Human retinal pigment epithelial cells from different donors continuously produce a vascular endothelial cell-stimulating factor into serum-free medium

Subretinal neovascularization (SRN) can progress to haemorrhage and disciform scarring, resulting in severe visual loss especially in age-related macular degeneration. Though the pathogenesis of SRN remains obscure, it has been postulated that retinal pigment epithelium (RPE) plays an important role in its occurrence. Clinical investigations (Hogan, 1972; Deutman and Grizzard, 1978; Sarks, 1980; Gartner and Henkind, 1982; Young, 1987) and experiments on animals (Henkind and Gartner, 1983; Korte et al., 1984; Mancini et al.,1986) have suggested that RPE can modulate choriocapillaris through a diffusible “vascular modulating factor”. Several research teams have found that cultured RPE cells from human and animal origins synthesize and release one or more angiogenic substance which has a mitogenic effect on the vascular EC (Orlidge and D’Amore, 1987; Schweigerer et al., 1987; Wong et al., 1988; Morse et al., 1989; Smets et al., 1990), although Glaser and colleagues have claimed that they have inhibitory activity (Glaser et al., 1985, 1987). Further analysis of these mitogenic substances is severely limited by the fact that only small quantities of conditioned medium (CM) can be prepared from individual HRPE cell lines in primary cultures and early passages. Our study, based on human cell cultures, has compared the ability of 14 HRPE cell lines, in early cultures and late passages, to produce and release vascular EC growth substance(s) into CM. Preliminary characterization and immunological examination of the active factor were carried out. The results suggest that the EC mitogen in HRPE-CM is distinct from basic fibroblast growth factor (bFGF), which has been found within the bovine RPE cells (Schweigerer et al., 1987).

Medium 199 (M199), fetal calf serum (FCS), antibiotics, trypsin-EDTA (0.05%-0.02%) solution and buffers were purchased from Gibco. Bovine fibroblast growth factor-acidic (aFGF), bovine FGF-basic (bFGF), polyclonal anti-aFGF, polyclonal anti-bFGF, collagenase (Clostridium histolyticum, type IA), trypsin and soybean trypsin inhibitor were obtained from Sigma. Tissue culture flasks and 48-well plates were from Costar, [³H]thymidine from Amersham and low-protein-binding filters (0.22 μm) from Millipore.

3-[(3-Cholamidopropyl)dimethylammonio]-1-propane sulfonate (CHAPS), Nonidet P-40 (NP-40) and Tween 20 were obtained from Sigma. Trichloroacetic acid (TCA) and sodium dodecyl sulfate (SDS) were from Merck. 3,3′,5,5′-tetramethyl-benzidine (TMB) was from Aldrich.

Human eyes from donors without known ocular disease were collected from the University Hospital of Antwerp. Fourteen HRPE cell lines were used for the experiments over a period of two years. HRPE cells were isolated and cultured in our laboratory as previously described (Smets et al., 1990; Vanden Berghe et al., 1992). For primary culture, the cells were seeded into 25 cm² tissue culture flasks (T-25 flask) at a density of 2 × 10⁶ cells per flask and cultivated in 8 ml M199 supplemented with 15% FCS and 100 i.u./ml penicillin (growth medium). The confluent cultures were dissociated with trypsin-EDTA solution and split at a ratio of 1:3 in the growth medium for subculturing.

HUVEC were harvested as described previously (Vanden Berghe et al., 1992) and grown in M199 containing 30% heatinactivated human adult serum (HAS) and 100 i.u./ml penicillin. The cells were identified as endothelium by positive staining for Factor VIII and EN-4 antigens using indirect immunofluorescence with monoclonal antibody specific to human Factor VIII or EN-4 (Jaffe et al ., 1973; Berneman et al ., 1989). Cultures between the first and fourth passage were used in the proliferation and mitogenic assays.

All cell cultures were incubated in a humidified atmosphere of 95% air/5% CO₂ at 37°C.

