Solution-Phase Chelators for Suppressing Nonspecific Protein−Metal

Anal. Chem. 2009, 81, 5008–5015

Solution-Phase Chelators for Suppressing Nonspecific Protein-Metal Interactions in Electrospray Mass Spectrometry Jingxi Pan,†,‡ Kun Xu,§ Xiaoda Yang,§ Wing-Yiu Choy,† and Lars Konermann*,† Departments of Chemistry and Biochemistry, The University of Western Ontario, London, Ontario, N6A 5B7, Canada, State Key Laboratory of Fine Chemicals, Dalian University of Technology, Dalian 116012, People’s Republic of China, and School of Pharmaceutical Sciences and National Research Laboratories of Natural and Biomimetic Drugs, Peking University, Beijing 100083, People’s Republic of China Protein-metal complexes may be transferred from solution into the gas phase by electrospray ionization (ESI), such that they can be directly analyzed by mass spectrometry (MS). In principle, therefore, ESI-MS represents a simple and elegant approach for gaining insights into the binding stoichiometry and affinity of these assemblies. Unfortunately, the formation of nonspecific metal adducts during ESI can be a severe problem, often leading to binding levels that are dramatically higher than those in bulk solution. Focusing on several calcium binding proteins as test systems, this work explores the suitability of different salts to serve as metal source. Despite their widespread use in previous ESI-MS studies, calcium chloride and acetate induce extensive nonspecific adduction. In contrast, much lower levels of artifactual metal binding are observed in the presence of calcium tartrate. In the case of high and intermediate affinity proteins, the resulting ESI-MS data are in excellent agreement with the calcium binding behavior in bulk solution. The situation is more challenging when studying proteins with very low affinities, but in the presence of tartrate qualitative information on protein-metal interactions can still be obtained. The beneficial effects of tartrate also extend to zinc binding experiments. This work does not directly explore the mechanism by which tartrate suppresses nonspecific metalation. However, it seems likely that weak chelators such as tartrate sequester metal ions within rapidly shrinking droplets during the final stages of ESI, thereby reducing nonspecific metal adduction to protein carboxylates. The use of tartrate and possibly other weak chelators will greatly enhance the reliability of future ESIMS studies on the interactions of proteins with divalent metal ions. Protein-ligand interactions play a central role in numerous biological processes. Many of the underlying binding events are based on relatively weak noncovalent contacts, where a protein * To whom correspondence should be addressed. Phone: 519-661-2111, ext 86313. Fax: 519-661-3022. E-mail: [email protected]. † The University of Western Ontario. ‡ Dalian University of Technology. § Peking University.

5008

Analytical Chemistry, Vol. 81, No. 12, June 15, 2009

receptor interacts with partners that can range from low molecular weight species all the way to large biopolymers. Various calorimetric, spectroscopic, and chromatographic techniques have been applied for identifying these binding partners and for measuring their affinities.1-5 Electrospray ionization (ESI) mass spectrometry (MS)6 plays a key role for monitoring protein-ligand interactions. ESI-MS may be used for indirect approaches such as hydrogen/deuterium exchange7,8 or covalent labeling techniques9,10 that report on structural changes of a protein in response to ligand binding. In addition, ESI-MS offers the opportunity to monitor protein-ligand interactions directly, i.e., by transferring the corresponding complexes into the gas phase such that they can be detected in the mass spectrum.11,12 Binding stoichiometries can be determined from the measured mass,13-19 and affinity estimates may be obtained from the ion intensity ratios of free versus bound receptor.20-24 The simple mix-and-measure nature of these experi(1) Schriemer, D. C. Anal. Chem. 2004, 76, 441A–448A. (2) Schermann, S. M.; Simmons, D. A.; Konermann, L. Exp. Rev. Proteomics 2005, 2, 475–485. (3) Entzeroth, M. Curr. Opin. Pharmacol. 2003, 3, 522–529. (4) Roda, A.; Guardigli, M.; Pasini, P.; Mirasoli, M. Anal. Bioanal. Chem. 2003, 377, 826–833. (5) Auer, M.; Moore, K. J.; Meyer-Almes, F. J.; Guenther, R.; Pope, A. J.; Stoeckli, K. A. Drug Discovery Today 1998, 3, 457–565. (6) Fenn, J. B. Angew. Chem., Int. Ed. 2003, 42, 3871–3894. (7) Powell, K. D.; Ghaemmaghami, S.; Wang, M. Z.; Ma, L.; Oas, T. G.; Fitzgerald, M. C. J. Am. Chem. Soc. 2002, 124, 10256–10257. (8) Zhu, M. M.; Rempel, D. L.; Du, Z.; Gross, M. L. J. Am. Chem. Soc. 2003, 125, 5252–5253. (9) Kamal, J. K. A.; Chance, M. R. Protein Sci. 2008, 17, 1–16. (10) Gerega, S. K.; Downard, K. M. Bioinformatics 2006, 22, 1702–1709. (11) Ganem, B.; Li, Y.-T.; Henion, J. D. J. Am. Chem. Soc. 1991, 113, 6294– 6296. (12) Katta, V.; Chait, B. T. J. Am. Chem. Soc. 1991, 113, 8534–8535. (13) Benesch, J. L. P.; Ruotolo, B. T.; Simmons, D. A.; Robinson, C. V. Chem. Rev. 2007, 107, 3544–3567. (14) Loo, J. A. Int. J. Mass Spectrom. 2000, 200, 175–186. (15) Smith, A. M.; Jahn, T. R.; Ashcroft, A. E.; Radford, S. E. J. Mol. Biol. 2006, 364, 9–19. (16) Zhang, Z.; Krutchinsky, A.; Endicott, S.; Realini, C.; Rechsteiner, M.; Standing, K. G. Biochemistry 1999, 38, 5651–5658. (17) Boys, B. L.; Kuprowski, M. C.; Konermann, L. Biochemistry 2007, 46, 10675–10684. (18) Hagan, N.; Fabris, D. Biochemistry 2003, 42, 10736–10745. (19) Heck, A. J. R. Nat. Methods 2008, 5, 927–933. (20) Jecklin, M. C.; Touboul, D.; Bovet, C.; Wortmann, A.; Zenobi, R. J. Am. Soc. Mass Spectrom. 2008, 19, 332–343. (21) Wang, W.; Kitova, E. N.; Klassen, J. S. Anal. Chem. 2003, 75, 4945–4955. 10.1021/ac900423x CCC: $40.75  2009 American Chemical Society Published on Web 05/13/2009

