AIF-independent parthanatos in the pathogenesis of dry age-related macular degeneration

Cell death of retinal pigment epithelium (RPE) is characterized as an essential late-stage phenomenon of dry age-related macular degeneration (AMD). The aim of this study was to elucidate the molecular mechanism underlying RPE cell death after exposure to oxidative stress, which occurs often because of the anatomical location of RPE cells. ARPE-19, an established RPE cell line, exhibited necrotic features involving poly (ADP-ribose) polymerase-1 (PARP-1) activation in response to hydrogen peroxide (H2O2). ARPE-19 cells were resistant to H2O2 when PARP-1 was depleted using siRNA or inhibited by a pharmacological inhibitor of PARP-1, olaparib. Our data suggest a causal relationship between PARP-1 activation and ARPE-19 cell death in response to H2O2. Next, we investigated downstream molecular events in PARP-1 activation. Increased mitochondrial depolarization, mitochondrial fission and alterations of the cellular energy dynamics with reduced NAD+ and ATP were observed in H2O2-treated ARPE-19 cells. H2O2-triggered mitochondrial dysfunction was inhibited by olaparib. Nevertheless, translocation of apoptosis-inducing factor (AIF), a biochemical signature for PARP-1-dependent cell death (parthanatos), was not observed in our study. Moreover, the depletion of AIF did not affect the amplitude of cell death, demonstrating the lack of a role for AIF in the death of ARPE-19 cells in response to H2O2. This feature distinguishes the type of death observed in this study from canonical parthanatos. Next, we examined the in vivo role of PARP-1 in a dry AMD animal model system. Histological analysis of the outer nuclear layer in the mouse retina revealed protection against sodium iodate (SI) following treatment with olaparib. Moreover, retina fundus and electroretinograms also confirmed such a protective effect in the SI-treated rabbit. Collectively, we report that AIF-independent PARP-1-dependent necrosis constitutes a major mechanism of RPE cell death leading to retinal degeneration in dry AMD.

Age-related macular degeneration (AMD) is the most common cause of blindness among the elderly. 1,2 AMD is classified into wet and dry forms; the dry form is more common than the wet form. Wet AMD is characterized by the generation of abnormal angiogenesis underneath the retina and leads to rapid vision loss. In contrast, retinal cells die progressively, displaying geographic atrophy (GA), in dry AMD. This gradual degeneration of retinal cells in GA patients also results in vision loss. 3,4 Fortunately, antiangiogenic therapeutics effectively delay the progression of wet AMD. 5,6 However, FDA-approved treatments for dry AMD are not available, although a few are now in clinical trials. Therefore, development of neuroprotective agents to maintain the remaining vision has been suggested as a future therapy for dry AMD. 7 Retinal pigment epithelium (RPE), a monolayer of pigmented cells, is located between photoreceptor cells and Bruch's membrane and maintains retinal homeostasis via the transport of nutrients and waste, thereby protecting photoreceptor cells. 8 The pathogenesis of dry AMD involves oxidative stress, mitochondrial dysfunction and inflammation. 9-13 RPE cells are prone to exposure to high-energy light and rich polyunsaturated fatty acids, which are readily oxidized through photonic activation. Due to their anatomical localization and metabolic function, RPE cells are continuously exposed to chronic and cumulative oxidative stress and are most severely damaged in progressive dry AMD. 14 RPE degeneration impairs retinal protective measures for the photoreceptor cells and results in their progressive death. To study the death mechanism of RPE cells, the human-derived RPE cell line, APRE-19, is often used as a cellular model upon oxidative stress [15][16][17][18] because these cells display properties that are commonly observed in RPE cells, such as morphological polarization and expression of the RPE-specific markers cellular retinaldehyde-binding protein and RPE65. 19 The sodium iodate (SI) model is used to further understand the mechanism of RPE loss in dry AMD pathogenesis because SI is an oxidizing compound with specific toxicity for RPE and leads to alterations in RPE functions. [20][21][22][23] SI-induced retinal degeneration has been reported in various animal species, including sheep, rabbit and mice, with varying dose and administration routes. 20,24,25 Moreover, SI damages the RPE through several mechanisms, including crossreactivity with melanin, which converts glycine into toxic glucoxylate, inhibition of energy production enzymes and ROS accumulation. [26][27][28] Therefore, we used SI-injected mice and rabbits to validate the in vivo role of PARP-1 in the pathogenesis of dry AMD.
