Molecular, Cellular, and Functional Heterogeneity of Retinal and Choroidal Endothelial Cells

Abstract

Purpose: To investigate the endothelial heterogeneity across distinct vascular beds in the inner and outer blood–retinal barriers.

Methods: We evaluated the molecular, cellular, and functional differences between primary human retinal endothelial cells (HRECs) and human choroidal endothelial cells (HCECs) in terms of angiogenic and vasculogenic properties, permeability, and transcytosis. Tube formation assay, cell migration assay, in vitro permeability assay, microfluidic sprouting assay, and transcriptome analysis were performed.

Results: HRECs showed higher proliferation and migration activity than did HCECs, whereas the tube formation ability was similar between HRECs and HCECs. Under angiogenic stimuli, HCECs displayed earlier sprouting angiogenesis, but the overall speed was faster and more stable in HRECs. HRECs expressed higher levels of adherens junctional proteins, whereas the tight junctional genes and transcytosis-related genes were more highly expressed in HCECs. Angiopoietin-2 was predominantly expressed in HRECs, but vascular endothelial growth factor (VEGF) receptors were more strongly expressed in HCECs. Platelet-derived growth factor subunit B (PDGFB) was more highly expressed in HRECs, which correlates to the lower degree of pericyte coverage in choroidal blood vessels.

Conclusions: Retinal and choroidal ECs showed significant cellular and molecular heterogeneities that correlated with their functional characteristics. Retinal ECs are vasculogenic with high migratory characteristics and faster angiogenic sprouting, and they are more responsive to VEGF-induced permeability. In contrast, choroidal ECs express high levels of transcytosis genes, and they are vasculogenic, rather proliferative, adept in generating tip cells, and less responsive to VEGF-induced permeability.

Microvascular endothelial cells (ECs) show organ specificity and tissue heterogeneity.¹ Moreover, even in the same organ, ECs show morphological and functional differences according to the location. As an example, the posterior parts of the eye have two distinct types of vascular beds—the retinal vascular network^2,3 and the choroidal blood vessels.^3,4 The retina and choroid are intact and separated by the thin collagenous Bruch membrane. The retina capillary ECs form continuous endothelium,⁵ and the choriocapillaris ECs form fenestrated endothelium with less pericyte coverage, which results in a higher rate of transvascular transport than that in retinal ECs.⁶ Each of these two heterogeneous capillaries has a distinct role in the two different blood–retina barrier (BRB) components: inner BRB and outer BRB.^6,7 

The retinal and choroidal blood vessels present different morphological and functional features. During embryonic development, the choroid is derived from different cell lines than the retina, which is derived from the neural ectoderm.⁶ Whereas retinal blood vessels are surrounded by neuroglial tissues, choroid vessels are supported by collagenous and elastic connective tissues that contain melanocytes, fibroblasts, and resident immunocompetent cells.⁶ Retinal capillaries form a trilaminar capillary plexus, but choriocapillaris is a highly anastomosed patch-like network of capillaries. The velocity of red blood cells in choriocapillaris is only about 77% of that in the retinal capillaries.⁸ 

The abnormal proliferation of retinal and choroidal blood vessels gives rise to vision-threatening diseases, including diabetic retinopathy and age-related macular degeneration. Whereas new retinal vessels respond well to anti-angiogenic treatments, choroidal neovascularization tends to mature without regression and persist despite the same treatment.⁹ Our group recently reported that perivascular coverage is different along the retinal and choroidal capillary, which is associated with differences in the changes of capillary diameter with aging and under diabetic conditions.¹⁰ Although the two vascular beds have been found to have distinct structural and functional features, there has been limited research on how the retinal and choroidal ECs differ at the cellular level. Here, we investigated the endothelial heterogeneity between retinal and choroidal ECs in terms of molecular, cellular, and functional characteristics. 

Methods

Acquisition of Human Donor Eyes

Human choroids and retinas were harvested from donor eyes that were left unused in corneal transplantation surgeries performed at the Asan Medical Center (Seoul, Korea). The study was approved by the Institutional Review Board of the Asan Medical Center (IRB 2014-0665). Eyes from two donors were used in this study: donor 1 (10-year-old male) and donor 2 (33-year-old female). The ECs from the donors’ eyes were not pooled, and all procedures and experiments were performed in an independent manner. 

