IGF-1–Induced VEGF and IGFBP-3 Secretion Correlates with Increased HIF-1α Expression and Activity in Retinal Pigment Epithelial Cell Line D407

Abstract

purpose. To examine insulin-like growth factor (IGF)-1 stimulation of expression of hypoxia inducible factor (HIF)-1α and secretion of vascular endothelial growth factor (VEGF) and IGF binding protein (IGFBP)-3 in human retinal pigment epithelial (RPE) cell line D407.

methods. D407 cells cultured in dishes or Transwell inserts were treated with cobalt chloride or varying doses of IGF-1. Whole cell lysates were assayed by immunoblot for HIF-1α expression, whereas conditioned medium was TCA precipitated and assayed by immunoblot for VEGF and ligand blot for IGFBP-3. Cells grown on coverslips were similarly treated and probed with antibodies to HIF-1α, VEGF, and IGFBP-3, and visualized by epifluorescence microscopy. Cells grown on Transwell inserts were probed with antibodies to the Na⁺/K⁺-ATPase α-1 subunit and either the alpha or beta subunits of the IGF-1 receptor and visualized in Z-section using confocal microscopy.

results. Immunoblot analysis of whole cell lysates from IGF-1–treated D407 cells revealed the upregulation of HIF-1α protein. Epifluorescence microscopy demonstrated a positive correlation between HIF-1α expression and nuclear localization, VEGF and IGFBP-3 synthesis and export, and IGF-1 action. Western and ligand blot analyses of RPE cell–conditioned medium indicated that IGF-1 induced increases in VEGF and IGFBP-3 secretion. Cells grown on Transwell inserts exhibited constitutive apical secretion of VEGF and IGFBP-3, which increased on apical or basolateral treatment with IGF-1. Confocal analysis of Transwell-cultured D407 cells confirmed the apical localization of the Na⁺/K⁺-ATPase α-1 subunit, characteristic of polarized RPE, with IGF-1 receptor α and β subunits exhibiting a nonpolarized distribution.

conclusions. IGF-1 stimulates increased HIF-1α expression as well as VEGF and IGFBP-3 secretion in D407 cells. Similar to their in vivo counterparts, D407 cells maintain reversed epithelial polarity. Apical secretion of VEGF and IGFBP-3 increases in response to either apical or basolateral IGF-1 stimulation consistent with the nonpolarized distribution of IGF-1 receptors.

Age-related macular degeneration (AMD) is characterized by the accumulation of drusen within Bruch’s membrane and macular hypo- or hyperpigmentation of the retinal pigment epithelium (RPE). ¹ A consequence of AMD is choroidal neovascularization (CNV), where newly formed blood vessels arising from the choroid enter the subretinal space, leak and bleed, leading to retinal detachment and photoreceptor death. ^{1 2 3 4 5 6} To date, argon laser photocoagulation is the principal, yet considerably ineffective treatment for CNV. ⁷ This underscores the reason this disease represents the most common cause of severe vision loss in elderly patients in developed countries. ² Identification of the mediators of ocular angiogenesis involved in the progression of CNV would provide important targets for the development of selective inhibitors. ^{7 8} Of particular interest to our studies is the RPE, a monolayer of highly specialized epithelial cells interposed between the retinal photoreceptors and the choroid that is central to photoreceptor survival and function. ^{9 10} With its apical surface in intimate contact with the light-sensitive cells of the retina, the RPE performs critical transport, barrier, and phagocytic support functions for the neural retina. ⁹ These functions require a characteristic apical distribution of certain proteins that are usually found on the basolateral membrane in other epithelia. A prime example and often used marker of RPE polarity is sodium/potassium ATPase (Na⁺,K⁺-ATPase), necessary for providing a Na⁺ rich environment appropriate for photoreceptor function. ^{11 12 13 14 15 16 17}  

The RPE is a major source of angiogenic (e.g., vascular endothelial growth factor, VEGF) and anti-angiogenic (e.g., pigment epithelium derived factor, PEDF) factors, thereby playing a central role in the modulation and progression of CNV. ^{18 19 20 21 22 23 24 25 26} Numerous animal models support a role for RPE-derived VEGF in the progression of CNV. ^{27 28 29 30 31 32 33 34} In addition to elevated VEGF levels in the vitreous, ²⁴ the RPE and surrounding subretinal membranes of CNV-afflicted retinas express increased levels of VEGF and its receptor KDR/flk-1 ^{35 36} (Amin RH, et al. IOVS 1995;36:ARVO Abstract 2565). A number of factors regulate VEGF production, foremost being hypoxia. Insulin-like growth factor (IGF)-1 stimulates VEGF expression in RPE cells. ²⁰ Punglia and coworkers ²⁰ have shown that serum and vitreous IGF-1 levels correlate with a wide variety of ischemic retinal disorders linked to neovascularization of the retina and iris. The RPE possesses receptors for IGF-1 and secretes both IGF-1 and IGF-2 as well as IGF binding proteins 3 and 6 ^{37 38 39 40 41 42} (Hunt RC, et al. IOVS 1987;28:ARVO Abstract 45). Along with the variable expression of these components in surrounding cells, the retina contains all components of a self-contained IGF-1/IGFBP (IGF binding protein) autocrine system capable of modulating normal retinal function and contributing to the pathogenesis of CNV through dysregulation. ³⁷  

In addition to growth, differentiation, proliferation, basement membrane degradation, and survival, IGF-1 secreted by the RPE plays an active role in regulating phototransduction through modulation of the rod light response. ^{37 43 44 45} Interestingly, the RPE also secretes IGFBPs into the interphotoreceptor matrix. ³⁷ Given their role as high affinity IGF antagonists, cell-specific expression of IGFBPs may allow for autocrine/paracrine control of IGF function in the pericellular microenvironment. ^{46 47 48 49} IGF-1R antagonism of IGF-1 suppresses retinal neovascularization and reduces the retinal endothelial cell response to VEGF, implicating an inhibitory role for IGFBPs by restricting IGF-1 access to receptors. ⁵⁰ A number of investigators have suggested that IGFBPs facilitate growth by augmenting IGF-1 interaction with receptors, thus potentiating a mitogenic response. ^{46 47 48 49} IGF-independent effects of IGFBPs in the retina have yet to be discerned. 

