Inhibition of retinal neovascularization by a PEDF-derived nonapeptide in newborn mice subjected to oxygen-induced ischemic retinopathy

Abstract

Retinopathy of prematurity (ROP) is a growing cause of lifelong blindness and visual defects as improved neonatal care worldwide increases survival in very-low-birthweight preterm newborns. Advancing ROP is managed by laser surgery or a single intravitreal injection of anti-VEGF, typically at 33-36 weeks gestational age. While newer methods of scanning and telemedicine improve monitoring ROP, the above interventions are more difficult to deliver in developing countries. There is also concern as to laser-induced detachment and adverse developmental effects in newborns of anti-VEGF treatment, spurring a search for alternative means of mitigating ROP. Pigment epithelium-derived factor (PEDF), a potent angiogenesis inhibitor appears late in gestation, is undetected in 25-28 week vitreous, but present at full term. Its absence may contribute to ROP upon transition from high-to-ambient oxygen environment or with intermittent hypoxia. We recently described antiangiogenic PEDF-derived small peptides which inhibit choroidal neovascularization, and suggested that their target may be laminin receptor, 67LR. The latter has been implicated in oxygen-induced ischemic retinopathy (OIR). Here we examined the effect of a nonapeptide, PEDF 336, in a newborn mouse OIR model. Neovascularization was significantly decreased in a dose-responsive manner by single intravitreal (IVT) injections of 1.25-7.5 μg/eye (1.0-6.0 nmol/eye). By contrast, anti-mouse VEGFA₁₆₄ was only effective at 25 ng/eye, with limited dose-response. Combination of anti-VEGFA₁₆₄ with PEDF 336 gave only the poorer anti-VEGF response while abrogating the robust inhibition seen with peptide-alone, suggesting a need for VEGF in sensitizing the endothelium to the peptide. VEGF stimulated 67LR presentation on endothelial cells, which was decreased in the presence of PEDF 336. Mouse and rabbit eyes showed no histopathology or inflammation after IVT peptide injection. Thus, PEDF 336 is a potential ROP therapeutic, but is not expected to be beneficial in combination with anti-VEGF

Introduction

Retinopathy of prematurity (ROP) is an important cause of childhood blindness globally, with rising incidence (Zin and Cole, 2013). While estimated to occur in 10% of extremely preterm infants in high-income countries, this may approach 20-40% in developing ones (Quinn, 2016). Infant retinal neovascularization in ROP often regresses spontaneously but in many cases, progression to critical decision points requires immediate control by laser ablation or local VEGF antagonism (Stark et al., 2018). Laser treatment destroys major peripheral parts of the retina and anti-VEGF in developing eyes leads to short-term delayed vascular growth and long-term decreased function with neuroretinal changes (Tokunaga, et al., 2014). Anti-VEGFs are sustained in low-volume infant circulation (Kong et al., 2015), and may cause systemic suppression of normal vascularization, and appropriate organ development (Clayton et al., 2008). Visual field damage and choroidal degeneration may lead to long term visual deficits. These concerns impel a search for therapeutic alternatives.

Pigment epithelium-derived factor (PEDF) is an endogenous anti-angiogenic and neuroprotective glycoprotein that maintains angiogenic balance in normal tissues, including the eye (Amaral and Becerra, 2010; He et al., 2015). Decreased PEDF in vitreous is associated with proliferative diabetic retinopathy (PDR) (Yokoi et al., 2007; McAuley et al., 2014), this extends to ROP since PEDF is absent in low birthweight preterm vitreous, but is detected in full term vitreous (Sugioka et al., 2017), consistent with placental appearance of PEDF occurring late in gestation (Loegl et al., 2016). Oxygen-induced ischemic retinopathy (OIR) models in newborn animals are considered representative of ROP (Hartnett, 2017), this was exacerbated in PEDF-null mice and attenuated in mice expressing PEDF transgene (Huang et al., 2008; Park et al., 2011a). Recombinant-human PEDF (20nM) strongly suppressed VEGF-induced retinal microvascular endothelial cell (EC) proliferation and migration, 2 μg PEDF by intravitreal (IVT) injection inhibited retinal neovascularization (NV) in mouse OIR by > 80% (Duh et al., 2002).

PEDF potently induces apoptosis in VEGF-stimulated, but not in quiescent EC (Volpert et al., 2002) through molecular pathways that involve Fas, PPARγ, ERK and Wnt signaling (Wang et al., 2009; Biyashev et al., 2010; Park et al., 2011b; Zhang et al., 2016). This activity has been localized to a PEDF 34-mer peptide, (Filleur et al., 2005), repeated subconjunctivial injection of which was required to decrease NV in rat OIR (Amaral & Becerra, 2010). Local delivery of 34-mer in neonatal mouse OIR was recently revisited, similarly requiring multiple IVT injections. However improved antiangiogenic activity and single dose IVT efficacy was achieved by formulating 34-mer with 0.1% collagen (Kim et al., 2019), suggesting suboptimal physicochemical properties of the peptide. Daily parenteral 34-mer administration inhibited NV in neonatal mouse OIR via decreased recruitment of circulating EC precursor cells (Longeras et al., 2012).

