Stage-Specific Association of Apolipoprotein A-I and E in Developing Mouse Retina

Abstract

purpose. To characterize the intercellular lipid transport systems in differentiating retina.

methods. Developing mouse retinas were evaluated for the expression of apolipoproteins (apoE, apoA-I) by Western blot analysis and reverse transcription–polymerase chain reaction (RT-PCR). They were compared with changes in the lipid content and association of retinal proteins, such as postsynaptic density protein 95, glial fibrillary acidic protein, and cellular retinaldehyde-binding protein. Intraretinal distribution of apolipoproteins and their receptors was examined by immunofluorescence and in situ hybridization of prenatal and postnatal retinal sections. In vitro culture of dissociated cells was also examined.

results. Although apoE is known to be present in the mature retina, the neonatal retina remarkably expressed apoA-I mRNA and protein. This protein was present until postnatal day (P)3, and its putative receptor, scavenger receptor class B-I, was present until P5 to P7. This state subsequently exhibited a dramatic switchover to an apoE-rich one, in parallel with the stratification. Whereas apoE was synthesized at low levels until P7, apoE mRNA was clearly concentrated in Müller glia cells, which extended long apoE-bound processes to the plexuses and contours of photoreceptor cells. These acceptor cells expressed LDL receptor–related protein 1 as a putative receptor. ApoE genes were not transcribed in ganglion cells, though they were associated with a high level of the protein throughout the development. ApoE protein in ganglion cells initially appeared to be synthesized by astrocytes but later were observed to be supplied from an extraretinal space.

conclusions. The present results document several new aspects of apoA-I and apoE in the developing retina. The switchover of the lipoprotein systems runs a parallel course with the differentiation.

Lipids play vital structural and functional roles in neural systems. ^{1 2 3 4 5 6} Ranging from very-low density lipoprotein (VLDL) to high-density lipoprotein (HDL), blood plasma contains diverse classes of lipoprotein particles. They consist of characteristic apolipoproteins such as apolipoprotein A-I (apoA-I) and apoE and carry characteristic contents of lipids. The brain, which is isolated by the blood-brain barrier, uniquely contains a high level of apoE-containing HDL-like lipoproteins. ^{7 8 9} With its hierarchical neuronal network, the retina also contains apoE. It is implicated in the formation of synapses because knockout of the apoE gene in mice is shown to result in a defective electroretinogram. ^{10 11}  

The major membrane receptors of apoE are LDL receptor (LDLR)–related protein (LRP) family proteins. In addition to the authentic LDLR, this family consists of LRP1, megalin, VLDL receptor (VLDLR), apoE receptor 2 (apoER2), and LRP5/6. ^{12 13 14 15 16 17} They are all implicated in the genesis of neural systems. ApoE also interacts with non-LRP membrane proteins, such as ATP-binding cassette subfamily A member (ABCA)-1 and scavenger receptor class B-I (SR-BI). ^{18 19 20} These proteins interact more canonically with apoA-I, the major component of plasma HDL. Both SR-BI and ABCA-1 participate in the intercellular transport of cholesterol and some phospholipids in many types of cells. ^{21 22} However, their presence during the formation of the retina is unknown. 

To assess the functional framework of lipid transport in retinal development, lipid distribution and lipoprotein expression must be studied. With the use of mass spectroscopic analysis, we have found that the diversity of fatty acids increases at postnatal day (P)5 to P7 and that of the polar head group after increases P10 (Nakanishi H, et al., manuscript in preparation). Here we report the expression of apolipoproteins in developing mouse retinas. Remarkably, before the massive expression of apoE, prenatal and neonatal retinas were associated with apoA-I. The distribution of apoA-I was homogeneous, whereas apoE production was concentrated in Müller cells, which made contact with the adjoining LRP1-expressing plexuses and photoreceptor cells. These dynamic changes in lipoprotein systems suggest their specific roles along the time schedule of phospholipid distribution and membrane organization. We discuss possible functional contexts of differential association of these apolipoproteins and receptors. 

Materials and Methods

Primary Antibodies and Other Reagents

Goat IgG against mouse apoE, LDLR, LRP1, megalin, LRP6, apoER2, SR-BI, ABCA-1 (Santa Cruz Biotechnology, Santa Cruz, CA) and apoA-I (Rockland, Gilbertsville, PA) was from the indicated distributors. Rabbit IgG against mouse apoE (Biodesign, Saco, ME), postsynaptic density protein 95 (PSD-95; UBI, Lake Placid, NY), glial fibrillary acidic protein (GFAP; DAKO, Glostrup, Denmark), SR-BI (Novus Biologicals, Littleton, CO), and mouse monoclonal antibodies against cellular retinaldehyde binding protein (CRALBP; Affinity Bioreagents, Golden, CO) and VLDLR (Santa Cruz) were from indicated distributors. Fluorescent secondary antibodies (Alexa Fluor 488- or 594-labeled, goat or donkey) were from Invitrogen (Carlsbad, CA). Other reagents not specified were from Wako Pure Chemicals (Osaka, Japan). 

