Clearance of lipid droplets by chimeric autophagy-tethering compound ameliorates the age-related macular degeneration phenotype in mice lacking APOE

ABSTRACT Age-related macular degeneration (AMD) is the leading cause of irreversible blindness among the elderly, and there is currently no clinical treatment targeting the primary impairment of AMD. The earliest clinical hallmark of AMD is drusen, which are yellowish spots mainly composed of lipid droplets (LDs) accumulated under the retinal pigment epithelium (RPE). However, the potential pathogenic role of this excessive LD accumulation in AMD is yet to be determined, partially due to a lack of chemical tools to manipulate LDs specifically. Here, we employed our recently developed Lipid Droplets·AuTophagy Tethering Compounds (LD∙ATTECs) to degrade LDs and to evaluate its consequence on the AMD-like phenotypes in apoe−/− (apolipoprotein E; B6/JGpt-Apoeem1Cd82/Gpt) mouse model. apoe−/− mice fed with high-fat diet (apoe−/−-HFD) exhibited excessive LD accumulation in the retina, particularly with AMD-like phenotypes including RPE degeneration, Bruch’s membrane (BrM) thickening, drusen-like deposits, and photoreceptor dysfunction. LD·ATTEC treatment significantly cleared LDs in RPE/choroidal tissues without perturbing lipid synthesis-related proteins and rescued RPE degeneration and photoreceptor dysfunction in apoe−/−-HFD mice. This observation implied a causal relationship between LD accumulation and AMD-relevant phenotypes. Mechanically, the apoe−/−-HFD mice exhibited elevated oxidative stress and inflammatory signals, both of which were mitigated by the LD·ATTEC treatment. Collectively, this study demonstrated that LD accumulation was a trigger for the process of AMD and provided entry points for the treatment of the initial insult of AMD by degrading LDs. Abbreviations: AMD: age-related macular degeneration; APOE: apolipoprotein E; ATTECs: autophagy-tethering compounds; BODIPY: boron-dipyrromethene; BrM: Bruch’s membrane; ERG: electroretinogram; HFD: high-fat diet; LD·ATTECs: Lipid Droplets·AuTophagy Tethering Compounds; LDs: lipid droplets; OA: oleic acid; OPL: outer plexiform layer; ROS: reactive oxygen species; RPE: retinal pigment epithelium


Introduction
Age-related macular degeneration (AMD) is a disease that results in progressive and unreversible vision impairment. The estimated global prevalence of AMD in 2040 is 288 million, the amount more than the whole number of aggressive cancer and over twice the number of Alzheimer disease [1]. There are two clinical types of AMD, atrophic (dry) and neovascular (wet) AMD, which are characterized by atrophic lesions and abnormal new blood vessels respectively. Most AMD begins as dry AMD, and in 10-20% of individuals, it progresses to wet type [2]. The widely used treatment for AMD is anti-VEGF drugs in the late stage of wet AMD. However, no treatments are available for slowing the progression of early AMD to late AMD or preventing dry AMD. Since dry AMD is the most common type of AMD, it is important to reveal the pathogenic mechanism in the early stages of AMD to identify potential therapeutic strategies to slow down the progression.
A clinical hallmark of AMD are drusen, which consist primarily of lipid droplets (LDs) accumulated under the retinal pigment epithelium (RPE) [3][4][5]. In the course of AMD progress, those drusen become bigger and denser and eventually lead to geographic atrophy of RPE. Evidence suggests a correlation between LD accumulation and AMD development: 1) LD accumulation was found both in RPE from aged mice and AMD donor patients [6,7]. 2) GWAS revealed AMD-associated genes that are highly relevant to lipids, such as APOE, ABCA1 and CETP [8]. 3) Several epidemiology studies reported a positive relationship between hyperlipemia and AMD [9][10][11][12]. Mechanistically, the "lipid wall" of Bruch's membrane (BrM) impedes the normal transport between RPE and choroidal vessels [5]. The role of RPE is to support and provide neighboring photoreceptors with a continuous supply of adequate energy-producing substrates [13]. Its functional failure results in the degeneration of photoreceptors and is related to AMD occurrence. However, it is unclear whether LD accumulation is the cause or a result of AMD. It is reported that LD accumulation causes phagocytosis defects in glial cells in the brain and contributes to neurodegeneration through elevated oxidative stress and inflammation [14,15], which are two major pathological events in many neurodegenerative diseases including AMD [16][17][18][19]. LD accumulation also impairs phagocytosis in APRE-19 cell lines [7]. Thus, we hypothesized that LD accumulation in RPE may contribute to AMD pathogenesis, possibly by elevating oxidative stress and inflammation.
