Conversion of 7-ketocholesterol to oxysterol metabolites by recombinant CYP27A1 and retinal pigment epithelial cells.

Of the different oxygenated cholesterol metabolites, 7-ketocholesterol (7KCh) is considered a noxious oxy-sterol implicated in the development of certain pathologies, including those found in the eye. Here we elucidated whether sterol 27-hydroxylase cytochrome P450 27A1 (CYP27A1) is involved in elimination of 7KCh from the posterior part of the eye: the neural retina and underlying retinal pigment epithelium (RPE). We first established that the affinities of purified recombinant CYP27A1 for 7KCh and its endogenous substrate cholesterol are similar, yet 7KCh is metabolized at a 4-fold higher rate than cholesterol in the reconstituted system in vitro. Lipid extracts from bovine neural retina and RPE were then analyzed by isotope dilution GC-MS for the presence of the 7KCh-derived oxysterols. Two metabolites, 3β,27-dihydroxy-5-cholesten-7-one (7KCh-27OH) and 3β-hydroxy-5-cholesten-7-one-26-oic acid (7KCh-27COOH), were detected in the RPE but not in the neural retina. 7KCh-27OH was also formed when RPE homogenates were supplemented with NADPH and the mitochondrial redox system. Quantifications in human RPE showed that CYP27A1 is indeed expressed in the RPE at 2-4-fold higher levels than in the neural retina. The data obtained represent evidence for the role of CYP27A1 in retinal metabolism of 7KCh and suggest that, in addition to cholesterol removal, the functions of this enzyme could also include elimination of toxic endogenous compounds.

by CYP27A1 in the reconstituted system in vitro and present evidence that this P450 also oxidized 7KCh in vivo in bovine RPE. Our results suggest that cytochrome P450mediated metabolism could be at least one of the mechanisms whereby RPE removes toxic 7KCh, thus expanding our understanding of the retinal signifi cance of CYP27A1.

Materials
Cholesterol and 7KCh were purchased from Steraloids, Inc. (Newport, RI); [25,26,26,26,27,27,27-2 H 7 ]cholesterol (cholesterol-D 7 ) and [26,26,26,27, H 5 ]27-hydroxycholesterol (27OH-D 5 ) were from Medical Isotopes (Pelham, NH); [25,26,26,26,27,27 . Human recombinant CYP27A1, mitochondrial redox partners adrenodoxin (Adx) and adrenodoxin reductase (Adr), and microsomal P450 oxidoreductase were expressed and purifi ed as described (18)(19)(20)(21). Bovine eyes were obtained from a local slaughterhouse 3-5 h after animals were sacrifi ced. The anterior segment was removed, and the vitreous humor was poured out to expose the retina. The neural retina was carefully separated from the RPE, placed in a tube, and washed one time (by centrifugation at 2,000 g ) with PBS. The remaining eye cup was incubated on ice with 5 ml of PBS containing 1 mM EDTA to chelate Ca 2+ and facilitate dissociation of the RPE from the basement membrane ( 22 ). In 30 min, the RPE was gently brushed off, aspirated, and subjected to repeated washes by differential centrifugation ( 23 ) until the supernatant was clear and colorless and the pelleted RPE cells were essentially free of the contaminating nonmelanotic cells, erythrocytes, and rod outer segments as assessed by brightfi eld microscopy. Fluorescent microscopy (excitation at 488 and emission at 515 nm) did not detect autofl uorescent rod outer segments, erythrocytes, or macrophages, thus confi rming the analysis done by light microscopy. The isolated RPE was either used immediately or fl ash-frozen in liquid nitrogen and stored at Ϫ 80°C. Cadaveric human eyes were processed previously ( 24 ). The RPE was gently scraped under the dissecting microscope with a crescent knife, aspirated with a glass microcapillary tube, and frozen. Assessment of the purity of these preparations by brightfi eld and fl uorescent microscopies did not reveal significant contaminations: only few nonmelanotic nonautofl uorescent cells of undetermined origin were present. Human tissue use conformed to the Declaration of Helsinki and institutional review at Case Western Reserve University. Eye specimens were obtained from de-identifi ed donors following informed consent of their families.

