Anal. Chem. 2001, 73, 751-757
Lectin and Carbohydrate Affinity Capture Surfaces for Mass Spectrometric Analysis of Microorganisms Jonathan L. Bundy* and Catherine Fenselau
Department of Chemistry and Biochemistry, University of Maryland, College Park, Maryland 20742
In a preliminary report (Bundy, J. L.; Fenselau, C. Anal. Chem. 1999, 71, 1460-1463), we demonstrated the use of lectin-derivatized surfaces to capture and concentrate complex carbohydrates as well as microorganisms from sample matrixes unamenable to direct MALDI mass spectrometry. Here, we extend the work to include samples representative of a wider variety of microorganisms of importance to human health and of enveloped viruses. In this study, lectins were immobilized directly to a membrane surface via primary amines. A complementary approach was also explored, using immobilized carbohydrates to capture bacteria via microbial lectins expressed on their surfaces. The carbohydrate-based surfaces were constructed by first immobilizing streptavidin to the membrane, followed by attachment of a commercially produced biotin/carbohydrate polymer. Acid treatment of the sample prior to mass spectrometric analysis permits the observation of protein biomarkers from the captured microbial samples in the 5-20 kDa mass range. Bacteria samples were detected from physiological buffers, urine, milk, and processed chicken samples using the biocapture probes. Viral samples were detected from culture based on glycoprotein moieties desorbed directly from the surface. The carbohydratebased system provided greater sensitivity than the lectin system, possibly due to the larger number of accessible saccharide ligands on the polymer. In the past decade, the development of matrix-assisted laser desorption/ionization (MALDI) has spurred renewed interest in the use of mass spectrometry to rapidly identify bacteria1 and viruses2 based on observation of biomarker compounds in the mass spectrum. Most reports have relied on extensive sample preparation to release and purify protein biomarker compounds from intact cells. In cases where whole cells were employed,3-5 they were taken from purified laboratory cultures. In real-world (1) Baar, B. L. M. v. FEMS Microbiol. Rev. 2000, 24, 193-219. (2) Thomas, J. J.; Bakhitar, R.; Siuzdak, G. Acc. Chem. Res. 2000, 33, 179-87. (3) Heller, D. N.; Fenselau, C.; Cotter, R. J.; Demirev, P.; Oltoff, J. K.; Honovich, J.; Uy, O. M.; Tanaka, T.; Kishimoto, Y. Biochem. Biophys. Res. Commun. 1987, 142, 194-199. (4) Chong, B. E.; Wall, D. B.; Lubman, D. M.; Flynn, S. J. Rapid Commun. Mass Spectrom. 1997, 11, 1900-1908. (5) Arnold, R. J.; Reilly, J. P. Rapid Commun. Mass Spectrom. 1998, 12, 630636. 10.1021/ac0011639 CCC: $20.00 Published on Web 01/12/2001
© 2001 American Chemical Society
situations, such as clinical diagnosis, food safety, and biological terrorism, the microorganisms would be present in a complex sample matrix (i.e., blood plasma, foodstuffs, or environmental debris), which would likely make direct analysis by mass spectrometry more difficult. Affinity techniques typically based on antigen/antibody interactions have previously been used by a number of investigators6-9 to capture and clean up protein analytes prior to analysis by MALDI mass spectrometry. Along the same lines, we have been developing affinity capture methods in our laboratory to enable the rapid, on-line concentration and cleanup of microbial samples prior to mass spectrometric analysis. Immunoaffinity-based techniques are also often used for the identification and characterization of microorganisms,10 but generally an antibody is only useful for the isolation of a given strain or species, limiting the utility of this approach when the capture of a number of microbial species is desired. In contrast, our approach has tested the hypothesis that protein/carbohydrate binding pairs can provide broad-band interactions to clean up and concentrate microorganism samples. Previously, lectins, carbohydrate binding proteins of nonimmune origin,11 have been shown to capture microbial samples via glycoconjugates exposed at or near the cell surface. Such interactions have previously been used to identify bacteria based on binding to a panel of lectins of varying carbohydrate specificity.12-14 Many bacteria also have lectins present on their cell surfaces. These proteins, termed microbial lectins or adhesins,15 play an important role in the initial stages of infection by mediating the interaction of pathogens with host cell surface glycoconjugates.16 Two examples of these proteins are the mannose binding adhesin of fimbriated Escherichia coli17 and the Lewis-b (Leb) binding (6) Hutchens, T. W.; Yip, T.