Anal. Chem. 1998, 70, 1555-1562
Direct Mass Spectrometric Analysis of in Situ Thermally Hydrolyzed and Methylated Lipids from Whole Bacterial Cells Franco Basile, Michael B. Beverly, Christy Abbas-Hawks, Curtis D. Mowry,† and Kent J. Voorhees*
Department of Chemistry and Geochemistry, Colorado School of Mines, Golden, Colorado 80401-1887 Ted L. Hadfield
Armed Forces Institute of Pathology, Building 54, CPS-m, Washington, D.C. 20306-6000
Fatty acid methyl esters (FAMEs) were generated in situ, during pyrolysis, from whole-cell bacterial samples and analyzed by mass spectrometry (MS). The FAME profiles obtained by an in situ thermal hydrolysis methylation (THM) step were compared with gas chromatography (GC) and MS analyses of the chemically extracted and methylated fatty acids. This correlation was based on the ability of each technique to differentiate a representative group of 15 bacteria at the species level as predicted by principal component analysis. All three analyses, GC/FAME, pyrolysis-MS/FAME, and in situ THM-MS/FAME differentiated the studied bacterial sample set into three discrete clusters. The bacteria comprising each cluster were the same for all three analyses, showing that taxonomic information of the lipid profiles was preserved in the PyMS/FAME and in situ THM-MS/FAME analyses of whole cells. Contributions from saturated, unsaturated, cyclopropyl, and branched bacterial fatty acids to the differentiation of microorganisms were identified for all three analyses. The in situ THM-MS/FAME approach is simple, requires small samples (∼2 × 106 cells/profile), and is rapid, with a total analysis time under 5 min/sample. The need for an automated, real-time, and field-portable instrument for the chemical analysis of bacterial components has sparked a multitude of research projects that apply instrumental methods of chemical analysis to microbiology (chemotaxonomy). At the present time, no single analytical technique can achieve the level of specificity, sensitivity, and selectivity required for a universal field-portable bacterial detection and identification system.1,2 New advancements in the miniaturization of several instrumental techniques such as mass spectrometry have overcome some of the difficulties in the development of a field-portable biological detector. However, simplification, miniaturization and † Present address: Sandia National Laboratories, Dept. 1823, P.O. Box 5800, MS 0343, Albuquerque, NM 87185-5800. (1) Program for the Third International Symposium on the Interface Between Analytical Chemistry and Microbiology: Analytical Chemistry in Environmental Microbiology. Knoxville, TN, March 12-15, 1995. (2) Program for the First Joint Services Workshop on Biological Mass Spectrometry. Baltimore, MD, July 28-30, 1997.
S0003-2700(97)00970-0 CCC: $15.00 Published on Web 03/15/1998
© 1998 American Chemical Society
automation of the bacterial sample preparation step remains as one of the challenges in the construction of a field-portable biodetector. Because of its success in the laboratory, lipid profiling has been considered as a viable approach for biodetection. The goal of this study was to simplify the lipid-profiling sample preparation step to make it amenable to field operation and compare the new methodology against established procedures. Extensive studies correlating lipid composition to taxonomic trends have been published,3,4 and bacterial lipid profiling has been applied to solve a variety of problems.5-7 This approach has become routine in the microbiology laboratory with the introduction of standardized methylation/gas chromatographic (GC) techniques that detect the extracted/methylated bacterial fatty acids (FAs).8 One commercially available instrument is the MIDI system.9 This GC-based FAME analysis consists of a multiplesample 60-min lipid extraction/methylation from whole bacterial cells followed by a 15-20-min chromatographic run. The resulting chromatograms or fatty acid methyl ester (FAME) profiles are then compared with an existing computerized database and possible matches are identified. Due to the lipid extraction process and chromatographic analysis, the GC/FAME or MIDI methodology is not amenable for field-portable operation. By eliminating the chromatographic step, our laboratory performed the direct analysis of the chemically extracted/methylated FAMEs from a representative set of microorganisms using pyrolysis mass spectrometry (Py-MS).10 Differentiation of microorganisms by Gram type, genera, and species was achieved based solely on 10 FAMEs and their electron ionization (EI) fragment ions. Analysis time was reduced from (3) Abel, K.; deSchmertzing, H.; Peterson, J. L. J. Bacteriol. 