Epithelial Membrane Protein 2 (EMP2) Blockade Attenuates Pathological Neovascularization in Murine Oxygen-Induced Retinopathy

Abstract

Purpose: Retinopathy of prematurity (ROP) results from postnatal hyperoxia exposure in premature infants and is characterized by aberrant neovascularization of retinal blood vessels. Epithelial membrane protein-2 (EMP2) regulates hypoxia-inducible factor (HIF)–induced vascular endothelial growth factor (VEGF) production in the ARPE-19 cell line and genetic knock-out of Emp2 in a murine oxygen–induced retinopathy (OIR) model attenuates neovascularization. We hypothesize that EMP2 blockade via intravitreal injection protects against neovascularization.

Methods: Ex vivo choroid sprouting assay was performed, comparing media and human IgG controls versus anti-EMP2 antibody (Ab) treatment. In vivo, eyes from wild-type (WT) mice exposed to hyperoxia from postnatal (P) days 7 to 12 were treated with P12 intravitreal injections of control IgG or anti-EMP2 Abs. Neovascularization was assessed at P17 by flat mount imaging. Local and systemic effects of anti-EMP2 Ab treatment were assessed.

Results: Choroid sprouts treated with 30 µg/mL of anti-EMP2 Ab demonstrated a 48% reduction in vessel growth compared to control IgG-treated sprouts. Compared to IgG-treated controls, WT OIR mice treated with 4 µg/g of intravitreal anti-EMP2 Ab demonstrated a 42% reduction in neovascularization. They demonstrated down-regulation of retinal gene expression in pathways related to vasculature development and up-regulation in genes related to fatty acid oxidation and tricarboxylic acid cycle respiratory electron transport, compared to controls. Anti-EMP2 Ab-treated OIR mice did not exhibit gross retinal histologic abnormalities, vision transduction abnormalities, or weight loss.

Conclusions: Our results suggest that EMP2 blockade could be a local and specific treatment modality for retinal neovascularization in oxygen-induced retinopathies, without systemic adverse effects.

Retinopathy of prematurity (ROP) contributes significantly to the burden of childhood vision impairment. In premature infants, high oxygen exposure is required for management of lung immaturity. This systemic hyperoxia exposure results in local vascular attenuation of the retinal vasculature (the first phase of ROP).¹ The abnormal cessation of blood vessel growth, coupled with increasing metabolic demand, leads to pathologic neovascularization (NV), which characterizes the second phase of ROP.² Although ROP is likely modulated by multiple environmental factors including growth, nutrition,^3–5 and inflammation,⁶ the best-characterized local signals that regulate retinal blood vessel growth are hypoxia-inducible factor 1 alpha (Hif1α) and vascular endothelial growth factor (VEGF).¹ The mainstay of ROP management is oxygen titration to avoid fueling the first phase of ROP, but oxygen provision is often necessary for lung immaturity and to decrease the risk of hypoxia-associated mortality.⁷ Late ROP is managed with anti-VEGF intravitreal injections, often in combination with laser surgery.⁸ However, ROP treatments, especially laser therapy, can increase risk of vision abnormalities.⁹ Moreover, given the ubiquitous nature of VEGF, theoretical systemic effects of anti-VEGF therapy could be detrimental and more specific therapies may mitigate these risks. 

Our group has demonstrated that epithelial membrane protein-2 (EMP2) is an upstream regulator of Hif1α-mediated Vegf-A induction.¹⁰ In the murine oxygen-induced retinopathy (OIR) model, we have shown that Emp2-KO mice are protected against oxygen-induced NV via downregulation of Hif1α at postnatal day 12 (P12) and downregulation of VEGF-A at postnatal day 17 (P17) despite hyperoxia exposure.¹¹ We also demonstrated in vivo that Emp2 regulates hypoxia-induced angiogenesis in adult retinal pigment epithelial cell lines¹⁰ through genetic modulation in cell lines. Furthermore, antibody-mediated targeting of EMP2 in cancer stem cells can block tumor-initiating capacity in vivo, potentially via Hif-mediated pathways.¹² Given this, our study aims to determine whether EMP2 blockade via Ab targeting is an effective treatment to decrease pathologic NV ex vivo and in vivo. We hypothesize that intravitreal injection of an anti-EMP2-Ab will decrease pathologic NV at P17 in OIR, without retinal or systemic toxicity. 

Methods

Ethical Approval

Animal studies conformed with established guidelines set by the National Institutes of Health and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Animal-related protocols were approved by the Animal Research Committee of the University of California Los Angeles. Wild-type C57BL/6 mice were purchased from Charles River Laboratories (Thousand Oaks, CA) through the UCLA Division of Laboratory and Animal Medicine. Mice were housed in standard conditions with 12:12-hour light-dark cycles and fed as desired with a standard rodent chow diet (Pico Lab Rodent Diet 20, Lab Diet, St. Louis, MO, USA) and water. 

OIR Model

C57BL/6 wild-type (WT) mice were bred, and pups were designated as postnatal (P) day 0.5 the morning that they were found to be delivered. Pregnant dams were fed a standard rodent chow diet during mating and throughout gestation and suckling. Litters were culled to eight pups, as needed, per published recommendations.¹³ Male and female pups were used. Mothers and their litters were assigned randomly to either normoxia or hyperoxia (75% oxygen) conditions. For those mice assigned to the hyperoxia group, OIR was induced by placing mothers and their litters in 75% continuous oxygen from P7 to P12 within an airtight chamber (Proox model 360; BioSpherix, Parish, NY, USA). While mice were in the chamber, chamber conditions were monitored twice daily, including temperature and humidity and health checks performed on the animals. On postnatal day 12, the mothers and their litters were taken out of the chamber and placed into normoxia (room air, 21% oxygen). If mothers exposed to hyperoxia were found to be stressed/in poor health, pups were fostered onto a nursing mother after being taken out of the chamber. Mice were observed for overall health and activity daily until being weaned at P21, and weights were taken from pups at days P12 and P17. 