The physiological consequence of hypoxia in a variety of cells is an increase in the level of hypoxia-inducible factor (HIF)-1α protein through a decrease in proteasomal degradation, resulting in HIF-1 heterodimer formation, and transcriptional activity at various hypoxia response element (HRE) containing promoters (e.g., VEGF). ^{51 52} In addition, even in the absence of a defined HRE, HIF-1α is essential for hypoxia-induced protein expression (e.g., IGFBP-3). ⁵³ Based on the growing body of evidence demonstrating IGF-1-induced HIF-1 activity, and thus the potential contributions of this growth factor to AMD, IGF-1–stimulated HIF-1α protein expression and VEGF and IGFBP-3 secretion were examined in the D407 retinal pigment epithelial cell line. Immunoblot analysis of whole cell extracts from IGF-1–treated D407 cells revealed the upregulation of HIF-1α protein levels. Epifluorescence microscopy of D407 cells revealed a positive correlation between HIF-1α expression and nuclear localization, VEGF and IGF binding protein-3 (IGFBP-3) synthesis and secretion, and IGF-1 action. Western and ligand blot analysis of RPE-conditioned medium demonstrated that IGF-1 increased VEGF and IGFBP-3 secretion. These findings demonstrated that, as seen for VEGF, IGF-1–induced IGFBP-3 secretion in RPE cells correlated with increased HIF-1α expression and nuclear localization. In addition, confocal analysis of polarized D407 cells cultured on Transwell inserts revealed apical Na⁺/K⁺-ATPase, whereas the IGF-1 receptor exhibited a nonpolarized distribution. In this context, D407 cells exhibited constitutive apical secretion of VEGF and IGFBP-3. Increased apical secretion of both proteins in response to apical or basolateral IGF-1 treatment was consistent with the immunolocalization of IGF-1 receptors on both apical and basolateral membranes. 

Methods

Materials and Reagents

D407 cells were generously provided by Richard Hunt (Department of Immunology and Pathology, University of South Carolina Medical School, Columbia, SC). The following purchases were made: fetal bovine serum (FBS) from Atlas Biologicals (Fort Collins, CO); Dulbecco’s modified Eagle’s medium (DMEM) from Sigma (St. Louis, MO); cell culture dishes from Falcon (Franklin Lakes, NJ); Transwell inserts from Corning Costar (Cambridge, MA). Bacterially produced recombinant human IGF-1 and nonglycosylated VEGF was generously provided by Genentech (San Francisco, CA). Cobalt chloride was purchased from Fisher Scientific (Fair Lawn, NJ); tunicamycin from Calbiochem (San Diego, CA); bicinchoninic acid (BCA) reagent for protein quantification from Pierce (Rockford, IL); VEGF polyclonal antibody from Chemicon (Temecula, CA); HIF-1α monoclonal antibody from Transduction Laboratories (Lexington, KY); HIF-1β monoclonal antibody from Novus (Littleton, CO); β-actin polyclonal antibody from Sigma; and Na⁺,K⁺-ATPase α-1 antibody from Upstate (Lake Placid, NY). HRP, FITC, and rhodamine-conjugated secondary antibodies were purchased from Chemicon; Neutravidin-horse radish peroxidase (neutravidin-HRP) from Pierce; Alexa Fluor 488 and Alexa Fluor 546 secondary antibodies from Molecular Probes (Eugene, OR); enhanced chemiluminescence (ECL) reagent from Amersham Biosciences (Clearbrook, IL); Biomax film from Kodak (Rochester, NY); fugene 6 from Roche (Indianapolis, IN); and Dual-Luciferase Reporter Assay System from Promega (Madison, WI). p2.1 was a generous gift from the laboratory of Gregg L. Semenza (Johns Hopkins University School of Medicine, Baltimore, MD), and pRL-SV40 Renilla (Promega) was provided by David T. Kurtz (Department of Cell and Molecular Pharmacology and Experimental Therapeutics, Medical University of South Carolina). All other chemicals were of reagent grade or higher. 

Tissue Culture

D407 cells were cultured in DMEM Base D-5030 with 10% FBS and 10 μL/mL penicillin-streptomycin solution. Unless otherwise stated, cells were maintained at 37°C in a humidified 5% CO₂–95% air incubator. 

IGF-1, CoCl₂, and Tunicamycin Treatments

D407 cells were seeded at a density per well of 4.3 × 10⁵, respectively, in 6-well (9.6 cm² area) plates. Confluent cells were serum starved (FBS was eliminated in all experiments) for 24 hours before treatment with IGF-1, CoCl₂, and tunicamycin as indicated. 

Transwells

D407 cells were seeded at a density of 6.5 × 10⁵ per insert on Transwell clear tissue culture–treated polystyrene membranes (4.7 cm² area, 6-well cluster plate; Corning Costar) and grown to confluency. In addition to electrical resistance, the tightness of the monolayer was evaluated by apical or basolateral addition of biotinylated IGF-1 and subsequent sampling of conditioned medium from each compartment after 4 hours. Cells were serum starved for 24 hours, followed by administration of IGF-1 either apically or basolaterally. Conditioned medium was sampled from each compartment after 12 hours as indicated. 