ments makes the direct ESI-MS approach very attractive. Also, because of the high sensitivity of modern mass spectrometers these interaction studies require only small amounts of sample. Multiple potential ligands to a given receptor can be screened in parallel.25 The direct ESI-MS approach also allows the interrogation of biomolecular assemblies by ion mobility26,27 or by various dissociation techniques.28,29 Potential pitfalls associated with direct ESI-MS interaction studies include the risk that complexes dissociate during ionization or ion sampling.30 Conversely, nonspecific interactions may be formed as ESI artifacts, giving rise to gas-phase assemblies that do not exist in bulk solution.31,32 Several strategies have been proposed to address these issues. For example, the detection of large multiprotein complexes can be facilitated by elevating the pressure in the first pumping stage.33-35 Certain solvent additives promote evaporative cooling, thereby enhancing the survival probability of some weakly bound species.36 Nonspecifically formed ESI artifacts may be identified by conducting studies in the presence of control proteins.37,38 Competition assays have found applications for monitoring protein-peptide interactions,39 and attempts have been made to correct the occurrence of nonspecific binding on the basis of statistical analyses.37,40 Correction methods have also been developed to account for possible differences in response factors for free and bound receptors.23 Despite these recent advances, the possible occurrence of false-negative and false-positive outcomes continues to hamper the general acceptance of direct ESI-MS measurements as a tool for monitoring protein-ligand interactions. Protein-metal complexes represent a particularly interesting research area. Numerous proteins contain bound metals, for example, in the catalytic cleft of enzymes.41 Other proteins undergo metal-induced conformational changes in response to (22) Daniel, J. M.; Friess, S. D.; Rajagopalan, S.; Wendt, S.; Zenobi, R. Int. J. Mass Spectrom. 2002, 216, 1–27. (23) Gabelica, V.; Galic, N.; Rosu, F.; Houssier, C.; De Pauw, E. J. Mass. Spectrom. 2003, 38, 491–501. (24) Frycak, P.; Schug, K. A. Anal. Chem. 2007, 79, 5407–5413. (25) Wigger, M.; Eyler, J. R.; Benner, S. A.; Li, W.; Marshall, A. G. J. Am. Soc. Mass Spectrom. 2002, 13, 1162–1169. (26) Ruotolo, B. T.; Giles, K.; Campuzano, I.; Sandercock, A. M.; Bateman, R. H.; Robinson, C. V. Science 2005, 310, 1658–1661. (27) van Duijn, E.; Barendregt, A.; Synowsky, S.; Versluis, C.; Heck, A. J. R. J. Am. Chem. Soc. 2009, 131, 1452–1459. (28) Wang, W.; Kitova, E. N.; Klassen, J. S. J. Am. Chem. Soc. 2003, 125, 13630– 13631. (29) Jurchen, J. C.; Williams, E. R. J. Am. Chem. Soc. 2003, 125, 2817–2826. (30) Robinson, C. V.; Chung, E. W.; Kragelund, B. B.; Knudsen, J.; Aplin, R. T.; Poulsen, F. M.; Dobson, C. M. J. Am. Chem. Soc. 1996, 118, 8646–8653. (31) Peschke, M.; Verkerk, U. H.; Kebarle, P. J. Am. Soc. Mass Spectrom. 2004, 15, 1424–1434. (32) Wang, W.; Kitova, E. N.; Klassen, J. S. Anal. Chem. 2005, 77, 3060–3071. (33) Tahallah, N.; Pinkse, M.; Maier, C. S.; Heck, A. J. R. Rapid Commun. Mass Spectrom. 2001, 15, 596–601. (34) Schmidt, A.; Bahr, U.; M., K. Anal. Chem. 2001, 73, 6040–6046. (35) Chernushevich, I. V.; Thomson, B. A. Anal. Chem. 2004, 76, 1754–1760. (36) Sun, J.; Kitova, E. N.; Klassen, J. S. Anal. Chem. 2007, 79, 416–425. (37) Sun, J.; Kitova, E. N.; Wang, W.; Klassen, J. S. Anal. Chem. 2006, 78, 3010– 3018. (38) Sun, J.; Kitova, E. N.; Sun, N.; Klassen, J. S. Anal. Chem. 2007, 79, 8301– 8311. (39) Touboul, D.; Maillard, L.; Gra¨sslin, A.; Moumne, R.; Seitz, M.; Robinson, J.; Zenobi, R. J. Am. Soc. Mass Spectrom. 2009, 20, 303–311. (40) Daubenfeld, T.; Bouin, A.-P.; van der Rest, G. J. Am. Soc. Mass Spectrom. 2006, 17, 1239–1248. (41) Ceccarelli, C.; Liang, Z. X.; Strickler, M.; Prehna, G.; Goldstein, B. M.; Klinman, J. P.; Bahnson, B. J. Biochemistry 2004, 43, 5266–5277.