For the production of conditioned media, HRPE cells were subcultured into 75 cm² tissue culture flasks (T-75 flask) at a density of 10⁶ cells in 20 ml growth medium. After six days cultures reached confluency with a density of 6 × 10⁶ to 8 × 10⁶ cells per flask. The cells were then washed extensively with maintenance medium (MM) (Vanden Berghe et al., 1986) and changed over to serum-free M199 (10 ml/flask) with 100 i.u./ml penicillin. After 48-hour incubation at 37°C in 5% CO₂, CM was collected and the flasks were replenished either with the same amount of fresh M199 or with trypsin-EDTA solution for subculturing. The primary culture and first six subcultures were only conditioned once. From the seventh to the twelfth passage, multiple collections were obtained at 48-hour intervals up to a month. CM was filtered through a low-protein-binding filter membrane immediately after collection and stored at −20°C for later use.

HRPE-CM was concentrated 20-fold by ultrafiltration in a 400 ml Amicon cell equiped with a Diaflo YM 5 membrane (M_r 5,000 cut-off). The filtrate (<5,000) remained unconcentrated. The concentrate (>5,000) was fractionated by further ultrafiltration in an 8 ml Amicon cell equipped with a Diaflo YM 30 or YM 100 (M_r 30,000 or 100,000 cut-off) membrane consecutively. All fractionated fractions were adjusted to the starting volume in order to compare the biological activities.

HRPE cells (14 × 10⁷) were conditioned 5 times consecutively, resulting in 1 l HRPE-CM. The cells were then dissociated, washed and pelleted by centrifugation at 2,000 g for 5 min. The cell pellet was resuspended in 10 ml of buffer (10 mM Tris-HCl, 1 M NaCl, pH 7.5). The cell suspension was frozen in liquid N₂, thawed at 15°C and subsequently ultrasonicated for 20 s; this treatment was repeated twice. The broken-cell suspension was centrifuged at 10,000 g for 30 min and the supernatant was further centrifuged at 40,000 g for 50 min. The final supernatant, a postribosome supernatant (PRS), containing soluble proteins, was diluted 1:100 in M199, to reach a total volume of 1 l. Both HRPE-CM and diluted PRS were concentrated 100-fold by Amicon ultrafiltration through a Diaflo YM 5 membrane. The concentrates were collected and filtered immediately through low-protein-binding filter membranes. Samples were stored at −20°C. The procedures were performed at 4°C unless stated otherwise.

HUVEC were seeded into 48-well plates at a density of 10⁴ cells per well (1 cm²/well) in M199 supplemented with 20% HAS and 100 i.u./ml penicillin. Six hours later, cells were washed with serum-free medium and M199, with or without diluted samples containing 2% HAS, was added. Plates were incubated in a humidified atmosphere of 95% air/5% CO₂ at 37°C for 90 hours. At this time 1 μCi of [³H]thymidine was added to each well and the plates were reincubated for 6 hours at 37°C. The cells were washed with warm MM and fixed in 0.5 ml of ice-cold 10% TCA at 4°C for 30 min. Triplicate microscopic (×100) cell counts were obtained from each well. After 10% TCA was removed, 0.5 ml of 0.1% (w/v) SDS in 0.2 M NaOH was added to each well. The plates were then incubated at 60°C for 2 hours. The solubilized DNA was measured by counting the incorporated radioactivity. The latter was measured by addition of the sample to 10 ml scintillation cocktail using a Packard liquid scintillation counter.

An enzyme-linked immunosorbance assay (ELISA) (Voller et al., 1976) for aFGF and bFGF was performed. Dilutions of 100-fold HRPE-CM and HRPE-PRS were coated onto 96-well microtiter plates. Bovine aFGF and bFGF were used as standards. The plates were incubated overnight at 4°C. After washing, a saturated solution of casein diluted 1:10 in PBS was used to prevent non-specific binding. Polyclonal rabbit anti-aFGF and anti-bFGF antibodies were diluted 1:700 and added, followed by incubation at room temperature (RT) for 1 hour. The amount of bound antibody was detected by adding 1:3,000 diluted anti-antibody (biotinylated goat anti-rabbit IgG) for 45 min at 37°C, followed by addition of a 1:1250 dilution of horseradish peroxidase (HRP)/avidin at 37°C for 30 min. Wells were stained with the HRP substrate TMB plus H₂O₂ in a phosphate/citrate buffer. The reaction was stopped after incubation at RT for 25 min by adding a M H₂SO₄ solution. The rate of reaction was measured at 450 nm in a microplate reader spectrophotometer.