Immunoblot and Ligand Blot Assays

Confluent serum-starved cells were treated with IGF-1 or CoCl₂ as indicated, and whole cell lysates were prepared using a modified radioimmunoprecipitation (RIPA) buffer containing 50 mM Tris-HCl pH 7.4, 1% Triton X-100, 150 mM NaCl, 10 mM EDTA, 1 mM phenylmethylsulfonyl fluoride (PMSF), 10 μg/mL aprotinin and leupeptin, 2 mM sodium orthovanadate, and 10 mM NaF. Protein content was determined by BCA assay, and 100 μg aliquots were solubilized in SDS sample buffer. VEGF and IGFBP-3 were quantified in conditioned medium after precipitation in 10% trichloroacetic acid (TCA), washing of the pellet with acetone, and solubilization in SDS sample buffer. Proteins so collected were resolved on 10% nonreducing polyacrylamide gels in the case of lysates and 12.5% polyacrylamide gels in the case of secreted proteins, transferred to nitrocellulose (Osmonics, Westborough, MA) with a TE-70 SemiPhor apparatus (Hoefer Scientific Instruments, San Francisco, CA), and subjected to ligand or immunoblot assays. For ligand blot analysis, protein containing nitrocellulose membranes were washed for 10 minutes at 23°C in Tris-buffered saline (TBS) containing 3% Triton X-100 and blocked for 1 hour with TBS containing 0.2% gelatin. Blots were probed overnight at 4°C with 10 ng/mL of tetrabiotinylated–IGF-1 (manuscript in preparation), followed by a 2-hour incubation at 23°C with 200 ng/mL Neutravidin-horse radish peroxidase (neutravidin-HRP) in TBS containing 0.1% Tween-20 and 0.1% BSA. 

For immunoblots, the nitrocellulose membranes were blocked for 1 hour in bovine lacto transfer technique optimizer (BLOTTO), a TBS solution containing 0.1% Tween and 5% instant nonfat dry milk (Saco Foods, Inc., Middleton, WI) milk protein (reviewed in Ref. ⁵⁴ ), before being probed with either 1 μg/mL VEGF polyclonal antibody or 1 μg/mL HIF-1α monoclonal antibody, 1 μg/mL HIF-1β monoclonal antibody, or 1:10,000 β-actin monoclonal antibody in BLOTTO. HRP-linked secondary antibodies diluted 1:5000 in BLOTTO were subsequently added for 2 hours. To reprobe HIF-1α immunoblots for HIF-1β or β-actin levels, antibodies were removed from the nitrocellulose via the application of Chemicon light stripping solution according to the manufacturer. Blots were visualized with the ECL reagent on Biomax film. Films were subsequently digitized to tiff format and band intensity quantified using NIH Image, version II; Bethesda, MD. 

Luciferase Assays

The transcriptional activity of HIF-1 was assayed with the pGL2 basic p2.1 enolase (ENO)1 promoter vector, which contains a 68-bp ENO1 promoter fragment encompassing a HIF-1 binding site downstream from the luciferase gene. ⁵⁵ Each well of subconfluent D407 cells was transiently cotransfected with 100 ng reporter plasmid and 50 ng pRL-SV40 renilla as a control for transfection efficiency. After 24 hours, cells were treated with 100 nM IGF-1 or 100 μM CoCl₂ in 500 μL per well of fresh serum-free medium. After an 18-hour incubation, cells were lysed in 100 μL per well of passive lysis buffer provided with the Dual-Luciferase Reporter Assay System. Cells were scraped, centrifuged for 10 minutes at 18,890g, 20 μL of supernatant per sample loaded on a 96-well plate, and processed for luciferase activity on the Victor ² 1420 Multilabel Counter (Perkin Elmer Life Sciences, Downers Grove, IL) using firefly and renilla luciferase buffers provided with the Dual-Luciferase kit. 

Epifluorescence Microscopy

D407 cells were grown to confluency on glass coverslips placed in 6-well plates. After 24 hours of serum starvation and indicated treatments, cells were fixed with 3% paraformaldehyde for 20 minutes at 23°C. Cells were permeabilized and nonspecific binding was blocked by incubation for 12 hours at 4°C in PBS containing 1% BSA and 0.1% Triton X-100. After three 5-minute washes with PBS, coverslip-cultured cells were incubated with antibodies to HIF-1α, VEGF, or IGFBP-3 at a concentration of 10 μg/mL in PBS containing 0.1% BSA overnight at 4°C. Coverslips were subsequently washed three times for 5 minutes each with PBS containing 0.1% BSA and incubated with Texas red-conjugated goat anti-mouse IgG (Chemicon) or FITC-conjugated goat anti-rabbit IgG (Chemicon) at a 1:100 dilution in PBS containing 0.1% BSA for 1 hour at room temperature in total darkness. Cells were incubated for an additional 10 minutes with Hoechst stain (1:80,000) in PBS, followed by two 5-minute washes with PBS. Coverslips were slide mounted in Fluorsave (Calbiochem), viewed on a Leica DMLB epifluoresence microscope (Leica, Exton, PA), captured with a Hamumatsu color chilled 3CCD camera (model #C5810; Bridgewater, NJ), and digitized to tiff format using AdobePhotoshop version 5.0.2 (San Jose, CA). 

Confocal Microscopy

D407 cells grown to confluency on Transwell filters were serum starved, fixed, permeabilized, and blocked as described above. After three 5-minute washes with PBS, filters were incubated with antibodies to Na⁺,K⁺-ATPase α-1 and either the alpha or beta subunit of the IGF-1 receptor (Santa Cruz Biotechnology, Santa Cruz, CA) at a concentration of 5 μg/mL in PBS containing 0.1% BSA overnight at 4°C. Filters were subsequently washed three times for 5 minutes each with PBS containing 0.1% BSA and incubated with Alexa Fluor 488 and Alexa Fluor 546 secondary antibodies at 5 μg/mL in PBS containing 0.1% BSA for 1 hour at room temperature in total darkness. After three 5-minute washes with PBS, filters were incubated for 10 minutes with Slowfade Light Antifade Kit equilibration buffer (Molecular Probes). Transwell filters were mounted between slide and coverslip in Slowfade Light Antifade Kit antifade reagent with glycerol buffer, and viewed in Z section on a Leica Total Confocal System, Spectral Prism 2, Acoustic Optical Beam Splitter (TCS SP2 AOBS) in the Hollings Cancer Center Molecular Imaging Laboratory. 