Isolation of Human Primary Cells

Primary cells of human retinal ECs (HRECs) and human choroidal ECs (HCECs) were both obtained from the donor eyes. The retina and the choroid were separated from fresh donor eyes after corneal buttoning. Both tissues were chopped and digested for 2 hours with 1 U Collagenase/Dispase (10269638001; Roche, Basel, Switzerland) in EBM-2 (CC-3156; Lonza, Basel, Switzerland) in a shaking incubator (200 rpm) at 37°C. The digested tissues were filtered with a 40-µm cell strainer (43-50040-51; pluriSelect Life Science, Leipzig, Germany), and the ECs were isolated using Dynabeads CD31 Endothelial Cell (11155D; Thermo Fisher Scientific, Waltham, MA, USA; mouse IgG1) and DynaMag-2 Magnet (12321D; Thermo Fisher Scientific) following the manufacturer's protocol. 

Cell Preparation and Culture

Human umbilical vein endothelial cells (HUVECs) were purchased from Lonza and were cultured in endothelial cell growth medium-2 (EGM2; CC-3162, Lonza). CD31-positive HRECs and HCECs were cultured in 1% gelatin-coated dishes with the EGM-2 MV Microvascular Endothelial Cell Growth Medium-2 BulletKit (CC-3202, Lonza). CD31-negative cells from the retina and choroid were also collected and cultured in Gibco Dulbecco's Modified Eagle's Medium (DMEM)/F-12 (10565018, Thermo Fisher Scientific) supplemented with platelet-derived growth factor subunit B (PDGFB), fibroblast growth factor (FGF), insulin-like growth factor (IGF), and epidermal growth factor (EGF). The cells were used at passage 2 for analyzing molecular parameters and performing functional assays. Cells were harvested at 70% confluency for molecular profiling; we avoided 100% confluence because some cells start to elongate and lose their pebble-like morphology at this point. 

Flow Cytometry Analysis for Endothelial Cell Purity

Freshly isolated cells of CD31 bead-positive ECs and CD31 bead-negative cells were partially collected for purity analysis. Collected cells in Dynabeads buffer were washed with cold PBS and incubated in fluorescence-activated cell sorter (FACS) buffer (10-mM HEPES with 10% fetal bovine serum (FBS) and 1 mg/mL d-glucose) with 0.2 µg of CD31 antibody of different clones (ab28364, Anti-CD31 Antibody; Abcam, Cambridge, UK; rabbit IgG) on ice for 30 minutes. Cells and CD31 antibodies were gently mixed every 5 minutes by tapping. Cells were washed and resuspended in FACS buffer and fixed with 4% paraformaldehyde (PFA) for 15 minutes at room temperature. Fixed cells were washed and incubated with Allophycocyanin (APC) AffiniPure F(ab′)₂ Fragment Donkey Anti-Rabbit IgG (H+L) (711-136-152; Jackson ImmunoResearch, West Grove, PA, USA) for 30 minutes in the dark at room temperature. After incubation, cells were washed and analyzed with BD FACSCanto II (BD Biosciencies, Franklin Lakes, NJ, USA) using a 640-nm laser. FACS data were analyzed using FCS Express 7 software (De Novo Software, Pasadena, CA, USA). 

qPCR Expression Analysis

RNAs collected from CD31-positive ECs and CD31-negative cells at passage 1 were used for endothelial cell verification, and those collected at passage 2 were used for molecular profile analysis. RNA isolation was performed manually using Invitrogen TRIzol Reagent (15596026, Thermo Fisher Scientific). cDNAs were synthesized from the RNAs using the PrimeScript 1st Strand cDNA Synthesis Kit (6110A; Takara Bio, Shiga, Japan). Quantitative polymerase chain reaction (qPCR) was performed using the iQ SYBR Green Supermix (1708880; Bio-Rad, Tokyo, Japan) and custom-ordered primers (Bioneer, Daejeon, Korea; sequences are provided in Supplementary Table S1) with Bio-Rad CFX Connect and CFX Maestro software. One-way ANOVA was performed with Tukey's multiple comparisons for post hoc analysis for each gene. 

