Isolevuglandins and Mitochondrial Enzymes in the Retina

We report the first peptide mapping and sequencing of an in vivo isolevuglandin-modified protein. Mitochondrial cytochrome P450 27A1 (CYP27A1) is a ubiquitous multifunctional sterol C27-hydroxylase that eliminates cholesterol and likely 7-ketocholesterol from the retina and many other tissues. We investigated the post-translational modification of this protein with isolevuglandins, arachidonate oxidation products. Treatment of purified recombinant CYP27A1 with authentic iso[4]levuglandin E2 (iso[4]LGE2) in vitro diminished enzyme activity in a time- and phospholipid-dependent manner. A multiple reaction monitoring protocol was then developed to identify the sites and extent of iso[4]LGE2 adduction. CYP27A1 exhibited only three Lys residues, Lys134, Lys358, and Lys476, that readily interact with iso[4]LGE2 in vitro. Such selective modification enabled the generation of an internal standard, 15N-labeled CYP27A1 modified with iso[4]LGE2, for the subsequent analysis of a human retinal sample. Two multiple reaction monitoring transitions arising from the peptide AVLK358(-C20H26O3)ETLR in the retinal sample were observed that co-eluted with the corresponding two 15N transitions from the supplemented standard. These data demonstrate that modified CYP27A1 is present in the retina. We suggest that such protein modification impairs sterol elimination and likely has other pathological sequelae. We also propose that the post-translational modifications identified in CYP27A1 exemplify a general mechanism whereby oxidative stress and inflammation deleteriously affect protein function, contributing, for example, to cholesterol-rich lesions associated with age-related macular degeneration and cardiovascular disease. The proteomic protocols developed in this study are generally applicable to characterization of lipid-derived oxidative protein modifications occurring in vivo, including proteins bound to membranes.

We report the first peptide mapping and sequencing of an in vivo isolevuglandin-modified protein. Mitochondrial cytochrome P450 27A1 (CYP27A1) is a ubiquitous multifunctional sterol C27-hydroxylase that eliminates cholesterol and likely 7-ketocholesterol from the retina and many other tissues. We investigated the post-translational modification of this protein with isolevuglandins, arachidonate oxidation products. Treatment of purified recombinant CYP27A1 with authentic iso [4]levuglandin E 2 (iso [4]LGE 2 ) in vitro diminished enzyme activity in a time-and phospholipid-dependent manner. A multiple reaction monitoring protocol was then developed to identify the sites and extent of iso [4]LGE 2 adduction. CYP27A1 exhibited only three Lys residues, Lys 134 , Lys 358 , and Lys 476 , that readily interact with iso [4]LGE 2 in vitro. Such selective modification enabled the generation of an internal standard, 15 N-labeled CYP27A1 modified with iso [4]LGE 2 , for the subsequent analysis of a human retinal sample. Two multiple reaction monitoring transitions arising from the peptide AVLK 358 (-C 20 H 26 O 3 )ETLR in the retinal sample were observed that co-eluted with the corresponding two 15 N transitions from the supplemented standard. These data demonstrate that modified CYP27A1 is present in the retina. We suggest that such protein modification impairs sterol elimination and likely has other pathological sequelae. We also propose that the post-translational modifications identified in CYP27A1 exemplify a general mechanism whereby oxidative stress and inflammation deleteriously affect protein function, contributing, for example, to cholesterol-rich lesions associated with age-related macular degeneration and cardiovascular disease. The proteomic protocols developed in this study are generally applicable to characterization of lipid-derived oxidative protein modifications occurring in vivo, including proteins bound to membranes.
The mitochondrial enzyme cytochrome P450 27A1 (CYP27A1) is expressed in many tissues where it catalyzes the C27-hydroxylation of cholesterol and other sterols, thereby playing an important role in the maintenance of cholesterol homeostasis, bile acid biosynthesis, and activation of vitamin D 3 (reviewed in Ref. 1). In specific organs, the functions of CYP27A1 could also include elimination of cytotoxic 7-ketocholesterol (2)(3)(4) and regulation of cholesterol homeostasis (5). Deficiency in CYP27A1 leads to cerebrotendinous xanthomatosis, a lipid storage disease with multiple manifestations, including premature atherosclerosis and retinal abnormalities (6 -8). Recently, CYP27A1 was found to be expressed in the retina (9,10) and was shown to be the major contributor to enzymatic degradation of cholesterol in this organ (11). Quantification of CYP27A1 in human retina revealed that one of its peptides, VVLAPETGELK 476 , was consistently under-represented by as much as 50% compared with other CYP27A1 peptides (10). However, this peptide under-representation was not observed in the human brain (10). We surmised the difference resulted from retina-specific post-translational modification due to the highly oxidative retinal environment and constant exposure to light (12)(13)(14).
