Long-chain and very long-chain polyunsaturated fatty acids in ocular aging and age-related macular degeneration.

Retinal long-chain PUFAs (LC-PUFAs, C12-C22) play important roles in normal human retinal function and visual development, and some epidemiological studies of LC-PUFA intake suggest a protective role against the incidence of advanced age-related macular degeneration (AMD). On the other hand, retinal very long-chain PUFAs (VLC-PUFAs, Cn>22) have received much less attention since their identification decades ago, due to their minor abundance and more difficult assays, but recent discoveries that defects in VLC-PUFA synthetic enzymes are associated with rare forms of inherited macular degenerations have refocused attention on their potential roles in retinal health and disease. We thus developed improved GC-MS methods to detect LC-PUFAs and VLC-PUFAs, and we then applied them to the study of their changes in ocular aging and AMD. With ocular aging, some VLC-PUFAs in retina and retinal pigment epithelium (RPE)/choroid peaked in middle age. Compared with age-matched normal donors, docosahexaenoic acid, adrenic acid, and some VLC-PUFAs in AMD retina and RPE/choroid were significantly decreased, whereas the ratio of n-6/n-3 PUFAs was significantly increased. All these findings suggest that deficiency of LC-PUFAs and VLC-PUFAs, and/or an imbalance of n-6/n-3 PUFAs, may be involved in AMD pathology.

cation to assure that no signifi cant macular pathology was present. Likewise, eyes from donors with a clinical history of AMD were inspected to confi rm that their macular characteristics correlated with clinical examinations recorded prior to death. After removal of the vitreous body, the whole retina and the RPE/ choroid layer were carefully separated with forceps. Wet sample weights were recorded for all collected tissues after blotting excess moisture. All samples were immediately blanketed with argon and stored at Ϫ 80°C until further analysis.
All analytical solutions such as methanol, hydrochloric acid, isopropanol, and n-hexane were GC-MS grade reagents and were purchased from Fisher Scientifi c (Pittsburgh, PA). All standards such as tridecanoic acid (C 13 ) and hentriacontanoic acid (C 31 ) were purchased from Sigma-Aldrich (St. Louis, MO). Silica gel glass-encased solid phase extraction cartridges (500 mg/6 ml) were purchased from Sorbent Technology (Atlanta, GA).

