Fibronectin Fragments Promote Human Retinal Endothelial Cell Adhesion and Proliferation and ERK Activation through α₅β₁ Integrin and PI 3-Kinase

Abstract

purpose. Extracellular matrix degradation is associated with neovascularization in diabetic retinas. Fibronectin fragments (Fn-fs) are generated during vascular remodeling. The effects of cellular fibronectin (Fn) and selected Fn-fs on adhesion, proliferation, and signal transduction in human retinal endothelial cells (HRECs) were characterized.

methods. Relative quantitative RT-PCR, flow cytometry, and immunocytochemistry determined integrin expression on HRECs. Adhesion was evaluated by coating plastic with Fn or Fn-fs of 45, 70, 110, or 120 kDa, and MTT conversion was used to measure proliferation and survival. Peptide inhibitors and blocking antibodies determined adhesive sites and integrins used for adhesion. Pharmacologic inhibitors and Western analyses were used to evaluate intracellular signaling.

results. HRECs produced significant levels of α₂, α₃, α₅, α_v, β₁, β₃, and β₅ integrin subunit mRNA. Flow cytometry of surface integrin expression revealed high levels of α₃, α₅, and β₁ and lower levels of α₁, α_v, β₃, and β₅. These results were confirmed by immunocytochemistry. For adhesion to Fn and Fn-fs. the α₅β₁ integrin was essential. Pharmacologic inhibitors of PI 3-kinase blocked adhesion to Fn and Fn-fs, whereas the mitogen-activated protein (MAP) kinase kinase (MEK) inhibitor PD98059 blocked phosphorylation. The 110- and 120-kDa Fn-fs showed a concentration-dependent increase in proliferation, whereas 500 ng of the 70 kDa Fn-f-induced proliferation. Addition of III1-C, a matrix assembly domain, increased the proliferative effect of these Fn-fs.

conclusions. Fn and its Fn-fs modulate HREC adhesion and proliferation through signal-transduction pathways involving coupling of the α₅β₁ integrin through PI 3-kinase. Mitogenic signals for endothelial cells from degraded extracellular matrix may contribute to the development of diabetic retinopathy.

Diabetic retinopathy is the leading cause of adult blindness in the United States. ¹ Proliferative diabetic retinopathy is associated with aberrant endothelial cell proliferation, neovascularization, vitreous hemorrhage, and traction detachment. Increased amounts of aberrant extracellular matrix (ECM) observed in retinal vessels of diabetic patients may contribute to the endothelial cell dysfunction that is characteristic of this disease. Retinal vessels of diabetic patients contain increased amounts of fibronectin (Fn) ^{2 3 4} and in vitro exposure of retinal endothelial cells to high glucose levels increases secretion of Fn. ⁵ Retinas of patients with diabetic retinopathy are positive for the splice variant ED (extra domain)-B Fn, a marker of angiogenesis, ⁶ suggesting a role for locally synthesized Fn in neovascularization. 

Fn is a multifunctional glycoprotein found in plasma and ECM that regulates cellular adhesion, migration, oncogenic transformation, wound healing, and hemostasis. ⁷ It exists as a 450-kDa dimer with subunits joined by a pair of disulfide bonds located near the carboxyl termini. The diverse biological activities attributed to Fn have been localized to specific regions of the molecule (Fig. 1) . These regions were identified by their binding affinity for specific molecules such as fibrin, factor XIIIa, gelatin-collagen, and heparin. In addition to binding domains, functional domains, such as the matrix assembly domain located near the amino terminus, and the cell-binding domain, which spans type III repeats 8 to 10, have also been extensively characterized. ^{8 9}  

Fn is highly susceptible to proteolysis, often generating fragments (Fn-fs) with greater or different biological activity than the parent protein. ^{10 11 12} Degradation of Fn occurs in the vicinity of cells undergoing neoplastic transformation, possibly due to expression of proteases by cancer cells. ¹³ We have demonstrated that latent matrix metalloproteinase (MMP)-2, secreted by human retinal endothelial cells (HRECs), exists as a complex with the 30-kDa N-terminal portion of Fn and that binding of this fragment to latent MMP-2 inhibits the activation of this latent protease. ¹⁴ This suggests the active involvement of Fn and Fn-fs in regulating proteolytic enzymes in diabetic neovascularization. We have shown that the 120-kDa N-terminal Fn-f is a potent mitogen for HRECs, inducing greater proliferation than fibroblast growth factor (FGF)-2. ¹⁴ Fn-fs induce cell proliferation, cause release of cytokines from vascular cells, ¹⁵ and modulate adhesion, spreading, and migration of vascular endothelial cells. ¹⁴ These studies suggest that the Fn and Fn-fs present in microvascular basement membranes may modulate neovascularization in the diabetic retina. 

Endothelial cells are anchorage dependent and require both adhesion to the ECM and growth factor stimulation for survival, growth, and differentiation. Many of the adhesive contacts are mediated through integrins, heterodimeric receptor molecules formed by ion-dependent, noncovalent binding of one α and one β transmembrane glycoprotein subunit. ¹⁶ The extracellular domain of each subunit binds to several ligands including the ECM proteins Fn, vitronectin (Vn), and collagen. 