The retina is known to contain high amounts of polyunsaturated fatty acids (PUFAs), 4 of which arachidonic acid is one of the most abundant (13). Arachidonyl phospholipids (PLs) have been shown to undergo free radical-induced oxidation in vivo to generate isolevuglandins (isoLGs) as demonstrated by the presence of isoLG-protein adducts in human serum (15). Most isoLG isomers such as iso [4]levuglandin E 2 (iso [4]LGE 2 ) are only generated via free radical-mediated oxidation (16). One particular isomer, levuglandin E 2 , may also form through rearrangement of prostaglandin H 2 , a product of cyclooxygenaseinduced oxidation (17). Regardless of origin, isoLGs represent a family of ␥-keto aldehydes (16 -18), which are highly reactive toward free primary amines such as the ⑀-amine of lysine residues in proteins. They form covalent pyrrole-derived adducts with greater avidity than most other lipid oxidation products such as 4-hydroxynonenal (19 -21). In the human eye, isoLGadducted calpain-1 has been shown to accumulate in the trabecular meshwork with this modification abolishing enzyme activity (22).
Existing studies to quantify isoLG post-translational modification in vivo have focused on measuring total lysine-isoLG adducts following complete proteolytic digestion to free amino acids (23). This approach is a useful diagnostic tool for measuring long term markers of oxidative stress (18). However, it does not indicate which proteins are affected or the lysine residues modified. Furthermore, MALDI-TOF mass spectra of tryptic digests of adducted proteins have only been reported for modification in vitro (24 -27). Identification of proteins modified in vivo has only been based on Western blot analysis of two-dimensional SDS-PAGE using antibodies against adducts and subsequent proteomic analysis (28). Although the proteins identified co-migrate with immunoreactivity against specific adducts, proteomic evidence establishing the in vivo presence of isoLG-derived modifications of specific aminoacyl residues of proteins in biological samples has never been reported.
Site-specific post-translational protein modifications can be detected and quantified using mass spectrometry-based multiple reaction monitoring (MRM) assay (29). MRM focuses only on specific m/z, although all other m/z are excluded. This allows selection of the appropriate precursor/product ion pairs, or transitions, for the analyte of interest thereby enhancing sensitivity and dynamic range. This methodology has proved particularly successful for profiling phosphorylation (30), acetylation (31), glycosylation (32), and ubiquitination (33) of soluble proteins. Quantification of other types of modification using this approach is rare (27) and does not extend to membrane proteins.
In this study, we demonstrate how a combination of in vitro studies utilizing purified enzyme of interest and authentic modifying reagent followed by MRM analysis of a biological sample with the 15 N-labeled form of the enzyme as an internal standard enables detection of unconventional protein modification. This work creates a paradigm for similar studies on other proteins. Moreover, it provides novel mechanistic insight into how oxidative stress/inflammation could impair cholesterol elimination from the retina and lead to formation of cholesterol-rich drusen, a hallmark of age-related macular degeneration (AMD) (34,35).

EXPERIMENTAL PROCEDURES
Materials-Ammonium chloride ( 15 N, 99%) was purchased from Cambridge Isotope Laboratories (Andover, MA). DC protein assay kit was from Bio-Rad. Sequencing grade trypsin and chymotrypsin were from Promega Corp. (Madison, WI). Complete EDTA-free protease inhibitor mixture tablets were purchased from Roche Applied Science. 1,2-Dilauroyl-snglycero-3-phosphocholine (DLPC) and 1,2-dioleoyl-snglycero-3-phosphoethanolamine (DOPE) were purchased from Avanti Polar Lipids (Alabaster, AL). Authentic iso [4]LGE 2 was synthesized as described and assayed by NMR (16). All other chemicals were purchased from Sigma. Post-mortem human eyes were obtained from the Georgia Eye Bank followed by isolation of retina as described previously (11) within 16 h from time of death. Human tissue use conformed to the Declaration of Helsinki and institutional review at Case Western Reserve University. The donor (PM28) was an 86-year-old, Caucasian male diagnosed with dry AMD in both eyes. Whole bovine eyes were obtained from local slaughterhouses, and the retinas were isolated as described previously (11).
Expression and Purification of CYP27A1-Natural isotopic abundance (NIA) and 15 N-labeled recombinant human CYP27A1 were expressed as described previously (10,36). The P450 concentrations were calculated from the CO-reduced difference spectra using an absorption coefficient of 91 mM Ϫ1 cm Ϫ1 between A 450 and A 490 (37).
