Epithelial–Mesenchymal Transition in Proliferative Vitreoretinopathy: Intermediate Filament Protein Expression in Retinal Pigment Epithelial Cells

Abstract

purpose. To improve our understanding of how retinal pigment epithelial (RPE) cells behave in vivo and to establish similarities with dedifferentiation and adaptive events observed in RPE cells cultured under simulated intraocular pathologic conditions. At the same time, to examine the origin of epithelioid-shaped and fibroblast/fusiform-shaped cells in epiretinal membranes (ERM) from proliferative vitreoretinopathy (PVR).

methods. Cells of ERM were studied by electron-immunocytochemical techniques, using simple, double, and triple immunostaining for cytokeratins (CK), vimentin (Vim), and glial fibrillary acidic protein (GFAP). Ultrastructural morphology analysis was also carried out. Adult human RPE cells were obtained and cultured with normal and pathologic vitreous from proliferative vitreoretinal disorders, subretinal fluid aspirates from retinal detachment, and normal human serum. Their cytoskeleton was fractionated at 7 (early cultures) and 24 (late cultures) days of culture, electrophoresed, immunoblotted for intermediate filament proteins, and quantified by densitometric analysis for each condition. Changes in phenotype characteristics were also evaluated.

results. Epithelioid-shaped and fibroblast/fusiform-shaped cells, resembling RPE cells, expressed CK–Vim–GFAP simultaneously as intermediate filament proteins in their cytoskeleton. RPE cells in culture also expressed CK–Vim–GFAP and changed from an epithelial shape to a migratory fibroblast/fusiform-shaped phenotype in the presence of subretinal fluid aspirates and pathologic vitreous from proliferative intraocular disorders. In simulated cultures of proliferative intraocular disorders, cells decreased or retained their CK7, CK8, and CK18, retained Vim, and increased CK19 and GFAP, while their mesenchymal morphology became clearer over time.

conclusions. Studies of intermediate filament proteins in vivo suggest that dedifferentiation occurs in RPE cells in ERM. Dedifferentiated RPE cells may be responsible for epithelioid-like and fibroblast/fusiform-like cells. Furthermore, changes in intermediate filament protein levels were observed in RPE cells in simulated cultures of proliferative intraocular disorders. These changes were linked to cells acquiring a mesenchymal migratory phenotype. Results indicate that the dedifferentiation of RPE cells occurs both in vivo and in vitro and that it can be explained as an epithelial-mesenchymal transition.

Epiretinal membranes (ERM) are the result of cellular proliferation and connective tissue formation on the surface of the retina. This proliferation can occur in several ocular disorders, ^{1 2 3} but is typically the main histopathologic attribute of proliferative vitreoretinopathy (PVR), which sometimes complicates the natural history or surgical treatment of rhegmatogenous retinal detachment. During the evolution of PVR, the contraction of ERM causes a marked distortion of the retina, resulting in a complex tractional retinal detachment, which is difficult to repair (for review see ^{Refs. 4 5 6} ). 

Although the pathogenesis of ERM formation is not fully understood, attempts have been made to determine the origin of cells in the membranes by using ultrastructural ^{1 2 3 7 8 9 10} and light-microscopy immunocytochemical ^{11 12 13} criteria. These studies have shown several types of cell (including retinal pigment epithelial [RPE] cells, glial cells, fibroblasts, and cells with myofibroblast transformation) to be involved in contractile cell phenomena observed in the PVR process. In vitreous and standard cultures, most of these cell types undergo phenotypic changes, and thus no longer resemble the normal cell populations from which they originate. ^{14 15 16} There is also evidence that several peptides and serum proteins stimulate cell migration, adhesion and proliferation, and tissue contraction. ^{17 18 19 20}  

Several immunocytochemical studies of preretinal membranes have been reported, but little ultrastructural evaluation was carried out. They provided useful criteria for cell identification, although they were only partially successful, because of discrepancies between the immunocytochemical results and ultrastructural studies in many ERM series. ^{1 9 10} Second, some cell types that proliferate in ERM show few distinguishing characteristics; they are usually described as fibroblast-like cells ^{1 2 3 8 9} or as an indeterminate type, ^{1 2 3 7 8 9 10} suggesting that their characteristics were lost or modified during proliferation. Third, dedifferentiation and adaptive events are commonly observed in cell cultures and several pathologies. ^{14 15 16 21 22 23}  

During embryogenesis, organogenesis, tumor invasion, and metastasis, and in some reparative processes, epithelial cells separate from the epithelium and develop a mesenchymal phenotype with migratory properties, accompanied by dramatic changes in the program of cell differentiation (for review see ^{Ref. 23)} . Thus, the differentiation state of some cells may undergo changes, indicating that substantial phenotypic plasticity is retained in specialized adult cells. ²⁴ These changes are shown not only by the loss of specific gene expression and the apico-basolateral polarity typical of epithelial cells but also by abrupt modifications in their cytoskeletal organization such as protein expression of intermediate filaments and functional changes in cell–to–cell adhesion and cell–to–extracellular matrix interactions. ^{25 26 27 28} The process of conversion between epithelial and mesenchymal cell differentiation programs has led to the concept of“ epithelial–mesenchymal transition.” ²⁹  

Previous histopathologic observations ^{1 2 3 7 8 9 10 14 15 16 30} have suggested that environmental changes (such as culture conditions or retinal detachment and PVR) could induce dedifferentiation of RPE cells and give rise to mesenchymal-like cells. In the present study, the phenotypic and behavioral transformations in adult human RPE cells in culture were compared with the cell morphology and intermediate filament protein expression in epithelioid- and fibroblast/fusiform-like cells observed in ERM from PVR. Results indicate that dedifferentiation and adaptive migratory processes occur in human RPE cells in vitro and that most epithelioid- and fibroblast/fusiform-shaped cells, frequently observed in ERM from PVR, are dedifferentiated RPE cells that had undergone epithelial–mesenchymal transition. 