Retinal Sections and In Vitro Primary Culture

SPF ICR albino mice were obtained from Japan Slc (Hamamatsu, Japan). All procedures concerning animal use were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Eyes were fixed in 4% paraformaldehyde (PFA) in PBS (10 mM Na phosphate, 150 mM NaCl, pH 7.2) and were transferred in graded concentrations of sucrose in PBS. Frozen 7-μm sections were prepared with the use of OCT compound (Sakura Fine Technologies, Tokyo, Japan). 

For in vitro cultures, retinas were isolated from P0 mouse in Earl’s balanced salt solution containing 10 mM Na HEPES, pH 7.2. After dissociation in PBS containing 0.125% trypsin and 0.04% DNAse I (Sigma-Aldrich, St. Louis, MO), cells were collected by centrifugation at 1000g for 10 minutes at 4°C. Cells were cultured on poly-l-lysine-coated coverslips at 1.5 × 10⁶ cells/cm² in DME/F-12 containing gentamicin, ITS (mixture of insulin, apotransferrin, and Se; Sigma-Aldrich) and 10% fetal bovine serum. During the culture in 5% CO₂ atmosphere, half the medium was replenished with fresh medium every 4 days. For immunofluorescence, cells were fixed in 3% PFA in PBS containing 8% sucrose for 20 minutes. 

Preparation of Retina Expanded on a Nitrocellulose Filter

Excised retinas were expanded on a black nitrocellulose filter (HABP; Millipore, Bedford, MA) that had been cleaned by sonication in excess water and PBS for several times. They were fixed in 3% PFA in PBS containing 8% sucrose for 20 minutes. 

Immunofluorescence and Other Microscopy Techniques

Retinal frozen sections were equilibrated in PBS for 30 minutes at room temperature. Specimens were treated with 50 μg/mL digitonin for 10 minutes and were blocked with 0.2% gelatin in PBS at room temperature. They were incubated with appropriately diluted antibodies (usually at 1/300 to 1/500) in PBS overnight and subsequently with fluorescent secondary antibodies at 1/200 for 1 hour at room temperature. Specimens were mounted with mounting medium (Vectashield; Vector Laboratories, Burlingame, CA) and observed in a confocal microscope (LSM; Carl Zeiss, Oberkochen, Germany). For video-enhanced microscopy, a BX51 fluorescence microscope (Olympus, Tokyo, Japan) equipped with an OrcaII cooled charge-coupled device camera (Hamamatsu Photonics, Hamamatsu, Japan) controlled by imaging software (Metaview; Nippon Roper, Tokyo, Japan) was used. Alexa Fluor 488–tagged poly(ethylene glycol)cholesteryl ether (A-PEG-Chol) was synthesized with the use of amino-derivatized PEG-Chol, which had been synthesized from PEG-Chol mesylated at the distal end. ²³ Sections were first incubated with anti–apoE IgG and secondary Alexa Fluor 594–labeled IgG, as described, without pretreatment with digitonin. They were then treated with 2 nM A-PEG-Chol for 20 seconds and were washed with PBS containing 10% fetal bovine serum that had been treated at 56°C for 30 minutes and centrifuged at 100,000g for 1 hour. After a final wash with PBS, specimens were mounted in PBS using a parafilm spacer. Specimens were observed in a video-enhanced microscope. 

Western Blot Analysis

Retina was isolated in PBS containing a protease inhibitor cocktail (P8340; Sigma-Aldrich). They were washed by low-speed centrifugation to remove blood. Protein concentration was determined with the use of a RC DC protein assay kit (Bio-Rad, Hercules, CA) using SDS. After separating the same protein amounts (20 μg) by polyacrylamide gel electrophoresis (PAGE), they were transferred onto a polyvinylidene diflouride (PVDF) membrane (Bio-Rad) blocked with 3% BSA-TBS–0.1% Tween 20 and probed with various antibodies. Results were visualized (ECL Plus; GE Healthcare, Piscataway, NJ) and analyzed (ImageMaster VDS-CL; GE Healthcare). 

