Application of TOF-SIMS with Chemometrics To Discriminate between

Faculty of Life Sciences, The University of Manchester, The Michael Smith Building, Oxford Road,. Manchester M13 9PT, U.K., and School of Chemical ...
0 downloads 0 Views 177KB Size
Anal. Chem. 2005, 77, 1740-1745

Application of TOF-SIMS with Chemometrics To Discriminate between Four Different Yeast Strains from the Species Candida glabrata and Saccharomyces cerevisiae H. Jungnickel,† E. A. Jones,‡ N. P. Lockyer,‡ S. G. Oliver,† G. M. Stephens,‡ and J. C. Vickerman*,‡

Faculty of Life Sciences, The University of Manchester, The Michael Smith Building, Oxford Road, Manchester M13 9PT, U.K., and School of Chemical Engineering and Analytical Science, The University of Manchester, P.O. Box 88, Manchester M60 1QD, U.K.

We present a TOF-SIMS analysis of the cell surface differences between four yeast strains from two species, Candida glabrata and Saccharomyces cerevisiae (haploid strains BY4742 and BY4741 and the derived diploid BY4743). The study assesses the suitability of TOF-SIMS analysis in combination with statistical methods (principal component analysis, Fisher’s discriminant analysis, and cluster analysis) for the discrimination between the four yeast strains. We demonstrate that a combination of these statistical methods identifies 34 ions, from a total data set of 1200, which can be used to distinguish between the four yeasts. The study discusses the assignments of surface cell membrane phospholipids for the identified ions and the resulting differences in the phospholipid pattern between the four yeasts, particularly in relation to ploidy and budding pattern. The method shows that fatty acids, phosphatidylglycerols, phosphatidylethanolamines, phosphatidylserines, and phosphatidylcholines, as well as cardiolipins, are of diagnostic importance. Recent studies have shown that TOF-SIMS analysis can be used to characterize the biochemical pattern, not only of artificial structures such as liposomes or pulmonary surfactants, but also of the surfaces of biological cells, like human skin fibroblasts or rat pheochromocytoma cells.1-4 Static secondary ion mass spectrometry has also been used in combination with multivariate statistical methods to discriminate between 28 bacterial strains, using their fatty acid profile resulting from surface analysis of intact cells.5 However, so far, no study has investigated whether * Corresponding author. E-mail: [email protected]. Tel: +44(0)161 3064544. Fax: +44(0)161 3064439. † Faculty of Life Sciences. ‡ School of Chemical Engineering and Analytical Science. (1) Cannon, D. M.; Pacholski, M. L.; Winograd, N.; Ewing, A. G. J. Am. Chem. Soc. 2000, 122, 603. (2) Harbottle, R. R.; Nag, K.; McIntyre, N. S.; Possmayer, F.; Petersen, N. O. Langmuir 2003, 19, 3698. (3) Roddy, T. P.; Cannon, D. M.; Meserole, C. A.; Winograd, N.; Ewing, A. G. Anal. Chem. 2002, 74, 4011. (4) Cliff, B.; Lockyer, N.; Jungnickel, H.; Stephens, G.; Vickerman, J. C. Rapid Commun. Mass Spectrom. 2003, 17, 2163 (5) Ingram, J. C.; Bauer, W. F.; Lehmann, R. M.; O’Connell, P.; Shaw, A. D. J. Microbiol. Meth. 2003, 53, 295.

