Active Scatter Factor (HGF/SF) in Proliferative Vitreoretinal Disease

Abstract

purpose. Hepatocyte growth factor/scatter factor (HGF/SF) possesses mitogenic, motogenic, and morphogenic properties and has recently been implicated in various retinal diseases. The role of HGF/SF in proliferative vitreoretinal disease was investigated.

methods. Sections of epiretinal membranes were stained immunohistochemically for cytokeratins, to identify HRPE cells, and for HGF/SF receptor (c-Met). Cultured HRPE cells were stained for c-Met and investigated for shape change in response to HGF/SF, by using image analysis. The dose–response relationship for HRPE cells to HGF/SF was investigated by a cell migration assay and the specificity of this response evaluated by a neutralization experiment. Subretinal fluid (SRF) and vitreous from patients with retinal detachment and proliferative vitreoretinopathy (PVR) plus vitreous from eyes obtained after death, eyes with macular hole, and eyes with proliferative diabetic retinopathy (PDR) were investigated for the presence of HGF/SF using an enzyme-linked immunosorbent assay (ELISA). HGF/SF activity was measured using an MDCK cell scatter assay.

results. HRPE cells in epiretinal membranes and in culture expressed c-Met. Cultured HRPE cells responded to HGF/SF by an epithelial-to-mesenchymal shape change and by cell migration, a response that increased with increasing concentrations of HGF/SF. This response was reduced in the presence of neutralizing antibody. There was evidence of HGF/SF in increasing concentrations in more severe PVR and in PDR when measured by ELISA, and, conversely, there was evidence of correspondingly decreasing HGF/SF activity when measured by MDCK cell scatter assay in these diseases.

conclusions. HGF/SF is present in normal and pathologic vitreous. HRPE cells respond by shape change and cell migration to HGF/SF. Concentrations of HGF/SF increase in proliferative vitreoretinal disease and increase in turn with increased severity of the disease, but HGF/SF bioactivity decreases (consistent with activator depletion). These findings are consistent with the hypothesis that HGF/SF may play a role in the HRPE mesenchymal transformation that typifies PVR.

In healthy eyes, retinal pigment epithelial (RPE) cells form a polarized monolayer adjacent to the photoreceptors and are involved in diverse activities that are essential to retinal homeostasis and visual function. ¹ In normal circumstances RPE cells are stationary and mitotically feeble. In contrast, they become activated and mobilized in proliferative vitreoretinopathy (PVR), which is the major complication of rhegmatogenous retinal detachment (RRD) surgery. Indeed, dissemination of migratory, proliferating RPE cells from their normal site on Bruch’s membrane to multiple loci on the detached neuroretina is thought to be a key pathologic event in the genesis of the complex epiretinal membranes associated with the development of PVR. ^{2 3 4 5 6 7 8}  

The basic pathomolecular mechanism by which the sedentary RPE cells become activated is still poorly understood, but the morphologic and functional alterations that take place have been described in detail in the experimental work of Machemer and Laqua in owl monkeys. ² Migratory and proliferative RPE cells undergo a phenotypic change so that many of them resemble fibroblasts. ^{2 6 9 10} The shape change associated with the fibroblastic alteration is so dramatic that it has been called metaplasia, ^{2 9 10} although epithelial-to-mesenchymal transition ^{6 11} may be more appropriate. 

Undoubtedly the RPE cell transition is modulated by soluble growth factors and cytokines. ^{5 12 13 14 15 16 17} Fibroblast growth factor (FGF), tumor necrosis factor (TNF)-α, interleukin (IL)-1, IL-6, interferon (IFN)-γ, transforming growth factor (TGF)-β, platelet-derived growth factor (PDGF), and insulin-like growth factor (IGF), among others, are elevated in PVR, ^{15 17} and it is known that most of these polypeptides are capable of modifying RPE behavior in numerous ways, including division, migration, matrix synthesis, enzyme production, and contraction. ^{15 18} Also, successful attempts have been made to locate, by immunohistochemistry, growth factors and cytokines within the microenvironment of the epiretinal membrane. ^{14 19} Although the list of growth factors and cytokines that may have a role in RPE cell activation and the development of PVR is long, there may be others with an important role that have not been investigated in detail so far. ²⁰  

One growth factor that may have a pivotal role in the activation of RPE cells from stationary cells to mobile and proliferating cells is hepatocyte growth factor, also known as scatter factor (HGF/SF). ^{21 22 23} Typically, it is associated with increasing the motility of various types of epithelium, ^{21 24} and the responsive target cells express the c-Met receptor. ²⁵ The term scatter factor was coined because the polypeptide promotes the dissociation or scattering of formed colonies of cultured epithelium to the extent that a bioassay for the factor was developed based on the factor’s ability to scatter cells such as Madin–Darby canine kidney epithelial (MDCK) cells. ^{21 22 23 24} It has been noted that during the scattering process, the rounded epithelioid cells invariably adopt a fibroblast or spindle shape. ²⁶ A very similar phenotypic change is undertaken by RPE in the early stages of PVR, ^{2 6} which can be considered to be an epithelial-to-mesenchymal transition. 

