Precision of Heavy–Light Peptide Ratios Measured by MALDI-TOF

Jan 18, 2012 - ... Suping Zhang , Yassene Mohammed , Andrea L. Palmer , Darryl B. Hardie , Juncong Yang , Andre M. LeBlanc , Christoph H. Borchers...
10 downloads 0 Views 688KB Size
Article pubs.acs.org/jpr

Precision of Heavy−Light Peptide Ratios Measured by MALDI-TOF Mass Spectrometry N. Leigh Anderson,*,†,‡ Morteza Razavi,§ Terry W. Pearson,‡,§ Gary Kruppa,⊥ Rainer Paape,¶ and Detlef Suckau¶ †

Plasma Proteome Institute, Washington, D.C., United States SISCAPA Assay Technologies, Inc., Washington, D.C., United States § Department of Biochemistry and Microbiology, University of Victoria, Victoria, British Columbia, V8W 3P6, Canada ⊥ Bruker Daltonics, Billerica, Massachusetts, United States ¶ Bruker Daltonik, Bremen, Germany ‡

ABSTRACT: We have investigated the precision of peptide quantitation by MALDI-TOF mass spectrometry (MS) using six pairs of proteotypic peptides (light) and same-sequence stable isotope labeled synthetic internal standards (heavy). These were combined in two types of dilution curves spanning 100-fold and 2000-fold ratios. Coefficients of variation (CV; standard deviation divided by mean value) were examined across replicate MALDI spots using a reflector acquisition method requiring 100 000 counts for the most intense peak in each summed spectrum. The CV of light/heavy peptide centroid peak area ratios determined on four replicate spots per sample, averaged across 11 points of a 100fold dilution curve and over all six peptides, was 2.2% (ranging from 1.5 to 3.7% among peptides) at 55 fmol total (light + heavy) of each peptide applied per spot, and 2.5% at 11 fmol applied. The average CV of measurements at near-equivalence (light = heavy, the center of the dilution curve) for the six peptides was 1.0%, about 17-fold lower CV than that observed when five peptides were ratioed to a sixth peptide (i.e., a different-sequence internal standard). Response curves across the 100-fold range were not completely linear but could be closely modeled by a power law fit giving R2 values >0.998 for all peptides. The MALDI-TOF MS method was used to determine the endogenous level of a proteotypic peptide (EDQYHYLLDR) of human protein C inhibitor (PCI) in a plasma digest after enrichment by capture on a high affinity antipeptide antibody, a technique called stable isotope standards and capture by anti-peptide antibodies (SISCAPA). The level of PCI was determined to be 770 ng/mL with a replicate measurement CV of 1.5% and a >14 000-fold target enrichment via SISCAPA-MALDI-TOF. These results indicate that MALDI-TOF technology can provide precise quantitation of high-to-medium abundance peptide biomarkers over a 100-fold dynamic range when ratioed to samesequence labeled internal standards and enriched to near purity by specific antibody capture. The robustness and throughput of MALDI-TOF in comparison to conventional nano-LC−MS technology could enable currently impractical large-scale verification studies of protein biomarkers. KEYWORDS: MALDI-TOF, quantitation, peptide, CV, internal standard



INTRODUCTION Thousands of candidate protein biomarkers have emerged from proteome studies in recent years,1,2 but so far none of these has been successfully translated into an FDA-cleared clinical test.3 Perhaps the most critical step in translating candidate markers into clinical diagnostics, the so-called verification step, requires accurate protein and peptide quantitation in large sample sets (usually plasma or serum). Recently the techniques of multiple reaction monitoring (MRM) mass spectrometry and the use of stable isotope labeled standards (SIS) have been combined to provide precise relative quantitation of proteotypic signature peptides representing numerous proteins of plasma and other samples of clinical interest.4,5 Such directed assays typically © 2012 American Chemical Society

provide greater measurement precision and much higher sample throughput than MS platforms used to achieve the broad proteome coverage necessary in biomarker discovery studies. In addition, directed MS-based assays offer substantial advantages in comparison to the antigen capture immunoassays (e.g., ELISA) used to produce most quantitative protein measurement to date: MRM-MS provides almost absolute structural specificity (via multiple parent/product ion transitions), true internal standardization (via stable isotope labeled peptide standards), and facile multiplexing of disparate peptide targets. Received: November 2, 2011 Published: January 18, 2012 1868

dx.doi.org/10.1021/pr201092v | J. Proteome Res. 2012, 11, 1868−1878

Journal of Proteome Research

Article

Table 1. Characteristics of Peptides Used protein name

SP #

short name

delta mass shift

target peptide sequence

mass (natural)

mass (labeled)

average MALDI peak area

thyroglobulin transferrin receptor protein C inhibitor, PAI3 ferritin light chain thyroglobulin CA-125 (mucin-16) mesothelin 295−586 LPS binding protein HER2/neu alpha-fetoprotein osteopontin HE-4

P01266 P02786 P05154 P02792 P01266 Q8WXI7 Q13421 P18428 P04626 P02771 P10451 Q14508

TgVIF TfR PCI FLC TgFSP CA-125 Mesoth LPS-BP HER2 AFP OPN HE-4

10 10 10 10 10 10 8 8 6 10 8 8

VIFDANAPVAVR GFVEPDHYVVVGAQR EDQYHYLLDR LGGPEAGLGEYLFER FSPDDSAGASALLR ELGPYTLDR LLGPHVEGLK LAEGFPLPLLK AVTSANIQEFAGCK GYQELLEK YPDAVATWLNPDPSQK CCSAGCATFCSLPNDK

1271.7 1672.8 1351.6 1607.8 1406.7 1063.5 1062.6 1197.7 1503.7 979.5 1801.9 1847.7

1281.7 1682.8 1361.6 1617.8 1416.7 1073.5 1070.6 1205.7 1509.7 989.5 1809.9 1855.7

13 152 11 906 10 589 8 547 8 516 5 916 2 082 1 393 531 428 308 307

peptide samples21 have yielded intra-assay peak-ratio CVs near 10% and in some cases as low as 5%.22 MS/MS-based quantification using isobaric labels showed similar variability levels.23,24 Normalization with a very similar but nonisotopically labeled internal standard gave CVs near 2−4%25 and in one case near 1% using a “seed-crystal” sample preparation approach.26 The wide range of reported CVs and the perceived difficulty of achieving reproducible sample preparation using MALDI matrices have contributed to a widespread assumption that peptide-level assays would ultimately migrate to an LCMRM platform when clinical laboratory precision (e.g., CVs ∼5%) is required, as is currently the case with immunosuppressants, steroid hormones, and other small molecule analytes. In this paper, we examine basic parameters of MALDI-TOF peptide quantitation using same-sequence isotopically labeled internal standards in simple mixtures, with the objective of assessing advantages and disadvantages compared to triplequadrupole MRM measurements for use in SISCAPA assays. Within certain limits, MALDI-TOF appears to offer equal or superior quantitative precision for peptide assays in addition to the well-known features of speed and convenience. In particular, we show that in cases where SISCAPA capture is so specific that it eliminates the need for an LC separation prior to MS analysis, MALDI-TOF provides an extremely simple and precise assay methodology that may be applicable for measurement of diagnostic proteins in clinical laboratory environments.

