Anal. Chem. 2004, 76, 1935-1941
An HPLC-MS Approach for Analysis of Very Long Chain Fatty Acids and Other Apolar Compounds on Octadecyl-Silica Phase Using Partly Miscible Solvents Korne´l Nagy,† Annama´ria Jakab,† Jeno Fekete,‡ and Ka´roly Ve´key*,†
Institute of Chemistry, Chemical Research Center, Hungarian Academy of Sciences, H-1025 Pusztaszeri u´ t 59-67, Budapest, Hungary, and Institute of General and Analytical Chemistry, Budapest University of Technology and Economics, H-1111 Szt. Gelle´ rt te´ r 4, Budapest, Hungary
A novel approach for analyzing underivatized very long chain fatty acids (C16-C26) and other apolar compounds such as triacylglycerols is described. It is based on reversed-phase HPLC separation followed by mass spectrometric detection. Partly miscible solvents are used for stepwise gradient elution starting with a methanol/water and ending with a methanol/n-hexane binary mixture. The developed technique does not need derivatization, and analysis is fast (fatty acids were separated in 2-min-long chromatograms) and robust. The developed method is also very sensitive; a quantitation limit in the low-picogram range was achieved for fatty acids. The separation mechanism and advantages of the suggested technique are discussed and illustrated in the case of blood analysis and plant oil characterization. Analysis of fatty acids is of great importance since they are present in food products, plant oils, living organisms, biodiesel, etc. Fatty acids (saturated and unsaturated) are particularly important as a family of metabolites in living organisms. These compounds are a major source of energy production in cells via β oxidation, tricarboxylic acid cycle, oxidative phosphorylation, and energy storage (ketone bodies).1 They also serve as precursors in the synthesis of other compounds such as triacylglycerols, phospholipids, cholesterol, bile acids, or steroid hormones via an acetyl-coenzyme A intermediate. Deterioration of fatty acid metabolism causes serious illnesses, such as sudden infant death syndrome and short-, medium-, and long-chain fatty acyl-CoA dehydrogenase deficiency.2-8 There is a large variety of analytical * To whom correspondence should be addressed. E-mail: vekey@ chemres.hu. Tel: +36-1-438-0481. Fax: +36-1-325-9105. † Hungarian Academy of Sciences. ‡ University of Technology and Economics. (1) Stryer, L. Biochemistry; W. H. Freeman and Co.: New York, 1995. (2) Yang, Y. J.; Choi, M. H.; Paik, M. J.; Yoon, H. R.; Chung, B. C. J. Chromatogr., B 2000, 742, 37-46. (3) Johnson, D. W. J. Inherited Metab. Dis. 2000, 23, 475-486. (4) Inoue, K.; Suzuki, Y.; Yajima, S.; Shimozawa, N.; Orii, T.; Kondo, N. Clin. Chem. 1997, 43, 2197-2198. (5) Bruns, A. Fett Wiss. Technol.-Fat Sci. Technol. 1988, 90, 289-291. (6) Zytkovicz, T. H.; Fitzgerald, E. F.; Marsden, D.; Larson, C. A.; Shih, V. E.; Johnson, D. M.; Strauss, A. W.; Comeau, A. M.; Eaton, R. B.; Grady, G. F. Clin. Chem. 2001, 47, 1945-1955. 10.1021/ac034944t CCC: $27.50 Published on Web 02/28/2004
© 2004 American Chemical Society