1740 Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

TOF-SIMS can be used in combination with multivariate statistical methods to discriminate between different yeast genera or even between closely related yeast strains. We have investigated the surface chemistry of intact cells from Candida glabrata and three Saccharomyces cerevisiae strains, the two haploid strains BY4741 and BY4742 and the diploid strain BY4743, using a primary Au+ ion beam. C. glabrata is a common saprophyte in humans, which may become pathogenic in immunocompromised patients,6,7 while S. cerevisiae is probably the best characterized eukaryotic organism and has been used by mankind for millennia in the production of bread, beer, and wine. FAB, LC-MS (ESI), and MALDI-TOF mass spectrometry analyses have all been used in the characterization of various microorganisms,8-18 and phospholipid patterns were used to distinguish between different Candida species15 and for the distinction between C. glabrata and S. cerevisiae.16,17 However, phospholipid distribution analysis using FAB mass spectrometry requires large amounts of cell mass (10-20 mg) in combination with prolonged extraction and analysis methods. Moreover, phospholipid contaminants from cell compartments, such as mitochondria, may also influence the overall result. LC-MS (ESI) and MALDI-TOF methods also use large amounts of cell mass (1 g for LC-MS and 100 mg-1 g for MALDI-TOF) for the extraction of membrane phospholipids in combination with extensive membrane purification steps together with time-consuming phospho(6) Bailey, J. E.; Kliegman, R. M.; Annable, W. L. Am. J. Dis. Children 1984, 138, 965. (7) Morris, J. T.; McAllister,; C. K. South. Med. J. 1993, 86, 356. (8) Tavana, A. M.; Drucker, D. B.; Boote, V. J. Appl. Microbiol. 1998, 85, 1029. (9) Tavana, A. M.; Korachi, M.; Boote, V.; Hull, P. S.; Love, D. N.; Drucker, D. B. J. Appl. Microbiol. 2000, 88, 791. (10) Heller, D. N.; Murphy, C. M.; Cotter, R. J.; Fenselau, C.; Uy, O. M. Anal. Chem. 1988, 60, 2787. (11) Aluyi, H. S.; Boote, V.; Drucker, D. B.; Wilson, J. M.; Ling, Y. H. J. Appl. Bacteriol. 1992, 73, 426. (12) Drucker, D. B. ACS Symp. Ser. 1994, 541, 18. (13) Drucker, D. B.; Abdullah, N. J. Appl. Bacteriol. 1995, 79, 219. (14) Drucker, D. B., Wardle, H. M., Boote, V. J. Bacteriol. 1996, 178, 5844. (15) Abdi, M.; Drucker, D. B.; Boote, V.; Korachi, M.; Theaker, E. D. J. Appl. Microbiol. 1999, 87, 332 (16) Mahmoudabadi, A. Z.; Boote, V.; Drucker, D. B. J. Appl. Microbiol. 2001, 90, 668. (17) Mahmoudabadi, A. Z.; Boote, V.; Verran, J.; Johnson, E.; Drucker, D. B. J. Appl. Microbiol. 2003, 95, 883. (18) Freifelder, D. J. Bacteriol. 1960, 80, 567. 10.1021/ac048792t CCC: $30.25

