Differential Mobility Spectrometry-Driven Shotgun Lipidomics

Aug 26, 2014 - AB SCIEX, 1201 Radio Road, Redwood City, California 94065, United States. •S Supporting Information. ABSTRACT: The analysis of lipids...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/ac

Differential Mobility Spectrometry-Driven Shotgun Lipidomics Tuulia P. I. Lintonen,† Paul R. S. Baker,‡ Matti Suoniemi,† Baljit K. Ubhi,§ Kaisa M. Koistinen,† Eva Duchoslav,‡ J. Larry Campbell,‡ and Kim Ekroos*,† †

Zora Biosciences Oy, Biologinkuja 1, Espoo, FI-02150, Finland AB SCIEX, 71 Four Valley Drive, Concord, Ontario, Canada § AB SCIEX, 1201 Radio Road, Redwood City, California 94065, United States ‡

S Supporting Information *

ABSTRACT: The analysis of lipids by mass spectrometry (MS) can provide in-depth characterization for many forms of biological samples. However, such workflows can also be hampered by challenges like low chromatographic resolution for lipid separations and the convolution of mass spectra from isomeric and isobaric species. To address these issues, we describe the use of differential mobility spectrometry (DMS) as a rapid and predictable separation technique within a shotgun lipidomics workflow, with a special focus on phospholipids (PLs). These analytes, ionized by electrospray ionization (ESI), are filtered using DMS prior to MS analysis. The observed separation (measured in terms of DMS compensation voltage) is affected by several factors, including the m/z of the lipid ion, the structure of an individual ion, and the presence of chemical modifiers in the DMS cell. Such DMS separations can simplify the analysis of complex extracts in a robust and reproducible manner, independent of utilized MS instrumentation. The predictable separation achieved with DMS can facilitate correct lipid assignments among many isobaric and isomeric species independent of the resolution settings of the MS analysis. This leads to highly comprehensive and quantitative lipidomic outputs through rapid profiling analyses, such as Q1 and MRM scans. The ultimate benefit of the DMS separation in this unique shotgun lipidomics workflow is its ability to separate many isobaric and isomeric lipids that by standard shotgun lipidomics workflows are difficult to assess precisely, for example, ether and diacyl species and phosphatidylcholine (PC) and sphingomyelin (SM) lipids.

L

identified and accurately quantified from total lipid extracts in a relatively short analysis time.6 The approach is also renowned for its simplicity and speed for providing comprehensive characterization of lipidomes without compromising data quality.7,8 Integration of automated sample preparation,9 sample introduction systems10 and software tools11,12 has further enabled the performance of such analyses at high throughput and sensitivity without compromising data quality.13 In fact, a recent study into the long-term performance and stability of molecular shotgun lipidomics14 has demonstrated remarkable robustness, meeting good-laboratorypractice method requirements without any need of postdata corrections. This demonstrates that molecular shotgun lipidomics is directly applicable to large scale studies and regulatory environments, such as in clinical diagnostics and food industry. Nevertheless, a limitation of direct infusion ESI-MS techniques can be the precise determination of isobaric and isomeric species, of which typically lipidomes are particularly rich. The classical approach taken to circumvent this challenge is the application of up-front species separation by liquid chromatography (LC). Although this has shown to be very

ipids are fundamental constituents that structurally and chemically regulate cell membranes, store energy, and can become precursors to bioactive metabolites.1 The lipids present, in concert with their local concentrations, determine the proper function of such central cellular events. Therefore, a defect in lipid regulation and metabolism can have deleterious outcomes on the cell or organism and can assist in the pathophysiology cascade of diseases. It is known that major diseases, such as coronary artery disease2 and infectious diseases,3 all have a lipid component in their epidemiology. Consequently, lipidome-wide analyses have become pivotal in revealing underlying biological processes and identifying new diagnostic biomarkers. Lipidomes of eukaryotic cells are highly complex, comprising thousands of diverse molecular species.4 Nearly 40 thousand unique structures of lipids have been catalogued in the most comprehensive lipid structure database (LIPID MAPS, http:// www.lipidmaps.org). This lipidome complexity is confounded by the fact that the absolute quantity of individual molecular lipids can differ up to several million-fold, depending on the matrix of origin. A popular and effective lipidomics workflow is molecular shotgun lipidomics, which involves direct infusion of lipid extracts into an MS using electrospray ionization (ESI).5 A major benefit of this technology is that hundreds of lipids at both the sum level and at the molecular level can be directly © 2014 American Chemical Society

Received: June 12, 2014 Accepted: August 26, 2014 Published: August 26, 2014 9662

dx.doi.org/10.1021/ac5021744 | Anal. Chem. 2014, 86, 9662−9669

Analytical Chemistry

Article

workflows and how such orthogonal separation improves the quality of the results and confidence in the identification of lipid species. Together, this workflow adds a new dimension in shotgun lipidomics producing more comprehensive and precise outputs.

