Immunomagnetic Isolation of Enterohemorrhagic Escherichia coli

Jul 6, 2005 - Molecular weight database searches using the Expert Protein Analysis System proved useful for confirmation of the organism's identity. B...
0 downloads 10 Views 268KB Size
Download PDF
Anal. Chem. 2005, 77, 5258-5267

Immunomagnetic Isolation of Enterohemorrhagic Escherichia coli O157:H7 from Ground Beef and Identification by Matrix-Assisted Laser Desorption/ Ionization Time-of-Flight Mass Spectrometry and Database Searches Mariela L. Ochoa† and Peter B. Harrington*

Center for Intelligent Chemical Instrumentation, Department of Chemistry and Biochemistry, Clippinger Laboratories, Ohio University, Athens, Ohio 45701-2979

A rapid (25 min) and facile method was developed for the isolation and identification of the enterohemorrhagic Escherichia coli (serotype O157:H7) in ground beef. The isolation method employed microscopic magnetic beads coated with antibodies covalently bonded to the surface that were specific to antigens of serotype O157. This selective preconcentration step was necessary because direct matrix-assisted laser desorption/ionization (MALDI) MS analysis of bacteria was not amenable, serving to isolate the bacteria from meat components and other nontarget bacteria. The immunomagnetic separation increased the sensitivity of the method and permitted the detection of bacteria in meat. MALDI time-of-flight MS furnished bacterial mass spectra that were useful for organism identification. Molecular weight database searches using the Expert Protein Analysis System proved useful for confirmation of the organism’s identity. Bacterial biomarkers from direct MALDI analysis of pure bacterial suspensions were consistently present in bacterial suspensions of buffer/tryptic soy broth (positive controls) and meat extract samples. The detection limits were 2 × 106 cells/mL for the experimental approach used herein. Cross-reactivity studies performed on three nontarget bacterial strains revealed that the immunomagnetic beads are specific only to E. coli strain serotype O157:H7, and there is no cross-reactivity with the other relatively innocuous strains studied. The use of microorganisms as weapons of mass destruction remains a threat throughout the world. Agencies under the department of Homeland Security seek refined protocols for the rapid characterization and identification of microorganisms. Contamination with foodborne and airborne pathogens presents an important cause of intentional or unintentional injury and death. The U.S. Centers for Disease Control and Prevention (CDC) estimated that 76 million foodborne illnesses occur each year in * To whom correspondence should be addressed. Tel: 740-517-8458. Fax: 740-593-0148. E-mail: [email protected]. † Present address: Noven Pharmaceuticals, Inc., 11960 SW 144th St., Miami, FL 33186.

5258 Analytical Chemistry, Vol. 77, No. 16, August 15, 2005

the United States, resulting in 325 000 hospitalizations and 5000 deaths.1 Therefore, the differentiation of pathogenic from nonpathogenic bacteria is important for defense against biological warfare and for maintaining healthy environments. Monitoring of airborne pathogens is necessary for hygiene control in medical facilities but also as a precaution against bioterrorist releases. Liquid and solid analyses are important for securing food and drink processes against intentional and accidental contamination with virulent microbes. Escherichia coli serotype O157:H7 is a foodborne causative agent with the most severe manifestations in young children and the elderly.2 It was first identified as a foodborne pathogen in 1982 from outbreaks of gastrointestinal illness in Michigan and Oregon as a consequence of contaminated hamburgers in fast food restaurants. In subsequent years, other cases of contamination with E. coli O157:H7 were reported in several countries, including England, Canada, Scotland, and the United States. All these instances afflicted several hundred people and caused up to 20 deaths.3 E. coli O157:H7 are found in the intestinal tract of healthy cattle, and the bacteria may contaminate the meat during slaughter and processing.2,3 The transmission of this foodborne pathogen occurs mainly through the consumption of certain type of foods including raw milk, raw vegetables, fermented beef, and most importantly undercooked beef.4 The ease with which these pathogenic bacteria can be spread in the food chain and the fact that only 10-100 cells are required to cause infections in humans5,6 are important reasons for establishing rapid and sensitive methods of detection. (1) Buzby, J. C.; Frenzen, P. D.; Rasco, B. Food and Rural Economics Division, Economic Research Service, U.S. Department of Agriculture: Washington, DC, 2001. (2) Tuttle, J.; Gomez, T.; Doyle, M. P.; Wells, J. G.; Zhao, T.; Tauxe, R. V.; Griffin, P. M. Epidemiol. Infect. 1999, 122, 185-192. (3) Labbe, R. G.; Garcia, S. Guide to Foodborne Pathogens; John Wiley and Sons: New York, 2001. (4) Blackburn, C. D.; McCarthy, J. D. Int. J. Food Microbiol. 2000, 55, 285290. (5) United States Department of Agriculture. Food Safety and Inspection Service Home Page. http://www.fsis.usda.gov, accessed April 2004. (6) Tortorello, M. L.; Gendel, S. M. Food Microbiological Analysis: New Technologies; Marcel Dekker: New York, 1997. 10.1021/ac0502596 CCC: $30.25

