Phenotype-Associated Changes in Retinal Pigment Epithelial Cell Expression of Insulin-Like Growth Factor Binding Proteins

Abstract

Purpose.: The objectives of this study were to evaluate retinal pigment epithelial (RPE) cells as a source of insulin-like growth factor binding proteins (IGFBPs) and to characterize biosynthetic changes associated with the cell phenotype and vitreous growth factor stimuli known to be present in fibrocontractive diseases.

Methods.: Early culture-associated changes in RPE phenotype were characterized by indirect immunofluorescence localization and Western blot analysis of cell lysates. IGFBP expression was evaluated by RT-PCR and Northern blot analysis of total RNA preparations.

Results.: Normal unperturbed RPE are immunoreactive for cytokeratin 18 and negative for cytokeratin 19, vimentin, and α-smooth muscle actin (αSMA). Early reactive RPE (7 days in culture) express cytokeratin 18, cytokeratin 19, and vimentin. Myofibroblastic RPE (35 days in culture) express cytokeratin 19, vimentin, and αSMA. RT-PCR studies revealed that normal RPE can produce IGFBP-2, -3, -4, -5, and -6 but not IGFBP-1. Early reactive and myofibroblastic RPE have detectable levels of message for IGFBP-3, -5, and -6. However, Northern blot analysis suggests that IGFBP-5 is the predominant binding protein produced. Finally, stimulation with biologically relevant quantities of IGF-I and IGF-II had no detectable effects on IGFBP expression.

Conclusions.: Changes in RPE phenotype are accompanied by dramatic changes in IGFBP expression profile, with IGFBP-5 the predominant binding protein produced by myofibroblastic RPE cells.

Despite significant advances in vitreoretinal microsurgical techniques in the past two decades, proliferative vitreoretinopathy (PVR) remains the leading cause of failure to repair rhegmatogenous retinal detachments. ^1–3 The pathogenesis of this potentially blinding complication is attributable to the activities of the cells that gain access to the vitreous cavity, including migration, proliferation, and extracellular matrix production. ^4–7 These activities lead to the formation of fibrocellular scars at the vitreoretinal interface from which tractional forces emanate and, ultimately, cause retinal detachment. ⁸ Among the different cell types implicated in PVR, immunocytochemical and ultrastructural studies of fibrocellular scars consistently report the presence of retinal pigment epithelial (RPE) cells. ^9–12 Evidence supporting a causal role for RPE also comes from tissue culture studies documenting the capacity of this cell type to proliferate, migrate, and generate tractional forces while progressively dedifferentiating from an epithelial to a fibroblast-like phenotype. ^13–16 When considered with animal studies involving intravitreal injection of cells, these results constitute compelling evidence that RPE-derived cells are an important effector population in PVR. ^17–19  

Studies of RPE cell behavior in vitro implicated a broad panel of growth factors as promoters of key cellular activities, including platelet-derived growth factor (PDGF), transforming growth factor-β (TGF-β), hepatocyte growth factor (HGF) and the insulin-like growth factors (IGFs). ^16,20–27 In addition to the two well-characterized ligands IGF-I and IGF-II, the IGF system is composed of at least six high-affinity insulin-like growth factor binding proteins (IGFBPs) that can potentiate or inhibit IGF activities depending on the IGFBP species, growth factor, and experimental system examined. ^28–34 A recently published study from this laboratory revealed the capacity of IGFs to stimulate RPE tractional force generation ¹⁶ at concentrations comparable to those detected in normal and diseased vitreous. ^35–40 Other studies using growth factor-neutralizing antibodies attributed the majority of contraction-promoting activity in PVR vitreous to IGF-related species. ^38,40 Although these observations provide compelling circumstantial evidence of IGFs functioning as inducers of RPE responses in PVR, the mechanistic origin of vitreous IGF activity is still unclear. Complicating this issue is our incomplete understanding of the role of vitreous IGFBPs. At least two IGFBP species, IGFBP-2 and -3, are known to be present in normal vitreous, and recently published data from this laboratory revealed the capacity of these proteins to modulate RPE responses to IGFs through growth factor binding and sequestration. ^16,41,42 Based on these findings, we postulate that IGFBPs are involved in regulating vitreous IGF bioavailability and place increased emphasis on understanding the origins of vitreous IGFBPs in normal and disease states. 

