If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Quantitatively and rapidly analyzing lipids is necessary to elucidate their biological functions. Herein, we developed a quantitative method for various lipid classes using supercritical fluid chromatography (SFC) coupled with a charged aerosol detector (CAD), providing high-throughput data analysis to detect a large number of molecules in each lipid class as one peak. Applying the CAD was useful for analyzing lipid molecules in the same lipid class with a constant response under the same mobile phase composition. First, we optimized the washing method for the diethylamine column, achieving baseline separation of lipid classes while maintaining good peak shapes. In addition, the CAD conditions (organic solvent evaporation and numerical correction of the CAD data) were optimized to improve the signal-to-noise ratio. We used an internal standard (ceramide phosphoethanolamine d17:1–12:0), which did not coelute with the lipid classes and showed high extraction efficiency. Based on a quantitative analysis of HepG2 cells, the concentration of lipid classes detected by CAD was adequate compared with that obtained by triple-quadrupole MS (QqQMS) in a previous study because the deviations of the concentrations were 0.6- to 2.3-fold. These results also supported the quantitative performance of SFC-QqQMS developed in our previous report.
Lipids act as components of cell membranes, biologically active substances, and energy storage. Thousands of distinct lipid molecules are generally classified based on differences in their backbone and polar head groups (
). To date, various biological functions of lipid classes have been reported. For example, phosphatidylcholine (PC) synthesized via cytidine triphosphate-phosphocholine cytidyltransferase in the intestinal epithelium is important for dietary lipid absorption of high-fat diets (
). Phosphatidylglycerol (PG) and phosphatidylinositol (PI), which are anionic pulmonary surfactant lipids, exert antiviral effects against the respiratory syncytial virus (
). TLC, which provides high-throughput capabilities and simple operation, is the first method used to investigate the biological functions of the lipid classes (
). It does not require specialized equipment, but the associated result is often poorly reproducible in terms of quantitation due to the size and depth of the spots. To improve quantitation for each lipid class with high-resolution separation, chromatographic methods have been developed using LC and supercritical fluid chromatography (SFC) coupled with various detectors. Similar to TLC, each lipid class is separated by LC or SFC using a normal-phase column with a high-polarity stationary phase. Compared with LC with a normal-phase column, SFC with a normal-phase column [especially a diethylamine (DEA)-bonded silica column] can achieve good baseline separation of each lipid class while maintaining good peak shapes (
). A fast and reliable method to acquire the imbalance of lipid classes at a glance is expected to be provided by applying the separation mode of SFC to a high-sensitive detector.
Although mass spectrometers allowed for the calculation of concentrations of lipid classes by summing the responses of individual lipid molecules, data analysis remained challenging to detect a large number of molecules in each lipid class (
). Except for mass spectrometers, other detectors have been developed to analyze various compounds during gradient condition. UV-visible absorption spectroscopy detectors and refractive index detectors are not suitable detectors because they cannot achieve comprehensive lipid analysis and gradient elution, respectively. To meet the aforementioned demands, universal detectors, such as evaporation light-scattering detectors (ELSDs) and charged aerosol detectors (CADs), represent promising alternatives (
Comparison of ultraviolet detection, evaporative light scattering detection and charged aerosol detection methods for liquid-chromatographic determination of anti-diabetic drugs.
). ELSDs and CADs measure the light scattering of dried particles and the amount of charge on dried particles, respectively, after removing the mobile phase through evaporation. They can analyze most compounds except for highly volatile compounds and are amenable to gradient elution because the mobile phase is evaporated prior to detection. While lipid analysis using LC-ELSD or LC-CAD has been reported in the past (
Improved separation and quantification of neutral and polar lipid classes by HPLC-ELSD using a monolithic silica phase: application to exceptional marine lipids.
J. Chromatogr. B Analyt. Technol. Biomed. Life Sci.2009; 877: 1815-1819
Major lipid classes separation of buttermilk, and cows, goats and ewes milk by high performance liquid chromatography with an evaporative light scattering detector focused on the phospholipid fraction.
