Anal. Chem. 1999, 71, 3226-3230
Identification of Bacterial Proteins Observed in MALDI TOF Mass Spectra from Whole Cells Ricky D. Holland,† Christopher R. Duffy,† Fatemeh Rafii,‡ John B. Sutherland,‡ Thomas M. Heinze,† Claude L. Holder,† Kent J. Voorhees,§ and Jackson O. Lay Jr*,†
Divisions of Chemistry and Microbiology, Food and Drug Administration, National Center for Toxicological Research, Jefferson, Arkansas 72079, and Department of Chemistry, Colorado School of Mines, Golden, Colorado 80401
Characteristic ions in the MALDI TOF mass spectra from bacterial cells have been associated with four known proteins. The proteins, observed both from cells and in filtered cellular suspensions, were isolated by HPLC and identified on the basis of their mass spectra and their partial amino acid sequence, determined using the Edman method (10-15 residues). The acid resistance proteins HdeA and HdeB give rise to ions near m/z 9735 and 9060 in MALDI TOF mass spectra from cells and from extracts of both Escherichia coli 1090 and Shigella flexneri PHS-1059. However, the proteins associated with proteolytic cleavage by the peptidase Lep, rather than the precursor proteins, were observed, both using cells and from cellular extracts. A cold-shock protein, CspA, was associated with the ion near m/z 7643 from Pseudomonas aeruginosa. Similarly, a cold-acclimation protein, CapB, was identified as the source of the ion near m/z 7684 in P. putida. This last protein was homologous with a known CapB from P. fragi. While these experiments involved the detection of known or homologous proteins from typical bacteria, this same approach could also be applied to the detection of unique proteins or biomarker proteins associated with other bacteria of public health significance. Chemotaxonomy represents one approach to the characterization of bacteria. With this method, classification is based upon the measurement of chemical constituents, either as unique components (often called biomarkers) or as contributors to distinct ratios of common chemical components. Chemotaxonomy can be used to differentiate species of bacteria or to differentiate strains of related organisms. In principle, chemotaxonomy using mass spectrometric, rather than chromatographic, or other timeintensive methods, could shorten analysis times enough to justify the greater expense of the mass spectrometry (MS) instrumentation. Use of MS for the characterization of bacteria was investigated as early as 1975 by Anhalt and Fenselau.1 In a series of studies, Fenselau and co-workers evaluated fast atom bombard* To whom correspondence should be addressed. Tel.: (870) 543-7288. Fax: (870) 543-7686. E-mail: jlaycnctr.fda.gov. † Division of Chemistry, Food and Drug Administration. ‡ Division of Microbiology, Food and Drug Administration. § Colorado School of Mines. (1) Anhalt, J. P.; Fenselau, C. Anal. Chem. 1975, 47, 219-225.
3226 Analytical Chemistry, Vol. 71, No. 15, August 1, 1999
ment, laser desorption, and plasma desorption mass spectrometry for the analysis of bacteria.1-5 In these early studies, the principal analytes were phospholipids or other small-molecule biomarkers for bacteria. The use of sample pyrolysis with pattern recognition has been the basis for the identification of intact bacteria using mass spectrometry.6 Chemotaxonomy has also been successfully demonstrated using GC/MS techniques. For example, Fox et al. have demonstrated the use of GC/MS for differentiation of two closely related pathogenic organisms, Bacillus anthracis and B. cereus,7 that are difficult to differentiate phenotypically or genotypically. In 1994, Cain et al.8 reported the use of MALDI TOF/MS to differentiate bacteria on the basis of the analysis of proteins isolated from disrupted cells. Proteins were isolated from crude cellular extracts with methanol. This work was novel because the bacterial biomarkers detected were proteins rather than the smaller molecules examined in the earlier pyrolysis or laser desorption studies. The analysis of bacterial cells, rather than extracted proteins, by MALDI/TOF MS can be viewed as a logical extension of prior experiments. Holland et al.9 successfully characterized bacteria, rather than extracts, on the basis of ions formed by MALDI