Spatial organization of lipids in the human retina and optic nerve by MALDI imaging mass spectrometry.

MALDI imaging mass spectrometry (IMS) was used to characterize lipid species within sections of human eyes. Common phospholipids that are abundant in most tissues were not highly localized and observed throughout the accessory tissue, optic nerve, and retina. Triacylglycerols were highly localized in accessory tissue, whereas sulfatide and plasmalogen glycerophosphoethanolamine (PE) lipids with a monounsaturated fatty acid were found enriched in the optic nerve. Additionally, several lipids were associated solely with the inner retina, photoreceptors, or retinal pigment epithelium (RPE); a plasmalogen PE lipid containing DHA (22:6), PE(P-18:0/22:6), was present exclusively in the inner retina, and DHA-containing glycerophosphatidylcholine (PC) and PE lipids were found solely in photoreceptors. PC lipids containing very long chain (VLC)-PUFAs were detected in photoreceptors despite their low abundance in the retina. Ceramide lipids and the bis-retinoid, N-retinylidene-N-retinylethanolamine, was tentatively identified and found only in the RPE. This MALDI IMS study readily revealed the location of many lipids that have been associated with degenerative retinal diseases. Complex lipid localization within retinal tissue provides a global view of lipid organization and initial evidence for specific functions in localized regions, offering opportunities to assess their significance in retinal diseases, such as macular degeneration, where lipids have been implicated in the disease process.


Human eyes
Human eyes (whole globes) were obtained from the National Disease Research Interchange (Philadelphia, PA) and the Lions Eye Institute (Tampa, FL). Retinal tissue used in this study was from the right and left eyes of nine individuals, including a 45-year-old Caucasian female, a 47-year-old Caucasian male, a 55-year-old Hispanic male, a 56-year-old Caucasian female, a 60-year-old Caucasian male, a 73-year-old Caucasian female, a 79-year-old Black female, an 81-year-old Caucasian male, and an 82-year-old Caucasian female ( Table 1 ). At procurement, globes were removed, wrapped in moist gauze, placed in "eye bank" plastic containers on wet ice in Styrofoam coolers, and sent to Louisiana State University for arrival at the laboratory by 10:00 AM the following day . Upon arrival, globes were immediately dissected and prepared (unfi xed) for cryosectioning, or tissue samples collected for mass spectrometry. Frozen sections were placed on glass coverslips and immediately stored at Ϫ 80°C. Frozen sections were then moved to plastic containers on dry ice and shipped to the University of Denver for delivery the following morning. These coverslips were kept frozen until processing for MALDI imaging began. Tissue samples for mass spectrometry were placed in 1.5 ml microtubes in methanol containing butylated hydroxytoluene (0.1%), and stored at Ϫ 80°C until extraction.

Preparation of human retinas
The anterior segment was removed, radial cuts were made to fl atten each eye, and a horizontal strip containing the optic nerve and macular region was obtained and embedded in modifi ed OCT. The preparation of modifi ed OCT has been described in detail previously ( 34 ). The embedded tissue was frozen, attached to a sectioning stub, and cryosectioned to obtain 20 m-thick vertical sections showing all layers of the retina, choroid, optic nerve, and sclera ( Fig. 1B ). Sections were attached to glass coverslips and stored at Ϫ 80°C until needed for MALDI imaging. Alternate sections from each block were placed on conventional microscope slides and stained with hematoxylin and eosin (H&E) for orientation purposes. Rod and cone density across the human retina varies with photoreceptor type and location; cones are concentrated within the central macular region while rods occupy the peripheral retina ( Fig. 1A ) ( 35 ). Regions of interest within the eye, such as retina, optic nerve, sclera, accessory tissue, and RPE/choroid complex, as well as specifi c retinal layers, were labeled. For example, Fig. 1B shows an H&E-stained section that represents a typical tissue sample in this study . In total, more than 30 individual sections of eye tissue from the nine human subjects listed above were analyzed in both positive and negative ion mode by MALDI IMS for this study.