powerful, it still suffers from the insufficient separation capacity to resolve every single lipid entity in a single analysis. Instead, a large battery of highly dedicated LC methods is typically required.15 Moreover, retention-time drift and variations in ion intensities16 are common drawbacks that, in concert with the underlying matrix effects (e.g., ion suppression), can lead to inaccurate quantification.13 Isobaric and isomeric species of total lipid extracts can also be resolved by shotgun lipidomics approaches. Here, lipid characteristic fragment ions, generated by MS/MS, are utilized to distinguish isobars.6 Moreover, ion trap MS3 analyses can be applied to determine positional isomers.10 However, potential pitfalls arise when isobaric species yield the same fragment ions. In this case, other fragment ions, if applicable, should be selected for proper identification and quantification. To minimize this, but also increase the MS response of the lipids of interest, selective sample extraction procedures,17 up-front lipid class separation9 and lipid chemistry inherent multidimensional approaches18 have been applied to reduce the sample complexity prior to the shotgun analysis. However, a common drawback of these approaches is that they can be laborious and thus preclude any high-throughput analysis. Ion mobility-mass spectrometry (IM-MS) has shown to be a powerful tool for the analysis of variety of compounds including lipids. 19 An alternative to this is differential mobility spectrometry (DMS),20 where ions are formed by ESI and transmitted at atmospheric pressure between two planar electrodes by a flow of gas. A high-voltage asymmetric waveform is applied across these electrodes, and the difference between the ion’s mobility during the high- and low-field portions of the waveform determines the exact trajectory taken. Separation of ions can be achieved as a function of the DC voltage required to focus the ions’ sawtooth trajectory toward the MS inlet. Thus, far, DMS has typically been employed to separate isobaric ions, which often convolve low-resolution MSbased quantitative experiments.21 However, DMS has also been used in higher resolution separations of structural isomers,22,23 stereoisomers,11 and tautomers.24 Beyond the DMS’s ability to separate ions based upon their native mass, charge, and structures, additional “chemical effects” can be introduced to these experiments. One such DMS-based workflow39 has recently been employed for the analyses of naphthenic acids.25 Separation of ions in DMS is not a simple correlation of increasing m/z with increasing compensation voltage (CV), like increasing drift time in ion mobility spectrometry (IMS)26 or in traveling wave ion mobility spectrometry (TWIMS).27 By adding volatile modifier molecules, like isopropanol, to the DMS transport gas, the solvation state of the analyte ions can be affected, thereby altering the apparent mobility of these species.20 This generally results in an increase in the peak capacity of the DMS experiments, which are otherwise hindered without such added modifiers.28 Here, we evaluated DMS in conjunction with molecular shotgun lipidomics. Of the chemical modifiers considered, npropanol permitted the best overall separation and data quality. This was shown to be highly reproducible and linear; independent of the mass spectrometry instrument used. Utilizing n-propanol we achieved nearly baseline separation of the analyzed lipid standards, representing different classes. Intramolecular clustering correlated with the type of phospholipid (PL) headgroup. We demonstrate the impact of DMS in existing targeted and untargeted shotgun lipidomics



EXPERIMENTAL SECTION Chemicals. Methanol, isopropanol, n-propanol, n-butanol, ultrapure water, ammonium acetate, acetic acid (all LC-MS grade), and 2,6-di-tert-butyl-4-methylphenol (BHT) were purchased from Sigma-Aldrich GmbH (Steinheim, Germany). Chloroform and n-heptane (HPLC grade) were purchased from Rathburn Chemicals Ltd. (Walkerburn, Scotland). All synthetic lipid standards were purchased from Avanti Polar Lipids Inc. (Alabaster, AL) except for d3-lyso-PC (LPC) 16:0 that was purchased from Larodan Fine Chemicals AB (Malmö, Sweden). Standard Preparation. A 5 μM synthetic lipid standard mixture of LPC 17:0, PC 17:0/17:0, phosphatidic acid (PA) 17:0/17:0, phosphatidylethanolamine (PE) 17:0/17:0, phosphatidylglycerol (PG) 17:0/17:0, phosphatidylserine (PS) 17:0/17:0, ether-linked PC (PC O-) 18:0/18:1, ether-linked phosphatidylethanolamine (PE O-) 18:0/18:1 and SM 18:1/ 12:0 was prepared in chloroform/methanol (1:2, v/v) containing 5 mM ammonium acetate. A serial dilution was prepared in following concentrations: 5, 1, 0.5, 0.1, 0.01, 0.001 μM, each with 1 μM of d3-LPC 16:0 in chloroform/methanol (1:2, v/v) containing 5 mM ammonium acetate for the calibration line. Samples. Fresh frozen human plasma from a healthy subject was purchased from Finnish Red Cross Blood Service (Helsinki, Finland). Dissected liver organ of ZDF-leprfa/crl male rat was purchased from Charles River (St. Germain Sur L’Arbesle, France) and after animal sacrifice the entire liver was immediately frozen and stored at −80 °C. See Supporting Information experimental section S-1 for more details. Lipid Extraction. Robotic-assisted 96-well sample preparation and extraction was performed using a Hamilton Microlab Star system (Hamilton Robotics AB, Kista, Sweden). A modified Folch protocol was applied to extract a broad lipidtype spectrum.9 See Supporting Information experimental section S-1 for more details. Mild-Alkaline Hydrolysis. Saponification of lipid extracts were performed according to Schnaar RL et al.29 See Supporting Information experimental section S-1 for more details. Differential Mobility Spectrometry−Mass Spectrometry. All experiments were conducted on QTRAP 5500 and QTRAP 6500 mass spectrometers (AB SCIEX, Concord, Canada) equipped with a SelexION DMS interface20 mounted in the atmospheric pressure region between the sampling orifice of the hybrid quadrupole linear ion trap mass spectrometer (QqLIT)30 and ESI source. The integrated system was controlled using Analyst 1.6.2 software. Several chemical modifiers, including n-propanol, 2-propanol, n-butanol, n-heptane, and chloroform were individually added to the curtain gas (transport gas for the DMS cell). Addition of a chemical modifier, 1.5% (v/v) in nitrogen, to the curtain gas was used to alter the DMS separation of the lipid ions. Resolving gas was employed to enhance the resolution of the DMS separation. Nitrogen was employed for the curtain gas, resolving gas, and MS/MS collision gas. A constant gas flow in 9663

