Novel Lipofuscin Bisretinoids Prominent in Human Retina and in a Model of Recessive Stargardt Disease*

Bisretinoid adducts accumulate as lipofuscin in retinal pigment epithelial (RPE) cells of the eye and are implicated in the pathology of inherited and age-related macular degeneration. Characterization of the bisretinoids A2E and the all-trans-retinal dimer series has shown that these pigments form from reactions in photoreceptor cell outer segments that involve all-trans-retinal, the product of photoisomerization of the visual chromophore 11-cis-retinal. Here we have identified two related but previously unknown RPE lipofuscin compounds. By high performance liquid chromatography-elec tro spray ionization-tandem mass spectrometry, we determined that the first of these compounds is a phosphatidyl-dihydropyridine bisretinoid; to indicate this structure and its formation from two vitamin A-aldehyde (A2), we will refer to it as A2-dihydropyridine-phosphatidyleth a nol amine (A2-DHP-PE). The second pigment, A2-dihydropyridine-eth a nol amine, forms from phosphate hydrolysis of A2-DHP-PE. The structure of A2-DHP-PE was corroborated by Fourier transform infrared spectroscopy, and density functional theory confirmed the presence of a dihydropyridine ring. This lipofuscin pigment is a fluorescent compound with absorbance maxima at ∼490 and 330 nm, and it was identified in human, mouse, and bovine eyes. We found that A2-DHP-PE forms in reaction mixtures of all-trans-retinal and phosphatidyleth a nol amine, and in mouse eyecups we observed an age-related accumulation. As compared with wild-type mice, A2-DHP-PE is more abundant in mice with a null mutation in Abca4 (ATP-binding cassette transporter 4), the gene causative for recessive Stargardt macular degeneration. Efforts to clarify the composition of RPE lipofuscin are important because these compounds are targets of gene-based and drug therapies that aim to alleviate ABCA4-related retinal disease.

Throughout the life of an individual, retinal pigment epithelial (RPE) 2 cells of the eye accumulate bisretinoid adducts that comprise the lipofuscin of these cells (1)(2)(3). The compounds form as a byproduct of light-mediated isomerization of the visual chromophore 11-cis-retinal. Accordingly, conditions that reduce the production of all-trans-retinal (atRAL) from 11-cisretinal photoisomerization, such as reduced serum vitamin A (4 -6), variants, or mutations in the visual cycle protein RPE65 (7)(8)(9) and inhibitors of RPE65 and 11-cis retinol dehydrogenase (10 -13), substantially reduce the formation of this material.
Up to the present time, at least 17 constituents of RPE lipofuscin have been identified chromatographically and characterized structurally; added to these are biosynthetic intermediates such as N-retinylidene-phosphatidylethanolamine (NRPE), A2PE and dihydropyridinium-A2PE (see Fig. 1, A and B) (14 -19). The first RPE lipofuscin constituent to be described was A2E (see Fig. 1A, inset). The pyridinium bisretinoid (14 -16, 20, 21) structure of A2E (C 42 H 58 NO; molecular weight, 592) was confirmed by extensive nuclear magnetic resonance studies (14) and by total synthesis (22). A2E formation begins in photoreceptor outer segments when atRAL, instead of being reduced to all-trans-retinol, reacts with phosphatidylethanolamine (PE) in a 2:1 ratio. Although the double bonds along the side arms of A2E are all in the trans (E) position, Z-isomers of A2E have double bonds at the C-13/14 (isoA2E), C-9/9Ј-10/10Ј, and C-11/11Ј-12/12Ј positions, and all are detectable in human and mouse RPE (16). These pigments exhibit absorbances in both the UV and visible regions of the spectrum (A2E: max , 439 and 338 nm; iso-A2E: max , 428 and 337 nm).
Another bisretinoid compound of RPE lipofuscin also absorbs in the short wavelength region of the visible spectrum (17,18,23). This pigment, all-trans-retinal dimer (atRAL dimer; max , 432 and 290 nm) forms from the condensation of two atRAL and is present in RPE lipofuscin as Schiff base conjugates with either PE or ethanolamine (atRAL dimer-PE and atRAL dimer-E, respectively) or as unconjugated atRAL dimer. The pigments atRAL dimer-PE and atRAL dimer-E absorb in the visible range at about 510 nm, a "red" shift relative to atRAL dimer that is attributable to protonation of the Schiff base linkage. Although A2E is a pyridinium salt containing a quaternary amine nitrogen that does not deprotonate or reprotonate (24), the protonation state of the Schiff base linkage in atRAL dimer-PE and atRAL dimer-E is pH-dependent (18).