Matrix application
The glass coverslips containing the tissue sections were attached to stainless steel MALDI plate inserts (OptiToF, Applied Biosystems) using copper tape, and the matrix was deposited by sublimation ( 36 ). The sublimation of 2,5-dihydroxybenzoic acid and DHAP matrix onto tissues has been described in detail previously ( 34 ).

MALDI IMS
A quadrupole-TOF tandem mass spectrometer with an orthogonal MALDI source (QSTAR XL; Applied Biosystems/MDS Sciex, Thornhill, Ontario, Canada) was used to acquire mass spectral data from which images were constructed. MALDI mass spectra were obtained using a solid state laser (355 nm) at an energy of 7.4 J and a pulse rate of 200 Hz for positive ion MALDI imaging, and an energy of 7.6 J and a pulse rate of 500 Hz for negative ion imaging. The energy and pulse rate of the laser were selected to maximize the lipid signal from the tissue slice. Both the positive and negative ion MALDI imaging experiments had an accumulation signals, promoting cell survival (6)(7)(8). Another unique feature of retinal lipids is the abundance of phospholipids containing two PUFAs esterifi ed to the glycerol backbone, where a very long chain (24-36 carbons) (VLC)-PUFA) is located at the sn-1 position and DHA is esterifi ed at the sn-2 position ( 9 ). The functional importance of VLC-PUFAs in the retina is not well understood but VLC-PUFAs are reduced in retinas undergoing age-related macular degeneration (AMD) ( 10 ), and a key role for VLC-PUFAs has been suggested in both phototransduction regulation and photoreceptor survival ( 11,12 ).
Retinal phospholipids also play an important role in the visual cycle and the conversion of all-trans retinal to 11-cis retinal. In some cases, the N-retinylidene-glycerophosphatidylethanolamine (PE) formed during the visual cycle reacts with a second molecule of all-trans -retinal instead of hydrolyzing to PE and all-trans -retinal, and leads to the formation of the cytotoxic compound N-retinylidene-N-retinylethanolamine (A2E) ( 13 ). This side reaction product can progressively accumulate with age in the RPE and may contribute to retinal degenerative diseases such as AMD ( 14,15 ). Additionally, the production of ceramides, which are bioactive sphingolipids that are critical mediators for initiating oxidative stress-induced apoptotic cell death of both photoreceptors and RPE cells, may lead to several ret inal diseases including retinitis pigmentosa and AMD (16)(17)(18)(19).
Retinal phospholipids have been well-characterized by gas chromatography-mass spectrometry (20)(21)(22) and electrospray mass spectrometry ( 23-26 ) of extracted tissue, but the regional distribution of phospholipid molecular species in the various regions within the human retina has remained uncertain. MALDI imaging mass spectrometry (IMS) has been developed as a powerful method for investigating the distribution of phospholipids within tissues ( 27 ). Phospholipid molecular ions are produced directly from a tissue section coated with MALDI matrix, and sequential mass spectra are acquired by rastering the laser beam across the tissue surface; images can be obtained by plotting the intensity of the analytes of interest as a function of the x-y coordinates. Previously, MALDI IMS has been applied to the porcine lens ( 28 ), to salamander ( 29 ), rat, and mouse retinas (30)(31)(32), and to mouse optic nerve tissue ( 33 ), with the major focus generally on the localization of glycerophosphatidylcholine (PC) lipid molecular ions. In the current study, MALDI IMS is used to report the localization of PC, PE, glycerophosphatidylinositol (PI), and glycerophosphatidylserine (PS) lipid molecular species, as well as sphingolipids present in the human retina.

