Oxidative stress induces Z-DNA-binding protein 1–dependent activation of microglia via mtDNA released from retinal pigment epithelial cells

Oxidative stress, inflammation, and aberrant activation of microglia in the retina are commonly observed in ocular pathologies. In glaucoma or age-related macular degeneration, the chronic activation of microglia affects retinal ganglion cells and photoreceptors, respectively, contributing to gradual vision loss. However, the molecular mechanisms that cause activation of microglia in the retina are not fully understood. Here we show that exposure of retinal pigment epithelial (RPE) cells to chronic low-level oxidative stress induces mitochondrial DNA (mtDNA)-specific damage, and the subsequent translocation of damaged mtDNA to the cytoplasm results in the binding and activation of intracellular DNA receptor Z-DNA-binding protein 1 (ZBP1). Activation of the mtDNA/ZBP1 pathway triggers the expression of proinflammatory markers in RPE cells. In addition, we show that the enhanced release of extracellular vesicles (EVs) containing fragments of mtDNA derived from the apical site of RPE cells induces a proinflammatory phenotype of microglia via activation of ZBP1 signaling. Collectively, our report establishes oxidatively damaged mtDNA as an important signaling molecule with ZBP1 as its intracellular receptor in the development of an inflammatory response in the retina. We propose that this novel mtDNA-mediated autocrine and paracrine mechanism for triggering and maintaining inflammation in the retina may play an important role in ocular pathologies. Therefore, the molecular mechanisms identified in this report are potentially suitable therapeutic targets to ameliorate development of ocular pathologies.

Oxidative stress, inflammation, and aberrant activation of microglia in the retina are commonly observed in ocular pathologies. In glaucoma or age-related macular degeneration, the chronic activation of microglia affects retinal ganglion cells and photoreceptors, respectively, contributing to gradual vision loss. However, the molecular mechanisms that cause activation of microglia in the retina are not fully understood. Here we show that exposure of retinal pigment epithelial (RPE) cells to chronic low-level oxidative stress induces mitochondrial DNA (mtDNA)-specific damage, and the subsequent translocation of damaged mtDNA to the cytoplasm results in the binding and activation of intracellular DNA receptor Z-DNA-binding protein 1 (ZBP1). Activation of the mtDNA/ZBP1 pathway triggers the expression of proinflammatory markers in RPE cells. In addition, we show that the enhanced release of extracellular vesicles (EVs) containing fragments of mtDNA derived from the apical site of RPE cells induces a proinflammatory phenotype of microglia via activation of ZBP1 signaling. Collectively, our report establishes oxidatively damaged mtDNA as an important signaling molecule with ZBP1 as its intracellular receptor in the development of an inflammatory response in the retina. We propose that this novel mtDNA-mediated autocrine and paracrine mechanism for triggering and maintaining inflammation in the retina may play an important role in ocular pathologies. Therefore, the molecular mechanisms identified in this report are potentially suitable therapeutic targets to ameliorate development of ocular pathologies.
Normal retina function is facilitated by an orchestrated interplay between several cell types. The retinal pigment epithelium (RPE) is a highly specialized monolayer of cells with pigmented microvilli, lining the Bruch's membrane located between the neural retina and the choroid in the eye. RPE cells perform several critical functions to maintain retinal homeostasis, including phagocytosis of photoreceptor outer segments, recycling vitamin A, and maintaining the blood-retinal barrier (1). Polarized, mature, and terminally differentiated RPE cells are mitotically quiescent under normal physiological conditions. The RPE is also characterized by an apicalbasolateral polarity, with the basolateral surface sitting on the Bruch's membrane and the apical microvilli interfacing with the photoreceptors outer segments. The polarized architecture of the RPE together with tight junctions between cells enables the RPE to form a blood-retinal barrier that is critical for transport of nutrients, ions, and water across the RPE and for directional secretion of growth factors and extracellular matrix components to help maintain structure and function of the photoreceptors and choroids (1,2). Thus, considering the critical role of RPE in maintenance of retinal homeostasis, dysfunctional RPE is observed in several retinopathies. The best studied example of RPE-initiated ocular pathology is age-related macular degeneration (AMD). Many experimental and clinical data support the view that damage to RPE cells by enhanced oxidative stress is a crucial factor in AMD development (3)(4)(5)(6). Morphological changes in the RPE are also frequently involved in hereditary retinopathies such as retinitis pigmentosa where progressive degeneration of rods and cones is directly linked with RPE dysfunction (7).
Damage to mitochondrial DNA (mtDNA) together with mitochondrial dysfunction is well recognized in AMD, but they are currently viewed as a passive downstream effect of oxidative/inflammatory injury (8)(9)(10)(11)(12)(13)(14). It is established that in comparison to nuclear DNA (nuDNA), mtDNA is highly susceptible to oxidative injury in several cell types, including post mitotic cells (15)(16)(17). Also, from an immunological point of view, since several characteristics of mtDNA bear a resemblance to bacterial origin (small size, multiple copies, methylation status), it is recognized by immune cells as "foreign" not "self" DNA when is released to cytoplasm or extracellular space (18). Damaged mtDNA liberated from necrotic cells was shown to induce inflammatory responses, mostly in immune cells, by binding to several distinct RNA/ DNA receptors in various diseases (18). However, the importance of these DNA/RNA receptors in ocular pathologies is just starting to emerge.
Activation of microglia, resident immune cells of the central nervous system including the retina, is directly linked with development of various ocular pathologies. Microglia play a critical role in homeostatic maintenance of the neuroretinal microenvironment (19). In the normal, healthy retina, microglia are located in the inner retina, the ganglion cell layer, and inner and outer plexiform layer, where they exhibit branched cellular processes responsible for immune surveillance. Trauma or noxious insults to the retina such as oxidative stress, hypoxia, or inherited mutations trigger microglia reactivity characterized by amoeboid morphology, increased proliferation, and migration to the site of injury (20). While the initial "beneficial" inflammatory response can rapidly enhance tissue repair and a return to homeostasis, chronic activation of microglia induces severe alterations in retinal integrity and aggravates neuronal death (19,20). It has been proposed that pathological microglia and RPE cell cross talk in the subretinal space results in alterations to the structure and physiology of the RPE layer, which in turn transforms the retinochoroidal interface into a favorable environment for the progression and advancement of several ocular pathologies, including AMD (21). Activated microglia induce RPE alterations that result in an elevated chemoattractant, proinflammatory, and proangiogenic environment, which increases the recruitment and activation of immune cells and fosters the growth of neovascular vessels into the retina. In hereditary retinopathies such as retinitis pigmentosa, chronically activated microglia are engaged in the phagocytosis of rod debris and also exacerbate photoreceptors and other adjacent cells by secretion of proinflammatory neurotoxic factors. The critical roles of microglia in neurodegeneration of retinal ganglion cells in glaucoma and diabetic retinopathies have also been recognized (19,21). Although substantial advances have recently been made in understanding mechanisms of microglia activation, the communication between microglia and other retinaspecific cell types is still not well defined.
Oxidative stress, RPE dysfunction, and activation of the proinflammatory phenotype of microglia are commonly observed in ocular pathologies. In this project, we investigated the effect of chronic oxidative stress on communication between RPE cells and microglia. Our study revealed a unique RPE-microglia cross talk mediated by extracellular vesicles (EVs) and illuminated the importance of autocrine and paracrine signaling mediated by oxidatively damaged mtDNA. Moreover, we identified a pathological role of damaged mtDNA through its interaction with the immune DNA receptor, Z-DNA-binding protein 1 (ZBP1), in the development of a pathological proinflammatory environment in the retina, which may serve as a novel therapeutic target for ocular pathologies.