directly from bacterial cells using conditions much like those used for the analysis of isolated proteins. MALDI/ TOF MS gave signals in excess of 10 000 daltons from cellular components that were presumed to be proteins. The successful identification of a small test set of blind-coded bacteria9 and the appearance, within a few months, of similar results in two other laboratories10,11 suggested that MALDI/TOF MS might be a useful tool for the characterization of bacteria. Subsequently, this (2) Heller, D. N.; Fenselau, C.; Cotter, R. J.; Demirev, P.; Olthoff, J. K.; Honovich, J.; Uy, M.; Tanaky, T.; Kishimoto, Y. Biochem. Biophys. Res. Commun. 1987, 142, 194-199. (3) Ho, B. C.; Fenselau, C.; Hansen, G.; Larsen, J.; Daniel, A. A. Clin. Chem. 1983, 29, 1349-1353. (4) Fenselau, C.; Cotter, R. J. Chem. Rev. 1987, 87, 501. (5) Heller, D. N.; Cotter, R. J.; Fenselau, C.; Uy, O. M. Anal. Chem. 1987, 59, 2806-2809. (6) DeLuca, S.; Sarver, E. W.; Harrington, P. D.; Voorhees, K. J. Anal. Chem. 1990, 62, 1465-1472. (7) Fox, A.; Rogers, J. C.; Fox, K. F.; Schnitzer, G.; Morgan, S. L.; Brown, A.; Aono, R. J. Clin. Microbiol. 1990, 546-552. (8) Cain, T. C.; Lubman, D. M.; Weber, W. J., Jr. Rapid Commun. Mass Spectrom. 1994, 8, 1026-1031. (9) Holland, R. D.; Wilkes, J. G.; Rafii, F.; Sutherland, J. B.; Persons, C. E.; Voorhees, K. J.; Lay, J. O., Jr. Rapid Commun. Mass Spectrom. 1996, 10, 1227-1232. (10) Claydon, M. A.; Davey, S. N.; Edward-Jones, V.; Gordon, D. B. Nature Biotechnol. 1996, 14, 1584-1586. 10.1021/ac990175v CCC: $18.00
© 1999 American Chemical Society Published on Web 06/16/1999
technique has been used to distinguish 25 different strains of Escherichia coli.12 Although many of the same ions were present upon reanalysis of bacterial cultures, the ratios of ion intensity values and the specific ions observed for bacteria almost always varied somewhat in replicate experiments.9-11 Part of this variation could be described as “method-related” differences in spectra obtained from otherwise identical bacteria. An interlaboratory comparison by researchers at the University of Alberta and at the U. S. Army’s Aberdeen Proving Ground facilities has demonstrated that methodrelated variation can be controlled.13 Unfortunately, another major source of variation in spectra from bacteria is associated with differences in the biology of the bacteria which is based on environmental conditions, and this is potentially much more difficult to control. For bacteria obtained from the environment rather than laboratory settings, variability in the MALDI TOF mass spectra, associated with the organisms’ response to the environment, may be much more problematic than method-related variation, precisely because of the ability of MS to detect such small changes in chemical composition. Characterization of the protein biomarkers in the MALDI spectra of bacteria may lead to a better understanding of these biology-based changes, and the identification of specific biomarker ions should facilitate the use of specific ions for chemotaxonomic purposes. In this work, we demonstrate the isolation and characterization of several proteins responsible for characteristic ions in the MALDI/TOF MS spectra from bacteria. Some of the preliminary work was reported in ref 14. EXPERIMENTAL SECTION Preparation of samples. The bacteria Shigella flexneri PHS1059, E. coli 1090, Pseudomonas aeruginosa, and P. putida were grown on tryptic soy agar for 24 h. Colonies were suspended in 1 mL of 2:1 double distilled water/acetonitrile (J. T. Baker, Phillipsburg, NJ) with 0.1% trifluoroacetic acid (Fluka Chemical, Ronkonkoma, NY) in 1.5 mL polypropylene microcentrifuge tubes to make a very cloudy suspension. The bacterial suspension was then centrifuged using a 5415C Eppendorf centrifuge (Eppendorf, Madison, WI) set at 600g for 2-3 min. Supernatants from S. flexneri and E. coli were then filtered (0.5-µm Millipore filter) and stored at 5 °C until HPLC analysis. The filtered bacterial extracts from P. aeruginosa and P. putida were vacuum evaporated to dryness and redissolved in 200 µL of 2:1 water/acetonitrile with 0.1% TFA prior to HPLC separation. Isolation and sequencing of proteins. The bacterial components from cell extracts were separated and isolated by HPLC using a 4.6 × 250 mm Vydac C18 protein/peptide column (Vydac, Hesperia, CA). About 100 µL of the bacterial extract (filtered suspension) was used for each injection onto the HPLC column. The mobile phases were water and acetonitrile with 0.1% TFA in both phases. The gradient was 20-90% acetonitrile (linear) over (11) Krishnamurthy, T.; Ross, P. L. Rapid Commun. Mass Spectrom. 