Isolation of mitochondrial phospholipids (PL) from bovine neural retina and RPE and preparation of PL vesicles for the enzyme assay
Each tissue ( ‫ف‬ 5 g) was homogenized with Tefl on grinder in 10 ml of 50 mM Tris-HCl (pH 7.4), containing 250 mM sucrose, 5 mM MgCl 2 , 1 mM phenylmethylsulfonyl fl uoride, 1 mM dithiotreitol, 100 g/ml of butylated hydroxytoluene (BHT), and a cocktail of protease inhibitors (Roche Complete EDTA-free, 1 tablet/50 ml buffer). Tissue homogenates were subjected to centrifugation at 1,000 g for 20 min to remove unbroken cell, nuclei, and cell debris followed by centrifugation of the resultant supernatant at 9,000 g for 20 min to pellet the mitochondria. Mitochondria were resuspended in 3 ml of the homogenization buffer rats, 7KCh was observed in both the RPE-choriocapillaris region and neural retina, and in the latter, 7KCh was immunolocalized to the regions with high mitochondrial content, the ganglion cell layer and photoreceptor inner segments ( 9 ). 7KCh is studied in the retina because of the possible connection to age-related macular degeneration ( 4 ), a major cause of blindness in the elderly in industrialized countries ( 10 ). Loss of vision in age-related macular degeneration is caused by degeneration of the neural retina and RPE in the macula ( 11 ), a region near the center of the retina.
7KCh is putatively a toxic compound, yet the mechanisms whereby it is formed and eliminated in vivo are not fully understood. Evidence was obtained indicating that iron is likely involved in generation of 7KCh in the neural retina ( 9 ). The source of iron was not determined, but ferritin and/or cytochrome c are suspected ( 4 ). The routes of 7KCh elimination from the retina are also under investigation. Four pathways are considered: i ) enzymatic oxidation by cholesterol 25-hydroxylase and cytochrome P450 enzymes 27A1 and 46A1 (CYP27A1 and CYP46A1); ii ) reduction by 11 ␤ -hydroxysteroid dehydrogenase type 1; iii ) effl ux involving transporters such as ABCA1 and ABCG5; and iv ) sulfonation by sulfotransferases (reviewed in Ref. 4 ). Of these four possible pathways, we favor enzymatic oxidation of 7KCh by CYP27A1, a ubiquitous sterol 27hydroxylase with relatively broad substrate specifi city and multiple physiological roles. Indeed, CYP27A1 is involved in cholesterol elimination from extrahepatic tissues, metabolizes bile acid intermediates in the liver, and activates vitamin D 3 in the kidney (12)(13)(14). CYP27A1 is also a mitochondrial protein whose immunolocalization in the retina overlaps with that of 7KCh: photoreceptor inner segments, ganglion cell layer, and RPE ( 15 ). Furthermore, 7KCh was found to be 27-hydroxylated in human macrophages and HepG2 cells ( 16,17 ); it was suggested to be the substrate for CYP27A1 based on the site of hydroxylation, reduction of its metabolism by the cells treated with several CYP27A1 inhibitors, and absence of 27-hydroxylated 7KCh metabolite in the medium when CYP27A1-defi cient human macrophages were incubated with exogenous radioactive 7KCh ( 16 ). These studies, however, did not assess whether enzymes other than CYP27A1 were also inhibited and whether CYP27A1 defi ciency affected the uptake of exogenous radioactive 7KCh by the cultured cells.
CYP27A1 is a strong candidate for being the major enzyme eliminating 7KCh in the retina. Yet, in previous studies by others, hydroxylated products of 7KCh were not detected either in monkey retina or photodamaged rat retina when retinal lipid extracts were analyzed by high performance liquid chromatography (HPLC) coupled to mass spectrometry (MS) based on atmospheric pressure chemical ionization ( 8,9 ). Therefore, in the present work, we investigated retinal metabolism of 7KCh by a highly sensitive analytical tool, isotope dilution gas chromatography-mass spectrometry (GC-MS) with selected ion monitoring (SIM) following in vitro studies evaluating 7KCh as a substrate for purifi ed recombinant CYP27A1. We show here that 7KCh was effi ciently hydroxylated several times addition of the NADPH-regenerating system to initiate the enzyme reaction, which was carried out for 2 min.

Spectral binding assay with purifi ed recombinant human CYP27A1
Titrations with cholesterol and 7KCh were carried out at 30°C as described ( 28 ). Sterols were added from 1 mM stock in 4.5% aqueous 2-hydroxypropyl-␤ -cyclodextrin to a 1-ml solution of 50 mM KP i , pH 7.2, containing 0.4 M CYP27A1, 1 mM EDTA, 0.5 M NaCl, and 10% glycerol. The K d values were calculated by nonlinear least-squares fi tting using the quadratic form of the singlesite binding equation ( 29 ): where ⌬ A is the spectral response at different substrate concentrations [S], ⌬ A max is the maximal amplitude of the spectral response, and [E] is the enzyme concentration.

