Identification of a Novel Lipofuscin Pigment (iisoA2E) in Retina and Its Effects in the Retinal Pigment Epithelial Cells*

Background: Macular degeneration implicates lipofuscin deposition in the retina. Results: Bisretinoid iisoA2E in the retina was characterized; excessive accumulation of iisoA2E was cytotoxic to retinal pigment epithelial cells. Conclusion: Pyridinium iisoA2E is a unique diretinal adduct and serves as a fluorescent biomarker of aberrant all-trans-retinal metabolism. Significance: Characterization of iisoA2E gives a more complete understanding of the biosynthesis of retinal bisretinoid lipofuscin. Lipofuscin accumulation in retinal pigment epithelial (RPE) cells of the eye implicates the etiologies of Stargardt disease and age-related macular degeneration, a leading cause of blindness in the elderly. Here, we have identified a previously unknown RPE lipofuscin component. By one- and two-dimensional NMR techniques and mass spectrometry, we confirmed that this compound is a new type of pyridinium bisretinoid presenting an unusual structure, in which two polyenic side chains are attached to adjacent carbons of a pyridinium ring. This pigment is a light-induced isomer of isoA2E, rather than A2E, referred to as iisoA2E. This pigment is a fluorescent lipofuscin compound with absorbance maxima at ∼430 and 352 nm detected in human, pig, mouse, and bovine eyes. Formation of iisoA2E was found in reaction mixtures of all-trans-retinal and ethanolamine. Excess intracellular accumulation of this adduct in RPE cells in vitro leads to a significant loss of cell viability and caused membrane damage. Phospholipase D-mediated phosphodiester cleavage of the A2PE series generated isoA2E and iisoA2E, in addition to A2E, thus corroborating the presence of isoA2PE and iisoA2PE that may serve as biosynthetic precursors of isoA2E and iisoA2E.


Lipofuscin accumulation in retinal pigment epithelial (RPE) cells of the eye implicates the etiologies of Stargardt disease and
age-related macular degeneration, a leading cause of blindness in the elderly. Here, we have identified a previously unknown RPE lipofuscin component. By one-and two-dimensional NMR techniques and mass spectrometry, we confirmed that this compound is a new type of pyridinium bisretinoid presenting an unusual structure, in which two polyenic side chains are attached to adjacent carbons of a pyridinium ring. This pigment is a light-induced isomer of isoA2E, rather than A2E, referred to as iisoA2E. This pigment is a fluorescent lipofuscin compound with absorbance maxima at ϳ430 and 352 nm detected in human, pig, mouse, and bovine eyes. Formation of iisoA2E was found in reaction mixtures of all-trans-retinal and ethanolamine. Excess intracellular accumulation of this adduct in RPE cells in vitro leads to a significant loss of cell viability and caused membrane damage. Phospholipase D-mediated phosphodiester cleavage of the A2PE series generated isoA2E and iisoA2E, in addition to A2E, thus corroborating the presence of isoA2PE and iisoA2PE that may serve as biosynthetic precursors of isoA2E and iisoA2E.
Vision depends on a biochrome that consists of a light-sensitive protein called opsin attached to a chromophore. The visual chromophore for most vertebrate opsins is 11-cis-retinal. After capturing a photon by an opsin pigment, 11-cis-retinal bound to lysine 296 of opsin isomerizes to all-trans-retinal (1,2). Restoration of light sensitivity to the bleached opsin pigment involves chemical reisomerization of all-trans-retinal back to 11-cis-retinal via an enzymatic process called the visual cycle (3). Most steps of the visual cycle occur within retinal pigment epithelial (RPE) 3 cells. An additional RPE cell function is to phagocytose the distal tips of photoreceptor outer segments, which are shed by photoreceptor cells on a daily basis (4,5). Based on these functions, RPE cells are extremely critical for the maintenance of photoreceptor viability. Ongoing shedding and phagocytosis of distal photoreceptor outer segments lead to gradual accumulation of fluorescent retinoids, lipids, and protein debris, called lipofuscin, in RPE phagolysosomes (6,7). Deleterious lipofuscin accumulation in the eye is considered to be one of the causative factors responsible for blindness in patients with retinal disorders, particularly age-related macular degeneration (8) and Stargardt disease (9,10). Throughout the life of an individual, RPE cells of the eye accumulate fluorescent bisretinoids, which constitute the lipofuscin of the cells. These pigments are derived from reactions of all-trans-retinal. The first RPE lipofuscin constituent to be described was A2E (Fig.  1A), an unprecedented pyridinium bisretinoid. Previous evidence revealed that excessive accumulation of A2E in RPE cells in vitro can mediate detergent-like effects on cell membranes (11) as well as lead to the alkalinization of lysosomes (12) and detachment of proapoptotic proteins from mitochondria (13). All of the double bonds along the side arms of A2E assume the trans (E) configuration, whereas isoA2E (Fig. 1B), an isomer of A2E, has one cis (Z) olefin at the C13C14 position and exhibits a long-arm absorbance maximum about 12 nm blue-shifted from that of A2E.
On the other hand, the visual cycle produces a high flux retinoid that can lead to elevated levels of toxic retinoid intermediates, in particular all-trans-retinal, which can cause photore-ceptor degeneration (14). Palczewski and co-workers (15) pointed out that all-trans-retinal, the precursor of bisretinoids, could trigger caspase activation and mitochondrially associated cell death and was implicated in the complex pathogenesis of Stargardt disease and age-related macular degeneration (16,17). In the absence of light exposure, 11-cis-retinal, but not all-trans-retinal, was determined to be the primary source of lipofuscin components (18).