techniques for the characterization, quantitation, and detection of fatty acids in biological matrixes. A good indication of the importance of this field is that over 3000 papers have been published in 2003 related to fatty acid analysis. A full literature review of this field is outside the scope of this paper; the interested reader is referred to basic textbooks and review articles.1,9-13 Since the discovery of the importance of triacylglycerols in nutrition and in health care, these fatty acid derivatives (major components of plant oil) also enjoy high attention.14-17 The two most common analytical methods for studying fatty acids and triacylglycerols are gas chromatography (GC)8,18-20 and high-pressure liquid chromatography (HPLC). The latter technique is increasingly used and is the subject of the present paper. The apolar character of fatty acids covers a wide range. For instance, the biologically most important fatty acids span from C16 to C26 carbon numbers having log P values between 6.96 and 12.06.21 (P being the partitioning ratio of the compound between 1-octanol and water). Without gradient elution, efficient separation of fatty acids with different chain lengths is unfeasible since it would require prohibitively long analysis time.22 In usual HPLC practice, gradient elution with linear solvent strength methods are conducted. To reduce separation time, the eluent (7) Johnson, D. W. Rapid Commun. Mass Spectrom. 1999, 13, 2388-2393. (8) Vreken, P.; van Lint, A. E.; Bootsma, A. H.; Overmars, H.; Wanders, R. J.; van Gennip, A. H. J. Chromatogr., B 1998, 713, 281-287. (9) Moser, H. W.; Bergin, A.; Cornblath, D. Biochem. Cell Biol. 1991, 69, 463474. (10) Arab, L.; Akbar, J. Public Health Nutr. 2002, 5, 865-871. (11) Arab, L. J. Nutr. 2003, 133 (Suppl 3), 925S-932S. (12) Gutnikov, G. J. Chromatogr., B 1995, 671, 71-89. (13) Byrdwell, W. C. Lipids 2001, 36, 327-346. (14) Jakab, A.; Heberger, K.; Forgacs, E. J. Chromatogr., A 2002, 976, 255263. (15) Jakab, A.; Nagy, K.; Heberger, K.; Vekey, K.; Forgacs, E. Rapid Commun. Mass Spectrom. 2002, 16, 2291-2297. (16) Neff, W. E.; Byrdwell, W. C. J. Liq. Chromatogr. 1995, 18, 4165-4181. (17) Mottram, H. R.; Woodbury, S. E.; Evershed, R. P. Rapid Commun. Mass Spectrom. 1997, 11, 1240-1252. (18) Vallance, H.; Applegarth, D. Clin. Biochem. 1994, 27, 183-186. (19) Zee, T.; Stellaard, F.; Jakobs, C. J. Chromatogr., A 1992, 574, 335-339. (20) ten Brink, H. J.; Stellaard, F.; van den Heuvel, C. M.; Kok, R. M.; Schor, D. S.; Wanders, R. J.; Jakobs, C. J. Lipid Res. 1992, 33, 41-47. (21) Compudrug. (22) Snyder, L. R. In High-Performance Liquid Chromatography, Advances and Perspectives; Horva´th, C., Ed.; Academic Press: New York, 1981; Vol. 1, p 207.
Analytical Chemistry, Vol. 76, No. 7, April 1, 2004 1935
strength of the second solvent must be high. Various eluents and eluent mixtures have been used for the separation of fatty acids, including 2-propanol. The latter can be successfully used for efficient elution of fatty acids and triacylglycerols from octadecylsilica phase.23,24 Insufficient light absorbance of fatty acids and interference with the absorbance of the applied solvents add to the difficulty of HPLC analysis of these compounds. For this reason, fatty acid determination in biological and