Diabetes Disrupts the Response of Retinal Endothelial Cells to the Angiomodulator Lysophosphatidic Acid

Anti-mouse and anti-rabbit horseradish peroxidase–conjugated antibodies and the anti-Src antibody were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). The extracellular signal–related kinase (Erk), phosphorylated Erk, myosin light chain (MLC)2, phosphorylated MLC2, and anti–phosphorylatedSrc antibodies and the mitogen-activated protein kinase kinase (MEK) inhibitors (PD98059 and U0126) were from Cell Signaling Technology (Beverly, MA). The pan-LPA receptor antagonist BrP-LPA {1-bromo-3(S)-hydroxy-4-[(palmitoyloxy)butyl]phosphonate)} was from Echelon Biosciences (Salt Lake City, UT). LPA (oleoyl-L-α-lysophosphatidic acid sodium salt), d-glucose, l-glucose, 2′,7′-dichlorofluorescein diacetate (DCFH-DA), N-acetylcysteine (NAC), and tiron were purchased from Sigma (St. Louis, MO). The Src inhibitors (Su6656 and PP2) and streptozotocin were purchased from Calbiochem (Gibbstown, NJ). The LPA receptor inhibitor Ki16425 was from Cayman Chemical (Ann Arbor, MI).

The institutional animal care and use committee of the Schepens Eye Research Institute approved all animal procedures. Wild-type C57BL/6J mice <10 weeks of age were fasted for 4–5 h before being intraperitoneally injected with streptozotocin (55 mg/kg) dissolved in citrate-buffered saline solution. The mice received streptozotocin for five consecutive days. Three days later, glycemia levels were checked. If glucose levels were >250 mg/dL, then animals were considered diabetic. Weight loss was controlled with 0.3 units of a 1:10 dilution of subcutaneous NPH insulin (Humulin N; Lilly, Indianapolis, IN) every other day. The effect of insulin was short, as the day after administration glucose levels were >250 mg/dL. After 1 month, the mice were killed and the retinas recovered for explants. Although glucose levels were not measured at the time of death, previous work (19) shows that in a C57BL/6J background, DM persists 1 month after streptozotocin injection.

The explants were isolated and cultured as previously described (20). Briefly, retinas were cut into 1-mm² pieces and placed in a collagen sandwich (80% collagen [Advanced Biomatrix], 20 mmol/L HEPES [Lonza, Walkersville, MD], 0.5 μg/mL fibronectin and laminin [both from Sigma], 2 mg/mL NaHCO₃, and 0.02 N NaOH). The collagen sandwich was overlaid with endothelial basal medium (EBM) (Lonza) supplemented with 10% horse serum (Lonza), 12 μg/mL bovine brain extract (Hammond Cell Tech, Windsor, CA), and 80 units/mL penicillin-streptomycin plus 25 ng/mL VEGF-A. Vessel formation was observed after 2–3 weeks of treatment. In some experiments, the medium was altered to achieve a final concentration of 25 mmol/L d- or l-glucose during the 2- to 3-week period of tube formation.

Human retinal endothelial cells (HRECs) were acquired from Cell Systems (Kirkland, WA). Primary bovine retinal endothelial cells (BRECs) were isolated as previously described (21). The cells were cultured in endothelial growth culture medium-2 and EBM (Lonza), respectively. For high glucose (HG) treatment, endothelial cells were cultured for 10 days in media supplemented with additional glucose to a final concentration of 25 mmol/L. Control cells were simultaneously treated with either 25 mmol/L l-glucose or 5 mmol/L d-glucose. Primary retinal endothelial cells were used between passages 10 and 13. All cells were maintained at 37°C with 5% CO₂.

Control and HG-pretreated BRECs were cultured as a monolayer in a 96-well plate. Once the cells reached confluence, they were rinsed twice with Krebs-Ringer solution and incubated in 10 μmol/L DCFH-DA for 45 min at 37°C. DCFH-DA is readily oxidized by reactive oxygen species (ROS) into fluorescent dichlorofluorescein. Intracellular formation of ROS was measured at 485/530 nm wavelength on a Synergy 2 plate reader (Biotek, Winooski, VT). To measure ROS in endothelial cells that had organized into tubes, the assay was done under low-serum (0.5%) conditions to prevent spontaneous regression. After tubes formed, they were washed twice with PBS and incubated for 1 h with 10 μmol/L DCFH-DA. Next, the tubes were washed two more times with PBS and observed on a Nikon TE-2000S inverted microscope. The pictures were taken with a 1-min exposure to fluorescent light at ×100 magnification.

