Development of a Protein Chip: A MS-Based Method for Quantitation

Jun 11, 2004 - Erin N. Warren,† Phillip J. Elms,† Carol E. Parker, and Christoph H. Borchers* ... Protein chip technology permits analysis of the ...
0 downloads 0 Views 391KB Size
Download PDF
Anal. Chem. 2004, 76, 4082-4092

Development of a Protein Chip: A MS-Based Method for Quantitation of Protein Expression and Modification Levels Using an Immunoaffinity Approach Erin N. Warren,† Phillip J. Elms,† Carol E. Parker, and Christoph H. Borchers*

Department of Biochemistry & Biophysics, University of North Carolina at Chapel Hill, CB# 7260, Chapel Hill, North Carolina 27599-7260

Protein chip technology permits analysis of the expression and modification status of numerous targeted proteins within a single experiment, mainly through the use of antibody-based microarrays. Despite recent improvements in these protein chips, their applications are still limited for a variety of reasons, which include technical challenges in fabrication of the antibody chips as well as the very low specificity achieved by current detection methods. We have developed a unique approach for relative and/or absolute quantitation of protein expression and modification based on the capture of epitope peptides on affinity beads, which can be used to develop a massspectrometry-based protein chip technology. This new method, which utilizes antibodies immobilized on beads for the capture of target peptides, instead of proteins, eliminates many of the problems previously associated with protein chips. We present here several proof-ofprinciple experiments examining model peptides by this technique. These experiments show that the method is capable of (i) detecting peptides bound to a single antibody bead, (ii) detecting peptides at low (fmol) levels, (iii) producing MS/MS data of suitable quality for protein identification via database searching or de novo sequencing, (iv) quantitating peptides affinity-bound to antibody beads, (v) specifically detecting target peptides in complex mixtures over wide dynamic ranges, and (vi) is compatible with a microarray format for high-throughput analysis. Because our novel method uses antibody beads instead of a derivatized capture surface, and peptides instead of proteins for affinity capture, it can overcome many of the pitfalls of previous protein chip fabrications. Therefore, this method offers an improved approach to protein chip technology that should prove useful for diagnostics and drug development applications. Existing proteomic techniques allow the qualitative and quantitative analysis of hundreds to thousands of proteins within a single experiment using multidimensional chromatography1,2 or * Corresponding author. Phone: (919)843-5310. Fax: (919)966-2852. E-mail: [email protected]. † These authors contributed equally to the manuscript.

4082 Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

electrophoresis3,4 combined with mass spectrometry. The major drawback of these techniques, however, is the low sample throughput resulting from the time-consuming chromatographic or electrophoretic procedures, making them impractical for clinical and epidemiological research that typically require a large number of samples.3 As such, protein chip technologies that simultaneously analyze the protein expression and modification status of numerous targeted proteins without an exhaustive protein separation prior to analysis are of major interest for biomedical-related research requiring high sample throughput. Current protein chip technologies rely on the immobilization of proteins, primarily antibodies, to the surface of the chip and the subsequent binding of proteins from an analyte sample to the chip.5 Both of these steps, however, present major technical challenges in protein chip technology for two main reasons. First, proteins are extremely sensitive to the physical and chemical properties of the surface substrate; thus, the same type of surface material may not be suitable for the immobilization of all proteins. Second, proteins can be easily denatured, raising the possibility that protein-ligand interactions can be disrupted through loss of native three-dimensional protein structure following adsorption of bait proteins to the microchip surface or during sample handling. Therefore, a protein chip that utilizes immobilized proteins, specifically antibodies, in a way that preserves their threedimensional structure also ensures preservation of the antibodyantigen interaction and is of major importance for future developments of this technology. Of equal importance is the development of a method that preserves the integrity of the analyte applied to the protein chip. To this end, analytical detection of peptides are a robust alternative to the detection of full-length proteins, as peptides are far less susceptible to denaturation than are the fulllength proteins typically used in assays of this type. Another major drawback of current protein chip technology, especially for antibody-based protein chips, is the low specificity of the detection methods.6 Since the current methods used are (1) Lin, D.; Tabb, D. L.; Yates, J. R., III. Biochim. Biophys. Acta 2003, 1646, 1-10. (2) McDonald, W. H.; Yates, J. R., III. Dis. Markers 2002, 18, 99-105. (3) Jenkins, R. E.; Pennington, S. R. Proteomics 2001, 1, 13-29. (4) Ong, S. E.; Pandey, A. Biomol. Eng. 2001, 18, 195-205. (5) Zhu, H.; Snyder, M. Curr. Opin. Chem. Biol. 2003, 7, 55-63. (6) MacBeath, G. Nat. Genet. 2002, 32 Suppl, 526-532. 10.1021/ac049880g CCC: $27.50

