retinal adhesion. Recent insight into the role of αvβ5 integrin in retinal adhesion gained by exploring RPE pigment and marker protein fractionation and localization in β5 integrin knockout mice has demonstrated an experimental strategy that may prove useful in identifying additional receptor proteins of both RPE and photoreceptor and their subretinal ligands in the near future. Our better understanding of the molecular mechanisms of retinal adhesion are much needed to develop new therapeutic strategies aiming to prevent irreversible vision loss following retinal detachment by providing a microenvironment in the subretinal space that maximally promotes recovery of both damaged photoreceptors and RPE cells.

PHOTORECEPTOR OUTER SEGMENT RENEWAL

Physiology of Outer Segment Disk Assembly and Disk Shedding

Photoreceptor neurons normally do not renew themselves in the adult mammalian retina. However, Bairati and Orzalesi first hypothesized in 1963 that both rods and cones continuously turn over their outer segment portions, replacing the membrane disks that carry rhodopsin/cone opsins and the phototransduction machinery [31]. Since Young in 1967 provided the first direct evidence for such a process, now termed outer segment renewal [32], such de novo formation of outer segment disks has been extensively studied in numerous species [33, 34].

Photoreceptor cells synthesize outer segment resident proteins and lipids mainly in their inner segment. Following translation in the rough endoplasmic reticulum, nascent polypeptides mature while passing through the Golgi apparatus and traffic to the outer segment from the trans-Golgi network packaged in transport vesicles [35, 36]. Vesicles traverse the connecting cilium to contribute to the assembly of new membrane disks by evagination at the proximal end of the outer segment [35, 37]. Cones principally differ from rods with respect to protein turnover: Cone disks do not separate from each other, allowing diffusion of newly synthesized and assembled proteins within the length of the outer segment [38, 39]. In contrast, individual rod disks vary in age, with those in the proximal end of the outer segment most recently generated and those at the distal tip facing the RPE most aged.

Outer segment protein biosynthesis follows a daily rhythm that is in part regulated at the transcript level, as amounts of rhodopsin and cone opsin messenger RNAs (mRNAs) are maximal at the onset of dark [40–42]. Circadian rhythms upregulate gene transcription in chick cone photoreceptors to raise iodopsin (red cone opsin) mRNA levels prior to the onset of the dark period [43]. It has not yet been reported whether rates of transcription or transcript stability increase to reach higher steady-state levels of specific mRNAs in mammalian retina. In amphibian retina, outer segment disk assembly does not occur evenly at all times but is regulated by illumination and circadian rhythms [33, 44, 45]. In contrast, illumination has little effect on rates of disk assembly in mouse rods [46].

Photoreceptor disk assembly is generally thought to be largely photoreceptor cell autonomous and independent of photoreceptor–RPE interactions. However, in the healthy retina, production of new POS disks is in a precise balance with elimination of most aged portions of outer segments to maintain constant length of outer segments with time. Young and Bok (1969) first demonstrated the disposal of distal, most aged

rod outer segment tips in packets of uniform size at the interface of outer segments with the RPE in frog retina, a process that became to be known as disk shedding [47]. Unlike disk assembly, shedding of POS distal fragments most likely requires active participation of directly underlying RPE cells. RPE apical microvilli ensheathe rod and cone outer segments, and their extensions can be observed in between shedding POS and the remainder of the outer segment [48]. Moreover, outer segments cease to shed their distal tips in detached Xenopus laevis retina [49].

Rod POS shedding is highly synchronized in the retina and occurs at light onset in most species, regulated by circadian rhythms in mammals [50, 51], by illumination in frogs (Rana pipiens) [52], and by a combination of both in Xenopus laevis [53]. In the mouse retina, each rod photoreceptor sheds a packet of disks comprising approximately 10% of its total length once a day [46]. Similar rates of rod shedding are documented for other species.

Newly synthesized protein distributes diffusely throughout the length of the outer segment of photoreceptor cones [54]. Therefore, distal cone outer segment disks do not contain most aged outer segment components as distal rod disks do. Nevertheless, cones shed distal tips of their outer segments in similar size fragments to rods regardless of species and retinal location [55]. In lizard, goldfish, and chick retina, cone photoreceptors shed after the onset of dark [56–58]. In ground squirrel retina, cones shed either following the onset of dark or in the middle of the dark period [59]. In the cat [60] and in the diurnal Nile rat [61] retina, cones, like rods, shed following the onset of light. Taken together, cones shed distal tips of their outer segments in synchronized bursts like rods, but there is considerable variability among species with regard to the timing of cone shedding.