Results
Detrimental effect of chronic low-level oxidative stress on metabolic status, cellular viability, and DNA damage of differentiated ARPE-19 cells In this study, we used differentiated human retinal pigment epithelial cells (ARPE-19) that have been shown to regain characteristics of primary RPE cells after long-term culturing under low-serum conditions (22). ARPE-19 cells were maintained for at least 3 months in medium containing 1% FBS and their differentiation was confirmed by Western blot analysis of cellular retinaldehyde-binding protein 1 (CRLBP), a protein known to be associated with the vision cycle and expressed only in fully differentiated RPE cells (22). We also tested the protein level of monocarboxylate transporter 3 (MCT3) that was shown to be expressed only in fully differentiated and polarized ARPE-19 cells (23). In agreement with these reports, unlike undifferentiated, proliferating ARPE-19 cells, the differentiated ARPE-19 and primary human RPE cells showed marked expression of both CRLBP and MCT3 (Figs. 1A and S1). In subsequent experiments, we used only fully differentiated ARPE-19 cells.
Virtually all ocular pathologies that progress over time are linked with chronic, low-level oxidative stress as a major contributing factor for the development of pathological characteristics. Although most studies to date utilized H 2 O 2 directly to induce oxidative stress, H 2 O 2 treatment induces bolus oxidative stress that is not physiologically relevant, particularly for studies of time-dependent pathologies. Therefore, to mimic chronic oxidative stress, we used a glucose oxidase (GOx) treatment that can enzymatically generate constant low levels of H 2 O 2 over an extended period of time (24). The amount of H 2 O 2 generated by GOx was measured with Amplex Red and compared with a standard curve generated with H 2 O 2 (Fig. S2A). We measured that 10 mU/ml of GOx generated 40 μM, 60 μM, and 100 μM of H 2 O 2 within 1, 3, and 6 h, respectively (Fig. S2B). Next, we compared the effects of GOx and direct H 2 O 2 exposure by analysis of metabolic status and necrosis in ARPE-19 cells, using MTT assay and release of lactate dehydrogenase (LDH), respectively, and detected a marked difference. As an example, 30 mU/ml of GOx (which generates 300 μM of H 2 O 2 within 6 h) reduced MTT by 50%, 80% and 95% at 2, 6, and 24 h, respectively, while a modest (less than 20%) decrease of MTT was measured only at 2 h post 300 μM H 2 O 2 treatment (Figs. 1B and S3A). Similarly, treatment with 100 mU/ml of GOx (which generates 1 mM H 2 O 2 within 6 h) induced a two-and fivefold increase of LDH release at 6 h and 24 h, while a twofold increase in LDH release was measured only at 2 h post 1 mM H 2 O 2 treatment (Figs. 1B and S3A). In addition, we tested the effects of GOx and H 2 O 2 treatments on DNA damage measured by a long amplicon semiquantitative PCR (LA-qPCR) method that allows for concurrent measurement of nuclear and mitochondrial DNA (nuDNA and mtDNA) damage (25). Similarly, as above, the dynamics of oxidatively induced DNA damage and its repair in differentiated ARPE-19 cells treated with GOx and H 2 O 2 is different. The 30 mU/ml of GOx (equivalent to 300 μM H 2 O 2 ) induced a 40% and 80-90% decrease of mtDNA integrity at 2, 6, and 24 h, respectively, together with a significant decrease of nuDNA integrity measured at 24 h (Fig. 1C). In comparison, direct treatment of ARPE-19 cells with 300 μM of H 2 O 2 has no effect on mtDNA and nuDNA integrity at any time points (Fig. S3B). Similarly, treatment of ARPE-19 cells with 100 mU/ml of GOx (equivalent to 1 mM H 2 O 2 ) induced an 80% decrease of mtDNA as early as 2 h, with a 60% decrease in integrity of nuDNA as early as 6 h (Fig. 1C). However, direct treatment with 1 mM H 2 O 2 induced reduction of mtDNA integrity by 25% only at 2 h, which is mtDNA activates microglia subsequently repaired by 6 h. At this concentration of direct H 2 O 2, no measurable changes in the integrity of nuDNA were observed (Fig. S3B). Together, these data show a marked different cellular response to bolus H 2 O 2 treatment and chronic oxidative stress induced by GOx. We detected detrimental effects of chronic, low-level oxidative stress at a relatively low concentration of H 2 O 2 . We calculated that at least one order of magnitude less H 2 O 2 generated by GOx has more significant effects on cellular viability as compared with direct H 2 O 2 treatment. This observation is particularly important in studies of age or time-dependent pathologies where the effects of prolonged exposure to low levels of oxidants are suspected to drive pathological changes. In subsequent experiments, we use a sublethal concentration of GOx (30 mU/ml) to investigate its effects in differentiated ARPE-19 cells.
Oxidatively damaged mtDNA binds to the cytoplasmic DNA receptor, Z-DNA binding protein 1 We previously reported that exposure of replicating lung epithelial cells to oxidative stress causes mtDNA-specific damage and its translocation to the cytoplasm where it then binds and activates ZBP1, inducing the expression of proinflammatory markers (16). Here, we confirmed that differentiated ARPE-19 cells are also highly susceptible to mtDNA damage from oxidative injury (Fig. 1C). Since chronic inflammation is commonly observed in many pathologies including age-related macular degeneration, glaucoma, diabetic retinopathies, and others, we tested the effects of chronic oxidative stress on intracellular translocation of mtDNA using proximity ligation assay (PLA). PLA detects the binary interaction between two epitopes in cellulo based on antibody specificity (26). First, we tested the interaction between mtDNA and mitochondrial transcription factor A (TFAM), which is known to stably interact with mtDNA in mitochondria under physiological conditions (27). We compared interactions between mtDNA/TFAM in response to 30 mU/ml of GOx and 300 μM H 2 O 2 treatment. While a marked reduction of the mtDNA/TFAM signal at 1 h was detected in ARPE-19 cells treated with GOx ( Fig. 2A), no effect on mtDNA/TFAM interactions post direct H 2 O 2 treatment was observed (Fig. 2B). These results further confirm different cellular outcomes to exposure of chronic and bolus oxidative stress. We hypothesized that GOx-induced mtDNA damage (Fig. 1C) caused the dissociation between mtDNA and TFAM ( Fig. 2A) resulting in translocation of damaged mtDNA to the cytoplasm. Therefore, at various time points, we tested GOx treatment on interaction between mtDNA and several intracellular DNA/RNA receptors (18). Using PLA we detected marked increased interaction between mtDNA and ZBP1 as early as 1 h, which was maintained at 6 h and 24 h post GOx treatment (Fig. 2C). We did not detect interaction between mtDNA and Toll-like receptor 9 (TLR9), cyclic GMP-AMP synthase (cGAS), or nucleotide-binding domain leucine-rich repeat family, pyrin-domain containing 3 (NLRP3) at any time points post GOx treatment (Fig. 2C). Since PLA depends highly on antibody specificity, we further confirmed intracellular localization of PLA signal (mtDNA/TFAM and mtDNA/ ZBP1) by costaining PLA slides with the mitochondrial marker, subunit B of ATP synthase (ATPB). As expected, we observed colocalization of mtDNA/TFAM PLA signal with mitochondria in control but not GOx-treated cells and lack of colocalization of mtDNA/ZBP1 PLA signal in mitochondria of GOx-treated cells (Fig. S4). Since PLA depends on antibodies specificity, we independently confirm the interaction between mtDNA and ZBP1 in response to prolonged oxidative stress, by expressing HA-tagged ZBP1 in differentiated ARPE-19 cells (Fig. 2D) and tested for the presence of mtDNA and nuDNA in ZBP1-HA immunoprecipitated samples (IPs). More than eightfold increase of the mtDNA in ZBP1 IPs was measured with qPCR under GOx treatment, but not in control cells ( Fig. 2E) with negligible amounts of nuDNA detected in IPs (Fig. S5). Together, these data showed that under chronic oxidative stress, oxidatively damaged mtDNA is translocated to the cytoplasm to stably bind cytoplasmic receptor ZBP1.