) showed that the sensitivity and precision of CAD is higher than those achieved by ELSD at the lower end of the calibration curve. Because the CAD response depends on the amount of particles, CAD can analyze lipid molecules in the same class with a constant response. In addition, large numbers of molecules in the same lipid class can be detected in a single peak when separating lipid classes. Therefore, quantitative analysis of lipid classes with high sensitivity and throughput is expected by determining the optimal conditions to connect SFC to CAD while using a gradient elution mode of SFC. In this study, we developed a quantitative method of lipid classes from biological samples using SFC-CAD. To maintain the high separation of each lipid class, we applied the same SFC conditions optimized in a previous study (
). We analyzed lipid classes in HepG2 cells, a human hepatoma cell line, and evaluated the quantitative performance of the developed analytical method.
MATERIALS AND METHODS
Chemicals and reagents
Ammonium acetate was obtained from Sigma-Aldrich (St. Louis, MO). LC/MS-grade methanol and distilled water were purchased from Kanto Chemical Co. (Tokyo, Japan). HPLC-grade chloroform was obtained from Kishida Chemical (Osaka, Japan). All synthetic lipid standards and SPLASH LIPIDOMIX Mass Spec Standard were purchased from Avanti Polar Lipids Inc. (Alabaster, AL). The stock concentrations of the synthetic lipid standards were as follows: 25 μg ml−1 lysophosphatidylcholine (LPC) 18:1(d7); 5 μg ml−1 lysophosphatidylethanolamine (LPE) 18:1(d7); 160 μg ml−1 PC 15:0–18:1(d7); 5 μg ml−1 phosphatidylethanolamine (PE) 15:0–18:1(d7); 30 μg ml−1 PG 15:0–18:1(d7); 5 μg ml−1 phosphatidylserine (PS) 15:0–18:1(d7); 10 μg ml−1 PI 15:0–18:1(d7); 7 μg ml−1 phosphatidic acid (PA) 15:0–18:1(d7); 30 μg ml−1 SM d18:1–18:1(d9); 100 μg ml−1 cholesterol (d7); 350 μg ml−1 cholesteryl ester (CE) 18:1(d7); 2 μg ml−1 monoglyceride 18:1(d7); 10 μg ml−1 diglyceride (DG) 15:0–18:1(d7); and 55 μg ml−1 triglyceride (TG) 15:0–18:1(d7)–15:0. CO2 (99.9% grade; Yoshida Sanso Co., Fukuoka, Japan) was used as the SFC mobile phase.
Cell culture
Human HepG2 cells were maintained in DMEM with 75 cm3 cell-culture flasks and incubated for 3 days at 37°C. After the cells were trypsinized, the number of cells was measured using cell counters to transfer approximately 1.0 × 105 to 1.0 × 106 cells to 10 cm dishes. The cells were then incubated for 3 days at 37°C. After washing with cold phosphate-buffered salts, the cells were quenched by adding 1,000 μl of −30°C methanol and recovered with clean tubes.
Sample preparation
Lipid extraction from HepG2 cells was performed using Bligh and Dyer's method with minor modifications described previously (
). Briefly, 1,000 μl methanol was mixed with 400 μl chloroform, vortexed at the maximum setting for 1 min, and extracted by ultrasonication for 5 min. It was then centrifuged at 16,000 g for 5 min at 4°C, and 700 μl of the supernatant was transferred to clean tubes. After mixing with 300 μl chloroform and 400 μl distilled water, the aqueous and organic layers were separated by centrifugation at 16,000 g for 5 min at 4°C. The organic layer was divided into two clean tubes for CAD and triple-quadrupole MS (QqQMS) analyses. For the CAD analysis, 75 μl of the lipid extract was dried under nitrogen gas, and 30 μl of the internal standard solution (final concentration: 50 μg ml−1) was added. For QqQMS analysis, 30 μl of the lipid extract was mixed with 15 μl of the SPLASH MS standard (final concentration: 20-fold dilution of stock) and 15 μl ceramide (Cer) d18:1(d7)–15:0 (final concentration: 0.2 μg ml−1).