1963, 85, 1039. (4) Shaw, N. In Advances in Applied Microbiology; Perlman, D., Ed.; Academic Press: New York, 1974; Vol. 17, pp 63-104. (5) Vestal, J. R.; White, D. C. BioScience 1989, 39, 535-541. (6) Odham, G.; Tunlid, A.; Westerdahl, G.; Larsson, L.; Guckert, J. B.; White, D. C. J. Microbiol. Methods 1985, 3, 331-334. (7) Breitschwerdt, E. B.; Kovdick, D. L.; Keene, B.; Hadfield, T. L.; Wilson, K. J. Clin. Microbiol. 1995, 33, 154-160. (8) Wayne, C. M. In Analytical Microbiology Methods: Chromatography and Mass Spectrometry; Fox, A., Morgan, L. S., Larsson, L., Odham, G., Eds.; Plenum Press: New York, 1990; pp 59-69. (9) Microbial ID, Inc. Operating Manual, Ver. 3.0, Newark, DE, 1993. (10) Basile, F.; Hadfield, T. L.; Voorhees, K. J. Appl. Environ. Microbiol. 1995, 61, 1534-1539.
Analytical Chemistry, Vol. 70, No. 8, April 15, 1998 1555
the 15-20-min chromatographic run to a less than 1-min mass spectral analysis. The limiting step in the overall analysis, however, remained the chemical extraction/derivatization of lipids from whole-cell bacterial samples, which required ∼60 min/ sample. Moreover, for methods requiring lipid extraction and derivatization, such as the MIDI system,9 a large number of cells, ∼4 mg of dry cells or 2.4 × 1010 cells, are required to achieve profiles with a useful signal-to-noise ratio (based on 5.9 × 1012 cells/g of dry weight6). A reduction in the number of cells needed to obtain a lipid profile would translate into shorter growth periods to obtain a sufficiently large bacterial culture suited for the analysis. This in turn would translate into a reduction of the total analysis time. However, the bacterial growth period prior to analysis remains the ultimate limiting step of all chemotaxonomic analyses. Hence, there is a need to reduce the assay time (sample preparation time + analysis time) as well as sample size for whole-cell FA profiling without compromising the taxonomic information content of the data. This has been achieved using in situ thermal hydrolysis/ methylation (THM) of bound and free FAs from whole bacterial cells.11 In situ THM of organic acids is accomplished by mixing the sample with a methylating agent such as tetramethylammonium hydroxide (TMAH) or phenyltrimethylammonium hydroxide (PTMAH). A review of several methylating agents available for in situ THM has been published.12 Upon mixing, a quaternary N-methylammonium salt of the carboxylic acid forms, and with heating, the methylation at the acidic hydrogen(s) site(s) in the sample takes place. Moreover, saponification of ester-bound FAs also takes place during the in situ methylation, hence the name thermal hydrolysis methylation or THM. This process requires only seconds to minutes and results in a versatile method to methylate a variety of organic compounds containing NH and OH functional groups. The simplicity of the in situ THM technique makes this approach ideal for an automated sample preparation step in the analysis of biological markers from whole bacteria in a field situation. In addition, unlike the traditional chemical derivatization step that generates relatively large volumes of solvent waste, the in situ THM methodology uses small volumes of reagent (microliters), decreasing the possibility of sample loss and contamination during the methylation procedure. The in situ THM assay coupled with GC analysis has been applied to the analysis of acids in wines, polyesters, waxes, and bacterial components,13 including lipids and spore biomarkers (e.g., dipicolinic acid) from whole bacterial cells.11,12,14 In these studies, the bacteria/TMAH sample was coated onto a Curie point pyrolysis wire and the FAMEs (and other methylated biomarkers) generated during pyrolysis were analyzed by GC/MS. The sample preparation step was significantly reduced, yet the total analysis time was limited by a chromatographic step ranging from 5 to 30 min. Previous studies where the chromatographic step was replaced by the direct mass analysis of glycerides and phospholipids15 and chemically extracted and methylated bacterial (11) Holzer, G.; Bourne, T. F.; Bertsch, W. J. Chromatogr. 1989, 468, 181-190. (12) Kossa, W. C.; MacGee, J.; Ramachandran, S.; Webber, A. J. J. Chromatogr. Sci. 1979, 17, 177-187. (13) Challinor, J. M. J. Anal. Appl. Pyrolysis 1996, 37, 185-187. (14) Dworzanski, J. P.; Berwald, L.; Meuzelaar, H. L. C. Appl. Environ. Microbiol. 1990, 56, 1717-1724.