Quantifi cation of sterols in bovine tissues
Fresh neural retina or RPE was weighed and homogenized in 10 vol (w/v) of 40 mM KP i , pH 7.2, containing 1 mM EDTA and 100 µg/ml BHT. Tissue homogenization was followed by centrifugation at 1,500 g for 15 min to remove cell debris. Supernatant aliquots ( ‫ف‬ 15 mg of total protein per tube) were used either for lipid extraction or enzyme assays. For the former, two tubes were used, each supplemented with the same amounts of cholesterol-D 7 (500 nmol) and 27OH-D 5 (1 nmol) but different amounts of 7KCh-D 7 (0.2 or 0.02 nmol) for quantifi cation of 7KCh and 7KCh-27OH, respectively. Lipids were extracted from tissue homogenates fi ve times by vortexing with 2.5 vol of chloroform/ methanol (2:1, V/V) for 1 min. The chloroform phases were combined and evaporated to dryness at room temperature under vacuum using a Savant SC2104 SpeedVac Concentrator (Thermo Scientifi c, Ashville, NC). Oxysterols were separated from cholesterol by solid phase extraction as described ( 30 ). Briefl y, the dried lipid extract was dissolved in 10 ml of 70% ethanol and applied to a Varian C18 SPE column (1000 mg, Varian Inc., Lake Forest, CA) equilibrated with 30 ml of 70% ethanol. The fl owthrough fraction was collected and combined with the eluate from the subsequent column wash with 40 ml of 70% ethanol. The combined eluates represented the oxysterol fraction. The cholesterol fraction was eluted next by washing the column with 20 ml of ethanol. The oxysterol and cholesterol fractions were evaporated to dryness in a SpeedVac Concentrator and methylated with 3 ml of freshly prepared ethereal diazomethane at room temperature for 30 min. After methylation, the fractions were dried under nitrogen, derivatized with 60 l of bis-(trimethylsilyl)trifl uoroacetamide/trimethylchlorosilane (TMS) at 60°C for 1 h, and subjected to GC-MS.
Saponifi cation was performed on a dried lipid extract as described ( 31 ). Briefl y, 2 ml of 20% KOH in methanol was added and vortexed to dissolve lipids. Then 2 ml of diethyl ether was added, followed by vortexing and fl ushing the solution with nitrogen to remove air. The solution was vigorously shaken on ice for 3 h in a capped glass tube, then the reaction was stopped by 20% acetic acid (2 ml). Hexane (2.5 ml) was added, and the upper phase was isolated and dried. The dried extract was dissolved in 10 ml of 70% ethanol and used for separation on a Varian C18 SPE column as described above.

GC-MS analysis
The derivatized samples were analyzed using an Agilent 5973N-MSD mass spectrometer equipped with an Agilent 6890 gas chromatograph. A DB-35ms (Agilent Technologies, Santa Clara, CA) ( ‫ف‬ 5 mg protein/ml) with the Tris-HCl concentration reduced to 10 mM, and then mixed with 6 ml of the Folch reagent (chloroform-methanol, 2:1, V/V) ( 25 ) by vigorous vortexing for 30 s. Phase separation was facilitated by centrifugation at 2,000 g for 10 min; the lower organic phase was removed, placed in a glass tube, and dried under the nitrogen gas. Subsequent separation of nonphosphorous lipids and PLs was as described ( 26 ). Briefl y, lipid extract was dissolved in 100 l of chloroform and applied on a dry 690 mg Sep-Pak Silica Cartridge (Waters Corporation, Milford, MA). The cartridge was fi rst washed with 20 ml of chloroform to elute nonphosphorous lipids and then with 30 ml of methanol to elute PLs. Both nonphosphorous lipids and PLs were dried under the nitrogen gas and redissolved in 2 ml of chloroform. Absence of cholesterol and other nonphosphorous lipids in the PL fraction was confi rmed by thin layer chromatography (TLC) on plates with silica gel using hexane/diethyl ether/ methanol/acetic acid (90/20/5/2, V/V/V/V) as resolving mobile phase; bands were detected by iodine. The PL concentration was measured using ammonium ferrothiocyanate as described ( 27 ). The calibration curve was generated using known amounts of 1-palmitoyl-2-oleoyl-sn-glycero-3-[phosphor-rac-1(1-glycerol)]. Stock solution of PL vesicles for reconstitution with CYP27A1 was prepared by evaporating 1 mg of PL solution in chloroform in a glass tube, then adding 1 ml of 40 mM potassium phosphate buffer (KPi), pH7.2, containing 1 mM EDTA, vortexing, and placing the glass tube in an ice-water bath that was sonicated for 15 min with a Digital Sonifi er (Branson, Danbury, CT) at a 50% amplitude.