Confirming the pathogenic role of LD and dissecting the relevant mechanisms have been extremely challenging partially due to a lack of chemical tools to reduce LD and wellaccepted animal models that mimic dry AMD with excessive LD accumulation. This study aimed to investigate the potential pathogenic role of LD accumulation in AMD development. The apoe −/− mice have been used in mimicking AMD for showing thickened BrM and photoreceptor dysfunction [20]. However, the LD accumulation in this model is not fully investigated, particularly the direct role of LD accumulation in the pathology of AMD is still unknown.
Macroautophagy/autophagy is a potent process in the clearance of various substrates including proteins and nonproteins. We have previously established the concept of autophagosome-targeting compounds (ATTEC) as a potential degradation strategy [21,22]. And we expanded our strategy to non-protein degradation by targeting LDs based on the concept of ATTEC [23,24]. In this study, we utilized our recently developed Lipid Droplets·AuTophagy Tethering Compounds (LD•ATTECs) to examine the role of LD accumulation in the pathology of AMD. Our goal is to obtain a comprehensive understanding of the LDs' contribution to the development of AMD and seek a new entry point for AMD early intervention.

apoe −/− mice fed with high-fat diet showed AMD-like phenotypes
apoe −/− mice (8-months-old) were fed with high-fat diet (HFD) for 4 months (apoe −/− -HFD), and the age-matched wild-type mice were fed with normal chew diet (con). We observed a much faster increase of the body weight in the apoe −/− -HFD group than the ones in the control group (Figure 1a-b). The serum cholesterol (TC) level and the serum triglycerides (TG) in the apoe −/− -HFD mice were over fivefold and threefold higher than those of controls, respectively (Figure 1c). TC and TG levels of the RPE/choroidal tissues from apoe −/− -HFD mice were over twofold higher than those in the control mice ( Figure 1d). We further visualized LDs with the LD probe boron-dipyrromethene (BODIPY) in retinal sections [25]. The apoe −/− -HFD mice exhibited significantly increased LDs in multiple layers in retina when compared with control mice (Figure 1e). LD accumulation in apoe −/− -HFD mice was also revealed by fundoscopy, presumably mimicking drusen deposits in AMD (Figure 1f).
The observed LD accumulation in apoe −/− -HFD mice was accompanied by RPE impairment, the initial insult in AMD pathology. RPE flat mounts revealed the loss of normal cuboidal appearance of RPE cells and an increased number of multinuclear larger cells in apoe −/− -HFD mice (Figure 1g,h). And we investigated the loss of RPE cells in apoe −/− -HFD mice by assessing CTNNB1 (catenin beta 1), marker of the RPE adherents complex [26], as measures of epithelial integrity. CTNNB1 signals exhibited increased cytoplasmic and decreased cell boundary location (Figure 1g), confirming the loss of RPE integrity in apoe −/− -HFD mice. Ultrastructure of transmission electron microscopy images in apoe −/− -HFD mice depicted loss of basal infoldings, which form the base to support the inner organelles and are the prominent indicator of RPE polarity (Figure 1j). In addition, vacuoles, large lipid-like granules, more undigested photoreceptor outer segments and swollen mitochondrial were found in the RPE of apoe −/− -HFD mice but nearly absent in the RPE of control mice (Figure 1i-n). Importantly, the mature RPE-characteristic markers including RPE65 (retinal pigment epithelium 65), KRT18 (keratin 18) and MERTK (MER proto-oncogene tyrosine kinase) were significantly reduced in apoe −/− -HFD mice (Figure 1o,p).