Materials and Methods

Reagents and Antibodies

Eagle’s minimum essential medium (EMEM) and fetal calf serum were obtained from Bio–Whittaker (Boehringer Ingelheim, Ingelheim, Germany). Bovine serum albumin (BSA; fraction V, essentially fatty acid-free), glycine, Triton X-100, Tween-20, D,L-lysine, glutamine, penicillin, streptomycin, sucrose, and ammonium chloride were supplied by Sigma–Aldrich (St. Louis, MO). Paraformaldehyde and glutaraldehyde fixatives, lead citrate, and uranyl acetate came from Merck (Darmstadt, Germany). Protein A conjugated to 15 or 10 nm colloidal gold (pA–Au 15 nm; pA–Au 10 nm) was purchased from Hans Slot (University of Utrecht, The Netherlands). All other reagents were of the highest purity available. 

Several primary rabbit polyclonal and mouse monoclonal antibodies (mAbs) for intermediate filament protein detection were used: glial fibrillary acidic protein (GFAP), cytokeratin (CK), and vimentin (Vim). For immunoelectron labeling, we used a rabbit anti-bovine GFAP polyclonal antibody (Dakopatts, Glostrup, Denmark) at 1:200 dilution, a mAbs IgG anti-Vim (clone V9; Boehringer Mannheim, Mannheim, Germany), mAbs IgG anti-CK8.13 (clone K8.13, ICN Biomedical, Costa Mesa, CA), and mAbs IgG anti-CK19 (Amersham, Buckinghamshire, UK), all at 1:30 dilution. The polyclonal anti-CK18 (42 kDa) was kindly provided by Oriol Bachs and Ricardo Bastos (University of Barcelona, Spain) and used at 1:25 dilution. For immunoblot analysis the dilutions used were as follows: GFAP at 1:2000; Vim, CK8.13, and CK19 at 1:1000; and CK18 at 1:500. Secondary antibodies conjugated with peroxidase against rabbit immunoglobulins or mouse immunoglobulins were obtained from Dakopatts. Anti-mouse immunoglobulins (IgG plus IgM) conjugated with 5, 10, or 15 nm colloidal gold were from British BioCell Research Laboratory (Cardiff, UK). 

Tissue Sample Preparation

Normal human eyes (n = 6), donated for corneal transplant in accordance with the Standardized Rules for Development and Applications of Organ Transplants, as defined in Spanish law, were obtained from the Eye Bank of the Barraquer Ophthalmological Center (Barcelona, Spain). Donors were between 42 and 85 years of age (mean age, 60 years). Human ERM (n = 29) were dissected and removed by appropriate intraocular vitreous forceps (Grieshaber, Switzerland) from patients with retinal detachment complicated by PVR who were undergoing intraocular surgery. All subjects were fully informed of the purpose of the intraocular surgery and research and gave their written consent. Specimens (n = 20) were immediately fixed in 0.1% glutaraldehyde and 2% paraformaldehyde in 0.1 M phosphate-buffered saline (PBS) solution (pH 7.4) for at least 12 hours at 4°C. They were then rinsed in PBS; when there was sufficient tissue (n = 11), fixed membranes were divided in two under a dissecting microscope. One half was placed in 2.5% glutaraldehyde and 2% paraformaldehyde in PBS and stored at 4°C for conventional ultrastructural microscopy. The remaining specimens were immersed in 2.1 M sucrose in PBS for 30 minutes, mounted on a metal stub, rapidly frozen in liquid nitrogen, and stored at −196°C before cryoultramicrotomy. 

For conventional electron microscopy, specimens (n = 11) were rinsed abundantly in PBS, post-fixed in 1% osmium tetroxide in 0.1 M phosphate-buffered (PB) solution (pH 7.4) for 1 hour, dehydrated in increasing graded concentrations of acetone, and then embedded progressively in resin (Spurr technique) for polymerization at 60°C. Ultrathin sections (50–75 nm) were obtained by conventional ultramicrotomy (OmU2; Reichert-Jung, Wein, Austria), placed on copper grids (200 mesh), and then stained with uranyl acetate and lead citrate solution for conventional transmission electron microscopy (Hitachi 800 MTi; Hitachi, Tokyo, Japan). 

The remaining ERM (n = 9) were fixed in 0.1% glutaraldehyde and 4% paraformaldehyde in PB solution for at least 12 hours at 4°C. They were rinsed in PBS and treated with 0.15 M ammonium chloride in PBS solution. ERM were then progressively dehydrated and embedded in Lowicryl K4M (Chemiche Werke Lowi, Waldkraiburg, Germany) for polymerization at −35°C. 