Other known constituents of RPE lipofuscin are generated by photooxidation. By mass spectrometry, the photooxidation products of A2E and atRAL dimer present as a series of peaks differing by increments of mass 16 beginning with Mϩ 592 (A2E) or Mϩ 552 (atRAL dimer) (18,25). The moieties generated by the addition of oxygens at CϭC bonds of these bisretinoid compounds include endoperoxides, furanoid oxides, and epoxides (25)(26)(27). These oxidized products are more polar than the parent compound, and mono-and bis-oxidized forms of A2E and atRAL dimer have been detected in RPE from human eyes and in eyecups from mice with null mutations in Abca4 Ϫ/Ϫ (18,25), the gene responsible for recessive Stargardt macular degeneration. It is also notable that unconjugated atRAL dimer is a more efficient generator of singlet oxygen than is A2E and is also a more efficient quencher of singlet oxygen (18).
Insight into the composition of RPE lipofuscin and the biosynthetic pathways by which these compounds form aids in an understanding of retinal diseases characterized by lipofuscin overload, particularly those associated with mutations in ABCA4 (ATP-binding cassette transporter 4) of photoreceptor cells (1)(2)(3). We report that a previously unrecognized bisretinoid molecule absorbing with maxima at 490 and 331 nm is detected at elevated levels in Abca4 Ϫ/Ϫ mice, a model of recessive Stargardt macular degeneration. This compound is also present in human RPE. By high performance liquid chromatography-electrospray ionization-tandem mass spectrometry (HPLC-ESI-MS/MS), with corroboration by Fourier transform infrared spectroscopy (FTIR), we determined that this molecule is a bisretinoid presenting with a noncharged dihydropyridine core (Fig. 1C). We propose a biosynthetic pathway by which this pigment may form and demonstrate that enzymatic removal of the phosphatidic acid portion of the molecule generates a second novel component of RPE lipofuscin.
Animals and Tissues-Albino Abca4/Abcr null mutant mice homozygous for Rpe65-Leu 450 were generated and genotyped for the Abcr null mutation and Rpe65-L450M variant by PCR amplification of tail DNA as previously reported (8). For Rpe65 genotyping, digestion of the 545-bp PCR product with MwoI restriction enzyme (New England Biolabs) yielded an undigested 545-bp band in the case of Met 450 , 180-and 365-bp fragments when the sequence corresponded to Leu 450 , and all three bands in heterozygous mice. Rpe65 rd12 , Balb/cByJ, and C57BL/6J were obtained from the Jackson Laboratory (Bar Harbor, ME). Human donor eyes were received within 12 h of death from the Eye-Bank for Sight Restoration (New York, NY). Fresh bovine eyes were obtained from Sierra for Medical Science (Whittier, CA).
Tissue Extraction and HPLC Analysis-Murine posterior eye cups (3-12 eyes/sample as indicated), human RPE/choroid (1 eye/sample), bovine RPE (1 eye/sample), and bovine neural retina (16 retina/sample; retinae washed twice after removal from the eyes) were analyzed. The tissues were homogenized in Dulbecco's phosphate-buffered saline using a tissue grinder in the presence of chloroform/methanol (1:1). Subsequently, the sample was extracted three times with addition of chloroform and centrifuged at 1,000 ϫ g for 5 min. After passage through a reversed phase (C18 Sep-Pak, Millipore) cartridge with 0.1% trifluoroacetic acid in methanol, the extract was concentrated by evaporation of solvent under gas and was redissolved in 50% methanolic chloroform (2 eyes/30 ml of solvent). An Alliance system (Waters Corp., Milford, MA) equipped with 2695 Separation Module, 2996 Photodiode Array Detector, 2475 Multi Fluorescence Detector was used for HPLC analysis. For compound elution, an Atlantis dC18 (3 m, 4.6 ϫ 150 mm; Waters) reverse phase column was used for the stationary phase and for the mobile phase a gradient of acetonitrile in water with 0.1% trifluoroacetic acid: 75-90% acetonitrile (0 -30 min); 90 -100% acetonitrile (30 -40 min); 100% acetonitrile (40 -80 min) with a flow rate of 0.5 ml/min. Detection by photodiode array was set at 430 and 490 nm. Peak area (V⅐s) was determined using Empower software. Molar quantity per eye was calculated using calibration curves constructed from known concentrations of synthesized standards and a molar extinction coefficient of 26,100 and by normalizing to the ratio of HPLC injection volume versus sample volume.