signaling events,42 or they can act as metallochaperones that shuttle inorganic cations to specific locations inside the cell.43 In addition, the incorporation of metal ions can enhance the stability of native protein structures.44 Metal binding may also occur in the context of detoxification.45 Protein-metal interaction studies by ESI-MS are complicated by the pronounced tendency of metal ions to undergo nonspecific binding. In the absence of additional information, it can therefore be difficult to determine whether the detected metalation levels reflect those in bulk solution. For instance, ESI mass spectra recorded in the presence of Na+ generally show pronounced sodium adduction, even for proteins that do not possess any metal affinity in bulk solution.46 Nonspecific sodium adducts are often highly prevalent in the commonly used positive-ion mode, where salt-free solutions would result in “clean” mass spectra that are dominated by multiply protonated [M + nH]n+ ions. In the presence of sodium, however, one or more protons can be replaced by Na+, leading to the formation of heterogeneous [M + mNa + (n - m)H]n+ clusters where m can adopt values between zero and n. The situation is analogous for many other inorganic cations. Various desalting techniques have been developed to cope with nonspecific metal binding, including online dialysis,47-50 reversed-phase high-performance liquid chromatography (HPLC),51,52 and size exclusion chromatography.53 Approaches of this kind can be beneficial for increasing the signal-to-noise ratio of ESI-MS data. However, these techniques are not usually helpful for studies aimed at characterizing the solution-phase metalation of proteins, because desalting will induce shifts in the corresponding binding equilibria. Nonspecific cation adduction can be suppressed to some extent through the addition of ammonium acetate (NH4(CH3COO)), an effect that may be caused by salt precipitation54 or the displacement of metal ions by NH4+, followed by loss of NH3 in the gas phase.55,56 Many proteins that bind metals in solution are acidic due to a high number of Asp and Glu residues, resulting in a negative overall charge in neutral solution.57 These polypeptide chains appear to be particularly prone to nonspecific cation adduction when analyzed by positive-ion ESI-MS. This undesired effect can (42) (43) (44) (45) (46) (47) (48) (49) (50) (51) (52) (53) (54) (55) (56) (57)

Dempsey, A. C.; Walsh, M. P.; Shaw, G. S. Structure 2003, 11, 887–897. Rosenzweig, A. C. Chem. Biol. 2002, 9, 673–677. Matthews, B. W. Acc. Chem. Res. 1988, 21, 333–340. Ngu, T. T.; Stillman, M. J. J. Am. Chem. Soc. 2006, 128, 12473–12483. Verkerk, U. H.; Kebarle, P. J. Am. Soc. Mass Spectrom. 2005, 16, 1325– 1341. Xu, N.; Lin, Y.; Hofstadler, S. A.; Matson, D.; Call, C. J.; Smith, R. D. Anal. Chem. 1998, 70, 3553–3556. Yi, S.; Boys, B. L.; Brickenden, A.; Konermann, L.; Choy, W. Y. Biochemistry 2007, 46, 13120–13130. Lion, N.; Gellon, J.; Jensen, H.; Girault, H. H. J. Chromatogr., A 2003, 1003, 11–19. Hannis, J. C.; Muddiman, D. C. Rapid Commun. Mass Spectrom. 1999, 13, 323–330. Niessen, W. M. J. Chromatogr., A 1999, 856, 179–197. Abian, J.; Oosterkamp, A. J.; Gelpi, E. J. Mass Spectrom. 1999, 34, 244– 254. Cavanagh, J.; Benson, L. M.; Thompson, R.; Naylor, S. Anal. Chem. 2003, 75, 3281–3286. Iavarone, A. T.; Udekwu, O. A.; Williams, E. R. Anal. Chem. 2004, 76, 3944–3950. Felitsyn, N.; Peschke, M.; Kebarle, P. Int. J. Mass Spectrom. Ion Processes 2002, 219, 39–62. Turner, K. B.; Monti, S. A.; Fabris, D. J. Am. Chem. Soc. 2008, 130, 13353– 13363. Hu, P.; Ye, Q.-Z.; Loo, J. A. Anal. Chem. 1994, 66, 4190–4194.

Analytical Chemistry, Vol. 81, No. 12, June 15, 2009

5009

be somewhat reduced by employing negative-ion mode.57,58 In the case of Ca2+ binding proteins, for example, negative-ion ESI produces [M + mCa - (2m + n)H]n- ions.57 Still, it remains uncertain how well the metal loading levels m seen under these conditions reflect those in bulk solution. For example, several negative-ion ESI-MS investigations reported that calmodulin (CaM) binds up to four Ca2+ ions,57,59,60 a finding that is consistent with the presence of four EF-hand binding sites.61 However, other experiments suggested that CaM can accommodate up to eight calcium ions.62-64 Even higher calcium binding levels had previously been reported on the basis of ESI-MS investigations carried out in positive-ion mode,65 although other workers reported a maximum of four Ca2+ under similar conditions.66 This brief summary demonstrates some of the challenges associated with protein-metal interaction studies by ESI-MS, highlighting the need for more reliable experimental protocols. ESI droplets released from the Taylor cone undergo rapid solvent evaporation and multiple fission events, ultimately resulting in nanometer-sized offspring droplets.67 Notably, evaporation continuously increases the concentration of salts and other solutes.68 Assuming that the charged residue model applies to proteins,55,69-71 it seems likely that nonspecific cation adducts are formed just before nanodroplet evaporation to dryness releases protein ions into the gas phase. This high-concentration environment results in a strong tendency for free charge carriers to undergo ion pairing, evidenced by the observation of ESIgenerated salt clusters.68 Metal cations within these terminal ESI droplets can undergo ion pairing with free anions, but metal binding to negatively charged protein carboxylates will occur as well. The cations that become nonspecifically attached to the protein in this way are in addition to those that might already be bound as the result of genuine solution-phase interactions. These considerations suggest that it should be possible to control the extent of nonspecific cation adduction by using salts with carefully chosen counterions as metal source in the protein solution. Specifically, we hypothesize that the formation of nonspecific adducts will be reduced in the presence of anions that exhibit a substantial metal affinity, such that metal binding to these species will be preferred over ion pairing with protein carboxy(58) Veenstra, T. D.; Johnson, K. L.; Tomlinson, A. J.; Naylor, S.; Kumar, R. Biochemistry 1997, 36, 3535–3542. (59) Hu, P.; Loo, J. A. J. Mass Spectrom. 1995, 30, 1076–1082. (60) Watt, S. J.; Oakley, A.; Sheil, M. M.; Beck, J. L. Rapid Commun. Mass Spectrom. 2006, 19, 2123–2130. (61) Park, H. Y.; Kim, S. A.; Korlach, J.; Rhoades, E.; Kwok, L. W.; Zipfel, W. R.; Waxham, M. N.; Webb, W. W.; Pollack, L. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 542–547. (62) Nemirovskiy, O. V.; Ramanathan, R.; Gross, M. L. J. Am. Soc. Mass Spectrom. 1997, 8, 809–812. (63) Pukala, T. L.; Urathamakul, T.; Watt, S. J.; Beck, J. L.; Jackway, R. J.; Bowie, J. H. Rapid Commun. Mass Spectrom. 2008, 22, 3501–3509. (64) Mathur, S.; Badertscher, M.; Scott, M.; Zenobi, R. Phys. Chem. Chem. Phys. 2007, 9, 6187–6198. (65) Lafitte, D.; Capony, J. P.; Grassy, G.; Haiech, J.; Calas, B. Biochemistry 1995, 34, 13825–13832. (66) Hill, T. J.; Lafitte, D.; Wallace, J. I.; Cooper, H. J.; Tsvetkov, P. O.; Derrick, P. J. Biochemistry 2000, 39, 7284–7290. (67) Kebarle, P.; Peschke, M. Anal. Chim. Acta 2000, 406, 11–35. (68) Juraschek, R.; Dulcks, T.; Karas, M. J. Am. Soc. Mass Spectrom. 1999, 10, 300–308. (69) Iavarone, A. T.; Williams, E. R. J. Am. Chem. Soc. 2003, 125, 2319–2327. (70) de la Mora, F. J. Anal. Chim. Acta 2000, 406, 93–104. (71) Kaltashov, I. A.; Mohimen, A. Anal. Chem. 2005, 77, 5370–5379.