The current sensitivity limitations of MRM applied to unfractionated sample digests of approximately 1 μg/mL for an average protein in plasma5 can be overcome using specific affinity capture of peptide6 or protein7,8 analytes. Specific peptide enrichment from complex digests,6 referred to by the acronym SISCAPA (for stable isotope standards and capture by anti-peptide antibodies), permits use of larger amounts of sample than the commonly used nanoflow LC−MS/MS platforms can otherwise utilize (increasing sensitivity by 1000-fold or more) and also significantly reduces matrix background, allowing shorter LC cycle times and higher throughput. With SISCAPA, the quantitative result is determined as the ratio of sample-derived peptide and a same-sequence stable-isotope labeled internal standard and not by the absolute amount of sample-derived peptide loaded. Thus, SISCAPA allows one to “tune” a multiplex assay by varying the amount of different antibodies (and hence the amount of the different peptides that are captured) in order to generate near-equal signal intensities for each peptide at the MS detector. This approach can “flatten” the stoichiometric distribution of a series of analytes having widely different concentrations and reduces the dynamic range required in the MS detector. While MRM-MS methods typically use ESI-triple-quadrupole instruments, MALDI ionization with TOF mass analysis provides another powerful means of analyzing peptide mixtures. MALDI-TOF instruments are generally simpler, more robust, and easier to operate than LC−ESI-MS/MS systems used for peptide MRM analysis (particularly when nano-LC is used), but so far they have been used primarily for assessment of peak patterns based on mass and semiquantitative peptide analysis using nonisotopic standards. The mass spectrometric immunoassay (MSIA) work of Nelson and colleagues on MALDI analysis of immunocaptured proteins and natural peptides9−11 and Borcher’s and colleagues work on the immunoMALDI (iMALDI) technique for peptide analysis12,13 provide encouraging indications that MALDI-TOF has the resolution and sensitivity to implement a variety of powerful biomarker measurement methods. The precision of MALDI-TOF peptide quantitation, however, has not been considered on a par with triple-quadrupole measurements. Label free studies for bottomup and top-down biomarker discovery have yielded CVs of 25− 30%, indicating the variability of the method when used without internal calibration standards in complex biological matrices.14−16 Peaks ratioed against a dissimilar internal peptide standard17−20 or ratios between d0 vs d3 acrylamide labeled



MATERIALS AND METHODS

Peptides

Twelve pairs of synthetic peptides (Table 1) were prepared by JPT Peptide Technologies GmbH (Berlin, Germany) in two forms: unlabeled (“L”, light or natural) and labeled (“H”, heavy, having a stable isotope labeled C-terminal K or R residue that adds 8.014 or 10.008 amu, respectively). Peptides were supplied by the vendor at greater than 95% purity. One of the peptides (VIFDANAPVAVR, the TgVIF peptide, where Tg refers to parent protein human thyroglobulin) was also synthesized in three differentially labeled forms by the University of Victoria-Genome BC Proteomics Centre (Victoria, BC, Canada). These included forms incorporating labeled C-terminal R (TgVIF+10: +10.008 amu mass increment compared to the natural peptide), C-terminal VR (TgVIF +16: +16.022 amu mass increment), or C-terminal VAVR (TgVIF+26: +26.043 amu mass increment). Upon receipt from the vendors, all peptides were dissolved in 30% acetonitrile/ 1869

dx.doi.org/10.1021/pr201092v | J. Proteome Res. 2012, 11, 1868−1878

Journal of Proteome Research

Article

(N.L.A.) and Angela Jackson (UVic-Genome BC Proteomics Centre, Victoria BC, Canada) to minimize cumulative dilution. Briefly, pooled human potassium/EDTA plasma from 10 males and 10 females was obtained from Bioreclamation Inc. (Westbury NY; Cat No. HMPLEDTA). To denature, reduce, and alkylate plasma proteins, a lyophilized mixture of 9 M urea (Ultra Urea, Sigma-Aldrich Ltd., St. Louis, MO), 0.05 M Tris (2-carboxyethyl)phosphine) (TCEP) (Bondbreaker neutral TCEP solution, Thermo Scientific, Rockford, IL), and 0.2 M Tris (Trizma preset crystals pH 8.1, Sigma-Aldrich) was prepared. One milliliter of the pooled plasma was added to the lyophilized urea/Tris/TCEP mixture, and after mixing on a Vortex mixer, the solution was incubated at room temperature for 30 min with occasional Vortex mixing. The proteins were then alkylated for 30 min at room temperature in the dark by adding iodoacetamide at 3-fold molar excess over plasma cysteines. The urea concentration was reduced to 1 M by dilution with 0.2 M Tris pH 8.1 (Trizma preset crystals pH 8.1) before addition of tosyl phenylalanyl chloromethyl ketone treated trypsin (TRTPCK; Worthington Biochemical Corp., Lakewood, NJ) to the solution to achieve a final protein/ trypsin ratio of 20:1. Digestion was allowed to proceed overnight in a 37 °C incubator. To stop the digestion, the sample was brought to room temperature and was treated with tosyl lysyl chloromethyl ketone (TLCK; Fluka Biochemica, Buchs, Switzerland) at 2-fold excess over trypsin and incubated for 30 min at room temperature. The sample was acidified by addition of concentrated formic acid and diluted in a 1:1 ratio in 0.1% formic acid. The digested sample was concentrated and desalted using 150 mg solid-phase extraction cartridges according to the manufacturer’s instructions (Oasis HLB, Waters, Milford, MA) and was lyophilized overnight. The lyophilized sample was reconstituted to the original plasma volume with PBS and brought to pH 7.4 using 1 M NaOH. To generate dilution curves for the SISCAPA experiment, the synthetic L and H PCI peptides were added to the digested human plasma at known concentrations. For the forward curve, the light peptide was serially diluted from 1000 to 1 fmol while the SIS peptide was spiked at constant 500 fmol in all samples. To generate the reverse curve, the SIS peptide was titrated from 1000 to 1 fmol while the light spike was kept constant at 500 fmol per sample. Thus, each sample contained 10 μL of digested human plasma and varying amounts of the spiked peptides. All samples were diluted to a final volume of 100 μL in PBS/0.03% CHAPS, the detergent added to prevent loss of small amounts of peptides due to adsorption to plastics used during sample handling. The SISCAPA peptide capture procedure was performed in microtiter plates (Cat. No. CA83007-596, Thermo KingFisher 96 KF plate, 200 μL; Thermo Scientific) on a KingFisher 96 bead handling robot (Thermo Fisher Scientific). The first plate held 1.43 μL of custom produced, 1.0 μm diameter, protein G Dynabeads (Product Number 300.14D, Invitrogen) in 200 μL of PBS/0.03% CHAPS (wash buffer). The washed beads were then transferred to the second plate containing 1 μg (in 100 μL of PBS/0.03% CHAPS) of a rabbit monoclonal antibody (RabMAb) selected for its high affinity for the PCI peptide (Epitomics, Inc.). The beads were incubated for 1 h with the antibodies before they were transferred to the third plate containing the human plasma digest with titrated levels of the spiked peptides. The bead−antibody complex was incubated with the plasma digest for 1 h, during which the target peptides were captured. The beads were then transferred through three

0.1% formic acid to 10 nM on the basis of weighed amounts determined by the vendors and aliquots sent for quantification by amino acid analysis (Advanced Protein Technology Centre, The Hospital for Sick Children, Toronto, Ontario). The peptide concentrations were then readjusted to 10 nM and stored as concentrated stock solutions at −70 °C. After peptide storage and just prior to use, MALDI-TOF MS analysis of peptides was performed to test for poststorage integrity and to assess the presence of alterations or modifications. Preparation of Peptide Curves