In addition to motogenic and shape-altering effects, the factor is also a mitogen. It has been found to be identical with HGF, which stimulates hepatocytes, and a number of other cells, to divide, ²⁷ and the two names have therefore been joined (HGF/SF), but if one name is preferred, it is usually HGF that prevails. Lacrimal gland–derived HGF/SF in tears may well modulate corneal epithelial cell proliferation, migration, and differentiation. ²⁸ Inside the eye, it may be that HGF/SF is a normal constituent of aqueous humor. ²⁹ HGF/SF is produced by corneal stromal keratocytes in wound healing, and the corneal epithelium is a target cell. ^{30 31} Corneal epithelium expresses the c-Met receptor, ³⁰ as do lens epithelial cells. ³² A complication of cataract surgery is the development of postsurgical scar-like tissue where the fibroblastic cells are thought to be derived from lens epithelium. It is therefore of relevance that HGF/SF may be an important growth factor in postcataract scar tissue development. ³³  

Recently, evidence has been published that suggests that RPE may express the c-Met receptor. ^{20 34} That being the case, HGF/SF may have a key role in the epithelial-to-mesenchymal transition of RPE cells that heralds the development of complex membranes in PVR, but obviously this can be hypothesized only if HGF/SF is present in the pathologic vitreous. The present study was undertaken to determine whether there was any HGF/SF in normal and PVR vitreous that could produce scattering of epithelium and shape changes in RPE cells. In addition we wanted to confirm whether the c-Met receptor was present on human cultured RPE (HRPE) cells and whether there were cells expressing the c-Met receptor as constituents of epiretinal membranes. 

Materials and Methods

Collection of Specimens

Core vitreous was obtained from six eyes after death (9.5–22 hours) by aspiration through the pars plana with a 23-gauge needle. Pathologic specimens of vitreous were obtained during standard pars plana vitrectomy procedures for macular hole, ⁵ RRD without PVR, ¹⁶ RRD with grade A or B PVR, ¹² RRD with grade C PVR, ³² idiopathic epiretinal membrane, ⁴ and proliferative diabetic retinopathy (PDR). ³⁶ Specimens were obtained from the aspiration tubing before infusion began to avoid sample contamination by infusion fluid. Subretinal fluid (SRF) was collected by aspiration through a sclerotomy, with a 2-ml syringe and a Brunswick (27-gauge) needle, from eyes undergoing surgery for RRD without PVR, ⁶ RRD with grade A or B PVR, ¹ or RRD with Grade C PVR. ¹ Specimens contaminated with blood from the choroidal plexus were excluded. The material obtained by each of these techniques was immediately transferred to a siliconized Eppendorf tube (Sigma, Poole, UK). Each tube was centrifuged at 10,000 rpm for 5 minutes and the supernatant frozen at −70° until analysis. 

Specimens of epiretinal membrane were obtained at vitrectomy from 13 patients with PVR (in one of whom RRD followed local resection of choroidal melanoma), 3 with macular pucker (2 idiopathic, 1 after proton beam irradiation for malignant melanoma of the choroid), and 2 with PDR. 

Cell Cultures

MDCK cells are an immortal line of transformed canine kidney epithelial cells that are particularly responsive to HGF/SF in culture. ²¹ MDCK cells were obtained from Porton Down tissue culture collection (PHLS Centre for Applied Microbiology and Research, Porton Down, Salisbury, UK) and grown in Dulbecco’s minimal essential medium (DMEM) with 5 ml glutamine, 5 ml penicillin-streptomycin and 10% fetal calf serum (FCS) in 150-ml flasks (Sterilin, Stone, UK). Flasks were placed in an incubator (LEEC, Nottingham, UK) at 37°C in 5% CO₂. MRC-5 cells are a transformed line of human fetal lung fibroblasts that produce HGF/SF in culture. ²¹ These cells were obtained from Porton Down and grown in 150-ml flasks in DMEM with 5 ml glutamine, 5 ml penicillin-streptomycin, and 10% FCS. To obtain conditioned medium containing HGF/SF, MRC-5 cells were grown to confluence and washed three times with phosphate-buffered saline (PBS) and serum-free medium (DMEM) added to the flasks. These flasks were incubated for 48 hours at 37°C in 5% CO₂. This medium was then removed, aliquoted into siliconized Eppendorf tubes (Sigma) and frozen at− 70° until required. HRPE cells were established in culture using standard techniques. ^{35 36} They were isolated from eyes with no known ophthalmic disease that were obtained up to but not exceeding 48 hours after death. The isolated cells were grown in Ham’s F10 medium supplemented with 20% FCS, 5 ml amphotericin B, and 5 ml penicillin-streptomycin and their purity confirmed by cytokeratin staining. Cells were grown in 25-cm² flasks until confluent and then passaged or frozen in dimethyl sulfoxide in liquid nitrogen. Cells between the third and eighth passages were used for all experiments. 