The tube assay was performed as previously described (20). Briefly, BRECs were placed in a collagen gel sandwich and treated with 2.5 ng/mL VEGF-A. Tube formation was observed 12 h later on a Nikon TE-2000S inverted microscope. Treatments were performed in fresh media supplemented with VEGF-A. To observe the effect of endogenous LPA, the assay was performed in EBM containing 10% horse serum. Low serum diminishes the ability of BRECs to produce LPA (12). Consequently, the effect of exogenous LPA (1 μmol/L) was studied in EBM with 0.5% horse serum. Regression was observed 16 h after the addition of VEGF-A. Normalized tube length is the ratio of final and initial tube length. Pharmacological treatments were added 12 h after the addition of VEGF-A, when the tubes were first observed.

Tubes were extracted from the collagen sandwich by dissolving it with type I collagenase (1 mg/mL; Sigma) for 30 min at 37°C. The liberated tubes were centrifuged, and the resulting pellets were washed with PBS and lysed in electrophoresis sample buffer (10 mmol/L EDTA; 4% sodium dodecyl sulfate; 5.6 mol/L 2-mercaptoethanol; 20% glycerol; 200 mmol/L Tris-HCl, pH 6.8; and 0.2% bromophenol blue). Proteins were resolved on a 10% sodium dodecyl sulfate–polyacrylamide gel and subjected to Western blot analysis. Conditioned media was collected from tube assays that contained low serum, normalized to the total amount of protein present in the sample, and then subjected to Western blot analysis.

The concentration of LPA in conditioned media samples was measured using a radioenzymatic method (22), which involves the use of recombinant 1-acyl-sn-glycerol-3-phosphate acetyltransferase (23). Briefly, lipids were butanol extracted and incubated in a reaction mixture containing recombinant LPA acyltransferase and [¹⁴C]oleoyl-CoA (Perkin-Elmer, Waltham, MA). Subsequently, the reaction products were resolved by thin-layer chromatography. Under these conditions, LPA was converted to radioactively labeled phosphatidic acid, which was identified by autoradiography and densitometrically quantified.

The Src cDNA was obtained from Open Biosystems and subcloned into the pLXSN vector. The pLXSN-Src construction was transfected into 293GPG cells. The media containing the virus was collected for 5 days, concentrated by centrifugation at 25,000g for 90 min, and used as previously described (20) to infect HREC. Endothelial cells were selected on the basis of proliferation in the presence of G418 (0.3 mg/mL).

Data were analyzed using ANOVA. Differences were considered statistically significant if P < 0.05.

The overall goal of this study was to test whether DM alters the responsiveness of retinal neovessels to LPA, which promotes vascular regression in the developing eye (12). For this purpose, we focused on vessels that grew out of retinal explants (Rex) isolated from mice that endured 1 month of streptozotocin-induced DM or nondiabetic siblings (non-DM). The vessels were composed of endothelial cells and appeared to be organized into lumen-containing structures (Supplementary Fig. 1). Vessel growth was VEGF-A dependent and equivalent for both groups (Supplementary Fig. 1D). LPA induced regression of non-DM Rex vessels, whereas DM-Rex vessels were resistant (Fig. 1). We conclude that LPA promotes the regression of nondiabetic retinal vessels and that DM inhibits this response. These observations suggest that DM stabilizes retinal neovessels.

We investigated whether HG in the ex vivo setting would recapitulate LPA resistance caused by DM. To this end, we repeated the experiment shown in Fig. 1 with explants from non-DM mice, except that the glucose concentration was either kept at 5 mmol/L or increased to 25 mmol/L (HG-Rex) (Fig. 2A). As an osmotic control, Rex vessels were also cultured in the presence of 25 mmol/L l-glucose (L-G Rex) (Fig. 2A). LPA promoted the regression of L-G Rex vessels but not of HG-Rex (Fig. 2A and B). The behavior of L-G Rex vessels was equivalent to that of non-DM Rex in a normal concentration (5 mmol/L) of glucose (results not shown). VEGF-A–dependent vessel formation was equivalent for all three groups (Supplementary Fig. 1E). These experiments demonstrate that although DM causes multiple metabolic changes, HG treatment is sufficient to model LPA resistance in Rex vessels.