Reagents
Polyvinyl alcohol 6-98, 2,5-dihydroxybenzoic acid, and 2,6-dihydroxyacetophenone (DHAP) were purchased from Sigma-Aldrich Chemical Co. (St. Louis, MO). OCT compound was obtained from Ted Pella, Inc. (Redding, CA). Polypropylene glycol, average molecular weight 2,000 g/mol, was purchased from Fisher Scientifi c (Fair Lawn, NJ). from accessory tissue (e.g., adipose and connective tissue) indicated unique ions in these regions of the ocular section ( Fig. 2B ) with the most intense ion at m/z 881.8. The positive ion MALDI image of m/z 881.8 suggested that this ion was present within the tissue behind the sclera (outside the eye ball) and surrounding the optic nerve ( Fig. 2E ). The black areas in the image within this region of tissue are cross-sections of blood vessels. The CID of the tissue-derived MALDI ion at m/z 881.8 (supplementary Fig. IA, Table 2 ) revealed product ions at m/z 625.5, 603.5, 599.5, and 577.5, a behavior that was consistent with this ion being the [M+Na] + of a TAG, specifi cally a mixture of TAGs that had 52 acyl group carbon atoms with two double bonds residing collectively in the three fatty acyl groups and denoted as TAG(52:2) ( 38 ). Additionally, other sodiated TAG lipids were found with very similar regional distributions to m/z 881. The presence of a wide variety of lipid molecular species in the optic nerve was revealed in the positive ion MALDI mass spectrum of this region of the ocular tissue section ( Fig. 2C ). The most abundant ion in the positive ion MALDI mass spectrum of the optic nerve was at m/z 782.6 [PC(34:1)+Na] and the MALDI image suggested that the lipid generating this ion was uniformly distributed in the optic nerve with a slightly time of 243 ms per image spot. The MALDI plate was moved at a rate of 12.75 mm/min and after each horizontal line was completed the plate was moved vertically 50 m. The lateral resolution of this MALDI IMS technique has been estimated to be approximately 50 m. The mass spectrometric data were processed using a specialized script for Analyst software (Applied Biosystems/MDS Sciex) at a mass resolution of 0.1 amu, and images were visualized using TissueView software (Applied Biosystems/MDS Sciex). Collision-induced dissociation (CID) of selected ions in both positive and negative ion modes was carried out after the MALDI imaging experiments on the same tissue slice using relative collision energy of 40 V with argon as collision gas. The shorthand notation used for lipid species in this paper follows those outlined by Liebisch et al. ( 37 ). The lipids observed in positive ion mode [PC, SM, and triacylglycerol (TAG)] were identifi ed to the lipid species level by CID and are indicated by the number of acyl carbons and the number of double bonds, for example PC(34:1). The lipids observed in negative ion mode (PS, PI, and PE) were identifi ed to the fatty acyl level by CID and are indicated by the fatty acyl groups esterifi ed to the sn-1 and sn-2 position, such as PE(18:0/22:6).

RESULTS
Rod and cone density across the human retina varies with photoreceptor type and location; cones are concentrated within the central macular region while rods occupy the peripheral retina ( Fig. 1A ). Regions of interest within the eye, such as connective tissue, sclera, optic nerve, RPE, choroid, and retina, as well as the retinal layers, are labeled in an H&Estained section which represents a typical tissue sample ( Fig. 1B ). In this study the localization of lipid molecular species was determined using MALDI IMS. The positive ion mass spectra in the mass range m/z 650-900 ( Positive ion data obtained for the MALDI IMS experiments revealed the distribution of PC, SM, and TAG lipids in the ocular tissue. The positive ion MALDI mass spectrum At procurement, globes were removed, wrapped in moist gauze, placed in eye bank plastic containers on wet ice in Styrofoam coolers at NDRI or the Lions Eye Institute, and sent to Louisiana State University for arrival at the laboratory by 10:00 AM the following day. Upon arrival, globes were immediately dissected and prepared (unfi xed) for cryosectioning, or tissue samples collected for mass spectrometry. Frozen sections were placed on glass coverslips and immediately stored at Ϫ 80°C. Frozen sections were then moved to plastic containers on dry ice and shipped to the University of Denver for delivery the following morning. These coverslips were kept frozen until processing for MALDI imaging began. Tissue samples for mass spectrometry were placed in 1. the sclera, none of the positive ion MALDI images described above indicated that any of the lipid molecular species observed in the optic nerve were exclusive to this region of ocular tissue.