© 2005 American Chemical Society Published on Web 02/04/2005

lipid enrichment procedures to guarantee phospholipids in the analysis are derived from the cell membrane ones and do not come from other cell compartments. To avoid such problems, this study tries to establish a rapid method for the analysis of small sample volumes (3 µL) using TOF-SIMS technology in combination with chemometric studies to discriminate between four different yeast sample sets. By analyzing phospholipid profiles of the yeast cell surface, this method avoids contamination by phospholipids that are part of the mitochondrial membrane or other intracellular membranes. It also avoids the confounding effects of free intracellular fatty acids, the complement of which is highly dependent on the physiological state of the cell. Our study should also be able to show whether the statistical methods used are capable of distinguishing between haploid (mating type ∆ and a) and diploid (mating type a/∆) cells of S. cerevisiae from stationary-phase cultures using a significantly reduced data set of less than 40 variables compared to other studies, which needed several hundreds of variables to achieve similar results. This represents a facile new method for chemotaxonomic studies of different yeast species, and of different strains within one yeast species, using a small set of variables. The diploid Saccharomyces strain BY4743 was selected because MATa/MAT∆ diploids show a polar budding pattern, while haploid (MATa or MAT∆ cells) exhibit axial budding pattern.18-20 Bud scars are chitin-rich ring structures that mark the region, where daughter cells separated from the mother.19,21 Some of the genes involved in different budding patterns have been shown to participate in membrane lipid metabolism, changing the phospholipid content of the cell membrane,22,23 but nothing is known about which specific phospholipids change with different budding patterns. This study will, therefore, show which changes in the phospholipid content of haploid and diploid cells are associated with the different budding behavior of the two strains and therefore can be used to distinguish between haploid and diploid Saccharomyces cells. Different phospholipid patterns are also responsible for drug resistance in S. cerevisiae and C. glabrata. The method used can highlight the major differences in cell membrane composition between S. cerevisiae and C. glabrata and may therefore indicate which phospholipids are responsible for the phenotypic differences between the two species. C. glabrata is a constitutively haploid strain of yeast, no diploid cells have ever been isolated, and it divides in an axial manner similar to haploid strains of S. cerevisiae.24 EXPERIMENTAL SECTION Sample Preparation. C. glabrata was acquired from DSMZ (German Collection of Microorganisms and Cell Cultures, Braunschweig, Germany), isogenic S. cerevisiae haploid (BY4742, MAT∆ his3∆1 leu2∆0 lys2∆0 ura3∆0; BY4741, MATa his3∆1 leu2∆0 met15∆0 ura3∆0), and S. cerevisiae diploid (BY4743, MATa/ (19) Flescher, E. G.; Madden, K.; Snyder, M. J. Cell Biol. 1993, 122, 373. (20) Lachenbach, P. A.; Mickey, M. R. Tectrometrics 1968, 10, 1. (21) Zahner, J. E., Harkins, H. A., Pringle, J. R. Mol. Cell Biol. 1996, 16, 1857. (22) Desfarges, L. P.; Durrens, H.; Juquelin, H.; Cassagne, C.; Bonneau, M.; Aigle, M. Yeast 1993, 9, 267. (23) Revordel, E.; Bonneau, M.; Durrens, P, Aigle, M. Biochim. Biophys. Acta 1995, 1263, 261. (24) Lachke, S. A.; Joly, S.; Daniels, K.; Soll, D. R. Microbiology 2002, 148, 2661