Insight into the RPE lipofuscin composition and the biosynthetic pathways by which these compounds form helps to understand retinal disorders caused by an overload of toxic lipofuscin. Here, we report that a previously unknown fluorescent lipofuscin pigment with absorbance maxima at 430 and 352 nm is present in eyecups of C57BL/6 and BALB/cByJ mice. A notable feature of this molecule was the similar A2E and isoA2E mass/charge ratio (m/z 592) yet with lower polarity and a ϳ15-nm red-shifted max value of the short arm versus that of A2E and isoA2E (ϳ337 nm). This fluorophore was also detected in human RPE. Interestingly, it is the light-mediated isomer of isoA2E, rather than A2E, and is much less susceptible to photocatalytic oxidation than A2E and isoA2E. Herein, we have detected and characterized a light-induced isomer of isoA2E with a unique pyridinium bisretinoid structure that differs from the cis-isomers of A2E. Cell-based assays elucidated possible cytotoxic effects of this compound in human RPE cells. Furthermore, we propose a biosynthetic pathway by which iisoA2E is formed and a mechanism by which isoA2E is converted into this isoform via light.

EXPERIMENTAL PROCEDURES
Materials-All-trans-retinal and dipalmitoyl-L-␣-phosphatidylethanolamine (DP-PE) were purchased from Sigma-Aldrich. Phospholipase D (PLD) from Streptomyces chromofuscus was purchased from Merck Millipore. A colorimetric lactate dehydrogenase (LDH) assay kit was purchased from Shanghai Biyuntian Biological Science & Technology Co., Ltd. (Shanghai, China). Ethanolamine and HPLC grade trifluoroacetic acid were obtained from Aladdin. HPLC grade acetonitrile and methanol (MeOH) were purchased from Fisher. All other chemical reagents were AR grade.
Animals and Tissues-C57BL/6, BALB/cByJ, and Rpe65 rd12 mice were purchased from the Shanghai SLAC Laboratory Animal Co. Ltd. (Shanghai, China) and raised under 12 h on/off cyclic lighting with an in-cage illuminance of 60 -90 lux. All procedures were approved by the Institutional Animal Care and Use Committee and complied with guidelines set forth by the Ophthalmology Branch of the Chinese Medical Association. Human donor eyes were received within 12 h post mortem from the Eye Center of the Second Affiliated Hospital of Zhejiang University for Sight Restoration (Hangzhou, ZJ, China). All procedures were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2000 (5). Informed consent was obtained from all patients prior to study inclusion. Fresh bovine and pig eyes were obtained from local slaughterhouses.
Tissue Extraction and HPLC Analysis-Murine posterior eyecups (8 -12 eyes/sample), human RPE/choroid (1 eye/sample), bovine RPE (1 eye/sample), pig RPE (1 eye/sample), and ARPE-19 cell tissues were analyzed. Tissues were homogenized in a v/v solution of phosphate-buffered saline (PBS) and 50% methanolic chloroform using a 7-ml scale glass tissue grinder and then transferred to a 15-ml conical centrifuge tube. Following centrifugation at 3000 ϫ g for 5 min, the organic layer was placed in a 25-ml round-bottom flask. Residues were collected from three consecutive chloroform extractions. After the removal of combined solvents in a rotary evaporator, the residual material was transferred into a 0.5-ml centrifuge tube with 50% methanolic chloroform and dried under argon gas. The resulting extract was redissolved in MeOH and centrifuged at 7500 ϫ g for 1 min. The supernatant was examined by reversephase HPLC using an Alliance System (Waters Corp., Milford, MA) equipped with a 2695 separation module, a 2998 photodiode array detector, and a 2475 multichannel () fluorescence detector. For chromatographic separation, an analytical scale Atlantis dC18 (3 m, 4.6 mm ϫ 150 mm) column was used with a gradient mobile phase composed of acetonitrile and water in the presence of 0.1% trifluoroacetic acid: 75-90% acetonitrile (0 -30 min), 90 -100% acetonitrile (30 -40 min), and 100% acetonitrile (40 -100 min) with a flow rate of 0.5 ml/min. Photodiode array detection was monitored at 430 nm. HPLC extraction and injection were carried out under dim red light. Integrated peak areas (V⅐s) were determined by Empower version 3 software. Molar quantity per eye was calculated using a calibration curve constructed from known concentrations of synthesized standard and a molar extinction coefficient of 11,500 (iisoA2E in MeOH at 430 nm) and by normalizing to the HPLC injection volume versus sample volume ratio.
Biosynthetic Reaction and HPLC Analysis-A mixture of alltrans-retinal (50 mg) and ethanolamine (4.8 mg) in 3 ml of ethanol (EtOH) was stirred in the presence of acetic acid (5 l) in a 5-ml round-bottom flask with a sealed ground glass stopper at room temperature under dim light for 2 days. For HPLC analysis, the reaction mixture was diluted with MeOH and injected into an Atlantis dC18 reverse-phase column (3 m, 4.6 ϫ 150 mm). The mobile phase was a gradient of acetonitrile in water with 0.1% trifluoroacetic acid: 75-90% acetonitrile (0 -30 min), 90 -100% acetonitrile (30 -40 min), and 100% acetonitrile (40 -100 min) with a flow rate of 0.5 ml/min. The photodiode array detector was set at 430 nm for eluent monitoring.