Enzyme-Linked Immunosorbent Assay

A noncompetitive sandwich enzyme-linked immunosorbent assay (ELISA), using a monoclonal anti-HGF antibody (A3.1.2; Genentech, San Francisco, CA), was established to provide a quantitative estimate of HGF/SF levels in appropriate vitreous samples. All wells of 96-well plates were coated with 100 μl of primary antibody (A3.1.2.) at a concentration of 5 ng/ml in coating buffer. Plates were incubated at room temperature for 16 hours and washed in a solution of 20% Tween in PBS (pH 7.6), and 200 μl blocking buffer (1% bovine serum albumin[ BSA] in PBS) was added. After further washing, specimens or standard concentrations (range, 0.5–100 ng/ml) of single-chain HGF (Genentech) in diluent solution (0.5% BSA in PBS) in a volume of 50 μl were added to each plate and incubated at room temperature for 2 hours. After another wash, 100 μl of secondary antibody (rabbit-raised polyclonal anti-HGF; Genentech) was added at a concentration of 1:5000 in a diluent solution. Plates were washed, 100 μl of a biotinylated amplification antibody (monoclonal anti-goat IgG; Sigma) was added at a concentration of 1:5000 in a diluent solution, and the plates were incubated for 1 hour. The plates were then washed, 100 μl avidin-horseradish peroxidase (Sigma) was added at a concentration of 1:6000 in a diluent solution, and the plates were incubated at room temperature for 1 hour. The Plates were washed and coated with 100 μl substrate solution o-phenylenediamine (Sigma) in 0.2 M citric acid titrated to pH 5 by potassium hydroxide. The reaction was stopped with 2 M sulfuric acid. Plates were read at 490 nm (model 312e plate reader; Biotek, Winooski, VT). For the purposes of this study, specimens were tested undiluted and at dilutions of 1:10 and 1:100 in diluent solution. 

HGF/SF Bioassay

Two techniques were used to indirectly assess HGF/SF activity in pathologic and postmortem specimens of vitreous and SRF. The first (qualitative assay) was intended to detect whether there was activity present and the second (semiquantitative assay) to provide some measure of the activity. 

The qualitative scatter assay of MDCK cell target colonies was conducted in a manner similar to that outlined in the literature. ²¹ Cells were seeded at 2500 cells per well (1 ml DMEM and 10% FCS) into 24-well plates (Sterilin) and incubated at 37°C in 5% CO₂. After 24 hours, in the presence of their normal culture medium of DMEM and 10% FCS, the cells were examined, and those wells that had typical colonies of cells in the central area of the well were photographed under the ×10 objective of a light microscope (Diaphot; Nikon, Melville, NY). Thereafter, 100μ l of vitreous, MRC-5–conditioned medium (MRC-5 CM), or plain DMEM was micropipetted into the wells. The central area of each well was reliably and repeatedly identified by virtue of using the darker center (which almost filled the field of view when using the ×4 objective lens). After a further 24-hour incubation, the central area of the well was identified and photographed again. An assessment of the colonies of cells was made and graded as scattered, nonscattered, or equivocal. This assessment was performed in a masked fashion using photographic prints. 

A semiquantitative assay was undertaken using the dilution assay first described by Stoker and Perryman. ²¹ For this assay serial doubling dilutions in DMEM of the test specimen (i.e., vitreous MRC-5 CM), which served as a positive control, and DMEM, which served as a negative control, were prepared in 96-well plates. Serial dilutions of each sample and control were made (range, 1:2–1:512). The plates were then incubated at 37°C in 5% CO₂ for 24 hours. Thereafter, the plates were fixed in ethanol and stained with hematoxylin. A subjective assessment of scattering of the MDCK cells was made for each well, on the basis of their tendency to form colonies. This assessment was not masked, because the plates were marked to indicate which solution they contained and at which dilution. The lowest concentration (i.e., the highest dilution) at which scattering was observed was used to calculate the concentration of scattering activity according to the published method of Stoker and Perryman. ²¹ The number of units of scattering activity per milliliter equaled the inverse of the dilution divided by the volume in each well (0.2 ml). For example, if scattering was present in a well of dilution 1:16, the scattering activity was 16 divided by 0.2 (the volume of the fluid in each well), giving a value of 80 units of scattering activity per milliliter. 

No attempt to control for either pH or osmolality was made in either of the scattering bioassays. In the case of the qualitative assay, a relatively small volume of test solution (100 μl of vitreous, DMEM, or MRC-5 CM) was added to the wells that contained 1 ml DMEM and 10% FCS. 

Image Analysis of Shape Change

The scattering assay of Stoker and Perryman ²¹ depends on the tight cohesive colony formation associated with MDCK cells. Unfortunately, HRPE cells did not form sufficiently tight colonies in our culture conditions to allow them to be effectively evaluated by this bioassay. However, HGF/SF also produces a characteristic change in the shape of susceptible target cells. ²⁶ An image analysis procedure was adapted to determine whether HGF/SF would produce a shape change in HRPE cells. MDCK cells were examined in a similar fashion to act as a positive control for the morphologic change. HRPE cells were seeded at 4000 per well in 96-well plates in Ham’s F10 and 10% FCS, glutamate, and antibiotics. In some wells, MRC-5 CM was present at a dilution of 1:4 in DMEM (positive control), and in others DMEM alone was used to act as a negative control for shape change. In addition, 12.5 ng/ml of recombinant HGF/SF (Genentech) was added to additional wells. MDCK cells were seeded at 3000 cells per well in DMEM with glutamate and antibiotics. MRC-5 CM was present in some wells. After a 24-hours incubation at 37°C in 5% CO₂, the plates were all washed with PBS, fixed in ethanol for 15 seconds, and stained with hematoxylin. Wells were screened, and randomly selected cells were analyzed (PrismView; Improvision, Coventry, UK) for cell shape. Between 301 and 568 cells were analyzed of each cell type. Cells were analyzed for roundness (4 × area/π length²), perfect circle (perfect, 1; thin and elongated, <1), and form factor (4 × π × area/perimeter²; perfect circle, 1; irregular margin, <1). Statistical analysis using the Waller–Duncan K Ratio t-test was performed by the software (PrismView). 

