Screening of stationary phase selectivities for global

Journal of Chromatography A, 1548 (2018) 76–82

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Screening of stationary phase selectivities for global lipid profiling by ultrahigh performance supercritical fluid chromatography Said Al Hamimi, Margareta Sandahl, Marina Armeni, Charlotta Turner, Peter Spégel ∗ Department of Chemistry, Centre for Analysis and Synthesis, Lund University, Lund, Sweden

a r t i c l e

i n f o

Article history: Received 2 November 2017 Received in revised form 6 March 2018 Accepted 13 March 2018 Available online 14 March 2018 Keywords: Mass spectrometry Lipidomics Design of experiments Blood serum

a b s t r a c t The performance of seven sub-2-!m particle packed columns (2-picolylamine, 2-PIC; charged surface hybrid fluoro-phenyl, CSH-FP; high strength silica C18 SB, HSS-C18 ; diethylamine, DEA; 1aminoanthracene, 1-AA; high density diol and ethylene bridged hybrid; BEH) was examined for lipid separation in ultra-high performance supercritical fluid chromatography (UHPSFC) coupled to quadrupole time-of-flight mass spectrometry. Based on the results of the column screening a method for profiling of multiple lipid species from the major lipid classes was developed. Stationary phases containing "-hydroxy amines, i.e. 1-AA, DEA and 2-PIC, yielded strong retention and poor peak shapes of zwitterionic lipids with primary amine groups, such as phosphatidylserines, phosphatidylethanolamines and its lyso forms. The BEH and HSS-C18 columns showed strong retention of polar and nonpolar lipids, respectively. The Diol column retained the majority of major lipid classes and also produced symmetric peaks. In addition, this column also produced the highest resolution within and between major lipid classes. An injection solvent composed of methanol:chloroform (1:2, v:v) and the addition of 20 mM ammonium formate in the mobile phase improved chromatographic separation and mass spectrometry detection in comparison to ammonium acetate or absence of additive. Finally, chromatographic and mass spectrometric parameters were optimized for the Diol column using a design of experiments approach. The separation mechanism on the Diol column depended on the lipid functionality and the length and degree of unsaturation of the acyl groups. The developed method could resolve 18 lipid classes and multiple lipids within each class, from blood serum and brain tissue in 11 min. © 2018 Elsevier B.V. All rights reserved.

1. Introduction Lipids present a heterogeneous group of substances, displaying a wide diversity in chemical structure and biological function [1]. They constitute the membranes surrounding cells and organelles, serve as fuel molecules and energy stores, and are involved in cell signalling and cell–cell interactions [2–4]. Lipid metabolism generates signals that regulate cellular processes such as cell growth, proliferation, differentiation, apoptosis, inflammation and motility [5,6]. Hence, perturbations in lipid metabolism are associated with a wide range of diseases, including type-2 diabetes, cardiovascular disease and cancer [7–9]. Lipidomics aims at a comprehensive determination of lipids in a biological material [10]. The diversity of lipids poses a great challenge on these analyses. Several different techniques, including nuclear magnetic resonance (NMR) and mass spectrometry (MS)

∗ Corresponding author. E-mail address: [email protected] (P. Spégel). https://doi.org/10.1016/j.chroma.2018.03.024 0021-9673/© 2018 Elsevier B.V. All rights reserved.

have been applied to this date [11]. Direct infusion MS, also called shotgun lipidomics, is impacted by ionization suppression and the inability to distinguish structural isomers [12,13]. Prior chromatographic separation reduces these problems. Reversed phase high performance liquid chromatography (RP-HPLC) is the most common, offering separation based on the acyl chain length and degree of unsaturation [14,15]. Normal phase liquid chromatography (NP-LC) provides separation of lipid classes, thereby offering a selectivity orthogonal to that obtained in RP-HPLC, but requires toxic solvents that are sometimes incompatible with MS detection, such as hexane or chloroform [16]. More recently, hydrophilic interaction liquid chromatography (HILIC) has been recognized as an MS compatible alternative to NP-LC [17]. However, separation of lipids with similar hydrophobicity, such as cholesteryl esters (CEs), triacylglycerol (TGs) and diacylglycerol (DGs) is challenging in HILIC [18]. Supercritical fluid chromatography (SFC) is an environmentally sustainable alternative to normal phase chromatography. The low viscosity of supercritical CO2 realizes faster mass transfer and improved separation efficiency, as compared to HPLC [19,20]. In

S. Al Hamimi et al. / J. Chromatogr. A 1548 (2018) 76–82

addition, SFC shows higher sensitivity for less polar lipid classes compared to HILIC and RP-HPLC [21]. Selectivity and separation efficiency can be modulated by changes in the temperature, pressure, the flow rate of the mobile phase and its composition. Several SFC methods have been developed for lipid profiling using conventional packed columns, revealing increased selectivity and resolution compared to RP-HPLC and NP-LC [22,23]. Further improvements in separation efficiency has been achieved using ultra-high performance supercritical fluid chromatography (UHPSFC) using columns with sub-2 !m particles [24]. UHPSFC columns with a variety of chemistries have recently been developed. BEH silica columns have been used for lipid class separation [25], BEH 2-EP for analysis of TGs and DGs [26] and HSS C18 for analysis of fatty acids (FAs) [27]. Orthogonality between BEH and HSS C18 was utilized for development of a separation method for intra-class species based on acyl chain hydrophobicity [28]. However, these columns have not been systematically evaluated for their applicability in lipidomics. In this work, we aim to screen the performance of seven sub-2 !m columns for lipid analysis using UHPSFC/electrospray ionization-quadrupole time-of-flight (ESI-QTOF)-MS.

