Ref-1 redox activity modulates canonical Wnt signaling in endothelial cells

Abstract

Ischemic retinopathies, including proliferative diabetic retinopathy (PDR) and retinopathy of prematurity (ROP), are characterized by abnormal retinal neovascularization and can lead to blindness in children and adults. Current treatments, such as intravitreal anti-VEGF injections, face limitations due to high treatment burden and variable efficacy, as multiple signaling pathways, beyond VEGF, contribute to retinal neovascularization. Previous studies demonstrate that targeting the redox-mediated transcriptional regulatory function of APE1/Ref-1 reduces pathological neovascularization. We aimed to identify novel signaling pathways regulated by Ref-1 redox activity utilizing RNA sequencing of human retinal endothelial cells (HRECs) treated with a Ref-1 redox inhibitor. We found that Wnt/β-catenin signaling was significantly downregulated after Ref-1 inhibition. Given the role of Wnt signaling in vascular pathologies, we investigated how Ref-1 regulates Wnt/β-catenin signaling. Ref-1 inhibition downregulated Wnt co-receptors LRP5/6 at both the mRNA and protein levels in endothelial cells, suggesting transcriptional regulation. Ref-1 redox inhibitors APX3330 and APX2009 reduced Wnt3a-induced nuclear β-catenin levels, decreased Wnt transcriptional activity by TOPFlash luciferase assay, and blocked hypoxia-induced Wnt/β-catenin activation in HRECs. In the oxygen-induced retinopathy mouse model of retinal neovascularization, Ref-1 specific inhibitor APX2009 reduced the expression of Wnt-related genes at sites of neovascularization. These findings reveal a novel role for Ref-1 redox activity in modulating Wnt/β-catenin signaling in endothelial cells and highlight the potential of Ref-1 redox activity targeted inhibitors as a novel therapeutic approach for retinal neovascular diseases by modulating multiple disease-relevant pathways.

Graphical abstract

Highlights

• •Activation of Wnt signaling is implicated in retinal neovascularization.
• •Ref-1 redox activity modulates expression of Wnt co-receptors LRP5/6.
• •Ref-1 redox activity regulates β-catenin localization and TCF7 activation.
• •Hypoxia induces canonical Wnt signaling activation in endothelial cells.
• •Ref-1 redox activity regulates the expression of Wnt-related genes in vivo.

Abbreviations:

APE1/Ref-1

Apurinic/apyrimidinic endonuclease 1/reduction-oxidation factor 1

bEnd

Brain-derived endothelial cells

CA-β-catenin

Constitutively active β-catenin

DEGs

Differentially expressed genes

DME

Diabetic macular edema

DR

Diabetic retinopathy

DVL2

Dishevelled 2

EV

Empty vector

FDR

False discovery rate

FEVR

Familial exudative vitreoretinopathy

FZD

Frizzled

GCL

Ganglion cell layer

HREC

Human retinal endothelial cells

HSQC NMR

Heteronuclear single quantum coherence nuclear magnetic resonance

INL

Inner nuclear layer

IVT

Intravitreal

LRP5/6

Low-density lipoprotein receptor-related protein 5 and 6

OIR

Oxygen-induced retinopathy

OPPG

Osteoporosis pseudoglioma syndrome

PDR

Proliferative diabetic retinopathy

RLU

Relative light units

ROI

Region of interest

ROP

Retinopathy of prematurity

TCF7/LEF1

Transcription factor 7/lymphoid enhancer binding factor 1

VEGF

Vascular endothelial growth factor

WaterLOGSY

Water-ligand observed via gradient spectroscopy

1. Introduction

The retina is one of the most metabolically active tissues, requiring a continuous supply of oxygen to support its high metabolic demands [1]. Disruptions to oxygen supply, such as hypoxia, can lead to retinal neovascularization [[2], [3], [4], [5], [6]]. Retinopathy of prematurity (ROP) and proliferative diabetic retinopathy (PDR) are two prominent ischemic retinopathies marked by aberrant vessel growth [4,[7], [8], [9]]. In advanced PDR, fibrovascular membranes (FVMs) can develop on the surface of the retina, which can lead to retinal detachments [10]. Current treatments for these blinding eye diseases include intravitreal (IVT) injections of anti-VEGF biologics, but these treatments may be accompanied by side effects (such as myocardial infarction, stroke, and non-ocular hemorrhage), have a high treatment burden due to socioeconomic barriers, and exhibit variable patient responses due to complex pathogenesis involving multiple pathways beyond VEGF signaling [[11], [12], [13]]. Thus, treatment paradigms targeting multiple disease-relevant pathways may offer better treatment options.

One potential therapeutic target is the redox function of apurinic/apyrimidinic endonuclease 1/reduction-oxidation factor 1 (APE1/Ref-1), a multifunctional protein with both AP endonuclease activity (APE1) and a redox-transcription activator role (Ref-1). Ref-1 regulates transcription factors through a thiol/sulfide exchange, reducing transcription factors to activate their DNA binding while becoming oxidized in the process [14]. Ref-1 has been shown to regulate numerous transcription factors through this process, including HIF-1α, NF-κB, STAT3, and several others [15]. Given the pro-angiogenic and pro-inflammatory roles of these transcription factors regulated by Ref-1, several studies indicate that Ref-1 is overexpressed in human neovascular eye diseases and mouse models of retinal neovascularization [16,17]. Targeting Ref-1 with Ref-1 redox inhibitors can have therapeutic effects in varied models of ocular neovascularization [12,[16], [17], [18], [19], [20]]. Excitingly, APX3330, a small molecule Ref-1 redox inhibitor, recently completed a Phase IIb clinical trial as an oral therapeutic for diabetic macular edema (DME) (NCT04692688) [21]. APX2009, a second-generation small molecule inhibitor of Ref-1, was developed based on structure-activity relationships and offers improved potency and bioavailability compared to clinical candidate APX3330 [22]. Importantly, APX2009 inhibits the redox activity of Ref-1 without inhibiting the DNA repair function of APE1/Ref-1 [17,23], and both APX3330 and APX2009 bind directly to Ref-1, as demonstrated by waterLOGSY and TROSY-HSQC NMR [17,24]. Given the clinical relevance of Ref-1 redox inhibitors, we sought to identify novel pathways and targets regulated by Ref-1 redox activity to further understand its role in neovascular eye diseases.

Dysregulation of Wnt/β-catenin signaling is often tumorigenic and has been implicated in vascular retinopathies, suggesting a role in mediating pathological neovascularization [[25], [26], [27], [28]]. Wnt signaling is initiated when a Wnt ligand binds to a frizzled (FZD) receptor, facilitating the recruitment of low-density lipoprotein receptor-related protein 5/6 (LRP5/6) to form a receptor complex. Subsequent phosphorylation of LRP5/6 sequesters the β-catenin destruction complex to the plasma membrane, allowing for β-catenin to accumulate, translocate to the nucleus, interact with transcription factors TCF7/LEF1, and initiate the transcription of Wnt target genes such as c-Myc, cyclin D1, and PPARγ. Activated Wnt signaling can mediate a variety of cellular pathways including proliferation, differentiation, and migration and is important in embryonic development and cancer pathogenesis. When Wnt signaling is not activated, β-catenin is phosphorylated by GSK3β and CK1α, leading to its ubiquitination and proteasomal degradation, and the transcription of Wnt target genes is reduced [29]. The non-canonical Wnt ligand Norrin binds a unique receptor complex of FZD4, LRP5, and Tspan12, thus activating the expression of genes involved in maintenance of the blood-retinal-barrier [30,31].

Mutations in the Wnt genes FZD4, LRP5, and Wnt ligand Norrin have been linked with blinding eye diseases that affect the vascularization of the retina, such as familial exudative vitreoretinopathy (FEVR), osteoporosis pseudoglioma syndrome (OPPG), and Norrie Disease [[32], [33], [34], [35], [36], [37], [38], [39]]. Elevated levels of nuclear β-catenin and LRP5/6 levels have been detected in the retinas of DR patients, and mutations in the FZD4 gene have been noted in some severe ROP patients [40,41]. In the mouse retina, loss of LRP5 impairs retinal vascular development, highlighting the importance of Wnt signaling in regulating vascular growth [42,43]. Additionally, elevated levels of FZD4 and LRP5 are detected at sites of neovascularization in the mouse oxygen-induced retinopathy (OIR) model [44]. On the other hand, Norrin mimics, or FZD4:LRP5 agonists, can reduce pathological neovascularization in mouse models of ROP and restore blood-retinal-barrier integrity [28,31]. Interestingly, Wnt ligands Wnt3a, Wnt7a, and Wnt10a are upregulated in mouse models of retinal neovascularization, while norrin levels are unchanged [45]. These findings underscore the complexity of Wnt signaling and the context dependent signaling mechanisms of Wnt signaling in pathological retinal neovascularization.

Given the established role of Ref-1 and aberrant Wnt signaling in retinal neovascularization, and our RNA-seq data (below) implicating APX2009 in regulation of Wnt genes, we became interested in exploring the potential link between Ref-1 redox activity and Wnt3a-induced Wnt signaling activation. In this study, we revealed that Ref-1 inhibition modulated multiple Wnt signaling components. We therefore examined if Ref-1 redox activity modulates canonical Wnt/β-catenin signaling in endothelial cells. We found that inhibition of Ref-1 redox activity reduces LRP5/6 expression at both the protein and mRNA level, suggesting Ref-1 transcriptionally regulates LRP5/6. Inhibition of Ref-1 redox activity reduced Wnt/β-catenin transcriptional activation, and hypoxia was found to activate Wnt/β-catenin signaling in endothelial cells, in turn suppressed by Ref-1 inhibition. In a mouse model of retinal neovascularization, inhibiting Ref-1 redox activity with novel small molecule APX2009 reduced the expression of Wnt genes at sites of neovascularization. These findings reveal a novel link between Ref-1 and Wnt/β-catenin signaling in endothelial cells, highlighting a new signaling pathway by which Ref-1 may modulate aberrant neovascularization in retinal diseases.