Sample preparation including total lipid extraction, fatty acid methyl ester formation, and sample cleaning with solid-phase extraction
Samples were removed from the Ϫ 80°C freezer and placed on ice. Tridecanoic acid (20 g) and hentriacontanoic acid (1 g) were added as the internal standards. Total lipid was extracted from whole retina or RPE/choroid tissue based on a previously described method ( 39 ). The samples were probe sonicated with 0.5 ml hexane:isopropanol (3:2) in a glass bottle. Then hexane:isopropanol (3:2) was added to a fi nal volume equivalent to 40 times the sample weight (i.e., 1 g in 40 ml of solvent mixture). The sample bottles sealed under argon with Tefl on-lined caps were bath sonicated for 20 min at room temperature. The extract was then centrifuged at 5,000 rpm for 5 min, and the upper layer was dried under vacuum.
The dried fi lm dissolved in 200 µl hexane and 2 ml 8% HCl/ MeOH was sealed under argon with Tefl on-lined caps and heated at 80°C for 4 h to form fatty acid methyl esters (FAMEs) ( 36 ), then cooled on ice. The cool reacted solution was extracted three times with 1 ml distilled water and 2 ml hexane. The hexane layers were combined and dried under vacuum.
Silica gel glass-encased solid-phase extraction cartridges were subsequently used to clean the FAME extracts. The cartridge was activated by 6 ml of hexane before loading samples. The crude FAME extract was dissolved in 200 µl of hexane and loaded onto the activated cartridge. Then 6 ml hexane was used to wash the cartridge, and the eluate was discarded. Then the FAMEs were eluted by 5 ml hexane:ether (8:2), and the eluate was evaporated to dryness under vacuum. The dry fi lm was dissolved in 200 µl of hexane and centrifuged for 3 min at 14,000 rpm to remove particles prior to GC-MS analysis. Then 1 µl of sample was injected into the GC-MS instrument for LC-PUFA analysis. The sample was dried under vacuum again and redissolved in 30 l of hexane. Then 5 µl samples were injected into the GC-MS instrument for VLC-PUFA analysis.
For LC-FA analyses, we used the following MS conditions (Method A): 1 µl from a 200 µl sample was injected into the GC-MS using a splitless mode; the septum purge was on; and the injector temperature was set at 200°C. The column temperature was programmed as follows: initial temperature 60°C; 5 degrees/ troretinogram (13)(14)(15)(16)(17)(18)(19)(20). Thus, it is clear that LC-PUFAs play an essential role in the development of vision during human early life ( 16,21 ). In later human life, some epidemiological studies of LCn-3PUFA intake on the prevalence of advanced age-related macular degeneration (AMD) suggest a protective relationship (21)(22)(23)(24)(25)(26)(27), and Hubbard et al. ( 28 ) showed a protective effect of DHA and EPA intake in autosomal dominant Stargardt disease. Their role in the prevention of inherited retinal degenerations such as retinitis pigmentosa is more ambiguous ( 29 ).
Although retinal LC-PUFAs have received extensive attention over the past decades, very long-chain PUFAs (VLC-PUFAs, C n>22 ) ( 30,31 ) are also found in mammalian tissues, but they occur in only a restricted number of organs such as retina, testes, thymus, and brain in a low amount and their saturated analogs are essential parts of the skin moisture barrier ( 32,33 ). Unlike DHA, AA, EPA, and their precursors, VLC-PUFAs are not present in a normal human diet ( 34 ), but they can be synthesized from precursors such as 22:4n-6 and 22:5n-3 via a biochemical pathway (supplementary Fig. I) featuring the enzymes of the elongation of very long-chain FAs family along with ␤ -oxidases and desaturases ( 31,(35)(36)(37). Defects in the elongation of the very long-chain FAs-4 ( ELOVL4 ) gene are associated with autosomal dominant Stargardt macular dystrophy in the heterozygous state and neonatal demise in transgenic mice due to loss of skin barrier function in the homozygous state. VLC-PUFAs are of particular interest, because they exhibit a unique hybrid structure combining a proximal end with typical saturated FA character and a distal end more characteristic of common PUFAs ( 35,36 ). They may play important roles in biological systems that cannot be performed by the more common saturated and unsaturated LC-FAs ( 35,36 ); however, their greater length and minor abundance make them unusually diffi cult to analyze ( 36 ), which means that VLC-PUFAs have often been overlooked since their discovery ( 33,38 ).
In this study, we optimized analytical methods to detect LC-PUFAs and VLC-PUFAs in human retina and retinal pigment epithelium ( RPE)/choroid. We applied these methods to explore possible retinal functions of LC-PUFAs and VLC-PUFAs, and we assessed how their levels change in ocular aging and AMD.

Materials and dissections
Human donor eyes were obtained from the Utah Lions Eye Bank within 26 h after death, and donor demographic data are shown in Table Tables 2 and 3 . lows: ion source temperature, 200°C; multiplier voltage, 1,182 V; solvent delay, 22 min, SIM mode with m/z 79, 108, and 150. Comparison was made by normalizing the each peak area with the internal standard and the retina or RPE/choroid sample weight. The formula to calculate the values in Tables 4 and 5 and Fig. 5 is as follows: In the fi rst formula, PA, RF, and W are the peak area of target peak, response factor, and sample weight with grams as the unit. In the second formula, PA a is the peak area of IS in the sample, and PA b is the peak area of fresh IS solution at the same concentration.