© 2004 American Chemical Society Published on Web 06/11/2004

mainly nonmolecular approaches, such as fluorescence, they cannot unambiguously distinguish between signal and noise. To avoid this problem, “sandwich assays” are utilized in which the protein-specific antibody is spotted on the array first for capture and a second antibody, specific to a different part of the target protein, is added for detection.7 This approach, however, requires two antibodies recognizing different parts of the protein while still avoiding spatial interference of the two antibodies. Moreover, high levels of background noise resulting from cross-reactivity of antibodies or from nonspecific binding decrease the overall sensitivity of these detection methods because nonspecifically bound proteins cannot be distinguished from specifically captured proteins. This is especially evident in the analysis of low abundance proteins in the presence of highly complex mixtures such as cell lysates.8 In these cases, the target signal is likely to be negligible compared to the background. To address these issues, we have developed a novel antibody bead technology which is based on standard antibody immobilization techniques, but which relies on the analysis of peptides, instead of proteins, and uses mass spectrometry for detection, identification, and quantitation of target peptides. As has been described previously, mass spectrometry is capable of directly analyzing epitope-containing proteins, and especially peptides, affinity-bound to antibody beads using matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS)9-12 and of sequencing these peptides by tandem MS.13 Peptides that have been previously differentially labeled with stable isotopes14 may be used for absolute and/or relative quantitation and can easily be distinguished from one another based on detectable mass shifts observed in the mass spectrum. This direct, MS-based approach for the analysis of peptides offers the advantages of speed, accuracy, sensitivity, and sequencing capability15 that are currently lacking in many of the other available protein chip technologies. Hutchens and Yip were one of the first to develop an MS-based affinity capture technique for the desorption of affinity-captured proteins using surface-enhanced affinity capture (SEAC) time-offlight mass spectrometry.16 In this work, they demonstrated the use of “affinity capture devices”, which were comprised of singlestranded DNA, immobilized on agarose beads, for the capture of the human protein, lactoferrin. The specific capture of this low abundance protein from a complex biological fluid permitted its detection by direct MALDI-TOF analysis of the affinity beads. This pioneering work demonstrated that affinity capture of proteins, (7) Templin, M. F.; Stoll, D.; Schrenk, M.; Traub, P. C.; Vohringer, C. F.; Joos, T. O. Trends Biotechnol. 2002, 20, 160-166. (8) Liotta, L. A.; Espina, V.; Mehta, A. I.; Calvert, V.; Rosenblatt, K.; Geho, D.; Munson, P. J.; Young, L.; Wulfkuhle, J.; Petricoin, E. F., III. Cancer Cell 2003, 3, 317-325. (9) Parker, C. E.; Papac, D. I.; Trojak, S. K.; Tomer, K. B. J. Immunol. 1996, 157, 198-206. (10) Jeyarajah, S.; Parker, C. E.; Summer, M. T.; Tomer, K. B. J. Am. Soc. Mass Spectrom. 1998, 9, 157-165. (11) Parker, C. E.; Tomer, K. B. Mol. Biotechnol. 2002, 20, 49-62. (12) Nelson, R. W. K. J. R.; Bieber, A. L.; Williams, P. Anal. Chem. 1995, 67, 1153-1158. (13) Raska, C. S.; Parker, C. E.; Sunnarborg, S. W.; Pope, R. M.; Lee, D. C.; Glish, G. L.; Borchers, C. H. J. Am. Soc. Mass Spectrom. 2003, 14, 10761085. (14) Krijgsveld, J.; Ketting, R. F.; Mahmoudi, T.; Johansen, J.; Artal-Sanz, M.; Verrijzer, C. P.; Plasterk, R. H.; Heck, A. J. Nat. Biotechnol. 2003, 21, 927931. (15) Aebersold, R.; Mann, M. Nature 2003, 422, 198-207. (16) Hutchens, T. W. Y. T. Rapid Commun. Mass Spectrom. 1993, 7, 576-580.

followed by mass spectrometric analysis, was feasible and alleviated the suppression effects commonly encountered during analysis of low-abundance proteins in complex mixtures. Furthermore, it showed that placement of affinity media directly on the MALDI target did not affect the signal quality that could be achieved, despite the irregular surface. Nelson et al. extended this idea by capturing a target protein from snake venom using immunoaffinity beads and then eluting the bound proteins from the beads onto the MALDI target prior to MS analysis.12 Absolute quantitation was achieved by introducing a known concentration of the purified target protein, myotoxin a, (which had been previously labeled by lysine conversion to homoarginine), as an internal standard into the initial analyte mixture. By comparison of the ion abundances observed following isolation of both labeled and unlabeled protein species by affinity capture and MALDI-MS analysis, the exact amount of myotoxin a present in the snake venom was ascertained. More recently, this idea of coupling affinity capture with direct mass spectrometric analysis of target proteins has been applied to the development of protein chip microarrays. One method used surface-enhanced laser desorption and ionization time-of-flight (SELDI-TOF) mass spectrometry as a method to successfully detect disease-related proteins affinity-bound to antibodies that have been adsorbed to a chemically modified surface.17-21 Another method, developed by Mirzabekov et al., used hydrogel technology for the immobilization of antibodies and other capture molecules to a microchip.22 The immobilized biomolecules were used for the capture of target proteins from analyte samples, permitting their on-chip mass spectrometric detection by MALDIMS. Despite the advancement these chips make toward a highly versatile and efficient protein chip technology, their requirement for specific plate surface chemistries on which to immobilize capture molecules pose a significant development challenge. Furthermore, the specialized materials needed for surface derivatization and the specialized equipment for analyzing the protein chips are not broadly available to the general scientific community, unlike our method, which can be used with any commercially available MALDI-MS/MS instrument. Another drawback of previous technologies of this type is that they use native proteins for capture, detection, and quantitative analysis. Because proteins are nonhomogeneous with respect to hydrophobicity and solubility, there may be a significant bias in detection of all proteins of interest from a complex biological sample. Furthermore, the combination of a protein chip with MSbased detection may create a bias against the detection of higher molecular weight proteins since mass spectrometric detection of (17) Jr, G. W.; Cazares, L. H.; Leung, S. M.; Nasim, S.; Adam, B. L.; Yip, T. T.; Schellhammer, P. F.; Gong, L.; Vlahou, A. Prostate Cancer Prostatic Dis. 1999, 2, 264-276. (18) Hlavaty, J. J.; Partin, A. W.; Kusinitz, F.; Shue, M. J.; Stieg, A.; Bennett, K.; Briggman, J. V. Clin. Chem. 2001, 47, 1924-1926. (19) Paweletz, C. P.; Trock, B.; Pennanen, M.; Tsangaris, T.; Magnant, C.; Liotta, L. A.; Petricoin, E. F., III. Dis. Markers 2001, 17, 301-307. (20) Vlahou, A.; Schellhammer, P. F.; Mendrinos, S.; Patel, K.; Kondylis, F. I.; Gong, L.; Nasim, S.; Wright Jr, G. L., Jr. Am. J. Pathol. 2001, 158, 14911502. (21) Petricoin, E. F.; Ardekani, A. M.; Hitt, B. A.; Levine, P. J.; Fusaro, V. A.; Steinberg, S. M.; Mills, G. B.; Simone, C.; Fishman, D. A.; Kohn, E. C.; Liotta, L. A. Lancet 2002, 359, 572-577. (22) Rubina, A. Y.; Dementieva, E. I.; Stomakhin, A. A.; Darii, E. L.; Pan’kov, S. V.; Barsky, V. E.; Ivanov, S. M.; Konovalova, E. V.; Mirzabekov, A. D. Biotechniques 2003, 34, 1008-1014, 1016-1020, 1022.