Immunohistochemistry

The epiretinal membranes were placed in acetone containing protease inhibitors (20 mM iodoacetamide and 2 mM phenylmethylsulfonyl fluoride) at −20°C overnight. After exchange in methyl benzoate, the specimens were processed into glycol methacrylate resin at +4°C (JB4; Polysciences, Warrington, PA), as previously reported. ³⁷ Sections were cut at 2 μm and mounted on 2% 3-aminopropyltriethoxysilane–coated glass slides (Sigma). HRPE cells and MDCK cells for immunohistochemistry were grown on eight-chamber tissue culture slides (LabTec, Nunc; Roskilde, Denmark) to preconfluence and fixed in precooled (−20°C) methanol (5 minutes) and acetone (2 minutes). ³⁸ Slides bearing tissue sections or cultured cells were rehydrated in PBS (pH 7.6), endogenous peroxide activity was blocked with 1% aqueous hydrogen peroxide, and nonspecific antibody binding was blocked with 10% normal swine serum in PBS. Incubation with primary antibody lasted 60 minutes (room temperature). The primary antibodies used in this study were polyclonal, raised in rabbits against portions of the HGF/SF c-Met receptor and characterized by Rong et al. ³⁹ (a kind gift from George F. van de Woude). The polyclonal antibodies were used at a 1:200 dilution in PBS. To determine whether RPE in the membranes express c-Met, sections subsequent to those processed for c-Met detection were labeled for cytokeratins with a wide-screening rabbit antiserum (diluted 1:100 in PBS) to detect RPE cells. ⁴⁰ Control preparations were processed with nonimmune rabbit serum. After washes with PBS, the preparations were incubated with biotinylated immunoglobulins F(ab)₂ fragment of sheep anti-mouse–rabbit IgG; Sigma) diluted 1:40 in PBS, washed again, and incubated with peroxidase-conjugated streptavidin (Dako) diluted 1:400 in PBS. Sites of complexes were stained brown using 3,3 diaminobenzidine tetrahydrochloride. Sections were counterstained using hematoxylin. 

Chemoattraction Assay

Chemoattraction studies on cultured HRPE cells were undertaken using 48-well microchemotaxis chambers (Neuro Probe, Gaithersburg, MD), as previously described by our group. ⁴¹ We studied the response to MRC-5 CM and to purified recombinant human HGF/SF (ICRF Cell Interactions Laboratory, Cambridge, UK), made up at the required range of activities in DMEM (18 ng/ml to 1.8 μg/ml). This fluid served as an alternative source of HGF/SF, ²¹ and soluble fibronectin (10 μg/ml; Sigma), a known attractant for RPE cells, ⁴² served as a positive control. In addition we combined MRC-5 CM with an anti-HGF/SF antibody to determine whether we could neutralize the action of HGF/SF in the conditioned medium. 

The attractants were placed in 25-μl volumes in the lower wells by micropipette. A polycarbonate membrane (Nucleopore, Pleasanton, CA), perforated with 10-μm diameter pores, and a rubber gasket covered the 48 wells, and a section containing the 48 upper wells was placed on top and secured. Preconfluent cultures of HRPE cells, still in log phase of growth, were removed from their flasks with 0.25% trypsin and 0.02% EDTA. The cells were pelleted and resuspended in DMEM so that there were 40,000 RPE cells in 50 μm of solution, and this volume was pipetted into each of the upper wells. The chamber system was incubated in 95% air and 5% CO₂ at 37°C for 5 hours, after which the polycarbonate membrane was fixed in ethanol and air dried. After 30 minutes in hematoxylin, the membranes were washed in water and mounted. Counts were made of the number of cells (nuclear counts under the ×100 objective of a light microscope) migrated onto the bottom surface of the stained membrane. Twenty high-power fields were evaluated from each well, representing approximately one fifteenth of the available area. 

Results

c-Met Receptor

HRPE cells in culture showed positive staining for c-Met using both the sp460 and C28 antibodies. Epiretinal membranes removed during vitreous surgery showed positive staining for c-Met in 14 of 18 specimens (Table 1) . In 10 of the 14 the presence of HRPE cells in the sections was confirmed by positive staining with either wide-spectrum anti-cytokeratin antibody or K18 antibody, and in 6 of these specimens, c-Met clearly was observed to be codistributed with HRPE cells (Table 1 , Fig. 1 ). There was a statistically significant correlation between the presence of HRPE cells and c-Met immunoreactivity in the membranes (Table 2 ; Fisher exact test, P < 0.05). 

