Novel role for α_vβ₅-integrin in retinal adhesion and its diurnal peak

Abstract

α_vβ₅-Integrin is the sole integrin receptor at the retinal pigment epithelium (RPE)-photoreceptor interface and promotes RPE phagocytic signaling to the tyrosine kinase Mer tyrosine kinase (MerTK) once a day in response to circadian photoreceptor shedding. Herein we identify a novel role for α_vβ₅-integrin in permanent RPE-photoreceptor adhesion that is independent of α_vβ₅’s function in retinal phagocytosis. To compare retinal adhesion of wild-type and β₅-integrin^−/− mice, we mechanically separated RPE and neural retina and quantified RPE protein and pigment retention with the neural retina. Lack of α_vβ₅-integrin with normal expression of other RPE integrins greatly weakened retinal adhesion in young mice and accelerated its age-dependent decline. Unexpectedly, the strength of wild-type retinal adhesion varied with a diurnal rhythm that peaked 3.5 h after light onset, after the completion of phagocytosis, when integrin signaling to MerTK is minimal. Permanent α_vβ₅ receptor deficiency attenuated the diurnal peak of retinal adhesion in β₅-integrin^−/− mice. These results identify α_vβ₅-integrin as the first RPE receptor that contributes to retinal adhesion, a vital mechanism for long-term photoreceptor function and viability. Furthermore, they indicate that α_vβ₅ receptors at the same apical plasma membrane domain of RPE cells fulfill two separate functions that are synchronized by different diurnal rhythms.

Ligation of integrin receptors promotes cellular functions such as substrate adhesion to the extracellular matrix, migration, and phagocytosis. Integrin heterodimers constitute a large family of at least 24 receptors that often share and bind several ligands (20). Studies of integrin function in tissue culture have revealed striking similarities between signaling pathways elicited by integrin receptors during adhesion and during phagocytosis (6). However, matrix adhesion and phagocytosis usually occur at different subcellular domains, at the attached basal surface and at the free apical surface, respectively. Thus adhesive or phagocytic function of a given integrin receptor may be determined by its polarized localization. Alternatively, adhesion and phagocytosis in a given cell type may utilize different integrin family members. For example, macrophages that adhere to laminin via β₁-integrins use apical α_vβ₃-integrin to phagocytose apoptotic cells; adhesion of these cells to the α_vβ₃-integrin substrate fibrinogen redistributes α_vβ₃ basally, diminishing apical phagocytosis (11).

α_vβ₅-Integrin is the only integrin receptor that localizes to the apical surface of the retinal pigment epithelium (RPE) (1, 10). RPE cells form the outermost layer of the retina. Their microvillus-rich apical domain faces the outer segment portions of photoreceptors. Activities of the RPE that are essential for function and survival of photoreceptors include removing aged fragments of photoreceptor outer segments (POS) shed once daily by apical phagocytosis and maintaining contact with intact POS at all times by apical adhesion.

Intraocular pressure and a net fluid transport from retina to RPE likely contribute to retinal adhesion. In addition, apical surface receptors of the RPE are thought to adhere to ligands of the interphotoreceptor matrix (IPM), a complex mix of proteins and proteoglycans that fills the subretinal space and ensheathes outer segment portions of rod and cone photoreceptors (14, 16, 18, 21). IPM proteoglycan rearrangement and RPE microvillus collapse are early responses to retinal detachment that, if persistent, result in RPE dedifferentiation and proliferation, POS degeneration, and photoreceptor cell death (5). Despite their obvious importance for photoreceptor survival and hence vision, we still know little about RPE surface receptors or IPM ligands that may directly mediate retinal adhesion.

The apical surface of the RPE is the sole site of α_vβ₅ expression in the retina. Mice lacking α_vβ₅-integrin develop age-related blindness, illustrating the importance of α_vβ₅ for photoreceptor maintenance (25). Wild-type mice maximize phagocytic activity by 2 h after light onset that triggers rod shedding and barely phagocytose at other times. In contrast, β₅-integrin^−/− (β₅^−/−) mice lack the daily phagocytic peak but maintain significant phagocytic activity at all times.