Similar to growth factor activation, engagement of integrins initiates kinase cascades that activate multiple growth-associated kinases including focal adhesion kinase (FAK), phosphoinositide 3-OH kinase (PI 3-kinase), src, and raf, and the mitogen-activated protein (MAP) kinase kinase (MEK). ¹⁷ Conversely, lack of attachment to the ECM through integrins induces anoikis (cell death by detachment). ¹⁶  

In this study, we determined the expression of integrin subunits on HRECs by relative quantitative RT-PCR, flow cytometry, and immunocytochemistry. We compared the effect of Fn and key Fn-fs on cell adhesion and cellular signaling pathways associated with adhesion, proliferation, and cell survival. Our data indicate that the Fn-fs examined bind to the same integrin (α₅β₁) as the parent protein and transduce signals to activate extracellular signal-regulated kinase (ERK) through a PI 3-kinase-dependent pathway. These data suggest, that the interaction of HRECs with Fn and its proteolytic fragments initiate common intracellular signaling events that contribute to adhesion and proliferation. 

Methods

Materials and Reagents

Cell culture reagents were obtained from Invitrogen (Carlsbad, CA), except for insulin-transferrin-selenium and endothelial cell growth supplement, which were purchased from Sigma-Aldrich (St. Louis, MO). Wortmannin, LY294002, U0126, and PD98059 were purchased from Calbiochem (La Jolla, CA); purified cellular Fn and 120-kDa Fn-f from Invitrogen; purified 110-kDa Fn-f from Upstate Biotechnology, Inc. (Lake Placid, NY); and purified 70- and 45-kDa Fn-fs from Sigma-Aldrich. Antiserum for detection of phospho-ERK (E10 monoclonal) was from Cell Signaling Technology (Beverly, MA) and total ERK polyclonal antibody was from Sigma-Aldrich. Horseradish peroxidase (HRP)-conjugated anti-rabbit, HRP-conjugated anti-mouse, and enhanced chemiluminescence (ECL) reagents were purchased from Amersham-Pharmacia Biotech (Piscataway, NJ). Azide-free blocking antibodies against integrins β₁ (PC4C10), β₃ (B3A), and α₅ (P1D6) were from Invitrogen. Dimeric blocking antibodies α_vβ₃ and α₅β₁ (LM609 and BMA5, respectively) and a second β₁ blocking antibody (6S6) were from Chemicon International (Temecula, CA). For flow cytometry analysis, anti-integrin α₁ and anti-major histocompatibility complex (MHC) W6.32 were obtained from American Type Culture Collection (ATCC; Manassas, VA), and anti-integrin α₂ (PIE6), anti-integrin α₃ (PIB5), anti-integrin α₅ (P1D6), anti-integrin α_v (VNR139), and anti-integrin β₄ (3E1) were purchased from Invitrogen. Anti-integrin α₆ was from Immunotech (Santa Clara, CA). A second anti-β₄ from Beckman-Coulter (Hialeah, FL) was also tested with flow cytometry. Anti-mouse, anti-rabbit, or anti-goat fluorescein isothiocyanate (FITC)-conjugated secondary antibodies were from Sigma-Aldrich. Recombinant human III1-C Fn-f was a kind gift from Alex Morla (University of Chicago, Chicago, IL) and additional III1-C was obtained from Sigma-Aldrich. 

Isolation and Culture of HRECs

Human eyes were obtained from the National Disease Resource Interchange (Philadelphia, PA) within 36 hours of death (n = 3 donors). Human retinal endothelial cells were prepared and maintained as previously described, ¹⁸ and cells between passages 3 and 5 were used for the present study. The identity of HRECs is typically validated by demonstrating endothelial cell incorporation of fluorescence-labeled acetylated LDL. ¹⁸  

Relative Quantitative RT-PCR

Relative quantitative RT-PCR was performed as previously described. ¹⁹ Total RNA was isolated with extraction reagent (TRIzol; GibcoBRL, Grand Island, NY), according to the manufacturer’s instructions. Reverse transcription was performed with 2 μg of RNA and reverse transcriptase (Superscript MMLU RNase H⁻; Invitrogen), according to the manufacturer’s instructions. PCR was performed on the cDNA with previously described primers ²⁰ or glyceraldehyde-phosphate dehydrogenase (GAPDH) as an internal control. Data were normalized to GAPDH. 

Flow Cytometry

A nonenzymatic dissociation was used to remove HRECs from culture dishes, to preserve the integrity of the cell surface molecules. Cells were washed four times with Ca²⁺/Mg²⁺-free PBS and were then incubated for 15 minutes in 2 mM EDTA in Ca ²⁺/Mg²⁺-free PBS at 37°C and washed four times with PBS. Cells were incubated for 30 minutes on ice with individual integrin antibody diluted in PBS containing 1% bovine serum albumin. Cells were washed and then incubated with the appropriate fluorescein-labeled secondary antibody for 30 minutes on ice. Nonimmune species and isotype-matched antibodies were used as negative controls. Appropriate FITC-conjugated secondary antibodies were added and incubated. After a final wash, cells were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer. Finally, samples were analyzed for integrin expression in flow cytometry performed at the University of Florida ICBR Core facility (FACScan; BD Biosciences, Lincoln Park, NJ) and quantified by plotting the relative fluorescence units as histograms over control readings. 