HRPE Cell Shape Change

HRPE cell cultures did not exhibit good colony formation in our cultures. HRPE cells in culture were of variable appearance, but they were predominantly epithelioid (Fig. 2A ). When they were exposed to MRC-5 CM, many of the HRPE cells took on a pronounced fibroblastic form that reached its peak at approximately 24 hours after initial exposure (Figs. 2A 2B) . The content of activated HGF/SF, produced by MRC-5 conditioning, varied between 4 and 40 ng/ml, with a mean of 36.6 ± 5.3 ng/ml (SD) for our various conditioning runs. The HRPE cells’ response was very consistent between batches, however, indicating that the range of concentration was maximal to supermaximal for the change-in-form effect. Recombinant human HGF/SF, in a similar range of concentrations, also was effective at producing an epithelioid-to-fibroblastic shape change in HRPE cells, but not to the same marked extent as was seen with MRC-5 CM. 

Image analysis was conducted on HRPE cells in an attempt to quantify the shape change, with the emphasis placed on MRC-5 CM with the optimum time of 24 hours. Comparison was made with response of MDCK cells to the same stimulus in identical conditions (Table 3) . The two cell types had much the same roundness and an almost identical form factor in control medium. After 24 hours in conditioned medium, the HRPE cells had a 25% decrease in roundness and a 40% decrease in form factor. MDCK cells did appear to become more fibroblastic, but image analysis showed an overall decrease in roundness of only 10% and a form-factor change of a mere 3% (Table 3) . A feature of the shape change was that some cells respond notably, others respond less, and others hardly respond at all. With HRPE cells the responders predominated whereas with MDCK cells, the good responders were swamped by poor responders. 

HRPE Cell Migration

HRPE cells migrated very readily through the pores of polycarbonate membranes located in 48-well chemotaxis chambers to MRC-5 CM. The response was dramatic: 50 times greater than background and 2.5 times greater than to an optimum concentration of soluble fibronectin that acted as a positive control. The recombinant HGF/SF was also an effective chemoattractant for cultured RPE at the concentrations investigated, with an increasing positive migratory response between 18 ng/ml and 1.8 μg/ml. There was no decrease in response with the higher concentrations despite use of nonphysiologically high levels of HGF/SF. Within the range of concentrations we examined, the response to recombinant HGF/SF was less than that to soluble fibronectin. Neutralization of the HGF/SF in conditioned medium using anti-HGF/SF antibody caused a statistically significant reduction (45%) in the migration response (Mann–Whitney, P = 0.03). The results of the HRPE cell migration experiment are presented in Figure 2C . 

Vitreous Scattering Activity

Unlike our HRPE cells, the MDCK cells in sparse culture formed compact colonies of rarely more than 50 and often fewer than 20 cells. When exposed to a source of HGF/SF for 24 hours, the colonies broke up, and the component cells scattered. The scattered MDCK cells had clear space between each cell, so that the component cells of the original colony covered an area many times larger than that occupied by the original colony. Initial investigations of specimens of vitreous from patients with PVR showed clear evidence (data not shown) that the specimens could scatter target MDCK colonies during the 24-hour test period (Fig. 3) , justifying the value of a more rigorous semiquantitative analysis. 

The results of the semiquantitative assay showed scattering of colonies of MDCK cells by all samples of the positive control MRC-5 CM. Dilution studies indicated an average activity of HGF/SF of approximately 150 units of scattering that was equivalent to approximately 30 ng/ml of active factor. Approximately 60% of our test specimens produced a scattering effect at the lowest dilution of the assay. Nearly 70% of PDR specimens showed activity with an average scattering of approximately two thirds that of MRC-5 CM (Table 4) . Of 19 patients who underwent vitrectomy for diabetic retinopathy, 7 had type 1 diabetes (mean scattering 130 units), and 12 had type 2 diabetes (mean scattering 77 units). This difference in scattering was not significant (Mann–Whitney, P = 0.4), and there was no correlation between type of diabetes and presence of scattering (Fisher exact test, P = 1.0). Control was categorized as good in 13 patients, poor in 4, and very poor in 2, with no correlation between good diabetic control and presence of scattering (Fisher exact tests, P = 1). Details of the 19 patients whose vitreous was examined are given in Table 5 . Patients with retinal detachment had approximately half the activity of conditioned medium, whereas patients with PVR had only one third. The incidence of responders also decreased, with 80% of the patients with detachment having positive results, whereas less than 50% of the patients with PVR grade C showed a response (Table 4) . 

