Cholesterol homeostasis in the vertebrate retina: Biology and pathobiology

Cholesterol is a quantitatively and biologically significant constituent of all mammalian cell membrane, including those that comprise the retina. Retinal cholesterol homeostasis entails the interplay between de novo synthesis, uptake, intra-retinal sterol transport, metabolism and efflux. Defects in these complex processes are associated with several congenital and age-related disorders of the visual system. Herein, we provide an overview of the following topics: a) cholesterol synthesis in the neural retina; b) lipoprotein uptake and intraretinal sterol transport in the neural retina and the retinal pigment epithelium (RPE); c) cholesterol efflux from the neural retina and the RPE; and d) biology and pathobiology of defects in sterol synthesis and sterol oxidation in the neural retina and the RPE. We focus, in particular, on studies involving animal models of monogenic disorders pertinent to the above topics, as well as in vitro models using biochemical, metabolic, and omic approaches. We also identify current knowledge gaps as well as opportunities in the field that beg further research in this topic area.


INTRODUCTION
Sterols represent a diverse class of biologically significant lipids that are found ubiquitously in all eukaryotic cells, primarily in the plasma membrane (1). Cholesterol is, by far, the dominant sterol normally found in mammalian cells and tissues. Maintaining optimal levels of cholesterol is requisite for normal cellular function and viability, and represents a delicate balance between endogenous de novo synthesis, exogenous uptake, and efflux of sterols. Mechanisms governing the process of cholesterol homeostasis have been investigated extensively, given the role they play both in normal biology as well as in several significant human clinical disorders, such as Alzheimer's disease, cardiovascular disease, and age-related macular degeneration (AMD) (2).
Herein, we review the mechanisms governing cholesterol homeostasis in the neural retina, and the pathological mechanisms that underlie certain ocular diseases where this homeostasis is disturbed.
Cholesterol in the central nervous system (of which the retina is a part) exists mostly in the unesterified form, in nerve myelin and in the plasma membranes of both neuronal and nonneuronal (e.g., glial) cells. The brain relies exclusively on its own de novo synthesis of sterols, since the blood-brain barrier excludes circulating lipoproteins (3). By comparison, sterol homeostasis in the retina is somewhat more complex, since its sterol pool is derived from both local de novo synthesis and extraretinal uptake. The retina has many advantages for experimental studies, such as its layered organization and ease of accessibility, and represents one of the beststudied parts of the central nervous system. Cholesterol and its biogenic sterol precursors can undergo both enzymatic and non-enzymatic oxidation, which is particularly relevant given the prooxidative environment of the retina, yielding a variety of oxysterol products, some of which are highly toxic to cells (4,5). Sterols and sterol metabolites may play causative roles in several neurodegenerative conditions, including certain retinopathies (5)(6)(7). by guest, on January 12, 2021 www.jlr.org

Downloaded from
In part, this review is an extension of a prior review in a similar Thematic Issue series in this journal published a decade ago (8). A subsequent review by other authors highlighted the role of sterol homeostatic processes in AMD (9). The scope of this review encompasses the following We will focus on studies that, for the most part, have been conducted using either laboratory animals (e.g., mice, rats) or animal-derived cells in vitro, augmented by the use of selective inhibitors of enzymes in the cholesterol synthesis pathway or by genetic manipulation of the genes encoding those enzymes. While these experimental systems are admittedly highly simplified compared to the complexity of human ocular anatomy and physiology, and cannot fully model multifactorial disorders such as AMD, they nonetheless have provided fundamental insights into cholesterol homeostasis in the vertebrate retina. In addition, the knowledge derived from such models has provided potentially useful tools for developing effective therapeutic interventions for human diseases that are caused by defects in the cholesterol pathway that impact the structure and function of the retina. by guest, on January 12, 2021 www.jlr.org Downloaded from detailed analysis of the specific role of HDL in age-related retinopathies (30).] This review discusses the role of sterol efflux in retinal physiology and pathophysiology.
LXR activation stimulates sterol efflux by upregulating the expression of ABC transporters (36,37). Mutations in a critical mitochondrial CYP450 enzyme, CYP27A1, causes Cerebrotendinous Xanthomatosis (CTX; OMIM# 213700), an autosomal recessive disorder clinically characterized by chronic diarrhea, neuronal dysfunction, early-onset of atherosclerosis, and visual system deficits such as cataract formation (38,39). The CYP450 enzymes act on a broad spectrum of substrates including sterols and oxysterols, generating biologically important metabolites.

