Anal. Chem. 2010, 82, 6687–6694
High-Throughput Phospholipidic Fingerprinting by Online Desorption of Dried Spots and Quadrupole-Linear Ion Trap Mass Spectrometry: Evaluation of Atherosclerosis Biomarkers in Mouse Plasma Aure´lien Thomas,†,| Julien De´glon,†,| Se´bastien Lenglet,‡ Franc¸ois Mach,‡ Patrice Mangin,† Jean-Luc Wolfender,§ Sabine Steffens,‡ and Christian Staub*,†,| Unit of Toxicology, CURML, Geneva University Hospitals, Geneva, Switzerland, Division of Cardiology, Department of Internal Medicine, University Hospital, Foundation for Medical Researches, Geneva, Switzerland, School of Pharmaceutical Sciences, University of Geneva, University of Lausanne, Geneva, Switzerland, and Swiss Center of Applied Human Toxicology, University of Geneva, Geneva, Switzerland This work presents a strategy for the evaluation of differences in plasma phospholipid content between atherosclerotic and healthy mice from micro volumes (2 µL) spotted on filter paper. Dried plasma spots (DPS) were directly desorbed into a triple quadrupole linear ion trap mass spectrometer using a homemade prototype, ensuring high-throughput analysis of dried spots without any sample pretreatment. Multiple positive and negative neutral loss and precursor ion scans were simultaneously acquired in a single loop, allowing oriented fingerprinting until 2700 potential species including phosphatidic acid (PA), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylglycerol (PG), phosphatidylinositol (PI), phosphatidylserine (PS), and sphingomyelin (SM) classes. The phospholipidic variations between 15 healthy and 15 atherosclerotic mice were evaluated using t tests, matrix reduction and merging, and principal component analysis (PCA) as a chemometric statistical approach. The discriminating ions in PCA analysis were qualitatively identified in an information dependent acquisition (IDA) manner using enhanced resolution and enhanced product ion scans. PCA demonstrates a clear clustering between healthy and diseased mice. Regarding the most relevant variables identified, this procedure has confirmed the role of SM and PS classes in atherosclerosis and has established potential biomarkers shown to be significantly up- or down-regulated with the disease. The results presented in this work demonstrate the sample processing and analysis potential of the developed strategy to evaluate new biomarkers and the state of a disease. * To whom correspondence should be addressed. E-mail: Christian.staub@ hcuge.ch. † Unit of Toxicology, CURML, Geneva University Hospitals. | Swiss Center of Applied Human Toxicology, University of Geneva. ‡ Department of Internal Medicine, University Hospital. § School of Pharmaceutical Sciences, University of Geneva. 10.1021/ac101421b 2010 American Chemical Society Published on Web 07/15/2010
Among the “omics cascade”, lipidomics is still considered an emerging field.1 However, many recent studies have shown that the investigation of lipids in biological systems is of major importance in the evaluation of new biomarkers or biochemical mechanisms of diseases.2-4 Phospholipids play important roles in the biochemistry of all living cells, either as the building blocks of membranes or regulators of processes such as homeostasis, metabolism, and signal transduction.5 This primordial physiological role is highlighted by the numerous diseases in which phospholipids are implicated, such as atherosclerosis, cancer, and Alzheimer’s disease.6,7 Atherosclerosis is one of the major diseases in humans and is the most common cause of death in western countries. It is induced by a chronic inflammatory response in the subendothelial space resulting from the interaction of oxidized low density lipoprotein (ox-LDL), monocyte-derived macrophages, T cells, and normal cellular elements.8,9 In patients with hypercholesterolemia, excess LDL infiltrates the artery at sites of hemodynamic strain. Oxidative and enzymatic modifications of LDL then lead to the release of inflammatory lipids that induce the immune response.10 Indeed, numerous studies have shown that modified phospholipids lead to reactive species which are able to promote atherogenesis by interacting with proteins, cells, and the immune system.11,12 (1) Zhu, C.; Hu, P.; Liang, Q.-L.; Wang, Y.-M.; Luo, G.