Inflammatory response triggered by activation of ZBP1 signaling by mtDNA
The role of ZBP1 is recognized in virally induced inflammatory responses and programmed cell death (PANoptosis), including regulated pyroptosis (programmed inflammatory cell death), necroptosis (programmed necrosis), and apoptosis (28). To investigate which ZBP1-mediated signaling pathway is activated by oxidatively damaged mtDNA (Fig. S6), we tested the interaction of ZBP1 with interferon regulatory factor 3 (IRF3), TANK-binding kinase 1 (TBK1), and kinases receptorinteracting protein 1 and 3 (RIPK1 and RIPK3) in differentiated ARPE-19 cells treated with 30 mU/ml of GOx for 2 h. We detected interactions between ZBP1 and RIPK1, IRF3, TBK1, but not with RIPK3, indicating activation of the noninflammasomemediated inflammatory response (Fig. 3A). Because activation of the cGAS/stimulator of interferon genes (STING) pathway was recently suggested in AMD (29), we tested the effects of GOx treatment on the interaction between STING and IRF3 or TBK1. The lack of interaction between these proteins rules out the activation of cGAS/STING pathway induced by chronic oxidative stress in ARPE-19 cells (Fig. 3B), which further supports a lack of interaction between mtDNA and cGAS (Fig. 2C). Moreover, the lack of activation of ZBP1-mediated cell death in GOx-treated To test our conclusion that prolonged exposure to low oxidative stress results in inflammation, we measured the expression of several proinflammatory markers in response to GOx treatment at 24 h and detected a robust increase of IL-1α and IL-8 mRNA levels (Figs. 4A and S7). To confirm the key role of ZBP1 in the expression of proinflammatory markers, we used mouse embryonic fibroblasts (MEFs) derived from wild-type (WT) and ZBP1 KO mice (Fig. S8). As observed with ARPE-19 cells, significant mtDNA-specific damage was measured in MEFs subjected to 30 mU/ml of GOx at 1 h (Fig. 4B). Most importantly, a significant reduction of IL-1α, and KC (mouse equivalent of human IL-8) expression was measured in GOx-treated MEFs derived from ZBP1 KO mice in comparison to WT (Fig. 4C). These data strongly support the key role of mtDNA-mediated activation of ZBP1 signaling in generating a proinflammatory environment in response to prolonged oxidative stress in differentiated RPE cells.