Vitreous ELISA Findings

SF/HGF was detectable by ELISA in all vitreous specimens including those obtained after death from donor eyes. In addition, HGF/SF was also found to be present in all SRF specimens assayed. The mean value for the five donor eyes was 2.9 ng/ml. The donors’ ages at death, gender, cause of death, and HGF/SF values for these five specimens are shown in Table 6 . The mean value for macular hole was 4.6 ng/ml. The values for donor eyes and macular hole were lower than for macular puckers (6.7 ng/ml). A progressive increase was seen in the amount of HGF/SF from retinal detachment specimens (3.3 ng/ml) to grade C PVR (13.0 ng/ml), but the most marked levels were in specimens from patients with PDR (16.8 ng/ml). Analysis of variance of the series of groups of vitreous specimens (Kruskal–Wallis, P = 0.0046) indicated that at least one of the groups had a significantly greater mean HGF/SF concentration. Further comparison of individual groups using the Mann–Whitney test demonstrated a significant difference between the PDR group and each of the following groups of specimens: donor (P = 0.0005), macular hole (P = 0.0084), RRD (P = 0.0006), and PVR C (P = 0.02). There was no difference between the PDR specimens and PVR grade A and B specimens (P = 0.11). There were no other significant differences detected. Mean values for these groups are illustrated in Figure 4 . The group of patients with diabetes comprised eight with type 1 and nine type 2. Mean ages in these two groups were 47 and 60 years, respectively (Table 7) . The proportion with good diabetic control was similar in both groups and was not associated with increased vitreous HGF/SF concentrations. There was no significant difference in values for HGF/SF concentration between the two types of diabetes (Mann–Whitney P = 0.09) or between those with good control when compared with those with poor or very poor control (Mann–Whitney P = 0.7). 

HGF/SF levels tended to be higher in SRF specimens (mean 27.3 ng/ml) when compared with specimens obtained from patients who underwent vitrectomy for retinal detachment, with or without PVR (9.7 ng/ml). No significant difference was shown between those specimens (Mann–Whitney, P = 0.37). MRC5-CM, as used in bioassay and migration experiments, was determined by ELISA and showed a presence of 36.3 ± 5.3 ng/ml (SD). 

Discussion

It has been shown in a previous study that the HGF/SF receptor c-Met is present in epiretinal membranes associated with PVR, ³⁴ but the numbers of specimens were few, and the authors were unable to determine which cell type in the membranes express the c-Met receptor. Our much larger series established that c-Met–positive cells were evident in most membranes. In addition, by using glycol methacrylate resin embedding, we had sufficiently thin sections to colocalize c-Met staining in one section, with cytokeratin (HRPE cells) staining in a subsequent section. As a result we were able to show that, much but not all, of the c-Met staining was associated with HRPE cell distribution in the membranes. Cultured HRPE cells are known to express c-Met ^{34 43} but on the basis of the present results, it is likely that HRPE cells in vivo are subject to the influence of HGF/SF if present in the immediate environment. 

HGF/SF is a mitogen on many cells including hepatocytes, ⁴⁴ but it has only a modest proliferative effect on HRPE cells at high concentrations (500 μg/ml). ^{34 43} It is a far more effective motogen. HRPE cells exhibited a powerful migratory response to MRC-5 CM and, to a lesser extent, to recombinant HGF/SF, which is in line with the results of others. ^{34 43} Among the more distinctive biologic effects of HGF/SF on target cells is scattering of epithelial cells from colonies ^{21 26} and alteration of the cells from an epithelioid to a fibroblastic morphology. ²⁶  

Colony formation was not a feature of our HRPE cells, but we were able to show, with an image analysis–based assay, that HRPE cells undergo a profound shape change when exposed to conditioned media rich in HGF/SF. We considered the loss of roundness and form factor change away from circular to be impressive, given that our wild-type cultured HRPE cell population was not synchronized, and that the original shape of the cells was therefore not very uniform. The process of sporadic cell division could have been a confounding factor. The low sensitivity of our assay was highlighted by the fact that the shape change of MDCK cells hardly registered, even though alteration was evident by eye, and MDCK cells are the gold standard for the shape-change response. ²⁶ The marked shape alteration of HRPE cells seen in the present study therefore adds weight to the proposal that HGF/SF may play a key role in the epithelial-to-mesenchymal shift of HRPE cells. The shift creates epithelial cells that look and act like fibroblasts, as is found in PVR epiretinal membranes. ^{2 3 4 5 6}  

Those tissue culture–based experiments in which we examined the scattering and shape change effects of HGF/SF both in its recombinant form and in conditioned medium were performed in the presence of FCS, because without it the MDCK and HRPE cells would not thrive. That the wells to which only the control solution was added did not show either scattering or a shape change suggests that there was no contribution from the FCS to either of these effects. 

We showed by ELISA that in normal vitreous after death there were high quantities of HGF/SF, with an average level from our samples of 2.9 ng/ml, more than 10 times higher than levels found in aqueous humor. ²⁹ Our ELISA levels of HGF/SF for specimens with macular hole and RRD are reasonably comparable to those published in the recent literature. ⁴⁵ However, in the only other PVR study Nishimura et al. ⁴⁶ found lower quantities of the growth factor in the vitreous (3.3 ng/ml) than our values of 9 ng/ml for PVR A and B and 13 ng/ml for PVR C. The overall trend for higher levels of HGF/SF in PVR than RRD vitreous and the highest level of all in the vitreous of patients with PDR is evident in our data and is apparent in the findings in the two previous studies. ^{45 46} HGF/SF is present in the blood ⁴⁷ and it is therefore not surprising that in disease associated with leaky retinal vessels, such as PDR, these would be associated with particularly high levels of the growth factor. 

