Quantitative performance of internal standard platforms for absolute protein quantification using MRM-MS Kerry Bauer Scott1, 2, *, Illarion V. Turko1, 2, Karen W. Phinney1

1Biomolecular Measurement Division, National Institute of Standards and Technology,

Gaithersburg, MD 20899

2Institute for Bioscience and Biotechnology Research, Rockville, MD 20850

ABSTRACT

Stable-isotope-labeling mass spectrometry involves the addition of known quantities of stable-

isotope labeled standards, which mimic native molecules, to biological samples. We evaluated

three conventional internal standard platforms (synthetic peptides, QconCAT constructs, and

recombinant proteins) for quantitative accuracy, precision, and inherent advantages and

limitations. Internal standards for the absolute quantification of three human cytokine proteins

(interferon gamma, interleukin-1 beta, and tumor necrosis factor alpha) were designed and

verified. Multiple reaction monitoring assays, calibration curve construction, and regression

analysis were used to assess quantitative performance of the internal standard platforms. We

also investigated a strategy for methodological improvement to current platforms using natural

flanking sequences. Data analysis revealed that full length protein standards have the broadest

quantitative reliability with accuracy being peptide-dependent for QconCATs and synthetic

peptides. Natural flanking sequences greatly improved the quantitative performance of both

QconCAT and synthetic peptide standards.

INTRODUCTION

Protein quantification methods based on the principle of multiple reaction monitoring (MRM) with

stable-isotope labeled standard peptides provide absolute protein quantification values. These

methods are based on the addition of known quantities of isotope-labeled internal standards into

biological samples. The internal standards are analogous to native molecules and quantification

is achieved by comparing ion signals from isotope-labeled and native peptides. These

methodologies are broadly applicable to clinical biomarkers, systems biology, and the

pharmaceutical industry; which require accurate, precise, and reproducible protein abundance

measurements. Several internal standard platforms exist for protein quantification, with the three

conventional internal standard platforms including synthetic peptides, concatenated peptides

(QconCAT)1, and recombinant proteins. Each method has its own strengths and limitations in

terms of production, application, and analytical performance.2

A principle limitation of synthetic peptide and concatenated peptide standards is that amino acid

sequences surrounding the tryptic cleavage sites are not identical between the standard and the

native target protein. QconCATs are assembled with peptides from multiple different proteins.

These peptides are ordered in the construct such that local primary sequences are preserved

when possible. However, sequence context is not always able to be preserved. This difference

between standard and analyte presents potential for quantitative error since the efficiency of the

tryptic digestion is influenced by the amino acids neighboring the cleavage sites.3

Inclusion of natural flanking sequences around the cleavage sites of peptides, to better mimic

the target protein, has become a new trend.4 Natural flanking sequences have been

incorporated into QconCAT constructs5–8 and synthetic peptides9–12 by different researchers. A

comparison of concatenated peptides with and without flanking sequences showed that the

presence of flanking regions improved quantitative accuracy5 whereas a study comparing

cleavable and non-cleavable synthetic peptides showed no difference in accuracy or precision.9

The length of natural flanking sequences in the literature has varied in the number of residues

included in the flanking region, between two to six amino acids. There is a need to

experimentally determine the optimal size of flanking sequence to match peptide release from

the native protein. Furthermore, determining the minimal flanking length necessary to emulate

native analyte signal is important as QconCAT constructs and synthetic peptides have size

limitations for expression and synthesis.

Reviews have broadly discussed advantages and disadvantages of various stable isotope

internal standards, but few experimental comparisons between the methods have been

performed. Brun et al compared three standards including recombinant proteins, QconCAT, and

synthetic peptides by LC-MS analysis.13 They reported the superior performance of

recombinant protein standards for quantification. Two comparisons of QconCAT and peptide

standards have been conducted; one study evaluated the fidelity of the standards in terms of

production.14 The other study compared the performance of different peptides in the QconCAT

from the synthetic peptides.15 Finally, a study evaluated both recombinant protein and synthetic

peptide standards with improved accuracy and precision using the protein standard.16

This is the first systematic, parallel comparison between recombinant protein, QconCAT, and

synthetic peptide standards designed to also assess the use of natural flanking sequences and

peptide digestion kinetics. For this study, we randomly selected three clinically relevant proteins

(interferon gamma (IFNG), interleukin beta (IL1B), and tumor necrosis factor alpha (TNFA)) to

address conceptual questions regarding internal standard strategies. We developed

recombinant proteins, QconCAT constructs, and synthetic peptides for the quantification of

three human cytokine proteins. Calibration curve construction and regression analysis was used

to assess the quantitative accuracy, precision, and inherent strengths and limitations. A

methodological improvement using natural flanking sequences demonstrated improved

performance for QconCAT standards and synthetic peptides.

