Ionization-Mass Spectrometry of

Anal. Chem. 2007, 79, 1115-1125

Matrix-Assisted Laser Desorption/Ionization-Mass Spectrometry of Hydrophobic Proteins in Mixtures Using Formic Acid, Perfluorooctanoic Acid, and Sorbitol Rachel R. Ogorzalek Loo*,† and Joseph A. Loo†,‡

Department of Biological Chemistry and Department of Chemistry & Biochemistry, Molecular Biology Institute, UCLAsDOE Institute for Genomics & Proteomics, University of CaliforniasLos Angeles, Los Angeles, California 90095

Three MALDI-MS sample/matrix preparation approaches were evaluated for their ability to enhance hydrophobic protein detection from complex mixtures: (1) formic acidbased formulations, (2) perfluorooctanoic acid (PFOA) surfactant addition, and (3) sorbitol addition. While MALDI-MS of Escherichia coli cells desorbed from a standard sinapinic acid matrix displayed 94 (M + H)+ ions, 119 were observed from a formic acid-based matrix with no more than 10 common to both. Formic acid matrix revealed many lipoproteins and an 8282 m/z ion proposed to be the abundant, water-insoluble ATPase proteolipid. Among the formic acid-based cocktails examined, the slowest rate of serine/threonine formylation was found for 50% H2O/33% 2-propanol/17% formic acid. Faster formylation was observed from cocktails containing more formic acid and from mixtures including CH3CN. Sinapinic, ferulic, DHB, 4-hydroxybenzylidene malononitrile, and 2-mercaptobenzothiazole matrixes performed well in formic acid formulations. Dramatic differences in mixture spectra were also observed from PFOA/ sinapinic acid, at detergent concentrations exceeding the critical micelle concentration, although these matrix cocktails proved difficult to crystallize. E. coli ions observed from these matrix conditions are listed in Tables S-1 and S-3 (Supporting Information). Similar complementarity was observed for M. acetivorans whole-cell mixtures. Including sorbitol in the sinapinic acid matrix was found to promote homogeneous crystallization and to enhance medium and higher m/z ion detection from dilute E. coli cellular mixtures. Profiling intact bacterial proteins by MALDI-MS has shown utility in detecting bioterrorist acts and identifying clinical isolates.1-15 It promises to streamline analyses related to antibacte* To whom correspondence should be addressed. E-mail: [email protected]. Phone: (310) 206-1484. FAX: (310) 206-7286. † Department of Biological Chemistry. ‡ Department of Chemistry & Biochemistry. (1) Cain, T. C.; Lubman, D. M.; Weber, W. J. Jr. Rapid Commun. Mass Spectrom. 1994, 8, 1026-1030. (2) Krishnamurthy, T.; Ross, P. L.; Rajamani, U. Rapid Commun. Mass Spectrom. 1996, 10, 883-888. 10.1021/ac061916c CCC: $37.00 Published on Web 12/09/2006

© 2007 American Chemical Society

rial drug discovery16,17 and identification of food-borne contamination.18 MALDI-MS mixture analyses also form the basis of mass spectrometric imaging technologies,19 diagnostics, and biomarker discovery strategies.20,21 Key limitations in MALDI-MS analyses of intact protein mixtures are the (1) difficulty in obtaining unambiguous protein identifications, (2) specification of inter- and intralaboratory reproducibility, (3) restrictions on quantitation, and (4) limited set of proteins detected. This study addresses the last limitation by expanding the range of methods revealing different protein assortments from complex mixtures. The extent of ambiguity in identifications assembled from peptide masses, tandem mass spectrometry, or both is usually (3) Holland, R. D.; Wilkes, J. G.; Rafii, F.; Sutherland, J. B.; Persons, C. C.; Voorhees, K. J.; Lay, J. O. Jr. Rapid Commun. Mass Spectrom. 1996, 10, 1227-1232. (4) Arnold, R. J.; Karty, J. A.; Ellington, A. D.; Reilly, J. P. Anal. Chem. 1999, 71, 1990-1196. (5) Arnold, R. J.; Reilly, J. P. Anal. Biochem. 1999, 269, 105-112. (6) Demirev, P. A.; Ho, Y.-P.; Ryzhov, V.; Fenselau, C. Anal. Chem. 1999, 71, 2732-2738. (7) Gantt, S. L.; Valentine, N. B.; Saenz, A. J.; Kingsley, M. T.; Wahl, K. L. J. Am. Soc. Mass Spectrom. 1999, 10, 1131-1137. (8) Saenz, A. J.; Petersen, C. E.; Valentine, N. B.; Gantt, S. L.; Jarman, K. H.; Kingsley, M. T.; Wahl, K. L. Rapid Commun. Mass Spectrom. 1999, 13, 1580-1585. (9) Lay, J. O. TrAC, Trends Anal. Chem. 2000, 19, 507-516. (10) Ryzhov, V.; Fenselau, C. Anal. Chem. 2001, 73, 746-750. (11) Madonna, A. J.; Basile, F.; Furlong, E.; Voorhees, K. J. Rapid Commun. Mass Spectrom. 2001, 15, 1068-1074. (12) Demirev, P. A.; Lin, J. S.; Pineda, F. J.; Fenselau, C. Anal. Chem. 2001, 73, 4566-4573. (13) Smole, S. C.; King, L. A.; Leopold, P. E.; Arbeit, R. D. J. Microbiol. Meth. 2002, 48, 107-115. (14) Fenselau, C.; Demirev, P. A. Mass Spectrom. Rev. 2002, 20, 157-171. (15) Jones, J. J.; Stump, M. J.; Fleming, R. C.; Lay, J. O. Jr.; Wilkins, C. L. Anal. Chem. 2003, 75, 1340-1347. (16) Ogorzalek Loo, R. R.; Du, P.; Loo, J. A.; Holler, T. J. Am. Soc. Mass Spectrom. 2002, 13, 804-812. (17) Wilcox, S. K.; Cavey, G. S.; Pearson, J. D. Antimicrob. Agents Chemother. 2001, 45, 3046-3055. (18) Holland, R. D.; Rafii, F.; Heinze, T. M.; Sutherland, J. B.; Voorhees, K. J.; Lay, J. O., Jr. Rapid Commun. Mass Spectrom. 2000, 14, 911-917. (19) Chaurand, P.; Schwartz, S. A.; Caprioli, R. M. Curr. Opin. Chem. Biol. 2002, 6, 676-681. (20) Petricoin, E. F., III; Ornstein, D. K.; Paweletz, C. P.; Ardekani, A.; Hackett, P. S.; Hitt, B. A.; Velassco, A.; Trucco, C.; Wiegand, L.; Wood, K.; Simone, C. B.; Levine, P. J.; Linehan, W. M.; Emmert-Buck, M. R.; Steinberg, S. M.; Kohn, E. C.; Liotta, L. A. J. Natl. Cancer Inst. 2002, 94, 1576-1578. (21) Howard, B. A.; Wang, M. Z.; Campa, M. J.; Corro, C.; Fitzgerald, M. C.; Patz, E. F., Jr. Proteomics 2003, 3, 1720-1724.