MAT∆ his3∆1/his3∆1 leu2∆0/leu2∆0 lys2∆0/LYS2 MET15/ met15∆0 ura3∆0/ura3∆0) have been described previously.25 All strains were grown in 100 mL of standard mineral medium5 at 30 °C (200 rpm) for 4 days until they had reached the stationary phase. The cells (1 mL) were harvested by centrifugation (2500 rpm for 3 min), and the biomass was thoroughly washed by adding consecutively 3 × 1 mL of saline (1% NaCl). A 100-µg sample of cell material was resuspended in 10 µL of saline and used for TOFSIMS analysis. TOF-SIMS Analysis. An aliquot of each sample (3 µL) was applied onto a clean silicon wafer to form a thin coating and was air-dried. The wafer is mounted onto a copper stub, which is inserted into the Bio-TOF-SIMS (described elsewhere26) via a fast entry lock, allowing introduction into the ultrahigh vacuum. A gold liquid metal ion gun (Ionoptika) is used as the primary ion source at 15 keV. It is capable of delivering a beam of 6 nA in dc mode onto the surface.27 A pulsed beam of width of 40 ns was used for TOF-SIMS analysis. Each spectrum was acquired by scanning the beam over a sample area 400 µm × 400 µm. Positive secondary ions were collected in the mass range up m/z 1200 using 106 Au+ pulses. Five repeat spectra were obtained at different points on each sample. Statistical Analysis. The acquired data were binned to 1 Da. Data processing was carried out with the statistical package SPSS+ (version 11.0) using the mass range between 200 and 1200 mass units to detect differences between the four yeast species. Ions lower than mass 200 were excluded from the study to avoid contaminating ions from salts, system contaminants, and other mineral medium components. Each acquired spectrum was then normalized, setting the peak sum to 100%. For ion selection, a MANOVA was performed using the four data sets. All ions, which showed 99% or higher variance differences, were selected for a first principal component analysis (PCA). To show that all four data sets could be separated, a Fisher’s discriminant analysis28,29 was performed using the six factors resulting from the PCA. The performance of the discriminant model was verified by applying the cross-validation procedure based on the “leave-one-out” crossvalidation formalism.30 Cluster analysis was performed with the factors resulting from the PCA,31 and using the Ward method in combination with the squared Euclidean distance as interval measure. RESULTS From the 34 ions, which were selected for the second PCA, 16 could be assigned to molecular ions of phosphatidylglycerols, phosphatidylethanolamines, phosphatidylserines, phosphatidylcholines, phosphatidylinositols, and cardiolipins using FAB-MS and TOF-SIMS library data8-14,32 (see Table 1). Using the same FAB-MS and TOF-SIMS library data, 18 ions could be tentatively (25) Brachmann, C. B.; Davies, A.; Cost, G. J.; Caputo, E.; Li J. C.; Heiter, P.; Boeke, J. D. Yeast 1998, 14, 115. (26) Braun, R. M.; Blenkinsopp, P.; Mullock, S. J.; Corlett, C.; Willey, K. F.; Vickerman, J. C.: Winograd, N. Rapid Commun. Mass Spectrom. 1998, 12, 1246 (27) Davies, N.; Weibel, D. E.; Blenkinsopp, P.; Lockyer, N.; Hill, R.; Vickerman, J. C. Appl. Surf. Sci. 2003, 203, 223. (28) Dillon, W. R.; Goldstein, M. Multivariate Analysis; Wiley: New York, 1984. (29) Everitt, B. Cluster Analysis; Social Science Research Council, Heinemann Educational Books Ltd.: Oxford, U.K., 1974. (30) Krazanowski, W. J. Tectrometrics 1990, 7, 81-98. (31) Ward, J. H. 1963, J. Am. Stat. Assoc. 1963, 58, 236

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

1741

Table 1. Suggested Molecular Ion Assignments of Some Higher Molecular Mass (450-1200 Da) Ions That Were Identified as Characteristic Ions To Discriminate between C. glabrata and Each of the Three S. cerevisiae Strains ion, m/z

proposed phospholipid

fatty acid side chain

484 496 508 512 513 519 534 571 679 719 731 733 842 843 854 1193

phosphatidylethanolamine phosphatidylcholine phosphatidylethanolamine phosphatidylethanolamine phosphatidylglycerol phosphatidylglycerol phosphatidylethanolamine phosphatidylglycerol phosphatidylglycerol phosphatidylglycerol phosphatidylinositol phosphatidylinositol phosphatidylserine phosphatidylinositol phosphatidylserine L-lysylcardiolipin

C 17:5 C 17:0 C 19:0 C 20:5 C 18:6 C 18:3 C 21:1 C 22:5 C 29:0 C 32:1 C 27:4 C 27:3 C 32:2 C 35:4 C 33:1 C 54:3

assigned to phospholipid fragments (results shown in Table 2). The first three principal components described 72.20% of the total variance. The discriminant analysis resulted in three discriminant functions for a correct sample classification into four groups. All three discriminant functions were significant: function 1 describing 55.8% of the common variance; function 2, 23.4%; and function 3, 20.8%. The discriminant function analysis resulted in no group mismatch for all 24 samples. All 24 single data points (samples) were assigned correctly to one of the haploid strains S. cerevisiae BY4741 or BY4742, the S. cerevisiae diploid strain BY4743, or the C. glabrata group (Figure 1). In the leave-one-out cross-validation,

Figure 1. Values of two discriminant scores (discriminant analysis) for the 24 yeast samples for 34 principal ions, which were selected to discriminate between S. cerevisiae (haploid strains BY4741 and BY4742) (BY71 and BY72), S. cerevisiae diploid strain BY4743 (BY73), and C. glabrata (Cglab). S. cerevisiae (haploid strain BY4741) group; +. group centroids.