Our ex vivo experiments did not address the cell type(s) that were affected by DM/HG and their contribution to LPA resistance in DM/HG-Rex vessels. Consequently, we tested whether retinal endothelial cells responded to HG and became LPA resistant in the absence of any other cell types. For this purpose, we used a previously described tube assay (20,24) in which endothelial cells organize into tubes in a VEGF-A–dependent manner.

We pretreated BRECs with 25 mmol/L glucose for 10 days (HG group). As an osmolality control, we simultaneously pretreated a second group of BRECs with 25 mmol/L l-glucose (L-G). A third group included cells cultured in normal conditions (5 mmol/L glucose) (NG). Subsequently, the cells were subjected to the tube assay. Tube formation was observed 12 h after VEGF-A addition (Fig. 3A). HG and L-G treatment did not influence VEGF-A–dependent tube formation (results not shown). The overall morphology of the tubes was similar, and there was no significant difference in the number of branch points in the three groups (results not shown). Addition of exogenous LPA (1 μmol/L) to the tubes after they had formed induced regression in both control groups, whereas the HG tubes were resistant (Fig. 3). This result reproduced the events observed in the ex vivo assay. While a higher dose of LPA (10 μmol/L) also induced regression of NG tubes, it had no effect on HG tubes; a lower concentration of LPA (0.1 μmol/L) failed to promote regression of NG tubes (results not shown). We conclude that HG is sufficient to induce LPA resistance in endothelial cells, without the input of the other cell types present in the retinal explant.

In summary, our findings show that 1) DM eliminates the ability of Rex vessels to regress in response to LPA (enhances their stability), 2) HG is sufficient to induce LPA resistance in Rex vessels and primary retinal endothelial cells, and 3) primary retinal endothelial cells can become LPA resistant independently of the effect of DM on other retinal cell types. Therefore, retinal endothelial cells are the target of HG-mediated LPA resistance.

The experiments in Figs. 1–3 demonstrate that DM/HG induces resistance to regression mediated by exogenous LPA. When the tube assay is performed in 10% serum, tubes undergo spontaneous regression because of the combined action of endothelial cell–produced LPA and LPA from the serum (12,18). Consequently, we tested whether HG also induced resistance under these spontaneous regression conditions. Supplementary Fig. 2A and B shows that control tubes formed in the presence of 10% serum spontaneously regressed, whereas HG tubes remained stable. The observation that an LPA receptor inhibitor (10 μmol/L BrP-LPA) prevented tube regression (Supplementary Fig. 2A and B) confirmed that regression of the control tubes was LPA dependent. We repeated these experiments with primary HRECs and observed similar results (Supplementary Fig. 3). VEGF-A promoted comparable formation of tubes, which regressed spontaneously in control conditions but not in the HG setting. Furthermore, LPA receptor inhibitors BrP-LPA and Ki16425 (10 and 5 μmol/L, respectively) inhibited regression (Supplementary Fig. 3). We conclude that exposing either BRECs or HRECs to HG inhibited spontaneous regression driven by the combination of endogenously produced and serum-supplied LPA.

Decreased endogenous LPA production was a conceivable explanation for the resistance to spontaneous regression observed in HG-treated cells. For instance, HG may inhibit production and/or secretion of autotaxin, which would reduce the amount of LPA and thereby prevent spontaneous regression. Relative to NG tubes, the amount of autotaxin in tubes and conditioned medium was reduced, and so was the quantity of LPA (Supplementary Fig. 2C–H). However, these parameters were also reduced in L-G tubes, which regressed as well as NG tubes (Supplementary Fig. 2A and B). Consequently, it did not appear that the small change in the quantity of autotaxin and LPA was a likely explanation for HG-induced resistance to spontaneous regression. Furthermore, such a mechanism would be unlikely to explain why HG tubes failed to regress in response to exogenously added LPA (Fig. 3). Therefore, we suspected that LPA resistance arose from interference with LPA-dependent signaling.