NMR Spectroscopy-One-dimensional ( 1 H and 13 C) and two-dimensional (HMBC, HSQC, 1 H-1 H COSY, and NOESY) NMR spectra were recorded on a BrukerAvance 500-MHz spectrometer in CD 3 OD. Chemical shifts (␦ in ppm) were referenced to the carbon (␦ C 49.15) and residual proton (␦ H 3.31) signals of MeOD. Data processing was performed using vendor-supplied software. The NMR DEPT experiment was conducted using a polarization-transfer pulse of 135°.
Biomimetic Photo-oxidation-Solutions of HPLC-purified A2E, isoA2E, or iisoA2E in water (200 M) containing 2% DMSO was irradiated by a 500-W xenon illuminant (20,000 lux) for 10 min. The extent of oxidization was tested by mass spectrometry and reverse-phase HPLC. A Finnigan LCQ Deca XP plus ion trap mass spectrometer (Thermo Finnigan, San Jose, CA) equipped with an electrospray ionization (ESI) interface and an ion trap mass analyzer was utilized. A syringe pump was used for direct loop injection of sample solutions with a flow rate of 5 l/min. Operating parameters in the positive ion mode were as follows: collision gas, ultra-high purity helium; nebulizing gas, high purity nitrogen; ion spray voltage, 4.5 kV; nitrogen sheath gas, 5 arbitrary units; capillary temperature, 275°C; capillary voltage, 15 V; and tube lens offset voltage, 30 V. The energy for collision-induced dissociation was 35%, and the isolation width of precursor ions was 2.0 Th. For HPLC, an Atlantis dC18 (3 m, 4.6 mm ϫ 150 mm) reverse-phase column was used for the stationary phase, and for the mobile phase, a gradient of acetonitrile in water was used with 0.1% trifluoroacetic acid (85-100% acetonitrile (15 min) and 100% acetonitrile (15-30 min) with a flow rate of 0.8 ml/min) monitored at 430 nm with a 30-l injection volume.
Treatment and Cellular Uptake of iisoA2E-iisoA2E was stored as a stock solution in DMSO (10 mM) and kept at Ϫ80°C in the dark. Cells were incubated with serial dilutions of iisoA2E (0.625, 1.25, 2.5, 5, 10, 20, and 40 M) for 1-3 days to assess cell viability. All cell-based experiments included untreated cells as controls. After incubation with 10 M iisoA2E for 5 days, ARPE-19 cells were vigorously washed with PBS to ensure that only intracellular iisoA2E remained and then were harvested for HPLC analysis. The cell lysate was extracted with MeOH and chloroform. For compound elution, an Atlantis dC18 (3 m, 4.6 ϫ 150 mm) reverse-phase column was used for the stationary phase, and a gradient of acetonitrile in water with 0.1% trifluoroacetic acid was set for the mobile phase: 85-100% acetonitrile, 0.8 ml/min, 15 min; 100% acetonitrile, 0.8 -1.2 ml/min, 15-20 min; 100% acetonitrile, 1.2 ml/min, 20 -40 min. Photodiode array detection was set at 430 nm.
Cell Viability Assay-Cytotoxicity was tested by a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) colorimetric assay (Roche Applied Science) that measures the ability of viable cells to degrade a yellow tetrazolium salt into purple formazan crystals. Briefly, 20 l of MTT reagent was added to 200 l of culture medium in each well. Following a 4-h incubation, 150 l of DMSO was added. After oscillating for 10 min, solutions were spectrophotometrically measured at 490 nm. Cell viability was expressed as a proportion of control optical density (OD).
Lactate Dehydrogenase Assay-LDH assay was performed by a LDH cytotoxicity assay kit (Biyuntian, Shanghai, China). Cytoplasmic LDH was measured in the media of RPE cultures grown to 80 -90% confluence in 96-well plates following a 72-h iisoA2E treatment and a specified interval of 24 h (LDH release period) after the addition of fresh serum-free media to iisoA2Econtaining cells. After the plate was centrifuged at 400 ϫ g for 5 min, cell-free supernatants were collected and assayed in a new 96-well plate at 120 l/well. Thirty minutes after reaction mixture addition, absorbance was measured at 490 nm. In each experiment, six replicates of each condition were included, and background levels were determined using medium that was not exposed to cells and subtracted from absorbance values obtained for each condition.
PLD-mediated Hydrolysis of the A2PE Series-The A2PE series (550 g) was dissolved in DMSO (45 l) and then added to 465 l of 40 mM MOPS buffer (pH 6.5) containing 100 units/ml PLD and 15 mM CaCl 2 . The mixture was incubated for 4 h at 37°C, extracted with 1:1 (v/v) chloroform/MeOH and chloroform, dried under argon, and redissolved in 100 l of 50% methanolic chloroform. HPLC analysis was performed and monitored at 430 nm. The mobile phase contained a gradient of MeOH in water with 0.1% formic acid (75-90% MeOH (0 -30 min), 90 -100% MeOH (30 -40 min), and 100% MeOH (40 -100 min)) with a flow rate of 0.5 ml/min. To simultaneously detect the A2PE series and cleavage products released from its PLD-mediated hydrolysis, an XBridge TM C8 column (5 m, 4.6 mm ϫ 150 mm) was used with an injection volume of 50 l and a total running time of 100 min.