Insights into cholesterol synthesis and uptake in the neural retina
Early in vitro experiments by Fliesler and Schroepfer, using either bovine retina cell-free homogenates (10,000 x g supernatant fraction that contains microsome and cytosol) (45) or intact whole bovine retinas in organ culture (46) (47,48). The neural retina showed [ 3 H] cholesterol formation within 6 h, with little accumulation in intermediates, and its formation was fully inhibited upon co-injection with lovastatin (47)(48)(49). In the same study, inhibiting the post-squalene phase of the pathway using NB-598 (an inhibitor of squalene 2-3 epoxidase (SQLE; Fig. 3 [At the time, this was a striking finding, because it had been assumed that the blood-retina barrier was similar to the blood-brain barrier in that it would exclude the uptake of blood-borne  (81), possibly due to the requirement for de novo non-sterol isoprenoid synthesis, e.g., dolichols (49). Further, a recent in silico modeling of retinal sterol homeostasis suggests that photoreceptors acquire sterol from exogenous sources, rather than mevalonate pathway (85).
Hence, taken together, these results suggest that de novo synthesis of cholesterol in photoreceptors, per se, is likely minimal. This is curious and unexpected, given the prodigious daily rate of membrane assembly and turnover of photoreceptor outer segment membranes, which requires a continuous supply of lipids (including cholesterol) as well as proteins (13,86).
Also, it should be appreciated that a significant level of lipid (including cholesterol) synthesis is

Intraretinal cholesterol exchange: Role of Müller glial cells
Intravitreal injection of radiolabeled amino acids, followed by tracking the fate of newly synthesized retinal proteins by light and EM-level autoradiography, first demonstrated the diurnal process of photoreceptor outer segment renewal and the intimate involvement of RPE cells in this process (13,86). A similar approach was adapted to demonstrate the continual the synthesis and distribution of Apolipoprotein E (ApoE), a signature protein associated with VLDL and IDL particles, in the rabbit neural retina, and in a primary Müller glial cell culture model (88).
Radiolabeled amino acids were faithfully incorporated into ApoE within 3-6 h following intravitreal injection. SDS-PAGE autoradiography of immunoprecipitated ApoE showed its presence mostly in the vitreous and in the neural retina, with little incorporation in the optic nerve (88). This agrees with the results of neuron-glia co-culture studies, which showed that cholesterol, but not ApoE, is required for retinal ganglion cell synaptogenesis (65,66). A study using ApoE knockout mice reported electrophysiological deficits, accompanied by possible dropout of Müller glial cells, by postnatal 25 weeks of age (95). In another study, the lack of ApoE also reportedly lead to an appreciable (three-fold) increase in retinal cholesterol content  (96). In the CNS, a major receptor for ApoE is LRP1 (LDL-related particle 1), which is required for synaptogenesis, oligodendrocyte progenitor cell differentiation, and myelination (97,98). LRP1 is expressed in primary RPE cells (99) This is due to the inability to "tag" the de novo synthesized sterol with a fluor and follow its trafficking, secretion and uptake, since sterols (unlike proteins) are not coded by genes. However, an alternative approach might be targeted deletion of enzymes of the post-squalene pathway in Müller glia, such as SC5D, DHCR24, or DHCR7, and follow-up assessment of uptake and incorporation of the biogenic cholesterol precursor into neighboring retinal neurons (e.g., photoreceptor cells) (See Fig. 4).