other matrixes conventionally needs derivatization to introduce chromophore or fluorophore groups.12,25-34 These procedures are time-consuming, and laborious steps and recoveries might be influenced by the matrix and the conditions applied. Application of mass spectrometry often overcomes detection problems in HPLC as it has excellent selectivity, specificity, and sensitivity at the same time. For these reasons, HPLC-MS is becoming more and more popular.35-41 Introduction of atmospheric pressure ionization techniques, such as electrospray and atmospheric pressure chemical ionization (APCI),13,39,42 made the HPLC-MS combination robust and relatively easy to use. Mass spectrometric detection, on the other hand, puts some restrictions on chromatography, especially in the selection of solvents, the use of buffers, and restrictions on the flow rate.35,43,44 Various additives commonly used in HPLC (tetrahydrofuran, etc.) cannot be used with HPLC-MS, as they suppress ion signals. The basic idea in the present paper is the use of partly miscible solvents and a stepwise gradient to achieve fast gradient elution. The main object was to develop a method for analysis of very long chain fatty acids (C16-C26) in dried blood spots to be used for medical screening. The technique can, furthermore, be efficiently used for analysis of other apolar compounds, illustrated in the case of di- and triacylglycerols in plant oil. EXPERIMENTAL SECTION Chemicals. HPLC grade water, methanol, n-hexane, acetic acid, and palmitoic (C16), stearic (C18), eicosanoic (C20), docosano(23) Kapoulas, V. M.; Andrikopoulos, N. K. J. Chromatogr., A 1986, 366, 311320. (24) Lin, J.; Woodruff, C. L.; McKeon, T. A. J. Chromatogr., A 1997, 782, 4148. (25) Toyo’oka, T.; Takahashi, M.; Suzuki, A.; Ishii, Y. Biomed. Chromatogr. 1995, 9, 162-170. (26) Akasaka, K.; Ohrui, H.; Meguro, H. Analyst 1993, 118, 765-768. (27) Yasaka, Y.; Tanaka, M.; Shono, T.; Tetsumi, T.; Katakawa, J. J. Chromatogr., A 1990, 508, 133-140. (28) Yamaguchi, M.; Hara, S.; Matsunaga, R.; Nakamura, M. J. Chromatogr., A 1985, 346, 227-236. (29) Nimura, N.; Kinoshita, T. Anal. Lett. 1980, 13, 191-202. (30) Mehta, A.; Oeser, A. M.; Carlson, M. G. J. Chromatogr., B 1998, 719, 9-23. (31) Inoue, H.; Ikeno, M.; Ishii, Y.; Tsuruta, Y. J. Chromatogr., A 1998, 816, 137-143. (32) Kuroda, N.; Taguchi, Y.; Nakashima, K.; Akiyama, S. Anal. Sci. 1995, 11, 989-993. (33) Kotani, A.; Kusu, F.; Takamura, K. Anal. Chim. Acta 2001, 1-8. (34) Gerard, H. C.; Moreau, R. A.; Fett, W. F.; Osman, S. F. J. Am. Oil Chem. Soc. 1992, 69, 301-304. (35) Vekey, K. J. Chromatogr., A 2001, 921, 227-236. (36) Vogeser, M. Clin. Chem. Lab. Med. 2003, 41, 117-126. (37) Siegel, M. M. Curr. Top. Med. Chem. 2002, 2, 13-33. (38) Marquet, P.; Lachatre, G. J. Chromatogr., B 1999, 733, 93-118. (39) Bogusz, M. J. J Chromatogr., B 1999, 733, 65-91. (40) Muck, W. Pharmazie 1999, 54, 639-644. (41) Gelpı´, E. J. Chromatogr., A 1995, 703, 59-80. (42) Gaskell, S. J. J. Mass Spectrom. 1997, 32, 677-688. (43) Espada, A.; Rivera-Sagredo, A. J. Chromatogr., A 2003, 987, 211-220. (44) Zhao, J. J.; Yang, A. Y.; Rogers, J. D. J. Mass Spectrom. 2002, 37, 421-433.