Statistical Analysis-Data were analyzed by one-way ANOVA and Newman-Keuls multiple comparison test (Prism 4, GraphPad Software, San Diego, CA).

Detection of Two Unidentified Lipofuscin Pigments in Mouse
Eyecups-Chloroform/MeOH extracts of posterior eyecups harvested from C57BL/6 and BALB/cByJ mice were reexamined by reverse-phase HPLC while monitoring eluents at 430 nm. As expected, HPLC profile peaks were readily assigned to A2E, isoA2E, and a previously unrecognized peak (URP) that exhibited absorbance maxima of 445 and 342 nm and a retention time (Rt) of 38.5 min (Fig. 2). In addition to A2E, isoA2E, and URP, an additional unknown peak adjacent to URP was also observed; this minor peak, different from URP, exhibited absorbance maxima at 430 and 352 nm and a Rt of 39.3 min (Fig.   2, A and B). To investigate the origin of these compounds in the visual cycle, the chromatogram generated utilizing eyecup extracts of C57BL/6 ( Fig. 2A) and BALB/cByJ (Fig. 2B) mice was compared with that of Rpe65 rd12 mice ( Fig. 2A, right inset), in which retinal chromophores were not formed in vivo (21). The injectants from the eyecups of Rpe65 rd12 mice were distinguished by an absence of bisretinoid components, such as A2E and isoA2E, that were present in C57BL/6 and BALB/cByJ mice. Additionally absent in the case of the Rpe65 rd12 mice were URP and the 430/352 nm peak at Rt ϳ39 min.
Detection of iisoA2E and URP in Human, Bovine, and Pig RPE-To test for tissue localization of iisoA2E and URP, fresh pig eyes and dissected RPE were obtained. In methanolic chloroform extracts of retinal pigment epithelia from a single pig eye (Fig. 2D), iisoA2E and URP were readily detected by reverse-phase HPLC. In addition, HPLC analysis of the extract from bovine retinal pigment epithelia also showed URP and iisoA2E, eluted at 39.5 min and 39.8 min, respectively (Fig. 2E).
To substantiate the consistency of iisoA2E in bovine RPE with synthesized standard, the HPLC chromatogram of bovine RPE extract was compared with that of the bovine RPE extract and exogenous iisoA2E mixture. As expected, the peak height/area of iisoA2E was significantly increased. Quantification by integrating peak areas demonstrated that the percentage of iisoA2E relative to URP in the extract/mixture increased by 63.7% (supplemental Fig. S1). More importantly, the human donor eye (age 55, female), dissected as retinal pigment epithelia with attached choroid to maximize the yield of RPE cells, revealed the presence of iisoA2E and URP (Fig. 2F).
HPLC Analysis of Biosynthetic Reaction Mixtures-Next we established whether URP and the 430/352 nm compound in mice were generated in a reaction mixture of all-trans-retinal and ethanolamine (Fig. 2C). Besides observing the peaks attributable to A2E and isoA2E, the elution profile revealed a peak with UV-visible absorbance maxima at 430 and 352 nm. ESI mass spectrometry (ESI-MS) analysis of the 430/352 nm pigment in positive ion mode disclosed an m/z peak at 592.5 (Fig.  2C, right inset) consistent with the m/z of A2E and isoA2E. This peak was identical to the 430/352 nm pigment in eyecup extracts of wide-type mice (Fig. 2, A and B) as judged by Rt, ESI-MS, and UV-visible absorbance. By contrast, the peak attributable to URP was not formed in the reaction mixture, as determined by reverse-phase HPLC (Fig. 2C).
Photoisomerization/Photo-oxidation of A2E, isoA2E, and 430/352 nm Pigments-To investigate the light-induced modification of A2E, isoA2E, or the 430/352 nm pigment, solutions of these fluorophores in water (200 M) containing 0.2% DMSO for solubility purposes were exposed to sunlight, room lamplight, and 430-nm blue light. Exposure of the A2E solution to either sunlight or room lamplight for 35-120 min yielded isoA2E and several minor cis-isomers, all with a molecular ion signal at m/z 592 in the mass spectra (supplemental Fig. S2), consistent with a previous report (20,22). Moreover, isoA2E was more hydrophobic in comparison with other cis-isomers of A2E, and products behind isoA2E were not monitored in the HPLC chromatograms of the mixture generated from light illumination of A2E. Importantly, subjecting the isoA2E solution to sunlight for 35 min yielded a less polar peak absorbing at 430 and 352 nm (Fig. 3, A and B). Given that Rt (ϳ12.5 min), UVvisible absorbance maxima, and m/z 592 of this peak were in accordance with that of the 430/352 nm pigment in extracts of mouse eyecups and biomimetic reaction mixtures, we confirmed that this 430/352 nm pigment was produced from lightmediated isomerization of isoA2E. Accordingly, we will refer to this molecule as iisoA2E; this nomenclature is selected to reflect a molecule that can form from facile isomerization of isoA2E. To explore whether light-induced isomerization of iisoA2E to yield isoA2E occurred, a solution of iisoA2E was illuminated with sunlight. As a result, HPLC analysis revealed five isomers of iisoA2E (I-V; Fig. 3, C and D) on the basis of UV-visible absorbance (Fig. 3E) and a molecular ion peak at m/z 592 by ESI-MS. All of these isomers were eluted in front of iisoA2E, indicative of higher polarity than the parent compound. The chromatographic peaks behind iisoA2E were not observed in the HPLC profiles (Fig. 3D) of the mixture generated from light illumination of iisoA2E, suggesting that irradiation of iisoA2E did not generate products less polar than the parent compound. It was noted that peak V eluted at 9.5 min and showed a weak shoulder at 295 nm in the absorbance spectrum as well as characteristic absorbance maxima at 430 and 339 nm (Fig. 3E, blue), indicating that this product was slightly altered by photo-oxidation (23). According to absorbance maxima and Rt, it was inferred that isoA2E, different from products I-V, was not generated from photoisomerization of iisoA2E. To further corroborate this conclusion, a solution of iisoA2E was illuminated with sunlight for 35 min and injected into HPLC with synthesized isoA2E. As expected from the overlay chromatograms (Fig. 3F), a clear peak (Rt ϭ 38.6 min) was assigned to synthesized isoA2E and a shoulder peak (Rt ϭ 38.9 min) was readily detected that corresponded to product II, confirming that conversion of iisoA2E into isoA2E by light was not feasible. Likewise, these isomerizations were also observed when A2E, isoA2E, and iisoA2E were illuminated for 5 min by monochromatic light (430-nm blue light) (supplemental Fig.  S3).