Analytical Chemistry, Vol. 79, No. 3, February 1, 2007 1115

specified statistically. When the intact protein mass measurement and identification analyses are directly linked, i.e., dissociation of an intact protein ion isolated within the mass spectrometer, the statistical evaluation is sufficient. However, profiling approaches that determine intact masses from complex mixtures, but follow with fractionation (e.g., HPLC or liquid-phase isoelectric focusing) to simplify the mixture for enzymatic digestion and protein identification, introduce additional ambiguity, namely, the possibility that multiple proteins from different fractions possess the target mass. Certainly the similarity in mass between abundant Escherichia coli proteins cspA, cspC, and rl29, (Mr 7272.09, 7271.17, and 7273.45, respectively; Swiss Prot accession numbers P15277, P36996, and P02429) has frequently led to ambiguity in profiling experiments.4,5,10,15,16,22,23 Moreover, this ambiguity is difficult to quantify. At best, databases reveal only a small fraction of possible species arising from co- and post-translational modifications, alternative splicing, and protein processing. Methods to reduce this ambiguity rely on obtaining higher accuracy, better resolved measurements by FTMS,15 or securing distinctive constraints for the intact species, such as the number of methionines, revealed by selenomethionine labeling.16 Not fully appreciated initially was the extreme interdependence between mass spectral profiles and sample provenance and treatment prior to mass analysis. Questions relating to intra- and interlaboratory reproducibility have been raised in MALDI and SELDI profiling applications.8,22,24,25 In retrospect, the variations observed in different spectra are unremarkable, reflecting problems common to analyses of complex “real” substances when few steps have been taken to remove interferences; a subject addressed in quantitative analysis texts.26 Because “lab” samples are relatively free of interferences, their analysis is simple and influenced by only a small number of variables. Analysis of real substances, influenced by innumerable variables, generally requires that samples be freed of interferences or that realistic goals and limits be specified with methods validated for the range of sample compositions anticipated. Several authors have discussed inter- and intralaboratory reproducibility for MALDI-MS.7,8,22,24 By carefully attending to sample and matrix preparation and deposition, and to consistent data acquisition protocols, acceptable reproducibility can be obtained and specified. Quantification by MALDI-MS has been examined in several studies.27,28 With incorporated internal standards, laboratories have demonstrated linearity in quantitation ranging from 1 to 3 decades of analyte concentration. Analyses of complex protein mixtures within biological matrixes, while permitted by the buffer and salt tolerance of MALDIMS, may yet be inadequate due to the limited range of species detected. Ion suppression, solubility, protein abundance, basicity, (22) Dai, Y.; Li, K.; Roser, D. C.; Long, S. R. Rapid Commun. Mass Spectrom. 1998, 13, 73-78. (23) Easterling, M. L.; Colangelo, C. M.; Scott, R. A.; Amster, I. J. Anal. Chem. 1998, 70, 2704-2709. (24) Williams, T. L.; Andrzejewski, D.; Lay, J. O.; Musser, S. M. J. Am. Soc. Mass Spectrom. 2003, 14, 342-351. (25) Diamandis, E. P. Clin. Chem. 2003, 49, 1272-1275. (26) Skoog, D. A.; West, D. M. Fundamentals of Analytical Chemistry, 3rd ed.; Holt, Rinehart and Winston: New York, 1976; p 582. (27) Nelson, R. W.; McLean, M. A.; Hutchens, T. W. Anal. Chem. 1994, 66, 1408-1415. (28) Gusev, A. I.; Wilkinson, W. R.; Proctor, A.; Hercules, D. M. Anal. Chem. 1995, 67, 1034-1041.

1116

Analytical Chemistry, Vol. 79, No. 3, February 1, 2007

and other factors limit the subset of observed proteins. For example, E. coli cell profiles largely reflect ribosomal and other basic proteins, excluding or underrepresenting many of the most abundant proteins in the cell. Employing different matrixes often fails to change the list of detected masses radically, although auspiciously chosen additive and solvent combinations may increase the number of observed masses somewhat (especially when cellular lysis is decoupled from the solvents employed for analysis; i.e., cells lysed prior to analyses). Altered culture media or growth phase often changes the appearance of mass profiles more than sample preparation does. MALDI-MS of full-length membrane proteins and compatible methods for hydrophobic polypeptides have been reported.29-34 Similar studies apply to ESI-MS analysis.35-39 Most of these efforts optimized the detection of isolated proteins, protein complexes, or overexpressed proteins, i.e., samples containing relatively few proteins. This study focuses on extending the range of proteins detectable from mixtures, emphasizing hydrophobic proteins, a class underserved at present. Solvents, matrixes, additives, and sample/matrix deposition techniques have been tailored to enhance the detection of hydrophobic proteins, radically changing the mass spectra obtained from cellular mixtures. Popular formic acid-basic solvent and matrix formulations are compared with regard to their ability to covalently modify proteins. Studies of unfractionated protein mixtures will benefit from these approaches, as will analyses of purified peptides and proteins and of intact proteins isolated in polyacrylamide gel slices. EXPERIMENTAL SECTION Lyophilized E. coli K12 cells were acquired from Sigma (EC1, Lot 59H8604, St. Louis, MO), as were all protein standards. The cells were resuspended in H2O, sonicated for 3-5 min in an ultrasonic bath, pelleted, and dried. (Sonication was not required to obtain MALDI profiles from these cells; however, we wished to ensure that cell lysis was decoupled from MALDI-MS solvent selection. Following sonication, cell pellets were dried and stored at -80 °C. Methanosarcina acetivorans (methanogenic archae), cultivated as single cells anaerobically in a 100 mM methanol, 200 mM NaClbased medium at 37 °C, were thoughtfully provided by R. Gunsalus (UCLA). Cells were pelleted, dried, and stored at -80 (29) Schey, K. L.; Papac, D. I.; Knapp, D. R.; Crouch, R. K. Biophys. J. 1992, 63, 1240-1243. (30) Rosinke, B.; Strupat, K.; Hillenkamp, F.; Rosenbusch, J.; Dencher, N.; Krueger, U.; Galla, H.-J. J. Mass Spectrom. 1995, 30, 1462-1468. (31) Allmaier, G.; Schaffer, C.; Messner, P.; Rapp, U.; Mayer-Posner, F. J. J. Bacteriol. 1995, 177, 1402-1404. (32) Barnidge, D. R.; Dratz, E. A.; Sunner, J.; Jesaitis, A. J. Protein Sci. 1997, 6, 816-824. (33) Green-Church, K. B.; Limbach, P. A. Anal. Chem. 1998, 70, 5322-5325. (34) Schaller, J. In Protein and Peptide Analysis: New Mass Spectrometric Applications; Chapman, J. R., Ed.; Humana: Totowa, NJ, 2000; Vol. 146, pp 425-437. (35) Schindler, P. A.; Van, Dorsselaer, A.; Falick, A. M. Anal. Biochem. 1993, 213, 256-263. (36) le Maire, M.; Deschamps, S.; Moller, J.; Le, Caer, J. P.; Rossier, J. Anal. Biochem. 1993, 214, 50-57. (37) Ogorzalek, Loo, R. R.; Dales, N.; Andrews, P. C. Protein Sci. 1994, 3, 19751983. (38) Schaller, J.; Pellascio, B. C.; Schlunegger, U. P. Rapid Commun. Mass Spectrom. 1997, 11, 418-426. (39) Whitelegge, J. P.; Gundersen, C. B.; Faull, K. F. Protein Sci. 1998, 7, 14231430.

°C for subsequent resuspension prior to analysis. Murine mitochondria, (pelleted, dried, and stored similarly) were a kind gift from B. Berhane, T. Vondriska, and P. Ping (UCLA). Sinapinic acid, ferulic acid, 2-MBT, and DHB (3,5-dimethoxy4-hydroxycinnamic acid, 4-hydroxy-3-methoxycinnamic acid, 2-mercaptobenzothiazole, and 2, 5-dihydroxybenzoic acid, respectively) were employed as matrixes with and without the detergent perfluorooctanoic acid (pentadecafluorooctanoic acid or PFOA) or the monosaccharide sorbitol. “Standard matrix” is defined as follows: sinapinic acid (saturated) in 33% CH3CN/67% H2O/0.1% TFA (v/v/v), or ferulic acid (saturated) or 2-MBT (saturated) in 50% CH3CN/50% H20/0.1% TFA (v/v/v). DHB was prepared in the latter solvent at concentrations of 20 or 80 mg/mL or as a saturated solution. “Formic acid-based matrix” or FA matrix is defined as a saturated matrix in 60% formic acid/40% CH3CN. Dried cells, estimated to contain 50% protein by weight, were resuspended in the desired solvent to ∼2.5 mg of dry cells/mL. For analysis of H2O-suspended cells, 1 µL was withdrawn and mixed with 9 µL of standard matrix. For examination of formic acid-based conditions, cells were resuspended and vortexed in FA (60% formic acid/40% CH3CN (v/v)) immediately prior to deposition, maintaining identical sample and matrix solvent compositions to prevent precipitation. The bacterial pellet dissolved completely in the FA formulation, but acid-labile bonds, e.g., Asp-Pro, are susceptible to cleavage in this solvent. Subtle differences in spotting technique impacted whether spectra could be obtained from FA formulations, with successful methods yielding larger matrix crystals. Because FA matrix spreads freely on the probe surface, only 0.2 µL of sample was applied, deposited by quickly touching the pipet tip to the probe. Immediately thereafter, 0.2 µL of FA matrix (ferulic acid, 2-MBT, or DHB) was deposited on top of the droplet, without additional mixing. (Sample spots allowed to dry prior to matrix addition did not reveal proteins as high in mass as did those that remained wet.) Because crystals did not form reliably with ferulic acid, appearing only 50% of the time, sample spots in that matrix were prepared in duplicate. For sinapinic acid FA matrix, crystal “seeding” prior to spotting of sample and matrix was essential to obtain spectra reliably. Thus, crystals were first seeded by spotting 0.2 µL of the FA sinapinic acid matrix onto the MALDI probe. After the matrix dried, 0.2 µL of sample and 0.2 µL of matrix were spotted without mixing as per the above protocol. Figure 1 illustrates matrix deposits obtained from the recommended procedures. Stock solutions of PFOA (1 or 2% w/v in H2O) were incubated at 37 °C for 15-30 min to dissolve the detergent completely. When employed, PFOA was added to both sample and matrix. D-Sorbitol was optionally added to 1.5% (w/v) of matrix solution. To minimize serine and threonine side chain formylation induced at high concentrations of formic acid, analyses were performed within 1 h of solvent addition. Little, if any, formylation or oxidation was observed in this time frame, although matrixspotted samples maintained overnight (even under vacuum) displayed significant formylation, as observed previously.40 These formic acid-based matrix solutions performed well for at least 1 week, stored in the dark at room temperature. In contrast, a (40) Ehring, H.; Stromberg, S.; Tjernberg, A.; Noren, B. Rapid Commun. Mass Spectrom. 1997, 11, 1867-1873.