2. Materials and methods 2.1. Chemicals and reagents Ultrapure CO2 (99.999%) was provided by Air Products (Amsterdam, Netherlands). LC–MS grade methanol, 2-propanol and acetonitrile were purchased from Scharlau (Barcelona, Spain). Heptane (99.5%) was purchased from Alfa Aesar (Karlsruhe, Germany). HPLC grade chloroform was from VWR (VWR Chemicals, France). Methyl tert-butyl ether (MTBE; 99.8%) and 3-sn-phosphatidylethanolamine (≥98%) were provided by SigmaAldrich (Steinheim, Germany). Porcine total lipid extract from brain was from Avanti Polar lipids (Alabaster, AL) and containing PC (12.6%), PE (33.1%), PI (4.1%), PS (18.5%), PA (0.8%) and unknown (30.9%; all in w/w). MS-grade ammonium formate and ammonium acetate were purchased from Sigma-Aldrich (St. Louis, MO). Water was purified using a Milli-Q purification system (Millipore, Billerica, MA).

2.2. Sample preparation Lipids were extracted as previously described in detail [29]. Briefly, serum (40 !L), was extracted with methanol (300 !L) and MTBE (1 mL) by gentle vortexing for 1 h (VX-2500 Multi Tube Vortexer, VWR, West Chester, PA). Then, water (250 !L) was added followed by 10 min vortexing and centrifugation (12000 × g, 10 min). The upper phase was transferred to an Eppendorf tube and dried under a nitrogen flow. Finally, samples were re-dissolved in 0.5 mL of methanol:choloroform (1:2, v:v), if not stated differently elsewhere, and centrifuged (12000 × g, 2 min). A pooled sample was created by pooling several lipid extracts and spiking with 3-sn-phosphatidylethanolamine and brain total lipid extract (50 !g/mL). This pool was aliquoted, stored at −80 ◦ C, and used throughout the study unless stated differently elsewhere. For investigations on the impact of sample solvent on chromatographic performance, pooled sample was evaporated to dryness and re-dissolved in 0.250 mL chloroform:methanol (1:1, v:v), chloroform:methanol (2:1, v:v), isopropanol:acetonitrile (1:1, v:v), isopropanol:acetonitrile:chloroform (1:1:1, v:v:v) or a mixture of all above mentioned solvents (1:1:1:1, v:v:v:v).

77

2.3. UHPSFC/ESI-QTOF-MS Samples were analysed on an Acquity UPC2 (Waters, Milford, MA). Seven columns (100 mm × 3 mm) were investigated (Supplementary material Fig. A1): Acquity UPC2 Torus DIOL (Diol), Acquity UPC2 Torus 2-PIC (2-PIC), Acquity UPC2 Torus DEA (DEA), Acquity UPC2 Torus 1-AA (1-AA), Acquity UPC2 CSH Fluoro-phenyl (CSHFP), Acquity UPC2 BEH (BEH) and Acquity UPC2 HSS C18 SB (HSS-C18 ) (Waters). The particle size was 1.7 !m, except for the HSS-C18 column in which the particle size was 1.8 !m. For the screening process, the injection volume was 1 !L, the flow rate was 1.6 mL/min, the column temperature 50 ◦ C, and the active back pressure regulator (ABPR) was set at 125 bar. Methanol containing 20 mM ammonium formate was used as mobile phase modifier using the following gradient: 0 min, 1% modifier; 18 min, 50%; 19 min, 50%; 19.5 min 1%; 21 min 1%. The injector needle was consecutively washed with methanol and chloroform after each injection. The Acquity UPC2 was connected to a Xevo 2G QTOF-MS (Waters). Two T-pieces (Waters) were used to control the back pressure at the column outlet and to infuse methanol (0.25 mL/min) as a makeup liquid. The MS was operated in positive and negative ESI mode with a scan range of m/z 80–1200. In positive ESI mode, the capillary voltage was 3.2 kV, the sampling cone voltage 34 V, the source temperature 120 ◦ C, the drying gas temperature 420 ◦ C, the cone gas flow 40 L/h and the drying gas flow 690 L/h. In negative ESI mode, the capillary voltage was 2.6 kV, the sampling cone voltage 42 V, the source temperature 120 ◦ C, the drying gas temperature 390 ◦ C, the cone gas flow 40 L/h and the drying gas flow 740 L/h. A collision energy ramp between 15 and 55 eV was used for MSE , targeted MS/MS and data dependent MS/MS in both positive and negative ESI. Data was acquired using MassLynx v4.1 (Waters).