In the present study, we set out to test whether lack of apical α_vβ₅-integrin alters retinal adhesion in β₅^−/− mice compared with strain-matched wild-type (β₅^+/+) mice. We first hypothesized that retinal adhesion may merely be altered at the time of peak β₅^+/+ phagocytosis as a consequence of altered β₅^−/− phagocytosis. However, we found a robust decrease in retinal adhesion in β₅^−/− mice at all times of day independent of RPE phagocytic activity. To our knowledge, α_vβ₅-integrin is the first RPE surface receptor directly implicated in retinal adhesion. Retinal adhesion declined with age in both β₅^−/− and β₅^+/+ mice, suggesting that the age-related vision loss in β₅^−/− mice is not caused by weak retinal adhesion alone. Finally, our experiments showed that retinal adhesion varied significantly with time of day, peaking daily 3.5 h after light onset in β₅^+/+ mice and to a lesser extent in β₅^−/− mice. These results reveal a diurnal rhythm of retinal adhesion in mammalian retina that is independent of the rhythm of retinal phagocytosis. Together, these results imply that α_vβ₅-integrin separately mediates both retinal adhesion and phagocytosis at the same plasma membrane domain of the RPE.

MATERIALS AND METHODS

Animals and tissue collection

β₅^−/− mice characterized in detail previously (19, 25) and β5^+/+ mice of the same genetic background (129T2/SvEmsJ; Jackson Laboratory, Bar Harbor, ME) were housed and bred under cyclic 12:12-h light-dark conditions (light onset at 0600) and fed ad libitum. All procedures involving animals were approved by the Weill Medical College Institutional Animal Care and Use Committee.

To quantify retinal adhesion, we modified a protocol described by Endo and colleagues (8). Mice were killed by CO₂ asphyxiation. Lens and cornea were swiftly removed from each enucleated eyeball in HEPES-buffered Hanks’ saline solution containing calcium and magnesium. Eyeballs were kept at room temperature to preserve retinal adhesion (8). After transferring an individual eyecup to an empty plastic dish, we performed a single radial cut toward the optic nerve, flattened the eyecup retina facing up, and peeled off the neural retina with forceps from one side of the cut to the other. We stored individual neural retinas and remaining eyecups separately at −80°C. We conducted these tissue harvests in β₅^+/+ and β₅^−/− mice from 1 to 21 mo of age or at different time points of the light-dark cycle.

Sample lysis

Individual whole eyecups or isolated neural retinas were solubilized in 50 mM Tris · HCl pH 7.5, 2 mM EDTA, 150 mM NaCl, 1% Triton X-100, 0.1% SDS, and 1% NP-40, freshly supplemented with 1% each of protease and phosphatase inhibitor cocktails (Sigma). After separation of insoluble material including melanin pigment by centrifugation, we quantified the protein content of cleared lysates using the Bradford colorimetric assay (4).

RPE pigment quantification

Melanin pigment granules were contained in the insoluble pellet after sample lysis. After washing the pellet in 50% ethanol-50% ether, we dissolved the pellet in 20% DMSO-2 N NaOH at 65°C. We measured absorbance at 490 nm of samples and of commercial melanin pigment (Sigma) at defined concentrations to calculate sample pigment concentrations. We divided pigment concentration by the protein concentration of individual samples to generate a normalized microgram of pigment per milligram of protein concentration in each sample. This accounted for differences among samples in neural tissue yield. Next, we calculated the means ± SE of all samples of each experimental condition. Some samples had outlying normalized pigment concentrations compared with the mean, either because of incomplete harvest and disintegration of the neural retina (abnormally low protein) or because of sample contamination with highly pigmented iris tissue (abnormally high pigment). These samples were excluded from analysis before calculation of final mean pigment values and SEs based on at least three independent retina samples for each condition. We used GraphPad Prism 4.0 software to test for significance with the Student’s t-test or ANOVA followed by the Bonferroni test as appropriate as indicated in Figs. 1 and 4.

Fig. 1.

