Salamander retina phospholipids and their localization by MALDI imaging mass spectrometry at cellular size resolution.

Salamander large cells facilitated identification and localization of lipids by MALDI imaging mass spectrometry. Salamander retina lipid extract showed similarity with rodent retina lipid extract in phospholipid content and composition. Like rodent retina section, distinct layer distributions of phospholipids were observed in the salamander retina section. Phosphatidylcholines (PCs) composing saturated and monounsaturated fatty acids (PC 32:0, PC 32:1, and PC 34:1) were detected mainly in the outer and inner plexiform layers (OPL and IPL), whereas PCs containing polyunsaturated fatty acids (PC 36:4, PC 38:6, and PC 40:6) composed the inner segment (IS) and outer segment (OS). The presence of PCs containing polyunsaturated fatty acids in the OS layer implied that these phospholipids form flexible lipid bilayers, which facilitate phototransduction process occurring in the rhodopsin rich OS layer. Distinct distributions and relative signal intensities of phospholipids also indicated their relative abundance in a particular cell or a cell part. Using salamander large cells, a single cell level localization and identification of biomolecules could be achieved by MALDI imaging mass spectrometry.

retina having large cells and an atmospheric pressure (AP) MALDI IMS system having a more focused laser beam (<10 m m) enabled us to identify phospholipids at cellular size resolution. The narrow laser spots (small yellow dots, Fig.  1 , arrow) are visible inside cells after toluidine blue staining. The AP MALDI MS instrument was equipped with a high resolution optical microscope that facilitated observation and selection of the target area of tissue samples for accurate analysis. ESI MS analyses of the salamander retina lipid extract also provided a complementary lipid profi le.

Chemicals and reagents
All solvents used were HPLC and LC-MS grade and other chemicals used were analytical grade (Wako Pure Chemicals, Japan). Ultrapure water was obtained from Milli-Q water system (Millipore, MA).

Preparation of retina section for MALDI IMS
A. mexicanum was anesthetized by 0.2% meta-aminobenzoic acid ethylester methanesulfonate (Sigma). Dissected eyeballs were rapidly frozen by liquid nitrogen-cooled isopentane. Five micronthick frozen serial sections were cut on a cryostat (CM3050S, Leica, 2 20°C) and thaw mounted on an indium tin oxide (ITO) transparent (80%, l = 550 nm) conductive fi lm (100 V , Tobi Co. Ltd, Kyoto, Japan). The ITO fi lm was pasted on the MALDI target plate using a double-sided conductive carbon tape. The carbon tape was pasted on the top of the ITO fi lm. After MALDI IMS data acquisition, sections were stained by toluidine blue and photographed.
Matrix (2,5-dihydroxybenzoic acid, DHB) was coated on the retina sections by an evaporation method ( 18 ): evaporation condition, pressure <5 Pa; temperature, 130°C; time, 10 min; and amount of matrix, 30 mg. The matrix crystals were uniformly deposited on the tissue. The crystal size was about 1-2 m m (supplementary Fig. IX). The thickness error of matrix on the tissue was about 10% from center to edge but <1% in the analyzed area (analyzed area was about 1/100 of total tissue area.).

Preparation of retina lipid extract
Eyeballs were dissected from an anesthetized A. mexicanum . After dissection, retina was immediately separated from eyeballs and quickly frozen in liquid nitrogen. Isolated retina was kept at 2 80°C until lipid extraction. A pair of salamander retinas was homogenized with chloroform/methanol (1:2, v/v, 3 ml) using a glass homogenizer and total lipids were extracted using the Bligh-Dyer method ( 19 ). The total crude lipid extract was dried under a gentle nitrogen stream and redissolved in methanol (1 ml). The crude lipid sample was stored at 2 80°C until MS analysis.

