Neuropilin-2 Deficiency Promotes Mitochondrial Dysfunction and NAD⁺-Dependent Cellular Senescence in Retinal Degeneration

Abstract

Purpose

Retinal pigment epithelial (RPE) dysfunction is a central pathological feature of retinal degenerative diseases, leading to irreversible vision loss. RPE dysfunction is substantially driven by mitochondrial impairment and senescence. However, the upstream regulators of these processes remain largely undefined. This study investigated the role of neuropilin-2 (NRP2) in RPE homeostasis and explored mitochondrial-targeted therapy with echinacoside (ECH) as a potential intervention for retinal degeneration.

Methods

RPE-specific Nrp2 conditional knockout mice were generated using AAV-VMD2-Cre and retinal morphology and function were assessed by fundus imaging, optical coherence tomography, histology, and electroretinogram (ERG). Cellular and metabolic phenotypes were examined in NRP2-deficient ARPE-19 cells and validated in vivo. RPE senescence, mitochondrial function, NAD⁺ metabolism, and sirtuin activity were analyzed, and the effects of ECH treatment were evaluated both in vitro and in vivo.

Results

RPE-specific Nrp2 deletion resulted in progressive RPE atrophy, photoreceptor loss, and impaired ERG responses. NRP2 deficiency led to mitochondrial elongation, elevated ROS, membrane depolarization, and reduced NAD⁺/NADH ratios. Decreased NAD⁺ levels were accompanied by downregulation of SIRT1/SIRT3 and increased protein acetylation, promoting RPE senescence. Restoring NAD⁺ levels or ECH treatment rescued mitochondrial dysfunction and reduced senescence markers in vitro. Furthermore, in vivo ECH administration preserved retinal structure and visual function in NRP2-deficient mice.

Conclusions

NRP2 is a critical regulator of mitochondrial integrity and NAD⁺ metabolism in the RPE. Its loss disrupts metabolic homeostasis, promoting RPE senescence and retinal degeneration. Targeting the NRP2–mitochondria-NAD⁺ axis with echinacoside represents a promising therapeutic strategy for preventing retinal degenerative diseases.

Cellular senescence is a hallmark of aging and a major contributor to age-related diseases.¹ In the retina, senescence and dysfunction of retinal pigment epithelial (RPE) cells, are closely associated with age-related macular degeneration (AMD), the leading cause of blindness in the elderly.² Although RPE degeneration is known to involve mitochondrial impairment and disrupted metabolic coupling with photoreceptors,³ the upstream molecular mechanisms that compromise RPE metabolic homeostasis remain poorly understood.

RPE cells sustain retinal integrity through intricate autocrine and paracrine signaling networks, including the VEGF pathway and its co-receptors.^4,5 Perturbations of these signaling impair RPE metabolism and promote oxidative stress, contributing to cellular senescence and retinal degeneration.^3,6 Identifying molecules that coordinate extracellular cues with intracellular metabolic programs is therefore critical to understanding AMD pathogenesis.

Neuropilin-2 (NRP2), a transmembrane co-receptor, for VEGF and other ligands, integrates diverse extracellular signals to regulate cellular metabolism, survival, and responses to metabolic stress.^7,8 These functions might be important for RPEs during the challenges of aging. While NRP2 has been studied in vascular biology and tumor progression,^9,10 its role in retinal metabolism and AMD pathogenesis remains unexplored. Given the shared features between RPE cells and epithelial tumor cells, including autocrine VEGF signaling and metabolic adaptability,^11,12 it is possible that NRP2 serves as a critical regulator linking extracellular cues to mitochondrial function in RPE cells.

Here, we identify NRP2 as a critical upstream regulator of RPE metabolic homeostasis. NRP2 depletion in RPEs disrupts mitochondrial and NAD^⁺ metabolism, leading to premature cellular senescence and progressive retinal degeneration. Furthermore, we demonstrate that the natural compound echinacoside can intercept this degenerative cascade by restoring both mitochondrial function and NAD^⁺ metabolism, providing proof-of-concept for metabolic intervention strategies. These findings reveal a previously unrecognized mechanism by which extracellular matrix signals govern mitochondrial energetics in the RPE and highlights NRP2 as a novel therapeutic target for retinal degeneration diseases.

Methods

Animals

All experiments were carried out in accordance with the guidelines of the Chinese Association for Laboratory Animal Science (CALAS) and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, and were approved by the Animal Care and Use Committee of Zhongshan Ophthalmic Center (Ethics ID: 2020-093). Mice were housed in individually ventilated cages on a 12-hour light/dark cycle at 22°C ± 2°C with 50%–60% humidity, with food and water access as desired.

Nrp2^flox/flox mice with a C57BL/6J background were purchased from Cyagen Biosciences (Strain Number: CKOCMP-18187-Nrp2-B6J-VC). Mice received subretinal injections of 1 µL AAV2/8 vectors (EGFP or EGFP-Cre driven by the hVMD2 promoter; OBiO Technology, Shanghai, China) using a Hamilton microsyringe with a 33-gauge needle. Successful bleb formation was confirmed, and animals with failed injections were excluded. AAV transduction efficiency and Cre specificity were validated (Supplementary Fig. S5). Animals were assigned to groups based on endpoints: (1) NRP2 knockout characterization at four weeks after injection, (2) long-term retinal degeneration assessment at eight weeks after injection, and (3) echinacoside treatment efficacy evaluation at four weeks after treatment.