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

4083

the larger proteins is typically of lower sensitivity and mass resolution. The use of peptides is advantageous because of their improved solubility compared to full-length proteins, and because specialized conditions are not needed to maintain their native conformation.23,24 This means that a broader range of cellular proteins may be available for analysis. In addition, the smaller size of peptides, ranging anywhere from 900 to 3200 Da, is more amenable to mass spectrometric detection and sequencing by tandem MS. Another recently described technique addresses the concerns of using protein analytes for capture and detection using immobilized antibodies. This method relies on the specific capture of peptides, instead of proteins, using antibody microarrays.25 However, this method, like the SELDI and hydrogel technologies, uses prefabricated chips that require substantial sample volumes to achieve sample coverage of the entire chip. Since this requires dilution of the sample, the overall concentration of target peptides is, therefore, decreased. This may necessitate the use of antibodies with lower dissociation constants, which have strong affinities for the targeted epitope peptides, potentially limiting the utility of this technology. Additionally, this method did not confirm the identity of the captured epitope-containing peptides by tandem MS as we have done in this manuscript. The method we have developed, while clearly derived from these previously described techniques, uniquely applies off-chip immunoaffinity capture of target peptides and couples it to tandem MS analysis for conclusive sequence confirmation of the captured peptides. This approach permits absolute and/or relative quantitation of these targeted peptides in parallel with their capture and detection, thereby yielding a truly novel approach for the fabrication of a more robust protein chip technology. The proof-of-principle experiments described in this paper make use of commercially available synthetic peptides to simulate the peptide products generated from a proteolytic digest. The 3x Flag antibody and anti-Flag M2 monoclonal antibody beads are used here as models to demonstrate the high levels of sensitivity, mass accuracy, and dynamic range that can be achieved within complex mixtures using our method. Relative and absolute quantitation is demonstrated by capture of AU1 and 3x Flag peptides, respectively, on their cognate antibodies that have been immobilized on agarose beads. Finally, we show the feasibility of this method for use in a microarray format by selectively capturing each of three different peptides from a mixture using antibody beads specific for each peptide, a miniature antibody microarray. The selective capture of each of the different peptides is concomitant with the absolute quantitation of one target peptide, 3x Flag, using a synthetic, isotopically labeled 3x Flag peptide as an internal standard. These results reflect a significant improvement over previous approaches used for protein chip technologies and illustrate the great potential of this method for use in biomedical research, clinical diagnostics, and drug development. (23) Suckau, D.; Kohl, J.; Karwath, G.; Schneider, K.; Casaretto, M.; Bittersuermann, D.; Przybylski, M. Proc. Natl. Acad. Sci. U.S.A. 1990, 87, 98489852. (24) Przybylski, M.; Jetschke, M.; Mak, M.; Schuhmacher, M.; Toth, G.; Szekely, Z.; Penke, B. J. Cell. Biochem. 1994, 33, 0-160. (25) Scrivener, E.; Barry, R.; Platt, A.; Calvert, R.; Masih, G.; Hextall, P.; Soloviev, M.; Terrett, J. Proteomics 2003, 3, 122-128.

4084

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

EXPERIMENTAL SECTION Materials. Compact reaction columns and column filters were purchased from USB Corporation (Cleveland, OH). Gibco Dulbecco’s 10x phosphate buffered saline (PBS) was purchased from Invitrogen Corporation (Carlsbad, CA). The 3x Flag peptide (MDYKDHDGDYKDHDIDYKDDDDK), Influenza hemagglutinin (HA, YPYDVPDYA) peptide, immobilized EZview Red anti-Flag M2 monoclonal antibody beads, immobilized anti-phosphotyrosine monoclonal antibody beads, immobilized anti-HA monoclonal antibody beads, bovine serum albumin (BSA), and acetic anyhydride were purchased from Sigma (St. Louis, MO). Hexadeuteroacetic anhydride was obtained from C/D/N Isotopes (Quebec, Canada). The kinase domain of the insulin receptor-3 (pKDIR, TRDIYETDYpYRK) peptide was purchased from Anaspec, Inc. (San Jose, CA). The AU1 peptide (DTYRYI) was purchased from Berkeley Antibody Company (Richmond, CA) and immobilized anti-AU1 monoclonal antibody beads were obtained from Covance Research Products (Berkeley, CA). Sequencing-grade modified trypsin was obtained from Promega (Madison Wisconsin). R-Cyano-4-hydroxycinnamic acid was obtained from Aldrich Chemical Co. (Milwaukee, WI). Heavy 3x Flag antibody was synthesized at the UNC Peptide Synthesis Facility using an L-lysine-R-NFMOC, -N-t-BOC purchased from Cambridge Isotope Laboratories (Andover, MA) at the C-terminal residue of the peptide. Because the peptide was synthesized with a carboxy amide at the C-terminus, the increase in mass from the unlabeled 3x Flag peptide was 7 Da, instead of 8 Da. Methods. (1) Immunoprecipitation Protocol. Immunoprecipitation of epitope-containing peptides using immobilized antibodies was performed as previously described,13 with the following modifications: Affinity binding was carried out in compact reaction columns after loading with an aliquot of antibody beads (0.5-5 µL) and washing the beads five times with 450 µL of 0.1x PBS (diluted from a 10x stock of Dulbecco’s PBS). Peptides or peptide mixtures were added in a 100-200-µL volume of 0.1x PBS to the beads and incubated for 2-4 h at room temperature with shaking at 1000 rpm using an Eppendorf Thermomixer. After the beads were washed four times each with 450 µL of freshly made 50 mM ammonium bicarbonate, beads were resuspended in a small volume of 50 mM ammonium bicarbonate (10-20 µL) and an aliquot of the beads (0.15-0.5 µL) was spotted directly onto the MALDI target as indicated in the Results. (2) Trypsin Digestion of BSA and Yeast Cell Lysate. For digestion of bovine serum albumin (BSA), 4 mg of BSA was digested with 20 µg of trypsin in 1 mL of freshly made 50 mM ammonium bicarbonate containing 8% acetonitrile overnight at 37 °C. For digestion of the cell lysate, a 1-L culture of Saccharomyces cerevisiae budding yeast was grown to stationary phase, harvested, and lysed in a buffer (∼30 mL) containing 25 mM HEPES, pH 7.5, 400 mM NaCl, and 10% glycerol. A 5-mL volume of the lysate was mixed with 20 µg of trypsin and digested overnight at 37 °C. All digests were immediately frozen at -80 °C and lyophilized using a LABCONCO (Kansas City, MO) Freeze-Dry System. The peptides were resuspended in fresh 50 mM ammonium bicarbonate and mixed with 3x Flag peptide at the indicated ratios for subsequent immunoprecipitation on anti-Flag M2 monoclonal antibody beads.