2.4. Optimization of chromatography and mass spectrometry settings A face centered central composite design (FC-CCD) with three center points was created in MODDE 10.1 (Sartorius Stedim Biotech, Malmö, Sweden) and used to examine the effect of column temperature (40–60 ◦ C), flow rate (1.0–1.8 mL/min) and back pressure at column outlet (110–160 bar), on the retention time (tR ), resolution (Rs ), peak height (h) and peak width (full width half maximum, FWHM) for a selection of the most abundant lipids in human plasma [30]. Another FC-CCD design was used to examine the impact of capillary voltage (Cap.V; 2–4 kV for negative ESI and 1.5–3.5 kV for positive ESI), cone voltage (Cone.V; 10–50 V), desolvation gas temperature (Des.T; 200–600 ◦ C) and flow rate of desolvation gas (Des.F; 300–900 L/h) on the peak areas of CE 18:3 and PE 38:3. All responses were mean centered and scaled to unit variance. Projections to latent structures (PLS) was used to calculate the models. The model was evaluated by investigation of coefficient and contour plots. The best analysis conditions were determined using simplex optimization.

2.5. Data analysis Raw data were processed using the open source software package MZmine 2.16 (http://mzmine.sourceforge.net/) [31]. Data were generated by targeted peak detection with a m/z tolerance of 0.5 mDa or 5 ppm and a retention time tolerance of 0.2 min. Data were visualized by principal component analysis (PCA) calculated in SIMCA-P+ 12.0.1 (Sartorius Stedim Biotech, Malmö, Sweden). Groups were compared using ANOVA with Tukeys test post hoc.

78


Fig. 1. Influence of UHPSFC column type on resolution (Rs ), retention time (tR ), peak height (h) and peak widths (full width half maximum, FWHM) for 54 lipids (Supplementary material Table A1). Data were analysed by principal component analysis (PCA), to visualize similarities and differences between the columns in a single figure. The score scatter plot (A) reveals a systematic difference in the performance of columns containing pure silica particles (BEH) and particles with non-polar ligands (HSS-C18 ) and polar ligands (DEA, 2-PIC, 1AA, Diol). The corresponding loading plot (B) reveals which factors are driving the clustering observed in the score plot. Hence, resolution is generally higher for most lipids on the Diol column. R2 = 0.87, Q2 = 0.72, for 4 components.

3. Results and discussion 3.1. Factor screening Initially, applicable temperature, back pressure at column outlet and flow rate ranges were explored using the 2-PIC column. The column temperature was evaluated at four levels (40, 50, 60 and 80 ◦ C). Resolution generally increased with temperature due to decrease in the density resulting in a decreased elution strength. The resolution of two early eluting homologues, CE 16:0 and CE 20:4, increased from 1.7 at 40 ◦ C to 3.4 at 60 ◦ C. In contrast, the resolution for two late eluting phosphatidylcholines (PCs), PC 32:0 and PC 40:6, decreased by about two fold from 3.9 to 2.2. At the highest temperature (80 ◦ C) the resolution of early eluting lipids was seriously compromised due to peak broadening. The maximum flow rate of the system was limited to 1.8 mL/min by the lowest temperature included in the design of experiments and the pressure limit of the system (414 bar). In conclusion, the range of factor settings applicable in the experimental design was determined to be temperature (40–60 ◦ C), flow rate (1.0–1.8 mL/min) and back pressure at column outlet (110–160 bar).

3.2. Mobile phase modifier and additives The dipole moment of CO2 is similar to that of hexane, i.e. negligible, which necessitates the use of modifiers to increase the elution strength for compounds that interact strongly with the stationary phase. Methanol is the most frequently used modifier, with the benefit of promoting dissolution of mobile phase additives, such as ammonium acetate or formate [25]. Additives, such as ammonium salts have been suggested to deactivate free silanol groups on the stationary phase, thereby preventing secondary retention mechanisms [32]. Hence, in the present study ammonium formate or acetate were added to the mobile phase and not to the make-up solvent. Further motivating the choice of these additives are their ability to promote formation of ammonium adducts [M+NH4 ]+ , thereby facilitating ionization of CEs, TGs and DGs [33]. Our results show that the signal-to-noise ratio (S/N-ratio) of lyso-PC 22:5 (LPC 22:5) decreased from 73 to 34 when ammonium formate was exchanged for ammonium acetate. The S/N-ratio for the less abundant lipid LPC 22:6 was 15 in presence of ammonium formate but fell below limit of detection (LOD) when ammonium acetate was used. FWHM of most peaks corresponding to polar lipids (late eluting) were lower with ammonium formate as compared to ammonium acetate. Ammonium formate also enhanced

the detectability of most lipid classes in positive ESI mode compared to ammonium acetate, particularly for late eluting lipids such as LPCs and phosphatidylserines (PSs). This effect can be attributed to a slightly lower pH and less ionization suppression with ammonium formate as compared to ammonium acetate. Our results are in agreement with one previous study [34], but conflicting results have also been obtained [25]. Increasing the concentration of ammonium formate from 10 mM to 20 mM improved the peak symmetry for zwitterionic lipids. Moreover, ionization efficiency for neutral ammonium adduct forming lipids, such as CEs, TGs and DGs, which are eluting at a low cosolvent percentage, improved. S/N-ratios were significantly decreased when the additive concentration was further increased to 30 mM. Hence, methanol with 20 mM ammonium formate was used in the subsequent analyses.