For echinacoside treatment, mice received daily oral gavage of echinacoside (PCS0849, 80 mg/kg; HerbSubstance, Chengdu, China) or vehicle starting immediately after AAV injection and continuing for four weeks. Comprehensive pharmacological properties of echinacoside are detailed in Supplementary Information S6.

All animals were age-matched (six to eight weeks old at injection) with equal male/female distribution across experimental groups. Sample sizes for each experiment are specified in the corresponding figure legends.

Fundus Imaging, Optical coherence tomography, and Electroretinogram Recordings

Fundus images were captured using Phoenix Micron IV (Phoenix-Micron, Inc, Bend, OR, USA). Optical coherence tomography (OCT) scans were acquired with 30 frames per scan, and outer nuclear layer (ONL) thickness was measured at 100 µm intervals from the optic disc using ImageJ. Scotopic electroretinography (ERG) was performed using the RetiMINER System (Chongqing IRC Medical Equipment Co., Ltd., Chongqing, China) after overnight dark adaptation. Flash intensities ranged from 0.001 to 10.0 log cd · s/m². A-wave amplitude was measured from baseline to trough, and b-wave amplitude from a-wave trough to peak.

Histology and Immunostaining

Eyes were fixed in 4% formalin, embedded in paraffin, and sectioned at 3 µm. Histological examination by hematoxylin and eosin (H&E) staining followed standard protocols. For immunostaining, eyes were fixed in 4% paraformaldehyde and sectioned at 10 µm. Sections were blocked with 5% BSA/0.5% Triton X-100, incubated with primary antibodies overnight at 4°C, followed by secondary antibodies for one hour at room temperature. Images were acquired using LSM 980 Airyscan (Zeiss, Oberkochen, Germany) or TissueGnostics Tissue FAXS Q^⁺ (TissueGnostics, Vienna, Austria). For RPE flatmounts, posterior eyecups were fixed, dissected, and stained with primary antibody overnight at 4°C. For each eye, three representative sections per mouse were analyzed by an experienced technician blinded to experimental groups.

Cell Culture and Gene Manipulation

ARPE-19 cells (Wuhan Pricella Biotechnology Co., Ltd., Wuhan, China) were cultured in DMEM/F12 with 10% FBS at 37°C with 5% CO₂. For RPE differentiation experiments, ARPE-19 cells were cultured in MEM-Nic differentiation medium¹³ for two weeks to promote RPE marker expression before experimental manipulations.

For CRISPR knockout, cells were transduced with lentiviruses expressing Cas9 and sgRNAs targeting NRP2 (sequences in Supplementary Table S3). Single clones were selected with puromycin (1 µg/mL) and validated by Western blot.

For inducible knockdown, cells were transduced with lentiviral vectors expressing tetracycline-inducible shRNA targeting NRP2 (sequences in Supplementary Table S4). Knockdown was induced with 1 µg/mL tetracycline for 72 hours. For drug treatments, cells were treated with K₃[Fe(CN)₆] (702587, 0.1 mM; Sigma-Aldrich Corp., St. Louis, MO, USA) or echinacoside (5 µM, PCS0849; HerbSubstance) for indicated times.

Senescence and Proliferation Assays

SA-β-galactosidase activity was detected using the Senescence β-Galactosidase Staining Kit (9860; Cell Signaling Technology, Danvers, MA, USA). Positive cells were quantified from five random fields per condition. Nuclear morphology was assessed by Lamin A/C and DAPI staining, with cells exhibiting nuclear blebs or extranuclear DNA fragments scored as abnormal. Proliferation was assessed using the VF 488 Click-iT EdU Kit (HY-K1087; MedChemExpress, Monmouth Junction, NJ, USA) according to the manufacturer's instructions.

Mitochondrial Function Assays

Mitochondrial morphology was visualized by TOMM20 immunostaining or transmission electron microscopy (TEM). For TEM, cells were fixed in 2.5% glutaraldehyde, post-fixed with 1% osmium tetroxide, and embedded in resin. Sections (100 nm) were imaged using Tecnai G2 Spirit. Mitochondrial mass, mitochondrial ROS, and membrane potential were assessed by flow cytometry using MitoTracker Green (M7514; Invitrogen, Carlsbad, CA, USA), MitoSOX (M36008; Invitrogen), and JC-1 (T3168; Invitrogen), or TMRM (T668; Invitrogen), respectively.

Mitochondria Isolation and NAD⁺ Measurement

Mitochondria isolation was performed by the Mitochondria Isolation Kit (ab288084; Abcam, Cambridge, MA, USA). NAD⁺/NADH ratios were measured using the NAD⁺/NADH Quantification Kit (ab65348; Abcam) and normalized to protein content.

RNA Sequencing

Total RNA was extracted using TRIzol and sequenced on the BGI G400 platform (BGI Group, Shenzhen, China). Reads were aligned and analyzed for differential expression (adjusted P < 0.05, fold change > 1.41). KEGG pathway enrichment was performed using standard databases. Heatmaps were generated using the pheatmap package in R with z-score normalized data and hierarchical clustering.

Western Blotting and RT-qPCR

Cells were lysed in RIPA buffer with protease/phosphatase inhibitors (HY-K0011/HY-K0023; MedChemExpress). Proteins were separated by SDS-PAGE, transferred to PVDF membranes, and probed with primary antibodies overnight at 4°C. Antibody details are in Supplementary Table S1. Full-length Western blot images with molecular weight markers for all antibodies used in this study are provided in Supplementary Figure S8.