Enzyme assays with purifi ed recombinant human CYP27A1
In studies assessing the ability to hydroxylate 7KCh, 0.02 M or 0.2 M CYP27A1 was reconstituted with 2 M 7KCh, 100,000 cpm of [ 3 H]7KCh, Adx and Adr used at 30-and 4-fold molar excesses over CYP27A1, respectively, and NADPH-regenerating system (1 mM NADPH, 1 U of isocitrate dehydrogenase, and 5 mM trisodium isocitrate). Enzymatic reactions proceeded at 37°C in 1 ml of 40 mM KP i buffer, pH 7.4, containing 1 mM EDTA, and were terminated by addition of 5 ml of methanol: chloroform (2:1, V/V). The two phases were separated by a 10 min-centrifugation at 2,000 g ; the lower phase was isolated and dried under nitrogen ( 25 ). The dried extract was then dissolved in 100 l of acetonitrile:methanol:H 2 O (50:20:30, V/V) and subjected to HPLC using a Shimadzu HPLC system (Tokyo, Japan) with a Nova-Pak C 18 column (3.9 × 150 mm, Waters, Millford, MA) connected to a ␤ -RAM radioactivity detector (INUS Systems Inc., Tampa, FL). The fl ow rate (1 ml/ min) was kept at 100% solvent A (acetonitrile:methanol:water, 50:20:30, V/V) for 5 min, then a linear gradient between solvent A and solvent B (100% methanol) over 15 min was used, after which the fl ow was kept at 100% solvent B for another 5 min. Peak areas corresponding to 7KCh and its metabolites were integrated and used to calculate the substrate metabolism.
To identify metabolites of 7KCh, a separate enzyme assay was carried out, in which only unlabeled 7KCh was used. The concentrations of CYP27A1 and 7KCh were 0.2 M and 2 M, respectively, and those of Adx and Adr were 6 M and 0.8 M, respectively; the reaction time was 5 min. The products were isolated by HPLC, followed by analysis by GC-MS.
To determine the rates of cholesterol and 7KCh hydroxylation, a third set of experiments was conducted in which 0.05 M CYP27A1 was added to 1 ml of 40 mM KP i , pH 7.4, containing 1 mM EDTA, 20 g of PLs, 30 M cold sterol substrate, ‫ف‬ 1 nM [ This result is consistent with previous studies by us and others showing multiple oxidations of cholesterol and 5 ␤cholestane-3 ␣ ,7 ␣ ,12 ␣ -triol by CYP27A1 to yield C27 alcohol, C27 aldehyde, and C27 acid and dependence of the product profi le on the enzyme concentration and reaction time ( 32,33 ).
To confi rm that 7KCh in the enzyme assay was oxidized by CYP27A1 at C27, we used GC-MS analyses. An equimolar mixture of authentic 7KCh, 7KCh-27OH, 7KCh-27COOH as well as cholesterol, 27OH, and 5-cholestenoic acid (27-COOH) was prepared. The sterols were subjected to methylation and trimethylsilylation, and column and chromatography conditions were identifi ed that enabled good separation of all six sterols and high sensitivity of detection (supplemental Fig. IA). The limit of detection of 7KCh-27OH in this study was ‫ف‬ 1 pmol/injection in SIM mode (supplemental Fig. IA, inset), which is higher than 15 pmol of 7KCh-27OH/injection, reported previously ( 16 ).
Following optimization of the conditions of GC-MS, CYP27A1 was again incubated with 7KCh under conditions that produced comparable amounts of the two metabolites (0.2 µM CYP27A1, 5 min reaction time, Fig. 1D ). The metabolites were isolated by HPLC and analyzed individually by GC-MS after methylation and trimethyl s ilylation. The HPLC peak putatively corresponding to 7KCh-27OH (retention time 13 min, Fig. 1A, B ) eluted as one peak during GC (  ( Fig. 2D ). We did not identify the smaller peaks in Fig. 2B because they were nonsterol compounds. Thus, similar to endogenous substrates cholesterol and 5 ␤cholestane-3 ␣ ,7 ␣ ,12 ␣ -triol, 7KCh could be oxidized CYP27A1 multiple times, yielding hydroxylated and carboxylated products. Identifi cation of 7KCh-27COOH and its MS characterization enabled future search of this metabolite in biological samples.

Comparison of 7KCh and cholesterol as substrates for CYP27A1
Next we determined affi nities of purifi ed recombinant CYP27A1 for 7KCh and cholesterol ( Table 1 ). The K d values were found to be similar (0.07 M for 7KCh and 0.14 M for cholesterol), indicating that in the inner mitochondrial membrane where CYP27A1 resides and cholesterol content is low ( 34 ), 7KCh could compete with cholesterol for the CYP27A1 active site. Then we determined the capillary column (30 m × 0.25 mm × 0.25 mm) was used for all analyses. The mass spectrometer was operated in the electron impact (EI) ionization mode. Gas chromatographic conditions were as follows: 2 µl sample was injected in splitless mode (inlet was kept at 270°C with the helium fl ow at 1.0 ml/min) at the initial 150°C. The oven was fi rst kept at 150°C for 2 min, ramped at 10°C/min to 310°C, and held for 20 min isothermally. The ion source fi lament was operated at 70 eV. The mass detector was set at 310°C. We utilized both scan (total ion monitoring, m/z range of 100-700 Da) and SIM modes and the optimal dwell time (10 msec per ion). The following ions ( m/z ) were monitored in the SIM mode: 368 (cholesterol); 375 (cholesterol-D 7 ); 417 (27OH), 422 (27OH-D 5 ), and 472 (7KCh); 479 (7KCh-D 7 ); 455, 470, 545, and 560 (7KCh-27OH); and 411, 426, 501, and 516 (7KCh-27COOH). For quantifi cation, calibration curves were generated using a fi xed concentration of the internal standard, deuterated sterol analog, and varying concentrations of the corresponding unlabeled sterol.