EXPERIMENTAL SECTION

Materials. All chemicals were purchased from Sigma-Aldrich (St. Louis, MO), unless otherwise

noted.

Synthetic Peptides. Synthetic peptides were synthesized by Biomatik (Cambridge, Ontario) in

their unlabeled, natural isotope form. The purity was ≥98 % for all peptides, therefore peptide

peak areas were not corrected for purity. Peptides were solubilized according to manufacture

guidelines and stock peptide solutions were prepared in water at 1 µg/µL. Absolute

concentrations were confirmed by amino acid analysis. The molecular weight of the peptides

was verified by LC-MS and all peptides had their expected monoisotopic masses.

QconCAT Design. Two QconCAT constructs were designed to contain 3 to 4 tryptic peptides of

IL1B, IFNG, and TNFA concatenated with either no flanking sequences or flanking sequences

of six amino acids on both the N-terminal and C-terminal sides. Tryptic peptides were selected

following experimental evaluation of LC-MS detectability.

Protein Expression and Purification. The amino acid sequence of IL1B, IFNG, TNFA, and

two QconCAT constructs were coded into the corresponding DNA sequence and incorporated

into the pET21a expression vector, with codon optimization for E. coli (Biomatik, Cambridge,

Ontario). The plasmids were transformed into One Shot BL21 (DE3) competent E. coli cells

(Invitrogen, Grand Island, NY) and grown in M9 minimal media at 37 °C until the optical density

(OD) reached 0.6 to 0.8 at 600 nm. Protein expression was induced by 0.5 mmol/L isopropyl β-

D-1-thiogalactopyranoside (IPTG). After 3 h of growth, the cells were harvested by

centrifugation at 5000 g for 10 min and resuspended in 0.1 mmol/L dithiothreitol (DTT) and

sonicated. Following centrifugation at 35000 g for 30 min, supernatant containing IL1B was

collected. Inclusion bodies containing IFNG, TNFA, and both QconCATs were resuspended in 8

mol/L urea and clarified by centrifugation at 10000 g for 10 min. Recombinant protein

purification was achieved based on the 6xHis-Tag on a Ni-NTA agarose column (Qiagen,

Valencia, CA). The binding, washing, and eluting buffers were 8 mol/L urea, 100 mmol/L

Na2HPO4, 10 mmol/L Tris-HCl at pH 8.0, 6.0, and 4.5, respectively. Buffer exchange was done

from 8 mol/L urea to 50 mmol/L ammonium bicarbonate (NH4HCO3) using an Amicon filter (3

kDa MWCO, Millipore, Billerica, MA). QconCAT proteins were purified using HisPur cobalt resin

(Thermo Scientific, Waltham, MA) using the gravity-flow method. Columns were equilibrated,

loaded, and washed with 50 mmol/L sodium phosphate, 300 mmol/L sodium chloride, and 10

mmol/L imidazole at pH 7.4. QconCAT proteins were eluted with 50mmol/L sodium phosphate,

300 mmol/L sodium chloride, and 150 mmol/L imidazole at pH 7.4. Eluted proteins were

precipitated with methanol/chloroform/water and resuspended in 0.1 % RapiGest (Waters,

Milford, MA). Protein concentration was determined using a BCA protein assay and bovine

serum albumin as a standard (Thermo Scientific). Protein expression and purification were

evaluated with SDS-PAGE and mass spectrometry analysis on a 4700 Proteomics Analyzer

MALDI TOF/TOF (AB Sciex, Framingham, MA).

15N Incorporation. Plasmid transformed DE3 cells were expressed in M9 minimal media with 1

g/L 15NH4Cl (Cambridge Isotope Laboratories, Andover, MA) as the sole nitrogen source and

purified as described. Stable isotope incorporation into full length proteins was determined at

the peptide level with multiple reaction monitoring (MRM) analysis. The pair transitions for the

light (unlabeled) and heavy (15N labeled) form of each peptide were monitored in a sample

containing 15N labeled IL1B, IFNG, and TNFA. Incorporation was calculated as the percentile of

the area of the labeled peak to the sum of the labeled and unlabeled peaks. The final isotope

incorporation is based on combined data for three peptides and was found to be greater than 99

% for all proteins, so no correction for protein concentration was made during data analysis.

Digestion Optimization. Equal amounts of IL1B, IFNG, and TNFA were combined and

digested with one of 8 methods: (1) 50 mmol/L NH4HCO3, pH 8.0/0.1 % RapiGest, (2) 50

mmol/L NH4HCO3, pH 7.0/0.1 % RapiGest, (3) 8 mol/L urea (urea samples were desalted using

Zebra spin columns (7 kDa molecular weight cutoff, Thermo Scientific) prior to enzymatic

digestion), (4) 80 % acetonitrile (acetonitrile was added prior to enzymatic digestion), or (5) 0.05

% SDS. In all cases, proteins were reduced with 5 mmol/L DTT at 56 °C for 60 min and

alkylated with 15 mmol/L iodoacetamide in the dark for 30 min prior to enzymatic digestion.

