Biochemical Journal

Research article

Novel atypical PKC inhibitors prevent vascular endothelial growth factor-induced blood–retinal barrier dysfunction

Paul M. Titchenell , Cheng-Mao Lin , Jason M. Keil , Jeffrey M. Sundstrom , Charles D. Smith , David A. Antonetti


Pro-inflammatory cytokines and growth factors such as VEGF (vascular endothelial growth factor) contribute to the loss of the BRB (blood–retinal barrier) and subsequent macular oedema in various retinal pathologies. VEGF signalling requires PKCβ [conventional PKC (protein kinase C)] activity; however, PKCβ inhibition only partially prevents VEGF-induced endothelial permeability and does not affect pro-inflammatory cytokine-induced permeability, suggesting the involvement of alternative signalling pathways. In the present study, we provide evidence for the involvement of aPKC (atypical PKC) signalling in VEGF-induced endothelial permeability and identify a novel class of inhibitors of aPKC that prevent BRB breakdown in vivo. Genetic and pharmacological manipulations of aPKC isoforms were used to assess their contribution to endothelial permeability in culture. A chemical library was screened using an in vitro kinase assay to identify novel small-molecule inhibitors, and further medicinal chemistry was performed to delineate a novel pharmacophore. We demonstrate that aPKC isoforms are both sufficient and required for VEGF-induced endothelial permeability. Furthermore, these specific, potent, non-competitive, small-molecule inhibitors prevented VEGF-induced tight junction internalization and retinal endothelial permeability in response to VEGF in both primary culture and in rodent retina. The results of the present study suggest that aPKC inhibition with 2-amino-4-phenyl-thiophene derivatives may be developed to preserve the BRB in retinal diseases such as diabetic retinopathy or uveitis, and the BBB (blood–brain barrier) in the presence of brain tumours.

  • atypical protein kinase C (aPKC)
  • blood–brain barrier (BBB)
  • blood–retinal barrier (BRB)
  • vascular endothelial growth factor (VEGF)


The BBB (blood–brain barrier) and BRB (blood–retinal barrier) require a well-developed TJ (tight junction) complex in the vascular endothelium to create the defined environment required for proper neuronal function. In retinal pathologies, such as AMD (age-related macular degeneration), DME (diabetic macular oedema), ROP (retinopathy of prematurity) and RVO (retinal vein occlusions), breakdown of the vascular endothelial TJ complex leads to vessel hyperpermeability, tissue oedema and loss of neural function [14]. Growth factors, such as VEGF (vascular endothelial growth factor), and pro-inflammatory cytokines, such as TNFα (tumour necrosis factor α), have been implicated in the pathophysiology of these diseases and directly contribute to the retinal vascular hyperpermeability, angiogenesis and inflammation that are clinically observed [59]. Indeed, antibodies against VEGF improve visual function in patients undergoing laser surgery for DME; however, this was only observed in ~50% patients [1012]. These results suggest a need for the identification of downstream regulators of BRB dysfunction, and the elucidation of common mechanisms associated with growth factor and pro-inflammatory cytokine signalling may provide an ideal therapeutic target.

Previous studies have focused on the role of cPKC [classical PKC (protein kinase C)] isoforms, and PKCβ in particular, in regulating VEGF-induced vascular permeability [1315]. Endothelial permeability involves VEGF activation of PKCβ leading to phosphorylation and reorganization of the TJ complex, increasing vessel wall permeability [1618]. Previous studies have demonstrated that VEGF increases the phosphorylation of the TJ protein occludin at multiple sites [19], including Ser490 [20], in a PKCβ-dependent manner [21]. Phosphorylation at Ser490 allows the subsequent ubiquitination and endocytosis of occludin and breaks in the TJ [20,22]. However, the PKCβ inhibitor ruboxistaurin failed to achieve FDA (Food and Drug Administration) approval for diabetic retinopathy, and endothelial cell culture studies demonstrate that inhibition of cPKC isoforms only partially attenuate the VEGF-induced endothelial permeability [19]. Furthermore, cPKC inhibition fails to prevent TNFα-induced permeability [23], a pro-inflammatory cytokine known to contribute to diabetic retinopathy [8,24] and uveitis [25]. Therefore elucidation of novel kinases that regulate vascular permeability downstream of growth factors and inflammatory cytokines may provide the optimal therapeutic targets for macular oedema associated with a wide range of ophthalmic disease.

The PKC serine/threonine kinases are members of the AGC superfamily and may be subdivided into three classes. The cPKCs require calcium and DAG (diacylglycerol) for activation and include PKC α, βI/II and γ. The nPKC (novel PKC) isoforms include δ, ϵ, η and θ and require only DAG for activation, whereas while the aPKCs (atypical PKCs), PKCλ (mouse)/ι (human) and ζ, require neither calcium nor DAG for activation [26]. Activation of PKCζ/ι may occur by phosphorylation of two residues at the C-terminus, Thr410/Thr412 and Thr560/Thr555 for PKCζ/ι respectively [27]. Activation of the PI3K (phosphoinositide 3-kinase) pathway leads to PDK1 (phosphoinositide-dependent kinase 1) activation that can directly phosphorylate Thr410/Thr412 on PKCζ/ι, liberating the pseudosubstrate domain and allowing autophosphorylation of Thr560/Thr555, which results in a fully active kinase [28].

In the present paper we report the requirement of aPKCs in VEGF-induced endothelial permeability. Additionally, we identify a class of small-molecule, non-competitive and specific aPKC inhibitors based on a phenyl-thiophene structural backbone. Altering aPKC isoform activity and content using overexpression and RNAi (RNA interference)-mediated knockdown experiments, coupled with pseudosubstrate peptide inhibition and the use of novel small-molecule inhibitors, demonstrate the requirement for aPKC to mediate VEGF-induced permeability in primary cell culture and the rodent retina. Furthermore, a recent study has demonstrated the effectiveness of aPKC inhibition in preventing TNFα-induced retinal endothelial permeability [23]. Therefore this new class of compounds may lead to the development of novel drugs that preserve the BRB by preventing vascular hyperpermeability in diseases associated with increased VEGF and TNFα expression, including diabetic retinopathy, uveitis and macular degeneration.



Recombinant human VEGF165 was purchased from R&D Systems. PKCζ/ι myristoylated pseudosubstrate inhibitor was purchased from Calbiochem. Small molecules were purchased from ChemBridge or Sigma–Aldrich, or synthesized by Apogee. The Live/Dead® viability kit (Invitrogen) was used to assess cell viability, according to the manufacturer's instructions.