In light of the consistent involvement of RPE in PVR and the proposed role of IGFBPs, it is important to evaluate binding protein production by these cells as a disease-contributing mechanism. Although there is evidence for local IGFBP production in normal ocular tissues, another uncharacterized source are the cells actually present in fibrocellular scars. ^16,41–45 Previously published studies reported variable IGFBP expression in isolated human RPE cell lines, and a more recent study reported ubiquitous expression of all six high-affinity IGFBPs. ^46–51 Morphologic studies of RPE cells in PVR scars suggest the presence of different RPE phenotypes, and, though the later stages of RPE dedifferentiation in culture are well characterized, little is known about the early, rapid responses to perturbing stimuli. ^11,15 With this in mind, the goals of this study were to define the intermediate RPE phenotypes and to evaluate the influence of phenotype on the production of all six high-affinity IGFBPs. The potential modulating effects of IGFs on IGFBP biosynthesis were also considered because it is now known that interactions between these growth factors and Müller cells can influence IGFBP transcription. ⁵²  

Methods

Isolation and Culture of Porcine RPE Cells

RPE cells were isolated from normal porcine eyecups and were characterized and maintained in culture using previously published methods. ^15,16 The methods used to procure animal tissues complied with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Institutional Animal Care and Use Committee at the University of Alabama at Birmingham. For these studies, primary RPE cultures (passage 1) were allowed to achieve confluence (approximately 2 weeks), after which the cells were subcultured weekly. For routine culture, the cells were maintained in Dulbecco's minimum essential medium (DMEM; Invitrogen Corporation, Carlsbad, CA) containing 20 mM N-2-hydroxyethylpiperizine-N-2-ethanesulfonic acid sulfate and 10% fetal bovine serum (FBS; Invitrogen Corporation). Growth media were changed weekly, and the cells were harvested for subculture or experimentation by using 0.05% trypsin and 0.02% EDTA (Invitrogen Corporation). 

Immunofluorescence Microscopy

Porcine ocular tissues were prepared and cryosectioned using previously published methods, but in this case the retinas were removed from the eyecups before fixation. ¹⁵ For immunolocalization of cytoskeletal proteins, RPE cells attached to coverslips were probed as previously described. ¹⁵ Cryosections (10 μm) and coverslips were blocked with 20% nonimmune goat serum (Jackson ImmunoResearch, West Grove, PA) in phosphate-buffered saline (PBS; 0.01 M Na₂PO₄, 0.15 M NaCl; pH 7.4) for 60 minutes at room temperature. Specimens were then probed with primary and secondary antibodies in PBS with 2% goat serum (Jackson ImmunoResearch) for 60 minutes each with three 5-minute washes between and after antibody treatments. Photomicrographs were taken with a microscope (Optiphot; Nikon, Tokyo, Japan) equipped with epifluorescence illumination and phase-contrast optics using a digital camera (RETIGA EXi; QImaging Corp., Burnaby, BC, Canada). Images were assembled into composite photographs with image management software (Photodeluxe; Adobe Systems, Inc., San Jose, CA). 