Normal vitreous samples (n = 5) were obtained from the normal eyes donated for corneal transplant. Pathologic vitreous (n = 12), from PVR and proliferative diabetic retinopathy (PDR) patients, was obtained during surgery under visual control by aspirating liquefied vitreous from the center of the vitreous cavity with a tuberculin syringe before opening the vitrectomy infusion. Subretinal fluid aspirates (n = 5) were obtained by external drainage. All samples (300–1000 μl) were centrifuged (13,000g for 5 minutes at room temperature), divided into aliquots, and then stored at −20°C. Sera from normal subjects (n = 10), previously tested for HIV, were obtained from the clinical laboratory (Barraquer Ophthalmological Center). 

Isolation and Culture of Human RPE Cells

RPE cells were isolated from normal human eyes at autopsy according to the customary method ³¹ with some modifications. Cells were transferred to a 25-cm² tissue culture flask (Corning, Corning, NY) and cultured at 37°C in a humidified atmosphere of 5% CO₂ and 95% air. Medium (EMEM with 15% fetal calf serum supplemented by 2 mM glutamine, 100 U/ml penicillin, and 100 mg/ml streptomycin) was changed after 24 hours of isolation and every 48 hours thereafter and passed by trypsinization when cells reached confluence. RPE cells were used in all parts of the studies between the third and fifth passages. 

Culture Treatment and Fractionation of Human RPE Cells

RPE cells were plated on 60-mm-diameter tissue culture dishes (Nunc, Napperville, IL) at a density of 4 × 10⁴ cell/cm² in complete fresh medium at 37°C. After 12 hours the medium was replaced by EMEM with 5% fetal calf serum supplemented by glutamine and antibiotics, and the specific treatment was added (vol/vol): 10% PVR vitreous; 10% PDR vitreous; 10% subretinal fluid aspirates; 10% normal human vitreous; 2% normal human serum. Thereafter, the medium with its specific conditions was changed every 4 days. Cells were cultured for 7 and 24 days in each condition before cell fractionation. RPE cells cultured in 5% fetal calf serum in complete medium were used as control cell culture. 

Cultured RPE cells were fractionated in cytoskeleton buffer containing Triton X-100 (10 mM Tris–HCl, 0.14 M NaCl, 1.5 M KCl, 5 mM EDTA, 0.5% Triton X-100 at pH 7.6), following the standard method. ³² Cytoskeleton intermediate filament proteins from cultured neonatal rat hepatocytes ²⁸ and a crude extract from murine brain cortex were also used as control. 

Immunoelectroncytochemical Procedures

For immunocytochemical procedures, consecutive serial ultrathin sections (70–90 nm) at −105°C were obtained by cryoultramicrotomy (Ultracut FC4D; Reichert–Jung). Ultrathin frozen sections were placed on gold grids (200 mesh), formvar-coated for TEM, and then maintained in PBS at 4°C before the immunoelectroncytochemical studies. We performed immunoelectroncytochemical staining, following standard procedure. ³³ Negative control sections were performed by omission of the primary antibodies. Results were observed in conventional TEM (Hitachi 600 AB; Hitachi). 

For double and triple immunoelectron labeling we studied specimens embedded in hydrophilic resin Lowicryl K4M. Ultrathin sections (60–85 nm) were obtained by conventional ultramicrotomy (Reichert–Jung), placed on gold grids (200 mesh), and formvar-coated for conventional TEM. Grids were hydrated in 0.1 M glycine–0.1 M PBS (PBSG) solution (2 × 5 minutes) and then blocked in 2% ovalbumin in PBSG solution for 30 minutes at room temperature. For double immunoelectron labeling, grids were incubated with a mixture of polyclonal and monoclonal antibodies, diluted in 1% ovalbumin in PBSG solution for 2 hours at room temperature in a humidified chamber. After washes (3 × 5 minutes) in PBSG solution the sections were incubated with a mixture of pA–Au 15 nm (for polyclonals) and anti-mouse IgG-Au 10 nm or 5 nm (for mAbs) for 1 hour at room temperature in a humidified chamber. After several washes, first in PBS (3 × 5 minutes) and then in double distilled water (10 to 15 × 5 minutes), grids were contrasted with 2% uranyl acetate solution and lead citrate. Control experiments were performed in parallel by omission of primary antibodies. The triple-staining procedure was carried out first with double labeling and then with grids fixed in 1% glutaraldehyde in PB solution for 5 minutes, washed abundantly in PBS, treated with 0.15 M ammonium chloride in PBS (2 × 5 minutes), washed again, and blocked in 2% ovalbumin in PBSG solution for 30 minutes at room temperature. The second step was then performed for the corresponding antibody and its colloidal gold labeling. Negative control sections were prepared for each step by omission of the respective antibodies, to observe anti-mouse IgG–Au specificity and pA–Au affinity for human tissues. Electron micrographs were obtained on a Hitachi 600 AB. 