-T. Rapid Commun. Mass Spectrom. 1993, 7, 576580. (7) Papac, D. I.; Hoyes, J.; Tomer, K. B. Anal. Chem. 1994, 66, 2609-2613. (8) Nelson, R. W.; Krone, J. R.; Bieber, A. L.; Williams, P. Anal. Chem. 1995, 67, 1153-1158. (9) Brockman, A. H.; Orlando, R. Anal. Chem. 1995, 67, 4581-4585. (10) Notermans, S.; Wernars, K. Int. J. Food Microbiol. 1991, 12, 91-102. (11) Lis, H.; Sharon, N. Chem. Rev. 1998, 98, 637-674. (12) Pistole, T. G. Annu. Rev. Microbiol. 1981, 35, 85-112. (13) Doyle, R.; Keller, K. Eur. J. Clin. Microbiol. 1984, 3, 4-9. (14) Slifkin, M.; Doyle, R. J. Clin. Microbiol. Rev. 1990, 3, 197-218. (15) Mirleman, D. Microbial Lectins and Agglutinins: Properties and Biological Activity; John Wiley and Sons: Chichester, U.K.,1986. (16) Finlay, B. B.; Falkow, S. Microbiol. Mol. Biol. Rev. 1997, 61, 136-169. (17) Lanne, B.; Olsson, B.-M.; Jovall, P.-A.; Angstrom, J.; Linder, H.; Marklund, B.-I.; Bergstrom, J.; Karlsson, K.-A. J. Biol. Chem. 1995, 270, 9017-9025.
Analytical Chemistry, Vol. 73, No. 4, February 15, 2001 751
adhesin (BabA) of Helicobacter pylori.18 Immobilized carbohydrates should also, therefore, be useful ligands for microbial biocapture. In a preliminary report,19 we demonstrated the feasibility of using lectin-derivatized surfaces for the capture of bacteria prior to analysis by MALDI mass spectrometry. The bacterium used in that study (E. coli 11775) was identified via phospholipids desorbed directly from the cell surface. Although these compounds are regarded as a useful biomarkers for bacteria,20,21 protein-based biomarkers are the focus of the majority of recent reports on microbial identification by mass spectrometry.22-24 A given microorganism expresses thousands of protein molecules with unique molecular masses that provide a greater diversity of compounds upon which to base identification. Protein biomarkers can also be used in conjunction with identification schemes based on computer interrogation of proteome database information.25,26 In the present work, we use lectin-based and glycoconjugatebased membrane biocapture surfaces constructed in our laboratory which permit the observation of protein biomarkers from both bacteria and enveloped viruses isolated from a variety of sample matrixes. Samples containing bacteria or virus are first allowed to incubate on the biocapture surface, which consists of a membrane derivatized either with lectin or biotinylated glycoconjugates. A series of washes are then performed to remove weakly bound contaminants. An acidic solution is then added, which facilitates the observation of proteins that are directly desorbed from the microorganisms on the surface of the analyte. EXPERIMENTAL SECTION Microbial Growth. E. coli (K-12 and O157:H7) and Salmonella typhimurium were grown in house using procedures previously reported.25 Stock cultures of E.coli K-12 were obtained from the American Type Culture Collection (Rockville, MD); other bacteria were obtained from the Center for Veterinary Medicine, U.S. Food and Drug Administration. (CAUTION: E. coli O157:H7 and S. typhimurium are pathogenic microorganisms that should be handled with due care. Biosafety Level 2 practices27should be in place in any laboratory handling these bacteria.) Viral Preparation. The enveloped virus sample used in this study (Sinbis AR-339) was the reconstituted infectious crude culture medium as received from the ATCC. The titer of this mixture was ∼2.1 × 106 plaque-forming units (pfu)/mL. (18) Boren, T.; Falk, P.; Roth, K. A.; Larson, G.; Normark, S. Science 1993, 262, 1892-1895. (19) Bundy, J. L.; Fenselau, C. Anal. Chem. 1999, 71, 1460-1463. (20) Heller, D. N.; Cotter, R. J.; Fenselau, C.; Uy, O. M. Anal. Chem. 1987, 59, 2806-2809. (21) Ho, Y.-P.; Fenselau, C. Anal. Chem. 1998, 70, 4890-4895. (22) Cain, T. C.; Lubman, D. M.; Walter, J.; Weber, J. Rapid Commun. Mass Spectrom. 1994, 8, 1026-1030. (23) Holland, R. D.; Wilkes, J. G.; Rafii, F.; Sutherland, J. B.; Persons, C. C.; Voorhees, K. J.; Lay, J. O. Rapid Commun. Mass Spectrom. 1996, 10, 12271232. (24) Welham, K. J.; Domin, M. A.; Scannell, D. E.; Cohen, E.; Ashton, D. S. Rapid Commun. Mass Spectrom. 1998, 12, 176-180. (25) Demirev, P. A.; Ho, Y.-P.; Ryzhov, V.; Fenselau, C. Anal. Chem. 1999, 71, 2732-2738. (26) Pineda, F. J.; Lin, J. S.; Fenselau, C.; Demirev, P. A. Anal. Chem. 2000, 72, 3739-3744. (27) Centers for Disease Control and Prevention/ National Institutes of Health. Biosafety in Microbiological and Biomedical Laboratories, 4th ed.; U.S. Government Printing Office: Washington, DC, 1999.