-A. Chin. J. Anal. Chem. 2009, 37, 1390–1396. (2) Kaddurah-Daouk, R.; Kristal, B. S.; Weinshilboum, R. M. Annu. Rev. Pharmacol. Toxicol. 2008, 48, 653–683. (3) Ekroos, K.; Janis, M.; Tarasov, K.; Hurme, R.; Laaksonen, R. Curr. Atheroscler. Rep. 2010, 12, 273-281. (4) Adibhatla, R. M.; Hatcher, J. F.; Dempsey, R. J. AAPS J. 2006, 8, 314– 321. (5) Dowhan, W. Annu. Rev. Biochem. 1997, 66, 199–232. (6) Hammad, L. A.; Wu, G.; Saleh, M. M.; Klouckova, I.; Dobrolecki, L. E.; Hickey, R. J.; Schnaper, L.; Novotny, M. V.; Mechref, Y. Rapid Commun. Mass Spectrom. 2009, 23, 863–876. (7) Maxfield, F. R.; Tabas, I. Nature 205, 438, 612–621. (8) Glass, C. K.; Wilztum, J. L. Cell 2001, 104, 503–516. (9) Libby, P. Nature 2002, 420, 868–874. (10) Hansson, G. K. N. Engl. J. Med. 2005, 352, 1685–1695. (11) Berliner, J. A.; Watson, A. D. N. Engl. J. Med. 2005, 353, 9–11. (12) Berliner, J. A.; Leitinger, N.; Tsimikas, S. J. Lipid Res. 2009, 50, 207–212.
Analytical Chemistry, Vol. 82, No. 15, August 1, 2010
6687
Lipidomic analysis is largely based on electrospray mass spectrometry even if other ionization modes were recently investigated such as matrix-assisted laser desorption ionization.13,14 Numerous analyzers have been employed for the analysis of lipids. Triple stage quadrupole (QqQ) is often employed in profiling approaches where a preselection of the target compounds is carried out.15,16 Indeed, these instruments are particularly suitable for quantitative requirements with the use of multiple reaction monitoring (MRM) mode. In fingerprinting approaches where global and sensitive detection is attempted,17 methods are generally developed using devices with higher resolution and sensitivity in full scan mode, such as TOF, orbitrap or FT-ICR.18,19 The recent developments of hybrid systems such as QqQLIT, QTOF, and LTQ-Orbitrap have led to remarkable progress in lipidomicbased platforms, especially by combining the advantages of different analyzers and by promoting the use of data-dependent acquisition in order to increase the amount of information obtained in a single run.20,21 Compared to conventional LC/MS-based lipid analysis,22,23 a more recent approach was developed from the direct infusion of extracted biological sample with only the use of mass spectrometry as analytical support.24-26 This approach, also called shotgun lipidomics, is particularly well-suited to lipid analysis because it is nondiscriminative and fast.25,27 Shotgun lipidomics was originally carried out on QqQ devices using neutral loss (NLS) and precursor ion (PIS) scans to quantitatively determine the lipidic composition.24,28 Indeed, these compounds present specific CID fragmentation patterns based on their headgroup structure, which allows the selective monitoring and identification of each class of phospholipids.13 One limitation of this technique is the small number of experiments that these devices can simultaneously acquire due to their low scanning speed.29 Some studies have been developed using multiple PIS, based on QTOF systems with q2 trapping technology, to reduce ion loss due to the low duty cycle of orthogonal TOF and to increase the sensitivity of these devices (13) (14) (15) (16) (17) (18)
(19) (20) (21) (22) (23) (24) (25) (26) (27) (28) (29)
Pulfer, M.; Murphy, R. C. Mass Spectrom. Rev. 2003, 22, 332–364. Zehethofer, N.; Pinto, D. M. Anal. Chim. Acta 2008, 627, 62–70. Ikeda, K.; Shimizu, T.; Taguchi, R. J. Lipid Res. 2008, 49, 2678–2689. Nakagawa, K.; Oak, J.-H.; Higuchi, O.; Tsuzuki, T.; Oikawa, S.; Otani, H.; Mune, M.; Cai, H.; Miyazawa, T. J. Lipid Res. 2005, 46, 2514–2524. Dettmer, K.; Aronov, P. A.; Hammock, B. D. Mass Spectrom. Rev. 2007, 26, 51–78. Boccard, J.; Grata, E.; Thiocone, A.; Gauvit, J.-Y.; Lanteri, P.; Carrupt, P.A.; Wolfender, J.-L.; Rudaz, S. Chemom. Intell. Lab. Syst. 2007, 86, 189– 197. Hasegawa, M.; Takenaka, S.; Kuwamura, M.; Yamate, J.; Tsuyama, S. Exp. Toxicol. Pathol. 2007, 59, 115–120. Graessler, J.; Schwudke, D.; Schwarz, P. E. H.; Herzog, R.; Shevchenko, A.; Bornstein, S. R. PLoS One 2009, 4, e6261. Thomas, A.; Hopfgartner, G.; Giroud, C.; Staub, C. Rapid Commun. Mass Spectrom. 2009, 23, 629–638. Oursel, D.; Lutelier-Bourhis, C.; Orange, N.; Chevalier, S.; Norris, V.; Lange, C. M. Rapid Commun. Mass Spectrom. 2007, 21, 1721–1728. Ivinova, P. T.; Milne, S. B.; Myers, D. S.; Brown, H. A. Curr. Opin. Chem. Biol. 2009, 13, 1–6. Brugger, B.; Erben, G.; Sandhoff, R.; Wieland, F. T.; Lehmann, W. D. Proc. Natl. Acad. Sci.U.S.A. 1997, 94, 2339–2344. Cui, Z.; Thomas, M. J. J. Chromatogr., B 2009, 877, 2709–2715. Welti, R.; Wang, X. Curr. Opin. Plant Biol. 2004, 7, 337–344. Han, X.; Jiang, X. Eur. J. Lipid Sci. Technol. 2009, 111, 39–52. Han, X.; Gross, R. W. Anal. Biochem. 2001, 295, 88–100. Ekroos, K.; Chernushevich, I. V.; Simons, K.; Shevchenko, A. Anal. Chem. 2002, 74, 941–949.
