A comprehensive method for lipid proﬁ ling by liquid chromatography-ion cyclotron resonance mass spectrometry

This work aims to combine chromatographic retention, high mass resolution and accuracy, MS/MS spectra, and a package for automated identiﬁ cation and quantitation of lipid species in one platform for lipidomic analysis. The instrumental setup elaborated comprises reversed-phase HPLC coupled to a Fourier transform ion cyclotron resonance mass spectrometer (LTQ-FT), and Lipid Data Analyzer (LDA) software. Data analysis for lipid species quantiﬁ cation in this platform is based on retention time, mass resolution of 200,000, and mass accuracy below 2 ppm. In addition, automatically generated MS/MS spectra provide structural information at molecular level. This LC/MS technology allows analyzing complex biological samples in a quantitative manner as shown here paradigmatically for murine lipid droplets having a huge surplus of triacylglycerol species. Chromatographic preseparation of the bulk lipid class alle-viates the problem of ion suppression of lipid species from other classes. Extension of 1D to 2D chromatography is possible, yet time consuming. The platform affords unambiguous detection of lipid species as low as 0.1‰ within major lipid classes. Taken together, a novel lipidomic LC/MS platform based on chromatographic retention, high mass resolution and accuracy, MS/MS analysis, and quantitation software enables analysis of complex samples as demon-strated for lipid droplets. and A comprehensive method for lipid proﬁ ling by liquid chromatography-ion cyclotron resonance mass spectrometry. J. Lipid Res. 2011. 52: internal Lipid Analyzer; LM, LIPID MAPS; LPE, lysophos-phatidylethanolamine; LPC, lysophosphatidylcholine; LTQ-FT, hybrid linear ion trap Fourier transform ion cyclotron resonance mass spectrometer; MG, monoacylglycerol; NL, neutral loss; PA, phosphatidic acid; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PI, phosphatidylinositol; PS, phosphatidylserine; Q-Trap, quadrupole linear ion trap mass spectrometer; SM, sphingomyelin; S/N, signal to noise; TG, triacylglycerol; 1D, one dimensional; 2D, two dimensional; 3D, dimensional. species is

Cyclohexane and chloroform were HPLC grade, and ammonium acetate and acetic acid were analytical grade, all obtained from Merck KGaA (Darmstadt, Germany). 2-Propanol was LC/MS grade and supplied by Fluka (Steinheim, Germany). Methanol and acetonitrile were LC/MS grade, methyl tert-butyl ether (MTBE) was HPLC grade, and 28% ammonia p.a. were all purchased from Sigma-Aldrich Chemie GmbH (Steinheim, Germany). Nitrogen (purity 5.0) was obtained from Air Liquide (Graz, Austria). Ultrapure water purifi ed by a Milli-Q Gradient system (Millipore, Bedford, MA) was used in all experiments (resistivity > 18 M ⍀ cm).

Isolation of lipid droplets from murine primary hepatocytes
A pool of lipid droplets isolated from hepatocytes of mice was split into three equal parts and used as representative biological examples for lipid droplets. Primary hepatocytes from C57 black mice (C57BL) were isolated according to literature ( 28 ). The hepatocyte pellet obtained was resuspended in 5 ml ice-cold disruption buffer (20 mM potassium phosphate, pH 7.4, 250 mM sucrose, 1 mM EDTA, 1 mM PMSF), and incubated cells were lysed under nitrogen cavitation at 800 psi for 10 min in a nitrogen bomb (Parr Instrument Co., Moline, IL). The resulting homogenate was centrifuged at 1,000 g for 5 min at 4°C to remove cell debris. The supernatant obtained was overlayed with buffer (50 mM potassium phosphate, pH 7.4, 100 mM potassium chloride, 1 mM EDTA, 1 mM PMSF) and centrifuged at 100,000 g for 1 h at 4°C. Lipid droplets concentrated in a white band at the top of the tube and were used for further experiments.