Tube Formation Assay

The tube formation assay was performed using growth factor-reduced Matrigel (356230; Corning, Inc., Corning, NY, USA). Briefly, a pre-chilled 96-well plate was coated with 50 µL of Matrigel and incubated at 37°C for 30 minutes. After the Matrigel was set, 150 µL of culture medium was mixed with the ECs (4 × 10⁴ cells/well) and then seeded into each well and incubated at 37°C for 4 hours. The formation of tubes was then imaged using an inverted microscope. One-way ANOVA was performed with Tukey's multiple comparisons for post hoc analysis for each parameter. 

Microfluidics Sprouting Assay

For in vitro microfluidics sprouting assay, the 3D Cell Culture Chip (AIM Biotech, Singapore) was used after modification of the manufacturer's protocol as previously described.¹¹ Briefly, a collagen solution (2 mg/mL at pH 7.4) was prepared using Corning Collagen I, Rat Tail (354236; Corning) and was used to fill a gel-filling inlet. The collagen gel-filled chip was incubated at 37°C for 30 minutes. Human plasma fibronectin (F0895; Sigma-Aldrich, St. Louis, MO, USA) was diluted in EBM-2 (fibronectin 50 µg/mL) and added to both channels, incubated at 37°C for 1 hour, and then flushed out using culture media (EGM-2 for HUVECs, EGM-2 MV for primary ECs). The cells were seeded in only one channel (6  ×  10⁴ cells/channel) and the opposite channel was filled with the EGM-2 MV medium overnight to produce a confluent monolayer. After 24 hours, a 20-ng/mL vascular endothelial growth factor (VEGF) gradient was generated by adding 20 ng/mL VEGF in the culture media of the cell-seeded channel and 40 ng/mL VEGF in the culture media of the opposite channel. To construct interstitial flow, the upper media chambers were loaded with 80 µL media, and the lower media chambers were loaded with 60 µL media. The media were changed every 24 hours with freshly added VEGF (20 ng/mL gradient) in the culture media, and the chips were imaged using an inverted microscope. One-way ANOVA was performed with Tukey's multiple comparisons for post hoc analysis for each parameter. 

Wound Closure Assay

ECs were seeded at a density of 1 × 10⁶ cells per well in 12-well plates to reach full confluency overnight. A long-traversing scratch was made using a 200-µL tip to scrape the cell monolayer. The wound opening was imaged at the marked region at 0, 12, and 24 hours after scratching. For migration analysis, the wells were treated with 5 µg/mL mitomycin C (MMC) 2 hours prior to the scratch. Two-way ANOVA was performed for cell type with MMC treatment and time window. Sidak multiple comparison was used for the post hoc analysis of each parameter. 

Trans-Well Permeability Assay

The trans-well permeability assay was performed using the In Vitro Vascular Permeability Assay kit (ECM644; MilliporeSigma, Burlington, MA, USA). ECs were seeded at a density of 2 × 10⁵ cells per insert and were cultured overnight in culture media (EGM-2 for HUVECs, EGM-2 MV for primary ECs) to obtain a complete monolayer. Cells were growth factor–starved and serum-starved using serum-free EBM-2 for 2 hours to remove the effect of basal VEGF. The media were changed to 3% FBS-supplemented EBM-2 containing 50 ng/mL VEGF 165 (H9166; Sigma-Aldrich) or the same volume of sterile water for vehicle control, and cultured at 37°C for 24 hours. The permeability assay was performed according to the manufacturer's protocol. The fluorescein isothiocyanate (FITC)–dextran permeability was measured and analyzed using the SpectraMax Gemini XPS Microplate Spectrofluorometer (Molecular Devices, San Jose, CA, USA) and SoftMax Pro 3.1.2 software (Molecular Devices). Welch's t-test was used to analyze VEGF-induced permeability changes in each cell. 

Tissue Cryosection and Immunofluorescence Staining

Human choroids and retinas were fixed in 4% PFA for 24 hours at 4°C. Fixed retina tissues and choroid tissues were dehydrated in 30% sucrose, embedded in optimum cutting temperature (OCT) compound, and stored at −80°C. Cryo-blocks were cut into 14-µm sections. For immunostaining, sections were washed to remove excess OCT compound and were incubated with 10% BSA for blocking. Tissue sections were incubated with antibodies (list provided in Supplementary Table S2), imaged using LSM 710 (Carl Zeiss, Oberkochen, Germany), and analyzed with manufacturer-provided ZEN 2.3 software. 