Samples were either subjected to digestion with trypsin at a 1:50 protein to enzyme mass ratio

at 37 °C or (6) 48 °C. Digestions were also performed with (7) trypsin at a 1:5 protein to enzyme

mass ratio or (8) trypsin/lys-C mix (Promega, Madison, WI) at a 1:25 protein to enzyme mass

ratio. All digestion reactions were quenched after 16 h by addition of formic acid to 2.5 %. Equal

amounts of 15N labeled IL1B, IFNG, and TNFA were digested following method (1) and used as

a spike-in standard.

Digestion Time Course. Equal amounts of mixed unlabeled and labeled IL1B, IFNG, and

TNFA were used to perform a time course digestion. Sample digestion was quenched at (0.5, 1,

2, 4, 8, 16, 24, and 48) h after incubation with trypsin/lys-C at 37 °C for the 14N proteins and

after a single 16 h incubation for the 15N protein mixture which was used as a spike-in standard.

All digestions were performed in the presence of 0.1 % RapiGest.

Calibration Curves. Calibrant sample preparation was completed by combining equimolar

amounts of purified, 15N labeled IL1B, IFNG, and TNFA at a range of concentrations to give

(10.0, 5.0, 2.5, 1.0, 0.5, and 0.25) pmol/µL of each. Unlabeled purified, full-length IL1B, IFNG,

and TNFA, QconCAT standards, and synthetic peptide standards were spiked-in each calibrant

sample at 2.5 pmol/µL. Calibrants and standards were combined prior to enzymatic digestion.

The digestion conditions used for protein quantification are as follows: 0.1 % RapiGest, 1:5

trypsin: protein mass ratio, and a 4 h incubation at 37 °C.

LC-MS/MS Analysis. Peptide separation and MRM analysis were performed on an Agilent

Technologies 1200 series coupled to an AB Sciex API 5000 triple quadrupole mass

spectrometer with a QJet ion guide. Separation of peptides was performed with an Agilent C18

HPLC column (2.1 mm x 150 mm, 3.5 µm) at a flow rate of 0.2 mL/min over a 20 min gradient

from 5 % to 45 % acetonitrile containing 0.1 % formic acid. Acquisition methods used the

following parameters: ion spray voltage of 3000 V, curtain gas of 69 kPa (10 psi), ion source gas

of 379 kPa (55 psi), and source temperature of 600 °C. Scheduled MRM was performed with a

60 s MRM detection window and 1 s target scan time. An initial list of MRM transitions were

experimentally screened for the three most intense transitions per peptide. A single run

consisting of a series of MRM transitions at differing collision energies (CE) and declustering

potentials (DP) was created by making incremental adjustments of the product ion m/z value at

the hundredth decimal place.17 The CE and DP values were varied by ±15 in 5 steps of 3

relative to the default equations CE = 0.036(Q1) + 8.8 for doubly and CE = 0.0544(Q1) - 2.4 for

triply charged precursors and DP = 0.0729(Q1) + 31.117. MRM data was also collected on an

Agilent 6460 QQQ LC/MS system with an Agilent Technologies 1200 series HPLC (Santa Clara,

CA). Acquisition methods were as follows: fragmentor 135 V, electron multiplier 500 V, and

capillary voltage 3500 V. Dynamic MRM scan type was used with a delta retention time of 60 s.

Collision energies were optimized for each peptide from the default Agilent equations: CE =

0.031(Q1) + 1 for doubly and CE = 0.036(Q1) - 4.8 for triply charged precursors as described for

the API 5000. Optimize instrument parameters are shown in Supplemental Information Table S-

1.

Data Analysis. MRM peak area integration was performed using Skyline 2.5, AB SCIEX

Analyst 1.6, or Agilent MassHunter Qualitative Analysis B.06 and an Excel spreadsheet was

used to calculate peak area ratios. Peak integration was manually inspected and adjusted if

necessary. The peak ratios from three transitions were averaged to yield the peptide ratios.

Representative peptide ratios were plotted versus digestion time to determine optimal digestion

incubation. Calibration curves were constructed by linear regression analysis of the peak area

ratios plotted versus the molar ratios of the calibrant to the standard. All experiments were

performed in duplicate with three replicate injections to assess error and reproducibility. Data is

represented as the mean ± SD.