Varying ratios of L and H versions of 12 peptides (Table 1) were prepared using two different mixing strategies performed with an Agilent Bravo automated liquid handling platform. In the first strategy (here referred to as “combination curve”), a nominally equimolar mix (based on amino acid analysis) of all L peptides and another of all H peptides was prepared at 10 pmol/μL, and from these, 10:1 (mix A) and 1:10 (mix B) stock mixtures of 27.5 fmol/μL were made in 5% acetic acid in order to serve as the first and last points of a dilution curve (covering a total ratio span of 100-fold). Intermediate dilutions were prepared by mixing equal volumes (in each case at least 35 μL for pipetting accuracy) of two prior solutions to create a concentration precisely halfway between the input concentrations, while maintaining the total amount of peptide (L + H) as a constant. Thus, the first stage mixed equal volumes of mixes A and B to yield a 1:1 L/H mixture at the center of the curve. Three additional equal volume mixing stages resulted in the generation of an 11-point curve containing the following L/H ratios: 10, 4.18, 2.38, 1.51, 1.23, 1, 0.81, 0.66, 0.42, 0.24, and 0.1. The combination curve strategy decreases the number of pipetting steps and avoids small volume transfers (thus decreasing the overall error) and ensures that the ends and middle of the curve are as accurate as possible. Two replicate curves were prepared, and each was further diluted in 5% acetic acid by 5-fold steps to yield four curves with 27.5, 5.5, 1.1, and 0.22 fmol/μL of each peptide (L + H). A wider dynamic range was explored in a second set of curves (“serial dilution curves”), in which two differently labeled versions of the TgVIF peptide were present in a 100:1 ratio (e.g., 250 fmol/μL of TgVIF+26 plus 2.5 fmol/μL of TgVIF +16), while a third isotopomer (TgVIF+10) was diluted 2-fold in 11 steps from 1581 to 1.5 fmol/μL (a 1000-fold dynamic range). Since the total amount of all forms of the peptide was held constant, the ratios of TgVIF+10 to TgVIF+26 ranged from 6.3 to 0.003 fmol/μL, a total span of 2047-fold over 11 points. MALDI Target Preparation for Standard Curves

Peptide dilutions prepared in 96-well plates were applied using the Agilent Bravo robot to α-cyano-4-hydroxycinnamic acid matrix (CHCA) spots on Bruker 384-position prespotted AnchorChip MALDI plastic targets (PAC)27,28 in quadruplicate (2 μL each dilution applied to four adjacent target spots). Sample droplets were allowed to stand for 3 min on the CHCA thin layer, after which 3 μL of 10 mM ammonium phosphate/ 0.1% TFA were added to each spot. The liquid droplets were then immediately removed by aspiration using the same pipet tips, and the target was dried in air. SISCAPA-MALDI-TOF Measurement of Protein C Inhibitor (PCI) in Digested Human Plasma

Human plasma was digested with trypsin using the method described by Kuzyk,5,29,30 as modified by one of the authors 1870

dx.doi.org/10.1021/pr201092v | J. Proteome Res. 2012, 11, 1868−1878

Journal of Proteome Research

Article

Figure 1. MALDI-TOF spectrum of nine pairs of peptides (H and L) in equimolar amounts (55 fmol total H + L of each peptide applied to the target). Three of the 12 applied peptides having very low signal strength (AFP, OPN, and HE-4, all containing C-terminal arginine) are outside the mass window of the figure and hence not shown.

10:1 while the combined amount of L + H peptides on the target remained constant. Two such replicate dilution curves were prepared at three loading levels: 5.5, 1.1, and 0.22 fmol/μL total L + H concentrations. Each of the resulting 88 samples was applied to four spots on each of three replicate PAC MALDI targets (2 μL per spot). Thus, the total amount of each peptide (H + L) applied to each spot was 55, 11, 2.2, or 0.44 fmol. Six additional peptide pairs (Table 1, bottom six rows) were included in the dilutions but excluded from this analysis because of low MALDI signal intensity (all of these peptides contained C-terminal lysine, whereas all of the six sequences included in the analysis contained C-terminal arginine, which is known to be favorable for MALDI analysis). Figure 1 shows the spectra of the nominal 1:1 mixture of L and H peptides. Small deviations from expected equal peak heights of H and L pairs are presumably due to variability in initial stock peptide concentrations as measured by amino acid analysis. Initially three data analysis methods were compared using the FlexAnalysis software: (1) centroid monoisotopic peak area, (2) centroid monoisotopic peak intensity, and (3) the intensity derived by the SNAP algorithm, which fits the cluster of peptide isotopomers. Coefficients of variation (CV, standard deviation divided by mean) at each point of the 11-point dilution curves (55 fmol level) were averaged, and finally these values were averaged over the six peptide pairs. The resulting average CV values were 2.2, 2.6, and 2.5% for the three methods, respectively. The centroid monoisotopic peak area was used for all subsequent analysis of high resolution (reflector) spectral data. Figure 2 shows the CVs for each peptide across four replicate spots from one of the dilution curves (curve 2) on one of the targets (target B) at the highest level (55 fmol H + L) of peptide applied. The overall average CV of L/H ratios, averaged across these six peptides and across the 11 dilutions, was 2.2% (ranging from 1.5 to 3.7% among peptides). The central five points of the curve (between 0.66 and 1.5 dilutions) yield an overall average CV across the six peptides of 1.1%.

wash steps for a total elapsed time of approximately 10 min. The first two wash plates contained 200 μL of PBS/0.03% CHAPS, while the third wash plate contained 200 μL of 100% acetonitrile. The fourth and last microtiter plate contained 13 μL of 0.1% formic acid to elute the target peptides from the antibody before the bead−antibody complex was removed from the elution plate. Three microliters of the eluate were spotted onto a Bruker brushed steel MALDI target. After the spots had dried, 2 μL of CHCA matrix (5 mg/mL in 70% ACN/0.1% FA) were spotted onto each sample.29 Data Acquisition

MALDI spectra were acquired on a Bruker Autoflex Speed MALDI-TOF MS in positive ion mode at 1 kHz laser repetition rate. Primary data sets for quantitation (“100 k method”) were obtained in reflector mode by summing good spectra (50 shots per laser position) until the strongest peak reached 100 000 counts, a process which generally took 20−30 s per spot (longer at lower peptide loads). A faster reflector data acquisition strategy (“quick reflector”) collected 2−8e4 counts in the strongest peak in ∼6 s per spot at the 55 fmol level. Linear mode data were collected to obtain 2−8e4 counts in the strongest peak in 10−15 s per spot at 55 fmol level. Data Analysis

Bruker FlexAnalysis software was used to apply various peak detection algorithms as described below and to calibrate the mass dimension using the test peptides as internal standards. Peak lists were exported into Microsoft Excel, where the desired peaks were selected on the basis of mass, and aggregate statistics were calculated using pivot tables.