Incubations with RPE homogenates
RPE homogenates ( ‫ف‬ 15 mg of total protein) were incubated either with 10 mM NADPH or with 10 mM NADPH plus 5 M Adr and 15 M Adx or with 10 mM NADPH plus 5 M P450 oxidoreductase. Incubations were carried out for 60 min at 37°C with constant shaking. Extraction of lipids and GC-MS analysis were as described above.

Enzyme assays with purifi ed recombinant human CYP27A1 and identifi cation of 7KCh metabolites
As it had not previously been shown that purifi ed CYP27A1 metabolizes 7KCh, two enzyme concentrations (0.02 µM and 0.2 µM) and different reaction times (up to 2 h) were tested in the in vitro reconstituted system, and then the product formation was assessed by HPLC. At 0.02 µM CYP27A1 and 5 min reaction time, only one product peak, putatively 7KCh-27OH, was seen during HPLC separation ( Fig. 1A ). Yet two product peaks were observed when the enzyme reaction proceeded for 2 h ( Fig. 1B ). The retention time of the second product peak was earlier than that of the putative 7KCh-27OH, indicating that it is the more polar metabolite. Consequently, by analogy with our previous studies investigating cholesterol as a substrate for CYP27A1 ( 32 ), we hypothesized that the second product was 7KCh-27COOH. The kinetics of the formation of the putative 7KCh-27OH and 7KCh-27COOH in the enzyme assay were different ( Fig. 1C, D ). At low CYP27A1 concentration (0.02 µM), the 7KCh-27OH content increased up to 30 min of reaction time but then started to decline. In contrast, the amount of 7KCh-27COOH was constantly increasing and reached the levels of 7KCh-27OH after 120 min of enzymatic reactions ( Fig. 1C ). At 10-fold higher CYP27A1 concentration (0.2 µM), 7KCh-27COOH became a predominant metabolite after only 15 min of the enzymatic reaction. After 60 min, almost all 7KCh was converted to 7KCh-27COOH ( Fig. 1D ). Thus, CYP27A1 can metabolize 7KCh in vitro and yield different products depending on the conditions of the enzyme assay: higher enzyme concentration and longer reaction time lead to the formation of the more polar metabolite.
were optimized for the formation of only one metabolite, 27-alcohol, with the product formation linear with time and enzyme concentration. Under these steady-state conditions, the rate of 7KCh hydroxylation was about 4-fold higher than that of cholesterol in both retinal and RPE rate of 7KCh and cholesterol hydroxylation in the in vitro enzyme assay. Mitochondrial PLs were isolated from bovine neural retina and RPE and used for reconstitution with CYP27A1 in the presence of saturating substrate and redox partner concentrations. The conditions of the assay  lack of a good internal standard. Saponifi cation of the RPE lipid extract led to signifi cant reduction of the 7KCh-27OH peak and the disappearance of the 7KCh-27COOH peak, indicating decomposition of these sterols even though they were hydrolyzed on ice. Hence, all subsequent studies have been carried out only on nonsaponifi ed extracts of RPE.
Three preparations of the RPE were additionally isolated, and all three showed similar but ‫ف‬ 5-fold lower content of 7KCh ( ‫ف‬ 8 pmol/mg protein) compared with the fi rst preparation ( Fig. 4 ). Also, no ion peaks corresponding to 7KCh-27OH and 7KCh-27COOH were observed in these preparations. Therefore, we used these samples and reconstituted them with NADPH and either mitochondrial Adr and Adx, which transfer electrons from NADPH to all mitochondrial P450s including CYP27A1, or the microsomal P450 oxidoreductase, which reduces all microsomal P450s. These experiments were carried out because tissue disruption and homogenization usually leads to quick oxidation of endogenous NADPH and, consequently, terminate the P450-mediated enzymatic reactions. These reactions, however, could be reinitiated if the homogenate is supplemented with NADPH and the P450 redox partners that may be at limiting amounts in a biological sample.