752
Analytical Chemistry, Vol. 73, No. 4, February 15, 2001
Table 1. Sequences of Carbohydrate Groups Used for Biocapture
Preparation of Lectin Affinity Probes. The probes were constructed using a commercially available affinity membrane (Gelman Ultrabind US450, Pall Gelman, Ann Arbor MI). The membrane was incubated for 2 h with 1 mg/mL solutions of the lectin (concanavalin A or wheat germ agglutinin; both purchased from Sigma, St. Louis, MO) in binding buffer (100 mM sodium acetate pH ) 7.4 containing 100 µM MnCl2, 100 µM MgCl2, and 100 µM CaCl2). The membrane was washed extensively (3 × 5 mL, with gentle agitation) with binding buffer to remove unbound lectin and then immersed for 1 h in 5 mL of 100 mM Tris buffer, pH ) 8.0, to block remaining reactive sites. The membrane was washed again with binding buffer and water and was stored in binding buffer at 4 °C until use. Preparation of Carbohydrate Affinity Probes. The production of the carbohydrate probes was a two-step process, consisting first of immobilizing streptavidin to the membrane surface, followed by complexing the biotinylated carbohydrates. The membranes (as described above) were first incubated with a 1 mg/mL solution of streptavidin (Sigma) in phosphate-buffered saline (PBS) for 2 h, followed by blocking using conditions identical to those used with the lectin membranes. The membranes were then washed with PBS, followed by incubation with 1 mg/mL solutions of the carbohydrate polymers (Glycotech, Rockville MD) in PBS. These polymers are composed of a polyacrylamide backbone decorated with biotin (5 mol %) and carbohydrate (20 mol %) with a molecular weight of ∼30 000. Polymers displaying five different carbohydrates (β-D-glucose, R-Dmannose, and the blood group antigens Lewis a, b, and x [Lea, Leb, and Lex]) were immoblized to separate membranes. Their sequences are shown in Table 1. Following incubation (1 h), the membranes were washed as before with PBS to remove weakly bound material and stored in fresh buffer at 4 °C until use. Microbial Sample Preparation. The previously grown bacteria were suspended in binding buffer, urine, or milk to a concentration of 5 mg of dry weight of cells/mL. This corresponded to a cell count of ∼109 cells/mL, as determined by the use of a Petroff-Hauser counting chamber.