Column washing
To prevent the reduction of the signal-to-noise ratio (S/N) by the column bleed, the column was washed before connecting with the SFC-CAD. Column washing was performed using a Nexera X2 system (Shimadzu Co., Kyoto, Japan) because distilled water is applied as the mobile phase. The LC conditions were as follows: mobile phase (A), acetonitrile with or without 0.1% formic acid (FA); mobile phase (B), distilled water with or without 0.1% FA; flow rate, 0.4 ml min−1; isocratic condition; 50% (v/v); column oven temperature, 50°C; and washing time, approximately 3 days. Before connecting the washed DEA column with the SFC, the solvent was replaced with 100% acetonitrile to prevent phase separation between the supercritical CO2 and distilled water.
Analytical conditions
Lipid analysis was performed using a Nexera UC system (Shimadzu Co.) with a Xevo TQ-S micro tandem mass spectrometer (Waters, Milford, MA) and a Corona Veo Charged Aerosol Detector (Thermo Fisher Scientific, Waltham, MA). The Nexera UC system and Corona Veo CAD were controlled using LabSolutions LCMS version 5.87 (Shimadzu), and the Xevo TQ-S micro tandem mass spectrometer was controlled by MassLynx software version 4.1 (Waters). The SFC and QqQMS analytical conditions were optimized previously (
). The SFC conditions were as follows: injection volume, 1 μl; mobile phase (A), supercritical fluid CO2; mobile phase (B) (modifier) and make-up pump solvent, methanol/water (95:5; v/v) with 0.1% (w/v) ammonium acetate; flow rate of mobile phase, 1.0 ml min−1; modifier gradient, 1% (B) (1 min), 1% to 65% (B) (11 min), 65% (B) (6 min), 65% to 1% (B) (0.1 min), and 1% (B) (1.9 min); temperature of column manager, 50°C; active back-pressure regulator, 1,500 psi; analytical time, 20 min; and columns, ACQUITY UPC2TM ethylene-bridged hybrid, 2-ethylpyridine, Torus 2-picolylamine, Torus 1-aminoanthracene, Torus DIOL, and Torus DEA (inner diameter of each: 100 × 3.0 mm; particle size: sub 1.7 μm; Waters) and Inertsil ODS-4 (inner diameter: 100 × 3.0 mm; particle size: 2 μm; GL Sciences, Tokyo, Japan). The QqQMS conditions were as follows: capillary voltage, 3.0 kV; desolvation temperature, 500°C; cone gas flow rate, 50 l h−1; and desolvation gas flow rate, 1,000 l h−1. To enhance the ionization efficiency for CE and TG, the flow rate of the make-up pump was set to 0.1 ml min−1 for QqQMS analysis. The final CAD analytical conditions were as follows: filter, 5.0 s; power function value (PFV), 1.80; range, 500 pA; output offset, 0%; ion trap, 20.2 V; pressure of nebulizer gas, 62.2 psi; charger voltage, 2.40 kV; charger current, 1.00 μA; and drying tube temperature, 95°C.
Method validation
To create calibration curves for the CAD analysis, standard mixtures were prepared at the following concentrations: 0, 1, 2, 5, 10, 20, 50, 100, 200, 500, and 1,000 μg ml−1. Ceramide phosphoethanolamine (CPE) d17:1–12:0 (final concentration: 50 μg ml−1) was spiked into each standard mixture as an internal standard. To create calibration curves for the QqQMS analysis, standard mixtures were prepared at the following concentrations: 0, 1, 5, 10, 50, 100, 500, 1,000, 5,000, 10,000, 50,000, and 100,000 ng ml−1. The SPLASH MS standard and Cer d18:1(d7)–15:0 (final concentration: 20-fold dilution of stock and 0.2 μg ml−1, respectively) were spiked into each standard mixture as internal standards.
Data analysis
Identification and quantitation of the lipid molecules from the QqQMS data were performed using MassLynx software version 4.1 as described previously (
). Lipid classes in the CAD data were identified based on the results of QqQMS data and quantitated via LabSolutions LCMS version 5.87.
RESULTS AND DISCUSSION
Overview of the analytical system
In this study, the SFC-CAD system was used to quantitate lipid classes. To change the mobile phase to supercritical CO2, a critical temperature (31.1°C) and critical pressure (7.38 MPa) must be maintained. The pressure-control system with a back-pressure regulator differs according to the SFCs (supplemental Fig. S1). In some SFCs, the flow is split between the active back-pressure regulator and detector (i.e., CAD or QqQMS) to regulate the pressure (supplemental Fig. S1A). In our previous study, the ratio of this flow was approximately 9:1 (
). The sensitivity of the CAD system depends on the amount of particles, so the sensitivity could be lowered by introducing only partial samples. To address this concern, we used the Nexera UC system, which can control the back pressure without separating the flow between the active back-pressure regulator and detector (supplemental Fig. S1B). Nexera UC can introduce the entire sample volume to the detector, so it is expected to analyze compounds with high sensitivity compared with other SFCs.