Lipid extraction
Lipid extraction was carried out as described ( 29 ). Briefl y, 2 ml of the lipid droplet suspension (TG concentration 2.17 mg/ml) was placed in a glass tube with a tefl on-lined cap. A volume of 3 ml methanol and then 10 ml MTBE was added, and tubes were shaken for 1 h at room temperature. Upon addition of 2.5 ml deionized water and shaking, phase separation was induced. The upper organic phase was collected, the lower aqueous phase reextracted with MTBE, and the upper phases were combined. The solvent was evacuated in a SpeedVac (Thermo Fisher Scientifi c, San Jose, CA), and lipids were redissolved in 4 ml chloroform/ methanol 1:1 (v/v).

Identifi cation and quantitation
Dilution factors need to be adapted to expected lipid class concentrations in the sample and to the instrument setup in use. In case of TG class determination lipid droplet samples were diluted 1:57 with chloroform/methanol 1:1 (v/v); in case of determination of all other lipid classes together 1:3 with chloroform/ methanol 1:1 (v/v). In both cases, the solvent was evacuated in a SpeedVac and lipids were resuspended in 100 µl chloroform/ methanol 1:1 (v/v) for further analysis. Then lipid extracts were spiked with multiple LM quantitative internal standards (IS) for at a resolution of 50,000, including two data-dependent MS/MS spectra per cycle ( 19 ). In addition, hydrophilic interaction chromatography (HILIC), a modifi ed normalphase chromatography coupled to LC/ESI-MS, allows analyzing cardiolipins (CL) and bis(monoacylglycero) phosphates (BMP) ( 22 ). Increased chromatographic selectivity is obtained by a two-dimensional (2D) separation system of normal-phase and reversed-phase chromatography. Lipid profi les from rat peritoneal surfaces have been reported recently after moving to an online 2D LC/MS noncommercial approach ( 15 ).
Taken together, most platforms use MS/MS spectra and exact mass ( 10 ), retention time and specifi c fragments ( 24 ), or retention time and exact mass ( 25 ). It is our goal to combine all these parameters for the highest possible level of confi dence in one platform and to link it to a custom-developed software package for automated identifi cation and quantitation ( 7 ).
Lipid droplets are long-neglected cell organelles, composed of a hydrophobic lipid core of triacylglycerols (TG) and sterol esters surrounded by a monolayer of phospholipids, sphingomyelins (SM), and cholesterol. They emerged as a central hub for TG metabolism, among many other metabolic and signaling features ( 26 ). The huge concentration difference between TG and the other lipid species ( 27 ) makes lipid droplets a highly challenging analytical matrix due to ion suppression effects affected by bulk TG. An effi cient chromatographic system and subsequent analysis of lipid species by an ion cyclotron is tested by analyzing this matrix. Moreover, the ultrahigh mass spectrometric resolution delivered by an ion cyclotron should help to reduce the complexity of lipid droplet samples, which would be a fundamental advantage over systems with lower resolution. ,000 resolution ( m/z 400) from m/z 400 to 1,050 in positive and from m/z 350 to1,050 in negative ESI mode. Helium was used as gas for linear ion trap collision-induced dissociation (CID) spectra. From the LTQ-FT preview scan, the four most abundant ions were selected in data-dependent acquisition (DDA), fragmented in the linear ion trap analyzer, and ejected at nominal mass resolution. The following parameters were used for positive and negative ESI-MS/MS experiments: normalized collision energy was 35%, the repeat count was 2, and the exclusion duration 60 s. The activation Q was at 0.2, and the isolation width 2. For positive ESI, spray voltage was set to 5 kV, and the tube lens offset was at 120 V. For negative ESI, spray voltage was Ϫ 4.8 kV, and the tube lens offset was Ϫ 87 V. The sheath gas fl ow was set to 50 arbitrary units, auxiliary gas fl ow to 20 arbitrary units, sweep gas fl ow to 2 arbitrary units, and the capillary temperature to 250°C.

