Plasma Kallikrein-Kinin System as a VEGF-Independent Mediator of Diabetic Macular Edema

Diabetic macular edema (DME) is a leading cause of vision loss among working-aged adults, with a global prevalence of ∼16% for people with history of diabetes mellitus (DM) of ≥20 years (1). This sight-threatening disease is characterized by thickening of the macula due to the accumulation of intraretinal and/or subretinal fluid and exudates, with associated impairment of central visual acuity. The pathogenesis of DME has been primarily attributed to retinal vascular hyperpermeability, which results in extravasation of plasma proteins and lipids into the macula, and increased hydrostatic pressure across the blood-retinal barrier (2). Although DME can occur at any stage of diabetic retinopathy (DR), its incidence is higher in patients with more advanced stages of the disease, including severe nonproliferative DR (NPDR) and proliferative DR (PDR) (3). These observations suggest that underlying retinal pathologies associated with DR, in combination with systemic factors, including hyperglycemia, hypertension, and dyslipidemia, contribute to DME (4).

Vascular endothelial growth factor (VEGF) contributes to the macular thickening and visual impairment associated with DME (5). VEGF concentrations in vitreous and aqueous humor are increased in PDR and DME compared with controls without advanced DR (6–8). Intravitreal injection of therapies that block VEGF actions, including ranibizumab, bevacizumab, and aflibercept, have emerged as effective primary treatments for DME (9). However, most studies to date (10,11) have reported that a significant proportion of DME patients, up to ∼50%, do not fully respond to anti-VEGF therapy.

DR exhibits aspects of inflammation, including vascular hyperpermeability, immune cell recruitment, and elevated expression of proinflammatory cytokines, which have been implicated in the development and progression of retinal pathologies (4). The kallikrein-kinin system (KKS) contributes to the inflammatory response to vascular injury and is a clinically significant mediator of vasogenic edema associated with hereditary angioedema (12,13). Reports from our group and others (14–18) have demonstrated that the KKS mediates vascular hyperpermeability, leukostasis, cytokine production, and retinal thickening in rodent models of DR. The proinflammatory effects of the KKS are mediated by plasma kallikrein (PKal), a serine protease derived from the abundant circulating zymogen plasma prekallikrein (PPK). Zymogen activation by FXIIa cleaves PPK into disulfide-linked heavy and light chains of catalytically active PKal, which in turn cleaves FXII to FXIIa to provide positive feedback and amplification of the KKS. The physiological mechanisms that mediate FXII and PPK activation are not fully understood; however, hemorrhage, activated platelets, and unfolded proteins (19–21) have been shown to activate the KKS. PKal cleaves its substrate high–molecular-weight kininogen (HK) to release the nonapeptide hormone bradykinin (BK). BK activates the BK B2 receptor (B2R) and is also proteolytically processed by carboxypeptidase N into des-Arg⁹-BK (DABK), which is a BK B1 receptor (B1R) agonist (22). Both B1R and B2R are G-protein–coupled receptors that are highly expressed in the retina (23), and retinal expression of B1R is increased in DM (15,16). While the plasma KKS has been implicated in DR (24,25), the levels of KKS components in DME vitreous and mechanisms that mediate the effects of the KKS on retinal vascular permeability (RVP) and retinal thickening are not yet available.

In this report, we quantified KKS components and VEGF and used mass spectrometry–based proteomics to characterize vitreous samples from patients with DME. The effects of the KKS on retinal vascular function and thickness in diabetic rodents were examined using Klkb1, inducible nitric oxide synthase (iNOS), and endothelial NOS (eNOS) knockout mice and measurements of retinal B1R and B2R responses. Our results suggest that upregulation of the intraocular KKS may contribute to a VEGF-independent mechanism of DME.

