Direct Comparison of the Migration of Three Cell Types Involved in Epiretinal Membrane Formation

Abstract

purpose. The purpose of the present study was to develop an accurate and sensitive migration assay to compare the migratory capabilities of retinal pigment epithelial (RPE) cells, retinal glial (RG) cells, and fibroblasts (the cell types crucial in epiretinal membrane [ERM] formation) under identical microenvironmental conditions and thus to identify potential target areas in ERM management.

methods. Cultured bovine RPE and RG cells and scleral fibroblasts (SFs) in both single and mixed cell type populations were induced to migrate in modified 48-well Boyden chambers. The labels used to distinguish between the cell types were latex microspheres and carmine particles. The chemoattractants used were fibronectin and PDGF, both of which are associated with epiretinal membrane development.

results. When migrating independently, all three cell types showed a positive response to fibronectin at an optimal concentration of 10 μg/mL. The RG cells migrated in a significantly greater number than the RPE cells (P < 0.05), but the differences in number of migrating cells between RG cells and SFs and RPE cells and SFs were not significant. When the cells were labeled and migrating together, it became clear that the RG cells consistently migrated in a higher number than the SFs (P ≤ 0.001) and both the SFs and RG cells showed a greater migratory response than the RPE cells (P ≤ 0.01).

conclusions. The mixed cell migration (MCM) test system is a simple and useful assay to distinguish between the migratory responses of different cell types in the same microenvironment. It has highlighted the pronounced migratory response of RG cells to standard chemoattractants and has cast some level of doubt that, in comparative migratory terms, RPE cells are particularly good responders.

Epiretinal membranes (ERMs) are scarlike, cellular sheets on the surface of the retina that complicate retinal detachment and trauma. When these scarlike ERMs complicate retinal detachment, they produce the condition known as proliferative vitreoretinopathy (PVR), ^{1 2} and their contraction causes further retinal detachment that can lead to blindness. The cellular composition of ERMs includes an inflammatory component, ³ but the main constituents are retinal glial (RG) cells, RPE cells, and scleral fibroblasts (SFs). ^{4 5 6 7 8 9 10 11 12} In that they are scarlike tissue, it is not surprising that many ERMs have a significant extracellular matrix (ECM) that is rich in collagen I and -III and has a particular abundance of fibronectin. ^{13 14 15 16} The formation of ERMs, at the interface of brain and body, is a complex biological event that combines the repair processes of neural tissue (gliosis) with those of connective tissue (fibrosis). ¹¹ At the vitreoretinal interface, the former is dominated by the Müller cell and perhaps other glial cells, ¹⁷ whereas the latter is distinguished by the presence of RPE cells and SFs. ¹⁸  

Although cell proliferation is important in PVR, other cellular events in wound healing are just as important. ¹⁹ Several biological processes, which include cell activation, migration, settlement, contraction, and synthesis of ECM, are often emphasized in general repair ²⁰ and in addition are known to be of no less significance to the formation of ERMs and the expression of their subsequent clinical effects. ^{11 18 19 21} Of these activities, migration of the component cells to the developing ERM and then within the ERM is crucial. However, much remains to be understood about the subtleties of these recruitment processes, particularly regarding migratory responsiveness and directional or nondirectional movement of the component cells within the confines and microenvironment of the scarlike tissue of the ERM. 

Accurate measurement of migration in response to chemoattractants can be made in vitro with a microchemoattraction chamber, whether in cultured cells in general ²² or, more specifically, in those ocular cells involved in the development of PVR. In the bulk of the migration studies that are relevant to the formation of ERM and PVR, microchemoattraction chambers have been able to identify which substances stimulate the migration of ERM-related cells and to determine the nature of the migratory attraction ^{23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38} (see also Table 1 ^{39 40 41 42 43} ). These studies however, reveal little about the relative migratory responsiveness of the various cell types involved in the formation of ERM in response to a given stimulant—that is, when stimulants provoke these cells to migrate, do the cells move in comparable numbers and, if not, which type is the most active? Such information would provide insight into the migratory capabilities and recruitment of the key cell types; data that may help in the understanding of the genesis, construction, and spread of ERMs; and indications of possible new therapeutic directions for the future. 