RESULTS AND DISCUSSION

Calibrant and Standard Characterization. Proteotypic peptides for quantification of each

cytokine protein were experimentally determined. A list of peptides was generated from an in

silico tryptic digestion and corresponding light and heavy transitions were calculated using the

OrgMassSpecR program (http://orgmassspecr.r-forge.r-project.org). Peptides were selected

based on signal intensity, lack of cysteine and methionine residues, uniqueness, and molecular

weight and screened for the three most intensive transitions per peptide. (Supplemental

Information Table S-2). These transitions were used to access the level of stable isotope

incorporation into the 15N-labeled calibrant proteins calibrants. The isotopic incorporation was

calculated from the percentile of the area of the labeled peak to the sum of the labeled and

unlabeled peaks based on three Q-peptides per protein. IL1B, IFNG, and TNFA were found to

have 99.9 % ± 0.1 %, 99.3 % ± 0.2 %, and 99.9 % ± 0.1 % isotopic incorporation, respectively.

These values were accepted as complete labeling and no correction for labeling efficiency was

made during data analysis (Supplemental Information Figure S-1).

The amino acid sequence of QconCAT Q1 includes ten Q-peptides, three from IL1B, three from

IFNG, and four from TNFA. Expressed QconCAT Q2 encodes for the same ten Q-peptides with

natural flanking sequences of six amino acids on both the C-terminal and N-terminal sides of the

peptides (Supplemental Information Figure S-2). The expression and the molecular weights of

the full length recombinant and QconCAT proteins were assess by SDS-PAGE and found at

their expected molecular weights with high purity (Supplemental Information Figure S-3).

Digestion Condition Optimization. It has been well established that digestion conditions

significantly influence the number of measureable peptides and individual peptide

concentrations, highlighting the need to identify reaction conditions that ensure complete and

reproducible digestion prior to accurate quantification.18–21 Several conditions for protein

digestion were evaluated for optimal peptide release including denaturant, pH, temperature,

enzyme, and enzyme concentration (Figure 1). Unfortunately, but not unexpectedly, a single set

of digestion conditions optimal for all peptides was not found. Instead, various digestion

parameters were found to yield the highest signal for the peptides analyzed. This

nonstoichiometry of protein digestions has been previously seen18 and is attributed to denatured

protein structure and the identify of amino acid residues surrounding the cleavage sites. For

example, the highest peak area ratio for peptide IPVALGLK was achieved using a 1:5 trypsin:

protein mass ratio while the trypsin/Lys-C mix yield the highest signal ratio for peptide

SLVMSGPYELK. Interestingly, these peptides are from the same protein, ILB1. There was also

variation between peptide digestion efficiency depending on the denaturant used. Either urea or

RapiGest as denaturant was optimal for most peptides, but one peptide showed enhanced

signal with SDS.

A pH change from 8.0 to 7.0 had the least affect on peak area ratios with an average ratio of

1.17 and less than 30 % variability among the peptides evaluated. None of the peptides showed

enhanced digestion in the presence of acetonitrile or incubation at 48 °C. Digestion with both

trypsin at a ratio of 1:5 and a trypsin/Lys-C mix at a ratio of 1:25 outperformed trypsin at a

protein ratio of 1:50. Overall, digestion with trypsin at a protein ratio of 1:5 lead to the maximal

number of peptides with peak area ratios greater than one and was used for calibration curve

generation. Denaturation with 0.1 % RapiGest was selected for calibration curve sample

preparation since the use of urea resulted in high variability in the extent of peptide digestion.

Peptide Release and Stability. In addition to digestion conditions affecting peptide quantities,

the dynamic production and decay characteristics of peptides during digestion has been shown

to impact quantification results.18,19,22 Time course digestion profiles for peptide release and

stability from IL1B, IFNG, and TNFA reveal various peptides from each protein to have different

formation and stability characteristics. The peptides DDKPTLQLESVDPK from IL1B, FFNSNK

from IFNG, and VNLLSAIK and IAVSYQTK from TNFA display rapid release, reaching

maximum signal by 4 h, and maintain steady signals with slow decay through 24 h of digestion

(Figure 2A). Peptides that are released rapidly and show stability over time are optimal

candidates for quantification by standard peptides as quantitative accuracy will be maintained.