Treatment of CYP27A1 with Iso [4]LGE 2 in the Absence of PLs and Subsequent Characterization-In all descriptions of treatment of CYP27A1 with iso [4]LGE 2 , molar excesses are expressed relative to protein lysine content (23 mol of lysine/ mol of CYP27A1). Purified recombinant NIA-CYP27A1 (5 nmol) was added to 50 mM potassium phosphate buffer (KP i ), pH 7.2, containing 100 M diethylenetriaminepentaacetic acid (DTPA) and a 10-fold excess of iso [4]LGE 2 (1.150 mol) in a final volume of 1 ml. The reaction mixture was kept at room temperature with gentle shaking for 120 min and then quenched with 100 mM glycine to neutralize unreacted iso [4]LGE 2 . A portion of the mixture was used for measuring the reduced CO difference spectrum and assay of enzyme activity. The remainder was flash-frozen in liquid nitrogen and stored at Ϫ80°C for further analysis by MS. The control incubation lacked iso [4]LGE 2 . Enzyme activity was assessed in the in vitro reconstituted system consisting of 200 pmol of CYP27A1, 6,000 pmol of adrenodoxin, 800 pmol of adrenodoxin reductase, 11 pmol of [ 3 H]cholesterol, and 38 g/ml DLPC. The conversion of cholesterol to 27-hydroxycholesterol was measured by HPLC with in-line ␤ detection as described previously (38).
For LC-MS/MS analysis, in-gel and in-solution digests were performed using either trypsin or chymotrypsin. For in-gel digest, NIA-CYP27A1 (15 pmol) was subjected to 10% SDS-PAGE. The gel was stained with Coomassie Blue, and a region corresponding to proteins with molecular mass between 50 and 60 kDa was excised, destained, reduced in 20 mM DTT for 30 min at room temperature, and alkylated in 100 mM iodoacetamide in the dark for 30 min at room temperature. The protein was then digested by chymotrypsin or trypsin for 22 h at 37°C at a protease to P450 ratio of 1:50 (w/w) in 50 mM NH 4 HCO 3 , pH 7.8 buffer. Peptides were extracted with 50% acetonitrile containing 5% formic acid, dried in a vacuum concentrator, and stored at Ϫ20°C. For in-solution digest, NIA-CYP27A1 (1 nmol) was mixed with 1 nmol of unmodified 15 N-CYP27A1. The protein solution was dialyzed against 50 mM KP i , pH 7.2, containing 100 M DTPA, 0.2 M NaCl, overnight to remove detergent and glycerol. The sample was concentrated to a volume of 5 l in an Amicon Ultracentrifugal filter device (regenerated cellulose, 50-kDa molecular mass cutoff, Millipore, Billerica, MA) followed by protein denaturation in 8 M urea, reduction in 10 mM DTT for 30 min at room temperature, and alkylation in 25 mM iodoacetamide in the dark for 30 min at room temperature. The protein was then digested by chymotrypsin or trypsin for 22 h at 37°C at a protease to P450 ratio of 1:50 (w/w) in 10 mM Tris-HCl, pH 8.0, containing 0.8 M urea. The digest solution was applied to Ultra-Micro PrepTips (C 18 reverse phase extraction pipette tip columns, The Nest Group, Southborough, MA) to remove residual salt and detergent. Peptides were dried in a vacuum concentrator and stored at Ϫ20°C. Peptide separations were performed on an Ultimate 3000 LC system with a C 18 Acclaim PepMap 100 column (0.075 ϫ 150 mm, Dionex, Sunnyvale, CA). Peptides were eluted over a 50-min gradient from 0 to 80% acetonitrile in water, containing 0.1% formic acid, at a flow rate of 300 nl/min. The column effluent was continuously directed into the nanospray source of the mass spectrometer, a hybrid Fourier transform ion cyclotron resonance (FTICR)/linear ion trap mass spectrometer (LTQ FT Ultra, Thermo Scientific, West Palm Beach, FL). The following parameters were used for all acquisition methods on the LTQ FT Ultra MS: an ion spray voltage of 2400 V and an interface capillary heating temperature of 200°C. Full mass spectra were acquired from the FTICR, and the tandem mass spectra (MS/MS) of the eight most intense ions were recorded by the linear ion trap in data-dependent mode with normalized collision energy of 35 eV, isolation width of 2.5 Da, and activation Q of 0.25.
Mascot Data Base Search to Identify Modified Peptides-Peptides were identified from LTQ FT Ultra MS/MS experimental data by generating peak lists with Mascot Daemon and. submitting these to the Mascot search engine, version 2.3.0 (Matrix Science, Boston). S-Carbamidomethylation of cysteine was set as a fixed modification, whereas oxidation of methionine (methionine sulfoxide) was set as a variable modification. Formulas corresponding to the various oxidation and dehydration states of iso [4]LGE 2 adducts (39) were manually entered into the Mascot residue modification data base and selected as variable modifications for lysine (supplemental Table S1, structures shown in supplemental Fig. S1). Mass tolerances were Ϯ15 ppm for precursor ions and Ϯ 0.8 Da for fragment ions. One missed cleavage site was allowed for trypsin (cleaves at residues Lys and Arg but not at Pro) and three missed cleavage sites for chymotrypsin (cleaves at Phe, Leu, Trp, and Tyr but not at Pro). Searches were restricted to a sequence data base containing only human CYP27A1 (Swiss-Prot number Q02318, residues 34 -531) because purified protein was used. Only peptides with a significant score greater than 13 (p Ͻ 0.05) according to Mascot's scoring algorithm were considered.