For RT-qPCR, RNA was extracted with TRIzol and reverse-transcribed using HiScript III (R312-02; Vazyme Biotech Co., Ltd, Nanjing, China). QPCR was performed using ChamQ SYBR Mix (Q411-03; Vazyme Biotech Co., Ltd) on a QuantStudio 7 Flex system. Expression was normalized to 18S using the ΔΔCt method. Primer sequences are in Supplementary Table S2.

Statistical Analysis

Data are presented as mean ± SEM from at least three independent experiments. Normality was assessed by Shapiro-Wilk test. For normally distributed data, Student's t-test (two groups) or one-way ANOVA with Tukey's post hoc test (multiple groups) was used. For non-normally distributed data, Mann-Whitney U test, or Kruskal-Wallis test with Dunn's post hoc test was applied. P < 0.05 was considered significant. Analyses were performed using GraphPad Prism 9.5 (GraphPad, San Diego, CA, USA).

Results

RPE-Specific Nrp2 Deletion Causes Retinal Degeneration

To investigate NRP2 function in RPE, we generated RPE-specific Nrp2 deletion by subretinally injecting AAV-VMD2-EGFP-Cre into Nrp2^flox/flox mice, while control group received AAV-VMD2-EGFP injection (Fig. 1A). Immunostaining confirmed efficient loss of NRP2 in RPE cells, with NRP2 expression retained in control RPE (Fig. 1B), validating the RPE-specific targeting strategy.

Figure 1.

RPE-specific Nrp2 deletion causes retinal degeneration. (A) Illustration of AAV subretinal injection route. AAV pAAV-VMD2-EGFP-T2A-CRE-WPRE (1 µL, 5E + 11 vg/mL) was subretinally injected to Nrp2^flox/flox mice. Mock vector pAAV-VMD2-EGFP-WPRE (1 µL, 5E + 11 vg/mL) was injected to control eye. All tissues were harvested two months after subretinal injection unless indicated otherwise. (B) Representative immunofluorescence staining images of RPE65 (green) and NRP2 (red) in eye tissue sections (10 µm in thickness) two months after subretinal injection of either AAV-GFP or AAV-GFP-Cre in Nrp2^flox/flox mice. (C) Representative fundus photography and OCT images of Nrp2^flox/flox mice after AAV injection for two months. The white arrow line represents the OCT scanning plane. Note that aggregates are pointed out with yellow arrows in the eye after AAV-GFP-Cre subretinal injection. (D) Quantification of the ONL layer from OCT images (C). Each dot represents a measurement point. n = 8 mice per group. Data are mean ± SEM. Multiple unpaired t tests. (E) Representative H&E staining images of eye tissue sections (3 µm in thickness) of Nrp2^flox/flox mice after AAV injection. Upper, middle, and lower panels represent the central, middle, and periphery regions of the retina, respectively. The pigment aggregates are indicated by the yellow arrows. (F) Quantification of the total retina layer, the ONL layer and the IS/OS layer from the H&E staining images of the central region (E). Data are mean ± SEM. Each dot represents one mouse. n = 5 mice per group. Mann-Whitney U test. (G) Quantification of the RPE aggregates on top of the RPE layer per 500 µm length of retina from H&E staining images (E). Data are mean ± SEM. Each dot represents 1 mouse. n = 5 mice per group. Mann-Whitney U test. (H–J) Representative immunofluorescence staining images of ZO-1 (red) in RPE-choroid complex flatmounts of Nrp2^flox/flox mice after AAV injection, and quantification of RPE cell density and RPE cell size. Upper and lower panels represent the central and periphery regions of the RPE-choroid complex, respectively. Data are mean ± SEM. Each dot represents one field of view of the RPE-choroid complex, and three fields of view were taken per mice. For the AAV-GFP group, n = 21. For the AAV-GFP-Cre group, n = 18. I, Welch's t test. J, For the central region, Mann-Whitney U test. For the periphery region, Welch's t test. (K, L) Quantification of ERG scotopic and photopic a-wave and b wave amplitudes of Nrp2^flox/flox mice after AAV injection. Data are mean ± SEM. n = 10 mice per group. Mann-Whitney U test. For all panels, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Two months after injection, fundus imaging revealed striking abnormalities in NRP2-deficient eyes, including pale spots and punctate pigmentation (Fig. 1C, right). In contrast, control eyes receiving AAV-VMD2-EGFP showed uniform GFP fluorescence and intact RPE morphology (Fig. 1C, left). OCT imaging further demonstrated an indistinct OS/RPE boundary, and hyperreflective RPE aggregates at the level of the RPE layer (Fig. 1C, right), suggesting accumulation of pigmented debris. These changes were accompanied by significant thinning of the ONL in the central retina (Fig. 1D), indicating secondary photoreceptor degeneration due to RPE dysfunction.