Reverse Transcription–Polymerase Chain Reaction

Total RNA was isolated from porcine liver, RPE in porcine eyecups, and RPE cell cultures established from two different animals using a commercial reagent (TRIzol; Invitrogen Corporation) according to the manufacturer's instructions and were stored at −70°C. Gene-specific oligonucleotide primers for IGFBP-1, -2, -3, -4, -5, and -6 previously designed in our laboratory were used for this study. ⁵² Successful primer design was confirmed with RNA harvested from normal porcine liver. For RT-PCR experiments, 1 μg total RNA, 20 pM each primer, and commercial RT-PCR reagents (Ready-To-Go RT-PCR Beads; GE Healthcare, Buckinghamshire, UK) were used in 50-μL reaction volumes. Reaction programs with positive and negative controls were performed on a thermocycler (MiniCycler model PTC-150; MJ Research, Watertown, MA) and included the optimized conditions previously reported. ⁵² PCR products were separated on 2% agarose gels, visualized with ethidium bromide, and photographed with an instant camera (Polaroid Corp., Cambridge, MA) equipped with a filter (Tiffen 40.5-mm Deep Yellow 15; Fisher Biotech, Pittsburgh, PA). 

Northern Blot Analysis

Total RNA preparations were separated in agarose/formaldehyde gels, transferred, and covalently bonded to nylon membranes using previously published methods. ⁵² RNA samples from porcine liver, retina-free eyecups and RPE cultures were optimized to produce glyceraldehyde-3-phosphate dehydrogenase (G3PDH) staining of the same visual intensities. Template cDNA for each IGFBP probe was prepared as described for RT-PCR but with an additional gel-purification step (Qiaquick; Qiagen, Valencia, CA). DIG-labeled cDNA probes were prepared according to the supplier's instructions (PCR DIG Probe Synthesis Kit; Roche Diagnostics, Indianapolis, IN). Blots were prehybridized in hybridization buffer (UltraHyb; Ambion) for 1 hour at 42°C and hybridized overnight at the same temperature with 10 pM final concentration of DIG-labeled cDNA probe. The blots were washed at 42°C with increasing sodium chloride/sodium citrate stringency; block, wash, and detection solutions were prepared per instructions (DIG Wash and Block Buffer Set; Roche Diagnostics). The secondary antibody (Anti-DIG-AP Fab fragments; Roche Diagnostics) was diluted at 1:20,000 and a detection agent was used (CDP-Star; Roche Diagnostics). Blots were exposed to film (Fuji Super RX; Tokyo, Japan) and developed. 

Other Reagents

Recombinant human growth factors IGF-I and -II were obtained from GroPep, Inc. (Adelaide SA, Australia). Other antibodies obtained from commercial suppliers included mouse monoclonal anti–cytokeratin-18 (clone CY-90; Sigma-Aldrich), mouse monoclonal anti–cytokeratin 19 (clone b170; Novocastra Laboratories Ltd, Newcastle, UK), mouse monoclonal anti–vimentin (clone V9; Dako A/S, Glostrup, Denmark), mouse monoclonal anti–α-smooth muscle actin (clone 1A4; Sigma-Aldrich), mouse monoclonal β-actin (clone AC-15; Sigma-Aldrich), and Cy 2-conjugated goat anti–mouse IgG (Jackson ImmunoResearch). 

Results

Characterization of Early Reactive RPE Phenotype in Culture

To explore phenotype-associated alterations in IGFBP biosynthesis profiles, we first characterized rapid-onset changes in cytoskeletal protein expression occurring after introduction into tissue culture. Cryosections of normal porcine posterior poles (Fig. 1A) represented time 0 or normal RPE, and the progression of changes toward the previously described fibrocontractive cell end point ¹⁵ were evaluated at 24 hours (Fig. 1B), 7 days (Fig. 1C), and 35 days in culture (Fig. 1D). Cytokeratin 18 reactivity was evident in normal (Fig. 1E) and 24-hour cultures (Fig. 1F), heterogeneous in 7-day cultures (Fig. 1G), and undetectable after 35 days (Fig. 1H). In contrast, RPE cells at time 0 and 24 hours were completely negative for cytokeratin 19 (Figs. 1I, 1J), whereas cells at 7 and 35 days were homogeneously positive for this protein (Figs. 1K, 1L). Similarly, normal RPE cells were negative for vimentin (Fig. 1M), there was little or no immunoreactivity at 24 hours (Fig. 1N), and all cells were positive at 7 and 35 days (Figs. 1O, 1P). Finally, normal and 24-hour cultures were negative for αSMA expression (Figs. 1Q, 1R), there was some heterogeneous localization at 7 days (Fig. 1S), and, as we reported previously, all cells were strongly positive at 35 days in culture (Fig. 1T). ¹⁵  

Figure 1.