The peptides SLVMSGPYELK from IL1B, LTNYSVTDLNVQR from IFNG, and

ANALLANGVELR from TNFA show rapid formation with continual signal decrease over the

remainder of the digestion (Figure 2B). Peptides that are released rapidly, but also exhibit rapid

signal decay will lead to variability in quantitative accuracy. Discrepancies between the synthetic

standard and sample peptides will arise, especially if longer digestion times are used or if the

standard peptide is added post-digestion.22

The peptides IPVALGLK from IL1B and DDQSIQK from IFNG exhibit slow production without

reaching a signal plateau after 24 h (Figure 2C). The sequences surrounding the cleavage sites

of these peptides have been previously shown to promote missed cleavages and are classified

as slow trypsin cleavage sites, explaining the slow formation.3,23,24 The amino acids flanking

IPVALGLK are rich in negatively charged, acidic residues (Asp and Glu) while peptide

DDQSIQK has a Phe residue at position P2, relative to the trypsin cleavage site. Peptides that

are slowly released from the target protein are not optimal for use as surrogate standard

peptides for quantification. The relative amounts of these peptides in the sample may be less

than the amount of synthetic standard peptide added. This will result in an over estimation of

target protein present in the sample.

Analytical Strategy. Calibration curves were constructed from the LC-MRM analysis of a series

of tryptic digests where 6 different concentrations of heavy isotope labeled proteins (IL1B, IFNG,

and TNFA) were spiked with a constant concentration of standard full length protein, QconCAT,

or synthetic peptide. Each sample was injected in triplicate and three transitions were

monitored for each peptide resulting in 9 ratio determinations per concentration. Linear

regression analysis was performed on the peak area ratios (heavy/light) versus concentration

ratios for all target peptides. Inferences regarding quantitative accuracy, precision, digestion

efficiency, and linearity, were used to comparatively evaluate each quantitative scheme.

Calibration curves were generated from experimental data sets collected on both an Agilent

6460 QQQ and an AB Sciex API 5000 triple quadrupole mass spectrometer. Supplemental

Information Table S-3 summarizes the calibration curve accuracy and linearity for all standards

and target peptides across both instrument platforms. Regression analysis results showed high

reproducibility across instrument platforms, but results from the 6460 QQQ are highlighted.

Furthermore, regression analysis based on individual peptide transitions showed high

correlation (r2 ≥ 0.998) (Supplemental Figure S-4).

Method Quantification Comparison. Full length protein standards resulted in slopes very

close to unity (slope accuracy of 0.58 to 1.15 accepted as unity) within the set confidence

interval and precise mean peak area ratios with a maximum CV of 13.3 % and median CV of 3.2

%. A slope of unity reveals accurate peptide quantification and complete digestion. The

calibration curves also had strong linearity with r2 ≥ 0.993 except for peptide IPVALGLK (Figure

3).

The QconCAT construct without natural flanking sequences (Q1) resulted in variable

performance depending on the peptide. Three peptide calibration curves for ANALLANGVELR,

FFNSNK, and DDQSIQK had slopes near unity (1.04 ± 0.01, 1.03 ± 0.03, 0.87 ± 0.06) and

strong linearity (r2 = 0.999, 0.997, 0.982) suggesting accurate quantification (Figure 3).

However, the calibration curves for several peptides within Q1 had slope values above or below

unity indicating reduced or enhanced digestion efficiency and inaccurate quantification (Figure

4). Peptides VNLLSAIK, IAVSYQTK, and LTNYSVTDLNVQR had slopes of 28.0 ± 0.33, 2.02 ±

0.09, and 2.03 ± 0.04, respectively, indicating inhibited release from the QconCAT compared to

the full length protein standard. Peptide IPVALGLK had a slope of 0.38 ± 0.02 signifying

enhanced release from the QconCAT compared to the full length protein.

Analysis of the amino acid sequences surrounding the tryptic cleavage sites proved insufficient

to reliably predict peptide release from the QconCAT.. The release of some peptides from the

QconCat, in comparison to the full length proteins, could be explained by primary sequence. For

example, peptide DDQSIQK has an adjacent Phe residue, known to promote missed cleavages,

in the native protein as well as the QconCAT resulting in similar digestion efficiencies and equal

peptide release.24 In the QconCAT, IAVSYQTK has an acidic Glu residue proximal to a tryptic

cleavage site which is known to promote missed cleavages.25 The cleavage sites in the full

length protein are surrounded by neutral residues, not known to influence trypsin digestion

efficiency, explaining the decreased yield observed for this peptide from the QconCAT.

However, sequence based explanations are not as clear for other peptides. LTNYSVTDLNVQR

has missed cleavage promoting Glu and dibasic residues near its cleavage site in the full length

protein and neutral amino acids in the QconCAT. Yet, release of this peptide is decreased in the

QconCAT compared to the calibrant protein. FFNSNK has a missed cleavage promoting dibasic

site in the native protein which is replaced by a neutral residue in the QconCAT leading to

incorrect predictions of enhanced peptide release from the QconCAT. Finally, release of

VNLLSAIK was reduced in the QconCAT even though the cleavage sites for VNLLSAIK are

surrounded by neutral residues in both the QconCAT and full length protein.