5010

Analytical Chemistry, Vol. 81, No. 12, June 15, 2009

lates. Such a strategy should be successful, as long as the newly formed cation-anion pairs do not associate with the protein. The current work tests the viability of this approach by using various salts as metal source in ESI-MS studies on Ca2+ and Zn2+ binding proteins. Tartrate (-OOC-CH(OH)-CH(OH)-COO-) is a weak chelator that is shown to be a particularly effective counterion, giving rise to the formation of gas-phase proteins with metalation levels that are very close to those in bulk solution. EXPERIMENTAL SECTION Materials. Bovine R-lactalbumin (R-LA), bovine β-lactoglobulin (β-LG), horse skeletal muscle holo-myoglobin (Mb), and bovine ubiquitin were purchased from Sigma (St. Louis, MO). Stock solutions of all proteins were extensively dialyzed against 10 mM NH4(CH3COO) before use. Chicken CaM and human prothymosin R (PR) were expressed and purified as described elsewhere.48,72 CaCl2, calcium acetate (Ca(CH3COO)2), calcium tartrate (Ca(OOC(CHOH)2COO), (+)-L-[2R,3R] stereoisomer), calcium citrate, ZnCl2, zinc acetate (Zn(CH3COO)2), ammonium tartrate, and ethylenediaminetetraacetic acid (EDTA) diammonium salt were obtained from Sigma and used as received. Zinc tartrate (Zn(OOC(CHOH)2COO)) was prepared by mixing equimolar amounts of ammonium tartrate and zinc acetate; calcium EDTA was prepared using the same method. All samples were analyzed at a protein concentration of 10 µM, and the pH was adjusted to pH 7.0 with ammonium hydroxide (Fisher Scientific, Nepean, ON, Canada). The ESI mass spectra of this work did not exhibit any time dependence, indicating that metal binding after mixing is very rapid for the systems studied here. Nonetheless, an equilibration period of ca. 10 min prior to data acquisition was used for all samples. Mass Spectrometry. All measurements were performed on an Ultima API (Waters/Micromass, Manchester, U.K.) quadrupole time-of-flight (Q-TOF) instrument, utilizing a standard Z-spray ESI source in negative-ion mode. A capillary voltage of 2 kV, cone voltage of 45 V, and rf lens 1 voltage of 40 V were found to be optimal. The desolvation and source temperatures were set to 120 and 80 °C, respectively. N2 was used as nebulizer and desolvation gas. Samples were infused into the mass spectrometer at a flow rate of 10 µL min-1 via a syringe pump (Harvard Apparatus, South Natick, MA). The mass spectrometer was calibrated with CsI in positive-ion mode. All data were acquired and analyzed using MassLynx software provided by the instrument manufacturer. Measured protein mass spectra were converted to mass distributions using the MaxEnt deconvolution routine. RESULTS AND DISCUSSION ESI-MS experiments were initially carried out on three wellcharacterized calcium binding proteins, with the aim of identifying conditions that best preserve their solution-phase metalation states. (i) R-LA contains one high-affinity calcium binding site.73,74 The protein can accommodate a second Ca2+ ion, but only in the presence of strongly elevated calcium concentrations (e.g., (72) Hu, J.; Jia, X.; Li, Q.; Yang, X.; Wang, K. Biochemistry 2004, 43, 2688– 2698. (73) Permyakov, E. A.; Berliner, L. J. FEBS Lett. 2000, 473, 269–274. (74) Stuart, D. I.; Acharya, K. R.; Walker, N. P. C.; Smith, S. G.; Lewis, M.; Philips, D. C. Nature 1986, 324, 84–87.