Chronic oxidative stress induces release of extracellular vesicles (EVs) containing large fragments of mtDNA from differentiated ARPE-19 cells
We tested whether susceptibility of mtDNA to prolonged oxidative stress (Fig. 1C) has effects on mitochondrial respiration. All major bioenergetic parameters, namely oxygen consumption linked with basal respiration, maximal respiration, and ATP production, were significantly reduced at 3 h and 24 h post GOx treatment, with a marked increase in proton leakage (Fig. S9). Interestingly, we also detected a timedependent decrease of mtDNA content in ARPE-19 cells treated with 30 mU/ml of GOx (Fig. 5A), suggesting that oxidatively damaged mtDNA is translocated to cytoplasm to bind ZBP1, but is also subjected to degradation or extrusion to the extracellular space. To test the latter, we measured the number of particles derived from control and GOx-treated differentiated ARPE-19 cells using Nanoparticle Tracking Analysis (NTA) and detected an almost tenfold increase in total particle count (Fig. 5B). Interestingly, time-dependent analysis of the particle sizes revealed that larger particles (150 nm) are released earlier and smaller (60 nm) are released at later time points following GOx treatment (Fig. 5, C and D). Importantly, 90% of released EVs are smaller than 200 nm, ruling out the presence of apoptotic bodies in the collected media of GOx-treated cells and independently confirming a lack of apoptosis in our experimental conditions (Fig. 5C). We also detected the same profile of EVs in precleared (three subsequent low-speed centrifugations) medium and post ultracentrifugation (Fig. S10A) and showed that almost all EVs present in culture medium can be sedimented mtDNA activates microglia by ultracentrifugation (Fig. S10B). To test whether protein aggregates are generated by GOx treatment as opposed to EVs, we treated precleared media with either active or denatured (inactive) Proteinase K. We found no significant decrease in total particle count under Proteinase K treatment, which serves as confirmation that the particles present are EVs released from differentiated ARPE-19 cells (Fig. 5E). Next, we compared amounts of mtDNA and nuDNA in EVs sedimented by ultracentrifugation and detected a threefold increase of mtDNA in EVs derived from GOx-treated cells as compared with control and only negligible amounts of nuDNA in both treated and control samples (Fig. 5F). The majority of mtDNA is present in sedimented EVs but not in the remaining medium (Fig. S10C). Then, we investigated the properties of the mtDNA by analyzing sequence specificity and fragment size of mtDNA present in EVs. Using qPCR, we detected increased amounts of the most of mtDNA-encoded genes (Fig. 6A) in EVs derived from GOx-treated ARPE-19 cells, in comparison to the control (Fig. 6B). This further confirmed enhanced amounts of mtDNA in EVs derived from GOx-treated cells. To investigate fragment sizes of the mtDNA in EVs, we designed primers to amplify mtDNA fragments of 200 bp up to 6 kb, localized in three different regions of the mtDNA genome (Figs. 6A and S11). We were able to amplify the majority of the mtDNA fragments, particularly in EVs derived from GOx-treated cells (Fig. 6, C and D). In addition, we also determined that IL-8, which exhibits increased expression in ARPE-19 cells treated with GOx (Fig. 4A), is released to the culture medium but not within EVs (Fig. 6E). Taken together, we concluded that low-level oxidative stress enhances the release of EVs and increase amounts of mtDNA fragments of various sizes (up to 6 kb) in EVs derived from differentiated ARPE-19 cells.
EVs containing mtDNA derived from oxidatively stressed, polarized ARPE-19 cells activate microglia Fully differentiated ARPE-19 cells are characterized by an apical-basolateral polarity (2). Therefore, we tested the amount of DNA in EVs released from the apical and basolateral sites of  7A). We measured two orders of magnitude more mtDNA than nuDNA in EVs released from control and GOxtreated cells (Fig. 7, B and C). We also measured markedly more mtDNA released from the apical site than from the basolateral site, which was further enhanced (twofold increase) in cells subjected to GOx (Fig. 7B). More nuDNA was found in EVs released from the basolateral site than from the apical site, but its amount was reduced in GOx-treated ARPE-19 cells (Fig. 7C). To test the importance of ARPE-19 cell polarity on EVs release, we also measured the amount of nuDNA and mtDNA in EVs released from apical and basolateral sites of undifferentiated, replicating ARPE-19 cells. Generally, as with polarized, differentiated ARPE-19 cells, more mtDNA than nuDNA was measured in EVs of undifferentiated cells (Fig. S12). However, opposite to polarized ARPE-19 cells, EVs isolated from the basolateral site of replicating ARPE-19 cells contained more mtDNA and its amount did not change post GOx treatment (Fig. S12A). Also, in contrast to differentiated ARPE-19 cells, more nuDNA was found in EVs released from the apical site, which decreased post GOx treatment (Fig. S12B). These data emphasize the importance of testing fully differentiated and polarized ARPE-19 cells that closely mimic the primary RPE cells. Together, we found a marked difference in the amount of mtDNA in EVs released from the apical and the basolateral site suggesting the possibility of different roles for EVs released from polarized differentiated ARPE-19 cells.
Since microglia activation has been linked to many ocular pathologies (19,20), we tested the effects of EVs released from RPE cells on microglia activation. ARPE-19 cells differentiated on transwell plates were subjected to 30 mU/ml GOx for 24 h, and EVs isolated from apical and basolateral sites were used to treat unstimulated HMC3 cells (human microglia), followed by measurement of the expression level of several cytokines (Fig. 7A). We detected a significant increase in the mRNA of proinflammatory markers (IL-1β, IL-6, IL-8, and TNFα) at 24 h in HMC3 treated with EVs isolated from the apical site of GOx-treated, but not in control ARPE-19 cells (Fig. 7D). In addition, EVs isolated from the basolateral site of control or

