Strategy for Comprehensive Profiling and ... - ACS Publications

Subscriber access provided by University of Rochester | River Campus & Miner Libraries

Article

A Novel Strategy for Comprehensive Profiling and Identification of Acidic Glycosphingolipids Using Ultra-High-Performance Liquid Chromatography Coupled with Quadrupole Time of Flight Mass Spectrometry Ting Hu, Zhixin Jia, and Jinlan Zhang Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.7b02023 • Publication Date (Web): 23 Jun 2017 Downloaded from http://pubs.acs.org on June 23, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

A Novel Strategy for Comprehensive Profiling and Identification of Acidic Glycosphingolipids Using Ultra-High-Performance Liquid Chromatography Coupled with Quadrupole Time of Flight Mass Spectrometry Ting Hu, Zhixin Jia, Jin-Lan Zhang* State Key Laboratory of Bioactive Substance and Function of Natural Medicines, Institute of Materia Medica, Peking Union Medical College & Chinese Academy of Medical Sciences, Beijing 100050, PR China *Corresponding author: Fax: +86 10 63017757; E-mail: [email protected]

ABSTRACT: Acidic glycosphingolipids (AGSLs), which mainly consist of ganglioside and sulfatide moieties, are highly concentrated in the central nervous system. Comprehensive profiling of AGSLs has historically been challenging because of their high complexity and the lack of standards. In this paper, a novel strategy was developed to comprehensively profile AGSLs using ultra-high-performance liquid chromatography-quadrupole time of flight mass spectrometry (UPLC-Q-TOF-MS). Ganglioside isomers with different glycan chains such as GD1a/ GD1b were completely separated on a C18 column for the first time to our knowledge, facilitated by the addition of formic acid in the mobile phase. A mathematical model was established to predict the retention times (RTs) of all theoretically possible AGSLs based on the good logarithmic relationship between the ceramide carbon numbers of the AGSLs in the reference material and their RTs. A dataset was created of 571 theoretically possible AGSLs, including the ceramide carbon numbers, RTs and high-resolution quasi-molecular ions. A novel fast identification strategy was established for global AGSL profiling by comparing the high-resolution quasi-molecular ions and RTs of the tested peaks to the dataset of 571 AGSLs. Using this strategy, 199 AGSL candidates were identified in rat brain tissue. MS/MS fragments were further collected for these 199 candidates to confirm their identity as AGSLs. This novel strategy was employed to profile AGSLs in brain tissue samples from control rats and model rats with bilateral common carotid artery (2-VO) cerebral ischemia. Forty AGSLs were significantly different between the control and model groups, and these differences were further interpreted.

ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

INTRODUCTION Acidic glycosphingolipids (AGSLs), which mainly consist of a ganglioside and sulfatide, are members of the sphingolipid class according to the LIPID MAPS classification system.1 Gangliosides are the most complex and diverse lipid species and are composed of a hydrophobic ceramide moiety and a hydrophilic glycan chain with one or more sialic acids.2 Over 1500 different gangliosides are contained in the LIPID MAPS database.1 Gangliosides, which account for over 10% of the total brain lipids,3 are components of the cell plasma membrane that are highly concentrated in the central nervous system.4 The main gangliosides within adult human brain tissue are reported to be GM1, GD1a, GD1b and GT1b.5 Sulfatides, also known as 3-O-sulfogalactosylceramides, are a class of sulfolipids that contain a sulfate group coupled to a galactosyl ceramide.6 The diversity of sulfatides mainly results from the variation in the structure of the fatty acyl chain and sphingoid base in the ceramide portion of the molecule. Sulfatides are primarily found in the outer leaflets of the plasma membranes of cells, which are also particularly abundant in brain tissue.6-8 Because of their high concentration in the central nervous system, the role of AGSLs in neurodegenerative diseases has garnered wide attention from scientists. AGSLs are intimately involved in the development of brain diseases such as Alzheimer’s disease, Parkinson’s disease and cerebral ischemia.9-12 AGSLs have already become an attractive target for biomarker discovery because of their physiological significance. Because of this physiological significance, significant research efforts in past decades have focused on the robust and reliable identification and quantification of AGSLs. Thin-layer chromatography (TLC) coupled with densitometry is a conventional method for the quantification of AGSLs.13,14 However, this method has been recently outpaced by other technologies due to its low resolution, which is inadequate for AGSL detection in complex matrixes such as bodily fluids and tissues. Enzyme-linked immunosorbent assay (ELISA) has also been explored for the quantification of gangliosides based on a variety of monoclonal and polyclonal antibodies that recognize specific types of sialic acids and linkages, but only in the presence of specific underlying glycan chains.13,15,16 All these methods quantify the total content of a ganglioside species that possesses the same glycan chain, but they cannot distinguish or quantify gangliosides with identical glycan chains but different ceramide structures. High-throughput methods with better accuracy and precision and wider coverage need to be explored. In recent years, mass spectrometry (MS) has been widely explored for the identification and quantification of AGSLs due to its high resolution, sensitivity and accuracy.17-21 To better quantify AGSLs in complex matrixes, a preliminary separation device can be coupled with MS to achieve a better analysis result. The coupling of liquid chromatography (LC) with MS via an electrospray ionization source (LC-ESI-MS) is a mainstream AGSL analysis method with excellent throughput, separation efficiency, sensitivity and selectivity. LC-ESI-MS can provide information on not only the ceramide lipid portion of the molecule but also the glycan group. Reverse phase (RP) columns, which separate AGSLs based on their fatty acyl chain and sphingoid base, are the most commonly used mode for AGSL separation. However, RP columns are unable to separate ganglioside isomers with different glycan chains such as GD1a and GD1b.22 Therefore, hydrophilic interaction chromatography (HILIC) columns were introduced to separate AGSL isomers with different glycan chains.22 However, AGSLs with the same glycan chain but different ceramide portions cannot be separated on HILIC columns. AGSLs that differ by 1 or 2 Daltons might interfere with