Isolation of Mitochondrial PLs-Isolation of mitochondria from bovine retinal tissue was adapted from an established protocol (40). Fresh retina (10g wet tissue) was homogenized with a Teflon pestle on ice in 20 ml of homogenization buffer consisting of 50 mM Tris-Cl, pH 7.4, containing 250 mM sucrose, 5 mM MgCl 2 , 1 mM PMSF, 1 mM DTT, 1 tablet of the protease inhibitor mixture per 50 ml, and 100 g/ml butylated hydroxytoluene. Homogenate was centrifuged at 1,000 ϫ g for 15 min, and the supernatant was transferred to a new tube and centrifuged at 9,000 ϫ g for 30 min. The supernatant was discarded, and the pellet was resuspended in 10 ml of homogenization buffer and centrifuged at 9,000 ϫ g for 20 min, again discarding the supernatant. The remaining pellet was resuspended in 5 ml of 10 mM Tris-Cl, pH 7.4, containing 250 mM sucrose, 5 mM MgCl 2 , 1 mM PMSF, 1 mM DTT, 1 tablet of the protease inhib-itor mixture per 50 ml, and 100 g/ml butylated hydroxytoluene. Total mitochondrial lipids were extracted by vortexing mitochondria in 32 ml of Folch reagent (chloroform/methanol, 2:1, v/v) for 30 min (41). To aid phase separation, 2 ml of 150 mM NaCl was added to the suspension, vortexed for 5 min, and centrifuged at 1,000 ϫ g for 10 min. The lower organic phase was transferred to a clean glass tube. The aqueous phase was re-extracted two more times by vortexing with 16 ml of Folch reagent for 30 min, adding 2 ml of 150 mM NaCl, and vortexing for 5 min followed by centrifugation at 1,000 ϫ g for 10 min, and transferring the organic phase to a clean glass tube. All organic phases were pooled and dried in a vacuum concentrator (Savant SC210A SpeedVac Concentrator, Thermo Scientific, Asheville, NC), yielding 0.054 g of lipids. The residue was then dissolved in 500 l of chloroform and loaded onto a 690-mg silica solid phase extraction cartridge (Sep-Pak Classic Silica, Waters) for separation (42). Nonphosphorous lipids were washed from the cartridge with 20 ml of chloroform. PLs were eluted with 30 ml of methanol and concentrated under vacuum, yielding 0.044 g. PLs were dissolved in methanol (10 mg/ml) and stored at Ϫ20°C. Purity of PLs and nonphosphorous lipids was assessed by silica thin layer chromatography with a mobile phase consisting of hexane/diethyl ether/methanol/acetic acid (90:20:5:2, v/v) and detected by staining with iodine vapor as described previously (43).
Reconstitution of CYP27A1 into Liposomes-Stock solutions of retinal mitochondrial PLs, a mixture of DLPC and DOPE (4:3, w/w), and DLPC alone were prepared by dissolving 10 mg of PLs in methanol. PLs (1.25 mg, ϳ1800 nmol) were dispensed into a glass tube, and the solvent was evaporated under nitrogen. PLs were hydrated for 10 min in 200 -500 l of 50 mM KP i , pH 7.2, containing 100 M DTPA, briefly vortexed, and sonicated for 30 min at 50% power (Digital Sonifier S-450D, Branson Ultrasonics, Danbury, CT) (44). Purified recombinant CYP27A1 (5 nmol) was added to the liposome solution and allowed to incorporate for 30 min at 25°C.