1996, 10, 1992-1996. (12) Arnold, R. J.; Reilly, J. P. Rapid Commun. Mass Spectrom. 1998, 12, 630636. (13) Wang, Z.; Russon, L.; Li, L.; Roser, D. C.; Long, S. R. Rapid Commun. Mass Spectrom. 1998, 12, 456-464. (14) Holland, R. D.; Rafii, F.; Holder, C. L.; Heinze, T. M.; Sutherland, J. B.; Voorhees, K. J.; Lay, J. O., Jr. Proceedings of the 46th ASMS Conference on Mass Spectrometry and Allied Topics, Orlando, FL, 1998; p 154.
30 min, held at 90% for 5 min, then back to 20% after 3 min. The flow rate (Varian 9012 HPLC pump, Varian Instrument Group, Sugarland, TX) was 1.0 mL/min. Components in the suspension were detected at 260 nm with a Varian 9050 detector. All major HPLC separated components were manually collected in 1.5-mL microcentrifuge tubes. The solutions containing components were vacuum-evaporated to dryness using a Savant SCA110A system (Savant Instruments, Inc., Holbrook, NY). The system was set at 65 °C. The dried components were dissolved in 10 µL of 0.1% aqueous TFA for analysis. Specific HPLC separated components containing the proteins of interest were identified on the basis of MALDI/TOF MS analysis of each. Each of the components showing a response for m/z 9060, and 9735 from S. flexneri were collected 20 times, and the HPLC separated components corresponding to the same mass were pooled. Similarly, an HPLC separated component containing m/z 9739 from E. coli was collected 20 times and pooled. For the P. putida and P. aeruginosa, HPLC separated components containing peaks at m/z 7643 and 7684, respectively, were collected. However, for these two bacteria, only two injections were made, from more concentrated samples. All these purified and pooled HPLC separated components were sent to Midwest Analytical, Inc., (St. Louis, MO) for automated Edman analysis. Proteins were identified using the BLAST program from the National Center for Biotechnology Information.15,16 MALDI TOF/MS Analysis. The matrix (R-cyano-4-hydroxycinnamic acid) (Sigma Chemical Co., St. Louis, MO) was dissolved in 2:1 water:acetonitrile with 0.1% TFA. The matrix was added beyond the saturation point, shaken vigorously, and centrifuged for 1 to 2 min at 14 000 rpm. The clear liquid was decanted for use. Fresh matrix was prepared daily. Bovine cytochrome c or myoglobin from horse heart (Sigma) were used as standards for instrument calibration or as internal mass standards. The calibration standards (1 mg/mL) were prepared in 0.1% TFA in water. The HPLC fractions and filtered bacterial extracts were analyzed using 2 µL from a 9:2 (µL each, matrix/sample) mixture. Unfiltered bacterial suspensions were analyzed using a 18:2 ratio. The calibration standard was prepared at a 27:2 matrix/analyte ratio mixture. For each matrix/analyte mixture, 2 µL was placed on separate 2-mm sample pins and allowed to air-dry. The samples were placed in a Vestec model YM200 (Vestec, Inc., Houston, TX) linear time-of-flight mass spectrometer. The samples were desorbed with a Nd:YAG laser with the frequency tripled at 355 nm. The acceleration voltage was set at 26 000 V. Detector response was monitored using a Tektronix TDS 520 digitizing oscilloscope (Beaverton, OR). Mass assignments were made using a commercial software package (Grams 386, Galactic, Inc., Salem, NH). RESULTS AND DISCUSSION Acid-Resistance Proteins. The first step in the identification of the proteins giving rise to signals in the MALDI/TOF mass spectra was their detection in cellular suspensions. Prominent ions were observed near m/z 9060 and 9735 in MALDI/TOF mass spectra from cells of S. flexneri, using the method described in (15) Altschul, S. F.; Gish, W.; Miller, W.; Myers, E. W.; Lipman, D. J. J. Mol. Biol. 1990, 215, 403-410. (16) Altschul, S. F.; Madden, T. L.; Schaffer, A. A.; Zhang, J.; Zhang, Z.; Miller, W.; Lipman, D. J. Nucleic Acids Res. 1997, 25, 3389-3402.