© 2005 American Chemical Society Published on Web 07/06/2005

Current detection methods involve immunological or nucleic acid procedures.7 Immunological methods utilize antibodies that are specific to antigens of the outer cell membrane of microorganisms. Enzyme-linked immunosorbent assay has been reported as the most commonly used immunological method. Nucleic acid procedures employ short segments of single-stranded complementary nucleic acid called primers. These primers are designed to detect specific genetic sequences of DNA or RNA in microorganisms using the polymerase chain reaction. However, these methods have inherent limitations such as the necessity of preenrichment steps to amplify microorganisms or the presence of intact nucleic acid sequences in the sample to be analyzed. Yet, the major limiting factor of current methods is the time for analysis, which can range from 18 to 24 h depending on the target analyte.7 Those time-consuming methods might delay and interfere with the Pathogen Reduction/Hazard Analysis and Critical Control Point (PR/HACCP), which is a highly automated monitoring system within the food industry.5 In an effort to meet the requirements of food testing, research is being directed toward rapid methods of detection (minutes instead of hours) that could be implemented in highly automated processing environments. The analysis of microbes in food is a challenging task because foodborne pathogens may be dispersed in low concentrations among different types of food and high concentrations of harmless background microflora.8 Therefore, there is a need for a preconcentration step, which separates the target bacteria from the microflora and the complex background of food products. Affinity capture techniques such as immunomagnetic separations (IMS) have been proposed as attractive methods to isolate and concentrate the analyte of interest from complex matrixes.9-11 In this paper, we used IMS to reduce the effect of extraneous components present in ground beef, which may contribute to signal suppression and lack of reproducibility. Microsized magnetic beads coated with adsorbed and affinity purified antibodies against E. coli serotype O157 covalently bound to the surface were used for the rapid and selective isolation of bacteria at the genus, species, and strain levels. After the antibodies bind to the target bacterial antigens, the cells-beads complex is separated from the food background by using a magnetic plate. Matrix-assisted laser desorption/ionization (MALDI) time-offlight mass spectrometry (TOF-MS) has emerged as a promising technique for the rapid, sensitive, and accurate detection of bacteria.12-18 The development of MALDI and electrospray ionization as well as improvements in TOF-MS have made possible the (7) Ellis, D. I.; Goodacre, R. Trends Food Sci. Technol. 2001, 12, 414-424. (8) Jaykus, L. A. ASM News 2003, 69, 341-347. (9) Madonna, A. J.; Van Cuyk, S.; Voorhees, K. J. Rapid Commun. Mass Spectrom. 2003, 17, 257-263. (10) Madonna, A. J.; Basile, F.; Furlong, E.; Voorhees, K. J. Rapid Commun. Mass Spectrom. 2001, 15, 1068-1074. (11) Liu, Y. C.; Li, Y. B. J. Microbiol. Methods 2002, 51, 369-377. (12) Anhalt, J. P.; Fenselau, C. Anal. Chem. 1975, 47, 219-225. (13) 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. (14) Wang, Z. P.; Russon, L.; Li, L.; Roser, D. C.; Long, S. R. Rapid Commun. Mass Spectrom. 1998, 12, 456-464. (15) Lay, J. O. TrAC-Trends Anal. Chem. 2000, 19, 507-516. (16) Holland, R. D.; Rafii, F.; Heinze, T. M.; Sutherland, J. B.; Voorhees, K. J.; Lay, J. O. Rapid Commun. Mass Spectrom. 2000, 14, 911-917. (17) Lay, J. O. Mass Spectrom. Rev. 2001, 20, 172-194. (18) Vater, J.; Gao, X. W.; Hitzeroth, G.; Wilde, C.; Franke, P. Comb. Chem. High Throughput Screening 2003, 6, 557-567.

Table 1. Summary of E. coli Strains Studied, Nutrients, and Physical Conditions Employed As Recommended by the ATCCa microorganism

ATCC no.

culture broth

culture agar

E. coli E. coli O157:H7 E. coli O55:K59 (B5):H6 E. coli

15223 43895 12014 14948

NB TSB NB NB

NA TSA NA NA

biosafety level 1 2 2 1

a Aerobic atmospheric conditions were employed for all strains of bacteria and 37 °C as incubation temperature.

ionization and analysis of biological molecules with high resolution and mass accuracy.19 MALDI TOF-MS furnishes protein biomarkers from intact bacterial cells, which facilitates the speed of the identification process. The MALDI spectral fingerprints provide identification of microorganisms and differentiation between species and strains of bacteria. Recently, the feasibility of combining IMS with MALDI TOFMS for the rapid detection of Salmonella choleraesuis was reported.10 Madonna and co-workers were able to isolate and detect this species from river water, human urine, and chicken blood. However, the isolation of the foodborne pathogen E. coli O157:H7 in complex matrixes such as ground beef followed by its detection with MALDI-MS has not been reported. This project aimed at investigating the rapid isolation and identification of this strain of E. coli from ground beef using IMS combined with MALDI TOF-MS. Because the genome of the enterohemorrhagic E. coli O157: H7 has been completely sequenced,20 it was possible to use protein molecular weight database searches for bacterial identification. The proton-bound proteins from the MALDI spectra were matched to those proteins predicted by the organism’s proteome to tentatively identify many protein biomarkers. Despite the fact that this organism proteome predicts ∼2900 proteins in the mass range 3-20 kDa,21 the small fraction of these that was observed in the MALDI mass spectra could be used to create fingerprint databases for the identification of E. coli O157:H7. EXPERIMENTAL SECTION Bacteria and Methods. Freeze-dried stock cultures of bacteria were purchased from the American Type Culture Collection (ATCC, Manassas, VA). Table 1 provides a detailed summary of the E. coli strains that were studied and conditions used for bacterial culture and growth. (CAUTION: E. coli serotypes O157: H7 and O55:K9 are Biosafety Level 2 organisms and should be handled with extreme care.) All microbiological procedures were carried out in a certified Biosafety Level 2 cabinet using proper sterilization procedures in compliance with Ohio University’s (19) Whitehouse, C. M.; Dreyer, R. N.; Yamashita, M.; Fenn, J. B. Anal. Chem. 1985, 57, 675-679. (20) National Center for Biotechnology Information (NCBI) search engine and retrieval system (Entrez). http://www.ncbi.nlm.nih.gov/genomes/static/ eub_g.html, accessed October 2003. (21) The ExPASy (Expert Protein Analysis System) proteomics server of the Swiss Institute of Bioinformatics (SIB). http://us.expasy.org/srs/, accessed October 2003.