Lipid and Protein Quantification

Lipids were extracted from isolated retinas in CHCl₃ by a standard method. ²⁴ Cholesterol and phospholipid concentrations were determined by use of the F-Cho E test assay kit (Wako; based on a method using cholesterol oxidase-peroxidase, N-ethyl-N-(2-hydroxy-3-sulfopropyl)-3,5-dimethoxyaniline and ascorbate oxidase) and phosphorus analysis, ²⁵ respectively. Protein amounts were determined as described. 

Reverse Transcription–Polymerase Chain Reaction Analysis of the Expression of apoE and apoA-I

Total RNA was extracted from isolated retinas (RNeasy; Qiagen, Hilden, Germany) according to the manufacturer’s recommended protocol. Single-stranded cDNA was prepared from the extracted total RNA using an oligo (dT) primer and reverse transcriptase (ReverTra Ace; Toyobo, Osaka, Japan). Synthesized cDNA (200 or 400 μg) was used as the template for the PCR amplification. Specific primers were apoE forward (5′-ATTAAGCTTATGAAGGCTCTGTGGGCCGTG-3′) and apoE reverse (5′-GAAGAATTCTCATTGATTCTCCTGGGCCAC-3′); apoAI forward (5′-CCCAAGCTTATGAAAGCTGTGGTGCTGGCC-3′) and apoAI reverse (5′-GCGGGATCCTCACTGGGCAGTCAGAGTCTC-3′); glyceraldehyde 3-phosphate dehydrogenase (G3PDH) forward (5′-ACCACAGTCCATGCCATCAC-3′) and G3PDH reverse (5′-TCCACCACCCTGTTGCTGTA-3′). PCR amplifications were performed with DNA polymerase (ExTaq; Takara, Shiga, Japan). PCR program parameters of apoE cDNA amplifications were 94°C for 10 minutes, followed by 25 cycles of 20 seconds at 98°C for DNA denaturing and primer annealing and 3 minutes at 68°C for extension. The reaction was terminated after 10-minute extension at 72°C. For apoA-I, parameters were the same except that the number of cycles was 30. PCR products were separated on 1% agarose gel and stained with ethidium bromide. Results were analyzed digitally (BioDoC-It; UVP, Upland, CA). 

In Situ Hybridization

Mouse eyes were fixed with 4% paraformaldehyde in diethylpyrocarbonate- treated PBS at 4°C for at least 8 hours. After immersion in 15% and 30% RNase-free sucrose/PBS and in OCT compound, frozen tissue blocks were sliced into 7-μm sections. A HindIII–BamHI and a HindIII–EcoRI fragment corresponding to a full-length of apoA-I and apoE cDNA, respectively, were ligated into the indicated site of the predigested phagemid vector (pBluescriptII SK+; Stratagene, La Jolla, CA). Digoxigenin-labeled antisense and sense RNA probes were generated by transcribing from the T3 and T7 promoters, respectively (Riboprobe Combination System T3/T7; Promega, Madison, WI) after linearization by digestion with HindIII, BamHI, or EcoRI. These RNA probes were hybridized on the retinal sections at 72°C overnight. We found that this temperature was optimal to rule out the nonspecific binding of the sense RNAs. The probe on the sections was detected using alkaline phosphatase–conjugated antidigoxigenin (Roche Diagnostics, Mannheim, Germany) by a blue 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium color reaction. 

Results

Association of Apolipoprotein A-I and Apolipoprotein E in Developing Retina

Retinal cells have highly complex membranes such as neuronal extensions and rod disks. To gain insight into membrane differentiation, isolated retinas were analyzed for phospholipid and cholesterol. The content of these lipids continuously increased until the retina reached maturity (P27). In contrast, the protein content peaked and reached plateau by approximately P7 (Fig. 1A) , indicating that the retinal membranes began to enrich lipids at approximately that time. At this time point, our mass analysis also indicated that phospholipids increased their chain polyunsaturation (Nakanishi H, et al., manuscript in preparation). 

In addition to lipids, Western blot and RT-PCR analyses indicated drastic changes in the expression of apolipoproteins (Figs. 1B 1C) . Remarkably, we detected apoA-I protein until P3. Its mRNA was present at P0, suggesting autogenous synthesis of the protein. As a putative receptor, SR-BI was detected until P5 to P7. In contrast, ABCA-1 was hardly detectable, even by an enhanced biotin-avidin technique (data not shown). Moreover, in parallel with the changes in lipids, we found that the expression of apoE protein and mRNA began to increase at P7 and P5, respectively (Figs. 1B 1C) . 