1556 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998
Table 1. Microorganisms Used in This Study index
Gram type
microorganism
A B E F G H I J L N O R S T U
+ + + + + + + + -
Bacillus cereus Bacillus circulans Bacillus subtilis Bacillus thuringiensis Enterococcus (Streptococcus) faecalis Listeria monocytogenes Staphylococcus aureus subsp. aureus Staphylococcus epidermidis Enterobacter aerogenes Proteus mirabilis Providencia stuartii Pseudomonas fluorescens Pseudomonas putrefaciens Pseudomonas stutzeri Serratia marcescens
FAs10 have shown no compromise of the taxonomic information content provided by the resulting lipid profiles. The next step in developing the in situ THM-MS methodology is the application of this technique to the direct analysis of FAME profiles from whole bacteria. The work presented here demonstrates the capability of in situ THM-MS as a viable sample preparation step for the analysis of FAME profiles from whole bacterial cells. A direct comparison is presented between in situ THM-MS FAME profiling of whole cells and both GC and Py-MS analyses of the chemically extracted FAMEs of the same bacterial sample set. Principal component analysis (PCA) was used to evaluate the various data sets generated by each approach. Taxonomic information and data reproducibility of this methodology will be discussed as well as its implementation as a rapid sample preparation step for an automated real-time field portable bacterial identification system. EXPERIMENTAL SECTION Bacterial Samples. The microorganisms used in this study are listed in Table 1. All microorganisms were grown on a tryptic soy agar at 37 °C for 18 h. Bacterial cells were isolated by washing with distilled water, and then split into two equivalent sets. For bacteria in one set (sample set A), lipids (i.e., FAs) were extracted and methylated as described by Miller.16 FAME samples were redissolved in 1:1 ether-hexane as needed due to solvent evaporation. Bacteria in sample set B were spun down and lyophilized. Suspensions of the lyophilized bacteria in the 108 CFU/mL (CFU, colony forming units) range were prepared in distilled water and were used within the same day. As with any live class-II biological sample, care must be exercised in handling them. All biological samples were handled with latex gloves and inside a class-II biosafety cabinet. No sharp instruments were employed during the preparation of the bacterial suspension. Reagents. TMAH (Sigma), 0.1 M in HPLC grade methanol, was used for all in situ THM reactions (CAUTION: TMAH is a strong base and methylating agent). TMAH was chosen for its high methylating efficiency and its low molecular weight blank. All standard straight-chain saturated, unsaturated, hydroxy-FAMEs (15) DeLuca, S. J.; Sarver, E. W.; Voorhees, K. J. J. Anal. Appl. Pyrolysis 1992, 23, 1-14. (16) Miller, L. T. J. Clin. Microbiol. 1982, 16, 584-586.