Table 1. Properties of Ca2+ Binding Proteins protein

MW (Da) (apo form)

major Ca2+ binding sites

R-LA(ref73) β-LG(ref76) CaM(ref77)

14 182 18 271 16 706

1 1 4

Kd (M)

pI

3.3 × 10-9 3.1 × 10-3 2.7 × 10-6 7.7 × 10-6 3.1 × 10-5 3.1 × 10-5

4.5 5.1 4

100 mM CaCl2).75 (ii) β-LG weakly binds to a single calcium.76 (iii) CaM has four EF-hand calcium binding sites with intermediate affinities.77 The maximum number of metal ions that can be bound to CaM remains somewhat unclear, as noted in the introduction. The properties of the three proteins are summarized in Table 1. Recognizing the occurrence of nonspecific interactions can be facilitated by conducting control experiments on “known nonbinders”.37,38 For this reason, Mb (17 568 Da, pI 7) was included as a test protein that does not show calcium affinity in solution. In agreement with earlier reports,57,58 it was found that experiments carried out in negative-ion mode generally resulted in less nonspecific metal binding than positive-ion ESI-MS. Negative-ion mode was therefore used for all data discussed below. Calcium Binding at Low Ammonium Acetate Concentration. A first set of experiments on R-LA, β-LG, CaM, and Mb was conducted in the presence of 2 mM NH4(CH3COO). ESI mass spectra recorded after extensive dialysis and in the absence of calcium salts are dominated by nonmetalated protein ions (first row of panels in Figure 1). An intense satellite peak due to residual calcium binding is seen only for R-LA (Figure 1A). Evidently, the intrinsic Ca2+ affinity of this protein is so high that the dialysis procedure used is insufficient for complete decalcification. This behavior is consistent with the Kd values listed in Table 1, which show that R-LA binds Ca2+ much more strongly than β-LG and CaM. Exposure of 10 µM R-LA to 200 µM CaCl2 results in electrosprayed protein ions that exhibit pronounced calcium adduction (Figure 1B). The spacing between adjacent peaks is 38 Da (40 2 Da), reflecting the fact that two protons are displaced upon binding of each divalent metal ion.57 The most intense peak in the R-LA spectrum corresponds to m ) 2, but the distribution extends up to at least m ) 10. Somewhat lower metal binding levels are seen in the presence of Ca(CH3COO)2, but the mass distribution retains its most intense peak at m ) 2 (Figure 1C). On the basis of the known presence of a single strong binding site in R-LA,73 signals with m > 1 must be ascribed to nonspecific adducts. A completely different behavior is observed when using Ca(OOC(CHOH)2COO) as metal source in the protein solution. Under these conditions almost all of the electrosprayed R-LA ions are bound to a single calcium ion, with only a very minor peak for m ) 2 (Figure 1D). Qualitatively similar results were obtained in Ca2+ binding experiments on β-LG (Figure 1F-H), with the exception that the most intense signal corresponds to nonmetalated protein (m (75) Chandra, N.; Brew, K.; Acharya, K. R. Biochemistry 1998, 37, 4767–4772. (76) Farrell, H. M.; Thompson, M. P. Protoplasma 1990, 159, 157–167. (77) Nemirovskiy, O.; E., G. D.; Gross, M. L. J. Am. Soc. Mass Spectrom. 1999, 10, 711–718.

) 0) for all conditions studied. The prevalence of free protein is consistent with the much lower calcium affinity of β-LG compared to R-LA. β-LG can dimerize in solution,78 but all the data discussed here are for monomeric protein ions which dominated the mass spectra under the experimental conditions used. Ca2+ binding to CaM causes a conformational change61 that is reflected by the presence of two coexisting charge state envelopes.79-82 In accord with previous work,62 low charge states (centered around 7- and 8-) were found to contain more calcium than highly charged protein ions (11- to 18-). Only the low charge state envelope was considered for the deconvolution analyses in this study, because inclusion of highly charged protein ions would result in unrealistic skewing in favor of partially unfolded solution-phase conformers.83 ESI mass distributions of CaM recorded in the presence of CaCl2 and Ca(CH3COO)2 resemble each other, with strong signals at m ) 4, 5, and 6 (Figure 1, panels J and K). In contrast, measurements carried out with Ca(OOC(CHOH)2COO) resulted in a dominant peak at m ) 4, as expected based on the presence of four EF-hand calcium binding sites.77 Pronounced nonspecific metalation with m > 10 is seen for the Mb control in the presence of CaCl2 (Figure 1N), whereas the extent of adduction is somewhat lower with Ca(CH3COO)2 (Figure 1O). Calcium-free Mb becomes the dominant signal when using Ca(OOC(CHOH)2COO) as metal source (Figure 1P). In summary, the data in Figure 1 reveal that pronounced nonspecific calcium binding occurs during ESI in the presence of CaCl2 and Ca(CH3COO)2. We point out that both of these salts have been widely used in previous ESI-MS-based calcium binding studies. The metalation levels seen for Ca(OOC(CHOH)2COO) are considerably lower, resulting in better qualitative agreement with the expected binding behavior for all four proteins. Calcium Binding at Increased Ammonium Acetate Concentration. As noted in the introduction, previous studies reported that nonspecific metal adduction during ESI can be reduced by using elevated concentrations of NH4(CH3COO).54-56 To test the effectiveness of this solvent additive, calcium binding experiments were repeated in the presence of 100 mM NH4(CH3COO), i.e., at a 50-fold higher concentration than for Figure 1. Inspection of the resulting data confirms that an increased NH4(CH3COO) concentration generally lowers the extent of calcium adduction (Figure 2), but the magnitude of this effect depends on the protein and the metal salt used. Adduct formation is significantly reduced for R-LA, β-LG, and Mb in the presence of CaCl2 and Ca(CH3COO)2 (Figure 2, panels B, C, G, H, Q, and R). In the case of CaM, however, the metal binding levels with chloride and acetate are quite similar to those obtained at low NH4(CH3COO) concentration (Figure 2, panels L and M). Importantly, the fact that substantial adduct formation persists for the “nonbinder” Mb strongly suggests that the metal binding behavior of the other proteins in the presence of CaCl2 and Ca(CH3COO)2 cannot be taken at face value. (78) Invernizzi, G.; Samalikova, M.; Brocca, S.; Lotti, M.; Molinari, H.; Grandori, R. J. Mass Spectrom. 2006, 41, 717–727. (79) Grandori, R. Protein Sci. 2002, 11, 453–458. (80) Kaltashov, I. A.; Eyles, S. J. Mass Spectrom. Rev. 2002, 21, 37–71. (81) Konermann, L. J. Phys. Chem. B 2007, 111, 6534–6543. (82) Watt, S. J.; Oakley, A.; Sheil, M. M.; Beck, J. L. Rapid Commun. Mass Spectrom. 2005, 19, 2123–2130. (83) Kuprowski, M. C.; Konermann, L. Anal. Chem. 2007, 79, 2499–2506.