An 18-mer within the 34-mer was also found to be antiangiogenic (Mirochnik et al., 2009), and we recently discovered a much smaller epitope within this sequence, and described modified 8 and 9 amino acid peptides, PEDF 335 and PEDF 336, that inhibited laser-induced chroidal neovascularization (CNV) via their IVT injection 2-5 days preceding laser treatment (Sheibani et al., 2019). These potent 8 and 9-mers defined an anti-angiogenic PEDF epitope corresponding to G-Y-D-L-Y-R-V in the parent peptide. Optimizing modifications gave the sequence adipic-Sar-Y-N-L-Y-R-V common to PEDF 335 and PEDF 336 (Sheibani et al., 2019). Their shared active motif, Y - - - R, is found in laminin (Ln) β1 peptide YIGSR. This pentapeptide was the first prototypic ligand of non-integrin laminin receptor, 67LR, also known as LAMR1, which plays an active role in tumor cell invasiveness and angiogenesis (Donaldson et al., 2000). While several receptors have been described as mediating PEDF antiangiogenic activity (He et al., 2015), including 67LR, LRP6 via Wnt signalling, and F1-ATP synthase (Park et al., 2011b; Deshpande et al., 2012), only 67LR, has thus far been shown to bind to the 34-mer and to a 25-mer fragment therein with high affinity (Bernard et al., 2009).

Here we examined PEDF 336, an N-terminal adipic 9-mer, in a mouse model of OIR. We examined dose-response to a single injection of peptide at the transition from high oxygen to room air, and compared efficacy with that of bevacizumab (anti-human VEGF mAb) and of anti-mouse VEGFA₁₆₄ (R&D Ab493), then to a combination of Ab493 and peptide.

2. Materials and Methods

2.1. Reagents

Antiangiogenic nonapeptide PEDF 336: adipic-Sar-Tyr-Asn-Leu-Tyr-Arg-Val-Arg-Ser-amide has recently been described as mitigating laser-induced CNV in mice (Sheibani, et al., 2019). Its sequence is derived from the PEDF 34-mer sequence …Gly-Tyr-Asp-Leu-Tyr-Arg-Val-Ser…, having Gly¹ replaced by sarcosine (Sar, N-methyl-Gly), and change of anionic Asp³ to Asn, net charge maintained via N-terminal capping as an adipic half-amide (HOOC-(CH₂)₄-CO-NCH3R) in carboxylate form. The carboxylic acid form was prepared using solid state peptide synthesis, then di-HCl salt formation by double lyophilization from 10 mM HCl. Peptide identity (MW 1,268.4) and purity were confirmed by mass spectrometry (parent ion M⁺: 1,268, M⁺/2: 635) and HPLC, indicating > 96% purity. Solutions of PEDF 336 from 7.5-100.0 mg/mL (6-80 mM) in 1 mL water have acidic pH 2-3.5, and were neutralized to pH 6-7 before further formulation, by addition of 6-80 μL 1M NaHCO3 to give final stocks from 7.5-90 mg/mL (6.0-72 mM). Further dilution with vehicle and water/saline was then carried out, and final concentrations were established via UV spectrum based on extinction of 2,200 M⁻¹ cm⁻¹ at 276 nm.

Clinical grade bevacizumab (Avastin, Genentech, 100 mg) was dissolved in 4.0 mL of sterile water per package insert. Avastin vehicle (AV) was obtained by ultrafiltration of 0.5-1.0 mL of this 25 mg/mL Avastin solution in a sterile Amicon ultracentrifuge filter, MWCO 10kDa. UV analysis showed that < 0.2% of the applied protein remained in this filtrate, this vehicle contains phosphate, trehalose being the main source of osmolarity. Peptide and antibodies were IVT-injected in injection-vehicle comprised of the above AV and and equal volume of isotonic saline. Small volumes of 2 M sterile NaCl or sterile water was used to adjust all test solutions to within +/−5% of the osmolarity of isotonic saline, via osmometry. Affinity-purified polyclonal goat IgG, anti-mouse VEGFA₁₆₄ was from R&D Systems (AF-493-NA, 100 μg), referred to as Ab493, was dissolved at 200 μg/mL in isotonic saline, then diluted with AV/saline (1:1; v:v) to give test solutions ranging from 2-50 μg/mL in the injection-vehicle used in all experiments, with final osmolarity adjustment as described. All test solutions were sterile filtered and stored at 4°C.