(Aldrich) and branched iso- and anteiso-FAMEs (Ultra Scientific) were used without further purification. All FAME standards were dissolved in methanol (10 mg/mL). Instrumental Procedures. An Extrel ELQ-400 triple quadrupole mass spectrometer with a Curie point pyrolysis inlet was used for all of the direct analyses. A description of the instrument has been previously published.17 All EI mass spectra were collected at 70-eV electron energy and a mass range of 180-360 Da. This mass range was selected to match the range of FAMEs detected by the GC-FAME analysis and to avoid the TMAH blank fragment ion ((CH3)2NdCH2)+ at m/z 58). Mass spectra for data analysis were generated by averaging scans 5 through 10 in the pyrogram. For collision-induced dissociation (CID) experiments, argon at a pressure of 10-5 Torr was used as the collision gas in quadrupole 2. Tandem MS experiments were only performed to confirm the identity and validity of the peak assignment, and they would not be required to analyze an unknown. Curie point pyrolysis was performed at 358 °C, unless noted otherwise, with a 50-ms rise time and a pyrolysis time of 10 s. A 5-cm-long glass transfer line (between the pyrolysis chamber and the ion source) and the ion source itself were both maintained at 150 °C. Sample Preparation. For the bacterial FAME extract samples, a 10-µL aliquot of the FAME solution was applied to the Curie point wire and air-dried at room temperature. The total mass of FAMEs deposited on the wire was estimated to be in the lowmicrogram range. Samples of standard compounds and/or whole cells for Curie point pyrolysis-MS analysis were prepared by placing 25 µL of solution or suspension onto a rotating wire followed by solvent evaporation under a stream of warm air. For in situ THM experiments, a 5-µL aliquot of 0.1 M TMAH11 was added onto the sample-coated Curie point wire. The resulting mixture of TMAH and analyte was then evaporated to dryness. Other investigators have found similar results by premixing whole cells with TMAH.18 This study was performed with lyophilized cells; however, equivalent results are obtained if the cells are collected directly from the agar plate with the Curie point wire and subsequent pyrolysis in the presence of TMAH. Pattern Recognition. PCAs were carried out with the RESOLVE software package developed at the Colorado School of Mines.19,20 Each mass spectrum was collected as a set of raw intensities. The data were normalized to total intensity and mean centered. Both PC score and loading plots were generated. The loadings or eigenvector coefficients indicate the magnitude of the contribution of a variable (in this case the mass spectral peaks) in comprising the projected score.21 For the chromatographic data, output reports generated by the MIDI system were digitized into a spreadsheet and converted into a RESOLVE file format (integers). The average equivalent chain length (ECL) (proportional to the FAME GC retention time8) for each of the FAMEs identified in the MIDI system were rounded to the nearest 1/100th decimal place. (17) DeLuca, S.; Sarver, E. W.; Harrington, P. deB.; Voorhees, K. J. Anal. Chem. 1990, 62, 1465-1472. (18) Holzer, G. Personal communication. (19) Harrington, P. D. B.; Street, T. E.; Voorhees, K. J.; Radicati di Brozolo, F.; Odom, R. W. Anal. Chem. 1989, 61, 715-719. (20) Harrington, P. D. B.; Voorhees, K. J. Anal. Chem. 1990, 62, 729-734. (21) Sharaf, M. A.; Illman, D. L.; Kowalski, B. R. In Chemometrics; Elving, P. J., Winefordner, J. D., Eds.; John Wiley & Sons: New York, 1986; Chemical Analysis, Vol. 82, p 202.
Figure 1. Pyrolysis-mass spectra of triglycerides: (a) C14:0/C14: 0/C16:0 without TMAH, (b) C14:0/C14:0/C16:0 with TMAH (C16:0 ME, FW 270; C14:0 ME, FW 242), and (c) C18:1/C18:1/C16:0 with TMAH (C18:1 ME, FW 296).