Figure 1. E. coli cells suspended in 2:1 formic acid/CH3CN spotted onto the sample plate with MBT, ferulic acid, and sinapinic acid (SA) matrix, as described.

colorless CH3CN solution of indole-acrylic acid turned purple following addition of concentrated formic acid; this instability arises perhaps from indole’s sensitivity to oxidation under acidic conditions. Reactivities of various formic acid-based cocktails, referred to as “Kim”, “Cohen”, “Feick”, “Cadene”, and FA were evaluated by dissolving lyophilized protein standards in them and depositing the mixture and matrix for MS. Cocktail/matrix compositions follow. Kim cocktail: The sample was dissolved in 25 µL of 70% CH3CN/30% formic acid; a saturated solution of matrix was prepared by dissolving sinapinic acid in 70% CH3CN/30% H2O or 70% CH3CN/29.9% H2O/0.1% TFA.41 Cohen cocktail: The sample was dissolved in 25 µL of 50% H2O/33% 2-propanol/17% formic acid, while a saturated solution of matrix was prepared by dissolving R-cyano-4-hydroxycinnamic acid in the same solvent. This composition has been referred to previously as FWI.42,43 Feick cocktail: The sample was dissolved in 25 µL of 50% formic acid/ 25% CH3CN/15% 2-propanol/10% H2O.44 Matrixes were prepared by dissolving either R-cyano-4-hydroxycinnamic acid (saturated) in 1:1 CH3CN/0.2% TFA (v/v) or sinapinic acid (saturated) in 1:2 CH3CN/0.15% TFA (v/v). This composition has been referred to previously as FAPH.40 Cadene cocktail: The sample was dissolved in 25 µL of 50% formic acid/33% 2-propanol/17% water (v/v/v). Cadene and Chait45 deposited samples in this cocktail onto the MALDI target with a modified thin-layer method. To sidestep our limited mastery of this deposition method and perform reproducible measurements, we instead spotted an aliquot of sample solubilized in the Cadene cocktail followed by an aliquot of matrix (saturated) in FAPH. 3:2 Formic acid/acetonitrile (FA): As described above, protein samples were dissolved in 60% formic (41) Kim, Y. J.; Freas, A.; Fenselau, C. Anal. Chem. 2001, 73, 1544-1548. (42) Cohen, S. L.; Halaas, J. L.; Friedman, J. M.; Chait, B. T.; Bennet, L.; Chang, D.; Hecht, R.; Collins, F. Nature 1996, 382, 589. (43) Cohen, S. L.; Chait, B. T. Anal. Biochem. 1997, 247, 257-267. (44) Feick, R. G.; Shiozawa, J. A. Anal. Biochem. 1990, 187, 205-211. (45) Cadene, M.; Chait, B. T. Anal. Chem. 2000, 72, 5655-5658.


1117

acid/40% CH3CN and saturated matrix solutions were prepared from the same composition. Sample and matrix deposition were performed as described previously. The matrix 4-hydroxybenzylidene malononitrile was obtained from Lancaster Synthesis (Pelham, NH). All other matrixes were purchased from Sigma/Aldrich/Fluka (Milwaukee, WI), while TFA (sequencing grade) and formic acid (98-100%, extra pure) were obtained from Pierce (Rockford, IL) and Reidel-de Hae¨n/ Aldrich (Milwaukee, WI), respectively. PFOA was acquired from TCI America (Portland, OR), while HPLC-grade water, 2-propanol, and acetonitrile were purchased from EM Science (Darmstadt, Germany). D-Sorbitol was obtained from Acros Organics (Morris Planes, NJ). MALDI mass spectra were obtained on an Applied Biosystems Voyager DE-STR time-of-flight mass spectrometer (Framingham, MA) operated in linear mode with 337-nm irradiation and positive ion detection. Safety Considerations. Caution! Formic acid is dangerously caustic to skin. Proper eye and skin protection are required when handling concentrated formic acid solutions. Reasonable care should be exercised in handling 2-MBT, because the rubber allergen is widely distributed in everyday products. RESULTS AND DISCUSSION Three approaches to sample/matrix preparation were examined with regard to their impact on MALDI mass spectra of mixtures: (1) the solvent system 60% formic acid/40% CH3CN; (2) the detergent perfluorooctanoic acid, CF3(CF2)6COOH; (3) the matrix additive sorbitol. The solvent system 60% formic acid/40% CH3CN was modeled after the solvents employed with the most difficult-to-solubilize membrane proteins and in extracting proteins from polyacrylamide gels, while perfluorooctanoic acid, CF3(CF2)6COOH, was shown by Ishihama et al.46 to be an exceptionally ESI-MS compatible detergent. Curious as to the detergent’s compatibility with MALDIMS, we examined concentrations near its critical micelle concentration (cmc; 9.7 mM ) 0.4% w/v). While lower detergent concentrations are more likely to be compatible with matrix crystallization, solubilization efficacy will be highest at or above a detergent’s cmc. Addition of fucose or fructose to MALDI matrixes has previously been shown to benefit MALDI-MS. We similarly investigated the utility of another sugar, sorbitol. Sorbitol is a monosaccharide alditol; i.e., it lacks ketone and aldehyde functional groups. Solvent System 60% Formic Acid/40% CH3CN (FA). Formic acid’s excellent capacity to solubilize proteins is well known.47,48 Tarr and Crabb49 applied formic acid/2-propanol gradients to purify bovine rhodopsin and cytochrome P450 by reversed-phase HPLC, while Sheumack and Burley50 extended the solvent combination to the HPLC separation of egg yolk proteins. ESI-MS spectra of membrane proteins have been obtained with formic acid.38,39 Heretofor intractable protein aggregates, resistant to urea and SDS, such as huntingtin, β-amyloid, and Lewy body (46) Ishihama, Y.; Katayama, H.; Asakawa, N. Anal. Biochem. 2000, 287, 4554. (47) Heukeshoven, J.; Dernick, R. J. Chromatogr. 1982, 252, 241-254. (48) Leeder, J. D.; Marshall, R. C. Text. Res. J. 1982, 52, 245-249. (49) Tarr, G. E.; Crabb, J. W. Anal. Biochem. 1983, 131, 99-107. (50) Sheumack, D. D.; Burley, R. W. Anal. Biochem. 1988, 174, 548-551.