Anti-EMP2 Abs

The fully human anti-EMP2 IgG1 mAb has been previously described in detail¹⁴ and has been shown to bind both human and murine EMP2. LakePharma Inc. was contracted to produce anti-EMP2 mAbs in bulk, according to their standard practices and were stored in 20 mM Histidine, 140 mM NaCl, 0.02% Polysorbate-80, pH = 6 at −80°C prior to thaw. Each antibody batch was validated by both electropherogram and for affinity against a biotinylated human EMP2 peptide consisting of the second extracellular loop of the protein (DIHDKNAKFYPVTREGSYGGSGSK; Invitrogen, Carlsbad, CA, USA) using ForteBio Octet. Affinities between 5–8 nM were considered acceptable. 

Choroid Sprouting Assay

A well-described microvascular mouse explant model, the choroid sprouting assay, was used to determine gross anti-angiogenic potential of the anti-EMP2 Ab.¹⁵ Briefly, after euthanasia of the animal, eyes from P12 C57Bl6 mice were enucleated immediately, sprayed with 70% ethanol, and then rinsed and kept in ice-cold PBS for dissection. The cornea and lens were removed, and the choroid-scleral complex from the midperiphery was separated from the retina. The choroid-scleral complex was then cut into 1 mm square pieces, leaving the retinal pigment epithelium on, and seeded into the bottom of wells coated with 30 µL of growth-factor reduced Matrigel (Corning Inc., Corning, NY, USA or BD Bioscience, Franklin Lakes, NJ, USA) in a 24 well plate. After seeding the choroid fragments, plates were incubated at 37°C with 5% CO₂ for 10 minutes to allow Matrigel to solidify, then medium was added and left in the chamber for 48 hours. Each well received medium composed of 500 µL of endothelial growth medium-2 (EGM-2). EGM-2 was made with the following: (1) “Bullet Kit” (Lonza Group, Basel, Switzerland), (2) endothelium basal medium (Lonza Group), (3) penicillin-streptomycin (5 mL per 500 mL of endothelium basal media), and (4) 5 mg/L plasmocin (InvivoGen cat. no. ant-mpp; InvivoGen, San Diego, CA, USA). After 72 hours, medium was changed every 72 hours in the wells with one of the following: (1) media alone, (2) media + 60 µg/mL of human control IgG (Sigma-Aldrich, cat. no. I-4506; Sigma-Aldrich Corp., St. Louis, MO, USA), (3) media + 30 µg/mL of anti-EMP2 Ab, or (4) media + 10 µg/mL of anti-EMP2 Ab. Photos of individual explants were taken using a Leica DMIL microscope and Nikon D5 camera (Nikon Inc., Melville, NY, USA). Sprout growth was imaged and assessed on day 9 after plating. For image analysis, images were opened in Adobe Photoshop. Images were processed by auto-contrast, inverted adjustment, and color range for sampled colors using the eye-drop function to select background (fuzziness set to 40), and then in “Quick Mask Mode,” undesired sections were erased. The image was then set to “inverse” and the number of pixels in the selection was recorded under the Histogram screen. Raw counts for the control, anti-EMP2 Ab at 10 µg/mL, and anti-EMP2 Ab at 30 µg/mL were normalized to raw counts from fragments from the same eye treated with media only. 

Intravitreal Injections

To evaluate the safety of intravitreal anti-EMP2 Ab, adult 3–4 month old WT mice were given intravitreal injections (IVI) of 1 µg/g of control IgG, 1 µg/g of anti-EMP2 Ab, or 1 µL of control buffer (20 mM Histidine, 140 mM NaCl, 0.02% polysorbate-80, pH = 6). IVIs with microscope guidance using 33-gauge needles (Hamilton, cat. no. 87908; Hamilton Company, Reno, NV, USA) were performed using sterile technique. Mouse eyes were treated with triple antibiotic ointment (Bacitracin zinc USP 400 units, Neomycin 3.5 mg and polymyxin B sulfate, USP 5000 units; Actavis Pharma, Inc. Parsippany, NJ, USA) after injection. Adult mice and their surgical sites were observed daily for five days after IVI, and electroretinography (ERG) performed one and four weeks after injections. 

To assess the best time point for IVI in the OIR model, WT mouse pups were injected with 0.5 µL diluted 0.001% Trypan Blue solution in PBS (Corning, Inc.) at P7 or P12 (n = 6–14) and compared to age-matched pups who received no IVI (n = 4), via ERG assessment at P23, and fundoscopy and optical coherence tomography at P25. 

To assess the efficacy of IVI of anti-EMP2 Ab in the OIR model, mice were exposed to OIR as described above then removed from the oxygen chamber at P12 and sedated with 10 µL of ketamine (1:20)/xylazine (1:200). If the eyelids were still fused, eyelids were surgically opened under sterile technique. With microscope guidance, 0.5 µL IVIs of 4 µg/g of control IgG, 1 µg/g of anti-EMP2 Ab, or 4 µg/g of anti-EMP2 Ab into the vitreous body were performed using 33-gauge needles under sterile technique. Mouse eyes were treated with triple antibiotic ointment after injection. Mice and their surgical sites were observed daily for five days after IVIs. An experimental timeline for drug administration, tissue collection, and imaging is depicted in Figure 1. 

Figure 1.

 

Diagram of experimental timeline.