Analytical Chemistry, Vol. 81, No. 12, June 15, 2009

5011

Figure 1. Deconvoluted ESI mass spectra of four proteins recorded under different solvent conditions in negative-ion mode. First column (A-D), R-LA; second column (E-H), β-LG; third column (I-L), CaM; fourth column (M-P), Mb. Spectra in the first row were recorded in the absence of calcium salts. Data in rows two, three, and four were acquired in the presence of CaCl2, Ca(CH3COO)2, and Ca(OOC(CHOH)2COO), respectively. All solutions contained 10 µM protein and 2 mM NH4(CH3COO); calcium salts were at a concentration of 200 µM. Labels 0, 1, 2, etc. denotes the metalation level m, i.e., the number of calcium ions corresponding to selected peaks.

Figure 2. Deconvoluted ESI mass spectra of R-LA, β-LG, CaM, and Mb as in Figure 1, except that the data were recorded at a 50-fold higher concentration of NH4(CH3COO) (100 mM). The last row of panels (E, J, O, and T) represents theoretical metalation levels, calculated from the Kd values of Table 1. A strategy similar to that of ref 77, assuming four sequential metal binding steps, was employed for modeling the calcium distribution in panel O. The Maple software package (MapleSoft, Waterloo, ON, Canada) was used for calculating concentrations of CaM bound to 0,..., 4 Ca2+ by solving the equations describing the four equilibria and conservation of mass.

The calcium binding levels measured in the presence of Ca(OOC(CHOH)2COO) do not show a pronounced dependence 5012

Analytical Chemistry, Vol. 81, No. 12, June 15, 2009

on the concentration of NH4(CH3COO) (Figure 2, panels D, I, N, and S), and they remain considerably below those seen for CaCl2

and Ca(CH3COO)2. Most importantly, comparison of these tartrate data with theoretical binding distributions based on published Kd values results in excellent agreement (Figure 2, panels E, J, O, and T). A small degree of residual nonspecific binding persists, even when using Ca(OOC(CHOH)2COO) as metal source. This is evident from the minor m ) 2 and m ) 1 signals for R-LA and Mb, respectively. On the basis of these data, we interpret the low-intensity peak at m ) 5 seen for CaM (Figure 2N) as a nonspecific adduct as well. Microcalorimetry provides indirect support for the presence of auxiliary low-affinity Ca2+ binding sites on CaM.84 Our findings suggest, however, that calcium binding to CaM with m > 4 that was observed in previous ESI-MS studies62-64 cannot be interpreted as direct evidence for such additional sites, but has to be attributed to nonspecific binding. Citrate and EDTA Salts as Calcium Source. The data discussed above reveal that ESI-MS experiments carried out in the presence of Ca(OOC(CHOH)2COO) result in metal binding levels that are in excellent agreement with the corresponding solution-phase data. In contrast, CaCl2 and Ca(CH3COO)2 induce pronounced nonspecific adduction. Tartrate binds Ca2+ more strongly (Kd ≈ 10-2 M)85 than acetate (the Kd of Ca(CH3COO)2 is around 0.3 M2).85 Cation-anion interactions for CaCl2 are even weaker than for Ca(CH3COO)2. It is an obvious next step to test whether nonspecific metal adduction during ESI can also be suppressed by using stronger chelators. A series of experiments was therefore conducted on the four test proteins in the presence of calcium citrate (Kd ≈ 10-3 M)86 and calcium EDTA (Kd ≈ 10-7 M).87,88 Citrate resulted in calcium binding levels very similar to those seen for tartrate. However, the spectra were somewhat complicated by the presence of mixed citrate/metal adducts with moderate signal intensities. Chelator/metal adducts were extremely pronounced in the case of EDTA, such that no information on calcium binding could be obtained under these conditions (data not shown). Thus, Ca(OOC(CHOH)2COO) outperforms CaCl2, Ca(CH3COO)2, calcium citrate, and calcium EDTA in its ability to preserve the solution-phase metalation levels of proteins during the ESI process. Metal Binding at Elevated Calcium Concentration. The experiments of Figures 1 and 2 employed 10 µM protein and 200 µM calcium salts. These concentrations cause virtually complete saturation of solution-phase binding sites with Kd values up to ca. 10-5 M (Figure 2, panels E and O). Unfortunately, a Ca2+ concentration of 200 µM is not sufficient for inducing substantial metal binding for weak binders such as β-LG, which has a Kd value around 3 mM (Figure 2I). Studies on the solution-phase metalation of this and other low-affinity proteins require metal concentrations much higher than 200 µM. Clearly, these conditions will promote the formation of nonspecific adducts during ESI, and it has to be evaluated if meaningful information can still be obtained. (84) Milos, M.; Comte, M.; Schaer, J.-J.; Cox, J. A. J. Inorg. Biochem. 1989, 36, 11–25. (85) Cannan, R. K.; Kibrik, A. J. Am. Chem. Soc. 1938, 60, 2314–2320. (86) Thomas, W. C.; Pickett, W. C. Am. J. Physiol. 1960, 199, 103–106. (87) Shimomura, O.; Shimomura, A. Biochem. J. 1984, 221, 907–910. (88) Durham, A. C. H. Cell Calcium 1983, 4, 33–46.

Figure 3. Deconvoluted ESI mass spectra of β-LG (A-C) and Mb (D-F) recorded in 100 mM NH4(CH3COO). CaCl2, Ca(CH3COO)2, and Ca(OOC(CHOH)2COO) were added at a 10-fold higher concentration (2 mM) than for Figure 2. All solutions contained 10 µM protein.