Histological analysis corroborated the in vivo imaging findings. H&E staining revealed RPE/photoreceptor layer aggregates and overall retinal thinning, primarily in IS/OS and ONL layers in NRP2-deficient eyes (Figs. 1E–G). Whole-mount immunostaining for the tight junction protein zonula occludens-1 (ZO-1), which delineates RPE cell boundaries, demonstrated enlarged, irregular RPE cells and reduced cell density in NRP2-deficient eyes, in contrast to the regular hexagonal mosaic pattern observed in controls (Figs. 1H–J), consistent with RPE atrophy. Functionally, scotopic ERG recordings demonstrated significantly decreased a-wave and b-wave amplitudes (Figs. 1K, 1L), confirming impaired visual responses. Collectively, these findings establish that RPE-specific NRP2 loss induces RPE atrophy and progressive retinal degeneration.

Neuropilin-2 Deficiency Induces Cellular Senescence in the RPE

To investigate mechanisms underlying RPE degeneration, we used inducible shRNA-mediated NRP2 knockdown in ARPE-19 cells. Knockdown cells exhibited characteristic morphological features of senescence, including enlarged and flattened cellular morphology (Supplementary Fig. S1A). We then performed SA-β-gal staining, which showed that significant SA-β-gal staining positive cells were elevated in NRP2 knockdown cells (Figs. 2A, 2B). Next, we checked the levels of p21, a classic senescence marker, and found a significant upregulation (Fig. 2C). Knockdown of NRP2 also resulted in upregulation of γ-H2A.X, a DNA damage response marker frequently elevated in senescent cells (Fig. 2C). We then checked the features of senescence cells. Cell proliferation was evaluated using the EdU (5-ethynyl-2′-deoxyuridine) incorporation assay, which measures DNA synthesis during S phase. NRP2-deficient cells exhibited markedly reduced proliferative capacity, as indicated by decreased EdU incorporation (Figs. 2D, 2E). Morphologically, senescent cells displayed enlarged cell size detected by flow cytometry (Figs. 2F, 2G). NRP2 knockdown cells also showed nuclear abnormalities, including nuclear blebbing and nuclear envelope disruption with cytoplasmic chromatin fragments (Fig. 2H, yellow arrows; quantified in Fig. 2I), as well as redistribution of H3K9me3-marked heterochromatin from the nuclear periphery to the nuclear interior (Figs. 2J, 2K). Although senescence-associated secretory phenotype (SASP) is another characteristic of cellular senescence, we performed RNA sequencing using NRP2 knockdown cells. Analysis using the SenMayo gene set revealed enhanced SASP gene expression, including pro-inflammatory cytokines (IL-6, IL-18), chemokines (CXCL2, CXCL16), and extracellular matrix remodeling enzymes (MMP3) (Fig. 2L), further confirming the senescent phenotype. KEGG pathway enrichment analysis confirmed cellular senescence as one of the top pathways (Fig. 2M; Supplementary Fig. S1B), validating the results we found above.

Figure 2.

Neuropilin-2 deficiency induces cellular senescence in the RPE. (A, B) Representative SA-β-gal staining images and quantification in shCTRL, shNRP2-1 and shNRP2-2 cells. Cells stained blue are considered positive cells. The ratio of positive cells to all cells in the field of view (×20 objective) was quantified. Data are mean ± SEM. Each dot represents one field of view. n = 5 fields per group. One-way ANOVA with Tukey post hoc test. (C) Western blot analysis of NRP2, p21 and γ-H2A.X expression in whole cell lysates from shCTRL, shNRP2-1 and shNRP2-2 cells. (D, E) Representative immunofluorescence staining images and quantification of EdU (green) and Hoechst (blue) in shCTRL, shNRP2-1 and shNRP2-2 cells under tetracycline induction. All cells are under seven days of tetracycline induction before being harvested unless indicated otherwise. Cells stained green are considered EdU positive cells. The ratio of EdU-positive cells to cells stained with Hoechst in each field of view (×40 objective) was quantified. Data are mean ± SEM. Each dot represents one field of view. n = 20 fields per group. One-way ANOVA with Tukey post hoc test. (F) Relative cell size distribution in shCTRL and shNRP2-1 cells measured by flow cytometry. X axis is FSC-A, which reflects cell size. (G) Quantification of (F). The control group was set as one. Data are mean ± SEM. Each dot represents 1 sample. n = 11 samples per group. Welch's t test. (H, I) Representative immunofluorescence staining images and quantification of Lamin A/C (green) and Phalloidin (violet) in shCTRL and shNRP2-1 cells. Note that blebbed nuclei are pointed with yellow arrows in the field of view stained with Lamin A/C. DNA fragments are pointed with yellow arrows in the field of view stained with DAPI, indicating a disruption of nuclear membrane integrity. The ratio of blebbed nuclear to total nuclei in the field of view (×63 objective) was quantified. Data are mean ± SEM. Each dot represents one field of view. n = 6 fields per group. Student's t test. (J, K) Representative immunofluorescence staining images and fluorescence intensity quantification of H3K9me3 (green) in shCTRL and shNRP2-1 cells. Red arrow line shows the path across the nucleus from which fluorescence intensity data was collected for the H3K9me3 stain. X axis, the distance (µm) corresponding to the path marked by the red line. Y axis, H3K9me3 fluorescence intensity. Two graphs indicate a different distribution of H3K9me3 marks across the nucleus. (L) Heatmap of RNA-seq data using the SenMayo gene set. The heatmap displays upregulated SASP-related genes across shCTRL, shNRP2-1 and shNRP2-2 cells. Rows represent individual genes, and columns represent three different groups. Gene expression is color-coded, with red indicating high expression and blue indicating low expression. n = 3 samples per group. (M) Dot plot showing the top enriched KEGG pathways in the cellular process category for shNRP2-1 versus shCTRL. The y-axis represents enriched pathways, whereas the x-axis shows the proportion of genes in each pathway. Dot size indicates the count of genes enriched in each pathway, and dot color represents the statistical significance expressed as −log¹⁰ (adjusted P value), with darker purple indicating higher significance. Note that cellular senescence pathway (highlighted in red) shows significant enrichment, suggesting that NRP2 knockdown affects senescence-related gene expression. (N) Representative SA-β-gal staining images in eye tissue sections (10 µm in thickness) of one month after subretinal injection of either AAV-GFP or AAV-GFP-Cre in Nrp2^flox/flox mice. (O) Representative immunofluorescence staining images of ZO-1 (green) and p16 (red) in RPE-choroid complex flatmounts of 1 month after subretinal injection of either AAV-GFP or AAV-GFP-Cre in Nrp2^flox/flox mice. For all panels, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