Primary REC (retinal endothelial cell) culture

Primary BRECs (bovine RECs) were isolated and cultured from fresh bovine eye tissue as described previously [29]. HRECs (human RECs) were from Cell Systems. For experimentation, RECs were grown to confluence and stepped down in 0% FBS (fetal bovine serum) or 1% FBS for 24 h with 100 nM hydrocortisone and treated with VEGF at 50 ng/ml where indicated. All experiments were performed with cells at passage 4–8.


Male Sprague–Dawley rats (Charles River Laboratories) weighing 150–175 g were used to evaluate retinal vascular permeability and TJ protein localization. Animals were housed under a 12-h light/dark cycle with free access to water and a standard rat chow. All experiments were conducted in accordance with the ARVO (Association for Research in Vision and Ophthalmology) Statement for the Use of Animals in Ophthalmic and Vision Research, and were approved and monitored by the Institutional Animal Care and Use Committee (IACUC) at the Penn State College of Medicine.

In vitro permeability assay

A permeability assay was performed as described previously [30]. The rate of flux of the substrate, Po, was calculated over the 4 h time course by the following equation [30]: Embedded Image where Po is in cm/s, FL is basolateral fluorescence, FA is apical fluorescence, Δt is the change in time, A is the surface area of the filter and VA is the volume of the basolateral chamber.

Overexpression of aPKC in primary RECs

BRECs were transfected with the aPKC expression plasmid containing the PKCζ isoform (pCMV-aPKCζ), gift from Dr A. Toker (Department of Pathology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, U.S.A.), using the nucleofection technique (Lonza) [19]. Alternatively, BRECs were transduced with adenovirus at 90% confluence on Transwell® filters with AdGFP [adenovirus containing GFP (green fluorescent protein); vector], a gift from Dr S. Abcouwer (Department of Biochemistry and Molecular Biology, University of New Mexico School of Medicine, Albuquerque, NM, U.S.A.), AdWTaPKCζ (wild-type aPKC containing the PKCζ isoform), AdKDaPKCζ (kinase-dead aPKC containing the PKCζ isoform with a K281W mutation), and AdCAaPKCζ (constitutively active aPKC containing the PKCζ isoform with a N-terminal c-Src myristoylation signal), gifts from Dr A. Garcia-Ocaña (Division of Endocrinology and Metabolism, University of Pittsburgh, Pittsburgh, PA, U.S.A.). Adenoviral transduction, which has been shown to successfully transduce primary BRECs [23], was performed at a MOI (mutiplicity of infection) of 10000–20000 to match the expression levels of transgene.

RNAi-mediated knockdown of aPKC isoforms in primary RECs

siRNAs (small interfering RNAs; siGENOME) were designed and purchased from Dharmacon using GenBank® accession numbers NM_001205955.1 (PKCι) and NM_001077833 (PKCζ) and transfected into BRECs at 100 nM using the nucleofection technique (Lonza) [19]. An in vitro permeability assay was performed at 72 h following transfection, according to procedure described above. The oligonucleotides used were: PKCζ Construct A, 5′-CCAUGAAGGUGGUAAAGAA-3′; PKCζ Construct B, 5′-UGUAAUGUCCCGAGGAAUA-3′; PKCι Construct A, 5′-GCAAUGAACACCAGGGAAA-3′; PKCι Construct B, 5′-CUGUAAAAGUCAAUGGUUA-3′; PKCι Construct C, 5′-AGAAAUCAGUCUAGCAUUA-3′; PKCι Construct D, 5′-UCCUUCAAGUCAUGAGAGU-3′; and Non-Targeting #3 siGENOME (Scramble).

aPKC isoform differentiation and profiling

RNA was isolated from BRECs using the RNeasy® kit (Qiagen) and cDNA was generated using the Verso cDNA Kit (Thermo Scientific). PCR was performed using Phusion II High Fidelity Polymerase (New England Biolabs). A 470 bp amplicon was generated with the primers 5′-GATGAGGATATTGACTGG-3′ and 5′-CCTGCCATCATCTC-3′, amplifying cDNA of both PKCι and PKCζ. The restriction enzymes Pst1 (PKCι-specific) and Stu1 (PKCζ-specific) were used to differentiate between the aPKC isoforms.

Cell lysis and immunoblot analysis

Cells were harvested in lysis buffer [20] and protein extracts were blotted using the NuPAGE system (Invitrogen) [31]. Membranes were immunoblotted using anti-FLAG (Cell Signaling Technology), anti-HA (haemagglutinin; Cell Signaling Technology), anti-ERK1/2 (extracellular-signal-regulated kinase 1/2; Cell Signaling Technology), anti-pERK1/2 (phosphorylated ERK1/2; Cell Signaling Technology), anti-pSer473 Akt (Akt phosphorylated on Ser473; Cell Signaling Technology), anti-Akt (Cell Signaling Technology), anti-GFP (Abcam), anti-pThr410/Thr412 PKCζ/ι (PKCζ/ι phosphorylated on Thr410/Thr412; Cell Signaling Technology), anti-pThr560/Thr555 PKCζ/ι (PKCζ/ι phosphorylated on Thr560/Thr555; Abcam), anti-aPKC (C-20 or H-1) (Santa Cruz Biotechnology) and anti-actin (Millipore) antibodies. Primary antibodies were detected by anti-rabbit HRP (horseradish peroxidase)-conjugated IgG with ECL (enhanced chemiluminescence) Advance (GE Healthcare) or anti-mouse alkaline-phosphatase-conjugated IgG with ECF (enhanced chemifluorescence; GE Healthcare).


Male Sprague–Dawley rats were anaesthetized with ketamine and xylazine (66.7 mg and 6.7 mg/kg of body weight, intramuscular injection), and a 32-gauge needle was used to create a hole for an intra-vitreal injection (2.5 μl/eye) using a 5 μl Hamilton syringe. Animals received an intra-vitreal injection of either vehicle (0.1% BSA/PBS) or VEGF (50 ng) for the time indicated. Retinas were excised and lysed in 1% Nonidet P40, 10% glycerol, 50 mM Tris/HCl (pH 7.5), 150 mM NaCl, 2 mM EDTA, 1 mM sodium orthovanadate, 10 mM sodium fluoride, 10 mM sodium pyrophosphate, 1 mM benzamidine and Complete™ protease inhibitor mixture. The lysate was centrifuged at 14000 g for 10 min, and the supernatant was transferred to another microcentrifuge tube. Then 750 μg of protein was subjected to a preclear with 100 μl of a 1:1 slurry of Protein G–Sepharose™ 4 Fast Flow (GE Healthcare) for 1 h. After a brief microcentrifugation, the supernatant was incubated with 5 μg of the anti-aPKC antibody (C-20) for 2 h. Protein G beads were added, followed by further incubation for 1 h. The beads were recovered by centrifugation at 1500 g for 1 min and washed four times with 1 ml of lysis buffer. Proteins bound to Protein G were eluted by boiling in Laemmli buffer for 5 min, and these were then used in the Western blot analysis, as described above.

