Multiple Products Monitoring as a Robust Approach for Peptide Quantification Je-Hyun Baek, Hokeun Kim, Byunghee Shin, and Myeong-Hee Yu* Functional Proteomics Center, Korea Institute of Science and Technology, Hawolgok-dong, Seongbuk-gu, Seoul 136-791, Korea Received October 11, 2008
Quantification of target peptides and proteins is crucial for biomarker discovery. Approaches such as selected reaction monitoring (SRM) and multiple reaction monitoring (MRM) rely on liquid chromatography and mass spectrometric analysis of defined peptide product ions. These methods are not very widespread because the determination of quantifiable product ion using either SRM or MRM is a very time-consuming process. We developed a novel approach for quantifying target peptides without such an arduous process of ion selection. This method is based on monitoring multiple product ions (multiple products monitoring: MpM) from full-range MS2 spectra of a target precursor. The MpM method uses a scoring system that considers both the absolute intensities of product ions and the similarities between the query MS2 spectrum and the reference MS2 spectrum of the target peptide. Compared with conventional approaches, MpM greatly improves sensitivity and selectivity of peptide quantification using an ion-trap mass spectrometer. Keywords: peptide quantification • product ion monitoring • multiple products monitoring, target verification
Introduction Verification of biomarker candidates requires the accurate quantification of target proteins in human body fluids, such as saliva, urine, and serum. Antibodies against these newly discovered candidates are frequently unavailable, and substitutes for antibody-based detection assays (i.e., Western blotting or ELISA) have been sought in clinical proteomics. Liquid chromatography-mass spectrometry (LC-MS)-based proteomics provides new opportunities for high-throughput analyses in biological and clinical research.1 Most current advancements in protein identification and quantification have adopted MS intensity-based quantification with isotope labeling, such as SILAC (stable isotope labeling by amino acids in cell culture)2 and ICAT (isotope-coded affinity tags),3 or label-free quantification.4 Concerns about MS intensity-based quantification include the following: (i) specificity issues arise when a MS2 spectrum for the target MS peak pair is lacking; (ii) MS scans with wide ranges tend to lose sensitivity; and (iii) reproducibility in label-free quantification is difficult to maintain between LCMS runs.5 Selected reaction monitoring (SRM), multiple reaction monitoring (MRM), and isobaric tags for relative and absolute quantification (iTRAQ) often overcome such limits using MS2 intensity-based quantification through the continuous monitoring of a selected precursor and its product ion.6-9 These methods have been successfully applied to a variety of biological applications, including elucidation of cellular signal* Corresponding author: Dr. Myeong-Hee Yu, Functional Proteomics Center, Korea Institute of Science and Technology, 39-1 Hawolgok-dong, Seongbuk-gu, Seoul 136-791, Korea. E-mail:
[email protected]. Fax: +82-2958-6919. 10.1021/pr800853k CCC: $40.75
2009 American Chemical Society
ing networks,10 analysis of virulence factors,11 and detection of potential biomarkers in human plasma.12 Nevertheless, the practical use of SRM and MRM in proteomics is less widespread than expected, since SRM and MRM require trial and error in selecting product ions of target peptides.8 Candidate peptides for target proteins and their representative MRM transitions must be selected first, and then the uniqueness (selectivity) and sensitivity of the selected MRM transitions must be validated by quantification analysis. Moreover, additional MS2 data are required for target confirmation because the SRM and MRM data do not contain sequence information. Several MS2-based quantification approaches have been developed. With the use of data-independent acquisition in shotgun proteomic analysis, Yates’ group performed automated quantitative analysis of complex peptide mixtures from tandem mass spectra using Census and RelEX software.13,14 The chromatograms reconstituted from their MS2 scans showed good signal-to-noise ratios and were more accurate in quantitative analysis than MS-based quantification. Additionally, Arnott et al. and Vathany et al. showed that quantification using tandem mass spectra enabled the selective detection of target proteins.15,16 However, these methods are not sufficiently robust in target peptide verification with respect to selectivity and sensitivity. In the current study, we introduce a new, robust approach for peptide quantification, termed multiple products monitoring (MpM), which is based on monitoring the majority of product ions obtained in the MS2 scan. MpM enhances the quantitative analysis of target peptides by avoiding the arduous step of selecting transitions to be monitored. In addition, MpM improves the sensitivity and selectivity of peptide quantification. Journal of Proteome Research 2009, 8, 3625–3632 3625 Published on Web 05/26/2009
research articles
Baek et al.