1936
Analytical Chemistry, Vol. 76, No. 7, April 1, 2004
ic (C22), tetracosanoic (C24), and hexacosanoic (C26) acids were obtained from Supelco-Sigma-Aldrich GmbH (Steinheim, Germany). 2H4-Docosanoic, 2H4-tetracosanoic, and 2H4-hexacosanoic acids were obtained from H.J. ten Brink (VU Medical Center, Amsterdam). Cold-pressed almond oil was obtained from a local grocery store and was used in diluted form (0.1% in methanol). A methanolic fatty acid stock solution was prepared containing 80 µmol/L palmitoic acid, 80 µmol/L stearic acid, 40 µmol/L eicosanoic acid, 20 µmol/L docosanoic acid, 20 µmol/L tetracosanoic acid, and 40 µmol/L hexacosanoic acid. These relative concentrations were selected as most appropriate for blood analysis. Sample Preparation. Blood samples were received either as blood spots dried on Guthrie cards or frozen blood. First 100 µL of isotope-labeled internal standard mixture containing 2H4docosanoic acid in 50 µmol/L, 2H4-tetracosanoic acid in 50 µmol/ L, and 2H4-hexacosanoic acid in 4 µmol/L concentration was transferred to 1.5-mL Eppendorf tubes, and the solvent was evaporated under nitrogen stream. The dried blood spot disk (containing ∼50 µL of blood) or 50 µL of blood was transferred to the Eppendorf tube containing the isotope-labeled internal standard. Then 360 µL of acetonitrile and 40 µL of 5 M hydrochloric acid were added. The mixture was heated at 80 °C for 1 h to achieve hydrolysis of the fatty acids from phospholipid, triglyceride, cholesterol, and other esters.3 This mixture was then extracted with 2 × 0.5 mL n-hexane. The n-hexane was evaporated under a nitrogen stream, and the residue was dissolved in 100 µL of methanol and used for analysis. For assessing the accuracy of quantitation, standard addition experiments were performed by spiking 50 µL of blood sample with fatty acids at six different enrichment levels, adding fatty acids corresponding to 250, 500, 750, 1000, 1250, and 1500 µL of stock solution. The spiked blood samples were prepared as detailed above. HPLC Instrumentation and Conditions. The HPLC system used consisted of a binary solvent delivery system (two isocratic Perkin-Elmer Series 200 LC pumps connected via a Scientific Systems, Inc. high-pressure mixer device), a Perkin-Elmer Series 200 autosampler (Norwalk, CT) equipped with a 10-µL sample loop and a Perkin-Elmer Sciex API 2000 triple quadrupole mass spectrometer (Toronto, Canada). A Purospher Star RP-18e (55 mm × 2 mm i.d., particle size 3 µm) column was used, purchased from Merck KGaA (Darmstadt, Germany). Experiments were carried out at room temperature; no column thermostat was applied. All solvents were degassed in an ultrasonic bath prior to use. Typical HPLC conditions used in the present work were the following: The column was conditioned by pumping methanol/ water 90/10 v/v% containing 0.2% acetic acid (solvent A) through the column for 10 min using a flow rate of 200 µL/min. The 10µL sample was injected. A stepwise gradient program was used switching from solvent A to solvent B (methanol/n-hexane 90/ 10 v/v%, also acidified by 0.2% acetic acid) 1 min after sample injection. Flow rate was 200 µL/min. After each analysis, the column was reequilibrated with mobile phase A at 200 µL/min for 10 min. High-speed separation of fatty acids and triacylglycerols was performed in the same way but the flow rate of solvent A was set to 300 µL/min and the flow rate of solvent B was set to
Figure 1. Typical HPLC-MS chromatogram of a fatty acid mixture (sample amount 68 ng for docosanoic acid). Flow rate is 200 µL/ min, and negative APCI ionization was used. Total ion chromatogram is shown.
700 µL/min. Note that this flow rate was obtained at a relatively low pressure (∼80 bar). In a series of experiments, the equilibration time between individual runs was shortened to 2 min, retention times becoming stationary after three injections. Mass Spectrometry. The mass spectrometer was interfaced to the HPLC system using an APCI source. APCI corona probe current was 4 µA in positive and 1 µA in negative ion modes. Vaporizer temperature was 500 °C. High-purity nitrogen was used as nebulizer and collisional gas. Mass spectra of fatty acids were acquired in the m/z ) 200-450 Th range at an orifice voltage of -86 V. Cold-pressed almond oil was studied at an orifice voltage of 70 V, acquiring the spectra in the m/z ) 100-1000 Th range. Using a methanol/n-hexane solvent mixture may be regarded as an explosive hazard with high-temperature APCI. Addition of methylene chloride to the eluent can minimize this problem. Note also that a high flow of inert (nitrogen) nebulizer and auxiliary gas is added to the solvent vapor in APCI, which cools the solvent vapor and reduces the risk of ignition. RESULTS AND DISCUSSION Method Development. Applying the HPLC parameters detailed in the Experimental Section and using mass spectrometric detection in negative APCI mode, a mixture of six fatty acids (C16-C26) yields the chromatogram shown in Figure 1. Retention times are quite short, and the fatty acids are well separated. To obtain the chromatogram shown in Figure 1. A 10-µL sample of stock solution was injected (containing ∼100 ng of each fatty acid; see Experimental Section for details) resulting in good signal-tonoise ratio. The mass spectra in negative ion APCI mode is very simple; the fatty acids show only one major peak corresponding to the respective [M - H]- ion (m/z ) 255 for C16, m/z ) 283 for C18, etc). The use of a slightly acidic methanol/n-hexane mixture (solvent B) is advantageous for mass spectrometry, in terms of both sensitivity and volatility of n-hexane. The methanol/n-hexane
Figure 2. High-speed HPLC-MS chromatogram of a fatty acid mixture (sample amount 68 ng for docosanoic acid). The column was conditioned in eluent A, which was switched immediately after sample injection to eluent B; flow rate was 700 µL/min. Total ion chromatogram obtained in negative APCI ionization is shown.
