A lipidomic screen of hyperglycemia-treated HRECs links 12/15-Lipoxygenase to microvascular dysfunction during diabetic retinopathy via NADPH oxidase.

Retinal hyperpermeability and subsequent macular edema is a cardinal feature of early diabetic retinopathy (DR). Here, we investigated the role of bioactive lipid metabolites, in particular 12/15-lipoxygenase (LOX)-derived metabolites, in this process. LC/MS lipidomic screen of human retinal endothelial cells (HRECs) demonstrated that 15-HETE was the only significantly increased metabolite (2.4 ± 0.4-fold, P = 0.0004) by high glucose (30 mM) treatment. In the presence of arachidonic acid, additional eicosanoids generated by 12/15-LOX, including 12- and 11-HETEs, were significantly increased. Fluorescein angiography and retinal albumin leakage showed a significant decrease in retinal hyperpermeability in streptozotocin-induced diabetic mice lacking 12/15-LOX compared with diabetic WT mice. Our previous studies demonstrated the potential role of NADPH oxidase in mediating the permeability effect of 12- and 15-HETEs, therefore we tested the impact of intraocular injection of 12-HETE in mice lacking the catalytic subunit of NADPH oxidase (NOX2). The permeability effect of 12-HETE was significantly reduced in NOX2−/− mice compared with the WT mice. In vitro experiments also showed that 15-HETE induced HREC migration and tube formation in a NOX-dependent manner. Taken together our data suggest that 12/15-LOX is implicated in DR via a NOX-dependent mechanism.

In addition to dyslipidemia, oxidative stress has attracted considerable interest as a key factor in mediating retinal vascular injury during DR. Our previous studies highlighted NADPH oxidase (NOX) as a major source of ROS generation in the retinas of experimental diabetic rodents and OIR models, as well as in retinal endothelial cells treated with high glucose (HG) or hypoxia (23)(24)(25)(26). Our recent study showed that inhibiting 12/15-LOX by baicalein reduced diabetes-induced ROS generation and NOX2 expression in mouse retina, suggesting the existence of an interconnected signaling between LOXs and NOX ( 22 ). However, the cross-talk between NOX and infl ammatory bioactive lipids remains ill-defi ned in DR.
Therefore, the current study has been taken up to characterize the effect of hyperglycemia on bioactive-lipid profi le in human retinal endothelial cells (HRECs) to determine whether endothelial cells are a potential source of 12/15-LOX-derived hydroxyeicosanoids, and whether these metabolites play a causative role in the pathogenesis of BRB dysfunction during diabetes. Furthermore, the present work aimed to gain insights into the underlying molecular mechanisms by portraying the causal relationship with NOX.