It should be asked whether the growth factor concentration in vitreal samples is reasonably representative of what is going on at the tissue level. We attempted to do this by trying to determine whether HGF/SF levels were substantially different in two separate fluid compartments: the vitreous and the subretinal space. Unfortunately, we were unable to obtain vitreous and SRF from the same patients, but comparison showed a threefold greater average in the SRF than the vitreous when samples from groups containing RRD and the various grades of PVR, were examined. The difference did not reach statistical significance, which was probably related to the small sample size and the large variation in the SRF growth factor levels. That there was a significant trend from RRD to PVR grade C for greater amounts of vitreal HGF/SF cannot be taken to imply functional and pathobiologic consequences. It does not hold, for example, that more evidence of HGF/SF in the media means more growth factor–induced bioactivity in target cells such as HRPE cells. The growth factor may be below threshold levels for bioeffect, the receptor status of the target cells may be insufficient, and so on, but of particular relevance is whether the HGF/SF is in its active form. HGF/SF is secreted by cells as an inactive single-chained precursor, ⁴⁸ and only after conversion to a heterodimer by proteolytic action does it bind strongly to the c-Met receptor and become biologically active. ^{47 49 50}  

The ELISA did not distinguish between inactive and active HGF/SF, but with our semiquantitative MDCK scatter assay ²¹ we had the opportunity to examine HGF/SF activity, rather than amount, in the vitreous samples. Vitreous from PDR patients, not unexpectedly on the basis of the ELISA, had the greatest scattering activity, but what was surprising was that the trend for HGF/SF activity progressed downward from RRD to severe PVR—quite the opposite of the ELISA results! Clearly, further work is needed to confirm these findings and to distinguish between single chain and heterodimeric HGF/SF, and that research is going on at present. 

Our working hypothesis, however, is that single-chain HGF/SF is produced continuously by local cells, ⁴³ but the activator is in short supply. Thus, the proportion of heterodimeric HGF/SF, plus consequent bioactivity, goes down with the advancement of PVR, despite an overall increase in HGF/SF levels. The hypothesis fits well with the pathobiology of PVR, given that a shape change of HRPE cells and their migration (although many factors are involved in this) are considered to be early events. ³ Studies of tissues other than the eye have shown that the HGF/SF activators are proteolytic and are induced in injured tissue. ⁵¹ Urokinase-type plasminogen activator from macrophages ⁵⁰ and a factor XII homologue from serum ⁵² have been proposed as potential activators of HGF/SF elsewhere in the body. It would be useful to test this hypothesis by performing two scattering assays using one specimen of vitreous, the HGF/SF concentration of which has been measured by ELISA, before and after the addition of an exogenous activator. Unfortunately, the volume of vitreous that we routinely obtain is not adequate to perform this experiment. 

Future research to identify activators in the vitreous of PDR, PVR, and RRD may shed more light on the role of HGF/SF in retinal scarring. 

 Supported by The Guide Dogs for the Blind Association, the Sir Jules Thorn Charitable Trust, and Foundation for the Prevention of Blindness.

 Submitted for publication August 5, 1999; revised February 9, 2000; accepted March 15, 2000.

 Commercial relationships policy: N.

 

Table 1.

 

Epiretinal Membranes Stained Using Antibodies to c-Met and to Keratin

Specimen	Cause of ERM	Anti-sp460	Anti-C28	Anti-Keratin	Colocalization
1	PVR	+	+	+	+
2	PVR	+	+	+	+
3	PVR	+	+	+	+
4	PVR	+	*	+	*
5	PVR	+	*	+	*
6	PVR	+	*	+	+
7	PVR	+	+	−
8	PVR	+	*	−
9	PVR	+	−	−
10	PVR	−	−	−
11	PVR	−	−	−
12	PVR	+	+	+	+
13	PVR, resection of MM	−	−	−
14	PDR	+	+	−
15	PDR	−	−	−
16	Idiopathic	+	+	+	+
17	Idiopathic	+	*	+	*
18	Radiotherapy for MM	+	*	+	*

 In 6 of 10 cases in which both c-Met and keratin staining were present, colocalization was demonstrated. Antibodies to c-Met were sp460 and C28 and to keratin were anti-cytokeratin and K18. ERM, epiretinal membrane; MM, malignant melanoma; −, no immunoreactivity;+ , immunoreactivity.

*  Not tested.

Figure 1.

 

Light micrographs of sections from an epiretinal membrane stained with the immunoperoxidase method for (A) cytokeratins and (B) c-Met. A layer of RPE cells is seen to express c-Met (arrows). Hematoxylin counterstain; magnification, ×400.

Table 2.

 

Correlation between Presence of Keratin and c-Met in Cells in Epiretinal Membranes

	Keratin Positive	Keratin Negative	Total
c-Met positive	10	4	14
c-Met negative	0	4	4
	10	8	18

 There is a significant correlation between those epiretinal membranes staining positive for c-Met and those staining positive for keratin (Fisher’s Exact Test: Yates Corrected χ² = 3.86; P < 0.05).

Figure 2.