RESULTS AND DISCUSSION The in situ THM of lipids from whole cells cannot alter the chemical structures of the individual FAs during the reaction and must be quantitative. These characteristics have been demonstrated for in situ THM-GC/MS11 of triglycerides and phospholipids using TMAH as the methylating agent. In situ THM in the GC assay of whole cells is done at atmospheric pressure, while the direct MS approach presented here was performed in vacuo. To assess the quantitative nature of the in situ THM procedure used in this study, a test was conducted using the triglycerides C14:0/C14:0/C16:0 (FW 750) and C16:0/C18:1/C18:1 (FW 859) pyrolyzed in the presence of TMAH. Figure 1 shows the Py-mass spectra of these triglycerides with and without TMAH. The Py-mass spectrum of the triglyceride without TMAH (Figure 1a) shows the intact thermal desorption of the molecule with subsequent homolytic cleavage of the ester bond at all three positions yielding the fragment ions at m/z 523 (loss of C14H27O2•) and at m/z 495 (loss of C16H31O2•). No free fatty acid molecular ions were observed at m/z 228 (C14:0) and/or m/z 256 (C16:0). In Figure 1b, the Py-mass spectrum of the triglyceride C14:0/ C14:0/C16:0 with TMAH showed that each glycerol-bound fatty acid was methylated to its corresponding FAME (C16:0 ME at m/z 270, and C14:0 ME at m/z 242). No fragment ions due to pyrolysis and/or EI of the intact triglyceride (shown in Figure 1a) were observed. In situ THM of the glycerol-bound fatty acids was quantitative and complete in that the 2:1 ratio of C14:0 to C16:0 in the triglyceride molecule was observed in the spectrum and that no fragment ions were observed from the original triglyceride. This also holds for glycerol-bound unsaturated fatty acids as is shown in Figure 1c for the triglyceride C16:0/C18:1/ C18:1. However, the 2:1 ratio of C18:1 to C16:0 is observed between the C16:0 ME molecular ion (m/z 270) and the C18:1 ME M-32 fragment ion (m/z 264). This was also corroborated with the triglyceride C16:0/C16:0/C18:1 (mass spectrum not shown). Analytical Chemistry, Vol. 70, No. 8, April 15, 1998
1557
Table 2. FAME Ions Useful for Py-MS Bacterial Identification
Figure 2. P. stuartii pyrolysis-mass spectra of (a) whole cell, (b) whole cell with TMAH, and (c) FAME extract.
In Situ THM-MS/FAME Analysis of Whole Bacteria. The in situ THM reaction of free and bound fatty acids from whole bacteria is illustrated in Figure 2 for the Gram-negative bacterium Providencia stuartii. Mass spectra from pyrolysis of the whole cell, whole cell with TMAH, and the chemically extracted/ derivatized FAMEs are shown. Py-mass spectra of whole cells (Figure 2a and b only) were obtained with the same number of cells, ∼2 × 106 cells; hence a direct comparison of bacterial profiles and absolute ion intensities between the two spectra is possible. The FAME sample mass spectrum shown in Figure 2c was obtained from the chemical derivatization and extraction of 1010 cells. The Py-mass spectrum of whole cells (Figure 2a) produced ions corresponding to unbound metabolic fatty acids and free FAs from the pyrolysis of phospholipids: palmitic acid (C16:0, FW 256), traces of myristic acid (C14:0, FW 228), and oleic acid (C18: 1, FW 282; M - 18 at m/z 264). Ions corresponding to FAMEs dominate the mass spectrum of whole bacteria pyrolyzed in the presence of TMAH (Figure 2b). Palmitic acid and myristic acid are now methylated and they appear at a mass 14 Da higher, m/z 270 and 242, respectively. Even though the S/N of the in situ THM analysis is lower than the whole cell only analysis (base peak), greater selectivity is achieved by detection of a wider range of FAs. Other FAMEs observed include C18:1 ME (m/z 296) and its fragment ions at M - 32 (m/z 264) and M - 74 (m/z 222), cyclopropyl-C17:0 ME (cyC17:0 ME, m/z 282) and its fragment ion at M - 32 (m/z 250), and cyC19:0 ME (m/z 310) and its fragment ions at M - 32 (m/z 278) and M - 74 (m/z 236). Fatty acid methyl esters and their EI-fragment ions identified in this work are summarized in Table 2. These FAME ions were identified by comparison of the bacterial lipid profiles to mass spectra of known standard compounds and to a NIST MS database. Many FAME-derived ions detected in the in situ THMmass spectrum (Figure 2b) match those observed in the mass spectrum of the bacterial FAME extract sample (Figure 2c). Ions listed in Table 2 are a collection of all the ions observed from bacteria listed in Table 1; hence not all ions would be observed in a particular bacterial mass spectrum. Moreover, as in the case of unsaturated FAMEs, the most intense ion is the fragment ion resulting from a loss of methanol, M - 32, and not its molecular ion. The higher background observed in Figure 2b is due to an 1558 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998
FAME
m/z
FAME
m/z
C12:0 ME
214
C14:0 ME
242
cyC17:0 MEa M - 31 M - 32 M - 74
282 251 250 208
C15:0 ME M - 31
256 225
C18:0 ME M - 31
298 267
aC15:0 ME M - 57
256 199
iC15:0 ME M - 43
256 213
C18:1 ME M - 31 M - 32 M - 74
296 265 264 222
C16:0 ME M - 31 M - 29b M -14 M - 14 M - 14 M - 14
270 239 241 227 213 199 185
C19:0 ME M - 31
312 281
cyC19:0 ME M - 31 M - 32 M - 74
310 279 278 236
C16:1 ME M - 31 M - 32 M - 74
268 237 236 194
C20:0 ME M - 31
326 295
C21:0 ME M - 31
340 309
C17:0 ME M - 31
284 253
C22:0 ME M - 31
354 323
iC17:0 ME M - 43
284 241
a Also C17:1 ME. b Fragments represent the carbomethoxy-homologous series.