1118


fibrils (implicated in Huntington’s, Alzheimer’s, and Parkinson’s diseases, respectively)51-53 succumb to formic acid-based solvents. Moreover, Vanfleteren54 demonstrated the value of 70% solutions of formic acid for extracting intact proteins from polyacrylamide gels. Feick and Shiozawa44 introduced formic acid/acetonitrile/ 2-propanol/water, 50/25/15/10 (FAPH) to elute hydrophobic proteins from polyacrylamide gels and as a solvent for size exclusion chromatography. Protein elution with FAPH and formic acid/2-propanol/water (FPH) was combined with mass analysis by Cohen and Chait,42,43 Ehring et al.,40 and Righetti’s laboratory.55,56 Despite its utility, formic acid’s reputation as a general protein solvent is heavily tarnished. Tarr,49 Feick and Shiozawa,44 and others have listed potential problems as follows: (1) N-formylation of amino groups yielding blocked amino terminii, misidentification of lysine residues, or both; (2) O-formylation of serine and threonine residues; (3) reaction with the indole ring of tryptophan residues; (4) cleavage of acid-labile bonds; e.g., Asp-Pro. The discussion that follows evaluates the extent to which formic acid’s value as a solvent is compromised and establishes that previous concerns of N-formylation are unnecessary. Previous reports have listed amino group modification as a potential consequence of formic acid solubilization, but study of the literature provided no direct experimental support for this side reaction. Shively and colleagues attributed unsuccessful Edman sequence analyses to variable amino terminal blockage, thought to arise from pyridine formate buffers employed in HPLC purification, although no systematic study was performed.57,58 However, studies by Kienhuis et al.,59 Scheumack and Burley,50 Tarr,49 and Vanfleteren54 found no evidence of N-terminal or lysine formylation. Lysine formylation has only been documented for the more severe conditions employed in performic acid oxidation.60 O-Formylation, the esterification of serine and threonine residues, is a well-known reaction in proteins exposed to high concentrations of formic acid.47,50,59,61 CNBr cleavage, when performed with the classic protocol (incubation in 70% formic acid with added CNBr) partially modifies serine, threonine, and C-terminal homoserine side chains.62-64 However, Scheumack and Burley50 noted that O-formylation occurred slowly in their gradient (51) Hazeki, N.; Tukamoto, T.; Goto, J.; Kanazawa, I. Biochem. Biophys. Res. Commum. 2000, 277, 386-393. (52) Klunk, W. E.; Pettegrew, J. W. J. Neurochem. 1990, 54, 2050-2056. (53) Pollanen, M. S.; Bergeron, C.; Weyer, L. J. Neurochem. 1992, 58, 19531956. (54) Vanfleteren, J. R. Anal. Biochem. 1989, 385-390. (55) Piubelli, C.; Galvani, M.; Hamdan, M.; Domenici, E.; Righetti, P. G. Electrophoresis 2002, 23, 298-310. (56) Galvani, M.; Hamdan, M.; Righetti, P. G. Rapid Commun. Mass Spectrom. 2000, 14, 1889-1897. (57) Shively, J. E.; Hawke, D.; Jones, B. N. Anal. Biochem. 1982, 120, 312322. (58) Levy, W. P.; Rubinstein, M.; Shively, J.; Del, Valle, U.; Lai, C.-Y.; Moschera, J.; Brink, L.; Gerber, L.; Stein, S.; Pestka, S. Proc. Natl. Acad. Sci, U.S.A. 1981, 78, 6186-6190. (59) Kienhuis, H.; Blasse, G.; Matze, J. Nature 1959, 184, 2015-2016. (60) Dai, J.; Zhang, Y.; Wang, J.; Li, X.; Lu, Z.; Cai, Y.; Qian, X. Rapid Commun. Mass Spectrom. 2005, 19, 1130-1138. (61) Narita, K. J. Am. Chem. Soc. 1959, 81, 1751-1756. (62) Goodlett, D. R.; Armstrong, F. B.; Creech, R. J.; van Breemen, R. B. Anal. Biochem. 1990, 186, 116-120. (63) Andrews, P. C.; Allen, M. H.; Vestal, M. L.; Nelson, R. W. In Techniques in Protein Chemistry III; Villafranca, J. J., Ed.; Academic Press: San Diego, CA, 1992; pp 515-523. (64) Duewel, H. S.; Honek, J. F. J. Protein Chem. 1998, 17, 337-350.

of formic acid/2-propanol/CH3CN, suggesting that means exist to limit the extent of covalent modification; e.g., judicious selection of solvent composition and minimal exposure to the reactive species. By collecting their formic acid-rich HPLC eluent in an ice-cooled flask containing degassed water, Bollhagen et al.65 fractionated a peptide mixture without enduring formylation or oxidation, demonstrating that temperature control is also helpful. It is possible that the solvent’s reputation was sullied by amino terminii blocked instead by O f N acyl shifts experienced by N-terminal serine or threonine residues upon subsequent exposure to higher pH, e.g., those employed in phenyl isothiocyanate coupling of Edman degradation. Acyl groups are known to migrate reversibly between the initial residue’s β-hydroxy and R-amino functional groups.66 It should be noted that N f O acyl shifts, induced by acid, are useful for deblocking proteins,67 while the O f N acyl shifts operative at higher pH have been blamed for blocking amino terminii.66 The specter of deamidation has been a concern when using formic acid, although sequence analyses of proteins and peptides exposed to high concentrations of formic acid have shown negligible deamidation of Asn and Gln.44,54 That observation is reasonable, considering that succinimide intermediate-driven deamidation occurs at neutral and high pH.68-70 In general, deamidation is catalyzed by base, heat, and ionic strength and is retarded by organic solvents.68 Under acidic conditions (pH 1-2), slow Asn deamidation occurs via a tetrahedral intermediate, arising from a water molecule’s nucleophilic attack on an O-protonated side chain carbonyl.68 By excluding water, we are less likely to endure significant deamidation. In hydrolyzing proteins in a 70 °C vapor of aqueous pentafluoropropionic acid, Gobom et al.71 observed asparagine and glutamine deamidation only when the water content exceeded 50%, also suggesting that deamidation may be minimal under our solubilization conditions. An additional consideration is the less frequently observed, aspartic acid dehydration. Mediated by the same cyclic imide as base-catalyzed deamidation, Asp dehydration occurs at neutral or somewhat acidic pH;70 it is less likely to occur at the low pH conditions employed here. The cyclic imides are easily hydrolyzed by aqueous neutral or alkaline solutions affording either R- or β-aspartate, indistinguishable by mass. Under acidic conditions, however, the cyclic imides are more stable, suggesting that nonhydrolyzed products may be observable by their 18-Da decremented masses. The Asp-130 succinimide of methionyl human growth hormone has been isolated and characterized by mass spectrometry.72 Two concerns regarding tryptophan (Trp) residues’ exposure to formic acid are conversion to 1-formyltryptophan and indole oxidation. However, preparation of formyl tryptophan requires (65) Bollhagen, R.; Schmiedberger, M.; Grell, E. J. Chromatogr., A 1995, 711, 181-186. (66) Iwai, K.; Ando, T. Methods Enzymol. 1967, 11, 263-282. (67) Wellner, D.; Panneersel, C.; Horecker, B. L. Proc. Natl. Acad. Sci. U.S.A. 1990, 87, 1947-1949. (68) Daniel, R. M.; Dines, M.; Petach, H. H. Biochem. J. 1996, 317, 1-11. (69) Geiger, T.; Clarke, S. J. Biol. Chem. 1987, 262, 785-794. (70) Johnson, B. A.; Shirokawa, J. M.; Hancock, W. S.; Spellman, M. W.; Basa, L. J.; Aswad, D. W. J. Biol. Chem. 1989, 264, 1 4262-14271. (71) Gobom, J.; Mirgorodskaya, E.; Nordhoff, E.; Hojrup, P.; Roepstorff, P. Anal. Chem. 1999, 71, 919-927. (72) Teshima, G.; Stults, J. T.; Ling, V.; Canova-Davis, E. J. Biol. Chem. 1991, 266, 13544-13547.