To test the functional effects of RPE degeneration, we assessed the retinal function using electroretinogram (ERG). The apoe −/− -HFD mice exhibited a significant reduction of both rod photoreceptor response (a-wave) and postphotoreceptor response (b-wave) compared to the control mice (Figure 1q-r). Taken together, apoe −/− -HFD mice exhibited LD accumulation, RPE degeneration, and retinal dysfunction, all of which mimic the AMD-like phenotypes.

LD·ATTEC treatment attenuated RPE degeneration in apoe −/− -HFD mice
Based on the concept of ATTEC, we recently developed a new LD-degrading strategy, LD·ATTEC, which tethers LD to autophagosome [23]. Thus, we utilized LD·ATTEC (C3) to clear LDs in the mice eyes in order to identify the pathological contribution of LD accumulation to RPE degeneration and subsequent retinal dysfunction. We treated mice intravitreally with C3 or PBS once per week for one month, starting at 3 months after feeding with the high-fat diet. At the end of the fourth month, mice were sacrificed for assessments. Treatment of C3 significantly decreased TC and TG in RPE/ choroidal tissues in the apoe −/− HFD mice compared with the control mice (Figure 2a). We further assessed the lipid reduction by lipidomic analysis in the RPE/choroidal tissues of apoe −/− HFD mice and found that various types of diacylglycerols and triacylglycerol were obviously decreased in the C3 treated group compared to the control ones (Figure 2b,c). The LD clearance could also be demonstrated by evidence of BODIPY staining in retinal section (Figure 2d,e). However, expression of FASN (fatty acid synthase) in RPE/choroidal tissues was not changed after C3 treatment (Figure S1A-C).
We then measured the RPE response after the C3 treatment. Loss of specialized RPE function markers such as RPE65, KRT18 and MERTK were rescued by the C3 treatment (Figure 2f-h). Consistent with the rescue of the RPEcharacteristic proteins, C3 treated mice exhibited the normal RPE morphology (Figure 2i). RPE flat mounts depicted an increased number of RPE cells by C3 treatment (Figure 2j), CTNNB1 staining also exhibited the normal cell boundary location ( Figure 2i). Finally, eyes from apoe −/− -HFD group exhibited much thicker BrM compared to the control group, and this phenotype was also rescued by C3 treatment (Figure 2k). The evidence above demonstrated that C3 mitigated LD accumulation and rescued RPE degeneration, suggesting a causal relationship between these two.

LD•ATTEC treatment ameliorated retinal function defects in apoe −/− mice fed with HFD
In AMD, the neural retina is prone to degenerate and lose its function as a consequence of RPE degeneration. In order to evaluate the effect of C3 on retinal function, C3 or the control compound (GW, which is part of the C3 compound and does not reduce lipids [23]) was intravitreally injected in one eye in apoe −/− -HFD mice. And the other eye was injected with PBS. The GWtreated mice exhibited similar rod responses in the two eyes, whereas the C3-treated mice exhibited much stronger response in the C3-injected eye than the PBS-injected eye (Figure 3a-d).