Moreover, the propensity of iisoA2E to undergo photo-oxidation, as do A2E and isoA2E, was investigated. Samples in water with 0.2% DMSO were irradiated by 500-W xenon light (20,000 lux). In the case of A2E and isoA2E, subsequent ESI-MS analysis (Fig. 4) demonstrated the formation of several higher molecular mass adducts differing by 16 mass units; the structure of the highest m/z species at 736 was that of a polyoxygenated species corresponding to the addition of nine oxygen atoms to the polyenic side arms (Fig. 4, D and E). Conversely, subjecting the mixture generated from iisoA2E light illumination to ESI-MS resulted in the release of five peaks that differ in m/z by 16. The m/z 672 species indicated that only five oxygen atoms were inserted into the carbon-carbon double bonds  DECEMBER 13, 2013 • VOLUME 288 • NUMBER 50

JOURNAL OF BIOLOGICAL CHEMISTRY 35675
residing in the side chains of iisoA2E (Fig. 4F). Relative abundances of A2E and isoA2E as starting compounds in the mass spectra were significantly decreased compared with that of their photo-oxidized species, whereas the relative abundance of iisoA2E as the parent molecule was still far more greater than that of its photo-oxidative products. With chromatographic monitoring at 430 nm, peak heights of A2E, isoA2E, and iisoA2E were considerably diminished when these compounds were irradiated for 10 min with xenon light (Fig. 4, G-L), and quantitation by integrating peak areas reflected that levels of A2E, isoA2E, and iisoA2E were reduced by 99.65, 99.95, and 73.5%, respectively. These data support the conclusion that photo-oxidation of iisoA2E is much less efficient than that of A2E and isoA2E under identical experimental conditions.
Identification of iisoA2E-iisoA2E, a reddish solid, presented a molecular formula of C 42 H 58 NO ϩ as determined by high resolution ESI-MS at m/z 592.4497 [M] ϩ (theoretically, 592.4513) requiring 15 degrees of unsaturation. As with A2E (supplemental Fig. S4) and isoA2E (supplemental Fig. S5), it was not possible to obtain an interpretable 1 H NMR spectrum of this compound when isolated from eyes because each extract contained trace amounts of iisoA2E. Biomimetic synthesis of iisoA2E and subsequent two-step HPLC preparation allowed us to obtain sufficient purified samples and enabled NMR studies (supplemental Figs. S6 -S12) to identify its full structure (Figs. 1C and 5). The structure of iisoA2E was observed to be very special as the long side chain in this pigment was attached to the orthocarbon residing in a pyridinium ring versus that of A2E and isoA2E, which extended into the meta-carbon (Fig. 1). The 125-MHz 13 C NMR spectrum in CH 3 OD with the aid of spectroscopic techniques, including DEPT135 o , HSQC, and HMBC, resolved 42 carbon signals, including 9 methyls, 8 methylenes, 11 quaternary carbons, and 14 methines. All pro-ton and carbon chemical shifts present in iisoA2E 1 H and 13 C NMR spectra were carefully assigned ( Fig. 5; also see numbering in Fig. 1). 1  Cell-based Assays of iisoA2E-It was first confirmed that ARPE-19 cells did not contain endogenous iisoA2E by reversephase HPLC (supplemental Fig. S13). After incubation of ARPE-19 cells with iisoA2E, cellular uptake was verified by analysis of intracellular granules. In a representative HPLC chromatogram generated from the extract of RPE cells receiving 10 M iisoA2E, a clear peak attributable to iisoA2E was observed that was not present in the extract of cells unexposed to iisoA2E. In addition to iisoA2E, we did not detect additional products in iisoA2E-containing cells, indicating that further   Light-induced oxidization of A2E, isoA2E, and iisoA2E. A-C, ESI-MS spectra of A2E, isoA2E, and iisoA2E before irradiation by 500-W xenon light. D-F, ESI-MS spectral analysis of A2E, isoA2E, and iisoA2E after 10-min irradiation with xenon light. The addition of oxygen atoms is evidenced by a series of molecular ion peaks that differ in m/z by 16. The peak at m/z 736 in D and E indicates the formation of nonaoxo-A2E/isoA2E, a compound that is probably a complex mixture of stereoisomers. G-L, A2E, isoA2E, and iisoA2E (200 M in water with 2% DMSO) were irradiated for 10 min by a 500-W xenon light (20000 lux). Non-irradiated and irradiated samples were subsequently analyzed by reverse-phase HPLC. As highlighted in red, the percentage of irradiated sample relative to non-irradiated samples based on chromatographic peak areas (V⅐s) negatively correlated with the extent of photo-oxidation. mAU, milliabsorbance units. DECEMBER 13, 2013 • VOLUME 288 • NUMBER 50 degradation of this pigment did not occur. Online HPLC fluorescence detection demonstrated that iisoA2E exhibited clearly fluorescent signals with excitation/emission of 430/600 nm (Fig. 6A). Based on the autofluorescent nature of this molecule, cells with an accumulation of iisoA2E manifested autofluorescence viewed by laser-scanning fluorescence microscopy. Confocal imaging in the horizontal plane at depths within the cells (Fig. 6, B-F) confirmed that exogenously delivered iisoA2E was internalized by ARPE-19 cells in culture. Autofluorescent intensity of internalized iisoA2E granules was augmented, increasing in concentration from 5 to 40 M (Fig. 6, C-F).