 

Evaluation of RPE phenotype changes in culture. Cryosections of normal porcine posterior poles (A, E, I, M, Q) and freshly isolated RPE maintained in culture for 1 day (passage 1; B, F, J, N, R), 7 days (passage 1; C, G, K, O, S), or 35 days (passage 4; D, H, L, P, T) were probed with antibodies against cytokeratin 18 (E–H), cytokeratin 19 (I–L), vimentin (M–P), or αSMA (Q–T). Unstained sections and cells were visualized by bright-field (A), phase-contrast (B–D), and indirect immunofluorescence microscopy (E–T). Some images (D, H, L, P, T) are presented at half the magnification of the others. Scale bars, 100 μm.

Immunofluorescence findings were confirmed using the same reagents in Western blot analysis of lysates prepared from parallel cultures. Extracts from normal eyecups, cell cultures at 7 and 35 days were first normalized to produce β-actin staining of similar visual intensities (not shown). Extracts from normal and 7-day cultures were positive for cytokeratin 18, whereas 35-day cultures were negative (Fig. 2A). Conversely, extracts from normal RPE were negative for cytokeratin 19 (Fig. 2B) and vimentin (Fig. 2C), but reactive bands for both proteins were detectable in 7- and 35-day cultures. Finally, under these conditions, αSMA was detected only in the 35-day culture lysates (Fig. 2D). Together, these studies indicate that profound changes in RPE phenotype occur within 7 days in culture and can be defined by differential expression of cytoskeletal proteins. For the purposes of this study, proliferating cells of the intermediate phenotype (culture day 7) are described as early reactive and cells at 35 days are referred to as myofibroblastic. 

Figure 2.

 

Changes in RPE cytoskeletal protein expression. Detergent-extracted proteins from freshly isolated RPE (lane 1) and RPE maintained in normal growth medium for 7 days (lane 2, passage 1) or 35 days (lane 3, passage 4) were evaluated by Western blot analysis for the presence of cytokeratin 18 (A), cytokeratin 19 (B), vimentin (C), and αSMA (D). The positions of the prestained molecular weight standards (kDa) are indicated to the left of each panel.

RPE Phenotype-Associated Changes in IGFBP Production

To define changes in IGFBP biosynthesis associated with phenotype, total RNA harvested from normal, early reactive, and myofibroblastic RPE maintained in growth medium were examined by RT-PCR using primers specific for IGFBP-1 through -6. RNA from normal porcine liver served as a positive control because this tissue is a reported source of all six binding proteins. ^29,53,54 Amplimers of the predicted size for IGFBP-1 were detected in liver reactions but not in RPE at any stages examined (Fig. 3). Message for IGFBP-2 and -4 was detected in normal RPE but not in early reactive or myofibroblastic phenotypes. In contrast, message for IGFBP-3, -5, and -6 was detected in RPE at all phenotypes tested. These data suggest that normal RPE can potentially produce five of the six high-affinity IGFBPs, whereas the early reactive and myofibroblastic cell repertoire is limited to three species. 

Figure 3.

 

RT-PCR evaluation of IGFBP expression. Total RNA preparations from porcine liver (lane 1), freshly isolated RPE (lane 2), early reactive (lane 3), and myofibroblastic (lane 4) RPE cells were evaluated by RT-PCR with IGFBP-specific primers. G3PDH and heat-inactivated reverse transcriptase served as positive and negative control reactions, respectively. DNA standards are in the unmarked lane on the left of each panel.