Furthermore, peptides found to have slopes not equal to unity were not limited to a specific

peptide class. Peptide IPVALGLK showed slow release, LTNYSVTDLNVQR decayed rapidly

after being released, and IAVSYQTK was a rapidly released, stable peptide. Based on these

results, the selection of peptides and peptide order in QconCAT construct design is not

completely predictable. Some peptides predicted to perform quantitatively well based on primary

sequence analysis near the cleavage sites showed poor quantitative accuracy and vice versa.

Quantification Using QconCATs with Flanking Sequences. Calibration curves and

regression analysis for a QconCAT construct with natural flanking sequences of six amino acids

(Q2) showed improved quantitative performance compared to Q1 (Figure 4). Peptides

VNLLSAIK, IAVSYQTK, and LTNYSVTDLNVQR, and IPVALGLK had slopes of 1.06 ± .02, 1.03

± 0.01, 1.09 ± 0.0, and 0.93 ± 0.03, respectively and r2 values ≥ 0.995. The mean peak area

ratios for Q2 had a maximum CV of 33.8 % and median CV of 3.5 %, slightly lower precision

than observed with the full length proteins. The inclusion of natural flanking sequences of six

amino acids resulted in slopes near unity for peptides that showed inaccurate quantification

when non-native residues appeared in close proximity to trypsin cleavage sites. These results

suggest that accurate quantification using QconCAT technology can be achieved with a few

caveats. Peptide release from the native proteins and QconCAT proteins must be equimolar as

peptides with enhanced or inhibited proteolytic release will have negative effects on quantitative

accuracy leading to overestimation or underestimation of protein concentration. This makes

QconCAT design and optimization of peptide order very important. However, it may not be

readily evident which amino acids will influence digestion efficiency or the number of residues

surrounding the cleavage site that should be considered. Inclusion of natural flanking

sequences of six residues within the QconCAT design enabled peptide release most similar to

the full length protein and allowed for accurate quantification of more peptides.

Quantification Using Standard Peptides with Cleavage Sites. Calibration curves were

constructed from a series of labeled calibrant proteins spiked with unlabeled synthetic peptides.

The synthetic peptides consisted of the native peptide or included natural flanking sequences of

variable length (Supplemental Information Table S-4). The natural flanking sequences provided

cleavable sites within the synthetic peptide to better mimic the cleavage characteristics of the

native peptide. Quantification by amino acid analysis was performed post resolubilization to

address two main concerns regarding the use of synthetic peptides, accurate concentration

determination and quantitative resolubilization. Two peptides, DDKPTLQLESVDPK and

IPVALGLK, with flanking sequences of 0 (tryptic peptide alone), 2, 4, or 6 amino acids were

added to the calibrants prior to digestion to account for peptide degradation during this process.

Regression analysis showed slopes equal to or near unity for peptide DDKPTLQLESVDPK

regardless of presence or length of flanking sequence. This peptide was an optimal candidate

for quantification, especially by synthetic peptides, because it displayed rapid release and slow

decay during the time course digestion experiments. The cleavage sites of peptide

DDKPTLQLESVDPK are also bordered by amino acids not known to promote missed cleavages

or influence tryptic digestion accounting for why similar levels of standard peptide with and

without flanking sequences were detected.

The calibration curve for synthetic peptide IPVALGLK +0 yield a slope of 0.04 ± 0.00 which

would result in an overestimation of protein concentration as the concentration of this standard

peptide was not dependent on tryptic digestion for peptide release and subsequent detection

(Figure 5). This peptide displayed slow peptide release in the time course study, not even

reaching a signal plateau after 24 h of digestion. The tryptic cleavage sites of IPVALGLK are

surrounded by acidic Asp and Glu residues which are known to promote missed cleavages in

tryptic digestions. Both peptides with flanking sequences of 2 amino acids showed incomplete

tryptic digestion with high signal intensity for missed cleavage of the N-terminal internal

cleavage site and low signal intensity for the non-cleaved and fully tryptic peptides

(Supplemental Information Figure S-5). Calibration curves were not constructed for this

synthetic peptide. Peptides with 4 and 6 residue flanking sequences lead to slopes of 0.15 ±

0.02 and 0.77 ± 0.06, respectively (Figure 5). The quantitative discrepancies associated with

peptides that are slowly released or rapidly decay may be compensated with the inclusion of

natural flanking sequences. The use of natural flanking sequences to incorporate a cleavage

site in synthetic peptides will provide more freedom in the types of peptides that can be used

with quantitative accuracy, such as peptides with these non-optimal digestion kinetics. This is

especially important for small proteins with inherently fewer tryptic cleavages from which to

select quantitative peptides. Our results show significant improvement in quantitative

performance for peptide IPVALGLK with the addition of flanking sequences 6 amino acids in

length. Longer flanking sequences may provide higher quantitative accuracy, but may be more

expensive and challenging to synthesize. The additional expense of synthesizing longer

peptides could be offset by minimizing the need to screen peptides for optimal digestion

efficiency and the increased likelihood that the peptides will provide accurate quantification,

limiting the need to synthesize additional peptides.