Whole Mount Imaging

As per published standard guidelines for the OIR model,¹³ the ocular globe was enucleated at P17 and fixed in 4% paraformaldehyde for 1.5 hours. After the eyes were washed, retinas were dissected under a surgical microscope (Leica S6D; Leica, Wetzlar, Germany), washed with PBS, and then blocked with blocking buffer for one hour. Blocking buffer was made with 20% fetal bovine serum, 2% goat serum, 1% Triton X-100, and 0.5% bovine serum albumin in PBS. The retina was then stained with Alex594-isolectin GS-IB4 (1:500; Invitrogen) in buffer at 4°C overnight, washed with diluent buffer and PBS three times, then flat mounted onto slides. Four peripheral incisions were made at 90° intervals to divide the retina into equally sized quadrants and then mounted with ProLong medium (Molecular Probes, Eugene, OR, USA) with a coverslip. Images were taken with an AxioCam CCD camera (Carl Zeiss) mounted to an inverted epifluorescence microscope (AxioVert 135, Carl Zeiss). For each retina, four to six images were taken at 4X magnification and pieced together using the DP2-BSW software (Olympus). 

Quantification of Neovascularization Phase of ROP

Given that IVIs were assessed as safe at P12, the P17 time point was evaluated to quantify peak NV. Images from whole mount processing of mouse retinas were processed using Adobe Photoshop and analyzed for percentage of the NV area, using published methods.¹³ Two masked scorers performed the quantification. The Magic Wand Tool was used to select areas of NV, using a threshold of 50. The number of pixels of NV was then quantified as a percentage of the whole retinal area in pixels, which was determined using the polygonal lasso tool. 

Retinal Tissue Collection and Immunohistochemistry

At the designated time points of interest, mice were euthanized using inhaled isoflurane (5–10 minutes). Routine H&E staining of the retina was performed by the UCLA Translational Pathology Core Laboratory using standard methods to evaluate overall tissue morphology in groups of P17 WT mice treated with control or anti-EMP2 Abs. Retina were fixed with 4% formaldehyde for 24 hours at room temperature, then embedded in paraffin, and sectioned to 4 to 6 µm thickness by the Translational Pathology Core Laboratory at UCLA. Images were captured using a Leica S6D microscope. 

Whole retina samples were also collected at P17 in OIR mice treated with IVI control IgG (n = 3) and mice treated with IVI anti-EMP2 Ab (n = 4). These samples were snap frozen and stored at –80°C for later processing for RNA sequencing. 

RNA Sequencing

Whole retinal samples from WT mice who received control IgG or anti-EMP2 Ab were processed for bulk RNA sequencing (3 biological replicates per group). Tissue was suspended and sonicated in lysis buffer, then homogenized with the Qia Shredder kit (Qiagen, Valencia, CA, USA). RNA extraction was performed with the Qiagen RNeasy Mini Kit (Qiagen), and samples were then processed at the UCLA Technology Center for Genomics and Bioinformatics. After quantification and testing for RNA degradation, samples were deemed adequate for sequencing. Libraries were prepared using the KAPA Stranded mRNA-Seq Kit using the following steps: mRNA enrichment and fragmentation, cDNA synthesis of the first strand with random priming followed by conversion of cDNA:RNA hybrid to ds-cDNA for second strand synthesis, and then dUTP incorporation into the second cDNA strand. After, end repair for generation of blunt ends, A-tailing, adaptor ligation then PCR amplification was performed. Different adaptors were used for multiplexing of the sample. Illumina's HeSeq 3000 was used for paired-end 2 × 65bp run, SAV used for data quality check, and Bcl2fastq v2.19.1.403 software used for demultiplexing. The raw RNA sequencing data has been deposited in NCBI's Sequence Read Archive¹⁶ and are accessible through SRA Series accession number PRJNA1096742. 

STAR v2.7.10b was used to align sequence reads to a genome index including both the genome sequence (GRCm39 mouse primary assembly), as well as the exon/intron structure of known mouse gene models (Gencode M31 comprehensive genome annotation). Strand specific gene counts for protein-coding genes from STAR built-in counter were compiled and independent filtering applied to remove genes with low counts across all samples. As previously described,¹¹ expression estimates were computed from the filtered counts matrix after correction for gene mappable length and per-sample sequencing depth, and reported in units of transcripts per million. Cell type proportion estimates were generated with the Gene Expression Deconvolution Interactive Tool using signatures from the Tabula Muris Reference database.^17,18 Count-based and variance-stabilized data (vsd) were used for ordination and differential and clustering analysis; DESeq2 was used for differential expression analysis.¹⁹ The vsd matrix in R was used for principal component analysis (prcomp function) in R.²⁰ For pairwise comparisons between control IgG or anti-EMP2 Ab samples, genes were classified as significantly up- or down-regulated using a Wald-adjusted P value <0.05. 

Blood Collection

Blood samples were collected at one month of age in mice: (1) treated with 4 µg/g of intravitreal control IgG at P12, (2) treated with 4 µg/g intravitreal anti-EMP2 Ab at P12, or (3) 2 hours after receiving 20 µg of anti-EMP2 Ab via intraperitoneal injection. A sample spiked with 20 µg of anti-EMP2 Ab was used as a positive control, and a negative control from a mouse given no injection. Serum was isolated via cardiac puncture. Blood was incubated at room temperature for 25 minutes in a microcontainer (BD Bioscience) in order to allow it to coagulate, and then it was spun in a centrifuge at 1500 g for 15 minutes. Serum was collected from the top, clear portion and stored at −80°C. 

Western Blot Analysis

Mouse serum 0.75 µL from each sample was mixed with Laemmli buffer. Proteins were separated on 4% to 20% SDS-PAGE gels (ThermoFisher Scientific, Grand Island, NY, USA) and then transferred onto nitrocellulose membranes. Membrane blocking was performed with 10% nonfat dry milk in tris-buffered saline solution with 0.1% Tween-20 (TBSt) and then probed in 5% nonfat dry milk in TBSt containing anti-human Fc (1:3000) (Jackson) and anti-transferrin (1:1,000) (Protein Tech) antibodies. Proteins were detected using horseradish peroxidase–labeled secondary antibodies (Southern Biotech, Birmingham, AL, USA), and bands were visualized with chemiluminescence using Crescendo Western Substrate (EMD Millipore, Burlington, MA, USA) on the Odyssey XF (LI-COR) developer. NIH ImageJ software was used for band density quantification. Samples were normalized to transferrin to correct for loading variation. At least three individual experiments were analyzed. 

