Differences in activation of intracellular signaling in primary human retinal endothelial cells between isoforms of VEGFA 165.

Purpose: There are reports that a b-isoform of vascular endothelial growth factor-A 165 (VEGFA165b) is predominant in normal human vitreous, switching to the a-isoform (VEGFA165a) in the vitreous of some diseased eyes. Although these isoforms appear to have a different ability to activate the VEGF receptor 2 (VEGFR2) in various endothelial cells, the nature of their ability to activate intracellular signaling pathways is not fully characterized, especially in retinal endothelial cells. We determined their activation potential for two key intracellular signaling pathways (MAPK, AKT) over complete dose-response curves and compared potential effects on the expression of several VEGFA165 target genes in primary human retinal microvascular endothelial cells (HRMECs).
Methods: To determine full dose-response curves for the activation of MAPK (ERK1/2), AKT, and VEGFR2, direct in-cell western assays were developed using primary HRMECs. Potential differences in dose-response effects on gene expression markers related to endothelial cell and leukocyte adhesion (ICAM1, VCAM1, and SELE) and tight junctions (CLDN5 and OCLN) were tested with quantitative PCR.
Results: Activation dose-response analysis revealed much stronger activation of MAPK, AKT, and VEGFR2 by the a-isoform at lower doses. MAPK activation in primary HRMECs displayed a sigmoidal dose-response to a range of VEGFA 165 a concentrations spanning 10-250 pM, which shifted higher into the 100-5,000 pM range with VEGFA 165 b. Similar maximum activation of MAPK was achieved by both isoforms at high concentrations. Maximum activation of AKT by VEGFA 165 b was only half of the maximum activation from VEGFA 165 a. At a lower intermediate dose, where VEGFA 165 a activated intracellular signaling stronger than VEGFA 165 b, the changes in VEGFA target gene expression were generally greater with VEGFA 165 a. Conclusions: In primary HRMECs, VEGFA 165 a could maximally activate MAPK and AKT at lower concentrations where VEGFA 165 b had relatively little effect. The timing for maximum activation of MAPK was similar for the isoforms, which is different from that reported for non-retinal endothelial cells. Although differences in VEGFA 165 a and VEGFA 165 b are limited to the sequence of their six C-terminal six amino acids, this results in a large difference in their ability to activate at least two key intracellular signaling pathways and VEGF-target gene expression in primary human retinal endothelial cells.

Introduction

Vascular endothelial growth factor-A 165 (VEGFA) is the isoform of VEGF that is primarily responsible for driving retinal vascular pathology in diabetic retinopathy (DR), age-related macular degeneration (AMD), and retinopathy of prematurity (ROP), through activation of VEGF receptor 2 (VEGFR2). A seminal analysis of the vitreous fluids of patients with diabetic retinopathy established the presence of elevated VEGF concentrations associated with this and other conditions [1]. That study contributed to the eventual development of intravitreal VEGF-blocking drugs to treat neovascularization and edema in AMD and DR, and the current exploration of their use for ROP [3-5].

Blockade of VEGFA activity is provided by the use of intravitreal drugs, including ranibizumab (Lucentis, Genentech), bevacizumab (Avastin, Genentech), pegaptanib (Macugen, Bausch & Lomb), and aflibercept (Eylea, Regeneron). Some VEGF-blocking drugs injected into the vitreous can enter the systemic circulation, and in the case of bevicizumab and aflibercept, they can substantially decrease serum VEGF concentration for several days [6]. There is interest in advancing VEGF-regulating therapies through more precise titration of the VEGF concentration or modulating specific VEGF-mediated signaling rather than the full blockade provided by current drugs. This field will continue to benefit from more complete knowledge of VEGF’s signaling mechanisms within retinal endothelial cells. Unfortunately, much previous VEGF research did not technically differentiate between the VEGFA isoforms, and in the case of the retina, there is far less investigation using retinal endothelial cells, and even less from the human retina.

In addition to endothelial cells, VEGF receptors are expressed by cells of the immune system, including early and late hematopoietic progenitor cells, dendritic cells, T-lymphocytes, and macrophages [7]. Among the three VEGF receptor tyrosine-kinases (R1, R2, R3), VEGFR2 is required for angiogenesis and vasculogenesis, and its expression is most abundant in endothelial and endothelial progenitor cells [8-10]. VEGFR2 binds several isoforms of VEGFA that are produced from alternative splicing [11-14]. The most frequently detected isoforms are VEGFA, VEGFA, and VEGFA, with 121, 165, and 189 amino acids, respectively. VEGFA is most diffusible, lacking the heparin-binding domains of VEGFA and VEGFA. VEGFA has even more affinity for heparin than VEGFA with an additional heparin-binding domain encoded by exon 6 [11,12]. VEGFA165 binds as a dimer to activate VEGFR2, and binds neuropilin 1 as a coreceptor. VEGFA lacks two regions required to bind neuropilin 1.

