Natural Flanking Sequences for Peptides Included in a Quantification

Dec 18, 2014 - Crystal S. F. Cheung , Kyle W. Anderson , Kenia Y. Villatoro Benitez , Mark J. Soloski , John N. Aucott , Karen W. Phinney , and Illari...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/ac

Natural Flanking Sequences for Peptides Included in a Quantification Concatamer Internal Standard Crystal S. F. Cheung,†,‡ Kyle W. Anderson,†,‡ Meiyao Wang,†,‡,§ and Illarion V. Turko*,†,‡ †

Institute for Bioscience and Biotechnology Research, Rockville, Maryland 20850, United States Biomolecular Measurement Division, National Institute of Standards and Technology, Gaithersburg, Maryland 20899, United States

‡

S Supporting Information *

ABSTRACT: Quantification by targeted proteomics has largely depended on mass spectrometry and isotope-labeled internal standards. In addition to traditionally used recombinant proteins or synthetic peptides, concatenated peptides (QconCATs) were introduced as a conceptually new source of internal standard. In the present study, we focused on assessing the length of natural flanking sequences, which surround each peptide included in QconCAT and provide for identical rates of analyte and standard digestion by trypsin. We have expressed, purified, and characterized a set of seven 15Nlabeled QconCATs that cover seven tryptic peptides from human clusterin with a length of natural flanking sequences ranging from none (+0) to six amino acid residues (+6) for each tryptic peptide. Individual QconCATs were mixed with recombinant human clusterin at a 1:1 molar ratio and digested, and the actual ratios for each combination of peptide/flanking sequence were measured with a multiple reaction monitoring assay. Data analysis suggested that natural flanking sequences shorter than +6 residues can cause a quantitative error because the random appearance of other amino acid residues in close proximity to trypsin cleavage sites has unpredictable consequences for the digestion rates of QconCATs.

M

fortunately occupy a large portion of QconCAT in addition to the essential Q-peptides. Therefore, it is important to determine experimentally the shortest necessary size of flanking sequences that guarantee identical rates of proteolytic excision of the peptides from the QconCAT and target protein. In the present study, we selected seven Q-peptides from recombinant human clusterin and supplemented these peptides with natural flanking sequences ranging from zero (+0) to six (+6) amino acid residues. These peptides were assembled in seven QconCATs in a way that only one specific Q-peptide with one specific length of natural flanking sequence appears in each QconCAT. MRM measurements of 1:1 molar mixtures of clusterin and individual QconCATs have demonstrated that at least +6 amino acid residue natural flanking sequences for each Q-peptide are necessary for reliable quantification.

ultiple reaction monitoring (MRM) is a commonly used technique for targeted protein quantification in complex biological samples. Quantification by MRM assays typically relies on a stable isotope-labeled internal standard added to the biological sample. In 2005, a new technology, termed QconCAT, was proposed to produce isotopically labeled internal standards for MRM mass spectrometry by genetic engineering.1,2 QconCAT stands for quantification concatamer and is an artificial protein expressed from a synthetic gene and composed of multiple tryptic peptides (called Q-peptides) from proteins targeted for quantification. As originally proposed,1,2 Q-peptides are directly concatenated in the QconCAT. A pitfall of direct concatenation is that natural amino acid sequences surrounding the sites of trypsin-catalyzed cleavage are not identical in the QconCAT and the target protein.3 Because the efficiency of tryptic digestion is likely to be influenced by other residues in close proximity to the cleavage site,4−7 differences in the sequence composition around the cleavage sites in QconCATs and target proteins can cause an error in quantification. To avoid potential quantitative discrepancy, QconCATs with natural flanking sequences framing every Qpeptide were first proposed by Kito et al.8 Subsequently, four or six amino acid residue flanking sequences have been used by different researchers.9−12 Nearly identical efficiency of trypsin digestion for the target protein and QconCAT with flanking sequences was validated using recombinant clusterin and 15Nlabeled clusterin QconCAT.10 Flanking sequences can un© 2014 American Chemical Society

■

EXPERIMENTAL SECTION Materials. The DC protein assay kit was from Bio-Rad Laboratories (Hercules, CA). Ammonium chloride (15N, 99%) was purchased from Cambridge Isotope Laboratories (Andover, MA). Human recombinant clusterin (purity >95%) was from ProSpec-Tany TechnoGene Ltd. (Ness Ziona, Israel). Nickel nitrilotriacetic acid resin was from Qiagen (Valencia, CA). Received: October 2, 2014 Accepted: December 18, 2014 Published: December 18, 2014 1097