LC-ESI MS analysis
ESI MS analysis was performed using a 4000 Q-TRAP® quadrupole-linear ion trap hybrid mass spectrometer (Applied Biosystems/MDS Sciex, Concord, ON, Canada) equipped with an Ultimate 3000 HPLC system (DIONEX Co., San Francisco, CA) or an ACQUITY Ultra Performance LC ® (UPLC) (Waters, Millford, MA) combined with an HTC PAL autosampler (CTC Analytics, Zwingen, Switzerland). The conditions used for the 4000 Q-TRAP were as follows. The sample (5 m l) was subjected to ESI MS analysis by fl ow injection without LC separation. The mobile phase composition was acetonitrile/methanol/ammonium formate 50 mM (45:50:5, pH 7.4) at a fl ow rate 10 m l/min. The peroxidation that is considered one of the causative factors of retinal diseases, such as age-related macular degeneration and diabetic retinopathy ( 9,10 ). Because lipids are considered important causative and diagnostic molecules for retina diseases, a complete knowledge of the lipid profi le of the normal retina is necessary ( 10 ).
Extraction of lipids from tissue followed by lipid characterization using HPLC MS and GC MS is the major direction in lipid research. However, spatial distribution of lipids in a given tissue is lost by the extraction method ( 1-3, 11, 12 ). To get information about spatial distribution of biomolecules in a given tissue, MALDI imaging mass spectrometry (IMS) is a versatile and robust method. IMS is gaining popularity for the analysis of biomolecules directly on tissue sections (13)(14)(15)(16)(17).
Using MALDI IMS, Hayasaka et al. ( 13 ) characterized four phosphatidylcholine (PC) molecules that were distributed in three distinct layers in a mouse retina section. The exclusive distribution of phospholipids containing DHA in the retinal OS and pigment epithelium has also been shown. Recently, their group also revealed fatty acid layer distribution in a mouse retina section using a nanoparticle-assisted laser desorption/ionization imaging mass spectrometry (nano-PALDI IMS) ( 14 ).
An experimental model animal, salamander ( Ambystoma mexicanum ), was selected to identify and visualize phospholipids in the retina section. Due to the salamander's large cells and a relatively simple nervous system, we envisage that it could be a suitable model animal for a single cell IMS study. As shown in Fig. 1 , the salamander retina section has a similar basic structural organization with that of the mouse ( 13 ). Retinal pigment epithelium (RPE), OS and inner segment (IS), outer nuclear layer (ONL), outer plexiform layer (OPL), inner nuclear layer (INL), inner plexiform layer (IPL), and ganglion cell layer (GCL) are easily recognized. In addition, the salamander retinal cells (average diameter 14.5 m m) are larger than that of the mouse (average diameter 6.5 m m), and every layer contains fewer cells than the mouse retinal layer. The salamander Identifi cation of PC species was based on their molecular mass ions ( m/z ), their various alkali metals adducts, and MS/MS analysis followed by a lipid database search ( 23 ). The imaging data and MS/MS data were obtained from the same retina section. When the product ions corresponded to specifi c neutral loss of trimethylamine, (N(CH 3 ) 3 , 59 Da) and of cyclophosphane ring (C 2 H 5 O 4 P, 124 Da) from a PC (supplementary Figs. X-XII) the m/z value of the PC precursor ion was searched against the lipid database. The data search provided total carbon number and unsaturated bond number included in the PC. Due to low population of precursor ions, we could not obtain detailed fatty acid composition of PC by MS/MS analysis. However, the corresponding PC with detailed fatty acid composition was unambiguously identifi ed by MS/MS data obtained by LC-MS analysis of the crude lipid extract.
Using the AP MALDI MS system, we ran three experiments independently using at least three tissue sections from three animals and found the data to be reproducible (supplementary Fig. XIV).

Salamander retina phospholipids composition
Both positive and negative (supplementary Fig. I) ions ESI mass detected phospholipids in the salamander retina crude lipid extract. The most abundant phospholipids signals at m/z 732.6, 760.6, 782.6, 788.6, 806.6, 834.6, 878.5, and 906.6 were observed in the mass spectrum ( Table 1 ), as shown in Fig. 2. Fatty acid composition in these lipid species varies from long-chain saturated fatty acids (FA 16:0) to very long-chain PUFAs (FA 36:6). A unique lipid (PC 44:12, m/z 878.5) containing DHA at both the sn1 and the sn2 positions of glycerol moiety was detected. PCs containing very long-chain PUFAs (PC 54:12, m/z 1018.8; PC 56:12, m/z 1046.7; PC 58:12, m/z 1074.9) were also observed in the crude lipid extract ( Table 1 ). ESI MS analyses of the crude lipid extract also revealed the presence of phosphatidylethanolamine, phosphatidylserine, and phosphatidylinositol (  range of the instrument was set at m/z 400-1200 at a scan speed of 1000 Da/s. The trap fi ll-time was set at 3 ms in the positive ion mode. Nitrogen was used as a curtain gas (setting of 10, arbitrary unit) and as a collision gas (set to high). The declustering potential was set at 20 V to minimize in-source fragmentation. Both Q1 and Q3 resolutions were set to unit mass. The collision energy used was varied according to the desired experiment. In order to identify the PC species, precursor ion and neutral loss scanning were performed; these methods are effective for the detection of phospholipids, because characteristic fragment ions or neutral losses are generated by collision-induced dissociation (CID).
The MS/MS analysis was performed during LC-ESI MS measurement in negative ion mode. The ion spray voltage was set at −4500 V. Neutral loss scanning was performed to identify the PC species as described previously ( 20,21 ). The characteristic fragmentation patterns of individual molecular PC species were determined by enhanced product ion scanning (EPI). These EPI experiments were performed as continuous data independent measurements for obtaining mass chromatograms of characteristic product ions. The collision energy was set at 45 eV. The scan range of the instrument was set at m/z 200-1000 at a scan speed of 1000 Da/s. The Q0 trapping was set to "ON". The linear ion trap fi ll-time was set at 10 ms. The declustering potential was set at −60 V. The resolution of Q1 was set to unit mass.