In vivo permeability assay

Male Sprague–Dawley rats were anaesthetized with ketamine and xylazine (66.7 mg and 6.7 mg/kg of body weight, intramuscular injection), and a 32-gauge needle was used to create a hole for an intra-vitreal injection (2.5 μl/eye) using a 5 μl Hamilton syringe. Animals received an intra-vitreal injection of either vehicle (PBS, 32% ethanol and 1.3% DMSO), VEGF (50 ng), aPKC inhibitor pro-drug (aPKC-I-PD) at 103.8 ng or 259.5 ng to yield an estimated final vitreous concentration of 10 μM or 25 μM, assuming a 30 μl vitreous volume, or a dichloro-substituted aPKC inhibitor (aPKC-I-diCl) at 247.5 ng to yield an estimated vitreous concentration of 25 μM. The experimental groups received both inhibitor and VEGF simultaneously. The animals recovered for 3 h and were then anaesthetized again for the permeability assay. BRB permeability was assessed by measuring retinal Evan's Blue dye accumulation [32].

Library screen for aPKC inhibitors and in vitro HT (high-throughput) kinase assay

The DIVERSet collection of 50000 compounds from Chembridge was screened for aPKC isoform inhibition using recombinant human PKCζ (100 μM) (Enzo Life Sciences) and CREBtide (100 μM) (Enzo Life Sciences) as a substrate. The Kinase-Glo luminescence kit from Promega was used to measure residual ATP following a 3 h incubation. Hits were defined as compounds that inhibited PKCζ activity by at least 50%, and were further characterized in dose–response assays to determine potencies. Further SARs (structure–activity relationships) were performed using the Kinase-Glo luminescence kit with PKCζ (125 ng/ml), ATP (0.1 μM) and CREBtide (25 μM). Specificity profiling was performed by Millipore using a 32P-radiolabelled kinase assay at the Km (app) for ATP.

Enzyme kinetic studies

The ADP Quest assay (Discover Rx) was used to determine the inhibitor mechanism of action. Briefly, for ATP competition, compounds were serially diluted and incubated for 1 h at 30°C with 500 ng/ml PKCζ and 100 μM CREBtide, in the presence of a serial dilution of ATP. Substrate competition was performed using similar conditions with a serial dilution of compound and 250 μM ATP and 500 ng/ml PKCζ, and a serial dilution of CREBtide. ADP formation was measured on a SpectraMax M5 (Molecular Devices) in kinetic mode reading the fluorescence at excitation/emission wavelengths of 530/590 nM every 2.5 min. The signal obtained was converted into a rate ([RFU−RFUcontrol]/time; RFU is relative fluorescence unit) and plotted against the substrate concentration. RFUcontrol is the signal obtained in the absence of kinase at the respective substrate competition. The data were fitted to the Michaelis–Menten equation using Prism software (GraphPad Software) to obtain the Km values: Embedded Image IC50 values were calculated using variable slope sigmoidal dose–response curve. Ki values were derived by plotting the effect of varying substrate concentration on enzyme activity in the presence of various concentrations of inhibitor [I]. The data was fitted using global non-linear regression for non-competitive inhibition with the following equation: Embedded Image Embedded Image

Chemical synthesis

5-Methyl-4-phenyl-2-(substituted-benzoylamino)-thiophene-3-carboxylic acid ethyl ester:substituted-benzoylamino thiophenes were prepared and verified by 1H-NMR and LC (liquid chromatography)–MS using a modified procedure [33].

2-Acetylamino-5-methyl-4-phenyl-thiophene-3-carboxylic acid ethyl ester: 40 mg, 34%; oil; 1H-NMR (500 MHz, [2H]chloroform), δ 0.79–0.82 (t, J 7.5 Hz, 3 H, CH3), 2.15 (s, 3 H,=CCH3), 2.29 (s, 3 H, CH3), 3.95–4.00 (q, J 7.5 Hz, 2 H, OCH2), 7.15–7.16 (d, J 5 Hz, 2 H, Ar-H), 7.31–7.37 (m, 3 H, Ar-H), 11.28 (s, 1 H, NH).

2-(2-Chloro-5-nitro-benzoylamino)-5-methyl-4-phenyl-thiophene-3-carboxylic acid ethyl ester: 138 mg (yellow powder), melting point, 88–90°C; Sample, 1H-NMR (500 MHz, [2H]chloroform), δ, 0.80–0.83 (t, J 7.5 Hz, 3 H, CH3), 2.22 (s, 3 H,=CCH3), 3.97–4.01 (q, J 7.5 Hz, 2 H, OCH2), 7.19–7.20 (d, J 5 Hz, 2 H, Ar-H), 7.34–7.41 (m, 3 H, Ar-H), 7.70–7.72 (d, J 10 Hz, 1 H, Ar-H), 8.31–8.33 (d, J 10 Hz, 1 H, Ar-H), 8.72 (s, 1 H, Ar-H), 12.16 (s, 1 H, NH); MS m/z (relative intensity) 444.31 (M+, 30), 445.31 (15), 446.31 (15).

2-Acetylamino-5-methyl-4-phenyl-thiophene-3-carboxylic acid ethyl ester: 1H-NMR (500 MHz, [2H]chloroform), δ, 2.14 (s, 3 H,=CCH3), 2.31 (s, 3 H, CH3), 7.38–7.40 (m, 2 H, Ar-H), 7.31–7.37 (m, 3 H, Ar-H), 11.13 (s, 1 H, NH); MS m/z (relative intensity) 275.048 (M+, 15), 276.058 (30).

2-Amino-4-(chlorophenyl) thiophene-3-3- carboxylate: MS m/z (relative intensity) 253.38 (M+, 30), 254.38 (15).

2-Acetylamino-4-(4-chloro-phenyl)-thiophene-3-carboxylic acid ethyl ester: MS m/z (relative intensity) 323.48 (M+, 35), 324.48 (20).

2-(2-Chloro-5-nitro-benzoylamino)-4-(4-chloro-phenyl)-thiophene-3-carboxylic acid ethyl ester: MS m/z (relative intensity) 464.38 (M+, 45), 465.38 (10). 2-(2-Chloro-5-fluoro-benzoylamino)-4-(4-chloro-phenyl)-thiophene-3-carboxylic acid ethyl ester: MS m/z (relative intensity) 437.46 (M+, 35), 438.46 (10).