dx.doi.org/10.1021/ac5021744 | Anal. Chem. 2014, 86, 9662−9669

Analytical Chemistry

Article

improves separation of lipids, it often requires extended analysis times, and still cannot separate lipids that are closely related in m/z space. Hence, shotgun lipidomics, including LC-based approaches, would benefit from reducing the complexity of the sample prior to analysis. Therefore, we sought another approach that employs gas-phase separations, examples of which have recently demonstrated separation of a few classes of PLs based upon chemical properties, via selective covalent34 and noncovalent35 modifications. Differential Mobility Spectrometry Exploits Chemical Effects. As an initial test, we assessed whether DMS can resolve PL classes, and what effect (if any) the presence of chemical modifiers would have on these separations. We prepared a mixture containing five different 17:0/17:0 phospholipid species and analyzed it on a QTRAP 6500 system using Q1 scanning in negative ion mode. With the separation voltage (SV) set at 4000 V, we ramped the compensation voltage (CV) under several conditions, with only nitrogen as the curtain/DMS transport gas (i.e., 0% added chemical modifier) or with one of either isopropanol, npropanol, n-butanol, heptane, or chloroform. For consistency, identical experimental settings were applied with the only exception that a higher DMS temperature was required for nbutanol. Only PC could be separated from the other PLs when no modifier is used (Supporting Information Figure S-1a). Heptane produces a very modest separation of the phospholipid classes whereas the separation improves using chloroform, facilitating separation of PC, phosphatidic acid (PA) and PE from phosphatidylglycerol (PG) and PS. Using nbutanol further improves the separation; however, the broad peak widths result in significant overlap between PC, PA, and PE. Interestingly, the use of n-propanol or isopropanol as the chemical modifier enables the highest resolution of all the monitored phospholipid classes (Supporting Information Figure S-1b and S-1c). Importantly, these alcohols permit a close to complete baseline separation of PC, PE, and PS, thereby circumventing the previously discussed isobar/isomer related challenges exemplified in Supporting Information Table S-2. These results demonstrate that chemical modifiers in the DMS transport gas lead to enhanced separation, which is in line with separations observed with analogous “lipids”−oil−sandsderived naphthenic acids.36 We also observed that the addition of resolving gas to the terminus of the DMS cell20 provides slightly extended residence times for the lipid anions in the DMS (i.e., from the standard residence time of ∼7 to ∼12 ms). This is essential as it increases the separation power of the process. Combined with the ion/molecule clustering between the lipid anions and the alcohol modifier, the added DMS residence time afforded the separation observed here. Notably, identical results were obtained irrespective on mass spectrometer used showing the reproducibility of the DMS systems (data not shown). Considering the Role of Intramolecular Solvation in DMS-Based Phospholipid Separations. Upon the addition of chemical modifiers to the DMS cell/environment, an ion’s optimal CV for transmission can shift, sometimes dramatically, compared to nitrogen-only environment.37 This shift, generally to more negative CV values, is postulated to result from the formation of ion/molecule clusters in the DMS cell, which have different mobility behavior than the ion in the absence of these solvent molecules. In addition, an ion’s structure has also been shown to play a key role in these ion/molecule interactions in

the DMS cell was achieved by the vacuum pumping of the MS system. The fundamental behavior of DMS devices is described elsewhere.31 The typical DMS-MS operating parameters employed in this study are listed in Supporting Information Table S-1, unless otherwise specified. The separation voltage (SV) was held at an optimum value while the CV was scanned. At each incremental value of CV, MS data was acquired either in Q1 (quadrupole 1) scan, precursor ion scan (PIS) or multiple reaction monitoring (MRM) modes by using optimized Q1/Q3 (quadrupole 3) m/z values for the particular lipid molecules.6,10 These data were plotted in the form of ionograms. The samples were infused at a flow rate of 20 μL/ min. Data Processing. The resulting DMS-MS data were processed in PeakView and LipidView software (AB SCIEX) together with in-house bioinformatic tools20 to extract the identified peak areas. Theoretical Methods. All calculations were performed using Gaussian 09 (Revision A.1)32 and the results visualized using GaussView 5.0.9. See Supporting Information experimental section S-1 for more details.



RESULTS AND DISCUSSION Lipidomes are recognized to be extremely complex, comprising many overlapping isobaric and isomeric species. For example, the following common brutto lipid species; phosphatidylserine (PS) 36:1, phosphatidylcholine (PC) 32:2 and phosphatidylethanolamine (PE) 40:7 coexist in many sample matrices, such as human plasma.32 The identical or close m/z values of each of these lipids hamper their precise determination by shotgun lipidomics (Supporting Information Table S-2). On the basis of the calculated mass of their respective molecular ions, they cannot be distinguished using quadrupole instrumentation in full scan mode. Although high mass resolving instrumentation, such as time-of-flight (TOF) and orbitrap mass spectrometers, permit distinguishing PE 40:7 from PC 32:2 and PS 36:1 in full scan mode (i.e., at resolving power >37 000), the latter two lipids will remain undistinguishable because of their identical masses. Thus, another means of analysis is required to differentiate these species. MS/MS is typically applied since lipid classes produces selective characteristic fragment ions (reviewed in Ståhlman et al.9). In this case, the headgroup fragment ions allow one to distinguish PS from PE. Acyl anions originating from attached fatty acids can further be utilized for the delineation of molecular species, which is, together with the headgroup information, the basis of molecular shotgun lipidomics.6 From this detailed information, most of the molecular lipids underlying the represented brutto species can be pinpointed. However, the broad existence of overlapping acyl ions complicates both the identification and quantification processes, which in the worst case become impeded. For instance, the acyl ion of m/z 253.2 would suggest only the presence of PC 16:1/16:1 (PC 32:2), if neither the PS headgroup ion, m/z 153.0, nor the m/z of 311.3 are present. However, the presence of PS 20:0−16:1 (PS 36:1) complicates the identification of PC 16:1/16:1, since the 16:1 signal distribution between both lipid species is unknown. In such case, fatty acyl ratios (i.e., sn-1/sn-2) could be utilized to estimate, not quantify, the presence of the PC species, if such a ratio dramatically exceeds a value of two.33 This complexity is one of the reasons that LC separation of lipids has been explored exhaustively. However, while LC 9664

