Combination of ESI and MALDI mass spectrometry for qualitative, semi-quantitative and in situ analysis of gangliosides in brain

Gangliosides are a family of complex lipids that are abundant in the brain. There is no doubt the investigations about the distribution of gangliosides in brian and the relationship between gangliosides and Alzheimer’s disease is profound. However, these investigations are full of challenges due to the structural complexity of gangliosides. In this work, the method for efficient extraction and enrichment of gangliosides from brain was established. Moreover, the distribution of gangliosides in brain was obtained by matrix-assisted laser desorption ionization (MALDI) mass spectrometry imaging (MSI). It was found that 3-aminoquinoline (3-AQ) as matrix was well-suited for MALDI MS analysis of gangliosides in negative ion mode. In addition, the pretreatment by ethanol (EtOH) cleaning brain section and the addition of ammonium formate greatly improved the MS signal of gangliosides in the brain section when MALDI MSI analysis was employed. The distribution of ganliosides in cerebral cortex, hippocampus and cerebellum was respectively acquired by electrospray ionization (ESI) MS and MALDI MSI, and the data were compared for reliability evaluation of MALDI MSI. Further, applying MALDI MSI technology, the distribution of gangliosides in amyloid precursor protein transgenic mouse brain was obtained, which may provide a new insight for bioresearch of Alzheimer’s disease (AD).

and sialic acid are varied. The structural complexity of these lipids is an impediment to achieving accurate and precise analysis in biosystem.
So far, mass spectrometry (MS) has been extensively employed to study gangliosides. Fast atom bombardment (FAB)-MS successfully determined the structures of both the lipids and the carbohydrate portions of gangliosides standards [9][10][11] . However, FAB is a relatively harsh and insensitive ionization technique, which can cause extensive precursor ion fragmentation that complicates mixture analysis. More gentle ionization modes such as matrix-assisted laser desorption ionization (MALDI) and electrospray ionization (ESI) have also been utilized in the analysis of gangliosides. Ivleva and coworkers established a vibrationally cooled MALDI ion source, and fragile gangliosides could be desorbed from thin-layer chromatography (TLC) plates without fragmentation 12 . Further, Taki established a method for transferring lipids separated on a TLC-plate to a poly-vinylidene difluoride (PVDF) membrane and direct mass spectrometric analysis of ganglioside molecular species from the human brain by MALDI MS 13 . Recent advances in MALDI MS techniques of lipid analysis have led to the direct analysis of tissue slices with the MALDI mass spectrometry imaging (MSI). Colsch et al. developed a new matrix preparation (2,6-dihydroxyacetophenone [DHA]/ammonium sulfate/heptafluorobutyric acid [HFBA]) to maximize the detection of all ganglioside species, and reported the differential distribution of ganglioside species in the rat brain's cerebrum by MALDI MSI, which was used to map and image gangliosides with detailed structural information and histological accuracy 14,15 . ESI is the major ionization method in MS for lipid analysis. For example, an effective method by liquid chromatography (LC) ESI tandem mass spectrometry (LC/ESI-MS/MS) were developed to separate and distinguish regioisomeric gangliosides, such as GM1a and -1b, GD1a, -1b, and -1c, and GT1a, -1b, and -1c 16 .
Although MS is a powerful analytical tool, each ionization method still has some limits. MALDI MS analysis has the disadvantages of rather poor reproducibility, mainly originating from the heterogeneity of the matrix-analyte crystals, which leads to MALDI MS being heavily criticized for its quantitative analysis 17 . In addition, the samples will be quickly consumed under laser irradiation on specific site, which lead to a discontinuous ion flow, and is adverse to qualitative analysis. Meanwhile, the ESI MS cannot be directly used for imaging. The desorption electrospray ionization (DESI) may be capable of in situ analysis 18 , however the convenience and spatial resolution for imaging is expected to be improved. In our work, we constructed a workflow, by combination ESI with MALDI mass spectrometry, for qualitative, semi-quantitative and in situ analysis of gangliosides in brain. We performed cross validation of data for method validation. We expect to maximize the advantages of each ionization method.