Immunocytochemistry

HRECs were grown to 60% confluence (10⁴ cells/cm²) in eight-well chamber slides (Laboratory-Tek, Naperville, IL) coated with 0.5 μg/mL Fn (Sigma-Aldrich). The cells were washed with PBS, fixed with 4% paraformaldehyde in 0.1 M phosphate buffer, washed with PBS, and incubated with 10% normal blocking serum in PBS for 30 minutes to suppress nonspecific binding of IgG. The blocking serum used was from the identical species in which secondary antibody was raised. Cells were then incubated with the primary antibody for 1 hour at 37°C and with the appropriate FITC-conjugated secondary antibody, diluted with 1.5% blocking serum in PBS for 30 minutes at 37°C. Nonimmune species and isotype-matched antibodies were used as negative controls. The cells were washed, mounted in glycerol/PBS, and photographed with a fluorescence microscope (Axiophot; Carl Zeiss; Thornwood, NY). 

Evaluation of Fn-fs’ Purity and Binding of Fn and Fn-fs to Tissue Culture Plastic

Fn and Fn-fs were obtained from three separate suppliers, because the fragments were not available from a single supplier. However, all fragments were purified by the manufacturer, by high-performance liquid chromatography, and were then reconstituted according to manufacture’s instructions and stored at −80°C in single-use aliquots. With each purchase of fragments we analyzed each fragment by SDS-PAGE and confirmed that each sample consisted of a single band by silver staining analysis. ²¹ Tissue culture dishes were coated overnight (4°C) with Fn or the Fn-fs (0.1–10 μg/mL in PBS). Parallel plates were also coated with 100 μg/mL of each Fn-f and washed twice with PBS. Protein attachment was detected by the bicinchoninic acid (BCA) method (Pierce, Rockford, IL), to ensure that each Fn-f bound to the plastic culture wells with the same avidity. Plates for adhesion and proliferation assays were blocked for 1 hour with 2% BSA in PBS before use. 

Adhesion of HRECs

HRECs were incubated for 2 hours in 96-well culture plates coated overnight with cellular Fn and the 45-, 70-, 110-, or 120-kDa Fn-fs diluted in PBS, as described earlier. HRECs were serum-starved for 24 hours, dissociated in 0.5× trypsin-EDTA, and washed two times in 1% BSA in 1:1 DMEM/F-12. Cells were incubated for 1 hour in 2% BSA in Dulbecco’s modified Eagle’s medium (DMEM)/F-12 (1:1) before assay and then plated (5 × 10⁴/well). They were allowed to adhere for 2 hours at 37°C, 5% CO₂ after preincubation for 10 minutes with RGDS, RGES, or III1-C peptide. Nonadherent cells were removed by washing twice with PBS. Concentrations of Fn-fs were chosen at which a dose-dependent effect was observed, rather than selecting the optimal concentration, to increase the likelihood of observing the effects of the added peptides. HREC proliferation was measured by a modified MTT assay, which measures the ability of live cells to use thiozolyl blue and convert it to dark blue formazan. Complete growth medium containing 0.5 mg/mL MTT was added, and MTT conversion was measured using a microplate spectrophotometer (absorbance, [A]₅₅₀–A₆₉₀). For antibody-blocking experiments, HRECs were processed as described earlier and preincubated with antibody to β₁, β₃, β₅, α₅β₁, or α_vβ₃ diluted to 10 μg/mL (except for α_vΒ₃, which was used at 100 μg/mL) in PBS containing 2% BSA for 10 minutes before being plated. Control cells were incubated with nonimmune and isotype matched species antibodies. In experiments in which kinase inhibitors were used, cells were incubated for 30 minutes in the presence of the inhibitors, and control cells were incubated in 0.01% dimethyl sulfoxide (DMSO) for experiments with wortmannin, LY294002, U0126, and PD98059. 

Kinase Activation Assays

Cells were serum-starved for 24 hours (1:1 DMEM/F-12), dissociated as described earlier, allowed to recover for 1 hour, and incubated in the presence of inhibitors (PD98059, LY294002, wortmannin or 0.01% DMSO as a control) for 30 minutes before stimulation. Cells were plated on Fn- and Fn-f-coated cell culture dishes (as described for adhesion assays) and allowed to adhere for 2 hours. Nonadherent cells were collected by centrifugation (800g) at 4°C. The assay was terminated with the addition of lysis buffer containing protease and phosphatase inhibitors (1% Triton X-100, 10 μg/mL aprotinin, 20 μM leupeptin, 1 μM E-64, 1 mM NaF, 200 μM sodium pervanadate, 1 mM dithiothreitol, 5 mM EDTA, and 25 mM Tris [pH 6.8]) and frozen at −20°C until assayed. Electrophoresis was performed according to the method described by Laemmli. ²² Proteins were fractionated on 10% polyacrylamide gels. Parallel gels were stained with Coomassie blue to verify loading, sample integrity, and protein separation. Proteins were transferred for 2 hours (50 V) from acrylamide gels to polyvinyl difluoride (PVDF) membranes for immunodetection. Membranes were blocked for 1 hour with 5% nonfat powdered milk in TTBS (25 mM Tris-HCl, 150 mM NaCl, 0.05% Tween 20 [pH 7.3]) and probed at room temperature (phospho-ERK, 1:1000 in TTBS, 6 hours or total ERK 1:40,000 in TTBS for 1 hour). HRP-conjugated anti-rabbit or anti-mouse was used for detection at a dilution of 1:1000. Secondary antibody incubations were for 1 hour, and membranes were washed three times in TBS between antibody incubations. Peroxidase activity was detected using chemiluminescence with 1- to 5-minute exposure times. Densitometry was performed on the film (Scion Image, Frederick, MD). Average background density was subtracted, and optical densities were plotted on computer (Origin; Microcal, Northampton, MA). 