Analytical Chemistry, Vol. 77, No. 16, August 15, 2005

5259

Figure 1. Flowchart for the IMS procedure to isolate the bacterial cells from meat extracts. Positive control bacterial suspensions were prepared by removing 108 cells from agar plates and placing them in a microcentrifuge tube. The bacterial cells were diluted with 900 µL of PBS and 100 µL of TSB.

Institutional Biosafety Committee22 and the Centers for Disease Control and Prevention Office of Health and Safety.23 Nutrient broth (NB, lot 3280588) was purchased from Difco (Sparks, MD). Nutrient agar (NA, lot 98H0898), tryptic soy agar (TSA, lot 053K0101), and tryptic soy broth (TSB, lot 101K0020), were purchased from Sigma-Aldrich (St. Louis, MO). Cells were rehydrated by dissolving the bacteria pellet in 5 mL of tryptic soy or nutrient broths, which were previously autoclaved at 121 °C for 15 min. The bacteria were allowed to rehydrate and grow in the infusion for 24 h at 37 °C. Subsequently, the bacteria were streaked in Petri dishes containing sterilized agar. The agar plates were transferred to an oven maintained at 37 °C for incubation of the bacterial cultures. After 24 h, the plates were removed from incubation and the bacteria were ready for analysis. Chemicals. Sinapinic acid (SA, lot 062K3722), ferulic acid (FA, lot 09012BA), R-cyano-4-hydroxycinnamic acid (CHCA, lot 122K3768), equine cytochrome c (lot 102K7053), equine myoglobin (lot 122K7057), and bovine trypsinogen (lot 81K7680) were purchased from Sigma Aldrich Chemical Co. (St. Louis, MO). Acetonitrile, HPLC grade (lot 013031) was purchased from Fisher Scientific (Fairlawn, NJ). Ionate trifluoroacetic acid, HPLC grade (TFA, lot BK45823) was purchased from Pierce Biotechnology, Inc. (Rockford, IL). (22) Ohio University Biosafety Program Manual, June 1, 1999; p 125. http:// www.ohiou.edu/ehs/docs/Biosafety_Manual.pdf, accessed May 2005. (23) Biosafety in Microbiological and Biomedical Laboratories (BMBL), 4th ed.; U.S. Government Printing Office, 1999; p 265. http://www.cdc.gov/od/ohs/ pdffiles/4th%20 µBBL.pdf, accessed May 2005.

5260 Analytical Chemistry, Vol. 77, No. 16, August 15, 2005

Determination of Bacterial Numbers. To quantify the amount of bacteria in each sample, the standard plate count technique was employed.24 Phosphate-buffered saline (PBS, 10 mM, pH 7.4, 0.05% Tween 20, lot 043K8205 Sigma) was used to dilute bacterial samples. A handheld electronic colony counter (Bel-Art Products, Cole-Parmer Instrument Co., Vernon Hills, IL) was used to count bacterial colonies. Sample Preparation for MALDI TOF-MS Analysis. For direct MALDI analysis, suspensions of pure bacterial cultures were used. A few colonies (108 cells) were removed from the agar plate by using a previously sterilized inoculating loop and were placed in a microcentrifuge tube. Bacterial cells were suspended in a solution containing 1 mL of 40:60 acetonitrile/TFA (0.1%, v/v). The bacterial suspensions were vortex mixed at intermediate speed for 2 min until a cloudy homogeneous solution was observed. Bacteria-spiked meat samples were prepared following the protocol proposed by the U.S. Department of Agriculture, FSIS.5 Ground beef was bought at a local supermarket. For each meat sample, 25 g of ground beef was weighed and placed in a filtered homogenizer bag (3M Microbiology, St. Paul, MN). The meat was spiked with 100 µL of bacterial infusion (refer to the protocol for reviving bacteria early in the Experimental Section), which resulted in a bacterial concentration of 2 × 104 cells. Bacteriaspiked samples were enriched with 225 mL of TSB and incubated at 37 °C for 20 h. Negative controls were also prepared by (24) Harley, J. P.; Prescott, L. M. Laboratory Exercises in Microbiology, 5th ed.; McGraw-Hill Science: New York, 1996.

Figure 2. MALDI positive ion mass spectra of whole E. coli cells showing mass range 3-15 kDa. Analysis was conducted using bacterial suspensions from pure cultures. (a) E. coli O157:H7 ATCC 43895; (b) E. coli ATCC 15223; (c) E. coli ATCC 14948; (d) E. coli O55 ATCC 12014. Each spectrum is plotted with the relative intensity as a function of m/z. The labels above each peak represent the ions’ molecular weight (top) and absolute intensity (bottom), which are provided to facilitate interspectral comparisons.