RESULTS

Precision of Peptide L/H Ratios Measured in Reflector Mode

Six pairs of unlabeled (L) and labeled (H) peptides (Table 1, top six rows) were measured by MALDI-TOF MS across a series of 11 mixtures in which L/H ratios varied from 1:10 to 1871

dx.doi.org/10.1021/pr201092v | J. Proteome Res. 2012, 11, 1868−1878

Journal of Proteome Research

Article

curve were no longer measured in every spot because of sporadic absence of signal from the lower of L or H peaks, while the central five points and central point gave average CVs of 3.0 and 2.9%, respectively. The same dilution curve samples were analyzed on three separate MALDI targets prepared sequentially on the same day. When the L/H ratios (average values over four replicate spots on each target) for a given peptide were compared across these three data sets and the resulting CVs averaged over the 11point curves, results for the six peptides at 55 fmol ranged from 3.6 to 5.7% (overall average 4.4%), while the central five points of the curve ranged from 1.5 to 3.3% (overall average of 2.0%). Data obtained from the same target when it was freshly prepared and again 3 months later yielded comparable values, with overall average CVs (across six peptides) between reads of 3.8% (whole 11 point curve) and 2.5% for the central five points (data not shown). When peak areas for peptides of different sequence were compared against one another, as an alternative to the use of stable isotope labeled internal standards, the resulting ratios showed a high level of variation. The CVs of L/H ratios for four replicate spots at the 1:1 (H = L) dilution (55 fmol L + H on target) averaged 0.9%, whereas when each of the five L peptides other than TgVIF was ratioed to TgVIF L peptide (instead of the respective H version), the CVs averaged 17.2%. The crosssequence ratios were extremely variable from target to target and run to run, suggesting that different peptides are fractionated to some extent across the prespotted matrix thin layers. It is evident that the response curves measured in the experiment (Figure 4) are not precisely linear. We found good fits to the response curves using a power law fit, and using this approach, all six peptides gave R2 values greater than 0.998. Alternative fits, including polynomial fits, performed poorly at low ratios. Specific peptides exhibited slightly different power fit parameters, indicating that separate calibrations are likely to be required for each peptide sequence. Exponents in the power fits ranged from 0.79 to 0.85, while linear coefficients ranged from 0.81 to 1.02. The accuracy of quantitative determinations based on L/H ratios was tested by two approaches. First, power law calibration curves were calculated from eight of the 11 dilutions (leaving out the 0.4, 1, and 2.4 ratio points), and then the inferred dilution values at those points were computed from the measured ratios (effectively an internal “jack-knife” calibration strategy). The deviations of the inferred values from the input L/H ratios ranged from 0.2 to 17.4%, averaging 4.1% across all six peptides and all three inferred points. In a second approach (external calibration), power law calibration parameters were calculated using all 11 points from the first dilution curve and applied to the measured peak area ratios from the second replicate dilution curve, yielding an average deviation across all six peptides of 5.4% from the input L/H ratio values. When fewer laser shots are collected in order to decrease acquisition time compared to the 100 k method used above, replicate CVs increase. Using a “quick” acquisition method yielding 5 000−30 000 counts for the highest peak instead of 100 000 (but requiring on average only 6 s per spot), average replicate spot CVs at 55 fmol across 11 dilutions and six peptides ranged from 3.7 to 6.7% for the three replicate MALDI targets, compared to 2.3−3.6% for data acquired in the 100 k mode.

Figure 2. Plots of CV (calculated across four replicate target spots) as a function of L/H ratio for the selected six peptides and the average of all six.

The average CV across peptides for the replicate spots at a ratio of 1.0 (midpoint of the curve) was 1.0%. Figure 3 shows the relationship between CVs of the L/H ratios and the minimum area of the L and H peaks used to

Figure 3. Relationship of L/H ratio CV to the minimum of the H and L peak areas involved in the ratio.

compute these ratios (55 fmol L + H applied per spot). Low CVs require large peak areas, as expected on the basis of counting statistics. In this experiment, minimum peak areas for both L and H peptides of about 7000 yield ratio CVs less than 2%. When lower amounts of peptides were loaded, replicate CVs increased (Table 2). At 11 fmol (total L + H for each peptide) Table 2. Coefficients of Variation for Various Portions of Standard Curves at Three Peptide Loadings whole 11 pt curve central 5 pts central 1:1 value total L/H measurements

55 fmol

11 fmol

2.2 fmol

2.3% 1.1% 1.0% 242

2.9% 2.3% 2.1% 250

3.4% 3.1% 164

full curve, central five points and central point CVs increase to 2.5, 1.8, and 1.1% respectively. At 2.2 fmol, the ends of the 1872

dx.doi.org/10.1021/pr201092v | J. Proteome Res. 2012, 11, 1868−1878

Journal of Proteome Research

Article

Figure 4. Calibration curves (MALDI-TOF response vs input L/H peptide ratio; combination curve dilution scheme) for PCI peptide on plate B, dilution 2 using centroid data, showing 1 SD error bars and power law fit parameters. The data is shown in both log−log (A) and linear (B) plots.

Linear TOF Mode

Data were also collected in linear TOF mode and analyzed using the SNAP algorithm (selected since in linear mode the isotopic peaks are not fully resolved). Overall, average CVs (across all 11 points of the curve and across all six peptides) were 1.5 times larger than those observed in reflector mode centroid-processed data at 55 fmol, and 3 times larger at 11 fmol. With the instrument and software parameters we used, reflector data appeared to be both more sensitive and more precise than linear mode data, and this is consistent with the general preference for reflector mode data whose higher resolution typically yields better signal-to-noise levels. Power fits to the linear mode response curves showed exponents of 0.93−1.05, much closer to 1.0 (i.e., more linear response curves) than those observed in reflector mode. Calibration using Labeled Isotopomers

A wider dynamic range strategy was tested using a more conventional serial dilution curve approach with three differentially labeled versions of the TgVIF peptide having 10, 16, and 26 amu added mass (denoted TgVIF+10, TgVIF+16, and TgVIF+26; Figure 5). A 1:100 mixture of TgVIF+16 and TgVIF+26 was added as internal standard, while TgVIF+10 was varied over a 2000-fold range (the sum of all peptides was kept constant). In this experiment, we made use of the greater multiplicity and more evenly spaced intensities of peaks in the isotopic envelope of the TgVIF+26 peptide to serve as broad range internal standards. Figure 6 shows the measured response curve for TgVIF+10 in relation to TgVIF+26, which is linear on a log−log scale over a span of ∼500-fold (TgVIF+10/TgVIF +26 from 6.3 to 0.012), with a toe below 0.012. In the wellbehaved region, a good fit (R2 = 0.998) was obtained with a power model (linear coefficient of 2.14 and exponent of 0.845) similar to that was seen in the previous experiment but extending over a significantly wider range. The average deviation between the calibrated and theoretical TgVIF+10/ TgVIF+26 ratios was 6.9%, indicating that perhaps the 10-stage serial dilution process was not as accurate as the combination curve approach. The isotopic envelope of the TgVIF+26 peptide, which includes four labeled amino acids, contains seven measurable peaks (Tg+24 through Tg+30). The ratios between each of these and the highest peak (TgVIF+26) were stable across the dilution curve, except at the final point (highest relative level of TgVIF+10) and appear to provide a workable internal standard curve spanning more than 100-fold in relative abundance.

Figure 5. MALDI-TOF spectrum of Tg+10, Tg+16, and Tg+26 peptides, showing distribution of peak heights for seven isotopomers of Tg+26.

Figure 6. Calibration curve for a 2000-fold serial dilution curve (Tg+10 peptide varying relative to Tg+26), including ratios of six Tg+26 isotopomers to Tg+26 (with 1 SD error bars).