Physiology of RPE Engulfment of Shed Outer Segment Fragments

The synchronicity of rod disk aging and shedding facilitated autoradiographic tracer studies that allowed Young and Bok in 1969 to demonstrate that RPE cells phagocytose shed rod POS [47]. Furthermore, the Royal College of Surgeons (RCS) rat strain has long provided an animal model of hereditary retinal degeneration caused by abnormal accumulation of POS in the subretinal space [62]. Mullen and LaVail generated and studied chimeric rats with mosaic RCS and wild-type RPE to demonstrate that the presence of wild-type RPE underlying RCS photoreceptor rods was sufficient to promote normal POS turnover. Therefore, POS accumulation in the RCS retina occurs as a consequence of deficiencies in the RPE rather than in photoreceptors [63]. The rapid and complete degeneration of the retina in the RCS rat illustrates the importance of the RPE and its phagocytic activity in POS renewal specifically and in retinal homeostasis in general.

The rhythmic and continuous nature of POS phagocytosis in the retina renders RPE cells the most active phagocytes in nature. The enormous task of the RPE cells in POS disposal becomes immediately obvious when one considers that each RPE cell in the mammalian retina faces numerous photoreceptor rods (e.g., ~45 in the peripheral rhesus monkey retina and a staggering 300 in the rat retina; [64]), each one of which sheds its distal tip containing about 100 disks every morning. Each RPE cell must therefore completely dispose of several thousand outer segment disks before its next phagocytic challenge. Since RPE cells do not turn over in the adult mammalian eye, each individual cell must clear its enormous phagocytic load promptly and efficiently every 24h over many decades.

The RPE cells in the healthy mammalian retina respond to circadian rod POS disk shedding with a vigorous phagocytic response and then cease phagocytosis until the next shedding event. Indeed, electron microscopy images of the retina–RPE interface acquired from samples of tissues harvested at times of POS shedding usually reveal POS that still remain attached to the outer segment or POS that were already internalized by the RPE. Observation of shed POS that reside in the subretinal space awaiting RPE engulfment is very rare. This suggests that RPE cells may be optimally prepared at light onset to respond to shed POS with prompt internalization. At other times, intact outer segment distal tips remain in immediate proximity to RPE apical projections without triggering phagocytic attack. Taken together, this suggests that the phagocytic function of the RPE may be tightly temporally regulated in the retina.

The rhythmic nature of RPE clearance of POS requires that RPE cells rid themselves of phagocytosed material before the next phagocytic challenge to avoid gradual accumulation of POS with time. Despite its importance in maintaining RPE long-term function, few studies have focused on digestion of POS. Lysosomes of different size and enzymatic content fuse with POS phagosomes by as early as 30 min after light onset [65]. Electron microscopy images show that the disk structure of engulfed POS in RPE phagosomes disappears gradually starting at around 2.5 h after light onset in light-entrained rats [65]. At the light microscopic level, numbers of discernible phagosomes in albino mouse RPE return to baseline levels approximately 3 h after light onset [66]. Finally, at least some of the final products of POS digestion, such as docosahexaenoic acid, are transported back for reuse by photoreceptors [67–71].

Molecular Mechanisms of Shedding and RPE Phagocytosis

Studies seeking to identify the molecular machinery used by photoreceptors and RPE cells for POS renewal have greatly benefited from the fact that RPE cells in tissue culture retain their phagocytic activity toward POS. Recording the binding and internalization kinetics of POS by RPE in culture ideally complements the classical microscopic characterization of outer segment uptake by RPE in vivo (Fig. 4). First, outer segment recognition cannot be studied separately from outer segment internalization in vivo because shed outer segments in the subretinal space are juxtaposed to the RPE surface whether or not they are recognized or bound by RPE receptors. Second, far greater numbers of RPE cells can be evaluated in each sample in assays in vitro than in tissue sections. Therefore, small but significant alterations in RPE phagocytic activity may be detected by in vitro assays that may be missed in the light and electron microscopy studies of postmortem tissues. Third, RPE cells in vitro can be studied following specific manipulation of their phagocytic mechanism by pharmacological compounds, recombinant proteins, protein overexpression or downregulation, just to name a few. Gainand loss-of-function approaches are well suited to unequivocally identify critical components of the RPE phagocytic machinery. Fourth, in vitro phagocytic challenge of RPE cells allows one to test directly the phagocytic activity of RPE cells toward POS, while altered photoreceptor shedding or IPM may indirectly alter the phagocytic activity by RPE cells in vivo. Importantly, however, RPE cells in vitro only provide a phagocytic assay system with relevance to RPE phagocytosis in vivo if cells are studied as differentiated, polarized epithelial monolayers that assemble their phagocytic machinery at their apical surface,