Mevalonate pathway activity in the retinal pigmented epithelium (RPE)
The above results pertain only to de novo cholesterol synthesis and uptake in the neural retina.
We will now specifically consider the mevalonate pathway in the RPE. While immunohistochemical analysis has shown the presence of HMGCR in human and murine RPE cells (61,80), investigating RPE cholesterol synthesis rates in vivo is extremely challenging, due to the technical difficulties involved in the metabolic approaches, and the need for targeted cell type-specific inhibition of the mevalonate pathway. As a potentially more tractable and fruitful alternative, RPE in vitro models of genetic diseases pertaining to the mevalonate pathway and related pharmacological models as well as the conventional metabolic approach may be of utility    (119).

Sequestration of cholesterol in storage depots
A key mechanism in cellular cholesterol homeostasis is the esterification of excess free cholesterol and its storage in lipid droplets (120,121). Three enzymes catalyze the esterification of acetyl-CoA acetyltransferases (ACAT1 and ACAT2) and lecithin:cholesterol acyltransferase (LCAT) (122). Both human and macaque retinas express ACAT1 and LCAT (79,123). ACAT1 immunolocalization also has been reported in the murine retina in the photoreceptor outer segment layer, outer plexiform layer, ganglion cell layer, and in the RPE (80). However, LCAT appears to localize to Müller glial cells, rather than to retinal neurons (124). Also, LCAT ablation does not lead to retinal degeneration or dysfunction (125). Inhibition of cholesterol efflux, such as observed in CYP27A1/46A1 double-knockout mice, with resultant increase in photoreceptor cholesterol content, exhibits ACAT1-dependent esterification of sterols in photoreceptors, notably in their outer segments (124). The latter finding is curious, considering the fact that, historically, extensive lipid composition analyses across multiple vertebrate species have failed to detect such molecules in purified photoreceptor outer segment membrane preparations (11).

Role of ATP binding-cassette transporters in retinal cholesterol efflux
ABC transporters ABCA1 and ABCG1 generally account for the majority of cellular sterol efflux, depending on tissue expression levels (24). ABCA1 and ABCG1 are expressed widely in most tissues, including brain, retina, and macrophages (31), while ABCG4 is predominantly expressed in the brain (126). ABCG1 additionally caters to efflux of sterols derived from OxLDL to HDL particles in macrophages, and plays a protective role in atherosclerosis (127) Curiously, the onset of retinal dysfunction upon inhibition of cholesterol efflux at the level of LXRα/β, or ABC transporters (also CYP hydroxylase knockouts-discussed later) leads to a slow retinal degeneration phenotype that manifests by about 1 year. This is unlike inhibition of retinal de novo cholesterol biosynthesis either through statin treatment (which most likely is due to protein prenylation compromise, rather than blockade of sterol synthesis), or as observed in the AY9944induced model of SLOS (see below), which leads to retinal degeneration on a timescale of weeks (75,(130)(131)(132). In sum, these observations independently indicate significant differences in the rates of retinal sterol synthesis and turnover, in agreement with the results of previous studies (59,78).

Cholesterol efflux in the RPE
ABCA1 and ABCG1 expression has been reported for the RPE, in addition to the neural retina (80,128,133  rats with Voriconazole, a CYP46A1 inhibitor, did not lead to retinal degeneration or altered darkadapted ERG responses (157). However, intraperitoneal injection of Voriconazole led to a significant decrease in retinal 24S-OH-Chol levels without concomitant changes in brain or serum levels within 5 days (158). This is very surprising, given that CYP46A1-synthesized 24S-OH-Chol is the predominant cholesterol elimination product in brain unlike the retina, which strongly depends on CYP27A1-dependent metabolism for cholesterol efflux (153,154,159,160).