Experimental Procedure
survey scan with values ranging between 300 and 2000 m/z. This scan was followed by 2-3 consecutive, full-range MS2 scans and 2-3 SRM MS2 scans (one or three microscans) with several fixed precursor m/z values (target peptide isolation width, 3 m/z; normalized collision energy, 35%; modified automatic gain control for MS2, 1.0 × 105). For absolute quantification using the MpM method, a known amount of isotopically labeled peptide (heavy peptide) was spiked into a plasma sample that was already digested, and MpM peak areas for heavy and light peptides were calculated. To quantify mTRAQ-labeled peptides using the MpM method, we used a narrower isolation width of the target precursor (1 m/z), considering the mass difference between the light and heavy pair. All MS and MS2 scans for the MpM and SRM methods were obtained using an LTQ XL linear ion trap in centroid mode. To increase the number of targets for quantification, we divided the LC-MS/MS run into several segments (each segment had a 5-min interval and included a maximum of six precursor m/z values for the target peptides in the duty cycle) or used the turboscan mode for one LC-MS/MS run. Data Searching and Search Parameters. The data files for the tandem mass spectra were generated with the extract-msn program (v.3) using Bioworks software (v3.2) and a minimum ion-count threshold of 15 and a minimum intensity of 100. The SEQUEST searches (TurboSequest v.27, rev 12) were individually performed without enzymatic restriction against protein sequence databases, including 180 contaminants: 261 878 entries in the Uniprot sequence database containing 48 proteins, 6939 entries in yeast, and 57 564 entries in IPI.human.v3.14. Mass tolerance for precursor ions with an average mass type was set to 3.0 amu, and that for product ions with a monoisotopic mass type was set to 1 amu. Search criteria included a variable modification of 16 Da for methionine oxidation. For mTRAQ-labeled samples, search criteria included variable modifications of 16 Da for methionine oxidation and 140 Da (light) or 144 Da (heavy) for each mTRAQ adduct. Trans Proteomic Pipeline (v.3.5) software from the Institute for Systems Biology (Seattle, WA), which includes the peptide probability score (P) programs, PeptideProphet and ProteinProphet,17 was used to validate peptide/protein identification (P g 0.9). Protein validation has a 86.6% sensitivity and a 0.7% error at P g 0.9.
Sample Preparation. The 48-protein mixture (Sigma S5697, proteomic dynamic range standard set), yeast lysate, and human plasma proteins were denatured in labeling buffer (6 M urea, 0.05% SDS, 5 mM EDTA, and 50 mM ammonium bicarbonate, pH 8.0) for 30 min, reduced with 3 mM Tris (2carboxyethyl) phosphine hydrochloride for 30 min, and alkylated with 5 mM iodoacetamide for 30 min while shaking at 50 °C. The protein concentration was determined using the Bradford assay. Protein samples were digested with trypsin (Promega, Madison, WI) at a protein-to-trypsin ratio of 50:1 (w/w). SDS and other reagents were removed from the digested protein sample using a mixed strong cation exchange cartridge (OASIS, Waters). Peptides were eluted by adding 5% ammonia in methanol, dried in a speed-vac, and dissolved in 0.4% acetic acid prior to analysis. The 48-protein mixture and isotopically heavy cytochrome c demo peptides (Cat. No. 300100) were donated by Sigma and Thermo Electron, respectively. The heavy peptide (MW ) 1174.62 Da) is 7 Da heavier than the light peptide, which has the sequence TGPNLHGLFGR (underlined leucine residue is composed of seven 13C atoms). The standard peptides were serially diluted in the 0.01-200 fmol range. mTRAQ Labeling of Enolase Peptides. Lyophilized enolase (100 µg) was dissolved in 20 µL of 0.5 M TEAB buffer (pH 8.5). The protein sample was digested with 2 µg of trypsin at 37 °C overnight and then divided into two tubes (each 50 µg) and dried in a speed-vac. Light and heavy mTRAQ reagents were dissolved in 50 µL of isopropanol, transferred into each tube, and mixed. The samples were adjusted to a pH greater than 8.0 with 0.5 M TEAB buffer (pH 8.5) and incubated for 1 h at room temparature. After labeling, excess reagents were removed by a mixed strong cation exchange cartridge (OASIS, Waters). Peptides were eluted by the addition of 5% ammonia in 45% H2O and 50% acetonitrile. The light and heavy mTRAQlabeled enolase peptides were mixed in 4:1, 2:1, 1:1, and 0.5:1 ratios and then spiked into a digested plasma sample. The samples were dried in a speed-vac and dissolved in 0.4% acetic acid prior to analysis. Liquid Chromatography and Mass Spectrometric Analysis. The standard peptides (0.01-200 fmol), a tryptic digest of the 48-protein mixture (1/50 