Page 2 of 16

Page 3 of 16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


each other in low-resolution MS because of the overlap of their isotope peaks if complete chromatographic separation is not attained. This issue might be the greatest problem related to the HILIC separation of AGSLs. High-resolution MS including Fourier-transform ion cyclotron resonance (FT-ICR) and quadrupole time-of-flight (Q-TOF) MS are usually employed for the identification of AGSLs according to high-resolution mass data and collision induced dissociation (CID) fragments.20,23 AGSLs can be detected both in positive and negative MS mode.19,23 Generally, AGSLs are better ionized in negative ESI mode because of the presence of the sialic acid or sulfuric acid groups. Usually, diagnostic fragments of the glycan chain can be obtained under optimized CID conditions. However, diagnostic fragments of the sphingoid base cannot be produced in negative CID.22,23 Thus, the structures of the fatty acyl chain or sphingoid base cannot be interpreted in negative mode. To date, only a few studies have simultaneously quantitated gangliosides and sulfatides using a single LC-ESI-MS method.20,22 Moreover, because of the lack of individual AGSL standards, most of the quantitation methods have covered only the abundant gangliosides or sulfatides in the reference material. LC-MS methods with a higher coverage for the simultaneous profiling of both gangliosides and sulfatides need to be explored. In this paper, a novel strategy was developed to simultaneously profile and identify AGSLs in biological samples using ultra-high-performance liquid chromatography-quadrupole time of flight mass spectrometry (UPLC-Q-TOF-MS). Isomers of gangliosides with different glycan chains, such as GD1a and GD1b, were completely separated on a RP column for the first time, facilitated by the addition of formic acid in the mobile phase. A novel fast identification strategy that integrates the application of ceramide carbon numbers, retention times (RTs) and high-resolution quasi-molecular ions was established for global AGSL profiling in complex matrixes. A total of 199 AGSLs were identified in rat brain tissue using this strategy, and the newly developed AGSL profiling strategy was employed for the quantification of AGSLs in brain tissue samples from control rats and model rats with bilateral common carotid artery (2-VO) cerebral ischemia.