Analytical Chemistry, Vol. 71, No. 15, August 1, 1999
3227
Figure 1. MALDI/TOF mass spectra of the supernatants from Shigella flexneri and Escherichia coli. (See text for conditions.)
Figure 3. MALDI/TOF mass spectra of isolated proteins 1 and 2 from Shigella flexneri. (See text for conditions.)
Figure 2. HPLC separation of proteins 1 and 2 from Shigella flexneri. (See text for conditions.)
ref 9 (data not shown). These same ions were also observed, using cells, with a closely related bacteria, E. coli. These two ions were detected in cellular supernatants from both bacteria (see Figure 1). The mass assignments for both components (m/z 9060 and 9735) showed a measured standard deviation of (5 daltons on the basis of triplicate analyses. (The precision in these mass measurements is typical of our results with other bacteria and proteins using this instrument.) Figure 2 shows an HPLC chromatogram from the same supernatant of a centrifuged suspension of S. flexneri. MALDI/TOF MS analyses of individual HPLC fractions were used to determine the retention times for target proteins. They were found in fractions corresponding to the peaks marked 1 and 2 in Figure 2, and their MALDI/TOF mass spectra are shown in Figure 3. Mass spectral analysis of the other major HPLC peaks associated with the separation (data not shown) revealed that many other proteins also observed in the MALDI spectrum from bacteria were present in the supernatant as well, but these have not yet been identified. For protein identification, peaks 1 and 2 were separated by HPLC, and the corresponding peaks were pooled. The two pooled samples were purified in a final HPLC step (Figure 4) and submitted to automated Edman analysis. On the basis of the first 10 amino acid residues from each protein, the identities were determined to be homologous with hns deletion-induced protein 3228 Analytical Chemistry, Vol. 71, No. 15, August 1, 1999
Figure 4. HPLC trace from the pooled samples of proteins 1 and 2 from Shigella flexneri. (See text for conditions.)
B (HdeB) (m/z 9060) and hns deletion-induced protein A (HdeA) (m/z 9735) from E. coli.17 Both of these proteins are associated with acid resistance.20 The first 10 N-terminal amino acids sequenced by the Edman method, for the HPLC component labeled 1 in Figure 2, were ANESAKDMTC, corresponding to a portion (residues 34-43) of the sequence for the HdeB precursor from E. coli strain K12. Figure 5a shows the complete sequence for E coli K12 protein HdeB where the bold portion (residues 34112) of the sequence depicts the smaller portion of a homologous protein actually collected from S. flexneri. The predicted mass for this sequence is 9064 daltons (Table 1), in good agreement with the measured mass of 9060 daltons (see Table 1, Figure 3b). Similarly, for the HPLC separated component corresponding to peak 2 in Figure 2, the first 10 amino acids sequenced by the Edman method were ADAQKAADNK, corresponding to a portion (residues 22-31) of the sequence for the HdeA precursor from (17) Link, A. J.; Robison, K.; Church, G. M. Electrophoresis 1997, 18, 12591313. (18) Yoshida, T.; Ueguchi, C.; Mizuno, T. J. Bacteriol. 1993, 175, 7747-7748. (19) Arnqvist, A.; Olsen, A.; Normark, S. Mol. Microbiol. 1994, 13, 1021-1032. (20) Yoshida, T.; Ueguchi, C.; Yamada, H.; Mizuno, T. Mol. Gen. Genet. 1993, 237, 113-122.