Results
The Work Flow for Qualitative, Semi-quantitative and in Situ Analysis of Gangliosides. Mass spectrometry (MS), in particular MALDI and ESI MS, has been extensively employed to study lipids [19][20][21] . MALDI MS has demonstrated its unique features, namely, no necessity of labeling, high sensitivity, high throughout, molecule-specific, and the capacity of in situ localizing a wide range of biomolecules simultaneously from a tissue specimen in one single run. However, MALDI MS analysis has the disadvantages of both qualitative and quantitative analysis, already mentioned in Introduction. In this regard, ESI MS could be a perfect complementary. In this work, we constructed a workflow, shown in Fig. 1, by combination ESI with MALDI MS, for qualitative, semi-quantitative and in situ analysis of gangliosides in brain. The ESI and MALDI MS complement and verify each other, and we expect to maximize the advantages of each ionization method and to obtain more comprehensive and more accurate data as far as possible.
To overcome the phenomenon of ion suppression, UHPLC was employed to separate lipids in brain extracts before ESI MS detection. The mass spectrum and several representative extracted ion chromatogram (XIC) were shown in Fig. 2. GD, GT and GQ series mainly appeared as the doubly charged molecular ions ([M-2H] 2− ); while GM series mainly appeared as deprotonated ions ([M-H] − ). By matching the precursor ion (mass error < 5 ppm) and product ions ( Supplementary Fig. S1-S8) obtained by ESI MS/MS with free online databases and standard gangliosides, gangliosides were identified, and the m/z value in the mass spectrum with corresponding deprotonated molecule ion and the mass accuracy after comparing the theoretical m/z value were indicated in Supplementary Table S1. The presence of the O-acetyl forms of the GD1 and GT1 gangliosides was consistent with the results reported in previous research 22 . The result showed that the major ceramide portions in brain were d18:1/18:0 and d18:1/20:0.
The peak area of extracted ion in XIC could be used for semi-quantitative analysis of the ganglioside. The opposing polarities of head group and ceramide tail of gangliosides required finely balanced proportions of organic solvents to concomitantly meet the requirements for efficient solubilization, chromatographic separation, and ionization. The optimized LC conditions were shown in "FTICR MS" in Methods below. Due to the high  mass accuracy and resolution, we could extract an exact mass ion chromatogram with a narrow mass window of 5 ppm. The retention time (RT) of gangliosides identified was also listed in Supplementary Table S1. The data indicate that the ganglioside was separated on the basis of the ceramide portion, and the gangliosides with long and less-unsaturated fatty acid chain in ceramide portion eluted later. The relative standard deviation (RSD) of peak area of every single ganglioside is less than 7%. The peak areas in XIC of each ganglioside in cerebral cortex, hippocampus and cerebellum were compared, and a different distribution of ganglioside was observed. For example, d18:1/18:0 GM1 (m/z 1544.8643) and d18:1/20:0 GM1 (m/z 1572.8964), is mainly present in cerebral cortex and hippocampus, less expressed in cerebellum; however, d18:1/20:0 GD1 (m/z 1863.9927) is mostly found in cerebral cortex, barely present in hippocampus and cerebellum. For these gangliosides together with GalNAcβ 1-4(NeuGcα 2-3)Galβ 1-3GalNAcβ 1-4Galβ 1-4Glcβ -Cer(d18:1/20:0) (m/z 1791.8146) and NeuAcα 2-8 NeuAcα 2-8NeuAcα 2-3Galβ 1-4Glcβ -Cer(d18:1/26:0) (m/z 1874.0472) and 9-OAc-NeuAcα 2-8NeuAcα 2-8Neu Acα 2-3Galβ 1-4Glcβ -Cer(d18:1/24:1(15Z)) (m/z 1886.0112), the detailed distribution information in cerebral cortex, hippocampus and cerebellum from C57BL/6 control mice, as representatives, was showed in Fig. 3.

MALDI MS Analysis of Gangliosides.
