Absolute Protein Quantification by Mass ... - ACS Publications

Subscriber access provided by CORNELL UNIVERSITY LIBRARY

Article

Absolute Protein Quantification by Mass Spectrometry: Not as Simple as Advertised Christopher M Shuford, James J Walters, Patricia M Holland, Uma Sreenivasan, Nadav Askari, Kevin B. Ray, and Russell P Grant Anal. Chem., Just Accepted Manuscript • Publication Date (Web): 12 Jun 2017 Downloaded from http://pubs.acs.org on June 13, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Absolute Protein Quantification by Mass Spectrometry: Not as Simple as Advertised Christopher M. Shuford1⧧⧧, James J. Walters2, Patricia M. Holland1, Uma Sreenivasan3, Nadav Askari4, Kevin Ray2, Russell P. Grant1 1. Laboratory Corporation of America Holdings, Center for Esoteric Testing, Burlington, North Carolina 2. MilliporeSigma, Saint Louis, Missouri 3. MilliporeSigma, Round Rock, Texas 4. MilliporeSigma, Rehovot, Israel ⧧Author

for Correspondence:

Christopher M. Shuford, Ph.D. Center for Esoteric Testing Laboratory Corporation of America Holdings Burlington, North Carolina, USA 27215 (336) 436-3366 [email protected]

Key Words Proteomics, Isotope Dilution Mass Spectrometry, Digestion, Stable Isotope Label, Internal Standard, Reference Standard, Calibration, Quantification

Page 1 of 36 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Abstract Stable-isotope labeled (SIL) tryptic peptides, cleavable SIL peptides, and a full-length SIL protein were compared for internal calibration (i.e., as internal calibrators) and external calibration (i.e., as internal standards) when quantifying three forms of unlabeled, human thyroglobulin (Tg) by bottom-up protein analysis. All SIL materials and human proteins were standardized by amino acid analysis to ensure traceability of measurements and allow confident assignment of accuracy. The three forms of human Tg quantified were 1) the primary reference material BCR®457 – a native protein purified from human thyroids – 2) a commercially available form also purified from human thyroids, and 3) a full-length recombinant form expressed and purified from a human embryonic kidney 293 cell-line. Collectively, the results unequivocally demonstrate the lack of commutability of tryptic and cleavable SIL peptides as internal calibrators across various bottom-up assays (i.e., denaturing/digestion conditions).

Further, the results demonstrate the potential

during external calibration for surrogate protein calibrators (i.e., recombinant proteins) to produce inaccurate concentration assignments of native protein analtyes by bottom-up analysis due to variance in digestion efficiency, which is not alleviated by altering denaturation/digestion stringency and indicates why protein calibrators may not be commutable in bottom-up protein assays. These results have implications regarding the veracity of “absolute” protein concentration assignments by bottom-up assays using peptide calibrators as well as protein calibrators given absolute accuracy was not universally observed. Nevertheless, these results support the use of recombinant SIL proteins as internal standards over SIL peptides due to their ability to better mimic the


Page 2 of 41

Page 3 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


digestion of human-derived proteins and mitigate bias due to digestion-based matrix effects that were observed during external calibration.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 41

Introduction Reference standards, or calibration standards (CSs), refer to those materials whose nominal concentration is used in value assignment of an unknown test sample. Internal standards (ISs), or rather internal normalizing standards, refer to those materials which are added to both calibration samples and unknown test samples for the purpose of reducing systematic bias (matrix effects) between the two sample types, as well as reducing imprecision due to random error. Routine isotope dilution mass spectrometry (IDMS) utilizes external calibration (EC) with internal standardization.

For example,

external calibration samples are commonly created with unlabeled CSs spiked at known quantities into a surrogate matrix (native for exogenous analytes, synthetic or depleted matrix for endogenous analytes), which are processed in parallel with the unknown test samples. Both external calibration samples and test samples receive an identical amount of stable isotopically-labeled (SIL) internal standard (IS) that serves to “normalize” for differences (i.e. matrix effects) between the sample and surrogate matrix, as well as between different samples of the same matrix type.1 It is entirely feasible to perform accurate EC without the use of ISs (e.g., ELISA) as concentration assignment is made relative to the unlabeled CSs that experience the same recovery losses as the endogenous analytes. However, in the presence of matrix effects between samples and calibration matrix, an appropriate IS serves to mimic the analyte during sample preparation, reducing systematic biases between matrices (i.e. “normalizing” or eliminating systematic error). Additionally, given the innate variability of ionization/detection during IDMS, an appropriate IS reduces the imprecision of response generation by, again, experiencing the