Biomimetic Reaction and HPLC Analysis-A mixture of egg-PE (7.28 mg, 1 equivalent) in 3 ml of chloroform and atRAL (6 mg, 2 equivalents) in 1 ml of methanol with formic acid (5 ml) was stirred at room temperature in a capped flask in the dark for 7 days. For HPLC analysis, the reaction mixture was concentrated in a rotavapor and injected into an Atlantis dC18 (3 m, 4.6 ϫ 150 mm; Waters) reverse phase column. HPLC analysis was performed with monitoring at 490 nm. The mobile phase was a gradient of acetonitrile in water with 0.1% trifluoroacetic acid: 75-90% (0 -30 min) acetonitrile; 90 -100% acetonitrile (30 -40 min); 100% acetonitrile (40 -80 min) with a flow rate of 0.5 ml/min. For elution from a Delta Pak C4 column (5 m, 3.9 ϫ 150 mm; Waters), the following gradient of acetonitrile in water with 0.1% formic acid was utilized: 0 -5 min; 75% acetonitrile, flow rate, 0.8 ml/min; 5-20 min, 75-100% acetonitrile; flow rate, 0.8 -1.2 ml/min; 20 -30 min, 100% acetonitrile, flow rate, 1.2 ml/min.  PA) and 15 mM CaCl 2 . The mixtures were incubated for 2, 3, and 6 h at 37°C and then extracted with 2:1 (v/v) chloroform/methanol containing 0.1% trifluoroacetic acid, dried under argon, and redissolved in 60 l of 50% methanolic chloroform. A2PE and A2E were detected by HPLC using the C4 column, with an injection volume of 30 l, 430 nm monitoring, the gradients of acetonitrile/water/trifluoroacetic acid described above, and a total running time of 50 min. To detect the cleavage product of PLD-mediated hydrolysis of A2-DHP-PE, the dC18 column was used with monitoring at 490 nm. The gradient of acetonitrile/water/trifluoroacetic acid was the same as described above, and the total running time was 100 min.
HPLC-ESI-MS/MS-HPLC-MS was performed using a Bruker Esquire TM 3000 ion trap mass spectrometer that was coupled on-line to an Agilent 1100 series HPLC. The mass spectrometer was equipped with ESI interface and ion trap analyzer operating in full scan mode from m/z 50 -2000. The same dC18 column and mobile phase was used as discussed above for HPLC. Mass spectral detection was achieved using atmospheric pressure electrospray ionization in positive and negative mode. Helium was employed as the nebulizer gas with heating gas flow rate at 8 liters/min and pressure at 25 p.s.i. The ion source temperature was 350°C. The capillary exit was set at 166.9 V with the skim voltage at 71.1 V. MS detection was carried out in alternating positive and negative ion modes or in MS2 mode. EsquireControl software permitted data acquisition and instrument control.
Spectroscopic Analysis-FTIR spectra were obtained using a Matson RS-1 series spectrometer with a deuterated triglycine sulfate detector. The spectrometer and the sample compartment were continuously purged with Praxair purge gas generator to remove CO 2 and moisture. The sample in methanol was deposited on a CaF 2 window, and methanol was rapidly evaporated, which created a dry film of the sample. This CaF 2 window was mounted on the sample compartment, and an identical clean CaF 2 window was placed on the reference compartment of a "shuttle" cell. The shuttle cell takes blocks of data alternately between the sample and reference compartments and then cycles, eliminating artifactual signal caused by temporal variations in CO 2 and moisture. All of the spectra were obtained at room temperature (ϳ22°C). The first set of spectra was obtained in about 1 h. Thereafter, three more sets were obtained. Because all the sets gave identical results, we concluded that the sample was stable throughout the experiment and added and averaged the spectra.
Gaussian-based Density Functional Theory: Geometrical Optimization of Two Model Compounds-The IR spectra of two model compounds having dihydropyridine (I) and dihydropyridinium (II) cores corresponding to A2-DHP-PE and dihydropyridinium-A2PE, respectively, were computed by density functional theory using the Gaussian 03 program (4). A B3LYP functional was employed along with the basis set 6-31Gϩ(d,p) to optimize the structures and calculate IR vibrational frequencies. At this level of density functional theory, and for the basis set used, a scaling factor of 0.9806 is suggested to compare the calculated spectra to those experimentally observed. Because our primary interest in this calculation is to differentiate between the pyridine and pyridinium cores, only the calculated C-N modes are discussed, and the scaled frequencies are quoted in the text.