Treatment of CYP27A1 with Iso [4]LGE 2 in the Presence of PLs and Subsequent Characterization-The P450/liposome mixture (5 nmol of P450) was incubated with a 2-fold molar excess of iso [4]LGE 2 in a final volume of 1 ml at room temperature with gentle shaking. Aliquots were taken at different time points, quenched, and used for analysis by the following: 1) the CO-reduced difference spectrum; 2) assay of enzyme activity; and 3) MRM. Enzyme activity was measured as described above except no extra DLPC was added. For MRM analyses, modified NIA-CYP27A1 (100 pmol) was mixed with an equimolar amount 15 N-CYP27A1 and digested with trypsin for 22 h at 37°C at a protease to P450 ratio of 1:50 (w/w) in 25 mM NH 4 CO 3 , pH 7.9. Peptide separations were performed on a nanoLC-2D (Eksigent, Dublin, CA) with a PicoFrit (75 m inner diameter, 10-m tip inner diameter, New Objective, Woburn, MA) column self-packed to a bed length of 12 cm with Reprosil-Pur 120 C18-AQ, 3 m resin (Dr. Maisch, GmbH, Germany). Peptides were eluted over a 42-min gradient from 13 to 31% acetonitrile in water containing 0.1% formic acid at a flow rate of 300 nl/min. The column effluent was continuously directed into the nanospray source of the mass spectrometer, a hybrid triple quadrupole/linear ion trap mass spectrometer (4000 QTRAP, ABI/MDS-Sciex, Carlsbad, CA). All acquisition methods used the following parameters: an ion spray voltage of 2100 V, curtain gas of 30 p.s.i., source gas of 9 p.s.i., interface heating temperature of 170°C, declustering potential of 76 V for ϩ2 precursor ions and 65 V for ϩ3 precursor ions, collision energy of 30 eV for ϩ2 precursor ions and 22 eV for ϩ3 precursor ions, and collision cell exit potential of 16 V for ϩ2 precursor ions and 13V for ϩ3 precursor ions. The dwell time for all transitions was 40 ms.
The mass spectrometer monitored three transitions per peptide. Selection of transitions was as described previously (45). The identities of the measured peptides were confirmed based on two parameters of the internal standard that was run under the same conditions: 1) the retention time of the three MRM peaks from a given peptide, and 2) the ratio among the three MRM peaks. The means Ϯ S.D. were calculated by treating the three transitions for each of the target peptides and the three experimental replicates all as independent measurements.
Preparation of 15 N-CYP27A1 Modified with Iso [4]LGE 2 -15 N-CYP27A1 (1 nmol) in 50 mM KP i , pH 7.2, containing 100 M DTPA was incubated with an equimolar amount of iso [4]LGE 2 in a final volume of 0.2 ml at room temperature with gentle shaking for 15 min. The reaction mixture was quenched by addition of 1 l of 100 mM glycine. The mixture was flashfrozen in liquid nitrogen and stored at Ϫ80°C.
Processing of the Human Retina-Retina was placed in 25 mM NH 4 HCO 3 and homogenized by sonication at 30 watts using three continuous 10-s cycles (Sonicator 3000, Misonix Inc., Farmingdale, NY). The total protein concentration was measured in the presence of 2% (w/w) SDS using the DC protein assay kit and bovine serum albumin as a standard. The homogenate was centrifuged at 153,000 ϫ g for 30 min, and the resulting total membrane pellet was frozen and stored at Ϫ80°C. For proteolytic hydrolysis, the total membrane pellet was thawed and resuspended in 25 mM NH 4 HCO 3 , pH 7.9, containing 0.2% (w/w) sodium cholate and supplemented with modified 15 N-CYP27A1 (25 pmol) as an internal standard. The samples were heated at 90°C for 5 min, cooled to room temperature, and treated with trypsin for 15 h at 37°C. The trypsin to P450 ratio was 1:50 (w/w). After trypsinolysis, the samples were centrifuged at 153,000 ϫ g for 30 min, and the supernatants were transferred to new tubes. Each supernatant was then treated with 0.5%, w/v, TFA for 30 min at 37°C and centrifuged again at 153,000 ϫ g for 30 min. The supernatants from the second centrifugation were transferred to new tubes again, mixed with an equal volume of acetonitrile, and dried using a Vacufuge (Eppendorf AG, Hamburg, Germany).