Physicochemical stability of the mitogenic activity in HRPE-CM was tested with various treatments. (1) Heat sensitivity: HRPE-CM was incubated at 56°C for 2 hours or 90°C for 1 hour. (2) Sensitivity to pH: 3 × 100 ml of 50 mM Tris-HCl, 100 mM NaCl buffer solutions were adjusted to pH of 2.5, 7.4 and 10.0 with 1 M HCl and 1 M NaOH, respectively. Three × 1 ml HRPE-CM were dialysed (M_r 3,500 cut-off) against the solutions with different pH values at RT for 3 hours and then neutralized by further dialysis against 200 ml M199, pH 7.4, for 3 hours with one medium change after 1 hour. (3) Trypsin sensitivity: HRPE-CM was incubated with 100 mg/l of bovine pancreatic trypsin at 37°C and the trypsin activity was stopped at 0 min or 360 min by the addition of 400 mg/l soybean trypsin inhibitor for 15 min and then stored in ice. (4) Stability to detergents: 0.001%, 0.01%, 0.1% and 0.2% of CHAPS, Tween 20 and NP-40 were, respectively, mixed with M199, with or without 50% HRPE-CM. The stability of the mitogenic activity in HRPE-CM after the various treatments was determined as described earlier. The untreated HRPE-CM served as a positive control (100% activity) while treated M199 served as a negative control. All samples were adjusted with M199 to a final concentration of 50% HRPE-CM before being used in the bioassays.

All experiments were performed in duplicate or triplicate. The results were calculated as means ± s.d. Each individual experiment was repeated three times. The proliferation and mitogenic effects of HRPE-CM on HUVEC are presented as cell numbers (cells/mm²) or [³H]thymidine incorporation into DNA (cts/min).

The isolated cells were defined as a pure RPE population according to their intact morphology and growth characteristics. In primary culture, HRPE cells were plated out at a density that approximated confluency and formed a polygonal cell monolayer with highly pigmented cells, without obvious cell division, similar to the RPE layer in vivo in man (Fig. 1A). In long-term subculture, HRPE cells lost pigmentation with continued cell replication, but did retain their homogeneous epithelioid morphology and contact inhibition character (Fig. 1B).

The mitogenic response of HUVEC to HRPE-CM was evaluated after adding different concentrations of HRPE-CM to HUVEC cultures in M199 with 2% HAS after 4 days. Starting with 100-150 cells/mm², a just confluent monolayer was usually obtained after 4 days in M199 with 30% HAS and in M199 containing an optimal concentration of HRPE-CM with 2% HAS. In control (M199 + 2% HAS), most cells were still alive after 4 days but no multiplication had taken place. Stimulation by HRPE-CM was thus confirmed by an increase of the cell number and/or an enhancement of the DNA synthesis of HUVEC as compared to the control. Fig. 2 demonstrates the dose-dependent stimulation in the concentration range of 5% to 70%(v/v) HRPE-CM, whereas higher concentrations of CM (70% to 90%) exhibited a plateau or a slight decrease in activity. Since maximal stimulation activity was usually achieved in the presence of 50% to 70% HRPE-CM, 50% HRPE-CM was chosen as the appropriate concentration to detect and compare the mitogenic effect of HRPE-CM on HUVEC growth in our experiments.

To determine whether the EC growth stimulation effect of HRPE-CM was related to individual donor factors or to environmental alterations during HRPE cell cultivation in vitro, we examined the conditioned media collected individually from the primary cultures and late passages of HPRE cell lines. Fig. 3 shows that the media containing 50% HRPE-CM, in all 14 cases (Table 1), had significant stimulatory effects on the growth of HUVEC. Furthermore, the ability of HRPE cells to produce and release this mitogen was not influenced by increasing their in vitro age (Fig. 3 and Fig. 4). We found that a confluent HRPE cell monolayer could be conditioned in serum-free M199 at 48-hour intervals up to a month without notable deterioration. When 50% HRPE-CM from each of the 15 collections was tested on HUVEC, the results showed a marked and stable stimulation profile (Fig. 5).