(3) Isotopic Labeling for Relative Quantitation. The isotopic labeling of peptides by acetylation with acetic anhydride and hexadeuteroacetic anhydride was performed as described previously by Hochleitner et al.26 Briefly, 10 µg of AU1 peptide was reacted for 15 min at room temperature with 5 µL of either acetic anhydride (forming H3-Ac-AU1) or hexadeuteroacetic anhydride (forming D3-Ac-AU1), using freshly made 50 mM ammonium bicarbonate to bring the reaction mixture to a final volume of 100 µL. Reactions were immediately frozen at -80 °C and lyophilized using a LABCONCO Freeze-Dry System. Samples were resuspended in 120 µL of 50 mM ammonium bicarbonate to give 100 pmol/µL stock solutions of H3-Ac- and D3-Ac-AU1 peptides. These stock solutions were mixed together at the indicated ratios (where 1 is equal to a concentration of 2.5 µM), added to anti-AU1 antibody beads for immunoprecipitation, and quantitated by direct MALDI-MS analysis of the beads on the MALDI target. (4) Absolute Quantitation. A 3x Flag peptide containing an isotopically labeled (heavy) lysine at its C-terminus was synthesized by standard methods at the UNC peptide synthesis facility as described under the Materials section. The 3x Flag peptide from Sigma was used as the nonlabeled (light) peptide. Both peptides were accurately weighed on a microbalance and dissolved in 1x PBS. The heavy and light labeled peptides were mixed at the indicated ratios (where 1 is equal to 1.4 µM), immunoprecipitated on anti-Flag M2 antibody beads, and quantitated by direct MALDI-MS analysis of the beads on the MALDI target. (5) Mass Spectrometric Analysis. All MALDI-MS experiments were performed on a MALDI-time-of-flight instrument (Reflex III) from Bruker Daltonics (Billerica, MA) using their Anchor-chip MALDI-target plates (600 µm/384 spot format). The tandem MS analyses were carried out on an Applied Biosystems Voyager 4700 (Framingham, MA) MALDI-TOF/TOF instrument. The nomenclature of Roepstorff and Fohlman,27 later modified by Biemann,28 was used to identify all product ions. The matrix used for all experiments was R-cyano-4-hydroxycinnamic acid (HCCA) after recrystallization with hot methanol. The solvent for HCCA was water/ethanol/formic acid (45/45/10, v/v/v) and was used as a saturated solution. For placement of one bead per spot on the MALDI target, 0.1x PBS was added to suspended antibody beads following immunoprecipitation and a Zeiss Invertoskop microscope at 15× magnification aided in the selective pipetting of one single bead from the bead slurry. Following placement of the antibody bead(s) on the target, 0.3-0.5 µL of HCCA matrix was added to the bead(s), and the spot was allowed to dry at room temperature. (6) Calculation of the Amount of 3x Flag Peptide on a Single Antibody Bead. Exactly determining the amount of affinity-bound peptide on a single antibody bead is difficult due to large variations in the bead diameters (from 40 to 165 µm) as well as the nonuniform distribution of antibodies immobilized on the beads. Using the binding capacity (0.6 mg/mL) as determined by the manufacturer for the anti-Flag M2 monoclonal antibody beads for a tested Flag epitope-tagged protein, bacterial alkaline phosphatase (MW: 47 kDa), we determined the molar binding capacity as 1.2 pmol of Flag-tagged protein bound per microliter (26) Hochleitner, E. O.; Borchers, C.; Parker, C.; Bienstock, R. J.; Tomer, K. B. Protein Sci. 2000, 9, 487-496. (27) Roepstorff, P.; Fohlman, J. Biomed. Mass Spectrom. 1984, 11, 601. (28) Biemann, K. Methods Enzymol 1990, 193, 886-887.

