induces cellular transformation. Curr Biol 1997; 7:794-7.

45. Qiu RG, Abo A, McCormick F, Symons M. Cdc42 regulates anchorage-independent growth and is necessary for Ras transformation. Mol Cell Biol 1997; 17:3449-58.

46. Barcellos-Hoff MH, Aggeler J, Ram TG, Bissell MJ. Functional differentiation and alveolar morphogenesis of primary mammary cultures on reconstituted basement membrane. Development 1989; 105:223-35.

47. Pujuguet P, Radisky D, Levy D, Lacza C, Bissell MJ. Trichostatin A inhibits beta-casein expression in mammary epithelial cells. J Cell Biochem 2001; 83:660-70.

48. Petersen OW, Nielsen HL, Gudjonsson T, Villadsen R, Rank F, Niebuhr E, Bissell MJ, Ronnov-Jessen L. Epithelial to mesenchymal transition in human breast cancer can provide a nonmalignant stroma. Am J Pathol 2003; 162:391-402.

49. Petroll WM, Barry-Lane PA, Cavanagh HD, Jester JV. ZO-1 reorganization and myofibroblast transformation of corneal endothelial cells after freeze injury in the cat. Exp Eye Res 1997; 64:257-67.

Lee, Mol Vis 2003; 9:624-634 <>
©2003 Molecular Vision <>
ISSN 1090-0535