2. Materials and methods

2.1. Animals

All animal experiments were approved by the Indiana University School of Medicine Institutional Animal Care and Use Committee and followed the “Animal Research: Reporting of In Vivo Experiments” (ARRIVE) guidelines and Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Visual Research. Wild-type female C57BL/6J timed pregnancy mice were purchased from Jackson Laboratory (Bar Harbor, ME, USA) and housed under standard conditions in the Indiana University Laboratory Animal Research Center (LARC). Pups of both sexes were used for experiments. Mice were under a 12h:12h light-dark cycle, temperature at 21 ± 1 °C, and humidity was 30–70 %. Animals were fed Teklad 2918X rodent chow and chlorinated water (via Hydropac) ad libitum. Treatments were randomly assigned by cage and animals.

2.2. Cells

Primary human retinal microvascular endothelial cells (HRECs) were obtained from Cell Systems, Inc. (lots 181.01.04.02.02, 181.03.03.02.02, 181.04.01.01.03 all from male subjects; #ABCRI 181, Kirkland, WA, USA) and used between passages 4 and 7. HRECs were cultured in endothelial basal medium (EBM-2; Cat. no. CC-4176; Lonza, Walkersville, MD, USA) supplemented with EGM-2 “Bullet Kit” (Cat. no. CC-4176; Lonza) to make EGM-2 growth media. Cells were monitored regularly for mycoplasma contamination.

bEnd.3 cells (ATCC) that were simultaneously transduced with two lentivirus vectors encoding TOPFlash reporter and Renilla luciferase were used between passage 6 and 20 to track Wnt/β-catenin transcription activation (See section “Generation of bEnd.3 cells stably transduced with TOPFlash reporter and Renilla luciferase” for details).

HREC siRNA transfection of siRNA for human Ref-1 and scrambled control was done as previously described [16]. In bEnd.3 cells, 50 nM of custom siRNA for murine Ref-1 (Thermo Fisher Scientific, Waltham, MA, USA) or scrambled control were transfected with Lipofectamine RNAiMAX for 48 h according to vendor protocol. Sequences of the siRNA for murine Ref-1 are 5′ GUCUGGUAAGACUGGAAUACCUA 3′ and scrambled are 5′ GUCUGGUAUCUCUGGAAUACCUA 3′.

Chemical structures, properties and specific binding to Ref-1 of APX3330, APX2009 and RN7-58 have been previously published [24,46,47]. APX2009, APX3330, or RN7-58 were dissolved in DMSO and at indicated concentrations added to HRECs or bEnd.3 cells for 24 h. Doses were chosen based on GI₅₀ values and previous studies [16,18,47]. For Wnt3a stimulation, cells were first stimulated with 200 ng/ml Wnt3a for 2 h before addition of indicated compounds. DKK1, an endogenous Wnt inhibitor, was used at a dose of 300 ng/ml when indicated. GSK-3β inhibitor CHIR99021 was used at a concentration of 2 μM according to previous studies [48,49]. For hypoxia stimulation, cells were placed at 1 % O₂ in an InvivO₂ 400 Physoxia Workstation (Baker, Sanford, ME, USA) for different lengths of time, as indicated in the text.

2.3. Generation of bEnd.3 cells stably transduced with TOPFlash reporter and Renilla luciferase

bEnd.3 cells (ATCC) were simultaneously transduced with two lentiviral vectors. The first vector encoded the TOPFlash reporter and provided puromycin resistance. This vector was generated from the plasmid 0166_pBARls_TOPFLASH_lenti (gift from Stephane Angers). The second lentiviral vector encoded Renilla luciferase under a constitutive CMV promoter, provided blasticidin resistance, and was generated from the plasmid pLenti_CMV_MCS-Renilla_IRES_BlastR. This plasmid was made by modifying Addgene 102343 (deposited by Ghassan Mouneimne [50] by adding restriction sites and inserting the Renilla coding sequence. Viral vectors were packaged and purified at the University of Minnesota Viral Vector and Cloning Core as follows. 200,000 bEnd.3 cells were seeded into a well of a 6-well dish. 24 h later, cells were simultaneously transduced with Renilla lentivirus (8.4 μl of 4 x 10⁸ TU/ml) and TOPFlash lentivirus (10 μl of 3.6 x 10⁸ TU/ml) and selected in high glucose DMEM, 10 % FBS, 1 % penicillin/streptomycin, 1 μg/ml puromycin and 10 μg/ml blasticidin. Untransduced control cells cultured in parallel died within 10 days under selection pressure. The surviving transduced cell population was expanded and maintained in the same medium. Selection pressure was removed only immediately before an experiment. The stably transduced cell population responded to recombinant norrin with > 5-fold change of TOPFlash activity until at least passage 20.

2.4. Dual luciferase assay

Firefly and Renilla luciferase activities were assayed using the Dual Luciferase Reporter Assay System (Cat. No. E1910; Promega Corp., Madison, WI, USA) according to vendor protocol. Renilla luciferase was used for normalization and all transfection experiments were repeated three times for three independent experiments.

2.5. RNA-seq library preparation and sequencing

RNA sequencing was performed at the Indiana University Center for Medical Genomics according to their standard protocols. Briefly, HRECs were treated either with DMSO or 10 μM APX2009 in 6-well plates for 24 h. After 24 h, cells were washed with PBS and extracted with TRIzol. Total RNA was evaluated for quantity and quality by Qubit and Agilent Bioanalyzer. 100 ng of total RNA was used for library preparation with a Biomek and the KAPA mRNA Hyperprep Kit (KK8581) (Roche, Basel, Switzerland). Illumina NovaSeq 6000 was used for sequencing with 100bp paired-end reads using the v1.0 reagent kit. Base calling and quality scoring were performed by Real-Time Analysis (RTA) v2 in Illumina NovaSeq 6000. The bcl2fastq2 conversion software was then used to convert base call (bcl) files to FASTQ files and trim adapter sequence at the same time.

2.6. Bioinformatics and data analyses

The reads were mapped to the human genome hg38 using STAR (v2.7.2a) [51]. RNA-seq was aligned with the following parameters: “---outSAMmapqUnique 60”. Uniquely mapped sequencing reads were assigned to GENCODE 31 gene using featureCounts (v1.6.5) [52] with “-p -Q 10-O” parameters. The data were filtered using read counts >10 in at least 5 of the samples, normalized using the TMM (trimmed mean of M values) method and subjected to differential expression analysis using edgeR (v3.20.8) [53,54]. DEGs were determined with a false discovery rate (FDR) of less than 0.05 and a fold change cutoff of 1.5. The dataset is archived to the Gene Expression Omnibus (GEO), accession GSE290559. To identify biological and molecular pathways associated with the DEGs, QIAGEN Ingenuity pathway analysis (IPA) software (QIAGEN Inc, https://digitalinsights.qiagen.com/IPA) was used [55]. A threshold of log(p-value) of 1.3 (p-value of 0.05) was set. Pathway diagrams showing DEGs were generated using IPA software. Heatmap and volcano plot diagrams were generated using SRplot [56].

Raw counts of gene expression in control retinas and human FVMs were obtained from GSE94019 [57]. Genes were analyzed as described previously [16]. P-values were adjusted for multiple testing using an FDR calculator with the Benjamini-Hochberg adjustment.

2.7. Constitutively active β-catenin lentivirus production and HREC transduction

Lentiviruses were produced in HEK239T cells cultured in D-10 media (DMEM supplemented with 10 % FBS, GlutaMax, penicillin, and streptomycin). Transgene plasmids used were either β-catenin (S33A, S37A, T41A, S45A)-pcw107-V5 (constitutively active [CA] β-catenin) (deposited by David Sabatini and Kris Wood; Addgene plasmid #64613) or pcw107 (empty vector, [EV]) (deposited by John Doench and David Sabatini; Addgene plasmid #62511) [58]. Cells were plated at 1.0 x 10⁷ cells per T75 flask in 12 ml of media. The next day, when cells were 80–95 % confluent, the cells were transfected with 20 μg of transgene plasmid, 20 μg of psPAX2 packaging plasmid (deposited by Didier Trono; Addgene plasmid #12260), and 2 μg of VSV-g envelope plasmid (deposited by Didier Trono; Addgene plasmid #12259) using the Promega ProFection Mammalian Transfection System in fresh medium. 24 h after transfection, the medium was changed to harvest medium (OptiPRO [Thermo Fisher Scientific], GlutaMax, 5 mM sodium butyrate, penicillin, streptomycin) and incubated for 24 h. The supernatants were harvested and centrifuged at 1200 rpm for 5 min. Vector containing supernatants were then clarified using 0.45 μm syringe filters, aliquoted, and stored at −80 °C.

For infection, HRECs were maintained in 6 cm dishes until they were about 80 % confluent. On the day of transduction, cells were washed with PBS, and 750 l of crude viral supernatant was added to the cells with 1.5 ml of EGM-2 complete medium and 750 l of low serum medium (Opti-MEM, Cat. No. 31985-070; Thermo Fisher Scientific, Waltham, MA, USA) with polybrene at a final concentration of 4 μg/ml. 24 h after transduction, the medium was replaced with fresh medium. 48 h after transduction, puromycin at a final concentration of 1 μg/ml was added to select for infected cells. Cells were maintained in EGM-2 complete media with 1 μg/ml puromycin.