EXPERIMENTAL SECTION Materials Organic solvents of HPLC grade or higher were purchased from Mallinckrodt Baker Inc. (Philipsburg, NJ, USA). Formic acid of HPLC grade was obtained from TEDIA Company, Inc. (Fairfield, OH, USA). Ammonium formate of MS grade was purchased from Sigma-Aldrich (St. Louis, MO, USA). Ganglioside reference materials including GD3, GM3, GQ1b, GD1a, GM1 asialo (GM1A), and GT1b were purchased from Cayman Chemical Co. (Ann Arbor, MI, USA). GM1 was purchased from Avanti Polar Lipid, Inc. (Alabaster, AL, USA). Reference materials of fucosylated GM1 (FGM1), GM2, GM4, and GD1b and total sulfatides from bovine brain were purchased from Matreya, LLC. (State College, PA, USA). Isotopic internal standards (ISs) including d3-GM3 (d18:1/18:0) and d3-sulfatide (d18:1/18:0) were also a product of Matreya, LLC.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Solutions and Standards Stock solutions of the AGSL reference materials, including GM1, GM2, GM3, GM4, GM1A, FGM1, GD1a, GD1b, GD3, GT1b, GQ1b, and total sulfatides, were prepared separately in chloroform/methanol/water (C/M/W, 2:1:0.1, v/v/v) to obtain a final concentration of 1 mg/mL for each component. Isotopic ISs including d3-GM3 (d18:1/18:0) and d3-sulfatide (d18:1/18:0) were dissolved in chloroform/methanol/water (C/M/W, 2:1:0.1, v/v/v) to obtain a 1 mg/mL IS stock solution for each isotopic IS. Both the standard solutions and isotopic ISs used in the whole experiment were prepared from the stock solution using methanol as the dilution solvent.

Construction of the 2-VO Rat Model Male Wistar rats (8 weeks old, 200±5 g body weight) were purchased from the Beijing Weitonglihua Experimental Animal Technology Co. Ltd. (Beijing, China). The rats were randomly divided into two groups (control group, n=8; model group, n=12) and acclimated for 2 weeks prior to modeling. The animals were housed under specific pathogen-free conditions (12 h light/12 h dark photoperiod, 23±2 °C, 55±5% relative humidity). The animal experiments were conducted in accordance with institutional guidelines and ethics and were approved by the Laboratories Institutional Animal Care and Use Committee of the Chinese Academy of Medical Sciences and Peking Union Medical College. After acclimation, the 2-VO model was constructed for the animals in the model group, and a sham operation was performed for the rats in the control group, as previously reported.24 Learning and memory abilities were examined using the Morris water maze task at three weeks after ligation, as previously described.25,26 The criteria for the evaluation of the 2-VO model were the same as those described by Yu et al.25 A total of 17 rats meeting the criteria (control group, n=7; model group, n=10) were screened. On the 58th day, all rats were sacrificed. After the blood was collected, brain tissues were collected and immediately frozen in liquid nitrogen. All biological samples were stored at -80 °C until further analysis.

Biological Sample Pretreatments Extraction of AGSLs from rat brain tissue was performed based on the Folch lipid extraction method.26 Brain tissue samples were first homogenized in an 8-fold volume of saline. An aliquot of 100 µl of brain tissue homogenate was added to the bottom of a 10 ml glass tube. Next, 100 µl of a 2 µg/ml IS solution in methanol was added to each tube. Then, 4 ml of chloroform/methanol (C/M: 2:1, v/v) was added for whole lipid extraction. After dispersion, the whole mixture was agitated for 20 min in an orbital shaker at room temperature. Then, the mixture was washed with 0.8 ml of water. After vortexing for 2 min, the mixture was centrifuged at 2000 rpm for 20 min to separate the two phases. The upper layer was removed and transferred to another glass tube. Another 0.9 ml of methanol and 0.8 ml of water were added to the lower chloroform phase in succession. After vortexing for 2 min, the mixture was centrifuged at 2000 rpm for 20 min to separate the two phases. The upper phase was pooled together with the previous upper layer and evaporated to dryness under a nitrogen stream. The dried residue was redissolved in 100 µl of


Page 4 of 16

Page 5 of 16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


methanol. The final mixture was centrifuged at 13,000 rpm for 10 min in a centrifuge maintained at 4 °C to induce precipitation of residual proteins. The supernatant was collected for instrumental analysis.

Method Validation The sensitivity of the UPLC-Q-TOF-MS analysis was evaluated using two isotopic ISs because of the lack of individual AGSL standards. Quantitation of the AGSL compounds was based on the isotopic IS method. To investigate the linearity, mixed AGSL standard solutions at 100, 200, 500, 1000, 2000 and 5000 ng/ml were diluted from the stock solution with appropriate amounts of methanol. An aliquot of 100 µl of each mixed AGSL standard solution was mixed with 100 µl of a 2 µg/ml IS solution in methanol and subjected to the same extraction procedure as that in the biological sample pretreatment. The precision was investigated using brain tissue homogenate. Six replicate samples of brain tissue homogenate were pretreated and analyzed as previously described. After calibration with the corresponding IS, the precision was calculated as the relative standard deviation (RSD) of the AGSL quantification data in six samples pretreated in parallel.