Data analysis
Identifi cation and quantitation of lipids was performed by LDA, a platform-independent Java application ( 7 ). Briefl y, the algorithm identifi es peaks in the three dimensional (3D) LC-MS data space (retention time, m/z , and intensity). It determines the peak borders in m/z as well as in time dimension, and integrates the intensities within the borders. Furthermore, the algorithm uses a theoretically calculated isotopic intensity distribution as peak selection criterion to improve the specifi city.

RESULTS
It is the aim of this study to establish an integrated MSbased analytical platform, relying on chromatography, ultrahigh mass resolution, MS/MS fragmentation, and automated data processing ( Fig. 1 ). We develop a hybrid approach by optimizing both reversed-phase-HPLC and subsequent MS parameters. In addition, we extend reversed-phase chromatography by a HILIC step (2D LC/ MS) that may enhance MS sensitivity for lipid classes by minimizing ion suppression. In this process, we test platform performance by analyzing profi les of lipid species present in pooled lipid droplets isolated from mouse hepatocytes.

Reversed-phase-HPLC development
The particular challenge is the high excess of TG in the samples. ESI was used because of its extraordinary sensitivity for monitoring lipid species of TG, diacylglycerols (DG), phosphatidylcholines (PC), lysophosphatidylcholines (LPC), phosphatidylethanolamines (PE), phosphatidylserines (PS), and SM in positive mode and phosphatidylinositols (PI) in negative mode. Phosphatidic acids (PA), phosphatidylglycerols (PG), CL, lysophosphatidylethanolamines (LPE), and ceramides (Cer) were searched for, but they are not detectable in positive or negative ionization mode. Cholesterol and cholesterol esters were not determined, as analysis by a separate silicabased, normal-phase chromatography method with postcolumn addition of a polar solvent is the preferred experimental approach ( 18 ).
The method developed here is a compromise between chromatographic resolution and running time as shown in Fig. 2 . The chromatogram reveals that in the fi rst 10 min, the most-polar lipids, like LPC, elute, followed by each lipid class to monitor. Five replicas of extracted lipid samples were prepared for lipid quantifi cation as well as for accuracy and precision analysis. Accuracy and precision evaluation was performed by spiking fi ve replicas of extracted lipid droplets with 350 pmol TG 17:0/17:0/17:0, 130 pmol PC 12:0/12:0, 270 pmol PE 12:0/12:0, and 220 pmol PS 12:0/12:0. For further determination of the dynamic range , three replicas were spiked with these four standards at nine concentration levels, each ranging from 0.02 to 328 µmol/µl per standard.

Chromatographic methods
High-performance liquid chromatography. The Accela HPLC system was equipped with a reversed-phase C18 column (reversed-phase C18; 100 × 1 mm i.d., 1.9 µm particle size), both from Thermo Fisher Scientifi c (San Jose, CA). Mobile phase A was 10 mM ammonium acetate containing 0.1% formic acid. Mobile phase B was acetonitrile/2-propanol 5:2 (v/v) containing 10 mM ammonium acetate and 0.1% formic acid. The binary gradient started with 35 to 70% B for 4 min, then was raised up to 100% B in another 16 min and further held for 10 min. The fl ow rate was 250 µl/min, the oven temperature was 50°C, and tray temperature 10°C. For analysis, 5 µl samples were injected. After each run, the column was fl ushed 5 min with 35% B before the next run was started.
High-performance liquid chromatography for 2D chromatographic separation. HILIC-HPLC was carried out in an Agilent 1100 HPLC system (Waldbronn, Germany) consisting of a degasser, a binary pump, a thermostated autosampler, and a thermostated column compartment equipped with a semipreparative column, fi lled with Nucleosil 100-5 OH, 250 × 10 mm i.d., 5.0 µm particle size (Macherey-Nagel, Düren, Germany). Mobile phase A was cyclohexane, and mobile phase B was 2-propanol/deionized water/acetic acid/28% of ammonia 86:13:1:0.12 (v/v). The HPLC fl ow rate was 1,000 µl/min at an isocratic composition of 10% A and 90% B, at 35°C oven and 5°C tray temperature. The injection volume was 100 µl. For online monitoring of lipid fractionation, the HPLC system was coupled in positive ESI mode to a 4000 quadrupole linear ion trap mass spectrometer (Q-Trap) (Applied Biosystem/MDS Sciex, Concord, ON, Canada) with a split of 1:21. Two fractions were manually collected, and each was subjected to desalting and concentration. For this, 10 ml deionized water was added to each fraction and the organic phase was collected. After reextraction of the aqueous phase with chloroform/methanol 1:1 (v/v), combined organic phases were dried in a SpeedVac and taken up again in 100 µl chloroform/methanol 1:1 (v/v) for further analysis by reversed-phase-HPLC as described before.

