implications

The principal extracellular lesions of age-related maculopathy (ARM), the leading cause of vision loss in the elderly, involve Bruch's membrane (BrM), a thin vascular intima between the retinal pigment epithelium (RPE) and its blood supply. With age, 80-100 nm solid particles containing esterified cholesterol (EC) accumulate in normal BrM, and apolipoprotein B (apoB) immunoreactivity is detectable in BrM- and ARM-associated lesions. Yet little evidence indicates that increased plasma cholesterol is a risk factor for ARM. To determine if RPE is capable of assembling its own apoB-containing lipoprotein, we examined RPE for the expression of microsomal triglyceride transfer protein (MTP), which is required for this process. Consistent with previous evidence for apoB expression, MTP is expressed in RPE, the ARPE-19 cell line, and, unexpectedly, retinal ganglion cells, which are neurons of the central nervous system. De novo synthesis and secretion of neutral lipid by ARPE-19 was supported by high levels of radiolabeled EC and triglyceride in medium after supplementation with oleate. Lipoprotein assembly and secretion is implicated as a constitutive retinal function and a plausible candidate mechanism involved in forming extracellular cholesterol-containing lesions in ARM. The pigmentary retinopathy and neuropathy of abetalipoproteinemia (Mendelian Inheritance of Man 200100; Bassen-Kornzwieg disease), which is caused by mutations in the MTP gene, may involve loss of function at the retina.

signals for transmission to the brain.The photoreceptors are supported by the retinal pigment epithelium (RPE), a monolayer with diverse functions including daily phagocytosis of the distal tips of photoreceptor outer segments (OS), and the choroid, a vascular bed with the body's highest blood flow.The choriocapillaris is a dense capillary plexus in the innermost choroid, and Bruch's membrane (BrM) is a thin vascular intima between the RPE and the choriocapillaris (Fig. 1B, arrowheads).In human retina, the macula subserves high-acuity vision and is vulnerable to age-related maculopathy (ARM), the major cause of vision loss among the elderly of industrialized societies.The most prominent histopathologic and clinical signs of ARM are extracellular lesions [drusen (Fig. 1E, F) and basal linear deposits (not shown)] in the RPE/BrM complex that ultimately impact vision by the photoreceptors (1,2).Choroidal neovascularization, an invasion of choriocapillaries across BrM and lateral spread within the plane of drusen and basal linear deposit (see 3), is the principal sight-threatening complication of ARM's obscure underlying degeneration.
Recent findings highlight a role for lipids and lipoproteins in this degeneration.These include a protective effect of the apolipoprotein E4 (apoE4) genotype in populations and the presence of apoB and apoE and histochemically identified lipids in aging-and ARM-associated drusen and deposits in human tissues (4-8) (Fig. 1E, F).The best established risk factor for early ARM is advanced age (9).A large age-related change in normal human retina is the accumulation of neutral lipids in BrM throughout adulthood (7,(10)(11)(12)(13).This material binds oil red O and contains esterified cholesterol (EC) (Fig. 1D) and triglyceride (TG) (7,(10)(11)(12).Furthermore, EC is associated with 80-100 nm diameter solid particles (Fig. 1C) that have surface and core substructure (12)(13)(14).Understanding the pathway(s) that lead to neutral lipid deposition in normal human BrM is critical for understanding the pathway(s) that lead to drusen and basal linear deposits in ARM.Studies of BrM lipid composition (11,12,15) implicate extravasation of plasma lipoproteins from the choroid or deposition from intraocular cells as candidate mechanisms.The RPE processes abundant lipids from ingested OS membranes (16).It synthesizes and secretes BrM constituents (17,18) and contains biosynthetic machinery for numerous druse-associated proteins (19,20), including apoE (6,21).However, a mechanism to account for the abundant EC and other neutral lipids in normal BrM is unknown.
Atherosclerotic cardiovascular disease and ARM feature apoB-and cholesterol-containing extracellular lesions in a vascular intima and shared risk factors of smoking and hypertension (22,23), among others.However, only 1 epidemiologic study of 23 since 1963 (24) links ARM with increased total plasma cholesterol (see supplementary table), a defining risk factor for atherosclerosis (25).Although not the strongest indicator of cardiovascular risk, total plasma cholesterol was the one measure common to all of these investigations.Many studies were small and used different ARM end points.However, if plasma lipoproteins were the major source of cholesterol in BrM and sub-RPE deposits, then a more consistent association should have been detected despite these limitations.This paradox could be reconciled if an apoB-containing lipoprotein of intraocular origin were the major source of early-arriving EC in BrM-and ARM-associated lesions, in contrast to apoB-containing lipoprotein from plasma that arrives early in incipient atherosclerotic plaques (26,27).This idea is credible, because native human RPE contains apoB mRNA and protein (8,19), and the thermal behavior of birefringent EC crystals in BrM differs from that in sclera, which accumulates EC from plasma apoB-containing lipoprotein (28).
If the RPE synthesizes, assembles, and secretes an apoBcontaining lipoprotein with a neutral lipid core, it should also express microsomal triglyceride transfer protein (MTP).This soluble heterodimer in the endoplasmic reticulum lumen is required for apoB-containing lipoprotein assembly (29, 30).Cells expressing apoB without MTP cannot secrete lipoproteins, and all cells expressing MTP also secrete apoB-containing lipoprotein (31)(32)(33)(34).MTP mutations resulting in a lack of functional protein cause abetalipoproteinemia (ABL; MIM 200100, Bassen-Kornzwieg disease), a rare autosomal recessive disorder.ABL features the absence of plasma apoB-containing lipoprotein, fat malabsorption, steatorrhea, acanthocytosis, ataxic neuropathy, and, importantly, a pigmentary retinopathy (31,35).To begin evaluating our hypothesis, we determined whether MTP was present in human neurosensory retina, native RPE, and ARPE-19, a spontaneously arising transformed cell line with differentiated properties (36)(37)(38).We then determined that ARPE-19 could secrete EC-containing particles, confirming these results in primary porcine RPE.Of a lipoprotein's surface and core compo- nents, we focused on EC, because it is localized exclusively within BrM (Fig. 1D).Our results have implications for ARM pathobiology, retinal cell biology, and ABL.