Functional changes in the retina are associated with changes in the retinal structures. Analyzing the thickness of retinal layers using optical coherence tomography showed that there were no significant changes among different groups, indicating that the global retinal cells loss was not apparent including photoreceptors ( Figure S1 d-f). We then explored the synaptic connectivity between photoreceptors and secondary neurons. Cross-sections of retina were immunolabeled with antibodies against CALB (calbindin), the marker of horizontal cells, and BSN (bassoon), which identifies synaptic ribbons in photoreceptors [27,28]. CALB staining in the retina of control group exhibited a punctate staining pattern in the outer plexiform layer (OPL) (Figure 3e). No significant changes were found in the control mice injected with C3, providing information regarding the safety of C3 in this experimental condition ( Figure 3e). We observed a notable retraction and reduction of horizontal cell neurites in apoe −/− -HFD group by comparison with control animals, with only a few dendrites visible ( Figure 3e). In contrast, apoe −/− -HFD mice treated with C3 exhibited a significant improvement in anatomical organization of dendrites and axons in horizontal cells (Figure 3e). In addition, immunoreactive puncta of BSN was relatively abundant in control retinas across the OPL, which revealed normal synaptic connection between photoreceptors and second-order retinal neurons. Few BSN-positive signals were evident in the retinas of apoe −/− -HFD group, and these changes were rescued by the C3 treatment (Figure 3e-g).
Another characteristic cellular hallmark of AMD is microglial activation [29]. We next stained retina with antibodies against the microglial marker AIF1/IBA1 (allograft inflammatory factor 1). AIF1-positive cells were rare in OPL of control mice, and no significant changes were observed in the number of AIF1positive cells in control mice treated with C3, indicating that C3 does not influence microglia activation in a normal context. However, a substantially increased number of AIF1-positive cells in the OPL were observed in apoe −/− -HFD group. This phenotype was also alleviated by C3 treatment (Figure 3e-h).
These results indicated that the LD·ATTEC mitigated photoreceptor connectivity reduction, microglial activation, rescuing the retinal dysfunction in apoe −/− -HFD mice.

LD·ATTEC lowered LDs and mitigated oxidative stress in ARPE-19 cells
We next investigated the potential mechanism that mediates the protective effect against AMD by LD clearance. Oxidative stress is one of the main causes of neurodegenerative diseases including AMD [30]. LD accumulation is reported to cause elevated oxidative stress in glial cells in the brain [15]. Therefore, we first assessed the potential effects of LDs on oxidative stress in RPE. In this study, LDs were induced in an RPE cell line ARPE-19 by extracellular oleic acid (OA) treatment according to the standard protocol [31]. We used high concentration of OA (400 μM) in this study to trigger oxidative stress. LDs were formed rapidly within the first 4 h and reached equilibrium at about 8 h after treatment, maintaining a plateau within 32 h ( Figure 4a). Accordingly, we treated cells with the LD•ATTECs (C3 and C4) or the control compound (GW) 8 h after induction. We found APOE knockdown (si-APOE) in ARPE-19 cells caused a significant increase of LD after OA treatment, whereas both the treatment with C3 and C4 significantly lowered LDs, with C3 exhibiting a stronger effect ( Figure 4b). Thus, C3 was utilized in the animal experiments as described above and subsequent experiments.
We then investigated whether lowering of LD could influence oxidative stress in ARPE-19 cells. We stained cells by CellRox and MitoSox to measure the total reactive oxygen species (ROS) and mitochondrial superoxide, respectively. In the si-APOE group, both total ROS and mitochondrial superoxide were upregulated compared with the DMSO group (Figure 4c GW, the LC3-binding moiety of LD·ATTEC alone was incapable of lowering LDs (Figure 4c-f). We observed a significant reduction of ROS by the C3 treatment but not the GW treatment, suggesting that degrading LDs mitigated the oxidative stress in ARPE-19 cells (Figure 4c-e).