6688
Analytical Chemistry, Vol. 82, No. 15, August 1, 2010
in this scan mode.29,30 However, this approach does not allow the simultaneous acquisition of NLS that is required for monitoring certain classes of phospholipids, such as PS or PE, due to the low abundance of headgroup fragments.31 In any lipids analysis approach, a sample preparation is required in order to purify complex biological samples before infusion. Lipids are generally extracted using the Bligh and Dyer procedure, consisting of a liquid-liquid extraction (LLE).32 Even if the loss of analytes in these procedures is quite small, they are nevertheless time-consuming in the workflow analysis. An interesting alternative approach relies on the online desorption of a dried spot33,34 to directly infuse the plasma samples previously spotted on filter paper into the MS. This approach integrates the spotted filter paper directly into the analytical system using a homemade desorption cell, coupling the well-known advantages of filter paper as sample support (storage, shipment, inhibition of infectious agent) and a procedure which does not involve sample pretreatment.33 Based on this technology, a recent homemade prototype, presented in this work, has been developed that allows the highthroughput analysis of 30 dried spots successively. The goal of this study is to evaluate potential phospholipid biomarkers of atherosclerosis in a model of apolipoprotein E deficient (ApoE-/-) mice using a high-throughput oriented fingerprinting strategy. This strategy consists of the automated desorption of a µ-plasma dried spot into a hybrid triple quadrupole linear ion trap mass spectrometer and chemometric analysis allowing the simultaneous analysis of phosphatidic acid (PA), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylglycerol (PG), phosphatidylinositol (PI), phosphatidylserine (PS), and sphingomyelin (SM) classes by multiple NLS and PIS from only 2 µL of plasma without any sample pretreatment. EXPERIMENTAL SECTION Chemicals and Reagents. LC grade acetonitrile (ACN) and methanol (MeOH) were purchased from Merck (Darmstadt, Germany). PI (16:0/16:0), PE (15:0/15:0), PS (14:0/14:0), PC (14: 0/14:0), and SM (d18:1/16:0) internal standards (IS) were obtained from Avanti Polar Lipids, Inc. (Alabaster, AL) and prepared in MeOH to a final concentration of 1 µg/mL. Animal Experiments. For evaluating the described procedure and looking for eventual biomarkers of atherosclerosis, we compared age-matched male wild-type (WT) C57BL/6 mice (n ) 15) with male ApoE-/- mice (n ) 15) fed with high cholesterol diet (HCD; 1.25% cholesterol) for 10 weeks. HCD feeding was started at the age of 10 weeks to induce atherosclerosis development. Analyses between the two groups were carried out in a randomized way. Plasma samples were obtained by microcentrifugation of venous whole blood. All animal studies have been approved by the local ethical committee. (30) Chernushevich, I. V.; Loboda, A. V.; Thomson, B. A. J. Mass Spectrom. 2001, 36, 849–865. (31) Schwudke, D.; Oegema, J.; Burton, L.; Entchev, E.; Hannich, J. T.; Ejsing, C. S.; Kurzchalia, T.; Shevchenko, A. Anal. Chem. 2006, 78, 585–595. (32) Bligh, E. A.; Dyer, W. J. Can. J. Biochem. Physiol. 1959, 37, 911–917. (33) De´glon, J.; Thomas, A.; Cataldo, A.; Mangin, P.; Staub, C. J. Pharm. Biomed. Anal. 2009, 49, 1034–1039. (34) Thomas, A.; De´glon, J.; Steimer, T.; Mangin, P.; Daali, Y.; Staub, C. J. Sep. Sci. 2009, 33, 873–879.
Figure 1. Scheme of the online dried spot automated system coupled with the QTrap MS. Enlargement shows the working of the clamp pistons and the location of the filter paper.
Figure 2. Mass spectrometry experiments for phospholipidic fingerprinting. Discriminant ions were determined by principal component analysis.