amount), yeast lysate (0.75 µg), a tryptic digest of plasma (2 µg), and mTRAQ-labeled light (25-200 fmol) and heavy (50 fmol) enolase peptides in a tryptic digest of plasma (2 µg) were subjected to LC-MS/MS analysis for MpM peptide quantification. Each peptide sample was loaded onto a C18 (Magic C18aq, Michrom BioResources, Auburn, CA)-packed trap column and separated using a capillary C18 column (20 cm × 75 µm) coupled with a nanospray tip. Peptides were eluted using a 30-min linear gradient of 5-35% solution B in a 60-min run (Solution A, 0.1% formic acid in H2O; Solution B, 0.1% formic acid in 100% acetonitrile). Elution was performed at a flow rate of 300 nL/ min using either the Eksigent MDLC or Agilent 1100 Nanopump system. Peptides were identified using the LTQ XL linear ion trap mass spectrometer (Thermo Finnigan, San Jose, CA). To identify peptides, we used an MS survey scan with values ranging between 300 and 2000 m/z with one microscan being followed by a data-dependent scan with a dynamic exclusion for the MS2 scans (isolation width, 3 m/z; normalized collision energy, 28-35%; exclusion duration, 5 min). To quantify the peptides using the SRM and MpM methods, we used an MS 3626
Journal of Proteome Research • Vol. 8, No. 7, 2009
Results Overall Scheme of Targeted LC-MS/MS for MpM. We performed a series of “discovery” runs on several protein samples, including yeast enolase, the 48-protein mixture, and human plasma, to generate a library of high-confidence MS2 spectra using an LTQ ion-trap mass instrument and searched the spectra using the SEQUEST algorithm. We then performed a targeted LC-MS/MS analysis on the LTQ ion-trap mass instrument in a data-independent manner by specifying a set of precursor m/z values for target peptides. In this analysis, we deliberately chose to acquire MS2 data on only the m/z values that would ultimately produce CID (collision induced dissociation) spectra on our target peptides. The searched spectral data in discovery runs for this inclusion list was used as the matching source (master spectra) for our subsequent analysis using the MpM algorithm as described below. The MpM program requires two types of experimental data (Figure 1). One type is the master MS2 spectra of the target peptide and its peptide sequences (dta, out, and xls formats), which have already been acquired and searched during the discovery
research articles
Multiple Products Monitoring Approach for Peptide Quantification
Figure 1. Overall scheme of MpM. Quantification of a target peptide is performed by comparing a master MS2 spectrum acquired during the discovery step with targeted MS2 spectra acquired from an inclusion list for the target precursor. The MpM chromatogram reconstituted from the MpM scores is used for quantification of the target peptide.
Figure 2. Scoring process using the MpM approach. Numbers in the black arrows indicate the normalized intensities of the product ions in the master MS2 spectrum with respect to the base peak (i.e., y5+). Numbers in the gray arrows indicate the normalized intensities of the product ions in the targeted MS2 spectra with respect to the “Top 1” in each targeted MS2 spectrum. Lined arrows at the bottom indicate the matched product ions (marked with asterisks in the targeted MS2 spectrum), while dashed arrows with question marks indicate unmatched ions. Scoring considers both the absolute intensity and the number of matched product ions in each target MS2 spectrum.
step. The other type is the targeted LC-MS/MS spectra for the target precursors (mzXML format). The combined results of the product ion intensity in the targeted MS2 spectra and spectral comparisons between the master and targeted MS2 spectra generate MpM scores. The MpM chromatogram that is reconstituted from the MpM scores was used for target peptide quantification. MpM is a quantitative method that uses the product ion intensity as the quantitative metric rather than precursor ion intensity (as is done in both SRM and MRM), but uses multiple product ions instead of a single product ion. MpM Score Algorithm. The MpM score is the product of the matched, fragment ion-intensity sum and a scale that represents the similarity of the query spectrum to the master spectrum. User-friendly executable software for both MpM scoring and quantification was implemented in Java language (Supporting Information, Figure S1). a. Selecting Product Ions from the Master MS2 Spectrum. During the prescoring step (Figure 2), a mass table was generated for the product ions which were previously identified as a-, b-, and y-ions in the master MS2 spectrum. This table includes a list of m/z values, their intensities, and their ranks (“Top N”) in terms of intensities of the product ions.