The positive ion MALDI mass spectrum of the retina indicated the presence of common, as well as some novel, phospholipid molecular species ( Fig. 2D ). The MALDI image of the most prevalent positive ion found in the retina, m/z 782.6 [identifi ed as PC(34:1)+Na above], indicated the presence of this ion with varying intensity throughout the entire retina, as well as the sclera, optic nerve, and accessory tissues ( Fig. 2G ). The most intense signal observed for m/z 782.6 in the retina occurred in the rod photoreceptor-dominant periphery, but this ion was also present in the cone photoreceptor-dominant macular region ( Fig. 2G ). Discrete retinal layers were not observed in the positive ion MALDI image of m/z 782.6, but the distribution of this ion was apparent within all the retinal layers. Many other ions present in the MALDI mass spectrum of the retina ( Fig. 2D ) were sodiated PC or SM lipids that contained saturated or monounsaturated fatty acids higher signal intensity at the optic disc ( Fig. 2G ). The presence of additional lipids in the optic nerve at m/z 810.6 [PC(36:1)+Na] and 832.6 [PC(38:4)+Na], which were identifi ed as sodiated PC lipids by CID ( Table 2 ), is indicated in Fig. 2C . The positive ion MALDI images of these sodiated PC ions also localized these lipids to the optic nerve, as well as sclera, and retina ( Fig. 2H , supplementary Fig. IIA). Additionally, the positive ion MALDI image of m/z 753.6 indicated that this particular ion was uniformly distributed throughout the optic nerve as well as in the retina and sclera, but with lower intensity, and that this particular lipid molecular species was not present at the optic disc ( Fig. 2F ). CID of m/z 753.6 yielded product ions at m/z 570.6 (neutral loss of phosphocholine), m/z 146.9 (sodiated cyclic 1,2-phosphodiester), and m/z 184.1 (phosphocholine) ( Table 2 ), and was consistent with the identifi cation of m/z 753.6 as SM(d18:1/18:0)+Na. Another sodiated SM at m/z 723.6, which was identifi ed as SM(d18:1/16:1)+Na by CID ( Table 2 ), occurred with a similar distribution described above in the optic nerve, sclera, and retina (supplementary Fig. IIB). Unlike the sodiated TAG lipids that were uniquely observed in the tissue behind Some of the lipid molecular ions present in the retina were highly localized and are described below. Within the inner retina, the neuronal portion proximal to the photoreceptor layer, a positive ion MALDI image of m/z 810. 6 [identifi ed as PC(36:1)+Na above] indicated that the signal for this lipid molecular species was strong, with similar intensity in both the peripheral and macular regions, at (  Fig. IIB-D). Additionally, none of these PC or SM lipids containing only saturated or monounsaturated fatty acids esterifi ed to the glycerol backbone were uniquely present in the retina. [PC(36:1)] (supplementary Fig. IIA, E), which were identifi ed by CID ( Table 2 ).
Within the photoreceptor layer, the positive ion MALDI image of m/z 856.6 indicated that this lipid was uniquely the optic disc, and within the astrocyte-rich optic nerve head and lamina cribrosa ( Fig. 2H ). Additionally, a similar localization in the inner retina was also observed in the MALDI images of m/z 832.6 [PC(38:4)+Na] and m/z 788.6 section ( Fig. 3B ). The negative ion MALDI image of the most abundant [M-H] Ϫ in accessory tissue, m/z 853.8, revealed that this particular ion was present only in the sclera ( Fig. 3E ). CID of m/z 853.8 resulted in a product ion at m/z 151.1 and a product ion resulting from the neutral loss of choline (59 amu) (  were not unique to the accessory tissues, but were widespread across the entire tissue section ( Fig. 3G ; supplementary Fig. IVA, B).
The negative ion MALDI mass spectrum of the optic nerve in the mass range m/z 650-1,000 indicated that a wide variety of lipid molecular species were present ( Fig. 3C ). The negative ion MALDI image of one of the most abundant ions observed in the optic nerve, m/z 726.5, indicated that the lipid generating this particular ion was located only in the optic nerve and was not present in the retina or the localized and was present with similar intensity in both the peripheral and macular regions ( Fig. 2I ). Also, this image indicates that this ion was not present at the optic disc. The CID spectrum of m/z 856.6 obtained directly from the tissue contained product ions at m/z 146.9, 184.1, and 673.6 (  Fig. IIF, G), which were identifi ed by CID ( Table 2 ).