harsher conditions; i.e., formic acid saturated in HCl,73,74 making the reversible modification unlikely in the absence of mineral acid. Oxidation of Trp and methionine (Met) are larger concerns. Cohen et al.75 observed Met oxidation (to Met-sulfoxide) from peptides mass-analyzed with freshly prepared R-cyano-4-hydroxycinnamic acid (HCCA) matrix in FPH. The extent of oxidation depended on the length of exposure to matrix solution and was independent of the source of formic acid. Attributed to trace peroxides present in the formic acid, oxidation was circumvented by aging HCCA-FPH matrix for at least 2 weeks prior to use or by spiking the matrix solution with 10 mM methionine. Some peptide bond cleavage must be accepted as a casualty of formic acid solubilization, but its impact is protein-dependent. Of 31 polypeptides tested by Feick and Shiozawa,44 only bovine albumin was markedly sensitive to acidic conditions. However, another study49 found cytochrome P450 to be particularly susceptible to formic acid, as all five Asp-Pro bonds were cleaved. Facile cleavages N- or C-terminal to aspartic acid, and especially between Asp-Pro residues are signatures of acid hydrolysis. Acid cleavage before or after Asp arises from nucleophilic attack by a side chain carboxylate onto a peptide backbone carbonyl, accessing a cyclic anhydride intermediate. To understand the potential impact of acid hydrolysis on our conditions, we turn to previous studies aimed at inducing hydrolysis. Li et al. exploited prolific aspartyl residue cleavages, achieved by incubating polyacrylamide gel slices in 2% formic acid at 108 °C, to identify proteins,76 However, no cleavage was observed at room temperature. Gobom et al.71 expanded on previous efforts by Tsugita, hydrolyzing lyophilized proteins or protein-embedded gel slices in a 70 °C vapor of 90% pentafluoropropionic acid/10% H2O for 1 h, affording internal backbone cleavages at Asp, Ser, Thr, and Gly and N- and C-terminal sequence ladders. These examples demonstrate the importance of temperature to effect useful hydrolysis rates. It is also notable that, at high acid concentrations, limited availability of water molecules to complete cleavage reactions decreases the hydrolysis rate, resulting in the need for long incubation times. Thus, a prescription to reduce hydrolysis is to prepare samples quickly50 and at low temperature.65 Analyzing Complex Mixtures. Panels A and B in Figures 2 compare MALDI-MS spectra below 15 kDa, obtained from lyophilized E. coli cells desorbed from sinapinic acid using standard and formic acid-based matrix conditions, respectively. Both conditions yield complex spectra. Figure 3 illustrates the medium and high molecular weight E. coli profiles detected from cells prepared with formic acid-based conditions. For the complete m/z range examined, standard matrix conditions displayed 94 ions attributed to singly charged species, 20 doubly charged ions, and 13 ions ascribed to sinapinic acid adducts. Formic acid-based conditions revealed 119 ions attributed to singly charged species, 20 doubly charged ions, and 9 ascribed to sinapinic acid adducts. (Evaluation was performed manually and included all features reproduced in multiple spectra for a given preparation.) The (73) Holmgren, A. Eur. J. Biochem. 1972, 26, 528-534. (74) Lundblad, R. L.; Noyes, C. M. Chemical Reagents for Protein Modification; CRC Press: Boca Raton, FL, 1984; Vol. 2. (75) Cohen, S. L.; Ward, G.; Tsai, P.-K. Proc. 47th ASMS Conf. Mass. Spectrom. Allied Topics, Dallas, TX, June 13-17. 1999. (76) Li, A.; Sowder, II, R. C.; Henderson, L. E.; Moore, S. P.; Garfinkel, D. J.; Fisher, R. J. Anal. Chem. 2001, 73, 5395-5402.


1119

Figure 2. MALDI mass spectra obtained from lyophilized E. coli cells desorbed from sinapinic acid using (A) standard and (B) formic acidbased matrix conditions. (C) Expansion of B from m/z 7000 to 7500 revealing heterogeneity characteristic of lipid acylation in lpp (P69776).

spectra are strikingly different; no more than 10 protein ions are common to both. We ascribe many of the peaks illustrated in Figure 2B to lipoproteins, based on their pattern of heterogeneity (Figure 2C) and on corresponding masses in predicted lipoproteins or in isolated E. coli lipoproteins analyzed previously by us and others.77,78 N-Terminal cysteines of mature, secreted proteins are predicted to be lipid-modified based on the “lipobox” motif within their signal peptide sequence, by automated prediction tools, or both.79 Swiss-Prot database (release 42) annotates 110 E. coli proteins as potentially lipid acylated; of these proteins, 11 are not found in K12 strains, and 17 are encoded on plasmids, leaving 82 to be considered. Additional lipoprotein predictions are provided by Gonnet et al.79

The 7.2-kDa protein ions apparent in Figure 2B,C correspond to the protein lpp (Swiss-Prot entry P69776), better known as Braun’s lipoprotein, murein lipoprotein, or major outer membrane lipoprotein.78 We have studied this protein previously from 1- and 2-D gels and are familiar with its acylation pattern, oxidation sensitivity, solubility, mass, and other characteristics. Based on the pattern of lipid acylation measured both here and from dried isoelectric focusing gels77 for the intact proteins lpp and osmE (P69776 and P23933, respectively), mature lipoprotein masses could be calculated by adding the lipid mass (experimentally determined as 788 Da for the dominant acyl species).16 At a stunning 750,000 copies per cell, 2/3 of which are free (not covalently bound to peptidoglycan),80 Braun’s lipoprotein is the major outer membrane protein and among the most abundant proteins in the E. coli cell. Its absence from MALDI spectra obtained from standard matrix preparations is quite remarkable, until one considers the hydrophobicity of lipid acylation. That lpp is readily observed with standard sinapinic acid matrix conditions in the virtual 2-D gel experiment,77 but not in the whole cell protein experiment, hints at the interplay (for a given solvent system) between a protein’s solubility, the solubility of other protein and non-protein components, and how those two factors influence incorporation within matrix crystals. Previously, Zhang et al. observed a strong, unusually broad signal at ∼7000 Da from an E. coli pellet extracted with 0.2% SDS/ 0.1% TFA after removal of water-soluble proteins.81 We suspect that by including SDS, they, too, observed Braun’s lipoprotein. Another series of ions (major species m/z 3410) was consistent with the cell envelope lipoprotein entericidin B (ecnB, P56549), for which a mass of 3409.5 Da is calculated for the major lipid acylated species. Interestingly, mass spectra did not reveal entericidin A (ecnA, P56548), the antidote protein to bacteriolytic entericidin B. Chromosomally encoded entericidin A and B constitute an antidote/toxin gene pair that together modulates membrane stability; thus, we expected both to be present.82 Ion

(77) Ogorzalek, Loo, R. R.; Cavalcoli, J. D.; VanBogelen, R. A.; Mitchell, C.; Loo, J. A.; Moldover, B.; Andrews, P. C. Anal. Chem. 2001, 73, 4063-4070. (78) Pittenauer, E.; Quintela, J. C.; Schmid, E. R.; Allmaier, G.; Paulus, G.; de Pedro, M. A. J. Am. Soc. Mass Spectrom. 1995, 6, 892-905. (79) Gonnet, P.; Rudd, K. E.; Lisacek, F. Proteomics 2004, 4, 1597-1613.

(80) Inouye, S.; Takeishi, K.; Lee, N.; DeMartini, M.; Hirashima, A.; Inouye, M. J. Bacteriol. 1976, 127, 555-563. (81) Zhang, N.; Doucette, A.; Li, L. Anal. Chem. 2001, 73, 2968-2975. (82) Bishop, R. E.; Leskiw, B. K.; Hodges, R. S.; Kay, C. M.; Weiner, J. H. J. Mol. Biol. 1998, 280, 583-596.

Figure 3. Medium (A) and high (B) m/z range E. coli ions detected from a formic acid-based preparation of sinapinic acid.

1120 Analytical Chemistry, Vol. 79, No. 3, February 1, 2007

suppression may play some role in ecnA’s absence. Although the first 23 amino acids differ at only 6 positions in these two small proteins, with ecnB possessing 4 additional amino acids, a pI of 4 is predicted for lipid-acylated ecnB, while a pI of 7 is predicted for ecnA. Also dominant in formic acid-based conditions is an 8.3-kDa ion, absent or very weak under standard conditions. Searching 8282 ( 4 Da returned no potential proteins, while searching 8413 ( 4 Da (to accommodate potential initiator Met excision) returned three proteins: yeeT (P64521, 8411 Da), ypeB (P56604, 8415 Da), and HNSB2 (P64467, formerly ydgT, 8417 Da). However, yeeT and ypeB are not methionine aminopeptidase substrates, but HNSB2 could be the observed ion. Nevertheless, the dominance of the 8282 Da ion under formic acid-based conditions, and its absence under standard conditions, led us to believe that the protein responsible for this ion should be extremely abundant and likely reputed as challenging to analyze. Consequently, we performed additional searches assuming +28 (formyl) or +42 (acetyl) modifications. Those additional searches suggested atpL (P68699), the ATP synthase C chain. This well-characterized integral membrane protein, one of the subunits comprising proton channels in E. coli, has a formylated amino terminus. Also referred to as the “ATPase proteolipid” due to its unusual solubility profile, mature atpL has Mr 8284 Da and an isoelectric point of 4.10. The very hydrophobic and very abundant E. coli protein is insoluble in water and was first purified from chloroform/methanol mixtures. Thus, we attribute the dominant 8282 Da ion observed by MALDI only with formic acid-based conditions to atpL, or possibly, HNSB2. Major E. coli K12 ions observed from this formic acid-based preparation are listed in Table S-1 (Supporting Information), along with potential assignments, acquired by searching the Swiss Prot database (versions 46 and higher) by intact molecular weight. (Potential assignments for ions observed from standard preparations have been presented previously.10) Mass tolerances employed were 0.05% for proteins below 10 kDa, 0.1% between 10 and 20 kDa, and 0.2% for proteins larger than 20 kDa. Plasmidborne proteins and proteins from non-K12 strains were excluded from consideration. Employing TagIdent83 over broad isoelectric point ranges, searches were performed in parallel, to accommodate initiator Met retention and excision. However, for proteins in which amino-terminal residues had been verified experimentally (e.g., by Edman degradation), only the verified Met status was considered. Other matrixes were also prepared with high concentrations of formic acid; e.g., Figure 4 illustrates spectra obtained from ferulic acid preparations. When applied in conventional solvents, ferulic acid’s detection of high mass proteins exceeds that of sinapinic acid; it has been employed advantageously with 17% formic acid to routinely produce protein signals beyond 20 kDa from Gram-negative and -positive bacteria.84-86 Its high mass advantage was similarly evident at higher formic acid concentrations, from which proteins to 100 kDa in size were observed. (83) Wilkins, M. R.; Gasteiger, E.; Sanchez, J.-C.; Appel, R. D.; Hochstrasser, D. F. Curr. Biol. 1996, 6, 1543-1544. (84) Nilsson, C. L. Rapid Commun. Mass Spectrom. 1999, 13, 1067-1071. (85) Madonna, A. J.; Basile, F.; Ferrer, I.; Meetani, M. A.; Rees, J. C.; Voorhees, K. J. Rapid Commun. Mass Spectrom. 2000, 14, 2220-2229. (86) Meetani, M. A.; Voorhees, K. J. J. Am. Soc. Mass Spectrom. 2005, 16, 14221426.