Electrophoresis and Western Blot Analysis

One-dimensional sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) was performed as previously described. ³⁴ Cytoskeleton proteins (5 μg) were mixed in a 1:1 (vol/vol) electrophoresis sample buffer (1% SDS, 10% 2-mercaptoethanol, 10% [wt/vol] glycerol, 0.001% bromophenol blue, 0.125 M Tris–HCl, pH 6.8), kept at 100°C for 5 minutes, and electrophoresed (Mini-Protean II 200/2.0 Electrophoresis Apparatus, Bio–Rad, Richmond, CA) for 1 to 2 hours at 75 to 100 V on a 10% polyacrylamide gel in SDS. Western blot analysis of proteins on nitrocellulose was performed using antibodies as described previously. ³⁵ SDS-PAGE proteins were transferred (Trans-Blot 200/2.0 Transfer Apparatus, Bio–Rad) at 20 V for 2 hours at 4°C onto a nitrocellulose sheet (HyBond-c, Amersham). Membranes were blocked for 1 hour in 3% nonfat dry milk in 10 mM PBS and then incubated overnight at 4°C with polyclonal or monoclonal antibodies in 0.05% Tween-20 in blockage solution. Secondary anti-rabbit or anti-mouse peroxidase–conjugated immunoglobulins (Dakopatts) diluted 1:2000 were then applied for 3 or 4 hours at room temperature. The blot was developed with ECL system (Amersham). 

Densitometric Quantification

A direct densitometric reading was performed with digitized images of autoradiograph sheets. Scanning densitometry was carried out with a Hewlett–Packard scanner (ScanJet 4c, Hewlett–Packard), and the signals were standardized and quantified by the Molecular Analyst program (version 1.4.1; Bio–Rad). 

Results

Simultaneous CK–Vim–GFAP Expression in Cells of ERM

Ultrathin sections obtained by cryoultramicrotomy provided acceptable resolution for comparison with samples prepared by conventional ultrastructural microscopy. RPE cells were identified by their polarity with numerous cytoplasmic and microvillous processes as well as with intracytoplasm–bounded melanin granules. Cells were occasionally arranged in a rosette-like configuration forming a false lumen, with microvilli facing it (Figs. 1 A and 1E). These cells were frequently, but not always, CK-positive (Figs. 1B 1C 1D) . Cells with a similar arrangement and characteristics expressed Vim to variable degrees (Figs. 1F 1G 1H 1I) . 

Double and triple immunogold labeling were carried out on ultrathin sections obtained in Lowicryl K4M inclusion. Epithelioid-shaped cells, with or without intracytoplasmic melanin granules, with abundant intermediate filament bundles and some polarization, were simultaneously immunoreactive for both CK–Vim (Fig. 2 A) and CK–GFAP (Fig. 2B) . These cells were frequently observed in cluster arrangement. 

Cells with glial-like ultrastructural features were often elongated or spindle-shaped and tended to polarize with microvillous processes. Bundles of microfilaments and intermediate filaments associated with well-developed organelles, such as mitochondria and rough endoplasmic reticulum, were usually noted. In these cells, GFAP and Vim were usually expressed simultaneously (Fig. 3 B) but not CK. However, simultaneous immunoreaction of GFAP–Vim was also observed in fibroblast/fusiform-shaped cells, with or without intracytoplasm–bounded melanin, which suggested cells similar to RPE ones. These cells were not polarized and microvillous-like processes were not observed (Fig. 3A) . These cell types were often surrounded by abundant collagen fibers. 

Triple immunogold labeling revealed that epithelioid-shaped cells with fusiform morphology with some cytoplasmic processes and abundant intermediate filaments were simultaneously positive for CK–Vim–GFAP (Fig. 4) . These cells often formed clusters. In the same way, triple simultaneous immunoreactivity was observed in fibroblast/fusiform-shaped cells, which preserved some cytoplasmic processes (Fig. 5) . These cells also formed clusters and had abundant intermediate filament bundles. Strict control experiments were carried out for simple, double, and triple immunolabeling by omitting primary antibodies and incubating immunogold particles of different diameters (Fig. 5A , inset). 

Subretinal Fluid Aspirates and Vitreous from PVR Change RPE Cell Phenotype

Our previous studies (unpublished) of cellular proliferation and growth showed that primary cultures of adult human RPE cells change their phenotype when deprived of fetal calf serum. Cultured cells retained their epithelial morphology and “cobblestone” arrangement for 3 to 4 weeks in the presence of 5% fetal calf serum. Cells changed their morphology and arrangement when cultured in concentrations of less than 5% fetal calf serum. Therefore, we grew RPE cells in medium (medium/antibiotics/l-glutamine), supplemented by 5% fetal calf serum as control culture, in early (7 days) and late (24 days) cultures, to compare the effects of each simulated condition (Fig. 6) . 

To evaluate changes in morphology, control cultures of RPE cells were grown with 10% (vol:vol) subretinal fluid aspirates, from retinal detachment patients, or 10% (vol:vol) vitreous from PVR patients. Significant changes in phenotype and cell arrangement were noted in both simulated conditions. Cells with fibroblast/fusiform and epithelioid shapes were observed in early cultures, and their number increased with time. Changes in cell shape were more pronounced when subretinal fluid was present than in PVR vitreous cultures. No changes in phenotype were noted in control cultured cells at 7 or 24 days (Fig. 6) . 