To exclude potential shRNA off-target effects, we generated a stable NRP2 knockout clone using the CRISPR-Cas9 technology. RNA sequencing using NRP2 knockout cells again identified cellular senescence among the most significantly enriched pathways (Supplementary Figs. S2A, S2B). Indeed, NRP2-null clones phenocopied knockdown cells across all senescence parameters, including increased SA-β-gal activity, upregulation of senescence markers, reduced proliferation, enlarged cell size, nuclear abnormalities, and enhanced SASP genes expression (Supplementary Figs. S2C–K). To further validate these findings in a more physiologically relevant cell model, we validated findings in differentiated ARPE-19 cells cultured in MEM-Nic medium with NRP2 knockdown (Supplementary Figs. S7A, S7B), which recapitulated the senescence phenotype (Supplementary Figs. S7C–E), demonstrating that NRP2-mediated senescence regulation occurs independently of differentiation status. In summary, these results demonstrate that NRP2 loss directly induces senescence in vitro.

We then investigated whether NRP2 knockout induces cellular senescence in vivo. RPE tissues from AAV-VMD2-Cre-injected Nrp2^flox/flox mice recapitulated the senescence phenotype, showing increased SA-β-gal-positive cells and elevated p16 expression one month after knockout, before photoreceptor degeneration (Figs. 2N, 2O). Together, these findings identify NRP2 as a previously unrecognized suppressor of RPE senescence, whose loss triggers premature cellular aging both in cell culture and in vivo.

Neuropilin-2-Depleted RPE Cells Exhibit Mitochondrial Dysfunction

We next sought to elucidate the mechanisms underlying the senescence phenotype. Cellular senescence is closely associated with mitochondrial abnormalities, including altered morphology, increased ROS production, and bioenergetic failure.^6,14 We next examined mitochondrial structure and function in NRP2-depleted cells.

TOMM20 immunostaining revealed altered mitochondrial morphology in shNRP2 cells compared with controls (Fig. 3A). TEM further demonstrated markedly thin and elongated mitochondria in both shNRP2 and CRISPR knockout cells (Figs. 3B, 3C; Supplementary Figs. S3A–C), indicating senescence-associated mitochondrial remodeling.

Figure 3.

Neuropilin-2-depleted RPE cells exhibit mitochondrial dysfunction. (A) Representative immunofluorescence staining images of TOMM20 (red) in shCTRL and shNRP2-1 cells. (B, C) Representative TEM images of mitochondria in shCTRL and shNRP2-1 cells and quantification. Data are mean ± SEM. Each dot represents one mitochondrion. For shCTRL group, n = 92. For shNRP2-1 group, n = 75. Mann-Whitney U test. (D, E) Flow cytometry analysis and quantification of Mitotracker green staining in shCTRL, shNRP2-1 and shNRP2-2 cells. X axis represents Mitotracker green fluorescence intensity. The control group was set as 1. Data are mean ± SEM. Each dot represents one sample. n = 6 samples per group. One-way ANOVA with Tukey post hoc test. (F, G) Flow cytometry analysis and quantification of MitoSOX staining in shCTRL, shNRP2-1 and shNRP2-2 cells. X axis represents MitoSOX fluorescence intensity. The control group was set as 1. Data are mean ± SEM. Each dot represents 1 sample. n = 8 samples per group. Welch ANOVA with Dunnett T3 post hoc test. (H, I) Flow cytometry analysis and quantification of JC-1 staining in shCTRL, shNRP2-1 and shNRP2-2 cells. X axis represents FITC fluorescence (green) intensity, indicating JC-1 monomers. Y axis represents PE fluorescence (red) intensity, indicating JC-1 aggregates. The percentage of JC-1 monomers related to control was quantified, and the control group was set as one. Data are mean ± SEM. Each dot represents one sample. n = 6 samples per group. Welch ANOVA with Dunnett T3 post hoc test. For all panels, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Functionally, flow cytometric analyses revealed altered mitochondrial homeostasis in shNRP2 cells. MitoTracker Green, which stains mitochondria independently of membrane potential, showed increased mitochondrial mass (Figs. 3D, 3E). MitoSOX Red detected elevated mitochondrial ROS levels (Figs. 3F, 3G). The JC-1 assay, which measures mitochondrial membrane potential through the ratio of JC-1 monomers to JC-1 aggregates, revealed reduced mitochondrial membrane potential in shNRP2 cells (Figs. 3H, 3I). The observed pattern of increased mitochondrial mass coupled with impaired mitochondrial function is consistent with previously reported characteristics of senescent cells.¹⁵ Similar mitochondrial defects were observed in CRISPR knockout cells (Supplementary Figs. S3D–I), confirming the reproducibility of this phenotype.