3.3. Column screening A generic gradient was selected as the basis for the column screening, aiming at identifying the column yielding the narrowest, most symmetric and intense peaks with a high resolution. The rationale for this being that column chemistry is the main factor controlling the performance of a separation in SFC [35]. However, individual optimization of conditions for all tested columns may yield additional benefits, although this would be highly time consuming. Retention time, peak height, FWHM, asymmetry factor and Rs , within and between lipid classes, were determined for selected lipids (Supplementary material Table A1), including FAs, TGs, DGs, monoacylglycerols (MGs), ceramides (Cers), CEs, hexosyl ceramides (HexCers), phosphatidylglycerols (PGs), sulfatides, sphingomyelins (SMs), PCs, LPCs, phosphatidylethanolamines (PEs), lyso PEs (LPEs), phosphatidylinositols (PIs) and phosphatidylserines (PSs) (chromatograms in Supplementary material Fig. A2). The score scatter plot from a PCA calculated on these data (described variation, R2 = 0.87; predictive ability, Q2 = 0.72) reveals a clear difference between columns. Columns with pure silica (BEH), nonpolar ligands (HSS C18 ) and polar ligands bonded on silica (DEA, 2-PIC, 1-AA and Diol) are well separated in the PCA score scatter plot (Fig. 1A). This clustering is driven by strong retention of nonpolar lipids such as CEs, TGs, DGs, and MGs on HSS-C18 and 1AA and strong retention of polar lipids on the BEH column (Fig. 1B). The elution order of nonpolar lipids is similar on 1-AA, Diol, 2PIC, DEA, CSH-FP and BEH (CEs < TGs < DGs < MGs). However, this order is altered on HSS-C18 (DGs < TGs < CEs < MGs). The stationary phase chemistry had a more pronounced impact on the elution


79

Fig. 2. Extracted ion chromatograms for PC 34:2, PI 38:4, SM 38:1, PS 36:1, PE 32:1, LPE 18:1 and LPC 16:0 on the investigated columns. Elution order and peak shapes vary between all columns. Zoom-ins are shown in Supplementary material Fig. A3.

order of later eluting polar lipids (Fig. 2). The BEH column showed the strongest retention of polar lipids, and hence the longest analysis time among investigated columns. This is likely due to strong interactions between lipids and polar silanol groups on the surface of the particles. The zwitterionic lipids with terminal primary amines, PEs, LPEs and PSs showed very strong retention and wide peaks on columns containing "-amino alcohol ligands, 2-PIC, DEA and 1-AA (Supplementary material Fig. A3). PS species, having an carboxylic acid functionality in addition to the amine and the phosphoryl group, also showed strong retention and distorted peaks on BEH, HSS-C18 and CSH-FP. A higher concentration of ammonium formate in the modifier ameliorated these problems for PS, but this was associated with a reduced S/N-ratio. Peaks containing nonpolar and polar lipids were generally wider on the HSS-C18 and BEH columns, respectively. PE, LPE and PS species showed low intensities on DEA, 2-PIC and 1-AA. For example, the intensity of PE 32:1 was about 25 times lower on 2-PIC than on Diol. Peaks, in particular those corresponding to strongly retained lipids, were more symmetric on the Diol, as compared to the other tested columns (Supplementary material Table A1). Moreover, the Diol column produced a larger number of baseline separated lipid classes (11 pairs as compared to 9 for the second best column, 2-PIC) and within lipid class separated lipid species (14 lipid pairs versus 8 for 2-PIC). A high between lipid class resolution is desirable in quantitative analysis using lipid class standards. However, a very narrow peak may produce high lipid concentrations, promoting lipid aggregation. Moreover, competition for space and charge may occur, compromising detection of low abundant lipids. Within lipid class resolution may hence allow for injection of more concentrated lipid samples, thereby facilitating analysis of low abundant lipid species [36,37]. This is exemplified by the peak area for one of the low abundant PE species that we could determine in our study, PE 34:0 [30], being linearly dependent on the within lipid class resolution (R2 = 0.86, p < 0.01). In general, separation of lipid classes follows a similar order as in normal phase chromatography, with nonpolar lipids eluting before the more polar species. However, retention also depended on the acyl chain length and degree of unsaturation (Fig. 3). Retention increased with both acyl chain length and degree of unsaturation. The influence of unsaturations on retention has been shown on numerous polar columns, including 2-PIC, Diol, BEH and DEA columns [38], and suggested to relate to the polarity

Fig. 3. Extracted ion chromatogram for FAs (A) and LPCs (B) on the Diol column. The elution order of FAs (C) and LPCs (D) depend on the acyl-chain length and degree of unsaturation.