mtDNA activates microglia
GOx-treated ARPE-19 cells had no effect with regard to changes in mRNA of inflammatory markers (Fig. 7D). Since EVs isolated from the apical site of GOx-treated ARPE-19 cells contained enhanced amounts of mtDNA (Fig. 7B), we also directly tested the effects of treatment with isolated mtDNA and nuDNA (1 μg) encapsulated within a lipid bilayer (Lipofectamine 2000) in naïve HMC3 cells. A robust increase in expression of TNFα was detected in HMC3 cells treated only with isolated mtDNA, but not nuDNA (Fig. 7E). Finally, we tested which intracellular DNA receptor in microglia binds mtDNA derived from RPE cells. We isolated mtDNA from ARPE-19 cells, amplified it in the presence of bromodeoxyuridine (BrDU), and treated naïve HMC3 cells with BrDU-labeled mtDNA encapsulated within a lipid bilayer (Lipofectamine 2000, Fig. 8A). Using the PLA approach, we detected an interaction between BrDU-labeled mtDNA with ZBP1 but not with cGAS, TLR9, AIM2, or NLRP3 (Fig. 8B). To further confirm microglia activation by isolated mtDNA, we detected that treatment of HMC3 with isolated mtDNA, but not nuDNA, caused enhanced generation of reactive oxygen species (ROS) in HMC3, as detected by MitoSOX staining (Fig. 8C). Together, these data revealed a previously unknown mechanism of intercellular communication by EVs derived from the apical site of RPE cells in activation of microglia. Moreover, our data point toward mtDNA as being a critical signaling molecule in RPE-microglia cross talk and ZBP1 acting as its intracellular receptor.
In summary, our study shows that chronic oxidative stress that is commonly observed in ocular pathologies results in the induction of mtDNA-specific damage that causes expression of proinflammatory markers in RPE cells via activation of ZBP1 signaling. Additionally, polarized RPE cells subjected to oxidative stress release EVs containing mtDNA that are capable of activation of proinflammatory phenotype of microglia. Based on these data, we propose a novel mtDNA-centered mechanism for the generation of a proinflammatory environment in the retina. In our model, low levels of oxidative stress generate a proinflammatory niche by activation of mtDNA/ZBP1 pathway in RPE while releasing EVs containing mtDNA within the retina, which then activates the proinflammatory phenotype of microglia. We suspect that chronic activation of pathological RPE-microglia cross talk affects retina homeostasis resulting in a decreased viability of photoreceptors, ganglion, and RPE cells, among others, contributing to retina dysfunction (Fig. 9).

mtDNA activates microglia
Together, we propose that the molecular mechanisms described in this study contribute to the development of several ocular pathologies and therefore are potentially novel targets for therapeutic interventions.