Column washing
Background noise is always observed in the CAD analysis because CAD can detect most compounds except for highly volatile compounds. To improve the S/N of the target compounds, it is necessary to reduce the background noise. First, the effect of the stationary phase of each column on the background response was explored (supplemental Fig. S2). We applied a gradient condition that was optimized in a previous study [1% (B) (0–1 min), 1%–65% (B) (1–12 min), 65% (B) (12–18 min), 65%–1% (B) (18–18.1 min), 1% (B) (18.1–20 min)] (
). When the column was not connected to the instruments, the background was stable with low intensities at all times. Similar background behavior was observed by connecting an ethylene-bridged hybrid column. However, columns with highly polar stationary phases showed high background levels regardless of functional group. It was possible that the high background was induced by column bleed caused by hydrolysis of the stationary phase, dissolution of the support material, and/or column degradation (
). We previously demonstrated that the DEA column achieved the best separation of each lipid class among six columns tested whose stationary phases exhibited high polarities as a result of column screening (
). The washing conditions of the DEA column were explored to reduce column bleed without affecting the stationary phase. Figure 1 shows a comparison of retention times (RTs) observed for the HepG2 cell extracts with the two DEA columns, which were washed using 50% (v/v) acetonitrile/water with and without 0.1% FA. For both washed conditions, CAD data backgrounds decreased to adequate levels. For the DEA column washed with 50% (v/v) acetonitrile/water, changes in the RTs were detected, whereas no significant change was observed for the column washed with 50% (v/v) acetonitrile/water containing 0.1% FA. From Fig. 1A, a decrease in RT was observed for acidic lipids [lysophosphatidic acid (LPA), lysophosphatidylserine (LPS), PA, and PS], and an increased RT was observed for lipids containing phosphocholine (i.e., quaternary ammonium cation). The former can be ascribed to the degradation of the stationary-phase amount (i.e., diethylamino groups), and the latter is associated with the possibility of enhanced interaction between the quaternary amino groups and silanol groups of the silica particles. It was suggested that during washing with 50% (v/v) acetonitrile/water, the diethylamino (basic) groups resulted in an alkali condition on the silica-particle surfaces to degrade the siloxane bonds anchoring the functional groups to the surfaces. This caused an undesirable change in the retention ability for SFC measurements, as reported with the aminopropyl-bonded phase for LC measurements (
). Because of the aforementioned result, it should be noted that the DEA columns must be washed and preserved with a proper organic solvent mixture containing a buffer at a pH of approximately 3 (e.g., using 0.1% FA) to perform the highly qualitative SFC measurements.
Fig. 1Determination of the washing method of the DEA column. Chromatograms of HepG2 cells analyzed by SFC-CAD are shown on the left of each panel, whereas the effects of retention behavior as a result of column washing are shown on the right of each panel. The retention behavior was investigated using a lipid standards mixture analyzed by SFC-QqQMS. The black squares show lipids containing phosphocholine; black triangles represent acidic lipids. A: 50% (v/v) acetonitrile/water without 0.1% FA. B: 50% (v/v) acetonitrile/water with 0.1% FA.