Animal preparation and experimental design
All procedures with animals were performed in accordance with the Public Health Service Guide for the Care and Use of Laboratory Animals (Department of Health, Education, and Welfare publication, National Institutes of Health 80-23) and Georgia Regents University guidelines. Six to eight-week-old male 12/15-LOX knockout (LOX Ϫ / Ϫ ) mice lacking the NOX2 subunit of NOX (NOX2 Ϫ / Ϫ ), and corresponding littermate controls, WT mice in C57BL/6J background (Jackson Laboratory, Bar Harbor, ME), were matched according to sex, age, and weight. Animals were given intraperitoneal injections of freshly prepared streptozotocin (45 mg/kg; Enzo Life Sciences) dissolved in 0.9% NaCl after 4 h fasting for fi ve consecutive days.
Mice with blood glucose levels >300 mg ⁄dl were considered diabetic. Six weeks after establishment of diabetes, retinal vascular permeability was evaluated using fl uorescein angiography. Thereafter retina samples were used for total albumin analysis by Western blot. For intravitreal injections, the procedure was essentially the same as previously described ( 27 ). To avoid uncontrolled intraocular pressure increase, the volume of intravitreal injections was limited to 1 l. 12-HETE was dissolved in ethanol and a working solution of 10× was prepared by diluting 0.32 l of stock solution (312 M) to 100 l with PBS, assuming the vitreous volume of mouse eye is ‫ف‬ 10 l ( 28 ). Then by injecting 1 l of this working solution, a 0.1 M vitreal concentration of 12-HETE was obtained. The vitreal concentration of ethanol was 0.032%. The volume of the injected solution apparently did not cause signifi cant pressure-induced retinal damage, because 0.032% ethanol-PBS-injected control eyes showed normal retinal morphology with no apparent apoptosis within 7 days. The dose of 12-or 15-HETE was chosen according to what was detected previously in the vitreous of patients with DR, 50 ng/ml ( ‫ف‬ 0.1 M) ( 21 ). ( 4,5 ). Likewise, cumulative risks of cataract and glaucoma, as well as local immunosuppression, are frequently associated with corticosteroid intravitreal therapy ( 6,7 ). Even with laser-based photocoagulation, the gold standard treatment for DR, side effects ranging from blurred peripheral vision to scotomas may occur ( 8 ). These therapeutic limitations impose the necessity for novel therapeutic interventions via unraveling the pathophysiology of DR.
In recent years, human studies have underscored the strong association between dyslipidemia and the development of DR. Clinically, it has been reported that the severity of retinopathy in type 1 diabetes was correlated positively with triglyceride level and negatively with the HDL concentration ( 9 ). Additionally, intensive dyslipidemia therapy signifi cantly slowed the progression of DR in type 2 diabetes over 4 years ( 10 ). Diabetic dyslipidemia is characterized by a shift in the fatty acid profi le with an increase in omega-6 PUFAs, in particular arachidonic acid (AA), released from membrane lipids by the activated phospholipase A2 (PLA2) ( 11,12 ). The released AA, in turn, can be converted to infl ammatory bioactive lipid mediators such as HETEs, leukotrienes, and prostaglandins via different enzymatic pathways including cycloxygenase, lipoxygenase (LOX), and cytochrome P450 ( 13 ).
Bioactive lipids activate specifi c signaling pathways that are implicated in cell proliferation, differentiation, and apoptosis. Sustained cellular response to increased bioactive lipids becomes pathological and results in chronic diseases such as cancer and atherosclerosis ( 14 ).
While there is irrefutable evidence for the role of dyslipidemia in the progression of DR ( 9,15 ), bioactive lipid metabolites have received limited attention. HETEs are major monohydroxylated AA bioactive lipid metabolites that are produced during infl ammatory and immunological reactions ( 16 ). The 12-and 15-HETEs are formed by human 15-LOX and its murine ortholog 12/15-LOX ( 17 ) in a variety of mammalian cells such as endothelial cells, eosinophils, and epithelial cells to exhibit distinct biological activities. These activities include stimulation of endothelial cell mitogenesis, vascular infl ammation, and mucus release from human airways (18)(19)(20). Recently, we have shown that pharmacological inhibition or deletion of 12/15-LOX reduces retinal NV in the oxygen-induced retinopathy (OIR) model ( 21 ). Moreover, we have demonstrated in both human patients and animal models that retinal expression of 12/15-LOX is robustly induced during diabetes ( 21 ). At the same time, its pharmacological inhibition dampened the levels of infl ammatory cytokines, reactive oxygen species (ROS) gener ation, and phosphorylated VEGF receptor 2 expression in the retinas of diabetic mice ( 22 ). However, the cellular source of these lipid metabolites is unknown and could be derived from retinal tissues, including the retinal vascular endothelial, glial, and pigmented epithelial cells, as well as from infi ltrated infl ammatory cells. Moreover, the role of 12-or 15 HETE in mediating the breakdown of barrier function, such as that observed in DR, has not been investigated thoroughly to date. 4°C. The next day the sections were incubated in Oregon greenlabeled anti-rabbit antibody. Sections were covered using DAPI mounting medium and images were obtained with confocal microscopy (LSM 510; Carl Zeiss).

Fluorescein angiography
The anesthetized mouse was placed on the imaging platform of the Phoenix Micron III retinal imaging microscope and Goniovisc 2.5% was applied liberally to keep the eye moist during imaging. Mice were administered 10 to 20 l 10% fl uorescein sodium and rapid acquisition of fl uorescent images ensued for ‫ف‬ 5 min, as previously described ( 33 ).