Collectively, these data demonstrate that NRP2 deficiency compromises mitochondrial integrity, leading to structural remodeling, mitochondria ROS accumulation and membrane depolarization. These mitochondrial defects likely serve as a key mechanistic link between NRP2 loss and RPE senescence.

NAD⁺ Depletion and Sirtuin Inactivation Link Mitochondrial Dysfunction to RPE Cellular Senescence

Mitochondrial dysfunction is tightly linked to redox balance and disruption of NAD⁺ metabolism, and declining NAD⁺ levels are a hallmark of cellular senescence that impairs sirtuin-mediated protein deacetylation.^16–18 To determine whether NAD⁺ dysregulation accompanied the mitochondrial defects in NRP2-deficient cells, we measured mitochondrial NAD⁺/NADH ratios and found a significant reduction in shNRP2 and CRISPR knockout RPE cells (Fig. 4A; Supplementary Fig. S4A).

Figure 4.

NAD⁺ depletion and sirtuin inactivation link mitochondrial dysfunction to cellular senescence. (A) Quantification of NAD⁺/NADH ratio in the mitochondria of shCTRL, shNRP2-1 and shNRP2-2 cells. The control group was set as one. Data are mean ± SEM. Each dot represents one sample. n = 6 samples per group. One-way ANOVA with Tukey post hoc test. (B–D) QPCR analysis on the indicated genes in shCTRL, shNRP2-1 and shNRP2-2 cells. The control group was set as 1. Data are mean ± SEM. Each dot represents one sample. n = 4 samples per group. (B–D) One-way ANOVA with Tukey post hoc test. (E–G) Western blot analysis of NRP2, QPRT, phosphorylated and total SIRT1, SIRT3, acetyl-histone H3 and acetyl-histone H2B expression in whole cell lysates from shCTRL, shNRP2-1 and shNRP2-2 cells. (H) Western blot analysis of p21 expression in whole cell lysates from shNRP2-1 cells with or without K₃[Fe(CN)₆] treatment. (I, J) Flow cytometry analysis and quantification of MitoSOX staining in shCTRL, shNRP2-1 and shNRP2-2 cells with or without K₃[Fe(CN)₆] treatment. X axis represents MitoSOX fluorescence intensity. The control group was set as one. Data are mean ± SEM. Each dot represents one sample. n = 6 samples per group. Welch ANOVA with Dunnett T3 post hoc test. (K, L) Representative immunofluorescence staining images and quantification of EdU (green) and Hoechst (blue) in shCTRL and shNRP2-1 cells. The ratio of EdU positive cells to cells stained with Hoechst in each field of view (×40 objective) was quantified. Data are mean ± SEM. Each dot represents 1 field of view. n = 10 fields per group. Welch ANOVA with Dunnett T3 post hoc test. For all panels, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

RT-qPCR analysis revealed broad transcriptional suppression of NAD⁺ biosynthetic genes involved in all three NAD⁺ biosynthesis pathways: the kynurenine, Preiss–Handler, and salvage pathway (Figs. 4B–D), with the key enzyme QPRT showing the most pronounced decrease. Western blot analysis validated the decrease of QPRT protein levels in both shNRP2 and CRISPR knockout cells (Fig. 4E; Supplementary Fig. S4B). This coordinate suppression across multiple pathways suggests a global disruption of NAD⁺ homeostasis rather than a pathway-specific defect.

Given that NAD⁺ serves as an essential cofactor for sirtuins, a family of NAD⁺-dependent deacetylases that regulate aging and stress responses,^18,19 we next examined whether NAD⁺ depletion affects sirtuin signaling in NRP2-deficient cells. Western blot analysis showed a marked reduction in phosphorylated SIRT1, SIRT1, and SIRT3 levels in NRP2-deficient cells (Fig. 4F) and concomitant elevation of global protein acetylation (Fig. 4G; Supplementary Fig. S4C). This pattern indicates compromised sirtuin-mediated protein deacetylation, a key mechanism through which NAD⁺ depletion can promote cellular senescence.

To determine whether NAD⁺ depletion acts as a causal driver of senescence rather than a secondary effect, we performed rescue experiments using K₃[Fe(CN)₆], a cell-impermeable oxidant that increases NAD⁺/NADH ratios by promoting NADH oxidation without supplying NAD⁺ precursors.¹⁷ Strikingly, K₃[Fe(CN)₆] treatment effectively restored mitochondrial NAD⁺/NADH ratios in NRP2-depleted cells compared to control cells (Supplementary Fig. S4D), demonstrating the efficacy of this metabolic intervention. K₃[Fe(CN)₆] treatment also significantly reduced p21 levels (Fig. 4H; Supplementary Fig. S4E), a critical senescence marker dysregulated in NRP2-deficient senescent cells. Furthermore, K₃[Fe(CN)₆] attenuated mitochondrial ROS accumulation (Figs. 4I, 4J), a critical driver of oxidative stress-induced senescence, and restored the proliferative arrest in NRP2-depleted cells (Figs. 4K, 4L). Together, these findings demonstrate that NRP2 deficiency disrupted mitochondrial function and impaired NAD⁺ metabolism, leading to reduced sirtuin activity and the induction of RPE cellular senescence. Restoration of NAD⁺/NADH ratio effectively reverses these phenotypes, establishing NAD⁺ depletion as a key mediator of NRP2-regulated mitochondrial senescence.