Oxidative stress and inflammation were increased in apoe −/− -HFD mice and were attenuated by LD·ATTEC treatment
We then investigated whether apoe −/− -HFD mice were subjected to increased oxidative stress. The levels of oxidative damage markers (4-HNE, MDA, 8-OHdG) [32] in RPE/choroidal tissues of apoe −/− -HFD mice were higher than those in the control group (Figure 5a-c). Consistent with this, the antioxidant enzymes SOD1 (superoxide dismutase 1) and HMOX1 (heme oxygenase 1) that protect mitochondrial from superoxide radicals were significantly reduced in apoe −/− -HFD mice (Figure 5d,e). The swollen mitochondrial observed in transmission electron microscopy of apoe −/− -HFD mice also suggested the mitochondria dysfunction (Figure 1k-n), which may contribute to oxidative stress in neurodegenerative diseases including AMD [33]. Besides oxidative stress, inflammation also plays a significant role in AMD pathogenesis [32,34]. Substantial evidence has demonstrated that LDs serve as platform for the synthesis of inflammatory mediators [15,35,36]. We then measured the levels of IL1B (interleukin 1 beta), IL6 (interleukin 6) and TNF (tumor necrosis factor), the proinflammation mediators in RPE/choroidal tissues. ELISA results of these inflammatory cytokines revealed an enhancement of inflammation in apoe −/− -HFD group compared with control group (Figure 5f-h). RT-PCR analysis further confirmed increased expression of proinflammatory genes (Il1b, Il6 and Tnf) in RPE/choroidal tissues from apoe −/− -HFD mice (Figure 5i-k). Treatment with C3 significantly inhibited the increase of oxidative markers (4-HNE, MDA, 8-OHdG) in apoe −/− -HFD mice and increased the protein level of antioxidant enzymes (SOD1 and HMOX1). Moreover, C3 treatment significantly attenuated the inducement of inflammation mediates as evidenced by both ELISA and RT-PCR results (Figure 5f-h).

Discussion
Utilizing LD-targeting and autophagosome-tethering compounds, we demonstrated LD accumulation in apoe −/− -HFD mice caused dry AMD-relevant phenotypes. LD·ATTEC treatment attenuated the AMD-like phenotypes possibly by alleviating the oxidative stress and inflammation. The results in this study demonstrated the causing role of LD accumulation in AMD occurrence and development, suggesting degrading LD as the entry point to prevent AMD.
AMD is a complicated and multifactorial disease whose pathogenesis is far from well understood, leading to a lack of therapeutic drugs. A number of factors contribute to the development of AMD, including aging, smoking, diet and genetics [37][38][39][40]. Multiple pathways including the complement system, extracellular matrix remodeling, inflammation, angiogenesis, and oxidative stress are believed to contribute to its pathogenesis [41]. Currently, various therapeutic approaches are being investigated, such as antioxidant agents, complement inhibitors, neuroprotective drugs, visual cycle inhibitors, gene therapy and cell therapy [42]. In spite of this, there is no meaningful approved treatment for AMD, especially for dry AMD. RPE degeneration is the primary impairment in AMD development and yet its pathological mechanism is largely unknown. And there is currently no available strategy in targeting RPE degeneration in AMD.
AMD is characterized by LD accumulation in the RPE or on BrM, a process known as drusen [43,44]. It is noteworthy that drusen are the first and major clinical signs in AMD patients, much earlier than the vision decline. Drusen leads to reduced blood flow and subsequent hypoxia and metabolic shift in retina, compromising the function of RPE and photoreceptors [45]. We hypothesized that LD accumulation was the initial pathology event in AMD, forming drusen and subsequent RPE degeneration, ultimately causing AMD development.
The relationship between LD accumulation and AMD is currently unknown, partially because of the paucity of animal models for dry AMD with LD accumulation. Since the multifactorial pathogenesis of AMD, there are no well-accepted animal models mimicking AMD. The current murine AMD models are mainly as follows: 1) Multiple genetically engineered mice including multiple genes such as CFH, CCL2, CLIC4, SOD1 [46][47][48][49]. 2) Immunized mice with carboxyethyl pyrrole [50]. 3) Laser-induced choroidal neovascularization [51]. 4) Senescence accelerated mice [52]. However, these models exhibit one or more features of AMD such as the RPE abnormality, focal photoreceptor degeneration, retinal dysfunction and choroidal neovascularization, but LD accumulation is rare.