To obtain the lowest background levels with the washed DEA column, we optimized the CAD conditions (i.e., nebulizing of the column eluate, organic solvent evaporation, and numerical correction of the CAD data). Background levels were reduced by lowering the nebulizer gas pressure. However, lipid analysis of the HepG2 cells showed that the S/N remained unchanged despite changing the nebulizer gas pressure (data not shown). It is likely that the constant S/N was due to the reduction of nebulizing efficiency. Therefore, the nebulizer gas pressure was set to 62.2 psi, which was the recommended level. By raising the drying tube temperature in the lipid analysis of HepG2 cells, background levels were reduced gradually while maintaining the signal values of the lipid classes. On the other hand, the intensity of the peak corresponding to the elution time of DG classes and cholesterol decreased by raising the drying tube temperature (Fig. 2). We investigated this peak using synthetic standards of DG 18:0–20:4 and cholesterol. By raising the drying tube temperature, the intensity of DG 18:0–20:4 did not decrease, but that of cholesterol decreased dramatically by approximately 16-fold (supplemental Fig. S3B–D). Vervoort et al. (
) reported that the response of some compounds similarly changed due to the drying tube temperature in ELSD analysis. This indicates that the response is reduced by thermal degradation of the compounds during evaporation. However, cholesterol can be easily quantitated using various assays (e.g., biochemical analysis), and it is not necessary to analyze it by SFC-CAD. Similar to cholesterol, unexpected compounds were assumed to be removed during evaporation even when coeluted with the target lipid class. Therefore, the drying tube temperature was set to 95°C, which was nearly the maximum level for the CAD instrument.
Fig. 2Improved S/N by raising the drying tube temperature. To evaluate the S/N of each lipid class in the biological samples, the lipid extract of HepG2 cells was analyzed using SFC-CAD. A: Mixed solvent (1:1 methanol/chloroform; v/v). B: Lipid extract of the HepG2 cells.
In the Corona Veo CAD, the PFV, which reduces background noise and improves the correlation coefficient of the calibration curve, is an important parameter for CAD data output. When using the CAD, it is thought that linearity is lost as concentrations increase beyond a certain point because the response per particle mass for aerosol charging decreases for particles with a diameter of ˃10 nm (
). It is well-known that the calibration curves of each compound are obtained using power functions (y = axb, where x is the concentration of the compound, y is the area value, and a and b are coefficients that depend on the chromatographic conditions and/or compounds). With increasing PFV, the signal current value increased significantly compared with that of the noise, as described by equation 1. The calibration curves of some compounds were not power functions but linear functions (y = ax + b), which improved the correlation coefficient.
(Eq. 1)
We further examined a synthetic PE 18:0–18:1 standard that elutes in the middle of the gradient. As the PFV increased, the S/N improved, and the calibration curve was linear with a high correlation coefficient while maintaining sensitivity and repeatability (Fig. 3). In this study, the PFV was set to 1.80.
Fig. 3Improvement of S/N by raising the PFV. A: Difference in the analyses of the HepG2 cell extracts using various PFVs. B: Calibration curve of the PE 18:0–18:1 synthetic standards at each PFV. Each calibration curve was prepared at the following concentrations: 0, 1, 5, 10, 50, 100, 500, and 1,000 μg ml−1. Error bars indicate the standard deviations of the analytical replicates (n = 3).
Nontargeted quantitation of lipid classes using hydrophilic interaction liquid chromatography-electrospray ionization mass spectrometry with single internal standard and response factor approach.
) used CPE d17:1–12:0, which was eluted between the PE and LPE classes, as the internal standard for lipidome analysis using hydrophilic-interaction LC coupled with ESI-MS. When using SFC with a DEA column, CPE d17:1–12:0 was similarly eluted between the PE and LPE classes (Fig. 4B). From the analysis of HepG2 cells, no compounds were detected at the same RT as CPE d17:1–12:0 (Fig. 4C). Its extraction efficiency was also investigated using 1 mg ml−1 of the internal standard extracted following the modified Bligh and Dyer's method (Fig. 4D). In addition, 1 mg ml−1 of the internal standard was diluted in accordance with the theoretically estimated concentration following the extraction process. The area values of these samples were compared, and their extraction efficiency was evaluated. No differences in these area values were observed, indicating adequate repeatability of extraction efficiency.
Fig. 4Investigation of the internal standard used for SFC-CAD analysis. A: Structure of CPE d17:1–12:0. B: Gradient condition of SFC (B pump, %) (top) and RTs of each lipid class (bottom). The opened circles show synthetic standards of each lipid class, and the closed circle represents the CPE d17:1–12:0. C: SFC-CAD data of the HepG2 cells and internal standard. D: Evaluation of extraction efficiency of CPE d17:1–12:0 using 1 mg ml−1 of the internal standard extracted three times (closed squares; E1–E3). The theoretical dilution rate in the extraction process was 20-fold, so 50 μg ml−1 of the internal standard was prepared (open square) for comparison. The error bars indicate the standard deviations of analytical replicates (n = 3). ISD, internal standard.