Electroretinography in Mice

For evaluation of retinal function, full-field electroretinograms (ERGs) were performed for the following controls: (1) mice at P23 after receiving no intravitreal injection, (2) 0.5 µL injection of 0.001% trypan blue at P7 or P12 to determine optimal timing of intravitreal injection,²¹ and (3) adult mice who received no injection. ERG was also performed on three- to four-month-old mice at 1 and 4 weeks after injection for toxicity evaluation: (1) who received 1 µg/g anti-EMP2 Ab injection, (2) who received 1 µg/g control IgG as a control, or (3) 1 µL control buffer injection. In addition, adverse effects after IVI in OIR mice was evaluated by ERG in control IgG- and anti-EMP2 Ab-treated mice at one month of age, approximately three weeks after exposure to OIR at P7-P12. 

After overnight dark adaptation, mice were given an intraperitoneal injection of saline solution containing ketamine (15 mg/kg body weight) and xylazine (7 mg/kg body weight) for anesthesia. ERGs were recorded with the Celeris electrophysiological system (Diagnosys, Lowell, MA, USA) as previously described.²² Pupil dilation was achieved using one drop each of 1% Tropicamide and 2.5% phenylephrine hydrochloride ophthalmic solutions, and a drop of lubricant eye gel (0.3%, GenTeal Tears; Alcon, Geneva, Switzerland) was applied to the corneal surface to ensure good optics and maintain corneal integrity. Body temperature was maintained at 38°C. Stimuli were delivered to both eyes simultaneously using a fiber-optic bundle with an embedded electrode at each fiber-optic tip. Signal processing was performed using software provided by the instrument manufacturer. Post data collection analyses were performed using custom software. Responses were computer-averaged for each stimulus condition. All stimuli were presented at 1 Hz except for the brightest flashes, where the presentation rate was slowed to 0.2 Hz. ERG responses were recorded to a range of intensities under dark-adapted conditions and the intensity–response functions were analyzed to extract V_max, which is the maximum saturated b-wave amplitude, for each stimulus condition. 

Fundoscopy, Optical Coherence Tomography, and Fluorescein Angiography in Mice

Fundoscopy, optical coherence tomography (OCT), and fluorescein angiography (FA) were performed as previously described.²³ Mice were anesthetized with a saline solution containing ketamine and xylazine, and pupils were dilated as described above. Mice were positioned on a movable platform to align their eyes with the imaging device. Ocular demulcent solution containing hydroxypropyl methylcellulose (Gonak; Akorn, Lake Forest, IL, USA) was then applied to the eye to ensure a smooth, transparent interface anterior to the cornea. Ultra-high-resolution spectral domain OCT (SD-OCT) was used to evaluate gross retinal abnormalities in mice who received IVIs of very diluted trypan blue at P7 or P12. A total of 100 b-scans were collected using the Bioptigen SD-OCT system, aligned and stacked to create a three-dimensional image of the retinal volume. A high-resolution horizontal b-scan was obtained by averaging and spatially aligning 20 individual b-scans. 

Fundus images were obtained using the Micron II retinal imaging microscope (Phoenix Research Laboratories, Inc., Pleasanton, CA, USA) immediately after OCT during the same anesthetic session. The mouse eye was aligned with the optical axis of the camera, with the objective lens positioned so that it touched the corneal surface (corneal applanation). A drop of lubricant eye gel (0.3%, GenTeal Tears) was applied to the corneal surface to maintain corneal integrity. Serial images were recorded to document the change in retinal and ocular appearance over the course of the study. 

Retinal angiography was performed using the same general fundus imaging procedure described above. Mice were anesthetized as described above and injected subcutaneously with 10% sodium fluorescein (Akorn Inc.) at a dose of 0.01 mL per 5–6 gm of body weight. Short-wavelength light (1_max = 487 nm) was used to excite the fluorescein, and a blocking filter (transmission <500 nm <0.1%) was placed in the optical pathway to prevent the excitation light from reaching the camera. A movie of approximately three minutes’ duration was recorded from which individual images and clips were extracted. Fluorescein angiography was performed in one-month-old OIR mice treated with IgG control or anti-EMP2 Ab. 

Statistical Analysis

Statistical analysis was performed using GraphPad Prism software (GraphPad Software Inc., La Jolla, CA, USA) for choroid sprout quantification, ERG measurements, whole mount neovascularization ratio, weight, and protein quantification by Western blotting. Data were tested for normality using the Shapiro-Wilk test or D'Agostino & Pearson testing. Data are presented as means ± SEM, unless otherwise indicated. For comparisons among three or more groups (choroid sprout quantification, ERG measurements, whole mount NV ratio, Western blot analysis), one-way ANOVA with Tukey correction was used for post-hoc comparisons. For comparisons between two groups (body weight), unpaired, two-tailed Student t-tests were used. P values <0.05 were considered statistically significant. 