Results

VEGF and IGFBP-3 Secretion

To examine IGF-1–stimulated increases in VEGF secretion, serum-starved D407 cells were incubated in the presence or absence of 100 nM IGF-1 for 1 to 24 hours. As shown in Figure 1A , VEGF secreted by D407 cells migrates on nonreducing SDS gels as a doublet of ≈42 and 44 kDa. This represents VEGF homodimers that are singly and doubly glycosylated at each of the single consensus N-linked glycosylation sites. Time course analysis revealed that VEGF secretion exhibited a lag phase of approximately 6 hours, becoming significantly greater at 12 and 24 hours (Fig. 1B) . 

To elucidate further the stimulatory role of IGF-1 on IGFBP-3 secretion in RPE cells, originally characterized by Feldman and Randolph, ⁵⁶ serum-starved D407 cells were incubated in the absence or presence of 100 nM IGF-1 and examined secretion incrementally over 24 hours. Conditioned medium was processed as described in Methods. As shown in Figure 2A , four bands ranging in size from ≈28 to 45 kDa and an additional two bands at 26 and 24 kDa were secreted by D407 cells and detected by ligand blot. An identical banding pattern was observed by immunoblot analysis using an IGFBP-3 polyclonal antibody (not shown), confirming that our ligand blots had detected IGFBP-3 (not shown). Given that IGFBP-3 contains three consensus sites for N-linked glycosylation, the pattern of the upper four bands most likely represented nonglycosylated IGFBP-3 and IGFBP-3 glycosylated at one, two, and three of these sites. To define the contribution of N-linked glycans to the observed banding pattern, D407 cells were incubated with tunicamycin, an inhibitor of N-linked sugar transfer from dolichol precursors. ⁵⁷ As shown in Figure 2A , tunicamycin treatment of unstimulated cells resulted in the expression of a single species of IGFBP-3 in conditioned medium that comigrated with recombinant nonglycosylated IGFBP-3. Bands representing mono-, di-, and triglycosylated IGFBP-3, ranging from 30 to 35 kDa, were absent. Similarly, IGF-1–stimulation of tunicamycin-treated cells resulted in the absence of glycosylated isoforms and a single-species nonglycosylated IGFBP-3. In addition, an increased accumulation of IGFBP-3 proteolytic fragments exhibiting unaltered electrophoretic migration was observed in tunicamycin-treated cells. Glycosylated, nonglycosylated, and proteolytic fragments were confirmed to be IGFBP-3 by immunoblot analysis (data not shown). 

Bands representing non-, mono-, di-, and triglycosylated IGFBP-3, along with IGF-1 binding proteolytic fragments were measured to quantify IGFBP-3 secretion in all subsequent figures. IGFBP-3 release from D407 cells exhibited a lag phase of approximately 6 hours before a sufficient quantity was detectable in the conditioned medium (Fig. 2B) . From that point, until the end of the assay (24 hours), IGFBP-3 secretion was approximately linear. IGF-1 addition significantly increased the level of IGFBP-3 secretion over unstimulated cells. Comparison of the data in Figure 2B indicates that D407 cells exhibited significant constitutive release of IGFBP-3 with 100 nM IGF-1 stimulating this level threefold. 

Stimulation of D407 cells with a series of IGF-1 doses for 12 hours resulted in a dose-dependent increase in VEGF secretion (Fig. 3) . Maximal VEGF secretion was observed at a concentration of 100 nM IGF-1, with a 15-fold induction of VEGF secretion over control. As shown in Figure 4 , D407 cells also exhibited an IGF-1 dose-dependent increase in IGFBP-3 secretion comparable to that seen for VEGF secretion. 

HIF-1α Protein Expression

VEGF is principally regulated by hypoxic stimulation, leading to the altered turnover of HIF-1α, the regulated binding partner present in the HIF-1 transcription complex. To establish an effect of IGF-1 on HIF-1α expression, serum-starved D407 cells were exposed to a maximal stimulatory dose of IGF-1 (100 nM) for 4 hours. This treatment resulted in increased HIF-1α protein expression (Fig. 5A) . As a positive control, cells were exposed to 100 μM CoCl₂, a chemical hypoxia-inducing agent. Close scrutiny reveals a doublet band at ≈120 kDa in untreated and IGF-1 treated cells, whereas a single, extremely intense band migrated with the lower band of the doublet and an absence of the upper band, in the case of CoCl₂ stimulation. Studies conducted by Cockman and colleagues ⁵⁸ demonstrate the slower migrating isoforms of HIF-1α to be ubiquitinated. This explanation accounts for the absence of the upper HIF-1α band when cells were treated with CoCl₂, an inhibitor of HIF-1α ubiquitination. ⁵⁹  

The blots were stripped and reprobed for HIF-1β, the constitutively expressed HIF-1α binding partner present in the HIF-1 heterodimer (Fig. 5B) . As expected, neither IGF-1 nor CoCl₂-induced changes in HIF-1β expression in D407 cells, demonstrating the specificity of these agents on HIF-1α expression and comparative increase in HIF-1α protein expression (Fig. 5C) . To establish the connection between HIF-1α protein expression and functional HIF-1 heterodimer formation, transcriptional activity was analyzed (Fig. 5D) . D407 cells were cotransfected with the p2.1 HIF-1 reporter plasmid and the pRL-SV40 control plasmid and subject to conditions described above for 18 hours. As expected, IGF-1– and CoCl₂-stimulated increases in HIF-1 transcriptional activity roughly correlated with increases in HIF-1α protein expression. 