The validation of the established analytical method was performed based on the optimized CAD conditions and internal standards. Prior to the experiment, we analyzed 100 μg ml−1 of each lipid standard (Fig. 5). The response was changed slightly with an increasing ratio of the modifier solvent despite analyzing the same concentration of the lipid standards mixture. It is well-known that the CAD response changes based on the composition of the LC mobile phase because the formation efficiency of micro droplets is affected by organic solvent content during nebulization. The CAD response also changed according to the ratio of the composition of CO2 and organic solvent for SFC. To maintain a constant response, an inverse gradient, which is a method for adding organic solvent in front of the detector, has been recommended for LC applications (
). However, it is impractical for SFC to apply this method because an additional CO2 bomb and pump would be needed. To address this issue, the calibration curves were obtained using the ratio of the peak area of the synthetic standards corresponding to each lipid class to that of CPE d17:1–12:0. The results of the analytical validation are summarized in Table 1. The LOD and LOQ were determined using the calibration curve, whose x axis was the concentration of each lipid synthetic standard and y axis was the S/N. Because the CAD response depended on the amount of particles, the linear ranges of each lipid class were similar except for those of acidic lipids such as LPA, LPS, PA, and PS (i.e., lipid classes whose peak shapes were tailing). It is assumed that these acidic lipids are retained on the stainless steel tube. On the other hand, the LOD and LOQ of each lipid class differed due to the changing background levels in response to the mobile-phase composition. By applying the PFV, calibration curves of almost all lipid classes were linear functions (y = ax + b) with high correlation coefficients (R2 >0.9886). LPA, lysophosphatidylinositol (LPI), LPS, and PI necessitated power functions (y = axb) to obtain high correlation coefficients (R2 >0.9915) and are presented as linear log-log plots (logy = blogx + loga). The optimal PFV for the linear function must be tuned for individual target compounds, but the PFV cannot be changed during sample analysis. Based on the calibration curves, we determined a correction factor to calculate the quantitative levels of each lipid class.
Fig. 5Analysis of the lipid synthetic standards using SFC-CAD. Standard solutions (100 μg ml−1) of each lipid synthetic standard were prepared in a mixed solvent (65:35:8 methanol/chloroform/distilled water; v/v/v). Dotted line shows the gradient condition of SFC (B pump, %). 1, DG 18:0–20:4; 2, Cer d18:1–17:0; 3, PC 17:0–17:0; 4, SM d18:1–17:0; 5, LPC 18:0; 6, PE 17:0–17:0; 7, CPE d17:1–12:0 (internal standard); 8, LPE 17:1; 9, PG 17:0–17:0; 10, lysophosphatidylglycerol (LPG) 17:1; 11, PI 18:0–20:4; 12, PS 17:0–17:0; 13, PA 17:0–17:0; 14, LPI 17:1; 15, LPS 17:1; and 16; LPA 17:0.
Each calibration curve was prepared at the following concentrations: 0, 1, 2, 5, 10, 20, 50, 100, 200, 500, and 1,000 μg ml−1. The calibration curves of LPA, LPI, LPS, and PI are presented as linear log-log plots.
Similarly, the response of individual lipid molecules was examined using synthetic PC standards. CAD response depends on the amount of particles under the same mobile-phase composition. In addition, individual lipid molecules of the same lipid class are eluted at similar RTs by SFC separation with the DEA column. The slopes, R2 values, LODs, LOQs, and linear ranges of the calibration curves showed similar results among the examined PC standards with different carbon numbers and double bonds (supplemental Table S1). Therefore, it is possible to quantitate lipid classes briefly by detecting their corresponding peaks.
The workflow for the quantitative analysis of lipid classes in biological samples was as follows: i) lipid extracts were prepared by liquid-liquid extraction, and prior to extraction, CPE d17:1–12:0, used for the normalization of extraction efficiency and CAD analysis, was added; ii) the lipid extract was analyzed with SFC-CAD; and iii) the concentration of lipid classes was calculated using the correction factor determined by calibration curves.
