Effect of subretinal injection on retinal structure and function in a rat oxygen-induced retinopathy model

Animals

Animal care and procedures adhered to the Institute for Laboratory Animal Research Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Utah. Time pregnant Sprague Dawley rats (embryonic day 13–17) were obtained from Charles River (San Diego, CA), kept under a 12 h:12 h light-dark cycle, given ad libitum access to food and water, and allowed to deliver the pups naturally. Dams were kept with their litters until postnatal day (P) 28.

Oxygen-induced retinopathy

Within 6 h of birth, neonatal rat pups were supplemented to litter sizes of 12–16 animals from dams with smaller litters born at the same time. Pups with dams were moved into the OxyCyler (BioSpherix, Parish, NY). Oxygen was cycled between 50% and 10% O₂ every 24 h. On P14, the litters were placed in room air until P32 (Figure 1A).

Subretinal injections

Initial procedures were performed to optimize subretinal injections. To assess the extent of the injections, some animals were euthanized by intraperitoneal injection of 100 mg/kg ketamine and 10 mg/kg xylazine after the procedures, and the eyes were dissected and examined. To determine the location of the subretinal fluid, some animals received subretinal injections of a pFmCD44 lentivirus tagged with green fluorescent protein (GFP) that transduces Müller cells at P8 and were imaged at P26 with a Micron IV camera (Phoenix Laboratories, Pleasanton, CA) camera. We chose P8 to administer subretinal injections, because this time point was used in different studies to allow sufficient time for lentiviral transduction in OIR [1,2]. Subretinal injection of 1 μl PBS (pH 7.4 without Ca²⁺ and Mg²⁺ (catalogue # 10010049, Gibco, Invitrogen, Life Sciences, Thermo Fisher Scientific, Rochester, NY) was chosen for this study, as it is a non-toxic alternative to 0.9% normal saline in intravitreal injections in the mouse [12].

At the beginning of the 50% O₂ cycle on P8, animals underwent subretinal injections. The pups were anesthetized with isoflurane using a metered device at an oxygen flow rate of 1-2 l/min, the eyelids were carefully cut open (along the eyelid fissure with a sharp razor blade) and the pupils were dilated with tropicamide eye drops 1% (Bausch + Lomb, Rochester, NY). A 30G needle was used to create a hole in the sclera adjacent to the limbus, and a blunt 33G cannula attached to a Hamilton syringe was advanced into the eye. A slight resistance to the needle indicated Bruch’s membrane was reached. Gently, 1 μl sterile PBS was injected (sPBS-treated). For the pilot-treated eyes, a 30G needle was used to create a hole in the sclera, and a 33G cannula was advanced a short distance into the intravitreal space avoiding the retina and without injecting fluid. Some animals remained untreated without eyelid opening (untreated) or received only opening of the eyelids followed by 30 s exposure to microscope light, which is the time typically needed to administer a subretinal injection (light-treated). Except the untreated or light-treated eyes, all incised eyes received erythromycin ophthalmic ointment 0.5% following the procedures. Litters were returned to the OxyCycler within 1.5 h of the beginning of the procedure.

Retinal imaging

At P26, the animals were anesthetized with ketamine and medetomidine (75/0.5 mg/kg), the pupils were dilated with tropicamide eye drops 1%, and the corneas were lubricated with 0.3% hypromellose gel (GenTeal Eye Gel, Alcon, Novartis, Fort Worth, TX). The objective of a Micron IV retinal camera (Phoenix Laboratories) was advanced near the cornea, and a fundus image was recorded. There was no significant difference in the weights of the animals in untreated or light-treated versus the pilot- or sPBS-treated animals at P28 (p=0.86). We did not observe subretinal bleeding in any of the eyes.