The vast knowledge of VEGF-mediated signaling is mostly derived from the study of non-retinal cell types, and is beyond the scope of this paper; however, readers can refer to the excellent review by Koch and Claesson-Welch [15]. In endothelial cells, several pathways are activated by VEGFA165 that affect proliferation, migration, survival, and endothelial permeability. Two of these pathways include MAPK (ERK1/2) and AKT (PKB), where MAPK is a dominant regulator of cell proliferation, and AKT modulates cell survival and permeability [16-19]. Autophosphorylation of VEGFR2 (Y1175) leads to RAS-independent activation of the phospholipase-C𝛾 (PLC𝛾)/protein kinase C/MAPK pathway [20,21]. Activation of VEGFR2 also leads to activation of the TSAD/phosphotidylinositol-3 kinase (PI3K)/phosphoinositide-dependent protein kinase (PDK)/AKT pathway [22,23]. AKT phosphorylates the BCL-2-associated death protein (BAB) and caspase-9 to block apoptosis and increase EC survival [24]. AKT also activates endothelial nitric oxide synthase eNOS (NOS3; Gene ID 4846, OMIM 163729) to modulate vascular permeability [25].

Recently, b-isoforms of VEGFA were described that differ from the previously known a-isoforms in the sequence of their C-terminal six amino acids: CDKPRR in VEGFAa becomes SLTRKD in VEGFAb [13,14]. This difference renders VEGFAb incapable of bringing neuropilin 1 into a receptor–ligand complex with VEGFR2 [26-28]. Similar to VEGFA, VEGFAb is reported to be less angiogenic than VEGFAa [13,29,30]. VEGFAb’s binding affinity for VEGFR2 itself is similar to that of VEGFAa in human umbilical vein endothelial cells (HUVECs) and less than VEGFAa in direct binding analysis in transformed human embryonic kidney cells 293 (HEK293) cell assays [31,32]. VEGFAb also displays weaker activation of VEGFR2 and ERK1/2 (MAPK) than VEGFAa in transfected porcine aortic endothelial (PAE) cells [31]. Perrin et al. reported a switch from mostly VEGFAb to mostly VEGFAa in the vitreous of patients with active diabetic retinopathy [33]. A similar shift, from VEGFA165b to VEGFA165a, was reported in vitreous humor from young patients with ROP [34]. A murine equivalent of isoform switching (VEGFAb to VEGFAa) was also reported in a mouse model of oxygen-induced retinopathy [35]. Several publications using isoform-specific antibodies have detected VEGFAb in human and mouse tissues, but their presence from RNA analysis in sheep hypothalamus is controversial [36].

Although single-dose studies using transfected cells or non-retinal endothelial cells indicate that VEGFAb activates VEGFR2 and MAPK less than VEGFAa, we do not know if this is true throughout a full dose range. We also wanted to determine full dose–response curves for activation of intracellular signaling in primary human retinal microvascular endothelial cells. We hypothesized that there is a significant difference in the full dose–response curves for activation of these pathways in human retinal endothelial cells between VEGFAa and VEGFAb. To address this, we used in situ assays to determined full dose–response curves for the activation of MAPK (ERK1/2), AKT, and VEGFR2 in primary HRMECs for VEGFAa and VEGFAb. The results demonstrate that there is a significant dose–response difference in the ability of these two isoforms to activate intracellular signaling kinases, and that these dose–response differences propagate to differences in VEGF–target gene expression in primary human retinal endothelial cells.

Methods

Certification of primary human retinal microvascular endothelial cells

Primary human retinal microvascular endothelial cells (HRMECs, genotype XY) were obtained from Cell Systems (Kirkland, WA) as passage 3 cells (catalog number: ACBRI-181). Cells used for experimentation were not used past passage 7. The endothelial character of the cell line was established by Cell Systems immunofluorescence testing: >95% positive by immunofluorescence for cytoplasmic VWF/Factor VIII, cytoplasmic uptake of Di-I-Ac-LDL, cytoplasmic CD31, and <1% by immunofluorescence for glial fibrillary acidic protein (GFAP), glutamine synthetase, neural/glial antigen 2 (NG2), and platelet-derived growth factor receptor beta (PDGFR-beta). The cell line was also subjected to short tandem repeat (STR) profile testing by cell systems at the master level (P1), performed by the laboratory analysis service of the American Type Culture Collection (ATCC, Manassas, VA). Seventeen STR loci plus the gender-determining locus, amelogenin, were amplified using the commercially available PowerPlex® 18D Kit from Promega. The cell line sample was processed using the ABI Prism® 3500xl Genetic Analyzer. Data were analyzed using the GeneMapper® IDX v1.2 software (Applied Biosystems). Appropriate positive and negative controls were used throughout the test procedure. The DNA profile (STR #12625 Analysis) was as follows: TH01: 9, 9.3; D5S818: 9, 13; D13S317: 8, 10; D7S820: 11, 12; D16S539: 11, 12; CSF1PO: 10, 12; Amelogenin: X, Y; vWA: 16, 19; TPOX: 8, 11. The ATCC test conclusions were that the submitted sample profile is human, but not a match to any profile in the ATCC STR database, as would be expected for primary HRMECs. Additionally, routine testing for retinal endothelial character at the author’s laboratory (Eye Research Institute, Oakland University) confirmed that the HRMECs express all three of the human Norrin receptor-complex genes FZD4 (Gene ID 8322, OMIM 604579), LRP5 (Gene ID 4041, OMIM 603506), and TSPAN12 (Gene ID 23,554, OMIM 613138), using real-time PCR analysis (data not shown). The HRMECs also demonstrated VEGF-mediated regulation of several VEGF target genes, as shown in the results.