Human vitreous samples were obtained from patients undergoing pars plana vitrectomy during surgery for macular hole (MH) and various stages of DR, including DME without any PDR findings such as preretinal proliferative membrane or neovascularization, and PDR without vitreous hemorrhage. VEGF levels were measured by ELISA (R&D Systems, Minneapolis, MN). Undiluted vitreous samples were separated by SDS-PAGE (vitreous volumes used: 5, 1, and 1 μL for PPK/PK, FXII/FXIIa, and Kininogen Heavy Chain [HC], respectively) and immunoblotted using primary antibodies against Kallikrein 1B (catalog #NBP1–87711; Novus Biologicals), FXII (anti-FXII HC, catalog #ab1007; Abcam), HK (anti-Kininogen HC [2B5], catalog #sc-23914; Santa Cruz Biotechnology). Results were visualized by enhanced chemiluminescence (20× LumiGLO; Cell Signaling Technology) and quantified using ImageJ, version 1.48 (National Institutes of Health), image analysis software. Western blot analysis of purified standards for PPK, FXII, and HK (Enzyme Research Laboratories) were also performed along with vitreous samples for the quantification. Vitreous proteomics was performed by tandem mass spectrometry, as described previously (26,27). Protein abundance was calculated using total spectral counting for each protein from the X! Tandem search results. For statistical comparisons between groups, samples that did not have two unique peptides for an identified protein were assigned a value of 1, corresponding to half of the minimum detection threshold of two unique peptides. Protein identifications (IDs) were converted to gene symbols using DAVID (28). This study was carried out with approval from the institutional review board and was performed in accordance with the ethical standards of the 1989 Declaration of Helsinki.

DM was induced by intraperitoneal injection of streptozotocin (STZ; Sigma-Aldrich, Milwaukee, WI) in male Sprague-Dawley rats (8 weeks old, Taconic Farms, Hudson, NY). Blood glucose values >250 mg/dL were considered diabetic. Experiments were conducted after 4 weeks of DM and compared with age-matched nondiabetic (NDM) controls.

Klkb1 knockout mice were described previously (29) and backcrossed against C57BL/6 mice for nine generations. Mice deficient in eNOS (B6.129P2-Nos3^tm1Unc/J) and iNOS (B6;129P2-Nos2^tm1Lau/J) and age-matched wild-type (WT) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). DM was induced by intraperitoneal injection of STZ (45 mg/kg) every 24 h for 5 days. Mice with blood glucose levels >250 mg/dL were considered diabetic, and retinal measurements were performed on eNOS^−/− and iNOS^−/− mice at 8 weeks and on Klkb1^−/− mice at 12 weeks of DM duration. Approval for the animal protocols was obtained from the Joslin Diabetes Center Institutional Animal Care and Use Committee. The procedures adhered to the guidelines from The Association for Research in Vision and Ophthalmology for animal use in research.

Intravitreal injections of BK, DABK, and VEGF in rats were performed, and RVP was measured by vitreous fluorophotometry, as described previously (14). Briefly, the eyes were dilated and received intravitreal injection of BK (2 μmol/L; Sigma-Aldrich), DABK (2 μmol/L; Sigma-Aldrich), VEGF (0.02 ng/mL in vitreous) (rhVEGF 165; R&D Systems), or PBS. Ten minutes after the injection, the rats were infused with 10% sodium fluorescein (300 µL/kg) through a jugular vein catheter. After an additional 30 min, vitreous fluorescein levels were measured by vitreous fluorophotometry, as previously described (19). l-N^G-nitro-l-arginine methyl ester (l-NAME) (100 mg/L), a nitric oxide synthesis inhibitor, was administered for 3 days prior to intravitreal injection via the drinking water. B1R antagonist des-Arg¹⁰-Hoe140 and B2R antagonist Hoe140 (AnaSpec, Fremont, CA) were systemically administered via micro-osmotic pump (subcutaneously implanted in the dorsal flank; model 1003D; ALZET) at 1 µg/µL/h, initiated 24 h prior to intravitreal injection. Bevacizumab (12.5 µg/eye) was administered to rats by intravitreal injection in the absence or presence of VEGF or human vitreous. RVP in mice was measured using Evans Blue permeation (30). Mice were anesthetized, received intravitreal injections with 1 µL of 20 μmol/L BK or DABK or PBS, and infused with Evans Blue dye (45 mg/kg) through a catheter in the jugular vein. The dye was allowed to circulate for 1 h prior to sacrifice. After tissue fixation by 10% formalin, the eyes were enucleated and the retinas were extracted. Retinas were incubated with dimethyl formamide overnight at 72°C after drying, and the resultant supernatant was used to determine Evans Blue dye content (14). Results were normalized by retinal tissue weight and levels of the dye in plasma.