Optical coherence tomography

Spectral domain optical coherence tomography (OCT) was recorded on P30 using Image-Guided OCT (Phoenix Laboratories). The animals were anesthetized using ketamine and medetomidine (75/0.5 mg/kg), the pupils were dilated with tropicamide eye drops 1%, and the corneas were hydrated with 0.3% hypromellose gel. The lens of the OCT was advanced near the cornea, and the retinal image was brought into focus. The measurement area was chosen approximately one diameter away from the optic nerve head. The thicknesses of the retinal layers (photoreceptor outer segments and outer nuclear layer (PRS + ONL), the outer plexiform and inner nuclear layers (OPL + INL), and the inner plexiform and ganglion cell layers (IPL + GCL) and the total retina) were each measured from captured images using Insight software (Phoenix Laboratories). As the OCT measurements were taken at the same distance to the optic nerve head, the data were not normalized to the total retinal thickness, which was not expected to vary substantially.

Flat mounted retinas and immunohistochemistry

The animals were humanely euthanized by intraperitoneal injection of 100 mg/kg ketamine and 10 mg/kg xylazine, followed by cervical dislocation, on P32, and then the eyes were enucleated and dissected free of connective tissue and ocular muscles. For the flat mounted retinas, the eyes were fixed in 4% paraformaldehyde (PFA) in PBS for 30 min and in HistoChoice tissue fixative (Sigma Aldrich, St. Louis, MO) overnight. The retinas were isolated, blocked with 10% normal goat serum in 0.5% Triton-X-100/PBS for 1 h at room temperature, stained with isolectin GS-IB4 conjugated with an Alexa Fluor 488 or 568 fluorophore for 48 h, and mounted with Fluoromount-G mounting media (SouthernBiotech, Birmingham, AL). Images of the flat mounted retinas were collected at a 4X magnification with an Olympus IX81 microscope and stitched together using MetaMorph software (Molecular Devices, Sunnyvale, CA). The total retinal area and the avascular retina and intravitreal neovascularization areas were determined using ImageJ software (NIH, Bethesda, MD). Data are displayed as percentages of the total retinal area: avascular/total retinal area (AVA) and intravitreal neovascular/total retinal area (IVNV). Because the AVA and the IVNV were not different in the light-treated eyes compared to the untreated eyes, the flat mounts for both conditions were combined into one group for statistical analyses.

For IHC, eyes were fixed for 1 h in 4% PFA in PBS, cryoprotected in 10% sucrose for 1 h followed by 30% sucrose overnight, and embedded in optimal cutting temperature compound (Tissue-Tek, Sakura, FineTek, Torrance, CA). From the tissue blocks, 12 μm cryosections were cut. The slides were washed with PBS, permeabilized with 10% normal donkey serum and 0.1% Triton-X 100 in PBS for 45 min, and stained with rabbit anti-opsin antibody (1:100 dilution, catalogue # AB5405, Millipore, Billerica, MA) at 4 °C overnight. Alexa Fluor 488 goat anti-rabbit immunoglobulin G (IgG; Invitrogen) and TO-PRO-3 (1 μM, Invitrogen) were incubated for 60 min at room temperature, and then the slides were mounted with Fluoromount-G mounting media. Images were taken at 20X magnification on an Olympus IX81 confocal microscope. The thicknesses of the photoreceptor outer segments and the outer nuclear layer, the outer plexiform and inner nuclear layers, and the inner plexiform and ganglion cell layers were determined in three different locations of each image using ImageJ and displayed as a percentage of the total retinal thickness to normalize for differences in retinal thickness between the central and peripheral retina.

Electroretinogram

For the scotopic ERGs, the rat pups were dark adapted overnight and handled under red light illumination throughout the procedure. The animals were anesthetized using 100 mg/kg ketamine and 10 mg/kg xylazine. The pupils were dilated with tropicamide eye drops, and the corneas were lubricated with 0.3% hypromellose gel. The animals were placed on a heated pad to maintain body temperature, the subdermal ground electrode was inserted at the base of the tail, and the subdermal reference electrode on the forehead midway between the eyes.