Decreased retinal pigment epithelium (RPE) pigment adherence to neural retina of β₅-integrin^−/− (β₅^−/−) mice. A and B: whole mount bright-field microscopy views of peeled-off retinas with exposed outer retinal surface harvested from 2-mo-old mice. Fields at bottom show enlargements of fields at top. Scale bars (top and bottom): 100 μm. A: wild-type (β₅^+/+) retina retains dense patches of RPE pigment. B: β₅^−/− retina retains significantly less RPE pigment. C: quantification of solubilized RPE pigment peeled off with neural retinas. At all ages tested, β₅^−/− retina yielded less RPE pigment than β₅^+/+ retina. Furthermore, decreasing amounts of RPE pigment attached to both β₅^+/+ and β₅^−/− retina with age. Bars represent mean ± SE relative yields of pigment of peeled-off retina; n = 3–5 individual retinas from 2–4 different mice. Significant differences of β₅^−/− yield compared with β₅^+/+ yield at the same age: *Student’s t-test (P < 0.05), **Student’s t-test (P < 0.01).

Fig. 4.

Attenuation of a diurnal peak in retinal adhesion in β₅^−/− retina. A: RPE pigment content in individual peeled-off neural retina samples from 2-mo-old β₅^−/− and β₅^+/+ mice killed at different times of day as indicated (light onset at 0600). Filled bars, β₅^+/+ samples; gray bars, β₅^−/− samples. Open bars show that the reduction in β₅^−/− pigment content compared with β5^+/+ pigment content is largest at the time of peak adhesion (0930). Bars represent mean ± SE relative yields of pigment of peeled-off retina; n = 3–5 individual retinas obtained from 2–4 different mice. Significant differences of pigment yield compared with yield at 0800 h (striped bars) of the same genotype: *ANOVA (P < 0.05), **ANOVA (P < 0.001). Student’s t-test indicated significantly less pigment in β₅^−/− samples compared with β₅^+/+ samples at all time points except 0630 (P < 0.05). B: representative immunoblots of individual peeled-off neural retina lysates harvested from 2-mo-old mice at times of day as indicated show significantly increased amounts of RPE65 in samples harvested at 0930 compared with other time points. Changes in other marker proteins did not reach significance. Note that RPE65 levels increased significantly in β₅^−/− samples at 0930 relative to 0800 but remained far below levels of RPE65 in β₅^+/+ samples at any time of day.

SDS-PAGE and immunoblotting

Immunoblotting analysis was performed only on samples that were included in pigment analysis, as described above. We separated sample lysates representing 15% of one mouse retina or 10% of a whole eyecup in reducing sample buffer on 10% SDS-polyacrylamide gels. After electrophoresis and protein transfer onto nitrocellulose membrane, we immunoblotted with primary antibodies to β₁-integrin (9EG7, provided by D. Vestweber, Max-Planck-Institute of Vascular Biology, Münster, Germany), inter-photoreceptor retinoid binding protein [IRBP; provided by B. N. Wiggert, National Institutes of Health (NIH), Bethesda, MD; Ref. 27], neural cell adhesion molecule (N-CAM; provided by E. Rodriguez-Boulan, Weill Medical College), RPE65 (provided by T. M. Redmond, NIH, Bethesda, MD), glial fibrillary acidic protein (GFAP) and ezrin (both from Sigma), α_v- and β₃-integrin (both from BD Pharmingen), β₅-integrin (Santa Cruz Biotechnology, Santa Cruz, CA), and Mer tyrosine kinase (MerTK; R&D Systems) and appropriate horseradish peroxidase-conjugated secondary antibodies followed by chemiluminescence detection (PerkinElmer). X-ray films were scanned, and signals of samples of the same experiment present on the same blot were quantified with NIH Image 1.63 software. We calculated β₅^−/− -to- β₅^+/+ ratios for each experiment, calculated mean ratios and SEs, and tested for significance with Student’s t-test.

RT-PCR

Total RNA was isolated from individual mouse eyecups with the RNAgent Total RNA isolation system (Promega) and treated with RQ1 DNAse I (Promega) using the manufacturer’s protocols. We reverse transcribed 1 μg of mRNA with the Reverse Transcription System (Promega) as instructed. We used cDNA templates to amplify integrin subunits with a custom MultiGene-12 RT-PCR profiling kit (Superarray Biosciences). We used aliquots of the same templates to amplify different regions of the β₅-integrin coding sequence, the neomycin resistance cDNA, and cyclophilin A as an internal control by RT-PCR with the following oligonucleotides: β₅-5′, forward 5′-ACCTCGTGTGAAGAATGCCTG and reverse 5′-CTGGTTA-GAGGCTGTGTACTC; β₅-3′, forward 5′-GGAAGTGAGGAAGCAGAGGGTGTCCCGGAACCG and reverse 5′-GACTGTCCCGGAAGCCCACGGGCCTCAAGG; neomycin, forward 5′-CCGGCCGCTTGGGTGGAGAGGC and reverse 5′-GGTCAGCCCATTCGCCGCCAAGC; and cyclophilin, forward 5′-TGGTCAACCCCACCGTGTTCTTCG and reverse 5′-GGTGATCTTCTTGCTGGTCTTGC. After separation of PCR products on 2% agarose 1× Tris-acetate-EDTA gels, we acquired digital pictures of ethidium bromide-treated samples on a GeneFlash apparatus (Syngene Bioimaging).