At P5 to P7, the inflection point of the association of characteristic proteins was reached (Fig. 1B) . Expression of postsynaptic density protein 95 (PSD-95), a protein enriched in excitatory synapses, ²⁶ increased with kinetics similar to that of apoE. A dendrite protein, microtubule-associated protein 2 (MAP2), was also similarly expressed (data not shown). Of two marker proteins of glia cells, glial fibrillary acidic protein (GFAP; astrocytes) ²⁷ was expressed with similar kinetics, and cellular retinaldehyde binding protein (CRALBP; Müller cells) became detectable at P5. These results suggested that apoE was expressed in parallel when the retina formed. 

Distribution of apoA-I and SR-BI over Whole Prenatal and Postnatal Retinas

Next, we examined the spatial distribution of apoA-I in retinal sections by immunofluorescence (Figs. 2A 2B 2C 2D) . In addition to the postnatal eyes, we examined prenatal ones. At embryonic day (E)15, 2 days after formation of the optic cup, very weak apoA-I immunoreactivity was present (Fig. 2A) . At E17, stronger immunoreactivity outlined the cells (Fig. 2B) . At P0 and P3, the overall immunoreactivity became weaker (Figs. 2C 2D) , and it became undetectable at P5 (data not shown). 

To identify apoA-I–producing cells, in situ hybridization was performed (Figs. 2I 2J 2K 2L) . To rule out nonspecific binding of the probe, incubation was initiated at 72°C. Under this condition, hybridization was significant only in the E17 section (Fig. 2J) . Nearly all cells expressed mRNA in a manner similar to that for protein expression. Although RT-PCR detected apoA-I mRNA in the P0 retina, specific in situ hybridization was not significant in this section or the E15 section (Figs. 2I 2K) . 

ApoA-I distribution was compared with that of SR-BI by double immunofluorescence (Figs. 2A 2B 2C 2D 2E 2F 2G 2H) . SR-BI expression also reached peak at E17. Fluorescence was more intense in the outer neuroblast layer (Fig. 2F) . From P0 onward, SR-BI immunoreactivity became weak, in a manner similar to that of immunoreactivity (Figs. 2G 2H) . These results suggested that regulation of expression of these two proteins were related, though not identical. 

Expression of apoE mRNA in Neuroblasts, Astrocytes and RPE, but Not Ganglion Cells in Early Postnatal Retinas

In contrast to that of apoA-I, apoE mRNA expression was not apparent in E17 retinas (Fig. 3J) . Remarkably, apoE mRNA was clearly detected at P0 (Fig. 3K) . It was expressed in the outer half of the retinal layer and in a smaller number of cells in the ganglion cell layer (Fig. 3K ; small arrows). ApoE mRNA was also expressed in RPE (Fig. 3K ; arrowheads). Cells expressing apoE mRNA in the ganglion cell layer were presumably astrocytes because enhanced fluorescence images revealed the overlap of apoE and GFAP immunoreactivities (Figs. 3G 3H) . However, the actual level of increase in apoE immunoreactivity in P0 retina was only slightly higher than in the control (Fig. 3D , red fluorescence; compare with Fig. 3A ). Moreover, we noticed that expression levels of apoE protein in the subretinal regions did not correlate directly with those of the mRNA. Even in an enhanced image shown by green fluorescence, neuroblasts in the outer retinal layer were still low in significant immunofluorescence (Fig. 3D) . In contrast, immunofluorescence was stronger in RPE and GCL, suggesting that the translation of apoE mRNA was differently controlled in these cells. Moreover, significant immunoreactivity in ganglion cells that lacked mRNA again suggested delivery of the protein from other cells. 

At P3 and P5, expression levels of apoE mRNA became clearly higher in cells that aligned in the middle retinal layer (Figs. 3L 3M) . However, apoE immunoreactivity in this region increased only moderately (Figs. 3E 3F) . Translation of the mRNA was therefore still not very active. Although we subsequently found centralized expression of apoE in Müller glia cells, the marker protein, CRALBP, was also not expressed in the central zone of the retina at these time points (data not shown). 

Although SR-BI was present (Fig. 3J) , no significant immunoreactivity of LRP family proteins—i.e., LDLR, LRP1, megalin, VLDLR, apoER2, and LRP6—was detected during these early time points (data not shown). 