Mass spectrometry
LTQ-FT mass spectrometry. A 7.0 Tesla hybrid linear ion trap Fourier transform ion cyclotron resonance mass spectrometer (LTQ-FT; Thermo Fisher Scientifi c, Bremen, Germany) equipped with an electrospray ion source was used. The instrument was operated in preview mode for parallel MS/MS spectra in the linear ion trap, while running the ion cyclotron in full scan mode at could infl uence ionization effi ciency. However, in a validation experiment with standard LM 6000 (eight deuterated TG species), we found that the HPLC gradient applied and TG structural features have only a minor impact [coeffi cient of variation (CV) of signal intensities of 8.4% only, data not shown].

Lipid identifi cation and quantitation
Accurate mass determination. The solvent chosen facilitates the formation of ammonia adducts [M+NH 4 ] + for TG in positive ESI to yield an elemental fi ngerprint of C x -H y O 6 N. The exact mass of better 2 ppm is suffi cient in most cases for unambiguous identifi cation of the elemental composition. The resolution of 200,000 at m/z 400 results in 90,000 at m/z 950, the upper limit for molecular mass of major lipid species in lipid droplets. This resolution does not provide baseline separation of C x H y O 6 N and 13 C 2 C x-2 H y O 6 N, but it is still good enough for peak top separation at 80% peak height. The instrument would be able to deliver much higher resolution, but the resolution chosen is a compromise because ion cyclotron sensitivity decreases with increasing resolution on the one hand, and acquisition time increases with increasing resolution, resulting in a slower duty cycle, on the other hand.

Determination of lipid molecular species by MS/MS fragmentation.
For interpretation of exact mass data in terms of fatty acid composition of lipid species and for confi rmation of identity by specifi c fragments, MS/MS spectra were acquired in DDA. In this setup, the cyclotron and linear ion trap operate as two separate instruments. This method allows acquisition of low-resolution MS/MS spectra in the linear ion trap on the most intense peaks in parallel to high-resolution spectra acquired by the ion cyclotron ( 30 ). The MS/MS coverage of lipid species attained in DDA is 66%. The method circumvents the need either to rerun the sample for obtaining MS/MS spectra on precursors identifi ed in a previous high-resolution run or to operate the instrument sequentially in high-resolution MS and subsequent MS/MS mode, which would result in a lower duty cycle. This is illustrated paradigmatically in an extracted ion chromatogram shown in Finally, in the retention time window from 20 to 30 min, bulk TG elute, the least-polar species. Notably, with the help of this reversed-phase-HPLC, the class of TG can be separated even in high amounts from the rest of the lipids.
Mobile phase composition and structural features (e.g., chain lengths and degree of unsaturation of TG species)   Tables I-IV, retention times demonstrate the occurrence of these very minor species by decreasing elution order of these compounds according to number of double bonds. Clearly, identifi cation of lipid species becomes possible by retention time even without availability of reliable MS/MS spectra as long as any other species of the same lipid class with the same carbon number has a reliable MS/MS spectrum. Fig. 2 reveals that in reversed-phase-HPLC/ESI-MS analysis, lipid species from different lipid classes overlap, particularly at retention times between 10 and 20 min. Owing to excellent ionization effi ciency in positive electrospray mode, PC species are a case in point. They exert a dominating ion suppression on other species of partially coeluting lipid classes. Thus, a chromatographic selectivity complementary to reversed-phase-HPLC is needed to attain separation of PC species from other polar lipids. Our choice is a diol-based stationary phase allowing for HILIC. The semi-preparative diol column is preconditioned once for 1 h with the isocratic solvent composition as described in Materials and Methods . The advantage of isocratic elution over gradient elution is that it requires no time-consuming preconditioning between individual runs and has better combinatorial restrictions arising from elemental composition allow for deducing the molecular species to be  ( 6,31 ). Constituent FA is determined by neutral losses in positive ionization mode and carboxylates in negative ionization mode.