mixture compared to water has low surface tension (helping to transfer target molecules to the gas phase) and low viscosity, so relatively high flow rates are accessible in HPLC. In these studies, APCI ionization was used as it showed an order of magnitude better sensitivity than electrospray both for fatty acids and for triacylglycerols. For fatty acids negative ion mode and for di- and triacylglycerols (having no acidic groups) positive ion mode were more advantageous, so these were used, respectively. The retention times shown in Figure 1 are quite short, but the chromatogram can be further speeded up by conditioning the column only for 2 min at 300 µL/min flow rate switching from eluent A to B immediately after injection and by increasing the flow rate of solvent B to 700 µL/min. The result is shown in Figure 2. Resolution between hexacosanoic and tetracosanoic acid (as defined above) is R ) 3.4, only slightly worse than that obtained in Figure 1 (R ) 4.5), while separation time is less than 2 min. The relatively high flow rates, however, reduce sensitivity by a factor of 5 so there is a tradeoff between sensitivity and analysis time. Several solvent combinations were tested to achieve optimal elution of fatty acids including water, methanol, acetonitrile, 2-propanol, and n-hexane. Methanol/water and methanol/nhexane solvent mixtures gave excellent results, and these were used in the present work. Both eluents were slightly acidified (0.2% acetic acid); the composition of solvents A (methanol/water) and B (methanol/hexane) was varied to obtain optimal elution of both the shortest (C16) and the longest (C26) fatty acids. Results are summarized in Table 1 showing retention times and peak areas. The table shows that retention times increase with decreasing mobile phase strength (e.g., changing methanol/water ratio from 9/1 to 8/2), as may be expected. With increasing retention time, the resolution also increases, although not significantly. Peak areas obtained from the reconstructed ion chromatograms differ significantly (Table 1) depending on solvent composition. Determining optimal solvent composition is always a tradeoff of various aspects, but for a general purpose, the best results were obtained Analytical Chemistry, Vol. 76, No. 7, April 1, 2004
1937
Table 1. Retention Time and Peak Area of Six Fatty Acids versus Mobile-Phase Compositiona eluent methanol/water methanol/n-hexane resolution of C24 and C26
composition 9:1 8:2 4.3
10:0 9:1 3.1
fatty acid palmitoic acid (C16) stearic acid (C18) eicosanoic acid (C20) docosanoic acid (C22) tetracosanoic acid (C24) Hexacolsanoic acid (C26)
9:1 10:0 5.3
8:2 9:1 5.1
retention time (min) 2.90 3.39 3.86 4.35 4.85 5.40
1.13 1.40 1.80 2.35 3.02 3.70
fatty acid] palmitoic acid (C16) stearic acid (C18) eicosanoic acid (C20) docosanoic acid (C22) tetracosanoic acid (C24) hexacoslanoic acid (C26)
9:1 9:1 4.8
2.85 3.37 3.99 4.65 5.37 6.16
5.15 6.25 7.42 8.70 10.11 11.86
4.44 5.07 5.70 6.41 7.56 8.62
peak area 2 468 446 3 836 231 2 874 667 2 398 553 2 749 012 5 307 337
3 534 408 4 525 911 3 228 727 2 579 835 2 396 475 3 548 659
2 697 475 3 497 895 2 559 561 2 245 375 2 422 789 4 202 234
2 505 169 4 009 442 2 490 617 2 264 827 2 573 579 3 914 343
1 793 797 2 955 403 1 750 317 1 421 982 1 449 354 2 318 432
a Data obtained on a Purospher Star RP-18e (55 mm × 2 mm i.d., particle size 3 µm) column. Resolution was calculated according to the equation R ) 2(tR - tRprev)/(w + wprev).