Quantitative analysis of lipid classes in HepG2 cells
In our preceding study, we successfully demonstrated a quantitative analysis of lipid classes and their molecules using SFC-QqQMS (
). ESI-MS is ideal for analyzing lipids because it can measure ionized compounds with high sensitivity. However, the quantitative results of lipid molecules observed by ESI-MS vary according to differences in equipment and/or facilities due to a number of reasons (e.g., the lack of suitable internal standards, extraction efficiency, ionization efficiency) (
Harmonizing lipidomics: NIST interlaboratory comparison exercise for lipidomics using Standard Reference Material 1950 Metabolites in Frozen Human Plasma.
). It was demonstrated that the ionization efficiency of lipid molecules mainly depends on the characteristics of their polar head groups and that the ionization efficiency hardly varied for the different fatty acid side chains (
Shotgun lipidomics: electrospray ionization mass spectrometric analysis and quantitation of cellular lipidomes directly from crude extracts of biological samples.
). To quantitate lipid molecules by normalizing ionization efficiency, we used a DEA column to sufficiently separate each lipid class. By adding internal standards for each lipid class, each lipid molecule was analyzed within an appropriate quantitation ranging from 64.9% to 103.5%. In addition, individual lipid molecules, including structural isomers, can be monitored using multiple reaction monitoring (MRM) transitions of QqQMS derived from their fatty acid side chains.
Although this analytical method could quantitate individual lipid molecules, several issues, including the throughput of data analysis, remain for estimating the concentration of lipid classes. Lipid classes can be quantitated by summing the concentration of lipid molecules in the same class. On the other hand, the MRM mode of QqQMS can analyze only target lipid molecules, so unexpected lipid molecules may affect the quantitation accuracy. Furthermore, we can normalize ionization efficiency by using internal standards for each lipid class, but the fragmentation efficiency of the MRM mode cannot be normalized. Of course, quantitative values are affected by unexpected compounds for the CAD. Compounds in the organic layer obtained by the modified Bligh and Dyer method consist mostly of lipids. In addition, the drying tube temperature was set to nearly the maximum level of the system to remove unexpected compounds during evaporation even if they were coeluted with the target lipid class. From this point of view, the accuracy of lipid class quantitation is expected to increase by using the CAD.
In this study, we compared the concentrations of lipid classes in HepG2 cells estimated by CAD and QqQMS to evaluate quantitative performance. Prior to this comparison, calibration curves were created for QqQMS analysis. To remove the effect of storage on quantitative levels of the lipid synthetic standards, the same lipid synthetic standards used in CAD analysis were used for QqQMS analysis. As internal standards for QqQMS analysis, a deuterium-labeled SPLASH MS standard was used (
). We previously used dodecanoyl- or heptadecanoyl-based synthetic standard mixtures as internal standards, but these lipid molecules were occasionally found in biological samples. Calibration curves of the QqQMS analysis were constructed using the ratio of the peak area of each lipid molecule to that of the internal standards of the representative lipid classes.
The study design for comparing quantitative levels of each lipid class between CAD and QqQMS data is shown in Fig. 6. To remove the effect of extraction efficiency of the internal standards, they were directly added to the lipid extracts. Figure 7 shows the concentration of lipid classes in the HepG2 cells detected by CAD and QqQMS. In Fig. 7C, the lipids were quantitated within a plausible range (PC: 0.6-fold; PE: 1.3-fold; PI: 0.6-fold; PS: 1.6-fold; SM: 2.2-fold; and DG: 2.3-fold and higher). From supplemental Fig. S3, DG and cholesterol were contained in the same peak in the CAD data. Because cholesterol could be quantitated by QqQMS, the DG concentration detected by CAD was estimated by subtracting the area value equivalent to the cholesterol concentration. It has been suggested that the effect on the quantitation of lipid classes by fragmentation efficiency of MRM is not immense. On the other hand, data-analysis throughput was dramatically improved because peak picking for all lipid molecules was not necessary. Unfortunately, some lipid classes detected by QqQMS (i.e., LPC, LPE, PG, PA, and Cer) were not identified using CAD because of the different sensitivity. It is recommended that lipid classes with low abundance should be quantitated using highly sensitive MS methods (
). In addition, the CE and TG classes were excluded as target lipid classes because it was impossible to separate them using the DEA column with a highly polar stationary phase (
). Therefore, to obtain their precise quantitative levels, biochemical analysis or lipidome analysis using reversed-phase LC/MS or SFC/MS is necessary (
Fig. 6Strategy for comparing the quantitative levels of the lipids detected by SFC-CAD and SFC-QqQMS. To remove the influence of the efficiency of lipid extraction, internal standards for SFC-CAD and SFC-QqQMS analyses were added to the lipid extract.