The Ganzfeld ERG was recorded in each eye using the UTAS Visual Diagnostic System with BigShot Ganzfeld (LKC, Gaithersburg, MD). As recordings were made unilaterally, each eye was recorded on separate days, either on P31 or P32 to allow complete dark adaptation in the fellow eye. A contact lens electrode was positioned on the cornea, and waveforms were recorded using white light flashes (5 ms duration, luminances ranging from −5.1 to 2.20 cd s m⁻², with 30 sweeps and 2 s intervals for −5.1 to −3.6, three sweeps and 10 s intervals for −2.4, three sweeps and 20 s intervals for −2.0, three sweeps and 30 s intervals for −1.6, three sweeps and 40 s intervals for −1.2, three sweeps and 60 s intervals for −0.2, and two sweeps and 120 s intervals for 1.2 to 2.2 cd s m⁻²).

Because corneal and lens opacification occurred during the time needed for anesthesia to record eyes on the same day, the focal ERG (fERG, image-guided focal ERG, Phoenix Laboratories) was recorded on P28 or P29 to permit clear media to evaluate the retinal function in each eye. The lens electrode was carefully advanced near the cornea, and the recording area in the central retina (spot size A 0.50 mm, approximately one diameter away from the optic nerve head) was chosen under red light illumination. White light flashes (5 ms duration, luminances ranging from −0.4 to 3.2 cd s m⁻², with 20 sweeps and 10 s intervals for −0.4 to 1.4, ten sweeps and 20 s intervals for 2.0, three sweeps and 60 s intervals for 2.6, and two sweeps and 2 s intervals for 3.2 cd s m⁻²) were used to record the fERG. Scattered light was confirmed not to contribute to the fERG response by aiming the highest intensity light stimulus at the optic nerve head (data not shown).

Raw data were stored on a personal computer for off-line analysis. The a- and b-wave amplitudes and latencies were determined (Figure 1B) and are displayed as means ± standard error of the mean (SEM). Note that a- amplitude is usually negative but is recorded as a positive absolute value for ease in comparison.

Statistical analysis

Previous studies observed the AVA mean ± standard deviation (SD) of 47.62±10.12 and 32.36±11.43 in two conditions analogous to the present study [1,13]. Assuming a similar effect size for the AVA, we required nine eyes in each condition to achieve 80% power using a two-sided alpha 0.05 comparison. The same studies observed an IVNV mean ± SD of 3.62±1.71 and 0.72±1.56. Assuming a similar effect size, we required seven eyes in each condition to achieve 80% power using a two-sided alpha 0.05 comparison. With clustering, if the study outcome in the unit of analysis (eyes) within a cluster (rat) is correlated, measured by the intraclass correlation coefficient (ICC), the sample size must be increased by the design effect, which includes the ICC in its calculation. If the ICC is 0, where the eyes are independent, adjustment of the design effect is not necessary. We assumed the ICC would be 0, or at least negligible, as the intervention was specific to a response in the eye itself, rather than systemic; therefore, we kept the original calculation of nine eyes per condition. For analysis of the AVA and the IVNV, therefore, nine flat mounted retinas were used per condition. The thickness of the retinal layers was measured in four untreated, seven pilot-treated, and five sPBS-treated retinas with OCT and in three locations in each of three images recorded in two sections from each of two animals per condition measured with IHC. Ganzfeld ERGs were recorded in nine eyes from five untreated animals, ten eyes from seven pilot-treated animals, ten eyes from eight sPBS-treated animals, and six eyes from three light-treated animals. fERG data were collected in nine eyes from five untreated animals, ten eyes from eight pilot-treated animals, eight eyes from eight sPBS-treated animals, and six eyes from three light-treated animals.