In vitro leukocyte adhesion assay
HRECs were grown to confl uence in 12-well plates then treated with or without 15-HETE (0.1 M) in the presence or absence of apocynin (30 M) or NAC (50 M), as well as with positive control, 10 g/ml of the endotoxin Escherichia coli lipopolysaccharide (Sigma-Aldrich, St. Louis, MO) for 24 h. Following the treatment, leukocytes (300,000-400,000 cells/well)

Microvascular HRECs
In vitro experiments were performed using cultured HRECs (Cell Systems Corporation). After the cells were 80-90% confl uent, they were serum starved (2% FBS) overnight, then treated with 15-HETE (0.1 M) with or without apocynin (30 M) or Nacetyl-L-cysteine (NAC) (50 M), or HG (D-glucose, 30 mM) in the presence or absence of 20 M of AA. The osmolarity of the control group in the HG experiments was adjusted using L-glucose. Cytotoxicity of tested inhibitors was assessed by MTT assay as previously described ( 29 ). Transcellular electrical resistance (TER) and cell migration were done using electric cell-substrate impedance sensing (ECIS) (Model 1600R, Applied BioPhysics) as previously described ( 30,31 ). The experiment was terminated after 24 h of 15-HETE treatment for analysis of leukocyte adhesion and tube formation. Conditioned media were used for multiplex assay of various cytokines as previously described ( 30 ). Meanwhile, the HG experiment was terminated 5 days after initiation of treatment and cell lysates were collected and analyzed for 15-LOX and 5-LOX protein expression using Western blot analysis. At the same time, collected cell lysates were assayed for PLA2 activity according to manufacturer's instructions (Cayman, Ann Arbor, MI). For lipidomic analysis, the cells were incubated in indicatorfree media for 5 days and then collected with cells together, sonicated, centrifuged to remove cell debris, and then frozen for the analysis in the Lipidomics Core Facility (Wayne State University, Detroit, MI) as described before ( 32 ).

Immunofl uorescence
Retinal paraffi n sections of human subjects with or without DR obtained from Capital Bioscience (Rockville, MD) were fi xed in 10% neutral buffered formalin. Following rehydration of the paraffi n section and two washes in PBS, sections were treated with proteinase K for 10 min and washed twice in PBS followed by blocking with 10% normal goat serum and then incubated with phospho-PLA2 antibody overnight in a humidifi ed container at

Data analysis
The results are expressed as mean ± SD. Differences among experimental groups were evaluated by using the two-tailed t -test or one-way ANOVA. When statistical differences were observed using ANOVA, a post hoc Tukey's test was performed to determine which groups differed.
The Bonferroni correction factor (BCF), which accounts for a large number of comparisons by reducing the ␣ level as to decrease the probability of obtaining a type I error, was used when analyzing lipid metabolite profi les among groups. BCF was calculated by dividing the ␣ level (0.05) by the number of bioactive lipid metabolites detected. Because 56 individual metabolites were detected, the new ␣ value used for individual metabolites was 0.0009 (0.05/56 ). A new ␣ level of 0.0055 (0.05/9) was also calculated for the top 15% of metabolites which are used for further analysis by measuring their profi le in the presence of exogenous AA as a substrate under hyperglycemic conditions.
Differences were considered statistically signifi cant if P < 0.05, except for lipid metabolites, which used the ␣ values calculated from the Bonferroni correction.
purchased from Sanguine Bioscience (Valencia, CA) and labeled with diluted LeukoTracker solution (Cell Biolab) for 60 min at 37°C, according to the manufacturer's instructions, were added to confl uent monolayer HRECs and incubated for 90 min at 37°C. Immediately before the assay, nonadherent cells were removed by washing three times with RPMI-1640 medium (Life Technology, Grand Island, NY). The adherent labeled leukocytes were counted under the inverted fl uorescence microscope at 480 nm/520 nm from three separate fi elds per well.

In vitro endothelial tube formation on Matrigel
Tube formation assay was done as described before ( 34 ). Briefl y, vehicle or 15-HETE (0.1 M) in the presence or absence of different inhibitors, was added to the appropriate wells and the cells were incubated at 37°C overnight. Tube formation was observed under an inverted microscope and the images were captured with a digital camera attached to the microscope. The measurement was done on three randomly chosen microscopic fi elds per well, and the mean of the three fi elds was used as a single observation. Tube formation was further quantifi ed by measuring the total length of tube-like cells using ImageJ software.