Retinal thickness was analyzed using a commercial spectral domain-optical coherence tomography (SD-OCT) system (840SDOCT System; Bioptigen, Durham, NC), as previously described (14). Animals received intravitreal injections, as described above. In addition, a subset of mice received intravitreal injections with VEGF and DC101 (GeneTex). In brief, rectangular volumes of each retina were obtained consisting of 1,000 A-scans by 100 B-scans over a 1.5 × 1.5 mm area centered upon the optic nerve head. Thickness was measured at 500 µm relative to the optic nerve head. A total of four caliper measurements were taken from each site, corresponding to the temporal, nasal, superior, and inferior quadrants of the retina. The measurements from the retinal pigmented epithelium to the nerve fiber layer were averaged to produce a single thickness value for each retina.

Bovine eyes were obtained from a local abattoir, and neonate rats were purchased from Taconic (Germantown, NY). Bovine retinal capillary endothelial cells (BRECs) and astrocytes were isolated as previously reported (31,32). Astrocytes were confirmed by immunocytochemical staining for glial fibrillary acidic protein. Cultured BRECs and ASCs were used between passages 4 to 7 and 3 to 5, respectively. Subconfluent cells were starved with endothelial basal medium (Lonza, Walkersville, MD) for BRECs or DMEM/F-12 for ASCs, containing 1% calf serum and subsequently treated with DMEM containing 25 mmol/L glucose, which were refed with fresh media every 24 h for up to 5 days. Total cell lysates were extracted, and Western blot analysis was performed to detect B1R (Enzo Life Sciences, Farmingdale, NY), B2R (BD Biosciences, San Jose, CA), iNOS (Ab3523; Abcam), eNOS (antiphospho-eNOS [Ser1177], catalog #9571, and anti-eNOS, catalog #9572; Cell Signaling Technology), and GAPDH (Santa Cruz Biotechnology). Band intensities were quantified and expressed as the percentage ratio to GAPDH.

All results were expressed as the mean ± SEM. The statistical significance of differences between groups was analyzed by two-tailed Student t test or Mann-Whitney U test. For in vivo experiments, group conditions were compared by one-way ANOVA using the Fisher least significant differences method for pairwise multiple comparisons. Differences were considered significant at P < 0.05.

Concentrations of KKS components and VEGF were measured in vitreous samples obtained by pars plana vitrectomy from patients with NPDR and DME (NPDR/DME), PDR, and MH (Supplementary Table 1). Within the PDR group, patients with combined PDR and DME versus those with PDR alone were not examined separately. Levels of PPK, FXII, and HK in undiluted vitreous samples were quantified by Western blotting and comparison of band intensities for corresponding purified protein standards (Fig. 1A–E and Supplementary Fig. 1A and B). Western blots of PPK and FXII revealed bands at ∼75 and 50 kDa, which are consistent with the zymogens and HCs of the activated enzymes, respectively (Supplementary Fig. 1A and B). PPK levels were significantly elevated in the vitreous from eyes with DME and PDR by approximately twofold compared with MH (P < 0.001) (Fig. 1B). Activated PKal, measured by HC levels, was increased by 11.0- and 10.4-fold, respectively, in the DME and PDR samples, relative to the average MH level (Fig. 1C). Comparable increases in FXII and HK HC were observed in NPDR/DME and PDR compared with MH vitreous (Fig. 1D and E and Supplementary Fig. 1B). In addition, increased levels of carboxypeptidase N were detected in vitreous from patients with NPDR/DME and PDR compared with MH, and hemoglobin was increased in PDR vitreous (Supplementary Fig. 1C). Concentrations of VEGF in vitreous with NPDR/DME (1,063.1 ± 285.3 pg/mL, P < 0.0001, n = 20) and PDR (1,107.8 ± 260.7 pg/mL, P < 0.0001, n = 24) were elevated compared with VEGF in vitreous from MH, which was below the detection limit for the assay (Fig. 1F).