dx.doi.org/10.1021/ac5021744 | Anal. Chem. 2014, 86, 9662−9669

Analytical Chemistry

Article

the DMS cell.37 For example, in the CV-ramping experiments conducted here (Supporting Information Figure S-1), the PL anions were transmitted through the DMS cell in the following order of increasing CV: PC, PA, PE, PG, and PS. We postulated that the DMS separates these anionic PLs in that order because each headgroup can, to varying degrees, participate in intramolecular charge solvation of the phosphate anion. If such intramolecular charge solvation is present, there should be less opportunity for intermolecular solvation of the phosphate anion with chemical modifier molecules added to the DMS cell. Ions that cluster more efficiently with the modifier should exhibit more negative CV shifts than ions that do not (e.g., where intramolecular solvation inhibits such cluster formation resulting in more positive CV shifts). To support this hypothesis, we calculated the lowest-energy structures for truncated models of each of the PL anions studied. These calculations show that the degree of intramolecular solvation of the phosphate anion increases in the same order as observed in Supporting Information Figure S-1 (data not shown). In addition to the evidence provided by the geometry optimization calculations, we also calculated the dipole moments for each PL (Supporting Information Figure S2), which correlated in a negative linear fashion with the optimal CV for transmission of each PL class (Figure 1). These

as well as the very close CV overlap of the PA and PE species (both having very similar gas-phase acidities).. In a second study, the addition of chemical modifiers to IMS buffer gas was used to attempt to enhance the separation of protonated diamines.39 However, some ions displayed less of a change in mobility in the presence of the modifiers, indicating less ion/ molecule cluster formation. The authors postulated that the formation of intramolecular salt bridges limited the interactions between modifier molecules and the ions that hindered formation of ion/molecule clusters. Shotgun Lipidomics Using DMS Facilitates More Precise Quantification. Showing that different PL classes can be resolved by DMS is an important first step. However, it is crucial that the separated species can be accurately quantified. To pinpoint this, we first investigated the raw detection response versus the concentration of the measured lipid. We found that the detection response was linear in the range of 2.5−3 orders of magnitude based on a serial dilution of PE 17:0/17:0 (Supporting Information Figure S-3) and with a lower linearity cutoff at approximately a raw peak area of 100 counts. Similar response curves and linearity cutoffs were also observed in other PL classes (data not shown). For accurate quantification, nonendogenous internal standards are typically added to samples prior to the extraction step. We have previously shown that class-selective single standards are applicable for absolute quantification of endogenous species in shotgun lipidomics.11,32 Therefore, we approached the quantitative assessment in a similar fashion using DMS and wanted to explore how the DMS results compared to standard molecular shotgun lipidomics.6 Because of the lack of available internal standards, the assessment was based on lyso-PCs (LPC); a dilution series of LPC 17:0 was prepared as above and containing a fixed amount of the deuterated d3-LPC 16:0. With DMS, the acquisition was performed by MRM; however, because MRM is not a preferred detection method for standard shotgun analyses, we chose to acquire this data by regular PIS.6 For optimal comparison, the instrument parameters were matched as closely as possible, with 10 scans summed in PIS mode to match the number of points measured for the MRM peak. The calibration lines obtained by DMS and standard shotgun lipidomics were in very good agreement, as shown in Figure 2. By performing linear regression analyses, both lines extend R2 values of 0.9 indicating linear response and this at least in the range of 2.5−3 orders of magnitude (from 0.01 to 5 μM). The lower limit of linearity is similar for both methods, although with DMS the cutoff might be slightly higher. This and the observed small deviation in the slopes, 1.88 with DMS and 1.66 without DMS, could potentially be due the technical discrepancies in the analyses. In context of the slopes, another reason behind the differences could be that LPC 17:0 and d3LPC 16:0 are almost baseline separated in DMS, while this is apparently not the case in standard shotgun lipidomics. Future studies are needed to pinpoint these aspects. We postulate that the lipid class selective internal standard approach13 is also applicable with DMS; however, for more accurate quantification, stable isotope labeled standards of the lipids of interest would be preferred. Negative Ion Mode Analysis of Extracts. Besides the desire to prove the DMS’ ability to be used for quantification, it is essential to prove that this technique can successfully interrogate complex biological samples. This entails confirmation that sample complexity does not alter the DMS-based

Figure 1. Correlation analysis of calculated headgroup dipole moment and obtained CV for individual PL standard from Supporting Information Figure S-1c. The dipole moment (D) of individual PL is shown on the y-axis and respective CV on the x-axis. An R2 value of 0.899 with significance (p-value) of 0.015 was obtained by linear regression analysis.