The RPE, a monolayer of cells associated with the tips of the photoreceptors at the back of the retina, contained a lipid at m/z 548.5 that demonstrated unique localization in the eye tissue ( Fig. 2J ). This lipid molecular ion displayed a higher intensity in the peripheral region compared with the macular region of the retina and was also notably absent at the optic disc. CID of m/z 548.5 yielded a product ion at m/z 264.4 (supplementary Fig. IC ( Fig. 2K ). Specifi cally, PC(36:1)+Na was observed within the retinal ganglion cell layer, the axons of which comprise the optic nerve, PC(40:6)+Na had panretinal localization but was only found within the photoreceptor layer, and Cer(d18:1/18:0)-H 2 O was specifi cally located in the RPE/choriocapillaris complex.
Negative ion data obtained from the MALDI IMS experiments revealed the distribution of many lipid classes, including PI, PS, and PE, in the ocular tissue. The negative ion MALDI mass spectrum of accessory tissue did not reveal many lipids that were unique to this region of the tissue of these particular lipids in the retina ( Fig. 3K ). Specifically, PS(18:0/22:6) was located throughout both the inner and outer retinal layers and was not specifi cally present in any one layer, PE(P-18:0/22:6) was specifi cally located in the retinal ganglion cells and the optic nerve head, and PE(18:0/22:6) was specifi cally located in the outer retinal layers, which include photoreceptors and the RPE cells ( Fig. 3K ). The retina is known to contain many dipolyunsaturated lipids, with the majority present as PC molecular species that contain VLC-PUFAs from the n-3 and n-6 series having 24-36 carbon atoms and four to six double bonds in the mass range of m/z 900-1,100 ( 11 ). The most abundant ions in the positive ion MALDI mass spectrum in the mass range of m/z 900-1,100 occurred at m/z 909.3, 951.3, 1009.3, and 1067.3 ( Fig. 4C ) and were determined to be nonlipid species due to the low mass defect and the CID data that was not indicative of lipids (data not shown). It was thought that these ions originated from the embedding compound and the tissue preparation process as the MALDI ion images indicate that these ions occur both on and off the tissue section (supplementary Fig. VA-D). The positive ion MALDI mass spectrum of the retina in this higher mass range ( m/z 900-1,100) revealed the presence of several ions with an even mass-to-charge ratio and a high mass defect of 0.6 or 0.7 amu ( Fig. 4C ), indicative of lipid molecular species. The ions at m/z 900.6 and 1,070.7 were collisionally activated after the MALDI IMS experiment and displayed product ions at m/z 147.0 and 184.1 and a product ion corresponding to neutral loss of 183 (phosphocholine) ( Table 2 ). This CID data indicates that the ions at m/z 900.6 and 1,070.7 were sodiated PC lipids and were therefore identifi ed as a dipolyunsaturated PC, PC(44:12)+Na, and a VLC-PUFA PC lipid, PC(56:11)+Na, respectively. The positive ion MALDI images of PC(44:12)+Na and PC(56:11)+Na indicate that these PC lipids were uniquely localized to the retina ( Fig. 4E, F ). When the positive ion MALDI images of PC(40:6)+Na (blue), PC(44:12)+Na (red), and PC(56:11)+Na (green) were merged, the colocalization of these lipids in the photoreceptor layer was apparent ( Fig. 4G ). Additionally, the presence of DHA-and VLC-PUFA-containing PCs in the photoreceptor layer was further confi rmed in this particular section of tissue where the photoreceptor layer was detached from the inner retina ( Fig. 4A, B , and arrows Fig.  4D-G ). This fortuitous tissue sectioning artifact allowed the photoreceptor layer to be distinguished and imaged/analyzed separately from the inner retina and aided in determining that DHA-containing and VLC-PUFA-containing PCs are indeed present in the photoreceptor layer. Additionally, similar distributions were also observed in the MALDI images of the VLC-PUFA PC lipids at m/z 1,040.7, 1,042.7, and 1,044.7 (supplementary Fig. VE-G), which were identifi ed as PC(54:12)+Na, PC(54:11)+Na, and PC(54:10)+Na ( Table 2 ).