Immunofluorescence labeling of retinal cryosections

Eight-micrometers-thick frozen sections from paraformaldehyde-fixed eyecups were prepared and stained with antibodies to RPE65 and ezrin and fluorescent secondary antibodies (Molecular Probes) as described previously (10). Images were acquired using a Leica TSP2 confocal microscopy system and recompiled using PhotoShop 7.0 software.

RESULTS

Decreased retinal adhesion in mice lacking α_vβ₅-integrin

Specific receptor-ligand interactions that mediate retinal adhesion are thus far unknown. However, earlier studies found that increased adhesion in amphibian retina precisely coincides with the time of daily peak phagocytosis (7, 24). β5^−/− mice that are deficient in α_vβ₅-integrin receptors lack the 0800 phagocytosis peak that follows circadian rod outer segment shedding in β₅^+/+ mice ~2 h after light onset at 0600 (25). To test whether reducing phagocytosis alters retinal adhesion, we determined the relative strength of retinal adhesion at 0800 in eyes of age-matched β₅^+/+ and β₅^−/− mice from 1 to 21 mo of age using established adhesion tests (3, 8). As a consequence of retinal adhesion, whole RPE cells or apical domains of RPE such as microvilli remain attached to the neural retina when the neural retina is peeled off a flattened eyecup. Melanin pigment of the RPE retrieved with isolated neural retina correlates with the extent of RPE attachment to the neural retina, which directly reflects the strength of adhesion between RPE and POS. We observed dense patches of RPE pigment on the outer surface of neural retina peeled off β₅^+/+ mouse eyecups at 0800 (Fig. 1A). In contrast, β₅^−/− retina isolated at 0800 retained dramatically less RPE pigment in all areas, indicating weakened retinal adhesion (Fig. 1B). In some β₅^−/− samples, more RPE pigment was retrieved at the periphery of the retina (data not shown; Ref. 3).

Quantification of solubilized pigment of neural retina samples and normalization to retinal yield showed that significantly less pigment attached to β₅^−/− retina than to β₅^+/+ retina at 0800, the time of the RPE phagocytosis peak (Fig. 1C). This was true for all ages tested, suggesting that lack of α_vβ₅-integrin receptors directly reduces retinal adhesion. Interestingly, pigment retrieved with the neural retina decreased with age in both β₅^+/+ and β₅^−/− mice. Average pigment content of 12-mo-old neural retina was 45% in β₅^+/+ mice and 34% in β₅^−/− mice compared with 1-mo-old retina of the same genotype (Fig. 1C). These data imply that retinal adhesion weakens with age even in β₅^+/+ mice. Lack of α_vβ₅-integrin receptors may accelerate this process: we retrieved 42% less melanin when harvesting neural retina from 2.5-mo-old β₅^−/− mice than from 1-mo-old β₅^−/− mice. In contrast, the same difference in age decreased melanin content of neural retina harvested from β₅^+/+ mice by only 26%.