of the column [39]. The Diol column was used in the subsequent investigations due to the greater peak symmetries, higher intensities, narrower peak widths and higher resolutions obtained on this column as compared to the other tested columns. 3.4. UHPSFC conditions Chromatographic parameters were optimized for the Diol column using a FC-CCD. The same wide gradient starting with low elution strength as was used in the column screening, was applied in the method optimization to ensure retention of weakly retained lipids. The aim of the optimization was to identify flow-rate, column temperature and back pressure at column outlet settings yielding the best resolution. Hence, models were calculated for the within lipid class resolution of one early eluting lipid pair TG 52:0/TG 54:3 (R2 = 0.82, Q2 = 0.71) and one late eluting pair PC 34:0/PC 38:4 (R2 = 0.85, Q2 = 0.66). Temperature and back pressure determine the density and viscosity of the mobile phase and thus the elution strength, and are hence expected to influence the chromatographic performance. A high back pressure at the column outlet reduced retention of early eluting lipids. This is due to the compressibility of pure CO2 (or CO2 with low co-solvent content), which results in an increased density of the eluent and consequently an increased solvation power early in the gradient. Consequently, retention of more polar lipids was less influenced due to the lower compressibility of the eluent containing a higher methanol fraction. Accordingly, a higher back pressure at the column outlet reduced resolution of early eluting lipids, but had no impact on resolution of late eluting lipids (Fig. 4). It must be noted that the co-solvent also directly impacts on the solvation of the analytes. A higher column temperature increased resolution of early eluting TG 52:0 and TG 54:3 (Fig. 4). Resolution increased from 1.8 to 2.6 when the temperature was increased from 40 to 60 ◦ C. On the other hand, resolution of the late eluting lipids, PC 34:0 and PC 38:4, decreased three-fold when the temperature was increased from 40 to 60 ◦ C. In addition, co-elution of SMs and PEs occurred

80


Fig. 4. Contour plots from a face centered central composite design (FC-CCD) describing the impact of chromatographic conditions (temperature, back pressure and flow rate) on resolution of TG 52:0/TG 54:3 and PC 34:0/PC 38:4. Temperature and flow rate has opposite effects on the resolution of early and late eluting lipid pairs. Pressure has a minor impact. Hence, an intermediate temperature and flow rate is optimal, confirming results obtained from the simplex optimization.

at elevated temperature (data not shown). Hence, an intermediate column temperature was deemed optimal, with the added advantage of providing a lower back pressure as compared to the lower temperatures. An increased flow rate reduced the resolution of weakly retained lipids, whereas it improved resolution of late eluting lipids. This may be explained by the increased back-pressure resulting from a higher flow rate, causing density to increase, thereby increasing the elution strength. Furthermore, the impact of pressure drop between the inlet and the outlet of the column is more pronounced early in the gradient due to the low co-solvent content. A high pressure drop has been shown to cause peak distortion [40]. Due to the difficulties in assigning the optimal condition we applied a simplex optimizer function to derive the best separation condition. The best separation conditions, showing maximal resolution, was found at a column temperature of 48 ◦ C, a back-pressure at the column outlet of 131 bar and a flow rate of 1.5 mL/min. 3.5. Injection solvent The sample solvent is known to impact on the peak shape [41]. It is recommended to dissolve the sample in a solvent that has less elution strength than the initial mobile phase, offering a means of pre-concentrating the sample in a narrow band at the top of the column. However, this is inevitably limited by restrictions in the solubility of the analytes in the solvent and the compatibility of the solvent with MS detection. In this study, five different solvent combinations, based on four different solvents, with polarity indexes ranging from 2 to 7 were tested [42] (Supplementary material Fig. A4). Peak area, FWHM and the asymmetry factor were evaluated for a weakly retained, CE 18:2, an intermediately retained, Cer 36:3 and a strongly retained, PS 36:2, lipid. The composition of the injection solvent was found to have a large impact on the peak

areas (Supplementary material Table A2). Isopropanol:acetonitrile (1:1) decreased the detected peak area of all species by 22–78% compared to chloroform:methanol (1:1). This was likely due to limited solubility of the lipids in this mixture, which was supported by observations of suspended particles. Isopropanol:acetonitrile (1:1) slightly reduced FWHM of early eluting peaks, resulting in an improved within class Rs of these, but caused broadening of late eluting peaks compared to the other tested solvents. Adding chloroform to this mixture generally increased peak areas and improved asymmetry factors. Overall, chloroform:methanol yielded the highest peak areas and the lowest FWHM. Increasing the ratio of chloroform to methanol to 2:1, had no impact on peak areas compared to the 1:1 ratio, but reduced peak width and improved asymmetry factors of all evaluated lipids (data not shown). A combination of all tested solvents reduced peak areas and resulted in poorer peak shapes as compared to chloroform:methanol (2:1, v:v) (Supplementary material Table A2.). Hence, chloroform:methanol (2:1, v:v) was found to be the best injection solvent. The influence of injection volume on the chromatographic performance was tested by injecting 1, 2, 3 and 5 !L of the lipid extract dissolved in chloroform:methanol (2:1, v:v) (Supplementary material Fig. A5). The peak area increased linearly (R2 = 0.991) and resolution was unaffected up to 3 !L. A further increase in injection volume resulted in a non-linear increase of the peak area and loss of resolution, particularly for the early eluting lipid classes. 3.6. MS conditions The diversity of lipid structures and change of mobile phase composition from 1 to 50% methanol during the run complicates optimization of the ion source parameters. To facilitate identification of the optimal settings, parameters were evaluated during chromatographic analyses. Models were calculated for