with 90/10 v/v% methanol/water (mobile phase A) and 90/10 v/v% methanol/n-hexane (mobile phase B). Gradient elution causes zone compression, which results (in an ideal case) in approximately the same peak widths for all compounds. A further feature of ideal gradient elution is that the retention time difference between different members of a homologous series is approximately constant.22 Figures 1 and 2 show that both conditions are approximately satisfied. Chromatography was also investigated under isocratic conditions using methanol/ water (9/1) and methanol/n-hexane (9/1) mixtures. The obtained chromatograms are shown in Figure 3A and B, respectively. These clearly show that the presented method is gradient elution and not simple isocratic elution with the second solvent. Dead time of the chromatographic system was determined as 0.52 min using 200 µL/min flow rate (Figure 1) and 0.30 min using 700 µL/min flow rate experiments (Figure 2), determined using 0.02 mg/mL L-serine in methanol. Retention time of the first peak (C16) is significantly longer (2.85 and 0.92 min, respectively) than the dead time. The developed technique can be used with various C18 columns, Purospher RP-18e, Purospher Star 100 RP-18e, LiChrospher 100 RP-18, and Novapac C-18 were tried. All columns gave acceptable results and qualitatively similar chromatograms; the best results were obtained by the Purospher Star 100 RP-18e column. Sensitivity of the developed method for fatty acids and its potential for quantitative studies were also tested. Docosanoic acid, for example, could be detected at 10-pg quantity with a signal-tonoise ratio better than 4:1 (using the peak-to-peak noise definition, observing the [M - H]- ion). The calibration curve in the 10 pg680 ng range showed good linearity: the regression coefficient (R2) was 0.999 and the mean relative standard deviation of data from the linear regression line was 12%. This suggests that quantitation of fatty acids down to the 10-pg level is possible, which makes this HPLC-MS technique competitive with time-consuming GC or GC/MS procedures. The developed HPLC-MS method is robust and well reproducible. These were checked by measuring the fatty acid standard 1938 Analytical Chemistry, Vol. 76, No. 7, April 1, 2004
Figure 3. Separation of fatty acids using isocratic methods in methanol/water 9/1 (A) and methanol/n-hexane 9/1 (B) solvent mixtures.
solution under different conditions and on different occasions. The composition of both solvent A (methanol/water) and solvent B (methanol/n-hexane) were varied in the range of 91/9-89/11. The average standard deviation of retention times was 0.14 min, the relative standard deviation being 3%. The intra- and interday repeatability (standard deviation) of the retention times were 0.03 (n ) 9) and 0.07 (n ) 4) min, respectively. The presented method involves an unusual solvent combination; the mechanism of separation can be tentatively described in
Figure 4. Individual ion chromatograms of six very long chain fatty acids obtained from a dried blood spot using negative APCI ionization and SIM. Following hydrolysis of the blood sample, no derivatization was needed; the target compounds were measured in their native form.
the following manner. Initially, the pores and interstitial volume of the column are filled with methanol/water mixture. Due to the high hydrophobicity (apolarity) of the injected compound (log P >6), all solutes will be adsorbed on the surface of the stationary phase inside the pores, so the retention factor will be high (log k >10). Assuming a sudden change in mobile-phase composition, the solvent composition in the interstitial volume would change immediately but this would be slow inside the pores. The dissolution rate will depend on the distribution constant of the solute in solvent A and solvent B. As the solvent strength in the pores increases, the retention factor of fatty acids will decrease (log k