The matrix is a critical factor for MALDI MS analysis, and therefore we first optimized the matrix for ganglioside analysis. Among the vast majority of commercially available matrices, DHB was a widely accepted matrix compound for the analysis of lipids since it provides an excellent signal-to-noise (S/N) ratio for the peaks of the analyte of interest and provides a weak background 23,24 . 3-AQ, first applied by Metzger et al. for the analysis of plant inulins, is described as superior to DHB in terms of sensitivity and resolution in negative ion mode 25,26 . 9-AA is recently developed for negative charged lipids detection 27 . nH has also been reported for negative-ion MALDI of neutral oligosaccharides 28,29 . Therefore, these matrices were chosen to screen for the best matrix for gangliosides analysis, and the mass spectra obtained by MALDI-FTICR MS analysis of mouse brain extracts were shown in Fig. 4. Not like ESI MS, the gangliosides purely appeared as deprotonated ions ([M-H] − ) by MALDI MS analysis. It was also observed that using 3-AQ as matrix, the mass spectrum with more gangliosides MS signals and more excellent S/N ratio was acquired. Moreover, some gangliosides MS signals, for example, the signals with m/z at 1791.8146, 1820.0051, 1843.9986, etc, which cannot be detected by UHPLC-ESI-FTICR MS were observed. By matching the precise precursor ion obtained by FTICR MS with free online databases and MALDI MS/MS spectra ( Supplementary Fig. S9), gangliosides were identified and listed in Table 1. These promising features allow the matrix 3-AQ to be well-suited for MALDI MS analysis of gangliosides in negative mode, and therefore the 3-AQ was used as the matrix for MALDI MSI analysis of ganglioside in brain section.
The sample preparation is important for MALDI MSI analysis. We first tried to use the 3-AQ as matrix for MALDI MSI analysis of ganglioside in brain section (Fig. 5B). Unfortunately, there was nearly no gangliosides signal but full of phosphatidylserines (PSs) and phosphatidylinositols (PIs). One work reported that the phospholipid, like phosphatidylcholine, would suppress other lipid's detection, and a NH 4 Ac wash removes contaminants in tissue, enhances the overall spectral quality, and benefits additionally in profiling of biological molecules in tissue 30 . Therefore, different solvents, including chloroform, methanol and ethanol were used to clean the tissue section. It was found that a pretreatment of cleaning by 70% EtOH for 30 s and 95% EtOH again for 30 s would greatly    (Fig. 5D). Based on the above, we had brain tissue section pretreated by 70% EtOH, 95% EtOH each for 30 s and 3-AQ/125 mM ammonium formate as matrix, which gave rise to more and higher gangliosides MS signals, like d18:1/18:0 GQ1(m/z 2418.1509) and d18:1/20:0 GQ1(m/z 2446.1805) (Fig. 5A). However, these pretreatments may cause problems. One concern is that the application of EtOH washing step would remove some gangliosides from brain section. So the EtOH after washing were collected, dried and re-suspended by 10 μ L of a mixture of methanol and water (1:1, v/v) for analysis by MALDI FTICR MS in negative ion detection mode by using 3-AQ as matrix. The result showed ( Supplementary Fig. S10) that no gangliosides were detected, showing that the application of EtOH washing step didn't remove gangliosides. Another concern is that the EtOH treatment would affect the distribution of gangliosides in the brain section. Therefore, the effect of EtOH on the distribution of gangliosides was carefully checked. The results were shown in Supplementary Fig. S11. The distribution of gangliosides with or without EtOH cleanup was almost identical. After cleanup by EtOH, the MS signals of gangliosides were much higher (Fig. 5A). Taken together, EtOH cleanup was adopted. Using MALDI MSI, the distributions of gangliosides in cerebral cortex, hippocampus and cerebellum of C57BL/6 control mice brain of about 20 gangliosides were acquired, and the mappings of gangliosides with m/z at 1544.8643, 1572.8964, 1791.8146, 1863.9927, 1874.0472 and 1886.0112, as representative, were shown in Fig. 3. These results were consistent with those obtained by UHPLC-ESI-FTICR MS, suggesting that the data obtained by MALDI MSI was reliable.

MALDI MSI Analysis of Gangliosides in Brain with Alzheimer's Disease (AD). In this work, ESI
and MALDI mass spectrometry were combined for the analysis of ganglioside composition and their corresponding change in brain. It was reported that ganglioside metabolism is closely associated with the pathology of Alzheimer's disease (AD) 5,7 . Based on these studies and our research purpose, amyloid precursor protein (APP) transgenic mouse model, a well-known AD mouse model, was used.