2D HPLC approach. The chromatogram in
Identifi cation of lipid species by retention time. Low abundant compounds sometimes do not generate reliable MS/ MS spectra. In this case, accurate retention in addition to accurate mass becomes an important criterion for identification. The method of choice is reversed-phase chromatography, where separation of lipid species within one class, expressed by equivalent carbon numbers ( 32 ), are mainly based on interaction between hydrophobic stationary phase and acyl carbon chains of the lipid. In fact, chain length and degree of unsaturation of constituent FA of the lipid matter. Consequently, species having FA of same chain lengths (same number of acyl carbons) can be separated by reversed-phase chromatography by their degree of unsaturation in the molecule; the chromatograms shown in Fig. 4 impressively illustrate such behavior. Elution times of TG with the same number of acyl carbons Application of reversed-phase-HPLC/MS to profi le lipid species from lipid droplets. As the gain in signifi cant improvements by 2D chromatography is low in relation to considerably more experimental effort and time needed, our method retention time reproducibility. For method development, standards are used as described in Materials and Methods . During HILIC of lipid droplet samples, the eluent of the fi rst 10 min is discarded, then lipids eluting between 10 and 24 min (TG, DG, PS, PE, and PI) are collected in fraction 1. The second fraction includes PC, SM, and LPC and elutes between 24 and 48 min ( Fig. 5 ). The subsequent liquid-liquid extraction after fractionation has two purposes. On the one hand, it is necessary to desalt fractions from exceeding ammonium acetate and transfer them into a solvent compatible with the second chromatographic dimension. On the other hand, concentrations of lipids in fractions could be increased by a factor of 140 by this procedure.
Our fi ndings indicate that the 2D method elaborated results in a higher sensitivity for certain lipid classes. The data shown in Fig. 6A demonstrate that DG species in this sample can be detected by the 2D approach with a higher sensitivity compared with the 1D approach, and the same is found for PE species ( Fig. 6B ).   Table IV). The lowest concentration points of the calibration curves are all at a signal-to-noise (S/N) ratio greater than 10:1, a value generally referred to in the literature, thus defi ning the limit of quantitation (LOQ). Accuracy and precision were determined for fi ve replicas spiked with these compounds. The former is within 20% deviation; the latter is below 10% relative standard deviation (supplementary Table IV). All species detected show an S/N ratio of at least 34:1 or greater, well above the limit of detection (LOD) defi ned at S/N 3:1. The mean retention time deviation of all acquired lipid species is 0.37% (n = 5).