Fig. 4. Fluorescence microscopy quantification of photoreceptor outer segment (POS) phagocytosis by retinal pigment epithelial (RPE) cells in vivo and in vitro. A–C Cryosections of eyecups from 2-month-old wild-type mice were labeled with rod opsin antibody B6-30 [105]. Maximal projections of confocal microscopy x–y sections representing tissue sections 5- m thick are shown. A High-magnification view of the RPE–POS interface shows intact rod outer segments and opsin-positive phagosomes (white) in RPE cells adjacent to RPE nuclei (gray). B and C Lowmagnification view of similar opsin signals without nuclei stain illustrates that these images can be used to count opsin phagosome numbers in the RPE. B At 1 h before light onset, the RPE cell layer shows few opsin-labeled phagosomes. C At 2 h after light onset, the RPE cell layer shows numerous opsin-labeled phagosomes, confirming the daily burst of rod POS phagocytosis by the RPE. D Maximal projection of confocal microscopy x–y sections of primary wild-type mouse RPE cells in culture 1 h after phagocytic challenge demonstrates vigorous uptake of fluorescent isolated bovine POS by RPE cells in vitro. Phagocytosed POS appears in white, RPE cell junctions stained with ZO-1 tight junction marker antibody appear in gray. All scale bars 20 m.

like RPE cells in vivo. Finally, while all evidence suggests that the phagocytic activity of polarized RPE cells in vitro retains the primary characteristics of the phagocytic activity of RPE cells in vivo, this is not the case for the nature of particle contact. In experimental phagocytosis assays, RPE cells must establish firm binding of isolated POS that is stable enough to withstand shear forces during sample processing, including vigorous washing steps. This is in sharp contrast to the contact of apical RPE receptors with shedding/shed POS in the subretinal space, where mechanical stress is absent and a stable binding event per se may not occur. Thus, comparison of in vivo and in vitro RPE phagocytosis counting fluorescenceor opsin antibody-labeled POS as illustrated in Fig. 4 are both required to fully elucidate the phagocytic machinery of the RPE.

The events that cause and accompany shedding of photoreceptor tips remain largely obscure. As outlined in the previous section, it is generally thought that RPE microvilli– outer segment interactions as well as photoreceptor intrinsic mechanisms contribute to

POS shedding. However, the molecular mechanisms that promote outer segment shedding remain yet to be uncovered. This includes the identification of putative “shed me” and “eat me” signals exposed by photoreceptor tips. Molecular changes likely designate the distal tip for shedding and may also serve as recognition signals triggering subsequent RPE phagocytosis. POS changes in surface structure or composition may promote recognition by RPE phagocytic receptors directly. Alternatively, POS changes may promote recruitment of ligands, concentrating or presenting it for engagement of RPE surface receptors. This scenario predicts that the same ligand–ligand complex binds to POS and to RPE receptors. Although there is no evidence for such ligand as yet for retina, a precedent exists. Most cells undergoing apoptosis flip the anionic phospholipid phosphatidylserine from the inner to the outer leaflet of the plasma membrane lipid bilayer [72]. External phosphatidylserine serves to signal the apoptotic process to phagocytic cells. Phagocytes may express surface receptors that recognize phosphatidylserine directly [73]. In addition, soluble phosphatidylserine-binding proteins form a bridge between apoptotic and phagocytic cells. It has not yet been demonstrated whether shedding POS in the retina exposes external leaflet phosphatidylserine. However, phosphatidylserine-enriched liposomes have been reported to affect in vitro outer segment uptake [74]. Furthermore, apoptotic cells and POS compete for binding by RPE cells and by macrophages in culture [75]. This strongly suggests that these different phagocytic particles share “eat me” signals. The only such signal conclusively identified for apoptotic cells is phosphatidylserine.

Much progress has been made over the past decade identifying the components that form the RPE phagocytic machinery. Most important, it is now clear that RPE cells use a mechanism to phagocytose shed outer segment that belongs to a group of related clearance mechanisms used by other cell types, such as macrophages, to phagocytose apoptotic cells [75, 76]. Professional phagocytic mechanisms like the RPE’s are saturable, receptormediated processes with distinct phases of particle recognition/binding, internalization, and digestion. All known phagocytic mechanisms employ multiple phagocyte receptors, five of which have been described in RPE cells as described below.

In vitro evidence suggests that pattern recognition receptors also associated with inflammatory phagocytosis may participate in POS uptake. Inhibiting a mannose-binding protein using antibodies reduces POS uptake by rat RPE in culture [77]. Toll-like receptor 4 has been shown to redistribute at the apical surface of human RPE cells in response to POS isolated from human but not from bovine retina and to induce a cytoplasmic signaling response in the RPE that may be involved in triggering POS clearance [78]. Little is known yet about the significance of these receptors for POS renewal in vivo.