To determine whether increased pigment levels in neural retina samples directly correlate with increased cell transfer and adhesion, we next determined levels of RPE- and retina-specific proteins in extracts of peeled-off neural retinas (Fig. 2A). Indeed, we detected higher levels of the RPE-specific protein RPE65 in β₅^+/+ retina extracts than in β₅^−/− retina extracts harvested from mice from 1 to 12 mo of age (see Fig. 2A for representative immunoblots and Table 1 for quantification). The same was true for ezrin, a major constituent of RPE apical microvilli (2). In contrast, we saw no difference in levels of the RPE phagocytosis receptor MerTK, possibly because MerTK expressed in the neural retina may obscure differences in RPE-derived MerTK content. However, β₅^−/− retina extracts of all ages contained lower levels of α_v-integrin protein than β₅^+/+ extracts and, as expected, no β₅-integrin protein (Fig. 2A and Table 1). Similar levels in β₅^+/+ and β₅^−/− neural retinal extracts of IRBP and GFAP confirmed that both extracts represented the same yield of IPM and neural retina (Table 1). Figure 2B shows that the marker proteins we selected, RPE65, ezrin, MerTK, and IRBP, were expressed at equal levels in β₅^−/− and β₅^+/+ whole eyecups. Low levels of RPE proteins RPE65 and ezrin in neural retina extracts therefore result from poor RPE-retina adhesion in α_vβ₅-integrin-deficient mice.

Fig. 2.

Decreased content of RPE proteins in peeled-off β₅^−/− neural retina lysates. A: representative immunoblots of peeled-off neural retina lysates harvested from 2.5-mo-old and 7-mo-old mice show reduced amounts of RPE65, ezrin, and α_v-integrin but similar amounts of Mer tyrosine kinase (MerTK), interphotoreceptor retinoid binding protein (IRBP), and glial fibrillary acidic protein (GFAP) in β₅^−/− lysates compared with β₅^+/+ lysates. As expected, only β₅^+/+ lysates contained β₅-integrin. For quantification of ratios of retrieved β₅^−/− protein to β₅^+/+ protein, see Table 1. Each lane shows lysate from a different retina. For 2-mo-old mice and 7-mo-old β₅^−/− mice, samples shown are from both retinas from the same animal. For 7-mo-old β₅^+/+ mice, lanes show retina lysates from 2 different mice. B: immunoblots comparing equal fractions of individual whole eyecups of 2-mo-old mice show similar protein expression levels of RPE65, ezrin, MerTK, and IRBP in β₅^−/− and β₅^+/+ lysates. Lysates from both eyecups of 1 β₅^+/+ and 1 β₅^−/− mouse are shown. Similar results were obtained by comparing β₅^−/− with β₅^+/+ lysates from 1-mo-old and 12-mo-old mice (data not shown).

Table 1.

Effect of age on marker protein yield in β₅^−/− neural retina extracts relative to β₅^+/+ extracts

Age	RPE65	Ezrin	MerTK	IRBP	αv-Integrin
1 mo	68 ± 4	64 ± 3	93 ± 4	96 ± 5	58 ± 15
2.5 mo	50 ± 16	53 ± 16	97 ± 12	94 ± 26	51 ± 18
7 mo	65 ± 4	64 ± 15	88 ± 2	98 ± 6	69 ± 8
12 mo	53 ± 8	ND	90 ± 12	104 ± 10	66 ± 13

Values (in %) are means ± SE β₅-integrin^−/− (β₅^−/−)-to-wild type (β₅^+/+) ratios for each protein. For each protein indicated, 3 separate experiments were performed. For each experiment, 1 retina from 1 mouse of each genotype was lysed and analyzed on the same immunoblot. Band intensities were quantified, and the β₅^−/−-to- β₅^+/+ protein ratio for each protein was calculated. Results from 3 independent experiments were combined to calculate mean ratios. At all ages, retinal pigment epithelium RPE65, ezrin, and α_v-integrin levels in β₅^−/− extracts were significantly lower than in β₅^+/+ extracts (Student’s t-test, P < 0.05). In contrast, Mer tyrosine kinase (MerTK) and interphotoreceptor retinoid binding protein (IRBP) were present at equal levels in β₅^−/− and β₅^+/+ samples. ND, not determined.