CONCLUSIONS

We are the first to report a comprehensive, side by side study of recombinant protein,

QconCAT, and synthetic peptide standards including the evaluation of natural flanking

sequences. Overall, full length proteins provided robust accuracy based on slopes of unity,

linearity, and high precision for all target peptides. This method also minimizes the evaluation

process of optimizing amino acid sequence patterns surrounding cleavage sites and evaluating

digestion efficiencies of candidate quantitative peptides. Moreover, the MRM assays developed

for this study could be utilized for the quantification of three clinically important proteins in a

myriad of applications.

ASSOCIATED CONTENT

Corresponding author

* Address: 9600 Gudelsky Dr., Rockville, MD 20850, E-mail: kerry.bauer@nist.gov

Notes

The authors declare no competing financial interest.

Acknowledgments

Certain commercial materials, instruments, and equipment are identified in this manuscript in

order to specify the experimental procedure as completely as possible. In no case does such

identification imply a recommendation or endorsement by the National Institute of Standards

and Technology nor does it imply that the materials, instruments, or equipment identified are

necessarily the best available for the purpose.

Supporting Information Available

Additional information as noted in text. This material is available free of charge via the Internet

at http://pubs.acs.org.

REFERENCES

(1) Beynon, R. J.; Doherty, M. K.; Pratt, J. M.; Gaskell, S. J. Nat. Methods 2005, 2 (8), 587–589.

(2) Brun, V.; Masselon, C.; Garin, J.; Dupuis, A. J. Proteomics 2009, 72 (5), 740–749. (3) Keil, B. Specificity of Proteolysis; Springer-Verlag: Berlin-Heidelberg-New York, 1992. (4) Chen, J.; Turko, I. V. Trends Anal. Chem. 2014, 57, 1–5. (5) Kito, K.; Ota, K.; Fujita, T.; Ito, T. J. Proteome Res. 2007, 6 (2), 792–800. (6) Nanavati, D.; Gucek, M.; Milne, J. L. S.; Subramaniam, S.; Markey, S. P. Mol. Cell.

Proteomics MCP 2008, 7 (2), 442–447. (7) Chen, J.; Wang, M.; Turko, I. V. Mol. Neurodegener. 2012, 7, 41. (8) Cheung, C. S. F.; Anderson, K. W.; Wang, M.; Turko, I. V. Anal. Chem. 2015, 87 (2),

1097–1102. (9) Kushnir, M. M.; Rockwood, A. L.; Roberts, W. L.; Abraham, D.; Hoofnagle, A. N.; Meikle, A.

W. Clin. Chem. 2013, 59 (6), 982–990. (10) Barnidge, D. R.; Hall, G. D.; Stocker, J. L.; Muddiman, D. C. J. Proteome Res. 2004, 3 (3),

658–661. (11) Ocaña, M. F.; Neubert, H. Anal. Biochem. 2010, 399 (2), 202–210. (12) Jiang, H.; Zeng, J.; Titsch, C.; Voronin, K.; Akinsanya, B.; Luo, L.; Shen, H.; Desai, D. D.;

Allentoff, A.; Aubry, A.-F.; Desilva, B. S.; Arnold, M. E. Anal. Chem. 2013, 85 (20), 9859–9867.

(13) Brun, V.; Dupuis, A.; Adrait, A.; Marcellin, M.; Thomas, D.; Court, M.; Vandenesch, F.; Garin, J. Mol. Cell. Proteomics MCP 2007, 6 (12), 2139–2149.

(14) Mirzaei, H.; McBee, J. K.; Watts, J.; Aebersold, R. Mol. Cell. Proteomics MCP 2008, 7 (4), 813–823.

(15) Zimmerman, T. A.; Wang, M.; Lowenthal, M. S.; Turko, I. V.; Phinney, K. W. Anal. Chem. 2013, 85 (21), 10362–10368.

(16) Kuhn, E.; Whiteaker, J. R.; Mani, D. R.; Jackson, A. M.; Zhao, L.; Pope, M. E.; Smith, D.; Rivera, K. D.; Anderson, N. L.; Skates, S. J.; Pearson, T. W.; Paulovich, A. G.; Carr, S. A. Mol. Cell. Proteomics MCP 2012, 11 (6), M111.013854.

(17) Sherwood, C. A.; Eastham, A.; Lee, L. W.; Risler, J.; Mirzaei, H.; Falkner, J. A.; Martin, D. B. J. Proteome Res. 2009, 8 (7), 3746–3751.