DOI: 10.1021/ac503697j Anal. Chem. 2015, 87, 1097−1102

Article

Analytical Chemistry

Figure 1. Characterization of QconCATs. (A) 100 pmol of purified QconCATs (from #1 to #7) were separated on 12.5% SDS-PAGE. The molecular mass standards are shown on the right. (B) MALDI spectra for two Q-peptides from QconCAT #1, which were used to calculate 15N incorporation by Isotopic Enrichment Calculator (http://www.nist.gov/mml/analytical/organic/isoenrichcalc.cfm). The MALDI spectra for other QconCATs (from #2 to #7) are summarized in Figure S-1 in the Supporting Information. (C) Full-size QconCAT expression was confirmed using an Agilent 6550 QTOF instrument. The slectrospray ionization MS spectrum of QconCAT #1 is shown. The inset shows a deconvoluted spectrum with observed average molecular mass of QconCAT #1. The spectra for other QconCATs (from #2 to #7) are summarized in Figure S-2 in the Supporting Information.

centrifugation at 4000g for 20 min and washed twice by 10 mL of a 15NH4Cl-containing M9 medium. Cells were then transferred to 500 mL of a 15NH4Cl-containing M9 medium. The expression was induced with 1 mmol/L isopropyl β-D-1thiogalactopyranoside at OD600 of 0.6−0.8 and incubated for an additional 3 h at 37 °C. Cells were divided into 10 portions and harvested by centrifugation at 4000g for 30 min. One portion of cells was resuspended in 20 mL of lysis buffer (50 mmol/L Tris·HCl, pH 7.5). Cells were disrupted by sonication and centrifuged at 20000g for 30 min. The supernatant was discarded. The pellet was resuspended in 3 mL of urea buffer (7 mol/L urea/100 mmol/L NaH2PO4/10 mmol/L Tris·HCl, pH 8.0), and 15N-labeled QconCAT was purified on a nickel nitrilotriacetic acid resin in batch mode. The protein concentration of purified 15N-labeled QconCAT was measured using the DC protein assay kit and bovine serum albumin as a standard. The Bio-Rad DC protein assay kit is a colorimetric assay based on well-documented Lowry assay and can be used for protein samples containing detergent. The purified 15NQconCAT was then aliquoted and kept frozen at −80 °C. To determine the molecular mass of QconCATs, liquid chromatography−mass spectrometry (LC−MS) analyses were performed in positive-ion mode with an Agilent 6550 quadrupole-time-of-flight (QTOF) mass spectrometer (Santa Clara, CA) coupled with an Agilent 1200 high-performance

Sequencing grade-modified trypsin was obtained from Promega Corp. (Madison, WI). RapiGest SF surfactant was obtained from Waters Corp. (Columbia, MD). All other chemicals were purchased from Sigma-Aldrich (St. Louis, MO). Expression, Purification, and Characterization of 15NLabeled QconCATs. Synthetic genes encoding seven QconCATs were synthesized by Biomatik Corp. (Cambridge, Ontario, Canada). The design of these QconCATs includes seven tryptic peptides (Q-peptides) from human clusterin with their natural flanking sequences on both sides of the Qpeptides. The length of the natural flanking sequences ranges from 0 to 6 amino acid residues for each Q-peptide. Each QconCAT has only one copy of a specific Q-peptide with a specific length of the natural flanking sequence. The synthetic genes were cloned into the NdeI/HindIII restriction sites of the pET21a expression vector in-frame to the C-terminal His6-tag. For expression, the plasmid was transformed into One Shot BL21(DE3) cells and cultivated at 37 °C in a M9 minimum medium containing 1 g/L 15NH4Cl as the sole nitrogen source. The initial inoculation started with 5 mL of LB media, and the cells were grown for 6−8 h at 37 °C. Cells were collected by centrifugation at 4000g for 20 min and washed once by 10 mL of a 15NH4Cl-containing M9 medium. Cells were then transferred to 50 mL of a 15NH4Cl-containing M9 medium and grown for 12−14 h at 37 °C. Cells were collected by 1098