Hyperglycemia alters bioactive-lipid profi le in HRECs
In order to evaluate the potential role of bioactive lipids in mediating the edematous phenotype observed in DR, we fi rst sought to characterize their profi le under diabetic milieu. In this regard, the lipidomic profi le of HRECs was screened for metabolites whose levels changed dramatically during hyperglycemia. Out of 126 bioactive lipids screened, 70 were not detectable ( Table 1 ), 47 did not change signifi cantly ( t -test, P > 0.05), and 9 metabolites were increased ( t -test, P < 0.05) under hyperglycemic conditions (30 mM D-glucose) compared with the normoosmotic control (5 mM D-glucose + 25 mM L-glucose) ( Fig. 1A ). However, after the Bonferroni correction, the fold changes in the level of eight metabolites among the nine increased were statistically not signifi cant, and only 15-HETE had a signifi cant fold increase (2.4 ± 0.4) with a P value of 0.0004 (BCF ␣ = 0.0009). Next, we narrowed our focus to confi rm these data, considering the top 15% of the upregulated metabolites detected in this screen . By measuring their profi le in the presence of exogenous AA as a substrate under hyperglycemic conditions, we found a signifi cant increase in AA-derived eicosanoids generated by the 12/15-LOX pathway including: 15-HETE (BCF ␣ = 0.0055, Tukey's post hoc, P = 0.001), 11-HETE (BCF ␣ = 0.0055, Tukey's post hoc, P = 0.0001), and 12-HETE (BCF ␣ = 0.0055, Tukey's post hoc, P = 0.001) in which 15-HETE again had the highest (5.48 ± 1.12) fold increase. On the other hand, the levels of other non-AAderived metabolites [8-, 13-, 16-, and 17-hydroxydocosahexaenoic acid (HDoHE) and 13-HODE] did not change signifi cantly in the presence of exogenous AA with the exception of 13-HODE, Fig. 1B .
Western blot analysis of LOXs from HRECs further corroborated data obtained by LC/MS metabolite measurement. Results revealed that 15-LOX protein expression was signifi cantly increased by 2.5-fold after treating HRECs with HG compared with control. Meanwhile, the protein expression level of 5-LOX did not change signifi cantly under the same treatment, Fig. 2A . Given the prominent increase in levels of 15-HETE under the hyperglycemia, we proceeded to investigate the propensity of 15-HETE to disrupt vascular barrier function in the HREC monolayer using real-time analysis of TER, an indicator of monolayer integrity. We tested the concentration of 15-HETE  pictures taken at constant interval of every mouse studied in each group. The fl uorescence intensity of FA per mouse retina was calculated by the ImageJ software after conversion to binary images then normalized to that of normal WT, which was arbitrarily set at 100. B: Western blot of total retinal albumin among the studied groups followed by densitometric analysis. Ratios of albumin band intensities relative to the actin for each group were compared with normal WT control, which was arbitrarily set at 1.0. Data shown for the comparison are the mean ± SD of four to six mice studied in each group.
that had been detected in the vitreous of patients with DR, ‫ف‬ 55 ng/ml (0.1-0.3 M) ( 21 ) and shown by us to elicit transcellular leakage of FITC-dextran in bovine retinal endothelial cells ( 22 ). Changes in TER were fi rst observed after 26 h of 15-HETE treatment and continued to decrease throughout the experiment period, Fig. 2B . These data presumably refl ect a PLA2-dependent pathway in releasing endogenous free AA as the mediator's precursor followed by the activation of metabolizing LOXs. To this end, we determined the in vitro endothelial PLA2 activity in response to diabetic conditions. As depicted in Fig. 2C , treatment of HRECs with hyperglycemia for 5 days stimulated the activity of PLA2 by 2-fold compared with control. This fi nding of increased PLA2 activity has been substantiated by the detection of phosphorylated PLA2 (active form) in diabetic human retinal sections ( Fig. 2D ). Phosphorylated PLA2 was mostly localized in the endothelial layer of the retinal vessels and perivascular retinal glial cells, as well as in the outer segment of the photoreceptors.