Results

Antibody-Mediated EMP2 Targeting Decreases Neoangiogenesis in the Choroid Sprout Assay

To determine whether an anti-EMP2 Ab is effective at reducing the degree of angiogenesis, we performed ex vivo choroid sprouting assays. Fragments of the choroid-scleral complex were treated with (1) media alone, (2) media + 60 µg/mL of control IgG, (3) media + 10 µg/mL of anti-EMP2 Ab, or (4) media + 30 µg/mL of anti-EMP2 Ab (Fig. 2A). Choroidal fragments cultured with 30 µg/mL of anti-EMP2 Ab showed decreased vessel sprouting when compared to the control fragments cultured in 60 µg/mL of control IgG or media culture alone (Fig. 2B) (normalized counts: media 1.00, control IgG 1.29 ± 0.12, anti-EMP2 Ab 10 µg/mL 1.07 ± 0.27, anti-EMP2 Ab 30 µg/mL 0.615 ± 0.11, P = 0.03). Decreased choroid sprouting was not seen in fragments treated with 10 µg/mL of anti-EMP2 Ab compared to controls (media alone or control IgG-treated fragments). Taken together, these results support that anti-EMP2 Ab at appropriate concentrations can reduce angiogenesis ex vivo. 

Figure 2.

 

Anti-EMP2 Ab decreases choroid sprouting ex vivo. (A) Representative images of choroid sprouts at day 9 after treatment with media alone, control IgG at 60 µg/mL, anti-EMP2 Ab at 10 µg/mL, or anti-EMP2 Ab at 30 µg/mL. Scale bars: 2 mm. (B) Graph representing decreased choroid sprouting quantified as normalized counts (relative to the media choroid sprout taken from the same eye) for sprouts treated with anti-EMP2 Ab at 30 µg/mL compared to those treated with media alone, control IgG at 60 µg/mL, or anti-EMP2 Ab at 10 µg/mL (P = 0.03; n = 13–15/group). Data are represented in graphs as means ± SEM.

Intravitreal Anti-EMP2 Ab Administration in Adult Mice Does Not Affect Retinal Structure or Function

Given that anti-EMP2 Ab was effective in decreasing angiogenesis ex vivo, we next wanted to test the safety of intravitreal anti-EMP2 Ab administration in vivo. In adult mice (ages three to four months), we administered one intravitreal dose of 1 µg/g anti-EMP2 Ab, 1 µg/g control IgG, or 1 mL control buffer and performed ERG at one and four weeks after the injection. We found that any IVI reduced the V_max at both time points (anti-EMP2 Ab one week after IVI: 228.98 ± 32.07 µV, n = 14; anti-EMP2 Ab four weeks post-IVI: 420.64 ± 64.55 µV, n = 9; control IgG one week after IVI: 273.03 ± 28.74 µV, n = 20; control IgG four weeks after IVI: 367.74 ± 39.12 µV, n = 14; control buffer one week after IVI: 214.59 ± 55.84 µV, n = 6; control buffer four weeks after IVI: 262.59 ± 53.90 µV, n = 5), compared to three- to four-month-old mice who did not receive an injection (no IVI: 584.83 ± 31.65 µV; n = 21) (P < 0.0001) and that this reduction in V_max after IVI improved with time for all injection types (Fig. 3). However, adult mice who received IVI anti-EMP2 Ab did not show statistically significant differences in V_max, compared to mice who received control IgG or control buffer injections at all time points (P > 0.05), implying no adverse effect or benefit of the anti-EMP2 Ab on retinal electrophysiological responses. 

Figure 3.

 

Intravitreal injection of anti-EMP2 Ab in adult mice does not result in abnormal retinal function or structure, compared to intravitreal injections of control IgG or control buffer. (A) Graph representing decreased V_max measured by ERG in adult mice who receive any intravitreal injection, compared to no injection. However, there are no differences in V_max measured by ERG in adult mice treated with 1 µg/g of anti-EMP2 Ab, 1 µg/g control IgG or 1 µL of control buffer (n = 5–21/group per time point). ERG data are represented in graphs as means ± SEM. *P < 0.05, ***P < 0.001, ****P < 0.0001. (B) Representative fundoscopy and OCT images of adult mice who received IVI anti-EMP2 Ab, control IgG and control buffer immediately after injection, one week after IVI and four weeks after IVI.

To determine the optimal time to administer the anti-EMP2 Ab in the OIR model, mice received dilute trypan blue IVIs at P7 or at P12. These time points were chosen to accommodate the OIR model, which requires continuous hyperoxia exposure from P7 to P12 and results in peak aberrant NV at P17.¹³ After IVI at P7, there was a significant reduction of V_max in mice at P23 compared to mice who did not receive IVI or received IVI at P12 (none: 568.6 ± 69.2 µV, P12: 511.6 ± 45.2 µV, P7: 308.7 ± 70.4 µV; n = 4-6/group; P = 0.035) (Fig. 4A). In addition, significant opacification of the lens was detected by fundoscopy and resulted in the inability to obtain high quality OCT at P25 (Fig. 4B). Mice who received IVI at P12 did not exhibit a difference in the average V_max from mice who received no IVI (P > 0.05) (Fig. 4A). In addition, mice who received IVI at P12 did not demonstrate opacification of the lens at P25 and by OCT imaging, the retinal layers appear grossly intact (Fig. 4B). 

Figure 4.

 

Intravitreal injection results in no gross adverse retinal effects when given at P12, but results in decreased V_max and corneal clouding if given at P7. (A) Graph representing decreased V_max measured by ERG at P23 in mice who received IVI at P7 compared to mice who received IVI at P12 or controls who received no injections (*P = 0.035, n = 4–16/group). ERG data are represented in graphs as means ± SEM. *P < 0.05. (B) Representative fundoscopy images (top row) and OCT images (bottom row) taken of mice at P23 who received no intravitreal injection, intravitreal injection at P7, and intravitreal injection at P12.