OX-LDL AND OXYSTEROLS ON THE NEURAL RETINA AND THE RPE
We have previously discussed the critical role oxysterols play as LXR agonists, thus facilitating ABC transporter-mediated cellular cholesterol efflux. Therefore, oxysterols act as regulators of cholesterol biosynthesis and metabolism (163,164). Elevated oxysterol levels in membranes leads to organelle dysfunctions like mitochondrial dysfunction, ER stress, and lysosomal membrane permeabilization (165). High-intensity light exposure leads to non-enzymatic lipid peroxidation and generation of 4-hydroxynonenal (4-HNE, a by-product of oxidation of omega-6 PUFAs) and, subsequently, oxidatively-modified retinal proteins (166). Similar conditions can promote the non-enzymatic oxidation of cholesterol and other sterols, generating cytotoxic oxysterols (167)(168)(169)(170)(171). In addition to conditions that cause photo-oxidative stress, alterations in iron homeostasis also have been shown to promote lipid peroxidation (172,173), including in the retina (174,175). The chemistry of oxysterols and their biophysical effects on the plasma membrane have been discussed above, as well as in other reviews (176,177). We will now discuss the uptake mechanisms of OxLDL and the biological effects of oxysterols and OxLDL in the RPE and the retina.

Treatment of RPE-J cells with
Cholesterol homeostasis of the outer retina, especially as pertains to sterol and lipid uptake and efflux across the RPE, plays a critical role in the pathogenesis of AMD (85,204). While rodent models serve as excellent biochemical systems to investigate sterol homeostasis and related monogenic disorders, the notable lack of a cone-rich fovea in rodents and other non-primate species presents a challenge in reliable modeling AMD in laboratory animals. Key insights into the pathogenesis of AMD arises from lipidomic and proteomic analysis of drusen deposits obtained from human AMD patient donor eyes (205). The major lipid constituents of sub-RPE drusen deposits include esterified cholesterol, sphingomyelin, and phosphatidylcholine (205,206 (138,208). These results are consistent with the proposed role for RPE sterol efflux in AMD pathogenesis. 7KChol levels in RPE-choroid of primates and humans increase with age, and is the major oxysterol constituent of drusen (195,209). Understanding the origins of drusen lipid content has been aided by analysis of the accompanying proteome (205).
Major peptide constituents of drusen include ApoE, ApoB, serum albumin (arising from blood), and also proteins possibly of RPE/retinal origin, such as complement factors (CFH, C3, C5), TIMP3, crystallins, and ApoA1 (205,210). These lipidomic and proteomic findings suggest that serum LDL and outer retina sterol efflux both contribute to the formation of drusen deposits. The drusen proteome also was found to contain lipid peroxide adducts, suggesting the involvement of oxidative stress in drusen formation (210)(211)(212). Although a recent large-scale study suggests a lack of correlation between AMD patient serum OxLDL levels and drusen formation (213), this cannot rule out a critical role for retinal oxidative biology in AMD pathogenesis. This is because simple ELISA assays to quantify serum OxLDL levels do not detect individual oxysterol species synthesized locally due to chronic oxidative stress in the retina (212,214). The latter is clearly evidenced by the accumulation of 7KChol and lipid peroxide adducts detected in drusen, and from the described biological effects of oxysterols on the RPE in vivo. These findings suggest that the oxysterols observed in drusen may arise from lipid peroxidation of already-formed drusen, rather than deposition of OxLDL.  (216). Cholesterol is required for phagosome retrograde trafficking (218,219). Phagosome maturation defects were observed in an iPSC-derived RPE in vitro model of SLOS, characterized by increase in cellular 7DHC levels (76). The sluggish degradation of phagocytized OS observed in the SLOS RPE model is also seen in genetic and pharmacological animal models of SLOS (76). The RPE in SLOS animal models also accumulates undigested OS and lipid droplets, and exhibits increased lipofuscin and A2E content (75,76). It should be noted that in SLOS in vivo models, sterol homeostasis in the RPE is altered both by the de novo synthesis of 7DHC and 7DHC-derived oxysterols, and their uptake via receptor-mediated endocytosis of LDL and OxLDL. However, rodent and in vitro experiments investigating the effects of A2E on RPE have made two fundamental assumptions: 1) that the RPE is homogeneous across the eye, and 2) that A2E is homogeneously distributed across the eye; hence, the presumption that data obtained from whole-eyecup cell isolations and lipid extractions are meaningful. These conditions may be true for rodents (220), but likely not for human and other foveated species (220)(221)(222)(223). Recently, foveal RPE cells isolated from human eyes have been determined to have unique properties compared to non-foveal RPE cells (224,225).