Evolving role of 12/15-LOX pathway in compromising retinal barrier function during diabetes
The aforementioned data from lipidomic analysis provide a rationale to explore the role of the 12/15-LOX pathway in diabetes-induced BRB breakdown. We used two complementary approaches for this investigation. The fi rst approach was aimed to assess whether knocking out retinal 12/15-LOX, an ortholog of human 15-LOX1 ( 17 ), would effi ciently reduce the pathological vascular permeability seen in DR. To achieve this goal, we used the welldefi ned 12/15-LOX (Alox15) knockout mice and matched congenic controls. First, these mice were regenotyped before the study to ensure that they had deletion of 12/15-LOX in the retina according to the Jackson Laboratory's protocol (data not shown). Next, streptozotocin was administered to ‫ف‬ 6-week-old mice and blood glucose was monitored. Mice were considered diabetic if their nonfasted glucose level was higher than 300 mg/dl. After 6 weeks, the magnitude of retinal vascular leakage was assessed via a clinical diagnostic technique, fl uorescein angiography. At this time point, fl uorescein angiography from WT diabetic mice exhibited a signifi cant elevation, almost 2.5-fold in fl uorescein hyper-fl uorescence compared with the nondiabetic littermates. In contrast, the increase in fl uorescein hyper-fl uorescence was significantly lowered to 1.9-fold by knocking out 12/15-LOX, Fig. 3A . To further confi rm the role of the 12/15-LOX pathway in mediating diabetes-induced vascular permeability, Western blot analysis of total retinal albumin was performed. As shown in Fig. 3B , the total retinal albumin content in diabetic WT mice was 4-fold higher than that in nondiabetic WT mice. However, knocking out 12/15-LOX resulted in a 59% reduction of total albumin in the retinas of diabetic mice.
Further evidence for the ability of 12/15-LOX to compromise endothelial barrier function was obtained from our second approach. In this approach, an in vivo experimental eye model was used in which normal mice were injected intravitreally with a 12/15-LOX-derived predominant murine metabolite, 12-HETE, into the right eye and vehicle into the left eye. Thereafter, the putative biological effect of 12-HETE per se on retinal permeability was photographically seen via fl uorescein angiography. As shown in Fig. 4A , a marked increase in fl uorescein leakage was observed in eyes receiving 12-HETE compared with the vehicle-injected contralateral eyes. Concordantly, Western blot analysis for total retinal albumin was performed to further corroborate 12-HETE's ability to induce vascular permeability. As shown in Fig. 4B , 12-HETE injection induced a signifi cant 2.8-fold increase in total retinal albumin compared with vehicle-injected controls.

The effect of 12-HETE on retinal vasculature was characterized by a marked increase in the infl ammatory response
We next sought to determine whether the deleterious effect of 12-HETE on retinal vasculature was associated with a pro-infl ammatory phenotype (increased adhesion molecule expression and leukocyte adhesion). To address this point, we determined the effect of 12-HETE injection on retinal levels of ICAM-1 and VCAM-1, well-established markers of endothelial dysfunction in infl ammatory conditions ( 35 ). Western blot analysis of ICAM-1 showed a 2.9-fold increase in 12-HETE-injected eyes compared with vehicle-injected contralateral eyes. Likewise, the retinal VCAM-1 level was higher (2.5-fold, P < 0.05) in 12-HETE-injected eyes than in control. Furthermore, the infl ammatory response to the intravitreally injected 12-HETE was determined by leukocyte infi ltration as assessed by CD45, a common infl ammatory leukocyte antigen. A signifi cant increase in CD45 immunoreactivity (1.4-fold) in the retinas of 12-HETE-injected eyes was observed compared with control ( Fig. 4C ). The relative fl uorescence intensity of FA per mouse retina was calculated by the ImageJ software then normalized as a percentage to that of vehicle-injected control, which was arbitrarily set at 100%. B: Western blot of total retinal albumin among the studied groups followed by densitometric analysis. Ratio of the albumin band intensity relative to actin for the 12-HETE-injected group was compared with the vehicle-injected control, which was arbitrarily set at 1.0. C: Western blot analysis of retinal ICAM-1, VCAM-1, CD45, and NOX2 after intraocular injection with either 12-HETE (0.1 µM), or vehicle followed by densitometric analysis. Ratios of band intensities of ICAM-1, VCAM-1, CD45, and NOX2, respectively, relative to the actin for 12-HETE-injected group were compared with the vehicle-injected control, which was arbitrarily set at 1.0. Data shown for the comparison are the mean ± SD and representative of four to six mice studied in each group.