all four ungrouped yeasts were correctly assigned to the right group. A Cluster analysis, using the Ward (or “increase in square sum”) method, was performed using these 34 ions. This method favors “spherical” clusters with equal variance and sample size. Ward’s method is based on minimizing the within-cluster sum of squares.31 The cluster analysis showed that the four yeasts could be separated using this method (see Figure 2). Every data subset correctly clustered to the haploid S. cerevisiae cluster BY4741 or BY4742, the S. cerevisiae diploid strain BY4743 cluster, or the C. glabrata cluster. When the next least scoring ion resulting from the first principal component analysis was omitted from the

Table 2. Suggested Ions Assignments of Lower Molecular Mass (200-450 Da) Fragment Ionsa

a Ions were identified as characteristic ions to discriminate among C. glabrata, S. cerevisiae haploid strains BY4741 and BY4742, and S. cerevisiae, diploid strain BY4743. R1 represents a fatty acid side chain, and R2 represents the phospholipid head group. All double bond assignments are tentative.

1742 Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

Figure 2. Dendrogram showing the results for the Ward cluster analysis using 34 principal ions, which were selected to discriminate between S. cerevisiae haploid strains BY4741 and BY4742 (BY71 and BY72), the S. cerevisiae diploid strain BY4743 (BY73), and C. glabrata (Cglab). Clusters are shown on a rescaled cluster distance of 25.

statistical analysis, the leave-one-out formalism resulted in one case mismatch, indicating that at least 34 ions are necessary to get a sufficient separation between the four data sets. Discriminant function analysis and cluster analysis were used to show that similar results can be achieved using two different methods, which provides assurance that the identified 34 ions can really be used for the separation of the four yeast data sets. While the discriminant function analysis tests the statistical validity of the prior group assignments, and establishes what distinguishes these various groups of clusters from one another, the cluster analysis uses mathematical procedures to first identify the candidate cluster solutions from all the possible cluster solutions. Combining both cluster analysis and discriminant function analysis provides an inductive technique for identifying distinct groups within the dataset. The assignment of the ions responsible for the separation of the four yeast species identified phosphatidylglycerols, phosphatidylethanolamines, phosphatidylserines, phosphatidylcholines, phosphatidylinositols, and ions resulting from phospholipid fragmentation. The ions m/z 395, 397, and 411 may result from fragmentation of phosphatidylglycerols, which are responsible for the separation of the four yeasts. Ion m/z 395 may result from the fragmentation of a phosphatidylglycerol with 11 carbon atoms and one additional double bond in one of the fatty acid side chains. In ion m/z 397, the fatty acid side chain may consist of 11 carbon atoms and no additional double bond. In ion m/z 411, the fatty acid side chain may consist of 12 carbon atoms without any additional double bonds. Ion m/z 284 may result from phospho(32) Vickerman, J. C.; Briggs, D.; Henderson, A. The Static SIMS Library, Version 3; SurfaceSpectra Ltd.: Manchester, U.K., 2001.