Sample preparation. Plasma samples (2 µL) of the WT and the knockout groups were spotted onto a filter paper card (Whatman, item no. 105355097, Dassel, Germany) with a micropipet (Eppendorf, Hamburg, Germany). The spots were then allowed to dry at room temperature for two hours before being packed in a sealable plastic bag. Air was expelled from the bag and it was set in a -20 °C environment. Before analysis, a 6 mm diameter disk containing a dried spot was punched out from the card and 2 µL of IS was added directly to the dried spot.
Online DPS Automated System. This work is focused on the direct desorption of filter paper into the mass spectrometer. To this end, an automated system has been developed to ensure the high-throughput analysis of dried spots (Figure 1). In the present prototype, 30 positions are available, set on a rotative inox plate. A mechanical clamp allows different positions to be chosen in a sequential way. When the clamp is locked on a position by pneumatic tightening of two pistons, a start signal activates the LC pump (Ultimate 3000 RS pump, Dionex, CA) via a remote Analytical Chemistry, Vol. 82, No. 15, August 1, 2010
6689
Figure 3. Data treatment workflow.
connection. The online desorption of the compounds from the filter paper is then carried out at a MeOH flow rate of 100 µL/ min. After the desorption step, a cleaning of the clamp, locked in the retracted position, is carried out by applying a MeOH flow rate of 2.0 mL/min for 1 min in order to avoid contamination of the next desorption. Mass Spectrometric Conditions. Equipment. Analyses were performed on a 5500 QTrap triple quadrupole/linear ion trap (QqQLIT) mass spectrometer equipped with a TurboIon-Spray interface (AB Sciex Concord, Canada). Data acquisition and analysis were performed using Analyst software version 1.5.1 (AB Sciex Toronto, Canada). MS Detection. The TurboIon-Spray interface was operated in both positive and negative ionization modes with a polarity switch delay of 50 ms. The source parameters, using nitrogen as curtain gas and nebulizer, were the following: 5.0 kV and -4.0 kV capillary voltage in positive and negative modes, respectively, 425 °C temperature, 20 psi curtain gas, 35 psi GS1, and 40 psi GS2. The MS/MS experiments consisted of the simultaneous acquisition of nine experiments combining PIS and NLS to monitor the different phospholipid classes in a single loop (Figure 2). Q1 and Q3 resolutions were set to give a 0.7 Da full width at half-maximum (fwhm) for both PIS and NLS. Each of the nine experiments was acquired from 650-950 Da using a 0.1 amu step size at 2000 Da/s in 0.15 s, affording the possibility of monitoring until 2700 potential species. The collision energy (CE) was set as 40 or 50 eV, depending on the phospholipid class. The duty cycle was 1.49 s, and the scans were acquired during 9 min in multiple channel acquisition (MCA) mode, which improves the ion statistics by summing the number of scans. 6690
Analytical Chemistry, Vol. 82, No. 15, August 1, 2010
Discriminant ions determined with principal component analysis (PCA) were then identified in a dependent information manner with the use of NLS or PIS as survey scans and two enhanced product ion scans (EPI) as dependent scans with a CE of 40 and 50 eV, respectively. An enhanced resolution scan (ER) is always monitored at 250 Da/s in order to obtain mass assignments along with dependent MS/MS experiments in the linear ion trap. EPI were performed in the linear ion trap over a mass range of m/z 100-950 at a rate of 10 000 amu/s using dynamic fill time (DFT). The IDA criteria were selected such as the two most intense peaks were acquired above a threshold of 500 counts and dynamically excluded for 75 s after one occurrence. The duty cycle of the method was 1.07 s and the analysis time was 5 min. Data Analysis. Data were treated with Markerview software version 1.2 (AB Sciex, Toronto, Canada). Data from each MS experiment was normalized using the sum of the total area of the peaks. Normalization by IS was also carried out and tested for the phospholipid classes for which they were available but did not improve the results. Peak alignment was tested with a mass tolerance of 0.1 amu. A mass tolerance of 0.5 amu was used to find peaks in the profile spectra with a threshold of 5.0e4 cps and a maximum of 5000 peaks. Each matrix from each MS experiment was first reduced by a t test, selecting the 10 most relevant variables corresponding to the 10 lowest p-values. The nine reduced matrices were then merged into a single matrix using a spreadsheet application before being resubmitted in Markerview for chemometric analysis. Cluster and discriminant ions between the two groups were then determined by PCA and t test. The most relevant ions were characterized by IDA and by the LIPID MAPS MS Prediction Tool (http://www.lipidmaps.org/tools/