2.8. Proliferation assay

Endothelial cell proliferation was assessed as previously described [18]. HRECs were seeded at 2500 cells per 100 l per well in a 96-well clear-bottomed black plate and incubated for 24 h at 37 °C and 5 % CO₂. After 24 h, medium was changed, and APX2009 was added in concentration range of 0.1 nM to 100 μM in 1 l per well (final DMSO concentration of 1 %). The plates were then incubated for 44 h at 37 °C and 5 % CO₂. Then, 11.1 l of alamarBlue (Cat. No. BUF012B; BioRad, Hercules, CA, USA) was added to each well, and fluorescence was measured 4 h later, 48 h after addition of compounds. The fluorescence was read using a Synergy H1 plate reader (BioTek, Winooski, VT) with excitation and emission wavelengths of 560 and 590 nm, respectively. The median growth inhibitory concentration (GI₅₀) and dose-response curves were calculated using GraphPad Prism.

2.9. Protein preparation and immunoblot analysis

For whole cell lysate immunoblots of HRECs or bEnd.3 cells, cells were washed with PBS and lysed with 1 % SDS supplemented with protease inhibitors. The lysed cells were then transferred to a microcentrifuge tube, sonicated on ice, heated at 95 °C for 5 min, and then centrifuged at 12,000×g for 2 min at 4 °C. For immunoblots of fractionated HRECs, cells were harvested and fractionated with the NE-PER Nuclear and Cytoplasmic Extraction Reagents according to vendor protocol (Cat. No. 78833; Thermo Scientific). Equal amounts of total protein (20 μg) from each sample were resolved and immunoblotting was done as previously described [16]. Antibodies used to detect proteins are indicated in the text. See Suppl.Table 1 for catalog numbers. The bands were quantified by densitometry using ImageJ.

2.10. Immunocytochemistry of HRECs

HRECs were seeded onto four-well chamber slides at 50,000 cells per well and allowed to attach overnight. HRECs were then treated with the compounds indicated in the text. Cells were then fixed in 4 % PFA in PBS for 20 min and permeabilized with 0.5 % Triton X-100 in PBS. Cells were blocked in 3 % BSA in PBS for 2 h and then incubated overnight at 4 °C with the antibodies indicated in the text (Suppl. Table 1) diluted in 1 % BSA in PBS. Secondary antibodies were diluted in 1 % BSA in PBS, and cells were counterstained with DAPI to visualize nuclear staining. Images were acquired using either the Plan-Apochromat 20x objective or Plan-Apochromat 63x objective of an LSM 700 laser scanning confocal microscope with ZEN imaging software (Zeiss, Thornwood, NY, USA).

2.11. RNA extraction and qPCR

For mouse retinas, following euthanasia, eyes were enucleated, and retinas were immediately dissected out, and two retinas were pooled per reaction. Retinas were kept in RNAlater at 4 °C until RNA extraction. HRECs were treated as mentioned in the text. The cells were then trypsinized and cell pellets were collected and frozen at −80 °C. Total RNA was isolated from cells and tissue using the RNeasy mini kit and Qiashredder according to manufacturer instructions (Qiagen, Germantown, MD, USA). cDNA was synthesized using the iScript cDNA synthesis kit (BioRad, Hercules, CA, USA) according to manufacturer’ protocol, using 0.5 μg RNA for reverse transcription. qPCR reactions were prepared using the TaqMan Fast Advanced Master mix and TaqMan probes. All TaqMan probe sets are listed in Suppl.Table 2. A ViiA7 Real-Time PCR system (Thermo Fisher) was used to perform qPCR under the following conditions: hold at 50 °C for 2 min, hold at 95 °C for 2 min, and 40 cycles of denature at 95 °C for 10 s and anneal/extend at 60 °C for 20 s. Quantification was performed by the ΔΔC_t method and normalized to HPRT1 and TBP. Relative expression was compared to a single control sample.

2.12. Oxygen-induced retinopathy mouse model

Mice were subjected to oxygen-induced retinopathy (OIR) as previously described [59,60]. Briefly, postnatal day 7 (P7) mice, with their nursing mothers, were exposed to 75 % O₂ (hyperoxia) in a hyperoxia chamber (Coy, Grass Lake, MI, USA) for five days (P7–P12) to initiate retinal vascular obliteration and returned to room air (normoxia) at P12. Upon exiting the hyperoxia chamber and returning to room air, P12 OIR mice received twice daily 12.5 mg/kg body weight i.p. injections of APX2009 or vehicle [Propylene Glycol, Kolliphor HS 15, Tween 80 (PKT)], with injections at least 8 h apart as previously described [16]. Mice received their last dose in the morning of P17 2 h before euthanasia.

2.13. Intravitreal injections

Intravitreal (IVT) injections were performed as previously described [16]. Briefly, P12 pups were anesthetized and 0.5 % tetracaine hydrochloride ophthalmic solution (Oceanside Pharmaceuticals, Mission Viejo, CA, USA) was used as topical anesthetic. After an incision was made at the nasal-temporal ora serrata with a 30-gauge insulin syringe needle, a sterile 33-gauge Hamilton syringe (0.5–5 μL volume) was used for injections. Mouse anti-VEGF₁₆₄ (Cat. No. AF-493-NA, Lot No. YU1419101, R&D, Minneapolis, MN, USA) (0.5 μL for a final dose of 5 ng/eye) was injected intravitreally. Eyes that had back flushing of injection were deemed unsuccessful and were excluded. After injections, triple antibiotic ointment (Vetropolycin; Dechra Pharmaceuticals, Fort Wayne, TX, USA) was applied and anesthesia was reversed with 0.1 mg/ml i.p. injection of atipamezole hydrochloride.

2.14. RNAscope multiplex in situ hybridization

RNAscope in situ hybridization (ISH) was conducted on formalin-fixed, paraffin embedded mouse retina sections following the manufacturer’ guidelines (ACD, Advanced Cell Diagnostics, Newark, CA, USA). Mouse eyes were enucleated, and the cornea was punctured with a 29G needle. The retinas were immersed in 10 % neutral buffered formalin (NBF) for 1 h before posterior eyecups were dissected out and fixed in 10 % NBF for 24 h at room temperature. After three 15-min washes in PBS, the posterior eyecups were dehydrated and embedded in paraffin using standard histological procedures. Tissues were cut into 5 μm sections with a microtome, and RNAscope in situ hybridization was carried out within a week of sectioning to minimize RNA degradation. Slides were deparaffinized, underwent antigen retrieval, and were treated with hydrogen peroxide and protease digestion. RNAscope Multiplex Fluorescent Assay was subsequently performed to hybridize probes specific to the RNA sequences. To detect mouse Apex1 mRNA, RNAscope Mm-Apex1 (Cat No. 1191081, lot # 22243A) was used, while RNAscope Mm-Fzd1 (Cat No. 404871, lot # 24138B) and RNAscope Mm-Fzd4 (Cat No. 404901, lot # 23082B) were used to detect mouse Fzd1 and mouse Fzd4, respectively. RNAscope Mm-Vegfa (Cat No. 312931, lot # 23082B) and RNAscope Mm-Lrp5 (Cat No. 315791, lot # 23082B) were used to detect mouse Vegfa and mouse Lrp5, respectively. The universal negative control probe targeting the DapB gene from Bacillus subtilis strain SMY (accession #EF191,151, Cat No. 320871) was used as a negative control. Positive controls using ACD’s mouse specific probes for Polr2a and Ppib (Cat No. 321881) were used. After hybridization of probes, a three-step amplification process was performed followed by development of channel specific signal and binding of 520 Opal Fluorophore (SKU FP1487001KT) (Apex1 and Fzd1 probes), 570 Opal Fluorophore (SKU FP1488001KT) (Vegfa and Fzd4 probe), and 690 Opal Fluorophore (SKU FP1497001KT) (LRP5 probe) at a 1:500 dilution (Akoya Biosciences, Marlborough, MA, USA). Sections were counterstained with DAPI and mounted with Fluoromount-G (SouthernBiotech, Birmingham, AL, USA). Slides were imaged using either the Plan-Apochromat 20x objective or Plan-Apochromat 63x objective on an LSM 700 laser scanning confocal microscope with ZEN imaging software (Zeiss). Image settings were standardized across all samples to accurately quantify and qualitatively compare sections.

RNAscope in situ hybridization was analyzed in ImageJ according to ACD protocol. Background signal was measured by averaging readings from at least five distinct areas per sample and channel, and the threshold was set to 3.5 times the average background intensity. The watershed algorithm was used to separate overlapping or touching puncta. Then, a region of interest (ROI) was drawn around the target area and the number of puncta and area of the ROI were quantified. The number of puncta was then normalized to the area of the ROI to yield a value representing the average number of RNA transcripts per unit area of retina.

2.15. Statistics

Values of p < 0.05 were considered statistically significant. Data are presented at mean ± SEM; n is defined in the figure legends. All experiments were repeated at least three times. Data sets were tested for normality and equal standard deviations. Statistical analyses were performed using GraphPad Prism v10.4 and statistical tests used are listed in figure legends.