Reduced levels of α_v-integrin protein but not transcript in β₅^−/− eyecups

Integrins form a large family of heterodimeric receptors comprised of α- and β-integrin subunits, and at least 24 αβ combinations exist in vertebrates (20). In a given cell, different integrin receptors may share extracellular ligands and have overlapping functions. For instance, we showed previously (11) that macrophages can use either α_vβ₃- or α_vβ₅-integrin to phagocytose apoptotic cells and isolated POS fragments. Therefore, we tested whether expression levels of integrin subunits other than β₅ differed between β₅^−/− and β₅^+/+ mouse eyecups. First, we compared mRNA levels of four α-and seven β-integrin subunits with a MultiGene-12 RT-PCR profiling kit. Figure 3A shows that levels of transcripts of these integrin subunits did not vary greatly between β₅^−/− and β₅^+/+ eyecups. Notably, this included transcripts of β₅-integrin, because β₅-specific primers of the kit amplified a fragment of the 5′ region of the β₅ cDNA located upstream of the neomycin insertion site of the β₅ targeting construct (Ref. 19 and personal communication with Superarray Biosciences). Because we could not obtain the precise sequence information of the proprietary primer sets included in the MultiGene-12 kit, we performed control RT-PCR amplifications testing β₅ cDNA levels in aliquots of the same templates that we used for the kit reactions. As expected, using primers amplifying a 5′ region that overlaps with the neomycin gene insertion yielded a product only from β₅^+/+ tissue (Fig. 3B). In contrast, only β₅^−/− samples contained the neomycin sequence (Fig. 3B). Finally, primers recognizing sequences of the 3′ end of the β₅ cDNA that is untouched by the neomycin insertion amplified products from both β₅^+/+ and β₅^−/− tissues (Fig. 3B). These data confirmed the ${\beta }_5^{23/-}$ genotype of our β₅-integrin-knockout mice. Furthermore, similar levels of cyclophilin control transcripts in both templates showed that transcripts of β₅-integrin disrupted by the neomycin cassette are present in β₅^−/− eye-cups at steady-state levels similar to those of intact β₅-integrin transcripts in β₅^+/+ eyecups.

Fig. 3.

Comparison of expression of integrin subunits and neural cell adhesion molecule (N-CAM) in β₅^−/− and β₅^+/+ eyecups. A: amplification of integrin-specific cDNA as indicated showed no gross difference in integrin subunit transcript levels when comparing templates prepared from individual 2-mo-old β₅^−/− and β₅^+/+ eyecups. B: amplification of β₅-integrin and neomycin cDNA regions as indicated. Only β₅^+/+ templates possess the 5′ region of the native β₅-integrin cDNA (5′ β₅) that is targeted to inactivate the β₅-integrin gene in β₅^−/− mice. Only β₅^−/− templates include the neomycin selection sequence (Neo). Both templates contain the 3′ region of the β₅-integrin cDNA (3′ β₅) that is untouched by neomycin insertion. Cyclophilin was used as an internal control to show equal amounts of cDNA in both templates. C: comparative immunoblots of eyecup lysates with primary antibodies as indicated show equal protein levels of β₁-integrin, β₃-integrin, and N-CAM in eyecup lysates prepared from 2-mo-old β₅^−/− and β₅^+/+ mice. In contrast, β₅^−/− lysates contained reduced amounts of α_v-integrin and, as expected, no β₅-integrin. Identical levels of MerTK in β₅^−/− and β₅^+/+ samples confirmed that equivalent fractions of eyecups were loaded. Lysates from individual eyecups of 2 different β₅^+/+ and β₅ ^−/− mice are shown.

Second, we compared expression levels of integrin subunit proteins between β₅^−/− and β₅^+/+ eyecups (Fig. 3C). Using MerTK as a loading control, we found that α_v-integrin protein in β₅^−/− eyecups decreased by 48 ± 4% compared with β₅^+/+ controls (mean ± SE of 3 individual eyecups from 3 different mice). In contrast, β₁- and β₃-integrins were present at similar steady-state levels in β₅^−/− and β₅^+/+ eyecups, although both may form integrin receptor dimers with α_v. Finally, β₅^−/− eyecups contained normal levels of the nonintegrin adhesion receptor N-CAM, whose apical localization in the RPE may depend on RPE-neural retina interaction, as it is lost in RPE in vitro (13). Together, these data indicate that specific reduction of α_v-integrin protein in β₅^−/− eyecups occurs through post-transcriptional mechanisms. Normal expression levels of other adhesion proteins in β₅^−/− eyecups suggest that α_vβ₅-integrin receptors may directly mediate retinal adhesion.