findings lend support to the theory that the greater the degree of intramolecular solvation of the PL anion, the less clustering that anion will undergo with chemical modifier molecules in the DMS cell. The ability of the PL anions to cluster with the negative charge is greater in PCs given the lack of intramolecular solvation of the anion site compared to the other PL head groups. For example, both PG and PS bear head groups that contain a hydroxyl functionality that can engage in intramolecular hydrogen bonding to the phosphate anion, which can serve to disrupt efficient ion/molecule clustering between the anions and the added alcohol molecules in the DMS cell. Hence, the anions will require more positive CVs for optimal transport through the DMS cell. This proposed mechanism is also supported by previous examples from the literature. For example, Thomas and coworkers sought to explain the relative gas-phase acidities of PLs38 and noted that the calculated lowest-energy structures of model PL anions displayed significant intramolecular solvation of the negative charge in the more acidic PL head groups. This finding correlates well with the relative CV shifts observed here, 9665

dx.doi.org/10.1021/ac5021744 | Anal. Chem. 2014, 86, 9662−9669

Analytical Chemistry

Article

Figure 2. y-axis show the peak area ratios obtained from serial dilution of LPC 17:0 and a fixed amount of d3-LPC 16:0 plotted against the LPC 17:0 concentration (x-axis) (see Experimental Section for more details). The R2 values obtained by linear regression analyses of respective raw calibration data (from 0.01 to 5 μM) is displayed (linear fit equations; DMS, Y = 1.869X + 0.1062 and no DMS, Y = 1.657X + 0.03867). In DMS experiment the lipids were monitored using the MRM transitions m/z 568.4 → 269.3 and m/z 557.4 → 258.2 respectively (filled circles) and in standard shotgun lipidomics by selecting these transitions in PIS of m/z 258.2 and m/z 269.3 respectively (open circles). Spectral peak areas were used for no DMS and ionogram peak areas for DMS MRM experiments. The acquisitions were performed in negative ion mode on a QTRAP 6500 system. d3-LPC 16:0 separated at approximately a CV of −18.5 V whereas LPC 17:0 separated 0.5−0.7 V later. Six replicates (n = 6) per dilution point including SEM (standard error mean) is represented. N.B. The log−log plot format is used to condense the visual form of our calibration data; the linear regression was performed on the raw data.

Figure 3. MRM traces of (a) the PL standards in solution and (b) selected PLs in complex rat liver extracts analyzed by DMS in negative ion mode using a QTRAP 5500 system. The MRMs corresponds to (a) LPC 17:0 (568.6/269.2), PC 17:0/17:0 (820.6/269.2), PA 17:0/ 17:0 (675.6/269.2), PE 17:0/17:0 (718.6/269.2), PG 17:0/17:0 (749.6/269.2), PS 17:0/17:0 (762.6/269.2) and (b) LPC 16:0 (554.6/ 255.2), PC 16:0−18:1 (818.6/255.2 and 818.6/281.2), PE 16:0−18:2 (714.7/255.2 and 714.7/279.2), PE 16:0−22:6* (762.5/255.2 and 762.5/283.3* (FA 22:6-CO2)), phosphatidylglycerol (PG) 16:0−18:1 (747.7/255.2 and 747.7/281.2) and PI 18:0−20:4 (885.6/283.3 and 885.6/303.3). The MRM transitions are presented as m/z of Q1/Q3. *Fragment ion of m/z 283.2 ([FA 22:6-CO2]−).

separation and resolution of the individual lipids in the sample. As anticipated, the CVs of monitored PL standards spiked into total lipid extracts of rat liver (Figure 3 upper panel) or human plasma (not shown) were not altered, with the separation (i.e., CV values and peak widths) reproduced similar to Supporting Information Figure S-1c. As an initial approach and demonstration purpose, we monitored a set of known molecular species present in rat liver. Figure 3 (lower panel) shows the MRM traces of a few selected molecular species representing different lipid classes present in rat liver.40 In line with the results of the standards, the endogenous molecular species were distinguished into their respective classes. Moreover, inclusion of phosphatidylinositol (PI) species shows that this class can also be readily distinguished and has the highest optimum CV of the monitored classes at approximately +5 V. Notably, the analysis of human plasma produced a closely resembling lipid class separation. Thus, the DMS can separate the PL classes in CV space in a manner independent of sample complexity or matrix type. The signals obtained for PS 16:0−18:0 (m/z 762.5/255.2 and 762.5/283.2) did not match the expected CV for PS. Conversely, in both MRM traces, the signals correspond to PE. Based on molecular mass, m/z 762.5, and measured acyl anions these signals refer to PE 16:0−22:6, since docosahexaenoic acid yields a fragment ion with m/z 283.2431 ([fatty acid (FA) 22:6CO2]−) that is isobaric with the acyl anion m/z 283.2642 of stearic acid FA 18:0. The presence of this lipid can be confirmed by MS/MS of m/z 762.5 producing the PE characteristic fragment ions corresponding to PE headgroup and acyl anions (data not shown) (see Supporting Information Table S-2). The same lipid has also been identified in the work of Retra and colleagues.41 Thus, the matching to the expected CV for respective lipid class allows determining the trans-