MALDI MS analysis
AP MALDI IMS data were acquired using a Mass Microscope ( 22 ). The Mass Microscope was developed by combining a LC-MS-IT-TOF (Shimadzu Corp., Kyoto, Japan) with a newly scanning AP MALDI system. The scanning AP MALDI system includes a high resolution optical microscope (40×) and a pulsed Nd:YAG laser (355 nm). The image resolution of the optical microscope is only a few micrometers. The laser irradiation area could be adjusted by defocusing the triplet lens, and the diffraction limit of the laser system is approximately 4 m m. These remarkable functions enabled us to detect the distribution of biomolecules on tissue sections at high spatial resolution. The following parameters were used for data collection: measurement mass range, m/z 550-1000; mass resolution, 10000; laser power, 0.11 m J; laser shot pitch, 8 m m; laser spot size, ‫ف‬ 7 m m; number of pixels, 65 × 65 ( Fig. 3 ) and 30 × 6 ( Fig. 4 ) (positive ion mode); number of laser irradiations per pixel, 80.

MALDI IMS data processing
In-house built software and MATLAB TM version 7.5 (Mathworks, Inc.) were used for raw data saving, processing, and interpretation of IMS data. For principle component analysis (PCA), the hundred most intense peaks were picked from m/z 550-1000 at a mass tolerance of 0.5 Da. Pareto scaling was applied to reduce the relative importance of large peaks. PCA was carried out, and several principal components with high score were selected. VARIMAX rotation was applied to them in order to obtain more interpretable data.
PCA of acquired MALDI IMS data provided several principal components with their contributed spectra. Among them, two signifi cant principal components were selected. Their contributed spectra and distribution maps for six major phospholipids are shown in Fig. 3 . All major lipids that were detected by MALDI MS were also observed in the crude lipid extract by ESI MS. Either PCA or manually picking up m/z values showed similar lipid distribution patterns in the retina section. As shown in Fig. 3   ).

Distribution of phospholipids in the retina section
PCA is one of the multivariate analyses and categorizes a number of correlated variables into several principal components that are not correlated with each other. By applying PCA and displaying each component's "score" in a score-plot, one can distinguish particular regions in it and know the constituent ions ( m/z ) ( 4 ). were detected in the OS and IS regions ( Fig. 4 ). These lipids were highly populated over 50% of the total pixels (30 × 6) in the given scanned area and showed their discrete distribution with a variable intensity over the scanned area. Among the three PC species, PC (38:6) and PC (40:6) were the most abundant lipids ( Fig. 4B ). Both lipids were localized on the basal part of photoreceptor cells whereas the other PC 36:4 was concentrated on the apical part. These lipids' localization is consistent with the whole retina imaging map, as shown in Fig. 3 . At the pixel 45, the signal intensity of PC 38:6 was relatively faint compared with PC 36:4 and PC 40:6 signals, indicating its low abundance at this position ( Fig. 4B , arrow). The relative intensity of each molecular ion ( m/z ) indicated its relative abundance in the particular cell or cell part at each pixel ( Fig. 4C ). Each circle (laser spot) could be considered to cover a single cell or a part of a single cell, and a relative lipid composition in a particular spot might indicate its lipids profi le. Empty spots/circles could be a result of low lipid content, poor matrix distribution, the mass spectrum. This lipid also showed similar alkali metal adducts formation in the OPL and INL layers. The other lipid also formed alkali metal adducts throughout the retinal layers in a similar fashion ( Table 3 ). It is worth mentioning that the adduct formation affi nity of metal ions with lipids containing PUFAs was somewhat different from lipids containing saturated or monounsaturated fatty acids (supplementary Table I). These results indicated that we observed lipid distribution throughout the retinal tissue rather than alkali metals distribution.