APP has been known to act as an important protein during the occurrence of AD. To explore the brain ganglioside alterations in AD, 5-month old mice (3 pairs), total 6 mice, were used for MALDI MSI analysis. The MALDI MSI results were very consistent and shown in Fig. 6. In the case of AD group, we can clearly find that 20 kinds of gangliosides in cerebellum were nearly disappeared. After careful observation, we also found that lots of gangliosides with m/z at 1572.8964, 1791.8146, 1820.0051, 1835.9602, 1858.0568, 1863.9927, 1878.0433, 1886.0112, H&E staining experiments were performed. The histopathological features (Fig. 7) from the examination showed that amyloid β (Aβ ) protein were accumulated in right cerebral hemispheres of APP transgenic mouse brain. In addition, compared with the cerebellum of control mouse, neuronal cells (blue dot) become significantly reduced in APP transgenic mouse. The damage originated from the amyloid β (Aβ ) protein accumulation in right cerebral hemispheres, and the lack of neuronal cells in cerebellum, might be the reason that the gangliosides were reduced or disappeared in right cerebral hemispheres and cerebellum of APP transgenic mouse brain, respectively. The low unsaturated gangliosides content in right cerebral hemispheres of APP mice may also be a result of Aβ -mediated lipid oxidation. Unsaturated lipids in the brain are readily attacked by free radicals because of their double bond content, becoming oxidized into lipid peroxides 31 . Aβ , the primary component of amyloid plaques, has been implicated in oxidative damage through lipid peroxidation in synaptic plasma membranes 32,33 . Indeed, increased lipid peroxidation has been observed in human AD, mild cognitive impairment, and in transgenic mouse models of AD [34][35][36] , suggesting that antioxidant defenses are impaired in AD 37 . These changes of gangliosides in different area of the brain could contribute to the development of the AD. The present data strongly suggest that application of this technology to gangliosides analysis from pathological brain may provide new insight for bioresearch of AD.

Discussion
A workflow by combination ESI with MALDI MS, for qualitative, semi-quantitative and in situ analysis of gangliosides in brain was developed. The extraction efficiency of gangliosides by Bligh and Dyer method was investigated, and the efficiency was from 35.5% to 89.2%, depending on the structure and property of the gangliosides. A FTICR MS was selected for this study due to the high mass accuracy and resolution, which greatly improved the accuracy of qualitative analysis. The sample preparation was optimized for MALDI MSI analysis, and it was found that 3-AQ (30 mg/mL)/ammonium formate (125 mM) as the matrix, and pretreatment by EtOH cleaning brain section could be used for MALDI MSI analysis of ganglioside in brain section. The result of gangliosides distribution obtained by MALDI MSI was confirmed by UHPLC-ESI MS, suggesting that the data obtained by MALDI MSI was reliable.
To explore the potential feasibility of our approach, the brain ganglioside alterations in Alzheimer's disease mouse were investigated. The distribution of gangliosides in brain was obtained, which firstly showed that gangliosides were almost nonexistent in the cerebellum of amyloid precursor protein (APP) transgenic mouse brain, in addition, most of gangliosides had lower distribution in right cerebral hemispheres of APP group. The altered distribution of gangliosides may be related with the damage originated from the amyloid β (Aβ ) protein accumulation in right cerebral hemispheres, and the lack of neuronal cells in cerebellum, which may provide a new insight for bioresearch of Alzheimer's disease (AD).
To the best of our knowledge, this is the first report to combine ESI with MALDI MS for analysis of gangliosides. The ESI and MALDI MS complement and verify each other to obtain more comprehensive and more accurate data as far as possible. This workflow is believed to maximize the advantages of ESI with MALDI MS and to complement or verify with each other, and is expected to provide great help for brain research.  from Sigma-Aldrich (St. Louis, MO, USA). Ultra-pure water was obtained from a Milli-Q purification system (Millipore Corporation, USA). All of the above materials were used as received without further purification.
Animals. The amyloid precursor protein (APP) transgenic mice and controls (C57BL/6 mice, 5 months old, Female, 6 pairs) were provided by Dr. Weidong Yong in the institute of Laboratory Animal Science, Chinese Academy of Medical Sciences. The mice were sacrificed by suffocation of CO 2 , and their brains were immediately surgically removed. The cerebral cortex, hippocampus and cerebellum tissues were manually separated and quickly frozen in liquid nitrogen. The whole brain from APP transgenic mice and controls were as well quickly frozen in liquid nitrogen, or fixed by 10% formalin solution for hematoxylin and eosin (H&E) staining. These tissues were stored at − 80 °C until use. The animal experiments were performed according to the 'Guide for the Care and Use of Laboratory Animals' and were approved by the Animal Care and Use Committee of the Chinese Academy of Sciences.