The correlations among levels of KKS components and VEGF in the vitreous from NPDR/DME and MH subjects were examined. PPK levels correlated with its substrate HK (PPK vs. HK: r = 0.656, P < 0.001; Supplementary Fig. 2A) but not with VEGF (VEGF vs. PPK: r = 0.168, P = 0.499; Fig. 1G). DME vitreous samples were ranked in order of their VEGF concentrations, and the levels of PKal in these samples are shown in Fig. 1H. PKal and VEGF levels in DME vitreous do not correlate (r = 0.266, P = 0.112). Moreover, these results revealed three groups of samples, including samples (ID #6–20) with high PKal and low VEGF, samples (ID #9–10 and 14–3) with elevated levels of both VEGF and PKal, and sample ID #13 with high VEGF and low PKal (Fig. 1H). Similarly, the correlations among the other KKS were higher than their correlations with VEGF (PPK vs. FXII: r = 0.329, P = 0.157; FXII vs. HK: r = 0.592, P = 0.006; VEGF vs. FXII: r = 0.130, P = 0.585; and VEGF vs. HK: r = 0.020, P = 0.934) (Supplementary Fig. 2B–E).

Vitreous proteomic abnormalities have been implicated in DME (19,26). We used proteomics to further characterize the DME vitreous proteome and correlations with PPK and VEGF concentrations. Mass spectrometry–based proteomics was performed on 10 vitreous samples from DME patients and on 5 vitreous samples from MH patients. DME samples used for proteomics are indicated in Fig. 1H. The list of proteins and corresponding numbers of spectral/peptide counts for each sample are shown in Supplementary Table 2. We identified a total of 167 proteins, including 39 proteins detected only in DME vitreous and 10 proteins detected only in MH vitreous (Fig. 2A). This proteome contained 138 proteins that were identified in three or more independent vitreous samples. The numbers of total spectral-peptide matches were used to compare the abundance of these proteins between DME and MH samples. Thirty proteins were increased in DME (fourfold or more, P < 0.001, vs. MH) in vitreous from DME compared with MH samples (Table 1). The concentrations of PKal and VEGF in the 10 DME vitreous samples used for proteomics are indicated in Fig. 1H. Correlation analyses for spectral-peptide counts for these proteins and PPK and VEGF concentrations were performed. This analysis shows that most of the proteins in Table 1 displayed higher correlative R values for comparison with PPK concentrations than with VEGF concentrations (Fig. 2B). In addition, we show that the concentrations of PPK and total protein correlate in DME vitreous (r = 0.618, P = 0.001), whereas VEGF concentration do not correlate with total vitreous protein concentration (r = 0.115, P = 0.499) (Fig. 2C and D). These data reveal marked heterogeneity in the concentrations of KKS components, VEGF, and total protein in vitreous samples from patients with NPDR and DME.

To further characterize the effects of DM on the intraocular KKS, we examined retinal B1R and B2R expression in rats with 4 weeks of STZ-induced DM and age-matched NDM controls. B1R levels were increased by 2.2-fold in DM rats compared with NDM controls (P = 0.001), whereas B2R was similarly expressed in DM and NDM rat retinas (Supplementary Fig. 3A). Next, we examined the effects of extracellular glucose concentration on B1R and B2R levels in cultured BRECs and rat brain astrocytes. B1R expression was increased by twofold to threefold in BRECs and astrocytes cultured in media containing 25 mmol/L glucose compared with cells maintained in media containing 5 mmol/L glucose (P < 0.01, n = 3) (Supplementary Fig. 3B and C). In contrast, incubation of BRECs and astrocytes in the presence of 25 mmol/L glucose did not alter B2R levels.