Human tissue
Human donor eyes with grossly normal maculas were obtained from the Alabama Eye Bank within 6 h of death (n ϭ 8, 65-85 years for RT-PCR; n ϭ 9, 30-92 years for Western blot; n ϭ 9, 38-89 years for immunohistochemistry).For immunohistochemistry, anterior segments of donor globes were removed by encircling cuts at the corneoscleral limbus.Posterior segments were immersed in 4% paraformaldehyde in 0.1 M phosphate buffer for 6-16 h and stored in 1% paraformaldehyde at 4 Њ C until used.To isolate retinal mRNA and protein, globes were incised at the equator, and anterior segments were removed with vitreous body attached.Small forceps were used to peel the retina from the RPE/choroid while the scleral eyecup was stabilized with another forceps.

Cell culture
ARPE-19 and HepG2 cell lines were obtained from the American Type Culture Collection at passage 22 and subjected to RT-PCR, Western blot, and secretion experiments after 2 passages.ARPE-19 cells were also provided by Dr. N. J. Philp (Thomas Jefferson University) (37).For RT-PCR and Western blot, cells were plated in T-75 flasks or on six-well plates and grown for 4 weeks in DMEM/F12 (1:1) supplemented with 10% fetal calf serum as described (36).Medium was changed twice weekly.HepG2 cells were grown on six-well plates in MEM containing 10% fetal calf serum for 5 days with a medium change every other day.Primary cultures of porcine RPE were prepared as described (39) and used 24 h after isolation.

RT-PCR and sequencing
Total RNA was isolated from human retina, RPE, ARPE-19 cells, and HepG2 cells as described (8).Total RNA from porcine retina, RPE, and liver was isolated with Trizol Reagent (Invitrogen, Carlsbad, CA).Human ileum mRNA was purchased from Stratagene.Primers used for RT-PCR are listed in Table 1 .To distinguish between amplified mRNA and genomic DNA, primers were designed to span intron boundaries except where noted.One-step RT-PCR was used for human apoB and porcine apoB and MTP, as described (8).Two-step RT-PCR was used for others.The first-strand cDNA was synthesized with Omniscript™ Reverse Transcriptase (Qiagen, Valencia, CA).The PCR Core System (Promega, Madison, WI) was used for PCR.The cDNA was denatured for 4 min at 94 Њ C before cycling.The reaction was amplified through 35 cycles of 45 s at 94 Њ C (denaturing), 45 s at 55-66 Њ C (annealing), and 1 min at 72 Њ C (extension), then incubated for 10 min at 72 Њ C. Cross-contamination of retinal and RPE mRNA was assessed by checking the expression of the retina-specific gene RHO in RPE mRNA and the RPE-specific gene VMD2 in neurosensory retina mRNA.All results for neurosensory retina and RPE were obtained in at least one sample without contamination.The primers 5 Ј -GAG AAA CTG ACT GCT CTC AC-3 Ј (sense) and 5 Ј -ATG ATA GTG CTC ATC AAG ACT T-3 Ј (antisense) were used to amplify a 234 bp human apoB cDNA fragment containing the editing site of apoB-48.The cDNA sequence was analyzed as described (8).