lipid fragmentation, where the phospholipid fragments into the headgroup and one hydroxy fatty acid side chain with 14 carbon atoms and no additional double bond (see Table 2). From the 13 fatty acids, 6 (undecenoic acid, dodecanoic acid, tridecadienoic acid, eicosatrienoic acid, tricosadienoic acid, tetracosanoic acid) have already been described for S. cerevisiae9,33,34 and undecanoic acid and heneicosenoic acid have been described for Candida humicola, Candida albicans and Candida tropicalis,33,34 while all others were not detected in these two yeast species before, but are already known as lipid constituents in nature. R-Hydroxy fatty acids were already identified in S. cerevisiae35 and Candida species.36 Comparison between C. glabrata and the Three S. cerevisiae Strains. From the 34 selected ions, 27 show significant differences in ion yields between C. glabrata and the three Saccharomyces strains. Using these 27 characteristic ions together with an (average ( standard deviation) comparison between the four yeasts, showed the following: (a) Cardiolipins. Cardiolipins (represented by ion m/z 1193) are more abundant in C. glabrata (see Figure 3) in comparison to all investigated Saccharomyces strains. (b) Phospholipids. C. glabrata showed lower levels of phosphatidylglycerols with one shorter fatty acid chain (C11:0, C12:0, C13:2, C14:4) in comparison to the diploid Saccharomyces strain BY4743. The short-chain phosphatidylglycerols (C18:3, C22:5) also show lower abundances in C. glabrata than in the diploid strain BY4743. Higher amounts of ion m/z 842 indicate significantly higher levels of phosphatidyl serine C32:2 in C. glabrata than in the diploid Saccharomyces strain BY4743, while the phosphatidylserines are similar to both parental strains BY4742 and BY4741 (see Figure 3). C. glabrata contains significantly lower amounts of the ethanolamine C19:0 compared with strain BY4743. C. glabrata had significantly higher levels of the phosphatidylinositols C27:4 and C35:4 than the diploid strain BY4743 (c) Fatty Acids. C. glabrata was found to contain lower amounts of undecenoic acid (C11:1) and consistently higher amounts of octadecapentenoic acid (C18:5) compared to all S. cerevisiae strains examined. For fatty acids with lower carbon atom chain lengths (C11-C14), C. glabrata always had lower amounts than the diploid Saccharomyces strain BY4743. DISCUSSION Using principal components analysis in combination with discriminant function analysis and cluster analysis as statistical tools resulted in the identification of 34 ions that are necessary and sufficient to discriminate between the four yeast strains: 16 ions of higher molecular mass (450 up to 1000 Da, Figure 3), which may predominantly result from molecular ions of different phospholipid species, and 18 ions which may result from phospholipid fragmentation. Since both the haploid S. cerevisiae strains and C. glabrata bud in an axial manner, while budding is polar in the diploid strain BY4743, the results show that polar budding may be accompanied (33) Homer, D. Meijeritieteellinen Aikakauskirja 1983, 41, 66-68. (34) Zalashko, M. V.; Andreevskaya, V. D.; Obraztsova, N. V. Gidroliznaya i Lesokhimicheskaya Promyshlennost 1977, 1, 8-10. (35) Mitchell, A. G.; Martin, C. E. J. Biol. Chem. 1997, 272, 28281. (36) Stodola, F. H.; Deinema, M. H.; Spencer, J. F. T. Bacteriol. Rev. 1967, 31, 194.

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

1743

Figure 3. Relative abundances of characteristic ions (450-1200 Da) used to separate four different yeasts: S. cerevisiae (haploid strain BY2742), C. galbrata, S. cerevisiae (diploid strain BY4743), and S. cerevisiae (haploid strain BY4741). The x-axes shows the selected ions; the y-axes shows the relative ion abundance (calculation: S% × 1000).