Intravitreal Anti-EMP2 Ab Administration Attenuates Pathologic Neovascularization in Oxygen-Induced Retinopathy

To now determine if intravitreal administration of anti-EMP2 Ab effectively reduces in vivo NV, mice exposed to OIR were treated at P12 with control IgG at 4 µg/g (control), anti-EMP2 Ab at 1 µg/g, or anti-EMP2 Ab at 4 µg/g. There was no difference in the P17 weights of mice treated with control IgG or anti-EMP2 Ab (control IgG: 5.1 ± 0.11 g, anti-EMP2: 5.0 ± 0.123 g; n = 37–44/group; P = 0.6). Using whole mount imaging, retinal sections were taken at P17 to assess NV in these groups. Mice treated with 1 µg/g of anti-EMP2 Ab demonstrated a decrease in the NV ratio at P17 (0.063 ± 0.013; n = 8) compared with the control retina treated with control IgG (0.084 ± 0.0099; n = 16), though this did not reach significance (p = 0.22) (Fig. 5). However, there was a 42% decrease in pathologic NV with a higher concentration (4 µg/g) of anti-EMP2 Ab compared to control IgG-treated mice (control IgG: 0.084 ± 0.0099, n = 16; anti-EMP2 Ab 4 µg/g: 0.049 ± 0.0045, n = 10; P = 0.034) (Fig. 5). Taken together, these results suggest that an anti-EMP2 Ab attenuates pathologic NV in the OIR mouse model and therefore may play a role in targeted therapy for oxygen-induced retinopathies. 

Figure 5.

 

Anti-EMP2 Ab intravitreal treatment results in attenuated neovascularization at P17 in murine OIR. (A) Representative whole mount images demonstrating decreased severity of neovascularization at P17 in mice treated with 4 µg/g of anti-EMP2 Ab (bottom) compared to control IgG-treated mice (top). Red lectin staining of endothelial cells outlines the retinal vasculature. Scale bars: 500 µm. (B) Graph representing decreased neovascularization ratio at P17 in mice treated with 4 µg/g of anti-EMP2 Ab compared to mice treated with 1 µg/g of anti-EMP2 Ab or IgG controls (P = 0.013, n = 8 = 16/group). Data are represented in graphs as means ± SEM.

Intravitreal Anti-EMP2 Ab Administration in Oxygen-Induced Retinopathy Down-Regulates Gene Expression Related to Angiogenesis and Eye Developmental Pathways, and Up-Regulates Genes Related to Fatty Acid Oxidation and Biosynthesis

To understand how EMP2 blockade attenuates NV in the OIR model, whole retina RNA sequencing was performed on mice exposed to OIR that were treated at P12 with control IgG at 4 µg/g and compared to those treated with anti-EMP2 Ab at 4 µg/g. There were 194 genes that were down-regulated and 259 genes up-regulated in OIR mice treated anti-EMP2 Ab compared to those treated with control IgG (Wald P value < 0.01, Supplementary File S1). When pathway analysis was run on differentially expressed genes, we obtained a complete segregation between pathways associated with down-regulated genes and up-regulated genes (Supplementary File S2). Pathways exhibiting strong down-regulation included vasculature development and eye developmental pathways (tissue morphogenesis and eye morphogenesis) (Fig. 6). Notable genes included in these pathways are Col1a1, Epas1, Wnt5a, Jag1, Nrp1, and Wnt5a. Other down-regulated pathways included neural crest cell differentiation, melanogenesis, visual perception, receptor signaling pathways, cellular response to growth factor stimulus, and extracellular matrix organization. Preferential up-regulation was observed for metabolic processes, including fatty acid oxidation, TCA cycle and mitochondrial pathways. Notable genes in these pathways include Acat1, Ech1, and Pparg. Other up-regulated pathways include nucleoside metabolic processes, propanoate metabolism and carbon metabolism. 

Figure 6.

 

Intravitreal EMP2 Ab targeting leads to downregulation of gene expression related to angiogenesis and eye development, and upregulation of gene expression related to fatty acid oxidation. Gene expression changes induced by Emp2 blockade (A) Functional enrichment network for all genes differentially expressed between anti-EMP2 Ab and control IgG-treated mice. Each node represents and individual gene ontology term. Ontologies with similar gene members are grouped by categories (node color) and labeled using a representative member. Node size is proportional to statistical significance (hypergeometric P value) as shown. Edge thickness is proportional to between-node similarity and reflects the overlap between genes annotated in both ontology terms. Only edges representing a Kappa similarity score > 0.3 are shown. Only significant ontology terms are shown (hypergeometric P value < 1e-02). The network is oriented to highlight the segregation between ontologies enriched in up and down-regulated genes after anti-EMP2 Ab treatment. (B) Expression heatmaps for genes classified as differentially expressed between anti-EMP2 Ab and control IgG-treated mice. Data for each biological replicate is displayed. Shown are names for selected genes in different functional categories.

Intravitreal Anti-EMP2 Ab Administration in Oxygen-Induced Retinopathy Does Not Have Adverse Local Retinal or Systemic Side Effects

To evaluate for retinal toxicity after intravitreal anti-EMP2 Ab in the OIR condition, ERGs were performed at one month of age in OIR mice treated with IVIs of the anti-EMP2 Ab or human IgG. Untreated mice were also used as technical controls. No statistically significant differences in V_max were observed between any of the groups (none: 286.9 ± 18.8 µV, control IgG: 310.1 ± 64.6 µV, anti-EMP2 Ab: 293.0 ± 24.2 µV; n = 4–10/group; P > 0.05) (Fig. 7A). Gross retinal morphology remained unaffected between groups. 

Figure 7.

 

Intravitreal injection of anti-EMP2 Ab results in no differences in V_max or retinal thickness in OIR mice, compared to those who received control IgG or no injection. (A) Graph representing decreased V_max measured by ERG at 1 month age in OIR mice treated with 4 µg/g of control IgG at P12 compared to mice treated with 4 µg/g of anti-EMP2 Ab at P12 or controls who received no injections (P = 0.88, n = 4 = 10/group). Data are represented in graphs as means ± SEM. (B) Representative retinal sections stained with H&E, showing no gross retinal abnormalities in OIR mice at P17 after treatment with IVI control IgG, or IVI treatment with anti-EMP2 Ab (top row). Scale bars: 50 µm. Graph showing no differences in total retinal thickness in OIR mice at P17 after treatment with IVI control IgG, or IVI treatment with anti-EMP2 Ab (P = 0.78; n = 4–6/group). Data are represented in graphs as means ± SEM. Representative fluorescein angiography images in OIR mice at one month age who received control IgG or anti-EMP2 Ab IVI (bottom).