81

efficiency was found in negative ESI mode (Supplementary material Fig. A6). Due to the opposite impact of some of the MS parameters on the behaviour of early and late eluting lipids, an attempt was made to achieve a constant composition of the eluent reaching the ionization source using an inverse make-up solvent gradient. However, this resulted in severe dilution of early eluting lipids. Moreover, also the peak areas of late eluting lipids were reduced. Finally, a simplex optimizer was applied to find settings resulting in high ionization efficiency for both early and late eluting lipids. The highest response in positive ESI mode were found for a capillary voltage of 3.2 kV, a cone voltage of 34 V, a desolvation gas temperature of 420 ◦ C and desolvation gas flow rate of 690 L/h. For negative ESI mode a capillary voltage of 2.6 kV, a cone voltage of 42 V, a desolvation gas temperature of 390 ◦ C and a desolvation gas flow rate of 740 L/h were found to yield the highest responses. Alternative settings may improve targeted analysis of either polar or nonpolar lipids as hinted in Fig. 5. Fig. 5. Coefficient plots from a face centered central composite design (FC-CCD) showing the influence of MS parameter settings on the peak area for the early and late eluting lipids CE 18:3 and PE 38:4, respectively. Cap. V, capillary voltage; Cone. V, cone voltage; Des. T, desolvation gas temperature; Des. F, desolvation gas flow rate.

two responses; the peak areas for one weakly retained lipid, CE 18:3 (R2 = 0.88, Q2 = 0.72) and one strongly retained lipid, PE 38:4 (R2 = 0.82, Q2 = 0.67). The early and late eluting lipids are differently influenced by the MS settings (Fig. 5). For both lipids the capillary voltage was without impact, whereas the cone voltage was found to be one of the most influential parameters. The peak areas of CE 18:3 and PE 38:3 were increased by 55% and 42%, respectively, when the cone voltage was increased from 10 to 45 V. The desolvation gas temperature had opposite impact on ionization efficiency for these lipids. CE 18:3 showed a decreased peak area with increasing desolvation gas temperature, whereas the peak area for PE 38:3 increased with desolvation gas temperature. This is likely to result from the high content of methanol in the mobile phase at the end of the gradient, requiring a higher temperature for efficient desolvation of PE 38:3. The ionization efficiency for PE 38:3 was further improved by a higher desolvation gas flow rate, whereas this parameter was ineffective in altering the response for CE 18:3. A similar influence of the investigated parameters on the ionization

3.7. Method validation and application Finally, the gradient was adapted to improve within and between lipid class resolution. A slow initial gradient (1–4% modifier in 3 min) improved the resolution between and within CEs and TGs. This was followed by a second steeper gradient (4–19% modifier in 1 min) and a third steep gradient (19–45% modifier in 3 min) to enhance resolution of polar lipids (Fig. 6). The repeatability and reproducibility of the method were evaluated for retention time and peak area for CE 18:2, Cer 36:2 and PS 36:3 in positive ESI mode (Supplementary material Table A3). Repeatability (within day n = 5) ranged from 0.5 to 1.2% RSD for retention time and 0.9–2.1% RSD for peak area. Reproducibility (between days n = 5) ranged from 0.8 to 1.4% RSD for retention time and 2.1–3.2% RSD for peak area. Finally, the method was applied to analyse lipids from the pure nonspiked serum extract (Supplementary material Fig. A7), affording detection of more than 180 lipid species within 15 lipid classes (Supplementary material Table A4). 4. Conclusion In this study, the selectivity of seven UHPSFC columns for separation of major lipid classes was evaluated. The Diol column showed improved resolution when compared with the columns having "amino alcohol ligands. The non-polar columns showed very poor

Fig. 6. Chromatograms obtained for serum spiked with polar lipid brain extract using the Diol column and the optimized method in positive (A) and negative (B) ESI mode, (ESI(+) and ESI(−), respectively).