Parallel Northern blot analyses were performed to evaluate the relative abundance of IGFBP-specific messages using the same total RNA preparations described and were limited to the three binding-protein species detected in all three RPE phenotypes. IGFBP message levels varied widely in porcine liver (Fig. 4), consistent with the reported abundance in this tissue. ⁵³ Message levels for IGFBP-3 were low in normal RPE and minimally detectable in early reactive and myofibroblastic cells. IGFBP-5 message levels decreased in early reactive phenotype and were most abundant in myofibroblastic RPE. Finally, IGFBP-6 message was evident in normal RPE and undetectable in the later phenotypes. We concluded that both early and late changes in RPE phenotype are accompanied by equally dramatic changes in IGFBP expression, resulting in a more restrictive production profile that appears to emphasize IGFBP-3 and IGFBP-5. 

Figure 4.

 

Northern blot analysis evaluation of IGFBP expression. Total RNA preparations from porcine liver (lane 1), normal (lane 2), early reactive (lane 3), and myofibroblastic (lane 4) RPE were probed in Northern blot analysis for the presence of message specific for G3PDH, IGFBP-3, -5, and -6. The positions of 28S (4.7 kb) and 18S (1.9 kb) ribosomal RNA subunits are indicated to the left of each panel.

IGF System Ligand Effects on RPE Production of IGFBP

The effects of growth factor deprivation and the two IGF system ligands were evaluated by incubating myofibroblastic RPE cell cultures in serum-free media with and without 5 × 10⁻¹⁰ M IGF-I and 1 × 10⁻⁹ M IGF-II. Recent studies from this laboratory demonstrated that these concentrations effectively promote RPE tractional force generation in culture ¹⁶ and are comparable to the concentrations of free growth factor reportedly present in diabetic vitreous. ⁵⁵ Total RNA samples harvested after 24 hours were examined by RT-PCR using IGFBP-specific primers. Interestingly, the profile of detectable message under all three experimental conditions appeared to be identical with that of myofibroblastic RPE maintained in growth medium containing 10% FBS (Fig. 5 compared with Fig. 3, lane 4). Similarly, Northern blot analysis of the three detectable mRNAs were also unchanged in that IGFBP-3 and IGFBP-6 were weak or undetectable whereas IGFBP-5 levels were comparatively high (Fig. 6). Together, these observations suggest that IGFBP production by myofibroblastic RPE tends to be constitutive in that it appears to be unaltered by serum or biologically active concentrations of IGF-I and IGF-II. 

Figure 5.

 

RT-PCR evaluation of IGFBP expression in response to exogenous IGF-I and IGF-II. Total RNA preparations from porcine liver (lane 1) and myofibroblastic RPE cells incubated in serum-free medium (lane 2) with IGF-I (lane 3) or with IGF-II (lane 4) were evaluated by RT-PCR using IGFBP-specific primers. G3PDH and heat-inactivated reverse transcriptase reactions served as positive and negative controls, respectively. DNA standards are in the unmarked lane on the left of each panel.

Figure 6.

 

Northern blot analysis evaluation of IGFBP expression in response to IGF-I and IGF-II. Total RNA preparations from normal RPE (lane 1) and myofibroblastic RPE cells incubated in serum-free medium (lane 2) with IGF-I (lane 3) or IGF-II (lane 4) were probed in Northern blot analysis for the presence of G3PDH, IGFBP-3, -5, and -6-specific message. The positions of 28S (4.7 kb) and 18S (1.9 kb) ribosomal RNA subunits are indicated to the left of each panel.

Discussion

The goal of this study was to evaluate the influence of cell phenotype on RPE production of IGFBPs, and our results suggest a dynamic relationship. RT-PCR evaluation detected message for five of the six high-affinity IGFBPs in normal RPE, with IGFBP-1 the only negative species. Changes in proliferating, early reactive RPE included loss of IGFBP-2 and -4 expression while IGFBP-3, -5 and -6 remained unchanged. Interestingly, despite the dramatic changes in phenotype, this profile remained unchanged in myofibroblastic RPE. Parallel Northern blot analysis to evaluate message abundance suggested that IGFBP-5 is the major binding protein expressed by myofibroblastic RPE. Finally, we did not detect changes in IGFBP expression in response to FBS or to biologically relevant concentrations of either IGF system ligand. 