Incubations with RPE homogenates
The ion peaks characteristic for 7KCh-27OH were only observed in the ion chromatograms of the extracts from the incubations with NADPH and a mitochondrial redox system ( Fig. 4 ). The 7KCh-27OH concentration ( ‫ف‬ 1 pmol/mg protein) was similar in all three preparations. No ion peaks characteristic for 7KCh-27OH were detected in the extracts from control incubations lacking NAPDH and a redox system or from incubations with NADPH and a microsomal redox partner P450 oxidoreductase ( Fig. 4 ). Thus, the RPE can metabolize 7KCh to 7KCh-27OH, and this enzymatic reaction is specifi c to a mitochondrial P450. To confi rm that CYP27A1 is active under the experimental conditions used, we measured its cholesterol hydroxylase activity and the levels of 27OH formed from cholesterol. The concentration of 27OH increased ‫ف‬ 260-fold after initiation of the enzymatic reaction with NADPH and Adr and Adx (2 versus 521 pmol/mg protein), and to a much lesser extent, ‫ف‬ 5-fold and 3-fold, respectively, in the presence of NADPH and NADPH plus the microsomal redox partner P450 oxidoreductase ( Fig. 4 ). Thus, even when the concentration of cholesterol in the RPE homogenate was four orders of magnitude higher than that of 7KCh, PLs ( Table 1 ). Along with the binding data, this result suggested that 7KCh has a potential to be a substrate for CYP27A1 in vivo, and it justifi ed our subsequent search of 7KCh metabolites in bovine neural retina and RPE.

Analysis of retinal and RPE extracts for the presence of 7KCh and its metabolites
Several precautions were taken to avoid artifi cial autooxidation of cholesterol to 7KCh during the workup of biological samples ( 35 ). Only tissues immediately isolated from fresh bovine eyes (3-5 h postmortem according to the slaughterhouse) were used, and BHT and EDTA were always included in the buffer for tissue homogenization. Lipid extracts were always evaporated at room temperature under vacuum, followed by the rapid separation of the oxysterol fraction from cholesterol by using solidphase extraction. To ensure high accuracy of the measurements, the best possible internal standard, 7KCh-D 7 , was used because its extraction and derivatization effi ciency is very similar to that of endogenous 7KCh. Also, to determine interpreparation variability, pooled samples of the neural retina and RPE were prepared four times, each time from a group of animals with different demographics (age, breed, and ratio of bulls to cows), fed either grass or grain, and sacrifi ced at two different slaughterhouses.
The fi rst preparation of the neural retina and RPE was analyzed for free (unconjugated) and total (conjugated) forms of sterols. The detector was programmed in a SIM mode for the ions characteristic for 7KCh, 7KCh-D 7 , 7KCh-27OH, and 7KCh-27COOH. In the neural retina, only molecular ions for 7KCh-D 7 ( m/z 479) and 7KCh ( m/z 472) were observed ( Fig. 3A ), which eluted as two major peaks at the retention time of the authentic 7KCh-D 7 and 7KCh, respectively. In this preparation, the concentration of free 7KCh in the neural retina was 25 pmol/mg protein, and it remained the same after cold saponifi cation, indicating lack of esterifi cation. Ion peaks for free 7KCh were also present in the RPE ( Fig. 3B ) but at a concentration higher than that in the neural retina (41 pmol/mg), with saponifi cation leading to further increase in the level of the sterol to 51 pmol/mg. Unlike the neural retina, the nonsaponifi ed extract of RPE also contained four ion peaks characteristic for 7KCh-27OH ( m/z 455, 470, 545, and 560) and 7KCh-27COOH ( m/z 411, 426, 501, and 516), which coeluted for each metabolite at the retention times of the authentic standards ( Fig. 3B ). The content of 7KCh-27OH was estimated based on the calibration curve using 7KCh-D 7 as internal standard and found to be 1 pmol/mg protein. We did not quantify 7KCh-27COOH because of the drial membrane was signifi cantly different from that in the homogenate. Incubations with the homogenates provide additional evidence for the involvement of CYP27A1 in the metabolism of 7KCh in the RPE.
CYP27A1 was able to metabolize 7KCh, yielding 7KCh-27OH. The latter is likely because mitochondria were not signifi cantly disrupted during homogenization and the ratio between cholesterol and 7KCh in the inner mitochon-  lizes 7KCh and oxidizes this sterol ‫ف‬ 4-fold more effi ciently than cholesterol, despite a similar affi nity for 7KCh and cholesterol. These data served as the biochemical basis for the search of the C27-hydroxylated metabolites of 7KCh in the retina.
From previous studies by others ( 8,9 ) we knew that 7KCh is present in the retina but its hydroxylated derivatives were not detected by HPLC-MS with atmospheric pressure chemical ionization mode. We also knew that 7KCh is one of the most diffi cult oxysterols to measure because even minor auto-oxidation of highly abundant cholesterol during sample isolation and workup will lead to falsely high levels of 7KCh and overestimation of its content by as much as a 1,000-fold ( 35 ). We were also aware that to accurately quantify 7KCh, as well as any other sterol, its deuterated analog should be added to a biological sample prior to the beginning of any manipulations. Therefore, to assay bovine neural retina and RPE for 7KCh and its C27-oxygenated metabolites, we employed a highly sensitive and accurate GC-MS methodology with SIM detection, used deuterated 7KCh as internal standard, and maintained anaerobic conditions during sample processing.
The measured levels of free 7KCh were in a low picomolar range; they were 1.6-fold higher in the RPE (41 pmol/ mg protein) than in the retina (25 pmol/mg of protein) as assessed by the measurements in one preparation. Similar to cholesterol and oxysterols formed enzymatically ( 36 ), saponifi cation did not signifi cantly increase the 7KCh content either in the neural retina or RPE, indicating that sulfonation or other types of esterifi cation are unlikely to be the major mechanisms by which retinal cells eliminate or reduce the toxicity of this metabolite. These data are consistent with undetectable retinal levels of mRNA for sulfotransferase 2B1b ( 4 ), one of the enzymes that could sulfonate 7KCh ( 37 ).
A notable difference between oxysterols generated enzymatically ( 36 ) and the nonenzymatically formed 7KCh was a signifi cant interpreparation variability in the levels of the latter. Of the four preparations tested, a 40 picomolar/mg protein concentration of 7KCh was detected only in one preparation of the RPE, whereas the other three preparations contained similar and much lower levels of 7KCh ( ‫ف‬ 8 pmol/mg protein). A 5-fold variation in the content of 7KCh in the RPE could be due to differences in animal housing, feeding, demographics, and duration of