If a truly direct comparison is to be made of the migration response of different cell types, then all the competing cell types should be evaluated in near identical conditions in a quantifiable assay. The 48-well microchemotaxis chamber provides a highly quantifiable migration setup in which evaluations generally are made between the migration of cells in different sets of wells in response to different attractants or to different concentrations of the same attractant (Table 1) . Because of variation and irregularities in, for example, the plastic walls of the well, the permeable membrane through which the cells migrate, and the fluid environment on either side of the membrane, the microenvironment in each of the wells of a given chamber is likely to be subtly different. Probably the environmental differences are even more marked if the wells of different chambers are to be analyzed in a comparative study. It seemed to us that identical migratory conditions could be guaranteed only if the competing cell types were mixed in equal numbers and seeded together into each of the same wells of the microchemoattraction chamber. With the proviso that although in the in vivo situation, the numbers of cells of different cell would not necessarily be equal, the mixture of cells would be in the same microenvironment and would have identical exposure to any soluble chemoattractants present. Any eccentricity in the proportions of the migrated cell types on the bottom surface of the porous membrane was taken to be a response difference that may have recruitment implications. 

Previously, no attempts, as far as we are aware, have been made to compare the migration of different cells directly in this manner, but an obvious difficulty is to be able to make accurate distinction between the different types of cell, before or after migration through the porous membrane. We overcome the identification problem by prelabeling our test cells with inert particulate markers. The markers were phagocytosed by the appropriate cells before the migration assay and then could be identified in the cytoplasm of the labeled cells. We call the technique the mixed cell migration (MCM) assay. 

The purpose of this study was to exploit the strengths of our MCM assay and to determine whether there were significant differences between the numbers of migrating RPE, RG, and SF cells when exposed to a standard chemoattractant. We considered that such studies were relevant to the understanding of the migratory events associated with recruitment of cells to an ERM and also to the ERM’s subsequent expansion. The main chemoattractant used in our investigation was fibronectin, which is an important constituent in ERMs. ⁴⁴ Also used, to a lesser extent, was platelet-derived growth factor (PDGF) which is known to play a part in the development of ERMs. ^{32 45} Both fibronectin and PDGF have an established history of use as chemoattractants in conventional chemoattraction chamber assays with cells appropriate to ERMs (Table 1) . 

Materials and Methods

Cell Culture

The RPE, RG, and SF cells were cultured from bovine eyes that were obtained from a local abattoir, and cultures were established within 6 hours (RPE and SF) or within 24 hours (RG) of death. The anterior segment of the eye was removed together with the vitreous, and each cell type was isolated in the following manner. The RPE cells were cultured according to the method of Basu et al., ⁴⁶ as modified in our laboratory. ³⁸ The RG cells and SFs were isolated according to methods devised in our laboratory. ^{47 48} The culture medium used throughout was MEM with 10% FCS, and the cells were grown in humidified 37°C incubators with an atmosphere of 5%CO₂-95% air. 

The purity of the cultures was evaluated by using immunocytochemical staining of cytoskeletal intermediate filaments. Samples of the primary cultures were grown on eight-well slides (LabTek, Naperville, IL), fixed in precooled (−20°C) methanol and acetone, and incubated with the following antibodies: The RPE cells were stained with an antiserum for bovine muzzle epidermal keratins (Dakopatts, Glostrup, Denmark), the RG cells with an antiserum for glial fibrillary acidic protein (GFAP; a gift from Usha Chakravarthy, Ophthalmology and Vision Science, Queen’s University, Belfast, UK) and the SFs with a monoclonal antibody for vimentin (VIM 13.2; Sigma, Poole, UK). 

Experiments were restricted for all our cultured cells to between passages 2 and 5, when all cell types had stable growth curves, and there was no evidence of cellular senescence. With RG cells, we know that the proliferation rate deteriorates beyond passage 5 ⁴² and that the RPE cells lose their hexagonal, cobblestone appearance after passage 6, which was also described by Del Monte and Maumenee. ⁴⁹  