Figure 4. MALDI-MS of E. coli cells desorbed from ferulic acid using formic acid-based matrix conditions.

Figure 5. MALDI-MS of E. coli obtained with the matrix MBT in 60% formic acid/40% CH3CN.

Highly soluble in hydrophobic solvents, 2-MBT was also prepared with formic acid/CH3CN. Figures 5 and S-1 (Supporting Information) display spectra obtained from E. coli cells and mouse heart mitochondria, respectively, employing 2-MBT. Mercaptobenzothiazole required lower laser fluence than sinapinic acid for desorption, while ferulic acid consistently revealed the highest mass proteins, as noted previously and evident in Figures 2-5.84,86,87 Other matrixes examined with formic acid included DHB and 4-hydroxybenzylidene malononitrile, both of which yielded spectra similar to those discussed, but without advantages. We rely on these matrix formulations as described above to supplement standard matrix conditions when profiling complex protein samples by MALDI-MS. We also applied these methods to tryptic digest samples in order to alter the selection of peptides displayed, reflective of previous efforts by Cohen and col(87) Bornsen, K. O. In Protein and Peptide Analysis: New Mass Spectrometric Applications; Chapman, J. R., Ed.; Humana: Totowa, NJ, 2000; Vol. 146, pp 387-404.


1121

Figure 6. MALDI mass spectra (8150-10000 m/z) from lyophilized E. coli cells desorbed from ferulic acid and 3:2 formic acid/CH3CN (A) immediately after deposition, and (B) several hours after deposition. Bars indicate patterns of formyl addition (+28 Da). The extent of formylation is protein-dependent.

leagues.75,88 Although previously unobserved tryptic peptides were exposed by our conditions, the sensitivity for peptides below 3000 Da in mass was reduced by at least 1 order of magnitude. Clearly our methods extend the range of both peptides and proteins accessed by MALDI profiling, but appear most suited to profiling protein mixtures. Protein Modification Is Time- and Solvent-Dependent. Samples exposed to high concentrations of formic acid are best analyzed shortly after formic acid addition, because the number of formate esters increases with time. Increased formylation (+28Da adducts) is apparent when comparing the Figure 6 spectra of E. coli cells analyzed immediately and several hours after solubilization and spotting. The reactivities of formic acid-based cocktails were evaluated by dissolving lyophilized protein standards in the compositions and depositing them immediately with matrix onto the sample stage. Mass analysis was performed at several time points after deposition. In addition, solubilized proteins were incubated at room temperature for 21 h, after which sample and matrix were deposited and analyzed. The four compositions, described in the Experimental Section and referred to as Kim, Cohen, Feick, and Cadene, are compared to the FA composition introduced in this study. Ehring et al.40 noted that protein formylation continues even after sample/matrix crystallization. For example, bovine ubiquitin (3 pmol) analyzed immediately after deposition in 3:2 FA with either sinapinic acid or MBT matrix displayed no formylation, yet analysis of the same spots 1.5 h later revealed approximately 68% unmodified, 18% formylated (+28 Da), and 14% of +42 Daincremented ubiquitin (oxidation and formylation), based on peak heights (Table S-2, Supporting Information). Interestingly, no +16Da contribution from oxidation alone was observed. Measurements performed with Kim’s and Feick’s compositions 1.5 h after deposition displayed similar amounts of unmodified protein, but with reproducibly more +42 than +28 Da modification present. (88) Ward, G.; Cohen, S. L.; Tsai, P. K. Proc. 48th ASMS Conf. Mass Spectrom. Allied Topics, Long Beach, CA, June 11-15, 2000.


Figure 7. MALDI mass spectra acquired from ubiquitin deposited 1 day after solubilization in (A) 60% formic acid/40% CH3CN, (B) Feick cocktail, 50% formic acid/25% CH3CN/15% 2-propanol/10% H2O, (C) Cadene cocktail, 50% formic acid/33% 2-propanol/17% H2O, and (D) Cohen cocktail, 50% H2O/33% 2-propanol/17% formic acid. Symbols signify peaks corresponding to multiply formylated forms of ubiquitin with (]) no oxidation (+0 Da), (b) single oxidation (+16 Da), (O) double oxidation (+32 Da), and (1) triple oxidation (+48 Da).

Again, there was little +16-Da-modified protein, hinting that the rate of oxidation is enhanced in formylated ubiquitin. Interestingly, no formylation was observed 1.5 h after matrix deposition with the Cohen cocktail. For all proteins examined, the extent of formylation was found to be independent of matrix, depending only on protein, cocktail, and length of incubation. Ubiquitin, deposited 1 day after solubilization in 3:2 FA, revealed proteins formylated (+28) from 1 to 9 times, and partially oxidized as revealed by a second ladder of multiply formylated peaks, offset by 16 Da (Figure 7a). Samples deposited with sinapinic acid, ferulic acid, and MBT in 3:2 FA displayed similar extents of formylation and oxidation. A similar pattern was obtained from the Feick formulation, but with slightly less formylation (Figure 7B). With the Cadene cocktail, equal in formic acid content to that of Feick, but lacking CH3CN, formylation was greatly reduced (Figure 7C). Dominant ions reflect singly oxidized ubiquitin with 0 or 1 formyl groups attached. Less intense ions reflect singly oxidized, doubly formylated ubiquitin (+72 Da) and the doubly oxidized, singly formylated protein (+60 Da). In spectra acquired with the Cohen mix (Figure 7D), the major peaks correspond to singly (+16) and doubly (+32) oxidized ubiquitin. Peaks consistent with singly formylated species oxidatively incremented by 32 and 48 Da in mass were also observed. This preparation showed less formylation than the Cadene mix, but more oxidation. Ubiquitin’s single methionine is expected to oxidize easily to the sulfoxide (+16 Da), whereas oxidation to the sulfone (+32) is not expected, generally requiring harsher conditions than those employed here (e.g., performic acid oxidation). Tryptophan and

tyrosine residues, of which ubiquitin possesses 0 and 1, respectively, are also prone to oxidation. Tyrosine residues were shown to convert to a form 34 Da heavier on exposure to the stronger oxidizer performic acid.89 Ubiquitin’s 3 serines and 7 threonines offer 10 hydroxyl sites for formylation, slightly more than the maximum number (9) of formyl adducts observed here. Horse heart myoglobin was also analyzed 1.5 h after solubilization and deposition with matrix. Again, the 3:2 mixture resulted in the most modification (addition of 0-2 formyl groups), while the Cohen recipe showed essentially none. In this same time frame, with cytochrome c, the 3:2 and Feick formulations yielded 90% unmodified protein and 10% M + 28, while no modification was observed with the Kim formulation. The propensity for protein formylation among the cocktails examined here is