RESULTS AND DISCUSSION
Characterization of CYP27A1 after Treatment with Iso [4]LGE 2 in the Absence of PLs-The effect of the iso [4]LGE 2 treatment (10-fold molar excess for 120 min) on functional properties of CYP27A1 was assessed using two assays: the COreduced difference spectrum, which reflects the content of the functional P450, and an enzyme assay, which reflects the ability of CYP27A1 to hydroxylate cholesterol in a reconstituted enzyme system. Only a peak at 420 nm was observed in the CO spectrum of the treated CYP27A1 indicating enzyme denaturation under the experimental conditions used. Also, no enzyme product, 27-hydroxycholesterol, was detected in the enzyme assay. In contrast, the P450 content and enzyme activity in control incubations were similar to those observed at the initiation of treatment. These data suggest that the iso [4]LGE 2 treatment leads to modification of CYP27A1 and is accompanied by enzyme denaturation and loss of activity. To confirm modification, the treated CYP27A1 was trypsinolyzed in-gel and in-solution followed by FTICR MS/MS. Peak lists from the acquired MS/MS spectra were submitted to the Mascot data base search engine. Evaluation criteria for identification of the modified peptides were as follows: 1) a significant peptide score as reported by Mascot's scoring algorithm (46); 2) the existence of y and b series ions upstream and downstream of the modified lysine in the tandem MS spectra; 3) detection under different types of hydrolysis; and 4) the existence of multiple adduct dehydration and oxidation states. Mascot analysis of the spectra from the in-solution tryptic digest (17% sequence coverage, encompassing 26% of the Lys residues) did not identify any modified peptides with a significant score. In contrast, analysis of the in-gel tryptic digest (72% sequence, 56% Lys coverage) identified two distinct peptides, AVLK 358 ETLR and VVLAPETGELK 476 SVAR, modified at Lys 358 and Lys 476 (Table 1). In both peptides, modified lysine residues were within the peptide and not at the C terminus, indicating that modification results in missed tryptic cleavage. Remarkably, the latter peptide is an expansion of VVLA-PETGELK 476 , which was consistently under-represented in our previous study assessing by MRM the CYP27A1 content in human donor retinas (10). To increase sequence coverage, modified CYP27A1 was digested with chymotrypsin in-gel (96% sequence, 100% Lys coverage) and in-solution (95% sequence, 100% Lys coverage). Mascot analysis identified only one additional peptide with modification at Lys 134 , NQRLLK 134 PAEAALY, for both in-gel and in-solution chymotrypsin digests, with the latter having a significant score (Table 1). Thus, based on our evaluation criteria, Mascot identified only 3 lysine residues out of 23 in CYP27A1 as modified by iso [4]LGE 2 . Lys 358 was shown previously to be  [4]LGE 2 Treatment of CYP27A1 Reconstituted into Different PLs-To model membrane insertion, purified recombinant CYP27A1 was reconstituted into liposomes of various composition as follows: bovine retinal mitochondrial PLs, a DLPC and DOPE mixture (4:3, w/w) which approximates the ratio of phosphatidylcholine to phosphatidylethanolamine found in human hepatic mitochondria (49), or DLPC which contains no free amines and is a standard model PL. In this set of experiments, the ratio of iso [4]LGE 2 was lowered to a 2-fold molar excess, and kinetics were investigated. Enzyme activity was decreased sharply by 75% at 5 min. Yet the P450 content decreased only by 25%, indicating that modification impaired activity without completely denaturing the protein. Both enzyme activity and P450 content plateaued after 15 min with little further reduction (Fig. 1). This is in contrast to the treatment with a 10-fold molar excess of iso [4]LGE 2 leading to complete P450 denaturation and loss of enzyme activity. Such effects of high iso [4]LGE 2 concentrations could be attributable to modification of additional, less reactive lysine residues resulting in peptides with multiple missed cleavage sites. Such peptides are difficult to unambiguously identify with Mascot because the number of false-positive hits increases as three or more missed cleavage sites are allowed. Furthermore, at high iso [4]LGE 2 concentrations some arginine and histidine residues could be modified. As the chemical structure of these modifications has not been demonstrated, Mascot search criteria could not be created. Consequently, these putative modifications were not investigated.
At each time interval, aliquots were withdrawn and spiked with the internal standard, 15 N-CYP27A1, in an amount approximately equimolar to NIA-CYP27A1. Proteins were subjected to proteolyses followed by quantitative assessment by MRM. The ratios of transition intensities between the NIA peptides and corresponding 15 N peptides were calculated and assumed to be equal to the initial mixing ratio of the proteins for NIA peptides not possessing modification and decreased for modified peptides proportionally to the extent of modification. Peptides that did not contain lysine residues and therefore could not be modified by iso [4]LGE 2 were analyzed first (Fig. 2,  Group I). Within an aliquot (Fig. 2, z axis), the ratio between these peptides varied up to 17% due to MRM quantification error (supplemental Table S2). The ratios also varied up to 19% from aliquot to aliquot (Fig. 2, x axis) likely due to pipetting errors. Because ratios were calculated independently for each time point, pipetting uncertainties did not contribute to uncertainties in quantification. Mixing ratios in group I were then used for calculating lysine residue modification in other peptides that were grouped based on sequence and observed changes in mixing ratio. Group II consisted of only one peptide, VVLAPETGELK 476 , in which the ratio for the 5-min reaction time was 0.69, whereas that in group I was 0.84 indicating that 18% of the peptide was modified (Fig. 2 and supplemental Table  S2). The extent of modification of this peptide further increased to 34 and 40% after 15 and 120 min, respectively. Thus, changes in both enzyme activity and peptide modification level off after 15 min of iso [4]LGE 2 treatment. Group III is composed of three lysine-containing peptides (WTRPVLPFWK 236 , YLDGWN-AIFSFGK 250 , and EIEVDGFLFPK 387 ) that showed either no modification or modification up to 16% (which is within experimental variation). This was substantially less than in group II. Finally, peptides that contain both lysine and methionine residues formed group IV because of possible background oxidation of the latter. SIGLMFQNSLYATFLPK 226 did not appear to be either modified or oxidized as indicated by no decrease in the ratio compared with group I. Conversely, NDMELWK 101 and LLK 134 PAEAALYTDAFNEVIDDFMTR seemed to have methionine oxidation because the ratios in the control aliquot were lower than in group I. DFAHMPLLK 354 was likely unmodified in the presence of retinal mitochondrial PLs as the mixing ratio was close to experimental error. Furthermore, the presence of methionine in the peptide introduced additional uncertainty in estimating the extent of modification.