Fig. 7Quantitation of lipid classes in the HepG2 cell extracts. A: Chromatogram of the HepG2 cell extracts obtained by SFC-CAD. Each peak was determined based on the result of lipidome analysis using SFC-QqQMS. B: Lipidome analysis of the HepG2 cell extracts detected by SFC-QqQMS. Separation behavior was evaluated using a box plot. The figures in brackets show the number of lipid molecules detected in the HepG2 cell extracts. C: Comparison of quantitative lipid levels analyzed by CAD and QqQMS. Open squares show the concentration obtained by SFC-CAD, and closed squares indicate the concentration obtained by QqQMS. The error bars indicate the standard deviations of the analytical replicates (n = 4).
We successfully developed a quantitative methodology for analyzing lipid classes in biological samples using SFC-CAD. Although the DEA column provided a baseline separation of lipid classes, high background levels in the CAD data were observed due to column bleed. The background was reduced to adequate levels while maintaining good separation of lipid classes by column washing with 50% (v/v) acetonitrile/water containing 0.1% FA. To further reduce background levels, the CAD conditions were optimized. By raising the drying tube temperature by 95°C, the background levels were reduced, and unexpected compounds were assumed to be removed. In addition, the S/N was improved, and a high correlation coefficient in the calibration curves was achieved by raising the PFV. CPE d17:1–12:0 was used as the internal standard because it did not coelute with the lipid classes of interest and showed high extraction efficiency. Lipid extracts of HepG2 cells were analyzed via CAD and QqQMS, where each lipid class was quantitated within an adequate range from 0.6- to 2.3-fold. These complementary results supported the quantitative performance of SFC-QqQMS analysis. Our quantitative analysis of lipid classes represents a potential tool for elucidating the biological functions of lipids.
Acknowledgments
The authors thank Dr. Kaori Taguchi (Thermo Fisher Scientific) for the continuous technical support of CAD.
Comparison of ultraviolet detection, evaporative light scattering detection and charged aerosol detection methods for liquid-chromatographic determination of anti-diabetic drugs.
Improved separation and quantification of neutral and polar lipid classes by HPLC-ELSD using a monolithic silica phase: application to exceptional marine lipids.
J. Chromatogr. B Analyt. Technol. Biomed. Life Sci.2009; 877: 1815-1819
Major lipid classes separation of buttermilk, and cows, goats and ewes milk by high performance liquid chromatography with an evaporative light scattering detector focused on the phospholipid fraction.
Nontargeted quantitation of lipid classes using hydrophilic interaction liquid chromatography-electrospray ionization mass spectrometry with single internal standard and response factor approach.
Harmonizing lipidomics: NIST interlaboratory comparison exercise for lipidomics using Standard Reference Material 1950 Metabolites in Frozen Human Plasma.
Shotgun lipidomics: electrospray ionization mass spectrometric analysis and quantitation of cellular lipidomes directly from crude extracts of biological samples.
This work was supported by Japan Agency for Medical Research and Development Grants 19gm0910010h0204, 19gm0910013h0003, and 19gm1010010s0202 (Y.I., T.B.); Japan Science and Technology Agency Grant JPMJAL1607 (Y.I., T.B.); and Japan Society for the Promotion of Science Grants-in-Aid 17H06304 (Y.I., T.B.) and 18H01800 (T.B.).
The online version of this article (available at http://www.jlr.org) contains a supplement.
32241-0&id=gr1.jpg){kind=link}
32241-0&id=gr2.jpg){kind=link}
32241-0&id=gr3.jpg){kind=link}
32241-0&id=gr4.jpg){kind=link}
32241-0&id=gr5.jpg){kind=link}
32241-0&id=gr6.jpg){kind=link}
32241-0&id=gr7.jpg){kind=link}