of beads. Assuming that there are ∼100 antibody beads per 1 µL of settled bead slurry,13 and dividing the molar binding capacity by the number of beads per microliter, results in a concentration of ∼12 fmol affinity-bound protein or peptide per antibody bead. RESULTS AND DISCUSSION Analytical Scheme. The analytical scheme outlining our proposed antibody microarray for the capture of epitope-containing peptides is shown in Figure 1. First, the predicted peptide products that would be generated by digestion with a sequence-specific enzyme, such as trypsin, are calculated from the target protein’s amino acid sequence. A predicted proteolytic peptide (of a molecular weight suitable for MS sequencing) is selected, and antibodies are raised against this peptide and immobilized on affinity beads. This enzyme is then used to digest the proteins in the sample, which can originate from a cell extract, tissue biopsy, or a biological fluid. Using sequence-specific proteases would generate predictable peptide products from the proteins in the sample, and these would include the predicted peptide against which the antibody was raised. For relative quantification, the peptides from the proteolytic digest of one proteome would be labeled with a stable isotope and compared to another sample containing peptides labeled with a different isotope. This labeling can be achieved by (a) 14N/15N methods using stable isotope-enriched cell culture media,29 (b) digestion with the 16O-H2O/18O-H2O method,30,31 or (c) stable labeling approaches such as acetylation using acetic anhydride/ hexadeuteroacetic anhydride26 (Figure 1A), or by ICAT.32 ICAT labeling requires that the targeted epitope-containing peptides contain cysteines and that the labeling does not interfere with antibody recognition of the epitope-containing peptide. For absolute quantitation, peptides identical to the target epitope-containing peptides would be synthesized using stable isotopically labeled amino acid residues and added as internal standards at a defined concentration prior to immunoprecipitation of the target peptides (Figure 1B). Absolute quantitation of the peptide reflects the amount of the intact target protein in the original sample. In contrast to relative quantitation, which is limited to the pairwise comparisons of a “control” and “experimental” sample set, this approach permits the quantitation of an unlimited number of proteins from samples obtained under a variety of experimental conditions. In our method, antibodies immobilized on agarose beads are used to capture epitope-containing peptides, thereby avoiding the need for immobilizing the antibodies directly on the microarray chip. In addition, direct, off-chip incubation of the antibody beads with the sample (Figure 1A,B) provides an aqueous environment for the antibodies that maintains the integrity of the capture molecules and requires a lower sample volume to incubate the beads than would be necessary to cover the surface of a prefabricated chip. Following immunoprecipitation, the immobilized antibodies, bound to the differentially labeled peptides, are placed on a (29) Oda, Y.; Huang, K.; Cross, F. R.; Cowburn, D.; Chait, B. T. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 6591-6596. (30) Reynolds, K. J.; Yao, X.; Fenselau, C. J. Proteome Res. 2002, 1, 27-33. (31) Yao, X.; Afonso, C.; Fenselau, C. J. Proteome Res. 2003, 2, 147-152. (32) Gygi, S. P.; Rist, B.; Gerber, S. A.; Turecek, F.; Gelb, M. H.; Aebersold, R. Nat. Biotechnol. 1999, 17, 994-999.

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

4085

Figure 1. Analytical scheme. (A) Relative quantitation. Two proteomes to be compared, control and experimental, are proteolytically digested and then differentially labeled with stable isotopes, such as acetic anhydride and hexadeuteroacetic anhydride as shown here. The differentially labeled peptides are then pooled and incubated with immobilized antibodies to immunoprecipitate the epitope-containing peptides of interest. (B) Absolute quantitation. Epitope-containing peptides synthesized using isotopically labeled amino acids are added to the proteolytic digest of a proteome of interest and subsequently incubated with the appropriate antibody beads to immunoprecipitate the epitope-containing peptides of interest. (C) After immunoprecipitation of differentially labeled peptides, antibodies are arranged in a microarray/spot format on the MALDItarget plate. (D) Adding MALDI matrix solution to the affinity-bound peptides enables the elution of the peptides from the immobilized antibodies, permitting MALDI analysis of the peptides. The relative abundances of the molecular ion signals corresponding to light (H3-labeled peptides for relative quantitation and unlabeled peptide for absolute quantitation) and heavy peptides (D3-labeled peptides for relative quantitation and isotopically labeled, synthetic epitope peptides for absolute quantitation) are used to quantify the amount of this protein in the original sample.

MALDI target in a microarray format (Figure 1C). Conceptually, each spot on the MALDI microarray could contain only a single bead, containing one type of antibody. The identity of the antibody on each bead or spot does not need to be known in advancesit can be determined from the identity of the captured epitopecontaining peptide, which is in turn determined from its molecular weight, and ultimately confirmed by tandem MS. The molecular weights and amino acid sequences of all of the epitope-containing peptides are known since the antibodies used in the microarray were raised against these selected peptides. Following placement of the antibody beads on the target, an acidic MALDI matrix solution is added on top of the beads, thereby permitting elution of the bound peptides and embedding them among the matrix crystals for subsequent MALDI-MS analysis. Using the ion abundances corresponding to the different isotopically labeled peptides from the sample and the standard, the amount of the protein in the original sample can be determined. Sensitivity of One Bead Analysis. To first determine if our new approach has sufficient sensitivity for protein chip applica4086 Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

tions, we used the 3x Flag peptide and anti-FLAG M2 antibody beads as a model system. The 3x Flag peptide was incubated at different concentrations (from 12 to 600 nM) with immobilized anti-Flag EZview M2 monoclonal antibodies. Figure 2 shows the MALDI-MS spectra (Figure 2B) obtained from the analysis of a single antibody bead placed on the MALDI target (Figure 2A) following incubation with 3x Flag at a concentration (12 nM) corresponding to approximately 12 fmol of peptide per bead. For better visualization of the antibody bead embedded in the HCCA crystal, Sigma EZview Red anti-Flag M2 affinity gel was used instead of white or noncolored affinity beads. In addition to ion signals originating from the matrix, the MALDI-MS spectra show a peak at m/z 2861.2720 that corresponds to the monoisotopic, singly charged ion signal of the 3x Flag peptide with a 43 ppm deviation from its theoretical mass using external calibration. The inset of the MALDI-MS spectrum in Figure 2B depicts the peak at m/z 2861.2720. A signal-to-noise ratio (S/N) of greater than 10 and a nearly baseline separation of the isotopes of the 3x Flag peptide were achieved.

Figure 2. Direct MALDI-MS analysis of 3x Flag peptide affinitybound to one anti-Flag antibody bead. (A) Image of a single affinity bead placed on MALDI target. For visualizing purposes, Sigma’s EZView Red anti-Flag M2 monoclonal antibody beads (bead marked by the arrow) are used. (B) Direct MALDI-MS of ∼12 fmol 3x Flag peptide affinity-bound on one antibody bead.