DOI: 10.1021/ac503697j Anal. Chem. 2015, 87, 1097−1102

Article

Analytical Chemistry

Q2 (TLLSNLEEAK), Q3 (SGSGLVGR), Q4 (IDSLLENDR), Q5 (ASSIIDELFQDR), Q6 (ELDESLQVAER), and Q7 (VTTVASHTSDSDVPSGVTEVVVK). They were then supplemented with natural flanking sequences ranging from 0 to 6 amino acid residues on both sides of each Q-peptide and were arranged in seven QconCATs in a way that only one specific Qpeptide with only one specific length of natural flanking sequence appears in each QconCAT (Table S-1 in the Supporting Information). Characterization of QconCATs. The purity of QconCATs was estimated by SDS-polyacrylamide gel electrophoresis (PAGE) (Figure 1A) and considered to be nearly 100% pure. No correction for the protein concentration of the QconCATs was made for further MRM analysis, and the total protein concentration for the purified QconCATs was used as mg/mL, as determined by DC protein assay. 15 N isotope incorporation was determined based on MALDI spectra of Q-peptides by Isotopic Enrichment Calculator (http://www.nist.gov/mml/analytical/organic/isoenrichcalc. cfm). Figure 1B shows representative matrix-assisted laser desorption ionization (MALDI) spectra for two Q-peptides from QconCAT #1. The MALDI spectra for these Q-peptides from other QconCATs (from #2 to #7) are summarized in Figure S-1 in the Supporting Information. Both peptides yielded consistent results, and the mean value based on two peptides and three analytical replicates was higher than 99% for each QconCAT (Table 1). These values were accepted as complete labeling, and no correction was applied to the data.

liquid chromatograph (HPLC; Santa Clara, CA). The QconCAT was eluted from an Agilent ProtID C18 nanochip (75 μm × 150 mm, 300 nm) over a 10 min gradient from 20% to 80% acetonitrile containing 0.1% formic acid at a flow rate of 400 nL/min. The acquisition method in positive-ion mode used a capillary temperature of 275 °C, a fragmentor of 180 V, a capillary voltage of 1950 V, and a m/z 500−2000 mass window. Mass deconvolution was later performed using MagTran 1.0 software. Processing of Samples. A total of 20 pmol of human recombinant clusterin, 20 pmol of individual QconCAT, and 20 μg of bovine serum albumin were mixed in 25 mmol/L NH4HCO3 buffer/1% sodium dodecyl sulfate (SDS)/10 mmol/L dithiothreitol. The mixture was incubated at room temperature for 60 min to allow reduction of cysteines and was then treated with 55 mmol/L iodoacetamide for another 60 min to alkylate the reduced cysteines. Alkylated samples were precipitated with chloroform/methanol13 to deplete salts, urea, and SDS from the samples. Protein pellets were then sonicated in 100 μL of 25 mmol/L NH4HCO3/0.1% RapiGest SF surfactant and treated with trypsin for 15 h at 37 °C. The substrate/trypsin ratio was 10:1 (w/w). After trypsin digestion, the samples were treated with 0.5% trifluoroacetic acid for 30 min at 37 °C to break down acid-cleavable RapiGest. The insoluble byproduct of RapiGest was then removed by centrifugation at 106000g for 30 min. After centrifugation, the supernatants were dried using a vacuum centrifuge (Vacufuge, Eppendorf AG, Hamburg, Germany). LC−MS/MS Analysis. Dried peptides were reconstituted in 30 μL of 3% acetonitrile/97% water containing 0.1% formic acid, and 5 μL was used for a single LC−MS/MS run. Instrumental analyses were performed on an Agilent Zorbax Eclipse Plus C18 RRHD column (2.1 mm × 50 mm, 1.8 μm particle) coupled to an Agilent 6460 triple−quadrupole LC− MS system (Santa Clara, CA). Peptides were eluted over a 35 min gradient from 5% to 50% acetonitrile containing 0.1% formic acid at a flow rate of 200 μL/min. The gradient settings were 5−10% solvent B in 5 min, 10−30% solvent B in 25 min, and 30−50% solvent B in 5 min and then returned to 5% solvent B in 5 min. Solvent A was water containing 0.1% formic acid, and solvent B was 95% acetonitrile, 5% water, and 0.1% formic acid. The acquisition method used the following parameters in positive-ion mode: fragmentor, 135 V; electron multiplier, 500 V; capillary voltage, 3500 V. The collision energy was optimized for each peptide using the default equation from Agilent: CE = m/z 0.036 − 4.8. The dwell time for all transitions was set at 50 ms.