by higher levels of short-chain phosphatidylglycerols and higher amounts of N-containing phospholipids phosphatidylserine C32:2 and phosphatidylethanolamine C19:0. Polar budding strains seem also to have higher amounts of phosphatidyl inositols than axial budding strains. This may reflect fundamental chemical differences associated with these two modes of yeast growth. The higher amounts of cardiolipins in C. glabrata in contrast to all three S. cerevisiae strains may indicate different drug resistance patterns in C. glabrata and S. cerevisiae strains. The results suggest that the method employed selects for rather uncommon fatty acids in order to distinguish between C. glabrata and the three Saccharomyces strains. Some of these unusual fatty acids may occur in only low quantities in the side chains of the phospholipids. These rare fatty acids have not been detected in previous studies of total lipid extracts from these two yeast species. The utility of these uncommon fatty acids in our analyses may indicate that their ratios are more stable and thus of more diagnostic significance. General Membrane and Biochemical Differences among the Four Yeasts. C. glabrata exhibited significantly higher amounts of cardiolipin than the three Saccharomyces strains investigated. Cardiolipins have already been identified as phospholipids in Candida albicans, Candida guilliermondii, Candida membranefaciens, and Candida (Yarrowia) lipolytica37,38,39,.40 This study identified, for the first time, cardiolipins in C. glabrata and showed them to be an integral part of the cell surface. C. glabrata showed increased levels of the phosphatidylserine C32:2 in comparison to the diploid S. cerevisiae strain. Phosphatidylserines selectively affect amino acid transport in yeast cells.41 Another study showed that phosphatidylserines were present in higher concentrations in Candida species, while they were decreased in Saccharomyces strains.16 Again, this indicates that (37) Abu-Elteen, K. H.; Whittaker, P. A. Mycopathology 1998, 140, 69. (38) Khaware, R. K.; Koul, A.; Prasad, R. Biochem. Mol. Int. 1995, 35, 875. (39) Lee, J. R.; Takenaka, H.; Takahashi, N.; Makise, M.; Yamaguchi, Y.; Tsuchia, T.; Mizushima, T. J. Biochem. 2002, 131, 541. (40) Trivedi, A.; Singhal, G. S.; Prasad, R. Biochim. Biophys. Acta 1983, 729, 85. (41) Kang, Z.; Jiang, K.; Bian, Y.; Cao, Y. Huaxue Tongbao 1982, 11, 655.

1744 Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

the method described here is able to identify significant differences between different yeasts. CONCLUSIONS TOF-SIMS in Combination with Chemometric Analysis Is Able To Discriminate among Four Different Yeasts. The study showed that 34 ions are sufficient to distinguish among the yeasts C. glabrata, the haploid S. cerevisiae (BY4742 and BY4741) strains, and the diploid S. cerevisiae (BY4743) strain. Discriminant function analysis and cluster analysis results showed that the four yeasts can be separated into four groups, each group representing a single yeast strain. Ion Assignments Are Possible as Phospholipid Molecular Ions or Phospholipid Fragment Ions. The 34 ions can be assigned to phospholipid molecular ions or phospholipid fragments, which are the predominant molecules of the cell membrane of yeast cells. From the tentatively assigned ions, five have already been described as major fatty acid side chains of cell membrane phospholipids for at least one of the two yeast species studied. Differences between the phospholipid pattern observed in earlier studies may be due to the fact that we used a defined mineral medium for yeast growth, while all the earlier analyses were performed with yeast cultures fermented in complex media. The results show that the ion assignments are reasonable, but also that the chemometric system does not use the most abundant ions, but only those accessible by bombardment of the cell surface. Moreover, in seeking consistent differences between strains, the chemometric analysis may focus on specific differences between the different yeast strains that are due to “uncommon” compounds among phospholipids and fatty acid side chains. Such compounds may have rather unusual structures (e.g., odd or hydroxy fatty acid side chains of the phospholipids) and may only be present in relatively small amounts in one of the examined sample sets that were chosen by the statistical system to distinguish between the strains. Therefore, this approach could also be used to highlight and target unusual fatty acids or phospholipids in yeasts, which may have pharmaceutical relevance.

Outlook. Ultimately, the method can be used in combination with the already established freeze-fracture method4 to examine biological material for membrane differences not only of the cell membrane but also inside the cell for either cell organelle membrane composition differences between different yeast species or different growth modes and physiological states within a species. The method could be used to gain new insights into the spatial distribution of phospholipids and phospholipid metabolites within living cells and into the mechanisms used to transport lipids between different cell organelles.

ACKNOWLEDGMENT H.J. was supported by a UK EPSRC Postdoctoral Mobility Grant to S.G.O.

Received for review August 14, 2004. Accepted December 20, 2004. AC048792T

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

1745