Figure 6 also shows the ratio of TgVIF+16 peptide to TgVIF +26, which should also be constant but instead shows increasing levels of interference over the four points of highest 1873

dx.doi.org/10.1021/pr201092v | J. Proteome Res. 2012, 11, 1868−1878

Journal of Proteome Research

Article

Figure 7. MALDI-TOF spectrum of peptides enriched from a spiked human plasma digest by antipeptide antibodies (EDQYHYLLDR specificity) on magnetic beads (SISCAPA workflow).

the samples: the absence of other highly ionizable HSA peptides (such as LVNEVTEFAK) indicates that the target peptide is enriched by at least 14 000-fold, and in most cases by substantially more, relative to other peptides in plasma digest. Removal of this HSA peptide by washing of the immunoadsorbent beads is currently a focus of our work since this will improve SISCAPA performance. Likewise, the presence of a sodiated version of QDTYHYLPF suggests that it may be beneficial to include an ammonium phosphate wash in SISCAPA-MALDI-TOF target preparation, as is done for the PAC targets. Using the ratio between centroid areas of the predominant peaks of unlabeled and labeled peptide EDQYHYLLDR, we analyzed two forms of standard curves (Figure 8). Plasma digest was spiked with varying levels of labeled peptide (ranging from 1 to 1000 fmol) and a constant level (500 fmol) of unlabeled peptide (forming a “reverse” curve), showing a working range spanning about 100-fold. The lowest point at which triplicate values were obtained (CV = 5.9%) is equivalent to ∼30 fmol spiked labeled peptide. An equivalent standard addition (“forward”) curve, in which labeled peptide was spiked at a constant level (500 fmol) and unlabeled peptide was spiked at levels ranging from 1 to 1000 fmol, yielded equivalent data except for the presence of about 170 fmol of unlabeled peptide derived from the endogenous PCI in 10 μL plasma (indicating about 770 ng/mL PCI protein in this sample before digestion). The average of CVs across all measured points of triplicate curves was 1.9% for the forward curve and 3.4% for the reverse curve, while the CV of the triplicate measurements of the three

TgVIF+10. This suggests that the isotopic envelope of TgVIF +10 contaminates TgVIF+16 when very high levels of TgVIF +10 are present, which is not surprising given the small (6 amu) mass increment between them. SISCAPA-MALDI-TOF Experiment

Antipeptide antibodies, as used in SISCAPA, can substantially enrich target peptides from an extremely complex peptide mixture such as a plasma digest. Figure 7 shows a typical MALDI-TOF MS spectrum of peptides eluted from magnetic beads that were coated with an antibody to target tryptic peptide EDQYHYLLDR (proteotypic for protein C inhibitor; PCI) and used to capture peptides from 10 μL of trypsindigested normal plasma. The target peptide and its associated labeled version (500 fmol spiked into plasma digest before antibody capture) were detected as the sixth and second most intense peaks, respectively. Two other peptides (identified by MALDI-TOF/TOF; data not shown) were enriched to comparable levels: QDTYHYLPF (derived from clusterin, detected in free and Na-adducted forms) and RHPDYSVVLLLR (derived from human serum albumin; HSA). The first of these is a “bonus” peptide that shares the sequence YHYL with the target and is therefore likely to be captured in an epitope-specific manner (the antibody used here can therefore perform SISCAPA assays for both PCI and clusterin). The albumin peptide, by contrast, appears as a frequent contaminant in different SISCAPA experiments and is likely to bind either to the magnetic beads or to rabbit IgG at a site different from the antigen binding site. The amount of HSA in 10 μL of plasma (40 μg = ∼7 000 000 fmol) is 14 000-fold greater than the 500 fmol of heavy target peptide spiked into 1874

dx.doi.org/10.1021/pr201092v | J. Proteome Res. 2012, 11, 1868−1878

Journal of Proteome Research

Article

peptides or even salt contaminants, is nonlinear and specific for each peptide, hence the necessity for pairing each peptide analyte with a same-sequence isotopically labeled internal standard version displaying comparable physicochemical properties such as ionization efficiency, detector response, affinity for matrix, and deposition characteristics during solvent evaporation on the MALDI plate. In the MALDI response curve studies, we attempted to minimize matrix crystallization effects in dilution curve studies by the use of prespotted AnchorChip targets (PACs) in which a thin layer of CHCA matrix, prepared under strict temperature and humidity control during the production process, acts as affinity substrate for peptide deposition. The solvents used for peptide dilution do not visibly affect the crystal morphology.28,31 The ammonium phosphate washing step additionally helps remove sample contaminants such as salts, further improving spectra quality.32 In addition, all spectra were acquired under conditions designed to scan a representative large proportion of the precisely defined matrix covered area (800 μm diameter) to obtain a well-averaged spectrum that is representative of the sample and minimizes position-dependent effects. The fact that robotic sample preparation was employed and matrix handling was avoided by use of PAC targets led to a minimum in pipetting variation and to the 5% accuracy level. It takes strong signals to achieve high precision. The best data presented here were obtained using an acquisition strategy requiring collection of at least 100 000 counts in the dominant peptide peak in each summed spectrum, which typically required 20−30 s at 1 kHz for sample spots to which 55 fmol of each peptide (L + H) had been applied. Dilution curves at the 11 fmol level showed CVs almost as good as at 55 fmol (average CVs of 2.5 vs 2.2%), but at 2.2 fmol not all points were measurable, and CVs near the midpoint of the dilution curve were 3-fold greater than at 55 fmol. This suggests that something between 2 and 10 fmol (H + L) of a reasonably well-ionizing peptide is needed for good precision in the procedure used here with prespotted PAC targets. The amount of peptide consumed in these analyses was of course substantially less than that applied to the target, as supported by the fact that two or more complete high-precision data sets were acquired from some targets and because only a portion of the peptide applied to the prespotted MALDI matrix spots is likely to have been retained when the sample fluid is removed a few minutes after application to the target. Preparations on PAC targets have been shown to capture only a fraction of peptide molecules in the analyte solution because the hydrophobic binding surface of the small matrix crystallites is limited. Hence, the absolute sensitivity of high-precision MALDI can likely be increased substantially from the PAC results presented here. The importance of signal strength in determining peptide ratio precision is related to counting statistics and applies to both of the peptide signals contributing to the L/H ratio. The limited ionization capacity of the MALDI process restricts the number of peptide ions generated compared to ESI, which is expected to inhibit the utility of MALDI in analyzing widely disparate peptide amounts. Two isotopomer peptides at equal abundances can both be collected at high signal levels (given enough laser shots), but if one peptide is 10-fold more abundant than the other, the less abundant is likely to exhibit substantially more noise, resulting in a less precise ratio. This effect is evident in the behavior of replicate spot CVs across our dilution curves: the best CVs occur near the 1:1 ratio points (e.g., the middle of Figure 2) and become progressively greater

Figure 8. Forward and reverse standard curves (measured vs input ratios of varying and constant versions of PCI peptide) resulting from SISCAPA enrichment of PCI peptide from 10 μL plasma digest.

lowest points on the curve (very closely approximating the endogenous peptide level) is 1.5%.