Changes in Intermediate Filament Protein Expression in Cultured RPE Cells

For immunoblot experiments we used a mAb anti-CK8.13 (clone K8.13), which recognized several subtypes of acid and basic cytokeratins, ³⁶ an antiserum directed against CK18 and mAb anti-CK19, which recognized cytokeratins 18 and 19, respectively. Adult human RPE cell cultures grown in medium (medium/antibiotics/l-glutamine) supplemented by 5% fetal calf serum were the control for intermediate filament protein expression experiments. In early cultures (7 days) RPE cells contained CK18 (42 kDa) and a very small expression of CK19 (40 kDa). CK7 (54 kDa) and CK8 (52 kDa) were identified with mAb anti-CK8.13. Vim (54 kDa) and GFAP (50 kDa) were also detected as cytoskeleton intermediate filament protein (Fig. 7 , lane 1). Thus, early control cultures expressed three types of intermediate filament protein: cytokeratins (CK7, CK8, CK18, CK19), Vim, and GFAP. In late cultures (24 days), a slight increase in CK18 was detected. However, levels of CK7, CK8, CK19, Vim, and GFAP were higher than in early cultures (Fig. 7 , lane 7). In the control culture, intermediate filament protein levels increased with time (Fig. 8) , which was considered a normal pattern of response of RPE cells grown in culture. 

Simulated condition was achieved when control cultures were supplemented (vol:vol) by 10% normal or pathologic vitreous (PVR and PDR), 2% normal human serum, or 10% subretinal fluid aspirates from retinal detachment patients. In early cultures (Fig. 7 , lanes 1 to 6, and Fig. 8 ), simulated conditions induced an increase in all intermediate protein levels, mainly observed with normal human serum (Fig. 7 , lane 2) and subretinal fluid aspirate (Fig. 7 , lane 3) cultures. In contrast, not all simulated conditions induced variations in the expression of intermediate filament proteins in late cultures (Fig. 7 , lanes 7 to 12, and Fig. 8 ). Increased levels of CK19 were detected in all conditions (Fig. 8) , but CK18, CK7, and CK8 levels decreased in the presence of normal human serum–supplemented (Fig. 7 , lane 8) and subretinal fluid aspirate–supplemented (Fig. 7 , lane 9) cultures. In contrast, levels of CK18, CK7, and CK8 were retained or increased in other simulated conditions, such as normal (Fig. 7 , lane 12, and Fig. 8 ) or pathologic vitreous (PVR and PDR; Fig. 7 , lanes 10 and 11, and Fig. 8 ). Interestingly, normal vitreous showed a marked decrease in CK19 levels (Fig. 7 , lane 12, and Fig. 8 ), whereas cultures supplemented with normal human serum showed a strong decrease in CK7 and CK8 (Fig. 7 , lane 8, and Fig. 8 ). Thus, it seems that normal vitreous tends to retain epithelial characteristics in the intermediate filament protein expression of RPE cells, but normal serum tends to induce changes in cytokeratin expression. 

The behavior of Vim and GFAP modifications was similar to that observed for cytokeratins in early cultures (Fig. 7 , lanes 1 through 6, and Fig. 8 ). All simulated conditions increased levels of these proteins, although some conditions, such as normal human serum (Fig. 7 , lane 2) and subretinal fluid aspirates (Fig. 7 , lane 3), produced higher levels than others. However, few modifications were found in late cultures (Fig. 7 , lanes 7 to 12, and Fig. 8 ). In simulated culture conditions, levels of Vim and GFAP were similar to the control. Vim and GFAP levels decreased in cultures supplemented with normal vitreous (Fig. 7 , lane 12, and Fig. 8 ). 

Discussion

Here we attempt to elucidate the behavior of RPE cells in vivo and to identify similarities with dedifferentiation and adaptive events observed in such cells cultured under simulated intraocular pathologic conditions. At the same time, we examine the origin of epithelioid-shaped and fibroblast/fusiform-shaped cells in ERM ^{1 2 3 4 5 6 7 8 9 10 11 12 13 14} from PVR. Cultures with normal human serum could indicate the breakdown of intraocular physiological barriers that allows the action of several circulating components in the intraocular milieu. Subretinal fluid from retinal detachment was used as a first step for the development of more complicated states of retinal detachment with PVR. RPE cells were also observed in the presence of normal and pathologic vitreous from chronic intraocular proliferative disorders such as PVR and PDR. In addition, we studied the differences in the cytoskeleton intermediate filament proteins under these conditions. Our results show that adult human RPE cells in vitro and in PVR undergo cellular dedifferentiation and changes in phenotype, suggesting an adaptive phenomenon. 

Coexpression of CK–Vim–GFAP in Epithelioid-like and Fibroblast/Fusiform-like Cells in ERM

Several studies ^{1 2 3 7 8 9 10} have attempted to determine the ultrastructural features of cells in ERM. Our ultrastructural data are consistent with those previously reported. However, because dedifferentiation usually occurs in cell migration and proliferation, ^{15 23 24 25 26 27 28 29} the identification of cell origin based only on morphologic criteria is inaccurate, and other approaches are required. We used intermediate filament proteins as specific internal markers in simple, double, and triple labeling in immunoelectrocytochemistry, and we also compared cell-type identification with ultrastructural morphologic features. Our observations showed the following: Cells with epithelioid-shaped features, with or without pigment, resembling RPE cells in their morphology, arrangement, or both, can express CK–Vim–GFAP simultaneously; cells with fibroblast/fusiform-shaped features, with or without pigment (which can resemble RPE cells in their morphology, arrangement, or both), can express CK–Vim–GFAP simultaneously; and cells with glial-like ultrastructural features had simultaneous labeling for GFAP–Vim but not for CK. 

Preembedding immunoelectroncytochemistry experiments demonstrated that there are cells in ERM with variable morphologic characteristics that did not express any marker (GFAP, Vim, or CK) as internal protein. ³⁷ The present study showed simultaneous expression of different intermediate filament proteins in“ undifferentiated” cells that had similar ultrastructural features and arrangement. Colloidal-gold postembedding immunoelectroncytochemistry, with triple labeling techniques in hydrophilic resins, was also an appropriate procedure for studying cell behavior in ERM. 