MCM Assays

The migration assay was performed in modified 48-well microchemoattraction chambers. ^{50 51 52} The chemoattractants used were soluble fibronectin from bovine plasma and PDGF from human platelets (both from Sigma). The optimal chemoattractant concentration for each cell type was determined as previously reported. ⁵³ The chemoattractants were placed in the lower wells of each chamber and covered by an upper set of wells separated by a gelatin-coated perforated polycarbonate membrane. ⁵¹ The cells to be tested for migration were prepared in suspension, as either a single cell type or as mixtures (see later), and introduced into the upper wells. Preliminary experiments had shown 40,000 cells per well with a 5-hour incubation period to be the most suitable number and assay period for all three cell types. After 5 hours, the permeable membrane was removed, fixed in ethanol, and stained with hematoxylin (Shandon Scientific Ltd., Runcorn, UK). Cells that had migrated through the pores could be visualized by light microscopy (Optiphot; Nikon, Tokyo, Japan). In initial attractant dose–response studies, each concentration was run in quadruplicate and each experiment performed at least three times. Studies were performed to determine the appropriate membrane pore size for subsequent MCM assays and pore sizes of 8, 10, and 12 μm (diameter) were evaluated. 

Flow Cytometry

We thought that if there were a substantial difference in size of any of the three cell types (RPE, RG, and SF) it might influence the rate at which the cells settle on the membrane out of suspension and such an asymmetry could prejudice the MCM assay. To check for size differences, three 25-cm² flasks of each cell type were grown to confluence in MEM with 10% FCS culture medium. The cells were removed by treatment with a solution of equal volumes of trypsin (0.25%) and disodium EDTA (0.02%), centrifuged twice at 100g for 10 minutes, and finally resuspended in phosphate buffered saline (PBS) at a concentration of 10⁵ cells per milliliter. The samples were then filtered through a 41-mm nylon filter (Spectrum Scientific Corp., Houston, TX) and analyzed in suspension in a flow cytometer (FACScan; BD Bioscience, San Jose, CA) at 488-nm illumination. Data from minimum samples of 10⁴ of each of the three cell types were collected on computer (Lysis II software; BD Bioscience) using the parameters of forward-angle light scatter (FSC) and 90°-angle light scatter (SSC). Dot plots of SSC against FSC expressed in quantitative arbitrary units (AU) from the photodiode were analyzed to demonstrate any possible relative differences in granularity (density) and cell volume, respectively, between the cell populations. 

Cell Labeling

The two cellular labels used were polystyrene (latex) microspheres 0.797 μm (0.8 μm) in diameter (Sigma) and carmine particles (Merck, Hoddesdon, UK; diameter range, 0.5–1.5 μm). Before use, weighed amounts of both labels were cleaned by centrifugation in sterile PBS (10 minutes, 75g) and then suspended in PBS containing 0.07% penicillin-streptomycin and a fungicide (0.07% amphotericin B) at 4°C. Titration experiments were performed to determine the particle concentration in the medium needed for phagocytosis and optimal incorporation of label into the cytoplasm of each cell type. Concentrations of latex microspheres and carmine particles were varied from as high as 1–5 × 10³ down to 25 to 50 particles per cell. After exposure times of either 16 or 24 hours, the cells were examined for degree and uniformity of cytoplasmic incorporation. 

Initial studies with the MCM assay were restricted to mixtures of two cell types, in which only one cell type in the mixture was labeled. The experiments were conducted in duplicate with the second experiment being identical with the first, except that the label was switched to the unlabelled cell type of the first experimental run. The switchover run provided a further check on any possible locomotion-stimulating or receptor-signaling effect of the label itself. At all times after labeling, care was taken to wash the cells clean of free particles before mixing with the unlabelled partner. The cells were mixed in equal numbers and seeded at a final total concentration of 40,000 cells per well and for these two cell type comparison studies, fibronectin was the only chemoattractant used. The final studies involved comparison of the three cell types and for these mixtures, once again, the seeding was 40,000 cells per well. The three-cell-type MCM assays involved both fibronectin and PDGF as chemoattractants. 

Estimation of Migration

Counts were made from the bottom surface (migration surface) of cells that had passed from the top to the bottom surface during the assay. Twenty fields were counted by using standard procedures from each well by light microscopy under a ×100 objective. ^{51 52} Intra- and interexperimenter errors were also calculated and found to be between 3% and 5% and 7% and 10%, respectively. Photography using bright field optics (Polyvar; Reichert-Jung, Leitz, Cambridge, UK) was undertaken so that features such as location of the labels and comparative gross morphology of the labeled and unlabelled cells on both sides of the membranes could be compared. Statistical analyses of the data were performed on computer (SPSS, ver. 9.0.0; SPSS Inc., Chicago, IL). Student’s t-test for independent samples (unpaired t-test) was used to compare two sets of normally distributed data. One-way analysis of variance (ANOVA) with a Bonferroni-type correction (Student-Newman-Keuls interval) was used for comparison of multiple data sets. 