Apoptosis and necrosis seem to be activated flexibly depending on the cell types and cellular context in the retina. 29 When apoptosis is inhibited in photoreceptor cells, regulated necrotic death predominates, as if compensating for the absence of apoptosis. In this case, the sum of cell death remains relatively static despite the altered ratio of regulated necrosis to apoptosis. This compensation provides an explanation for therapeutic failure with single blockage of   apoptosis to prevent retinal cell death. Therefore, necrotic death in the retinal cells has been studied extensively, and a combination therapy of apoptotic and necrotic inhibitors seems to be promising for the protection of retinal cells. Poly (ADP-ribose) polymerases (PARPs) constitute a large family of enzymes that catalyze the transfer of ADP-ribose units onto target proteins. In humans, 17 members of the PARP family have been identified and share a conserved catalytic domain. 30 PARP-1 is thought to have a critical role in cellular physiology because the majority (490%) of PAR polymer synthesis derives from PARP-1. 30-32 PARP-2 is the closest homolog to PARP-1, displaying 69% similarity in the catalytic domain, and PARP-1 and -2 are responsible for most of the PAR polymer synthesis. 33 PARP-1 has multiple cellular roles, acting both in cell survival and in cell death pathways in cell type-and stimulus-dependent manners. 34,35 Activated PARP-1 under oxidative stress consumes NAD+ and depletes cellular ATP, eventually leading to cellular energy collapse. 36 Moreover, PARP-1 activation results in the translocation of apoptosis-inducing factor (AIF) from mitochondria to the nucleus, fragmenting DNA. 37,38 Accumulation of such biochemical events completes PARP-1-mediated necrotic death of cells, parthanatos, which has been implicated in various age-related neurodegenerative diseases. 39,40 This observation led us to examine whether parthanatos is also involved in the pathogenesis of dry AMD, a prevalent senile eye disease.
Here, we demonstrate that PARP-1-mediated necrosis constitutes a substantial portion of the death of ARPE-19 cells in response to hydrogen peroxide (H 2 O 2 ) insult. Mitochondrial dysfunction was observed downstream of PARP-1 activation. However, AIF was irrelevant in this death process, indicating that this type of cell death is distinct from canonical parthanatos in which AIF translocation follows PARP-1 activation. Moreover, we also validated the in vivo role of PARP-1 in retinal degeneration using Si-injected mice and rabbits. Collectively, this study reports the presence of an AIF-independent parthanatos pathway in the pathogenesis of dry AMD.  Figure 1). In contrast, ARPE-19 cells died in dose-and timedependent manners at pathophysiological concentration ranges reported in diverse diseases 41,42 (Figures 1a and b). Because ARPE-19 cells are proximal to the choroid, we  examined cell death following exposure to H 2 O 2 in other cells  distal to the choroid, retinal ganglion cells (RGC-5). RGC-5  cells were more sensitive to H 2 O 2 compared with ARPE-19 cells (Supplementary Figure 2), suggesting that distance from the choroid might correlate with retinal cellular sensitivities to oxygen. Next, we performed flow cytometry to elucidate the predominant form of death in ARPE-19 cells following exposure to H 2 O 2 . Propidium iodide (PI) singlepositive cells progressed to PI/Annexin V double-positive cells in a time-dependent manner ( Figure 1c). Conversely, the population of Annexin V single-positive cells were not altered throughout the assessed time period. Moreover, caspase-3 was not activated by up to 1 mM H 2 O 2 ( Figure 1d). This finding implies that necrotic death occurred under the above conditions. In another set of experiment, Annexin V-positive cells increased with caspase-3 activation following the administration of staurosporine (STS), demonstrating that the apoptotic machinery in ARPE-19 cells was intact (Figures 1c and d). Furthermore, z-VAD treatment did not protect ARPE-19 cells from the H 2 O 2 insult, indicating that apoptotic death did not occur ( Figure 1e). Necrotic death in ARPE-19 cells was further confirmed using confocal microscopy ( Figure 1f) and by the release of high-mobility group box 1, a marker of necrosis, into the medium (Figure 1g) Figure 3e).