Instrumental Analysis An Agilent 1290 series UPLC system coupled with an Agilent 6550 Q-TOF mass spectrometer (Agilent Technologies, Santa Clara, CA, USA) with a dual AJS electrospray ionization (dAJS-ESI) source was used for AGSL analysis. An Agilent Zorbax Eclipse Plus C18 column (2.1×50 mm, 1.8 µm particle size) was employed for chromatographic separation. Mobile phase A consisted of water with 0.2% formic acid and 10 mM ammonium formate. Mobile phase B consisted of methanol with 0.2% formic acid and 10 mM ammonium formate. The flow rate was 0.4 ml/min, and the column temperature was controlled at 45 °C. The injection volume was 10 µl. The solvent program was as follows (linear gradient): 0-3 min, 80-85% B; 3-17 min, 85-100% B; 17-19 min, 100-100% B; 19-19.01 min, 100-80% B; and 19.01-20 min, 80-80% B. The acquisition parameters for positive Q-TOF-MS detection were as follows: drying gas flow=14 l/min, drying gas temperature=200 °C, nebulizer=20 psi, sheath gas flow=11 l/min, and sheath gas temperature=250 °C. The spray voltage was 3,000 V. The mass range was from 600 to 2000 Daltons in the full scan data acquisition step. Targeted CID MS/MS data were acquired in positive mode with the collision energy (CE) set at 20, 40 and 60 eV.

Data Processing and Statistical Analysis Agilent MassHunter Q-TOF Qualitative Analysis B.07.01 and Quantitative Analysis B.07.01 software programs were employed for data processing. Identification of the AGSL compounds was based on their high-resolution MS data, their theoretical retention characteristics on the C18 column and their CID MS/MS data. Each AGSL compound was quantified as the peak area ratio of the AGSL and its corresponding IS from the corresponding extracted ion chromatograms (EICs). For the gangliosides, d3-GM3 (d18:1/18:0) was used as the calibration IS, and for the sulfatides,



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

d3-sulfatide (d18:1/18:0) was used as the calibration IS. IBM SPSS 21 (Armonk, New York, USA) was used for the bilateral t-test. SIMICA-P+ 12.0.1 (Umetrics AB, Sweden) was employed for principle component analysis (PCA). The open source tool MetaboAnalyst 3.0 (http://www.metaboanalyst.ca/) was employed for heat map generation.

RESULTS AND DISCUSSION Separation of Ganglioside Isomers with Different Glycan Chains on the RP Column It is commonly believed that ganglioside isomers with different glycan chains, such as GD1a and GD1b, can not be distinguished using an RP column via hydrophobic interactions.22 However, the RP column has been the most widely used in separation methods for the analysis of AGSLs because of its good separation efficiency and excellent compatibility. The exploration of isomer separation will be of great importance in the development of ganglioside profiling. Dozens of C8 and C18 columns with different lengths, inner diameters, particle sizes and producers were tested for the separation of AGSLs. An Agilent Zorbax Eclipse Plus C18 column (2.1×50 mm, 1.8 µm particle size) gave the best separation and peak shapes of the AGSLs after careful optimization of the elution gradient. Ammonium acetate or ammonium formate are the most commonly reported mobile phase additives for the separation of AGSLs. However, these additives are not capable of separating ganglioside isomers with different glycan chains. Different mobile phase additives were tested in this work to better separate the ganglioside isomers with different glycan chains. Finally, a combination of ammonium formate and formic acid luckily enabled the complete separation of ganglioside isomers with different glycan chains on the RP column. To our knowledge, this is the first report of the complete separation of ganglioside isomers with different glycan chains on a RP column. Gangliosides possessed a weak acid property because of the existence of sialic acids in the chemical structures. Lowering PH of the mobile phase through the addiction of formic acid may increase the value of separation factors of solutes with weak acid properties, which resulted in an increase of the selectivity.27 In addition, the dissociation of gangliosides with acid property was suppressed by the addition of formic acid in the mobile phase, and then improvement of their chromatographic retention and peak shape on RP column can be achieved. As shown in Figure 1, the selectivity of GD1a/ GD1b on C8 and C18 column was significantly improved by the addition of formic acid in the mobile phase. Resolution of RP column was determined by the column efficiency, selectivity and capacity factor (equation 1 in the Supporting information). Increase of column selectivity will unquestionable improve the resolution of RP column. Therefore, separation resolution of gangliosides can be improved by the addiction of formic acid in the mobile phase. As shown in Figure 1 (EICs in black color), peaks of GD1a and GD1b were almost overlapped under the mobile phase with only addiction of ammonium acetate. However, complete separation of GD1a 36:1/GD1b 36:1 and GD1a 38:1/GD1b 38:1 was achieved on C18 (Agilent Zorbax Eclipse Plus C18, 2.1×50 mm, 1.8 µm) and C8 (Agilent Poroshell 120 EC-C8, 2.1×50 mm, 2.7 µm) column with the addition of both formic acid and ammonium acetate in the mobile phase. The LC