Figure 4. Spectra obtained from the nine PIS and NLS experiments for the determination of the different phospholipid classes. Analysis was performed on 2 µL of mouse plasma using the online DPS automated system coupled with the QTrap 5500. The PIS 184, NLS 185, PIS 168, NLS 141, PIS 241, NLS 87, NLS 171, PIS 196, and NLS 420 was used to monitored PC, PS, SM, PE, PI, PS, PG, PE, and PA, respectively.
index.html), which confirmed the number of radyl carbons and of double bonds in the detected phospholipids. The data treatment workflow is presented in Figure 3. RESULTS AND DISCUSSION Online DPS Automated System. Compared to conventional lipid analysis, in which biological samples are purified by LLE, the procedure described here combines the advantages of filter paper (i.e., storing, shipment, and sample volume) with a no sample pretreatment process since the dried spot is directly integrated into the homemade prototype, ensuring the direct desorption of the compounds into the mass spectrometer (Figure 1). To this end, the dried spots are beforehand set into the rotative inox plate and then a sequence, synchronized with the mass spectrometer, is programmed for the successive analysis of the samples (see desorption process section). When the automated clamp is locked on a plate position, desorption of phospholipids from the filter paper is ensured by a flow rate of organic solvent delivered by a LC pump. Different solvents, such as MeOH or ACN, were tested and the best results in terms of sensitivity were obtained with MeOH (data not shown). Furthermore, the use of organic solvents has been shown to be particularly suitable for dried spot desorption since it leads to the precipitation of proteins on the filter paper.33 Different flow rates were tested, including 50, 100, 200, 300, 500 µL/ min. Even if the flow rates did not change the obtained spectra, a flow rate of 100 µL/min was ultimately selected to increase
the desorption time and to obtain a profile quite close to a conventional infusion over several minutes (about 10 min) and with a convenient desorption delay (Figure 3). In order to evaluate the procedure, spectra obtained from dried spot desorption and from infusion of the same extracted plasma by the Bligh and Dyer protocol were compared.32 No differences were found between the two spectrum profiles (Supporting Information Figure S-1). Furthermore, the sensitivity obtained from each method is very similar despite the difference in plasma volume; only 2 µL was spotted on the filter paper compared to 100 µL used for LLE extraction. This may be explained by the general efficiency of the dried spot desorption process, in which a solvent is continuously percolated during a fixed time combined to a no-dilution procedure.33,34 Beyond the saving time, the developed prototype limits the risk of carryover between successive desorptions because a retracted clamp position is programmed between each desorption to clean the system using a high solvent flow rate (i.e., 2 mL/min for 1 min). Automated system carryover was investigated by desorbing a blank filter paper after desorption of a dried spot. The results showed that no carryover was observed with our procedure (lower than 0.01%, data not shown). Generally, the use of filter paper is associated with whole blood analysis. However, the advantages of filter paper in terms of stability, shipment, and required biomaterial quantity have led to use of this support for other biomatrices such as plasma (particularly when red blood cells have to be removed), urine, Analytical Chemistry, Vol. 82, No. 15, August 1, 2010
6691
Figure 5. PCA score plot (PC1 × PC2) of the data obtained for the diseased (triangle) and healthy (circle) mouse groups.