Results
Thereafter, we examined whether parthanatos, another type of regulated necrosis, participates in H 2 O 2 -induced death of APRE-19 cells. We found that H 2 O 2 treatment induced the synthesis of cellular poly ADP-ribose (PAR) polymers  (Figures 2g and h). Furthermore, we analyzed the subtype of PARPs that was activated by H 2 O 2 in ARPE-19 cells using various PARP inhibitors because PAR polymers can be synthesized by various PARP family members. The reagents 3AB and DPQ, inhibitors of PARP-1 and -2, prevented H 2 O 2 -induced ARPE-19 cell death, but UPF-1069 (a selective PARP-2 inhibitor) and XAV-939 (a selective PARP-5 inhibitor) did not ( Figure 2i). Collectively, these data indicate that ARPE-19 cell death in response to H 2 O 2 is mediated mainly by PARP-1. Next, we examined whether PARP-1-mediated death occurs in ARPE-19 cells in response to other oxidative stresses. In ARPE-19 cells, olaparib blocked 1-methyl-3-nitro-1-nitrosoguanidine (MNNG)-induced death but not tert-Butyl hydroperoxide (t-BHP)-and rotenone-induced death (Supplementary Figure 4a). These results indicate that MNNG triggers PARP-1-mediated death, but t-BHP and rotenone do not. PARP-1 activation by MNNG was further confirmed using western blot analysis of PAR (Supplementary Figure 4b).
Taken together, our data show that PARP-1 is not a common mediator of APRE-19 cell death in response to diverse oxidative stresses.
AIF is dispensable for H 2 O 2 -induced necrotic death through PARP-1 activation in ARPE-19 cells. We next examined the translocation of AIF into the nucleus, a biochemical signature of parthanatos, in H 2 O 2 -insulted ARPE-19 cells. AIF remained in the mitochondria upon H 2 O 2 insult by fluorescence intensity profile analysis (Figures 3a and b). Subcellular fractionation analysis using differential ultracentrifugation to determine the relative distribution of AIF also revealed no changes in the localization of AIF upon H 2 O 2 insult ( Figure 3c). Moreover, we also did not observe the genomic DNA fragmentation in H 2 O 2 -exposed ARPE-19 cells, further confirming lack of AIF translocation to the nucleus (Supplementary Figure 5). Nevertheless, mitochondrial fission was observed, revealing a punctate morphology in ARPE-19 cells ( Figure 3a). Next, we investigated whether AIF translocates into the nucleus in another type of retinal cells, RGC-5, because ARPE-19 cells and RGC-5 cells showed differential sensitivities to H 2 O 2 (Supplementary Figure 2). We also examined AIF translocation in another cell type, SH-SY5Y and mouse embryonic fibroblast (MEF), upon H 2 O 2 . AIF was released from the mitochondria and translocated into the nucleus in RGC-5, SH-SY5Y and MEF cells following the H 2 O 2 insult and olaparib blocked AIF translocation (Supplementary Figure 6) Figure 7). Our findings indicate the presence of a novel necrotic pathway that is distinctive from canonical parthanatos in H 2 O 2 -exposed ARPE-19 cells.
PARP-1 activation triggers mitochondrial dysfunction and cellular energy collapse in response to H 2 O 2 insult. We examined whether PARP-1 activation damages mitochondria in H 2 O 2 -insulted ARPE-19 cells because mitochondrial defects are observed in the RPE of AMD eyes. 48  in ARPE-19 cells following exposure to H 2 O 2 . Therefore, our data clearly demonstrate that H 2 O 2 damages mitochondria through PARP-1 activation in ARPE-19 cells. Next, we measured the cellular NAD+ and ATP levels following exposure to H 2 O 2 . The levels of NAD+ and ATP significantly decreased in response to H 2 O 2 , and this decrease was prevented by the pharmacological inhibition of PARP-1 in ARPE-19 cells (Figures 4d and e). These results indicate that activated PARP-1 by H 2 O 2 triggers cellular energy depletion. Collectively, our findings suggest that PARP-1 activation upon an H 2 O 2 insult provokes mitochondrial dysfunction and leads to energy failure with declines in cellular NAD+ and ATP in ARPE-19 cells.