The reaction of all-trans retin al with PE to form a Schiff base is the initial step in formation of the toxic bis -retinoid, A2E ( 14 ), which has been found to accumulate in the RPE. Because this positive ion is of relatively low molecular weight, it seemed possible that it could be observed by sclera ( Fig. 3F ). CID of this lipid ion yielded product ions at m/z 281. , occurred with a similar distribution described above for m/z 726.5, and were only present in the optic nerve (supplementary Fig. IVC-F). Furthermore, these lipids were identifi ed by CID directly from the optic nerve tissue ( Table 3 ).
The negative ion MALDI mass spectrum of the retina indicated a wide variety of phospholipid molecular species ( Fig.  3D ). The MALDI image of one of the most abundant negative ions found in the retina, m/z 885.6, which was identifi ed above as PI(18:0/20:4), indicates that this particular ion is found in the optic nerve, retina, and sclera ( Fig. 3G ). Furthermore, the negative ion MALDI image of m/z 834.5 indicated that this molecular species was present in the retina, except at the fovea, and at the optic disc ( Fig. 3H ) Table 3 ) and was identifi ed as PS(18:0/22 :6). Another lipid at m/z 750.5 had a similar distribution mentioned above and was present in both the optic nerve and the retina (supplementary Fig. IVG), and this ion was identifi ed by negative ion CID as PE(P-18:0/20:4) ( Table 3 ).
The distributions of some of the lipid molecular species observed in the negative ion MALDI images of the retina were highly localized. The negative ion MALDI image of m/z 774.5 indicated that this ion was observed at the optic disc and within the inner retina, except for a notable absence of signal intensity at the fovea ( Fig. 3I ) ( Table 3 ), and was identifi ed as PE(P-18:0/22:6) ( 45 ). Very similar localization was also observed in the negative ion MALDI image of m/z 766.5 (supplementary Fig. IVH), which was identifi ed as PE(18:0/20:4) by CID ( Table 3 ). The negative ion MALDI image of m/z 790.5 indicated that this lipid molecular species was found only in the photoreceptor layer, including the fovea, but not at the optic disc or the optic nerve ( Fig. 3J ) ( Table 3 ), and was identifi ed as PE(18:0/22:6). This lipid is thought to be present uniquely in the photoreceptor cells due to its specifi c localization in the outermost retinal layers, but not in the optic nerve, optic disc, or innermost retinal layers where photoreceptors are absent. Furthermore another lipid at m/z 762.5 had a similar distribution to PE(18:0/22:6) and was only present as a narrow line of signal intensity across the length of the retina, but not within the optic nerve or optic disc (supplementary Fig. IVI). This ion at m/z 762.5 was identifi ed by negative ion CID as PE(16:0/22:6) ( Table 3 ).
The  ( 4,21,22,47,48 ), however the exact molecular species and diverse phospholipid classes of esterifi ed DHA has not been previously described. MALDI IMS readily revealed the location of DHAcontaining phospholipids and indicated that most were only observed in the photoreceptor layer of the retina ( Figs. 2I, 3J ). However, MALDI IMS revealed that there were some DHA-containing phospholipids that were either in both the ganglion cell layer and the RPE/choriocapillaris complex, such as PS(18:0/22:6) shown in Fig.  3H , or highly localized to the ganglion cell layer and the optic disc, like PE(P-18:0/22:6) shown in Fig. 3I . In addition to the glycerophospholipids that contain this omega-3 MALDI IMS at m/z 592.5. A signal for this ion was localized to the RPE and clearly separated from the photoreceptor layer as indicated by the ion abundance at m/z 856.6 [PC(40:6)+Na] and the neuronal layer indicated by m/z 810.6 [PC(36:1)+Na] ( Fig. 5 ). However, the signal intensity of this ion was too low to obtain CID data and the identifi cation of this ion as A2E is tentative. The molecular ion of A2E was also observed in the RPE in a MALDI IMS study in mice that directly correlated lipofuscin with A2E ( 46 ).