Cell culture and antibodies

The EndoGRO-MV Complete Media Kit for culture of microvascular endothelial cells without VEGF was obtained from Millipore (Burlington, MA). EndoGro basal medium was supplemented with rhEGF, L-glutamine, heparin sulfate, and ascorbic acid according to the kit instructions. Additionally, supplementation with fetal bovine serum (FBS) and hydrocortisone hemisuccinate (1 µg/ml) was dependent upon the assay requirements. Recombinant human VEGFA165a and VEGFA165b were obtained from R&D Systems (Minneapolis, MN). Odyssey Blocking Buffer and an Odyssey Infrared Imager were purchased from LI-COR Biosciences (Lincoln, NE). The antibodies used for in situ labeling of HRMECs (in-cell western, ICW) and western blotting are summarized in Table 1. Mini Protean TGX gels were purchased from Bio-Rad (Hercules, CA).

Table 1

Specific antibodies used and sources.

Antibody	IgG Species	Company	Cat. No.
Phospho-AKT (P-Ser473)	Rabbit	Cell Signaling Technologies	4058
AKT(pan)	Rabbit	Cell Signaling Technologies	4691
Phospho-p44/42 MAPK (Thr202/204)	Rabbit	Cell Signaling Technologies	4370
p44/42 MAPK (pan)	Rabbit	Cell Signaling Technologies	4695
Phospho-VEGFR2 (Tyr1175)	Rabbit	Cell Signaling Technologies	2478
Beta-Actin	Mouse	Millipore	MAB1501
Anti-mouse IgG IRDye 800CW	Goat	LI-COR	32,210
Anti-rabbit IgG IRDye 680RD	Goat	LI-COR	68,071
Anti-rabbit IgG IRDye 800CW	Goat	LI-COR	32,211
Anti-mouse IgG IRDye 680RD	Goat	LI-COR	68,070

Immunoblotting of activated MAPK and AKT in HRMECs

HRMECs were grown in 100 mm dishes that had been precoated with attachment factor (Cell Systems, Kirkland, WA). The medium was EndoGRO (No VEGF, Millipore Sigma, Burlington, MA) with 5% FBS to establish confluency. When the cells were confluent, the media were replaced with fresh media with or without VEGFA isoforms. After incubation for 10 min, the cells were washed with ice-cold PBS (1X; 9.8 mM Na2HPO4, 2.5 mM NaH2PO4, 137 mM NaCl, 2.7 mM KCl, pH 7.4, Fisher Scientific, Pittsburg PA), and the dishes were scraped to detach the cells. The cell suspension was centrifuged at 1,000 × g for 5 min, and the cells were reconstituted in radioimmunoprecipitation assay (RIPA) cell-lysis buffer (150 mM NaCl, 1% Triton-X-100, 0.1% sodium dodecyl sulfate (SDS), 50 mM Tris, pH 8.0, 10 mM sodium fluoride, 1 mM sodium orthovanadate, and complete protease inhibitor cocktail (1 tablet/10 ml)). The cells were sonicated or vortexed every 10 min while kept on ice for 30 min. The cell lysate was collected after centrifuging at 14,000 × g for 15 min at 4 °C. The protein concentration was measured using Pierce BCA Protein Assay (Thermo Fisher Scientific, Waltham MA). The samples were prepared with Laemmli sample buffer and loaded onto (4–15%) gradient gels for SDS–PAGE (SDS–PAGE) electrophoresis. After electrophoresis, the gels were equilibrated in cold transfer buffer and transferred to polyvinylidene difluoride (PVDF) membranes overnight. The membranes were blocked with Odyssey Blocking Buffer and incubated with appropriate rabbit primary antibodies (Table 1) along with mouse monoclonal actin antibody, which was used for normalization. After washing with PBS, 0.1% Tween-20 (4 × 5 min), the membranes were incubated with secondary antibodies (goat anti-rabbit IRDye 680RD and goat anti-mouse IRDye 800CW) for 30–60 min. The membranes were washed and scanned on an Odyssey Infrared imager (LI-COR).