DISCUSSION The results reported here demonstrate that MALDI-TOF MS spectra can yield extremely precise measurements of ratios between labeled and unlabeled same-sequence peptides over a dynamic range of 100−500. The average CVs of L/H peak area ratios determined on replicates (n = 4, same MALDI target) at 11 points along a 10:1 to 1:10 (100-fold) dilution curve ranged from 1.5 to 3.7% for the six peptides investigated here (average 2.2%) when 55 fmol of peptide was applied to prespotted targets. The average replicate CV for the six peptides at the nominal 1:1 L/H ratio at the center of the curve was 1.0%. When the same samples were measured on replicate MALDI targets, or when the same target was retested after a 3 month delay, the average replicate spot CVs across the 11 points and six peptides were 4.4 and 3.8%, respectively. These values compare favorably with the precision of most reported LC− ESI-MRM studies of similar L/H peptide ratios5,29,30 and suggest that under certain circumstances MALDI-TOF can be an extremely precise tool for readout of peptide ratio assays in relatively low-complexity samples. Same-sequence isotopically labeled internal standard peptides are clearly required to attain this high precision. When we examined ratios of five different sequence peptides to a sixth, instead of each against its own labeled version, replicate CVs at the 1:1 dilution were much higher (17.2 vs 0.9%) and extremely variable. This result is consistent with the common perception that “quantification is difficult with MALDI”, which we show is correct if same-sequence internal standards are not used. Various factors contribute to the variation in ratios against a different sequence peptide. It seems likely that peptides could fractionate across the target spot because of differential affinity for matrix crystals of different form or orientation or to differential transport or matrix interaction effects during drying of the spot. Indeed, the converse assumption that all peptide sequences would behave identically during deposition from aqueous solution on a heterogeneous hydrophobic support seems farfetched. Such effects could lead to large variations in peptide peak intensity as a function of the laser spot position on the analysis surface. Clearly, the concentration response of peptides in MALDI spectra, particularly in a mixture with other 1875

dx.doi.org/10.1021/pr201092v | J. Proteome Res. 2012, 11, 1868−1878

Journal of Proteome Research

Article

experiment, we also investigated the use of multilevel internal standards comprising seven isotopomer peaks in the envelope of a peptide containing three labeled amino acids (Tg+26). Inclusion of such a heavily labeled standard peptide would allow selection of at least one standard (H) peak from the isotopic envelope whose area is within 10-fold of that of any analyte (L) peak across a 2000-fold range, improving precision. In the future, we will explore data acquisition methods allowing collection of specified numbers of counts for specific massdefined peaks in the MALDI spectrum. Our initial objective in investigating MALDI-TOF MS peptide quantitation was to determine its potential utility as a detector in SISCAPA assays, where one or a few target peptides are enriched from complex sample digests, along with a samesequence labeled internal standard, by specific antipeptide antibodies. The resulting semipurified analyte peptides (Figure 7) approximate the simple test mixtures used in our dilution curve experiments and thus should be simple enough to avoid issues associated with the limited ionization capacity of MALDI. In the example used here, a specific rabbit monoclonal antibody created to bind EDQYHYLLDR (derived from protein C inhibitor, a medium abundance plasma protein) was used to capture this peptide and a same-sequence labeled internal standard from a digest of unfractionated human plasma. At its endogenous level measured here (170 fmol peptide in 10 μL digest, equivalent to ∼700 ng/mL of the parent protein in the plasma), high replicate precision was obtained: 1.5% CV for measurements of the L/H ratio. The reverse (H peptide varying) dilution curve (Figure 8) indicates that the lower limit of detection of the SISCAPA-MALDI-TOF assay was approximately 10-fold below the endogenous level (∼70 ng/mL for PCI). An equivalent SISCAPA-LC-MRM assay (using nano-LC with triple-quadrupole MRM MS detection on a 4000 QTRAP instrument) gave a level of 184 fmol in 10 μL of digest (manuscript in preparation), in good agreement with the SISCAPA-MALDI-TOF measurement, but has an LOD approximately 1000-fold below the endogenous level for PCI. This comparison suggests that the current SISCAPAMALDI-TOF assay, which has not been optimized for sensitivity, is probably about 100-fold less sensitive than the optimized SISCAPA-LC-MRM platform. While PCI at its normal level is much more abundant than many low-abundance clinical biomarkers such as troponin-I or PSA, there is nevertheless a wide range of known and potential plasma biomarker proteins accessible to SISCAPA assays of this sensitivity, including a substantial number of widely used clinical analytes.34 A second potential limitation of the MALDITOF approach for peptide quantitation compared to LC-MRM lies in its limited specificity: MALDI-TOF resolves the peptides in a single high-resolution dimension (singly ionized parent mass), whereas typical MRM systems employ an LC separation plus two MS dimensions (parent and fragment masses) and can use ratios between multiple such transitions to identify and reject interferences. Other hybrid detection approaches such as MALDI-MRM and LC-MALDI have been developed and could potentially improve the specificity of a SISCAPA-MALDI approach; however, the former is not widely available, and the second sacrifices the principal throughput benefits of SISCAPAMALDI-TOF, for which reasons we have not considered them here. The specificity of MALDI-TOF quantitation in SISCAPA assays therefore rests heavily on the removal of interfering (same-mass) peptides by the antibody affinity capture step,

in both directions away from 1:1. However, it is clear that, using modern MALDI instruments containing high repetition-rate lasers, a very large number of laser shots can be accumulated quickly to improve statistics. Some previous studies of MALDI quantitation appear to have been limited in precision by an artificial constraint on signal quality: a tendency to collect a “nice-looking” spectrum rather than one driven by counting statistics. The degradation of precision away from 1:1 L/H ratios has the effect of limiting the available dynamic range of precision measurements by MALDI-TOF MS, absent compensating increases in peptide amount or spectral acquisition time. Despite this limitation, it is possible in some situations to focus attention on a specific abundance range where greatest precision is required. If, for example, the H peptide is added at the clinical decision level of a biomarker assay, then the L analyte can be most precisely measured at the clinical decision level, thus providing greatest precision where it is needed most. In terms of protein level sensitivity, a sample containing 2.2 fmol (H + L) of a 1:1 L/H ratio, where the L peptide is derived from a 50 kD protein, represents 1.1 fmol of L peptide, 55 pg of the parent protein, or 55 μL of a sample containing the protein at 1 ng/mL. At this level, replicate CVs measured with pure peptides averaged 3.4%. L/H ratio curves were not precisely linear in our studies, as shown in Figure 4, but could be modeled very precisely by a power fit, yielding calibration curves across the full 100-fold range with R2 values of 0.998 or better. This nonlinearity could be due to instrument response characteristics or to MALDI ionization factors. Instrument response characteristics are influenced by parameters such as digitizer offsets, detector voltage settings, and applied laser power, and joint adjustments of these parameters may be required to maximize linearity and fitability of calibration curves under various conditions. MALDI ionization factors are only partially understood and relate to charge competition in the gas phase of the MALDI source. The relative gas phase basicity of peptides can be critical for peptide protonation in MALDI,33 since the number of available protons is limited, and could lead to complex behaviors. The fact that the six peptides used here yielded only slightly different fit parameters suggests that such effects are small but indicates the need to determine parameters for each sequence independently. We tested the precision of peptide-specific power fits to calibrate our L/H ratio measurements, either calibrating three points using parameters computed from the remaining eight within a curve or else calibrating all measurements on a second replicate curve using all 11 points on the first curve samples. The resulting corrected values yielded average deviations from the theoretical value of 4.1 and 5.4%, respectively, including both dilution and measurement errors. To achieve this precision, we carried out the preparation of peptide dilution curves and their application to prespotted MALDI targets using a pipetting robot. Using this approach, we expect that assay results calibrated from measured response curves can have accuracies comparable to many of the best clinical tests in use today. Wider dynamic range measurements are possible by MALDI, as evidenced by the results of a 2000-fold dilution study by conventional serial dilution (Figure 6). Though the results showed greater deviation from linearity, it was clear that reasonable estimates of L/H ratios could be obtained outside a 100-fold span, albeit at reduced precision on account of the reduced counts in the weaker of two ratioed peaks. In this 1876