OxLDL and oxysterols in the neural retina
Lipid peroxidation in rod OS membranes occurs non-enzymatically by formation of free radicals through ferrous (Fe 2+ ) ion-mediated Fenton reaction (226). In vitro assays have demonstrated 7KChol has been implicated in several age-related disorders (44,164,247). The predominant oxysterol found in the OxLDL formed in vivo (generated by copper-or iron-catalyzed oxidation of LDL) is 7KChol; other significant oxysterols include 7αβ-OH-Chol and 5,6α/β-epoxy-Chol (248).
In vitro studies on the effects of exogenous OxLDL utilizes copper-or iron-catalyzed oxidation of LDL to generate OxLDL. The types of oxysterols found in the neural retina of albino rats subjected to bright light conditions suggest involvement of the Fenton reaction in their formation (248). By contrast, oxysterol levels are minimal in control animals (not subjected to intense light exposure), and are attributable to basal levels of enzymatic (e.g., CYP27A1) and/or non-enzymatic sterol oxidation. The immunolocalization of 7KChol in photodamaged retinas is qualitatively comparable to that of H-and L-ferritin (a local, endogenous source of ferrous ions) in the neural retina, i.e., the inner segment, inner plexiform, and the ganglion cell layers (248)(249)(250)(251). The lack of 7KChol metabolites in photodamaged retinas (at 48 h) may be due to the relatively rapid rate of mitochondrial and extramitochondrial generation of 7KChol, compared to the rate of its mitochondrial metabolism by CYP27A1, consistent with the relatively slow retinal sterol turnover rate, and the rate of cholesterol accumulation in the CYP27A1 knockout mouse (203,248,249,252). Elevated 7KChol levels have been observed in the RPE-choroid of rats with laser-induced choroidal neovascularization (CNV); the CNV was dramatically prevented by pre-treatment with sterculic acid (196). In a knockout mouse model of hemochromatosis (a recessive human "iron overload" disease), cholesterol and oxysterol accumulation has been attributed to decreased Age-dependent accumulation of 7KChol in murine and monkey ocular tissues has been observed predominantly in Bruch's membrane and choroid, and induces VEGF formation through an NFk-B-mediated pathway (195,209,257). 7KChol in Bruch's membrane serves as a chemoattractant, promoting microglial/macrophage migration, infiltration, and activation (197,203,257). Retinal microglia undergo cell death at the same dose range of 7KChol as other retinal cell types (255,257). 7KChol-induces lipid droplet accumulation in microglia, as well as their polarization to the pro-inflammatory M1 state (257). In vitro studies suggest that CD36-mediated 7KChol and OxLDL uptake induces inflammasome formation in endothelial cells and in retinal pericytes, the cellular components of the inner blood-retinal barrier (257,258). LOX1 OxLDL receptor is expressed in

Lessons learned from animal models with pharmacological disruption of cholesterol synthesis
To date, only few animal models exhibit altered cholesterol homeostasis with chronic elevation of oxysterol and OxLDL levels. Serum OxLDL levels (specifically as involves 7KChol) are elevated in Ldlr -/and ApoE -/mice, which offer tractable animal models for studying atherosclerosis (265,266). Elevated intraocular pressure (modeling glaucoma) induces increased retinal CYP46A1 activity, with resultant increase in 24-OH-Chol levels (267). However, none of these models exhibits chronic elevation of oxysterol levels, as is observed in diseases such as SLOS. The AY9944-induced rat model of SLOS has provided some key insights into the role of oxysterols (and possibly, by inference, OxLDL) in retinal degeneration (75). A unique feature of 7DHC, which accumulates in all bodily tissues and fluids of human SLOS patients and in SLOS animal models, is that it is the mostly readily oxidizable organic molecule known (74). 7DHC undergoes enzymatic and non-enzymatic (free radical-induced) oxidation, at rates 200-fold faster than cholesterol, and even seven-fold faster than DHA oxidation (74). This explains the formation and buildup of lipid peroxides and 7DHC-derived oxysterols in various tissue, including retina, brain, liver, and blood, in the SLOS rat model, and patient samples (68,261,268).