The major risk factor for AMD is aging. Besides, the risk for AMD has been linked to several lipid-related genes including APOE, LIPC, CETP, and ABCA1 [53]. The widely supported hypothesis about the origin of drusen, the hallmark of AMD patients, is that it comes from RPE which secreted excess lipids from dietary and outer segment, suggesting the relationship of hyperlipidemia with AMD [5]. Some epidemiology studies reported the positive relationship between hyperlipemia and AMD [9][10][11]54,55]. Thus, in this study we utilized a mouse model that combined three known AMD risk factors: aging, hyperlipidemia, and APOE genotype. It is reported that apoe −/− mice exhibited retinal dysfunction at 13months age [20]. In our model, apoe −/− -HFD mice exhibited photoreceptor dysfunction earlier (11-months age), indicating that high-fat diet accelerated the process. In our study, apoe −/ − -HFD mice exhibited abundant lipid in both serum and RPE/ choroid tissues. And RPE degeneration in response to lipid accumulation was observed in apoe −/− -HFD mice. Photoreceptors are among the body's most energyconsuming organs and sustained by RPE. When RPE is incapable of performing this function, it causes photoreceptor dysfunction, which in turn leads to AMD [13]. In this study, apoe −/− -HFD mice exhibited photoreceptor connectivity reduction, microglial activation and retinal dysfunction, while the whole retinal thickness was not changed, suggesting the early pathology of dry AMD phenotype.
Statins, the widely used cholesterol-lowering drugs, are reported to reduce cholesterol synthesis in cultured RPE cells [56]. In some clinical studies, those statins reduce the risk of AMD [57][58][59]. However, others have reported no statistically significant results about the association between statin and AMD development [60][61][62]. The possible reasons of these contradictory results are as follows. Besides the systemic circulation source of cholesterol, RPE secreted locally is possibly the dominant source of cholesterol [56,63]. Whether statins have an ocular effect depends greatly on their ability to penetrate the blood retinal barrier. A statin's lipophilicity affects its ability to penetrate the blood retinal barrier. However, among all the studies published evaluating statin therapy for AMD, they have grouped all statins together. Importantly, triglycerides are among the lipids that are deposited in BrM in addition to cholesterol [43]. Besides, statins lower cholesterol by inhibiting the mevalonate pathway, which also influences endothelial function, the inflammatory and coagulation [64]. And statins are shown to inhibit macrophage activation and reduce the expression of VEGF and C-reactive protein, which are believed to be risk factors for AMD [65,66]. Thus, the effect of statins in AMD cannot be excluded by these physiological processes. Other reported lipid-lowering agents in AMD models including miRNA which target ABCA1, a cholesterol efflux protein in AMD [67]. However, multiple transporter proteins participate in lipid efflux of AMD including ABCA1, ABCG1, ADIPOR1, MFRP [68]. And the multi-target role of miRNA cannot be excluded. Therefore, the direct lipid-targeting agents are needed in investigating the causal relationship of lipid accumulation with AMD.
Strategy for non-protein degradation is challenging because they can't be ubiquitinated using approach such as "PROTAC". Another potent degradation process is autophagy, which is widely distributed in eukaryotic cells with limited selectivity. Hence, harnessing the power of autophagy for degradation may offer new opportunities in new pharmaceutical agents. ATTECs, our recently developed degrading strategy, function in a ubiquitination-independent manner, and tethers the targets with autophagosomes through their direct binding to LC3 [21]. Thus, they could be used to degrade both protein and non-protein targets. Accordingly, we have designed LD•ATTECs, which tether LD to autophagosomes for degradation via autophagy, a known mechanism for LD clearances [23]. LD•ATTECs are able to degrade lipids including both triglycerides and cholesterol [23]. By targeting LDs directly, LD•ATTEC could be used as a potent agent in investigating the relationship between LD accumulation and AMD. In the present study, LD·ATTEC effectively lowered LDs both in ARPE-19 cells and RPE/choroid tissues of apoe −/− -HFD mice. Lipid-lowering by LD·ATTECs improved photoreceptor connectivity reduction and retinal function. As we reported before, LD·ATTECs lower LDs without dramatically changing the proteome and have no obvious influence on proteins involved in lipid synthesis or hydrolysis [23]. In our present study, both the gene and protein levels of lipid synthase were not changed by LD·ATTEC treatment. Additionally, the retinal thickness was not changed after treatment, verifying the safety of the LD·ATTEC under our experimental conditions. To our knowledge, this is the first study to use an effective LD-targeting compound in a dry AMD model, and we have successfully rescued the AMD-like phenotypes in mice with this compound. Since drusen are not only a clinical hallmark for AMD, but also for other dense deposit diseases including glomerulonephritis [69][70][71]. Notably, neurodegenerative diseases including Alzheimer disease and Parkinson disease, exhibit LD accumulation in disease pathology [72]. Our strategy could possibly be expanded to these diseases. Since autophagy is capable of degrading various types of substrates such as proteins and non-proteins like organelles, pathogens, and nucleic acids. Theoretically, ATTEC could expand the scope of degradation strategy.