The fractionated HRPE-CM was tested to estimate the approximate size of the molecule(s) responsible for the EC-mitogenic effect. The EC growth-promoting activity in HRPE-CM appeared to be due to a soluble molecule with a relative molecular mass higher than M_r 30,000, and most likely between 30,000 and 100,000, since the major activity was retained in these two fractions (Fig. 6).

We compared the activity of 100-fold HRPE-CM with HRPE-PRS. The results show that HRPE-PRS was able to stimulate the HUVEC growth dose-dependently, as shown in Fig. 7A.

Polyclonal rabbit antibodies against bovine aFGF and bFGF were determined for cross-reactivity with 100-fold HRPE-CM and HRPE-PRS using ELISA. Fig. 7B,C demonstrates that only HRPE-PRS was recognized by anti-bFGF IgG, whereas 100-fold HRPE-CM gave a negative signal to both anti-aFGF and anti-bFGF IgGs, when RPE-PRS and RPE-CM were tested at a range of concentrations that induced an identical EC-mitogenic activity (Fig. 7A).

Preliminary characterization of HRPE-derived EC growth factor(s) was performed by using various treatments on the HRPE-CM and then testing its activity on the growth of HUVEC. The data show (Table 2) that this factor was resistant to extremes of pH (2.5 to 10.0) but sensitive to heat and trypsin. Furthermore, the factor was not dialysable (M_r 3,500 cut-off) and unable to pass through membranes of M_r 5,000 and 30,000 cut-off (Fig. 6). In addition, the effect of three kinds of detergents on HUVEC was examined. NP-40 had the highest cytotoxic effect even when diluted 1:1000 in 50% HRPE-CM while 0.01% Tween 20 was not cytotoxic but still inhibited the bioactivity. The mitogenic effect of HRPE-CM could be fully retained when mixed with 0.01% CHAPS, a known non-cytotoxic detergent and useful for stabilizing growth factors (Matuo et al., 1988).

In the present study we investigated the mitogenic activity of RPE-CM on vascular EC using human material in particular, since it should more closely reflect the physiological or pathological conditions in man.

Four research teams, up to now, have demonstrated the EC-mitogenic activities of RPE-CM from human origin. However, their findings were controversial, reported as inhibition (Glaser et al. 1985, 1987), stimulation (Wong et al., 1987, 1988; Smets et al., 1990) or no effect (Bryan and Campochiaro, 1986; Campochiaro et al., 1989) on the proliferation of vascular EC. As animal RPE cells showed only stimulatory activities (Orlidge and D’Amore, 1987; Schweigerer et al ., 1987; Morse et al., 1989), donor factors could be responsible for the contradictory results. We, therefore, examined 14 RPE cell lines isolated from apparently normal human eyes. The CM collected from these RPE cell lines and different passages was tested individually on HUVEC. Our data indicate that the ability of HRPE cells to produce the EC growth-promoting substance(s) was not related to donor factors like sex, age and cause of death. In addition, this ability was retained when HRPE cells were in a long-term culture, since the CM from primary cultures and late passages demonstrated identical stimulatory activities on the growth of HUVEC.

Contact inhibition is a common growth characteristic of epithelial cells. In our laboratory a confluent monolayer of HRPE cells could be maintained for several months, by refreshment with M199 containing 5% FCS once a fortnight. This is consistent with the observations of Flood et al. (1982) and Stramm et al. (1985). Furthermore, a confluent monolayer of HRPE cells could be conditioned in serum-free medium and harvested at 48-hour intervals many times without visible deterioration. Surprisingly, all these CM stimulated the growth of HUVEC to the same extent. This result may reflect the in vivo finding that the RPE monolayer remains quiescent with respect to replication while very active in metabolism, including the production of mitogenic substances.