Because oxidative stress produced by ROS is a common element linking diverse mechanisms mediating vascular injury in DM (25,26), we investigated the role of ROS in HG-mediated LPA resistance. Quantification of ROS produced by monolayers of control and HG BRECs revealed that ROS was elevated in HG BRECs (Fig. 4A). Similarly, more ROS was visible in tubes organized from HG BRECs (Fig. 4B). These results suggested that resistance to regression was related to an elevated level of ROS. Indeed, the ROS scavenger NAC (27,28) reduced ROS levels (Fig. 4A and B) and restored spontaneous regression in HG tubes (Fig. 4C and D). Similar results were obtained with tiron, another antioxidant (Supplementary Fig. 4). The LPA receptor inhibitor BrP-LPA prevented regression of NAC-treated tubes, indicating that regression was LPA dependent. This idea was reinforced by the observation that NAC failed to promote regression of tubes in low (0.5%) serum, in which the LPA level is below the amount necessary for regression (results not shown). This set of experiments demonstrates that elevation of ROS was a key component of HG-induced LPA resistance.

ROS activates Src-family kinases (SFKs) in multiple cell types, including human pulmonary (29), coronary, and umbilical vein endothelial cells (30). Our laboratory demonstrated that activation of Src prevents tube regression (20). To test the impact of HG on Src activity, we assessed phosphorylation of Src’s Y416 in lysates prepared from tubes. Phospho-Src was significantly higher in HG tubes (Fig. 5A and B). Furthermore, Src inhibitors (Su6656 and PP2) promoted tube regression, which was blocked by coaddition of BrP-LPA (Fig. 5C and D). Moreover, we observed that a modest overexpression of Src induced LPA resistance in HRECs (Fig. 6). We conclude that increased Src activity is a key component of HG-induced LPA resistance.

Since SFK-mediated stability of tubes is dependent on the Erk pathway (20), we considered their involvement in HG-mediated resistance to LPA. In tubes organized from HG BRECs, Erk phosphorylation was elevated and could be suppressed with NAC (Fig. 7A–D). Furthermore, two different MEK inhibitors rescued LPA-dependent regression (Fig. 7E and F). These results reveal that HG activates the Erk pathway and that this event is required for LPA resistance. Finally, since activation of the Erk pathway antagonizes RhoA/rho-associated protein kinase (ROCK) (20,31), which is required for LPA-induced regression (12,20), we compared phosphorylation of a ROCK substrate, MLC2, in HG and control tubes. As expected, MLC2 phosphorylation was reduced in HG tubes (Supplementary Fig. 5).

In summary, our results show that LPA resistance of HG-treated endothelial cells involves elevated ROS that engages Src and the Erk pathway and thereby antagonizes essential LPA-dependent signaling events. These observations suggest that HG reprograms signaling and thereby alters the response of retinal endothelial cells to angiomodulators such as LPA.

Our next step was to test whether the mechanism mediating HG-induced LPA resistance described in the tube assay was relevant to DM-Rex vessels. Indeed, LPA-mediated regression of DM-Rex vessels was observed in the presence of either NAC or inhibitors of Src or MEK (Fig. 8A and B). No regression was observed if the inhibitors were added alone (Fig. 8B). We propose the following model to explain how DM/HG blunts the responsiveness of endothelial cells/vessels to LPA (Fig. 8C). Hyperglycemia elevates the level of ROS, which activates SFKs that promote the Erk pathway. The Erk pathway antagonizes RhoA/ROCK, which is required for LPA-mediated regression (12,20).

We found that DM induced resistance to LPA-mediated regression of retinal neovessels. Furthermore, HG mimicked DM-induced LPA resistance in both ex vivo retinal neovessels and in vitro tubes organized from primary retinal endothelial cells. Finally, LPA resistance resulted from DM/HG-induced reprogramming of signaling pathways that were required for LPA-dependent regression (Fig. 8C).