Intravitreal injection of 12-HETE increased retinal expression of NOX2
Previous studies, including those from our laboratory, have pointed out that NOX-derived ROS have a central role in both endothelial and leukocyte activation, keys of the retinal infl ammatory paradigm ( 25,36 ). Therefore, a causal relationship between 12-HETE and NOX in mediating endothelial dysfunction was investigated. To accomplish this, we fi rst assessed the question of whether the retinal expression of NOX2 is enhanced by intravitreal injection of 12-HETE or not. As shown in Fig. 4C , the NOX2 protein level was increased 1.7-fold in the retinas of 12-HETE injected eyes compared with contralateral controls, implying an existing causal relationship.

12/15-LOX lipid metabolites induce HREC activation/ dysfunction in a NOX-dependent manner
Guided by the aforementioned results, the role of the NOX-dependent pathway in 15-HETE-mediated endothelial dysfunction was tested in depth using a cell culture model of HRECs. With this model, incubation with either 12-HETE or 15-HETE signifi cantly activated endothelial cells to release multiple pro-infl ammatory cytokines and chemokines. Among these, IL-6, IL-8, IL-17, and MCP-1 appear to dominate ( Fig. 5A ). To ensure that 12-or 15-HETE-activated endothelial cells exhibited features of microvascular dysfunction characterized by augmentation of polymorphonuclear leukocyte (PMN) adherence, we performed leukocyte adhesion assays. Exposure of HRECs to 15-HETE for 24 h signifi cantly augmented the number of PMNs adhering to HRECs (5.9-fold increase) compared with control. Next, we tested the role of NOX-derived ROS in 15-HETE-induced leukostasis by using apocynin, a NOX inhibitor, to inhibit ROS generation, as well as NAC (a ROS scavenger) to eliminate ROS as they were generated. When apocynin or NAC was added before 15-HETE, the number of adherent leukocytes induced by 15-HETE dropped back to normal ( Fig. 5B ). To ensure that these effects were not caused by nonspecifi c cytotoxicity of apocynin or NAC, we assessed cell viability in HRECs after 24 h exposure to apocynin or NAC, using the MTT assay. Apocynin or NAC, at the concentration used, did not affect HREC viability (94 ± 4% vital cells), indicating that the decrease in the number of adherent leukocytes to 12-HETE-injected WT littermates ( Fig. 8 ). Taken together, these results confi rm that NOX is involved in the 12-HETEinduced disruption of the inner BRB integrity.