Concentrations of CYP27A1 in the RPE
Quantifi cations were carried out as described ( 24 ) using multiple reaction monitoring and individual samples from four human donors. These samples were isolated during the course of our previous studies in which we measured CYP27A1 expression in the neural retina; in the present work, we determined the CYP27A1 levels in the RPE. Depending on the donor, the enzyme concentration per milligram of total protein was 2-to 4-fold higher in the RPE than in the neural retina ( Table 2 ), thus confi rming that CYP27A1 is indeed expressed in the RPE and providing an explanation as to why we did not detect metabolites of 7KCh in the neural retina. For each GC-MS injection, the same amount of total protein was used to extract lipids either from the neural retina or RPE. In the RPE, the levels of 7KCh metabolites were close to the limits of detection. Hence, in the neural retina, where both 7KCh and CYP27A1 are present at concentrations lower than in the RPE (per mg of protein), less 7KCh metabolites are likely generated, and these amounts may be below the limits of detection of the utilized protocol. Thus, our study does not exclude metabolism of 7KCh by CYP27A1 in the neural retina. Instead, it suggests that a higher sensitivity is required to investigate metabolism of 7KCh in this anatomical element. Note that normalizations based on total protein do not refl ect local concentrations of CYP27A1 in the neural retina and RPE. The neural retina is a multilayered and multicellular organ with uneven expression of CYP27A1 as assessed by immunohistochemistry ( 15 ), whereas the RPE represents one layer of the same cell type. Therefore, concentrations of CYP27A1 in some retinal cells could be comparable to those in the RPE.