Results

Migration through Membranes with Different Pore Diameters

A standard stimulant of 10 μg/mL fibronectin stimulated significant migration of RG, RPE, and SF cells through all three pore sizes of 8, 10, and 12 μm diameter. SFs migrated through the three pore sizes equally well (ANOVA, P ≥ 0.05). There was no significant difference between the migration of RG and RPE cells when 10- and 12-μm diameter pores were compared (P ≥ 0.05). However, there was a significantly lesser migration of these two cell types through 8-μm diameter pores compared with that through 10-μm pores. The restriction of migration was more pronounced for RPE cells (77% reduction; P < 0.0002) than for RG cells (43% reduction; P ≤ 0.0004). On the basis of these results, membranes with a pore diameter of 10 μm were used throughout the study. 

Flow Cytometry

Dot plots of SSC against FSC for RG, RPE, and SF cells in suspension showed no relative shift up or down in value for any of the three cell populations. Histograms of FSC impulse peak heights showed mean values (±SD) of 473.1 ± 69.7 AU, 494.0 ± 132.4 AU, and 428.3 ±74.9 AU for RG, RPE, and SF, respectively, with no significant differences between them (unpaired t-test, P ≥ 0.05). 

Migration of the Three Cell Types Independently

Fibronectin stimulated the migration of RG (Fig. 1A) , RPE (Fig. 1B) and SF (Fig. 1C) cells. The migration was significantly greater than background values (ANOVA, P < 0.05). The numbers of migrating cells per well per 5 hours reached a peak for all three cell types between 5 and 10 μg/mL fibronectin (Figs. 1A 1B 1C) . The maximal concentration for migration for RG and RPE cells was 10 μg/mL, and there was no significant difference between 5 and 10 μg/mL fibronectin for the SFs. Consequently, 10 μg/mL fibronectin was used as the standard chemoattractant in the subsequent migration experiments. The results of all the migration tests were pooled (Fig. 2) . Ten migration tests were performed for each cell type with at least three wells examined (maximum examined, six) in each experiment. At the optimum concentration of 10 μg/mL fibronectin, there was a 7% difference in migration between the SFs and the RPE that was not significant (unpaired t-test, P > 0.2; Table 2 ). 

Migration of the Three Cell Types after Exposure to Latex Microspheres

Unlabelled cells could be easily distinguished, and labeling had no detectable effect on cell morphology (Fig 3A) . After exposure to latex microspheres the cells incorporated the particles, but within a given culture, some cells incorporated the label more readily than others. At optimal concentrations and exposure times, effective labeling reached 99% of RG and SF cells, whereas the RPE cells were slower but effective labeling level of more than 90% was reached. Further increases in the concentration of microspheres and exposure times were ineffectual. After optimal labeling, we were able to show that the presence of microspheres had no adverse effect on the ability to migrate. There was no significant difference between the number of cells, with or without microspheres, that had migrated to a fibronectin attractant. This held true for all three cell types (Fig. 4) . 

The three cell types were mixed together in pairs, and comparisons were made for each of the three possible permutations. It became clear that although the cells were settled in the migration wells in equal numbers, the number of cells that migrated were not equal (Fig. 5) . Significantly greater numbers of RG cells (162%; P ≤ 0.01) and SFs (122%; P ≤ 0.01) were counted on the migration surface of the permeable membrane when paired against RPE cells, regardless of which cell type was labeled with microspheres (Figs. 5A 5B) . More RG cells migrated than SFs when these two were assayed together (Fig. 5C) , but the difference was only marginally significant when RG cells were labeled (P < 0.04) and not significant when SFs were the labeled cells (Table 2) . 

Migration of the Three Cell Types, Individually and in Combination, after Labeling with Carmine Particles

Incorporation of carmine into the cytoplasm of the three cell types was associated with no obvious effect on cell morphology (Figs 3B 3C) . After exposure to carmine for 24 hours at a density of 50 particles per cell, migration runs were conducted for SF, RG, and RPE cells, comparing each cell type, with and without labeling. For each cell type, significantly fewer cells containing carmine migrated compared with those without the label (P ≤ 0.01; Fig. 6A ). However, when labeling density was reduced to 25 particles of carmine per cell, incorporation in SFs was still approximately 90%. In addition, there was no significant difference in the number of migrating cells when mixed with unlabelled SFs (Fig. 6B) . Because 25 particles per cell was marginally migration stimulatory in RG and RPE cells, SFs were always labeled with carmine in the appropriate MCM experiments. 