DISCUSSION
The current therapeutic interventions to treat DR still rely heavily on controlling systemic hyperglycemia. However, many diabetic patients develop retinopathy despite having a good control ( 37 ). Furthermore, the recent targeted treatments strategies, corticosteroids, and anti-VEGF therapies, as well as laser photocoagulation, are limited by their off-target effects. Therefore, it is worthwhile to explore new therapeutic avenues to improve DR. To the best of our knowledge, the current study provides the fi rst preclinical evidence pertaining to the involvement of 12/15-LOX-derived bioactive lipid metabolites in the pathogenesis of vascular barrier breakdown in the early stages of DR. Effectively, our study is the fi rst to screen the bioactive lipid profi le in HRECs under hyperglycemic conditions. The key fi ndings of this screen are: 1 ) The products of the 12/15-LOX pathway were signifi cantly upregulated under hyperglycemic conditions with 15-HETE exhibiting the most signifi cant increase. 2 ) 15-HETE activates retinal endothelial cells through the NOX system leading to increases in leukocyte adhesion, hyperpermeability, and finally NV, the cardinal signs of DR.
LOX is an enzyme that catalyzes the addition of oxygen to PUFAs containing 1,4-pentadiene structure. The most common analogs are 5-LOX, 12-LOX, and 15-LOX, with the number indicating at which carbon the oxygen is inserted. Of note, species-specifi c differences between orthologous LOX isoforms have been described for the murine 12-LOX, which is an arachidonate 15-LOX in humans. This difference indicates that care should be taken if experimental data on LOX metabolism are being transferred from one species to another ( 38,39 ). Accordingly, 15-HETE-activated HRECs was indeed consecutive to inhibition of NOX activity, but not to cell death.
In addition to stimulation of leukostasis by 15-HETE, it also promoted retinal endothelial angiogenesis. This is clearly depicted in two representative in vitro angiogenic assays, capillary-like structure formation, and cell migration ( Figs. 6, 7 ). As shown in Fig. 6A and quantifi ed in Fig.  6B , the length of capillary-like structure formed by HRECs in the presence of 15-HETE was signifi cantly increased by 4-fold compared with control. To investigate the contribution of NOX in the formation of capillary-like structure induced by 15-HETE treatment, NOX inhibitors were added to HRECs before 15-HETE treatment. Thereafter, the mean length of tubes between cells was measured. We found that apocynin, as well as NAC, signifi cantly abrogated 15-HETE-induced tube formation. Similarly, 15-HETE signifi cantly increased HREC migration rate, an early step in angiogenesis. This is graphically shown in Fig.  7A , where the capacitance of all treated cells was increased from ‫ف‬ 1 nF to 3 nF, wounding phase, then recovered at differential rates during the migration, healing phase. The mean migration rate of HRECs treated with 15-HETE was signifi cantly higher ( P < 0.01) than the rate obtained with vehicle-treated control. Whether or not NOX is involved in 15-HETE-induced endothelial cell migration was then investigated using two complementary approaches. First, a NOX inhibitor, apocynin, signifi cantly inhibited 15-HETEinduced HREC migration without affecting cell viability ( Fig. 7A ). Second, HRECs were transiently transfected with NOX2 siRNA or scrambled siRNA ( Fig. 7B, C ) and then treated with 15-HETE as before. NOX2 siRNA, but not scrambled siRNA, significantly inhibited 15-HETEinduced HREC migration ( Fig. 7D ).
To further understand the contribution of NOX in mediating endothelial barrier dysfunction induced by 12/15-LOX lipid metabolites, we injected 12-HETE intravitreally into NOX2 Ϫ / Ϫ mice. At 1 week after injection, retinal fl uorescein leakage of NOX2 Ϫ / Ϫ mice was statistically less than  we have used 12-HETE in all murine experiments versus 15-HETE in all human cell experiments. Although a large body of data indicates that LOX plays a role in the pathogenesis of various diseases, including cancer and atherosclerosis, novel physiological roles continue to emerge. In relation to retinal vasculopathies, LOXs present themselves as attractive candidates for therapeutic targeting. This input has originated partly from pathological studies showing the elevated level of 15-HETE in epiretinal membranes of proliferative vitreoretinal and proliferative DR patients ( 40 ). These initial observations have been supported by LC/MS analysis of HETEs in biopsied vitreous samples from diabetic subjects ( 41 ), and reinforced in postmortem human retinas by additional pathological studies showing a marked increase in the protein levels of both leukocyte and platelet 12/15-LOX in diabetics compared with nondiabetics ( 21 ). Furthermore, it has been demonstrated that 5-or 12-LOX deletion reduces diabetes-induced leukostasis. However, deletion of 5-LOX, but not 12-LOX, reduced capillary degeneration ( 42 ), suggesting that 5-, 12-, and 15-HETE are each required for different stages of DR. Additionally, we have previously demonstrated in a model of OIR that 12/15-LOX and its products are important regulators of retinal NV through modulation of VEGF and pigment epithelium-derived factor expression ( 21 ). In the present study, we provide evidence for the role of the 12/15-LOX pathway in compromising retinal barrier function early during diabetes. Hence, inhibition of 12/15-LOX could potentially benefit the two main hallmark features of DR, i.e., barrier function disruption and pathological NV, which independently can lead to loss of vision among diabetics.