Some data clustering existed in the data set. Pups from the same litter were randomized to different conditions, or study groups, to avoid clustering by litter within any condition. In this way, the groups were balanced on any possible litter effect. For IHC, within an eye, the three locations in each of the three images from two sections were averaged to obtain a single data point, or summary measure, per eye so clustering was avoided. Clustering remained for Ganzfeld and focal ERGs when both eyes were used in some pups. To account for this clustering of eyes within the same pup, all statistical comparisons between conditions were performed using a mixed-effects linear regression, with the eyes nested within the pup. For graphing, SEMs were obtained using post-test marginal estimation from the mixed-effects models, so that the SEMs correctly accounted for the data clustering. This statistical modeling was done using Stata-14 statistical software (StataCorp, College Station, TX).

Given that the Ganzfeld ERG and fERG were measured at several light intensities, we used the primary-secondary approach to address multiple comparisons [14-16]. We selected the highest light intensity as the primary outcome and then tested our research hypothesis at just this highest intensity. In this way, there was only one comparison; thus, there was no need for a multiple comparison adjustment. The lower light intensities were considered secondary comparisons, which are merely descriptive, and nonconfirmatory. As no hypothesis was tested with them, there was no need for multiple comparison adjustment for them. The primary-secondary approach to multiple comparisons is a commonly used approach in randomized controlled trials reported in The New England Journal of Medicine (e.g, [17]).

Delivery of subretinal injections and fundus images

The fundus images of untreated, pilot-treated, or sPBS-treated eyes in P26 animals had similar appearances (Figure 2A–C). The estimated extent and location of retinal involvement from the subretinal injections were determined in eyes from pups that were euthanized after injection and examined for the extent of retinal detachment or that had received subretinal injections of a pFmCD44 lentivirus tagged with GFP that transduces Müller cells at P8 and imaged at P26 to assess the location of transduction with the fluorescence of the Müller cell endfeet visualized with the Micron IV camera. From these experiments, green fluorescence was determined in about one third of the posterior retina in approximately 80% of the injected eyes (Figure 2D–F).

Retinal vascularization and structure

Compared to the untreated eyes, the AVA was increased in the pilot-treated eyes and more so in the sPBS-injected eyes at p32 (Figure 3A,B). The IVNV was increased in the sPBS-treated eyes, but not in the pilot-treated eyes, compared to the untreated controls (Figure 3A,C). No statistically significant difference in the IVNV was found between any of the remaining groups (Figure 3).

The OCT on P30 and immunohistochemically labeled cryosections on P32 showed reductions in the thickness of the PRS + ONL in the pilot- and sPBS-treated eyes compared to the untreated eyes. No statistically significant differences were detected in the OPL + INL, IPL + GCL, or total retinal thickness between any of the treatment groups (Figure 4 and Figure 5).

Electroretinography

The amplitudes of the a- and b-waves in the scotopic Ganzfeld ERG were statistically significantly reduced in the pilot-treated eyes compared to the untreated eyes. The b-wave amplitude was relatively increased in the sPBS-treated eyes compared to the pilot-treated eyes and was not statistically significantly different from in the untreated eyes. The a-wave amplitude was statistically significantly smaller in the sPBS-treated eyes compared to the untreated eyes, but there was no statistically significant difference compared to the pilot-treated eyes (Figure 6). To determine whether the subretinal PBS injections affected retinal function in the injection region, we recorded focal ERGs in the central retina within an approximate one disc diameter of the optic nerve head. The amplitudes of the a-wave of the focal ERG were reduced in the pilot- and sPBS-treated eyes compared to the untreated eyes, but the amplitudes of the b-wave in the focal ERG were similar in all groups (Figure 7).

The eyelids must be opened in P8 rats from the OIR model to perform injections. To determine whether suppression of the a- and b-wave amplitudes in the scotopic ERGs of the pilot-treated eyes was due to light exposure, some eyes had eyelid opening followed by 30 s exposure to the microscope light only. The light-treated eyes had similar amplitudes of a- and b-waves on the focal and Ganzfeld ERGs compared to the untreated eyes, showing that this time duration of light exposure did not affect retinal function (Figure 8).