Mosaic distribution of phospholipids in the OS and IS regions
It was diffi cult to differentiate the distribution of lipids in the IS, OS, and RPE layers because of their intercrossing. To detect specifi c lipid composition on the IS and OS layers of the salamander retina, MALDI raster scanning was focused on the IS and OS layers ( Fig. 4A , area in circles). Three major phospholipids containing PUFAs at m/z 820.5 (PC 36:4), 844.5 (PC 38:6), and 872.5 (PC 40:6)  composed of lipids containing mainly PUFAs ( 5-7 ). The phototransduction process involving conformational changes of rhodopsin is largely facilitated by a suffi ciently fl exible lipid bilayer, and the lipid bilayer fl uidity is mostly dependent on the composition of phospholipids' fatty acyl groups. Phospholipids composing PUFAs like DHA, with low gel-to-fl uid properties ( 26 ), provide sufficient fl uidity of the lipid bilayer and play signifi cant roles in the phototransduction process ( 8 ). Thus, the specifi c presence of lipids containing PUFAs in the OS and IS layers indicates that these lipids form the fl exible lipid bilayers and their different compositions might indicate their cell and cell part specifi city in the OS and IS layers. However, identifi cation of particular cell types through their lipid profi les is beyond the scope of this present work. Further research is necessary to clarify the cell specifi c lipid profi le. The large cells of the salamander allowed us to identify and localize lipids in the retinal layers at cellular size resolution. By considering the tissue section thickness (5 m m), which is smaller than the salamander retinal cell (14 m m), a single cell level analysis could be achieved, as shown in Fig. 4 . absence of matrix, or laser irradiation on an intercellular space.

DISCUSSION
Phospholipids content and composition of the salamander retina lipid extract showed high similarity with that of the rat retina lipid extract ( 12 ). Detection of lipids containing long-and very long-chain PUFAs that associated with rhodopsin was consistent with bovine retina lipids ( 25 ). Major lipid species along with their fatty acid compositions were identifi ed on the basis of their m/z , MS/MS data, and database search ( 23 ) ( Tables 1, 2 ). Among them, six major lipids were detected and localized in the AP MALDI IMS measurement. PCA of MALDI IMS data found the same phospholipids peaks as detected in the crude lipid extract by ESI MS. Thus, PCA is a useful tool to fi nd signifi cant molecular peaks from principal components without previous knowledge of molecules.
Lipids containing PUFA acyl moieties were observed mostly in both the OS and the IS layers, which is consistent with previous fi ndings ( 13 ). The OS layer harboring rhodopsin, where phototransduction process occurs, is  ESI MS data of the salamander retina crude lipid extract are complementary to MALDI MS data obtained directly from the salamander retina section. It is useful to have ESI and MALDI MS methods in combination to get a relatively complete lipid profi le of tissue/cell, particularly where MALDI MS could not provide useful MS/MS information. The ESI MS analysis showed a much richer lipid profi le of the salamander retina than the MDLDI MS. In this study, The highly focused laser beam gave imaging capacity within a small area (7 m m). However, the major limitations of the small laser spot are signifi cant decrease of ion population of mass signals and an increase in experimental time. With a signifi cant reduction of ion population from a 7 m m spot, MS/MS data analysis is more diffi cult. Identifi cation of low abundant lipids was also impractical due to very low intensity of mass signals. we focused only on the major PC analysis by MALDI MS. Analysis of other classes of lipids (e.g., PE, PS, and PI) along with low abundant lipids deserves further research. It is suggested that these lipids undergo cleavage during ionization from matrix (DHB) ( 27 ). Thus, their analysis could be achieved using other matrix and negative ion mode MALDI.
In conclusion, identifi cation and molecular profi ling of a particular cell or cell types using MALDI IMS would be an interesting future research area, and the salamander, having large cells, could be an ideal model animal for IMS analysis and profi ling of biomolecules in a given tissue and/or cell.