3. Results

3.1. Gene expression analysis reveals Wnt signaling under Ref-1 redox control

To investigate novel signaling pathways under Ref-1 redox control in HRECs, we performed mRNA sequencing on HRECs treated with 10 μM Ref-1 redox inhibitor APX2009 or DMSO vehicle control. Out of 12,285 genes analyzed, there were 1805 DEGs, with 1236 DEGs significantly downregulated and 569 significantly upregulated DEGs with cutoffs indicated in the methods (Fig. 1A). To understand molecular pathways that might be under Ref-1 redox control in HRECs, we subjected the DEGs to IPA pathway and network analysis. Top downregulated pathways after Ref-1 redox inhibition included generic transcription, mitochondrial biogenesis, apoptosis signaling, Wnt/β-catenin signaling, TNFR1 signaling, and JAK-STAT signaling (Fig. 1B). Top upregulated pathways included protein ubiquitination, pre-mRNA and rRNA processing, autophagy, cell cycle checkpoints, p53 signaling, and degradation of β-catenin by the destruction complex (Fig. 1C). Given Ref-1’s known roles in transcriptional regulation and cell cycle regulation, many of these pathways were not surprising. But it was noteworthy that Wnt/β-catenin signaling and degradation of β-catenin were among the top down-regulated and up-regulated signaling pathways, respectively. Canonical Wnt/β-catenin signaling DEGs demonstrate that inhibition of Ref-1 redox activity downregulated genes associated with Wnt signaling activation and upregulated genes associated with Wnt signaling inhibition (Fig. 1D–E). To validate the relevance of Wnt signaling in human vascular retinopathies, we analyzed the expression of Wnt pathway components in CD31⁺ cells from human FVMs and control retinas from an existing RNA-seq study (GSE94019) [57] and found Wnt genes to be overexpressed in endothelial cells of FVMs compared to endothelial cells in control retinas (Fig. 1F–I). Given the role of Wnt/β-catenin signaling in vascular retinopathies, we chose to further explore the effects of Ref-1 redox inhibition on Wnt/β-catenin signaling.

Fig. 1.

Differential gene analysis reveals Wnt/β-catenin signaling is under Ref-1 redox control. A, volcano plot of up-regulated (red) and down-regulated (blue) DEGs after treatment with 10 μM APX2009 in HRECs with an FDR cut off of 0.05 (-log₁₀FDR of 1.3) and 1.5-fold change. B and C, select top down-regulated (B) and up-regulated (C) pathways after APX2009 treatment. D, heatmap demonstrating expression of canonical Wnt signaling genes affected after APX2009 treatment, represented by raw Z-score. Five biological replicates per treatment group. E, canonical Wnt/β-catenin signaling pathway. DEGs from RNA sequencing are bolded and superimposed on targets in the pathway. Blue indicates downregulated genes, light red represents upregulated genes, and white indicates non DEGs. F–I, RNA-seq analysis of CD31⁺ cells from human FVMs and CD31⁺ cells from control retinas (GSE94019) [57] indicates that (F) CTNNB1 (encoding β-catenin), (G) TCF7 (encoding TCF7), and Wnt target genes (H) MYC (encoding c-Myc) and (I) CCND1 (encoding cyclin D1) are overexpressed in endothelial cells of human FVMs compared to normal retinas. F-G analyzed by Welch’s t-test; H–I analyzed by Mann Whitney test; t-tests adjusted for multiple testing (see methods for analysis); n = 4–9 per group. ∗p < 0.05. Data represented as mean ± SEM. FPRM = fragments per kilobase of transcript per million mapped reads.

3.2. Inhibition of Ref-1 redox activity decreases expression of LRP5/6

Based on the RNA-seq data, we aimed to validate if Ref-1 redox activity modulates canonical Wnt/β-catenin signaling. LRP5 and LRP6 are critical co-receptors necessary for initiation of canonical Wnt/β-catenin signaling, but they are not involved in non-canonical Wnt signaling [61,62]. Given this, we explored the effects of Ref-1 redox inhibition on LRP5 and LRP6 expression. To validate our RNA-seq findings, we used the second-generation Ref-1 redox inhibitor APX2009, as well as clinical candidate APX3330. The doses of APX2009 and APX3330 were selected based on previous studies that have established effective doses of these compounds in endothelial cells [[16], [17], [18]]. To stimulate Wnt/β-catenin signaling, we pretreated select groups of HRECs with Wnt3a, a Wnt ligand that activates canonical Wnt signaling in endothelial cells and is implicated in regulating endothelial cell proliferation and migration [26,[63], [64], [65], [66]]. The amount of Wnt3a was selected based on optimization studies that determined that 200 ng/ml of Wnt3a was an effective dose to activate Wnt signaling by inducing nuclear accumulation of β-catenin, and based on previous published studies [30,66].

APX2009 and APX3330 significantly decreased mRNA levels of LRP5 and LRP6 in HRECs both with and without Wnt3a stimulation (Fig. 2A–B; Suppl. Fig. 1A–B). RN7-58, a structural analog of APX2009 that binds to Ref-1 but exerts no effects on Ref-1 redox activity, thus acting as a negative control [67], did not affect mRNA levels of LRP5 and LRP6 (Suppl. Fig. 2A–B). APX2009 and APX3330 also induced dose-dependent decreases in the protein expression of LRP5 and LRP6 in HRECs with and without Wnt3a stimulation (Fig. 2C–J). Ref-1 redox inhibitors or Wnt3a stimulation did not affect Ref-1 protein levels in HRECs (Suppl. Fig. 1C–F). Inactive analog RN7-58 did not affect LRP5/6 protein expression in HRECs (Suppl. Fig. 1G–J). Following up on the RNA sequencing data, we also examined if Ref-1 redox inhibition affects protein expression of DVL2 or FZD receptors involved in angiogenesis. We found that APX2009 does not affect the protein levels of FZD1 or FZD5 but does slightly decrease DVL2 protein expression (Suppl. Fig. 2). Further, we validated that our loading control, vinculin, is not affected by Wnt3a stimulation or Ref-1 inhibition in HRECs (Suppl. Fig. 3). Together, these results indicate that Ref-1 redox activity regulates LRP5/6 mRNA and protein expression in HRECs.

Fig. 2.

Inhibition of Ref-1 redox activity decreases expression of LRP5/6. A and B, qPCR of LRP5 (A) and LRP6 (B) in HRECs treated with vehicle (DMSO), 20 μM APX3330, or 10 μM APX2009 for 24 h with and without 200 ng/ml Wnt3a pretreatment indicates that Ref-1 redox inhibitors decreased mRNA levels of LRP5 and LRP6. One-way ANOVA with Tukey’s post hoc tests; n=3 biological replicates. C–F, immunoblots of HRECs treated with indicated doses of APX2009 with and without 200 ng/ml Wnt3a pretreatment and probed for LRP5 (C–D) or LRP6 (E–F) and vinculin (loading control). APX2009 dose-dependently decreased protein levels of LRP5 and LRP6 in HRECs. One-way ANOVA with Tukey’s post hoc tests; n=3 biological replicates. G-J, immunoblots of HRECs treated with indicated doses of APX3330 with and without 200 ng/ml Wnt3a pretreatment and probed for LRP5 (G–H) or LRP6 (I–J) and vinculin (loading control). APX3330 dose-dependently decreased protein levels of LRP5 and LRP6 in HRECs. One-way ANOVA with Tukey’s post hoc tests; n=3 biological replicates. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001. Data represented as mean ± SEM.

3.3. Inhibition of Ref-1 redox activity decreases nuclear β-catenin

To determine if Ref-1 redox activity has functional effects on canonical Wnt signaling, we sought to determine if Ref-1 redox inhibitors decrease nuclear β-catenin. HRECs were pretreated with 200 ng/ml Wnt3a for 2 h before they were treated with Ref-1 redox inhibitors or vehicle. Wnt3a stimulation increased nuclear β-catenin, while Ref-1 redox inhibitors decreased nuclear β-catenin (Fig. 3A–B). Wnt3a stimulation and Ref-1 redox inhibitors did not affect nuclear or cytoplasmic Ref-1 levels, consistent with previous work (Fig. 3C–F) [16]. Furthermore, cytoplasmic β-catenin levels were not affected (Fig. 3D–E). Mean fluorescence intensity quantification of immunocytochemistry confirmed increased nuclear β-catenin with Wnt3a stimulation and showed significantly reduced nuclear β-catenin with Ref-1 redox inhibition (Fig. 3G–H). Together, these results demonstrate that Wnt3a-induced increases in nuclear β-catenin are, in part, dependent on Ref-1 redox activity.

Fig. 3.

Inhibition of Ref-1 redox activity decreases nuclear β-catenin. HRECs were stimulated with 200 ng/ml Wnt3a for 2 h before being treated with 10 μM APX2009, 20 μM APX3330, or vehicle (DMSO) for 24 h. A-F, HRECs were fractionated and nuclear and cytoplasmic localization of β-catenin was assessed by immunoblot. A-B, Wnt3a stimulation induced nuclear localization of β-catenin, and Ref-1 redox inhibitors APX2009 and APX3330 decreased nuclear β-catenin. C, nuclear Ref-1 levels were unchanged with treatments. D-F, Cytoplasmic β-catenin and Ref-1 levels were unchanged with treatments. One-way ANOVA with Tukey’s post hoc tests; n=5 biological replicates. G-H, HRECs were immunostained for Ref-1 (green), β-catenin (magenta), and DAPI (blue). Mean fluorescence intensity [68] of nuclear β-catenin indicates that inhibition of Ref-1 redox activity decreases nuclear β-catenin. Yellow arrows indicate increased nuclear β-catenin with Wnt3a stimulation. One-way ANOVA with Tukey’s post hoc tests; n=53–58 cells across three biological replicates. Scale bars = 10 μm. ∗p < 0.05; ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Data represented as mean ± SEM.