Page 6 of 16

Page 7 of 16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


conditions were carefully optimized for both C18 and C8 column separation, and the final parameters were shown in the Supporting Information. Addition of formic acid in the mobile phase resulted in significantly improvement of gangliosides resolution on the RP column, leading to the first report of complete separation of ganglioside isomers with different glycan chains on the RP column in this work. High resolution MS data of rat brain sample acquired under two different mobile phase conditions was both induced for AGSL database searching according to the method described below. As we expected, the addition of formic acid in the mobile phase could significantly increased the AGSL candidates numbers detected in rat brain samples. A comparison of this method and the previously reported AGSL profiling methods was shown in Table S-1, this method had significant advantage in the chromatographic separation efficiency. Both separation of ganglioside isomers with different glycan chains and separation of AGSL homologues can be achieved under this optimized chromatographic condition.

Figure 1. EICs obtained using a mobile phase containing 10 mM ammonium formate (black peaks) or 10 mM ammonium formate and 0.2% formic acid (red peaks). Complete separation of ganglioside isomers with different glycan chains was achieved through the addition of formic acid in the mobile phase. (A) EICs obtained using an Agilent C18 column. (B) EICs obtained using an Agilent C8 column.

Comprehensive Identification of AGSLs in Reference Materials by UPLC-Q-TOF-MS Because of the lack of individual standards, high-resolution MS, which can provide both the high-resolution quasi-molecular ion and high-resolution MS/MS data, was the best choice for AGSL identification in the complex mixture. In this work, the AGSLs in the reference materials were comprehensively analyzed using positive-mode UPLC-Q-TOF-MS. Because of the deficiency of diagnostic fragments for the sphingoid base in negative-mode CID, positive mode MS was employed for AGSL identification. For each AGSL species with an identical glycan chain, high-resolution MS data were collected as the reference material in positive mode via UPLC-Q-TOF-MS in the first step. A computationally generated precursor ion m/z mass list ([M+H]+ or [M+2H]2+) including all theoretically possible AGSLs with ceramide carbon numbers that varied from 32 to 46 and degrees of unsaturation that varied from 0 to 2 was used to search



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 16

the MS data. Only those peaks with a mass error of less than 15 ppm compared to the theoretical m/z value were accepted as AGSL candidates. All the AGSL candidates with accurate m/z values and RTs were applied for targeted MS/MS data acquisition. By resolving the targeted MS/MS data, structural information, including the glycan chain, fatty acyl chain and sphingoid base can be determined. Therefore, the false positives of AGSL candidates can be excluded, and the structures of the final candidates can be elucidated. A total of 304 AGSLs was identified from the reference materials, and the compound numbers for each AGSL species are shown in Table S-3. The CID MS/MS fragmentation rules applied for the AGSLs are described in detail in the Supporting Information. Figure 2 shows the CID MS/MS mass spectrum of GD1a 38:1.

(A)

Y3α

Y4α Z4α

NeuAc

O

Gal

B1α

Cer d18:1-H2O Intens. x10 3 6 4

O

C3α Y2β

O

Z0

Glc

O

C1β C4

Cer 38:1

C5

B4 NeuAcB 1β

B5

Gal-Glc-Ger38:1-H2O

454.1512

Cer38:1

900.6639

NeuAc-Gal-Glc-Ger38:1-H2O

576.5639

292.1008

GalNAc

Gal

Z2β

1103.75557

NeuAc-Gal-H2O

NeuAc-H2O 274.0908 NeuAc

B 3α

O

Z1

GalNAc-Gal-Glc-Cer38:1-H2O

292.2980

264.2660

C2α B 2α

Y0

Y1 Z2α

GalNAc

O

C1α

Cer d20:1-H2O

Y2α Z3α

1191.7667

NeuAc-Gal-GalNAc-H2O

204.0859

Gal-GalNAc

657.2296

Gl c-Ger38:1-H2O [M +2H]2+

366.1369

2

738.6167

GalNAc-NeuAc-Gal-Glc-Cer38:1--H2O

933.5084

1394.8453

0 100

200

300

400

500

(B)

600

700

800

900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 m/z

Intens.