OXYSTEROL DISTRIBUTION
Early investigations of tissues, including the retina, utilized freeze-fracture electron microscopic analysis of filipin-induced "pit" formation in cell membranes to determine the distribution of spatially-distinct, cholesterol-rich membrane domains (subsequently thought to equate to "lipid rafts") (286)(287)(288). In addition, the ability of digitonin to form one-to-one molecular complexes with cholesterol, followed by treatment with electron-dense contrast reagents, was leveraged to enable observation of sterol distribution by electron microscopy (289,290). Filipin is also an endogenously fluorescent molecule (λexc = 360 nm, λem = 480 nm) (291), which has prompted its use as a probe to study cellular cholesterol distribution using fluorescence microscopy, although its utility is somewhat limited by its susceptibility to rapid photobleaching. Alternatively, fluorescent analogs of cholesterol, such as cholestatrienol or BODIPY-derivatized cholesterol, can be utilized to study cholesterol trafficking and distribution, but require exogenous addition of the fluorescent analog. Similarly, fluor-tagged choleratoxin-B, which binds to GM1 (a ganglioside enriched, along with cholesterol, in lipid rafts) may be used to analyze lipid raft distribution (291). However, these strategies cannot be utilized for live-cell imaging, due to the cytotoxic effects of these reagents.
Hence, the information obtained from the more classical approaches mentioned above is limited to spatiotemporal snapshots of cholesterol distribution in cells and tissues at a given point in time, rather than affording dynamic monitoring of cholesterol trafficking and distribution in real time.
Perfringolysin O (PFO) is a membrane pore-forming bacterial Ɵ-toxin that binds to membrane cholesterol through its carboxy terminal D4 domain (292). Biotinylated PFO, in conjunction with streptavidin-gold, has been used to monitor cellular cholesterol distribution by electron microscopy (293). Fluor-tagged PFO D4 peptide also allows for monitoring of cholesterol distribution on the exofacial leaflet of cell membranes (294). Recently engineered chimeric D4-based probe (mutant D4 domain of PFO, D4H, tagged with mCherry) is available as expressible, non-cytotoxic probe, which recognizes cholesterol on the cytosolic face of the membrane (294)(295)(296). Sterol distribution in cells also may be investigated using biorthogonal analogs of cholesterol, i.e., analogs which can be integrated into a biological system without altering its natural biology, and allows monitoring of esterified and free cholesterol without cholesteryl esterase treatment, unlike filipin or mCherry-D4H. An alkyne analog of cholesterol, C19-alkyne cholesterol (or simply "eCholesterol") is amenable to high-resolution microscopy imaging upon derivatization with a fluorescent azide using copper(I)-catalyzed alkyne-azide cycloaddition (297).
Similar biorthogonal oxysterols, such as 25-OH-[C19-alkyne]-cholesterol (or simply, 25-OH-eCholesterol), have been developed, allowing for monitoring cellular distribution of oxysterols of interest (297,298). Adaptation of such engineered probes, or "click chemistry" in vitro in primary RPE and retinal neurons provides insights into the role of cholesterol in specialized cellular processes, such as RPE phagocytosis and neuronal synaptogenesis.

CONCLUDING REMARKS
To date, the fundamental processes involved in establishing and maintaining cholesterol homeostasis in the vertebrate retina appear to be fairly well-understood, particularly under normal physiological conditions (See Fig. 3)   [Modified and adapted, with permission, from (299)].