Whole chickens were purchased at a local grocery store and kept on ice until arrival at the laboratory. The carcass of each chicken was washed with 100 mL of sterile PBS (Sigma) as previously reported.28 A portion of the rinse was reserved to determine whether the samples had any existing microbial contamination; the balance was used to prepare spiked samples, as described above. The liquid remaining in the chicken packaging (package exudate) was also collected and divided in the same manner. For sensitivity studies, dilutions of the well-mixed stock suspensions were made in the appropriate sample matrixes. Solutions of virus in culture medium (EMEM, containing 5% fetal bovine serum) were used for biocapture without any further sample preparation. Capture of Microorganisms. The biocapture surfaces were cut to the approximate size of a sample well (3 × 1.5 mm) on a 20-position Kratos (Manchester, U.K.) MALDI sample plate and affixed to the plate using double-sided tape. One microliter of bacterial or viral sample along with an equal volume of binding buffer was deposited on each capture surface on the plate. The samples were then transferred to a home-built humidity chamber and incubated at room temperature for 2 h to ensure maximal binding. After binding, samples were washed extensively (6 × 2.0 µL) with water to desalt and remove nonspecifically bound material. Bacterial samples were then treated with 1 µL of 5% aqueous trifluoroacetic acid (TFA) to lyse the cells, promoting the appearance of protein biomarkers. Viral samples were treated with 50% acetic acid solution to disrupt the virion, as previously reported29 followed by addition of 1 µL of MALDI matrix (see below). Mass Spectrometry. MALDI mass spectra were obtained on a Kratos Kompact 4 time-of-flight instrument. The mass spectrometer was equipped with pulsed ion extraction and a 337-nm N2 laser. It was operated in the positive linear mode, with the extraction voltage set to 20 kV. The MALDI matrix used was 100 mM sinapinic acid (Aldrich, Milwaukee WI) in 70:30 (v/v) acetonitrile/0.1% aqueous TFA. To ensure accurate mass calibration of bacterial samples, internal calibration with bovine insulin and equine cytochrome c was employed. To maximize sensitivity, these internal calibrants were not used in dilution studies. For viral samples, mass calibration was external using the molecular ion of bovine serum albumin deposited on an unmodified membrane. Mass spectra were typically the average of 50 laser shots taken across the sample surface. RESULTS AND DISCUSSION Lectin Capture of Bacterial Samples. In our previous work, we employed self-assembled monolayers of the alkanethiol dithoibis(succinimdylpropionate) (DSP) to immobilize lectins to a surface that was subsequently used for microbial capture. One drawback to this approach is that the relatively short linker length on the monolayer compound effectively limited the number of protein molecules that could be immobilized.30 Our initial work also relied on the use of phospholipid biomarkers for detection (28) Chen, S.; Yee, A.; Griffiths, M.; Yu, K. Y.; Wang, C.-N.; Rahn, K.; Grandis, S. A. D. J. Appl. Microbiol. 1997, 83, 314-321. (29) Thomas, J. J.; Falk, B.; Fenselau, C.; Jackman, J.; Ezzell, J. Anal. Chem. 1998, 70, 3863-3867.
of microorganisms. The number of molecules of a given protein in a microorganism is much less than the number of phospholipid molecules on the cell surface. We explored a number of alternative surfaces and binding chemistries such as activated dextrans immobilized on gold, latex microspheres, and various membranes seeking to raise our sample density to a level that allowed the consistent observation of protein biomarkers. One surface that proved to be useful was Gelman Utrabind US450, an activated polyethersulfone that covalently immobilizes proteins to its surface via their primary amines. When lectins were immobilized on this membrane, sufficient numbers of bacteria could be captured to produce a spectrum in the mass range of interest. MALDI mass spectra obtained from E.coli O157:H7 using a Con A lectin-based membrane probe and from controls are depicted in Figure 1. A suite of peaks in the 5-10 kDa range is observed with this and the other bacteria used in this study. The spectra of bacteria obtained after capture are, as expected, somewhat different from those obtained with other methods. It was previously demonstrated that MALDI mass spectra of bacteria vary based on growth condition and sample preparation.31 Therefore, any system for identification should be based on a technique other than simple pattern matching, unless growth and sample handling are both carefully controlled. Since the genome of E. coli has been fully sequenced, it is possible to tentatively identify several of the peaks based on the masses for entries in the SWISS-PROT protein database. The masses observed in this spectrum were entered into a database searching program under development26,32 which identified the microorganism as E. coli. One important observation was that the addition of 5% aqueous TFA was necessary for the detection of these proteins. A spectrum recorded of E. coli O157:H7 without this treatment is shown in Figure 1b, which shows no signal arising from the microorganisms in the region of interest. Typically, whole microorganisms are suspended in solutions containing relatively large amounts of organic solvent, such as actetonitrile, prior to mass spectrometric analysis. This is assumed to effect a gentle lysis of the cell, releasing protein biomarkers. The samples used in this work were not suspended in such solvents, so it was necessary to add this step before analysis. One might expect that the membrane would exhibit a degree of nonspecific binding that might make it undesirable for such work. This was addressed in a control where E. coli O157:H7 was captured and washed under identical conditions with an unmodified membrane, with the results shown in Figure 1c. The only signal observed is likely from detector noise, leading us to believe that nonspecific binding is not a significant factor. Concanavalin A, specific for glucosides and mannosides, has previously been shown to bind to a variety of Gram-negative microorganisms12 and as such represents a logical choice for a broad-band capture surface for such samples. We then immobilized another lectin, wheat germ agglutinin (WGA), which is specific for N-acetylglucosamine residues, and challenged it with the microorganisms employed in this study, to see if it provided (30) Brockman, A. H.; Orlando, R. Rapid Commun. Mass Spectrom. 1996, 10, 1688-1692. (31) Arnold, R. J.; Karty, J. A.; Ellington, A. D.; Reilly, J. P. Anal. Chem. 1999, 71, 1990-6. (32) http://infobacter.jhuapl.edu/servlets/MicroorganismIdentification.