2.2. OIR Model

All the animal studies were performed in accordance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research and approved by the Institutional Animal Care and Use Committee of the University of Wisconsin School of Medicne and Public Health (IACUC assurance number D16-00239). For these studies C57BL/6J mice (Jackson Laboratory, Bar Harbor, ME) were used. One week-old (postnatal day 7; P7) and the mother were exposed to 75% oxygen for 5 days as previously described (Jamali et al. 2017). Following hyperoxia treatments, P12 mice were transferred to room air, and half of the litter received 1 μL of test material and the other half received the AV/saline vehicle control in both eyes by IVT injection. Mice were sacrificed 5 days later (P17), retinal flatmounts were prepared and stained with anti-collagen IV antibody (AB756P; EMD Millipore, Burlington, MA), and images were obtained using a Zeiss microscope. The non- perfused areas and degree of vascularization were determined as previously described (Jamali et al. 2017).

2.3. VEGF and PEDF effects on 67LR expression in mouse EC

Mouse retinal and choroidal EC were prepared and propagated as previously described by us (Su et al., 2003; Fei et al., 2014). Cells were plated on glass coverslips and allowed to grow for 1 day in regular growth medium. Cells were then serum-starved using serum-free growth medium overnight. The next day cells were incubated with recombinant mouse VEGF (50 ng/ml; 450-32, Peprotech, Rocky Hill, NJ) in serum-free growth medium for 24 h. In another set of experiments, cells were incubated with VEGF and with or without 336 peptide (10 μM; in serum-free growth medium) for 24 h. The expression of 67LR on the cells was assessed by immunofluorescence using Alexa 647-tagged monoclonal antibody (Laminin-R (H2), Santa Cruz Biotech). Briefly, coverslips were washed with PBS, and cells were fixed using 4% paraformaldehyde for 10 min on ice. Coverslips were washed with TBS three times for 5 min each and blocked by 1% BSA in TBS for 30 min at room temperature. The cells were incubated with the conjugated anti-67LR (1:100, diluted in 1% BSA in TBS) at 4°C overnight. Coverslips were washed and mounted. The images were captured in digital format using a Zeiss microscope (Carl Zeiss, NY). The mean fluorescence intensities were quantified in at least 8 high power fields using ImageJ. The data represent Mean ± SD from 3 biological replicates.

2.4. Safety testing in rabbit eyes

Both eyes of New Zealand white rabbits, under anesthesia, were injected intravitreally with 30 μL solutions of 1:1(v:v) AV:saline injection-solution, or Avastin:saline, or Avastin:saline containing from 3 mM to 36 mM PEDF 336 (final concentration). For the latter, freshly neutralized 2× PEDF 336 stock solutions at 6.0, 18.0, 36.0 and 72.0 mM were prepared in isotonic saline. Each peptide stock solution was then diluted 2-fold by mixing with an equal volume of Avastin (25 mg/mL). In one cohort (20 rabbits) eyes were resectioned and harvested after euthanasia at day 4, while eyes from a second cohort (22 rabbits) were similarly obtained at 14 days postinjection. Four eyes in each cohort received each test solution, no individual rabbit received identical test solutions in both eyes. IVT injections (30 μL) were carried out in 3-4 kg rabbits acclimated for at least 7 days. Final peptide concentrations for injection were 3.75, 11.25, 22.5, and 45.0 mg/mL (3, 9, 18, or 36 mM). Controls included 12.5 mg/mL Avastin (50% saline by volume) and vehicle-only for each cohort. For the test set evaluated at 4 days after injection (40 eyes), final doses of PEDF 336 per eye were 0.11, 0.34, 0.68 mg (90, 270 and 540 nmol) while another test set (44 eyes), evaluated at 14 days after injection, additionally received 1.36 mg/eye (1,080 nmol) in 4 eyes. Based on 3.5 μL P12 mouse IVT volume (Fortmann, et al, 2018) and a 1.4 mL average of estimated IVT rabbit volumes (Ahn et al., 2016), we proposed a 400-fold proportional dose in rabbit vs. newborn mouse. The highest rabbit dose (1.36 mg) is 360× the highly efficacious 3.75 μg delivered to mouse, essentially 90% of an equivalent dose after volume correction. Eyes were fixed in formaldehyde, embedded in paraffin and pupil - optic nerve sections obtained and stained with hematoxylin and eosin using standard techniques as previously described (Melgar-Asensio et al., 2018; Sheibani et al., 2019).

2.5. Safety testing in mouse eyes

Newborn mice not exposed to hyperoxia were examined for PEDF 336 safety-histopathology. On postnatal day 12 mice received 1 μL injections of 3.0 or 6.0 mM PEDF 336, dissolved in a 1:1 (v:v) mixture of AV:saline or AV:saline vehicle control in both eyes. Mice were sacrificed at P17 and eyes were subjected to histopathological evaluation as previously described (Zaitoun 2018, et al.; Sheibani et al., 2019).

2.6. Statistical analysis

Statistical differences between groups were evaluated with the One-way ANOVA followed by Tukey’s multiple comparison test using GraphPad Prism version 5.4 for Windows (GraphPad Software, La Jolla, CA). Statistical differences were confirmed with Bonferroni’s comparison of selected pairs of columns and Student’s unpaired t-test (two-tailed). Mean ± SD is shown. P< 0.05 was considered significant.