Dose–response analysis of intracellular signaling in primary HRMECs

HRMECs were seeded into black 96-well plates (5,000/well) that had been coated with attachment factor. The cells were grown to confluence using fully supplemented EndoGRO-MV media for 4–5 days. The cells were serum starved overnight using EndoGRO-MV without hydrocortisone. The cells were treated with VEGFAa or VEGFAb for various lengths of time after which the treatments were immediately removed and replaced with 4% paraformaldehyde. They were fixed for 20 min at room temperature followed by permeabilization with PBS, 0.1% Triton X-100 (10 min). The cells were blocked by incubation with Odyssey Blocking Buffer (Li-Cor) for 1.5 h at room temperature and incubated with primary antibodies (1:200) for either 2 h at room temperature or overnight at 4 °C. Rabbit antibodies were used against the proteins of interest, and a mouse monoclonal anti-beta-actin antibody was used for normalization. The cells were washed with PBS, 0.1% Tween-20 (5 × 5 min) and incubated with secondary antibodies, goat anti-rabbit IRDye 800CW and goat anti-mouse IRDye 680RD (1:750) for 45 min at room temperature. After washing with PBS 0.1% Tween-20 (5 × 5 min), the plates were scanned on an Odyssey Imager (Li-Cor). For time–response and dose–response experiments, the doses of VEGFAa and VEGFAb were assayed in quadruplicate wells, and dose–response experiments were repeated three times to confirm reproducibility of relative dose–response curves for activation of MAPK (phospho-Thr202/Tyr204), AKT (phospho-Ser473), and VEGFR2 (phospho-Tyr1175). Data were fit to a four-parameter log-logistic response function Equation (1) at the 95% confidence level for each dose x using the Drc package for R [37]. The parameters fit were b, the steepness of the curve at e the median effective dose (ED50), with c and d the lower and upper limits of the response. The 95% confidence level for fitting was used to produce curves:

Quantitative PCR analysis of HRMEC gene expression

Primary HRMECs were grown to confluence in six-well plates. After the desired treatment, the cells were trypsinized. Total RNA was isolated using the Absolutely RNA Miniprep kit (Agilent, Santa Clara, CA). The cells were homogenized in 20 µl of lysis buffer. Homogenization was accomplished using conical pellet pestles in 1.5 ml microfuge tubes, with a handheld rotary tool (Bel-Art, Wayne, NJ). First-strand cDNA was synthesized by reverse transcribing 500 ng of total RNA per sample using either the AffinityScript qPCR DNA Synthesis kit (Agilent) or the LunaScript RT SuperMix Kit (NEB # E3010(S/L), Ipswich, MA) with Oligo-dT priming. The reaction conditions were according to the manufacturer’s instructions. For AffinityScript: 25 °C for 5 min, 42 °C for 20 min, 95 °C for 5 min, and 10 °C for 10 min. For LunaScript RT: 25 °C for 2 min, 55 °C for 10 min, and 95 °C for 1 min. All compared samples were processed using the same reagent set. Stock first-strand cDNA preparations were stored at −70 °C and were not used for analysis after a maximum of three freeze–thaws. The duplex reaction format was used with FAM-labeled probe and primer pairs for the gene of interest and VIC-labeled probe and primer-limited pairs for tata-binding protein (TBP, Gene ID 6908; OMIM 60075) as the normalizer gene (ThermoFisher, Waltham, MA). For real-time PCR reactions, sample first-strand cDNA was diluted fivefold with deionized water, and 2 µl added to 18 µl of Master Mix for 20 µl PCR reactions. Triplicate reactions were used for each sample. Master Mix chemistries were either 2X Gene Expression Master Mix (ThermoFisher, Applied Biosystems, Waltham, MA), or the Luna Universal Probe qPCR Master Mix (2X) Gene Expression (New England BioLabs # M3004L, Ipswich, MA), both mixes with the Rox reference dye option. Reactions were run on either an Mx3000P real-time PCR system using MxPro software or an AriaMx Real-time PCR System using the AriaMx HRM QPCR Software (Agilent). Gene expression assays were evaluated for high PCR efficiency using a dilution series of HRMEC cDNA to ensure validity of using the delta-delta Ct method for comparing relative gene expression. Each replicate reaction was internally normalized relative to endogenous TBP gene expression. The specific assay probe sets used for gene expression analysis are listed in Table 2.

Table 2

Taqman gene expression probes for HRMEC gene expression.

Gene	Probe Set #	Spans Exons
CLDN5	Hs01561351_m1	1–2
ICAM1	Hs00164932_m1	2–3
OCLN	Hs00170162_m1	5–6
SELE	Hs00174057_m1	4–5
TBP	Hs00427620_m1	3–4
VCAM1	Hs01003372_m1	6–7

Results

Preliminary observations

The cell-based experimental results reported here were inspired by observations made during preliminary testing for a different project to develop an intravitreal injection model using VEGFAa to cause dilation of primary retinal veins, which was reported to occur in the Long Evans rat [38]. That model typically involves a high dose of VEGFA. Using fluoresceine angiography, and SD-OCT imaging, it was noted that similar effects on the vasculature could be induced with VEGFAb or VEGFAa. Although the observational studies did not use large numbers of rats, they are included as supplemental observations (Appendix 1) because they brought to our attention that the dose–responses for intracellular pathway activation were not fully known in primary human retinal endothelial cells.