Table 1. Properties of QconCATs 15

QconCAT #1 #2 #3 #4 #5 #6 #7

N incorporation (%) 99.7 99.6 99.6 99.7 99.7 99.7 99.5

± ± ± ± ± ± ±

0.08 0.08 0.08 0.04 0.08 0.14 0.51

calcd mass (Da)

obsd mass (Da)

ppm

19427 19459 19301 19286 19430 19472 19265

19425 19425 19382 19365 19471 19472 19231

103 1747 4197 4096 2110 N/A 1765

To determine the molecular mass of the QconCATs, LC− MS analyses were performed in positive-ion mode with an Agilent 6550 QTOF instrument coupled with an Agilent 1200 HPLC. Figure 1C shows a representative spectrum and deconvolution for QconCAT #1. The related spectra for other QconCATs (from #2 to #7) are summarized in Figure S2 in the Supporting Information. Overall, the observed mass of each QconCAT was consistent with the calculated mass (Table 1), indicating full-length expression of the QconCAT constructs. Accordingly, the calculated average molecular mass was used to convert mg/mL concentration of QconCAT into the mol/L concentration, which was further used to mix QconCAT and recombinant clusterin at a 1:1 molar ratio. Measured Ratios as a Function of the Length of Natural Flanking Sequences. Three MRM transitions for each Q-peptide were selected as previously described14 and are tabulated in Table S-2 in the Supporting Information. These transitions were treated as independent measurements, each resulting in a ratio value for the corresponding Q-peptide. Measurements were performed with three analytical replicates for each of two biological replicates resulting in the total n = 18. The average value for each natural flanking sequence variant of

■

RESULTS AND DISCUSSION Design of QconCATs. The selection of Q-peptides has size and composition restraints. Peptides that contain less than eight amino acids are typically omitted because of small m/z values. Peptides with Cys or Met are also avoided because of the existence of various oxidation entities, which could introduce variations into the quantification. In the present study, human clusterin (apolipoprotein J) was selected as a model protein because this secreted protein has an average mass (52.5 kDa) and multiple Cys/Met but allows the selection of up to nine tryptic peptides, which comply with Q-peptide selection criteria. As a first step, the abundance of these Q-peptides in the tryptic digest of recombinant clusterin was determined and seven Q-peptides were selected based on the signal intensity. These seven Q-peptides were named Q1 (EIQNAVNGVK), 1099

DOI: 10.1021/ac503697j Anal. Chem. 2015, 87, 1097−1102

Article

Analytical Chemistry

Figure 2. Measured ratios. Recombinant human clusterin was mixed with individual QconCATs at a 1:1 (pmol/pmol) ratio resulting in seven separate samples. MRM assay for these samples was used to determine the actual measured ratios for each individual Q-peptide with different lengths of natural flanking sequences. (A) Measured ratios (mean ± SD, n = 18) plotted versus the length of the natural flanking sequence for all seven Qpeptides. Blue dashed lines border a section of the plot for ratios in the range of 1.0 ± 30%. (B) Data for Q2 (shown in red) used as a representative example to demonstrate how shorter natural flanking sequences (shown in black) cause an appearance of random nonoriginal amino acid residues (shown in blue) in close proximity to trypsin cleavage sites. Related data presentation for all of the other Q-peptides are summarized in Table S-3 in the Supporting Information.

sequence itself. This observation prompted us to analyze the actual amino acid sequences surrounding trypsin cleavage sites because shorter natural flanking sequences brought other random amino acid residues into close proximity to the trypsin cleavage sites. As a representative case, the data for Q2 are shown in Figure 2B. Related data for all other QconCATs are summarized in Table S-3 in the Supporting Information. Figure 2B allows a side-by-side comparison of the ratios for Q2, and these ratios depend on the changing pattern of amino acid residues surrounding trypsin cleavage sites, with natural flanking residues shown in black and random residues shown in blue. In general, these other random amino acid residues can have negative, positive, or no effect for trypsin digestion and subsequent peptide release. Indeed, +1 and +2 flanking combinations for Q2 do not differ from the +6 natural flanking sequence. However, +0 and +6 combinations have an approximately 2-fold difference. The same is true between the +4 and +6 combinations. Some of these deviations can be explained based on the so-called Keil rules (reviewed in refs3 and 17). For example, when +6 (KTNEERKTLLSNLEEAKKKKEDA) is compared to +0 (QIKTLIKTLLSNLEEAKCRSGSG), it is apparent that omitting most positively and negatively charged residues on both sides of Q2 in the +0 combination improves the rate of trypsin digestion, resulting in a higher yield of Q2 from QconCAT versus recombinant