Retinal thickness, as measured by H&E staining of retinal sections also revealed no significant differences between groups (control IgG: 194.5 ± 8.3 µm, anti-EMP2 Ab: 190.7 ± 10.2 µm; n = 4–6/group; P > 0.05) (Fig. 7B). FAs in IgG-injected OIR mice qualitatively demonstrate more vessel tortuosity compared to anti-EMP2 Ab-injected OIR mice who demonstrate less aberrant vascularization (Fig. 7B). 

To determine whether there were systemic side effects from IVIs with the anti-EMP2 Ab, the change in weight from P12 to P17 was compared across control IgG injected and anti-EMP2-injected mice. There was no significant difference in weight change between the control IgG-injected group (0.49 ± 0.19 g, n = 7) and the anti-EMP2 Ab group (0.56 ± 0.12 g, n = 10) (P > 0.05) (Fig. 8A). There are conflicting reports about the systemic accumulation of Abs given intravitreally,^24,25 with some groups showing leakage of these biologics particularly in the disease setting. To determine whether anti-EMP2 Ab is prone to systemic leakage, animals were injected intravitreally and sera collected over time to detect the reagent. There was minimal detection of the anti-EMP2 Ab in the serum of mice treated by intravitreal injection (0.13 ± 0.03; n = 3) compared to control mice who received control IgG injection (0.28 ± 0.14; n = 3) (P = 0.3278) (Fig. 8B). 

Figure 8.

 

Intravitreal injection of anti-EMP2 Ab results in no differences in body weight change from P12 to P17 in OIR mice, compared to those who received control IgG or no injection, and is minimally detectable in systemic circulation. (A) Graph showing no differences in change in body weight from P12 to P17 in OIR mice treated with IVI control IgG compared to those treated with anti-EMP2 Ab (P = 0.74; n = 7–10/group). Data are represented in graphs as means ± SEM. (B) Representative western blot images of human heavy chain (Fc) detection at 150 kDa and transferring (Rb) at 70 kDa in mice who received 4 µg/g IVI anti-EMP2 Ab, IVI control IgG, compared to negative controls who received no IVI, and positive controls (IP injection with 20 µg of anti-EMP2 Ab) (top row). Graph demonstrating almost no detection of human Fc as a percentage of transferrin in mouse serum after IVI anti-EMP2 and no difference compared to IVI control IgG group (P = 0.32; n = 3/group). Data are represented in graphs as means ± SEM.

Discussion

There is a need for more specific therapies targeting aberrant neovascularization in retinopathy of prematurity. Although anti-VEGF agents have shown efficacy in improving ROP outcomes in human neonates, significant concerns remain regarding the off-target effects of anti-VEGF exposure in the neonatal population.²⁶ Intravitreally administered anti-VEGF agents have been detected in systemic circulation within one day and suppression of serum VEGF lasts for eight to 12 weeks.²⁷ This persistence raises concerns about the effects of anti-VEGF on developing organs in the neonate. In animal models, this concern has been documented with various anti-VEGF agents, with some intravitreally administered agents showing detrimental effects on body weight gain, the noninjected eye, and renal blood vessels.²⁸ In addition, anti-VEGF agents administered intravitreally do leave the possibility of recurrence of ROP,²⁷ adding to the burden of follow-up needed for infants who receive anti-VEGF injections for ROP. 

Previous studies have shown that Emp2 is significantly upregulated in the neuroretina at P17 in mice exposed to hyperoxia. Moreover, Emp2 knockout mice demonstrate decreased pathologic neovascularization when compared to wild-type mice exposed to hyperoxia.¹¹ Other work has also shown that EMP2 is a regulator of VEGF in adult retinal pigment epithelial cell lines in vitro,¹⁰ which is a known regulator of normal vascular development throughout the entire body,²⁹ suggesting that targeting EMP2 may be a novel method to reduce neoangiogenesis following hyperoxia exposure. To determine whether targeting EMP2 can be utilized for therapy, we performed ex vivo and in vivo testing of an experimental fully human anti-EMP2 IgG1 mAb to attenuate neovascularization in choroid sprouting assays and murine oxygen-induced retinopathy. We have found that anti-EMP2 Ab treatment safely and effectively reduces pathologic neovascularization in the OIR model, without evidence of significant systemic effects or detection in systemic circulation. Importantly, compared to targeting VEGF, EMP2 has a limited tissue distribution,^14,30 allowing for more specific tissue control of neoangiogenesis. 

We have previously described that the mechanisms by which Emp2 genetic knock-out mice attenuate NV may be via downregulation of angiogenic signals, including Hif1α-induced VEGF expression.¹¹ Interestingly, in this study, our RNA sequencing results suggest that other angiogenic pathways may also be downregulated by Ab-mediated EMP2 blockade. Col1a1 has been described as marker gene of a subset of pericytes, which are recruited toward endothelial cells and aggravate hypoxia-induced apoptosis.³¹ Epas1 ± mice are HIF-2α-haplo insufficient and demonstrate protection against retinal neovascularization in the OIR model.³² Jag1 expression, alongside Wnt/β-catenin activation, via the Frizzled-7 pathway has been demonstrated as a potential regulator of pathological retinal NV in the OIR model as well.³³ Non-canonical Wnt5a expression in pericytes and endothelial cell expression of CYCR61-CTGF-NOV 1 (CCN1) regulate each other, and their signaling between pericytes and endothelial cells regulate physiologic and pathologic angiogenesis in murine OIR.³⁴ Neuropilin-1 (Nrp-1) downregulation via galectin-1 silencing (adenoviral-Gal-1RNA interference) in the OIR has been associated with reduced retinal NV.³⁵ Nrp-1 has been described as an obligate receptor for proangiogenic mononuclear phagocyte chemotaxis, necessary for the innate immune response to retinal ischemia driving neovascularization in the OIR model.³⁶ EMP2 blockade in the retina appears to have broad antiangiogenic effects via pericyte- and immune cell-endothelial interactions that underlie retinal neovascularization. 