HREC Proliferation Measured by MTT Conversion

The effect of Fn-fs on cell proliferation was measured by MTT conversion. Proliferation was measured using 10⁴ cells/well seeded in 96-well microtiter plates coated with proteins as indicated above. HRECs were incubated for 96 hours in DMEM/F-12 (1:1) supplemented with insulin-transferrin-selenium and 1% endothelial cell growth supplement. Next, 10 μL of MTT (5 mg/mL) cells was added to each well (0.5 mg/mL, final concentration). Plates were returned to the incubator after addition of MTT. The assay was terminated by aspiration of the medium with beveled needle, and MTT was solubilized in 100 μL isopropanol. MTT conversion was measured with a microplate spectrophotometer (A₅₅₀–A₆₉₀). 

Statistical Analysis

Data were analyzed by ANOVA on computer (Origin Microcal). Each concentration was compared with control levels and the difference considered significant at P < 0.05. 

Results

A schematic representation of Fn and the regions contained in each of the five Fn-f examined is shown in Figure 1 . The 70-kDa Fn-f contains the amino terminal fibrin-heparin and the collagen-binding domain. The 45-kDa Fn-f is contained within the 70-kDa fragment and includes the collagen-binding domain and the matrix assembly domain. The 120-kDa Fn-f consists of the first 11 type III repeats and contains the cell-binding domain. The 110-kDa Fn-f comprises the first nine type III repeats and contains a portion of the cell-binding domain. III1-C Fn-f is a fragment of the first type III repeat and contains a matrix assembly domain. 

Integrin Expression of HRECs

Relative RT-PCR and flow cytometry were used to determine which integrins were expressed by HRECs and available to bind Fn. HRECs produced significant levels of α₂, α₃, α₅, α_v, β₁, β₃, and β₅ integrin subunit mRNA. Table 1 details the mRNA expression for integrin subunits in HRECs, normalized to GAPDH. The α_v and β₃ integrin mRNAs were expressed at the highest levels, consistent with the role of this integrin dimer in endothelial cell adhesion to various ECM molecules in other vascular beds. Flow cytometry was then used to determine whether mRNA levels were representative of the surface integrin expression. Figure 2 details the cumulative results of flow cytometry analysis of surface integrin expression by HRECs. Integrins containing subunits α₃ or α₅ were present at high levels on the cell surface. Integrin subunit α_v was also expressed, as was α₁, although to a lesser degree than either α₃ or α₅. The β₁ subunit was also highly expressed, and to a lesser degree β₃ and β₅. All the integrins detected by RT-PCR were also detected by flow cytometry, except α₁ mRNA that was not detected by RT-PCR but was detected by flow cytometry and immunocytochemistry. The immunocytochemical localization of the α₅ and β₁ integrin subunits on HRECs is shown in Figures 3A and 3B , respectively. Integrin subunits α₃ and β₁ were also immunolocalized and reacted with similar intensity, whereas integrin subunits α₁, α_v, β₃, and β₅ reacted with less intensity (not shown). Negative controls did not stain (not shown). 

HREC Adhesion to Fn and Fn-fs

Adhesion was supported on plates coated with cellular (c)-Fn and the 70-, 110-, and 120-kDa Fn-fs (Fig. 4) . The 110-kDa fragment supported the greatest adhesion up to 10 μg/mL, whereas the 70- and 120-kDa Fn-fs and c-Fn supported similar levels of adhesion but less than the 110-kDa Fn-f. As expected, the 70-kDa fragment was not as potent at promoting Fn adhesion as were the Fn-fs containing the cell-binding domain. Preincubation with the matrix assembly-promoting III1-C peptide ²³ potentiated adhesion to all Fn-fs and to c-Fn (Fig. 4) . The 5-kDa fragment did not support adhesion, even in the presence of the III1-C peptide, and it was not capable of blocking adhesion to the other fragments tested or to c-Fn (not shown). In addition, adhesion to all substrates was blocked by preincubation with the RGDS peptide, which uncouples RGD-dependent integrin binding, but not the control peptide RGES (Fig. 4) . Adhesion to the 110-kDa Fn-f was only partially blocked by RGDS, indicating either increased affinity of this fragment for the integrin or the presence of an alternate cell-binding domain or integrin partner. Adhesion to the 70-kDa fragment, which does not contain the cell-binding domain, was also blocked by RGDS. It has been suggested, that this is due to cross-competition between the RGD sequence and a “second” integrin-binding domain on the N-terminal fragment. When this RGD-binding domain of the 70-kDa Fn-f is occupied, the cells are unable to bind through the second integrin-binding site in the N-terminal fragment. ²⁴ These data indicate that HRECs adhere to Fn and Fn-fs through integrins and require the presence of the RGD sequence, regardless of whether binding occurs directly at the cell-binding domain. 