Western blot
Protein extracts of human retina, RPE, ARPE-19 cells, and HepG2 cells were prepared and Western blot analysis was performed as described (8).Polyclonal rabbit anti-bovine MTP 97 kDa large subunit (40) was a gift from Dr. J. R. Wetterau at Bristol-Meyers-Squibb (used at 1:500).Monoclonal anti-human apoB [1D1, recognizing an N-terminal epitope between amino acids 474 and 539 (41)] was a gift from Dr. R. Milne at the University of Ottawa (used at 1:1,000).Secondary antibodies, horseradish peroxidase-conjugated donkey anti-rabbit IgG, and donkey anti-

Immunohistochemistry
Cryosections of human retina and choroid were prepared as described (8).Except where noted, reagents were purchased from Vector Laboratories (Burlingame, CA).For localizing apoB, MTP, and monocarboxylate transporter 3 (MCT3) in human RPE, sections were bleached before staining as described (42).Briefly, cryosections were removed from Ϫ 20 Њ C, heated for 30 min at 50-55 Њ C, and rinsed with PBS.Slides were incubated in 0.05% potassium permanganate for 25 min, rinsed for 5 min in distilled water, flooded with 35% peracetic acid in a humidified container for 15 min, and washed in distilled water for 10 min, all at room temperature.Next, sections were heated at 50-55 Њ C for 30 min and incubated with the primary antibodies anti-apoB (Biodesign, Saco, ME; 1:500), anti-MTP (1:100), and anti-MCT3 (1:100; a gift from Dr. N. J. Philp) followed by biotinylated secondary antibodies, peroxidase-conjugated streptavidin, and the VIP Peroxidase Substrate Kit.Appropriate controls were processed with stained sections.For localizing apoB, MTP, and Brn3a in neurosensory retina, immunohistochemistry was performed as described (43).Briefly, sections were incubated overnight at 4 Њ C with primary antibodies against apoB (1:1,000), MTP (1:100), and Brn3a (Chemicon, Temecula, CA; 1:100).Next, biotinylated secondary antibodies and the ABC-Peroxidase Kit or the ABC-Alkaline Phosphatase Kit were applied.Immunoreactivity was visualized with the VIP Peroxidase Substrate Kit for MTP and apoB and the Blue Alkaline Phosphatase Substrate Kit III for Brn3a.For double labeling, sections were incubated first with anti-Brn3a and alkaline phosphatase substrate, then with anti-apoB or MTP and VIP substrate.For apoB and MTP fluorescence double labeling, sections were incubated with biotinylated secondary antibody and allophycocyanin-conjugated streptavidin (eBioscience, San Diego, CA) for apoB (at 1:500) and FITC-conjugated goat antirabbit IgG (Jackson Immunoresearch) for MTP.

Extraction of lipids from cells and medium
Lipids were extracted from culture medium using chloroformmethanol-water (14:7:1) (44).Aliquots of organic phase were evaporated under nitrogen and solubilized in isopropanol.Lipids were extracted from PBS-washed cells on six-well plates using two 30 min incubations with 1.0 ml of hexane-isopropanol (3:2, v/v).Aliquots of pooled extracts were evaporated under nitrogen, solubilized in isopropanol, vortexed, and sonicated for 15 min.Remaining cellular material was digested in 0.1 N NaOH/0.1% SDS to determine protein content (DC Protein Assay Kit; Bio-Rad).

Enzymatic fluorimetric assay for total and unesterified cholesterol
Total cholesterol (TC) and unesterified cholesterol (UC) mass in extracts of native RPE and cultured cells was determined by enzymatic fluorimetric assay as described (12,45).The resulting fluorescence was read at 325 nm excitation wavelength and 415 nm emission wavelength using a model F-2500 Fluorescence Spectrophotometer (Hitachi).The EC value was defined as the difference between the TC and UC values.UC solubilized in isopropanol was used as a standard.

Conditioned medium experiments
ARPE-19 cells were cultured on six-well dishes for 2 weeks.After cells were washed, 1 ml of DMEM/F12 containing 0.4 mM oleic acid and 1.5% FFA-free BSA, prepared as described (46), was added to each well.After 24 h of incubation, the medium was collected and centrifuged at 5,000 rpm for 5 min to remove cellular debris.The supernatant was withdrawn for lipid extraction and analysis of cholesterol mass and composition.