Novel Lipofuscin Pigment iisoA2E in Retina
Next, the health of ARPE-19 cells with accumulated iisoA2E was examined. 1-3 days after introducing iisoA2E to cultures, cell viability was tested by MTT assay. Concomitant loss of cell viability was not observed with iisoA2E concentrations ranging from 0.625 to 20 M within 1 day (Fig. 6G). However, when the concentration of iisoA2E increased to 40 M, a significant loss of cell viability was detected. After 2 days, loss of cell viability was not detected at 0.625 M but was significantly observed at higher concentrations (1.25-40 M). After 3 days, a significant loss of cell viability was observed at 0.625 M. However, at 0.625 M and 1.25 M, loss of cell viability was not significantly associated with incubation time, whereas cell viability by MTT assay clearly decreased with time at concentrations starting from 2.5 M. To test the ability of iisoA2E amassed in RPE cells to perturb membrane integrity, evidence of membrane damage was assayed by investigating release of cytoplasmic LDH into the culture medium. It was found that RPE cells exposed to 0.625, 1.25, 2.5, 5, 10, and 20 M iisoA2E did not exhibit elevated LDH levels after 72 h of treatment, followed by 24-h incubation in fresh medium (Fig. 7). A significant increase in LDH levels was observed when cultures accumulated iisoA2E from the 40 M concentration.
Enzymatic Cleavage of the A2PE Series Releases A2E, isoA2E, and iisoA2E-A2PE was a compound mixture with phosphatidic acid moieties that vary in fatty acids in vivo (25). DP-PE is  a phospholipid (C 37 H 74 NO 8 P) component of the cell membrane. To exclude the formation of A2PE with variable fatty acid compositions, all-trans-retinal was reacted with DP-PE to yield an A2PE species (dipalmitic A2PE), which contained a dipalmitic acid. Interestingly, HPLC analysis of the reaction mixture-derived A2PE sample showed a cluster of chromatographic peaks rather than a single peak attributable to dipalmitic A2PE, suggesting that this material may be a compound mixture (an A2PE series) in which A2PE, isoA2PE, and iisoA2PE were included (Fig. 8A). Further evidence identifying the analyte of interest as a mixture of A2PE and its isomers was also provided by ESI-MS in positive ion mode (Fig. 8A, right  inset). The mass spectrum only exhibited prominent ion peaks at m/z 1223.3 and 1245.5, corresponding to dipalmitic A2PE (C 77 H 125 NO 8 P, calculated as 1222.914) and the Na ϩ adduct of the molecular ions of dipalmitic A2PE (C 77 H 124 NNaO 8 P, calculated as 1244.896). In experiments aimed at determining whether isoA2PE and iisoA2PE are formed in the reaction of all-trans-retinal with DP-PE and act as the substrates for PLD, we incubated the A2PE series with PLD and observed a reduction in all chromatographic peaks corresponding to the A2PE series together with the appearance of four more hydrophilic peaks in the HPLC profile at 21.97, 22.74, 24.13, and 24.83 min, respectively (Fig. 8B). On the basis of absorbance spectra and by co-injection with synthetic A2E, isoA2E, and iisoA2E, the first three peaks were readily identified as A2E, isoA2E, and iisoA2E, respectively (Fig. 8B, left inset). These observations corroborate the fact that besides A2PE, isoA2PE and iisoA2PE are generated in vivo as the immediate precursors of isoA2E and iisoA2E, with isoA2E and iisoA2E being released from isoA2PE and iisoA2PE by enzymatic hydrolysis (Fig. 8, C-E).