3.4. Ref-1 redox activity regulates Wnt-mediated transcriptional activation

To determine if Ref-1 redox activity mediates transcriptional activity of TCF7, a key target of Wnt/β-catenin signaling, we utilized bEnd.3 cells that were co-transduced with lentiviral vectors expressing the TOPFlash reporter to detect Wnt/β-catenin pathway activation and constitutively active Renilla luciferase. We first verified that inhibition of Ref-1 redox activity with APX2009 decreases LRP5 protein expression in bEnd.3 cells, while negative analog RN7-58 does not (Suppl. Fig. 4A and B). We were unable to detect LRP6 expression in this cell line. Dual luciferase assay revealed that Wnt3a stimulation induced Wnt-mediated transcriptional activation, and APX2009 and APX3330 dose-dependently decreased Wnt/β-catenin pathway activation (Fig. 4A–B). RN7-58 demonstrated no effects on Wnt/β-catenin transcriptional activation (Suppl. Fig. 4C). To further verify that inhibition of Ref-1 redox activity decreases Wnt-mediated transcriptional activation, we examined if Ref-1 redox inhibitors decrease the expression of TCF7 target genes in HRECs. Wnt3a stimulation upregulated the mRNA transcript levels of MYC, CCND1, and PPARD, while APX2009 decreased levels of these transcripts (Fig. 4C–E). APX3330 decreased the expression of MYC and CCND1 (Fig. 4C–D). Ref-1 gene expression was unaffected by treatments (Fig. 4F).

Fig. 4.

Ref-1 redox activity regulates Wnt-mediated transcriptional activation. A-B, dual-luciferase assay of bEnd.3 cells co-transduced with TOPFlash reporter lentivirus and Renilla luciferase lentivirus demonstrates that APX2009 (A) and APX3330 (B) dose-dependently decrease Wnt/β-catenin pathway activation after pre-treatment with 200 ng/ml Wnt3a. 300 ng/ml DKK1 (endogenous Wnt inhibitor) was used as a positive control. All comparisons are made against vehicle + Wnt3a (white bars). C-E, Wnt3a stimulation upregulates Wnt/β-catenin transcriptional targets MYC (C), CCND1 (D), and PPARD (E) in HRECs. Ref-1 redox inhibitors APX2009 and APX3330 decrease the expression of MYC and CCND1 (C–D), while APX2009 also decreases the expression of PPARD (E). F, Wnt3a stimulation and Ref-1 redox inhibitors do not affect mRNA expression of APEX1 in HRECs. G, bEnd.3 cells were pre-treated with 2 μM GSK3 inhibitor CHIR99021 for 2 h and then treated with 10 μM APX2009, 10 μM RN7-58, or 300 ng/ml DKK1 for 24 h. Dual-luciferase assay demonstrates that APX2009 inhibits CHIR99021-induced Wnt activation. One-way ANOVA with Tukey’s post hoc tests; n=3 biological replicates. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Data represented as mean ± SEM; ns = no significance. RLU = Relative light units.

While both APX3330 and APX2009 bind and inhibit Ref-1 activity, APX2009 is more potent [17,47]. Thus, we chose to focus only on APX2009 to further investigate the relationship between Ref-1 redox activity and canonical Wnt signaling, given the similar activity observed with both APX2009 and APX3330. Considering our observation that inhibition of Ref-1 redox activity decreases LRP5/6 expression, we hypothesized that this is the main mechanism by which Ref-1 redox activity regulates Wnt signaling. To investigate this, we utilized a GSK3 inhibitor, CHIR99021, to activate Wnt signaling downstream of LRP5/6. CHIR99021 induced Wnt transcriptional activation, and surprisingly, APX2009 decreased CHIR99021-induced Wnt transcriptional activation, but to a lesser extent than to what we observed with direct Wnt3a stimulation (Fig. 4G). Inactive control analog RN7-58 and endogenous Wnt inhibitor DKK1 did not inhibit CHIR99021-induced Wnt transcriptional activation. Collectively, these data provide evidence that Ref-1 redox activity modulates Wnt-mediated transcriptional activation partly by modulation of LRP5/6 receptor expression, but may be regulating Wnt/β-catenin signaling through other mechanisms in addition to LRP5/6 receptor regulation.

3.5. Ref-1 knockdown modestly decreases Wnt3a-induced nuclear β-catenin

Since Ref-1 has multiple functions, we sought to determine if Ref-1 redox activity is the major function of Ref-1 that is responsible for regulation of Wnt signaling. To determine this, we utilized Ref-1 siRNA to knockdown all functions of Ref-1 in HRECs. After stimulation with 200 ng/ml Wnt3a for 2 h and treatment with siRNA for Ref-1 (gene name APEX1) (siRef-1) or scrambled (Scr), we examined the nuclear levels of β-catenin following fractionation. We verified that we achieved at least 75 % knockdown of Ref-1 following siRef-1 (Fig. 5A–B, D-E). Wnt3a-induced increases in nuclear β-catenin, and siRef-1 modestly decreased nuclear β-catenin levels (Fig. 5A–C). Cytoplasmic β-catenin levels were unchanged (Fig. 5D–F). Immunocytochemistry and mean fluorescence intensity quantification confirmed increased nuclear β-catenin with Wnt3a stimulation and showed significantly less nuclear β-catenin with siRef-1 (Fig. 5G–H). These results demonstrate that knockdown of Ref-1 decreases Wnt3a-induced increases in nuclear β-catenin.

Fig. 5.

siRef-1 decreases Wnt3a-induced increases in nuclear β-catenin. HRECs were stimulated with 200 ng/ml Wnt3a for 2 h before being treated with 25 nM of scrambled (Scr) or Ref-1 siRNA (siRef-1) for 48 h. A-F, HRECs were fractionated and nuclear and cytoplasmic localization of β-catenin was assessed by immunoblot. A-B, siRef-1 knockdown of Ref-1 resulted in at least 75 % knockdown of nuclear Ref-1; A and C, Wnt3a-stimulation induced nuclear localization of β-catenin, and siRef-1 modestly decreased nuclear β-catenin. D-F, Cytoplasmic β-catenin levels were unchanged; siRef-1 treatment knocked down at least 75 % of cytoplasmic Ref-1 levels. One-way ANOVA with Tukey’s post hoc tests; n=3 biological replicates. G-H, HRECs were immunostained for Ref-1 (green), β-catenin (magenta), and DAPI (blue). Mean fluorescence intensity [68] of nuclear β-catenin indicates that siRef-1 decreases nuclear β-catenin. Yellow arrows indicate increased nuclear β-catenin with Wnt3a stimulation. One-way ANOVA with Tukey’s post hoc tests; n=51–53 cells across three biological replicates. Scale bars = 10 μm. ∗p < 0.05; ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Data represented as mean ± SEM.

3.6. Ref-1 knockdown does not decrease Wnt transcriptional activation

To further explore the effects of siRef-1 on Wnt/β-catenin signaling, we explored if siRef-1 affects LRP5/6 expression in HRECs. Interestingly, we observed only modest decreases in LRP5/6 expression with knockdown of Ref-1 with siRef-1 (Fig. 6A–E), with decreases in receptor expression only significant under Wnt3a stimulated conditions, despite at least 80 % knock down of Ref-1 (Fig. 6C–F). We completed similar experiments in bEnd.3 cells, confirming that siRef-1 knocked down Ref-1 at least 75 % (Fig. 6G–I) and that siRef-1 exhibited similar effects on LRP5 protein expression as in HRECs (Fig. 6G and H). Surprisingly though, we observed that siRef-1 slightly increased Wnt/β-catenin transcriptional activation (Fig. 6J). Together, these findings suggest that other functions of Ref-1, besides its redox regulatory activity, may act to repress Wnt/β-catenin signaling, at least in bEnd.3 cells.

Fig. 6.

Knockdown of Ref-1 modestly decreases LRP5/6 expression but does not decrease transcriptional activation of Wnt/β-catenin signaling. A-F, HRECs were stimulated with 200 ng/ml Wnt3a for 2 h before being treated with 25 nM scrambled (Scr) or Ref-1 siRNA (siRef-1) for 48 h. A-B, D-E, immunoblots reveal that siRef-1 modestly decreases LRP5 and 6 expression under Wnt3a-stimulated conditions. C and F, quantification of Ref-1 bands in A and D, respectively, demonstrate at least 80 % of Ref-1 protein was knocked down. G-I, bEnd.3 cells were stimulated with 200 ng/ml Wnt3a for 2 h before being treated with 50 nM Scr or siRef-1 for 48 h. G and H, immunoblot and quantification reveals that siRef-1 decreases LRP5 protein levels under Wnt3a-stimulated conditions. I, quantification of Ref-1 bands in G reveal that at least 75 % of Ref-1 was knocked down with siRef-1. J, Dual luciferase assay of bEnd.3 cells stimulated with 200 ng/ml Wnt3a and treated with 50 nM of Scr or siRef-1 for 48 h. siRef-1 increased Wnt/β-catenin transcriptional activation under Wnt3a-stimulated conditions. One-way ANOVA with Tukey’s post hoc tests; n=3 biological replicates. ∗p < 0.05; ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Data represented as mean ± SEM. RLU = Relative light units.