A Pigment with Absorbance Maxima at 490 and 331 nm Is
Detectable in Abca4 Ϫ/Ϫ Mouse Eyecups-The Abca4 null mutant mouse serves as a model of autosomal recessive Stargardt disease and has been shown to accumulate the bisretinoid compounds of RPE lipofuscin in abundance (8,10,17,18,28,29). Accordingly, we analyzed chloroform/methanol extracts of posterior eyecups of Abca4 Ϫ/Ϫ mice (age, 5 months) by reverse phase HPLC while monitoring at 490 nm. As expected, peaks in the HPLC profile could be readily assigned to previously identified lipofuscin pigments including A2E, isoA2E, and atRAL dimer-PE (8,14,17,18). With an analytical running time of 80 min, we also readily detected a previously unidentified peak that exhibited absorbance maxima of 490 and 331 nm and a retention time (R t ) of 46.2 min; this peak eluted immediately in front of atRAL dimer-PE ( Fig. 2A). Online fluorescence monitoring showed that when considered relative to the corresponding absorbance and with emission monitored at 630 nm and excitation at 430 nm, the fluorescence exhibited by the 490/331 nm peak of interest was of greater intensity than A2E, isoA2E, and atRAL dimer-PE (Fig. 2B). To test the origin of this compound as a vitamin A aldehyde adduct, we compared the chromatographic profile generated using eyecup extracts from Abca4 Ϫ/Ϫ mice ( Fig. 2A) to that from Rpe65 rd12 mice (age 2 months) (Fig. 2C) that do not generate 11-cis-retinal and atRAL chromophores (30). As shown in Fig. 2C, when the eluents were monitored for absorbance at 490 nm, the samples from the Rpe65 rd12 mice were distinguished by an absence of bisretinoid constituents such as A2E, isoA2E, and atRAL dimer-PE that are present in the Abca4 Ϫ/Ϫ mouse ( Fig. 2A). Also absent in the case of the Rpe65 rd12 mice was the 490/331 nm peak at R t of ϳ46 min.
HPLC Analysis of Biosynthetic Reaction Mixtures-Our next aim was to establish whether the compound with 490/331-nm absorbance in the mouse eyecups was generated in reaction mixtures of atRAL and PE (Fig. 2D). Egg-PE was used as the starting material because, having fatty acids of varying carbon lengths and degrees of saturation, it is more complex than, for instance, dipalmitoyl-L-␣-phosphatidylethanolamine. After incubating the reactants in a 1:2 ratio (egg-PE:all-trans-retinal) for 7 days, we analyzed the mixture by reverse phase HPLC using a dC18 column and gradient of acetonitrile and water (with 0.1% trifluoroacetic acid). Accordingly, the elution profile revealed a peak with UV-visible absorbance maxima at 490 and 333 nm; this peak also eluted with the same R t (46.0 min) as the 490/331-nm absorbance peak in the mouse eyecup extract ( Fig.  2A). Again, this peak eluted just before atRAL dimer-PE (R t , 47.5 min); unconjugated atRAL dimer exhibited a R t of 75.5 min. When analyzing the same reaction mixtures using a dC4 column (Fig. 2D, inset), atRAL, and NRPE, precursors of the A2PE/A2E biosynthetic pathway were readily detectable on the basis of UV-visible absorbance as was the product A2PE; the latter compound does not elute from a dC18 column.