The viability of metal binding studies at elevated calcium concentration was tested in ESI-MS experiments on β-LG in the presence of 2 mM Ca2+ salts and 100 mM NH4(CH3COO) (Figure 3). Measurements carried out in the presence of CaCl2 or Ca(CH3COO)2 show extensive β-LG metalation (Figure 3, panels A and B). Importantly, strong metal binding is also observed for the Mb control (Figure 3, panels D and E), thus revealing that nonspecific adduction is highly prevalent. In fact, the distributions of β-LG and Mb in Figure 3, panels A/D and B/E, are so similar that one would not be able to tell from these data that only one of them is a calcium binder in solution. A small difference between the two proteins is apparent only when using Ca(OOC(CHOH)2COO) (Figure 3, panels C and F). Some nonspecific calcium binding to Mb persists (Figure 3F), but the effect is dramatically reduced when compared to Figure 3, panels D and E. The intensity ratio between the m ) 0 and m ) 1 peaks of β-LG in Figure 3C is 1.3, which is close to the value of 1.5 that is expected on the basis of its Kd value (Table 1). Despite this satisfactory agreement, Figure 3 demonstrates that quantitative ESI-MS-based calcium binding studies on proteins with dissociation constants around 3 mM and above are challenging due to the persistence of nonspecific adducts caused by the required high metal concentrations. Realistically, only qualitative information can be obtained for these weakly binding proteins, even when using Ca(OOC(CHOH)2COO) as metal source. Experiments conducted with millimolar concentrations of CaCl2 and Ca(CH3COO)2 are not meaningful. Ca2+ Binding to Prothymosin r. Having carried out metalation experiments on three extensively studied calcium binding proteins, we now shift our attention to a system that has not been characterized to the same degree. Human PR (11 985 Da, pI 3.5) is an intrinsically disordered protein. Almost half of its sequence is composed of acidic residues (34 Glu and 19 Asp). PR is involved in a wide variety of biological processes such as cell proliferation, apoptosis, and regulation of oxidative stressprotecting gene expression.89,90 The interaction of PR with several (89) Pineiro, A.; Cordero, O. J.; Nogueira, M. Peptides 2000, 21, 1433–1446.

Analytical Chemistry, Vol. 81, No. 12, June 15, 2009

5013

Figure 4. Deconvoluted ESI mass spectra recorded in 100 mM NH4(CH3COO). Panels A-H are for PR. (A) no metal added; (B) CaCl2; (C) Ca(CH3COO)2; (D) Ca(OOC(CHOH)2COO); (E) no salt added; (F) ZnCl2; (G) Zn(CH3COO)2; (H) Zn(OOC(CHOH)2COO). Panels I-L are for ubiquitin, using the same solvent conditions as for the respective PR data in panels E-H. All metal salts were added at a concentration of 2 mM. Peaks marked by triangles correspond to deacetylated PR; asterisks denote chloride adducts. Labels 0, 1, 2, etc. denote the metalation level m, i.e., the number of calcium or zinc ions corresponding to selected peaks.

target proteins is mediated by Zn2+ binding to the C-terminal portion, resulting in partial folding of this region.90,91 However, the zinc affinity of the protein is rather low, with apparent Kd values that are probably on the order of 1 mM.48,91 Weak binding of PR to calcium has been reported as well. 48,91 The number of binding sites, as well as the exact affinities for zinc and calcium, remain unknown. Because of the low metal affinity of PR, calcium binding experiments were carried out under the same conditions as for Figure 3, i.e., in the presence of 2 mM metal salts and 100 mM NH4(CH3COO). Exposure of the protein to CaCl2 or Ca(CH3COO)2 results in metal binding distributions with maxima at m ) 2 (Figure 4, panels B and C). As in the case of β-LG, these spectra resemble those of the Mb control (Figure 3, panels D and E), such that information on possible PR-Ca2+ interactions in solution cannot be obtained. Only in the presence of Ca(OOC(CHOH)2COO) does PR exhibit a dramatically higher metalation level (Figure 4D) than Mb (Figure 3F), thus confirming that PR does indeed undergo specific calcium binding in solution. The accurate determination of Kd values from the data in Figure 4D is hampered by (i) a lack of knowledge regarding the number of Ca2+ binding sites and (ii) the fact that a noticeable level of nonspecific binding persists (Figure 3F). (90) Karapetian, R. N.; Evstafieva, A. G.; Abaeva, I. S.; Chichkova, N. V.; Filonov, G. S.; Rubtsov, Y. P.; Sukhacheva, E. A.; Melnikov, S. V.; Schneider, U.; Wanker, E. E.; Vartapetian, A. B. Mol. Cell. Biol. 2005, 25, 1089–1099. (91) Chichkova, N. V.; Evstafieva, A. G.; Lyakhov, I. G.; Tsvetkov, A. S.; Smirnova, T. A.; Karapetian, R. N.; Karger, E. M.; Vartapetian, A. B. Eur. J. Biochem. 2000, 267, 4745–4752.