Our work is not the only example in which DM/HG reprograms signaling pathways within vascular cells. Geraldes et al. (32), reported that DM and HG led to persistent activation of protein kinase C-δ in retinal capillaries and pericytes, which precipitated nuclear factor-κB–independent apoptosis of retinal pericytes (32). Furthermore, reestablishing normoxia by insulin administration did not correct the elevated protein kinase C-δ activity, which indicates that DM/HG-induced reprogramming of cell signaling is persistent. This phenomenon was also observed in studies presented herein; neovessels that formed from retinas of DM mice were LPA resistant, even after 2–3 weeks of culture in normoglycemic conditions. Thus, LPA resistance appears to be a manifestation of metabolic memory.

The majority of new PDR therapies have focused on controlling the angiogenic effects of VEGF. Although the use of these new drugs induces regression of preretinal neovascularization in a significant number of patients, anti-VEGF therapy does not benefit all patients. For instance, Adamis et al. (6) reported that after Macugen treatment, 37.5% of PDR patients still present retinal neovascularization. Other studies indicate that 75% of diabetic retinopathy patients treated with Bevacizumab continue to present residual neovascularization (7). Unpublished work by our laboratory indicates that LPA is present in vitreous of nondiabetic patients and promotes regression of tubes. Taken together, these observations suggest that pathogenesis of PDR results from not only a rise in the level of proangiogenic agents in vitreous but also a resistance of retinal neovessels to regression factors such as LPA.

The HG-initiated ROS/Src/Erk pathway mediating LPA resistance (Fig. 8C) provides multiple targets for potential new therapies for PDR. For instance, antioxidants are an attractive choice, especially in light of the fact that ROS levels are elevated in vitreous of diabetic retinopathy patients and increase as the disease progresses to PDR (33). However, dietary antioxidant therapies for diabetic retinopathy have led to ambiguous results (26). Although animal studies were promising, the results in human studies were contradictory. For instance, oral antioxidant supplementation decreases ROS activity in plasma and is associated with delayed diabetic retinopathy progression in type 2 diabetes (34). On the other hand, Millen et al. (35) showed that dietary supplementation with vitamins A and E had no effect on diabetic retinopathy progression. The reason for these discrepancies is not clear. Not being able to measure ROS levels and hence the efficiency of treatment on the desired target represents a major roadblock in the area of research.

The studies presented herein are consistent with the idea that inhibition of Src activity may promote regression of retinal neovessels and thereby be a potential therapy for patients with PDR. Support for this idea includes the observation that topical administration of Src inhibitors significantly reduces vascular leakage induced by intravitreal injection of VEGF (36). Importantly, there are tyrosine kinase inhibitors with the capacity of targeting Src (37,38) that are Food and Drug Administration approved for the treatment of leukemia and renal cell carcinoma (dasatinib and sunitinib, respectively) (37,39). Our findings beg the question of whether these agents would be beneficial for management of patients with PDR.

This study was supported by grants from the Juvenile Diabetes Research Foundation (I-2008-905 to A.K.), the National Institutes of Health (DK-083336 to E.I.), the Pew Latin American Fellows Program in the Biomedical Sciences, the American Diabetes Association Mentor-Based Minority Fellowship (7-09-MI-04), and Fundacion Mexico en Harvard (to J.A.).

No potential conflicts of interest related to this article were reported.

J.A. designed research, performed the experiments, and wrote the manuscript. R.M. measured LPA in conditioned media of tubes and provided critical input for the manuscript and experimental results. E.I. made the initial observation that HG induced LPA resistance in retinal endothelial cells. A.K. designed the research and wrote the manuscript. A.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors are very grateful for the excellent technical assistance of Daniel Lorenzana (St. Catherine of Siena Medical Center) and Alzbeta Godarova (Department of Cancer and Cell Biology, University of Cincinnati). Chiara Gerhardinger, Michael Goodridge, and Jessica Lanzim from the Schepens Eye Research Institute provided critical input for the care of diabetic mice. Dr. Kazuhiko Kume from Kumamoto University kindly donated the plasmid encoding LPA acyltransferase (23). The VEGF-A used for this study was a generous donation from the Biological Resources Branch Preclinical Repository of the National Cancer Institute. The authors also thank Dr. Timothy Kern (Case Western Reserve University) for providing plasma from non-DM and DM mice.