dx.doi.org/10.1021/pr201092v | J. Proteome Res. 2012, 11, 1868−1878

Journal of Proteome Research

Article

(4) Anderson, L.; Hunter, C. L. Quantitative mass spectrometric multiple reaction monitoring assays for major plasma proteins. Mol. Cell. Proteomics 2006, 5 (4), 573−88. (5) Addona, T. A.; Abbatiello, S. E.; Schilling, B.; Skates, S. J.; Mani, D. R.; Bunk, D. M.; Spiegelman, C. H.; Zimmerman, L. J.; Ham, A. J.; Keshishian, H.; Hall, S. C.; Allen, S.; Blackman, R. K.; Borchers, C. H.; Buck, C.; Cardasis, H. L.; Cusack, M. P.; Dodder, N. G.; Gibson, B. W.; Held, J. M.; Hiltke, T.; Jackson, A.; Johansen, E. B.; Kinsinger, C. R.; Li, J.; Mesri, M.; Neubert, T. A.; Niles, R. K.; Pulsipher, T. C.; Ransohoff, D.; Rodriguez, H.; Rudnick, P. A.; Smith, D.; Tabb, D. L.; Tegeler, T. J.; Variyath, A. M.; Vega-Montoto, L. J.; Wahlander, A.; Waldemarson, S.; Wang, M.; Whiteaker, J. R.; Zhao, L.; Anderson, N. L.; Fisher, S. J.; Liebler, D. C.; Paulovich, A. G.; Regnier, F. E.; Tempst, P.; Carr, S. A. Multi-site assessment of the precision and reproducibility of multiple reaction monitoring-based measurements of proteins in plasma. Nat. Biotechnol. 2009, 27 (7), 633−41. (6) Anderson, N. L.; Anderson, N. G.; Haines, L. R.; Hardie, D. B.; Olafson, R. W.; Pearson, T. W. Mass spectrometric quantitation of peptides and proteins using Stable Isotope Standards and Capture by Anti-Peptide Antibodies (SISCAPA). J. Proteome Res. 2004, 3 (2), 235−44. (7) Nelson, R. W.; Nedelkov, D.; Tubbs, K. A.; Kiernan, U. A. Quantitative mass spectrometric immunoassay of insulin like growth factor 1. J. Proteome Res. 2004, 3 (4), 851−5. (8) Ackermann, B. L.; Berna, M. J. Coupling immunoaffinity techniques with MS for quantitative analysis of low-abundance protein biomarkers. Expert Rev. Proteomics 2007, 4 (2), 175−86. (9) Nelson, R. W.; Krone, J. R.; Bieber, A. L.; Williams, P. Mass spectrometric immunoassay. Anal. Chem. 1995, 67 (7), 1153−8. (10) Kiernan, U. A.; Addobbati, R.; Nedelkov, D.; Nelson, R. W. Quantitative multiplexed C-reactive protein mass spectrometric immunoassay. J. Proteome Res. 2006, 5 (7), 1682−7. (11) Kiernan, U. A.; Nedelkov, D.; Niederkofler, E. E.; Tubbs, K. A.; Nelson, R. W. High-throughput affinity mass spectrometry. Methods Mol. Biol. 2006, 328, 141−50. (12) Jiang, J.; Parker, C. E.; Fuller, J. R.; Kawula, T. H.; Borchers, C. H. An immunoaffinity tandem mass spectrometry (iMALDI) assay for detection of Francisella tularensis. Anal. Chim. Acta 2007, 605 (1), 70−9. (13) Reid, J. D.; Holmes, D. T.; Mason, D. R.; Shah, B.; Borchers, C. H. Towards the development of an immuno MALDI (iMALDI) mass spectrometry assay for the diagnosis of hypertension. J. Am. Soc. Mass Spectrom. 2010, 21 (10), 1680−6. (14) Maltman, D. J.; Brand, S.; Belau, E.; Paape, R.; Suckau, D.; Przyborski, S. A. Top-down label-free LC-MALDI analysis of the peptidome during neural progenitor cell differentiation reveals complexity in cytoskeletal protein dynamics and identifies progenitor cell markers. Proteomics 2011, 11 (20), 3992−4006. (15) Neubert, H.; Bonnert, T. P.; Rumpel, K.; Hunt, B. T.; Henle, E. S.; James, I. T. Label-free detection of differential protein expression by LC/MALDI mass spectrometry. J. Proteome Res. 2008, 7 (6), 2270−9. (16) Toyama, A.; Nakagawa, H.; Matsuda, K.; Ishikawa, N.; Kohno, N.; Daigo, Y.; Sato, T.-A.; Nakamura, Y.; Ueda, K. Deglycosylation and label-free quantitative LC-MALDI MS applied to efficient serum biomarker discovery of lung cancer. Proteome Sci. 2011, 9, 18. (17) D’Imperio, M.; Della Corte, A.; Facchiano, A.; Di Michele, M.; Ferrandina, G.; Donati, M. B.; Rotilio, D. Standardized sample preparation phases for a quantitative measurement of plasma peptidome profiling by MALDI-TOF. J. Proteomics 2010, 73 (7), 1355−67. (18) Floreani, A.; Navaglia, F.; Rizzotto, E. R.; Basso, D.; Chiaramonte, M.; Padoan, A.; Petridis, I.; Cazzagon, N.; Testa, R.; Marra, M.; Plebani, M. Mass spectrometry measurement of plasma hepcidin for the prediction of iron overload. Clin. Chem. Lab. Med. 2011, 49 (2), 197−206. (19) Pang, R. T. K.; Johnson, P. J.; Chan, C. M. L.; Kong, E. K. C.; Chan, A. T. C.; Sung, J. J. Y.; Poon, T. C. W. Technical evaluation of

which must be critically evaluated (for example, by forward and reverse curves as shown here) to detect interferences. Additional potential limitations of MALDI-TOF in comparison with LC−MRM-MS involve dynamic range (100−1000 vs 1e4−1e5) and peptide multiplexing (5−10 vs 50 peptides demonstrated). Because of these limitations, significant further improvements in SISCAPA-MALDI-TOF will be needed before the platform will be suitable for the full range of biomarker assays, i.e., those in the ng/mL level and below. For assays amenable to MALDI-TOF quantitation, the method can provide important advantages in throughput, convenience, and cost. Given the availability of several automated SISCAPA workflows,35,36 including the one used here requiring ∼30 min of robot time per 96 samples, and the ability to carry out precise quantitative measurements by MALDI at a rate of ∼30 s per sample spot, it is clear that substantial sample throughput should be achievable. For example, 384 samples would require ∼2 h of total robot time for SISCAPA processing and ∼3.5 h of mass spectrometer time, implying that ∼1000 samples per day could be practical if all components of sample processing (e.g., tryptic digestion) and data analysis are effectively optimized. While LC-MRM platforms currently retain important advantages in SISCAPA assays for low-abundance biomarkers, our results indicate that high precision quantitation of numerous high- and mediumabundance biomarkers is practical in large sample sets using SISCAPA with MALDI-TOF detection. The relative simplicity of this approach in comparison to lengthy shotgun proteomics workflows should enable wider dissemination of such specific assays in biomarker research.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS N.L.A., T.W.P., and M.R. acknowledge support through U24 Grant CA126476 from the NCI Clinical Proteomic Technology Assessment for Cancer (CPTAC) program. We thank the Natural Sciences and Engineering Research Council of Canada (NSERC) for a Discovery Grant to T.W.P., the Government of British Columbia through the Pacific Century Graduate Scholarship Program for support of M.R., and the staff at the University of Victoria − Genome BC Proteomics Centre for their help and support. We are especially grateful to Angela Jackson for her expertise and unfailing help with trypsin digestions and to Agilent Technologies for the loan of a Bravo automated liquid handling platform.