Mixed Migration of Labeled RPE, RG, and SF Cells

Microscopy showed that the three cell types could be distinguished easily after comigration (Fig 3D) . Based on the findings in our earlier studies for the three-cell-type MCM assay, SFs were labeled with carmine, RG cells were labeled with microspheres, and RPE cells were unlabelled. In these experiments, there were significant differences in migration between the three cell types (Table 2) , with the RG cells migrating in the greatest number and the RPE cells being the least efficient (P ≤ 0.001; Fig. 7A ). In addition, in the alternative experiments in which the RG and RPE cell labeling was reversed but the SFs were once again labeled with carmine, comparable results (Fig. 7B) were recorded. In another MCM study, fibronectin was replaced by PDGF as the chemoattractant (Fig. 8) . All three cell types showed a dose-dependent response to PDGF, with dose optima that were very similar, all within the range of 25 to 50 ng/mL. At the three most responsive concentrations of 10, 25, and 50 ng/mL, RG cells migrated in the greatest number, whereas RPE cells were the most sluggish. The maximum response of RG cells was 71% greater than that of SFs, but more than 700% greater than the RPE cells (P ≤ 0.05 and P ≤ 0.001, respectively). 

Discussion

The three key cell types in formation of ERMs (RG, SF, and RPE) have been shown to be active migratory cells, and their migration is a crucial event that is central to development and maturation of ERMs. ^{11 16 54} Not only is directed and nondirectional locomotion to an attractant essential to the cell recruitment needed at the site of ERM formation, ^{10 27} but also it is a necessary part of the process of cell-mediated contraction. ¹¹ Contraction forces generated by developing ERMs result in retinal pucker that eventually leads to fixed retinal folds and complex retinal detachment. ^{12 18}  

We set out to compare the migratory response of RG, SF, and RPE cells to known chemoattractive stimulants that are present in ERMs ^{55 56} and are thought to be effective on our test cells (see Table 1 ). To standardize comparison and minimize variation we developed the MCM assay, which is a variant of the microchemoattraction assay but allows the test cells to be mixed and migrated in the same well. After migration through the permeable membrane, subsequent identification depended on labeling the cytoplasm of test cells with either microspheres or carmine particles at doses that on the one hand gave reasonably comprehensive labeling (within a 5%–10% error), but on the other hand did not affect key areas of cell behavior. At this level of phagocytosis and cytoplasmic incorporation the materials were effectively biologically inert. 

It may be argued that labeling in this manner is too complex a procedure compared, for example, with immunohistochemical identification of the relevant cell types. Although there are no particularly unique identifiers for SFs, RG cells can be distinguished by immunolabeling of the glial structural protein GFAP, ⁴⁷ and RPE are highlighted by antibodies raised against wide-ranging cytokeratins. ⁵⁷ Work in our own laboratory also has been directed toward immunohistochemical staining of migratory cells. ³⁹ It has to be said that our previous immunohistochemical studies conducted on cells adherent to the porous polycarbonate membranes ³⁹ were fraught with technical difficulties. Essentially, the major problems are that there is a variable loss of cells from the polycarbonate membrane during the more protracted immunostaining procedures, and the polycarbonate membranes tend to wrinkle, making counting of migratory cells difficul; and uniform immunostaining over a large batch of experimental wells problematic. Once particle labeling was standardized, however, the MCM assay differed little from a regular microchemoattraction assay, except that an image analysis alternative for automating the counting of migrated cells was no longer a reasonable option. 

Our MCM assay showed quite clearly that when the cells occupied the same microenvironment and were subjected to an identical stimulus (whether that stimulus was fibronectin or PDGF) the result was invariably the same. RG cells migrated in far greater numbers than SFs, whereas SFs were far more efficient than RPE cells. When cell types were run separately in response to a fibronectin stimulant, the difference between SFs and RPE cells was not apparent, although the dominance of RG cells over the other two was still obvious. As we went from pairs of cell types to experimental runs involving all three cell types in the same well, the MCM assays progressively highlighted the differences between the three (see Table 2 ). 