Concentrated Production of apoE Protein in Growing Müller Cells and RPE during Stratification

In the P7 retina, apoE immunoreactivity outlined elongating cell bodies that distributed in the central zone (Fig. 4B) . Expression of apoE mRNA was progressively accumulated in the same region (INL; Figs. 4J 4K 4L 4M ). From P10 onward, expression of the protein in these cells also clearly increased. Remarkably, the protein was further localized to the outer plexiform layer (OPL), outer nuclear layer (ONL), and contours of photoreceptor cells (Figs. 4D 4F 4H) . Lack of apoE mRNA in these structures (compare with Figs. 4K 4L 4M ) indicated that an intercellular distribution mechanism began to function. 

To assess this mechanism, we tried to determine the identity of apoE in plexuses and photoreceptor cells. We compared the distribution pattern of apoE immunoreactivity with that of CRALBP. In P7 retina, an indication of coexisting immunoreactivities in INL cells was apparent (Figs. 4B 4C) . From P10 onward, the overlap was seen not only in this region but also in OPL, photoreceptor cells, and some extensions in the inner plexiform layer (IPL; Figs. 4D 4E 4F 4G 4H 4I ), suggesting that apoE in these structures was directly delivered by Müller cell processes. 

From P15 onward, strong apoE immunoreactivity was present in the RPE (upper perimeters in Figs. 4F 4H ). This timing was, however, later than the emergence of apoE in OPL and the inner part of ONL (P10), suggesting that apoE synthesized by RPE was not transported to the plexuses. In contrast, its distribution to outer segment (OS) was very likely occurring. 

In addition to these structures, we localized strong binding of the anti–CRALBP mAb (clone B2; Ref. ²⁷ ) to the innermost part (Fig. 4C) . Such intense immunofluorescence was not seen at P27 (Fig. 4I) . Its identity is unknown. 

Expression of LRP1 in Maturing Retinas

In parallel with the infiltration of apoE, we found that LRP1 began to be expressed in many retinal substructures. Although this protein was hardly detectable before P7 (a result at P5 is shown in Fig. 4N ), it was present in OPL and on the contours of many neurons and photoreceptor cells from P10 onward (Figs. 4O 4P 4Q) . Some apoE/LRP1 immunoreactivities were particulate, particularly in the GCL of the P15 retina (Fig. 4P) . From P15 onward, it was also enriched in OS (Figs. 4P 4Q) . Double immunofluorescence experiments revealed that all these structures were associated with apoE immunoreactivity (data not shown but compare with Figs. 4D 4F 4H ). In contrast, we observed no significant immunoreactivity of other LRP family proteins—i.e., LDLR, megalin, VLDLR, LRP6, and apoER2 (data not shown). These results suggested that LRP1 was exclusively expressed as the major receptor of apoE in the stratifying retina. It was likely that Müller cells and neuronal cells coupled through this pairing. 

Accumulation of Fluorescent Derivative of Cholesterol in Particulate apoE

We have previously shown that a cholesterol derivative of fluorescent poly(ethylene glycol) (PEG-Chol) spontaneously distributes in lipid membranes and is eventually accumulated in cholesterol-rich ones. ²⁸ By high-magnification video-enhanced microscopy, we suggested that microdomains containing ligand-bound EGF receptors are rich in cholesterol. ²⁸ By using Alexa Fluor 488–tagged PEG-Chol (A-PEG-Chol), we asked whether particulate apoE protein enriched cholesterol. To this end, sections were treated shortly with A-PEG-Chol without permeabilization of the membrane. The excess probe was removed by heat-treated serum from which denatured lipoprotein particles were cleared by ultracentrifugation. 

At P7, the accumulation of A-PEG-Chol fluorescence was not significant, though apoE immunoreactivity distributed in a spotlike pattern (Figs. 5A 5D) . At P10, however, on the somas of Müller cells, A-PEG-Chol fluorescence was enriched in spots (Fig. 5E , arrows). Notably, many of them were colocalized by apoE immunoreactivity (Fig. 5B) . Fluorescence was from the cell surface because pretreatment with digitonin for surface permeabilization resulted in strong intracellular A-PEG-Chol fluorescence (data not shown). These results suggested that cholesterol was enriched in much membrane-bound apoE. 

In P15 retina, A-PEG-Chol fluorescence was also accumulated, though intense overall fluorescence, which suggested an increase in the average cholesterol level, made the discrimination more difficult (Figs. 5C 5F) . In addition, because of intense fluorescence, we could not specify the enrichment of cholesterol in a spot-like pattern in the plexus region (data not shown). 