Since the effects of the KKS on vascular permeability are primarily mediated by the BK system (22), we examined the effects of BK and DABK on RVP and retinal thickness in rats with 4 weeks of DM along with age-matched NDM controls. The effects of intravitreal injection of BK and DABK in DM rats were examined over a time course of 40 min to 48 h (data not shown). BK rapidly increased RVP (at 40 min), and this response gradually decreased at later time points. In contrast, the effect of DABK on RVP was delayed and was highest at 48 h after intravitreal injection. The effects of BK and DABK on RVP were further characterized at 40 min and 48 h, respectively. We observed a small increase in RVP in DM rats compared with NDM rats at 40 min (P < 0.0001) and 48 h (P < 0.0001) after intravitreal injections with PBS vehicle. Intravitreal injection of BK increased RVP after 40 min in both NDM rats (3.6-fold, P = 0.001) and DM rats (2.1-fold, P < 0.0001) compared with PBS-injected controls, and levels returned to baseline levels at 48 h postinjection (Fig. 3A). In contrast, intravitreal injection of DABK did not affect RVP in the short term at 40 min but increased RVP at 48 h postinjection in DM rats (twofold, P < 0.0001), an effect not observed at 48 h in NDM controls. The effects of BK and DABK on RVP in DM rats were blocked by Hoe140 (B2R antagonist) and des-Arg¹⁰-Hoe140 (B1R antagonist) administered by subcutaneously implanted mini-osmotic pumps (Fig. 3B). The effects of intravitreal injection of VEGF on RVP were also investigated. VEGF increased RVP by 2.6-fold compared with vehicle-injected eyes at 40 min (P = 0.004), and this response returned to baseline at 48 h postinjection (Fig. 3C). The effect of VEGF on RVP measured at 40 min postinjection was blocked by coadministration with bevacizumab, but not by combined B1R/B2R antagonism.

The effects of intravitreal injection of BK and DABK on retinal thickness and retinal vessels in DM rats were examined by SD-OCT imaging at 6, 24, and 48 h postinjection compared with baseline. Representative B-scans show retinal structure at 24 and 48 h after BK and DABK injections, respectively, compared with vehicle-injected eyes (Fig. 3D).

Intravitreal injection of saline vehicle produced a small increase in retinal thickness at 24 and 48 h postinjection (5.46% and 5.81% increase, respectively; P < 0.05, n = 5). Intravitreal injection of BK rapidly increased the retinal thickness at 6 and 24 h postinjection (20.1% increase, P < 0.01), and thickness tended to decrease at 48 h but was still thicker than baseline (11.9% increase, P < 0.01). DABK increased retinal thickness at 24 h (13.7%) and 48 h (15.8%) postinjection (Fig. 3D).

The potential effects of VEGF and BK peptides in human vitreous on the RVP were examined in the rat model of RVP described above. We selected two human vitreous DME samples, as indicated in Fig. 1H, patient ID #12, with relatively high PKal and low VEGF (PK↑ vit; PPK 91.6 nmol/L [18.1-fold increase in PKal], VEGF 114 pg/mL) and patient ID #13, with low PKal and high VEGF (VEGF↑ vit; PPK 25.2 nmol/L [2.3-fold increase in PKal], VEGF 2,864 pg/mL), and a control MH vitreous with low levels of both PKal and VEGF. Ten microliters of human vitreous fluid was injected into the vitreous of DM rats in the presence or absence of coinjections with bevacizumab or the combination of Hoe140 and des-Arg¹⁰-Hoe140. At 40 min postinjection, PK↑ vit did not increase RVP, whereas VEGF↑ vit significantly enhanced RVP (1.41-fold, P < 0.05) compared with vitreous from the MH patient. The VEGF↑ vit–induced response was blocked by bevacizumab (P < 0.01) but not affected by Hoe140/des-Arg¹⁰-Hoe140 (Fig. 4A). In contrast, at 48 h, RVP was increased by intravitreal injection of PKal↑ vit (1.6-fold, P < 0.01), which was suppressed by Hoe140/des-Arg¹⁰-Hoe140 (P < 0.01), but not by bevacizumab (P = not significant [NS]). RVP that was increased at 40 min by intravitreal injection of VEGF↑ vit returned to baseline at 48 h postinjection (P = NS vs. MH) (Fig. 4B).