Image Analysis

Brightfield microscopic images were obtained using an inverted microscope (IX70; Olympus, Tokyo, Japan) with the DP controller software. The morphometric analyses of images were performed using Java-based imaging software (ImageJ 1.52p; National Institutes of Health, Bethesda, MD, USA). The images were calibrated with scale bars of the images acquired before measuring the areas and lengths. 

Experimental Replicates and Statistical Analysis

All experiments were performed with three individual culture replicates for HRECs and HCECs harvested from each individual donor eye. Because HUVECs are the most widely used cell type for in vitro EC research, all analyses were performed by using HUVECs as reference for quantification in order to reduce data differences from the technical and conditional differences as we used two different donor eyes, and the experiments were performed in an independent manner. The values are presented as mean ± standard error (SE). Statistical significance was determined using Welch's t-test for comparisons between two groups and ANOVA for comparisons among three groups. All graph generation and statistical analyses were performed using Prism 8.0 (GraphPad, San Diego, CA, USA), and statistical significance was set at P < 0.05. 

Results

We first analyzed the similarity and differences between the cells from two donors (donor 1 and donor 2) and found that almost all parameters that we analyzed showed similar tendencies, such as higher expression in HCECs than in HRECs, in both donors (Supplementary Fig. S1, Supplementary Table S3). 

HRECs and HCECs Have Similar Morphologies But Different Molecular Profiles

To perform molecular and functional analysis of the retinal and choroidal ECs, we isolated HRECs and HCECs from the human donor eyes (Fig. 1A). PECAM1 (CD31) and VEGFR2 gene expression in the isolated cells and PLVAP gene expression in HCECs were analyzed to ensure that the isolated cells were ECs in comparison with the non-endothelial cells collected during the isolation process (Fig. 1B). The purity of isolated HRECs and HCECs was analyzed using FACS by measuring the proportion of CD31-positive cells over CD31-negative cells. The purity of HRECs was 95.1% and that of HCECs was 93.9% (Supplementary Fig. S2). 

Figure 1.

 

Isolation of endothelial cells from human retina and choroid. (A) A schematic diagram of the isolation of endothelial cells from fresh donor eyes using CD31(+) magnetic beads. (B) The mRNA expression levels of PECAM1 and VEGFR2 in CD31-positive endothelial cells and CD31-negative cells isolated from the retina. (C) The mRNA expression levels of PECAM1, VEGFR2, and PLVAP in CD31-positive endothelial cells and CD31-negative cells isolated from the choroid. Error bars indicate standard deviations from qPCR loading replicates.

HRECs and HCECs showed similar morphologic characteristics in terms of shape and size, which were similar to the well-known endothelial cell line, HUVECs. All three cell populations showed the endothelial cell-specific cobblestone morphology up to passage 2 (Fig. 2A). We then analyzed the endothelial-related mRNA expressions in HRECs and HCECs and found that the expression levels of VEGF receptor 2 (VEGFR2) and VEGF receptor 3 (VEGFR3) were higher in HCECs than in HRECs (Fig. 2B), whereas the expression of the pericyte interaction gene PDGFB was stronger in HRECs (Fig. 2C). In terms of tip cell–related genes, the tip cell–forming gene DLL4 was higher in HCECs whereas the tip cell–expressed angiopoietin 2 (ANGPT2) gene was higher in HRECs (Fig. 2D). The levels of transcytosis-related genes, especially the fenestration-related gene PLVAP and the caveolin-1 (CAV1) gene, were higher in HCECs (Fig. 2E). In terms of junctional genes, the level of adherens junction–related gene PECAM1 was higher in HRECs (Fig. 2F), whereas the levels of tight junction-related genes were higher in HCECs (Fig. 2G). 

Figure 2.

 

HRECs and HCECs have similar morphologies but different molecular profiles. (A) The morphologies of HUVECs (passage 4), HRECs (passage 2), and HCECs (passage 2) (magnification, 20×). Scale bar: 100 µm. (B–G) mRNA expression analysis of VEGF receptors (B) and genes related to pericyte interaction (C), tip cell function (D), transcytosis (E), adherens junction (F), and tight junction (G). Error bars indicate standard error (one-way ANOVA with Tukey's post hoc analysis for multiple comparisons; n = 6 per group, three different cultures each from donor 1 and donor 2). *P < 0.05, **P < 0.01, ***P < 0.001.