3. Results

3.1. OIR inhibition by PEDF 336

IVT injections in newborn mice were always 1 μL, using peptide solutions of 1.0, 3.0 or 6.0 mM. Thus, eyes received a single injection, on the day of transition from high-to-normal oxygen, of 1.25, 3.75 or 7.25 μg/eye of peptide corresponding to 1, 3 or 6 nmol. Peptide inhibition of OIR is shown on Figure 1 A,B,C. The lowest dose, 1.25μg/eye, gave a statistically significant decrease (p = 0.04) of approximately 25% in neovascular area (NV) as a percent of either total area or of vascularized area. Vessel obliteration (VO) was not significantly altered at this dose. At the dose of 3.75 μg/eye (3nmol) the NV, as a percent of total area or as percent of neovascular area, was decreased by more than 50% (P values from 0.002 to 0.005). There was also statistically significant 30-50% reduction of VO area (P< 0.006). OIR inhibition was reproducibly observed at this dose and was maintained or improved at the highest dose of 7.5 μg/eye (6 nmol). The latter gave > 55% inhibition of NV as percent of total or vascular area (P = 0.003), with VO decreased very significantly (P = 0.0007) by 70%. Vessel obliteration is a response to exposure to hyperoxia, and is generally associated with severity of NV during room air exposure. Maximum NV occurs at P17 and this time point is used for assessment of neovascularization/vascular tufts in the retina. VO is normally measured at P12 prior to the room exposure and sensing of the ischemia. The VO, as measured in P17 mice here, also may indicate the aspects of normal retinal vascularization of vaso-obliterated area in response to ischemia. Thus, reduced vaso-obliterated area indicates normal vascularization of retina under different treatment conditions. PEDF 336 displayed dose-dependent efficacy in OIR inhibition over the 6-fold range tested.

Figure 1.

Retinal antiangiogenic activity of PEDF 336 peptide. Mice (P7 pups and mother) were subjected to OIR as detailed in Methods. After 5 days exposure to hyperoxia mice were brought to room air and divided to 2 groups. Each group received a single intravitreal injection (1 μL) of vehicle (control) or peptide (treated) at different concentrations: A (1 mM), B (3 mM), C (repeat of B; 3 mM), and D (6 mM). Mice (P17) were sacrificed, wholemount retinas were prepared and stained with anti-collagen IV antibody to visualize the retinal vasculature, and images were captured in digital format for assessments of vessel obliteration and neovascularization (n≥ 4 mice per group; *P< 0.05, **P< 0.01, and ***P< 0.001).

3.2. Anti-VEGF effects on OIR

Anti-VEGF effects on OIR were first examined with anti-human VEGF (Avastin), then polyclonal anti-mouse VEGF (Ab493). Avastin did not significantly inhibit any parameter at 6 or 12.5 μg per mouse eye, although there is a trend to lowered NV at the higher dose approaching, but not reaching statistical significance. Thus, response to Avastin was considered minimal in our model, as shown in Figure 2. Polyclonal Ab-493 was then examined with 1 μL doses ranging from 2-50 ng/eye. There was no effect seen, compared to controls, at 2 or 5 ng doses (data not shown). Figure 3 A,B,C show the effects of Ab-493 at doses of 12.5, 25, and 50 ng per eye. Only the 25 ng dose gave statistically significant decreases in NV (P=0.026, % total; P=0.021 % vascular) of approximately 60%. This dose similarly decreased VO (p=0.042). However half this dose (Fig. 3A) gave no statistically significant reduction in any of the above parameters. Notably, doubling the active dose (Fig. 3C) to 50 ng produced no signs of lowered OIR. Thus while active at a single dose (25 ng/eye), Ab-493 displayed no dose-response when halving or doubling the active dose, indeed a steep V-shaped relationship was noted.

Figure 2.

Retinal antiangiogenic activity of human anti-VEGF antibody. Mice were subjected to OIR as detailed in Figure 1 legend. The control group received vehicle and treatment group received a single injection (1 μL) of anti-human VEGF (Avastin): A (6 mg/ml) and B (12.5 mg/ml). The degree of vessel obliteration and neovascularization was assessed from wholemount images as detailed in Methods (n≥ 4 mice per group).

Figure 3.

Retinal antiangiogenic activity of mouse anti-VEGF antibody. Mice were subjected to OIR as detailed in Figure 1 legend. The control group received vehicle and treatment group received a single injection (1 μL) of anti-mouse VEGF: A (12.5 μg/ml), B (25 μg/ml), and C (50 μg/ml). The degree of vessel obliteration and neovascularization were assessed from wholemount images (n≥ 4 mice per group; **P< 0.01).