Next, we tested the effects of VEGF receptor 2 (VEGFR2) blockade using DC101, an anti-VEGFR2 neutralizing antibody, on both VEGF- and BK-induced retinal thickening in WT C57BL/6 mice. Retinal thickness was measured by SD-OCT at baseline, and then mice received 1-μL intravitreal injections of VEGF (10 ng) and BK (20 μmol/L), in the absence or presence of DC101. At 24 h postinjection, retinal OCT measurements were repeated and the changes in thickness were determined. VEGF and BK similarly increased retinal thickness at 24 h by 15.2% and 13.6%, respectively (P < 0.001), compared with baseline (Fig. 5). Coinjection of DC101 at doses of 0.36 and 0.73 μg reduced VEGF-induced retinal thickness in a dose-responsive manner to 12.3% (P < 0.001) and 2.3% (P = NS) compared with baseline, respectively. Coinjection of 0.73 μg DC101 did not alter the effects of BK on retinal thickness (12.3%, P < 0.001) compared with baseline. Coinjection of DC101 did not alter retinal thickness in control eyes receiving PBS vehicle.

PPK, encoded by the Klkb1 gene, is the major source of proinflammatory levels of BK (22). Previous studies (14) have shown that PKal inhibition reduces RVP in diabetic rats. To further evaluate the role of PPK in DR, we examined the effects of Klkb1 deficiency on RVP and retinal thickness in WT and Klkb1^−/− mice with 3 months of DM compared with NDM controls. Klkb1 deficiency did not alter body weight or blood glucose levels for DM or NDM groups, compared with age-matched WT littermate controls (Supplementary Fig. 4A and B). RVP, quantified using Evans Blue dye permeation, was increased in DM female (400 ± 63%, P < 0.001) and male (138 ± 31%, P < 0.001) WT mice compared with NDM WT controls (Fig. 6A and B). NDM WT and NDM Klkb1^−/− mice of the same sex displayed similar RVP; however, we observed increased basal RVP in NDM male mice compared with NDM female mice. RVP in female DM Klkb1^−/− mice was 78 ± 6% (P < 0.001) less than that in female DM WT mice (Fig. 6A). Similarly, the DM-induced increase in RVP in male Klkb1^−/− mice was 58 ± 11% (P = 0.033) less than that in male DM WT mice (Fig. 6B).

Previous studies have reported that DM causes retinal neurodegeneration in mice (33). Using SD-OCT, we observed 3.4 ± 0.6% and 5.8 ± 0.8% decreases (P < 0.01) in retinal thickness in DM WT mice compared with NDM WT mice for female and male mice, respectively (Fig. 6C and D). The DM-associated decrease in retinal thickness in female Klkb1^−/− mice was 1.6 ± 0.6%, which was thicker than that observed in female WT mice (204.9 ± 1.3 vs. 198.9 ± 1.2 µm, P = 0.002). In contrast, retinal thicknesses in male DM Klkb1^−/− and DM WT mice were similar.

Increased nitric oxide level contributes to the action and vascular permeability of BK in DR (17,34,35); however, the role of eNOS and iNOS in the effects of the KKS on the retina are not yet available. We examined the role of nitric oxide in BK- and DABK-induced RVP using l-NAME, which is hydrolyzed to N^G-nitro-l-arginine, a competitive inhibitor of eNOS, iNOS, and neuronal nitric oxide synthase. Systemic administration of l-NAME in DM rats blocked both the short-term effect of BK and the delayed effect of DABK on RVP (Fig. 7A). The effects of BK and DABK on eNOS phosphorylation were examined using BRECs. The addition of BK to the culture media increased eNOS phosphorylation at Ser1177 within 5 min, and maximum phosphorylation was observed after 30 min. In contrast, the incubation of BRECs with DABK did not alter eNOS phosphorylation at this site (Fig. 7B). The potential role of eNOS on BK-induced RVP was examined in WT and eNOS knockout mice in the absence and presence of 8 weeks of DM. In WT mice, BK increased RVP by 2.1-fold and 1.7-fold at 40 min after intravitreal injection compared with vehicle-injected eyes (P < 0.01) in NDM and DM mice, respectively, whereas intravitreal injection of BK did not alter RVP in eNOS-deficient mice (with/without DM) (Fig. 7C).