Immunocytochemistry and confocal microscopy

Cells were grown to confluence on plastic coverslips and then serum-starved with Endogro [ZO (zona occludens)-1], without Endogro (occludin) for 24 h. The ommitance of Endogro when staining for occludin is required to maximize occludin staining at the periphery of endothelial cells, which has been demonstrated previously [22]. Cells were treated as indicated and then fixed with 1% paraformaldehyde for 10 min at room temperature (21°C) followed by permeabilization with 0.2% Triton X-100 for ZO-1 or pre-extracted with high-sucrose buffer fixed with ethanol on ice for 30 min for occludin [34]. Following blocking with 10% goat serum (ZO-1) or 10% BSA (occludin) in 0.1% Triton X-100, cells were stained with a rat monoclonal anti-ZO-1 antibody or polyclonal rabbit anti-occludin antibody and fluorescently imaged as described previously [23]. Occludin localization in retinal vessels was assessed by immunohistochemistry in whole retinas, as described previously [17], and displayed as collapsed serial images. Imaging was accomplished using a Leica confocal microscope and imaging software (TCS SP2 AOBS; Leica).

Statistical analysis of data

All studies were performed in duplicate or triplicate and presented either as a compilation of multiple independent experiments or representative of multiple experiments. Unless otherwise stated, statistical analysis was carried out using Prism software from GraphPad using one-way ANOVA with the Neuman–Keuls post-hoc analysis or Student's t test. A P value of less than 0.05 was considered statistically significant. The sample size is indicated in the Figure legends.


VEGF treatment activates aPKC isoforms in rodent retina and primary endothelial cells

In order to investigate whether aPKC isoforms contribute to VEGF-induced retinal permeability in vivo, Sprague–Dawley rats were intravitreally injected with VEGF and their retinas were excised and probed for autophosphorylation of PKCζ/ι at Thr560/Thr555. VEGF induced aPKC autophosphorylation within 15 min and was maximal at 30 min, with an approximately 3-fold increase relative to the sham injection (Figure 1A). The PI3K-dependent priming phosphorylation site, Thr410/Thr412, was probed using a phospho-specific antibody against pThr410/Thr412 PKCζ/ι [28]. VEGF increased phosphorylation of this residue within 15 min (Figure 1B) and returned to basal levels following longer time points. Further mechanistic studies were carried out using primary BRECs to further define the contribution of aPKC signalling to VEGF-induced endothelial permeability. First, conservation of the VEGF-induced activation for aPKC isoforms was determined. BRECs were treated with VEGF for 15 min followed by Western blotting using phospho-specific antibodies. VEGF activates aPKC isoforms, as measured by a 2-fold increase in phosphorylation at Thr410/Thr412 with a more modest, but significant, increase at Thr560/Thr555 (Figure 1C). Robust VEGF intracellular signalling was verified in these cells, demonstrated by a significant increase in phosphorylated ERK1/2 (Figure 1C).

Figure 1 VEGF activates aPKC isoforms in both the rodent retina and in primary RECs

(A) VEGF was intravitreally injected for the indicated time and retinas were excised from Sprague–Dawley rats. aPKC was immunoprecipitated where indicated and immunoblotted for the autophosphorylated residues pThr560/Thr555. (B) VEGF was intravitreally injected for 15 min and retinas were excised from Sprague–Dawley rats. aPKC was immunoprecipitated where indicated and immunoblotted for pThr410/Thr412. Successful immunoprecipiation was verified with immunoblotting for aPKC. For (A) and (B) quantification of the results (means±S.E.M.) from three independent experiments is shown and are expressed relative to the control with a total of n≥8. *P<0.05 and **P<0.01. (C) BRECs were treated with VEGF (50 ng/ml) and lysates were subjected to immunoblotting for pThr410/Thr412 and pThr560/Thr555 PKCζ/ι, aPKC, pERK1/2 and ERK1/2 as described in the Experimental section. Quantification of the results is shown and is expressed as the mean relative to the control. Values are means±S.E.M. n≥8; *P<0.05 and **P<0.01. Ab, antibody; IB, immunoblot; IP, immunoprecipitation; MW, molecular mass.

aPKC isoforms contribute to VEGF-induced retinal endothelial permeability

To determine the role of aPKC isoforms in VEGF-induced permeability, genetic manipulation of aPKCζ expression was performed. An expression plasmid for FLAG-tagged wild-type aPKCζ was transfected into BRECs (Figure 2A), and the cells were grown to confluence on 0.4 μm Transwell® filters. VEGF treatment of control cells increased the permeability of the monolayer to 70 kDa RITC–Dextran 1.5–2.0-fold, an effect that was significantly potentiated with the overexpression of wild-type aPKCζ (Figure 2A). Furthermore, BRECs were transduced with recombinant adenoviruses containing a wild-type PKCζ (AdWTαPKCζ), a kinase-dead PKCζ mutant (AdKDαPKCζ) and a constitutively active mutant of PKCζ (AdCAαPKCζ) (Figure 2B). Overexpression of AdKDαPKCζ completely prevented the VEGF-induced permeability to 70 kDa RITC–Dextran in primary endothelial cells (Figure 2C). Additionally, AdCAαPKCζ alone was sufficient to significantly augment basal permeability in BRECs compared with AdGFP- or AdWTαPKCζ-transduced cells, demonstrating that overexpression of an active aPKC isoform is sufficient to increase permeability in RECs without stimulus (Figure 2D).

Figure 2 aPKC kinase activity is both sufficient and required for VEGF-induced retinal endothelial permeability in primary culture

(A) Primary BRECs were transfected with the plasmid pCMV-FLAG-aPKCζ or empty vector where indicated and grown to confluence on 0.4 μm Transwell® filters. Immunoblot analysis using the anti-FLAG antibody showed successful expression of the transgene. After 24 h serum deprivation, cells on filters were treated with 50 ng/ml VEGF as indicated. Permeability to a 70 kDa RITC–Dextran tracer was measured over a 4 h time course. (B) BRECs were transduced with recombinant adenoviruses and an immunoblot using an HA tag was performed to demonstrate successful transduction. Total aPKC content was monitored to demonstrate the extent of transgene overexpression compared with the control. (C) BRECs were grown on Transwell® filters as described above and infected with AdKDaPKCζ 6 h prior to a 24 h serum deprivation. Permeability to a 70 kDa RITC–Dextran tracer was measured over a 4 h time course following VEGF treatment. (D) BRECs were grown on Transwell® filters as described above and infected with AdWTaPKCζ and AdCAaPKCζ. Permeability to a 70 kDa RITC–Dextran tracer was measured over a 4 h time course following VEGF treatment. All results are expressed as the mean relative to the control with a total of n≥8. Values are means±S.E.M. Average Po values for control and AdGFP were 2.0×10−6 (cm/s) and 1.47×10−6 (cm/s). ***P<0.001 and *P<0.05. IB, immunoblot; MW, molecular mass.