The scavenger receptor family member CD36 facilitates apoptotic cell clearance, fatty acid transport, and cell–matrix interactions by recognizing structurally defined lipid peroxidation products [79–82]. Blocking CD36 with CD36 antibodies partially inhibits outer segment engulfment by RPE cells in vitro [83]. Clustering of CD36 is sufficient to alter the rate with which RPE cells in culture internalize surface-bound outer segment [84]. In vitro assays suggest that CD36 clustering may activate intracellular signaling processes that ultimately target the internalization mechanism of the RPE. Lack of functional or morphological retinal abnormalities in CD36 knockout mice suggests that CD36 may be dispensable for POS renewal in the healthy eye, at least in mice kept under standard vivarium conditions. However, acute high-intensity illumination generates oxidized

phospholipids in rat retina that are specifically recognized by CD36 [85]. In vitro, these ligands impair POS phagocytosis by RPE cells in a CD36-dependent manner. Notably, rats that harbor a deletion variant in the scavenger receptor CD36 were found to be more sensitive to intense light-induced retinal damage [86]. Thus, CD36 signaling may not be essential for RPE phagocytosis under basal conditions but may be involved in outer segment clearance under conditions that increase oxidative stress in the retina.

A deletion in the coding region of the gene for the receptor tyrosine kinase Mer (MerTK) abolishes RPE phagocytosis in the RCS rat [87, 88]. MerTK deficiency also causes rapid retinal degeneration and defective apoptotic cell clearance by macrophages in MerTK kinase-deficient transgenic mice [76, 89]. MerTK-deficient RCS RPEs in culture retain the ability to recognize and bind outer segment but fail to internalize surface-bound outer segments [90, 91]. Despite the critical importance of MerTK for outer segment phagocytosis in vitro and in vivo, its downstream effectors in RPE cells remain to be identified.

The permanent RPE cell lines RPE-J (rat), ARPE-19 and d407 (both human), and mouse, rat, and human RPE in primary culture employ the integrin adhesion receptor αvβ5 to recognize and bind isolated POS in experimental phagocytosis assays [28, 92– 94]. The αvβ5 integrin is essential for POS binding in vitro as RPE cells derived from β5 integrin knockout mice in primary culture largely fail to bind isolated POS [95]. RPE cells in β5 integrin knockout mice in vivo retain basal levels of phagocytic activity but lack the characteristic burst of phagocytosis upon early morning rod shedding (Fig. 5A) [95]. Moreover, αvβ5 integrin deficiency is sufficient to cause age-related vision loss in β5 integrin knockout mice accompanied by excessive accumulation of lipofuscin lipid storage bodies in the RPE, a cardinal feature of RPE aging and disease [95]. These findings illustrate that αvβ5 integrin receptors of the RPE are critical for retinal function.

Furthermore, studies comparing wild-type RPE cells with RPE cells lacking αvβ5 receptor in vivo and in vitro indicated that αvβ5 integrin receptor engagement initiates intracellular signaling processes in the RPE that promote POS engulfment. Integrin receptors do not possess intrinsic enzymatic activity. However, integrin cytoplasmic domains assemble cytosolic signaling proteins and often bind different sets of proteins depending on receptor occupancy. Indeed, phagocytic challenge with POS rapidly increases the presence of multiple tyrosine-phosphorylated proteins associated with apical αvβ5 integrin in cultured RPE cells [96]. One of these proteins is focal adhesion kinase (FAK) [96]. FAK is a cytoplasmic tyrosine kinase that relays integrin signals [97]. Dramatic reduction of POS internalization by RPE cells in which FAK is specifically inhibited demonstrates that FAK is an important mediator of RPE phagocytic signaling downstream of αvβ5 integrin. MerTK tyrosine phosphorylation is thought to reflect MerTK activity. Strikingly, silencing FAK signaling in RPE cells abolishes MerTK tyrosine phosphorylation induced by POS phagocytosis [96]. MerTK is thus a target of FAK signaling in RPE cells. Furthermore, a strict temporal regulation of both FAK and MerTK activities precisely coincides with circadian shedding of rod outer segments in intact mouse retina [95]. These synchronized signaling events are completely abolished in αvβ5 integrindeficient retina [95] (Fig. 5B). These data provide the first direct evidence of a crosstalk of the recognition receptor αvβ5 integrin and the internalization receptor MerTK: The αvβ5 integrin-dependent signaling via FAK controls the efficiency of RPE phagocytosis by promoting peak MerTK activity at the time of rod outer segment shedding.