Figure 6. Amino acid sequence of (a) P. aeruginosa CspA and (b) P. fragi CapB. Figure 5. Amino acid sequence of E. coli K12 (a) HdeB and (b) HdeA. Table 1: Proteins Identified from Bacterial Extracts bacterium
protein
HPLC tr (min)
observed massa
expected massb
mass differencec
S. flexneri S. flexneri E. coli P. aeruginosa P. putida
HdeA HdeB HdeA CspA CapBd
10.9 10.5 10.9 7.3 8.2
9735 9060 9739 7643 7684
9739 9064 9739 7605
-4 -4 0 38
a From suspended cells, in daltons. b For the protonated molecule, predicted from the sequences in Figures 5-8, in daltons. c Difference between the expected and observed masses, in daltons. d Homologue of P. fragi CapB; the actual mass for the P. putida protein is unknown.
E. coli. The expected mass, based on the protein sequence shown in bold (residues 22-110, Figure 5b), is 9739, in good agreement with 9735 daltons, the observed mass (Figure 3a). The segments of proteins identified in our study are both part of larger precursor proteins. They are consistent with the expected proteolytic cleavage by the signal peptidase, Lep, between residues 33 and 34 for HdeB and residues 21 and 22 for HdeA.17 These proteins are encoded by genes on the hns-dependent expression AB operon (hdeAB),18 which is regulated by the stationary-phasespecific sigma factor (σs).19 Although the function of the two proteins is not known,18,20 they are associated with the acidresistance phenotype21 which allows bacteria to survive acidic conditions in the stomach. Both of these proteins isolated from S. flexneri were homologous to proteins associated with E. coli K12. This is consistent with phylogenetic studies reporting that these bacteria might even be classified as a single species.22 Nevertheless, to provide additional evidence that these two components were actually homologous proteins, one of them was also characterized from E. coli. The protein believed to be HdeA (identical to peak 2 from S. flexneri) was isolated and characterized from E. coli exactly as described above. Analysis of the corresponding E. coli HPLC separated component (peak 2) by Edman degradation gave the same amino acid sequence observed with S. flexneri. Moreover, the mass determined by MALDI was 9739 daltons (Table 1), in perfect agreement with the predicted nominal mass. (An HPLC separated component (m/z 9060, peak 1) with the same retention behavior as HdeB has also been isolated from E. coli but not (21) Waterman, S. R.; Small, P. L. C. Mol. Microbiology 1996, 21, 925-940. (22) Wang, R. F.; Cao, W. W.; Cerniglia, C. E. Mol. Cell. Probes 1997, 11, 427432.
sequenced.) Thus, the HdeA protein gives rise to the same signal in the MALDI/TOF MS spectra of E. coli and S. flexneri using either cells or extracts, and this evidence, as well as the phylogenetic studies cited above, strongly suggests that the same is true for the HdeB protein. Interestingly, the acid resistance proteins reported in this work and in our preliminary report14 were not identified in a recent study using a peptide mapping approach, even though they were detected.23 Our work suggests two possible reasons why these proteins might not have been identified by peptide mapping. First, the proteins might have been resistant to proteolysis. In our laboratory, we have been unable to produce good fragments from HdeA or HdeB using trypsin (data not shown). Second, as noted above, proteolytic cleavage by Lep resulted in smaller proteins than were expected from the genome database. Deviations from the masses or sequences predicted on the basis of the genome may confound attempts to identify proteins exclusively on the basis of mass spectrometry measurements, unless such changes are accounted for in some way. For example, a study of the E. coli proteome indicated that 60% of the proteins encoded in the genome were proteolytically processed.17 In this study, the Edman method was critical to the identification of these proteins. Cold-Induced Proteins. On the basis of the observation that MALDI/TOF mass spectra of bacteria change upon storage at 5 °C,24 the presence of cold shock proteins was anticipated. A protein isolated from P. aeruginosa, with a mass of 7643 daltons, was collected in the same manner as the proteins described above (Table 1). This same mass was observed both from cells and HPLC separated fractions. However, the MALDI/TOF MS spectra from these HPLC fractions also