3.7. Wnt signaling is activated under hypoxia and reduced by inhibition of Ref-1 redox activity

Because there is evidence that canonical Wnt signaling is activated in neovascular retinal diseases [41,42,45,69,70], and because hypoxia is one of the main drivers underlying retinal neovascularization [[71], [72], [73]], we asked if canonical Wnt signaling is activated under hypoxic conditions in endothelial cells. We have previously shown that Ref-1 redox activity regulates hypoxia-induced endothelial cell proliferation [16], but whether or not that is linked to Wnt signaling remains elusive. HRECs were exposed to hypoxia (1 % O₂) for various time points, fractionated, and subcellular localization of β-catenin was examined by immunoblot. Nuclear β-catenin time-dependently increased under hypoxic conditions, with nuclear β-catenin being significantly upregulated at 24 h compared to room air (0 h hypoxia) (Fig. 7A–B). Cytoplasmic β-catenin was also strongly upregulated after 24 h in hypoxia (Fig. 7C–D). Next, we asked if inhibition of Ref-1 redox activity with APX2009 decreases hypoxia-induced nuclear β-catenin. HRECs were exposed to hypoxia or room air and treated with 10 μM APX2009 or vehicle (1 % DMSO) for 24 h. Hypoxia increased nuclear β-catenin levels, while APX2009 restored hypoxia-induced nuclear β-catenin levels back to baseline (Fig. 7E–F). APX2009 also restored hypoxia-induced cytoplasmic β-catenin levels (Fig. 7G–H). Dual luciferase assay of bEnd.3 cells exposed to hypoxia for 24 h (Fig. 7I) or 48 h (Fig. 7J) revealed that Wnt/β-catenin transcriptional activation is modestly increased after 24 h hypoxia and strongly increased after 48 h hypoxia and APX2009 restored hypoxia-induced Wnt/β-catenin transcriptional activation. RT-qPCR of Wnt transcriptional targets in HRECs revealed that MYC and PPARD were induced by hypoxia, and APX2009 restored these transcript levels (Fig. 7K–L). Interestingly, CCND1 was not induced by hypoxia, and APX2009 did not significantly affect its transcript levels in this context (Fig. 7M). Together, these results indicate that hypoxia induces Wnt/β-catenin signaling in endothelial cells, and inhibition of Ref-1 redox activity reduces hypoxia-induced Wnt/β-catenin signaling.

Fig. 7.

Wnt signaling is activated under hypoxia and inhibition of Ref-1 redox activity reduces hypoxia-induced Wnt/β-catenin signaling. A, immunoblot of nuclear fractions of HRECs subjected to hypoxia (1 % O₂) for 0 h (21 % O₂), 1 h, 4 h, or 24 h and B, quantification demonstrates that nuclear β-catenin is upregulated in HRECs exposed to hypoxia for 24 h. C, immunoblot of cytoplasmic fractions of HRECs subjected to hypoxia (1 % O₂) for 0 h (21 % O₂), 1 h, 4 h, or 24 h and D, quantification, demonstrates that cytoplasmic β-catenin is upregulated in HRECs exposed to hypoxia for 24 h. E, immunoblot of nuclear fractions of HRECs subjected to hypoxia (1 % O₂) or room air (21 % O₂) and treated with APX2009 or vehicle for 24 h and F, quantification, demonstrates that APX2009 inhibits hypoxia-induced increases in nuclear β-catenin. G, immunoblot of cytoplasmic fractions of HRECs subjected to hypoxia (1 % O₂) or room air (21 % O₂) and treated with APX2009 or vehicle for 24 h and H, quantification, demonstrates that APX2009 inhibits hypoxia-induced increases in cytoplasmic β-catenin. I-J, bEnd.3 cells co-transduced with TOPFlash reporter lentivirus and Renilla luciferase lentivirus were exposed to hypoxia (1 % O₂) or room air (21 % O₂) and treated with APX2009 or vehicle for 24 (I) or 48 (J) hours. Hypoxia induced Wnt/β-catenin pathway activation modestly at 24 h (I) and strongly at 48 h (J) (note different y-axis scales). K-M, RT-qPCR of HRECs exposed to hypoxia (1 % O₂) or room air (21 % O₂) and treated with APX2009 or vehicle for 24 h reveals that TCF7 transcriptional targets MYC (K) and PPARD (L) are induced by hypoxia and APX2009 decreases their mRNA levels. (M), TCF7 transcriptional target CCND1 was not induced by hypoxia, and APX2009 had no effect on CCND1 mRNA transcript expression. Two-way ANOVA with Tukey’s post hoc tests; n=3 biological replicates. ∗p < 0.05; ∗∗p < 0.01, ∗∗∗∗p < 0.0001. Data represented as mean ± SEM. RLU = Relative light units.

3.8. Canonical Wnt signaling is not the only signaling pathway by which Ref-1 regulates endothelial cell proliferation

To determine if inhibition of the canonical Wnt signaling is a major pathway by which Ref-1 redox inhibitor modulates endothelial cell proliferation, we utilized a lentiviral vector to constitutively activate β-catenin. To do this, we utilized a vector in which four key phosphorylation residues in β-catenin are mutated to alanines to prevent phosphorylation and degradation of β-catenin (Suppl. Fig. 5A) [58,74]. HRECs infected with constitutively activated β-catenin (CA-β-catenin) or empty vector control (EV) lentiviruses were verified by immunoblot. CA-β-catenin transduced cells exhibited a V5-reactive band at 93 kDa, the expected molecular weight of tagged active β-catenin, and exhibited more active (non-phosphorylated) β-catenin levels compared to EV transduced cells (Suppl. Fig. 5B). Next, we did a proliferation assay with these lines to determine if CA-β-catenin affects the anti-proliferative properties of APX2009. Interestingly, CA-β-catenin did not significantly affect the GI₅₀ values of APX2009 in HRECs compared to EV transduced HRECs or un-transduced HRECs (Un-transduced HRECs GI₅₀ APX2009: 2.3 μM; EV GI₅₀ APX2009: 1.6 μM; CA-β-catenin GI₅₀ APX2009: 1.6 μM) (Suppl. Fig. 5C). These data suggest that canonical Wnt signaling is not the only signaling pathway by which Ref-1 regulates endothelial cell proliferation.

3.9. Inhibition of Ref-1 redox activity decreases the expression of Wnt target genes in the whole retina of OIR mice

Given the evidence above demonstrating the ability of Ref-1 redox inhibitors to decrease canonical Wnt/β-catenin signaling activation in endothelial cells, we sought to determine if Ref-1 redox inhibitor APX2009 would reduce the expression of Wnt genes in vivo. To do this, we utilized the OIR mouse model. P7 neonatal pups and nursing mothers were placed at 75 % O₂ (hyperoxia) for 5 days. On P12, the pups and nursing mothers were returned to room air (21 % O₂). This transition causes relative hypoxia, inducing the activation of pro-angiogenic factors that lead to abnormal vessel growth to support the energetic demands of the retina [60], plus upregulation of Ref-1 [16]. Pups received twice daily i.p. injections of 12.5 mg/kg APX2009 or vehicle. We have previously shown that twice daily i.p. injection of 12.5 mg/kg APX2009 significantly decreases retinal neovascularization in the OIR mouse model [16]. After one day of relative hypoxia and one day of APX2009 treatments (P13), we investigated if Wnt target genes were upregulated in OIR mice compared to room air mice, and if APX2009 could decrease the expression of Wnt target genes to see if inhibition of Ref-1 redox activity affects Wnt signaling mechanisms that are responsible for induction of neovascularization. RT-qPCR on whole retinas demonstrated that Vegfa was induced at P13, and APX2009 decreased Vegfa (Fig. 8A). Further, Wnt target genes Myc and Ccnd1 were upregulated in P13 OIR mouse retinas and were decreased with APX2009 treatment (Fig. 8B–C). Wnt target gene Ppard was slightly induced in P13 OIR retinas, although not significantly. APX2009 decreased Ppard expression compared to P13 OIR mice receiving vehicle control (Fig. 8D). Lrp5 expression was not affected by OIR or APX2009 in P13 OIR whole retinas (Fig. 8E).

Fig. 8.

Wnt target genes are induced in P13 OIR retinas and APX2009 decreases the expression of Wnt target genes. A, Vegfa, a pro-angiogenic gene, was induced in OIR retinas, and APX2009 reduced Vegfa expression. B–C, Wnt target genes Myc (B) and Ccnd1 (C) were upregulated in P13 OIR retinas and APX2009 decreased their expression. D, Wnt target gene Ppard was not significantly induced in P13 OIR retinas but was reduced with APX2009 treatment. E, Wnt receptor gene Lrp5 was not significantly affected by OIR or APX2009 in whole P13 OIR retinas. One-way ANOVA with Tukey’s post hoc tests; n=3 biological replicates; each data point represents 2 pooled OIR retinas; ∗p < 0.05; ∗∗p < 0.01. Data represented as mean ± SEM.