DISCUSSION
The lipid biochemistry present in the human retina is remarkable from many aspects, not the least of which is the unique accumulation of DHA (22:6) esterifi ed to various suggesting a dual role for ceramide in regulating cellular homeostasis.
MALDI IMS studies of nonhuman retinas have also revealed stratifi cation of specifi c phospholipids. In salamander retinas, PC(32:1), PC(32:0), and PC(34:1) are present in the outer and inner plexiform layers, and PC(36:4), PC (38:6), and PC(40:6) are found within the photoreceptor inner and outer segments, and the RPE ( 29 ). Due to the large size (10 m or larger in diameter) of salamander photoreceptors this study was able to show that PC(36:4), PC(38:6), and PC(40:6) were differentially present within a single rod outer segment cross-section. Unfortunately, single 1.5-2.0 m diameter rods found in human retinas cannot be detected by the method used in the current study, which has a resolution of 50 m. A MALDI-based study of mouse retinas also demonstrated that PC(16:0/18:1) was localized to the inner nuclear layer (INL) and outer plexiform layer, PC(16:0/16:0) to the outer nuclear layer (ONL) and photoreceptor inner segment, and PC(16:0/22:6) and PC (18:0/22:6) to the photoreceptor outer segment and RPE ( 30 ), demonstrating 22:6 enrichment within photoreceptors as previously reported ( 4 ).
The accumulation of lipofuscin within the RPE occurs throughout life, is considered an indication of aging, and may correlate with the onset and/or progression of AMD ( 54,55 ). While the complete makeup of lipofuscin is not known, it is highly enriched in bis-retinoids where the retinoid visual cycle is most active ( 13 ). The most completely characterized of the intermediate retinaldehyde adducts is A2E. In animal models with increased photoreceptor retinaldehyde, lipofuscin and A2E are also increased, suggesting that the A2E precursors, phosphatidylethanolamine and retinaldehydes, accumulate in the RPE from phagosomes and undergo incomplete degradation with gradual formation and accumulation of A2E. Evidence associates A2E with retinal degeneration ( 56 ). Recent studies utilizing MALDI imaging to compare lipofuscin and A2E localization in human and mouse RPE found good correlation in murine RPE ( 45 ), but only minimal correlation within human RPE ( 57 ). This would suggest that A2E metabolism may vary between human and mouse, and that variation in photoreceptor distribution may play a role in localization.
An additional MALDI IMS study utilized mice and rats to characterize lipids of the optic nerve ( 33 ). Mass spectra of optic nerve sections were generated from m/z 700-970 in both positive and negative ion mode, and peaks identifi ed, revealing a wide variety of PC, PS, SM, and Cer lipids. Specifi cally, the sulfatide (d18:1/18:0) was localized to a region corresponding to the meninges around the nerve bundle and the sulfatide (d18:1/24:0) within the optic nerve, PC(18:1/18:0) was found around the optic nerve as it exited the retina, and PC(16:0/16:0) near the optic nerve head-vitreous interface. PC(18:0/20:4) was identifi ed in the meninges, the connective tissue surrounding the optic nerve, and within the wall of a blood vessel. Several PS and PI lipids were also imaged, indicating unique labeling patterns for each lipid, demonstrating the specifi city of MALDI IMS in revealing retinal lipid distribution. fatty acid, there were also VLC-PUFAs produced by the elongation of DHA to fatty acyl chains containing 26-40 carbon atoms by the enzyme elongase of very long chain fatty acid 4 (ELOVL4) in the retina ( 49 ). While the structure of these VLC-PUFA-containing phospholipids has been previously established ( 22 ), determination of their exact location required careful dissection to uniquely isolate individual cellular structures within the retina. Our MALDI IMS studies did demonstrate the presence of VLC-PUFA-containing PC lipids exclusively in the photoreceptor layer, but the signal intensity of these ions was not robust due to the relatively low abundance of these lipids in general in the retina ( 10 ).