PARP-1 participates in retinal degeneration in a dry AMD mouse model. To validate the role of PARP-1 in the pathogenesis of AMD in vivo, SI-injected mice were used as an animal model. 27,49 The administrative procedure is  Figure 5a. The synthesis of PAR polymers was triggered in mice upon SI injection, reaching a maximum at day 1 ( Supplementary Figures 10a and b), and this synthesis was blocked by olaparib (Figures 5b and c). Next, we performed a histological analysis using hematoxylin and eosin staining to investigate the effect of PARP-1 on retinal PARP-1 inhibition preserves the physiological function of the retina in rabbits following SI insult. To validate the in vivo role of PARP-1 in the physiological function of the retina, we performed a fundoscopic examination to assess the geographic atrophy in live animals. Fundoscopy showing the optic nerve regions was used in SI-injected rabbits because the large size of rabbit eyes facilitates this procedure. The administrative procedure is schematized in Figure 6a, and a representative fundus image is shown in Figure 6b. The retina fundus of SI-injected rabbits appeared to be brighter than that of the control rabbits, representing RPE loss. The brightness of the retina fundus was moderated by olaparib, implying that PARP-1 mediates the RPE loss in SI-injected rabbits (Figure 6b). Next, we evaluated the function of the retina using electroretinography in SIinjected rabbits to confirm the effect of PARP-1 in the pathogenesis of AMD. The amplitude of the A-wave was drastically reduced, demonstrating functional damage of the photoreceptors of SI-injected rabbits (Figures 6c and d). 50,51 Similarly, the amplitude of the B-wave also decreased, revealing dysfunction in the bipolar cells of SI-injected rabbits. 52,53 Olaparib conserved the amplitude of A-and B-waves similar to the control. The parameters of the electroretinogram (ERG) potential are summarized in Table 1. Collectively, these results show that PARP-1 impairs visual function in the pathogenesis of dry AMD.

Discussion
This study shows that the PARP-1-mediated necrotic pathway is distinct from that of the canonical parthanatos, serving as a novel mechanism of RPE loss in the pathogenesis of dry AMD: The mechanism underlying oxidative stress-induced RPE loss in AMD pathogenesis remains unclear. Consistent with a previous study reporting low expression levels of the DNA fragmentation factor 45/40, 16 we did not observe features of apoptosis in H 2 O 2 -treated ARPE-19 cells. Endogenous caspase-8 expression was also decreased in ARPE-19 cells compared with that in other ocular cells. 16 Such a downregulation of apoptotic components might serve as a survival strategy to compensate for the rapid turnover of RPE cells. 54 An attenuated cellular apoptotic potential would shift the cellular death pathway toward necrosis, in which substrates of caspases, including RIPK1 and PARP-1, have key roles, providing one explanation for the predominance of necrosis in ARPE-19 cells in response to oxidative stress.
It is noteworthy that PARP-1-mediated necrosis is activated, whereas RIPK1-mediated necroptosis is quiescent between the two regulated necrotic pathways. Inactivation of the necroptotic machinery in H 2 O 2 -exposed ARPE-19 cells seems to result from the lack of RIPK3 expression. Failure of necrosome formation due to the RIPK3 deficiency would   preclude the completion of necroptosis, irrespective of RIPK1 expression. Such a restriction would be advantageous for longevity in post-mitotic RPE cells with limited regeneration potential. [55][56][57] In contrast, PARP-1 participates in the cellular repair process and in necrosis. Therefore, retention of PARP-1 activity would serve dual measures for repair and cell death.
Our study has shown that genetic depletion of PARP-1 removed most PAR polymers in H 2 O 2 -treated ARPE-19 cells. Moreover, the pharmacological inhibition of PARP-1 sufficiently protected ARPE-19 cells against H 2 O 2 . Thus, PARP-1 has a major role in PAR polymer synthesis under oxidative stress in ARPE-19 cells. PARP-1 is highly conserved, especially in the contiguous 50-amino-acid sequence, the signature motif of PARP, in the catalytic domain, which displays 100% conservation in vertebrates and the most abundantly expressed isoforms among the PARP family members, supporting the importance of PARP-1. 30,31,32,58,59 Therefore, our data support the use of PARP-1 as a target for the treatment of dry AMD.
RPEs located proximal to the choroidal vessels are frequently exposed to oxygen and thereby might have adapted to cope with oxidative stresses. The lack of AIF translocation into the nucleus upon exposure to H 2 O 2 in ARPE-19 cells would interrupt one death pathway: apoptosis. Therefore, the blockage of AIF translocation would protect against nuclear damage. In contrast, RGCs distal to the choroidal vessel would not have to be rigorous to resolve alterations caused by oxidative damage. This observation seems to provide a physiological explanation for the translocation of AIF in RGC-5 cells in response to oxidative damage. Therefore, we speculate that differential sensitivities to H 2 O 2 among retinal cells might derive from their anatomical positions.