REFERENCES

(1) Polanski, M.; Anderson, N. L. A list of candidate cancer biomarkers for targeted proteomics. Biomarker Insights 2006, 2, 1−48. (2) Lee, B. T. K.; Liew, L.; Lim, J.; Tan, J. K. L.; Lee, T. C.; Veladandi, P. S.; Lim, Y. P.; Han, H.; Rajagopal, G.; Anderson, N. L. Candidate List of yoUr Biomarker (CLUB): A web-based platform to aid cancer biomarker research. Biomarker Insights 2008, 3, 65−71. (3) Anderson, N. L. The clinical plasma proteome: a survey of clinical assays for proteins in plasma and serum. Clin. Chem. 2010, 56 (2), 177−85. 1877

dx.doi.org/10.1021/pr201092v | J. Proteome Res. 2012, 11, 1868−1878

Journal of Proteome Research

Article

MALDI-TOF mass spectrometry for quantitative proteomic profiling. Clin. Proteomics J. 2004, 1, 259−70. (20) Bublitz, R.; Kreusch, S.; Ditze, G.; Schulze, M.; Cumme, G. A.; Fischer, C.; Winter, A.; Hoppe, H.; Rhode, H. Robust protein quantitation in chromatographic fractions using MALDI-MS of tryptic peptides. Proteomics 2006, 6 (13), 3909−17. (21) Thompson, L.; Turko, I.; Murad, F. Mass spectrometry-based relative quantification of human neutrophil peptides 1, 2, and 3 from biological samples. Mol. Immunol. 2006, 43 (9), 1485−9. (22) Kiernan, U. A.; Phillips, D. A.; Trenchevska, O.; Nedelkov, D. Quantitative mass spectrometry evaluation of human retinol binding protein 4 and related variants. PLoS One 2011, 6 (3), e17282. (23) Schmidt, A.; Kellermann, J.; Lottspeich, F. A novel strategy for quantitative proteomics using isotope-coded protein labels. Proteomics 2005, 5 (1), 4−15. (24) Shirran, S. L.; Botting, C. H. A comparison of the accuracy of iTRAQ quantification by nLC-ESI MSMS and nLC-MALDI MSMS methods. J. Proteomics 2010, 73 (7), 1391−403. (25) Kroot, J. J. C.; Laarakkers, C. M. M.; Geurts-Moespot, A. J.; Grebenchtchikov, N.; Pickkers, P.; van Ede, A. E.; Peters, H. P. E; van Dongen-Lases, E.; Wetzels, J. F. M.; Sweep, F. C. G. J.; Tjalsma, H.; Swinkels, D. W. Immunochemical and mass-spectrometry-based serum hepcidin assays for iron metabolism disorders. Clin. Chem. 2010, 56 (10), 1570−9. (26) Gutierrez, J. A.; Dorocke, J. A.; Knierman, M. D.; Gelfanova, V.; Higgs, R. E.; Koh, N. L.; Hale, J. E. Quantitative determination of peptides using matrix-assisted laser desorption/ionization time-offlight mass spectrometry. BioTechniques 2005, No. Suppl, 13−7. (27) Schuerenberg, M.; Luebbert, C.; Eickhoff, H.; Kalkum, M.; Lehrach, H.; Nordhoff, E. Prestructured MALDI-MS sample supports. Anal. Chem. 2000, 72 (15), 3436−42. (28) Gobom, J.; Schuerenberg, M.; Mueller, M.; Theiss, D.; Lehrach, H.; Nordhoff, E. Alpha-cyano-4-hydroxycinnamic acid affinity sample preparation. A protocol for MALDI-MS peptide analysis in proteomics. Anal. Chem. 2001, 73 (3), 434−8. (29) Kuzyk, M. A.; Smith, D.; Yang, J.; Cross, T. J.; Jackson, A. M.; Hardie, D. B.; Anderson, N. L.; Borchers, C. H. MRM-based, multiplexed, absolute quantitation of 45 proteins in human plasma. Mol. Cell. Proteomics 2009, 8 (8), 1860−77. (30) Yocum, A. K.; Gratsch, T. E.; Leff, N.; Strahler, J. R.; Hunter, C. L.; Walker, A. K.; Michailidis, G.; Omenn, G. S.; O’Shea, K. S.; Andrews, P. C. Coupled global and targeted proteomics of human embryonic stem cells during induced differentiation. Mol. Cell. Proteomics 2008, 7 (4), 750−67. (31) Fenyo, D.; Wang, Q.; DeGrasse, J. A.; Padovan, J. C.; Cadene, M.; Chait, B. T. MALDI sample preparation: The ultra thin layer method. J. Visualized Exp. 2007, 3, 192. (32) Smirnov, I. P.; Zhu, X.; Taylor, T.; Huang, Y.; Ross, P.; Papayanopoulos, I. A.; Martin, S. A.; Pappin, D. J. Suppression of alpha-cyano-4-hydroxycinnamic acid matrix clusters and reduction of chemical noise in MALDI-TOF mass spectrometry. Anal. Chem. 2004, 76 (10), 2958−65. (33) Baumgart, S.; Lindner, Y.; Kuhne, R.; Oberemm, A.; Wenschuh, H.; Krause, E. The contributions of specific amino acid side chains to signal intensities of peptides in matrix-assisted laser desorption/ ionization mass spectrometry. Rapid Commun. Mass Spectrom. 2004, 18 (8), 863−8. (34) Hortin, G. L.; Sviridov, D.; Anderson, N. L. High-abundance polypeptides of the human plasma proteome comprising the top 4 logs of polypeptide abundance. Clin. Chem. 2008, 54 (10), 1608−16. (35) Whiteaker, J. R.; Zhao, L.; Anderson, L.; Paulovich, A. G. An automated and multiplexed method for high throughput peptide immunoaffinity enrichment and multiple reaction monitoring mass spectrometry-based quantification of protein biomarkers. Mol. Cell. Proteomics 2009, 9 (1), 184−96. (36) Anderson, L.; Pope, M.; Jackson, A.; Pearson, T. W.; Werner, P.; Miller, C. Automation of a SISCAPA magnetic bead workflow for protein biomarker quantitation by mass spectrometry using the agilent bravo automated liquid handling platform; Application Note 5990−

7360EN; Agilent Technologies: Jan 25, 2011. http://www.chem.agilent. com/en-US/Search/Library/_layouts/Agilent/PublicationSummary. aspx?whid=70397&liid=5698.

1878

dx.doi.org/10.1021/pr201092v | J. Proteome Res. 2012, 11, 1868−1878