Activation of MAPK and AKT in primary HRMECs by VEGFA165a and VEGFA165b

To examine the effect of VEGF isoforms on intracellular signaling pathways, we chose primary HRMECs and initially used immunoblotting to test activation of the MAPK and AKT pathways by both isoforms. These pathways were examined as two of the intracellular signaling kinases implicated in mediating the effects of VEGFA on endothelial proliferation and the blood–retinal barrier. Under a high dose (100 ng/ml, 5,300 pM) for maximum activation, both isoforms of VEGFA increased the amounts of the active form of MAPK (Figure 1A) and the active form of AKT (Figure 1B). From three separate full experimental repeats, the average activation was about 2.7-fold and 1.6-fold (Figure 1C) for MAPK and AKT, respectively. Although the treatments with VEGFAa and VEGFAb always increased the activation of both kinases, the magnitude of the increase varied between experiments as illustrated by the comparison of two of the three experiments on the same gel (Figure 1A,B).

Figure 1

MAPK and AKT are activated in HRMECs by VEGFA165a and VEGFA165b. A: Immunoblots of total MAPK and activated MAPK (phospho-MAPK) 10 min after treatment with VEGFAa (Va), VEGFAb (Vb), or media control (C). Two representative experiments are shown. B: Immunoblots of total AKT and activated AKT (phospho-AKT) 10 min after treatment with VEGFAa (Va), VEGFAb (Vb), or media control (C). Two representative experiments are shown. C: Fold activation of MAPK and AKT from three immunoblotting experiments. P values are shown, t test relative to media controls (onefold).

Dose–response for the activation of MAPK by VEGFA165a and VEGFA165b

Although repeated immunoblotting experiments confirmed activation of MAPK and AKT by both VEGFA isoforms, the experimental variability introduced by protein extraction, protein assays, electrotransfer, band shape, and image analysis was judged to be unsuitable for dose–response analysis. To address the variability, new in situ assays, or ICW assays, were developed using the same kinase-specific antibodies while removing the sample processing required for immunoblotting. This also permitted faster fixation of cells at the desired time, the use of many doses, and the use of several biologic replicates per dose.

The maximum activation of the MAPK (ERK1/2) pathway in primary HRMECs occurred at 10 min for treatment with VEGFAa and VEGFAb, and returned to control levels by 90 min, using a dose of 2 ng/ml (1,050 pM). See Figure 2A. The activation of VEFGAa was statistically significantly stronger than that of VEGFAb at 10 min (t test, p<0.01).

Figure 2

VEGFA165a strongly activates MAPK in primary HRMECs at doses where VEGFA165b has little effect. Activation of MAPK was measured with in situ immunofluorescence with a phosphospecific antibody (phospho-Thr202/Tyr204) for p44/42. A: Activation of MAPK at three time points after the addition of 1,000 pM VEGFAa or VEGFAb. Bar shows standard deviation. (t test, ***p<0.001 relative to untreated control, n=8 biologic replicates.) B: Dose–response curves for activation of MAPK. The median effective dose (ED50) was 73 pM for VEGFAa and 1,015 pM for VEGFAb. Bars indicate the 95% confidence interval for data fit to the four-parameter log-logistic function. n=4 biologic replicates per dose.

We next performed dose–response analysis for the activation of the MAPK at 10 min with cells preadapted to lower serum. Eleven doses of each isoform, with four biologic replicates per dose, were evaluated, and the dose–response data for MAPK activation were fit to the log-logistic four-parameter dose–response function using the Drc package in R [37] (Figure 2B). Activation of MAPK by VEGFAb showed a typical sigmoidal dose–response. Compared to VEGFAb, the dose–response curve for VEGFAa displayed a strong allosteric shift to a binary-like activation response. VEGFAa was more potent for activation of MAPK at lower doses. A similar maximum level of MAPK activation could be achieved with the highest dose tested (10,000 pM) using either isoform. Activation of MAPK by VEGFAb was only 10% of that generated by VEGFAa using a dose of 250 pM. The ED50 was 73 pM for VEGFAa and 1015 pM for VEGFAb for the experiment shown. The experiment was repeated three times confirming the large difference in ED50 between the two isoforms. Differences between the ED50 values were typically in the 800 to 1,000 pM range.

Dose–response for the activation of AKT activation by VEGFA165a and VEGFA165b

Maximum activation of the AKT pathway in HRMECs occurred 30 min after treatment with VEGFAa and 15 min after treatment with VEGFAb (Figure 3A) using what was thought to be a maximum activating dose (100 ng/ml, 5,300 pM). Dose response analysis was performed for AKT activation using these respective maximum time points, and the data were fit to the four-parameter log-logistic response function (Figure 3B). With VEGFAa, AKT activation in HRMECs displayed a much steeper dose–response curve than for VEGFAb. The ED50 values were 53 pM for VEGFAa compared to 126 pM for VEGFAb. The maximum activation of AKT generated by the b-isoform was less than half that obtained from treatment with the a-isoform. Repetition of the experiment confirmed similar relative responses.