a Q-peptide was normalized to the value obtained for the respective six amino acid natural flanking sequences of that Qpeptide. Normalization was performed to simplify evaluation of the effect of the flanking sequence length on each Q-peptide as well as to allow for a comparison between Q-peptides. Data in Figure 2A are presented as mean ± SD and plotted versus the length of the natural flanking sequence for all seven Q-peptides. Recombinant clusterin and QconCAT were mixed at a 1:1 molar ratio. If the presence of a natural flanking sequence is not essential for accurate quantification, then all measured ratios should be around 1.0 with a coefficient of variation (CV) in the range of 30% (shown with blue dashed lines in Figure 2A). CV in the range up to 30% is typical for MRM assays of proteins15,16 and covers uncertainties associated with measurements of total protein concentrations, pipetting during analyte/ standard mixing, and biological and analytical replicates. Some data presented herein fit well in this range. For example, Q7 shows little ratio variation associated with the length of the natural flanking sequence, and even the absence of a natural flanking sequence yields a ratio (1.05 ± 0.13) similar to the ratio measured for the presence of a six amino acid residue natural flanking sequence (1.0 ± 0.10). However, we have also observed multiple examples where measured ratios significantly deviate from the expected 1.0 ± 30%. These deviations did not correlate predictably with the length of the natural flanking 1100

DOI: 10.1021/ac503697j Anal. Chem. 2015, 87, 1097−1102

Article

Analytical Chemistry

QconCATs is often used for “relative” comparison of two (or more) biological samples. Acquiring quantification with up to ±2-fold error in both biological samples will not change their relative comparison in side-by-side experiments and can justify support of QconCATs without natural flanking sequences. However, there are several applications where truly absolute quantification is important, for example, interlaboratory comparisons, the discovery of post-translational modifications, a comparison of protein expression levels in different tissues/ species, etc. For all of these applications, it would be better to use at least +6 natural flanking sequences on each side of every Q-peptide being included in the QconCAT.

clusterin. Overall, it gives a measured ratio 0.45 ± 0.05 for the +0 combination. However, measured ratios for many other combinations cannot be so simply explained. For example, the difference in flanking sequences between the +4 and +5 combinations is small and difficult to relate to the Keil rules, while the difference in ratios is approximately 3-fold. Sequence-based analysis of other Q-peptides (Table S-3 in the Supporting Information) concurs with trends described here for Q2 and can be summarized in two assumptions. First, inclusion of +6 residue natural flanking sequences on Qpeptides closely mimics the native region in proximity to trypsin cleavage sites and provides the most accurate quantification of the target protein. Second, shorter natural flanking sequences could still be quantitative, and on some occasions (such as Q4, Q5, Q6, and Q7 in the present study), even no natural flanking sequence (+0 combination) performed quantitatively well (Figure 2A). However, this is not a guarantee because random residues brought into close proximity of trypsin cleavage sites can have unpredictable effects on the efficiency of trypsin digestion. Work by others has attempted to provide structural and mechanistic effects for cleavage site efficacy as a rationale for deviations from expected Keil rules, such as greater trypsinolysis of exposed sites within loop structures and decreased trypsin cleavage fidelity near negatively charged pockets created by bringing acidic residues into close proximity in the folded conformation rather than spaced out within the linear amino acid sequence.17 These contradictions to the Keil rules on a structural basis cannot be applied to our observations because both clusterin and QconCATs were denatured with SDS, thereby reducing the structural effects on trypsinolysis. While a minor degree of secondary structure may be present as a result of incomplete denaturation by SDS, it is unlikely that it will significantly impact cleavage by trypsin and, therefore, our observations can mostly be attributed to the primary amino acid sequence. Because variation between trypsin cleavage fidelity in native proteins and QconCATs is not completely predictable, several precautionary steps may be considered to reduce inconsistencies in quantification. First, and perhaps the simplest method to avoid error in measurement, is to include six or more natural amino acid residues on both sides of every Q-peptide. Inclusion of six or more natural amino acids will enable more reliable relative quantification, approaching absolute quantification, and will be appropriate for measurements within similar samples performed by consistent protocols. Second, and perhaps more laborious, is to optimize the natural flanking sequence length by conducting experiments similar to those provided herein to determine the appropriate length of the natural flanking sequence to yield a consistent 1:1 ratio between QconCAT and the analyte for each Q-peptide. Optimization would be preferred for QconCATs produced for interlaboratory comparisons and comparisons of protein concentrations between tissues or species where more accurate quantitative data are imperative. A third precaution that should always be considered when using peptides to quantify protein is to use three or more Q-peptides per target protein to reduce peptide-to-peptide variance and to indicate anomalies in poorly performing Qpeptides for their omission. An important observation in the present study is that the randomly occurring error of quantification associated with natural flanking sequences shorter than +6 residues was not greater than approximately 2-fold. Although the term “absolute” quantification is broadly used, MRM quantification with