Identification of A2-DHP-PE-Molecular mass and structural characterization of the 490/333 nm intermediate was carried out utilizing HPLC-ESI-MS/MS, a soft ionization method. With chromatographic separation on a dC18 column and ESI operating in negative ion mode, the 490/336-nm peak eluting with a R t of 46.0 min presented in the MS scan as a prominent chloride adduct ion [M ϩ Cl] Ϫ at m/z 1259.8, corresponding to the chlorinated molecular ion of a phosphatidyl-dihydropyridine bisretinoid compound (C 77 H 126 ClNO 8 P, calculated as 1258.891) (Fig. 3); we will refer to this compound as A2-dihydropyridine-phosphatidylethanolamine (A2-DHP-PE). The observed m/z 1259.8 was consistent with a phosphatidic acid moiety having dipalmitic acid as fatty acid constituent. With respect to the identification of the m/z 1259.8 peak as a dihy- dropyridine compound, it is important to note that were it a dihydropyridinium compound bearing a positive charge conferred by a quaternary amine nitrogen, signals would not have been observed in negative ion mode. Interestingly, the m/z 1259.8 peak was accompanied by a m/z ϩ16 peak (m/z 1275.8) attributable to mono-oxidized species. As we have previously shown for other bisretinoid lipofuscin compounds (18,25,26), oxidation of the compounds would have most likely occurred by oxygen insertion along the polyene chain. Also detected was a chloride adduct ion [M ϩ Cl] Ϫ at m/z 933.6 accompanied by an m/z ϩ16 peak; however, a peak with the same m/z value was not present when operating in positive ion mode ( Fig. 4A and see below), and at this time we do not know the identity of this peak.
Evidence for the identification of the analyte of interest as the phosphatidyl-dihydropyridine bisretinoid, A2-DHP-PE, was also provided by HPLC-ESI-MS in positive ion mode (Fig. 4A).
Here the mass spectrum exhibited prominent peaks at m/z 1246.6 and 1262.6, corresponding to the Na ϩ adduct of the molecular ions of A2-DHP-PE (C 77 H 126 NNaO 8 P, calculated as 1246.912), and mono-oxidized A2-DHP-PE (C 77 H 126 NNaO 9 P, calculated as 1262.907). ESI-MS/MS was subsequently performed on the parent Na ϩ adduct ion at m/z 1246.6 (Fig. 4B). Fragmentation of the parent ion by electron impact produced an intense peak at m/z 1245.7 that reflected the loss of 1 m/z unit from the parent ion (i.e. [A2-DHP-PE-1H ϩ Na] ϩ ) (C 77 H 125 NNaO 9 P, calculated as 1245.904) and that corresponded to the Na ϩ adduct of the fragment ion, A2PE [A2PE ϩ Na] ϩ . The production of the A2PE molecule as fragment ion was consistent with the loss of one proton from the dihydropyridine ring of A2-DHP-PE, resulting in aromatization of the ring and formation of A2PE. Two additional fragment ion peaks at m/z 1109.6 and 973.5 were indicative of the loss of one and two ␤-ionyl rings, respectively, from the parent ion at m/z 1246.6. The absence of an A2-DHP-PE attributable peak in the MS/MS spectrum upon MSinduced loss of one proton, together with the production of A2PE as fragment ion, confirmed that the compound of interest was not a dihydropyridinium conjugate but rather a dihydropyridine conjugate. In this way, aromatization of A2-DHP-PE under electron impact likely models the chemical transformation occurring with the action of dehydrogenating agents.
Corroboration of Structure by FTIR-To provide further insight into the structure of this fluorescent phosphatidyl-dihydropyridine bisretinoid conjugate, FTIR spectroscopy was performed on the isolate from the synthetic reaction mixture that exhibited 490/336-nm absorbance and R t of 46.0 min. The FTIR spectrum exhibited absorbance bands at 2926, 2856, 1682, 1458, 1362, 1203, and 1136 cm Ϫ1 (Fig. 5A). The two strong bands at 2926 and 2856 cm Ϫ1 and the peak at 1458 cm Ϫ1 reflect C-H and CϭC stretch vibrations, respectively; these peaks are attributable to C-H and CϭC bonds located within the conjugated systems that comprise the dihydropyridine ring and retinoid-derived arms of the molecule. Esters have characteristic strong absorptions that can account for the peak at 1682 cm Ϫ1 . Importantly, the bands at 1136, 1203, and 1362 cm Ϫ1 are characteristic of C-N stretch and thus could be assigned to the three C-N bonds in the dihydropyridine core of the compound (31). Conversely, the absence of a strong band at ϳ1600 cm Ϫ1 attributable to a CϭN stretch (32) was consistent with a dihydropyridine, not a dihydropyridinium ring, within the bisretinoid compound of interest. This conclusion was further corroborated by Gaussian-based density functional theory. Specifically, the IR frequencies calculated for the dihydropyridine model (Fig. 5B, panel I) using density functional theory gave frequencies of 1138, 1208, and 1401 cm Ϫ1 for the C-N stretch and 1583 and 1664 cm Ϫ1 for the CϭN stretch in the pyridinium model ( Fig. 5B, panel II). Therefore the experimentally observed frequencies of 1136, 1203, and 1362 cm Ϫ1 in the IR spectrum together with the absence of clear peaks around 1600 cm Ϫ1 unambiguously confirmed that the core of the molecule was not a pyridinium but rather a dihydropyridine.