Interestingly, our retinal RNA sequencing data also demonstrated that EMP2 Ab-mediated blockade resulted in upregulation of fatty acid oxidation pathways and mitochondrial electron transport in OIR. Fatty acids, especially polyunsaturated fatty acids (PUFAs) like arachidonic acid (ARA) and docosahexaenoic acid (DHA), have been studied in retinopathy of prematurity.³ Decreased systemic levels of ARA and DHA are associated with worse ROP,^4,37 and enteral supplementation of ARA and DHA in extremely preterm infants reduces the risk of severe ROP by 50%.³⁸ In animal studies, PUFAs not only serve as important structural components for neuroretinal cell membranes, but also play a role in inflammation and oxidative stress in retinal diseases.³⁹ Acyl-coenzyme A:cholesterol transferase 1 (Acat1), a cholesterol-metabolizing enzyme, is upregulated in microglia in the OIR model, and treatment with the ACAT1 inhibitor, K604, inhibited retinal neovascularization.⁴⁰ ECH1, an enzyme involved in fatty acid oxidation, was identified as an activator of NV using a proteomics approach in the murine OIR model and validated as highly up-regulated in hypoxic human retinal endothelial cells.⁴¹ However, our findings that EMP2 blockade increases Acat1 and Ech1 expression, not decreases them, may contribute to the incomplete attenuation of neovascularization observed in our study. Last, peroxisome proliferator-activated receptor gamma (Pparγ) is upregulated by EMP2 blockade. Pparγ has wide-reaching effects, regulating insulin sensitivity, glucose homeostasis, lipid metabolism, oxidative stress, inflammation and angiogenesis. Pparγ agonists have been demonstrated in experimental models of retinal NV to prevent vascular inflammation and leakage. In OIR, retinal Pparγ expression increases from P17 to P20, inducing M2 microglial polarization and inhibition of inflammatory cytokines, resulting in spontaneous regression of NV.⁴² EMP2 blockade may fuel expression of Pparγ and, in turn, protect against oxidative stress and promote an anti-inflammatory and anti-angiogenic environment, thereby attenuating NV in OIR. 

Our study tested a limited range of anti-EMP2 Ab doses, and although we did not detect any serious systemic or visual adverse effects in mice, a broader range of doses should be tested to optimize efficacy while limiting side effects. Moreover, studies on how long the Ab is detectable in the vitreous would inform optimal timing and the need for repeat doses, as is sometimes indicated in clinical treatment of ROP with anti-VEGF agents. However, this study demonstrates that intravitreal targeting of EMP2 in the retina during phase 1 of ROP may attenuate the severity of NV in phase 2 of ROP in a more specific and safer fashion that anti-VEGF agents. Last, although the murine OIR model has fidelity to the active phases of vessel loss and neovascularization seen in human ROP, it does not recapitulate all aspects of the multifactorial nature of ROP seen in humans. Moreover, the vascular abnormalities seen in murine OIR are reported to resolve by ∼P23, while a broader range of long-term sequelae are reported in human ROP. We did note that, at P17, although there were no differences in weight between control-IgG and anti-EMP2 Ab treated mice, all mice had weights on the low end of ideal weight for this model. It has been reported, in both the murine OIR model and human ROP, that poor postnatal weight gain is associated with more severe neovascularization.^43-45 We postulate that the low weight seen in our P17 mice were due to the stress of intravitreal injections at P12, which may have impaired pup nursing in the periprocedural and postprocedural recovery period. Therefore the efficacy of anti-EMP2 Ab treatment to mitigate NV severity compared to IgG-injected controls may have been in the setting of worse NV, albeit in both groups. Despite these limitations, the murine OIR model remains an important model in which to manipulate genetic aspects of retinopathy and test preclinical therapeutic agents. 

Future studies focused on uncovering the mechanisms by which anti-EMP2 exerts its protective effect on OIR-induced NV are important. Studies are underway to establish the efficacy of anti-EMP2 Ab treatment in blockade of pro-angiogenic factor gene and protein expression from specific cell populations after OIR exposure, such as endothelial and retinal neuronal cells. These studies could inform mechanistic evaluations in vitro. Moreover, functional studies of whether anti-EMP2 Ab treatment may ameliorate breakdown of the blood-retina barrier could provide insight into how EMP2 modulates permeability. 

Acknowledgments

Supported by the NEI R01 EY032561 (to A.C., M.W., L.G.), by NIH/NCI P50-CA211015 (M.W.), NCI R01 CA163971 (M. W.) and NEI P30 EY 00331 (S.N.; PI: David Williams). This work is also supported by an Unrestricted Grant from Research to Prevent Blindness, Inc. to the Department of Ophthalmology at UCLA. 

Disclosure: B. Aguirre, None; M-C. Lin, None; E. Araujo, None; C-H. Lu, None; D. Casero, None; M. Sun, None; S. Nusinowitz, None; J. Hanson, None; K. Calkins, Fresenius Kabi Mead Johnson Nutrition (C), Baxter (C), Prolacta (C), Mead Johnson Nutrition (S); L. Gordon, anti-EMP2 mAb antibody (P); M. Wadehra, anti-EMP2 mAb antibody (P); A. Chu, None