DISCUSSION
This study was undertaken to further elucidate cholesterol metabolism in the retina and the role of CYP27A1 in this process. Previously we measured retinal and the RPE content of cholesterol and sterol metabolites generated enzymatically ( 36 ). In the present work, we focused on 7KCh, the nonenzymatic product of cholesterol oxidation, and investigated its possible elimination by CYP27A1. By using in vitro assays, we demonstrated for the fi rst time that purifi ed recombinant CYP27A1 binds and metabo- compared with the 27COOH/cholesterol ratio may refl ect either more effi cient metabolism of 7KCh relative to cholesterol or less effi cient elimination of 7KCh-27OH relative to 27COOH. The former would be in agreement with the results of our in vitro studies evaluating cholesterol and 7KCh as the substrates for CYP27A1, as well with prior investigations by others showing cytotoxic, pro-infl ammatory, and apoptotic effects of 7KCh on the RPE-derived cells ( 4,15,44 ). 27-Hydroxylation seems to reduce toxicity of 7KCh as demonstrated by survival of the RPE cells treated with 7KCh-27OH at concentrations and time in which 7KCh essentially killed the entire culture ( 15 ). Thus, high 7KCh-27OH/7KCh ratio in a cell would be consistent with attenuated toxicity of 7KCh-27OH relative to 7KCh. Indeed, if a metabolite is as toxic as the parent compound, it will not be accumulated at comparable levels in a cell. The ability of the RPE to metabolize 7KCh was also confi rmed in the incubations with the homogenates from the preparations where the 7KCh content was low ( ‫ف‬ 8 pmol/ mg protein) and its metabolites could not be detected by GC-MS. 7KCh-27OH was formed only when the homogenates were supplemented with the mitochondrial P450 redox system. In these experiments, ‫ف‬ 1/5 of 7KCh was converted to 7KCh-27OH, but only ‫ف‬ 1/200 of cholesterol was converted to 27OH. More effi cient metabolism of 7KCh relative to cholesterol suggested that CYP27A1 should be seriously considered as a contributor to 7KCh elimination from the RPE. Yet studies of other possible elimination pathways are required to assess quantitative signifi cance of 27-hydroxylation in the overall metabolism of 7KCh in the RPE. Also, it is important to identify the carriers of 7KCh-derived oxysterols inside and outside the cell. Oxysterol-binding protein was found to interact with 7KCh and 27OH ( 45,46 ), and oxysterol-binding proteinrelated protein 2 was found to bind 7KCh ( 47 ). 27OH and 27COOH were shown to spontaneously diffuse across plasma membranes and become associated outside the cell either with HDL (27OH) or albumin (27COOH) ( 48,49 ). Conceivably, upon exit from the mitochondrion, 7KCh-27OH could interact with one of the oxysterol transport proteins and be delivered to the plasma membrane, which it quickly traverses, followed by complex formation with either HDL or albumin. This mechanism would be concordant with the fi nding that the neural retina and RPE express many proteins necessary for HDL lipid transport, for example, apolipoprotein A1 ( 50 ), the major component of HDL. Immunohistochemistry localization on monkey retina found apolipoprotein A1 in Bruch's membrane, which represents the interface between the RPE and choroid, as well as on the apical side of the RPE (this side faces the interphotoreceptor matrix and neural retina) and in the interphotoreceptor matrix surrounding the photoreceptor outer segments ( 50 ). Thus, 7KCh metabolites can be accepted by HDL on either side of the RPE. Consequently, they can diffuse from the RPE through either apical or basal sides. Yet passive diffusion usually requires a concentration gradient between the donor cell membrane and the extracellular acceptor. Because of the intensity of the choroidal blood fl ow, the HDL-7KCh me-exposure to sunlight. Studies of the photodamaged rat retinas demonstrated that the levels of 7KCh are increased 6-fold when animals are exposed to intense, constant light ( 9 ). The fi rst preparation of the neural retina and RPE was the only one obtained in late spring, whereas the other three preparations were isolated throughout autumn. Despite interpreparation variability, a range of the 7KCh concentrations in our study was still signifi cantly lower than that reported in monkey and rat tissues: 0.21 pmol/nmol of free cholesterol in bovine neural retina versus 1-1.6 pmol/nmol of free cholesterol in untreated monkey and rat retinas; and 0.07-0.36 pmol/nmol of free cholesterol in bovine RPE versus 5-8 pmol/nmol of free cholesterol in monkey RPE-choriocapillaris ( 8,9 ). These differences could represent interspecies variations and be pertinent to the way the quantifi cations were carried out. In our study, sterol concentrations were determined by plotting the ratio of the ion peak areas (endogenous sterol/deuterated internal standard) against the calibration generated separately for each sterol of interest. In sterol quantifi cations in monkeys and rats, the concentrations were derived from the peak areas specifi c to cholesterol and 7KCh after normalization to the peak area of the internal standard ␤ -sitosterol ( 8,9 ). In any case, the range of the 7KCh concentrations in bovine neural retina and RPE (0.07-0.36 pmol/nmol cholesterol) was more comparable to the 7KCh levels in hamster brain ( ‫ف‬ 0.05 pmol/nmol cholesterol) ( 38 ) with higher concentrations in the neural retina possibly refl ecting its unique oxidative environment (reviewed in Ref. 4 ).
Increased levels of 7KCh in the fi rst preparation of bovine RPE enabled the detection of its metabolites, 7KCh-27OH and 7KCh-27COOH, whose identifi cation was based on the coelution of four fragment ions specifi c for 7KCh-27OH and 7KCh-27COOH at the retention times of the authentic standards. The sites of oxygenation in 7KCh as well as expression in the RPE strongly suggest that 7KCh-27OH and 7KCh-27COOH are the enzymatic products of CYP27A1, consistent with our in vitro studies using purifi ed recombinant enzyme. Furthermore, it is conceivable that 7KCh could be formed in the mitochondrion because the mitochondrial respiratory chain is known to produce reactive oxygen species, which in turn can oxidize cholesterol to yield 7KCh and other products ( 39,40 ). Recent studies also suggest that in the retina cholesterol oxidation could involve ferritin, which colocalizes with 7KCh ( 9 ). Ferritin was shown to release bound iron upon light exposure ( 41 ) and induce lipid peroxidation in photoreceptor outer segments ( 42 ). This mechanism is supported by identifi cation of a new form of ferritin, mitochondrial ferritin, which localizes in the mitochondrial matrix ( 43 ).
The steady-state levels of 7KCh-27OH in bovine RPE were ‫ف‬ 1/40 of those of 7KCh, and the metabolite to substrate ratio would presumably have been even higher if we could have quantifi ed 7KCh-27COOH. But even when assessed based on 7KCh-27OH, this ratio is more than two orders of magnitude higher than the ratio of the other CYP27A1 metabolite and substrate pair, 27COOH and cholesterol ( ‫ف‬ 1/14,000) ( 38 ). Such high 7KCh-27OH/7KCh ratio