Inhibition of HREC Adhesion to Fn and Fn-fs by Integrin Antibodies

Expression data were used to select candidate integrins for antibody blocking experiments. We first tested blocking antibodies against the β subunits that were expressed (Fig. 5) . Anti-β₁ blocked adhesion to the 70- (Fig. 5A) , 110- (Fig. 5B) , and 120-kDa (Fig. 5C) Fn-fs and to c-Fn (Fig. 5D) . Blocking antibodies against β₃ and β₅ integrins did not block adhesion to any of the proteins tested. 

The β₁ subunit is capable of forming dimers with several α-subunits. The α₅β₁ dimer is relatively selective for Fn, although Fn can also bind the vitronectin-selective α_vβ₃ integrin under certain experimental conditions. Anti-α_vβ₁ ²⁵ did not block adhesion to any of the fragments tested (data not shown). Preincubation with anti-α₅β₁ (10 μg/mL), blocked adhesion to the 70-, 110-, and 120-kDa Fn-fs and Fn-f (Fig. 6) . Although some promiscuity exists in integrin-ECM interactions, preincubation with anti-α_vβ₃ (up to100 μg/mL) did not alter adhesion to any of the fragments tested (Fig. 6) . 

PI 3-Kinase Activity in Adhesion

We next examined the intracellular signaling pathways activated by integrin engagement to promote adhesion. Cells were pretreated with wortmannin or LY294002 (inhibitors of PI 3-kinase) and allowed to adhere to each Fn-f or c-Fn for 2 hours. Wortmannin and LY294002 partially blocked adhesion to all fragments, indicating a role for PI 3-kinase in HREC adhesion (Fig. 7) . Wortmannin (100 nM) was more potent than LY294002 (50 μM) in inhibiting adhesion, but considerable variability existed between donors in the degree of observed inhibition with these compounds. When cells were pretreated with U0126 and PD98059 inhibitors of MEK/ERK (Fig. 8) , adhesion to the 110- and 120-kDa fragments was partially blocked by both inhibitors (Figs. 8B 8C) . U0126 (1 μM) was more potent than PD98059 (50 μM) in inhibiting adhesion. U0126 blocks both active and inactive MEK/ERK, suggesting that active (U0126-inhibited but not PD98059-inhibited) MEK/ERK activity may be necessary for adhesion to the 110- and 120-kDa Fn-fs. These data suggest a role for ERK in adhesion that may be mediated solely by the cell-binding domain. 

ERK Activation by Adhesion to Fn and Fn-fs

Adhesion to cellular Fn or the 70-, 110-, and 120-kDa Fn-fs increased activation of ERK, whereas the 45-kDa Fn-f modestly increased activation when compared with untreated cells (Fig. 9A) . Quantification of the p42 ERK isoform is shown in Figure 9B . Inhibitors of PI 3-kinase were next tested for their ability to inhibit ERK activation. Both wortmannin and LY294002 reduced activation of ERK, as did the MEK inhibitor PD98059 (Fig. 9B) . The 44-kDa ERK isoform was present at lower levels in HRECs and was below the limit for linear detection by this method, which may explain the differential effects on the 44- and 42-kDa isoforms. These data indicate that ERK activation by Fn and Fn-fs occurs through a pathway that is dependent on both PI 3-kinase and MEK. 

HREC Proliferation Measured by MTT Conversion

Cells plated on Fn- and Fn-f-coated dishes (0–10 μg/mL) in serum-free medium showed concentration-dependent proliferation in response to all but the 45-kDa Fn-f. We have demonstrated that the 45-kDa Fn-f inhibits proliferation. ¹⁴ Exposure of HRECs to the 110- and 120-kDa Fn-f resulted in a concentration-dependent increase in the number of cells (Figs. 10B 10D , respectively) The 70-kDa Fn-f, at low concentrations, had no effect on proliferation, but at concentrations in excess of 500 ng, the 70-kDa Fn-f induced proliferation (Fig. 10A) . In contrast, exposure of HRECs to c-Fn had no effect on proliferation (Fig. 10D) . Similar to the adhesion process, stimulation of proliferation on different Fn-fs and c-Fn was significantly potentiated by the III1-C peptide containing a matrix assembly site (Fig. 10) . The III1-C fragment alone did not have an effect on MTT conversion (data not shown). 