incubating the same amount of meat used for bacterial samples and using the same experimental conditions with the exception that no bacteria were added to the meat. Positive controls (bacteria suspended in buffer/TSB solution), meat extract suspensions, and negative controls were employed for the IMS procedure. Positive control samples were prepared by dissolving 108 cells in 100 µL of TSB and 900 µL of PBS. Meat extract bacterial suspensions were prepared as follows. A 1-mL aliquot of meat filtrate was centrifuged at a relative centrifugal force of 2000 × g for 10 s, which ensures that heavy particulates and debris from meat settle at the bottom of the centrifuge tube, and the bacterial cells are suspended in the liquid.25 The pellet was discarded, and the supernatant was used for analysis. A 200µL aliquot of the supernatant was placed in a clean microcentrifuge tube and dissolved in 800 µL of PBS. Three matrixes were initially examined to assess the best experimental conditions: FA, CHCA, and SA. SA furnished the best signal-to-noise ratio and highest signal intensity. SA matrix was prepared at a concentration of 10 mg mL-1. All solutions were homogenized on a vortex mixer for 2 min before analysis. For MALDI analysis, a 2-µL aliquot of bacterial suspension was placed on the MALDI sample probe, air-dried, and codeposited with 1 µL of SA. The samples were air-dried before analysis. The calibration standards consisted of a protein mixture containing cytochrome c (MW 12 360), myoglobin (MW 16 951), and trypsinogen (MW 23 980). Working standards of concentration 10 pmol mL-1 were prepared from stock standards. A solution containing 3 µL of cytochrome c, 5 µL of myoglobin, and 5 µL of trypsinogen working standard solutions was mixed with 13 µL of SA matrix solution. This solution was vortex mixed for 30 s, and 1 µL of standard/matrix mixture was placed onto the MALDI sample probe. Immunomagnetic Separations. Dynabeads anti-E. coli O157 (lot F25300, Dynal Biotech, Brown Deer, WI) were supplied in a (25) Stannard, C. J.; Wood, J. M. J. Appl. Bacteriol. 1983, 55, 429-438.

Figure 3. Evaluation of the total variance among the MALDI mass spectra of four E. coli strains by PCA. PCA score plot accounts for 73% of the total variance for the first two principal components (PC). For the axis labels of the score plot, the values in parentheses represent the variance accounted for by PC 1 (abscissa) and PC 2 (ordinate) expressed in percentage followed by the absolute eigenvalue. The 95% confidence ellipse is given around each cluster of scores. Each letter represents principal component scores obtained from the MALDI spectra. (A) ATCC 12014; (B) ATCC 14948; (C) 15223; (D) ATCC 43895.

suspension of PBS (pH 7.4) with 0.1% bovine serum albumin and 0.02% sodium azide. Figure 1 gives a flowchart showing the steps involved in the immunomagnetic separation procedure. This approach was simple, rapid (25 min), required very small amounts of sample, and is amenable to automation. Thirty microliters of magnetic beads was added to the bacterial suspensions, and the microcentrifuge tubes containing these suspensions were loaded into the magnetic rack. Bacterial suspensions and beads were incubated at room temperature for 20 min in a vortex mixer set at the lowest speed to achieve a very slow mixing motion. While incubating, 360° manual rotations facilitated a better interaction of the bacterial sample with the magnetic beads. At the end of the incubation period, the magnetic plate was placed in the rack to concentrate the beads-bacteria complex on a side of the microcentrifuge tubes. Carefully, so as not to disturb the beads, the supernatant was removed by aspirating the liquid with a 1-mL micropipet. The supernatant (waste) was saved for MALDI analysis to evaluate the recovery of the method and to assess any cross-reactivity with nontarget bacteria. The beadsbacteria complex was resuspended in 50 µL of high-purity water (18.2 MΩ‚cm resistivity, Millipore Milli-Q Plus, Catalog No. CPMQ004R1, Millipore Corp., Billerica, MA) and mixed to obtain a homogeneous solution. The IMS procedure for negative controls was carried out as stated previously. MALDI TOF-MS Analysis. All mass spectra were furnished by a M@LDI LR by Micromass Ltd. (Hertsfordshire, U.K.) timeof-flight mass spectrometer, which is equipped with a nitrogen UV laser (337-nm wavelength) and a time-lag focusing (delayedextraction) ion source. The instrument was operated in positive polarity and linear mode. The following parameters were set in the mass spectrometer for MALDI spectra acquisition: accelerating voltage 15 kV, pulse voltage 1.6 kV, laser firing rate 5 Hz, 10 laser shots per spectrum, time-lag focusing 499 ns, matrix Analytical Chemistry, Vol. 77, No. 16, August 15, 2005

5261

Figure 4. MALDI positive ion mass spectra of E. coli O157:H7 from three different samples showing the mass range 3-18 kDa: (a) direct MALDI of whole cells; (b) MALDI mass spectrum of beads-bacteria complex after IMS of a positive control sample; (c) MALDI mass spectrum of beads-bacteria complex after IMS of a meat extract sample; (d) MALDI mass spectrum of a negative control sample after the IMS. Each spectrum is plotted with the relative intensity as a function of m/z. The labels above each peak represent the ions’ molecular weight (top) and absolute intensity (bottom), which are provided to facilitate interspectral comparisons.

suppression delay 2 kDa, microchannel plate detector 1.9 kV, and sample period 0.5 ns. For each spectral acquisition, the laser was fired at different spots in the sample well and each mass spectrum represents an average of ∼60 scans. The spectra were processed using Masslynx 4.0 Global Mass Informatics Maldi V4.0 SP1 Micromass. The raw spectra were smoothed, background subtracted, and externally calibrated using the three-point calibration method with the molecular ions for myoglobin, trypsinogen, and cytochrome c. Suspensions of pure bacterial cultures, control samples, and meat extracts were analyzed with seven replicates and using a random block design. A search was conducted by protein molecular weight and organism using the Sequence Retrieval System (SRS) module in the SwissProt/TrEMBL database, which is the Expert Protein Analysis System (ExPASy) of the Swiss Bioinformatics Institute.21 Protein molecular weights found in the MALDI mass spectra were tentatively matched to biomarker peaks contained in the SwissProt results query form for identification of the bacterial main biomarkers. Note that these assignments are putative and could be erroneous due to biological processes occurring within the cell. The proteins identified with this method would correspond to the reported values when the proteins in the database are in the exact form and with the exact level of processing as the proteins in the cell. RESULTS AND DISCUSSION For direct MALDI analysis, the bacterial suspensions were prepared by suspending the bacterial cells in TFA and acetonitrile. The thin cell walls of Gram-negative bacteria and, in this case, E. coli, rupture upon treatment with organic solvents, releasing their 5262 Analytical Chemistry, Vol. 77, No. 16, August 15, 2005