The differences in the expression of endothelial-related genes were also analyzed in tissue sections (Supplementary Fig. S3). Similar to the results for HRECs and HCECs, the expression of CD31 (PECAM1) was higher in the retina capillary than in the choriocapillaris (Supplementary Fig. S3A); in contrast, the expression of CD31 in large choroid vessels was similar to that in the retina. Tie2 expression was higher in the retinal vessels than in the choroidal vessels (Supplementary Fig. S3C), whereas gene expression in molecular profiles did not show a significant difference between HRECs and HCECs (Fig. 2C). Similar to the gene expression analysis results in HRECs and HCECs, ANGPT2 expression was higher in retinal vessels than in choroidal vessels (Supplementary Fig. S3D). VEGFR2 and PLVAP expression was higher in choroid vessels, similar to the gene expression results. Although the expression of tight junction protein ZO-1 was higher in HCECs (Fig. 3G), the expression levels in retinal capillary vessels and choriocapillaris did not show a significant difference (Supplementary Fig. S2F). 

Figure 3.

 

HRECs and HCECs have similar vasculogenic activities. (A) Brightfield microscopic images of tubes formed by endothelial cells 4 hours after seeding on Matrigel and skeletonized images for analysis (magnification, 4×). Scale bar: 500 µm. (B–E) Profiles of the tubes generated by HRECs and HCECs in comparison with HUVECs in terms of number of nodes (B), number of loops (C), length of the tubes (D), and area of the loops (E). Error bars indicate standard error (one-way ANOVA with Tukey's post hoc analysis for multiple comparisons; n = 6 per group, three different cultures each from donor 1 and donor 2). *P < 0.05, **P < 0.01, ***P < 0.001.

HRECs and HCECs Have Similar Vasculogenic Activities

To evaluate the vasculogenic functions of the ECs, HRECs and HCECs were analyzed using a tube formation assay at 4 hours after seeding (Fig. 3A). HRECs and HCECs did not show significant differences in the number of nodes (Fig. 3B) and loops (Fig. 3C), which were significantly lower than those in HUVECs. The tube length was similar among HRECs, HCECs, and HUVECs (Fig. 3D). The loop areas of HRECs and HCECs did not show a significant difference, but both were significantly larger than those of HUVECs (Fig. 3E). 

HRECs Sprout Faster and HCECs Make More Tip Cells

We performed an angiogenic sprouting assay using a microfluidics chip while imposing a VEGF gradient with interstitial flow from the upper channel to the lower channel. The VEGF gradient induced cell growth from the VEGF-low channel to the VEGF-high channel through a hydrogel channel by forming tip cells and sprouting (Fig. 4A). The results were measured at 1 and 3 days after the induction of a VEGF gradient (Fig. 4B). HRECs sprouted faster than did HCECs (Figs. 4C, 4E, 4F), but the number of tip cells was higher in HCECs (Fig. 4D). HRECs and HCECs both showed higher responsiveness to VEGF-induced angiogenic sprouting than HUVECs. 

Figure 4.

 

HRECs sprout faster and HCECs make more tip cells. (A) Schematic diagram of a microfluidics chip with a VEGF gradient and interstitial flow. (B) Brightfield images of sprouting endothelial cells and skeletonized images for analysis (day 1 magnification, 20×; day 3 magnification, 10×). Scale bar: 200 µm. Red dots indicate the leading tip cells. (C) Average sprouting length; statistical significance is not indicated. (D–F) Number of tip cells at day 3. Cell growth length has been analyzed at day 1 (E) and at day 3 (F). Error bars indicate standard error (one-way ANOVA with Tukey's post hoc analysis for multiple comparisons; n = 6 per group, three different cultures each from donor 1 and donor 2). *P < 0.05, **P < 0.01, ***P < 0.001.

HRECs Show a Migratory Behavior and HCECs Show a Proliferative Behavior

In vitro wound closure assay was performed to analyze the proliferative and migratory behaviors of the ECs. Wound closure was analyzed at 12 hours and 24 hours after scraping, and HRECs showed significantly faster and more potent wound closing compared with HCECs and HUVECs (Fig. 5A). Specifically, at 12 hours, the scraped opening was almost closed in HRECs whereas more than 50% of the openings remained in HCECs and HUVECs; at 24 hours, the scraped opening was fully closed in HRECs, but more than 30% of the openings remained in HCECs and HUVECs (Fig. 5C). 

Figure 5.