Cohen < Cadene < Kim < Feick < 3:2 FA

That both the Cohen and Cadene mixtures avoid CH3CN seems key to their reduced rate of formylation, as discussed below. Although oxidation was apparent with many of these formulations and most pronounced in the Cohen mix, it can likely be circumvented by including scavengers, such as methionine, tryptophan, phenol, thioglycolic acid, thiodiglycol, or 2-mercaptoethanol. The extent of formylation increases with length of exposure to formic acid and continues for samples crystallized in matrix, and even on sample targets within the mass spectrometer. More formylation is observed from cocktails containing higher concentrations of formic acid and from the acetonitrile-containing cocktails examined here (3:2 FA, Kim, and Feick). We attribute the latter effect to the small equilibrium constant of acid-catalyzed Fischer esterifications, magnifying the importance of the formic acid/hydroxyl ratio. Given that esterifications can be driven to completion by removing water, we should expect that a solubilization cocktail with only formic acid and acetonitrile would promote formylation. Detergent PFOA. For membrane proteins, detergent inclusion is often essential to their isolation and purification. It is also important for their mass analysis.29-31,34,37,38,41,87 Here we study the ability of a detergent to enhance detection of protein components from complex mixtures. Similar efforts have been applied previously to enzymatic digest products (e.g., inclusion of octyl glucoside90 or sodium dodecyl sulfate91 to reveal peptides >2 kDa for comprehensive peptide mapping). Meetani and Voorhees86 evaluated nonionic saccharide, N,N-dimethyldodecylammonium N-oxide and zwittergent 3-12 detergents with five MALDI matrix systems for whole cell bacterial analyses, while Nilsson84 reported on N-octyl β-D-glucoside’s utility for Helicobacter pylori analyses. Van Adrichem et al.92 recommended the nonionic detergents Triton X-100 or Tween 80 as additions for MALDI-MS protein fingerprinting of mammalian cells and culture supernatants. (89) McLachlin, D. T.; Chait, B. T. Anal. Chem. 2003, 65, 6826-6836. (90) Cohen, S. L.; Chait, B. T. Anal. Chem. 1996, 68, 31-37. (91) Tummala, R.; Green-Chruch, K. B.; Limbach, P. A. J. Am. Soc. Mass Spectrom. 2005, 16, 1438-1446. (92) van Adrichem, J. H. M.; Boernsen, K. O.; Conzelmann, H.; Gass, M. A. S.; Eppenberger, H.; Kresbach, G. M.; Ehrat, M.; Leist, C. H. Anal. Chem. 1998, 70, 923-930.

We were also interested in examining detergent concentrations near the cmc, hoping to keep hydrophobic proteins soluble for a greater length of time during matrix crystallization, potentially enhancing their matrix incorporation. We hypothesized that, relative to detergent-free conditions, dramatic differences in mixture spectra could be obtained from detergent concentrations exceeding the cmc. For the detergent examined here, PFOA, the cmc is 9.7 mM (0.4% w/v). It should be noted that the tradeoff in employing such high PFOA concentrations is that matrix deposition becomes extraordinarily difficult and efforts to obtain crystals capable of yielding good MALDI spectra can be frustrating. Figure S-2 (Supporting Information) compares spectra obtained from M. acetivorans cells with and without PFOA-containing matrix. Remarkably, only two major peaks are common to both spectra! Below 15 kDa, 83 singly charged ions (excluding matrix adducts) were observed from the standard preparation, while 26 were observed from the PFOA preparation. Above 15 kDa, 8 singly charged ions were observed from the standard preparation, and 33 were observed from the PFOA mixture. The standard preparation yielded ions to 28 kDa, while the PFOA-based medium revealed ions to 64 kDa. The PFOA preparation excelled at higher m/z in the intensity and number of M. acetivorans ions detected. Figure 8 compares MALDI-MS profiles for E. coli lyophilized cells analyzed with and without PFOA. Figure 8A was obtained without detergent (but with TFA), while spectra B-D were acquired from a PFOA-containing mixture. Similar results were obtained with three detergent concentrations: 1, 0.5, and 0.25% PFOA, but lower concentrations yielded spectra resembling those obtained without detergent. The two preparations complement one another, as is apparent from only six ions being common to spectra A and B. Table S-3 (Supporting Information) lists the major ions observed from the PFOA-based preparation, along with potential assignments, acquired by searching the Swiss Prot database by intact molecular weight, as described previously. Lipid-acylated and other hydrophobic proteins from E. coli whole cells were observed from the detergent-containing mixtures. Ions were easily attributed to lipid-acylated proteins based on their characteristic pattern of heterogeneity, as described earlier. Correlating the observed molecular weights to those calculated for the 82 nonplasmid encoded E. coli K12 lipoproteins and to those predicted by Gonnet et al.,79 we ascribed 3.4-, 7.2-, 10.8-, 5.9-, and 7.0-kDa ions to the membrane proteins ecnB (P56549), lpp (P69776), and osmE (P23933) and to hypothetical membrane proteins yifL (P39166) and ygdI (P65292), respectively. Lipid acylation in ygdI has been verified experimentally by Matsuyama.79 We were unable to ascribe a lipoprotein to the 6.9-kDa cluster of ions (Figure 8D), despite their apparent lipoprotein signature. This species could arise from oxidative cleavage of the unsaturated lipid in lpp. Cohen has demonstrated such cleavage for phospholipid standards air-dried on MALDI targets.93 Alternatively, this failure may arise because (1) the peaks are unrelated and correspond to distinct proteins, (2) the peaks correspond to a known or predicted lipoprotein that has been C-terminally truncated (to retain the N-terminal lipid acylation), or (3) the peaks correspond to a so-far unpredicted lipoprotein. Note that three of (93) Cohen, S. L. Anal. Chem. 2006, 78, 4352-4362.


1123

Figure 8. MALDI mass spectra of E. coli K12. (A) Lyophilized cells suspended in H2O (∼2.5 mg of dry cell/mL) were mixed with an equal volume of sinapinic acid (saturated) in 1:2 CH3CN/0.15% TFA (v/v). An aliquot of the mixture was spotted onto the MALDI probe. (B) Lyophilized cells (∼2.5 mg of dry cells/mL) suspended in 1% PFOA (w/v) were mixed with an equal volume of sinapinic acid (saturated) in 1:2 CH3CN/1% PFOA (w/v). An aliquot of the cell/PFOA/matrix mixture was spotted onto the MALDI probe, immediately followed by an equal volume of standard matrix (sinapinic acid (saturated) in 1:2 CH3CN/0.15% TFA (v/v). (C) Lower m/z range for 1% PFOA condition (same preparation as in B). (D) Expanded view of 5.8-7.1-kDa range in 1% PFOA condition (same preparation as in B). The ??-labeled species is discussed in the text.

the five lipoproteins (ecnB, lpp, osmE) found here were also observed from formic acid preparations. In contrast to the dominant atpL ion observed with formic acid preparations, only a weak ion was observed at m/z 8285 for the putative proteolipid P68699. Acyl carrier protein (ACP, P0A6A8), with 1.5 × 106 molecules/ cell, is the most abundant protein in E. coli.94 Despite this cytosolic protein’s abundance, it is rarely observed as an intact protein by MALDI-MS from unfractionated whole cell mixtures, although it has been detected in whole cell trypsin digests of E. coli.95 apoACP, a 77-residue protein with Mr 8508 Da, and pI 3.98, requires a phosphopantetheine prosthetic group to be covalently bound to a serine residue in order to function. The mature protein, holoACP, has Mr 8847 Da, consistent with the base peak of Figure 8B. A search of E. coli K12 nonplasmid proteins predicted to weigh 8847 Da yielded only one other candidate protein, holE (DNA polymerase III, P0ABS8, Mr 8846 Da, pI 9.15). A search of Mr 8978 (accounting for unannotated initiator methionine excision) suggested no other proteins. We believe that m/z 8847 most likely arises from ACP, because of the protein’s abundance and unusual properties, as discussed below. Our previous MALDI-MS analyses of isolated holo-ACP (data not shown), confirmed its mass and demonstrated that the protein yields excellent MALDI-MS spectra from standard preparations and protein loads. These observations, in concert with the protein’s extreme abundance, suggest that its absence from spectra of whole cell mixtures reflects suppression. Acyl carrier protein is known to precipitate readily from dilute acetic acid solutions.96 Thus, protocols solubilizing cell pellets in aqueous acid may precipitate ACP prior to sample deposition, and TFA-containing matrixes may exacerbate suppression effects when mixed with cell suspensions. In precipitating quickly from acidic matrix/ protein mixtures, ACP may suffer reduced matrix incorporation relative to that of other proteins. The strong dependence of ACP signals on solution conditions and on the detergent additive may also reflect its atypical solubility profile, namely, that it is unusually soluble in isopropanol.96 (94) Matthews, C. K.; Van Holde, K. E. Biochemistry; Benjamin/Cummings Publishing: Redwood City, CA, 1990. (95) Pribil, P.; Fenselau, C. Anal. Chem. 2005, 77, 6092-6095. (96) Pugh, E. L.; Kates, M. Biochim. Biophys. Acta 1994, 1196, 38-44.