Nuclear translocation of AIF following PARP-1 activation is the signature of parthanatos, and AIF is required as an executioner of parthanatos. 37,[60][61][62][63] The lack of AIF translocation with PARP-1 activation shown in our study has also been found in human renal proximal tubule epithelial cells. 64 Furthermore, the lack of involvement of AIF in executing PARP-1-mediated death has been reported in MEF cells following exposure to MNNG. 65 Considering the dual role of AIF, that is, mitochondrial protection and DNA fragmentation in the nucleus, AIF seems to function flexibly, depending on the context such as the cell type or the type of stimulus.
Several genes are involved in the pathogenesis of dry AMD. Moreover, clinical analysis of dry AMD patients has revealed a dysregulation of complement-associated genes. 66 Therefore, it will be important to investigate the clinical relevance of PARP-1 in patients with dry AMD. Abundant PARP-1 inhibitors that have already been developed as therapeutics for other diseases would serve as repurposed drug candidates for the treatment of dry AMD.  Gene silencing with siRNA. Small interfering RNA (siRNA) oligonucleotides were purchased from Bioneer (Daejeon, Korea) with sequences targeting RIPK1 #2 (5′-CACACAGUCUCAGAUUGAU-3′), AIF #1 (5′-GCAAGUUACUUAUCAAGCU-3′) and PARP-1 #3 (5′-GGAGGGUCUGAUGAUAGCA-3′). ARPE-19 cells were transfected with 200 nM of the indicated siRNA or scRNA using Lipofectamine RNAiMAX reagent (Thermo Fisher Scientific) according to the manufacturer's instructions. The effects of siRNA on the indicated protein levels were examined by western blot analysis.

Animal model
Mice: C57BL/6 mice (male, 9 weeks old, weight range: 23-26 g) were purchased from Central Lab Animal (Seoul, Korea). All mice were maintained in the animal facility of Chungnam National University (Daejeon, Korea) and acclimatized to a light schedule of alternating 12-h periods of light and dark with free access to food and water for at least 1 week before the experiment and experimental duration. All animal studies were conducted in accordance with the institutional guidelines for the care and use of laboratory animals. All mice were divided into 4 groups; control (n = 6), control-olaparib (n = 6, olaparib 15 mg/kg, i.p), SI-vehicle (n = 6, SI 30 mg/ kg, i.p), SI-olaparib (n = 6, SI 30 mg/kg, i.p and olaparib 15 mg/kg, i.p). The administrative procedure is shown in Figure 5a.
Rabbits: Chinchilla rabbits (male, 3 months, 3 kg) were purchased from Yonam University (Chonan, Korea). All rabbits were maintained in the animal facility of the Catholic University of Korea (Seoul, Korea) and were acclimatized to a light schedule of alternating 12-h periods of light and dark with free access to food and water for at least 1 week before the experiment and experimental duration. All rabbits were divided into 4 groups: control (n = 3); control-olaparib (n = 3, olaparib 15 mg/kg, i.p); SI-vehicle (n = 3, SI 15 mg/kg, i.v); and SI-olaparib (n = 3, SI 15 mg/kg, i.v and olaparib 15 mg/kg, i.p). The administrative procedure is shown in Figure 6a.
Protein extraction from the retina. Retinas from enucleated eyes were homogenized in lysis buffer as stated above. Homogenates were centrifuged at 12 000 × g for 10 min at 4°C. Total protein concentrations were determined by the Bradford method. The same volume of proteins was subjected to western blot analysis.
Hematoxylin and eosin staining. Enucleated eyes were prefixed in 4% paraformaldehyde in PBS at room temperature for 20 min, and the lens was extirpated. Next, the samples were incubated in 4% paraformaldehyde for 12 h and embedded using routine procedures. After embedding, retinal cross sections were prepared with a thickness of 5 μm. The slices were dewaxed, stained with hematoxylin for 5 min, and restained with eosin for 5 min. The samples were observed under an optical microscope (Leica Microsystems, Wetzlar, Germany) and imaged with a slide scanner (Motic Electronic, Xiamen, China).