bile and lysed tissues,35-37 particularly in disease studies on animal models where biomaterial is often limited and where numerous biological tests are carried out.34 For example, by spotting only 2 µL of plasma on filter paper, only 10-20 µL of whole blood is required for microcentrifugation which is compatible with capillary puncture. Stability of phospholipids in dried plasma spots was tested by analyzing spots stored at ambient temperature and at -20 °C over three months. No significant differences were observed between the spectra of the different phospholipid classes, which is in agreement with the well-known stability of filter paper for biological samples.38,39 Furthermore, as discussed above, no significant differences were observed between spectra obtained from dried spot desorption and from infusion of the same extracted plasma by the Bligh and Dyer protocol. These results demonstrated that no phospholipid oxidation (appearance of peaks with a mass shift of 16 Da in the spectrum) occurs on the filter paper during the drying time due to the presence of air. Mass Spectrometric Detection. Triple stage quadrupole MS has shown to be particularly suitable for phospholipid profiling due to selective scan detection, such as NLS and PIS.13 Indeed, study of certain classes, such as PS, required the use of NLS and may not be analyzed by PIS. Furthermore, NLS is sometimes more efficient than PIS in term of sensitivity and spectral information as is the case here for PE, PA, and PG, owing to the fact that fragmentation does not lead to abundant headgroup fragment ions.31 One major limitation of these instruments is their low scanning speed, limiting the analysis to only a single NLS or PIS at a time.40 This drawback necessitates repeat analysis for profiling (35) Michopoulos, F.; Theodoridis, G.; Smith, C. J.; Wilson, I. D. J. Proteome Res. 2010, 9, 3328-3334. (36) D’Avolio, A.; Simiele, M.; Siccardi, M.; Baietto, L.; Sciandra, M.; Bonora, S.; Di Perri, G. J. Pharm. Biomed. Anal. 2010, 52, 774–780. (37) Burggraf, S.; Do ¨rho ¨fer, B.; Olgemo ¨ller, B. Clin. Chem. 2007, 53, 1387– 1389. (38) De´glon, J.; Lauer, E.; Thomas, A.; Mangin, P.; Staub, C. Anal. Bioanal. Chem. 2010, 396, 2523–2532. (39) Li, W.; Tse, F. L. S. Biomed. Chromatogr. 2010, 24, 49–65. (40) Stahlman, M.; Ejsing, C. S.; Tarasov, K.; Perman, J.; Boren, J.; Ekroos, K. J. Chromatogr., B 2009, 877, 2664–2672.