proteins. Pure bacterial suspensions and bacteria recovered from magnetic beads will form a solid solution with the matrix upon mixing and air-drying. When the matrix-analyte crystals are irradiated with the UV laser, the excess energy absorbed by the matrix causes ionization of analyte molecules, which are then analyzed by the mass spectrometer.26 Determination of Statistical Differences of Four E. coli Strains by Direct MALDI Analysis. Whole-cell suspensions of four E. coli strains were analyzed in SA as a matrix: ATCC 12014 (serotype O55), 14948, 15223, and 43895 (serotype O157:H7). Their MALDI mass spectra are given in Figure 2, which by inspection shows that the four strains can be distinguished from one another by the presence of different biomarkers. To ascertain whether the strains could be distinguished by the presence of distinct biomarkers, using principal component analysis (PCA)27-31 to evaluate differences in the MALDI mass spectra. The results showed that these four strains could be differentiated by MALDI analysis because clusters pertaining to each E. coli strain resulted in Figure 3. The PCA score plot accounted for 73% of the total (26) Cotter, R. J. Time-of-Flight Mass Spectrometry: Instrumentation and Applications in Biological Research; American Chemical Society: Washington, DC, 1997. (27) Jolliffe, I. T. Principal Components Analysis; Springer-Verlag: New York, 1986. (28) Wold, S.; Esbensen, K.; Geladi, P. Chemom. Intell. Lab. Syst. 1987, 2, 3752. (29) Massart, D. L.; Vandeginste, B. G. M.; Deming, S. N.; Michotte, Y.; Kaufman, L. In Chemometrics: A Textbook; Kaufman, L., Ed.; Elsevier: New York, 1988; Vol. 2, pp 339-370. (30) Henrion, R. Chemom. Intell. Lab. Syst. 1994, 25, 1-23. (31) Malinowski, E. R. Factor Analysis in Chemistry, 2nd ed.; John Wiley & Sons: New York, 1991.

To investigate whether there was a trend in the ion abundance distribution for the major protein peaks found in these three samples, Figure 5 was generated. Overall, the distribution of ion abundance with molecular mass of proteins found in buffer/TSB samples is comparable to that from direct analysis. This trend might be expected because the bacterial cells were suspended in similar environments (buffer and organic solvents). In fewer instances, bacterial proteins from meat extracts showed comparable or slightly higher ion abundances than those for buffer/ TSB and direct analysis samples. For meat samples, the most abundant biomarker was found in the mass range 7265-7273 Da. The fact that bacterial proteins detected in meat gave lower ion abundances and more peaks than the other two types of samples might be related to meat nutrients triggering the production of other proteins. A more diverse set of proteins competing for charge during the MALDI ionization process results in lower ion abundance because the reservoir of charge furnished by the matrix is distributed among a larger number of proteins. Two-way ANOVA with interaction was performed for 13 peaks that were consistent in the MALDI mass spectra of the two samples in which E. coli O157:H7 cells were analyzed with the IMS procedure (positive controls and meat extracts). Before ANOVA was applied, the MALDI mass spectra of the two samples were processed in the following manner. Approximately 60 spectra/sample spot were read for each replicate. The spectra were baseline corrected, mass aligned, and interpolated to the same mass scale. The average spectrum was calculated from the 60 spectra/sample. Each average spectrum was normalized to unit vector length. All spectra were mass aligned to the average spectrum for the data set of the two samples and 14 spectra (7 replicates/sample). The m/z values were obtained by finding the maximum ion intensity in a 20-Da window around the peak list value for each spectrum. Two-way ANOVA with interaction was applied to these m/z values. Table 2 gives the summary of two-way ANOVA. The results revealed that the type of sample analyzed with IMS (positive controls and meat extracts) is not significant, and therefore, the same peaks were detected in these samples. The different biomarker peaks did differ, which would be expected because the peaks were all resolved from one another. The interaction term was significant, which indicates that some of the peak m/z values may differ significantly based on the type of sample (e.g., buffer or meat extract). These findings are important for devising intelligent algorithms that can furnish the molecular masses of several biomarkers pertaining to a bacterial species, which can be used for the detection of microbes in food products.

Figure 5. Ion abundance distribution for main biomarkers of E. coli O157:H7 from direct analysis, positive control, and meat extract samples in the MALDI spectra. Peaks shown have been matched to ExPASy biomarkers. Confidence intervals at 95% were calculated based on seven replicate analyses.

variance for the first two principal components. The 95% confidence ellipses obtained from the Hotelling T2 distribution are given for the scores of each strain. MALDI Analysis of E. coli O157:H7 from Pure Cultures, Positive Controls, and Meat Extracts. Microscopic beads coated with antibodies specific for the E. coli serotype O157 were used to isolate and concentrate E. coli O157:H7 from buffer/TSB and meat extract solutions. MALDI analyses were performed on three different samples containing the enterohemorrhagic E. coli O157:H7, and the MALDI mass spectra are given in Figure 4. The bacteria were recovered using the immunomagnetic step from suspensions of this strain in buffer/TSB (positive control, b) and meat extracts (c). The results were compared to direct MALDI analysis (a), whereby the immunomagnetic step was omitted. In addition, the MALDI mass spectrum of a negative control sample (d) was included for reference. This sample was prepared by following the same protocol as the meat extract samples only no bacteria were added. Note the presence of more biomarkers in the MALDI mass spectrum of meat extracts when compared with those for positive controls and direct analysis. Our hypothesis is that meat induces the production of more proteins by the bacteria than the buffer/ TSB solution in the mass range 3-20 kDa. The negative control suggests that the magnetic beads and the meat do not produce a background signal that could interfere with the MALDI mass spectra of bacteria.