82


retention of polar lipids. The method developed on the Diol column provided within lipid class separation, which is not commonly achieved in NP-LC or HILIC, in addition to between lipid class separation, which is unfeasible with RP-LC. Duality of interests The authors have no conflicts of interest to declare. Acknowledgements This work was supported by Lund University Antidiabetic Food Centre (LU-AFC, Vinnova) and The Swedish Research Council Formas. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.chroma.2018.03. 024. References [1] E. Fahy, D. Cotter, M. Sud, S. Subramaniam, Lipid classification, structures and tools, Biochim. Biophys. Acta 1811 (11) (2011) 637–647. [2] G. van Meer, D.R. Voelker, G.W. Feigenson, Membrane lipids: where they are and how they behave, Nat. Rev. Mol. Cell Biol. 9 (2) (2008) 112–124. [3] S.P. Colgan, Lipid mediators in epithelial cell–cell interactions, Cell. Mol. Life Sci. 59 (5) (2002) 754–760. [4] W. Tong, D. Shah, J.F. Xu, J.A. Diehl, A. Hans, M. Hannink, G.Y. Sun, Involvement of lipid mediators on cytokine signaling and induction of secretory phospholipase A(2) in immortalized astrocytes (DITNC), J. Mol. Neurosci. 12 (2) (1999) 89–99. [5] Y.A. Hannun, L.M. Obeid, Principles of bioactive lipid signalling: lessons from sphingolipids, Nat. Rev. Mol. Cell Biol. 9 (2) (2008) 139–150. [6] J.R. Krycer, L.J. Sharpe, W. Luu, A.J. Brown, The Akt-SREBP nexus: cell signaling meets lipid metabolism, Trends Endocrinol. Metab. 21 (5) (2010) 268–276. [7] K.A. Balogun, Lipid metabolism and the risk factors of cardiovascular disease: implication of dietary omega-3 polyunsaturated fatty acids, Appl. Physiol. Nutr. Metab. 41 (2) (2016) 223. [8] I. Raz, R. Eldor, S. Cernea, E. Shafrir, Diabetes: insulin resistance and derangements in lipid metabolism. Cure through intervention in fat transport and storage, Diabetes Metab. Res. Rev. 21 (1) (2005) 3–14. [9] C.R. Santos, A. Schulze, Lipid metabolism in cancer, FEBS J. 279 (15) (2012) 2610–2623. [10] M.R. Wenk, The emerging field of lipidomics, Nat. Rev. Drug Discov. 4 (7) (2005) 594–610. [11] M. Li, L. Yang, Y. Bai, H.W. Liu, Analytical methods in lipidomics and their applications, Anal. Chem. 86 (1) (2014) 161–175. [12] G. Isaac, Electrospray ionization tandem mass spectrometry (ESI–MS/MS)-based shotgun lipidomics, Methods Mol. Biol. 708 (2011) 259–275. [13] X. Han, R.W. Gross, Shotgun lipidomics: electrospray ionization mass spectrometric analysis and quantitation of cellular lipidomes directly from crude extracts of biological samples, Mass Spectrom. Rev. 24 (3) (2005) 367–412. [14] T. Cajka, O. Fiehn, Comprehensive analysis of lipids in biological systems by liquid chromatography-mass spectrometry, Trends Anal. Chem. 61 (2014) 192–206. [15] J. Willmann, H. Thiele, D. Leibfritz, Combined reversed phase HPLC, mass spectrometry, and NMR spectroscopy for a fast separation and efficient identification of phosphatidylcholines, J. Biomed. Biotechnol. 2011 (2011). [16] P.M. Hutchins, R.M. Barkley, R.C. Murphy, Separation of cellular nonpolar neutral lipids by normal-phase chromatography and analysis by electrospray ionization mass spectrometry, J. Lipid Res. 49 (4) (2008) 804–813. [17] C. Zhu, A. Dane, G. Spijksma, M. Wang, J. van der Greef, G. Luo, T. Hankemeier, R.J. Vreeken, An efficient hydrophilic interaction liquid chromatography separation of 7 phospholipid classes based on a diol column, J. Chromatogr. A 1220 (2012) 26–34. [18] E. Maciel, M. Leal, A. Lillebø, P. Domingues, M. Domingues, R. Calado, Bioprospecting of marine macrophytes using MS-based lipidomics as a new approach, Mar. Drugs 14 (3) (2016) 49. [19] D.R. Gere, R. Board, D. McManigill, Supercritical fluid chromatography with small particle diameter packed columns, Anal. Chem. 54 (4) (1982) 736–740. [20] F. Li, Y. Hsieh, Supercritical fluid chromatography-mass spectrometry for chemical analysis, J. Sep. Sci. 31 (8) (2008) 1231–1237.