PKCι mediates VEGF-induced permeability in primary RECs

To investigate which aPKC isoforms are specifically expressed in our primary retinal endothelial model, and to avoid the complications associated with differences in primer efficiency, homologous primers were designed to amplify PKCζ and PKCι that contain unique restriction sites within the amplicon to differentiate between the two aPKC isoforms. Following cDNA library generation, these homologous aPKC primers were used to amplify aPKCζ/ι. Restriction digests were performed to identify which aPKC isoforms were expressed and to determine the relative stoichiometric ratio of expression in BRECs. Digestion with Pst1 (PKCι-specific) completely digested the 470 bp amplicon of aPKC and Stu1 (PKCζ-specific) failed to digest this amplicon (Figure 3A), suggesting that PKCι is the primary aPKC expressed in BRECs. Both restriction enzymes were shown to be active against lambda DNA (results not shown). Multiple siRNAs were designed and generated to target the aPKC isoforms in BRECs. Multiple PKCζ-specific siRNAs failed to knockdown aPKC protein content in BRECs, further supporting that PKCι is the primary isoform expressed (Supplementary Figure S1A at Importantly, three siRNA duplexes targeting PKCι significantly decreased aPKC protein content within 72 h (Figure 3B). PKCι Constructs A, C and D resulted in a robust knockdown (~60–80%) in aPKC content (Figure 3B). BRECs were transfected with PKCι Constructs A, C and D and VEGF-induced permeability was assessed. All three constructs prevented the VEGF-induced increase in retinal endothelial permeability (Figure 3C).

Figure 3 PKCι mediates VEGF-induced retinal endothelial permeability in BRECs

(A) PCR was performed using homologous primers for aPKCζ/ι that contained a unique restriction enzyme site to differentiate between the isoforms (Pst1, PKCι-specific) and (Stu1, PKCζ-specific). The amplicon generated was then subjected to restriction enzyme digestion. (B) BRECs were transfected with either 100 nM Scramble or PKCι Constructs A/C/D and subjected to immunoblot analysis for aPKC. Actin served as a loading control. (C) BRECs were transfected with either Scramble or with PKCι Construct A/C/D for 72 h. Permeability to a 70 kDa RITC–Dextran tracer was measured over a 4 h time course following VEGF treatment. The results are expressed as the mean, relative to the control±S.E.M., n≥3. Average Po values for Scramble were 6×10−7 (cm/s) *P<0.05, **P<0.01 and ***P<0.001. IB, immunoblot; MW, molecular mass.

Peptide inhibition of aPKC isoforms prevents VEGF-induced retinal endothelial permeability

To determine whether aPKC kinase activity is essential for the VEGF-induced increase in endothelial permeability, a PS (pseudosubstrate) peptide inhibitor of aPKC isoforms (aPKC-PS) was used to block the kinase activity of the enzyme. This inhibitor is a myristoylated peptide corresponding to the auto-inhibitory PS domain of aPKC isoforms [35]. BRECs were grown to confluence on Transwell® filters and cells were treated with the indicated dose of aPKC-PS for 30 min prior to VEGF treatment. Permeability to 70 kDa RITC–Dextran was measured over 4 h, starting 30 min after the addition of VEGF. The aPKC-PS inhibited VEGF-induced permeability in a dose-responsive manner (Figure 4A). A repeat experiment using 50 nM aPKC inhibitor demonstrated a significant reduction in the VEGF-induced BREC permeability (Figure 4B).

Figure 4 aPKC peptide inhibitor blocks the VEGF-induced increase in endothelial permeability

(A) BRECs were grown to confluence on 0.4 μm Transwell® filters and then stepped-down for 24 h. BRECs were treated for 30 min with the indicated concentration of aPKC peptide inhibitor (aPKC-PS) prior to 30 min treatment with 50 ng/ml VEGF where indicated. Permeability of the monolayer to a 70 kDa RITC–Dextran tracer was measured. n≥4. (B) BRECs were treated with 50 nM aPKC peptide inhibitor for 30 min prior to a 30 min treatment with 50 ng/ml VEGF. n≥4. Permeability was measured as described in (A). Results are expressed as the mean relative to the control ±S.E.M. ***P<0.001 and *P<0.05.

Identification and characterization of small-molecule phenyl-thiophene inhibitors of aPKC isoforms

Although aPKC-PS is a potent inhibitor of aPKC isoforms, it is impractical to use as a drug because of poor bioavailability and pharmacodynamic profile characteristics of PKC peptide inhibitors. Therefore, in order to identify novel small-molecule inhibitors of aPKC for potential therapeutic intervention, a 50000 member chemical library from Chembridge was screened for compounds that inhibit recombinant PKCζ kinase activity in an in vitro assay. Initially a library screen of compounds was performed at a concentration of 100 μM using purified recombinant PKCζ and 25 μM CREBtide as a PKC substrate. The Kinase-Glo luminescence kit was used to measure the residual ATP concentration following incubation for 3 h at room temperature. Hits were defined as the compounds that inhibited recombinant PKCζ activity by at least 50%, and these were further characterized in dose–response assays to determine their potencies and specificities. A total of 14 compounds with IC50 values of 100 μM or less were identified, representing a 0.03% hit rate, and a group of compounds with molecular masses below 500 which showed structural similarity were identified. A number of the compounds contained a phenyl-thiophene core structure; therefore additional screening was focused on phenyl-thiophene derivatives and IC50 values were determined in order to elucidate a pharmacophore (Figure 5). From these studies, three drugs were identified for additional studies, a PD (pro-drug) for its favourable bioavailability profile for in vivo administration (aPKC-I-PD) and two stable derivatives with the lowest IC50 value from this class of compounds, one with a dichloro-substituted phenyl ring (aPKC-I-diCl), and a similar molecule with a dimethoxy-substituted phenyl-thiophene (aPKC-I-diMeO) (Figure 5B). From the list of compounds in Figure 5(A) an initial SAR was performed to identify the core pharmacophore required for in vitro activity. From this analysis it was deduced that a primary amine functionality at the R2 position is required for activity (Figure 5B). aPKC-I-PD possesses an amide bond at this position that probably is susceptible to protease cleavage within a cellular environment. The loss of the hydroxyiminoethyl at R1, coupled with the amide bond cleavage, was determined as necessary for an active compound in vitro.

Figure 5 Identification of novel small-molecule phenyl-thiophene inhibitors of aPKC isoforms

(A) SARs were performed on phenyl-thiophene derivatives and the IC50 was determined using an in vitro luminescence-based kinase assay against PKCζ using 200 ng/ml PKCζ, 25 μM CREBtide substrate and 0.1 μM ATP. IC50 values were calculated using Prism Software with values fitted to a sigmoidal dose–response using a variable slope (n=6). (B) A pharmacophore of a aPKC inhibitor noting essential substitutions, and the structure of aPKC-I-PD, aPKC-I-diCl and aPKC-I-diCl compounds are displayed where indicated. Compounds with Chembridge ID numbers (#) are provided and n/a (not applicable) applies to compounds synthesized in collaboration with Apogee (see the Experimental section). aPKC-I, aPKC inhibitor.