Role for α_vβ₅-integrin in a diurnal rhythm of retinal adhesion independent of the rhythm of retinal phagocytosis

The data above show that loss of the synchronized peak of rod POS phagocytosis coincides with decreased retinal adhesion at 0800 in β₅^−/− retina. At other times of day, phagocytosis continues at equal levels in β₅^−/− retina, while phagocytosis is essentially absent in β₅^+/+ retina (25). To determine whether phagocytic activity and retinal adhesion directly correlate, we compared β₅^+/+ and β₅^−/− retinal adhesion at different times of day before and after light onset at 0600. Strikingly, RPE melanin partitioning with the neural retina indicated that retinal adhesion in β₅^+/+ mice varied with time of day, with a distinct peak at 0930 (Fig. 4A). Thus retinal adhesion in β₅^+/+ mice was greatest not at the same time as, but subsequent to, the 0800 h peak POS phagocytosis (Fig. 4A). Retinal adhesion in mice lacking α_vβ₅-integrin was significantly reduced at all time points tested compared with β₅^+/+ retinal adhesion (Fig. 4A). Even in the absence of α_vβ₅, β₅^−/− retinal adhesion was greater at 0930 than at 0800. However, β₅^−/− adhesion differed most from β₅^+/+ adhesion at 0930, indicating that α_vβ₅-integrin receptors contribute to the synchronized increase to maximum retinal adhesion in normal retina (Fig. 4A). Immunoblots of neural retina extracts probed for RPE and retina markers as shown earlier (Fig. 2) supported the results of the melanin quantification. We previously showed (25) that levels of RPE65 protein in whole retina extracts do not vary with time of earlier day. However, maximum levels of RPE65, ezrin, and β₅-integrin in β₅^+/+ neural retina samples harvested at 0930 h were 1.46 ± 0.14-, 1.44 ± 0.18-, and 1.57 ± 0.06-fold the levels in samples harvested at 0800 (means ± SE of 3–5 individual retinas from 2 or 3 different mice). This clearly demonstrated diurnal variation of retinal adhesion in wild-type mice (Fig. 4B). Furthermore, comparative immunoblotting confirmed consistently weaker retinal adhesion and an attenuated phagocytic peak in β₅^−/− mice (Fig. 4B).

Similar peak and nonpeak localization of RPE proteins in β₅^+/+ and β₅^−/− RPE

It is well known that the composition or spatial organization of molecules in the IPM changes with time of day (28). We therefore tested whether changes in RPE subcellular distribution may contribute to the differential content of RPE65 or ezrin in neural retina extracts that we detected in our adhesion assays. However, the micrographs shown in Fig. 5 demonstrate that neither RPE65 nor ezrin changed its localization in the RPE with time of day (Fig. 5). Furthermore, the distribution of these marker proteins did not differ between β₅^+/+ and β₅^−/− RPE (Fig. 5). Together, our results therefore identify a diurnal rhythm of retinal adhesion in mammalian retina that depends on α_vβ₅-integrin.

Fig. 5.

Similar distribution of RPE65 and ezrin in β₅^−/− and β₅^+/+ RPE in situ independent of time of day. Double immunofluorescence labeling of retinal cryosections with antibodies to RPE65 (A–D) and ezrin (A′–D′). Each field shows a representative maximal projection of 1 μm of tissue for comparison of RPE65 and ezrin localization in β₅^+/+ and β₅^−/− RPE in the eye at 0930 and at 1700 as indicated. Control fields show nonspecific immunofluorescence signals obtained in labeling tissues with secondary antibodies only (E and F are controls for RPE65; E′ and F′ are controls for ezrin). Scale bars are 8 μm.

DISCUSSION

In this study, we demonstrate a novel role for α_vβ₅-integrin receptors in rhythmic retinal adhesion that is synchronized with the light-dark cycle. To our knowledge, there is no prior report that retinal adhesion varies with time of day in mammalian retina. We considered whether diurnal changes in RPE pigment or protein distribution might contribute significantly to the diurnal differences in their fractionation with neural retina that we interpret as a measure for retinal adhesion. Indeed, melanosomes distribute toward the apical surface of mouse RPE in response to light (12). However, our experiments show that retinal adhesion in wild-type mice increased by 58% from 2 h to 3.5 h after light onset, whereas Futter and colleagues (12) found that the fraction of RPE melanosomes localizing within apical microvilli decreases from 15.5% at 2 h to 5% at 3.5 h after light onset. Furthermore, subcellular distribution of RPE65 and ezrin did not change from time of peak to nonpeak retinal adhesion in either β₅^+/+ or β₅^−/− RPE in vivo. Finally, melanin quantification of the neural retina samples correlated very closely with partitioning of the RPE-specific cytoplasmic protein RPE65 with the neural retina. We therefore conclude that differences in retinal adhesion rather than marker mobility are responsible for the differential fractionation of marker pigment and protein we detect.