Metabolic labeling experiments
ARPE-19 cells were cultured for 2 weeks, maintenance medium was removed, and cells were washed twice with PBS.Serum-free DMEM containing 5 Ci/ml [ 3 H]oleate (5,000 dpm/nmol) bound to 1.5% FFA-free BSA was added.After 24 and 48 h, medium was collected and centrifuged at 5,000 rpm for 5 min to remove cellular debris, and the supernatant was withdrawn.Lipids were extracted from medium and PBS-washed cells (see above).Extracts were evaporated under nitrogen, solubilized in chloroform, and applied to LHPKD silica gel 60A TLC plates (Whatman).Plates were developed in petroleum ether-diethyl etheracetic acid (84:15:1), sprayed with 3% copper acetate in 8% phosphoric acid solution, and heated to reveal bands.Chloroform-solubilized oleic acid, UC, cholesteryl oleate, triolein, and trilinolenin served as standards.Bands corresponding to EC, TG, and phospholipid were excised, and radioactivity was quantified by liquid scintillation counting.Blank areas between bands were counted as background.A band at the UC position was not reliably separated from diglycerides in this system.Results from two to three wells are reported for one experiment.

Density gradient ultracentrifugation and negative stain electron microscopy
trifuge using an SW41Ti rotor.Supernatant (0.2 ml) was drawn from the tube top.Lipoprotein particles in the d р 1.21 g/ml fraction were visualized as described (47) with modifications.Samples were dialyzed overnight in 0.125 M ammonium acetate, 2.6 mM ammonium carbonate, and 0.26 mM tetrasodium EDTA (pH 7.4, 4 Њ C) and then mixed with equal volumes of 2% potassium phosphotungstic acid.A droplet was placed in the center of a Formvar-coated copper grid freshly decharged in 100% ethanol and dried for 30 s.The droplet edge, with concentrated particles, was viewed on a JEOL 1200 electron microscope.Samples from the d р 1.21 g/ml fraction and medium of nonsupplemented cells were also examined.

Expression of apoB and MTP in human retina
To test the possibility that neurosensory retina (19), in addition to RPE (8), contains mRNA transcripts for apoB, we performed RT-PCR analysis of RNA obtained from neurosensory retina, native RPE, ARPE-19 cells, and human hepatoma HepG2 cells as a positive control.Figure 2A shows RT-PCR products consistent with the predicted 419 bp in all lanes.ApoB-100 and apoB-48 are products of the same gene, with apoB-48 arising from posttranscrip- tional C-to-U substitution editing of apoB-100 mRNA by apoB-editing complex-1 (apoBEC-1) (48).To determine if apoB mRNA from human retina and RPE specifies apoB-100 or apoB-48, a 234 bp RT-PCR product containing the RNA-editing site was sequenced from both 5 Ј and 3 Ј ends.Sequencing revealed that apoB cDNA from human retina, RPE, ARPE-19 cells, and HepG2 cells contains a CAA codon, signifying apoB-100, and not a TAA stop codon, as in human ileum (Fig. 2B).Accordingly, we detected apoBEC-1 mRNA in ileum but not in retina, RPE, or HepG2 cells (Fig. 2A).
To test whether RPE and retina express mRNA for MTP, RT-PCR was performed using total RNA isolated from retina, RPE, ARPE-19 cells, and HepG2 cells as a positive control.RT-PCR products at the expected 698 bp size were generated in all lanes.In addition to transcripts for the 97 kDa large subunit MTP unit, all lanes showed transcripts at 715 bp for protein disulfide isomerase, the smaller subunit of the MTP heterodimer (Fig. 2A).
To determine if human retina and RPE express apoB and MTP proteins, detergent-extracted retina, RPE, ARPE-19, and HepG2 were subjected to SDS-PAGE and probed with a monoclonal antibody against human apoB (Fig. 2C) and a polyclonal antibody against bovine MTP 97 kDa large subunit (Fig. 2C, D).ApoB was detected in retina, RPE, ARPE-19, and HepG2, as indicated by bands at 512 kDa with faint bands at lower molecular masses.The large subunit of MTP was detected in all samples, indicated by prominent bands at 97 kDa (Fig. 2D).