3.3. Combined PEDF 336 and Ab493 in OIR

PEDF 336 was formulated in a mixture with Ab493 where their concentrations were 3.75 mg/mL (3.0 mM) peptide and 25 μg/mL (0.17 μM) protein. Thus a 1 μL IVT injection simultaneously delivers 3.75 μg of peptide and 0.025μg anti-mouse VEGF (3.0 nmol and approx. 0.017 nmol, respectively). Each of these alone had been shown to significantly decrease OIR (Fig. 1B, Fig. 3B). As seen in Figure 4, the combination treatment showed decreases in NV similar to either treatment alone but only displayed the modest statistical significance associated with Ab493 but not the robust P values seen with this dose of PEDF 336 alone. We conclude that VEGF availability, abrogated by anti-VEGF, is required for peptide efficacy. This is consistent with the VEGF requirement reported for PEDF antiangiogenic activity (Duh, et al., 2002) and the original selection of PEDF 336 through its apoptosis of VEGF-activated EC, while unactivated EC are not induced to apoptosis (Sheibani et al., 2019). The activity seen in Figure 4 also sugests that PEDF 336 does not interfere with antibody (Ab-493) mitigation of OIR.

Figure 4.

Retinal antiangiogenic activity of anti-mouseVEGF alone and combined with PEDF 336 peptide. Mice were subjected to OIR as detailed in Figure 1 legend. The control group received vehicle and treatment group received a single (1 μL) injection of anti-mouse VEGF alone, or mixed with 3 mM PEDF 336 peptide: (A) anti-mouse VEGF (25 μg/ml) or (B) anti-mouse VEGF(25 μg/ml) and PEDF 336 peptide (3 mM). The degree of vessel obliteration and neovascularization was assessed from whoiemount images (n≥ 4 mice per group; *P< 0.05).

3.4. Expression of 67LR in mouse EC

We previously optimized PEDF peptides against human microvascular EC (HMVEC) measuring their VEGF-dependent induction of apoptosis. Since anti-mouse VEGFA₁₆₄ (Ab493) appeared to abrogate mouse retinal NV protection by PEDF 336, we examined the effect of VEGF on peptide induction of mouse EC 67LR expression and how PEDF 336 affects this expression. Both retinal and choroidal EC showed basal expression of 67LR, which was more prominent in choroidal EC. This is consistent with the enhanced sensitivity of these cells to PEDF peptide at 50-100 μM compared to retinal EC (Sheibani et 2019). We observed enhanced expression of 67LR in retinal and choroidal EC incubated with VEGFA₁₆₄. In addition, this enhanced 67LR expression in choroidal EC was attenuated in the presence of PEDF 336 (10 μM; Fig 5). Furthermore, we recently observed VEGF enhancement of tumor cell surface presentation of 67LR (not shown). Thus, 67LR displayed on the cell surface membrane may be similarly affected by VEGF in cells that express this receptor. We propose that the interaction of PEDF or its peptide with 67LR promotes receptor internalization (decreased cell surface expression) initiating the downstream events that lead to loss of viability. Exposure of VEGF treated EC, displaying elevated 67LR, to PEDF 336 resulted in a decrease in expression of 67LR on the cells, supporting the notion of peptide dependent removal of 67LR from the cell surface and enhanced signaling.

Figure 5.

Expression of 67LR on the choroidal and retinal EC. Choroidal and retinal EC were subjected to immunofluorescence assessment of 67LR expression under various conditions as detailed in Methods. (A and B) Immunofluorescence representative images of control cells or those incubated with VEGF and/or VEGF plus PEDF 336. Please note the basal expression of 67LR in these cells, which is more prominent in choroidal EC, and its decreased expression on choroidal EC incubated with PEDF 336. (C-E) The quantitative assessment of data. Mean fluorescence intensities under various conditions were assessed using at least 8 high power fields (×200). Scale bar = 200 μm, *P < 0.05, **P < 0.01, ***P < 0.001.

3.5. Safety

The lowest effective dose of PEDF 336 tested in mice was 1.25 μg/eye (1 nmol). Rabbit eyes have approximately 400-fold greater vitreous volume than newborn mouse eyes, 1,400 μL vs. 3.5 μL (Ahn et al., 2016; Fortmann et al., 2018). Thus, a corresponding rabbit eye dose, corrected to volume, would be 500 μg/eye (400 nmol). Rabbit eyes received from 110-1,360 μg, about 0.22-to 2.7-fold the minimum active dose in mice as adjusted to rabbit eye volume. Higher peptide dosing was precluded by limitation of peptide solubility and maintaining low volume delivery into the rabbit eye. No abnormal pathology and no signs of inflammation attributable to peptide was observed in any rabbit eye, where injection of each test agent was replicated in four eyes at both 4 days and at 14 days post injection. The 30 μL peptide injections in rabbit eyes also contained 12.5 mg/mL Avastin (375 μg delivered). No effect of Avastin:saline was noted compared to AV:saline controls.