Decreased Association of apoE in Astrocytes during Late Maturation

Throughout development, ganglion cells were associated with apoE without expressing mRNA. We further studied the association of apoE in vertical sections and in explants that had been spread over nitrocellulose filters (Fig. 6) . In vertical sections, the number of cell bodies with GFAP immunoreactivity appeared similar from P5 to P27 (Figs. 6B 6D 6F 6H) . In contrast, the level of apoE gradually decreased (Figs. 6A 6C 6E 6G) . A similar change was observed on the GCL face of the retinal explants (Figs. 6I 6J 6K 6L) . At P7, strong apoE immunoreactivity associated with thin, long astrocyte processes that extended over the less fluorescent ganglion cell layer (Fig. 6I) . At P15, density of GFAP-bound processes decreased; many were devoid of apoE immunoreactivity, suggesting that the role of astrocytes as a source of apoE declined as the retina matured. 

Expression of apoE in Primary Retinal Culture

To determine whether the differential expression of apoE in the two types of glia cells were intrinsically programmed, we examined the primary culture of cells dissociated from P0 retina (Fig. 7) . GFAP-expressing cells were present as early as day 3 of in vitro culture (Div3) and Div5 (Figs. 7A 7C) . Expression of apoE in such cells reached peak at Div7 (Figs. 7B 7D 7F) . However, its immunoreactivity became weak at Div14 (Fig. 7H) . 

In contrast, CRALBP immunoreactivity was weakly detected in round cells until Div5 (Figs. 7I 7K) . From Div7 onward, CRALBP-expressing cells began to extend and were associated with particulate apoE immunoreactivity (Figs. 7M 7N 7O 7P) . These results suggested that apoE expression was programmed in the differentiation of these two types of glia cells. 

Discussion

Association of Lipoprotein Systems and Lipids in Developing Mouse Retina

To achieve harmonized growth and differentiation of cells, the distribution of membrane lipids must be regulated. Analysis of the intercellular distribution of lipoproteins and the lipid constituent is an important part of the efforts to understand cellular activities supported by membranes. We have been trying to specify lipids and lipoproteins in developing retina by various methods. Here, we found that the postnatal mouse retina remarkably exhibits a nonoverlapping presence of apoA-I/SR-BI and apoE/LRP1 during development. ApoA-I and apoE genes are independently present in distinct gene clusters. ²⁹ Our RT-PCR and in situ hybridization analyses also suggest differential transcriptional regulation (Figs. 2 3 4) . The switchover from apoA-I/SR-BI to apoE/LRP1, which took place at approximately P5 to P7, corresponded to the time of increase in retinal complexity. Not only the content (phospholipids and cholesterol; Fig. 1 ) but also the diversity (Nakanishi H, et al., manuscript in preparation) of membrane lipids began to increase. These results indicate that the switchover of lipoprotein systems and quantitative/qualitative changes in membrane lipids run parallel courses with retinal development. 

Association of apoA-I/ SR-BI in Undifferentiated Neuroblasts

Interaction between SR-BI and apoA-I has been shown to mediate the selective influx of cholesteryl ester and the efflux of free cholesterol in various cells. ^{30 31 32} It is likely that, after docking of a lipidated apoA-I at the cell surface, this pair of proteins plays a nutritional role by exchanging cholesterol and certain phospholipids. ^{30 33 34 35} Although apoE and LRP1 are implicated in cellular differentiation by the activation of various intracellular signals, apoA-I/SR-BI does not appear to trigger such a reaction. Interestingly, SR-BI in a neural system was shown to uniquely participate in cell-cell and cell-substrate interactions. ^{36 37} These aspects suggest that apoA-I/SR-BI expression may be beneficial for facilitating the synchronized growth of early neuroblasts. 

ApoE/LRP1 Interaction in the Differentiating Retina

In situ hybridization revealed that wide expression of a low level of apoE mRNA in neuroblasts shifted to accumulation in cells that differentiated in Müller cells (Figs. 3 4) . Concomitantly, most neurons and photoreceptor cells expressed LRP1 as a putative receptor (Figs. 4O 4P 4Q) . Early studies suggested the presence of LRP1 in mature rat retina, ³⁸ though the presence of other LRP family proteins had not been examined. We here specified that only LRP1 was expressed as the major protein. In addition to these findings, accumulated A-PEG-Chol fluorescence suggested the need for cholesterol in particulate apoE (Fig. 5) . It has been suggested that glia cell–derived cholesterol is necessary for synapse formation in vitro. ^{39 40} One of these in vitro experiments used a conditioned medium from a culture of retinal glia cells. ⁴⁰ Our present study suggests that such medium may contain apoE-containing lipoprotein from Müller cells (Figs. 6 7) . 