One of the most potent PKCζ inhibitors, aPKC-I-diMeO, was selected for mechanism-of-action studies owing to its improved solubility in aqueous environments. A competition assay was performed to determine the mechanism-of-action for this class of compounds. By measuring ADP formation under increasing ATP concentrations at various doses of inhibitor, it was determined that aPKC-I-diMeO significantly altered Vmax without affecting Km, with a Ki of 7±5 μM (Figure 6A). Furthermore, a similar competition assay was performed against CREBtide, a short peptide PKC substrate, and the peptide substrate also failed to compete with the aPKC-I-diMeO inhibitor and restore the Vmax (Figure 6B). Thus 2-amino-4-phenyl-thiophenes are non-competitive inhibitors of PKCζ.

Figure 6 Mechanism-of-action and specificity characterization of aPKC inhibitors

(A) ATP competition assay using ADP quest to measure initial velocities with 500 ng/ml PKCζ and excess CREBtide substrate or (B) with excess ATP. Ki was determined as described in the Experimental section. (C) IC50 profiling was performed with aPKC-I-diCl and aPKC-I-diMeO using a radiolabelled kinase assay at Km (app) for ATP. (D) aPKC-I-diMeO at 100 μM (10-fold Ki) was screened against 20 kinases of the AGC superfamiliy at the Km (app) for ATP using a radiolabelled kinase assay. (E) BRECs were pretreated for 30 min with 0.3 μM aPKC-I-diMeO and then stimulated with VEGF (50 ng/ml) for 15 min. Lysates were subjected to immunoblotting for pERK1/2, ERK1/2, pSer473-Akt and Akt as described in the Experimental section. IB, immunoblot; MW, molecular mass.

aPKC-I-diCl and aPKC-I-diMeO were screened against other PKC isoforms to determine class specificity using a radiolabelled kinase assay at the Km (app) for ATP. aPKC-I-diCl was 5–10-fold more specific towards the aPKC isoforms compared with the cPKCs (α and β) and over 10–20-fold more specific compared with the novel class (δ and ϵ) (Figure 6C). aPKC-I-diMeO improved the specificity against the cPKC isoforms, having a 25–50-fold lower IC50 compared with α and β, while also maintaining a 25-fold lower IC50 towards the novel class (δ and ϵ) (Figure 6C). These compounds do not exhibit specificity within the aPKC class, which share significant homology with similar IC50 values for PKCι and PKCζ. To determine whether aPKC-I-diMeO possesses significant inhibitor activity towards other kinases, 20 AGC superfamily kinases sharing the most similar sequence homology with PKCs, were screened at 100 μM, ~20-fold the Ki for aPKC isoforms. aPKC-I-diMeO has limited inhibitory activity to these other kinases, with only a modest reduction in two other PKC isoforms tested (Figure 6D). Additionally, the αPKC-I-diMeO does not inhibit cPKC activity in cell culture. ERK phosphorylation, which is downstream of cPKC in endothelial cells, was unaffected by aPKC-I-diMeO treatment (Figure 6E). Furthermore Akt, a downstream mediator of PI3K signalling in endothelial cells, was also unaffected by aPKC-I-diMeO treatment (Figure 6E). Finally, aPKC-I-diMeO inhibited an active kinase fragment devoid of regulatory domains (amino acids 211–592) of recombinant PKCζ as effectively as it inhibited full-length recombinant PKCζ, demonstrating that it acts within this region of the kinase (results not shown). Collectively, these studies demonstrate that phenyl-thiophene derivatives are potent inhibitors of aPKC isoforms with high specificity, and a pharmacophore has been delineated.

Phenyl-thiophenes prevent VEGF-induced vascular endothelial permeability

The effectiveness of these phenyl-thiophenes in preventing VEGF-induced endothelial permeability was determined. Primary BRECs were grown to confluence on 0.4 μm Transwell® filters as above and pretreated with 10 or 25 μM aPKC-I-PD for 30 min prior to treatment with 50 ng/ml VEGF. The permeability of the monolayer to 70 kDa RITC–Dextran was measured and the Po was determined. At both doses the PKCζI-PD was able to completely block the VEGF-induced increase in endothelial permeability with micromolar potency (Figure 7A). Furthermore, aPKC-I-diCl blocked VEGF-induced permeability in BRECs at an approximately 100-fold lower concentration (Figure 7B). The dimethoxy-substituted aPKC-I-diMeO displayed similar potency as the aPKC-I-diCl in its ability to effectively block VEGF-induced endothelial permeability (Figure 7C). Dose–response efficacy curves were determined and aPKC-I-diMeO failed to significantly prevent VEGF-induced permeability below 10 nM (Supplementary Figure S2A at, whereas aPKC-I-PD failed to block VEGF-induced permeability in the nanomolar range (Supplementary Figure S2B). Measures of BREC viability at 24 and 48 h revealed no evidence of cell death after treatment with aPKC-I-PD up to 30 μM, aPKC-I-diCl up to 300 nM or aPKC-I-diMeO up to 300 nM (Supplementary Figure S3 at

Figure 7 Phenyl-thiophene inhibitors of aPKC isoforms prevent VEGF-induced retinal endothelial permeability in primary culture

BRECs were grown to confluence on 0.4 μm Transwell® filters and then serum deprived for 24 h. BRECs were treated for 30 min with (A) aPKC-I-PD, (B) aPKC-I-diCl or (C) aPKC-I-diMeO prior to 30 min treatment with 50 ng/ml VEGF where indicated. Permeability of the monolayer to a 70 kDa RITC–Dextran tracer was measured. Results are expressed as the mean relative to the control with a total of n≥8, ±S.E.M. The average Po value in (A) was 2.43×10−6 (cm/s), (B) was 2.49×10−7 (cm/s) and (C) was 1.46×10−6 (cm/s). ***P<0.001, **P<0.01 and *P<0.05.