 

Photomicrographs of cultured HRPE cells. Cells were fixed in ethanol and stained with hematoxylin. HRPE cells cultured in the (A) absence or (B) presence of HGF/SF. HRPE cells are clearly more bipolar and elongated and have more processes in (B), whereas the fan-shaped cytoplasmic ruffles characteristic of epithelioid HRPE cells are far more in evidence in (A). Magnification, ×200. (C) Response of HRPE cells (cells migrating per high-power field, HPF) in a Boyden migration chamber to the following agents: MRC-5 CM containing HGF/SF, MRC-5 CM with anti-HGF/SF antibody at a concentration of 2 μg/ml (CM + HGF antibody), and increasing concentrations of recombinant HGF/SF. DMEM acted as negative control and fibronectin (10 μg/ml) as a positive control. Error bars: 1 SD. The reduction in response to MRC-5 CM with addition of anti-HGF/SF antibody was statistically significant (Mann–Whitney, P = 0.02).

Table 3.

 

Results of Image Analysis of Cell Shape

Parameter	Cell	Control Medium	Conditioned Medium	HGF/SF
Roundness	MDCK	0.328*	0.295*
	HRPE	0.350, †	0.264, †	0.328, †
Form factor	MDCK	0.280	0.274
	HRPE	0.277, ‡	0.159, ‡	0.246, ‡

 MDCK and HRPE cells were cultured in the presence of control media (DMEM), MRC-5 CM, or recombinant HGF/SF for 24 hours; fixed in ethanol; stained with hematoxylin; and analyzed.

*  ,

†  ,

‡   P < 0.05; Waller Duncan K Ratio t-test.

Figure 3.

 

Phase-contrast photomicrograph of MDCK cells in culture. (A) Colony of MDCK cells, before addition of vitreous specimen, cells formed a tight colony. (B) Same area of culture dish photographed 24 hours after the addition of vitreous specimen, showing separation of the cells. Magnification, ×400.

Table 4.

 

HGF/SF Concentrations Measured by the Scattering Assay in Vitreous Specimens

	Number of Patients	Scattering Detected (n)	Units of Scattering
Postmortem donor	1	0	—
Diagnosis
Macular hole	1	0	—
RRD, no PVR	10	8 (80.0%)	72 ± 23.3
RRD, PVR A or B	6	4 (66.7%)	40
RRD, PVR C, including macular pucker	19	9 (47.4%)	60 ± 11.5
PDR	19	13 (68.4%)	97.5 ± 27.8
All specimens	56	34 (60.1%)	—
MRC-5 conditioning	7	7 (100%)	148.6 ± 32.3

 The postmortem donor specimen was obtained from a 67-year-old man whose cause of death was given as ischemic heart disease. Data for units of scattering are means ± SEM. Details of the patients who underwent vitrectomy for PDR are given in Table 5 .

Table 5.

 

Details of Patients with PDR from Whom Specimens Were Obtained for the Semiquantitative Scattering Assay

Specimen	Age	Type of Diabetes	Duration of Diabetes (y)	Control	Units of Scattering
67	63	1	29	Good	80
73	64	1	30	Good	0
74	80	2	11	Good	80
78	60	2	19	Good	45
86	43	2	15	Very poor	40
87	77	2	21	Good	40
89	25	1	13	Very poor	80
91	46	1	18	Good	10
101	67	2	16	Good	40
104	63	2	11	Good	20
108	64	2	7	Good	64
118	58	1	21	Good	0
124	84	2	15	Good	80
129	49	2	4	Poor	593
132	38	1	20	Poor	320
141	57	2	25	Good	0
156	60	2	11	Poor	60
161	28	1	16	Good	160
176	63	2	5	Poor	0

 Control was categorized as good, poor, or very poor by the most recent hemaglobin A1C level before the performance of the biopsy.

Table 6.

 

Details of the Postmortem Donor Specimens Used in the HGF/SF ELISA

Specimen	Gender	Cause of Death	Age at Death (y)	HGF/SF Concentration (ng)
1	M	Head injury	20	2.0
2	M	Subarachnoid hemorrhage	23	2.6
3	F	Cerebrovascular accident	70	2.7
4	M	Ischemic heart disease	49	2.7
5	M	Ischemic heart disease	55	5.2

 Results of this assay are given in Figure 4 .

Figure 4.

 

Mean HGF/SF concentrations in vitreous for each diagnostic category, as measured by ELISA. Error bars: SEM. There was a statistically significant difference between the mean value for the PDR group when compared with that of the donor (postmortem), macular hole, RRD, and PVR C groups.

Table 7.

 

Details of the Patients with Diabetes from Whom Vitreous Specimens Were Obtained for the HGF/SF ELISA

Specimen	Age (y)	Type of Diabetes	Duration of Diabetes (y)	Control	HGF/SF Concentration (ng/ml)
156	61	2	11	Poor	6
176	63	2	5	Poor	11.3
185	31	1	10	Good	27.3
194	36	1	18	Poor	49.9
197	46	2	2	Good	24.1
206	53	2	4	Good	6.6
223	62	2	14	Poor	11
228	51	1	30	Very poor	10.1
247	66	2	26	Good	3.7
256	63	2	26	Poor	8.6
262	58	1	25	Poor	12.2
292	35	1	14	Poor	54.4
307	78	1	30	Poor	3.9
309	69	2	8	Poor	19.3
317	39	1	20	Good	16.2
326	50	1	18	Good	18.6
337	56	2	20	Poor	4.6

 See Table 5 for explanation of control classification.

The authors thank Lisa Heathcote and Penny Hogg for technical assistance.