Statistical analysis
Quantitative data were expressed as mean ± SEM, and they were analyzed for statistical signifi cance using one-way ANOVA in an SPSS statistical package (SPSS Inc., Chicago, IL, wersion 12.0). The number of samples used in each group is presented in the fi gure legends. P < 0.05 was defi ned as the level of signifi cance. min to 170°C; 1 degree/min to 180°C; 2 degrees/min to 240°C; 4 degrees/min to 290°C; and a hold at 290°C for 5 min. Transfer line temperature was 290°C. Helium was used as the carrier gas at a fl ow rate of 1.0 ml/min. MS conditions were as follows: electron ionization (EI) mode; ion source temperature, 200°C; multiplier voltage, 1,182 V; solvent delay, 5 min. All data were obtained by collecting the full-scan mass spectra within the scan range of 50-650 amu. Compounds were identifi ed by comparing their mass spectra with those in the National Institute of Standards and Technology (NIST) library. Authentic reference compounds were used to calculate the weight percentage of every peak.
For VLC-PUFAs, we used the following MS conditions (Method B): 5 µl from 30 µl of sample was injected onto the GC-MS using a splitless mode; the septum purge was on; and the injector temperature was set at 200°C. The column temperature was programmed as follows: initial temperature, 60°C; 10 degrees/min to 240°C; 1 degree/min to 290°C; and 290°C for 5 min. Transfer line temperature was 290°C. Helium was used as the carrier gas at a fl ow rate of 1.5 ml/min. Both liquid chemical ionization (LCI) and EI modes were used to identify VLC-PUFAs, whereas only the EI mode was used for quantifi cation. The MS conditions for LCI were as follows: ion source temperature, 180°C; multiplier voltage, 1,182 V; solvent delay, 22 min; selected ion monitoring (SIM) mode with molecular weights plus one ion; acetonitrile fl ow rate, 0.1 µl/min. The MS conditions under EI were as fol-LC-FAs were identifi ed and quantifi ed by MS using the full-scan EI mode (Method A), because all information about LC-FAs in the NIST library was acquired in this manner. Unfortunately, EI is less valuable for identifi cation of VLC-PUFAs, because they all exhibit the same base peak of m/z 79 ( 36 ), no standards are commercially available, and they are not a part of the NIST library. Therefore, we used the LCI mode to identify VLC-PUFAs. Unlike the EI mode, the base peak of each VLC-PUFA obtained under the LCI mode corresponded to the molecular weight of the lipid (supplementary Fig. III). Furthermore, n-3 and VLCn-6PUFAs can be distinguished under EI mode based on the ratio of 108 and 150 m/z ( 41 ) (supplementary Fig. IV). We set the mass de tector to be off at the fi rst 22 min due to the very high concentration of the LC-FAs that come out before 22 min and saturate the detector. Quantifi cation of the VLC-PUFAs was performed using SIM of the EI mode (Method B).
In principle, the internal standard should have not only similar chemical structure and characteristics with the target compounds, but also it should be absent in the samples. So far, there is no ideal unsaturated internal standard for VLC-PUFA analysis on the market. Therefore, we chose hentriacontanoic acid, a saturated VLC-FA, as the best available internal standard for the measurement of VLC-PUFAs in method B.

RESULTS
The characteristics of the various donor eyes such as the time point of tissue collection, the ocular condition (AMD or normal), sex, age, and tissue weight are described in Table 1 . Retina and RPE/choroid samples were grouped into three age categories: young (age range 12-22 years, 16.4 ± 3.6), middle (age range 30-49 years, 38.2 ± 6.2), and old (age range 70-78 years, 74.0 ± 3.4) and age-matched AMD (age range 70-87 years, mean age ± SD, 77.2 ± 6.4).

Method development
Because these lipids, especially VLC-PUFAs, are diffi cult to analyze, we had to optimize the assay conditions. Although TLC is a commonly employed method for purifi cation of FAMEs formed after transesterifi cation ( 33,36,40 ), it is cumbersome, solvent consuming, and time intensive. To circumvent these disadvantages, TLC was replaced by solidphase extraction in our study to purify the samples following total lipid extraction and methyl esterifi cation. Cholesterol is the main endogenous substance that can interfere with the analysis of FAME. Therefore, we chose silica gel glassencased cartridges to remove it (supplementary Fig. II). Silica gel cartridges function in the same way as TLC but with higher recovery and speed, which helps to minimize lipid peroxidation that could occur during TLC analysis.
In RPE/choroid, the concentration of LA (peak 8) was signifi cantly higher in the old age group relative to the middle age group ( P < 0.05). Our results further showed that DHA and adrenic acid were signifi cantly different between the old age group and the age-matched AMD group. In particular, DHA and adrenic acid were higher in the old age group relative to the age-matched AMD group ( P < 0.05), whereas DHA peaked in the middle age group ( P < 0.05). The ratios of AA/DHA and n-6/n-3 in age-matched AMD retinas (6.41 ± 0.41 and 7.97 ± 0.51) were signifi cantly higher than the old age control group (3.90 ± 0.19 and 5.22