To investigate the mechanism underlying LD-lowering's protective effect on apoe −/− -HFD mice. We focused on oxidative stress and inflammation, which play the central role in neurodegenerative diseases including AMD. LD accumulation is reported to cause oxidative stress and inflammation in glial cells in brain, and may contribute to neurodegenerative diseases including AMD [15]. Oxidative damage markers, for example, 8-OHdG, correlates well with drusen, and is elevated globally in ocular tissues in AMD patients [32,34]. In AMD, the overreactive microglia would release various proinflammatory mediates, including IL1B, IL6 and TNF [29]. Accumulated oxidative stress and inflammation during aging contribute to RPE dysfunction and degeneration [50]. We reported here that oxidative stress was enhanced in apoe −/− -HFD mice as evidenced by the increased oxidative damage marker (4-HNE, 8-OHdG, MDA) and the reduced antioxidant enzyme (SOD1 and HMOX1). Besides, microglial cells were activated and the pro-inflammation mediates (IL1B, IL6 and TNF) were upregulated in RPE/choroidal tissues, indicating the increased inflammation in apoe −/− -HFD mice. Following treatment by LD·ATTEC, oxidative damage markers and pro-inflammatory cytokines decreased, and antioxidant enzymes increased. Similar results have been reported that LDs reduce phagocytosis and increase oxidative stress in ARPE-19 cells [7]. Van Den Brink et al. have also reported that the inhibition of LD accumulation in Drosophila suppresses photoreceptor degeneration by lowering ROS [73]. Our observations complemented these findings in AMD mice model.
However, there are some disadvantages to intravitreal injection, including the invasive and repeated procedure. Thus, the lesion-targeting strategy could be pursued in the future. The pharmacokinetic and pharmacodynamic properties need to be improved before clinical use, possibly achieved by selective LD-binding warheads, higher affinities and smaller size.
In summary, this study utilized LD-targeting and autophagosome-tethering compounds to identify the causal relationship between LD accumulation and AMD, providing entry points for the treatment of the initial insult of AMD by degrading LD.

Reagents and cell culture
Listed below are regents in this study: DMSO (Sigma-Aldrich, D2650), GW5074 (Selleck, S2872), OA (Sigma-Aldrich, O7501), BODIPY (Thermo Fisher Scientific, D3922). ARPE-19 cells were purchased from ATCC and cultured according to generally accepted methods. The cultured medium was Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12 supplemented with 10% fetal bovine serum (Thermo Fisher Scientific, 11330032). The treatment was similar to what we described previously [23]. Briefly, ARPE-19 cells were plated at 50% confluence and incubated for 24 h before treatment. Then OA (400 μM) was added to induce LDs in cells, and 10% BSA (Sigma-Aldrich, B2064) was used as the control. C3 (10 μM) and the control compound (GW5074, 10 μM) were added into the medium 8 h after OA induction for another 24 h. Antibodies and primers used in this study were listed in tables.