(18) Lowenthal, M. S.; Liang, Y.; Phinney, K. W.; Stein, S. E. Anal. Chem. 2014, 86 (1), 551–558.

(19) Van den Broek, I.; Smit, N. P. M.; Romijn, F. P. H. T. M.; van der Laarse, A.; Deelder, A. M.; van der Burgt, Y. E. M.; Cobbaert, C. M. J. Proteome Res. 2013, 12 (12), 5760–5774.

(20) Loziuk, P. L.; Wang, J.; Li, Q.; Sederoff, R. R.; Chiang, V. L.; Muddiman, D. C. J. Proteome Res. 2013, 12 (12), 5820–5829.

(21) Brownridge, P.; Beynon, R. J. Methods San Diego Calif 2011, 54 (4), 351–360. (22) Shuford, C. M.; Sederoff, R. R.; Chiang, V. L.; Muddiman, D. C. Mol. Cell. Proteomics MCP

2012, 11 (9), 814–823. (23) Ye, M.; Pan, Y.; Cheng, K.; Zou, H. Nat. Methods 2014, 11 (3), 220–222. (24) Monigatti, F.; Berndt, P. J. Am. Soc. Mass Spectrom. 2005, 16 (1), 13–21. (25) Siepen, J. A.; Keevil, E.-J.; Knight, D.; Hubbard, S. J. J. Proteome Res. 2007, 6 (1), 399–

408.

Figure Legends

Figure 1. Effect of digestion condition on peptide release. Data points represent the peak area

ratio of the unlabeled proteins (14N) to the spike-in labeled proteins (15N) digested with the

condition indicated. Peak area ratios were normalized to condition (1). Conditions include: (1)

0.1 % RapiGest, (2) pH 7.0, (3) 8 mol/L urea, (4) 0.1 % SDS, (5) 80 % acetonitrile, (6) 48 °C, (7)

1:5 trypsin: protein, and (8) 1:25 trypsin/Lys-C: protein. Data and error bars display mean and

standard deviation for three transitions per peptide and duplicate experiments.

Figure 2. Time course analysis showing peptide release and stability for 9 peptides from ILB1

(peptides DDKP, SLVM, and IPVA), IFNG (peptides FFNS, LTNY, and DDQS), and TNFA

(peptides VNLL, IAVS, and ANAL) during proteolytic digestion. Peptides are grouped by class:

(A) rapid release and stable over time, (B) rapid release and subsequent signal decay, and (C)

slow release. Normalized peak area ratios of the unlabeled proteins (14N) to the spike-in labeled

proteins (15N) are plotted versus incubation time. Data and error bars represent the mean and

standard deviation for three transitions per peptide and duplicate experiments.

Figure 3. Calibration curves and regression analysis from a series of digests containing

recombinant protein (Top) or QconCAT protein without flanking sequences (Bottom) spiked into

samples with varying concentrations of 15N isotope labeled calibrant proteins. Measured peak

areas (averaged across three transitions and three replicates for a total of nine measurements

per concentration) were plotted against the expected molar ratios. The slope (m) and intercept

(b) from the regression analysis are shown with standard error for a 95 % confidence interval.

Figure 4. (A) Cartoon version of QconCAT constructs. The top QconCAT is a concatenation of

the natural peptide sequences with N-terminal Met and C-terminal His6-tag (shown in blue). The

bottom QconCAT consists of the same peptides with 6-amino acid long, N-terminal and C-

terminal natural flanking sequences (shown in purple) concatenated into the overall sequence.

(B) Calibration curves and regression analysis from a series of digests containing QconCAT

protein without flanking sequences (Top) or QconCAT protein with natural flanking sequences

of 6 amino acids (Bottom) spiked into samples with varying concentrations of 15N isotope

labeled calibrant proteins. Measured peak areas (averaged across three transitions and three

replicates for a total of nine measurements per concentration) were plotted against the expected

molar ratios. The slope (m) and intercept (b) from the regression analysis are shown with

standard error for a 95 % confidence interval.

Figure 5. Calibration curves and regression analysis from a series of digests containing

synthetic peptides without (+0) or with natural flanking sequences of +4 or +6 amino acids

spiked into samples with varying concentrations of 15N isotope labeled calibrant protein.

Measured peak areas (averaged across three transitions and three replicates for a total of nine

measurements per concentration) were plotted against the expected molar ratios. The slope

and standard error for a 95 % confidence interval for peptide DDKPTLQLESVDPK are 1.09 ±

0.02, 1.21 ± 0.03, 1.02 ± 0.02 and for peptide IPVALGLK are 0.04 ± 0.00, 0.15 ± 0.02, 0.77 ±

0.06 for synthetic peptides +0, +4, and +6, respectively.