parency of the selected MRMs and identifies the extent of contamination and prevents possible latent misidentifications. Since the separation is selective and robust, we predicted that single Q1 or enhanced MS scans would be beneficial as highly rapid and sensitive lipid profiling approaches. An extracted Q1 acquisition of the same rat liver sample as above is shown in Figure 4a. The obtained CV values enable us to identify the lipids present unambiguously despite the low mass resolution of Q1. Notably, this is not feasible by a Q1 analysis without DMS. The DMS-Q1 analysis directly resolves the example given in Supporting Information Table S-2, showing the presence of PS 36:1 and PE 40:7, but the lack of PC 32:2 in rat liver (m/z of 788). This high selectivity and the removal of background noise by DMS provide sufficient sample complexity reduction, resulting in the easy-to-interpret mass spectrum. Although not pursued, we expect that this will improve the quantification of the detected species as the amount of contaminating species and isotopic overlaps are dramatically diminished. Taking a closer look at the DMS-Q1 results, we observed that the DMS not only offers separation of lipid classes but also some separation at the level of individual species. For example, the separation profile of PCs shows that the separation is dependent on the length and degree of unsaturation of attached fatty acids (Figure 4b). The more carbons the FAs contain the higher CV is needed for optimal separation. This result in a clear separation order of PC 34:1−PC 36:1−PC 38:1. In contrast, a higher degree of unsaturation requires lower CV for optimal separation, thus resulting in a separation order of 3 double bonds (DB)−2 DB−1 DB. The results further suggest that the separation window is wider for species having higher 9666

dx.doi.org/10.1021/ac5021744 | Anal. Chem. 2014, 86, 9662−9669

Analytical Chemistry

Article

Figure 4. Ionograms corresponding to (a) PLs and selective (b) PCs of rat liver extracts acquired by Q1 analysis in negative ion mode using a QTRAP 5500 system. Size of the squares shows the ion intensity. (a) The obtained molecular masses (shown) and CVs provide the proposed lipid identifications. (b) The narrow CV range between −8 and −2 V shows the separation order of the individual brutto PCs with their median CV presented.

number of total carbons in their FA tails. Similar findings have been observed in the DMS-based analyses of naphthenic acid samples.36 Positive Ion Mode Analysis of Extracts. Several lipids are preferably analyzed in positive ion mode due to their chemical nature, including PCs and sphingomyelins (SM), which contain a positively charged quaternary ammonium of the choline headgroup. Collision-induced dissociation of the cations of PCs and SMs yields the characteristic fragment ion of the phosphorylcholine moiety (m/z 184.1) that has been frequently selected for quantitative profiling in unprocessed lipid extracts.42 However, since monoprotonated molecular ions of PCs have even nominal masses, whereas ions of SMs have odd nominal masses, the ability to differentiate these two lipid classes is hampered on low-mass resolution instruments. Therefore, we evaluated if DMS was capable of separating PCs from SMs prior to MS analysis to improve the measurement of these classes. Figure 5 demonstrates the DMS separation of PC from SM species in human plasma analyzed by PIS of m/z 184.1 in positive ion mode. Using npropanol as chemical modifier enabled us to separate the abundant PC class from SM at their baseline levels. The SM species present in human plasma were unambiguously identified despite the presence of the abundant PCs in the sample. Moreover, this profile represents the expected SM composition of human plasma, verified by standard molecular shotgun lipidomics analysis of the same samples that have been treated by saponification to remove the esterified PCs but not the SMs by mild alkaline hydrolysis.29 Thus, DMS permits precise measurement of endogenous PCs and SMs directly from complex extracts. Notably, as observed in the negative ion mode analysis (Figure 4) the individual SM (and PC) species are separated in the DMS cell (data not shown). Based on the closely overlapping endogenous SM profiles obtained with and without DMS (Figures 5b and 5c), our results suggest that CV has no effect on the response factor of the measured lipids. Moreover, this is further strengthened by the identical PIS of m/z 184.1 profiles (containing both PC and SM) obtained by standard

molecular shotgun lipidomics analysis compared to the sum profile of Figure 5a (sum of CV from −3 to 6 V) (Supporting Information Figure S-5). Although more detailed examinations are required, our observations indicate as mentioned above that the lipid class selective internal standard approach13,14 is also applicable with DMS. Another challenge in PIS of m/z 184.1 analysis is the precise discrimination and quantification of ether-linked PCs as these species closely overlap with diacyl PCs with FAs containing odd numbers of carbons.7 While this challenge can be circumvented using high mass resolving instrumentation,34 other strategies often have to be undertaken on lower-resolving power systems. By using the DMS, we were able to resolve alkenyl PC O-18:0/ 18:1 from diacyl PC 17:0/17:0 (Supporting Information Figure S-4). The ether bond at the sn-1 position of ether-linked phosphatidylcholine (PC O-) necessitates a CV value 1.5 V less that the diacyl PC to restore its trajectory through the DMS cell. A similar separation is observed for PE O-18:0/18:1 and PE 17:0/17:0, suggesting that the separation is determined, in part, by the ether-bond in addition to the headgroup. Comparable results could be obtained in negative ion mode, which further supports this finding. Even though the monitored alkenyl and acyl species are not isobaric (i.e., they differ by ∼10 Da), the good separation achieved in this example suggests that similar differentiation could also apply to isobaric species. We postulate that Q1 or MRM types of analyses are more preferred for DMS shotgun type of workflows. Moreover, enhanced separation through optimization of chemical modifier conditions (e.g., flow rates, chemical composition), pressure of resolving gas, and utilization of metal adducts, such as lithium43 or silver,44 might provide additional resolution for their precise determination even in PIS mode.