One observation with potential pathologic importance is the apparent loss of IGFBP-2 expression in early reactive RPE phenotype. Based on established ability of IGFBP-2 and -5 to inhibit IGF-induced responses, loss of these binding proteins at a local level would yield a net increase in growth factor activity and could be a contributory mechanism to pathologies involving the IGF system tractional force generation by responsive cells such as RPE and Müller cells. ^16,56 In addition, if one accepts the premise that RPE-derived myofibroblasts are present in PVR scar tissues, it seems likely that they contribute IGFBP-5 to the intravitreal environment, adding to the growth factor-binding capacity of that milieu. Given that IGFBP-5 is a potent inhibitor of RPE responses to IGF-I and -II, ¹⁶ on the surface the net effect would be self-limiting in its responsiveness to these IGF ligands. However, the true functional significance of IGFBP-5 expression may be more complex. The subject of a recent and relatively comprehensive review, IGFBP-5 is thought to play important roles in controlling cell survival, differentiation, and apoptosis in different tissue systems through IGF-dependent and independent mechanisms. ⁵⁷ However, we are aware of no studies demonstrating the presence of IGFBP-5 in vitreous; hence, the potential role of this protein in posterior pole disease remains speculative. 

These results obtained with porcine cells and tissues both compare and contrast with previous studies of IGFBP production by human RPE. Randolph et al. ^46,47 used RNase protection assays ⁴⁶ and then Northern blot analysis ⁴⁷ to examine IGFBP production by established (passages 10–20) primary cultures of human RPE. Their results were similar to our RT-PCR survey in that they consistently detected message for IGFBP-3 and -6. They did not, however, detect message for IGFBP-5. A more recent study by Yang and Chaum ⁵⁸ used PCR to evaluate IGFBP production in 13 RPE cultures at passages 2 to 21. Their results were also similar to ours in that the PCR and assays were consistently positive for IGFBP-3, -5, and -6. Unfortunately, from the experimental design it was not possible to ascertain the abundance of IGFBP-5 message relative to the other binding proteins. Their results also differed from ours in that PCR overamplification generated amplimers consistent with IGFBP-1, -2, and -4. Although these discrepancies may be a result of species-related differences, when one considers the more restricted expression profile reported by Randolph et al., ^46,47 differences in assay sensitivities seem more likely the cause. In either case, the Northern blot analyses suggest that functionally significant IGFBP production by myofibroblastic RPE is probably limited to IGFBP-3 and -5. Several other studies of permanent or immortalized cell lines reported expression of IGFBP-2 or IGFBP-3, which is also consistent with our findings for freshly isolated, normal RPE. ^49,50,59,60 However, because information allowing direct comparisons between these cell phenotypes is lacking, it seems unwise to draw conclusions from the similarities. 

Finally, our studies of RPE phenotype plasticity in culture indicated that even very early changes can be defined by cytoskeletal protein expression dynamics. Acquired protein expression seen with early reactive RPE in culture mimics that described in the early stages of age-related macular degeneration, ⁶¹ whereas myofibroblastic RPE in later culture stages show similarities with the ubiquitous fibroblastic cells associated with PVR scars. ¹² Of interest, de novo expression of cytokeratin 19 by early reactive RPE cells is consistent with previously published studies associating this change with migration and proliferation, activities considered key to PVR pathobiology. ^62,63 Based on this result, cytokeratin 19 may prove to be a useful marker to identify fibroblast-like cells of RPE lineage in disease. 

Footnotes

 Supported by the International Retinal Research Foundation, National Institutes of Health Grant EY013258, and Research to Prevent Blindness.

Footnotes

 Disclosure: S. Mukherjee, None; J.L. King, None; C. Guidry, None

Footnotes

 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.