In frog retina, light onset simultaneously and directly increases both retinal adhesion and POS shedding/RPE phagocytosis (7). In contrast, our time course study reveals that maximal retinal adhesion in mice occurs 1.5 h after maximal POS phagocytosis. Moreover, in preliminary experiments, we found that diurnal changes in retinal adhesion proceed on time even in constant darkness in mice that were previously adapted to normal dark-light fluctuations (data not shown). Like murine POS shedding and phagocytosis, murine retinal adhesion may thus be regulated by circadian rhythms.

Lack of α_vβ₅-integrin abolishes the daily rhythm of POS phagocytosis that restricts RPE phagocytic activity to a period of ~2 h following light onset in normal retina (25). Herein we have demonstrated that lack of α_vβ₅-integrin weakens but does not eliminate RPE-POS adhesion at all times of day and additionally attenuates its synchronized daily fluctuation. Lack of α_vβ₅-integrin receptors may decrease retinal adhesion directly. We found no evidence for expression changes in β₅^−/− retina of integrin subunits other than β₅’s partner subunit α_v or of the cell-cell adhesion receptor N-CAM, whose apical polarity in the RPE requires interaction with photoreceptors (12). Importantly, the difference between β₅^−/− and β₅^+/+ adhesion was largest precisely at the time of peak adhesion. We conclude that α_vβ₅-integrin contributes to retinal adhesion at all times and is particularly required for strengthening retinal adhesion 3.5 h after light onset.

Phagocytic and adhesive functions of α_vβ₅ receptors at the apical surface of the RPE may be independent of each other, because both are defective immediately after maturation of the retina in β₅^−/− mice. A precedent for independent regulation of retinal adhesion and POS phagocytosis exists: vitiligo mice, which carry a mutation in the microphthalmia transcription factor gene (23), display early-onset retinal detachment that is likely due to a primary defect in retinal adhesion (3). However, vitiligo RPE cells in situ phagocytose POS with normal diurnal rhythm, albeit less efficiently than do wild-type RPE cells (22, 26). Thus α_vβ₅-integrin in vitiligo RPE may function normally in POS phagocytosis but may not function in, or may not be sufficient for, retinal adhesion.

We previously showed (9, 25) that α_vβ₅-integrin receptors at the apical surface of the RPE initiate a signal transduction pathway via focal adhesion kinase (FAK) that activates the essential phagocytosis receptor MerTK precisely in time for peak phagocytosis 2 h after light onset. It appears unlikely that FAK and MerTK signaling also promote peak adhesion subsequent to phagocytosis, because activities of both kinases in the retina sharply decline before retinal adhesion increases (25). Rather, α_vβ₅ receptors at the same apical surface of the RPE may exist in two independent functional pools that utilize distinct downstream signaling pathways to promote their two distinct functions: POS phagocytosis and retinal adhesion. The highly synchronized α_vβ₅-dependent activities at the RPE-photoreceptor interface provide the ideal in vivo model system to test this intriguing possibility.

Dependence of rhythmic retinal phagocytosis and adhesion on α_vβ₅-integrin suggests that the IPM ensheathing apical RPE microvilli contains ligand proteins for α_vβ₅-integrin that remain to be identified. Both RPE and photoreceptor cells contribute to the IPM that consists of an elaborate and regionalized network of glycoproteins and proteoglycans (16, 17, 21). The α_vβ₅-integrin ligand vitronectin is synthesized by RPE cells in vivo and in vitro but localizes mostly to the basolateral surface of the RPE rather than to the IPM in the retina (15). Interestingly, light-to-dark transition directly stimulates changes in molecular conformation or regional distribution of IPM components in rat retina (28). Studies are under way to identify ligand-α_vβ₅ interactions in retinal adhesion and phagocytosis and to determine whether diurnal changes in ligand availability may contribute to the timely regulation of α_vβ₅ functions in the retina.