Since DABK effects on RVP were blocked by l-NAME but did not alter eNOS phosphorylation (Fig. 7A and B), we examined the effects of DABK on iNOS. Intravitreal injection of DABK increased iNOS expression in DM rat retinas compared with vehicle-injected eyes. This effect of DABK on iNOS was not observed in NDM rats (Fig. 7D). Furthermore, we examined the effect of DABK on iNOS protein levels in astrocytes in vitro. DABK induced increased expression of iNOS in astrocytes after 36 and 48 h of stimulation (P < 0.01 vs. untreated controls) (Fig. 7E). To examine the potential role of iNOS on DABK-induced RVP, we examined the effects of intravitreal injection of DABK in DM iNOS knockout mice. Intravitreal injection of DABK increased RVP at 48 h postinjection in WT mice (2.0-fold vs. vehicle-treated WT mice, P < 0.01), whereas DABK did not alter RVP in iNOS-deficient mice (Fig. 7F).

In this report, we show that the KKS is activated in vitreous from DME patients. Vitreous from subjects with relatively high concentrations of PKal and low concentrations of VEGF induce RVP via the BK pathway, whereas VEGF mediates this response for vitreous with high VEGF concentrations and low PKal concentrations. These data demonstrate heterogeneity among a cohort of DME subjects in regard to vitreous protein composition and activity of vitreous fluid. Since KKS components are abundant plasma proteins, and PPK levels correlate with other abundant plasma proteins in the vitreous, it is likely that the increases in DME vitreous are due to plasma protein extravasation as a result of blood-retina barrier dysfunction. Indeed, the proteins that we identified to be increased in DME vitreous (Table 1) have been previously identified in human plasma (http://www.plasmaproteomedatabase.org) (36). The concentrations of these proteins in normal plasma have been estimated by Farrah et al. (37) using spectral counts and are listed in Table 1. We show that PPK, but not VEGF, concentrations correlate with most vitreous protein increases identified by proteomics and total vitreous protein concentration. Moreover, PKal concentrations are elevated in most DME vitreous samples, whereas VEGF displays a greater range of concentrations among these vitreous samples. Increased intraocular concentrations of VEGF have been attributed largely to retinal ischemia/hypoxia, and there is a known wide range of VEGF concentrations among DME vitreous samples (7,38). The wide variability among vitreous VEGF concentrations and the extent of retinal edema imply that additional factors may play important roles in mediating some cases of DME (9,19). Increased RVP and plasma protein extravasation in DR has been attributed to a variety of mechanisms, which are not limited to VEGF (2). Intravitreal injection of KKS components increases RVP and retinal thickening in diabetic animals (14,16,19). While BK generated from HK is a primary effector pathway of the KKS, the DME vitreous also contains components of the complement and coagulations systems, such as complement C3, thrombin, and plasminogen, which may contribute to proinflammatory effects downstream of PKal and FXIIa (39). The temporal events and relative contributions of the VEGF and PKal pathways in DME may depend on the underlying retinal pathologies, such as levels of retinal ischemic, and the extent of vascular injury and inflammation. We propose that KKS extravasation in DR is mediated by multiple factors and that high levels of intraocular KKS activity exacerbates RVP, which overwhelms fluid resorptive mechanisms, resulting in retinal thickening. Moreover, our findings suggest that increased intraocular PKal may contribute to an incomplete or delayed response to anti-VEGF therapies for patients in whom intravitreal KKS activity is high. Additional VEGF-independent pathways and retinal structural changes such as disorganization of the retinal inner layers are also likely to contribute to the range of clinical responses observed after anti-VEGF treatment of DME (40–42).