DISCUSSION
A growing body of clinical and laboratory evidence demonstrates that anomalous lipofuscin accumulation is implicated in the pathogenesis of RPE cell degeneration in patients with agerelated macular degeneration, recessive Stargardt disease, and other forms of macular disease (26 -28). Here, we have isolated and structurally characterized iisoA2E, a new type of pyridinium bisretinoid of RPE lipofuscin. It was found that the C15 proton of A2E and isoA2E is substituted with a long side chain, whereas the long hydrophobic retinal arm of iisoA2E is attached to C16 rather than C15 (Fig. 1). In the case of iisoA2E, the extension of two polyenic side arms into the adjacent carbon atoms of the pyridinium ring narrows the distance between long and short side chains and probably strengthens steric hindrance of active carbon-carbon double bonds, thereby alleviating the extent of oxidation arising from light (29,30). Thus, we presume that the ability of iisoA2E to resist photo-oxidation may be partly attributed to its unusual structure (Fig. 1C).
Evidence indicates that two side arms of A2E have nine carbon-carbon double bonds that are strongly susceptible to light. Both long and short polyenic side chains beside the pyridinium   (15-30 min) overlaid with that of synthesized iisoA2E (blue). C-E, PLD cleaves the phosphodiester bond in A2PE, isoA2PE, and iisoA2PE to release A2E, isoA2E, and iisoA2E, respectively. mAU, milliabsorbance units.
ring in A2E undergo a series of chemical processes, including cis-isomerization, photo-oxidation, and photo-cleavage during irradiation with blue light to produce cis-isomers, epoxides, furanoid oxides, cyclic peroxides, and cleavage products (23,(31)(32)(33)(34). Several lines of investigation have established that there is a mutual conversion between A2E and isoA2E when each pigment is illuminated with light (20,35), which is in accord with our findings by HPLC (Fig. 3, A and B). Thus, lightinduced isomers of isoA2E are very similar, but not identical, to that of A2E. It is especially interesting that isoA2E undergoes light-mediated isomerization to yield iisoA2E. Irradiation of iisoA2E with light yields several isomers in which A2E and isoA2E do not exist, indicating that the unprecedented structure of iisoA2E (Fig. 1C) does not allow it to be converted into either A2E or isoA2E. The compounds attributable to peaks I-V in the HPLC chromatograms (Fig. 3D), different from light-induced isomers of A2E and isoA2E, represent cis/transisomers of iisoA2E at carbon-carbon double bonds. In addition to isoA2E, a Z-isomer of A2E at the C13C14 double bond, several additional cis double bond isomers of A2E are clearly formed in human RPE (22), whereas the structures of the minor cis-isomers of A2E have so far not been corroborated due to the inability to obtain sufficient samples for NMR studies (22). It has been confirmed that the C13C14 double bond of all-transretinal (Fig. 1D) is most prone to isomerization (36,37), which suggests that 13-cis-retinal may coexist with all-trans-retinal in vivo, albeit at low levels. Accordingly, isoA2E is the most prominent isomer of A2E when it is generated in vivo (Fig. 2), in the reaction mixture (Fig. 2C), and by A2E light exposure.
Wu et al. (19) have cultivated a scheme for A2E/A2-DHP-PE biosynthesis. An initial reaction between phosphatidylethanolamine and all-trans-retinal generates N-retinyl-phosphatidylethanolamine (NRPE), a Schiff base conjugate that is the ligand for the photoreceptor-specific ATR-binding cassette transporter ABCA4 that is mutated in recessive Stargardt macular degeneration (10, 38 -41). It is postulated that NRPE FIGURE 9. Proposed iisoA2E formation cascade. A, the proposed biosynthetic pathway by which iisoA2E, A2E, isoA2E, and A2-DHP-PE form in the retina. All-trans-retinal released from activated rhodopsin after photoisomerization of ground state 11-cis-retinal reacts with phosphatidylethanolamine (PE) in the disk membrane to produce the NRPE Schiff base that undergoes a [1,6]-proton tautomerization to generate PAE. After reaction with a second molecule of all-trans-retinal and 6-aza-electrocyclization, dihydropyridinium A2PE is generated. This intermediate readily undergoes a 1,3-H shift and hydrogen atom elimination to give A2-DHP-PE or can eliminate two hydrogens to form A2PE. Hydrolysis of the A2PE phosphate ester yields A2E. Because the C13C14 double bond of all-trans-retinal is most prone to cis-isomerization, yielding 13-cis-retinal, this chromophore is considered to coexist with all-trans-retinal in vivo at low levels. All-trans-retinal reacts with PE to generate NRPE. After NRPE undergoes a [1,6]-proton tautomerization to PAE, the intermediate reacts with 13-cis-retinal (not all-trans-retinal). Following 6-aza-electrocyclization, dihydropyridinium isoA2PE forms and undergoes facile auto-oxidation to isoA2PE. Ester hydrolysis removes phosphatidic acid from isoA2PE to produce isoA2E. As an alternative, after PAE reacts with 13-cis-retinal to form a 13-cis-iminium salt, the latter undergoes a hydrogen shift, C13C14 double bond cleavage, and 6-aza-electrocyclization to generate an intermediate bearing a three-membered ring that readily opens. Following opening of the three-membered ring at the C14C15 single bond and shift of C13-H to C15, dihydropyridinium iisoA2PE is generated. Facile auto-oxidation of this intermediate with the loss of two hydrogen atoms gives rise to iisoA2PE, which is cleaved by hydrolysis of the phospholipid to release iisoA2E. A2E and isoA2E can be interconverted by light. B, proposed mechanism by which isoA2E is converted into iisoA2E by light. isoA2E undergoes C13C14 double bond cleavage and a shift of 6-H to C13 to produce an intermediate with a three-membered ring. Photochemical ring opening and intramolecular hydrogen shift occur in this intermediate, leading to formation of iisoA2E. PLD, phospholipase D.