Discussion

Increased serum levels of Fn occur in diabetes, and increased amounts of Fn are detected in basement membranes of eyes in diabetes. Normalization of blood sugar levels corrects the elevated Fn and other serum abnormalities in individuals with diabetic retinopathy and delays or prevent development of the disease in individuals with no diabetic retinopathy. However, pancreatic transplantation and subsequent correction of hyperglycemia does not stabilize or reverse the retinopathy once it is established in an individual. Rather, retinopathy continues to progress in many individuals who have complete correction of diabetes after pancreatic transplantation. A possible explanation for this observation is that, although the metabolic milieu rapidly changes, endothelial cells continue to be exposed to the altered ECM that contains increased amounts of proteins such as Fn and Fn-f. In the current study, HRECs in culture expressed an increased amount of Fn when exposed to high-glucose conditions (30 mM glucose), ²⁶ and HRECs generated Fn-fs in vitro that modulated the activity of MMP-2 and affected proliferation and migration. 

Endothelial cells responded to increased Fn by increasing expression of an integrin subtype that binds Fn (e.g., α₅β₁). ²⁷ Overexpression of the α₅β₁ Fn receptor reduces cell migration. ²⁸ Thus, increased synthesis and availability of selected integrins may mediate the proliferative effects of matrix, further confirming that integrin binding with ECM components activates intracellular pathways implicated in growth regulation. ^{28 29} Depending on the context, integrins can transmit signals that permit or inhibit growth. ³⁰ In the current study, relative quantitative RT-PCR, flow cytometry and immunocytochemistry detected multiple integrins expressed by HRECs. However, there were disparities between the mRNA levels and the cell surface expression detected by flow cytometry. Although this may be a result of differential antibody affinity, the data raise the intriguing possibility that integrin levels in HRECs are subject to posttranscriptional regulation. 

In the adhesion experiments, preincubation with the matrix assembly promoting III1-C peptide potentiated adhesion to all fragments, and antibodies to β₁ blocked adhesion to Fn and all Fn-fs except at the highest concentration of the 70-kDa Fn-f tested. However, although the pattern of adhesion was identical in these experiments, the magnitude of adhesion observed was different between experiments (Figs. 4 5) . This suggests the possibility of donor-to-donor variation in experiments using primary cell cultures. However, results were very similar, but not identical, between donors. 

In the current study, the β₁ integrin-blocking antibodies did not completely inhibit adhesion and were unable to block cell binding to the highest concentration of 70-kDa Fn-f tested. The 70-kDa fragment has been characterized as a matrix assembly domain. We observed this effect with two sources of antibody (Invitrogen and Chemicon) and conclude that the 70-kDa Fn binds to sites other than those that are blocked by the antibodies. In contrast, the antibody raised against the α₅β₁ dimer, was a potent inhibitor of adhesion to all Fn-fs tested, including the 70-kDa Fn-f. It is possible that this is again a result of different antibody affinities; however, it may also result from β₁-independent adhesion in the absence of functional β₁ subunits. 

Fn-fs containing the RGD cell-binding domain induced proliferation that was comparable to the 120-kDa or greater than the 110-kDa Fn-f. In addition, the 70-kDa Fn-f, which does not contain the cell-binding RGD sequence, also induced HREC proliferation. Because the 45-kDa fragment is contained within the 70-kDa Fn-f and the 45-kDa Fn-f does not support proliferation or adhesion, the adhesive and mitogenic portion of the 70-kDa protein must reside within the first 30 kDa of the N terminus. This 30-kDa N-terminal fragment is generated by HRECs in culture and binds to MMP-2. ²¹  

Ligation of integrins activates tyrosine kinases and small guanosine triphosphatases (GTPases) necessary for the reorganization of actin required for cell spreading. ³¹ This is followed by Rho-dependent activation of cell contractility, resulting in formation of actin stress fibers, clustering of α₅β₁ integrin and assembly of mature focal adhesions. 

A striking correlation exists between the actin cytoskeleton and Fn matrix organization, suggesting that the Fn matrix may be a potential modulator of actin organization functioning to influence cell signaling and growth. The assembly of focal adhesions is associated with the reorganization of Fn into fibrils. ^{32 33} Exogenous full length Fn, as well as its 70-kDa amino terminal region of Fn, colocalize with Fn fibers. The interactions between the amino terminal region of Fn and the cell surface is the initial step in the assembly of exogenous Fn into extracellular matrix and is one of the intermolecular homophilic binding events critical for Fn polymerization. ^{25 34 35} These data suggest that binding of Fn amino terminus to endothelial cells has important cytoarchitectural as well as functional consequences and that there is an intimate relationship between Fn matrix assembly and cells growth control. This Fn matrix assembly requires the activity of the integrins, α₅β₁. Fn matrix assembly also depends on self-association sites within Fn, in addition to the N-terminal 70-kDa region. The III1-C fragment is also thought to be particularly important for the proper alignment of Fn molecules during matrix assembly. ^{36 37 38 39 40}  

III1-C was also found to induce spontaneous in vitro disulfide cross-linking of Fn, to increase binding of cells to Fn and to enhance matrix assembly. ⁴¹ Herein, we demonstrate that in HRECs, III1-C alone had no effect but when added to Fn and Fn-fs of 110, 120, and 70 kDa, it increased proliferation, as measured by MTT conversion. Cells can adhere and spread on III1-C, and both integrins and cell surface proteoglycans mediate this adherence. As is the case with most of the Fn-fs examined, the biological effect of III1-C is dependent on the cell type being examined, the concentration of the fragment, the presence of other Fn-fs, whether the fragment is used for coating of the culture dish or is added in solution to already adherent cells. 