Immunofluorescence and BODIPY staining
For BODIPY staining in vitro, the procedure was the same as we described previously [23]. For LD measurements using Incucyte (Essen Bioscience, IncuCyte S3, USA), the images were automatically captured every 0.5 h in a 96-well plate. The quantification was performed by the Incucyte Analyzer software. For ROS staining, probe-containing culture medium (2.5 μM; CellRox Deep Red; Thermo Fisher Scientific, C10422) was added to the dish after discarding the cell culture medium, then the cells were incubated for 20 min (37°C). Following that, probe-containing culture medium (5 μM; MitoSox Red; Thermo Fisher, M36008) was added and cells were incubated for 10 min (37°C). Next, Hoechst 33,342 (Thermo Fisher Scientific, H3570) was added and incubated for 5 min (37°C). The medium was discarded and cells were rinsed with live cell imaging solution (Thermo Fisher Scientific, A14291DJ) three times. Images were immediately taken by confocal scanning laser microscope (LSM 880, Zeiss, Germany).

RPE flat mount staining
Eyes were fixed in 4% paraformaldehyde (Sigma-Aldrich, 158127) in 0.1 M phosphate buffer, pH 7.4 (PB) for 0.5 h before dissection. RPE/choroid tissues were separated from retinas and fixed for 2 h. Tissues were blocked in 1% BSA (Bio Basic, A600332) with 0.05% Triton X-100 in PBS for 1 h in room temperature. We incubated RPE/choroid tissues with CTNNB1 (listed in Table 1) overnight at 4°C. Following that, tissues were washed and incubated in the secondary antibody (listed in Table 1) and phalloidin (1:1000; Abcam, 176753) for 1 h at room temperature in the dark. Followed by washing in PBS (Cytiva, SH30256.01) for three times, the sections were carefully mounted as described above. Images were taken using a Leica TCS SP8 confocal laser-scanning microscope. Four images from the central zone of RPE flat mount were taken randomly from five flat mounts per group.

Erg
Animal experiments in this study were performed adhering to the NIH guidelines for care and use of laboratory animals and with the approval of the Animal Care and Use Committee of Eye & ENT Hospital. B6/JGpt-Apoe em1Cd82 /Gpt mice (apoe −/− -mice; GemPharmatech, T001458) and C57BL/6JGpt (control mice; GemPharmatech, N000013) were used in this study. apoe −/− mice were fed with HFD (Research Diets, D12492) for 3 months and age-matched control mice were fed with normal chow for 3 months, then one eye was intravitreally injected with C3 or GW and the other eye was treated by PBS (Cytiva, SH30256.01) once a week for 4 weeks. The retina function was detected by ERG. After overnight dark adaptation, mice were anesthetized under dim red light, pupils were dilated using 1% atropine sulfate. Under illumination with dim red light, gold ring electrodes were placed on the central cornea. The grounding electrode was placed subcutaneously on the tail and the reference electrode was inserted into the chin of mice. Anesthetized animals were maintained at 37°C in absolute darkness. A series of stimulus intensities (0.01, 0.1, 1.0, 3.0,10.0, 30.0 cd•s/m 2 ) was applied for dark-adapted ERG. Five consecutive stimuli were recorded and averaged for each light level. Stimuli administration and data collection were performed using Espion Electrophysiology System (Diagnosys LLC, USA).

Measurements of TC and TG
The procedure was the same as we described previously [23]. The TG and TC levels in the serum and RPE/choroidal tissues were determined based on instructions from the manufacturer (TG: Nanjing Jiancheng Bioengineering Institute, A110-1-1; TC: Nanjing Jiancheng Bioengineering Institute, A111-1-1).

Statistical analysis
Statistical comparisons between two groups were conducted by the unpaired two-tailed student's t-test. Statistical comparisons among multiple groups were conducted by one-way ANOVA tests and post hoc Tukey's test for the indicated comparisons. All statistics were performed in GraphPad Prism 9 software. Significance was established at p < 0.05.