This analysis was repeated for CYP27A1 reconstituted into liposomes composed of a mixture of DLPC and DOPE or DLPC alone (Figs. 3 and 4 and supplemental Tables S3 and S4). In all three types of PLs, Lys 476 was the most extensively modified as assessed by quantitative disappearance of the peptide VVLA-PETGELK 476 . This is in agreement with direct observation of the modified peptide VVLAPETGELK 476 (-C 20 H 26 O 3 )SVAR in the FTICR MS experiments. Moreover, groups II and III peptides showed higher degrees of modification when CYP27A1 was incorporated into DLPC. Most group III peptides showed no modification in mitochondrial PLs and the DLPC/DOPE mixture, which both contained primary amines. This suggests that the modification is PL-dependent, which is likely due to a combination of differing CYP27A1 membrane insertion modes Values are expressed as a percentage relative to control (CYP27A1 incubated without iso [4]LGE 2 ). Enzyme activities are the average of three experiments Ϯ S.E. and utilization of iso [4]LGE 2 for both protein and PL modification.
To generate Lys 134 -and Lys 354 -encompassing peptides, which do not contain methionine, and a Lys 358 peptide, which was unobservable under tryptic hydrolysis, we mixed NIA-CYP27A1 modified in the presence of mitochondrial PLs with an equimolar amount of 15 N-CYP27A1 and performed chymotryptic hydrolysis. The average mixing ratio was calculated using the following two peptides: LRNSQPATHPRIQHPF 431 and LFPK 387 NTQF (Fig. 5 and Table S5). Three peptides, LK 134 PAEAALY, K 134 PAEAALY, and NQRLLK 134 PAEAALY, were monitored and showed unchanged ratios indicating that Lys 134 was not modified under these conditions. Lack of modification at Lys 134 may be due to the lower molar excess of isoLG and residue inaccessibility after insertion into the liposome. Ratios for LK 354 AVLK 358 ETL were lower by at least 30% (supplemental Table S5) in comparison with the average mixing ratio, suggesting modification. Modification is likely assignable to Lys 358 because of direct observation of adducted Lys 358 in the FTICR MS experiment and the small change in mixing ratios for the Lys 354 peptide in the presence of retinal mitochondrial PLs in the tryptic hydrolysis experiments. K 476 SVARIVL ratios   Fig. 2. Data labels have been rounded to one digit after the decimal, and some labels have been omitted for clarity. For all data points and statistical treatment, see supplemental Table S3. were also lower by at least 20%, in agreement with the results of the tryptic hydrolysis.
Overall, the data demonstrate that PL-reconstituted CYP27A1 possesses two lysine residues, Lys 358 and Lys 476 , that interact even with a 2-fold molar excess of iso [4]LGE 2 . Modification leads to missed cleavage and an increased m/z, thus lowering the yield of Lys 358 -and Lys 476 -containing peptides. These studies justified our subsequent experiments aimed at identification of iso [4]LGE 2 modification in the retina.
Preparation and Characterization of the Iso [4]LGE 2 -modified 15 N-CYP27A1 Internal Standard-To unambiguously detect iso [4]LGE 2 -CYP27A1 adducts in a biological sample, a 15 N-labeled internal standard with modified Lys 358 and Lys 476 should be used. This standard should have minimal modification at flanking lysine residues, which could lead to complex peptide patterns through multiple missed cleavage sites resulting from adduction. Therefore, we further lowered the excess of iso [4]LGE 2 over protein lysine content to a 1:1 molar ratio.  Table S6). The list of transitions was generated by the R software package OrgMassSpecR. We indeed observed multiple transitions for both peptides adducted with iso [4]LGE 2 (data not shown) indicating modification still occurred with an even lower amount of iso [4]LGE 2 .
Modified 15 N-CYP27A1 was then mixed with an equimolar amount of NIA-CYP27A1 treated with a 2-fold molar excess of iso [4]LGE 2 for different times in the presence of mitochondrial PLs. The mixtures were hydrolyzed with trypsin and monitored   Table 2). This analysis indicated that the MRM characteristics of modified 15 N-CYP27A1 were similar to that of modified NIA-CYP27A1 and that the former could be used for detection of this adduct in biological samples.