Previous studies have demonstrated that treatment of cells with III1-C inhibits lysophosphatidic acid-mediated actin organization and tyrosine phosphorylation. ⁴² The ability of III1-C to affect cytoskeletal function has been attributed to its ability to either disrupt preexisting Fn matrices or to stimulate increased Fn matrix deposition, ⁴³ depending on its concentration. The III1-C fragment can partition to caveolin-enriched microdomains and allow signals from the ECM to be transmitted to the interior of the cell to modulate growth and contractility, and thus it is not surprising that III1-C has been shown to modulate such complex processes as cell proliferation, angiogenesis, and tumor metastasis. 

The ability of each Fn-f to induce proliferation is correlated with its ability to support adhesion. Adhesion appears to be mediated by the same integrin (α₅β₁) for each Fn-f through a PI 3-kinase-dependent pathway. Integrin activation by Fn and Fn-fs leads to activation of the proliferation-associated kinase ERK through a pathway that is dependent on PI 3-kinase. Activation of both pathways are required for cell survival; however, although each fragment activates both pathways to a similar degree, there are differences in actin distribution in cells adherent to the fragments (Grant MB, unpublished observations, 2001). Differences between Fn-fs may exist in activation of other pathways, such as those coupled to paxillin phosphorylation ²⁴ that regulate cytoskeletal dynamics. 

ERK activation by Fn can occur through multiple pathways. Association of ECM components with the β integrin subunit activates a signal transduction cascade, resulting in autophosphorylation of FAK on Tyr³⁹⁷, which creates a binding site for Src-family kinases. ^{44 45} Src phosphorylates multiple constituents of the focal adhesion complex, including the docking protein p130^CAS and FAK itself (e.g., Tyr⁹²⁵). ⁴⁵ Src-dependent phosphorylation of FAK at Tyr⁹²⁵ creates an SH2 docking site for the recruitment of Grb2 and Sos, thereby linking integrins to the ras/raf/ERK cascade. ⁴⁵  

FAK-dependent activation of PI 3-kinase may also require Src kinase activity. ⁴⁶ The alternate pathway for activation of ERK by integrin engagement involves coupling of the α subunit to activation and phosphorylation of Shc. ^{30 47} Shc phosphorylation creates an SH2 binding site for Grb-Sos and links integrins to the ras/raf/ERK pathway in a manner that is independent of FAK. ^{47 48 49 50 51} Shc activation is required for integrin-stimulated proliferation and may cooperate with sustained ERK activation by FAK to cause cell cycle progression. ^{47 51 52} Therefore, the Fn-fs may modulate ERK signaling through different pathways that converge upstream of ERK. For example, the 70-kDa fragment increases FAK phosphorylation (∼70% of the phosphorylation observed with c-Fn) but does not induce paxillin phosphorylation. ²⁴ This indicates either incomplete activation of FAK or differential coupling of the integrin to the transduction machinery inside the cell. Future experiments will be designed to determine the divergent pathways activated by Fn-fs that contain the cell-binding domain compared with Fn-fs that do not contain the RGD sequence. 

Earlier studies have shown that cell adhesion to III1-C results in robust ERK1/2 activation and that this effect is blocked by integrin-blocking antibodies. ^{35 53} Our observations support these findings and suggest a possible involvement of Fn assembly as a prerequisite for cellular adhesion and proliferation. 

In summary, Fn and its proteolytic fragments modulate HREC adhesion and proliferation through similar signal-transduction pathways involving coupling of the a₅β₁ integrin though PI 3-kinase. Thus, signals from the degraded extracellular matrix provide mitogenic signals for HRECs, which may contribute to the development of diabetic retinopathy. 

 Supported by National Eye Institute Grants EY12601 and EY07739.

 Submitted for publication July 31, 2002; revised October 14 and October 31, 2002; accepted November 8, 2002.

 Disclosure: S.H. Wilson, None; A.V. Ljubimov, None; A.O. Morla, None; S. Caballero, None; L.C. Shaw, None; P.E. Spoerri, None; R.W. Tarnuzzer, None; M.B. Grant, None

 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.

 

Fn monomer modular organization and location of fragments analyzed. ED, extra domains (alternatively splice regions or sites); CS, cell-specific domains.

Table 1.

 

Relative Quantitative RT-PCR Detection of Integrin mRNA Levels

Subunit	mRNA Level
α2	0.459
α3	0.457
α5	0.142
αv	0.839
β1	0.392
β3	0.933
β5	0.484

Figure 2.

 

Surface integrin expression in HRECs. Flow cytometry was used to determine the surface expression of integrin subunits. (A) A representative profile for β₁. (B) Quantification of three separate experiments.

Figure 3.

 

HRECs stained positive for (A) α₅ and (B) β₁ subunits by immunofluorescence. Note the punctate staining. Scale bar, 20 μm.