Several unique features of the retina could be directly observed by the abundance of specifi c lipids. In tissues localized behind the eye, a high abundance of neutral lipids such as TAG molecular species could be observed, but there was little if any PUFA content in this region rich in adipose tissue ( Fig. 2E ). In the structures containing neuronal cells, for example, the ganglion cell layer and optic nerve head, a relatively high concentration of phospholipids containing fatty acyl chains derived from saturated and monounsaturated precursors was observed ( Fig. 2G, H ). These lipid molecular species, including the phosphatidylcholine lipid, PC(34:1), are typically observed as the most abundant molecular species in most other neuronal tissues such as the brain ( 50,51 ). Additionally, SM ( Fig. 2F ) lipids were also readily apparent in the retinal layers containing the neuronal cells and the optic nerve. The phospholipids that were unique to the photoreceptor layer of the retina contained DHA in several classes of glycerophospholipids including PC and PE. The highly unsaturated species of PI and PS, containing esterifi ed DHA and arachidonate, were not found to be highly localized and were observed in all of the retinal layers ( Fig. 3G, H ). In the RPE/choriocapillaris complex, lipid mediators such as the bioactive sphingolipid ceramide could be detected. Ceramide has been shown to play a role in cell cycle arrest ( 52 ) and, conversely, is critical as a key mediator of oxidative stress-induced apoptosis in retinal photoreceptor cells ( 17,18 ). Of interest is increasing evidence that suggests that the phosphorylation of ceramide to ceramide-1phosphate regulates cell proliferation and apoptosis ( 53 ), Generally, lipid characterization by MALDI IMS in ocular tissue has been limited to mass spectra that extend to roughly m/z 900, eliminating the possibility of detecting and identifying the VLC-PUFA-containing PC lipids, which occur at very low concentrations in photoreceptors. Electrospray mass spectrometry analysis of retinas [salamander ( 29 ), rat ( 11 ), and human ( 10,24 )] has revealed the presence of VLC-PUFA-containing PC lipids, but localization of these lipids by MALDI imaging has not been presented. It is now understood that 22:6 is required for a healthy retina (6)(7)(8). It is important in homeostatic regulation of photoreceptor cell integrity and serves as the precursor for the potent mediator neuroprotectin D1. It plays a role in managing neuroinfl ammation, and diminished 22:6 has been linked to macular degeneration. It can also be the precursor molecule for the generation of VLC-PUFAs which are tightly associated with opsin in photoreceptor outer segment disk membranes. Moreover, a mutation in the ELOVL4 gene produces a dysfunctional enzyme that is incapable of elongating PUFAs ( 12,58 ). Thus, it is extremely important to characterize the long chain and very long chain PUFAs of the retina and to determine their localization in young and aging human retinas, and in retinas exhibiting degeneration. MALDI imaging provides the means with which this can be accomplished.
While unique lipid biochemistry can be determined with this technique, application of this to study disease processes is limited to the availability of human eyes. The information presented in this manuscript, obtained from nine human subjects, revealed virtually identical distributions of major phospholipids in the inner retina, the photoreceptor cell layer, and the RPE/choriocapillaris complex, but this number of human subjects would be insuffi cient to ascertain subtle changes in lipid biochemistry attributable to disease processes. Nonetheless, MALDI IMS holds considerable promise if an appropriate number of age-matched control (normal) and ocular-diseased eyes could be obtained and imaged in the same study. This approach of determining the molecular species of phospholipids present in retinal tissue provides a global view of complex lipid content and evidence for unique biochemical processes taking place in localized regions.
In conclusion, the unique remodeling of omega-3 PUFAs within human retinal tissues was readily observed by MALDI IMS. The advantage of MALDI IMS is that the location of many different lipids, including glycerophospholipids, sphingolipids, and TAGs, can be readily assessed in tissue slices and relative abundance of these lipids can be determined based upon the intensities of their observed signals as long as they are within the same phospholipid family ( 59 ). Unfortunately, the 45-82 year time-span represented by the small number of specimens analyzed in this study was too small to determine age-specifi c differences. However, the results of this study revealed unique biochemical events related to cells within specifi c tissue regions of the retina and perhaps hold the promise that such an analytical approach could be used to assess the localization of lipids in age-related studies or in different diseases such as macular degeneration, retinitis pigmentosa, and diabetic retinopathy,