6692
Analytical Chemistry, Vol. 82, No. 15, August 1, 2010
multiple lipids classes or the use of MS segments.40,41 Even if the use of MS segments is possible, it is not convenient with the present sample delivery system since analysis is not based on an infusion but instead on a desorption where the signal decreases during time. Furthermore, the low scanning speed requires analysis time to be extended to obtain enough scans to merge while maintaining sufficient sensitivity.31 In order to overcome these drawbacks, electronic improvements in the scanning speed and sensitivity of the QTrap 5500 system were applied here in the analysis of phospholipids. Figure 2 shows the developed MS loop allowing the simultaneously monitoring of six phospholipids classes and sphingomyelins by combining nine PIS and NLS in both positive and negative modes, although a QqQ was employed. In this loop, the QqQ scan speed was set at 2000 Da/s with a 50 ms delay for polarity switching and a duty cycle of only 1.49 s. In comparison, the scanning speed was shown to be 4 times faster than on QTrap 4000 despite a reduction in the polarity switch by a factor of 14.42 Figure 4 presents the nine mass spectra obtained after the automated desorption of a 2 µL DPS directly into the MS. As discussed above, the reduction of duty cycle is crucial when multiple QqQ scanning are simultaneously acquired in order to obtain sufficient sensitivity in a limited analysis time. This is emphasized when unit resolution is set on the quadrupoles for maintaining good selectivity.31 Although the different scans were monitored simultaneously at unit resolution in the present method, the analysis time was decreased by a factor of 3 compared to other procedures based on direct infusion on QqQ devices, even if the system did not operate with a nanoelectrospray ion source.41 Indeed, it is obvious that the use of this kind of ion source will allow for a further decrease in analysis time by acquiring proportionally each experiment during only a few seconds rather than one minute as in the present work. (41) Yang, K.; Cheng, H.; Gross, R. W.; Han, X. Anal. Chem. 2009, 81, 4356– 4368. (42) Le Blanc, J. C. Y.; Hager, J. W.; Ilisiu, A. M. P.; Hunter, F.; Zhong, F.; Chu, I. Proteomics 2003, 3, 859–869.
Table 1. Presentation of the 10 Most Discriminant Ions Obtained by Chemometric Analysis and Their Phospholipidic Classesa regulation during ranking of T-values disease (disease/ loadings ions (m/z) classes (absolute values) health ratio) 1 2 3 4 5 6 7 8 9 10
687.77 737.79 727.17 804.12 837.25 782.14 792.75 837.55 784.14 947.64
SM SM PS PS PS PS SM PG PS PG
33.6 27.1 20.4 18.9 18.2 17.9 17.2 16.9 15.4 15.3
2.9 3.6 5.2 0.3 3.2 0.6 0.5 5.9 1.7 3.7
a Disease/health ratios were determined by calculating the ratios between the average values of the absolute responses obtained for the ions in the two sample groups. The T-value represents a measure of how well the variable distinguishes the two clusters determined by PCA.
Identification of atherosclerosis biomarkers. Fingerprinting analysis combined with PCA treatment are generally carried out in full scan experiments with high resolution devices.6,17 However, full scan experiments may lead to the scarcity of data available due to the large variety of interferences present in complex biomatrices even if a chromatographic dimension is added to the analysis.43 For these reasons, NLS or PIS can be particularly advantageous since they will lead to an oriented fingerprinting of ions sharing a same fragmentation pattern, which is convenient for a nondiscriminant phospholipid analysis. However, the goal of fingerprinting analysis is to find the most relevant biomarkers for evaluation of the state of a disease among multiple candidates. Working with multiple MS experiments leads to multiple associated data matrices and multiple PCA results.