CONCLUSIONS The use of DMS in conjunction with direct infusion driven lipid analyses provides a beneficial dimension for shotgun lipidomics workflows. Its ability to discriminate isobaric and closely related lipids allows DMS to provide a comprehensive lipidomic output with properly assigned identification from even simple low 9667

dx.doi.org/10.1021/ac5021744 | Anal. Chem. 2014, 86, 9662−9669

Analytical Chemistry

Article

Notes

For research use only. Not for use in diagnostic procedures. The trademarks mentioned herein are the property of AB Sciex Pte. Ltd. or their respective owners. AB SCIEX is being used under license. The authors declare no competing financial interest.



ACKNOWLEDGMENTS We thank all members of our laboratories and partnered organizations for valuable discussions and suggestions, and the Finnish Funding Agency for Technology and Innovation (Tekes) for financial support.



(1) Shevchenko, A.; Simons, K. Nat. Rev. Mol. Cell Biol. 2010, 11, 593−598. (2) Tarasov, K.; Ekroos, K.; Suoniemi, M.; Kauhanen, D.; Sylvanne, T.; Hurme, R.; Gouni-Berthold, I.; Berthold, H. K.; Kleber, M. E.; Laaksonen, R.; Marz, W. J. Clin. Endocrinol. Metab. 2014, 99, E45−52. (3) Morita, M.; Kuba, K.; Ichikawa, A.; Nakayama, M.; Katahira, J.; Iwamoto, R.; Watanebe, T.; Sakabe, S.; Daidoji, T.; Nakamura, S.; Kadowaki, A.; Ohto, T.; Nakanishi, H.; Taguchi, R.; Nakaya, T.; Murakami, M.; Yoneda, Y.; Arai, H.; Kawaoka, Y.; Penninger, J. M.; Arita, M.; Imai, Y. Cell 2013, 153, 112−125. (4) van Meer, G. EMBO J. 2005, 24, 3159−3165. (5) Han, X.; Gross, R. W. Proc. Natl. Acad. Sci. U. S. A. 1994, 91, 10635−10639. (6) Ekroos, K.; Chernushevich, I. V.; Simons, K.; Shevchenko, A. Anal. Chem. 2002, 74, 941−949. (7) Schuhmann, K.; Almeida, R.; Baumert, M.; Herzog, R.; Bornstein, S. R.; Shevchenko, A. J. Mass Spectrom 2012, 47, 96−104. (8) Simons, B.; Kauhanen, D.; Sylvänne, T.; Tarasov, K.; Duchoslav, E.; Ekroos, K. Metabolites 2012, 2, 195−213. (9) Stahlman, M.; Ejsing, C. S.; Tarasov, K.; Perman, J.; Boren, J.; Ekroos, K. J. Chromatogr. B: Anal. Technol. Biomed. Life Sci. 2009, 877, 2664−2672. (10) Ekroos, K.; Ejsing, C. S.; Bahr, U.; Karas, M.; Simons, K.; Shevchenko, A. J. Lipid Res. 2003, 44, 2181−2192. (11) Ejsing, C. S.; Duchoslav, E.; Sampaio, J.; Simons, K.; Bonner, R.; Thiele, C.; Ekroos, K.; Shevchenko, A. Anal. Chem. 2006, 78, 6202− 6214. (12) Herzog, R.; Schuhmann, K.; Schwudke, D.; Sampaio, J. L.; Bornstein, S. R.; Schroeder, M.; Shevchenko, A. PLoS One 2012, 7, e29851. (13) Jung, H. R.; Sylvanne, T.; Koistinen, K. M.; Tarasov, K.; Kauhanen, D.; Ekroos, K. Biochim. Biophys. Acta 2011, 1811, 925−934. (14) Heiskanen, L. A.; Suoniemi, M.; Ta, H. X.; Tarasov, K.; Ekroos, K. Anal. Chem. 2013, 85, 8757−8763. (15) Owen, J. S.; Wykle, R. L.; Samuel, M. P.; Thomas, M. J. J. Lipid Res. 2005, 46, 373−382. (16) Wang, S. Y.; Kuo, C. H.; Tseng, Y. J. Anal. Chem. 2013, 85, 1037−1046. (17) Ejsing, C. S.; Sampaio, J. L.; Surendranath, V.; Duchoslav, E.; Ekroos, K.; Klemm, R. W.; Simons, K.; Shevchenko, A. Proc. Natl. Acad. Sci. U. S. A. 2009, 106, 2136−2141. (18) Han, X.; Yang, K.; Gross, R. W. Mass Spectrom. Rev. 2012, 31, 134−178. (19) Kliman, M.; May, J. C.; McLean, J. A. Biochim. Biophys. Acta 2011, 1811, 935−945. (20) Schneider, B. B.; Covey, T. R.; Coy, S. L.; Krylov, E. V.; Nazarov, E. G. Int. J. Mass Spectrom. 2010, 298, 45−54. (21) Jasak, J.; Le Blanc, Y.; Speer, K.; Billian, P.; Schoening, R. M. J. AOAC Int. 2012, 95, 1768−1776. (22) Barnett, D. A.; Purves, R. W.; Ells, B.; Guevremont, R. J. Mass Spectrom. 2000, 35, 976−980. (23) Blagojevic, V.; Chramow, A.; Schneider, B. B.; Covey, T. R.; Bohme, D. K. Anal. Chem. 2011, 83, 3470−3476.