increase in the chemical noise and the methylation and detection of other biomarkers such as methylated DNA nitrogen bases22 (guanine with 3-methylations at m/z 193 and with 4-methylations at m/z 207). Because the sample preparation is performed at the same time the sample is introduced into the MS (during the pyrolysis process), the sample preparation time is reduced considerably, from 1.5 h for the MIDI system to 5 min for the in situ THM-MS whole-cell analysis. Moreover, sample and analyte losses are reduced, as well as the possibility of sample contamination, when compared to the chemical extraction/methylation assay. The in situ THM-mass spectrum shown in Figure 2b was generated with ∼2 × 106 cells, instead of 1010 cells used in the chemically extracted FAME assay as recommended in the MIDI system.9,16 It should be noted that the 2 × 106 cells analyzed in this work were obtained by delivering a 25-µL aliquot of a 108 cells/mL stock bacterial suspension. Lower cell densities can be detected, however, at the expense of increased sample preparation time (increase in both sample loading and solvent evaporation time). Previous in situ THM-GC/MS FAME analyses of whole cells used about 5 × 106 cells11 and 3 × 107 cells.14 The amount of FAMEs detected can be estimated from known percent lipid composition in bacteria.23,24 For 2 × 106 cells (assuming 4 × 10-13 g of dry weight/cell), an estimated 1.0-0.75 (22) Abbas-Hawks, C.; Hadfield, T. L.; Voorhees, K. J. Rapid Commun. Mass Spectrom. 1996, 10, 1802-1806. (23) Bergey’s Manual of Systematic Bacteriology, Krieg, N. R., Ed.; Williams & Wilkins: Baltimore, 1984; p 378.
Table 3. Comparison of FAMEs Detected in P. fluorescens and B. cereus by in Situ Py-MS/FAME and GC/FAME Analysis (MIDI) FAME
GC/FAME extract % FAME
C10:0 C10:0 3-OH C12:0 C13:0 iso C12:0 2-OH C12:0 3-OH C14:0 C15:0 C16:1 w7c C16:1 w5c C16:0 C17:0 iso C17:1 w8c C17:0 cyclo C17:0 C18:1 w7c/w9t/w12t C18:0 C19:0 iso C19:0 cyclo
Pseudomonas fluorescens 0.13 3.41 10.33 0.09 0.82 4.72 0.75 0.16 22.25 0.08 28.79 0.24 0.10 17.65 0.15 9.04 0.37 0.25 0.58
C12:0 iso C13:0 iso C13:0 anteiso C14:0 iso C14:0 C15:0 iso C15:0 anteiso C16:1 w7c alcohol C16:0 iso C16:1 w7C C16:0 C15:0 2-OH C17:1 iso w10c C17:1 iso w5c C17:1 anteiso C17:0 iso C17:0 anteiso C18:0 c
Bacillus cereus 0.44 11.19 0.69 3.20 3.35 39.02 2.63 0.83 4.19 8.98 3.24 1.54 4.24 5.73 0.71 7.03 0.54 ndc
in situ Py-MS ion detecteda,b (m/z)
214 (45)
236 (25) 270 (100) 250 (40) 264 (45) 298 (