In LRP1-deficient mouse, hypersensitivity, severing, and dystonia were manifested. ¹⁶ LRP1 is therefore strongly suggested to be necessary for synapse formation. The authors suggested that LRP1 formed complexes with PSD-95 and N-methyl d-aspartate receptor (NMDR) in cholesterol-rich lipid rafts. In the present study, expression of PSD-95 increased after P5 (Fig. 1) . Our results imply that Müller cells may escort LRP1-expressing cells for synapse formation by supplying apoE and cholesterol. 

ApoE/LRP1 was also highly expressed in OS (Fig. 4) . The major source of apoE appears to be RPE. Recent studies have documented that drusen, which are the prominent clinical and histopathologic sign of age-related maculopathy of humans, contain a wide variety of apolipoproteins—apoE, J, C-I, B, A-I, A-II, C-II, and C-III. ^{41 42 43} Drusen contain lipoproteinlike particles that do not resemble plasma lipoproteins. ⁴² Studies using RT-PCR and immunofluorescence strongly suggest that the sources of these proteins include RPE and neurosensory retina. ^{42 43} Formation of the plaques is thought to involve deregulated distribution or the expression of these apolipoproteins. ^{41 42 43} Interestingly, distribution patterns of apoE and A-I in the drusen often do not overlap, ⁴³ suggesting different timing of synthesis or distribution of these proteins. It may be interesting to address the spatiotemporal regulation of apolipoprotein expression in aged model animals. 

Distribution of apoE in Ganglion Cell Layer

The present results suggested that ganglion cells did not express apoE mRNA throughout development. They were, however, constantly associated with the protein, which should be derived from other cells (Figs. 3 4) . In early postnatal retina, astrocytes appeared to supply the protein. Although many of these cells ceased to produce apoE (Figs. 6 7) , ganglion cells were still associated with the protein. In adult rabbit retina, apoE secreted by Müller cells is suggested to reappear in the vitreous fluid and eventually to associate with optic fibers of ganglion cells. ⁴⁴ In the present study, particulate apoE appeared to bind to optic fibers (Fig. 6) . Further study may assess a similar mechanism in developing mouse retina. 

In vitro, apoE-containing lipoprotein derived from brain glia cells increased the rate of axon extension in retinal ganglion cells. ⁴⁵ This effect was prominent when the medium was added to distal axon and appeared to depend on LRP signaling. When the lipoprotein is added to cell bodies, growth is not enhanced. ⁴⁵ If a similar regulation operates in the developing retina, the apoE from different cellular sources may exert differential effects. However, conventional apoE-containing lipoprotein particles in the blood plasma are unlikely used by growing retinal neurons. We observed no apoA-I, by far the major component of apoE-containing HDL, in mature retina (Fig. 1and data not shown). 

In brain, cholesterol-rich apoE-containing lipoprotein particles are released after lipidation mediated by ABCA-1. ^{46 47 48} Such apoE is produced by astrocytes and microglia. In retina, however, ABCA-1 was hardly detectable. Whereas ABCA-1 is the most efficient vehicle for transmembrane movement of cholesterol, release of apoE still occurs in ABCA-1(–/–) cells. ⁴⁷ Interestingly, the composition of apoE-containing lipoproteins was recently shown to vary considerably in different producer cells. ^{49 50} Moreover, the degree of aggregation and the conformation of apoE were varied by different bound phospholipids. ^{51 52 53} They affect interactions of this protein with LDL receptor and cholesterol acyltransferase differently. ^{51 52 53} ABCA-1-mediated transfer of cholesterol to HDL does not require a high phospholipid content in the particle but that by SR-BI does. ³⁴ Further study involving the characterization of the retinal apoE-containing particle is necessary for defining its role in development. 

The present study reports that apoA-I and apoE emerge with distinct developmental timing. Our finding may advance its use to control retinal diseases. 

²  Present affiliation: Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, Maryland.

 Supported by research grants from Terumo Life Science Foundation and Research Foundation for Opto-Science and Technology.

 Submitted for publication August 2, 2006; revised October 18 and November 1, 2006; accepted January 25, 2007.

 Disclosure: S. Kurumada, None; A. Onishi, None; H. Imai, None; K. Ishii, None; T. Kobayashi, None; S.B. Sato, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Changes in proteins and lipids in the developing mouse retina. (A) Retinas (10 isolates on average in each assay) were analyzed for phospholipid, cholesterol, and total protein amounts. Average amounts per retinal sheet (out of three independent experiments) are plotted. (B) Western blot analysis of apoE, apoA-I, SR-BI, and retinal proteins PSD-95, GFAP, and CRALBP. The same total protein amounts were analyzed. ApoA-I and SR-BI were present at early postnatal days. In contrast, apoE and other proteins increased the amount during the late stage. (C) The same amounts of RNA extracted from isolated retina at various postnatal days were analyzed for apoE and apoA-I mRNA by RT-PCR. ApoA-I mRNA was detectable at P0. In contrast, the amount of apoE mRNA gradually increased. G3PDH was used as control.