To examine further the role for aPKC isoforms on steady-state barrier regulation, a dose–response curve with the phenyl-thiophenes was performed. BRECs were plated on Transwell® filters as above and treated with aPKC-I-PD at doses ranging from 10 to 0.1 μM for 30 min before the addition of the fluorescent tracer. The compound significantly decreased the permeability of the BREC monolayer at a dose as low as 1 μM (Supplementary Figure S4A at This basal effect of reducing permeability could also be observed in HREC monolayers with the aPKC-I-diCl molecule (Supplementary Figure S4B). These data demonstrate that aPKC isoforms play an important role in barrier homoeostasis in endothelial monolayers.

aPKC inhibition prevents disorganization of TJ proteins following VEGF treatment

VEGF treatment rRECs (rat RECs) and retinal vasculature alters the TJ complex and induces internalization of the TJ proteins occludin and ZO-1 [16,18,36]. The ability of the phenyl-thiophene derivatives to prevent the VEGF-induced reduction in TJ border staining was examined. BRECs were grown to confluence on coverslips and pretreated with 10 μM aPKC-I-PD or 100 nM aPKC-I-diCl for 30 min before treatment with 50 ng/ml VEGF for 60 min. Cells were fixed and stained with antibodies against ZO-1 (Figure 8A) or occludin (Figure 8B) and were visualized by confocal miscroscopy. VEGF decreased the border staining and continuity of ZO-1 and occludin labelling at the cell border as predicted. Pretreatment of cells with both aPKC inhibitors blocked the VEGF-induced redistribution of TJ proteins and preserved continuous border staining, suggesting that inhibition of aPKC acts to preserve barrier properties by preventing a net movement of TJ proteins from the plasma membrane into the cytoplasm and preventing the formation of TJ breaks.

Figure 8 Novel small-molecule inhibitors of aPKC isoforms prevent the effect of VEGF on breakdown of the retinal endothelial TJ complex

BRECs were grown to confluence on coverslips and treated with 10 μM aPKC-I-PD or 100 nM aPKC-I-diCl 30 min prior to 50 ng/ml VEGF. Cells were fixed 60 min after VEGF treatment and stained with primary antibodies against (A) ZO-1 or (B) occludin. Cells were imaged with confocal microscopy and are shown as the maximum projection of serial stacks. Images are representative of several similar fields. Scale bar=10 μm.

aPKC inhibtors block the VEGF induction of retinal vascular permeability in vivo

To determine whether aPKC inhibitors block retinal vascular permeability in vivo, we tested the ability of aPKC-I-PD and aPKC-I-diCl to block the VEGF-induced extravasation of Evan's Blue dye in the retina. Sprague–Dawley rats (150–175 g) received 5 μl intravitreal injections of either vehicle, or a final estimated vitreous concentration of 25 μM PKCζI-PD (Figure 9A) or PKCζI-diCl (Figure 9B), 50 ng of VEGF or aPKC inhibitor plus VEGF as indicated. Treatment with VEGF caused an approximate 60–70% increase in accumulation of Evan's Blue in the retina (Figures 9A and 9B). Administration of 25 μM of either aPKC inhibitor prevented the VEGF-induced increase in retinal Evan's Blue dye accumulation (Figure 9A and 9B). Dose-dependence was observed as administration of 10 μM aPKC-I-PD partially blocked BRB breakdown in response to VEGF (Supplementary Figure S5 at

Figure 9 aPKC inhibitors block the VEGF induction of retinal vascular permeability in vivo

Sprague–Dawley rats were injected intravitreally with (A) aPKC-I-PD at 25 μM or (B) aPKC-I-diCl at 25 μM with or without 50 ng of VEGF per eye and compared with vehicle injection. After 3 h, rats received a femoral vein injection of 45 mg of Evans Blue/kg. After 2 h, animals were perfused with citrate/paraformaldehyde buffer for 2 min, retinas were removed, dried and Evans Blue was extracted with formamide. Evans Blue was quantified on a spectrophotometer and normalized to plasma levels measured pre-perfusion. Permeability was calculated and expressed as μl of plasma/g of dry weight per h of circulation. Results are expressed as the mean relative to the control ±S.E.M., n≥8 per group. *P<0.05. (C) Retina flat mounts were prepared as described in the Experimental section and immunostained for occludin. Images are representative of several fields and are shown as the maximum projection of serial stacks obtained from confocal imaging. Scale bar=50 μm.

To determine the effects of aPKC isoform inhibition on the retinal vasculature TJ complex, retinal flat mounts were prepared and immunolabelled with occludin following the intravitreal injection of VEGF and/or aPKC-I-diCl. In the vehicle-injected eyes, occludin border staining was intense and continuous; however, upon VEGF treatment, the immunoreactivity was decreased and a discontinuous border-staining pattern was observed (Figure 9C). Upon aPKC-I-diCl co-injection, occludin border staining was preserved and there were no longer instances of a discontinuous occludin staining (Figure 9C). Of note, functional ERGs (electroretinograms) were performed 5 h after vehicle or aPKC-I-diCl (25 μM) injections to determine whether the small-molecule aPKC inhibitors affected retinal function. There was no statistical difference in either B- or A-wave amplitude in either light- or dark-adapted Long Evan's rats (Supplementary Figure S6 at Moreover, there was no evidence of morphological defects or retinal cell death following aPKC inhibitor injections (Supplementary Figure S7 at These results, combined with the cell-culture studies, demonstrate that inhibiting aPKC in RECs prevents VEGF-induced breaks in the TJ complex and subsequent retinal vascular permeability, without causing measurable retinal toxicity or functional defects.


Macular oedema contributes to the pathophysiology of a number of retinal diseases, including diabetic retinopathy, ischaemic retinopathies and uveitis, among others [37]. It is now well-established that VEGF, a potent vascular permeabilizing agent, contributes to retinal macular oedema, particularly in diabetic and ischaemic retinopathies. Previous research has focused on the contribution of cPKC isoforms downstream of VEGF signalling. Hyperglycaemia and AGEs (advanced glycation end-products) activate cPKC contributing to VEGF expression or release. Downstream of VEGF, cPKC isoforms are implicated in numerous diabetic vascular complications [13]. In particular, PKCβ contributes to diabetes-induced endothelial proliferation and permeability, and oral administration of a specific PKCβ inhibitor, ruboxistaurin, prevents these pathological outcomes [14,15]. Recent clinical trials suggest that ruboxistaurin delays sustained moderate visual loss in diabetic retinopathy [38]. Unfortunately, these inhibitors are only partially effective at blocking VEGF-induced permeability in RECs [19] and have not yet achieved FDA approval as therapeutics. In addition to VEGF, pro-inflammatory cytokines, such as TNFα, also contribute to diabetic retinopathy disease pathogenesis [9], and compounds that prevent both the angiogenic and inflammatory components of retinopathy may prove most effective [39]. Thus the discovery and generation of novel therapies that prevent or improve these vasculopathies are warranted.