Also examined was safety of the two highest doses of peptide 336, administered to newborn mice by IVT injection, with no antibody included. Figure 6 shows representative eyes evaluated for histological and morphological changes, without OIR. Histological evaluations showed no abnormalities in organization of the retinal layers and lacked any sign of inflammatory responses. Wholemounts were also prepared to examine adverse effects on postnatal retinal vascular development and integrity. We observed no adverse effects on these parameters under all evaluated conditions. Thus, no evidence of toxicity to newborn mouse eyes was found at the highly efficacious dose of 6 nmol per eye (7.5 ng/eye), harvested 5 days after injection. Figure 6A shows representative images from histological sections through the optic nerve at different magnifications. The average retinal thickness was obtained by determining retinal thickness on either side of the optic nerve every 100 μm. The quantitative assessment of data, representing similar retinal regions from control and treated eyes, are shown in Figure 6B. No significant differences were noted in thickness of retinas prepared from control and treated animals. Figure 6C shows retinal flatmount stained with anti-collagen IV to evaluate the development and integrity of retinal vasculature in control and treated animals. No dramatic changes in development, organization, and apparent density of retinal vasculature was noted in retinas from treated animals compared with control.

Figure 6.

Evaluation of adverse impact of treatment with PEDF 336 or anti-mouse VEGF. Mice (P12) kept in room air were divided to four groups (n≥ 5 mice per group) receiving 1 μL intravitreal injection of vehicle (control), different doses of PEDF 336 peptide (3 mM or 6 mM) or anti-mouse VEGF (25 μg/ml). Animals were sacrificed 5 days later and one eye was used for histological evaluation (A) and retina thickness analyses (B), and the other eye was used for retinal wholemount preparation to evaluate the integrity of retinal vascularization. The retinal sections thickness were determined as detailed in Methods. No significant differences were noted among the gropus. (n≥ 4 mice per group; P> 0.05). Please note no structural abnormalities in retinal integrity (A) or its vascularization were observed (C).

Discussion

We have previously reported inhibition of laser-induced CNV in adult mice by IVT injection of antiangiogenic 8 and 9 amino acid peptides, derived from PEDF (Sheibani, et al., 2019). Here this is extended for the 9 mer, PEDF 336, to mitigation of neovascularization in newborn mouse during OIR induced by transition from high to ambient oxygen. Neonatal mouse OIR is a highly reproducible model for study of retinal neovascularization in which exposure to high oxygen results in retinal vessel obliteration (VO) with attenuated retinal neovascularization (Kim et al., 2016). Return of animals to normal oxygen at P12 results in an ischemic response from which pathological retinal NV is believed to arise from elevated reactive oxygen species generated by NADPH oxidase (Wang et al., 2014; Wilkinson-Berka et al., 2014) formed under hyperoxia, the ensuing NV then peaks at P17. We observed the significant attenuation of retinal NV at P17 by IVT- injected PEDF peptide P336 on P12. We have previously shown that choroidal EC function is significantly different from that of retinal EC. For example choroidal EC are fenestrated and respond to lack of thrombospondin-1, an endogenous inhibitor of angiogenesis, in a different manner than retinal EC (Fei, et al., 2014). P336 is effective in attenuating neovascularization in both choroid and retina, based on our previously reported CNV mitigation (Sheibani et al., 2019) and the OIR results described here, taken together. This implies that PEDF similarly affects both types of neovascular condition.

A number of neonatal rodent OIR studies employing hyperoxic exposure from P7-P12, followed by room air have monitored retinal PEDF. In rats retinal PEDF underwent an initial increase (> 3-fold) at P12 but is diminished to control level by P14, with 2-fold further decrease at P16 when the VEGF/PEDF ratio was maximal (Gao, et al., 2001). In a similar mouse model western blot showed lowered PEDF at P17 (Al-Shabwarey et al., 2011). A rat OIR study reported a similar time-dependent pattern with PEDF (>3-fold) maximal decrease at P17, remaining below control even at P22 (Lei et al., 2015).

OIR in mice led to decreased glutamine synthetase and Gln transporter, with increased IL-1β at P17, compared to controls, where 1 μL injection of PEDF (2 μg) at P12 or P14 normalized these to untreated values (Wang et al., 2015). While this is more related to protection from neuronal damage and inflammation than angiogenesis, the results further support the need to compensate a PEDF deficit from P12-P14. Since PEDF drops precipitously in just 2 days after return to room air (Gao et al., 2001), sustaining antiangiogenic activity over the ensuing 5-day period may be essential to preserving normal vascular development since PEDF reaches a nadir at P16-17. This may be clinically impactful in the context of late gestational appearance of PEDF (Sugioka et al, 2017; Loegl et al., 2016). Indeed, ROP evolves in about 50% of extremely early gestational age neonates (ELGANs), then clinical intervention hinges on a very urgent 48h decision where critical ROP progression fails to reverse in approximately ½ of ROP patients from 29 -31 weeks postmenstrual age (Miller et al., 2014; Stark et al., 2018).