Figure 2.

 

Confocal double-immunofluorescence pairs of apoA-I (A–D) and SR-BI (E–H) and in situ hybridization of apoA-I mRNA (I–L) in prenatal (E15, E17) and postnatal (P0, P3) retinas. ApoA-I and SR-BI proteins associated with similar cells, though the fluorescence intensity profiles in E17 retina were not completely identical. In situ hybridization occurred in the entire cell population at E17 (J), whereas those at E15 (I) and P0 (K) were low. (L) Result with sense RNA in P0 retina. Typical results out of three independent experiments are shown. UDN, undeveloped neuroblasts; GCL, ganglion cell–rich layer; RPE, retinal pigmented epithelial cell layer. Bars, 20 μm.

Figure 3.

 

Immunofluorescence and in situ hybridization of apoE (B–G, I–M), GFAP (H), and SR-BI (O) in prenatal and early neonatal retinas. Detection sensitivities of confocal immunofluorescence are the same in red (A–D) and green (D–F) color groups, respectively. (A) Negative reaction using anti–ABCA-1 antibody on P0 retina is shown. (D) Independent results using the same primary antibody but different secondary ones. (B–D, red) Low apoE immunoreactivity in prenatal retinas became only slightly higher in P0 retina. (D–E, green) In neonatal retinas, it increased in UDN, GCL, and RPE cells. (G, H) Video-enhanced double immunofluorescence of apoE and GFAP partially overlapped in the inner part of P0 retina. (I–M) ApoE mRNA was not significantly present in prenatal retinas but was present in postnatal retinas. It was enriched in astrocytelike cells (K, small arrows) and pigmented epithelial cells (arrowheads) and in the outer retinal layer. Expression of apoE mRNA was concentrated in the middle retinal zone at P3 and P5 retinas. ApoE mRNA was not expressed in ganglion and amacrine cells. (N) A control result in P0 using a sense sequence. (O) Immunoreactivity of SR-BI in P5 retina. Bars, 20 μm.

Figure 4.

 

Confocal double-immunofluorescence results of apoE (A, B, D, F, H) and CRALBP (C, E, G, I), in situ hybridization of apoE mRNA (J–M) and confocal immunofluorescence of LRP1 (N–Q) in stratifying retina. Imaging sensitivities are the same in each color group. (A) Image representing association of apoE in P3 retina. (B, D, F, H) ApoE immunoreactivity was progressively localized to various retinal substructures. (J–M) ApoE mRNA was expressed only in cells in INL and RPE. (N–Q) Immunoreactivity of LRP1 was hardly detectable at P5 but was associated with various subretinal structures from P10 onward. Typical results from three independent experiments are shown. Bars, 20 μm.

Figure 5.

 

Video-enhanced double fluorescence pairs showing distribution of apoE immunoreactivity (A–C) and A-PEG-Chol (D–F) on the surfaces of INL cells. At P10, many spots of A-PEG-Chol fluorescence were associated with apoE immunoreactivity (arrows). At P15, fewer spots of A-PEG-Chol were discriminated against more intense fluorescence. Bar, 5 μm.

Figure 6.

 

Distribution of apoE (A, C, E, G, I, K) and GFAP (B, D, F, H, J, L) in retinal sections (A–H) and explants (I–L). Confocal double-immunofluorescence photographs are shown at fixed imaging sensitivity and optical thickness. ApoE immunoreactivity was reduced at P15 and P27. Processes associated with GFAP immunofluorescence in P27 section were not associated with apoE. (I–L) Ganglion cell faces of the expanded retina were examined. Bars, 20 μm.

Figure 7.

 

Distribution of GFAP (A, C, E, G), CRALBP (I, K, M, O), and apoE (B, D, F, H, J, L, N, P) studied by confocal double immunofluorescence in primary retinal cultures. Cells isolated from P0 retinas were cultured on coverslips for indicated days. ApoE immunoreactivity was significant in cells associated with GFAP since Div3 (D, F) but was reduced at Div14 (H). In contrast, it became significant in cells with CRALBP immunoreactivity only at Div7 (N) and Div14 (P). Bars, 20 μm.

The authors thank Yoshinori Fujiyoshi and Yoshinori Shichida for their support and comments.