The sensitivity achieved here (∼12 fmol, at S/N = 10) is comparable to that achieved by Nelson et al.12 (∼50 fmol, at S/N = 10) for the MALDI determination of myotoxin a, a 4823 Da polypeptide from snake venom, after affinity purification and elution. This early work, however, used only the molecular weight of this peptide for identification. In addition, our sensitivity is several orders of magnitude better than that reported for the hydrogel method,22 where 0.1-1 pmol of sample was required. In addition, the hydrogel technique required application of a much higher concentration of analyte (∼200 µM) to the protein chip when compared with our peptide-based method in which accurate and specific detection of peptides was demonstrated using initial concentrations as low as 12 nM. Our results demonstrate the feasibility of direct MALDI-MS analysis of peptides affinity-bound on a single antibody bead placed on the MALDI-target. It also shows that the approach is capable of detecting low femtomole amounts of peptide and that the high resolution and high mass accuracy of the TOF mass spectrometer are maintained. Thus, this approach fulfills three main analytical requirements for use as a protein chip: high mass accuracy, high resolution, and sensitivity in the low femtomole range. Absolute Specificity. In addition to accuracy, resolution, and sensitivity, the problem of specificity must also be addressed during the development of a new protein chip technology. Specificity is especially significant for the analysis of low abundance proteins in the presence of complex mixtures, such as cell

lysate or serum, where proteins exist in a very wide dynamic range. Even if the highly abundant proteins have low affinities for the antibody beads, the detection signal from these more abundant, nonspecifically bound proteins may still dominate and mask the signal of the target protein, thus leading to a false negative result. Mass spectrometry has the capability to distinguish between ion signals originating from nonspecifically bound peptides and those from epitope-containing peptides on the basis of their mass. In addition, mass spectrometric detection permits sequencing of the peptides, thus providing absolute identification of the detected ion signals and, thus, absolute detection specificity. To demonstrate the specificity of the MS/MS approach described here, a fixed amount of 3x Flag peptide (100 fmol) was combined with varying concentrations of BSA tryptic peptides at molar ratios ranging from 1:1 to 1:105 (Figure 3A,B and data not shown) of 3x Flag peptide to BSA peptides or added to a tryptic digest of a yeast cell lysate (Figure 3E). The 3x Flag peptide was then immunoprecipitated and the beads were analyzed by MALDIMS, and finally by MALDI-MS/MS, to examine the specificity achieved at the different molar ratios tested. As can be seen from the MALDI-MS spectra in Figure 3A,B, 3x Flag peptide can be detected at m/z values of 2860.993 and 2861.107 (both within 55 ppm or less of the theoretical peptide mass), corresponding to 102 and 105 molar excesses of BSA peptides, respectively. Accurate detection of the 3x Flag peptide is achieved despite an increase in the number and intensity of BSA peptides nonspecifically bound to the antibody beads. Moreover, the insets in Figure 3A,B demonstrate that isotopic resolution of the targeted peptides can be maintained during MALDI analysis of the peptide following bead placement on the target. Figure 3C illustrates the MS/MS sequencing of 3x Flag peptide immunoprecipitated from a 105 M excess of BSA peptides using tandem mass spectrometry. These data, acquired from direct analysis of the anti-Flag antibody beads, is of sufficient quality to permit database searching or de novo sequencing in order to positively confirm the identity of the target peptide. Similar results can be seen from the immunoprecipitation of the 3x Flag peptide from an even more complex mixture of peptides consisting of a 5 × 105 (w/w) excess of peptides from a yeast cell lysate tryptic digest (Figure 3E). These results illustrate both the additional specificity that can be achieved using an MS-based method for the detection of affinity-bound peptides and how the MS results can be used to identify an antibody from its antigenic peptide. Parts D and F of Figure 3 present graphical representations of the “chemical noise” generated in the MALDI-MS spectra by nonspecifically bound peptides from BSA or cell lysate, respectively, as compared to the signal intensity of our target 3x Flag peptide. Each bar represents the percentage of the total ion abundance contributed by nonspecifically bound peptides or the 3x Flag peptide. Had these experiments utilized nonmolecular approaches, like fluorescence or radioactivity that are less capable of distinguishing between signal and noise, the signal from our target peptide would have been overwhelmed by the chemical noise generated from nonspecifically bound peptides, potentially leading to a false negative result. Our MS-based approach for analysis of affinity-bound peptides provides unequivocal selectivity and specificitysthe target signal can be distinguished from Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

4087

Figure 3. Proof of principle. Specificity in a BSA and yeast cell lysate digest. MALDI-MS spectra of 100 fmol 3x Flag peptide affinity-bound to anti-Flag antibody beads purified from (A) a 100 M excess of BSA and (B) a 105 M excess of BSA. The insets in (A) and (B) show isotopic resolution of the 3x Flag peptide during the direct analysis of affinity-enriched peptides on the MALDI plate. (C) Direct MALDI-MS/MS analysis of anti-Flag antibody beads following immunoprecipitation of 3x Flag peptide from a 105 M excess of BSA (see (B)). (D) Bar chart illustrates the relative contributions (as a percentage of the total ion abundance) of the 3x Flag peptide and BSA peptides to the total ion abundance from MALDI-MS spectra analyzing different ratios of 3x Flag peptide to BSA peptides. (E) Direct MALDI-MS analysis of anti-Flag antibody beads following immunoprecipitation of 3x Flag peptide from a 5 × 105 (w/w) excess of cell lysate tryptic digest. (F) Bar chart illustrates the relative contributions (as a percentage of the total ion abundances) of the 3x Flag peptide and cell lysate peptides to the total ion abundances of the MALDI-MS spectra in (E).