PEDF 336 is adipic-Sar-Tyr-Asn-Leu-Tyr-Arg-Val-Arg-Ser-amide (compare to PEDF sequence ..Gly-Tyr-Asp-Leu-Tyr-Arg-Val-Arg-Ser..). In CNV inhibition studies of this peptide and the related 8-mer, PEDF 335 (adipic-Sar-Tyr-Asn-Leu-Tyr-Arg-Val-Pro-ethylamide), we replaced an internal Asp residue with neutral Asn, while appending adipate to the N-terminus to enable prodrug linkage to polymeric carriers (Sheibani, et al., 2019). PEDF 336 was selected initially by its potency in the induction of apoptosis in microvascular endothelial cells (EC) activated with VEGF, but not affecting EC without added VEGF (Sheibani et al., 2019). PEDF 336 displayed dose-responsive efficacy mitigating OIR in newborn mice receiving from 1.25-7.5 μg/eye (1-6 nmol), a 6-fold concentration range, with robust statistical significance at 3.75 and 7.5 μg/eye as seen in Figure 1.

Since single injection of anti-VEGF is an accepted mode of therapy for ROP, it was of interest to determine whether peptide mitigation of OIR would be additive to that of anti-VEGF. Bevacizumab, an anti-human VEGF antibody is known to bind poorly to rodent VEGF (Lu and Adelman, 2009), as we confirmed here by showing its poor efficacy in OIR inhibition at 6.0 or 12.5 μg/eye (p=0.24, p=0.08; Figure 2) although a very high dose of 25 ug/eye in mouse OIR has been reported modestly active, with p=0.006 (Hollanders et al., 2015). Anti-mouse VEGF₁₆₄ polyclonal antibody (Ab-493) is reported efficacious against rodent OIR at doses ranging from 5-200 ng/eye (Liu et al., 2013; Heidurschka et al., 2019) also active at 1 ng/eye when combined with a synthetic angiogenesis inhibitor (Sulaiman et al., 2016). In our examining a range of doses of Ab-493 spanning 2-50 ng/eye, only a single dose, 25 ng/eye, showed efficacy (P value from 0.02-0.04) inhibiting OIR. Halving or doubling that dose did not give statistically significant inhibition, the higher dose displaying especially poor efficacy. We have no explanantion for the discordance of our findings with those reported over a wider range of doses, although workers reporting efficacy at 200 ng/eye questioned why regular vascularization occurs when VEGF is fully neutralized, suggesting overlapping roles for other cytokines (Heidurschka et al., 2019).

Anti-mouse VEGF, Ab-493 alone (25ng) gave VO, NV/total area, NV/vascular area of 46%, 40% and 39% of control respectively (P values: 0.042, 0.026, 0.021) as shown in Figure 4A. When Ab-493 was tested in parallel, combined with 3.75 μg peptide, the reduction in OIR parameters was actually slightly poorer (Fig. 4B): VO, NV/total area, NV/vascular area 54%, 53%, 52% of controls (P values: 0.059, 0.022, 0.016). Thus, addition of peptide did not enhance efficacy even though 3.75 μg peptide alone (Fig. 1B, 1C) gave 53%, 49% and 47% of control values of the same parameters with P values ranging from 0.0021 to 0.005. Simple additivity of effect would have predicted combination values of 20%−30% of control, or at least improvement of P values found with antibody alone, approaching those of peptide alone. Absent any such observation we conclude that peptide makes little or no contribution to mitigating OIR in the presence of anti-VEGF. The lack of improved P values in the combined treatment suggests that the robustly efficacious peptide is unable to display its full activity in the presence of anti-VEGF, and that neutralization of VEGF abrogates OIR inhibition by PEDF 336. On the other hand peptide combination does not interfere with Ab-493 mitigation of OIR. This is consistent with our finding that human EC in confluent culture underwent peptide-dependent apoptosis only when preceded by VEGF exposure (Sheibani et al., 2019).

Further rationale for a VEGF requirement enabling PEDF 336 efficacy is seen in our proposed peptide mechanism below and related experiments (Fig. 5).

It was observed that 67LR appeared in retinas of neonatal mice during OIR (Stitt et al., 1998) it was then reported that both EC and tumor cells present cell surface 67LR during growth and cell spreading, but that this is strongly decreased by contact inhibition (Donaldson et al., 2000). OIR in neonatal mice was subsequently inhibited by daily parenteral injection of either of two small 67LR binding peptides (10 mg/kg/day, P12-P19), one presenting YIGSR, the second having an EGF sequence YSGDR (Gebarowska, et al., 2002). Both antiangiogenic PEDF 335 and PEDF 336 peptides contain the sequence: ..Tyr-Asn-Leu-Tyr-Arg-Val.. , (YNLYR), the same prototypic 67LR binding motif of YIGSR in Ln β1 chain. We propose that PEDF, having the same motif, YDLYR, in its α-helical 34-mer likewise targets 67LR (Bernard et al., 2009), as do our small, potent peptides. It is noteworthy that 67LR is recognized as a redox sensor which increases at the plasma membrane and drives cell extravasation with hyperoxia (Vilas-Boas et al., 2016). 67LR has also be shown to mediate hypoxic stress-induced cell production of H₂O₂ by NADPH oxidase, attenuation of this was seen in ligation of 67LR by EGCG or anti-67LR (Gundimeda et al., 2012). We hypothesize that VEGF increase, following transition to normoxia, promotes presentation of 67LR to enhance matrix invasion and oxidative damage. Ligation of 67LR by YIGSR or EGCG has been shown to promote apoptosis of tumor cells via 67LR endocytosis in vesicles (Kumazoe et al., 2013). We suggest that there may be a similar mechanism in VEGF-activated EC, whereby increased 67LR induces their apoptosis when ligating PEDF 34-mer or its Y - - - R peptide derivatives.