5014

Analytical Chemistry, Vol. 81, No. 12, June 15, 2009

Unlike for the data in Figure 2, no attempts were therefore made to interpret these calcium binding data within a numerical model. Nonetheless, a qualitative comparison of the PR distribution of Figure 4D (peaking at m ) 1) with that of β-LG (Figure 3C) shows that the metal affinity of the former is considerably higher. Accordingly, our data imply that at least one calcium ion in PR is bound with Kd < 3 mM, whereas possible additional sites likely have lower affinity. Zn2+ Binding to Prothymosin r. The data discussed so far show that tartrate minimizes the occurrence of nonspecific calcium binding during ESI under all conditions studied. In contrast, CaCl2 and Ca(CH3COO)2 were found to result in highly distorted metal distributions, especially for proteins with low binding affinity that require millimolar salt concentration. It is interesting to explore if the beneficial effects of tartrate extend to other metals. To address this question the Zn2+ binding properties of PR were probed in the presence of 2 mM ZnCl2, Zn(CH3COO)2, and Zn(OOC(CHOH)2COO) in 100 mM NH4(CH3COO). The metal-anion interactions of the last two salts are characterized by Kd values of 0.1 M2 and 2 × 10-3 M, respectively.85 These values are somewhat smaller than those quoted above for Ca(CH3COO)2 and Ca(OOC(CHOH)2COO). Attempts to use Mb as nonbinding control for zinc binding studies were unsuccessful due to the very low signal-to-noise ratio of the resulting spectra (not shown). Data of much higher quality were obtained for ubiquitin (8565 Da, pI 7), which does not have any Zn2+ binding affinity in solution either. Adduction to ubiquitin was most pronounced in the presence of Zn(CH3COO)2, which shows a metal distribution with a maximum at m ) 1 (Figure 4K). Data recorded in the presence of ZnCl2 were complicated by the presence of mixed metal/ chloride satellite peaks (Figure 4J). Analogous to our calcium binding studies, the least amount of nonspecific adduction to the control protein was seen for Zn(OOC(CHOH)2COO) (Figure 4L). Zinc binding levels for PR were substantially higher than for ubiquitin in all cases studied. As expected, PR metalation was more pronounced for Zn(CH3COO)2 (Figure 4G) and ZnCl2 (Figure 4F) than for Zn(OOC(CHOH)2COO) (Figure 4H), an observation that reaffirms the tendency of tartrate to suppress nonspecific adduction. Comparison of the data in panels H and D of Figure 4 reveals that the binding affinity and/or the number of metal binding sites on PR is higher for Zn2+ than for Ca2+. This finding is consistent with previous chromatographic experiments91 as well as rapid-desalting ESI-MS and NMR data.48 CONCLUSIONS Numerous earlier ESI-MS-based calcium binding studies have used CaCl2 or Ca(CH3COO)2 as metal source. The current work demonstrates that both of these salts result in extensive nonspecific Ca2+ adduction, such that the measured protein metalation levels are dramatically higher than those in bulk solution. In contrast, the extent of nonspecific metalation is strongly reduced when using Ca(OOC(CHOH)2COO). For the protein concentration used here (10 µM), the presence of 200 µM Ca2+ causes virtually complete saturation of binding sites with Kd values up to ca. 10-5 M. The use of Ca(OOC(CHOH)2COO) as metal source under these conditions ensures that ESI-MS data recorded in negative-ion mode provide an accurate reflection of the corresponding solution-

phase binding equilibria. Metalation levels measured in the presence of Ca(OOC(CHOH)2COO) are largely independent of the NH4(CH3COO) concentrations in the range studied here, but this solvent additive offers the possibility of conducting experiments at elevated ionic strength (e.g., 100 mM for Figures 2-4) which may be beneficial for some proteins. Low-affinity Ca2+ binders with Kd values in the range of 10-3 M require millimolar metal concentrations for inducing substantial binding in solution. The use of CaCl2 or Ca(CH3COO)2 in these cases is futile because of pronounced adduct formation. Ca(OOC(CHOH)2COO) at millimolar concentration also results in some nonspecific binding; however, a qualitative assessment of metal affinities through pairwise comparisons with other proteins is still possible. The superior performance of tartrate salts in ESI-MS-based metal binding experiments is not specific to Ca2+ but also extends to Zn2+ (and possibly other divalent cations). We are not aware of any previous studies that have reported on the beneficial effects of tartrate salts for protein-metal binding experiments by ESI-MS. This work does not directly examine the mechanism by which tartrate salts suppress nonspecific metalation during ESI. Yet, our data are consistent with the scenario proposed in the introduction, where a chelator undergoes ion pairing with free metal ions in rapidly shrinking solvent droplets, shortly before proteins are released into the gas phase as charged residues. According to this proposal, sequestration of metal cations at a late stage of the ESI process can prevent or reduce metal binding to carboxylates on the protein surface. The cation affinity of chloride and acetate is too weak for sufficient ion paring to occur, thereby increasing the likelihood of nonspecific metal adduction to the protein. Within the proposed mechanism, the metal affinity of tartrate is adequate for providing effective ion pairing once evaporation has produced a sufficiently high salt concentration within the ESI droplet. On the other hand, the tartrate affinity is low enough to avoid any disturbance of protein-metal binding equilibria in bulk solution. Strong metal chelators such as EDTA are ineffective, because they permanently bind Ca2+ and other divalent cations. Unlike in the case of citrate, calcium tartrate complexes do not have a tendency to remain associated with the protein after desolvation. Future studies are required to confirm whether this

proposed mechanism really forms the basis of the observed tartrate behavior. While the experiments for this work were in progress, a study by Turner et al.56 reported the use of chelators in the gas phase for reducing the extent of nonspecific cation adduction to nucleic acids. That work employed a double-sprayer setup, where chelators and adduct-infested analytes were exposed to each other during long-term ion storage in an rf hexapole. Although the approach by Turner et al.56 is certainly interesting, previous work indicates that nonspecific adducts may be just as strongly bound in the gas phase as genuine solution-type ligands.28,92 Hence, it is not clear if gas-phase metal transfer events can reliably differentiate these two types of binding partners, although Turner et al.56 present an example where this seems to be the case. The approach developed in this work relies on the prevention of nonspecific protein-cation pairing through the use of a chelator in solution, thereby avoiding any assumptions regarding the relative strength of specific versus nonspecific interactions in the gas phase. In addition, the technique proposed here is much easier to implement, since it involves simple mixing of protein and metal chelator in solution. Double sprayers or long-term ion storage devices are not required. Thus, it appears that the use of weak solution-phase chelators such as tartrate as described in the current work should be beneficial for a wide range of future protein-metal interaction experiments by ESI-MS. ACKNOWLEDGMENT This work was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC), the Canada Foundation for Innovation (CFI), the Province of Ontario, the Canada Research Chairs Program, the New Century Excellent Talents in University (China, 07-0134), and the National Natural Science Foundation of China (20603003).

Received for review February 25, 2009. Accepted April 17, 2009. AC900423X (92) Yin, S.; Xie, Y.; Loo, J. A. J. Am. Soc. Mass Spectrom. 2008, 19, 1199–1208.

Analytical Chemistry, Vol. 81, No. 12, June 15, 2009

5015