Separation of LC-FAs from retina and RPE/choroid
Full scan (50-650 amu) MS under EI mode was used to detect LC-FAs. The chromatogram is shown in Fig. 1 . Twenty-six LC-FAs ranging from 14:0 to 24:0 were identifi ed in retina and RPE/choroid after comparison with the NIST library and standards. Their retention times, structures, and common names are listed in Tables 2 and 3 . The retention time variation of individual peaks for different runs within 1 day was ± 0.1 min.
Oxidative damage is postulated to be involved in AMD ( 49 ). The multiplicity of double bonds in DHA renders it extremely sensitive to free radical damage during oxidative stress ( 50 ). Crabb et al. demonstrated that carboxyethyl pyrrole (CEP) protein adducts, the oxidative protein modifi cations generated uniquely from the docosahexaenoate-containing lipids such as DHA and adrenic acid ( 51 ), are more abundant in ocular tissues from AMD retina than from normal retina ( 52 ). CEP biomarkers have been shown to enhance the AMD predictive accuracy of genomic AMD biomarkers ( 53,54 ). The increase of CEP protein adducts in AMD retinas indicates that increased oxidation of DHA and adrenic acid that likely occurs in AMD eyes may lead to a reduction in DHA and adrenic acid, which is consistent with our study results. Dunaief et al. ( 55,56 ) reported that retinas and RPE/choroid from AMD donor eyes exhibit marked accumulation of iron stores compared with age-matched normal donor tissue, which may lead to increased Haber-Weiss reactions signifi cant. The ratios of n-6/n-3 VLC-PUFAs and the sum of all VLC-PUFAs in old age group retinas (0.83 ± 0.09 and 3203 ± 535) were signifi cantly decreased compared with the ratio and sum in age-matched AMD retinas (1.24 ± 0.10 and 1355 ± 268) ( P < 0.05).

VLC-PUFAs in RPE/choroid
The results of the RPE/choroid VLC-PUFA analysis are shown in Table 5 and Fig. 4B . C 24 -C 26 VLC-PUFAs were detectable in individual RPE/choroid samples, whereas C 30 -C 34 VLC-PUFAs were detectable only in pooled samples from the same age group. C 28 VLC-PUFAs were undetectable even in pooled samples, suggesting that their levels were quite low. The C 24 -C 26 VLC-PUFAs in RPE/ choroid were around one-third of that in retina, whereas C 30 -C 34 VLC-PUFAs were around one-tenth. Although there are no statistics provided in this portion of the study because samples had to be pooled, the trends of change of VLC-PUFAs in RPE/choroid with aging and AMD were similar to retina, peaking in middle age donors and severely decreased in AMD donors.