Vinores et al. ³⁸ observed coexpression of keratin–GFAP in some ERM cells, which suggested that an adaptive process with different intermediate filament protein coexpression could be a generalized event in ERM formation. Our data support this observation. Furthermore, we conclude that the simultaneous expression of CK–Vim–GFAP occurred mainly in dedifferentiated RPE cells. ERMs often have cells of glial origin, ^{1 2 3 10 11 12 13} which in culture may adopt an even greater variety of phenotypes. ^{16 19} We only observed cells with glial-like ultrastructural features coexpressing GFAP–Vim, indicating that both proteins could be expressed by glial cells in vivo. Müller cells can express GFAP under normal conditions and synthesize Vim in response to retinal injury or degeneration. ^{39 40 41 42} Similar transition of intermediate filament proteins from Vim to Vim/GFAP has been reported in the development of retinal astrocytes and radial glia in rats, ^{41 43 44 45} considered as a specific marker for glial differentiation. 

Morphology and Intermediate Filament Protein Expression of RPE Cells In Vitro

Previous studies ^{1 2 3 4 5 6 7 8 9 10 11 12 13 14 16 17 18 20 46 47} showed the role of RPE cells in the development of vitreoretinal disorders and traction. RPE cells in humans and other mammals frequently lose many of their epithelial characteristics (melanin synthesis, polarized cobblestone shape and junctional complexes, keratin expression) when subcultured for longer periods. ^{30 48 49 50 51} We studied adult human primary cultures that were subcultured a limited number of times. Abrupt changes to epithelioid- and fibroblast/fusiform-shaped cells were observed when 10% subretinal fluid (acute condition) or vitreous of PVR (chronic condition) was added to the culture medium. Interestingly, morphologic changes in vitro were more pronounced when subretinal fluid was added. This represents the initial breakdown of blood–retinal barriers, which allows circulating components, such as macromolecules and growth factors among others, entry to the intraocular milieu. ^{4 5 6 17 18 19 20 33} Furthermore, human RPE cells cultured in the presence of vitreous underwent morphologic changes, whereas the amount of CK and Vim decreased over time. ¹⁶ ERM can be understood as a late stage in the clinical evolution of PVR, in which cellular anchorage and interaction with this matrix determine stability in cellular events. ^{4 5 6}  

Human RPE cells in culture express Vim and several keratins, such as CK7, CK8, CK18, and CK19. CK18 and CK19 seem to be expressed only by migratory and proliferating cells. ^{30 49 52} Vim was not detected in RPE cells in situ in normal eyes, but RPE cells may acquire this protein in response to intraocular diseases, and it is abundant in cells in culture. ⁵³ Interestingly, subretinal fluid and serum induced keratins, Vim, and GFAP in the early days of culture, whereas vitreous from PVR did not alter the amounts of Vim and GFAP over time. In PVR conditions keratins tended to decrease, whereas CK19 was constant or increased. Altered keratin subtypes in premalignant or malignant epithelial lesions showed that increased CK8, CK18, and CK19 correlated with infiltrating characteristics and invasive ability. ^{54 55} These keratins were also associated with migration in cultured human RPE cells, but Vim did not interfere with their mobility. ^{52 53} In PVR, detached RPE cells migrate into the vitreous after undergoing changes to fibroblast/fusiform-like cells, which correspond to a migrating phenotype. Cells produce a collagenous matrix that leads to a fibrous membranogenic process with contractile properties. ^{4 5 6 20 46 47} Our data suggest a series of dynamic functional changes in cultured cells, which add up to a morphologic transformation: In PVR, after retinal detachment, RPE cells lose contact with their anatomic basement membrane; RPE cells, as in pathologic conditions simulated in vitro, temporarily lose intercellular junctions responsible for cell–to–cell contact and their cohesiveness and can even change phenotype. In parallel, RPE cells change their program of intermediate filament protein expression, increasing the Vim and GFAP induced first by subretinal fluid and later by PVR vitreous. At the same time, keratins (like CK7, CK8, and CK18) tend to decrease, whereas CK19 increases or remains constant, which may help the cytoskeleton acquire migratory characteristics. We conclude that the behavior of cultured adult human RPE cells under these particular simulated pathologic conditions is similar to that observed in the intraocular milieu during the clinical evolution of PVR. These results taken together may reveal the cell changes observed during the conversion of epithelium to mesenchyme. ^{23 24 25 26 27}  

Intermediate filament protein expression in cells involved in PVR is not being fully understood. In PVR (characterized by poorly understood multifactorial events) the interaction between cells and substrate is an important phenomenon before, during, and after ERM formation; cell metabolism and protein expression always respond to specific intraocular condition and modifications, thus regulating their functional characteristics. ^{14 15 16 39 44} Our results indicate that some cell types, characterized on the basis of morphologic criteria as epithelioid-like and fibroblast/fusiform-like cells, can simultaneously express CK–Vim–GFAP as intermediate filament proteins and be dedifferentiated RPE cells. Internal protein studies suggest a particular dynamic behavior in the intermediate filament proteins of cultured RPE cells, which indicates dedifferentiation and adaptive changes in cell phenotype toward a mesenchymal migratory morphology. Our results also support the treatment of RPE cells in simulated cultures as a model for pharmacological research and environmental response. Epithelial–mesenchymal transition may also characterize the complex functional and metabolic changes in growing cells during PVR. 