Figure 4.

 

Adhesion to the 70- (A), 110- (B), and 120-kDa (C) Fn-fs and c-Fn (D). Ninety-six-well cell culture dishes were coated, and cells were allowed to adhere for 2 hours at 37°C, 5% CO₂ after preincubation for 10 minutes with RGDS, RGES, or the III1-C peptide (10 μg/mL). Cells were washed with PBS, and adherent cells were quantified by MTT conversion in complete growth medium. For this assay, concentrations of Fn-fs were chosen for which a dose dependence had been observed, rather than selecting optimal concentrations. The cells adhered to all the fragments, and the higher the concentration of the fragment the greater the adherence. Note the enhancement by the III1-C and the expected inhibition by RGDS and absence of inhibition by the control peptide RGES. Data represent the mean ± SD from a single donor (n = 4; *P < 0.05; ANOVA, when compared with control data). Similar results were obtained in replicate experiments with primary cultures obtained from different donors.

Figure 5.

 

Integrin antibody β₁, but not β₃ or β₅, blocked adhesion to the 70- (A), 110- (B), and 120-kDa (C) Fn-fs and c-Fn (D). Cells were pretreated for 10 minutes with blocking antibodies (10 μg/mL) and plated on 96-well coated culture dishes. After 2 hours at 37°C, 5% CO₂, cells were washed with PBS, and adhesion was quantified by MTT conversion in complete growth medium. Data represent the mean ± SD, from a single donor (n = 4; *P < 0.05; ANOVA, when compared with control data). Experiments were repeated with similar results with primary cultures obtained from different donors and different sources of antibody.

Figure 6.

 

Adhesion to the 70- (A), 110- (B), and 120-kDa (C) Fn-fs and c-Fn (D) was blocked by antibodies against the α₅β₁ but not the α_vβ₃ integrin. Cells were pretreated for 10 minutes with blocking antibodies (10 μg/mL for α₅β₁, 100 μg/mL for α_vβ₃) and plated on 96-well coated culture dishes. After 2 hours at 37°C, 5% CO₂, cells were washed with PBS, and adhesion was quantified by MTT conversion in complete growth medium. Data represent the mean ± SD (n = 4, *P < 0.05; ANOVA, when compared with control data). Similar results were obtained when cells were incubated in the presence of 10 μg/mL α_vβ₃ antibody.

Figure 7.

 

Adhesion to the 70- (A), 110- (B), and 120-kDa (C) Fn-fs and c-Fn (D) was partially blocked by the PI 3-kinase inhibitors wortmannin and LY294002. Cells were pretreated for 30 minutes with 100 nM wortmannin or 50 μM LY294002 and allowed to adhere for 2 hours. Cells were then washed twice with PBS, and adhesion was quantified by MTT conversion in complete growth medium. Data represent the mean ± SE of results in four replicate experiments using four different donors (D, *P < 0.05; ANOVA, when compared with control data).

Figure 8.

 

Adhesion to the 110- (B) and 120-kDa (C) fragments was partially blocked by the MEK inhibitors U0126 and PD98059, but adhesion to the 70-kDa fragment and c-Fn was not blocked. Cells were pretreated for 30 minutes with each inhibitor and allowed to adhere for 2 hours. Nonadherent cells were washed off with two changes of PBS, and adhesion was quantified by MTT conversion in complete growth media. Data represent the mean ± SE of four replicate experiments using four different donors (*P < 0.05; ANOVA, when compared with control data).

Figure 9.

 

ERK activation by cellular adhesion to Fn and Fn-fs. HRECs were treated for 30 minutes with 0.01% DMSO, PD98059 (50 μM), LY294002 (50 μM), or wortmannin (100 nM) and allowed to adhere to substrates coated with c-Fn or Fn-fs for 2 hours, as described for adhesion assays. Cells were harvested in buffer containing Triton X-100 and protease and phosphatase inhibitors. Nonadherent cells were collected by centrifugation and combined with the adherent fraction. ERK activation was measured by Western blot with an antibody raised against the hyperphosphorylated form of the enzyme (A, top). Blots were stripped and probed with an antibody that recognizes total ERK (A, bottom) to verify equal loading. Cells pretreated with PD98059, LY294002, or wortmannin showed reduced ERK activation when compared with the control (B). Data represent the mean optical densities of two replicate Western blots minus the mean optical density of nonadherent cells under the same condition. Similar results were obtained using cells derived from a second donor.

Figure 10.

 

Proliferation of HRECs in response to the 70- (A), 110- (B), and 120-kDa (C), and Fn-fs and c-Fn (D), alone or in combination with III1-C. Ninety-six well dishes were coated with Fn and Fn-fs, and cells were incubated for 96 hours in growth medium without serum either with or without 50 μM III1-C. Viability was measured using MTT conversion. Data represent the mean ± SE of results in cells from two donors assayed in quadruplicate (*P < 0.05, when compared with uncoated wells; **P < 0.05 when compared with control data at the same concentration; ANOVA).

The authors thank Margaret I. Davis for her valuable contribution to the studies and the manuscript.