Analytical Chemistry, Vol. 73, No. 4, February 15, 2001
753
Figure 2. MALDI mass spectra of Sinbis virus captured from culture media using (a) Con A lectin probe. The major peak in the spectrum corresponds to the Sinbis glycoproteins. (b) Same sample captured on an unmodified membrane. BSA is bovine serum albumin.
Figure 1. MALDI mass spectra of E. coli O157:H7 captured from binding buffer using a Con A lectin probe: (a) with pretreatment with 5% TFA; (b) without pretreatment with TFA; (c) control spectrum run on an unmodified membrane.
any advantages. All four bacteria were detectable by mass spectrometry with the WGA probes. The only species employed in this study that demonstrated appreciably higher signals with a WGA probe was S. typhimurium (data not shown). Lectin Capture of Viral Samples. Enveloped viruses such as Sinbis and VEE, which have glycoprotein moieties on the surface, should also be amenable to capture via lectin probes. These glycoproteins contain N-linked saccharides of the highmannose type33 and therefore would be expected to bind to the Con A based probes. 754 Analytical Chemistry, Vol. 73, No. 4, February 15, 2001
A MALDI mass spectrum of a sample of Sinbis virus captured directly from culture on a Con A membrane is shown in Figure 2. The virus culture solution contains a significant amount of other proteins, predominately bovine serum albumin (BSA), which is the predominant peak observed when this sample is analyzed using an unmodified membrane (Figure 2b). When the sample is analyzed after capture and washing on a Con A membrane (Figure 2a), there is still a modest amount of signal arising from the BSA. However, the most intense peak is from the unresolved Sinbis glycoproteins E1 and E2 (∼51 200 Da). One interesting observation is that only glycoprotein signal is observed, even though there is an equimolar amount of nucleocapsid protein (∼29 300 Da) present in each virion. This may be explained by the fact that the glycoproteins are located on the viral surface, as opposed to the nucleocapisd proteins, which are located inside the viral envelope. Therefore, these proteins may be more easily desorbed from captured samples. Another possibility is that the conditions used do not release this protein from the virion. Capture of Bacterial Samples with Carbohydrate-Based Probes. Lectin-based probes constructed on the Gelman membranes still did not provide the level of sensitivity desired when dealing with samples captured from complex matrixes. We (33) Feldmann, H.; Will, C.; Schikore, M.; Slenczka, W.; Klenk, H. D. Virology 1991, 182, 353-356.
Figure 4. MALDI mass spectra of E. coli O157:H7 captured from urine using a Lex biocapture probe: (a) 106 cells on probe; (b) ∼105 cells on probe. Spectra depicted in (b) normalized to intensity of (a)
Figure 3. MALDI mass spectra of samples captured using a Lex modifed carbohydrate probe (a) E. coli O157:H7, (b) E. coli K-12, and (c) mass spectrum of a Lex probe without bacteria.
therefore sought to explore an alternative system based on an immobilized carbohydrate polymer that might afford additional binding sited for microbes. Since the carbohydrate-based probes have a multitude of binding ligands attached to each polymer backbone, there are more potential binding sites available, as well as greater conformational flexibility than the lectin-based probes.34 The three microbial species used in this study exhibited a variety of binding responses when challenged with the five different carbohydrates. As an example, Figure 3 shows MALDI
mass spectra of E. coli O157:H7 and E. coli K-12 obtained using a Lex probe. S. typhimurium did not exhibit an appreciable signal when challenged with this carbohydrate. One significant concern we had with these probes was that the acidic conditions might cause the appearance of interfering signal in the m/z range of interest, since it has been demonstrated previously that acidic conditions may disrupt the avidin/biotin interaction.35 However, when a mass spectrum was taken of a carbohydrate probe that was subjected to the same sample preparation conditions, save the addition of bacteria (Figure 3c), no interfering signal was observed. An illustration of the enhanced sensitivity achieved with these surfaces is shown in Figure 4, which presents the mass spectra obtained after capture of E. coli O157:H7 from urine with a Lex carbohydrate probe. Biomarkers arising from the bacteria are clearly visible in Figure 4b, which represents ∼105 cells added to the probe, an order of magnitude lower than observed using a lectin-based probe (data not shown). One complication in using lectin-based probes for analysis of bacteria in complex samples is that the sample matrix (blood, urine, etc.) may contain appreciable amounts of glycoconjugates, which would be captured by a lectin-based probe. Chicken package exudate is likely to contain red blood cells, which are (34) Bovin, N. V. Glycoconjugate J. 1998, 15, 431-46. (35) Schriemer, D. C.; Li, L. Anal. Chem. 1996, 68, 3382-7.