Next, the direct effects of specifi c 12/15-LOX-derived metabolites, 12-or 15-HETE, on retinal vasculature were demonstrated by two complementary approaches. First, the intravitreal injection of 12-HETE engendered many of the features characteristic of early DR, including pro-infl ammatory response and edema. Second, consistent with previous fi ndings ( 43 ), 12-or 15-HETE enhanced several in vitro endothelial cell activities that are relevant to barrier function and angiogenesis, including reduced resistance, adhesion response to PMNs, migration, and tube formation. cytoplasmic subunits, p47phox and p67phox. Our previous studies using animal and tissue culture models have shown that NOX2 is expressed at low levels in normal retinas and in retinal endothelial cells maintained in control conditions, but is substantially increased in retinal vessels of animals with diabetic or ischemic retinopathy and in retinal endothelial cells exposed to HG or hypoxia (23)(24)(25)(26). Furthermore, ROS derived from this enzyme stimulate diverse redox signaling pathways leading to angiogenesis-related gene induction as well as endothelial cell migration and proliferation ( 45 ). Therefore, in gaining more insight into the mechanisms of eicosanoid-induced angiogenesis, we expanded upon these studies to portray the causal relationship between NOX and 12-or 15-HETE. Our in vitro studies demonstrated a couple of notable fi ndings: First, the potential angiogenic activity of 12-or15-HETE in mediating both retinal endothelial cell migration as well as tube formation is regulated in a NOXdependent manner. Second, we broadened the scope to include other prominent features associated with endothelial dysfunction, such as the secretion of infl ammatory cytokines and PMN adhesion. 12-HETE or15-HETE activates retinal endothelial cells toward a pro-infl ammatory phenotype characterized by increased release of IL-6, IL-8, and MCP-1, key regulators of leukocyte recruitment, and the subsequent augmentation of leukocyte adherence through a NOX-dependent mechanism. Together, our results point out that retinal microvascular changes that occur after 12-or 15-HETE challenge are mediated by NOX activity. Figure 9 illustrates the interplay between these pathways and cross-talk with other cellular processes during DR.
While the introduction of anti-VEGF therapies, ranibizumab and afl ibercept, has shifted the treatment paradigm for DR from laser photocoagulation in favor of this targeted approach, there are still major unmet needs and Accordingly, the direct relationship between HRECs and eicosanoids has been explored through in vitro study. Previously, Cheranov et al. ( 44 ) reported that 15-HETE induced migration and tube formation of HRECs through Src-mediated Rac1 activation. Rac1, in nonphagocytic cells, is a component of the multi-subunit enzyme, NOX, which is composed of the catalytic subunits NOX2 (formerly known as gp91phox) as well as p22phox, and the Ϫ / Ϫ mice injected intravitreally with 12-HETE. The relative fl uorescence intensity of FA per mouse retina was calculated by the ImageJ software and normalized as a percentage to 12-HETE-injected WT mice, arbitrarily set at 100%. Data shown are the mean ± SD and representative of four to six mice studied in each group. Fig. 9. Cascade events involved in the pathogenesis of DR: Hyperglycemia activates the PLA2 to release AA from the retinal cell membrane. AA is then converted to 12-or 15-HETE that generates ROS through NOX, creating a status of oxidative stress. This oxidative stress leads to the activation of retinal endothelial cells through various infl ammatory signaling pathways, leading to leukocyte adhesion, hyperpermeability, and ultimately NV (the cardinal signs of DR).
gaps in the understanding of underlying biological processes. The potential long-term consequences of intraocular VEGF suppression on the retina have to be taken into account, because VEGF is a survival factor for retinal neurons ( 46 ). Thereby, targeting a pathway that has crucial roles in compromising the vascular integrity in early DR would provide long-term retinal vascular benefi ts without affecting neighboring neurons. Our current and previous studies ( 22 ) suggest 12/15-LOX as a potential therapeutic target that mediates microvascular dysfunction in DR, probably via a NOX-dependent mechanism. Therefore, inhibition of the activated retinal 12/15-LOX system would allow the occurrence of a baseline level of VEGF to exert its direct neuroprotective effects and thus might provide a therapeutic strategy in modulating pathological pathways of DR when the VEGF level is not correlated with disease severity.