Table 2. Summary of Results for Two-Way ANOVA with Interactiona Performed on the m/z for Main Biomarkers Found in E. coli O157:H7 Analyzed with IMS of Positive Control Samples, and Meat Extracts

a

ANOVA

source of variation

SS

df

biomarker peaks sample type interaction residual

108

6.91 × 6.34 × 10-1 1.08 × 102 3.32 × 102

12 1 12 156

total

6.91 × 108

181

MS

F

×107

2.71 × 0.298 4.21

5.76 0.634 8.96 2.13

107

P-value

F crit

0 0.586 9.98 × 10-6

1.81 3.90 3.90

ANOVA was performed at 95% confidence level and seven replicate analyses.

Analytical Chemistry, Vol. 77, No. 16, August 15, 2005

5263

Table 3. Biomarkers Found in the MALDI Mass Spectra of E. coli O157:H7 from IMS of Meat Extract Samples and Direct Analysis of Pure Cultures Tentatively Matching SwissProt/TrEMBL Database Proteins M (match)

(MH+)a meat extracts

direct analysis

3205(2) 3635(2) 4160(3) 4766(4) 4868(3) 4971(3)

ND 3639(2) 4166(2) 4772(2) 4872(3) 4976(2)

5592(3)

ND

6085(3)

ND

6407(3) 6852(3)

6410(1) 6858(2)

7139(4)

7152(2)

7267(4)

7272(3)

7327(3)

7332(2)

7471(4) 7697(4)

7479(3) 7709(3)

7864(4)

7866(2)

8316(5)

8323(3)

9179(3)

ND

9523(5)

9530(4)

9727(5)

9735(4)

9932(5)

9947(5)

10378(5)

ND

10661(7)

10662(13)

11174(7)

ND

12159(7)

12178(4)

ExPASy b b b *4890c 4864 *5096 *5098 5591 5587 6094 6411 6855 6856 7158 7159 7265 7271 7272 7273 7332 7334 *7612 7698 7716 7855 7871 *7992 *7995 *8000 *8001 8311 8315 8323 8325 8327 8328 9179 9190 9190 9525 9533 9535 9732 9742 9743 *9857 *9859 *9856 9938 9955 10387 *10495 10651 10675 11185 *11306 *11293 12164 12165 12179

protein description

pI

Z4042 protein putative RNA 30S ribosomal protein S22 Pyr BI operon leader peptide Z0870 protein, membrane component, involved in transport Z0393 protein Rep protein E1, component of extrachromosomal circular DNA, involved in DNA replication initiation 50S ribosomal protein L30 protein ycaR carbon storage regulator 50S ribosomal protein L35 putative DNA-binding protein putative tail-fiber protein cold-shock protein cspC (cytoplasmic) cold-shock protein cspA (cytoplasmic) 50S ribosomal protein L29 cold-shock protein cspE (cytoplasmic) putative lipoprotein Rzl PTS system, glucitol/sorbitol- specific IIB component putative holin protein of prophage CP-9330 sorbose-permease PTS system IIID putative galactosamine-6- phosphate isomerase 50S ribosomal protein L31 Holin protein Z3731 protein putative endolysin stability putative hemolysin expression (encoded on plasmid pO157) putative regulatory of cell division precursor of major outer membrane lipoprotein protein yjbJ Z0726 protein Z2618 protein YjgJ protein 30S ribosomal protein S16 toxin 2B-subunit (fragment) Z3165 protein putative regulatory protein DNA-binding protein HU-R hydrogenase isoenzyme formation protein hypC excisionase Shiga toxin 1B subunit Shiga toxin type II variant B subunit partial putative phage tail protein encoded by prophage CP-933R Z0404 protein 9932(5) putative transcriptional regulatory Z1625 protein 10 kDa chaperonin (cytoplasmic) sugar fermentation stimulation protein B integration host factor β-subunit. Specific DNA-binding protein functions in genetic recombination, transcriptional and translational control 50S ribosomal subunit protein L25 50S ribosomal protein L24 Ela B protein Z3025 protein (N-acetyltransferase) 50S ribosomal protein L7/L12 (binding site for several factors involved in protein synthesis, essential for accurate translation) Z3516 protein ATP-dependent Clp protease adaptor protein clpS.