Detection of Iso [4]LGE 2 -CYP27A1 Adducts in Human Retina-Initially, the total membrane pellet from a portion of the human neural retina was supplemented with modified 15 N-CYP27A1 and treated with trypsin. Unfortunately, MRM analysis revealed transitions only from the internal standard indicating a need for sample enrichment compatible with MRM workflow. Therefore, the total membrane pellet was separated by preparative 10% SDS-PAGE, and a Mini Whole Gel Eluter (Bio-Rad) was used to collect 14 fractions following the manufacturer's protocol. The fractions were re-analyzed by 10% SDS-PAGE to determine the molecular weight range of each (supplemental Fig. S2). Fractions 7 and 8, encompassing ϳ50 -65-kDa proteins, were pooled for further analysis based on the molecular mass of CYP27A1 (57 kDa). This pooled fraction was precipitated by chloroform/methanol, treated with trypsin, and subjected to MRM. A list of transitions monitored is summarized in supplemental  (Fig. 6). All the transitions were eluted at the same retention time. The observed y6 ϩ 1 and y4 ϩ 1 product ions encompassed regions of the precursor ion upstream and downstream of the modification, respectively. This is the first direct proof of isoLG modification in human retina and the first peptide mapping and sequencing of an in vivo isoLG-modified protein.
The (C 20 H 26 O 3 )-isobaric adduct that we detected in human retina could arise either from modification by iso [4]LGE 2 , which is generated nonenzymatically via free radical oxidation of arachidonyl phospholipids (21), or from modification with the isoLG isomer LGE 2 , which could also be generated by the cyclooxygenase pathway (reviewed in Ref. 18). The latter reflects an inflammatory process and may certainly have occurred in the retinal sample of our 86-year-old donor afflicted by age-related macular degeneration, a disease whose pathogenesis involves chronic inflammation (50). Although identification of the specific isoLG isomers involved in modifi-  cation of CYP27A1 in the retina is underway in this laboratory, the deleterious effects of the isoLG adduction on enzyme function is consistent with retinal abnormalities developed in individuals with CYP27A1 deficiency (51). Furthermore, because of age-related systemically elevated levels of oxidative injury, isoLGs could also modify CYP27A1 in vascular endothelium and macrophages, where it is abundant and plays an antiatherogenic role by eliminating cholesterol (52,53) and the auto-oxidation product, 7-ketocholesterol (2, 3), which is implicated in initiation of apoptosis and inflammation. IsoLG modifications in vascular endothelium and macrophages would explain premature atherosclerosis observed in individuals lacking CYP27A1 (51). IsoLGs promiscuously modify protein amino groups. Therefore, it is likely that isoLG modification affects not only CYP27A1 but other proteins, including those involved in maintenance of cholesterol homeostasis. Such modification could interfere with protein function, as suggested by this study and/or trigger other events as suggested by previous studies. IsoLG modification was shown to increase resistance to proteosomal degradation (22) and elicit immunogenic (24) and cytotoxic (54) responses to isoLG-adducted molecules, thus leading to impaired cellular health. We propose that increased oxidation of PUFAs represents a common link between aging/oxidative stress/inflammation and AMD as well as cardiovascular disease (Fig. 7). Consequently, availability of methodologies that enable identification of individual isoLG-modified proteins and sites of modification on the protein molecule will significantly facilitate investigation of this link and enhance our understanding of the mechanisms underlying widespread diseases such as AMD and atherosclerosis.
In summary, in this study, we investigated whether CYP27A1, mediating cholesterol removal from extrahepatic tissues, is modified by oxidized arachidonic acid in vitro and in the human retina. We first established that iso [4]LGE 2 treatment of purified recombinant CYP27A1 in solution abolished enzyme activity and led to the formation of iso [4]LGE 2 adducts. Next, we reconstituted CYP27A1 in different PLs and ascertained that the enzyme is also modified even when other biological amines were present and iso [4]LGE 2 concentrations were lower. MRM analysis revealed that modification in CYP27A1 occurred mostly at three residues, Lys 134 , Lys 358 , and Lys 476 , of which Lys 358 is known to interact with the redox partner and is important for enzyme activity (47). Then 15 N-CYP27A1 was generated, treated with iso [4]LGE 2 , and confirmed to be a suitable internal standard. Finally, a sample of AMD-afflicted human retina was supplemented with iso [4]LGE 2 -modified 15 N-CYP27A1, enriched, and shown by MRM to contain isoLGmodified NIA-CYP27A1. This finding enhances our understanding of the significance of PUFA oxidation, provides novel insight into mechanisms contributing to AMD and cardiovascular disease, and creates the basis for our future studies investigating whether CYP27A1-isoLG levels could serve as marker for both oxidative stress and impaired cholesterol elimination. The analytical protocol developed in this work is applicable to the investigation of other proteins and other types of modification and is expected to have broad application.