■

CONCLUSIONS Although natural flanking sequences for Q-peptides are not directly used for quantification and occupy a large portion of the QconCAT, it appears that at least +6 natural amino acid flanking residues are certainly necessary for reliable quantification. It is evident from our observations that trypsin cleavage is not solely dependent on the length of the natural amino acid flanking sequence in the +0 to +6 residue range but rather on the properties of amino acid residues found within that range. Furthermore, the effects of amino acid composition within the flanking sequence cannot be predicted entirely by Keil rules of trypsin cleavage. We suggest performing experiments similar to those performed herein to optimize each Q-peptide flanking sequence to confirm accurate quantification or to simply include +6 or more natural amino acid flanking residues.

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Present Address

§ M.W.: Analytical Biotechnology, MedImmune LLC, Gaithersburg, MD 20878.

Notes

The authors declare no competing financial interest.

■

REFERENCES

(1) Beynon, R. J.; Doherty, M. K.; Pratt, J. M.; Gaskell, S. J. Nat. Methods 2005, 2, 587−589. (2) Pratt, J. M.; Simpson, D. M.; Doherty, M. K.; Rivers, J.; Gaskell, S. J.; Beynon, R. J. Nat. Protoc. 2006, 1, 1029−1043. (3) Chen, J.; Turko, I. V. Trends Anal. Chem. 2014, 57, 1−5. (4) Hubbard, S. J. Biochim. Biophys. Acta 1998, 1382, 191−206. (5) Yen, C. Y.; Russell, S.; Mendoza, A. M.; Meyer-Arendt, K.; Sun, S.; Cios, K. J.; Ahn, N. G.; Resing, K. A. Anal. Chem. 2006, 78, 1071− 1084. (6) Thompson, R.; Hodgman, T. C.; Yang, Z. R.; Doyle, A. K. Bioinformatics 2003, 19, 1741−1747. (7) Lawless, C.; Hubbard, S. J. OMICS 2012, 9, 449−456. (8) Kito, K.; Ota, K.; Fujita, T.; Ito, T. J. Proteome Res. 2007, 6, 792− 800. (9) Nanavati, D.; Gucek, M.; Milne, J. L.; Subramaniam, S.; Markey, S. P. Mol. Cell. Proteomics 2008, 7, 442−447. (10) Chen, J.; Wang, M.; Turko, I. V. Mol. Neurodegener. 2012, 7, 41. (11) Chen, J.; Wang, M.; Turko, I. V. Anal. Chem. 2013, 85, 303− 307. 1101

DOI: 10.1021/ac503697j Anal. Chem. 2015, 87, 1097−1102

Article

Analytical Chemistry (12) Chen, J.; Huang, R. Y. C.; Turko, I. V. Anal. Chem. 2013, 85, 6011−6017. (13) Liao, W.-L.; Turko, I. V. Anal. Biochem. 2008, 377, 55−61. (14) Liao, W.-L.; Heo, G.-Y.; Dodder, N. G.; Pikuleva, I. A.; Turko, I. V. Anal. Chem. 2010, 82, 5760−5767. (15) Rivers, J.; Simpson, D. M.; Robertson, D. H. L.; Gaskell, S. J.; Beynon, R. J. Mol. Cell. Proteomics 2007, 6, 1416−1427. (16) Chen, C.-D.; Wang, C.-L.; Yu, C.-J.; Chien, K.-Y.; Chen, Y.-T.; Chen, M.-C.; Chang, Y.-S.; Wu, C.-C.; Yu, J.-S. J. Proteome Res. 2014, 13, 2818−2829. (17) Vandermarliere, E.; Mueller, M.; Martens, L. Mass Spectrom. Rev. 2013, 32, 453−465.

1102

DOI: 10.1021/ac503697j Anal. Chem. 2015, 87, 1097−1102