6.04 9.86 11.04 11.44 5.55 7.94 10.87

5264 Analytical Chemistry, Vol. 77, No. 16, August 15, 2005

10.96 4.91 8.16 11.78 9.94 5.98 6.82 5.57 9.98 8.06 7.74 11.36 9.16 9.85 6.25 9.46 9.40 7.98 9.58 6.25 9.46 9.45 9.30 5.44 4.29 9.10 5.63 10.54 8.64 4.85 9.05 9.57 4.17 9.20 7.77 7.74 10.71 7.85 8.47 9.34 5.15 9.89 9.34 9.60 10.21 5.35 6.63 4.60 9.70 4.94

Table 3. (Continued) M (match)

(MH+)a meat extracts

direct analysis

12633(8)

12651(7)

12648 12653 12663 12664

13975(7)

ND

13993

15382(10)

15400(7)

ExPASy

13966 13968 15408 *15527 *15534

protein description

pI

PTS system fructose-like IIB component 2 (membrane component) protein yfiA, belongs to the σ 54 modulation protein family putative head-tail adaptor putative IS encoded protein within prophage CP-9330, involved in DNA transposition holo-[acyl-carrier-protein] synthase (transferase activity, involved in fatty acid biosynthesis and lipid synthesis, (cytoplasmic) chemotaxis protein cheY (transmission of sensory signals, cytoplasmic) flagellar basal-body rod protein flgC DNA-binding protein H-NS biopolymer transport exbD protein (inner membrane protein) minor tail protein

5.36 6.19 10.02 9.38 8.91 4.89 5.23 5.44 4.70 5.06

a Experimental values represent the mean of seven replicate analyses. The precision of the experimentally determined molecular weight is reported as confidence intervals at 95%. b No match was found. c *Determined as posttranslational modified N-terminal Met loss.

Figure 6. Recovery studies with E. coli strains ATCC 12014, 14948, and 15223. MALDI mass spectra of beads-bacteria complexes and waste solutions after the IMS performed on meat extract samples containing these E. coli strains. Each spectrum is plotted with the relative intensity as a function of m/z. The labels above each peak represent the ions’ molecular weight (top) and absolute intensity (bottom), which are provided to facilitate interspectral comparisons. The specificity of the magnetic beads to E. coli O157:H7 was evaluated. The magnetic beads did not cross-react with relatively innocuous E. coli strains as can be observed from the MALDI mass spectra of the beads-bacteria complexes.

Identification of Protein Biomarkers in E. coli O157:H7 from Meat Samples. Twenty-eight peaks were studied in the mass range 3-20 kDa corresponding to E. coli O157:H7 recovered from bacteria-spiked meat extracts. These peaks were compared to biomarkers obtained by direct MALDI analysis of whole-cell pure bacterial suspensions (Figure 4a,c). The position of these

peaks is highly reproducible if the same experimental approach is followed. The mass assignments for these protein peaks gave a measured 95% confidence interval of 2-13 Da on the basis of seven replicate analyses. The experimental conditions used in this project permitted the observation of protein biomarkers from whole-cell bacterial samples in the mass range from 3 to 20 kDa. Analytical Chemistry, Vol. 77, No. 16, August 15, 2005

5265

In this region, peaks represent singly protonated protein molecules, while other components of the cell membrane such as lipooligosaccharides and peptidoglycans, are typically found below 3500 Da.32 No mass spectral peaks were observed above 20 kDa. Results of the database search using the experimental molecular weights as search parameters are summarized in Table 3, which gives tentative matches for the MALDI protein peaks detected in meat and in pure cultures, their description, and pI values. Several possible matches resulted in hypothetical proteins, which have been predicted in the organism’s genome, but there is not experimental evidence that they are expressed in vivo.21 The search was also conducted with posttranslational modification N-terminal methionine cleavage, indicated as starred matches in Table 3. It has been reported that the cleavage of the N-terminal Met residue is the most common posttranslational modification for prokaryotes, and it has been estimated that up to 50% of E. coli proteins undergo this transformation.33 Note that, there is a 131-Da mass difference between these posttranslational modified proteins and the experimental value.34 For biomarkers at m/z 3205(2), 3635(2), and 4160(3) for which no matches were found, it is speculated that these proteins may have originated outside of the cell or they may be secondary metabolites. These proteins are not accounted for by the genome.34 The database proteins matching the experimentally observed masses correspond mainly to proteins from the inside of the bacterial cells with the exception of biomarker at 8316 ( 5 Da, which matched a precursor of a major outer membrane lipoprotein (8323 Da). Nine proteins were matched by mass to ribosomal proteins in the 50S and 30S subunits. These proteins have high abundance because up to 45% of the mass of E. coli cells corresponds to ribosomes and up to 21% of the cell’s protein content is ribosomal.35 All of the ribosomal proteins are very basic (i.e., pI greater than 9) with the exception of the 12 164-Da 50S ribosomal protein L7/L12 (pI 4.60 in Table 3). Basic proteins are more favorable for protonation during the MALDI process.36 The description of the possible matches is also given in this table as well as their roles and locations in the cell. Although all proteins play an essential role in cellular function, some of the ones identified have well-known roles such as regulation of cell division, ATP and sugar metabolism, transcription initiation, transferase and endonuclease activities, protein transport, and DNA transposition, as well as regulation of DNA transcription and translation processes.21 In addition, subunits of the protein Shigalike toxin, which is responsible for the pathogenic properties of the enterohemorrhagic E. coli, were also identified. Cross-Reactivity Studies. To investigate the extent of crossreactivity in this method, we analyzed three nontarget bacterial strains as well as mixtures of them. E. coli ATCC 12014 (serotype O55), 14948, and 15223 were analyzed separately using immunomagnetic beads anti-E. coli O157. Figure 6 gives the MALDI mass spectra of beads-bacteria complexes and waste solutions, which were retained to investigate the recovery of the bacteria from the (32) Gibson, B. W.; Engstrom, J. J.; John, C. M.; Hines, W.; Falick, A. M. J. Am. Soc. Mass Spectrom 1997, 8, 645-658. (33) Hirel, P. H.; Schmitter, J. M.; Dessen, P.; Fayat, G.; Blanquet, S. Proc. Natl. Acad. Sci. U.S.A. 1989, 86, 8247-8251. (34) Leenders, F.; Stein, T. H.; Kablitz, B.; Franke, P.; Vater, J. Rapid Commun. Mass Spectrom. 1999, 13, 943-949. (35) Arnold, R. J.; Reilly, J. P. Anal. Biochem. 1999, 269, 105-112. (36) Ryzhov, V.; Fenselau, C. Anal. Chem. 2001, 73, 746-750.