3.10. Inhibition of Ref-1 redox activity reduces expression of Wnt genes at sites of neovascularization in the OIR mouse model

Next, we examined if Wnt genes were induced in P17 OIR retinas and if APX2009 decreased their expression, since P17 is at the peak of retinal neovascularization in the OIR mouse model. Whole P17 OIR retinas were subjected to RNAscope in situ hybridization to observe spatial distribution of Wnt mRNA transcripts in the OIR retina. We observed no significant differences in the mRNA levels of Lrp5 in the whole P17 OIR retina (Fig. 9A and B), but we observed significant increases in Lrp5 at the pre-retinal areas with neovascular tufts and in the retinal ganglion cell layer (GCL) of the OIR retina (Fig. 9A and C). APX2009 normalized the mRNA transcript levels of Lrp5 at sites of neovascularization at the GCL compared to OIR mice receiving vehicle. As a positive control, a group of mice received a single IVT injection of mouse anti-VEGF₁₆₄ antibody on P12. Anti-VEGF also normalized the levels of Lrp5 in the GCL (Fig. 9C). We also examined the spatial expression of Fzd4, since it has been implicated in retinal neovascularization [34,75]. We observed increases in Fzd4 in the GCL and neovascular tufts of OIR mice receiving vehicle, and APX2009 and anti-VEGF decreased Fzd4 in the GCL (Fig. 9D–F). Fzd4 in the whole retina remained unchanged (Fig. 9E). As a positive control, we examined the expression of Vegfa, since it is well known that Vegfa is strongly induced in P17 OIR retinas, and anti-angiogenic treatments decrease Vegfa expression [[76], [77], [78]]. Vegfa was induced in whole P17 OIR retinas, in P17 OIR retinas at the GCL and neovascular tufts, and in the VEGF-producing Müller glia in the INL. APX2009 and anti-VEGF decreased Vegfa in the whole retina, in the GCL/pre-retinal area, and at the Müller glia in the INL (Fig. 9G–J). We observed no changes in the expression of Fzd1 or Apex1 in these mice (Suppl. Fig. 6). We further verified our RNAscope in situ hybridization techniques with positive and negative controls (Suppl. Fig. 7). Together, these results demonstrate that inhibition of Ref-1 redox function minimizes retinal neovascularization associated with either decreasing or preventing the upregulation of Lrp5, Fzd4, and Vegfa at the sites of neovascularization.

Fig. 9.

Inhibition of Ref-1 redox activity reduces expression of Wnt genes at sites of neovascularization in the OIR mouse model. A-C, RNAscope in situ hybridization of Lrp5 in P17 neonatal OIR mouse retinas that received vehicle, APX2009, a single IVT injection of mouse anti-VEGF₁₆₄ antibody, or mice that remained at room air (control). B, no total changes in Lrp5 expression in whole retina observed. C, OIR mice exhibited increases in Lrp5 at the GCL, associated with sites of neovascularization. Mice treated with APX2009 or anti-VEGF demonstrated decreased Lrp5 expression, comparable to room air. D-F, RNAscope in situ hybridization of Fzd4 in OIR mice. E, no total changes in Fzd4 expression in whole retina observed. F, OIR mice exhibited increases in Fzd4 at the GCL, the site of neovascularization. Mice treated with APX2009 or anti-VEGF demonstrated decreased Fzd4 expression, comparable to room air. G-J, RNAscope in situ hybridization of Vegfa in OIR mice. H–J, OIR mice exhibited increases in Vegfa in the whole retina (H), at the GCL near the site of neovascularization (I), and in the INL where Müller glia (major VEGF producers) cell bodies are located (J). H-J, Mice treated with APX2009 or anti-VEGF demonstrated decreased Vegfa expression, comparable to room air. One-way ANOVA with Tukey’s post hoc tests; n = 12–24 per group; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001. Data represented as mean ± SEM. Blue = DAPI; magenta = Lrp5 (A), Fzd4 (D), Vegfa (G). Scale bars = 20 μm; 10 μm in insets. GCL = ganglion cell layer; INL = inner nuclear layer; ONL = outer nuclear layer.

4. Discussion

In this study, we present evidence demonstrating a novel link between Ref-1 redox activity and canonical Wnt/β-catenin signaling in endothelial cells (Fig. 10). Our study is the first to explore the mechanism by which Ref-1 redox activity modulates canonical Wnt signaling in the retina, although some earlier work noted crosstalk between Ref-1 signaling and Wnt signaling in cancer cell lines [79,80]. This study further highlights the potential of Ref-1 redox inhibitors as novel therapeutics for vascular retinopathies by modulating multiple signaling pathways involved in aberrant neovascularization.

Fig. 10.

Proposed mechanism by which Ref-1 redox activity modulates canonical Wnt signaling in endothelial cells. In the presence of a Wnt ligand, LRP5/6 and a Frizzled receptor form a complex, allowing for β-catenin to accumulate in the cytoplasm, translocate to the nucleus, and initiate Wnt-mediated transcription. Target genes such as CCND1 (encoding Cyclin D1), PPARD (encoding PPARδ), and MYC (encoding c-Myc) are upregulated. Hypoxia can induce Wnt signaling activation. In the presence of a Ref-1 redox inhibitor, such as APX2009, Ref-1 redox activity is inhibited, and Ref-1 is unable to reduce transcription factors (TF). We propose that Ref-1 directly reduces a TF responsible for LRP5/6 expression. With Ref-1 inhibition, LRP5/6 levels decrease, which prohibits Wnts from forming a complex with LRP5/6 and a Frizzled receptor. β-catenin is then degraded, and Wnt-mediated transcription is reduced. Wnt target genes CCND1, PPARD, and MYC are decreased.

Using transcriptomics, we discovered Wnt/β-catenin to be modulated by Ref-1 redox activity. Top down-regulated pathways altered by Ref-1 redox inhibition included JAK-STAT signaling, TNFR1 signaling, PPARα/RXRα signaling, apoptosis signaling, mitochondrial biogenesis, and generic transcription, all of which have been previously characterized under Ref-1 redox control in other cell types [47,[81], [82], [83], [84], [85]]. Interestingly, Wnt/β-catenin signaling and β-catenin independent Wnt signaling both came up as top down-regulated pathways as well. The top up-regulated pathways identified after Ref-1 redox inhibition included degradation of β-catenin by the destruction complex, in addition to other pathways known to be up-regulated following Ref-1 redox inhibition, including p53 signaling, cell cycle regulation, and autophagy [17,81,82,86]. Furthermore, we revealed that Wnt signaling genes are overexpressed in endothelial cells of PDR FVMs compared to endothelial cells in control retinas, underlining the role of Wnt signaling in pathological neovascularization. While the observed increases in CTNNB1 (encoding β-catenin) in both PDR FVMs and in HRECs following APX2009 treatment may seem counterintuitive, we hypothesize that the increases in CTNNB1 following APX2009 treatment may be a compensatory mechanism due to inhibition of Wnt signaling, while the increases in CTNNB1 and downstream Wnt target genes MYC and CCND1 in PDR FVMs suggest that the endothelial cells in the FVMs may be promoting overactivation of Wnt signaling. Following these results, we sought to further understand the link between Ref-1 redox activity and Wnt signaling.

We validated that Ref-1 redox inhibitors (clinical candidate APX3330 and more potent second-generation compound APX2009) decreased LRP5/6 mRNA and protein expression. Given this, we hypothesize that Ref-1 redox activity modulates a redox-regulated transcription factor that binds to the promoters of LRP5/6. This factor remains to be identified in future experiments. Our RNA-sequencing results also demonstrated that APX2009 decreased the mRNA levels of several Frizzled receptors and other Wnt signaling components. However, APX2009 did not affect the protein levels of Frizzled 5 and Frizzled 1. Interestingly, high doses of APX2009 modestly decreased the expression of Dishevelled 2 in HRECs. Loss of Dishevelled 2 decreases neovascularization in the OIR mouse model [45], further highlighting the anti-proliferative potential of Ref-1 redox inhibitors via inhibition of the Wnt signaling pathway.

Importantly, we validated that Ref-1 levels (mRNA and protein) are not affected by Wnt3a stimulation or Ref-1 redox inhibitors, confirming that effects we observed on the Wnt signaling pathway are due to inhibition of Ref-1 redox activity, and not due to total Ref-1 levels. Furthermore, we validated that RN7-58, a structural analog of APX2009 that exhibits no inhibition of Ref-1 redox activity [47,67], does not affect LRP5/6 levels or transcriptional activation of Wnt signaling by TOPFlash luciferase assay. These experiments further confirmed that the inhibitory effects on Wnt signaling we observed are due to on-target inhibition of Ref-1 redox activity.

To functionally validate our findings, we demonstrated that Wnt3a stimulation induced nuclear localization of β-catenin, and Ref-1 redox inhibitors APX2009 and APX3330 decreased nuclear localization of β-catenin, with APX2009 showing higher potency consistent with findings in other assays [87,88]. Wnt transcriptional targets MYC, CCND1, and PPARD were potentiated with Wnt3a stimulation, and were decreased with APX2009 and APX3330 treatment. Although PPARD did not significantly change with APX3330 treatment, our results still reflect the improved potency of APX2009 compared to APX3330. Other reports have noted that Ref-1 redox activity facilitates cyclin D1 expression, further validating our results [79,89].

We hypothesized that Ref-1 redox activity modulates Wnt signaling via regulating expression of LRP5/6. To test this, we utilized CHIR99021, a small molecule that induces Wnt signaling by inhibiting GSK3β. Interestingly, we observed that Ref-1 redox inhibition still downregulated CHIR99021-mediated Wnt transcriptional activation, although not to the same extent as Wnt3a-mediated activation. This suggests that Ref-1 redox activity modulates Wnt signaling via more than regulation of LRP5/6 alone. Given that our RNA-seq showed changes in β-catenin destruction complex components, this necessitates future exploration of other mechanisms by which Ref-1 redox activity could modulate Wnt signaling.

We investigated if knockdown of Ref-1 would offer similar results to Ref-1 redox inhibition and observed a modest decrease in nuclear β-catenin and no effects on cytoplasmic β-catenin, suggesting that knockdown of Ref-1 prevents shuttling of β-catenin from the cytoplasm to the nucleus and exhibits no effects on proteasomal degradation of β-catenin. We have previously shown that Ref-1 redox inhibitors do not affect proteasome activity [16]. Furthermore, knockdown of Ref-1 exhibited no inhibitory effects on Wnt transcriptional activity by TOPFlash luciferase assay. This could be due to dynamic effects of Ref-1 knockdown on Wnt signaling or variations in TCF variants that affect the repression or activation of Wnt-mediated transcriptional activation [[90], [91], [92], [93], [94]]. Furthermore, knockdown of Ref-1 knocks down not only Ref-1’s redox regulatory function, but also its other functions – DNA repair, RNA processing, chaperone activity – thus introducing confounding variables and inducing cell stress. Currently, there are no specific, non-toxic inhibitors for APE1/Ref-1’s DNA repair activity. CRISPR-modified cell lines that target specific functions of APE1/Ref-1 could be utilized in the future to help clarify the role of APE1/Ref-1’s functions in Wnt signaling.