Detection of A2-DHP-PE in Human and Bovine RPE-We have previously demonstrated that the bisretinoid compounds that form the lipofuscin of RPE cells undergo synthesis in photoreceptor outer segments of neural retina and are deposited in RPE secondarily (15,16). Thus to test for the tissue localization of A2-DHP-PE, we obtained bovine eyes and dissected RPE and neural retina separately. In extracts of RPE from a single bovine eye (Fig. 7A), A2-DHP-PE was readily observed. A2-DHP-PE was also detected in neural retina when the tissue extracts from 16 bovine eyes were pooled and concentrated (Fig. 7A, inset).
The differences in RPE and neural retina with respect to detection levels of A2-DHP-PE are consistent with the process of outer segment disc renewal that is known to involve disc shedding and RPE cell phagocytosis (33). As a result, the bisretinoids in photoreceptor cell outer segments are kept at lower levels, whereas accumulation occurs in RPE. The harvest from human donor eyes (age, 54 and 64 years) obtained as RPE with attached choroid to maximize the yield of RPE cells also revealed the presence of A2-DHP-PE (Fig. 7B).
To compare the levels of A2E and A2-DHP-PE, we HPLC-quantified both pigments in extracts of RPE harvested from the two human eyes and found the ratio of A2E:A2-DHP-PE to be 1.11 (54 year old) and 1.15 (64 year old) (A2E and isomers, 2601 and 2714 pmol/eye, respectively; A2-DHP-PE, 2350 and 2358 pmol/eye, respectively).
Enzymatic Cleavage of A2-DHP-PE Releases A2-DHP-E-We previously showed that atRAL dimer-PE and A2PE undergo enzymatic hydrolysis whereby PLD-mediated removal of the phosphatidyl groups generates atRAL dimer-E and A2E, respectively (15,16,34). In experiments aimed at determining whether A2-DHP-PE is also a substrate for PLD, we incubated A2-DHP-PE with PLD for 3 h at 37°C and observed a reduction in the A2-DHP-PE peak together with the appearance of a more polar peak at ϳ16.1 min ( max , 432 and 332 nm) (Fig. 8, A and B). This cleavage product will be referred to as A2-DHP-E (Fig. 8F). A2PE was also hydrolyzed to A2E as expected (Fig. 8, D and E). In this in vitro assay, PLD-mediated cleavage of A2-DHP-PE was less efficient than PLD-mediated hydrolysis of A2PE. For instance, after 2 h of incubation the decrease in the A2PE chromatographic peak area was 72%, whereas A2-DHP-PE was diminished by 21%. After 3 h the reduction was 84% for A2PE and 42% for A2-DHP-PE and after 6 h 97 and 56%, respectively (percent reductions calculated from averages determined in two experiments). Significantly, on the basis of UV-visible absorbance ( max , 432 and 332 nm) and an R t (16 min) that corresponded to the R t of the cleavage product in the PLD/A2-DHP-PE mixture, we identified A2-DHP-E in the HPLC elution profile generated with some eye extracts from Abca4 Ϫ/Ϫ mice (Fig. 8C) and C57BL/6J mice (data not shown). Like A2-DHP-PE, A2-DHP-E has not previously been identified in retinal extracts. It should be noted that a peak attributable to A2-DHP-E was not detected in all mice eyes; this peak was also not detected in RPE from single human and bovine eyes (Fig. 7). The reasons for this will be the subject of later study.