The ion observed at m/z 8892 is also of interest. Intact mass searches suggest three candidate proteins, ydhI (P64471), yahM (P75692), and yjbO (P32696), but yjbO matches only with Met excision, disfavored for its Met-Leu amino terminus. The proteins ydhI and yahM require initiator Met retention, expected for ydhI (Met-Lys) and, unlikely, but possible for the yahM Met-Ala amino terminus. Based on our previous discussion of acyl carrier protein, we also propose acetyl-acyl carrier protein (acetyl-ACP, P0A6A8) as a candidate for m/z 8892, with acetylation of the phosphopantetheine sulfur. Thioester linkages of acyl-ACPs are expected to be stable under the MALDI preparation conditions employed here and were confirmed in our previous MALDI-MS analyses of acetyl- and malonyl-ACP generated in vitro (data not shown). Acetyl-ACP accumulates to high levels when fatty acid biosynthesis is down regulated.97 In summary, m/z 8892 is proposed to arise from ydhI (P64471) and/or acetyl-ACP (P0A6A8); although yahM (P75692) cannot be completely excluded from consideration. These conditions for detecting hydrophobic proteins revealed ∼25 proteins larger than 25 kDa, with the largest protein detected being 95 kDa. In contrast, a standard preparation of the same sample yielded a single 35-kDa protein in that size range, although with superior mass resolution. Matrix Additive Sorbitol. Saccharide matrix additives have been described in several reports. Wilkins and colleagues98 included fucose in matrix preparations mass-analyzed by Fourier transform-ion cyclotron resonance mass spectrometers. In developing protocols for quantitative analyses, Gusev et al.28 found a fucose comatrix to enhance spectral reproducibility, signal intensity, and mass resolution. Moreover, it enhanced signal persistence (number of laser shots delivering useful data from a given laser position). Distler and Allison99 reported similar benefits in analyses of complex oligonucleotide mixtures, as did Sequenom researchers.100 Addition of fucose reduced fragmentation in (97) Heath, R. J.; Rock, C. O. J. Biol. Chem. 1995, 270, 15531-15538. (98) Koester, C.; Castoro, J. A.; Wilkins, C. L. J. Am. Chem. Soc. 1992, 114, 7572-7574. (99) Distler, A. M.; Allison, J. Anal. Chem. 2001, 73, 5000-5003. (100) Shahgholi, M.; Garcia, B. A.; Chiu, N. H. L.; Heaney, P. J.; Tang, K. Nucleic Acids Res. 2001, 29, e91/91-e91./10.

PAMAM dendrimers101 and oligonucleotides.99,100 Fucose and fructose were found to be the most effective matrix additives for DNA analysis.100 Gluckmann and Karas102 observed saccharide additives that markedly increased the initial velocity of insulin (M + H)+ ions desorbed from DHB, as well as others that reduced velocities. Fucose is a monosaccharide aldose, i.e., an aldehyde-containing reducing sugar, while fructose is a monosaccharide ketose, (ketone-containing reducing sugar). Aldoses and ketoses can react with amines (e.g., proteins) to form Schiff bases, and both types of sugars can cyclize. In contrast, sorbitol is a monosaccharide alditol corresponding to reduced glucose (C6H14O6). Alditols are polyhydroxy sugars (nonreducing) that do not cyclize. The reduced reactivity of sorbitol, relative to fucose and fructose, will be advantageous when it is employed as an additive. Sorbitol enhances medium and higher mass performance, as illustrated in Figure S-2 (Supporting Information), which compares E. coli spectra with and without sorbitol. Sorbitol, 1.5% (w/v) promoted homogeneous crystallization, the benefit of which was especially noticeable at lower protein concentrations, reducing the search for “sweet spots”. With adequate protein concentrations and freshly prepared matrix, little benefit was observed from sorbitol inclusion, although there was no downside. However, for lower protein concentrations or older sinapinic acid matrix preparations, it was extremely helpful, delivering successful analyses otherwise unattainable. Sorbitol should be useful to laboratories tasked with analyzing protein mixtures at variable concentrations and with variable contaminants. For “one-off” sample analysis, saccharide addition may not offer any enhancement over careful sample preparation. Nor is not clear if it benefits analyses of purified proteins presented in MALDI-compatible solutions. However, experiments profiling complex mixtures seeking better reproducibility, automation, or both should consider saccharide addition. CONCLUSIONS While most work evaluating the role of different matrixes, solvents, and preparation methods in MALDI-MS analyses of complex protein mixtures has focused on detecting the maximum number of ions, we describe methods to detect a different subset of proteins from mixtures. Analyses of intact proteins from E. coli K12 cells yielded 94 singly charged ions from standard sinapinic acid matrix versus 119 from a formic acid-based preparation, yet the spectra were strikingly different with fewer than 10 proteins common to both. Notably, the formic acid preparations revealed numerous lipoproteins and an 8282 m/z ion, proposed to be the abundant, water-insoluble ATPase proteolipid, atpL (P68699), absent under standard MS conditions. More β-hydroxyl side chain formylation was observed from cocktails containing higher concentrations of formic acid and CH3CN. The propensity for (101) Peterson, J.; Allikmaa, V.; Subbi, J.; Pehk, T.; Lopp, M. Eur. Polym. J. 2002, 39, 33-42. (102) Gluckmann, M.; Karas, M. J. Mass Spectrom. 1999, 34, 467-477.

formylation was found to be, Cohen43,75 < Cadene45 < Kim41 < Feick44 < FA. Comparing standard and PFOA-based matrix preparations of E. coli whole cell mixtures revealed only six peaks common to both preparations. E. coli lipoproteins were readily revealed from the PFOA-containing matrixes, as was a dominant ion at 8847 m/z, which we attribute to ACP (P0A6A8). Despite these encouraging results, we found that employing the high PFOA concentrations needed to exceed its cmc (0.4% w/v) could frustrate efforts to obtain crystals capable of yielding good MALDI spectra, underscoring the need for other surfactants or preparations. Adding 1.5% w/v sorbitol to sinapinic acid matrix promoted homogeneous crystallization and enhanced high mass ions. It was beneficial in analyzing dilute samples, but with adequate protein concentrations, little benefit was observed. Because alditol saccharides cannot react with amines to form Schiff bases, sorbitol may be preferable to fucose or fructose as a matrix additive. We believe that (1) the striking differences between bacterial spectra obtained under standard conditions from those obtained with formic acid or PFOA, as well as (2) our observation that, when purified, lpp and ACP are easily analyzed from a standard sinapinic acid preparation, point out the role of solubility in ion suppression, as proteins compete to be incorporated into matrix crystals. The matrix preparations introduced here access different proteins than those generally revealed by whole cell profiling experiments, increasing the number of data points available from which to ascribe (or confirm) bacterial identity, and increasing the range of components available to be monitored or screened. ACKNOWLEDGMENT We thank Robert P. Gunsalus, Unmi Kim, and Kristy Sandler of the University of CaliforniasLos Angeles Department of Immunology & Microbiology for providing the M. acetivorans cultures employed in these studies. We are also grateful to Beniam Berhane, Thomas Vondriska, and Peipei Ping of the University of CaliforniasLos Angeles Department of Physiology for providing murine mitochondria. R.R.O.L. acknowledges Philip Andrews (University of Michigan) for inspiration and for his appreciation of formic acid and detergents. The authors acknowledge support from the U.S. Department of Energy for funding the UCLAsDOE Laboratory of Genomics & Proteomics (DEFC03-87 ER60615), the David Geffen School of Medicine at UCLA, and the UCLA Molecular Biology Institute. The UCLA Mass Spectrometry and Proteomics Technology Center was established by a generous gift from the W.M. Keck Foundation. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review October 10, 2006. Accepted October 29, 2006. AC061916C


1125

Ionization-Mass Spectrometry of

Recommend Documents