Lipid Extraction. Frozen tissues, including cerebral cortex, hippocampus and cerebellum, were separately homogenized by using Tissuelyser-24 (Shanghai Jingxing Experimental Technology, Shanghai). Gangliosides were extracted by Bligh and Dyer method 22,38,39 . In brief, 100 μ L of homogenate (water solution, totally 40 mg of the tissue) were added into 750 μ L of mixture of chloroform-methanol (1:2, v:v), and then incubated in ice for 5 min. 250 μ L of chloroform and 250 μ L of H 2 O were then separately added. After vortexing for 1 min and centrifugation (9184 g, 5 min, room temperature), the upper phase was collected and the lower phase was re-extracted by adding 350 μ L of H 2 O. The upper phase were combined and dried with N 2 . The gangliosides extracts were re-suspended in 100 μ L of H 2 O for MS analysis.
Tissue Sectioning. Frozen brain tissue from the APP transgenic mice and controls was fixed atop a drop of saline on the cutting stage. All brains were sectioned at 12 μ m thickness using a Leica CM1950 cryostat (Leica Microsystems GmbH, Wetzlar, Germany) at − 18 °C and thaw mounted onto indium tin oxide (ITO) coated glass slides (Type Ι 1.1 mm/100ea, HST Inc., Newark, NJ, USA.). Tissue sections were washed in 70% EtOH for 30 s, and then in 95% EtOH for 30 s. The glass slides were then placed into a vacuum desiccator and dried for approximately 1 h before matrix application. Finally, 3-AQ (30 mg/mL) was dissolved in CH 3 CN-H 2 O (8:2, v:v) containing 125 mM HCOONH 4 , and sprayed on the tissue section by homemade electrospray-based matrix deposition device. The tissue section was used for MALDI MSI analysis. FTICR MS. FTICR mass spectrometric analysis was performed with a Bruker SolariX mass spectrometer equipped with a 9.4 T superconducting magnet and SmartBeam ™ laser optics. Mass calibrations were performed externally using ganglioside standards listed in Chemicals and Reagents. External ion accumulation was used in negative ion mode over a mass range of 200-3000 m/z with a resolution of 130000 at m/z 400. SolariX Control software was used for data acquisition.
For MALDI MS analysis, 160 laser shots were acquired per sample. For MALDI MSI analysis, mass spectra were acquired across the entire sample section with a SmartBeam II laser operating at 1000 Hz, a laser focus of 25 μ m, 3 scans acquired from each matrix spot, and a raster width of 200 μ m. The device parameters for both MALDI MS and MALDI MSI were chosen as follows: plate offset voltage, 100 V; deflector plate voltage, 180 V.
For ESI MS analysis, both the nebulizer and dry gases were nitrogen. Typical operating parameters were set as follows: capillary voltage, 3000 V; dry gas flow rate, 4.0 L/min; dry gas temperature, 200 °C; nebulizer gas flow rate, 1.0 bar; syringe flow rate, 180.0 μ L/h. A ultra high performance liquid chromatography (UHPLC) instrument, Ultimate 3000 (Thermo), was used for loading samples and to separate gangliosides before ESI MS detection. The mobile phase A was isopropanol-acetonitrile-formic acid (90: 10: 0.1, v/v/v) containing 10 mM ammonium formate. The mobile phase B was acetonitrile-water-formic acid (70: 30: 0.1, v/v/v) containing 10 mM ammonium formate. A BEH C18 column (1.7 μ m, 2.1 mm ID × 100 mm, Waters) was used for separation of gangliosides. The column was maintained at 65 °C. The UHPLC separations were 20 min per sample using the following scheme: Structural Identification. High-resolution MS spectra were used to distinguish different gangliosides, and the identification of gangliosides was achieved by precisely matching mass with free online databases, including Lipidmaps (http://www.lipidmaps.org) (accessed July 2015) and HMDB (http://www.hmdb.ca) (accessed July 2015) [40][41][42] . The mass error was set at 5 ppm. In addition, standard lipids were also used for assistance of identification of the gangliosides. MS/MS spectra obtained by collision induced dissociation (CID) were further used for confirmation of the structure of the gangliosides.