Analytical Chemistry, Vol. 73, No. 4, February 15, 2001
755
Figure 5. MALDI spectra of S. typhimurium captured using a Lea modified probe from (a) chicken package exudate (background peak masses in italics), (b) control spectrum of unspiked exudate (normalized to same intensity), (c) chicken rinse, (d) control spectrum of unspiked rinse (normalized to same intensity), and (e) reference spectrum from binding buffer.
glycosylated. We challenged the carbohydrate probes with chicken rinsings and package exudates that had been spiked with S. typhimurium, an organism that is implicated in many cases of poultry related food poisoning.36 The MALDI mass spectra obtained of these spiked samples, using a Lea probe, are shown in Figure 5a and c. In both cases, a series of peaks characteristic of the bacteria S. typhimurium were readily observed, as depicted in Figure 5e, a spectrum taken after capture in buffer. Control (36) Ekperigin, H. E.; Nagaraja, K. V. Vet. Clin. North Am. Food Anim. Pract. 1998, 14, 17-29.
756 Analytical Chemistry, Vol. 73, No. 4, February 15, 2001
spectra were also obtained of the two sample matrixes without added bacteria. The control spectrum of the unspiked package exudate, shown in Figure 5b, exhibits some background, which may be due to preexisting bacterial contamination of this sample. Microscopic examination of this sample revealed the presence of bacteria, although their identity could not be confirmed from database searching of the mass spectrum. When these same samples were exposed to a Con A lectin probe, mass spectra showing signals that could be attributed to S. typhimurium were only obtainable from the carcass rinsings.
probe. The Con A probe exhibits no apparent signal arising from the microorganism. As in the chicken samples, a reasonable explanation for this observation is that all of the carbohydrate binding sites on the lectin are blocked by glycoconjugates present in the milk. The Leb probe, on the other hand, provides mass spectral peaks that are characteristic of S. typhimurium. CONCLUSIONS The membrane-based immobilization methods presented here provided higher capture density than the surface construction based on gold-thiolate linkages. This higher capture density allowed the observation of protein biomarkers from both bacteria and virus samples. The data presented here demonstrate that both lectin- and carbohydrate-based biocapture surfaces are useful analytical tools for the isolation, concentration, and cleanup of microorganisms from complex sample matrixes for MALDI analysis. Carbohydrate capture surfaces maintain their affinity from microorganisms and are less readily blocked by contaminants in the samples studied here.
Figure 6. MALDI spectra of S. typhimurium captured from milk using (a) Leb modified probe (only known peaks arising from S. typhimurium are labeled) and (b) Con A lectin probe (normalized to same intensity as (a)).
It is plausible that carbohydrates present in the sample matrix block carbohydrate binding sites on the lectin. Spectra are shown in Figure 6 of S. typhimurium-spiked milk samples that were incubated with both a Con A probe and a Leb
ACKNOWLEDGMENT The authors thank Drs. Mary Carson and David Wagner of the FDA for providing stock cultures of E. coli O157:H7 and S. typhimurium, Dr. Yeoun Jin Kim for helpful discussions about the Sinbis virus, and Ms. Danying Zhu for the microorganisms used in these studies. Financial support for this work was provided by the Defense Advanced Research Projects Agency (DARPA), the Joint Center for Food Safety and Applied Nutrition (JIFSAN), operated by the Food and Drug Administration and the University of Maryland, and the Applied Physics Lab of the Johns Hopkins University. Portions of the work described here were previously presented at the 47th and 48th Conferences of the American Society for Mass Spectrometry.
Received for review September 28, 2000. Accepted November 17, 2000. AC0011639
Analytical Chemistry, Vol. 73, No. 4, February 15, 2001
757