DISCUSSION
In this study, LC-FAs and VLC-PUFAs were systematically analyzed with regard to age and AMD status in human donor eyes with special emphasis on VLC-PUFAs, because low levels have recently been associated with ELOVL4 -related retinal eye disease (42)(43)(44). No age-dependent change in the relative concentrations of the 26 LC-FAs was observed in total When we analyzed RPE/choroid LC-PUFAs, we noted some data with more variability relative to retina, perhaps due to the presence of contaminating blood lipids. LA in RPE/choroid was signifi cantly higher in the old age group compared with the young and middle age groups, suggesting that LA might accumulate in RPE lipofuscin with age after being delivered to the RPE from the diet ( 61 ). Lower DHA in RPE/choroid than retina may be due to selective recycling of DHA into photoreceptor outer segment lipids ( 21,62,63 ). DHA and adrenic acid in RPE/choroid were higher in the normal old age group than in the AMD group, which was similar to fi ndings in the retina. Such similarity between retina and RPE/choroid is likely due to RPE phagocytosis of approximately 10% of the photoreceptor outer segment material daily, because each photoreceptor renews its outer segment every 10 days ( 61,64 ).
C 24 Fig. I). Thus, the decrease of C 24 and C 26 VLC-PUFAs in AMD retinas compared with age-matched normal retinas is consistent with the decrease of 22:4n-6 and 22:5n-3. Many C 28-C 34 VLC-PUFAs were signifi cantly decreased in AMD retinas relative to age-matched normal retinas. This may be due to the lower level of precursors such as 22:4n-6 and 22:5n-3 in AMD retina or impaired enzymatic processing. C 28-C 38 VLC-PUFAs can be converted from EFAs through the consecutive enzymatic activities of desaturases, elongases, and ␤ -oxidation enzymes in mammals ( 31,35,36 ) (supplementary Fig. I), but elongation requires ELOVL4, whereby n-6 and n-3 PUFAs undergo oxidative degradation, resulting in generation of the corresponding active aldehydes such as CEP. Thus, accumulation of iron stores in AMD eyes may lead to oxidative degradation of PUFAs and subsequent CEP generation, which is consistent with our study results and Crabb et al. (51)(52)(53).
Besides oxidative damage, infl ammation is also postulated to be involved in AMD ( 49 ). When infl ammation happens, cytokines and chemokines stimulate phospholipase A2 and cyclooxygenases, which results in a breakdown of membrane glycerophospholipids with the release of DHA. Then, DHA is metabolized to resolvins and neuroprotectins, which can prevent infl ammation by inhibiting transcription factor nuclear factor B , pro staglandins, leukotrienes, and thromboxanes, preventing cytokine secretion and modulating leukocyte traffi cking ( 57,58 ). Therefore, in our study, reduction of DHA in AMD retinas may also be the result of antiinfl ammatory processes.
The ratio of AA/DHA is increased in AMD ocular tissue compared with the age-matched normal group, suggesting an imbalance in pro-and anti-infl ammatory processes in AMD retina. In accordance with our fi ndings, a lower retinal n-6/n-3 PUFA ratio has a protective effect against pathological angiogenesis ( 59 ). All of these results suggest that oxidative stress and infl ammation play an important role in the etiology of AMD ( 60 ), whereas a reduction in DHA and adrenic acid and an increase in the ratios of AA/DHA and n-6/n-3 may be biomarkers of AMD. retinas, which can reduce VLC-PUFAs in AMD retinas compared with normal retinas. The increased ratio of n-6/n-3 VLC-PUFAs in AMD retinas relative to normal retinas may also be due to differences in the relative levels of their n-6 and n-3 precursors.
In RPE/choroid, C 24 and C 26 VLC-PUFAs were detectable in all individual samples, whereas C 30-34 VLC-PUFAs were at such low concentrations that they were detectable only in pooled samples, but they still exhibited which uniquely exists in retina, sperm, skin, testes, thymus, and brain ( 36 ) (supplementary Fig. I). Reduced levels of VLC-PUFAs are found in retinas of animal models of autosomal dominant Stargardt macular dystrophy, a macular dystrophy that results from truncating mutations of the ELOVL4 gene ( 42 ), and the Met299Val variant in the ELOVL4 gene has been reported to be possibly associated with AMD risk ( 66 ). Furthermore, just like DHA, more oxidization of VLC-PUFAs may also occur in AMD Fig. 4. Comparisons of C 24 -C 34 VLC-PUFAs in young age, middle age, and old age and age-matched AMD donors. A and B are the levels of C 24 -C 34 VLC-PUFAs in the whole retina and RPE/choroid, respectively, from young age, middle age, old age, and age-matched AMD donors. Statistical analysis was performed using SPSS statistical software. Statistical signifi cance was determined by one-way ANOVA. §, Signifi cant differences ( P < 0.05) between young and middle age group; ¤, means signifi cant differences ( P < 0.05) between middle and old age group; *, signifi cant differences ( P < 0.05) between old and age-matched AMD group. In B, there are no statistics provided, because each group's samples had to be pooled. IS: internal standard. current study utilized predominantly early and intermediate stage AMD eyes, and higher sensitivity GC-MS methods should permit study of regional variations of VLC-PUFAs in macular and peripheral regions of the retina. Our results support the potential value of interventions to increase retinal VLC-PUFAs and to decrease n-6/n-3 ratios in the prevention and treatment of AMD.
We thank Robert E. Anderson, PhD and Richard S. Brush, PhD for sharing insights on VLC-PUFAs analytical methods. trends with age similar to retina. The RPE must phagocytize approximately 10% of the photoreceptor outer segment material daily insofar as each photoreceptor renews its outer segment every 10 days ( 61,64 ), and each RPE cell underlies as many as 200 photoreceptors and may, therefore, ingest an equivalent of 20 entire outer segments per day ( 61,64 ). This suggests that RPE/choroid may share the same C 30-34 VLC-PUFAs with retina. This sharing between RPE/choroid and retina may explain our detection of C 30-34 VLC-PUFAs in RPE/ choroid at lower concentrations with distributions similar to overlying retina even though ELOVL4 is absent in RPE/choroid ( 36 ). Interestingly, C 28 -C 34 VLC-PUFAs were below the limit of detection even in combined AMD RPE/choroid samples, correlating well with our retinal fi ndings.
In summary, we have demonstrated that the human retina and RPE/choroid exhibit substantial alterations in PUFA levels with age and that these changes are magnifi ed in AMD donor eyes. These alterations are especially prominent in lipids of the VLC-PUFA series in AMD eyes where these lipids are severely reduced in the retina and undetectable in the RPE/choroid, and n-6/n-3 ratios are increased. The underlying causes for these changes with age and disease remain to be elucidated, but they may originate from dietary alterations, decreased enzymatic synthesis, or accelerated breakdown of LC-PUFAs and VLC-PUFAs. Future studies should include donor eyes with a wider range of well-characterized AMD pathologies, because the d Average = peak area in pooled sample/n; young age group: n = 5, middle age group: n = 6, old age group: n = 5, age-matched AMD group: n = 8. No statistics are provided due to only one pooled sample in each group.