5266 Analytical Chemistry, Vol. 77, No. 16, August 15, 2005

Figure 7. Cross-reactivity studies with a mixture of innocuous strains and the pathogenic E. coli O157:H7. MALDI positive ion mass spectra showing the mass range 3-19 kDa of (a) direct MALDI analysis of E. coli ATCC 43895 and (b) beads-bacteria complex after IMS of a meat extract sample containing E. coli strains ATCC 14948, 15223, and 43895. Each spectrum is plotted with the relative intensity as a function of m/z. The labels above each peak represent the ions’ molecular weight (top) and absolute intensity (bottom), which are provided to facilitate interspectral comparisons. The consistency of protein molecular weight peaks associated with these spectra reveals that contamination with relatively innocuous strains does not affect the binding of the enterohemorrhagic E. coli O157:H7 to the magnetic beads.

magnetic beads and their specificity to E. coli O157:H7. The IMS and MALDI analyses were performed on meat extract samples containing nontarget bacterial cells using the same procedure. The bacterial biomarkers were successfully identified in the waste solution. The MALDI mass spectra of the beads-bacteria complex showed no significant signal above the baseline noise. To evaluate the performance of the IMS, the MALDI spectra of the waste solution after IMS was analyzed and compared to the MALDI spectra of pure bacterial cultures without IMS. The same bacterial biomarkers were successfully identified in both instances. Because clinical, environmental, or food samples often contain more than one bacterial strain, it was investigated whether a mixture of target E. coli O157:H7 and nontarget E. coli strains would affect the performance of the magnetic beads toward E. coli O157:H7. Meat samples containing the enterohemorrhagic E. coli spiked with strains ATCC 14948 and 15223 were analyzed, and the MALDI spectra are given in Figure 7. The top spectrum (a) pertains to the direct MALDI of E. coli ATCC 43895 and the bottom spectrum (b) corresponds to the beads-bacteria complex after the IMS step of a meat extract sample. The consistency of the molecular weight peaks associated with both spectra reveals that contamination with relatively innocuous strains does not affect the binding of the pathogenic E. coli to the magnetic beads. The waste solution from the IMS was investigated and its spectrum compared to that for direct analysis of E. coli strains 14948 and 15223. Figure 8 gives the direct MALDI spectra of E. coli 15223 in (a), the waste solution in (b), and direct MALDI of E. coli 14948 in (c). The isolation of the enterohemorrhagic E. coli was successfully accomplished even when high concentrations of other innocuous strains were concomitantly present in meat samples because some of the biomarker peaks from the direct analysis of these bacteria appeared on the MALDI spectra of the waste solution.

Figure 8. Cross-reactivity studies with a mixture of innocuous strains and the pathogenic E. coli O157:H7. MALDI positive ion mass spectra of (a) direct MALDI analysis of E. coli ATCC 15223, (b) waste solution after the IMS of a meat extract sample of E. coli O157:H7 spiked with strains ATCC 14948 and 15223, and (c) direct MALDI analysis of E. coli 14948. Each spectrum is plotted with the relative intensity as a function of m/z. The labels above each peak represent the ions’ molecular weight (top) and absolute intensity (bottom), which are provided to facilitate interspectral comparisons. The waste solution reveals a MALDI pattern that is similar to that for strains 14948 and 15223. Peaks that are consistently present in the waste solution are indicated with arrows.

CONCLUSIONS These studies demonstrated the feasibility of using IMS and MALDI TOF-MS for the rapid specific isolation and detection of the foodborne pathogen E. coli O157:H7 in complex matrixes such as ground beef. The immunomagnetic approach and MALDI sample preparation provide a total analysis time of 25 min. In addition, both IMS and MALDI are amenable for automation in cases where high sample throughputs are required. The concentration of bacterial suspensions used for the proposed method was 2 × 108 colony-forming units (cfu)/mL. Studies performed on

detection limits revealed that, for our experimental approach, 2 × 106 cfu/mL could still be used to reproducibly provide a signalto-noise ratio greater than 3. These detection limits were comparable with those reported by Madonna et al.10 Preliminary results and the ease with which bacterial cells could be isolated and detected in ground beef make this method worthy of considering as an alternative detection method. We believe that this approach merits consideration as a powerful analytical tool that could be implemented in the food industry. However, the successful implementation of this method depends on advances in proteomic research and the availability of the target organism’s genome sequence because database searches are a key component for the identification of protein biomarkers. Because with this method the limiting step is the incubation period, which was 20 h, future studies shall focus on optimizing the sample preparation step so that the incubation time for meat samples could be reduced. This optimization in turn would lead to much lower detection limits and faster analysis time. The applicability of this method to the analysis of other food products shall also be investigated. ACKNOWLEDGMENT This work was presented in part at the 55th Pittsburgh Conference and Exposition on Analytical Chemistry in Chicago IL, March 2004. The Research Corporation is thanked for the Research Opportunity Award. The Center for Chemical Instrumentation and the Department of Chemistry and Biochemistry at Ohio University are thanked. Preshious Rearden, Ping Chen, Leanna Kishler, and Ornella Smilla are gratefully acknowledged for their helpful comments and suggestions. This work is dedicated to the memory of Mark John. Received for review February 11, 2005. Accepted June 3, 2005. AC0502596

Analytical Chemistry, Vol. 77, No. 16, August 15, 2005

5267