contained several smaller ions, with lower masses corresponding to up to six fewer amino acids. Prior to the Edman sequence analysis, the significance of these ions was not clear. The first 10 amino acids from this sample gave the sequence RQNGTVKWFN, which is an exact match for residues 4-13 from the P. aeruginosa cold shock protein A (CspA, see Figure 6a). The absence of residues 1-3 (MSN) is attributed to degradation, starting during protein isolation and continuing until the sequence was determined. However, the major component in both cells and in the initially collected HPLC fractions is believed to be intact CspA. The expected mass, including the three missing N-terminal residues, is 7605 daltons, whereas the major observed mass was 7643 daltons (Table 1). The mass that is higher (23) Dai, Y.; Liang, L.; Roser, D. C.; Long, S. R. Rapid Commun. Mass Spectrom. 1999, 13, 73-78. (24) Holland, R. D.; Burns, G.; Persons, C. C.; Rafii, F.; Sutherland, J. B.; Lay, J. O., Jr. Proceedings of the 45th ASMS Conference on Mass Spectrometry and Allied Topics, Palm Springs, CA, 1997; p 1353.
Analytical Chemistry, Vol. 71, No. 15, August 1, 1999
3229
by 38 daltons attributed to the detection of a potassium adduct ion rather than a protonated molecule, although some other form of protein modification cannot be ruled out. A cold-acclimation protein was also identified as one of the components detected in the MALDI/TOF mass spectra from P. putida. An HPLC component was collected having a mass of 7684 daltons, also corresponding to a major ion in the spectra obtained from cells (data not shown). The sequence observed from this component, on the basis of the Edman degradation, was SNRQKGTVKWFNDEK, corresponding to 14 of 15 expected amino acids in residues 2-16 (M, the first residue is not typically detected) from a known P. fragi cold-acclimation protein B (CapB, Figure 6b). The observed sequence corresponds to exchange of a lysine for a threonine in the sixth residue of the sequence. The predicted mass for this protein, assuming no other mutations, is 7727 daltons. This is not in exact agreement with the observed mass, 7684 daltons, suggesting other differences between this CapB and the P. fragi CapB protein. The identification of this protein as a CapB homologue was relatively straightforward with both the molecular weight data and the protein sequence data. CONCLUSION The primary objective of this work was to isolate and identify specific proteins associated with characteristic ions in spectra obtained using the very rapid MALDI/TOF MS method of analysis (9-13). This was accomplished using bacterial suspensions; ions prominent in MALDI spectra from cells were also observed using MALDI to analyze bacterial suspensions and subsequently in spectra from HPLC isolated fractions. Table 1 gives a list of the
3230 Analytical Chemistry, Vol. 71, No. 15, August 1, 1999
proteins identified, their HPLC retention times, the expected masses (based on the Edman derived sequences shown in the figures), and the experimentally observed masses. The identity of the proteins was deduced on the basis of the amino acid sequences and the measured protein masses. The use of these ions as biomarkers should be significantly enhanced by their identification, especially since they have now been associated with specific bacterial genes. We have identified two acid resistance proteins, HdeA and HdeB, from both E. coli and S. flexneri and identified the specific masses in the MALDI/TOF mass spectra from cells (and cell extracts) associated with these two important proteins. Similarly, a cold-shock protein in P. aeruginosa and coldacclimation protein from P. putida were identified. The same approach could probably be applied to the detection of antibioticresistance, or heat-shock proteins, on the basis of the characterization of unique proteins detected from antibiotic- or heat-resistant bacteria. This approach holds much promise because so many proteins can be detected simultaneously in a single MALDI/TOF MS experiment; only the unique or biomarker proteins need actually be sequenced in the more time-intensive sequencing steps. Identification of biomarker proteins will also aid the study of bacterial physiology, as the profiles of specific proteins can be followed under various conditions of stress.
Received for review February 8, 1999. Accepted May 5, 1999. AC990175V