Treatment of serum-starved D407 cells with IGF-1 for 4 hours resulted in a dose-dependent increase in the HIF-1α doublet with maximal effect occurring at 10 nM IGF-1 (Fig. 6) . To determine the temporal parameters of IGF-1–induced HIF-1α protein expression, cells were stimulated with 100 nM IGF-1 over a 24-hour time course. As shown in Figure 7 , D407 cells exhibited a consistent IGF-1–induced increase in HIF-1α protein expression as early as 2 hours, reaching greatest significance at 4 to 6 hours, and remaining elevated until 12 hours, before dropping off to basal levels by 24 hours. Though difficult to distinguish in comparison to Figures 5 and 6 , HIF-1α resolved as a doublet in Figure 7 as well. β-Actin levels were unaffected by treatment with IGF-1. 

To determine whether the increase in HIF-1α protein expression translated into increased HIF-1α translocation to the nucleus, cells grown on coverslips were stimulated with IGF-1, followed by fixation and immunolocalization of HIF-1α (Figs. 8 and 9) . VEGF (Fig. 8) and IGFBP-3 (Fig. 9) were also immunolocalized in these cells. As shown in both figures, IGF-1 stimulation led to increased HIF-1α protein levels compared to control. It is also apparent in these figures that IGF-1 induced an increase in the expression of VEGF and IGFBP-3. 

D407 Cell Polarity

The polarity of VEGF and IGFBP-3 secretion in confluent D407 cells seeded on Transwell inserts was then examined. Immunoblot analysis of conditioned medium sampled from both the upper and lower compartments revealed that IGF-1–stimulated VEGF secretion principally occurred at the apical pole, irrespective of whether IGF-1 was added to the upper or lower chamber (Fig. 10) . Similarly, ligand blot analysis of conditioned medium indicated that IGF-1–stimulated IGFBP-3 secretion predominantly occurred at the apical domain, regardless of which chamber received IGF-1 (Fig. 11) . These results indicate that secretion of VEGF and IGFBP-3 is primarily via the apical secretory pathway, while IGF-1 receptors themselves lack a polarized distribution. To test this possibility, confluent D407 cells seeded on Transwell inserts were fixed and IGF-1 receptor α subunits were immunolocalized using the Na⁺,K⁺-ATPase α-1 subunit as a control. Confocal Z-section analysis revealed a nonpolarized distribution of IGF-1 receptor α subunits (Fig. 12A) whereas the Na⁺,K⁺-ATPase α-1 subunit was exclusively localized to the apical plasma membrane (Fig. 12B) . These findings are consistent with the known reversed epithelial polarity of RPE cells. ^{11 12 13 14 15 16 17}  

Discussion

The distribution and effects of the IGF-1 receptor on HIF-1α protein levels relative to VEGF and IGFBP-3 secretion was examined in the RPE cell line, D407. In this cell line, IGF-1 stimulated a time- and dose-dependent increase in HIF-1α, the regulated member of the HIF-1 heterodimer. Nuclear localization of HIF-1α, as revealed by epifluorescence microscopy and promoter activity, assayed through transfection with the p2.1 HIF-1 reporter plasmid, confirmed the formation of productive HIF-1 heterodimers. Under normoxic conditions, HIF-1α is maintained at low levels by a degradation process involving the ubiquitin-proteosome system. ^{60 61} Treins and colleagues ⁶² reported that insulin stimulation of HIF-1α translation is regulated by a phosphatidylinositol 3-kinase (PI 3-K)-dependent signaling pathway in ARPE-19 cells. In a prior study, ⁶³ this group reported that insulin- and IGF-1–stimulated VEGF expression proceeds via different signaling pathways in NIH 3T3 cells. Whereas insulin stimulates PI 3-K/protein kinase B (pkB), induction by IGF-1 involves ERK/mitogen-activated protein kinase (MAPK). In contrast, Fukuda and colleagues ⁵⁹ reported that IGF-1 induces expression of HIF-1α through both PI 3-kinase and MAP kinase pathways in HCT116 human colon cancer cells. 

Although the signaling cascade leading to IGF-1–induced HIF-1α expression is still intensely debated, it is well established that the VEGF promoter contains hypoxia response elements (HREs), activation of which results from the binding of HIF-1 (reviewed in Ref. ⁶⁴ ). Similarly, a connection, though tenuous, has been established between HIF-1 activity and IGFBP-3 protein expression. Work by Feldser and colleagues ⁵³ demonstrates that though the IGFBP-3 promoter lacks an obvious HRE, IGFBP-3 gene expression is markedly reduced in HIF-1α–deficient cells under hypoxic conditions. These findings suggest some other level of regulation, possibly an indirect effect. 

Although the specific signaling pathways involved in IGF-1–induced VEGF and IGFBP-3 secretion were not examined, the present study demonstrated that IGF-1 stimulated increased HIF-1α protein expression and HIF-1 activity that correlated with increased secretion of VEGF and IGFBP-3 in a time- and dose-dependent manner. The timecourse discrepancy between IGF-1–induced HIF-1 protein expression (4 hours) and the secretion of VEGF and IGFBP-3 (6 hours) most likely was due to the lag between increase in HIF-1α levels stimulating VEGF and IGFBP-3 promoter activity and the subsequent process of transcription, translation, and protein modification necessary before secretion allowed for accumulation of these proteins in the conditioned medium where they were measured. Yet, the continued increase in VEGF and IGFBP-3 at points late in the timecourse when HIF-1 levels returned to baseline suggests that whereas HIF-1 may be involved, other signaling cascades downstream of IGF-1 may stimulate expression of these secreted proteins. ⁶⁵  