[21] M. Lísa, E. Cífková, M. Khalikova, M. Ovˇcaˇcíková, M. Holˇcapek, Lipidomic analysis of biological samples: comparison of liquid chromatography, supercritical fluid chromatography and direct infusion mass spectrometry methods, J. Chromatogr. A 1525 (Suppl. C) (2017) 96–108. [22] J.W. Lee, S. Nishiumi, M. Yoshida, E. Fukusaki, T. Bamba, Simultaneous profiling of polar lipids by supercritical fluid chromatography/tandem mass spectrometry with methylation, J. Chromatogr. A 1279 (2013) 98–107. [23] T. Bamba, J.W. Lee, A. Matsubara, E. Fukusaki, Metabolic profiling of lipids by supercritical fluid chromatography/mass spectrometry, J. Chromatogr. A 1250 (2012) 212–219. [24] L. Nováková, A. Grand-Guillaume Perrenoud, I. Francois, C. West, E. Lesellier, D. Guillarme, Modern analytical supercritical fluid chromatography using columns packed with sub-2 !m particles: a tutorial, Anal. Chim. Acta 824 (2014) 18–35. [25] M. Lisa, M. Holcapek, High-throughput and comprehensive lipidomic analysis using ultrahigh-performance supercritical fluid chromatography-mass spectrometry, Anal. Chem. 87 (14) (2015) 7187–7195. [26] Q. Zhou, B.Y. Gao, X. Zhang, Y.W. Xu, H.M. Shi, L.L. Yu, Chemical profiling of triacylglycerols and diacylglycerols in cow milk fat by ultra-performance convergence chromatography combined with a quadrupole time-of-flight mass spectrometry, Food Chem. 143 (2014) 199–204. [27] M. Ashraf-Khorassani, G. Isaac, P. Rainville, K. Fountain, L.T. Taylor, Study of UltraHigh Performance Supercritical Fluid Chromatography to measure free fatty acids with out fatty acid ester preparation, J. Chromatogr. B Anal. Technol. Biomed. Life Sci. 997 (2015) 45–55. [28] J.W. Jones, C.L. Carter, F. Li, J. Yu, K. Pierzchalski, I.L. Jackson, Z. Vujaskovic, M.A. Kane, Ultraperformance convergence chromatography-high resolution tandem mass spectrometry for lipid biomarker profiling and identification, Biomed. Chromatogr. 31 (3) (2017), e3822-n/a. [29] V. Matyash, G. Liebisch, T.V. Kurzchalia, A. Shevchenko, D. Schwudke, Lipid extraction by methyl-tert-butyl ether for high-throughput lipidomics, J. Lipid Res. 49 (5) (2008) 1137–1146. [30] O. Quehenberger, A.M. Armando, A.H. Brown, S.B. Milne, D.S. Myers, A.H. Merrill, S. Bandyopadhyay, K.N. Jones, S. Kelly, R.L. Shaner, C.M. Sullards, E. Wang, R.C. Murphy, R.M. Barkley, T.J. Leiker, C.R. Raetz, Z. Guan, G.M. Laird, D.A. Six, D.W. Russell, J.G. McDonald, S. Subramaniam, E. Fahy, E.A. Dennis, Lipidomics reveals a remarkable diversity of lipids in human plasma, J. Lipid Res. 51 (11) (2010) 3299–3305. [31] T. Pluskal, S. Castillo, A. Villar-Briones, M. Oresic, MZmine 2: modular framework for processing, visualizing, and analyzing mass spectrometry-based molecular profile data, BMC Bioinf. 11 (2010) 395. [32] A. Cazenave-Gassiot, R. Boughtflower, J. Caldwell, L. Hitzel, C. Holyoak, S. Lane, P. Oakley, F. Pullen, S. Richardson, G.J. Langley, Effect of increasing concentration of ammonium acetate as an additive in supercritical fluid chromatography using CO2-methanol mobile phase, J. Chromatogr. A 1216 (36) (2009) 6441–6450. [33] R. Harkewicz, E.A. Dennis, Applications of mass spectrometry to lipids and membranes, Annu. Rev. Biochem 80 (2011) 301–325. [34] L. Novakova, M. Rentsch, A.G.G. Perrenoud, R. Nicoli, M. Saugy, J.L. Veuthey, D. Guillarme, Ultra high performance supercritical fluid chromatography coupled with tandem mass spectrometry for screening of doping agents. II: analysis of biological samples, Anal. Chim. Acta 853 (2015) 647–659. [35] T. Bamba, N. Shimonishi, A. Matsubara, K. Hirata, Y. Nakazawa, A. Kobayashi, E. Fukusaki, High throughput and exhaustive analysis of diverse lipids by using supercritical fluid chromatography-mass spectrometry for metabolomics, J. Biosci. Bioeng. 105 (5) (2008) 460–469. [36] J.C. Van De Steene, W.E. Lambert, Comparison of matrix effects in HPLC-MS/MS and UPLC-MS/MS analysis of nine basic pharmaceuticals in surface waters, J. Am. Soc. Mass Spectrom. 19 (5) (2008) 713–718. [37] X. Han, R.W. Gross, Shotgun lipidomics: electrospray ionization mass spectrometric analysis and quantitation of cellular lipidomes directly from crude extracts of biological samples, Mass Spectrom. Rev. 24 (3) (2005) 367–412. [38] F. Jumaah, R. Jedrkiewicz, J. Gromadzka, J. Namiesnik, S. Essen, C. Turner, M. Sandahl, Rapid and green separation of mono- and diesters of monochloropropanediols by ultrahigh performance supercritical fluid chromatography-mass spectrometry using neat carbon dioxide as a mobile phase, J. Agric. Food Chem. 65 (2017) 8220–8228. [39] K. Sakaki, Supercritical fluid chromatographic separation of fatty acid methyl esters on aminopropyl-bonded silica stationary phase, J. Chromatogr. A 648 (2) (1993) 451–457. [40] K. Kaczmarski, D.P. Poe, A. Tarafder, G. Guiochon, Pressure, temperature and density drops along supercritical fluid chromatography columns. II. Theoretical simulation for neat carbon dioxide and columns packed with 3-mum particles, J. Chromatogr. A 1250 (2012) 115–123. [41] V. Abrahamsson, M. Sandahl, Impact of injection solvents on supercritical fluid chromatography, J. Chromatogr. A 1306 (2013) 80–88. [42] T. Cajka, O. Fiehn, Toward merging untargeted and targeted methods in mass spectrometry-based metabolomics and lipidomics, Anal. Chem. 88 (1) (2016) 524–545.