NeuAc -H 2O

x10 3 6

NeuAc

274.0908

Cer d20:1 -H 2O

4

GD1a 38:1

CE: 20 eV

292.1008

292. 2980

2 0 Intens.

NeuAc -H 2O

x10 3

MS/MS

CE: 40 eV

274.0904

7.5

Cer d20:1 -H 2O 292.2980

5.0

NeuAc Cer d18:1 -H 2O

2.5

Cer d20:1

292.1008

310.3066

264.2672

0 Intens.

Cer d20:1 -H 2O

x10 3

10

11 RT/min

12

CE: 60 eV

292.2980

4 NeuAc -H 2O Cer d18:1 -H 2O 274.0894

2

NeuAc

Cer d20:1

292. 100 9

310.3071

264.2687

0 250

255

260

265

270

275

280

285

290

295

300

305

310

315

m/z

Figure 2. (A) MS/MS mass spectrum of GD1a 38:1 in CID mode. The structure of each fragment ion was labeled in blue font on the corresponding m/z value. All fragment ions appeared as singly charged species, except for the precursor ion (m/z 933.5084, [M+2H]2+). The fragmentation rule of GD1a 38:1 was shown on the top right corner. (B) Abundance variation of the GD1a 38:1 diagnostic ions related to the sialic acid and sphingoid base under CEs of 20 eV, 40 eV and 60 eV. The red peak to the left corresponds to the EIC of GD1a 38:1.

Mathematical Model Established Based on the Logarithmic Relationship between the Ceramide Carbon Numbers of AGSLs and their RTs Based on a careful analysis of the characteristics of the AGSL compounds identified in the


Page 9 of 16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


reference materials, the ceramide carbon numbers of the AGSL had a good logarithmic relationship with their RTs on the RP column, with all the regression coefficients larger than 0.9993 (Table S-2). A mathematical model was established based on this logarithmic relationship to predict the RTs of all theoretically possible AGSLs. This model was also established for AGSL RT calibration after small changes were made to the pipeline of the UPLC system and to the mobile phase, with a deviation of less than 1%. The model could then be used to assist in the identification of AGSLs in biological samples. Here, the GM1 species was taken as an example to refine and establish the mathematical model. A total of 13 GM1 compounds were identified in the GM1 reference material based on their high-resolution quasi-molecular ions and CID MS/MS data obtained in the previous step. Table S-4 lists the information about the GM1 compounds identified in the GM1 reference material. These GM1 compounds were used to construct the mathematical model. A new relative retention time (RRT) was calculated for each GM1 compound by dividing its RT by the RT of GM1 36:1, which was selected as the reference compound because of its higher relative abundance in the GM1 reference material. An excellent logarithmic relationship was observed between the ceramide carbon number and the corresponding RRT for the GM1 compounds with identical double bonds. The logarithmic relationships among the GM1 compounds with 0, 1 and 2 double bonds are shown in Figure S-1A, B and C, respectively. The calculated relative retention time (CRRT) was determined for each GM1 compound by substituting the carbon number into the corresponding logarithmic equation. The accuracies of these logarithmic relationships were also investigated by dividing the CRRTs by the RRTs. All the accuracies were between 99-101%, reflecting the excellent logarithmic relationships between the ceramide carbon number and the corresponding RRT.