To our surprise, constitutively active β-catenin did not change APX2009's growth inhibitory effects on HRECs. Because Ref-1 regulates other cell signaling pathways involved in endothelial cell proliferation (such as NF-κB, STAT3, p53, etc.), this indicates that inhibition of Wnt signaling is not the sole modality by which Ref-1 redox activity modulates endothelial cell proliferation. However, constitutively active β-catenin might still have effects on other proangiogenic endothelial cell functions beyond proliferation, such as tube formation or migration, which requires future investigation.

Given the potential for Ref-1 redox inhibitors to be therapeutics for ischemic retinopathies, we sought to determine if hypoxia induces Wnt/β-catenin signaling and confirmed this in endothelial cells. To our knowledge, this is the first time the cellular localization and expression of β-catenin has been characterized in hypoxic endothelial cells. Previous evidence indicates that hypoxia activates Wnt signaling in cancer cell lines, supporting our findings [95,96]. The strong upregulation of cytoplasmic β-catenin is intriguing and opens an avenue of future exploration. Both β-catenin and HIF-1α utilize CBP/p300 as co-activators for their transcriptional activation [97,98]. Could this be why we observed delayed increases in β-catenin under hypoxic conditions and more pronounced Wnt transcriptional activation after chronic hypoxia exposure? Further studies that explore the kinetics and mechanisms of hypoxia-mediated Wnt activation could answer questions on how exactly Wnt signaling becomes activated under hypoxia. In addition, Ref-1 forms complexes with CBP/p300 in the context of HIF-1α and STAT3 signaling [99], suggesting an additional link between Ref-1 and β-catenin via CBP/p300 that requires further investigation. Literature also points to increases in hypoxia driving NADPH oxidase 1 (NOX1) production of reactive oxygen species in endothelial cells, which can influence Ref-1 activity and lead to the promotion of endothelial cell proliferation and vascular remodeling [100,101]. Thus, perhaps not only Ref-1 itself but the redox milieu in general may influence Wnt signaling in endothelial cells. This requires further investigation.

To explore Ref-1 redox inhibition effects in vivo, we utilized the OIR mouse model. In the OIR mouse model, ischemia drives aberrant retinal neovascularization, with most pronounced neovascularization at P17. Following P17, this neovascularization naturally regresses and eventually resolves [102]. Early in relative hypoxia (P13), we revealed that Vegfa and Wnt-target genes are upregulated in the OIR retina, contributing to the pro-angiogenic environment driving ischemic neovascularization. Ref-1 redox inhibition downregulated the expression of Wnt target genes in the whole retina, but not the expression of Lrp5. Likely, Ref-1’s redox regulation of LRP5/6 is cell type dependent, thus explaining why we do not see major changes in Lrp5 in the whole retina after Ref-1 redox inhibition. But on P17, at the peak of neovascularization, we observed increases in Lrp5 at the ganglion cell layer and epiretinal layers, the sites of neovascularization. APX2009 and anti-VEGF attenuated OIR-induced expression of Lrp5 at the sites of neovascularization.

Fzd4 is overexpressed in retinal neovascularization, is required for angiogenesis [34,103], and has been shown to be upregulated under hypoxia [104]. We therefore explored the effects of Ref-1 redox inhibition on Fzd4 expression. Fzd4 was upregulated at sites of neovascularization in the OIR mouse model, and Ref-1 redox inhibition restored Fzd4 levels back to baseline. This suggests that Ref-1 redox activity might regulate ischemia-driven upregulation of Fzd4, even though it wasn’t one of the top Wnt DEGs in our RNA-seq analysis, but it requires further investigation. Previous research supports our findings: Lrp5 and Fzd4 are upregulated in neovascular tufts and loss of LRP5 decreases neovascularization in OIR [45]. Anti-VEGF has been shown to decrease Lrp5 and Fzd4 expression, further corroborating our findings [105].

Validating our in vitro data, we observed no changes in Fzd1 expression in the OIR model (consistent with previous data) [103] or after Ref-1 redox inhibition. Furthermore, we observed no changes in Apex1 expression in the OIR mouse model or with Ref-1 redox inhibition. Although we have previously published that Ref-1 protein levels are overexpressed in the OIR mouse model compared to controls [16], here the mRNA transcripts of Apex1 remain unchanged. This is not surprising as at P17, retinal neovascularization has peaked, and the retinal vessels are preparing to regress, thus not necessitating high transcription of Ref-1 to drive neovascularization.

We observed overexpression of Vegfa in the whole retina, at sites of neovascularization (ganglion cell layer/epiretinal layer), and the inner nuclear layer, where cell bodies of VEGF-producing Müller glia reside. VEGF₁₆₄, the major isoform of VEGF-A in mice, has been observed to peak at P17 in OIR mice compared to room air, supporting our observations [77]. Consistent with our previous studies, Ref-1 redox inhibition restored Vegfa levels back to baseline [16].

Given the role of Wnt signaling in vasculature development [42], future studies are needed to explore the effects of Ref-1 redox inhibition on development. Promisingly, P5 neonates in a murine model of necrotizing enterocolitis (NEC) that received twice daily i.p. injections of APX3330 until P9 showed improved NEC outcomes without causing weight loss or reduced white blood cell count, suggesting that perhaps Ref-1 redox inhibitors could be tolerated in neonates [106]. We also did not explore any differences between sexes in this study, although previous studies have shown that APX2009’s therapeutic effects in OIR are not sex-dependent [16].

Overactivation of Wnt signaling has been observed in proliferative retinopathies, while loss-of-function mutations in Wnt signaling leads to hypovascularization, indicating a complex and critical role of Wnt signaling in endothelial cell function [[40], [41], [42], [43], [44]]. While Wnt agonists are currently in clinical trials for diabetic macular edema (a complication of DR characterized by blood-retinal-barrier dysfunction) (NCT06571045), these therapeutics potentiate Norrin signaling, a non-canonical Wnt ligand that promotes healthy blood-retinal-barrier function. Future work is needed to explore the effects of Ref-1 redox inhibition on Norrin signaling and the effects on blood-retinal-barrier integrity. Interestingly, Ref-1 redox inhibitor APX3330 recently completed a Phase IIb clinical trial for DR and diabetic macular edema (NCT04692688). APX3330 was safe and prevented worsening of disease [107], suggesting that Ref-1 redox inhibition does not negatively affect genes involved in blood-retinal-barrier integrity.

In conclusion, this study provides direct evidence that Ref-1 redox activity modulates canonical Wnt signaling in endothelial cells. Hypoxia, one of the key drivers of neovascular retinal disease, induces canonical Wnt signaling activation in endothelial cells, suggesting that the activation of Wnt signaling could potentiate neovascularization. Our identification of Ref-1 regulation of Wnt signaling further adds to the therapeutic potential of Ref-1 redox inhibitors for neovascular eye diseases. Modulation of endothelial cell Ref-1 redox activity may not only be advantageous for vascular retinopathies, but for other diseases such as cancers in which pathological angiogenesis plays a role.

CRediT authorship contribution statement

Gabriella D. Hartman: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization. Kamakshi Sishtla: Writing – review & editing, Validation, Methodology, Investigation. Eyram K. Kpenu: Writing – review & editing, Validation, Investigation, Formal analysis. Mahmut Mijit: Writing – review & editing, Validation, Investigation. Anbukkarasi Muniyandi: Writing – review & editing, Methodology, Investigation. Ha-Neul Jo: Writing – review & editing, Resources, Methodology. Harald J. Junge: Writing – review & editing, Supervision, Resources, Conceptualization. Aaron Shaw: Resources, Investigation. Daniela Bischof: Writing – review & editing, Supervision, Resources. Sheng Liu: Writing – review & editing, Investigation, Formal analysis, Data curation. Jun Wan: Writing – review & editing, Supervision, Investigation, Formal analysis, Data curation. Mark R. Kelley: Writing – review & editing, Writing – original draft, Visualization, Supervision, Resources, Methodology, Funding acquisition, Conceptualization. Timothy W. Corson: Writing – review & editing, Writing – original draft, Visualization, Supervision, Resources, Methodology, Funding acquisition, Conceptualization.

Funding sources

This work was supported by the National Eye Institute [grant numbers R01EY031939, R01EY024261, R01EY033316, and F31EY035171]; the National Cancer Institute [R01CA167291, R01CA254110, and R01CA282478]; and the National Heart, Lung, and Blood Institute [R01HL140961]; the Riley Children’s Foundation; the IU Simon Comprehensive Cancer Center [P30CA082709]; the Canada Foundation for Innovation; the Natural Sciences and Engineering Research Council [RGPIN-2025-04563]; the Retina Research Foundation; and a Challenge Grant from Research to Prevent Blindness, Inc.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests. Mark R. Kelley reports a relationship with Apexian and Opus Genetics that includes: consulting or advisory and equity or stocks. Timothy W. Corson has a patent licensed to Apexian and Opus Genetics. Mark R. Kelley has patents licensed to Apexian and Opus Genetics. Other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.