chemical noise by MALDI-MS and its identity can be confirmed by tandem mass spectrometry. Relative and Absolute Quantitation. Our proposed protein chip technology is well-suited for the relative quantitation of target 4088 Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

epitope-containing peptides from protease-digested protein samples. In these experiments, peptides from two samples are differentially labeled to allow for the enhanced detection, identification, and quantitation of specific target peptides. To test the relative

quantitation method outlined in Figure 1A, we used AU1 peptide (m/z ) 830.4048) as our model target peptide and subjected it to differential labeling using either acetic anhydride or hexadeuteroacetic anhydride. This leads to the incorporation of CH3O (H3Ac) or CD3O (D3-Ac), respectively, into the AU1 peptide. The incorporation of different labels into the AU1 peptide results in a detectable mass shift of 3 Da between the control (H3-Ac) relative to the treated (D3-Ac) samples. Following labeling of the samples, the differentially labeled, epitope-containing peptides were mixed in various ratios (1:1, 1:3, 3:1, 1:10, or 10:1, H3-Ac-AU1:D3-Ac-AU1) and affinity-purified on anti-AU1 antibody beads. An aliquot of the beads (∼0.5 µL) was placed onto the MALDI target and subjected to MALDI-MS analysis as shown for the 1:10 and 10:1 AU1 peptide mixtures (Figure 4A, B). The abundances of the identical, but differentially labeled AU1 peptides observed during direct MALDI-MS analysis of anti-AU1 antibody beads is proportional to the actual concentrations of the differentially labeled peptides in the initial analyte mixture. Therefore, the ion abundance ratio of the differentially labeled target peptides can be used to determine the amount of the peptide in each of the two samples undergoing analysis relative to one another. A comparison of the relative ion abundances of the differentially labeled AU1 peptides reveals that the experimentally determined ratios of H3-Ac-AU1 and D3-Ac-AU1 monoisotopic ion abundances correlate very well with the theoretical ratios of the peptides applied to the affinity beads. Figure 4E (closed circles) shows a linear correlation (R2 ) 0.98) between the theoretically and the empirically determined H3-Ac-AU1/D3-AcAU1 (light/heavy, L/H) ratios over 2 orders of magnitude (from 1:10 to 10:1). Relative quantitation of peptides in this manner provides a relatively inexpensive and straightforward procedure for the direct comparison of two different samples such as a “control” versus an “experimental” sample. However, sample loss may occur at the various derivatization steps, which are disadvantages when targeting peptides/proteins already found in low abundance in biological samples. To address this concern, as well as to provide a means for determining the exact concentration of a given peptide (and therefore the original protein) in a sample, we have developed a method for the absolute quantitation of peptides using this method. For absolute quantitation, isotopically labeled peptides are synthesized using isotopically labeled amino acid residues, and known amounts of these labeled peptides are added to the peptide mixture as internal standards prior to immunoprecipitation with antibody beads (Figure 1B). No further modification of the sample is required, which helps to reduce sample loss. In these experiments, an isotopically labeled 3x Flag peptide (MDYKDHDGDYKDHDIDYKDDDDK) was synthesized that incorporated an isotopically labeled lysine residue at the Cterminus of the peptide (see Materials and Methods). Because the labeled (heavy) 3x Flag peptide behaves essentially identically to the unlabeled (light) 3x Flag peptide, the ratio of their ion abundances in the MALDI-MS spectrum can be used to quantitate the amount of unlabeled 3x Flag peptide originally present in the sample. In this experiment, light 3x Flag and heavy 3x Flag peptides were mixed in various molar ratios (light: heavy 3x Flag) and immunoprecipitated using anti-Flag beads. Analysis of an

aliquot of these beads spotted directly on the MALDI target reveals that the monoisotopic ion abundances of the two peptides can be distinguished from one another by a mass difference of 7 Da and, in each case, directly reflect the ratio of the two peptides in the original sample prior to bead incubation (Figure 4C,D, and data not shown). Figure 4E (open squares) shows a linear correlation (R2 ) 0.99) between the theoretically and the empirically determined light 3x Flag/Heavy 3x Flag ratios over 2 orders of magnitude (from 1:10 to 10:1). This proof-of-principle experiment demonstrates the power of mass spectrometry for the direct spectrometric analysis of peptides affinity-bound to antibody beads and its usefulness in the absolute quantitation of peptides by use of an internal standard. Model Protein Chip. To validate this approach for use in a microarray format, we performed a miniaturized protein chip experiment utilizing three different target peptides and a single internal standard. The 3x Flag peptide served as our target for absolute quantitation and was mixed in equimolar amounts (200 nM) with two other peptides including the influenza hemagglutinin peptide and a peptide from the kinase domain of the insulin receptor (pKDIR) that contains a phosphotyrosine residue. The heavy 3x Flag peptide was added at an equivalent concentration and was used as an internal standard for absolute quantitation. This peptide mixture was sequentially incubated with 0.5 µL each of anti-HA, anti-phosphotyrosine, and anti-Flag monoclonal antibody beads and each bead type (0.15 µL aliquot) analyzed by MALDI-MS for detection of affinity-bound peptides. Figure 5A shows the MALDI spectrum of the peptide mixture prior to incubation with antibody beads in which ion signals can be distinguished at m/z values corresponding to the molecular weight of each of the four peptides. The light and heavy 3x Flag peptides did not exhibit equivalent ion abundances even though their concentrations in the initial peptide mixture were equal. Because the heavy 3x Flag peptide was synthesized with an isotopically labeled lysine at the most C-terminal amino acid position, a carboxy amide instead of a carboxylic acid was generated at the C-terminal end of the peptide. This subtle difference between the light and the heavy 3x Flag peptides may explain the relatively greater suppression of the heavy 3x Flag ion signal during MALDI analysis when analyzed in a mixture containing other peptides with intense ion signals. These suppression effects are diminished, however, when both 3x Flag peptides are isolated from the mixture by immunoprecipitation, as shown in Figure 5D. This MALDI spectrum shows identical, monoisotopic ion signals for both the heavy and light 3x Flag peptides, demonstrating a direct correlation between their molecular ion abundances and their concentrations in the original peptide mixture. Similar examples of the specificity of this method are seen in immunoprecipitation experiments with other types of antibody beads. In Figure 5B a peak at m/z 1102.497 was detected, corresponding to the molecular weight of the HA peptide. Another peak, at m/z 1124.497, immunoprecipitates with the expected HA peptide and corresponds to its sodium adduct. Figure 5C demonstrates that antibodies immobilized on agarose beads may be used for the specific detection of peptides/proteins with posttranslational modifications such as phosphorylation. Other than some background ion signals from the matrix, the anti-phosphoAnalytical Chemistry, Vol. 76, No. 14, July 15, 2004