“Top 1” is the most intense ion of all identified product ions in the master spectrum. The number of top-ranked ions is determined as (n × 2 - 1), where n is the number of peptide backbone cleavage sites. The number corresponding to the b1 ion is not included in Top N because, theoretically, it is not observed after dissociation. Charge states are not considered because the number of the observed product ions of charges greater than +2 was not significant. The intensities of the topranked ions in the master spectrum were normalized with respect to the intensity of “Top 1”. b. Matching Step. Targeted MS2 spectra within the defined precursor mass window (default ) (0.1 m/z) were isolated from the targeted LC-MS/MS run. Full-range targeted MS2 spectra were compared with the master MS2 spectrum one by one (Figure 2, Matching step). In each comparison, the ions (m/z) of the master mass table were matched with their counterparts in the targeted MS2 spectrum. Note that while the “Top 1” in the master spectrum is generally the base peak, the “Top 1” in a target MS2 spectrum is not necessarily the base peak of the target spectrum. After matching, the target peak intensities of the matched ions are normalized with respect to the “Top 1” in the target MS2 spectrum. To avoid useless matching with nonspecific spectra, a target spectrum is excluded if the intensity of the “Top 1” peak in the target spectrum is less than 1% of the intensity of the base peak in the target MS2 spectrum (the default mass window for seeking the “Top 1” is (0.6 m/z). c. Scoring Step (Intensity-Based Scoring). Scoring of each targeted MS2 spectrum is performed using a score function that considers both the absolute intensity and number of matched product ions in the target MS2 spectrum (Figure 2). The equation, MpM score ) ωnmi
∑ (I
c,t
× ωr)
where Ic,t is the absolute intensity of a cross-matched product ion (c) in a target MS2 spectrum t, ωr is a weight function that relates the similarities of the normalized intensities of the matched ions, and ωnmi is a second weight function that relates the number of matched ions between the query spectrum and the master spectrum. All weights for MpM scoring range between 0 and 1. The first weight is defined as ωr ) 1 - |ip,m ic,t|, where ip,m is the relative intensity of a product ion (p) in the master mass table and ic,t is the relative intensity of the cross-matched ion (c) in the targeted MS2 spectrum. The second weight function (ωnmi) considers the number of matched ions between the master and target MS2 spectra and assigns any unmatched target spectra an extremely low weight factor, reducing the MpM score substantially. The second weight function is defined as a sigmoidal function, ωnmi ) 1/[1 + exp{-(xnmi - x0)/R}], where xnmi (0-1) is the fraction of matched ions divided by theoretically maximum number of ions (described above as 2n - 1; n is the number of peptide backbonecleavage sites), x0 is a constant related to the center point of the sigmoidal curve, and R is a second constant related to the curve shape (Figure 3). As shown in Figure 3B and 3C, the optimal values of the two constants were x0 ) 0.7 and R ) 0.6, which were used as default values in the MpM algorithm. If xnmi is greater than 0.7 (i.e., if the cross-matched value is >70% between the master and target MS2 spectra), the initial MpM score is almost entirely preserved (ωnmi approaches 1). If the xnmi is less than 0.7, the MpM score diminishes exponentially. Journal of Proteome Research • Vol. 8, No. 7, 2009 3627
research articles
Baek et al.
Figure 4. The linearity and sensitivity of SRM and MpM. Three LC-MS/MS replicates were performed with varying amounts of heavy cytochrome c peptide (0.01-200 fmol). The peptide sequence (MW ) 1174.62 Da) is TGPNLHGLFGR (underlined leucine residue is composed of seven 13C atoms). A triply charged precursor ion of the standard peptide (393.01 m/z) was used for linearity analysis. The base peak (y92+, 509.48 m/z) was used in SRM analysis (isolation width of 2 m/z) and the product ions from Top 1 to Top 19 were used for MpM analysis. Figure 3. Properties of the sigmoidal weight function in similarity matching between the master and target MS2 spectra. (A) Sigmoidal curve showing the relationship between Xnmi (fraction of the matched ions) and ωnmi. The dashed line indicates the center point (ωnmi ) 0.5) of the curve, which yielded a default value of x0 ) 0.7. (B) The optimal x0 constant (0.7) shows the target specificity of the high weights of the target peptide peak (RT ) ∼16.6 min), while the lower value (0.3) of x0 yields high weights of the nontarget peptides. (C) A magnified view of a peptide elution profile (the dashed line) is shown along with MpM chromatograms generated by three different constant values of R (0.2 for 4, 0.6 for b, and 0.9 for 3). On the basis of close similarities between the shape of the MpM chromatogram and the elution profile of the target peptide, the optimal value of 0.6 was determined for constant R.
A sigmoidal function was adopted instead of a linear function (Figure S2) because the increased similarity in matching between the master and targeted MS2 spectra resulted in a significant increase in the number of matched product ions (Figure S3). d. Parameters. All parameters can be adjusted in the option window of the MpM program. It is also possible to import and export files with user-defined parameters. The parameter file contains tolerance of precursor and product ions, the number of top-ranked product ions (“Top N”), sigmoidal weight constants, noise thresholds, and a method for calculating the area under the peak. The default parameters present in the current MpM program are optimized for low-resolution, iontrap mass spectrometry. Depending on the mass accuracy of the mass spectrometer, sample complexity, and the amount of target peptide (