Biotinylated IGF-1 ligand blot and IGFBP-3 immunoblot analyses of medium from tunicamycin-treated cells suggested that D407 cells primarily secrete glycosylated isoforms of IGFBP-3. The bands which appeared below intact IGFBP-3 in response to IGF-1 stimulation were also detected by IGFBP-3 immunoblots, indicating they are active fragments of IGFBP-3. The migration of these proteolytically cleaved fragments remained unchanged in response to tunicamycin treatment, indicating they are nonglycosylated. Considering that the three sites of glycosylation occurred in the mid-region of IGFBP-3, our findings imply these fragments consist of either amino or carboxyl terminal fragments. ⁶⁶ Based on immunocytochemical analysis, increased intracellular VEGF and IGFBP-3 levels were detectable before the accumulation of these proteins in the medium. Together, these findings extend the initial work of Randolph et al. ⁶⁷ and Punglia et al., ²⁰ demonstrating IGF-1–induced increases in VEGF and IGFBP-3, respectively. 

Previous work by Blaauwgeers and coworkers, ⁶⁸ using primary human RPE cells, demonstrated that VEGF secretion is primarily basolateral, increasing under hypoxic (1% O₂) conditions. In contrast, our studies and those of Marmorstein et al. ⁶⁹ indicate that VEGF is preferentially secreted apically by polarized RPE cells. IGF-1 stimulation or adenoviral expression of VEGF leads to enhanced apical secretion of VEGF secretion in each study, respectively. ⁶⁹ This discrepancy may be due to several factors. The Blaauwgeers study used transepithelial resistance (TER) to determine the tightness of their RPE cell monolayers. Though TER provides an estimation of junctional integrity, it provides little information as to imperfections across the monolayer. Hence, the addition of the extracellular compound ¹⁴C mannitol or labeled protein, as used in the present study, acts as a better determinant of junctional integrity across the entire filter. Considering the lower compartment holds anywhere from four to six times the volume of the upper compartment, simple diffusion may explain the observed basolateral accumulation of VEGF reported by Blaauwgeers. ⁶⁸ In contrast, the strongly apical accumulation of VEGF found in our studies of D407 cells and of Marmorstein et al. ⁶⁹ using the established polarized RPE cell line RPE-J, presents VEGF accumulating substantially against this diffusion gradient. Further, whereas Blaauwgeers used transmission electron microscopic examination of microvilli and tight junctions to characterize epithelial cell polarity, confocal microscopy was used in the present study to immunolocalize Na⁺,K⁺-ATPase. A distinguishing feature of the RPE is the localization of Na⁺,K⁺-ATPase to the apica1 surface. This pattern of distribution has been referred to as the RPE “reversed polarity.” ^{11 15} A major difficulty in the study of RPE cell polarity is the absence of apical Na⁺,K⁺-ATPase polarity in many primary and permanent cell culture systems. ^{70 71} In the present studies, apical immunolocalization of Na⁺,K⁺-ATPase was revealed by confocal microscopy, thus providing confidence that D407 cells cultured on Transwells represent a valid model for the study of IGF-1 receptor distribution and the polarity of VEGF and IGFBP-3 secretion. This may, in part, be due to the constituents of the medium, as Hu and Bok ⁷² reported that DMEM with high glucose (25 mM) is sufficient to support the differentiation of human RPE cells into functionally polarized monolayers. Immunolocalization of the IGF-1 receptor to both the apical and basolateral membranes in cells exhibiting apical immunolocalization of Na⁺,K⁺-ATPase suggests that IGF-1 coming from distinct regions of the retina can regulate RPE cell function. In support of a nonpolarized IGF-1 receptor distribution, application of IGF-1 at either the apical or basolateral surface caused an increase in apical secretion of VEGF and IGFBP-3. Secretion of VEGF and IGFBP-3 into the apical chamber indicated that both proteins were sorted through the apical secretory pathway. In addition to VEGF, apical release of IGFBP-3 by RPE cells may have important implications in the regulation of IGF-1/IGF-2 autocrine and/or paracrine functions at the RPE and photoreceptor layers, given that IGF action may be inhibited ⁷³ or enhanced ^{74 75} by IGFBP-3. As such, fluctuations in IGF-1, IGF-2, or IGFBPs may have significant implications on RPE proliferation and migration after choroidal capillary invasion and subsequent leakage of circulatory IGFs from choroidal vessels. ^{35 36 67 76 77} Consequently, dysregulation of the IGF-1 system at the level of the subretina may contribute to both the changes in RPE morphology and increases in angiogenic factor secretion, consistent with CNV. In addition, considering the nonpolarized distribution of the IGF-1 receptor and the ability of IGF-1 dosed to the lower Transwell compartment to stimulate VEGF and IGFBP-3 secretion, sub-RPE choroidal vessel invasion and circulatory IGF leakage may contribute to changes in RPE morphology and angiogenic factor secretion conducive to the development of CNV. Future studies aimed at defining changes in apical versus basolateral IGF-1/ IGF-2 levels will provide insight into this issue. 