 Reprint requests: R. P. Casaroli—Marano, Department of Cell Biology, University of Barcelona, Avda. Diagonal 645, Barcelona (08028), Spain.

 Supported in part by grants from FISS/90-0391 (SV) and a grant (RPCM-89) from the Barraquer Institute (RPCM) (Barcelona, Spain).

 Submitted for publication October 5, 1998; revised March 12, 1999; accepted April 12, 1999.

 Proprietary interest category: N.

 

Figure 1.

 

Ultrathin sections of ERM from PVR. Ultrastructure of RPE cells and morphologic characteristics of CK8.13–positive cells. (A) Conventional ultrathin sections; (B, C, and D) Cryoultrathin sections. (A) RPE cells with epithelioid-like characteristics showed some polarization, with cytoplasmic processes (microvillous processes) in their apical domain (asterisk). Sometimes, a few intracytoplasm–bounded melanin granules (arrow) are present. They were arranged in clusters, in which cells were joined by junctional complexes (arrowheads). (B) Sections of the same sample were incubated with mAbs anti-CK8.13 and cells with similar morphologic and nuclear-shaped characteristics labeled for 10 nm immunogold (arrows). (C) Low-power magnification micrograph showed a pigment-laden (arrowhead) cell with epithelioid morphology immunoreactive for CK8.13 (arrows). (D) High-power magnification of bracketed area observed in (C) shows 10 nm immunogold labeling (arrows) for CK8.13 and intracytoplasm–bounded melanin granule (arrowhead). Scale bar, (A) 2.5 μm; (B and C) 1μ m; and (D) 0.5 μm. Ultrathin sections of ERM from PVR. Ultrastructure of RPE cells and morphologic characteristics of Vim–positive cells. (E) Conventional ultrathin section; (F, G, H, and I) Cryoultrathin sections. (E) RPE cells were recognized by their cuboidal shape and polarity; these cells were sometimes seen in a rosette-like arrangement with microvilli facing a false lumen (asterisks). (F) Polarized epithelioid-like cells contained melanin granules and apical cytoplasmic processes (asterisk), with a generalized labeling for Vim (arrow). (G) Epithelioid-shaped cells in a similar rosette-like arrangement showing false lumen structures (asterisk) were positive for Vim. (H) High-power magnification of bracketed area in (G) showed generalized immunoreactivity for Vim. (I) High-power magnification of bracketed area seen in (F) showed 10 nm immunogold labeling (arrow) for Vim. Scale bar, (E) 5 μm; (F and G) 2 μm; (H) 1 μm; and (I) 0.5 μm.

Figure 2.

 

Ultrathin sections of ERM from PVR. Double immunogold labeling on sections from Lowicryl K4M. (A) Epithelioid-shaped cells with some polarization and few cytoplasmic processes (arrows) with well-defined intracytoplasm–bounded melanin granules (asterisk) and abundant intermediate filament bundles were immunoreactive for CK18 and Vim simultaneously. Bracketed area seen in (A) was magnified in the micrograph shown in the inset, which has a strong labeling for 15 nm immunogold (CK18) and 5 nm immunogold (Vim) scattered in intermediate filament bundles (arrowheads). (B) Epithelioid-shaped cells with some polarization (compare with Fig. 1G ) with similar intracytoplasmic characteristics observed in (A) was simultaneously immunoreactive for CK8.13 and GFAP. (Inset) Micrograph magnifies the marked area of (B). Intermediate filament bundles have 15 nm immunogold and 5 nm immunogold labeling (arrowheads) for CK8.13 and GFAP, respectively. Scale bar, (A, B) 1 μm; (insets) 0.5μ m.

Figure 3.

 

Ultrathin sections of ERM from PVR. Double immunogold labeling on sections from Lowicryl K4M. (A) Fibroblast-shaped cell, in a fusiform morphology and surrounded by abundant collagenic matrix (col), without polarization or cytoplasmic processes and with intracytoplasm–bounded melanin granules (asterisk), was immunoreactive for GFAP and Vim simultaneously. Bracketed area in (A) is magnified in the upper inset micrograph and the marked area within it is magnified in the bottom inset. Labeling (arrowheads) with 15 nm immunogold for GFAP and 5 nm immunogold for Vim was observed. (B) Cells with glial-like features exhibiting some cytoplasmic processes and abundant mitochondria (m) often had simultaneous immunoreaction for GFAP and Vim. They were surrounded by moderate collagen matrix (col). The marked area in (B) is magnified in inset micrograph. Labeling for GFAP (15 nm, arrowheads) and for Vim (5 nm, arrow) is observed. n, nucleus. Scale bars, (A, B and upper inset in A) 1 μm; (bottom insets in A and B) 0.5 μm.

Figure 4.

 

Ultrathin sections of ERM from PVR. Triple immunogold labeling on sections from Lowicryl K4M. (A) Epithelioid-shaped cells with fusiform morphology and some cytoplasmic processes (arrows) with abundant intermediate filaments were simultaneously immunoreactive for CK8.13, GFAP, and Vim. Some cells in this cluster had well-defined intracytoplasm–bounded pigment granules (asterisk; compare with Fig. 1A ). (B) Bracketed area in (A) shows abundant intermediate filament bundles with strong immunolabeling for gold particles. (C) The magnified area marked in (B) shows simultaneous triple immunogold labeling for GFAP (15 nm, arrowhead), CK8.13 (5 nm, arrow), and Vim (10 nm, large arrow). n, nucleus. Scale bars, (A) 1 μm, (B) 0.5 μm, and (C) 0.25 μm.