Figure 3

VEGFA165b is a weaker activator of AKT in primary HRMECs compared to VEGFA165a. A: Activation of AKT was measured with in situ immunofluorescence with a phosphospecific antibody for Ser473-AKT at five time points after the addition of 20 ng/ml (1,050 pM) VEGFAa or VEGFAb. Bar shows standard deviation. (t test, ***p<0.001 relative to untreated control, n=4 biologic replicates.) B: Dose–response curves for activation of AKT. The median effective dose (ED50) value for AKT activation by VEGFAa was 53 pM compared to 126 pM for VEGFAb. Bars indicate the 95% confidence interval for data fit to the four-parameter log-logistic function. n=4 biologic replicates per dose.

Dose–response for the activation of VEGFR2 by VEGFA165a and VEGFA165b

To test for relative differences in dose–response at the level of the receptor itself, we examined the dose–response for activation of VEGFR2. Maximum activation of the receptor was fast, within less than several minutes, as quickly as cells could be processed for treatment and fixation. The 5 and 10 min experiments were similar in relative response. Therefore, 10 min were used for the dose–response experiments, and data were fit to the four-parameter log-logistic response function as shown in Figure 4. The VEGFAa treatment achieved maximum activation by about 500 pM, whereas VEGFAb caused little if any activation at that dose. The ED50 for activation by VEGFAa was 254 pM compared to 1,192 pM for the VEGFAb. Furthermore, the maximum activation by VEGFAb was 73% of that obtained with VEGFAa. The relative pattern was confirmed with a repeated experiment (data not shown).

Figure 4

Differences in the activation of intracellular signaling between VEGFA165a and VEGFA165b in primary HRMECs begin at the receptor VEGFR2. VEGFAa was a stronger activator of VEGFR2 compared to VEGFAb. The median effective dose (ED50) for activation by VEGFAa was 254 pM compared to 1,192 pM for VEGFAb. Bars indicate the 95% confidence interval for data fit to the four-parameter log-logistic function. n=4 biologic replicates per dose.

Isoform effects on expression leukocyte-docking protein genes

We first examined gene expression from 1 to 24 h after treatment with high doses (100 ng/ml, 5,300 pM) of VEGFAa and VEGFAb to determine the optimal time point for monitoring any effects on the expression of ICAM1 (Gene ID 3383; OMIM 147840), SELE (Gene ID 6401; OMIM 131210), VACAM1 (Gene ID 7412; OMIM 192225). A dose of 5,300 pM was likely a saturating concentration and more than what would be experienced in vivo, to find a suitable time point when gene expression could be subsequently compared using physiologically relevant doses. All three of these genes displayed increased expression after treatment with both isoforms of VEGFA (Figure 5A,C,E). These trends were confirmed with repeated experiments (data not shown). Of the genes examined, the increasing expression trend of ICAM1 was more variable in timing, not always maximum at 6 h, sometimes later by 24 h (data not shown).

Figure 5

VEGFA165a has a stronger effect on leukocyte–endothelial cell adhesion gene expression in primary HRMECs than VEGFA165b. Relative gene expression was measured with quantitative PCR (qPCR). A: Expression of the ICAM1 gene at 1, 3, 6, and 24 h after treatment with VEGFA. Confluent human retinal microvascular endothelial cells (HRMECs) were treated with a saturating high dose of VEGFAa or VEGFAb (100 ng/ml, 5,300 pM). B: Expression of the ICAM1 gene comparing VEGFAa and VEGFAb at low, intermediate, and high doses measured after 6 h. C: Expression of the SELE gene at 1, 3, 6, and 24 h after treatment with VEGFA. D: Expression of the SELE gene comparing VEGFA165a and VEGFA165b at low, intermediate, and high doses measured after 6 h. E: Expression of the VCAM1 gene at 1, 3, 6, and 24 h after treatment with VEGFA. F: Expression of the VCAM1 gene comparing VEGFAa and VEGFAb at low, intermediate, and high doses measured after 6 h. (Triplicate assays, error bars show standard deviation. ANOVA (ANOVA): compared to media control *p<0.05, **p<0.01; compared to VEGFAa †p<0.05, ††p<0.01).

Selecting a fixed treatment time of 6 h, and guided by the activation dose–response curves for MAPK and AKT, we tested low (2 ng/ml, 100 pM), intermediate (19 ng/ml, 1,000 pM), and high (95 ng/ml, 5000 pM) doses to test for differences in the effects on leukocyte-docking gene expression. For all three leukocyte-docking protein genes (ICAM1, SELE, and VCAM1), VEGFAa increased their expression at the intermediate dose of 1,000 pM, where VEGFAb still had very little effect (Figure 5D,F,G).

Isoform effects on expression of tight junction protein genes

The effects of VEGFAa and VEGFAb on the expression of the two key tight junction protein genes, CLDN5 (Gene ID 7122; OMIM 608102), OCLN (Gene ID 100,506,658; OMIM 602876), were also compared. These genes encode the tight junction proteins CLDN5 and OCLN; see Figure 6. Both isoforms of VEGFA reduced the expression of these genes by the 6-h time point (Figure 6A,C). We next compared low (2 ng/ml, 100 pM), intermediate (19 ng/ml, 1,000 pM), and high saturating doses (95 ng/ml, 5,000 pM) at the fixed 6 h time point. VEGFAb was less effective at suppressing the expression of CLDN5 and OCLN compared to VEGFAa (Figure 6D,F).