DISCUSSION
We have advanced schemes for the biosynthesis of both A2E and the atRAL dimer family of compounds (14,15,17,18). In the case of A2E, we proposed that an initial reaction between the membrane phospholipid PE and atRAL would generate NRPE, a Schiff base conjugate that is probably the ligand for ABCA4 (29,(35)(36)(37), the photoreceptor-specific ATP-binding cassette transporter that is mutated in recessive Stargardt macular degeneration (38). We postulated that NRPE would undergo a [1,6[-proton tautomerization generating the phosphatidyl analogue of enamine. After reaction with a second molecule of all-trans-retinal, we suggested that an iminium salt would form and following electrocyclization, a phosphatidyl dihydropyridinium molecule (dihydro-pyridinium A2PE). We envisioned that given the drive to aromaticity, the latter molecule would automatically eliminate two hydrogens to yield A2PE, a phosphatidyl pyridinium bisretinoid. Mass spectrometry was utilized to confirm the structure of NRPE and A2PE (15) with ESI-mass spectrometric analysis of the reaction mixture of dipalmitoyl-L-␣-phosphatidylethanolamine and all-trans-retinal, revealing prominent peaks at m/z 958.6 and 1222.9, corresponding to the protonated molecular ions of NRPE and A2PE, respectively. Moreover, fast atom bombardment tandem mass spectrometry with collision-induced dissociation mass spectrometric analysis revealed product ions m/z 551.4 and 408.2 for NRPE, whereas one product ion at m/z 672.8 represented the phosphoryl-A2E fragment of A2PE. Experiments demonstrating the release of A2E upon phospholipase D-mediated cleavage of the phosphatidylethanolamine-bisretinoid A2PE established A2PE as the immediate precursor of A2E (15,16). Now with the isolation of A2-DHP-PE, a compound that we suggest could form from dihydropyridinium-A2PE with proton transfer/elimination and minimal electronic reorganization, we report that there is an alternative pathway for formation of an RPE lipofuscin pigment (Fig. 9). Specifically, the thermodynamic drive for the intermediate dihydropyridinium-A2PE to undergo oxidation (39) can lead to a 1,3-hydrogen shift and one hydrogen elimination (40) to form the stable uncharged dihydropyridine compound, A2-DHP-PE or can lead to the elimination of two hydrogens and the generation of A2PE, the immediate precursor of A2E.
The aromatic ring exhibited by A2PE exhibits exceptional thermodynamic and chemical stability, and like A2E, A2PE is a quaternary pyridinium salt, the positive charge on the pyridinium nitrogen being neutralized by a counter ion, probably chloride. The stability of A2-DHP-PE is also attested to by its detection in mouse eyecups and in human and bovine RPE and by our data demonstrating that A2-DHP-PE accumulates with age. A future objective will be to determine conditions that favor either one-or two-hydrogen elimination from dihydropyridinium-A2PE and to establish whether A2-DHP-PE can undergo further transformation to form A2PE. In this regard it is of interest that both in wild-type mouse eyecups and in normal human RPE, the ratio of A2E to A2-DHP-PE was found to be ϳ1:1; in Abca4 Ϫ/Ϫ mice however, the content of A2E was greater than that of A2-DHP-PE. This finding could reflect either accelerated formation of A2E versus A2-DHP-PE in Abca4 Ϫ/Ϫ mice or greater loss of A2-DHP-PE such as could occur because of photooxidation. The extent to which conversion of A2-DHP-PE to A2-DHP-E accounts for this difference is also not known. These are areas of study that we will pursue. Several lines of investigation have established that the bisretinoid precursors that constitute RPE lipofuscin originate in photoreceptor outer segments. For example, the formation of [ 14 C 2 ]A2PE was measured in outer segments isolated from excised whole retinas submitted to [ 14 C 2 ]ethanolamine incorporation and irradiation to release endogenous all-trans-retinal. A2PE also formed in outer segments incubated with exogenous all-trans-retinal (15,16). In addition, bisretinoid precursors have been identified in outer segments isolated from Abca4 Ϫ/Ϫ mice (28), and mass spectrometric analysis verified that A2PE is at least one of the autofluorescent pigments that accumulates in the orange-colored degenerating photoreceptor outer segment debris in Royal College of Surgeons rats (1,16), a strain having an inability to phagocytose shed outer segment membrane. These bisretinoid pigments likely also account for the lipofuscin-like autofluorescence that can be visualized in the photoreceptor cell membrane in recessive Stargardt disease and in some forms of retinitis pigmentosa (41)(42)(43). Because of continuous replacement of outer segment discs, however, these compounds may not reach levels sufficient for  JULY 24, 2009 • VOLUME 284 • NUMBER 30 HPLC detection in healthy neural retina unless tissue samples are pooled and concentrated as we have done here.