Localization of apoB and MTP in human RPE
To localize apoB and MTP in highly pigmented native human RPE, we treated cryosections with potassium permanganate and peracetic acid to bleach melanin (49).ApoB and MTP immunoreactivity was detected in moderately sized clumps in RPE cell bodies ( Fig. 3A, C , re-   (19), and HepG2 (H) cells.Total RNA was isolated and RT-PCR was performed as described in Materials and Methods.Expected lengths of RT-PCR products are listed in Table 1.ACAT1 was detected in human retina, RPE, ARPE-19, and HepG2.ACAT-2 was detected in HepG2 only.M, 100 bp DNA ladder; arrowhead, 500 bp.spectively), in contrast to control sections treated with the corresponding control immunoglobulins (Fig. 3B, D, respectively).Little apoB and MTP was detectable in unbleached sections (data not shown).To verify that this result was unrelated to the bleaching procedure, we examined immunoreactivity for MCT3, a basolaterally located, RPE-specific monocarboxylate transporter (37).Prior bleaching did not perturb the location of this antigen, and no intracellular granules were revealed (Fig. 3E, F).The coarse particulate labeling for apoB and MTP differed in size, shape, and distribution from both melanin granules (Fig. 3F) and autofluorescent lipofuscin granules (Fig. 3G).We conclude from these results that apoB and MTP are present in RPE, likely in the same subcellular compartment.

Expression of ACAT in retina and RPE
The above data support the hypothesis that EC-containing particles in human BrM could be particles with neutral lipid cores assembled and secreted by the RPE.To determine the capacity of RPE for esterifying cholesterol to long-chain fatty acids, we examined the expression of genes encoding ACAT-1 and ACAT-2, enzymes that catalyze cellular cholesterol esterification.In neurosensory retina, native RPE, ARPE-19, and HepG2, we detected the expected 413 bp RT-PCR products of ACAT-1 ( Fig. 4 ).Despite our use of three different intron-spanning probes, the 494 bp RT-PCR product of ACAT-2 was detected only in HepG2 (Fig. 4).

Synthesis and secretion of neutral lipid-containing particles by ARPE-19
Because we found identical mRNA transcripts and apoB and MTP proteins in ARPE-19 cells as well as native RPE, we analyzed EC mass and composition in ARPE-19 cells.We compared the results with HepG2 cells, a widely used hepatocyte model system for apoB-containing lipoprotein secretion.Table 2 shows that ARPE-19 contains EC, although at much lower absolute (3.5 g/mg protein) and relative (14% of TC) levels than HepG2 (14.6 g/mg and 48.7%, respectively).Table 3 shows that by electrospray ionization mass spectrometry (ESI/MS), HepG2 contained 10-to 26-fold higher mass of cholesteryl palmitate, cholesteryl linoleate, cholesteryl oleate, and cholesteryl stearate than ARPE-19 and similar mass of cholesteryl arachidonate and cholesteryl docosahexanoate.Of these esters, both cell lines were relatively enriched in cholesteryl oleate, much more so in HepG2 than in ARPE-19.
To assess the ability of ARPE-19 cells to secrete EC, cells were incubated with serum-free medium containing oleic acid bound to BSA, and EC concentration in the conditioned medium was determined.In a series of three experiments, we detected 491 Ϯ 35 g/mg cell protein of TC, 234 Ϯ 11 g/mg cell protein of UC, and 257 Ϯ 25 g/mg cell protein of EC in conditioned medium.Thus, 52% of cholesterol recovered was esterified.In separate experiments, conditioned medium was subjected to density gradient ultracentrifugation.Esters were found only in the d р 1.21 g/ml fraction of conditioned medium and not in any fraction of nonconditioned medium (data not shown).Table 4 shows that cholesteryl esters were detectable by ESI/MS in conditioned medium at levels 3-to 12fold higher than in nonconditioned medium, depending on ester identity, with cholesteryl oleate most abundant.
To assess the de novo synthesis of lipids, ARPE-19 cells were metabolically labeled with [ 3 H]oleate-BSA and lipids were detected with TLC and scintillation counting.Table 5 shows representative results from five independent experiments in which ARPE-19 cells were supplemented with radiolabeled oleic acid for 24 or 48 h.Radioactivity was high in bands corresponding to phospholipid, TG, and EC, relative to intervening bands.The medium contained more newly synthesized TG than EC at both time points.Twice as much radiolabeled TG and EC appeared in the medium after 48 h of incubation than after 24 h of incubation, consistent with sustained oleate uptake and conversion into TG and EC.In contrast, levels of newly synthesized phospholipid in the medium changed little during this time.To determine if EC in the d р 1.21 g/ml fraction of conditioned medium of oleate-supplemented ARPE-19 was associated with spherical particles, we used negative stain electron microscopy.Particles were detectable in medium ( Fig. 5A ) at a low abundance consistent with the low concentration of EC.For comparison, Fig. 5B shows plasma VLDL from a slightly hypertriglyceridemic subject at the same magnification.In contrast, particles were not detected in any fraction of nonconditioned medium.