The relatively feeble migratory response of RPE cells to fibronectin and to PDGF was not suspected by us before this study and was not apparent in our examination of the available literature (see Table 1 ). Quite the reverse is true: The migratory vigor of RPE cells in the development of PVR has been assumed and often emphasized in previous tissue culture studies, ³² experimental animal models of PVR, ^{6 58} histopathologic studies of ERMs, ^{35 59} and reviews. ^{16 18 60} An explanation may be that RPE cells respond far better to ERM-associated chemoattractants other than the two we used in this study. Of course there is a multiplicity of known and potential chemoattraction signals in the environment of a complex retinal detachment ^{32 55 56} ; however, the two we chose are essential and relevant attractants. We chose PDGF and fibronectin to demonstrate that the difference in migration was not a stimulant-dependent phenomenon produced purely by fibronectin. It should be said that fibronectin is present in some abundance in the ERM tissue, ¹⁵ the vitreous, ^{21 36} (mean concentration, 12.96 ± 6.3 μg/mL [SD], range, 6.4–24.8) ³⁶ and subretinal fluid (mean concentration, 20.9 ± 31.4 μg/mL; range, 0.9–111.1) ⁶¹ of patients with PVR. Although PDGF is known to be produced and released locally in the region of ERMs, ³¹ it has also been detected in the vitreous of patients with PVR (mean, 55.1± 35.45 pg/mL; range, 0–105). ⁶² On balance, our findings seem to cast some level of doubt on the universally held point of view that RPE, in comparison with other cells, have a pronounced migratory response to key chemoattractant signals generated in the area of a forming ERM and may explain, in part, some findings regarding the role of RPE cells in experimental ERMs after lipopolysaccharide treatment. ⁶³  

The intrinsic weaknesses of the MCM assay are that the different test cells must thrive in the same culture medium and are required to have more or less the same dose–response optimum produced by the chosen migratory stimulant. In practice, the former is far less of a restriction than the latter. The need for clear proof that the label has no influence on migration is obvious but means that each new cell type and label needs extensive preliminary work-up. Our choice of labels was based on the known biological inertia and high phagocytic incorporation of carmine particles ^{64 65} and latex microspheres, ^{66 67} but there is no reason why other alternative particulate materials should not be used instead. Label transfer between cells has been a problem in other systems ⁶⁸ but in the short time frame of the MCM assay, such a transfer of label was not a problem for us. Single cells with double label, a classic sign of label transfer, were looked for but never found. 

We have found that an important strength of the MCM assay is to distinguish clearly between the recruitment capabilities of different cell types in a completely standardized migratory microenvironment. It provides a true baseline for comparison. In addition, there are other potential strengths of this assay system. It provides a particularly powerful way to identify a relatively weakly responding or nonresponding cell to a given migratory stimulus, which in essence is what we have done with RPE cells in the present study. We consider that the assay, with its identical conditions for each cell type, may also be a valuable means of evaluating the effects of a migratory inhibitor such as TGF-β ^{69 70} particularly, if the inhibitor is cell-type selective. Finally, the microenvironment of the MCM assay provides the conditions for examining autocrine, juxtacrine, and paracrine influences. Also, it provides a means of distinguishing between receptor- and nonreceptor-evoked migration with appropriate and selective antibody neutralizations. Several of these types of investigation are on our current program of research into the dissection of behavioral properties of key cell types involved in the formation of ERMs. 

 Supported by the Royal National Institute for the Blind, Foundation for the Prevention of Blindness, and the Guide Dogs for the Blind Association.

 Submitted for publication October 11, 2001; revised March 18, 2002; accepted April 1, 2002.

 Commercial relationships policy: N.

 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.

 

Table 1.