In the present paper we report a novel function for aPKC isoforms in regulation of TJ breaks and vascular permeability in response to VEGF. Furthermore, we identify a novel class of aPKC small-molecule inhibitors that prevent VEGF-induced BRB breakdown and describe a pharmacophore for these inhibitors. The novel compounds act as non-competitive inhibitors in respect to both ATP and substrate binding, with high specificity for the aPKC isoforms. Furthermore, no evidence of endothelial cell toxicity or functional retinal defects were observed with inhibitor treatment. Importantly, the inhibitors were effective at completely blocking VEGF-induced 70 kDa RITC–Dextran permeability in cell culture and albumin permeability in the retina in vivo. Previous work from our laboratory demonstrates a role for aPKC in TNFα-induced permeability, which was blocked by the aPKC-I-PD [23]. Taken together, these results demonstrate a novel aPKC inhibitor class that effectively blocks both growth factor- and inflammatory cytokine-induced vascular permeability.

Previous work implicates aPKC isoforms in endothelial barrier breakdown and disassembly following a number of pathological insults. Signalling through aPKC is required for TNFα-induced barrier disruption [23], thrombin-induced endothelial permeability [40,41] and ischaemia-induced BBB dysfunction [42]. The results of the present study demonstrate a necessity for aPKC signalling in VEGF-induced barrier destabilization and vascular permeability. Alternatively, previous data suggests that aPKC isoforms are important for barrier assembly and junctional formation evident from the PKCλ knockout [43]. How aPKC promotes both assembly and disassembly of the junctional complex is not completely understood and probably depends on location, specific signalling pathways and the degree of junctional assembly. A similar role in both pro- and anti-barrier properties exists for the Rho family of GTPases [44,45]. Indeed RhoA promotes barrier destabilization when activated by VEGF [46] or downstream of GEF (guanine-nucleotide-exchange factor)-H1 [47], but spatially restricted activation of RhoA by p114RhoGEF promotes junction formation [48]. The specific downstream mediators and direct substrates of aPKC signalling and their role in barrier function in RECs will be the basis of future investigations. Importantly, the requirement for aPKC isoforms in VEGF-induced endothelial permeability is shown in the present study through the use of multiple genetic and pharmacological manipulations.

The high degree of sequence homology of the aPKC isoforms makes it difficult to determine which isoforms contribute to a specific disease phenotype without the use of genetic loss-of-function experiments. Although data from systemic knockout animals suggests distinct phenotypes, there is emerging evidence that aPKC isoforms may play redundant roles. For instance, aPKC isoforms share a redundant mechanism in insulin-simulated-glucose uptake in adipocyte and muscle cells [49,50]. The results of the present study indicate that the predominate isoform expressed in BRECs is PKCι, where we demonstrated using multiple siRNA oligonucleotide duplexes that knockdown of PKCι is sufficient to prevent VEGF-induced permeability. Furthermore, our novel small-molecule inhibitors inhibit both aPKC isoforms with no degree of specificity. The degree of isoform-specific contribution to retinal vascular permeability in animals will be elucidated in future studies. Importantly, we demonstrated that aPKC isoforms are novel downstream targets of VEGF and novel small-molecule inhibitors of this class of kinases are effective at preventing the deleterious effect of VEGF-induced permeability.

Although the biology and emerging significance of the aPKC isoforms has become apparent, the lack of readily available potent and specific small-molecule inhibitors has hindered both preclinical and clinical studies on this class of kinase. Studies have identified limited small-molecule inhibitors of aPKC; however, some of these compounds lack specificity and potency [5156]. Recent evidence using a class of aPKC inhibitors has been shown in some models of Type II diabetes to be able to correct the metabolic abnormalities of the disease [57]. In the present study, aPKC was demonstrated as necessary for VEGF-induced hyperpermeability and a non-competitive, highly specific aPKC inhibitor pharmacophore was identified. The phenyl-thiophene inhibitor demonstrated a low micromolar Ki with similar potency for aPKC isoforms. These compounds prevent both VEGF and inflammatory cytokine-induced endothelial permeability and may be developed as novel therapies for the treatment of macular oedema resulting from various ocular diseases, and may be useful in treatment of BBB dysfunction during inflammation, cerebral injury and brain tumours.


Paul Titchenell researched data, contributed to the discussion and wrote the paper. Jeffrey Sundstrom contributed to the discussion and reviewed/edited the paper prior to submission. Jason Keil, Cheng-Mao Lin, Charles Smith and David Antonetti researched data, contributed to the discussion and reviewed/edited the paper prior to submission.


This work was supported by the National Institutes of Health [grant number EY012021 (to D.A.A.) and core grants EY07003 and DK020572], the JDRF (Juvenile Diabetes Research Foundation) (to D.A.A.), the Jules and Doris Stein Professorship from Research to Prevent Blindness (to D.A.A.) and the Fight for Sight Research Foundation (to P.M.T.).


We greatly appreciate the gifts of pCMV-FLAG-aPKCζ from Dr Alex Toker, and AdWTaPKCζ, AdKDaPKCζ and AdCAaPKCζ from Dr Adolfo Garcia-Ocaña. We thank all members of the Penn State Retina Research Group for their insights and thoughtful suggestions, and in particular Ellen B. Wolpert (Penn State University) and Edward Felinski (Penn State University) for their technical support and expertise. We would also like to thank Dr Lynn Maines (Apogee) and all of Apogee for their insights and technical support.

Abbreviations: AdCAaPKCζ, constitutively active aPKC containing the PKCζ isoform with a N-terminal c-Src myristoylation signal; AdGFP, adenovirus containing green fluorescent protein; AdKDaPKCζ, kinase-dead aPKC containing the PKCζ isoform with a K281W mutation; AdWTaPKCζ, wildtype aPKC containing the PKCζ isoform; aPKC, atypical protein kinase C; aPKC-I-diCl, dichloro-substituted aPKC inhibitor; aPKC-I-diMeO, dimethoxy-substituted aPKC inhibtor; aPKC-I-PD, aPKC inhibitor pro-drug; BBB, blood–brain barrier; BRB, blood–retinal barrier; BREC, bovine retinal endothelial cell; cPKC, classical protein kinase C; DAG, diacylglycerol; DME, diabetic macular oedema; ERK1/2, extracellular-signal-regulated kinase 1/2; FBS, fetal bovine serum; FDA, Food and Drug Administration; GEF, guanine-nucleotide-exchange factor; GFP, green fluorescent protein; HA, haemagglutinin; HREC, human retinal endothelial cell; p, phosphorylated; PI3K, phosphoinositide 3-kinase; PKC, protein kinase C; PS, pseudosubstrate; REC, retinal endothelial cell; RFU, relative fluorescence unit; RNAi, RNA interference; SAR, structure–activity relationship; siRNA, small interfering RNA; TJ, tight junction; TNFα, tumour necrosis factor α; VEGF, vascular endothelial growth factor; ZO, zona occludens


View Abstract