 

HRECs show a migratory behavior and HCECs show a proliferative behavior. (A) Wound closure assay without MMC treatment that tests both proliferative and migratory functions. The wound area was evaluated at 12 hours and 24 hours post-scraping. (B) Wound closure assay with pretreatment of MMC before scraping. (C–F) Wound opening areas without MMC treatment (C, E) and with pretreatment of MMC (D, F) were measured and analyzed at 12-hour intervals. Error bars indicate standard error (two-way ANOVA test; n = 6 per group, three different cultures each from donor 1 and donor 2). *P < 0.05, **P < 0.01, ***P < 0.001. (G–I) Wound opening areas at 12-hour intervals with or without MMC treatment in HUVECs (G), HRECs (H), and HCECs (I). Error bars indicate standard error (two-way ANOVA test; n = 6 per group, six replicate assays from donor 2). ***P < 0.001 versus MMC(–).

Considering that wound closure can be mediated by both the proliferation and migration of cells, we performed the assay again by adding MMC treatment, a cell proliferation inhibitor, to solely analyze the migratory function of the ECs (Fig. 5B). Even after MMC treatment, the scraped opening was almost fully closed in HRECs at 24 hours (Figs. 5E–5I); on the other hand, the wound-closing capability of HCECs had significantly decreased after MMC treatment, with more than 60% of the opening remaining at 24 hours. These results indicate that HRECs and HCECs are largely characterized by migratory and proliferative behaviors, respectively. 

HRECs Show High Permeability in Response to VEGF Stimulus

Changes in cell permeability in response to VEGF stimulus were analyzed using a Transwell permeability assay. ECs were seeded on collagen-coated Transwell inserts to form full monolayers. The cells were then treated with extra VEGF, and FITC–dextran was added to assess the degree of permeability (Fig. 6A). The cells were stained to confirm the breakage of the monolayer (Fig. 6B). At 24 hours after VEGF stimulus, HRECs showed a significant increase in cell permeability, whereas HUVECs and HCECs did not show significant differences after VEGF stimulus (Fig. 6C). 

Figure 6.

 

HRECs show high permeability in response to VEGF stimulus. (A) Schematic diagram of permeability assay under VEGF stimulus. (B) Staining of cells prior to fixation for confirmation of the breakage of the monolayer. (C) Fluorescence measurement of FITC–dextran after 24 hours of VEGF stimulus. Error bars indicate standard error (Welch's t-test or Student's t-test; n = 6 per group, six different culture assays from donor 2). *P < 0.05.

Figure 7.

 

Molecular, cellular, and functional heterogeneities of HRECs and HCECs. A schematic illustration of the relationships between molecular profile differences and functional characteristics. (A) An illustration of retinal and choroidal vessels where the HRECs and HCECs were isolated. (B) HRECs were migratory, and HCECs were proliferative. (C) VEGF responsive permeability was higher in HRECs than in HCECs. (D) HCECs developed more tip cells with low fidelity, but HRECs developed stable tip cells with high angiopoietin-2 (Ang2) expression under the VEGF gradient. (E) HRECs showed high pericyte coverage with PDGFB expression, and HCECs showed fenestrations with diaphragms with high PLVAP expression.

Discussion

Here, we have demonstrated that retinal ECs and choroidal ECs show heterogeneities at the cellular level. Retinal ECs are vasculogenic with high migratory characteristics and faster angiogenic sprouting, and they are more responsive to VEGF-induced permeability. In contrast, choroidal ECs express high levels of transcytosis genes, and they are vasculogenic, rather proliferative, adept in generating tip cells, and less responsive to VEGF-induced permeability (Fig. 7). 

The molecular profiles of HRECs and HCECs largely explain their differences in functional assays. Retinal ECs expressed higher levels of PECAM1 than did choroidal ECs, which in turn upregulates Rho activation and induces cell migration¹² for faster wound healing. Contrary to our expectation, genes related to tight junction showed higher expression levels in HCECs than in HRECs, although the exact expression levels of tight junction proteins on the cell surface were not evaluated in this study. However, the higher gene expression levels can be understood for faster replacement¹³ of junctional proteins in choroidal ECs, and may also explain the higher degree of paracellular transport of ions¹⁴ through a lower density of tight junction proteins in retinal ECs to support retinal neurons. 