ApoB/MTP expression and neutral lipid secretion by porcine RPE
Because the above experiments showed lipid secretion from a transformed cell line, we sought evidence that the apoB/MTP system was detectable in cultured primary RPE cells.The pig eye is advantageous for retina research, although it lacks a macula, because its large size and dual retinal vasculature resemble those of humans.To test whether porcine RPE and retina express mRNA for apoB and MTP, RT-PCR was performed, using porcine liver as a positive control.Figure 6 shows that the expected RT-PCR products for apoB and MTP, 351 and 336 bp, respectively, were found.Table 6 shows results representative of three metabolic labeling experiments using oleate-BSA and porcine RPE, demonstrating newly synthesized lipids in cells and medium at much higher levels (per microgram of cell protein) than in ARPE-19 (Table 5).

Localization of apoB and MTP in human neurosensory retina
The results presented to this point support our hypothesis regarding RPE, but Fig. 2 also indicates strong signals for apoB and MTP message and protein in neurosensory retina.To localize these proteins, cryosections of normal human macula were probed with polyclonal antibodies ( Fig. 7A ).Immunoreactivity was found in retinal ganglion cell (RGC) layer cells (Fig. 7B) and a few regularly spaced cells of the inner nuclear layer (INL; Fig. 7C).To determine the identity of these cells, double-label experiments were performed using a monoclonal antibody to Brn3a, a RGC marker (50) (Fig. 7D, E).All apoB-positive and MTPpositive cells in the RGC layer and INL were also Brn3apositive.ApoB and MTP immunoreactivity formed large clumps within RGC perikarya and primary dendrites.We performed double-label immunofluorescence experiments using allophycocyanin-and FITC-tagged secondary antibodies to localize apoB and MTP, respectively.This experiment revealed that the same coarsely particulate clumps contained both antigens (Fig. 7F-H).In sections bleached to reveal RPE labeling (Fig. 3), RGC labeling was abolished (data not shown).