The preferentially apical secretion of VEGF by RPE cells into the subretinal space at first appears counterintuitive to the progression of choroidal neovascularization, considering that the tight junctional barrier established by these epithelial cells should prevent the free diffusion of VEGF between these cells toward the choroid. If CNV leads to the upward migration of the choroidal vasculature, presumably driven by a VEGF chemoattractant gradient through Bruch’s membrane, past the epithelial layer, and into the subretinal space, apically released VEGF secretion provides the driving signal. In support of this model, choroidal neovascularization was found to develop in a transgenic mouse model where VEGF was overexpressed in RPE cells. ⁷⁸ In addition, recombinant adenoviral gene delivery of VEGF, specifically targeting the RPE, leads to CNV in the rat. ³² Of particular note in these studies was the lack of retinal neovascularization. This stands in striking contrast to the targeted, opsin promoter–driven expression of VEGF by photoreceptors, which leads to retinal, rather than choroidal, neovascularization. ^{79 80 81} This may be explained by VEGF secretion into upper retinal layers. Any VEGF accessing the subretinal space would be antagonized by anti-angiogenic factor secretion (i.e., PEDF) preventing angiogenesis at the subretinal space. ^{19 82 83} RPE cells secrete the anti-angiogenic factor PEDF. ^{18 20} Dysregulation of the balance between angiogenic and anti-angiogenic factors may contribute to the progression of retinal vascular disease, including choroidal neovascularization (reviewed in Ref. ⁸⁴ ). Alternatively, transcytosis of apically secreted VEGF to the basolateral domain would limit VEGF accumulation in the subretinal space while providing access to the Bruch’s membrane and delivery to the choroid. This, in turn, would stimulate choroidal neovascularization and the upward migration of new vessels toward the higher VEGF concentration. 

It has been suggested that because VEGF and its receptors are co-localized to fibroblasts, endothelial, and RPE cells, altered autocrine and/or paracrine VEGF loops may be involved in the progression of experimental choroidal neovascularization (discussed in Ref. ³⁶ ). On the other hand, VEGF may induce changes in tight junction structure and function in the RPE, similar to that seen in endothelial cells. ⁸⁵ This, in turn, may lead to the paracellular diffusion of VEGF toward the choroid, facilitating choroidal neovascularization and the spread of capillary growth toward and/or through the RPE layer into the subretinal space. This would serve to explain the consistent, low levels of secreted VEGF and IGFBP-3 measured in the basolateral compartment in the present study. In addition, the breakdown of tight junctions might explain how elevated levels of subretinal VEGF lead to the progression of CNV in various animal models. 

In summary, IGF-1 stimulated the expression and nuclear translocation of HIF-1α and the secretion of VEGF and IGFBP-3 in both a time- and dose-dependent manner. Constitutive secretion of VEGF and IGFBP-3 via the apical secretory pathway was observed. The nonpolarized distribution of the IGF-1 receptor in polarized RPE cells is consistent with observed increases in apical secretion resulting from either apical or basolateral application of IGF-1. Taken together, these results provided further evidence for a role of an IGF-1 autocrine/paracrine system in the retina both in terms of normal ocular physiology as well as in the progression of CNV. Future studies will be designed to elucidate the roles of reduced oxygen tension and retinal cytokines on HIF-1α expression and alternatives to HIF-1 in promoting VEGF and IGFBP-3 secretion in the RPE and their influence on the progression of CNV. Studies at the cellular level will provide important insights into the mechanisms underlying the pathologies observed in the animal models of CNV. This will lead to a better understanding of the pathogenesis of this disease and to better treatments for this leading cause of blindness. 

 Supported in part by National Institutes of Health Grant CA78887 (SAR), and Department of Defense Grant N6311601MD10004 to the Hollings Cancer Center (SAR).

 Submitted for publication June 5, 2003; revised November 21, 2003 and April 2, 2004; accepted April 13, 2004.

 Disclosure: M.G. Slomiany, None; S.A. Rosenzweig, None

 A portion of this work was presented at the 85^th Annual Meeting of the Endocrine Society, June 2002, San Francisco, CA, the 1^st GH-IGF Joint Symposium, October 2002, Boston, MA, and the Association for Research in Vision and Ophthalmology (ARVO), May 2003, Fort Lauderdale, FL.

 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.

 

Figure 1.

 

Time course of IGF-1–stimulated VEGF secretion by D407 cells. (A) Confluent D407 cells grown in 6-well plates were serum starved for 24 hours before treatment with 100 nM IGF-1 or carrier BSA alone. After 12 hours, conditioned medium was processed for immunoblot analysis. Blot shown is representative of three or more experiments. (B) Duplicate chambers of confluent D407 cells grown in confluent 6-well plates were serum starved for 24 hours before treatment with 100 nM IGF-1 or carrier BSA alone for incubations ranging from 1 to 24 hours. At each time point, conditioned medium was processed for immunoblot and densitometric quantification. Error bars: SD in secretion between duplicate wells. The densitometrically quantified immunoblot is representative of three experiments. Significant differences in VEGF secretion are noted (*P < 0.05).

Figure 2.

 

Time course of IGF-1–induced IGFBP-3 secretion by D407 cells. (A) Confluent D407 cells grown in 6-well plates were serum starved for 24 hours before treatment with 100 nM IGF-1, 1 μg/mL tunicamycin, or carrier BSA alone. After 12 hours, conditioned medium was processed for ligand blot analysis. Arrowhead indicates position of recombinant, nonglycosylated VEGF. The blot is representative of three experiments. (B) Duplicate chambers of confluent D407 cells grown in 6-well plates were serum starved for 24 hours before treatment with 100 nM IGF-1 or BSA as described in the legend to Figure 1 . Error bars: SD between duplicate wells. The densitometrically quantified ligand blot is representative of three experiments. Significant differences in IGFBP-3 secretion are noted (*P < 0.02).

Figure 3.

 

IGF-1 dose–dependent increase in VEGF secretion by D407 cells. Triplicate wells of confluent D407 cells grown in 6-well plates were serum starved for 24 hours, followed by treatment with increasing concentrations of IGF-1 ranging from 10 pM to 1 μM in fresh serum-free medium. After a 12-hour incubation, conditioned medium was subjected to immunoblot and densitometric quantification. Error bars: SD in secretion between triplicate wells of D407 cells. The densitometrically quantified immunoblot is representative of three experiments. Significant differences in VEGF secretion are noted (*P < 0.001, **P < 0.05).

Figure 4.