Figure 6

VEGFA165a has a stronger effect on the expression of tight-junction-protein genes in primary HRMECs than VEGFA165b. Relative gene expression measured with quantitative PCR (qPCR). A: Expression of the CLDN5 gene at 1, 3, 6, and 24 h after treatment with VEGFA. Confluent human retinal microvascular endothelial cells (HRMECs) were treated with a saturating high dose of VEGFAa or VEGFAb (100 ng/ml, 5,300 pM). B: Expression of the CLDN5 gene comparing VEGFA165a and VEGFA165b at low, intermediate, and high doses measured after 6 h. C: Expression of the OCLN gene at 1, 3, 6, and 24 h after treatment with VEGFA. D: Expression of the OCLN gene comparing VEGFAa and VEGFAb at low, intermediate, and high doses measured after 6 h. (Triplicate assays, error bars show standard deviation. ANOVA (ANOVA): compared to media control *p<0.05, **p<0.01; compared to VEGFAa †† p<0.01).

Discussion

Although the role of VEGFA in vascular development, tumorgenesis, and hypoxia has been investigated for almost three decades, less is known about the functional differences between VEGFA isoforms. Recently, isoform-specific analysis reported that most VEGFA in normal vitreous was VEGFAb, with the ratio changing to mostly VEGFAa in eyes with active diabetic retinopathy and ROP [33,34]. The relative expression of VEGFAa and VEGFAb can be regulated by alternative splicing of the VEGFA pre-mRNA, which is regulated by phosphorylation of the ASF/SF2 splicing factor by SR-protein kinase (SRPK1/2) [39]. This raised the possibility that modulation of this isoform ratio might be exploited for therapeutic purposes. Modulation of the splicing between the VEGFAa and VEGFAb mRNAs was demonstrated using SRPK inhibitors in human RPE cells, and topical infusion of these inhibitors reduced neovascularization in a mouse laser-induced choroidal neovascularization (CNV) model [40]. Another group used intravitreal injection of a different SRPK inhibitor to inhibit neovascularization in the mouse CNV model [41]. An antibody specific for the C-terminus of VEGFAa, which does not bind VEGFAb, can also inhibit the proangiogenic effects of VEGFAa [42].

Specific blockade of the VEGFAa protein, or suppressing VEGFAa expression in favor of VEGFAb, could be a therapeutic strategy if there are significant differences in the ability of these isoforms to activate intracellular signaling in the retinal endothelium. The preference for examining signaling differences using primary HRMECs is supported by studies that revealed differences between endothelial cells derived from different organs. For example, viral delivery of murine recombinant-VEGFb attenuated inflammatory-response damage in a murine model of ulcerative colitis, yet it exacerbated damage in the blood–brain barrier in a model of focal cerebral ischemia [43-45]. Extensive gene expression profiling has also demonstrated that HUVECs have greater expression of embryonic genes, and other differences, compared to endothelial cells from the choroid and neural retina. Significant differences also exist between human retinal endothelial cells and human choroidal endothelial cells in the expression of genes that are related to endothelial function and neovascularization [46].

We began with immunoblotting to examine the activation of MAPK and AKT in primary HRMECs. Although VEGFAb does not bind the neuropilin 1 coreceptor, it does bind to VEGFR2 with similar or less affinity as VEGFAa depending on the cell model and method used [31,32]. Regardless of relative binding affinity, VEGFAb is reported to bind and activate VEGFR2 based on immunoblotting studies [31,47,48]. Using immunoblotting, we found that MAPK and AKT were activated to a similar extent by VEGFAa or VEGFAb, using a dose of 100 ng/ml (5,300 pM) in primary HRMECs. Previous single-dose immunoblotting studies in different cell types reported that VEGFAb activated VEGFR2 and MAPK less than VEGFAa, and that the maximum activation of MAPK by VEGFAb occurred later than VEGFAa in PAE cells and bovine aortic endothelial cells using a dose of 2,600 pM [31,47]. Another study of human pulmonary microvascular endothelial cells (HPMECs) found that the a-isoform was a stronger activator of VEGFR2 and MAPK (ERK1/2) than the b-isoform at a dose of 20 ng/ml (1,050 pM) [48].

We found that MAPK and AKT were activated less by VEGFAb at the lower dose (20 ng/ml, 1,000 pM; data not shown). However, it was noted that although activation was detected in repeated experiments, the relative magnitude of activation was variable using immunoblotting. Immunoblotting involves numerous processing steps, and a substantial number of cells are required for a single sample; therefore, we decided that in situ assays would be superior for generating full dose–response curve data.