DISCUSSION
The recognition that aging-and ARM-associated extracellular deposits contain apoB, apoE, and EC, and the presence of abundant EC-rich particles in normal human BrM, a vessel wall, underscore the potential role for lipoproteins in retinal health and disease, yet the lack of evidence for an association of ARM with increased plasma cholesterol is striking.Here, we demonstrate mRNA and protein for MTP in RPE and, unexpectedly, in RGC, providing a potential explanation for these findings.Immunohistochemical experiments suggest that apoB and MTP are colocalized in the same subcellular compartments within human RPE and RGC, presumably the endoplasmic reticulum.Stacks of rough endoplasmic reticulum typify long axon neurons, including RGC, but RPE has little rough and abundant smooth endoplasmic reticulum.Data are representative of five independent experiments and are expressed as means Ϯ SD.TG, triglyceride.Cells were incubated with 5 Ci/ ml [ 3 H]oleate (5,000 dpm/nmol) bound to 1.5% fatty acid-free BSA.Confirming the exact identity of the subcellular compartment will require further experimentation.
In addition to MTP, we also detected mRNA transcripts for apoB, protein disulfide isomerase, and ACAT-1, but not APOBEC-1 or ACAT-2, in human retina, RPE, and ARPE-19.The complete transcriptome of retina-and RPEexpressed genes is a stated research goal (51), and the genes we report have not yet appeared in high-throughput surveys.Expression of essential RPE-specific genes is considered low relative to regulatory and housekeeping genes (52).However, MTP transcripts were reported for whole mouse eye (GenBank AV360595; H. Konno, K. Aizawa, S. Akahira et al., unpublished observations).We confirmed MTP mRNA and protein in murine retina/RPE (C-M.Li and X. Zhang, unpublished observations), in addition to our current result on porcine retina.MTP is the hallmark of cells that secrete lipoproteins with neutral lipid cores; therefore, we used metabolic labeling to ensure that neutral lipid synthesis and secretion by the ARPE-19 cell line and primary culture of porcine RPE were possible.Regarding ACAT, the ratio of ACAT proteins and mRNA transcripts differs among organs.By either measure, ACAT-1 predominates in liver and ACAT-2 predominates in intestine (53,54).Retinal expression of ACAT-1 is consistent with the finding that ACAT-1 deficiency causes extensive UC deposition in brains of severely hypercholesterolemic mice (55).Expression of only one ACAT in a lipoprotein-secreting cell, if confirmed by other technologies, would be unusual.
We detected apoB message and protein in native RPE.Intraocular mRNA transcripts for apoB were first detected for combined RPE/choroid (19) and then localized, with apoB protein, to native RPE (8).ApoB was detected in drusen by immunofluorescence (8) but not by proteomics (56), likely because the chloroform-methanol extraction used in that investigation removed core lipids and precipitated the amphipathic apoB peptide (57).In our report, ARPE-19 extracts contained a small but readily detectable high molecular mass band at the same position as in HepG2 extracts.We infer from its location, the sequencing of cDNA derived from apoB mRNA, and the lack of evidence for apoBEC-1 mRNA transcripts that this band represents apoB-100.We cannot exclude the possibility, however, that apoB-48, or another novel truncate, exists in the eye without using antibodies specific to apoB-48.However, because of the choice of cells, culture conditions, suboptimal lipid loading (see below), or inherently low secretion by RPE, we did not produce sufficient lipoprotein in culture medium to permit the extensive subfractionation for biochemical analysis of protein and lipid components that is possible for HepG2 (58,59); thus, we were unable to demonstrate synthesis and secretion of apoB itself.
Despite the lack of secreted apoB, we did detect spherical lipid particles in the d р 1.21 g/ml fraction of medium conditioned by oleate-stimulated ARPE-19, supporting but not proving the idea that BrM particles are nascent lipoproteins.Particles were sparse, consistent with low EC concentration, and they were smaller than solid BrM particles in situ (12,13), suggesting lower TG content.Although similar in size to plasma VLDL, ARPE-19-associated particles appeared more variable in diameter, possibly reflecting heterogeneity in FFA uptake or processing efficiency.Furthermore, differences in the amount of radiolabeled oleate incorporated into phospholipid at 24 and 48 h were minimal compared with those incorporated into TG and EC, suggesting that particle volumes after 48 h of incubation may be larger than particle volumes after 24 h, with proportionately smaller surface areas.Although ARPE-19-conditioned medium contained EC enriched in cholesteryl oleate and poor in cholesteryl docosahexanoate, BrM EC composition in vivo may differ, as the RPE regularly processes docosahexanoate-rich OS and BrM is exposed to plasma extravasated from the choroid.Notably, human RPE secretes apoE in vitro (21), 30% of which occurs in a fraction (d ϭ 1.193-1.256g/ml) that overlaps the EC-containing fraction of ARPE-19-conditioned medium.Thus, a potential setting for RPE apoE is the surface of a secreted apoB-containing lipoprotein (60).Finally, it is uncertain whether particles are secreted as such or form in the medium from secreted free apolipoproteins.For example, nascent apoA-I-containing high density lipoprotein particles that are discoidal and EC-poor [EC/TC, 0.16-0.22(58,61)] become spherical after acquiring EC in plasma via LCAT activity.Because native RPE and ARPE-19 express mRNA transcripts for LCAT (C-M.Li, unpublished observations), similar metabolic remodeling may occur in culture.We think this unlikely, because the observed particles were EC-rich (EC/TC, 0.52) and spherical rather than discoidal.Furthermore, hepatic LCAT is believed to be active only in plasma.These uncertainties will be resolved using an improved culture system.
Determining the predominant FFA input to an RPE lipoprotein in vivo is an important goal for future study, with implications for retinal cell biology and ARM pathogenesis.The relative contributions of exogenous and de novo synthesized FFA to hepatic and intestinal lipoproteins are being actively investigated (62) using oleate-stimulated HepG2 cells.FFA delivery to neurosensory retina and RPE is best understood for docosahexanoate (63), preferentially enriched in neural membrane and photoreceptor OS phospholipids.The RPE is capable of both docosahexanoate biosynthesis (64) and LDL receptormediated apoB-containing lipoprotein uptake (65,66).Thus, the relative contributions of exogenous and endogenous FFA to RPE secretory products must also be determined.However, the signature RPE lipid-processing activity is daily, light-synchronized phagocytosis of OS distal tips, so a potential source of exogenous FFA unique to RPE is phospholipids derived from these membrane-rich organelles.Our results with oleate-BSA indicate that phagocytosis is not required for neutral lipid secretion.However, lipoprotein assembly and secretion as a mechanism to release FFA from OS phospholipids as TG is an attractive hypothesis.Limited data on mammalian RPE support this notion, as intracellular TG in rat RPE cells increase for 4 h and decline over 24 h after OS phagocytosis in vivo (67) and in vitro (68).That OS phagocytosis, a uniquely eyespecific activity, should contribute lipids to ARM-associated lesions is a widely held expectation (69), but an output mechanism from RPE has never been specified.With an apoB-containing lipoprotein a plausible pathway for constitutive basolateral neutral lipid secretion, ARM lesion formation and its aftermath of inflammation and neovascularization can be reconceptualized as an atherosclerosis-like response-to-retention progression (26) with a local lipoprotein source.Testing this new hypothesis will be possible with an improved RPE culture system and animal models based on targeted deletion or overexpression of MTP.
Evidence for retinal MTP expression provides a basis to reexamine ABL from a novel perspective.ABL retinopathy involves macula and/or periphery and includes pigmentary changes, reduced electro-oculograms from RPE, reduced electroretinograms from cones and rods, predilection for angioid streaks (fractures in BrM) (70), and RPE cells with trilaminar inclusions in the subretinal space (71).For two decades, ABL retinopathy was attributed to systemic insufficiency of vitamin E transported by dietary apoB-containing lipoprotein (72).However, retinopathy persists or progresses despite vitamin E supplementation, even in patients treated from infancy to early adulthood (35,71), and patients with undetectable plasma vitamin E can have normal retinas (73).Perhaps the clinical response to vitamin supplementation in ABL patients is variable, because MTP mutations directly cause a loss of function in retina.Furthermore, our unexpected finding of MTP and apoB in RGC, a neuron of the central nervous system, is potentially important for ABL neuropathy.Among retinal neurons, RGC is the most commonly affected in hereditary disorders of lipid metabolism (74).It is widely believed that the brain does not express apoB, and apoB-containing lipoproteins are undetectable in human cerebrospinal fluid (75).Unique among apolipoproteins, apoB is essential for brain development, and mice with absent or reduced apoB have neural tube defects (76,77).Seeking MTP gene expression in brain with methods sensitive to low-abundance message is now warranted.
In summary, our results strengthen circumstantial evidence implicating apoB-containing lipoprotein assembly as an important retinal function, a plausible mechanism to place neutral lipid in BrM, and a candidate pathway involved in forming lipid-rich lesions in ARM.These data, along with the recently established cardiac apoB-containing lipoprotein (78), further widen the apoB/MTP system's role in mammalian physiology beyond classic functions of plasma TG transport.