A Novel Fast Identification Strategy for Global AGSL Profiling The established logarithmic relationships were further applied to identify AGSLs in biological samples. Only 13 GM1 compounds were identified in the GM1 reference material, with the EICs for compounds containing 0, 1, and 2 double bonds shown in Figure 3A, B and C, respectively. However, more GM1 compounds should have existed in the brain tissue samples. The theoretically possible GM1 compounds with ceramide carbon numbers ranging from 32 to 46 and double bonds ranging from 0 to 2 are listed in Table S-5. The CRRTs were determined for all the GM1 compounds by inputting the carbon number into the corresponding logarithmic equation. The predicted retention time (pRT) of each theoretical GM1 compound was calculated as the product of the CRRT and RT of the reference compound (RT=10.207 in this example). The exact m/z values and their corresponding pRTs were introduced to search the MS data from the biological sample. Only peaks with mass errors less than 15 ppm compared to the theoretical m/z value and real RTs (rRTs) within ±3% deviation of the pRT were accepted as GM1 candidates in the biological samples. The GM1 candidates were further confirmed using targeted MS/MS data. As shown in Table S-5, all the errors between the pRTs and rRTs were less than 1%, which suggested that the pRTs of GM1 compounds obtained from the mathematical model fit well with their rRTs. Twenty-four GM1 compounds were identified in the brain tissue based on this mathematical model, 11 of which could not be detected in the GM1 reference material. The EICs



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

of the GM1 compounds containing 0, 1, and 2 double bonds identified in rat brain tissue were shown in Figure 3E, F and G, respectively. Logarithmic relationships between the ceramide carbon numbers and their RTs were observed for all the AGSLs but were not restricted to the GM1 species. A reference compound with a higher relative abundance in the reference material should be assigned to each AGSL species. By inputting the RT of the reference compound into the mathematical model in Microsoft Excel, the pRTs of all theoretically possible AGSLs of the same species can be calculated. Using this process, a dataset of 571 AGSLs including the ceramide carbon numbers, RTs and high-resolution quasi-molecular ions was assembled. In addition to the 304 AGSLs identified in the reference materials, 267 of 571 AGSLs were predicted using this novel strategy. This dataset can be ultimately used for the preliminary assessment of AGSLs in biological samples. Only peaks with mass errors of less than 15 ppm compared to the high-resolution quasi-molecular ions and rRTs within ±3% deviation of the pRTs were accepted as AGSL candidates in the biological samples. A total of 199 compounds were identified as AGSL candidates in the rat brain tissue in this preliminary assessment, and over one-third of them could not be detected in the reference materials. The total ion chromatogram (TIC) and the MS spectra of rat brain analysis were shown in Figure S-10. Targeted CID MS/MS data were further collected in positive mode for these 199 compounds for their final confirmation and demonstrated that all were indeed AGSL compounds. Therefore, the mismatch rate of this novel strategy was zero with mass error set as 15 ppm and rRT deviation set as 3%, all the candidates predicted by this fast identification strategy were indeed AGSL compounds. Both extending and narrowing the mass error or rRT deviation would bring the mismatch. Details of the mismatch rate were described in the “Mismatch rate discussion” section of the Supporting Information. As shown in Table S-1, more AGSL compounds can be detected from biological samples in this work compared to the previously reported methods, and this was owing to the excellent chromatographic separation efficiency and the novel fast identification strategy for global AGSL profiling. The structural information, mass accuracy and RTs of these 199 AGSLs were listed in the Microsoft Excel file in the Supporting Information. All the pRTs predicted by this mathematical model were within 1% deviation of the rRTs. The numbers of AGSL compounds distributed in each species were shown in Figure S-2. A workflow indicating the procedures for AGSL identification in biological samples was shown in Scheme 1. This novel strategy is especially suitable for efficient and accurate AGSL identification in complex matrix and can effectively identify AGSLs with lower abundances. In order to compare this novel strategy with the conventional RT prediction method, a Quantitative Structure-Retention Relationship (QSRR) model was built for each AGSL specie to predict the AGSLs in biological samples using the Discovery Studio (V4.0). For the QSRR model building, the Merck Molecular Force Field (MMFF) was first used for the energy optimization of each AGSL. After that, the 2 dimension (2D) properties including the ALogP, Molecular_Weight, Num_H_Donors, Num_H_Acceptors, Num_RotatableBonds, Num_Rings, Num_AromaticRings, Molecular_FractionalPolarSurfaceArea were calculated for each AGSL. The QSRR model was built based on the relationship existed between the 2D properties and the RRT of AGSLs using the Multiple Linear Regression Model. Here, the GM1 species was also taken as an example to clarify the QSRR model. All the 13 GM1 compounds identified in the GM1 reference material were included for the QSRR model built using Discovery Studio with regression curve, equation and correlation coefficient shown in Figure S-3. The QSRR model was then employed for the RT