 

IGF-1 dose–dependent increase in IGFBP-3 secretion by D407 cells. Triplicate wells of confluent D407 cells grown in 6-well plates were serum starved for 24 hours, followed by treatment with increasing concentrations of IGF-1 ranging from 10 pM to 1 μM in fresh serum-free medium. After a 12-hour incubation, conditioned medium was subjected to ligand blot and densitometric quantification. Error bars: SD in secretion between triplicate wells of D407 cells. The densitometrically quantified immunoblot is representative of three experiments. Significant difference in IGFBP-3 secretion are noted (*P < 0.1, **P < 0.01).

Figure 5.

 

Effect of IGF-1 and CoCl₂ treatment on HIF-1α stability and activity in D407 cells. Confluent D407 cells were serum starved for 24 hours before treatment with 100 nM IGF-1 or 100 μM CoCl₂ in fresh serum-free medium. After a 4-hour incubation, cells were lysed and protein concentrations were determined. Lysates (100 μg) were solubilized in SDS sample buffer, reduced with DTT, resolved on a 10% acrylamide gel, transferred to nitrocellulose, and probed for HIF-1α (A) and HIF-1β (B). Bands were detected using HRP-conjugated secondary antibodies and the ECL reagent. Results shown are representative of three or more experiments. (C) Relative intensity of HIF-1α in relation to HIF-1β control for each treatment. (D) After transfection with p2.1 luciferase and pRL-SV40 renilla for 24 hours in serum-free medium, D407 cells were incubated for 18 hours with 100 nM IGF-1 or 100 μM CoCl₂. Lysates were processed for dual luciferase activity. Relative luciferase activity was calculated by dividing p2.1 luciferase by pRL-SV40 renilla activity. Error bars: SD between triplicate wells. The plot is representative of three or more experiments. Significant differences in HIF-1 reporter expression are noted (*P < 0.06, **P < 0.05).

Figure 6.

 

Dose-dependent effect of IGF-1 on HIF-1α protein expression in D407 cells. (A) Duplicate dishes of confluent D407 cells were serum starved for 24 hours before treatment with a battery of IGF-1 doses (10 pM–1 μM) in fresh serum-free medium. After 4 hours of incubation, cells were lysed, protein content determined, and lysates treated as described in the legend to Figure 5 and probed for HIF-1α and β-actin. The blot is representative of three experiments. (B) Relative intensity of HIF-1α in relation to β-actin control for each treatment. Error bars: SD between duplicate wells. Significant differences in HIF-1α protein expression are noted (*P < 0.05, **P < 0.01).

Figure 7.

 

Time course of IGF-1–induced HIF-1α protein expression in D407 cells. Duplicate dishes of confluent D407 cells were serum starved for 24 hours before treatment with 100 nM IGF-1 in fresh serum-free medium. At the times indicated, cells were lysed and protein contents were determined. (A) Lysates (100 μg) were solubilized in SDS sample buffer and analyzed as detailed in the legend to Figure 5 for HIF-1α content. The blot is representative of three experiments. (B) Relative intensity of HIF-1α in relation to β-actin control for each treatment. Error bars: SD between duplicate wells. Significant differences in HIF-1α protein expression are indicated (*P < 0.05, **P < 0.005).

Figure 8.

 

Microscopic analysis of IGF-1–induced HIF-1α and VEGF protein expression in D407 cells. Confluent D407 cells grown on coverslips were serum starved for 24 hours before treatment with 100 nM IGF-1 in fresh serum-free medium. After a 4-hour incubation, cells were fixed, blocked, permeabilized, and probed for HIF-1α and VEGF with Texas red and FITC conjugated secondary antibodies, respectively (A and B) or secondary antibodies alone (C). Cells were visualized by differential interference contrast (DIC). The images are representative of three or more experiments.

Figure 9.

 

Microscopic analysis of IGF-1–induced HIF-1α and IGFBP-3 protein expression in D407 cells. Confluent D407 cells grown on coverslips were serum starved for 24 hours before being treated as described in the legend to Figure 8 , with the exception of being probed for IGFBP-3. The images are representative of three or more experiments.

Figure 10.

 

Polarized secretion of VEGF by D407 cells. Confluent D407 cells grown on Transwell inserts were serum starved for 24 hours before treatment with 100 nM IGF-1 to either the upper or lower Transwell compartment. After a 12-hour incubation, conditioned medium from upper and lower compartments was subjected to immunoblot and densitometric quantification. Error bars: SD in secretion between the indicated number of Transwells. The densitometrically quantified immunoblot is representative of three or more experiments. Significant differences in apical versus basolateral VEGF secretion are noted (*P < 0.05, **P < 0.005), ***P < 0.0002).

Figure 11.

 

Polarized secretion of IGFBP-3 by D407 cells. Confluent D407 cells were grown and treated as described in the legend to Figure 11 . The densitometrically quantified ligand blot is representative of three or more experiments. Significant differences in apical versus basolateral IGFBP-3 secretion are shown (*P < 0.05, **P ≤ 0.005).

Figure 12.

 

Polarity of the IGF-1 receptor and Na⁺,K⁺-ATPase α-1 subunit expression. Confluent D407 cells grown on Transwell inserts were serum starved for 24 hours before being fixed, blocked, permeabilized, and probed with anti-IGF-1 receptor α subunit with Alexa Fluor 488 secondary and with anti-Na⁺,K⁺-ATPase α-1 with Alexa Fluor 546 secondary. Slide-mounted Transwell filters were visualized by confocal microscopy in Z-section at 630X magnification with 3X zoom. D407 cells probed for IGF-1 receptor α subunit (A) and Na⁺,K⁺-ATPase α-1 (B). The confocal images are representative of three experiments.

The authors thank Robert G. Gourdie, Jane Jourdan, and Ralph J. Barker for epifluorescence microscopy support, and Margaret M. Kelly of the Hollings Cancer Center Molecular Imaging Core for help with the confocal microscope.