Fig. 3 .
Fig. 3. ApoB and MTP immunoreactivity in bleached RPE of normal human macula.Immunoreactivity is purple.Arrowheads in B, D, and F delimit bleached RPE.Bar ϭ 20 m.A: Goat anti-human apoB labels clumps in RPE.B: No clumps are visible in a section treated with goat IgG.C: Rabbit anti-bovine MTP labels similar clumps in RPE.D: No clumps are visible in a section treated with rabbit IgG.E: The basolateral location of monocarboxylate transporter 3 is unaffected by bleach treatment.Label in outer segments (out of focus) is nonspecific.F: No basolateral labeling is apparent in a bleached section treated with rabbit IgG, but nonspecific outer segment labeling remains.G: Melanin granules in a section that was unbleached and unexposed to antibodies are smaller than apoB-and MTP-positive clumps.Arrows indicate the same cells in G and H. H: Autofluorescent lipofuscin granules in the same section as in G are smaller and more numerous than apoB-and MTP-positive clumps.

Fig. 6 .
Fig.6.Expression of apoB and MTP in porcine retina (R), RPE (P), and liver (L).Total RNA was isolated from porcine retina, RPE, and liver as described in Materials and Methods.Expected lengths of RT-PCR products are listed in Table1.To check genomic DNA contamination, control reactions that lacked reverse transcriptase were performed for MTP.Primers for apoB span a 960 bp intron.M, 100 bp DNA ladder; arrowhead, 500 bp.

TABLE 1 .
The sequences of PCR primers and expected fragment sizes

TABLE 2 .
TC and EC mass in ARPE-19 and HepG2 cells

TABLE 3 .
Mass and composition of cholesteryl esters in ARPE-19 and HepG2 cells Data are representative of three independent experiments performed in triplicate and are expressed as means Ϯ SD.http://www.jlr.org/content/suppl/2005/03/17/M400428-JLR20Supplemental Material can be found at:

TABLE 4 .
Composition of cholesteryl esters in medium of oleate-BSA-supplemented ARPE-19 Data for mass and normalization are representative of three independent experiments performed in triplicate.All data were normalized to cholesteryl oleate.The fold increase in cholesteryl ester over nonconditioned medium represents the mean of three experiments.

TABLE 5 .
New synthesis and secretion of phospholipid, TG, and EC by oleate-BSA-supplemented ARPE-19 cells

TABLE 6 .
New synthesis and secretion of phospholipid, TG, and EC by oleate-BSA-supplemented porcine RPE cells Data are representative of three independent experiments and are expressed as means Ϯ SD.Cells were incubated with 5 Ci/ml [ 3 H]oleate (5,000 dpm/nmol) bound to 1.5% fatty acid-free BSA.