We tested this hypothesis by exposing choroidal and retinal EC to VEGF followed by testing both cell expression and turnover of 67LR by PEDF 336. Immunofluorescence detection of 67LR using Alexa-647-tagged anti-67LR showed basal expression of the receptor on these cells. These results demonstrated that mouse choroidal and retinal EC displayed increased 67LR presentation after VEGF treatment (Fig. 5A), becoming sensitized to peptide. Addition of PEDF 336 then decreased VEGF-enhanced 67LR presentation on the choroidal EC (Fig. 5B,C) and retinal EC (not shown), supporting the proposed PEDF 336 antiagiogenic mechanism based on 67LR endocytosis following peptide ligation.

Current treatment of rapidly advancing ROP most commonly utilizes laser ablation surgery, although a significant fraction ELGANs with ROP are treated with a single injection of anti-VEGF or a combination of the above, both approaches are less than ideal, with poor long-term sequelae (Liegl et al., 2016; Rivera et al., 2017). VEGF antagonism is controversial for a number of reasons, it is noteworthy that IVT aflibercept reduced neovascular tufts in mouse OIR but increased the avascular retinal area at P17, delayed vascular growth corresponded to decreased ERG amplitudes (at P21 and P42) and structural changes in the retinal layers that persisted (at P42), despite vascular recovery (Tokunaga, 2014).

A major comorbidity of ROP is bronchopulmonary dysplasia (BPD) (Singh et al., 2019), and lung metalloprotease levels were elevated in hyperoxic rats by a single dose of IVT bevacizumab, indeed the antibody promoted alveolar hemorrhage at P23 following hyperoxia and at P45 under intermittent hypoxia (Valencia et al., 2017). Thus treatment alternatives such as described here merit preclinical comparison with VEGF inhibitors. Small PEDF peptides (Sheibani, et al. 2019) are readily synthesized and may be amenable to subcutaneous delivery in OIR models as reported efficacious for 6-8 days parenteral treatment with a laminin octapeptide or with PEDF 34 mer (Gebarowska, et al., 2002; Longeras et al., 2012). The latter has displayed PPARγ antagonism, decreasing Wnt/β-catenin signalling (Tsai et al., 2014) which are thought to drive BPD of prematurity (Lecarpentier, et al., 2019). Parallel therapeutic control of neonatal BPD and ROP might thus be achieved by systemic administration of PEDF peptides, during a critical few weeks period in premature infants when endogenous PEDF has not yet achieved full term levels. Since both of these comorbidities can be modeled by hyperoxic exposure in neonatal animals, the attempted amelioration of each by parenteral PEDF peptide treatment merits preclinical evaluation. Intravitreal safety studies here including single IVT injections of > 1 mg PEDF 336 in rabbit, along with 0.375 mg Avastin, and of 7.5 μg peptide alone in neonatal mouse showed no histological abnormalities or signs of inflammation. This is consistent with the lack of any ERG perturbation as we reported for similar dosing in adult mouse eyes (Sheibani et al., 2019). The apparent benign nature of the peptide suggests that systemic dosing can be well tolerated.

Conclusions

The PEDF-derived nonapeptide, PEDF 336, previously shown to inhibit CNV by IVT injection is efficacious in mitigating oxygen-induced ischemic retinopathy when injected at the transition to a normal air environment. It displays robust dose-response but is not additive with mouse anti-VEGF, thus free VEGF may be rquired for the peptide’s antiangiogenic effect.

Highlights.

• PEDF-derived 9-mer peptide, PEDF 336, previously shown to inhibit laser-induced CNV in mice was here tested in oxygen-induced ischemic retinopathy

• IVT injection of 1 μL PEDF 336 (1.25 -7.5 μg/eye) in neonatal mouse eyes at transition from 75% O₂ to room air (P12) significantly reduced NV and VO at P17

• While PEDF 336 displayed dose-responsive efficacy, anti-mouse VEGFA₁₆₄ was active only at 25 ng/eye, but not at half or twice this dose.

• Combination of both peptide and anti-VEGF showed only the antibody efficacy but not the robust peptide activity, suggesting the latter may require free VEGF.

• PEDF 336 contains a 5-mer epitope, Y- - - R, known to bind laminin receptor, 67LR, which we found increased on choroidal EC after their exposure to VEGF