Discussion
Our studies identified a central role of oxidatively damaged mtDNA and its intracellular receptor, ZBP1, in activation of microglia and generation of a proinflammatory environment in retina. This mtDNA-centered autocrine and paracrine signaling occurred in response to chronic oxidative stress, a type of insult that is considered to drive several ocular pathologies such as AMD. Additionally, our data highlight that fundamentally different cellular responses are generated by prolonged low levels of enzymatically generated H 2 O 2 versus bolus H 2 O 2 treatment, suggesting it is critical to consider the importance of the dynamics of oxidatively induced damage. Fully differentiated RPE cells can tolerate exposure to high levels of bolus H 2 O 2 but are highly susceptible to chronic low doses of H 2 O 2 . We hypothesize that the differential outcomes of bolus versus chronic oxidative stress are most likely due to the kinetics of ROS generation. Similar modeling of chronic oxidative stress in analysis of time-developed phenotype has been previously proposed in the ocular field (30). Furthermore, prolonged exposure to low levels of oxidants triggered mtDNA-mediated signaling that is absent in cells treated with bolus amounts of oxidants. Therefore, our report highlights mtDNA as a novel key signaling molecule in response to chronic oxidative stress conditions. There are several reports detecting greater damage to mtDNA than to nuDNA in retinal cells in aging animals and ocular pathologies. Increased amounts of 8-hydroxy-2 0 -deoxyguanosine (8-oxoG), the most frequently occurring DNA lesion, together with increased numbers of DNA breaks and deletions, were reported in mtDNA from the retina of old mice (10). It was proposed that this DNA damage is caused by age-related decrease in the expression of DNA repair enzymes and reduction in DNA repair capacity in aging retinas (10). We previously detected age-related and tissue-specific decrease of DNA repair capacity in mitochondria, along with a decrease in expression of antioxidant enzymes as an underlying cause of age-dependent accumulation of mtDNA damage (31). Increased levels of mtDNA damage in human aged macular, but not the peripheral region of retina, in combination with decreased DNA repair capacity were also reported previously (14). In contrast, another report showed no difference in the amounts of mtDNA damage between the macular and peripheral region in aged human donor eyes, but significant mtDNA damage was found in RPE cells (32). It was shown that the level of mtDNA damage positively correlates with the grading level of AMD (14). An increased amount of mtDNA deletion was suggested to be a result of the aging process, while increased numbers of DNA damage (DNA breaks, DNA lesions) were linked with AMD (13). It was also reported that mtDNA damage accumulates eight times faster than its nuclear counterpart and that in AMD there is an increased enhancement of mtDNA damage (13). All these reports clearly indicate a specific accumulation of mtDNA damage, DNA breaks, and DNA lesions, particularly in RPE mtDNA activates microglia cells during aging, a process that is further accelerated in AMD. These reports also strongly support testing the effects of chronic oxidative stress such as that generated by GOx treatment, used in this report, in cultured cells as a model closely reflecting age-and time-related changes.
Numerous reports demonstrate that mtDNA is particularly susceptible to oxidative stress and that oxidative mtDNA damage persists longer than in nuDNA (31,(33)(34)(35)(36)(37)(38)(39)(40). Some of the contributing factors include a reduced number of mtDNA repair pathways (41), enzymes involved in mtDNA repair (42)(43)(44), the close proximity to the mitochondrial respiratory chain, the main intracellular source of ROS (45). Indeed, we found that mtDNA is a primary target for oxidative stress also in fully differentiated ARPE-19 cells. Both bolus and chronic exposure to H 2 O 2 primarily induce damage to mtDNA, followed by damage to the nuDNA (Figs. 1 and S3B). While damage induced by bolus H 2 O 2 treatment can to some extent be efficiently repaired, GOx-induced time-dependent exposure to oxidants resulted in an increase of mtDNA damage at a rate exceeding mitochondrial DNA repair capacity, causing mitochondrial dysfunction that includes an increase of mitochondrial-derived oxidants, which in turn may further accelerate damage to nuDNA (Fig. S9). Consequently, we detected intracellular translocation of oxidatively damaged mtDNA to the cytoplasm where it stably binds the DNA/RNA receptor ZBP1 (Fig. 2). It was reported that mtDNA liberated from mostly necrotic cells induces inflammatory responses in immune cells by binding to Toll-like receptor 9 (TLR9), cyclic GMP-AMP synthase (cGAS), and nucleotide-binding domain leucine-rich repeat family, pyrin-domain containing 3 (NLRP3) inflammasome (18,(46)(47)(48)(49)(50)(51)(52)(53). ZBP1 was initially identified as an intracellular receptor specific to viral and bacterial DNA and RNA causing activation of type I interferon (28,54) and NF-kB (28,(54)(55)(56)(57) signaling. Recent reports showed the critical role of ZBP1-mediated signaling in the development of skin (58,59) and bowel (60) sterile inflammation; however, these reports lacked identification of endogenous nucleic acids that bind and trigger ZBP1 signaling. Our previous report (16) showed for the first time that ZBP1 binds to oxidatively damaged mtDNA and triggers expression of proinflammatory markers in viable lung epithelial cells. Recent reports showed activation of the cGAS-STING pathway by mtDNA liberated from mitochondria and translocated to the cytoplasm in response to pyrimidine deficiency and defective mitophagy, ultimately leads to activation of downstream STINGdependent inflammatory response (61,62). Our current report shows that exposure of differentiated RPE cells to chronic oxidative stress results in mtDNA damage and its translocation to cytoplasm where it binds and activates ZBP1 without activating cGAS-STING signaling (Figs. 2 and 3). Therefore, an immediate question arises: how does mtDNA released from mitochondria become distinctly recognized by different intracellular DNA/RNA receptors? At this point, we speculate that such differential interactions may be due to the specific type of damage induced by the GOx on mtDNA, or the size of mtDNA fragments released to the cytoplasm; however, these hypotheses need future investigation.
The role of RPE in the development of ocular pathology is extensively studied particularly in connection with AMD. The proper maintenance of mitochondrial homeostasis in RPE cells is crucial since they utilize oxidative phosphorylation for most of their energy production (63). It has been proposed that altered RPE cell metabolism could create a bioenergetic crisis in the retina that drives toward AMD pathology (64). Two independent studies showed reduced oxidative phosphorylation in RPE cells cultured from human donors with AMD when compared with RPE cells isolated from healthy donors (65,66). Additionally, we have herein detected significant mtDNA damage and remodeling of the major mitochondrial bioenergetic parameters, including a decrease in oxygen consumption linked with ATP production in differentiated ARPE-19 cells subjected to chronic oxidative stress (Figs. 1C and S9). Based on our data, we postulate that mitochondrial dysfunction including mtDNA damage may be an active source of retinal dysfunction rather than its passive consequence. We propose that damaged mtDNA serves as an important molecule in autocrine and paracrine-mediated signaling including mtDNA activates microglia activation of microglia. It was shown that prolonged microglial activation is detrimental to the cell, causing extensive ROS and inflammatory marker secretion including IL-1β, IL-6, TNFα. The generation of proinflammatory environments fostered by reactive microglia negatively affects retinal homeostasis, leading to cell death of photoreceptors and RPE dysfunction (20). Enlarged, chronically activated amoeboid microglia were found in the retina of AMD subjects (67). Our data showed that RPE-derived EVs containing mtDNA are a potent inducer of microglial activation (Fig. 7), and that persistent, low-level oxidative stress in RPE cells causes a substantial increase in mtDNA containing EV release to the surrounding environment over time (Fig. 5B). Thus, we propose that chronic activation of RPE-microglia communication via EVs significantly contributes to the generation of a detrimental proinflammatory and pro-oxidative niche in the retina. In fact, reactive microglia induce NLRP3 inflammasome activation in RPE cells that can subsequently induce RPE degeneration via caspase-1 mediated pyroptosis (68). ZBP1 is implicated as an upstream regulator of necrotic cell death (69); therefore, we hypothesize that chronic activation of microglia could in turn enhance mtDNA damage in RPE cells causing chronic activation of ZBP1 and subsequent retina degeneration. It is worth noting that while the GOx used in our experiments induces chronic oxidative stress, the amount of generated H 2 O 2 plateaus by 6 h, most likely due to the decreased enzymatic activity of GOx (Fig. S2). These conditions clearly activated ZBP1-mediated expression of proinflammatory markers without significant decrease of cellular viability. However, we cannot exclude the possibility that constant exposure to oxidants, e.g., by persistently activated microglia, could trigger activation of ZBP1-mediated PANoptosis at later time points.
Finally, our data revealed that fully differentiated, but not replicative RPE cells secrete EVs containing different cargo from the apical and basolateral sites. It was reported that bolus H 2 O 2 treatment of RPE cells causes secretion of neuroprotective αB-Crystallin from the apical site via EVs (70), while stimulation of polarized RPE cells differentially secretes IL-6 and IL-8 from apical and basolateral sites (71). In addition, significantly more EVs were secreted from the apical than the basolateral site of polarized porcine RPE cells (72). We found enhanced amounts of mtDNA present in EVs isolated from differentiated RPE cells that is further enhanced by oxidative stress (Figs. 5 and 7). In addition, we performed initial characterization of the mtDNA derived from RPE cells. Interestingly, although we found a lack of mtDNA sequence specificity, we discovered that large mtDNA fragments are present in EVs released from RPE cells (Fig. 6). Most importantly, our study discovered a previously unknown mechanism of communication mediated by the RPE-derived EVs in cross talk with microglia, which emphasizes the key role of mtDNA. The significance of EVs in cell-to-cell communication, immune modulation, and neovascularization among other functions is well established in cancer, but their role in ocular pathologies is yet to be fully understood (73,74). In this report we focus on mtDNA, but differential content of EVs released from polarized RPE cells need further investigation since they may contain other important mediators in communication with microglia and other cell types present in the retina. This seems to be particularly important since marked increase of EVs released from GOx-treated cells was measured. Thus, as this point we cannot exclude possibility that mtDNA is only one of the factors present in EVs that activate microglia. We also did note a time-dependent change in the size of EVs released from differentiated RPE cells following GOx exposure (Fig. 5, C and D), highlighting the dynamic nature of the oxidative stress response and perhaps indicating a shift in cargo content, which has yet to be identified. This observation also needs further analysis. Altogether, we have described novel underlying molecular mechanisms that may potentially drive the development of chronic inflammation that frequently occurs in ocular pathologies. In our model, chronic low levels of oxidative stress generate a proinflammatory niche in RPE cells by activation of the mtDNA/ZBP1 pathway and subsequent release of EVs containing mtDNA to the apical site, then activation of the proinflammatory phenotype of microglia, which in turn will decrease viability of photoreceptors, ganglion cells, and RPE cells, among others (Fig. 9). Our novel findings highlight various key players in the activation of retinal inflammation and expose new molecular pathways that may be used as target candidates in ocular therapies.