Elevated nitric oxide production has been implicated in the pathogenesis of DR via multiple mechanisms (43). Mice deficient in eNOS display increased iNOS expression, nitric oxide levels, and DR compared with diabetic WT controls (35). Mice with iNOS deficiency are protected from DM-induced RVP and retinal capillary degeneration (44,45). B1R and B2R are expressed in vascular, glial, and neuronal cells, and the mechanisms that couple the KKS to nitric oxide production differ at the cellular level. BK stimulates eNOS phosphorylation in BRECs, whereas the B1R agonist DABK induces iNOS expression in astrocytes. Astrocytes are distributed on the surface of the inner retina, and retinal capillaries are surrounded by the astrocyte foot process, which can swell and release inflammatory cytokines that increase blood-retina barrier permeability (46). These findings suggest that the KKS may exert its effects on RVP by inducing nitric oxide production in both glial and vascular cells. We show that BK similarly induces RVP in DM and NDM rats, suggesting that B2R may contribute to RVP in the presence of KKS activation in the vitreous. In addition, we show that B1R levels and DABK responses in the retina are increased in diabetic rats compared with NDM controls. These findings are consistent with those of previous reports (15,16,47) showing that B1R mRNA expression and responses are increased in DM and inflammation. In DM, proinflammatory cytokines may contribute to increased DABK action via upregulation of B1R and iNOS expression, which could provide additive effects to the B2R/eNOS pathway on RVP (34). Although selective inhibition of B1R could potentially provide some protective effect for retinal dysfunction in DR, this approach would not address the potential contributions of B2R or effects mediated by PKal substrates other than HK (18,29). However, the blockade of PKal could reduce the proinflammatory effects of the intravitreal KKS while preserving the potential beneficial ocular effects of BK generated by tissue kallikrein (Klk1) (48).

Taken together, our results suggest that the PKal-mediated KKS exerts effects on DR at two levels. In the early stages of DR, a low level of KKS activation increases RVP via both retinal B1R and B2R. This DM-induced vascular hyperpermeability is blocked by Klkb1 deficiency and the administration of either a PKal inhibitor (14) or B1R antagonists (15,17). Although this early increase in RVP may contribute to the pathogenesis of DR, it is not sufficient to cause retinal thickening. The transition to retinal edema, including DME, appears to require further upregulation of vascular permeability factors, such VEGF and PKal, which are markedly increased in the vitreous during advanced stages of DR compared with DM alone (6). Since many individuals are not fully responsive to anti-VEGF DME therapy (10), PKal inhibition in these patients may provide an opportunity to further reduce macular thickness and improve vision.

In summary, we show that the KKS is increased in human DME vitreous, that Klkb1 deficiency in mice decreases DM-induced RVP, and that the effects of the BK on retinal thickening do not require VEGFR2. Both B1R and B2R contribute to the effects of the KKS on retinal thickening and RVP, and these responses are mediated by eNOS and iNOS, respectively. These findings suggest that the intraocular KKS may contribute to DME via a VEGF-independent mechanism.

Acknowledgments. The authors thank Leilei Sun and Xiaohong Chen (Joslin Diabetes Center, Boston, MA) for their assistance with the proteomic analyses.

Funding. This project was supported in part by JDRF grant 17-2011-251; the Massachusetts Lions Eye Research Fund; National Institutes of Health Awards EY-019029, DK-36836, and T32-EY-007145; and Japan Society for the Promotion of Science Postdoctoral Fellowships for Research Abroad.

Duality of Interest. L.P.A. and E.P.F. are cofounders of KalVista Pharmaceuticals Ltd. The Joslin Diabetes Center holds U.S. patent 8,658,685 on methods for the treatment of kallikrein-related disorders. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. T.K. designed and performed the experiments and assisted in writing the manuscript. A.C.C., N.M., and Q.Z. performed the experiments and edited the manuscript. K.F. contributed to the collection of vitreous samples. T.I. supervised the clinical work. L.P.A. contributed to the study design, reviewed the data, and edited the manuscript. E.P.F. conceived and designed the overall study, reviewed the data, and wrote the manuscript. E.P.F. 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.