Figure 5.

 

Ultrathin sections of ERM from PVR. Triple immunogold labeling on sections from Lowicryl K4M. (A) Fibroblast/fusiform-shaped cells that conserved very few cytoplasmic processes (arrows) and abundant intermediate filaments showed simultaneous labeling for CK8.13, GFAP, and Vim. Inset micrograph represents control experiment omitting primary antibodies and incubating immunogold particles of different diameters in two steps, as described in the Materials and Methods section. (B) Bracketed area in (A) is magnified, and regions with abundant intermediate filament bundles were strongly labeled for GFAP (arrowhead), CK8.13 (arrows), and Vim (large arrow). (C) The magnified area marked in (B) showed triple immunogold labeling for GFAP (15 nm, arrowheads), CK8.13 (5 nm, arrows), and Vim (10 nm, large arrow). n, nucleus. Scale bars, (A) 1 μm, (inset in A, B, and C) 0.5 μm.

Figure 6.

 

Phenotype of adult human RPE cells in early (7 days) and late (24 days) primary cultures observed in phase–contrast microscopy. Control cells were cultured with 5% fetal calf serum (FCS 5%; control cultures). Cellular morphology and arrangement were similar during 24 days of observation. Control cultures with 10% subretinal fluid aspirates (SRF 10%) from patients with retinal detachment showed dramatic changes in cell morphology. In both early and late cultures, epithelioid (arrows) and fibroblast/fusiform (arrowheads) phenotypes were seen. Similar morphologic changes were observed when control cultures were supplemented by 10% vitreous from PVR (PVR 10%) patients. However, epithelioid (arrows) and fibroblast/fusiform (arrowheads) phenotypes were less pronounced. Scale bar, 30 μm.

Figure 7.

 

Western blot analysis of intermediate filament protein expression of adult human RPE cells in simulated culture. (Top) Cells were cultured with control medium (C, lanes 1 and 7), 2% normal human serum (HS, lanes 2 and 8), 10% normal human vitreous (V, lanes 6 and 12), or 10% pathologic vitreous (PVR, lanes 4 and 10; PDR, lanes 5 and 11) and 10% subretinal fluid aspirates (SRF, lanes 3 and 9) from retinal detachment, as described in the Materials and Methods section. They were fractionated, and samples of homogenates of early (7 days, lanes 1 through 6) and late (24 days, lanes 7 through 12) cultures were processed for SDS–PAGE and immunoblotted using antibodies against different cytokeratins (CK18, CK19, and CK8.13: CK7 and CK8), Vim, and GFAP. Intermediate filament proteins extracted from cultured neonatal rat hepatocytes (h, lane 13) and a crude extract from mouse brain cortex (b, lane 14) were used as controls. A representative experiment of three is presented. (Bottom) Bands from autoradiographic sheets were analyzed by densitometry and quantified. Absolute values were standardized and then normalized according to the values achieved from the control conditions (value = 0) for early (7 days, lane 1) and late (24 days, lane 7) cultures. Positive values represent relative gain, whereas negative values represent relative loss in protein levels. Increased levels of intermediate protein expression were observed in early simulated culture conditions (lanes 1 through 6).

Figure 8.

 

Comparative analysis of intermediate filament protein levels of adult human RPE cells in simulated cultures. Early (7 days) and late (24 days) cultures were subjected to different simulated conditions, as described in Figure 7 . Bands from autoradiographic sheets (Fig. 7 , top) were analyzed by densitometry and quantified. Absolute values were then standardized and compared, and relative percentages were obtained and diagrammed. Maximum absolute densitometric value was considered as 100% for each culture. The percentage variability of intermediate filament protein levels in early and late cultures was compared for different cytokeratins (CK18, CK19, and CK8.13: CK7 and CK8), Vim, and GFAP. In simulated culture conditions, levels of CK7, CK8, CK19, Vim, and GFAP were higher than control in early cultures. With time, levels of CK7, CK8, and CK18 tend to decrease and CK19 and GFAP tend to increase, whereas cells retain their Vim levels. C, control culture; V, normal vitreous; HS, normal human serum; SRF, subretinal fluid aspirates; PVR, vitreous from PVR; PDR, vitreous from PDR.

We thank David García for help with cell cultures and Montserrat Bigas with results analysis; the “Servicios Científico-Técnicos de la Universidad de Barcelona,” especially Ana Rivera, for technical assistance in preparing electron microscopy specimens; Robin Rycroft for expert assistance in correcting this manuscript; and Alfredo Muiños, Rafael Barraquer, Francisco Mateus–Márquez, Carlos D. Heredia and Daniel Vilaplana (“Centro de Oftalmología Barraquer,” Barcelona, Spain), Borja Corcóstegui (“Instituto de Microcirugía Ocular,” Barcelona, Spain), Prof. Jean Haut, Claire Monin, and Yannick LeMer (“Center Hospitalier National d’Ophthalmologie des Quinze–Vingts,” Paris, France) who all kindly provided specimens for these studies.