To determine the dose–response for activation of MAPK and AKT, in situ ICW assays were developed that permitted the use of multiple replicate doses with a limited supply of primary cells in a 96-well plate format. The resulting dose–response curves confirmed that there were substantial differences in the activation of MAPK and AKT between VEGFAa and VEGFAb in primary HRMECs. In the case of MAPK, the ED of VEGFAa was 900 pM less than the ED50 for VEGFAb. VEGFAa caused near maximum activation of MAPK at a concentration of 250 pM, while VEGFAb had little effect at 250 pM. Higher doses of VEGFAb could activate MAPK to a similar maximum level, and the timing for maximum activation of MAPK was similar for both isoforms in HRMECs, occurring by 10 min and decreasing by 30 min. This relative timing was different from that reported for HUVECs where the maximum concentration of active MAPK occurred 30 min after treatment with VEGFA165b [31].

For AKT, the difference in ED between VEGFAa and VEGFAb was not as large as seen for MAPK, just twofold greater for VEGFAb compared to VEGFAa. However, the activation of AKT by VEGFAb was substantially less than from VEGFAa over the entire dose–response range. In contrast to MAPK activation, the time to maximum activation of AKT was earlier with VEGFAb (15 min) compared to the VEGFAa (30 min). Although the differences in the ED50 values were not as large as seen for MAPK activation, the maximum activation of AKT obtained with VEGFAb was only 50% of the maximum activation generated by VEGFAa. This was another difference from that reported in a non-retinal endothelial cell type, human pulmonary endothelial cells (HPECs), where AKT activation is the same for VEGFAa and VEGFAb at a dose of 20 ng/ml (1,050 pM) in HPECs [48].

The differences noted above in the activation of MAPK and AKT between HRMECs and non-retinal endothelial cells suggest that VEGF-mediated intracellular signaling patterns are not completely universal between endothelial cells from different organs. An evaluation of primary endothelial cells from the human retina should be used whenever possible to study the retinal context. Altogether the dose–response curves for the activation of VEGFR2, MAPK, and AKT in primary HRMECs were consistent with the model proposed by Whitaker et al., in which VEGFAa activates VEGFR2 more intensely when it can bind the coreceptor neuropilin 1, to form a larger activation complex [49].

Based on the primary HRMEC dose–response data for MAPK activation, VEGFAa was more potent for activating intracellular signaling than VEGFAb at lower doses, which fall into a range of total VEGFAa concentrations previously reported in human vitreous. Vitreous concentrations for VEGFA have been measured with various techniques in several disease conditions. The most elevated VEGFA is the VEGFA isoform, which diffuses slowly, and the vitreous has protease activity; therefore, it is possible that the concentrations measured in the vitreous are somewhat lower than in the neural retina itself. With that stated, vitreal concentrations of 25 pM to more than 400 pM have been reported for proliferative diabetic retinopathy by several studies [1,50-54]. Other studies have reported 10 pM in diabetic macular edema [55], 100 to 450 pM in retinopathy of prematurity [34,56] and an average of 430 pM, with as high as 580 pM, in central retinal vein occlusion [57]. It is possible that elevations of retinal VEGFAa concentration to several hundred picomolars could increase the activation of endothelial intracellular signaling above its normal baseline. In contrast, the much lower activation potential of VEGFAb in this concentration range would leave MAPK activation close to normal if most of the VEGFA existed as the b-isoform.

We also found that VEGFAb could affect the expression of the same target genes that are also regulated by VEGFAa in primary HRMECs. These included genes involved in leukocyte–endothelial cell adhesion and the formation of tight junctions. Although the 5,000 pM dose of VEGFAb could activate MAPK and AKT more strongly than the intermediate 1,000 pM dose, it is unlikely that the 5,000 pM concentration would be experienced in vivo. For that reason, we suggest that the effects on gene expression at 1,000 pM were most informative. We found that the CLDN5 gene was particularly susceptible to repression by both isoforms of VEGFA compared to OCLN in this cell type. VEGFA-mediated activation of the VEGFR2 receptor is known to disrupt adherin junctions and causes β-catenin-mediated repression of the expression of the CLDN5 gene in endothelial cells [58]. For all three leukocyte-docking protein genes (ICAM1, SELE, VCAM1), VEGFAa increased their expression at the intermediate dose of 1,000 pM, where VEGFAb had less effect. We concluded that there were substantial dose–response differences in the regulation of VEGFA target genes, with VEGFA165a proving to be the stronger effector of gene expression overall.

Conclusion

The results indicated that differences in the last six amino acids between VEGFAa and VEGFAb have a significant effect on the relative activation of MAPK and AKT within primary HRMECs. VEGFAa activated both pathways at lower concentrations where VEGFb had little effect. The greater potential of VEGFAa to affect HRMECs also extended to changes in the expression of genes required for leukocyte-docking and tight-junction structure. Large dose–response differences in activation of these pathways exist in ranges of elevated VEGFAa concentrations that have been reported in the vitreous of diseased eyes. Key aspects of MAPK and AKT activation, such as timing and maximum activation, were different in HRMECs compared to those reported for some other non-retinal endothelial cell types. These results would support the concept that specific blockade of VEGFA165a, or modulating the VEGFA165b/VEGFA165a ratio, could be useful therapeutic strategies for retinal diseases involving elevated VEGFA concentrations.