4089

Figure 4. Proof-of-principle: relative and absolute quantitation. Direct MALDI-MS analysis of AU1 antibody beads following immunoprecipitation of isotopically labeled AU1 peptides at different ratios. (A) H3-Ac-AU1:D3-Ac-AU1 peptide ratio 1:10, (B) H3-Ac-AU1:D3-Ac-AU1 peptide ratio 10:1. Direct MALDI-MS analysis of anti-Flag antibody beads following immunoprecipitation of light and heavy 3x Flag peptides at defined ratios. (C) Light 3x Flag:heavy 3x Flag ratio 1:10. (D) Light 3x Flag:heavy 3x Flag ratio 10:1. (E) Logarithmic plot of the observed versus the theoretical ratios of the monoisotopic ion abundances of light and heavy epitope-containing peptides. Theoretical ratios of the heavy to light peptides were determined based on their concentration in the original peptide mixture that was applied to the affinity beads. Observed ratios of the heavy and light peptides were determined based on their monoisotopic ion abundances in the MALDI-MS spectrum. Relative quantitation of AU1 peptide (closed circles) and absolute quantitation of 3x Flag peptide (open squares) at various light/heavy peptide ratios.

tyrosine beads specifically captured the pKDIR peptide, which contains a phosphotyrosine residue, from the four-component peptide mixture. Work done previously13 has shown the specificity of the anti-phosphotyrosine antibody used in these experiments. In this work, anti-phosphotyrosine antibody beads were used to 4090 Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

specifically capture the pKDIR peptide from a solution containing both phosphorylated (pKDIR) and nonphosphorylated (KDIR) forms of the peptide, resulting in the capture of only the pKDIR peptide as evidenced by direct MS analysis of the beads. Furthermore, analysis of the supernatant following bead incubation

Figure 5. Proof-of-Principle: mini protein chip experiment. Equivalent amounts of HA, pKDIR, light 3x Flag and heavy 3x Flag peptides were combined and incubated sequentially with anti-HA, anti-phosphotyrosine, and anti-Flag monoclonal antibody beads. An aliquot of each bead type was spotted directly on the MALDI target for MS analysis. (A) MALDI-MS spectrum of peptide mixture prior to bead incubation. (a) HA peptide, theoretical m/z ) 1102.4733; (b) pKDIR peptide, theoretical m/z ) 1702.7478; (c) light 3x Flag peptide, theoretical m/z ) 2861.1484; (d) heavy 3x Flag peptide, theoretical m/z ) 2868.1484. (B) MALDI-MS spectrum from direct analysis of anti-HA antibody beads. (C) MALDIMS spectrum from direct analysis of anti-phosphotyrosine beads. (D) MALDI-MS spectrum from direct analysis of anti-Flag antibody beads.

resulted in a signal for the KDIR peptide only, indicating that the anti-phosphotyrosine beads specifically captured all of the pKDIR peptide present in the solution. Analogous to the AQUA method described by Gerber et al. using stable isotopically labeled phosphopeptides,33 this method allows absolute quantitation of the phosphorylation status of individual phosphorylation sites. However, in contrast to the AQUA approach, the method described in this manuscript does not require LC-MS/MS with selected reaction monitoring, and therefore permits more rapid analysis and a higher compatibility with high throughput analyses. CONCLUSIONS The proof-of-principle experiments described here use direct MALDI-MS and MALDI-MS/MS to examine model peptides affinity-bound to antibody beads. These results demonstrate the feasibility of this type of protein chip for the absolute quantitation of specific epitope-containing peptides, which are derived from (33) Gerber, S. A.; Rush, J.; Stemman, O.; Kirschner, M. W.; Gygi, S. P. Proc Natl Acad Sci U S A 2003, 100, 6940-6945.

their original protein of interest, and also for the detection of peptides containing specific post-translational modifications. We have shown that our method is capable of detecting target peptides present at femtomole levels after capture by a single antibody bead. In addition to a high level of sensitivity, our method is capable of absolute detection specificity because MALDI-MS/MS sequencing can confirm the identity of the captured peptide. Because the data produced is of suitable quality for protein identification via database searching or de novo sequencing, any unexpected peptides bound to the affinity beads can also be identified. This sensitivity and selectivity is combined with the ability to simultaneously quantitate the peptide of interest by relative and/or absolute quantitation techniques. We have shown that, with this method, detection and quantitation of target peptides in complex mixtures can be achieved over a wide dynamic range and is not compromised by the presence of nonspecifically bound peptides. The affinity-based approach described here offers more versatility than previously described techniques, in that it is readily compatible with high throughput microarray formats and can be used with any MALDI-MS or MS/MS instrument. Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

4091

The high mass accuracy, resolution, sensitivity, and specificity of this MS-based method makes this type of protein chip very well-suited for applications that require the detection of a wide array of protein/peptide targets. It is also suitable for the detection of proteins/peptides that contain post-translational modifications indicative of a pathological state, which may be present at very low levels within the cell, making it an ideal technology for use in biomedical research, clinical diagnostics, or drug development applications.

4092

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

ACKNOWLEDGMENT This work was supported in part by a gift from an anonymous donor to the UNC Michael Hooker Proteomics Core Facility for research in proteomics and by the NIH Cancer Center Core Support Grant (CA16086). Received for review January 19, 2004. Accepted April 22, 2004. AC049880G