Mitochondrial-Targeted Therapy With Echinacoside Prevents NRP2 Deficiency-Associated Retinal Degeneration

The findings above demonstrate that mitochondrial dysfunction and NAD⁺ depletion are central to NRP2 deficiency-induced senescence. To explore whether pharmacological intervention targeting mitochondrial function could rescue RPE degeneration, we tested ECH, a natural phenylethanoid glycoside previously reported to exert mitochondrial protective and anti-senescent effects.^20,21

In NRP2-depleted RPE cells, ECH markedly reduced mitochondrial ROS accumulation (Figs. 5⁶A, 5B), restored mitochondrial membrane potential (Figs. 5C, 5D), and enhanced cell proliferation (Figs. 5E, 5F), indicating recovery of mitochondrial function and suppression of cellular senescence.

Figure 5.

Mitochondrial-targeted therapy with ECH prevents NRP2 deficiency-associated retinal degeneration. (A, B) Flow cytometry analysis and quantification of MitoSOX staining in shCTRL, shNRP2-1 and shNRP2-2 cells with or without ECH treatment. X axis represents MitoSOX fluorescence intensity. The control group was set as one. Data are mean ± SEM. Each dot represents one sample. n = 6 samples per group. Welch ANOVA with Dunnett T3 post hoc test. (C, D) Flow cytometry analysis and quantification of JC-1 staining in shCTRL, shNRP2-1 and shNRP2-2 cells with or without ECH treatment. X axis represents FITC fluorescence (green) intensity, indicating JC-1 monomers. Y axis represents PE fluorescence (red) intensity, indicating JC-1 aggregates. The ratio of JC-1 monomers to all cells were quantified, and the control group was set as 1. Data are mean ± SEM. Each dot represents one sample. n = 6 samples per group. Kruskal-Wallis test. (E, F) Representative immunofluorescence staining images and quantification of EdU (green) and Hoechst (blue) in shCTRL and shNRP2-1 cells with or without ECH treatment. The ratio of EdU positive cells to cells stained with Hoechst in each field of view (40 × objective) was quantified. Data are mean ± SEM. Each dot represents 1 field of view. n =10 fields per group. Kruskal-Wallis test. (G) Timeline of AAV subretinal injection and the four-week ECH/PBS gavage for Nrp2^flox/flox mice. Arrows indicate AAV subretinal injection at week 8, followed by four-week ECH/PBS gavage. Eyeballs were collected at week 12 for further analysis. (H) Representative fundus photography and OCT images of gavage after AAV subretinal injection of Nrp2^flox/flox mice. The white arrow line represents the OCT scanning plane. Note that aggregates are pointed out with yellow arrows in the PBS gavage group, but nearly no aggregates were found in the ECH gavage group. (I) Quantification of the ONL layer from OCT images (H). Each dot represents a measurement point. For Cre + PBS group, n = 10 mice per group. For Cre + ECH group, n = 12 mice per group. Data are mean ± SEM. Mann-Whitney U test. (J, K) Quantification of scotopic ERG a-wave and b-wave amplitudes in Nrp2^flox/flox mice after subretinal AAV injection and subsequent gavage treatment. Data are mean ± SEM. n = 6 mice per group. Mann-Whitney U test. (L) Representative immunofluorescence staining images of ZO-1 (red) in RPE-choroid complex flatmounts in Nrp2^flox/flox mice after subretinal AAV injection and subsequent gavage treatment. (M, N) Quantification of RPE cell density and RPE cell size from (L). Data are mean ± SEM. Images were captured in the central regions of the RPE-choroid complex. Each dot represents one field of view, and three fields of view were taken per mice. For Cre + PBS group, n = 18. For Cre + ECH group, n = 24. Mann-Whitney U test. (O) Representative immunofluorescence staining images of ZO-1 (green) and p16 (red) in RPE-choroid complex flatmounts of Nrp2^flox/flox mice after subretinal AAV injection and subsequent gavage treatment. For all panels, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 6.

Schematic representation of the NRP2-mitochondria-NAD⁺ axis in RPE senescence and the protective effect of Echinacoside against retinal degeneration. NRP2 deficiency disrupts mitochondrial homeostasis by increasing ROS levels and reducing membrane potential, which in turn lowers the NAD⁺/NADH ratio and drives cellular senescence. Additionally, echinacoside effectively targets the NRP2-mitochondria-NAD⁺ axis, thereby preventing NRP2 deficiency-driven cellular senescence and preventing retinal degeneration, offering a viable mitochondrial-targeted senotherapeutic approach.

We next evaluated its therapeutic potential in vivo. RPE-specific Nrp2 knockout mice were established via subretinal AAV injection and treated systemically with ECH (Fig. 5G). Fundus imaging revealed that ECH treatment decreased retinal pale spots and pigment aggregation (Fig. 5H, upper panel), indicating improved RPE integrity. OCT analysis showed that ECH mitigated ONL thinning and retinal degeneration (Fig. 5H, lower panel; Fig. 5I), suggesting structural preservation of the photoreceptor layer. Functional assessment by scotopic ERG demonstrated that ECH restored both a-wave and b-wave amplitudes (Figs. 5J, 5K), suggesting improved visual function. Histologic and immunofluorescent analyses further revealed normalization of RPE morphology, reduced cellular enlargement (Figs. 5L–N), and decreased p16 expression (Fig. 5O), demonstrating attenuation of RPE senescence in vivo.