Glucose oxidase (GOx) and hydrogen peroxide (H 2 O 2 ) treatment
To generate constant, low-level oxidative stress, cells were treated with various concentration of glucose oxidase, GOx mtDNA activates microglia (SIGMA #G2133-50KU). GOx was resuspended in phosphate buffered saline (PBS) to prepare 100X stock that was diluted into to the culture medium. To generate bolus oxidative stress, cells were treated with various concentrations of H 2 O 2 (EMD Millipore #HX0635-3).

MTT assay
To measure changes in the metabolic activity of differentiated ARPE-19 cells, 30,000 cells were seeded per well in a 96-well plate and differentiated in a medium containing 1% FBS for 3 months. Next, a solution containing 1:10 of culture medium to MTT reagent (3 mg/ml, Millipore #475989) was incubated in a humidified incubator at 37 C with 5% CO 2 for 1 h. Culture medium was removed, and cells were resuspended in DMSO (Fisher #D128). The absorption was measured at 590 nm using a SpectraMax (Molecular Devices) microplate reader with SoftMax Pro 5.3 software. Four technical replicates were measured for each readout.

LDH assay
Necrotic cell death was assayed by the quantification of activity of lactate dehydrogenase (LDH) release to the culture medium. Briefly, 30,000 cells were seeded per well in a 96-well plate and differentiated in a medium containing 1% FBS for 3 months. Fifty microliters of culture medium was mixed with 50 μl of 200 mM Tris pH 8.0 (SIGMA #T3253), 50 μl of 50 mM Lithium Lactate (SIGMA #L1500), and 50 μl of NAD/ PMS/INT solution (SIGMA #N0632, #P9625, #I8377) followed by absorbance reading at 490 nm over 15 min using the SpectraMax (Molecular Devices) microplate reader with SoftMax Pro 5.3 software. Four technical replicates were measured for each readout.

ZBP1 immunoprecipitation (IP)
Differentiated ARPE-19 cells (1 million cells) were transfected with 100 ng of pCAGGS-HA-hZBP1 or empty expression vector (generous gift from Dr Taniguchi, University of Tokyo, Japan) using Lipofectamine 2000 (Thermo Fisher Scientific, #11668019). After 48 h, cells were washed once with PBS and treated with PBS (control) or GOx (30 mU/ml) for 1 h. Cells were washed with PBS, and DNA-protein cross link was performed by addition of formaldehyde (0.75%) for 10 min, followed by addition of glycine (125 mM) for 10 min. Cells were washed twice with PBS and lysed using NP-40 buffer (150 mM NaCl, 50 mM Tris-Cl pH8.0, 1% NP-40) for 30 min on ice followed by brief centrifugation to separate the insoluble fraction. Protein concentration was determined using DC Protein Assay (BioRad #5000111) according to the manufacturer's recommendation. In total, 200 μg of total cell extract was used for IP using Pierce Magnetic HA-Tag IP/co-IP Kit (Thermo Scientific, #88838) according to manufacturer's recommendations. Total DNA from IPs was isolated using DNeasy Blood & Tissue Kit (QIAGEN #69506) according to manufacturer's recommendations. The amounts of mtDNA and nuDNA were analyzed by qPCR with CFX96 TouchTM Real-Time PCR Detection System (Bio-Rad) and Maxima SYBR Green/ROX qPCR Master Mix (Thermo Scientific #K0221) with the following primers: mtDNA (located in COXIII gene): FW 5 0 -TGACCCACCAATCACATGC-3 0 , RV 5 0 -ATCACATGGCTAGGCCGGAG-3 0 ; nuDNA (located in ACTB gene): FW 5 0 -CATGTACGTTGCTATCCAGGC-3 0 , RV 5 0 -CTCCTTAATGTCACGCACGAT-3 0 . We used the following thermal cycle: 95 C for 10 min, 40 cycles at 95 C for 15 s, and 60 C for 1 min.