Dihydropyridine Bisretinoids in RPE Lipofuscin
As is also the case for the other phosphatidyl-bisretinoids A2PE and atRAL dimer-PE, A2-DHP-PE serves as a substrate for PLD, the resulting phosphate hydrolysis releasing A2-DHP-E and presumably phosphatidic acid. Cleavage of A2-DHP-PE can take place in the eye, because we also detected the product A2-DHP-E in mouse eyecups, albeit at low levels. Moreover, the cleavage presumably occurs within RPE, because A2-DHP-E was not detected in concentrated extracts of 16 pooled bovine neural retina (data not shown), whereas A2-DHP-PE was. The precursor A2PE is also cleaved within RPE lysosomes (generating A2E) (15,16,34), and this cleavage appears to be quite facile because A2E is always a substantial peak in RPE extracts, and A2PE is present at relatively low levels. For instance in a sample of pooled RPE from four bovine eyes, A2E and A2PE levels were found to be 279.8 and 23.1 pmol/eye, respectively. 3 Conversely, the relative abundance of A2-DHP-PE in mouse eyecups and human and bovine RPE indicates that A2-DHP-PE is more refractory to cleavage. This is also true for atRAL dimer-PE (18). In future experiments we will investigate the reasons for these differences.
Several observations support the conclusion that the compound identified in this work has a noncharged dihydropyridine core (Fig. 1C). First, because signals were obtainable from this compound with ESI working in negative ion mode, it was apparent that the compound was not charged; the observed mass (m/z 1259.8) was consistent with a chloride adduct phosphatidyl-dihydropyridine bisretinoid. Second, the generation of A2PE upon electron impact-induced aromatization of the analyte of interest demonstrated that the compound had only one more hydrogen than A2PE. Moreover, the experimentally determined FTIR frequencies were consistent with C-N bonds in the core of the molecule rather than CϭN bonds. In contrast to the relative stability of dihydropyridines, it is well known that dihydropyridinium molecules such as dihydropyridinium-A2PE are disadvantaged thermodynamically, and to the best of our knowledge, compounds having dihydropyridinium moieties at their core have not been isolated.
Using time-dependent density functional theory, we previously predicted the UV-visible absorbance spectrum of dihydropyridinium-A2PE to have maxima at 494 and 344 nm; similar maxima ( max , 490 and 330 nm) were determined experimentally (19) (Fig. 1B). The 330-nm absorbance is attributable to the conjugation system in the molecule having five double bonds and a ␤-ionone ring, a structure that is the same as all-trans-retinol. Thus it is not surprising that the 330-nm absorbance of dihydropyridinium-A2PE is similar to the peak absorbance of all-trans-retinol ( max , 325 nm). Like dihydropyridinium-A2PE, A2-DHP-PE has two arms, and the absorbance maxima are also similar to dihydropyridinium-A2PE. Interestingly, however, the electronic transition assignments for the chromophores of A2-DHP-PE and dihydropyridinium-A2PE are reversed. As shown in Fig. 1, the conjugation system present within the long arm of A2-DHP-PE extends into the dihydropyridine ring, thereby allowing for a system with six double bonds. The short arm of A2-DHP-PE also extends into the dihydropyridine ring giving five conjugated double bonds. With this arrangement, it is clear that the 490-nm absorbance can be assigned to the long arm of A2-DHP-PE and the 333-nm absorbance to the short arm. Not so, however, for dihydropyridinium-A2PE; here the conjugation system present in the long arm terminates before the pyridinium ring giving five double bonds to which the 330-nm absorbance is attributed, whereas the short arm extends into the pyridinium ring providing for a conjugation system of six double bonds and allowing for the 490-nm absorbance. It is perhaps notable that A2E and A2-DHP-E have very similar absorbance maxima; yet their phosphatidyl-bisretinoid precursors (A2PE in the case of A2E and A2-DHP-PE in the case of A2-DHP-E) have absorbance maxima that are relatively far apart. Quantum mechanical calculations (19,44) could provide insight into the electronic configurations underlying these absorbance differences.
In the development of approaches to the treatment of macular degeneration, attention is being given to therapies that reduce the formation of RPE lipofuscin pigments (6, 10 -12). These strategies include gene therapies based on adeno-associated virus-mediated (45) and lentiviral vector-mediated (46) delivery of the wild-type gene and the administration of compounds that limit the visual cycle (6,10,12,13). A2E is the compound employed as therapeutic end point measure in these preclinical studies. A2E remains the best known and most characterized of the bisretinoid lipofuscin pigments in RPE. However, the identification of components in addition to A2E is fundamentally important, at least because this knowledge increases awareness of the total burden placed on the RPE cell by the deposition of this retinoid-derived material. These compounds also differ in their properties (2) and can be expected to impact on RPE cells in a variety of ways. Improved understanding of the biosynthetic pathways by which these bisretinoid pigments form could also facilitate the introduction of yet other novel approaches to therapy.