undergoes a [1,6]-proton tautomerization generating a phosphatidyl analog of enamine (PAE). Following reaction with a second molecule of all-trans-retinal, an all-trans-iminium salt is suggested to form. Now with the isolation of iisoA2E, a compound that we suggest could form from a reaction of PAE with one molecule of 13-cis-retinal, rather than all-trans-retinal, with proton transfer/elimination, single/double bond cleavage, and minimal electronic reorganization (Fig. 9). The condensation reaction of PAE and 13-cis-retinal generates a 13-cis-iminium salt, which would go through a 6-H to C13 shift, cleavage of the C13C14 double bond, and 6 aza-electrocyclization, thereby leading to formation of an intermediate bearing a three-membered ring. After opening of the three-membered ring (42,43) at the C14C15 single bond and shift of C13-H to C15, dihydropyridinium iisoA2PE forms, characteristic of the attachment of the long side arm to C16. Subsequent auto-oxidation eliminates two hydrogen atoms from this intermediate to give iisoA2PE, which hydrolyzes to release iisoA2E. As an alternative, 13-cis-iminium salt undergoes 6 aza-electrocyclization to yield dihydropyridinium isoA2PE. After auto-oxidation and a two-hydrogen loss, isoA2PE is generated and subsequently produces isoA2E by hydrolysis of the isoA2PE phosphodiester bond.
We have demonstrated that PLD-mediated cleavage of the A2PE series generated A2E, isoA2E, and iisoA2E, thereby confirming the presence of isoA2PE and iisoA2PE in the DP-PE/ all-trans-retinal reaction mixture. The data also manifest that iisoA2PE may serve as the precursor of iisoA2E, together with isoA2PE as a precursor of isoA2E in the biosynthetic pathway, and corroborate, at least in part, our proposed biogenic scheme (Fig. 9A). For light-mediated mechanisms by which isoA2E is converted into iisoA2E (Fig. 9B), we proposed that isoA2E undergoes a 6-H to C13 shift and cleavage of the C13C14 double bond to generate a tautomer that contains a three-membered ring. As is known, three-membered rings readily open and participate in nucleophilic addition reactions (42,43); ring opening of the three-membered ring at the C14C15 single bond and shift of C13-H to C15 yield iisoA2E.
Next, we tested the stability of iisoA2E in polar solvents (see supplemental Methods and Results and Fig. S14). By HPLC, we elucidated that iisoA2E was safe in DMSO and MeOH, whereas EtOH and chloroform alter this pigment. Interestingly, iisoA2E altered by either EtOH or chloroform could be recovered by MeOH within 30 s (supplemental Fig. S15). Because biomimetic synthesis of iisoA2E ran in EtOH, we believe that a part of iisoA2E previously formed in the condensation reaction was probably altered. Indeed, the iisoA2E synthetic yield increased by 9.5% after using MeOH to recover altered iisoA2E. This study also provides insight into the proper usage of iisoA2E and facilitates the establishment of a correct process for its measure in the eyes.
We also detected an additional unrecognized RPE lipofuscin component, URP. As a neighbor of iisoA2E in the HPLC chromatograms (Fig. 2), URP appears more abundant than iisoA2E in human RPE and eyecups of C57BL/6 and BALB/cByJ mice. It should be noted that a peak attributable to URP was not detected in all mouse eyes and bovine RPEs. We have corroborated that URP is not a light-induced isomer of iisoA2E based on absorbance spectra and HPLC Rt values. Like other bisretinoid compounds, URP has two absorbance maxima ( max 342 and 443 nm) in its UV-visible spectrum, whereas the 342 nm absorbance intensity appears very weak. At this time, the structure of URP is not likely to be identified given that in vitro synthesis of this compound is still not established to obtain sufficient samples for NMR studies. In addition, we HPLCquantified the levels of iisoA2E in each type of eye: C57BL/6 mice (0.52 pmol/eye, average content, four samples, 8 eyecups/ sample); BALB/cByJ mice (0.33 pmol/eye, average content, four samples, 8 eyecups/sample); pig (20.72 pmol/eye, average content, two samples, 1 RPE/sample); bovine (35.44 pmol/eye, average content, two samples, 1 RPE/sample); female human (45.25 pmol/eye, age 55, one sample, 1 RPE/sample); and male human (80.82 pmol/eye, age 49, one sample, 1 RPE/sample). With cell-based assays, we have confirmed that excessive accumulation of iisoA2E in RPE cells precipitates cell death (Fig. 6) and causes membrane damage (Fig. 7). However, the levels at which iisoA2E is presented in these experiments are much higher than what is expected in vivo. Based on the obvious discrepancies, iisoA2E may not play an important role in RPE cell damage in vivo, but photo-oxidation of this adduct is likely to cause increased cytotoxic activity (44 -46). As a vitamin A aldehyde-derived compound, iisoA2E serves as a fluorescent biomarker of aberrant all-trans-retinal metabolism (47), and its characterization will give a more complete understanding of the biosynthesis of this class of compounds in the retina. Careful identification of RPE lipofuscin components is fundamentally important because this knowledge increases awareness of the total burden placed on RPE cells by the deposition of this material. Improved understanding of the biosynthetic pathways of RPE lipofuscin pigments could open avenues toward additional therapies based on limiting the formation of these adverse compounds.