Cellular bioenergetics
Analysis of mitochondrial respiration was assessed using the XF24 Extracellular Flux Analyzer (Agilent). Briefly, ARPE-19 cells were seeded on cell culture microplates (60,000/ well) and differentiated for 3 months using 1% FBS-containing medium. For analysis of mitochondrial respiration, cells were washed twice with low-glucose DMEM medium pH 7.4 supplemented with L-Glutamine (2 mM, GIBCO #25030149) and sodium pyruvate (0.33 mM, SIGMA #S8636). After 1 h incubation at 37 C in CO 2 -free incubator, the oxygen consumption rate (OCR) after treatment with oligomycin (1.5 μg/ml, SIGMA #O4876) was used to assess ATP production rate and the OCR after treatment with carbonyl cyanide-4-trifluoromethoxy phenylhydrazone (FCCP, 0.5 μM, SIGMA #C2920) to assess maximal mitochondrial respiratory capacity. Antimycin A (2 μg/ml, SIGMA #A8674) and rotenone (2 μM, SIGMA #R8875) were used to inhibit the flux of electrons through complex III and I, to detect residual nonmitochondrial OCR, which is considered to be due to cytosolic oxidase enzymes. We used four technical replicates for each readout.

Extracellular vesicle (EVs) isolation and characterization with nanoparticle tracking analysis (NTA)
We isolated EVs derived from 300,000 differentiated ARPE-19 cells. EVs released to the culture medium were isolated using the ultracentrifugation approach as described in (77). This method is commonly used to isolate "small EVs" including exosomes and microvesicles. Briefly, to isolate exosomes/microvesicles, we performed three subsequent centrifugation cycles to separate cellular debris from EVs: 2000g 10 min, 20,000g 20 min and 100,000g for 2 h, all at 4 C. The pellet of EVs obtained was resuspended in PBS for subsequent assays. The number and size of EVs were analyzed with NTA using NanoSight NS300 (Malvern Panalytical). Briefly, 1 ml of conditioned media samples was collected followed by differential centrifugation (1000g 5 min, 5000g 10 min, 10,000g 30 min, at 4 C) to clear cell debris and large vesicles before analysis or ultracentrifugation (100,000g at 4 C for 2 h). In some experiments, post ultracentrifugation supernatant was carefully removed and placed in a separate tube while the EV pellet was resuspended in 0.22 μm-filtered PBS to maintain the initial concentration. Samples were then recorded by NTA (3 capture intervals of 60 s per sample with camera level 7, threshold 2). To rule out sedimentation of protein aggregates, in control experiments, we treated EVs with active or inactive Proteinase K (0.1 mg/ml) for 30 min at 37 C before NTA. We inactivated Proteinase K by incubation at 95 C for 15 min. Three technical replicates were used for each readout.

Enzyme-linked immunosorbent assay (ELISA)
The amount of TNFα in cultured medium of HMC3 microglia was determined using human TNF ELISA Set (BD-Bioscience #555212) according to the manufacturer's recommendation. In total, 300,000 HMC3 cells were seeded per well in 12-well plates and treated with mtDNA or nuDNA for 24 h using Lipofectamine 2000 (Thermo Fisher Scientific, #11668019), according to manufacturer's recommendations. The amount of IL-8 in medium and isolated EVs derived from 300,000 differentiated ARPE-19 cell were determined using human IL-8/CXCL8 DuoSet ELISA (R@D System, #DY208-05).

MitoSOX red staining
A mitochondrial superoxide indicator was used to detect mitochondrial ROS production. Briefly, HMC3 (40,000 cells/ well) were seeded in Lab-Tek II chamber slides (Nalge Nunc International, #154526PK) and incubated at 37 C, 5% CO 2 humidified incubator overnight. The cells were treated with 100 ng of isolated from ARPE-19 cell mtDNA or nuDNA encapsulated within Lipofectamine 2000 (Thermo Fisher Scientific, #11668019), for 1 h followed by addition of 5 μM of MitoSOX Red dye (Invitrogen #M36008) for 10 min. For analysis, cells were washed three times with PBS, and the dye's specific fluorescence was visualized using a Nikon eclipse 80i inverted microscope with Photometric CoolSNAP HQ2 camera and NIS-Elements BR 3.10 software.

Analysis of mtDNA in EVs
Total DNA was isolated from EVs sedimented from 300,000 differentiated ARPE-19 cells. PCR primers were designed using the NCBI Primer-BLAST tool with reference mtDNA (accession number NC_012920.1). For each PCR reaction (25 μl), 2 μl of sample DNA and 0.5 μl of each primer (10 mM) were used together with other components according to manufacturer's recommendations for Standard Taq and Long Amp Taq DNA Polymerases (both from New England BioLabs, #M0273S, M0323S). Using a Thermocycler C1000 Touch (BioRad), the PCR programs include initial denaturing step at 95 C for 3 min followed by 32 cycles of 95 C for 30 s, 55 C for 30 s, and 68 C for 1 min, and final extension at 68 C for 5 min. We used 1 min extension step for amplicons 200 bp-1 kb and 6 min for amplicons 3-6 kb. The PCR products were separated using 1% agarose gel and visualized using G:Box with GeneSnap software. We used following primers: site I All statistical analyses were performed using GraphPad Prism software. Except where otherwise stated, presented data are based on at least three biological replicates (cells at different passage number or differentiated while in different passage number; n ≥ 3). In addition, most of the techniques have their technical replicates as indicated above. After performing a normality test of obtained data, we used t-tests to determine statistical significance between two groups. In the analysis of more than two groups, after performing normality tests, a One-Way ANOVA Kruskal-Wallis test was used. Significant differences are denoted as: *p < 0.05, ** p < 0.01.

Data availability
All data supporting the results presented herein are available from the article paper.
Supporting information-This article contains supporting information.
Acknowledgments-We express our gratitude to Clemmie J. White-Matthews and Katherine Kaus PhD for their editorial help. We are very thankful to Praveena Gupta PhD, OD, FAAO, for comments and suggestions. Conflict of interest-The authors declare no conflict of interest in regard to this manuscript.