In contrast to the results of Glaser et al. (1985, 1987), we have been unable to confirm the inhibiting effect of HRPE-CM on neovascularization either in the cell culture system or in the chick embryonic yolk sac membrane model (data not shown). The possibility remains that differences in cell species and assay systems used to determine the mitogenic activity of HRPE-CM may account for these discrepancies. As Gospodarowicz et al. (1988) reported that HUVEC contained a receptor for EGF, but bovine aortic endothelial cells (BAEC) did not, the use of FBAEC (fetal BAEC) may result in different biological responses to the mitogenic substances present in the human RPE-CM. Moreover, Glaser et al., (1985, 1987) tested HRPE-CM on FBAEC in the absence of serum, while in our experiments a minimal concentration of HAS (2%) was always added to the test media. This difference is important because it has been reported that serum or plasma is generally required for the replication of BAEC in vitro (Gajdusek and Schwartz, 1984). HUVEC showed even more dependency, cells becoming detached in serum-free medium (Jaffe, 1984; Yang and Vanden Berghe, unpublished results).

To date, the identity of the EC growth substance(s) in RPE-CM is unclear. Since Schweigerer and colleagues (1987) documented the finding of a bovine RPE cellderived bFGF immunoreactive and bioactive on the stimulation of EC growth, it has been postulated that bFGF might be responsible for the EC-mitogenic activity of RPE-CM (Wong et al., 1988; Morse et al., 1989). However, bFGF seems unlikely to be released directly into the CM, as it is a cell-associated mitogen (Folkman and Klagsbrun, 1987; Vlodavsky et al., 1987; Gospodarowicz et al., 1988) that lacks the hydrophobic signal sequence governing secretion (Abraham et al., 1986; Jaye et al., 1986). It is, therefore, important to rule out the possibility that the intracellular mitogens may have contaminated the CM via damaged cell membranes or lysis of dead RPE cells. In this study we have demonstrated that RPE cell extracts (RPE-PRS) dosedependently crossreacted with the antibody specific for bFGF, while RPE-CM showed a complete lack of recognition of both aFGF and bFGF. Furthermore, the mitogen appeared heat-labile and trypsin-sensitive but very stable to extremes of pH. These, together with the fact that the bioactive molecule could not pass through M_r 30,000 cut-off membrane, that it is a fairly high relative molecular mass polypeptide and is distinct from FGFs. Our observations strongly suggest that the EC growth-promoting activity present in HRPE-CM cannot be created by contamination with intracellular mitogens, e.g. bFGF.

The ability of normal RPE cell in vitro to synthesize and continuously secrete an EC growth factor may reflect the in vivo role of RPE in serving as a “trophic station” responsible for the embryonic development, maturation and maintenance of choriocapillaris (Henkind and Gartner, 1983). On the other hand, enhanced EC growth-promoting activity exhibited by intracellular disordered RPE cells in vitro (Smets et al., 1990) may explain the occurrence of SRN. The latter usually appears subsequent to the RPE damage and alterations that are often observed in age-related macular degeneration (Hogan, 1972; Deutman and Grizzard, 1978; Young, 1987). The possibility that the RPE-derived EC growth substance(s) may be involved in the etiology of SRN is an interesting hypothesis that emphasizes the importance of the identification of this mitogen, which may have a great value for further clinical investigations. The availability of adequate quantities of HRPE-CM has allowed us to carry out a biochemical study on the purification and identification of this RPE cell-secreted EC mitogen. The results of that study will be published elsewhere in the near future.

In conclusion, this study has demonstrated that RPE cells, derived from normal human eyes, can continuously synthesize and secrete a vascular EC mitogen into serum-free medium. The ability of HRPE cell lines to produce this mitogen seem unrelated either to in vivo donor factors or to in vitro cell life span. These findings make it possible to collect adequate quantities of CM of RPE cells obtained from a single pair of human eyes for biological, biochemical and metabolic studies of the EC growth factor. Furthermore, its striking characteristics (resistance to extremes of pH, a relative molecular mass higher than 30,000, no antigenic epitopes of aFGF or bFGF) indicate that the EC mitogen in HRPE-CM is distinct from FGFs.