Page 10 of 16

Page 11 of 16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


prediction of all the theoretically possible GM1 compounds with results shown in Table S-6. The errors between the pRTs and rRTs using QSRR model (Table S-6) were significantly larger than those using the logarithmic relationship based mathematical model (Table S-5). As shown in Table S-6, 11 of 24 GM1 compounds identified in rat brain tissue had errors larger than 1% using the QSRR model. From this perspective, mathematical model established based on the logarithmic relationship between the ceramide carbon numbers and their RTs possessed better accuracy, and can be better used for the global AGSLs profiling in biological sample compared to the QSRR model. Moreover, the mathematical model based on the logarithmic relationship was easy to established and can be accomplished in Microsoft Excel without the using of professional software. (A)

(B)

11

(C)

1

1 2

2 22

1. 2. 3. 4. 5. 6.

33 1. GM1 GM136:0 34:0 1. 2. 2. GM1 GM138:0 36:0 3. GM1 GM140:0 38:0 3.

GM1 GM1 GM1 GM1 GM1 GM1

3

34:1 35:1 36:1 37:1 38:1 40:1

3 4

1. 2. 3. 4.

5

4

GM1 36:2 GM1 38:2 GM1 40:2 GM1 42:2

6

0

5

10 10

15 15

RT/min RT/min

0

0

(E)

10

15

GM1 GM1 GM1 GM1 GM1 GM1

34:0 36:0 37:0 38:0 40:0 42:0

3

4

5

10

15

RT/min

(G)

1. GM1 GM1 34:0 1. 34:1 2. GM1 GM1 35:0 2. 35:1 3. GM1 GM1 36:0 3. 36:1 4. GM1 GM1 37:0 4. 37:1 5. GM1 GM1 38:0 5. 38:1 6. GM1 GM1 39:0 6. 39:1 7. GM1 GM1 40:0 7. 40:1 8. GM1 GM1 41:0 8. 41:1 9. GM1 GM1 42:0 9. 42:1 10. GM1 GM1 43:1 43:0 10. 11. GM1 GM1 44:1 44:0 11.

2

5

RT/min

(F) 1

1. 2. 3. 4. 5. 6.

5

1

1

2

2 1. 2. 3. 4. 5. 6. 7.

3 4 5 6

6

GM1 36:2 GM1 38:2 GM1 39:2 GM1 40:2 GM1 42:2 GM1 43:2 GM1 44:2

3 4 5 6

7

7 8 9 10

0

5

10

15

11

RT/min

0

0

5

10

15

5

10

15

RT/min

RT/min

Figure 3. EICs of GM1 compounds identified in the GM1 reference material (A, B, C) and rat brain tissue (E, F, G). (A, E) GM1 compounds with no double bonds in the ceramide portion. (B, F) GM1 compounds with 1 double bond in the ceramide portion. Peak 3 shown in red corresponds to the reference compound for the GM1 class. (C, G) GM1 compounds with 2 double bonds in the ceramide portion. Peak information is listed in the left corner of each panel.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Scheme 1. Flow chart of the novel fast identification strategy for global AGSL profiling.

AGSL Profiling of Brain Tissue in the 2-VO Rat Model The newly constructed UPLC-Q-TOF-MS method was further employed for the quantification of AGSLs in rat brain tissue. The quantification performance of the new method was investigated, and the results are shown in the Supporting Information. The satisfactory results achieved for the linearity and precision guaranteed the reliability of the AGSL quantification. The method was employed for AGSL profiling in rat brain tissue in control rats and 2-VO model rats. A total of 199 AGSLs, with information listed in the Supporting Information, were identified and quantified in rat brain tissue. The majority of the detected AGSLs contained C18- and C20-sphingosine in their ceramide portion. The content distribution of each ganglioside species is shown in Figure S-4A. As shown in the figure, GD1a, GD1b, GM1 and GT1b were the most abundant ganglioside species in rat brain tissue. The total content of GD1a, GD1b, GM1 and GT1b accounted for 86.1% of the total ganglioside content. The results achieved in this work were highly consistent with the results reported in previous works.5,28 Considering sulfatide, the content distribution of these AGSL species is shown in Figure S-4B. As shown in the figure, the sulfatide and gangliosides concentrations were almost the same in the brain tissue. Statistical analysis was carried out using the AGSL quantification data. A bilateral t-test was first conducted to sieve the most obviously changing AGSLs among the control and 2-VO model groups. Forty AGSL compounds were significantly different (P

Strategy for Comprehensive Profiling and ... - ACS Publications

Recommend Documents