Transcytosis-related genes such as PLVAP and CAV1 were reported to be higher in choroidal ECs, which explains choriocapillaris fenestrations^6,15 considering that PLVAP protein also forms choriocapillaris fenestration diaphragms and a higher number of caveolae in ECs.¹⁵ The high expressions of transcytosis-related genes and the larger number of transvascular transport channels—fenestrations and caveolae—add to our understanding of the role and function of choriocapillaris in the outer BRB such as metabolic support of the retinal pigment epithelium and waste clearance.¹⁶ 

The expression of VEGFR3 on the vascular endothelium regulates the basal permeability of ECs by modulating VEGF/VEGFR2 signaling,¹⁷ and the expression levels of VEGFR3 are related to the permeability of retinal and choroidal vasculatures. The lower level of VEGFR3 expression in retinal ECs indicates a higher basal permeability, and, under VEGF stimulus, retinal ECs show significant increases in permeability, which is in line with our results. Interestingly, a previous study showed that VEGFR3, a lymphatic vessel marker,¹⁸ is expressed in choriocapillaris,¹⁹ which suggests that choroidal ECs may have lymphatic-like characteristics. We speculate that high expression of VEGFR2 and Tie2 in choroidal ECs may contribute to its relatively weaker responsiveness to anti-VEGF agents.²⁰ 

HRECs and HCECs showed different angiogenic characteristics, as do retinal and choroidal vasculatures in physiological and pathological conditions.^21,22 Choroidal ECs showed increased tip cell generation, which is likely due to higher expression levels of VEGFR2 and DLL4 genes.^23,24 Conversely, the low ANGPT2 gene expression indicates less prominent tip cell characteristics,²⁴ which explains the low tip-cell fidelity of choroidal ECs. In contrast, retinal ECs expressed higher levels of ANGPT2 and had higher and more stable tip-cell characteristics. These characteristics recapitulate the different types of angiogenesis that occur in retinal and choroidal vasculatures, in which retinal capillary vessels mainly show sprouting angiogenesis and choriocapillaris vessels show intussusceptive angiogenesis.^25,26 

The interaction between ECs and pericytes is important for maintaining the stability and function of vessels.^27,28 The PDGFB gene, which produces the pericyte growth factor PDGFB, is expressed in ECs to recruit pericyte and maintain pericyte ensheathment.^29–31 The EC-to-pericyte ratio is 1:1 in retinal capillary vessels and 6:1 in choriocapillaris,^32,33 and this difference in pericyte coverage of endothelium supports our finding that retinal ECs showed higher PDGFB expression than did choroidal ECs. 

Our study had several limitations. We used subcultured cells rather than freshly isolated cells or in vivo tissues. Because the microenvironments of the retina and choroid cannot be fully mimicked in the in vitro condition, we could not adjust the experimental conditions for differences in the environmental niche. We also did not use growth factors for retina-specific or choroid-specific conditions. To overcome these limitations and validate the physiological relevance of our study, we additionally analyzed freshly fixed tissue sections of the same donor eye. Moreover, we analyzed the mRNA expression level and not the protein level for assessing the functional characteristics of the ECs. Despite these limitations, our analysis of the molecular differences of retinal and choroidal ECs that correlate to their functional characteristics can help us understand the two different BRB physiologies and related pathogeneses. 

In conclusion, we found that retinal ECs and choroidal ECs were significantly different in terms of molecular and cellular characteristics, which may add to our understanding of the functional differences between the vascular endothelial cells of the inner BRB and the outer BRB under physiological and pathological conditions. Our results suggest a new perspective on retinal choroidal vascular bed–specific disease therapeutics. 

Acknowledgments

The authors thank the members of the Department of Ophthalmology at Asan Medical Center for their help in the acquisition of fresh donor eye retina–choroid tissue. The statistical analysis was performed in consultation with the Department of Clinical Epidemiology and Biostatistics at the University of Ulsan College of Medicine, Asan Medical Center, Seoul. 

Supported by a National Research Foundation of Korea grant funded by the Korean government (2018R1A5A1025511) and by a grant from the Asan Institute for Life Sciences, Asan Medical Center, Seoul, Korea (2023IP0097-1). The sponsor or funding organization had no role in the design or conduct of this research. This study was awarded the Members-in-Training outstanding poster prize (Anatomy and Pathology/Oncology section) at the 2022 ARVO annual meeting. 

Disclosure: S.J. Kim, None; J.S. Lim, None; J.H. Park, None; J. Lee, None