 

Migratory Activity of Fibronectin and PDGF on RG, RPE, and Ocular Fibroblasts

Attractant	Cell Type	Migration Type	Concentration Range	Reference
Fibronectin	Immature rat retinal glia	HX	1–20 μg/mL	De Juan et al. 31
	Rat retinal glia	No response	0.1–30 ng/mL	Harvey et al. 28
	Human RPE	CX, HX	50–100 μg/mL	Campochiaro et al. 23
	Human RPE	CX	5–20 μg/mL	Robey et al. 39
	Ocular fibroblasts	CX	5–20 μg/mL	Joseph et al. 29
PDGF	Rat retinal glia	CX	0.1–10 ng/mL	Harvey et al. 28
	Immature rat retinal glia	CX	10–30 ng/mL	De Juan et al. 31
	Human retinal glia	CK	1–10 ng/mL	Uchihori and Puro 40
	Human RPE	CX	50–100 ng/mL	Campochiaro and Glaser 24
	Human RPE+ TNF-α	CX	20 ng/mL	Jin et al. 41
	Ocular fibroblasts	CX	1–30 ng/mL	Kamiyama et al. 42
	Ocular fibroblasts	CX	1–50 ng/mL	Kim et al. 43

 HX, haptotaxis; CX, chemotaxis; CK, chemokinesis.

Figure 1.

 

Dose–response curves of the migration of (A) RG, (B) RPE, and (C) SF cells produced by a range of concentrations of soluble fibronectin (5–100 μg/mL) in serum-free culture medium. Each point represents the mean (±SEM) of cell counts in three wells. The experiments were repeated at least three times.

Figure 2.

 

Pooled results of the single-cell-type migration of RG, RPE, and SF cells in response to the optimal concentration of 10 μg/mL of fibronectin in serum-free medium. The experiment was repeated 10 times for each cell type and each bar represents the mean (±SEM) of cell counts in at least three wells.

Table 2.

 

Percentage Changes in Migration of RG, SF, and RPE Compared with Each of the Other Three Cell Types

Migration Assay Type	Chemoattractant	%Increase in Migration
				RG/SF	SF/RPE	RG/RPE
Single	sFn 10 μg/mL	62	−7	50*
Paired	sFn 10 μg/mL	58, †	122, †	162*
Triple	sFn 10 μg/mL	88, ‡	212, ‡	487, ‡
Triple	PDGF 25 ng/mL	71, ‡	404, ‡	764, ‡

 Unpaired t-test significance levels:

*   P < 0.05,

†   P < 0.01,

‡   P < 0.001.

Figure 3.

 

A light micrograph showing (A) migratory RPE cells labeled with latex microspheres (arrows) and unlabelled SFs on a polycarbonate migration membrane. (B, C) SFs stained with hematoxylin on the upper surface (B) and the lower surface (C) of a migration membrane and labeled with carmine. (D) Migrating cells on the lower surface of a chemotaxis membrane. The RG cells were unlabelled, the RPE were labeled with latex microspheres (black arrows), and the SF were labeled with carmine (white arrows).

Figure 4.

 

Migration of (A) RG, (B) RPE, and (C) SFs in response to 10 μg/mL fibronectin in serum-free medium, with and without a label of latex microspheres. The labeling concentration of microspheres was 5 × 10³ per cell for the RG and RPE cells and 10³ for the SFs for a period of 24 hours. Each bar represents the mean (±SEM) of cell counts in six wells. Each experiment was repeated at least three times.

Figure 5.

 

Paired migration of the three cell types in response to 10 μg/mL of fibronectin. The RG cells were paired with the RPE (A), the RPE with the SFs (B), and the RG with the SFs (C). The label of latex microspheres was switched to the other cell type for the control experiments. Each bar represents the mean (±SEM) of cell counts in 12 wells. Each experiment was repeated twice.

Figure 6.

 

Histograms of the migration of SFs in response to 10 μg/mL fibronectin in serum-free medium, with and without a label of carmine particles. The labeling concentrations of carmine were 50 (A) and 25 (B) particles per cell for 24 hours. Each bar represents the mean (± SEM) of cell counts in six wells. Each experiment was repeated at least three times.

Figure 7.

 

Triple migration of the three cell types in response to 10 μg/mL fibronectin. The RG cells were unlabelled. The RPE cells were labeled with latex microspheres and the SFs with carmine (A). In the control experiment (B) the RG cells were labeled with latex microspheres, the RPE cells were unlabelled, and the SFs were again labeled with carmine. Each bar represents the mean (±SEM) of cell counts in 12 wells. Each experiment was repeated three times.

Figure 8.

 

Dose–response curve of the migration of RG (labeled), RPE (unlabelled), and SF (labeled with carmine) cells in response to a range of concentrations of PDGF (0.1–100 ng/mL). Each column represents the mean (±SEM) of cell counts in six wells. Each experiment was repeated twice.