Discussion

Cellular senescence is increasingly recognized as a key driver of RPE dysfunction and retinal degenerative diseases, yet the upstream molecular regulators remain poorly defined. In this study, we identified NRP2 as a critical regulator of RPE mitochondrial function and NAD⁺ homeostasis. Loss of NRP2 disrupted these processes, leading to mitochondrial dysfunction, premature RPE senescence, and progressive retinal degeneration. Importantly, restoring mitochondrial function with a natural compound echinacoside prevents RPE senescence and preserves retinal structure and visual function. These findings uncover a previously unrecognized NRP2-mitochondria-NAD⁺ axis, highlighting how extracellular signaling coordinates intracellular metabolic resilience and providing proof-of-concept for mitochondrial-targeted therapeutic strategies in retinal degeneration (Fig. 6). We employed ARPE-19 cells for mechanistic studies because of their practical advantages, including unlimited availability and ease of genetic manipulation. Although primary human RPE or iPSC-derived RPE would provide greater physiological relevance, donor variability and technical complexity were beyond the scope of this study. Nevertheless, we validated key findings in differentiated ARPE-19 cells, which showed consistent results. Critically, findings were confirmed in an in vivo conditional knockout mouse model, demonstrating NRP2-mediated senescence regulation in native RPE.

Mechanistically, NRP2 depletion disrupted mitochondrial architecture and bioenergetic function, accompanied by marked NAD⁺ depletion and downregulation of key NAD⁺ biosynthetic enzymes. Restoration of NAD⁺/NADH balance using K₃[Fe(CN)₆] or recovery of mitochondrial function with echinacoside reversed senescence phenotypes, establishing the causal role of the NRP2–mitochondria–NAD⁺ axis in RPE aging. The decreased NAD⁺/NADH ratio is a well-established hallmark of mitochondrial dysfunction and cellular senescence across multiple cell types.^15,22 In dysfunctional mitochondria, impaired electron transport chain activity leads to NADH accumulation and a collapsed NAD⁺/NADH ratio.²³ NAD⁺ serves as a central metabolic cofactor that underpins mitochondrial oxidative phosphorylation, redox homeostasis, and DNA repair, and its loss further exacerbates mitochondrial dysfunction.²⁴ In RPE cells, which rely heavily on mitochondrial metabolism to support outer-segment phagocytosis and nutrient recycling, NAD⁺ loss critically compromises ATP production and elevates ROS generation, thereby promoting RPE senescence.²⁵ Moreover, NAD⁺ depletion reduces the activity of sirtuins—particularly SIRT1 and SIRT3—which are essential for maintaining mitochondrial biogenesis, mitophagy, and antioxidant defense.^26–28 Thus NRP2 deficiency-induced NAD⁺ depletion and sirtuin suppression likely establish a detrimental feed-forward cycle of mitochondrial damage and cellular senescence.

Although the precise molecular pathway linking NRP2 to mitochondrial NAD⁺ metabolism requires further investigation, we propose that NRP2 may act through its well-established role as a co-receptor for multiple receptor tyrosine kinases, including VEGFR, PDGFR, and c-Met, thereby modulating downstream signaling pathways such as PI3K/AKT.^29,30 Given that AKT regulates key transcriptional programs governing mitochondrial biogenesis (e.g., PGC-1α, FOXO),³¹ we speculate that NRP2 loss during aging may disrupt these pathways, contributing to impaired mitochondrial function and NAD⁺ depletion in RPE. Future studies are warranted to elucidate the precise signaling mechanisms connecting NRP2 to metabolic homeostasis in RPE.

From a translational perspective, our findings demonstrate that mitochondrial-targeted pharmacological intervention can prevent RPE senescence. Echinacoside, a natural phenylethanoid glycoside with reported antioxidant and SIRT1-activating properties, restored mitochondrial function and preserved retinal structure and function in NRP2-deficient models. Importantly, systemic administration via oral delivery achieved therapeutic efficacy, suggesting practical feasibility for chronic treatment of retinal degeneration. Unlike senolytics or senomorphics that remove or suppress senescent cells,^32,33 ECH prevents senescence onset through metabolic restoration, offering a complementary strategy to preserve tissue function during aging.

Conclusions

Our findings identify NRP2 as an upstream regulator of RPE metabolic integrity and define the NRP2–mitochondria–NAD⁺ axis as a novel link between mitochondrial dysfunction and cellular senescence. Pharmacological intervention with echinacoside effectively interrupts this axis, preventing senescence-driven retinal degeneration and offering a viable mitochondrial-targeted senotherapeutic approach. Future work should evaluate the long-term efficacy and safety of echinacoside in aging models. Given the high degree of NRP2 conservation between mice and humans, and the validation of safety and tolerability of echinacoside in Phase I clinical trials,³⁴ our mechanistic and therapeutic findings in mouse models hold translational promise for humans. Future validation in higher-order animal models such as non-human primates would further support clinical translation. Collectively, this study establishes a NRP2-NAD⁺ dependent mechanistic and therapeutic framework for targeting mitochondrial metabolism to preserve retinal health.