Owing to the number of experiments carried out in this work, it was important to combine data into one final matrix for submission to chemometric analysis and to globally interrogate samples (Figure 3). Because Markerview software is limited to evaluate only one experiment at a time, this matrix was built by merging the 10 most relevant variables of each experiment (i.e., the 10 lowest p-values obtained with a t test) using an external spreadsheet application. Elimination of redundant signal and matrix merging caused an increase in the explained variance of the PCA axes from about 50% for each experiment to more than 86% for the combined matrix in the first projection plan (PC1 × PC2). As demonstrated in Figure 5, PCA treatment unambiguously allows for the separation of diseased mice from healthy mice along PC1, which describes about 83% of the total data variability. Diseased and healthy mice presented a negative and positive coordinate, respectively, on PC1. The 10 most discriminating variables, and thus the most relevant biomarkers, were determined according to their absolute loading in PC1 combined with results of the t test and are presented in Table 1. As depicted, this table is primarily composed of species belonging to SM and PS classes that may be correlated with the potential role of these classes in atherosclerosis. Indeed, plasma SM have been suggested as a risk factor for coronary heart disease and the inhibition of their biosynthetic pathway was shown to be a promising therapeutic target for treatment of the disease.44 In addition, PS exposure is known to be the best characterized signal for the apoptic cell recognition by macrophages that play a crucial role in plaque development and rupture.45 If these classes have been demonstrated to have a potential role in the disease, then the developed procedure may highlight influential species of these classes. Figure 6a presents the ER and EPI spectra obtained by IDA for the confirmatory analysis of variable 1 in Table 1. Elucidation of fragmentation pathway, combined with the use of a prediction tool,
Figure 6. Enhanced product ion spectrum of the variable 687 with a proposal of the fragmentation pathway (a) and profile plot showing the relative response of the ion 687 across all samples (b). The corresponding enhanced resolution spectrum is framed. Analytical Chemistry, Vol. 82, No. 15, August 1, 2010
6693
allowed the identification of the SM N16:0 compound. Indeed, EPI demonstrated specific fragments at m/z 507.5, 255.8, and 283.4 corresponding to the C16:0 acyl chain.6 As shown in Figure 6b, this species was shown to be up-regulated during atherosclerosis compared to the WT mice. Furthermore, the EPI spectrum of the m/z 737.79 in Table 1 shows a somewhat similar fragmentation pathway to N16:0 SM (Supporting Information Figure S-2). This compound was established to be a chloro-N16:0 SM, confirmed by tool prediction and evaluation of isotope pattern. This finding can be corroborated with recent studies showing that oxidative stress induces formation of reactive species such as myeloperoxidase (MPO). MPO is known to generate HOCL, which is able to form chlorinated phospholipids and particularly chlorinated sphingomyelin.46 The possible role of these species in atherosclerosis will be investigated in further studies. CONCLUSION This work presents an oriented fingerprinting strategy for phospholipid analysis with an alternative approach ranging from sample processing to MS analysis and adapted data treatment. The overall method demonstrated promising results and perspec(43) Wagner, S.; Scholz, K.; Donegan, M.; Burton, L.; Wingate, J.; Wo¨lkel, W. Anal. Chem. 2006, 78, 1296–1305. (44) Park, T.-S.; Panek, R. L.; Rekhter, M. D.; Mueller, S. B.; Rosebury, W. S.; Robertson, A.; Hanselman, J. C.; Kindt, E.; Homan, R.; Karathanasis, S. K. Atheroslcerosis 2006, 189, 264–272. (45) Schrijvers, D. M.; De Mayor, G. R. Y.; Herman, A. G.; Martinet, W. Cardiovasc. Res. 2007, 73, 470–480. (46) Nusshold, C.; Kollroser, M.; Ko¨feler, H.; Rechberger, G.; Reicher, H.; Ullen, A.; Bernhart, E.; Waltl, S.; Kratzer, I.; Hermetter, A.; Hackl, H.; Trajanoski, Z.; Hrzenjak, A.; Malle, E.; Sattler, W. Free Radical Biol. Med. 2010, 48, 1588–1600.
6694
Analytical Chemistry, Vol. 82, No. 15, August 1, 2010
tives for research of phospholipid biomarkers associated with disease states. Indeed, by enabling to monitor until 2700 potential phospholipid species from only 2 µL of plasma without any sample pretreatment or chromatographic separation, the oriented fingerprinting procedure has highlighted several SM and PS phospholipids that are significantly up- or down-regulated with the disease, providing eventual mechanistic insight into atherosclerosis. To the best of our knowledge, the online dried spot automated system presented here is the first device allowing the complete automation of desorption of biological samples previously spotted on a filter paper disk. Beyond direct desorption analysis, this process may find future applications in numerous biomedical and pharmaceutical fields owing to the ubiquity of filter paper analysis. Furthermore, the large improvement of QqQ electronics in terms of sensitivity and scan speed will bring new opportunities for fingerprinting or screening analyses based on NLS and PIS experiments even if chromatographic separation is employed. ACKNOWLEDGMENT We thank AB Sciex and especially Dr. Hartmut Wuster for kindly providing Markerview software. S.S. is supported by grants from the Swiss National Science Foundation, Swiss Life, and Wolfermann-Na¨geli Foundation. SUPPORTING INFORMATION AVAILABLE Figures S-1 and S-2. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review May 31, 2010. Accepted July 2, 2010. AC101421B