DISCUSSION
Excellent LC/MS methods were developed based on Orbitrap and LTQ-FT mass spectrometry, although to date, they are restricted exclusively to analysis of phospholipids ( 19,20,33 ). LC/MS analysis of lipid species by low-resolution multiple reaction monitoring (MRM), applicable only to a preset list of expected lipids, was also described ( 34,35 ). In this article, however, a method for comprehensive profi ling of lipid species is presented that enables data acquisition in one HPLC run per ionization polarity. Our platform provides full scan data of high-resolution ion cyclotron resonance mass spectrometry. Untargeted MS/MS data, which can be used in an untargeted manner in subsequent bioinformatics analyses, are also provided. This is an important aspect when high-throughput data of choice is reversed-phase HPLC/MS. This method allows identifi cation of lipid species by exact mass, MS/MS spectra, and retention time in one platform ( Fig. 1 ). Exact mass and retention time are then used by LDA for identifi cation and peak integration as described previously ( 7 ), whereas MS/MS spectra are inspected manually for confi rmation of identity and fatty acid analysis of lipid molecular species. If MS/MS spectra for some species are not available, confi rmation of identity can be deduced from retention time shifts ( Fig. 4 ); a practical example of the latter is given in Fig. 7 . As a result of these procedures, the platform is able to identify lipid species as low as 0.1‰ of the respective base peak of a given lipid class with a high degree of certainty, even beside bulk amounts of a few lipid species. Thus, we were able to identify 103 minor TG species in addition to 19 major TG species, with each of the minor TG species contributing less than 10‰ of the total amount of TG. The total number of lipids identifi ed accounts for 122 TG, 41 DG, 28 PC, 18 PE, 4 PS, 9 PI (PC compared with the latter three being the dominant phospholipid class), 13 LPC, and fi nally, 7 SM species. Etherlinked phospholipids (e.g., plasmalogens) were not detected. Fig. 8 shows the molar proportion of each lipid class relative to total lipid amount in the pooled samples. All individual lipid species found in pooled lipid droplets are shown in supplementary Tables I-III, which   base peak in the respective lipid class. In fact, the platform is a valuable tool for detection of changes occurring in very low concentration ranges of a lipid species compared with the total amount of lipid species in a given sample.
The limitation of reversed-phase-HPLC is separation of PC from other phospholipid classes. This could be overcome by adding a second chromatographic dimension with complementary selectivity. A separator linking the two chromatographic dimensions for automated analysis was developed recently, but it is not commercially available ( 15 ). In the case of lipid droplet analysis reported here, we carried out such an approach by manually linking the two chromatographic dimensions, and we found enhanced sensitivity for detection of PE and DG species (enhancement of PS species was not signifi cant). Obviously, in positive ESI mode, these lipid classes benefi t particularly from PC removal by the fi rst chromatographic dimension.
Shotgun lipidomics of chromatographically not separated, overlapping M+2 peaks requires isotopic correction ( 8 ). This is not necessary here, due to chromatographic resolution and the extremely high mass resolution delivered by the ion cyclotron. Several combinations of chromatography (retention time), high mass accuracy, and MS/MS information are addressed in the literature ( 15,19,20,25 ). But the combination of all these parameters, including bioinformatic tools, integrated in one platform applicable to a wide variety of lipid classes is the novel concept realized in this work. All of this results in high identifi cation certainty, sensitivity, and selectivity, and it contributes signifi cantly to unambiguous detection of minor lipid species. generation in "omics" sciences is required. Flexibility of the platform offers a further advantage (i.e., integration of "unexpected" lipid species in a sample into ongoing analysis). Without any need to rerun a sample, this method requires only reanalysis by LDA with a database expanded by the unexpected lipid species.
Clearly, shotgun and fl ow injection lipidomic analyses are closest to high-throughput data generation, as proven by successful application to biological entities having a more or less common lipid class distribution ( 8,10,36 ). But in samples containing bulk amounts of a certain lipid class, preseparation is mandatory ( 13 ) due to ion suppression exerted in MS on minor components of the lipidome that would remain undetected. Taking the example of lipid droplet analysis, we demonstrate here that reversedphase-HPLC coupled to MS monitoring is the method of choice for profi ling glycerolipids, glycerophospholipids, and sphingolipids, even in the presence of a huge excess of TG species. From the quantitative point of view, unambiguous detection of lipid species is possible at a concentration that is four orders of magnitude lower than the