Figure 5. Separation of PC and SM in human plasma samples by DMS in positive ion mode on a QTRAP 6500 system represented by (a) TIC from PIS m/z 184.1 experiment and (b) PIS of m/z 184.1 profile of SM separated between CV of 2.5−5 V in panel a. (c) SM representation of a saponified aliquot of the same sample monitored by PIS of m/z 184.1 using standard molecular shotgun lipidomics.7 DP was set to 60 V and CV step size to 0.1 V.

mass resolution full scan analyses and irrespective of sample complexity. Retention of linear response from DMS-enabled instrumentation further offers the possibility to quantify the monitored species. This lends itself to simplified profiling assays without compromising lipid identification - despite the use of lower-mass resolving power MS. In addition, given the instrument-independent performance of the DMS technology, one can generalize DMS practices and interchange easily between different shotgun workflows. Taken together, the addition of DMS to current shotgun lipidomics workflows provides a highly complementary tool that could help pave the way to higher-resolution lipidome analyses.



ASSOCIATED CONTENT

S Supporting Information *

Additional material as described in the text. This material is available free of charge via the Internet at http://pubs.acs.org



REFERENCES

AUTHOR INFORMATION

Corresponding Author

*E-mail: kim.ekroos@zora.fi. Tel: +358-40-7448997. 9668

dx.doi.org/10.1021/ac5021744 | Anal. Chem. 2014, 86, 9662−9669

Analytical Chemistry

Article

(24) Campbell, J. L.; Le Blanc, J. C.; Schneider, B. B. Anal. Chem. 2012, 84, 7857−7864. (25) Headley, J. V.; Peru, K. M.; Barrow, M. P. Mass Spectrom. Rev. 2009, 28, 121−134. (26) Woods, A. S.; Ugarov, M.; Egan, T.; Koomen, J.; Gillig, K. J.; Fuhrer, K.; Gonin, M.; Schultz, J. A. Anal. Chem. 2004, 76, 2187− 2195. (27) Kim, H. I.; Kim, H.; Pang, E. S.; Ryu, E. K.; Beegle, L. W.; Loo, J. A.; Goddard, W. A.; Kanik, I. Anal. Chem. 2009, 81, 8289−8297. (28) Shvartsburg, A. A.; Isaac, G.; Leveque, N.; Smith, R. D.; Metz, T. O. J. Am. Soc. Mass Spectrom. 2011, 22, 1146−1155. (29) Schnaar, R. L. Methods Enzymol 1994, 230, 348−370. (30) Guna, M.; Biesenthal, T. A. J. Am. Soc. Mass Spectrom. 2009, 20, 1132−1140. (31) Krylov, E. V.; Nazarov, E. G.; Miller, R. A. Int. J. Mass Spectrom. 2007, 266, 76−85. (32) Quehenberger, O.; Armando, A. M.; Brown, A. H.; Milne, S. B.; Myers, D. S.; Merrill, A. H.; Bandyopadhyay, S.; Jones, K. N.; Kelly, S.; Shaner, R. L.; Sullards, C. M.; Wang, E.; Murphy, R. C.; Barkley, R. M.; Leiker, T. J.; Raetz, C. R.; Guan, Z.; Laird, G. M.; Six, D. A.; Russell, D. W.; McDonald, J. G.; Subramaniam, S.; Fahy, E.; Dennis, E. A. J. Lipid Res. 2010, 51, 3299−3305. (33) Hsu, F. F.; Turk, J. J. Am. Soc. Mass Spectrom. 2003, 14, 352− 363. (34) Stutzman, J. R.; Blanksby, S. J.; McLuckey, S. A. Anal. Chem. 2013, 85, 3752−3757. (35) Pham, H. T.; Julian, R. R. Anal. Chem. 2014, 86, 3020−3027. (36) Noestheden, M. R.; Headley, J. V.; Peru, K. M.; Barrow, M.; Burton, L. L.; Sakuma, T.; Winkler, P.; Campbell, J. L. Environ. Sci. Technol. 2014, 48, 10264−10272. (37) Campbell, J. L.; Zhu, M.; Hopkins, W. S. J. Am. Soc. Mass Spectrom. 2014, 25, 1583−1591. (38) Thomas, M. C.; Mitchell, T. W.; Blanksby, S. J. J. Am. Soc. Mass Spectrom. 2005, 16, 926−939. (39) Fernandez-Maestre, R.; Wu, C.; Hill, H. H., Jr. Rapid Commun. Mass Spectrom. 2012, 26, 2211−2223. (40) Retra, K.; Bleijerveld, O. B.; van Gestel, R. A.; Tielens, A. G.; van Hellemond, J. J.; Brouwers, J. F. Rapid Commun. Mass Spectrom. 2008, 22, 1853−1862. (41) Leek, J. T.; Scharpf, R. B.; Bravo, H. C.; Simcha, D.; Langmead, B.; Johnson, W. E.; Geman, D.; Baggerly, K.; Irizarry, R. A. Nat. Rev. Genet. 2010, 11, 733−739. (42) Brugger, B.; Erben, G.; Sandhoff, R.; Wieland, F. T.; Lehmann, W. D. Proc. Natl. Acad. Sci. U. S. A. 1997, 94, 2339−2344. (43) Yang, K.; Zhao, Z.; Gross, R. W.; Han, X. PLoS One 2007, 2, e1368. (44) Maccarone, A. T.; Duldig, J.; Mitchell, T. W.; Blanksby, S. J.; Duchoslav, E.; Campbell, J. L. J. Lipid Res. 2014, 55, 1668−1677.

9669

dx.doi.org/10.1021/ac5021744 | Anal. Chem. 2014, 86, 9662−9669