Page 5 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


variability within a sample that the analyte experiences during ionization and subsequent MS detection. However, the terms “internal standards” and “calibration standards” are synonymous in most bottom-up proteomic assays as “known” amounts of tryptic SIL peptides (tSILs) are commonly spiked into test samples post-digestion to calculate analyte concentrations by internal calibration (IC). IC is a unique situation where the CS is added to the test sample (without use of an EC curve), thereby serving the purpose of both CS and IS – in essence behaving as an internal calibrant. IC has the benefit of eliminating matrix effects that can occur with EC, but both EC and IC ideally use a CS with identical physiochemical properties to the analyte. During IC, the CS must also be mass discriminated relative to the analyte to enable interference-free detection by MS and, consequently, IC is sometimes referred to as “surrogate analyte” calibration because the CS must be, by definition, a surrogate or analogue of the endogenous analyte due to the labeling.2 When the CS and endogenous analyte have different physiochemical properties there arises the potential for quantitative inaccuracy. The conventional hierarchy of standards (CS and IS) during bottom-up protein analysis has been established to be full-length SIL proteins (SIL proteins, best), enzymatically cleavable SIL surrogates (cSILs), and tryptic SIL peptides (tSILs, worst).3 Despite early reports that IC with tSILs routinely provide “absolute” (i.e., accurate) protein concentration assignments,4,5 it is now generally accepted that accuracy of IC with tSILs requires complete digestion of the target protein into the signature peptide (or perhaps more aptly, “surrogate peptide” in applications of tryptic peptide calibrators).6 Enzymatically cleavable



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

surrogates, such as concatemers,7 cleavable peptides,8 or partial proteins,9 have consequently been utilized during IC to facilitate absolute accuracy by normalizing variance in enzymatic digestion, particularly by emulating the native protein sequence in the flanking regions on either termini of the signature peptide(s).10,11 However, the lack of secondary or higher-order structure has proven to be a limiting factor for cSILs given the primary structure (i.e., sequence) is only one determinant governing digestion efficiency.12,13 In comparative studies with tSILs or cSILs, SIL proteins have routinely demonstrated superior performance with respect to both accuracy and precision during IC.13–15 Nonetheless, both tSILs and cSILs continue to be used for IC in lieu of SIL proteins given their greater accessibility and affordability. However, the burden of proof needed to demonstrate complete digestion or absolute accuracy with tSILs or cSILs exclusively for IC has rarely been demonstrated.4 Often, relative accuracy has been inferred via agreement, or simply correlation,16 between multiple signature peptides rather than comparison of the absolute concentration assignment to an absolute reference value. Indeed, the few reports striving for the latter level of quality have often made observations contradicting the claim of absolute accuracy during IC with tSILs or cSILs.13–15 Additionally, it is yet unclear whether recombinant full-length proteins digest with the same efficacy as native, human protein and, thus, whether or not recombinant proteins can routinely be used as surrogate calibrators to accurately quantify endogenous protein biomarkers. Recombinant protein standards are more readily obtained as a highly purified material ideal for preparation of calibrators, yet different post-translational modifications (PTMs) and/or conformations may alter the way in which they are denatured and digested relative to their endogenous counterparts, potentially undermining their ability to provide Page 6 of 36 ACS Paragon Plus Environment

Page 6 of 41

Page 7 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


absolute accuracy during EC (unlabeled proteins) or IC (SIL proteins). Studies touting the veracity of recombinant SIL proteins for IC have done so by measuring the equivalent unlabeled recombinant protein (prepared using the same recombinant system as the SIL protein, but without isotopically labeled amino acids) rather than by measuring a natively expressed protein.13,17 Indeed, accurate quantification of biotherapeutic drug levels are also held as evidence for the veracity of absolute quantification using SIL proteins, yet many of these examples also utilize the recombinant drug as the CS.15,18 This work aims to definitively address the fundamental questions related to calibration and internal standardization for bottom-up protein analysis using higher order amino acid analysis (AAA) as a foundation for accuracy assessment through quantification of three purified sources of human thyroglobulin (Tg). Specifically, these studies aim to determine if recombinant proteins as surrogate CS or IS provide absolute (accurate) quantification of native proteins and to clarify in what circumstances tSILs and cSILs may fail to achieve absolute quantification of native proteins when used as surrogate CSs or ISs. Tg was selected as the model protein in these studies primarily due to the commercial availability of multiple sources of purified, human-derived material that could be contrasted against invitro expressed protein. Given its large size (>330 kDa monomer) and large number of PTMs, Tg also potentially provides an increased challenge with respect to the bottom-up protein quantification as it likely has multiple proteoforms which could have unique digestion properties not only between recombinant and native forms, but also between different human sources.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Materials & Methods Chemicals & Reagents All materials unless otherwise noted were from MilliporeSigma (St. Louis, MO). Chicken serum was from Invitrogen (Carlsbad, CA), OptimaTM LC-MS/MS grade water and acetonitrile were from Fisher Scientific (Carlsbad, CA), and Zwittergent 3-16 was from Calbiochem® (Billerica, MA). Recombinant human serum albumin (rHSA) was expressed and purified from rice and provided by MilliporeSigma. Pooled human serum used in these studies was generated in-house from de-identified, remnant serum specimens that had each been tested and confirmed to be devoid of any significant amount of endogenous human Tg (99% (Supporting Information, Figure S-1 and Table S-1). Synthetic tSIL and cSIL peptides (Supporting Information, Table S-2) were purchased from New England PeptideTM Page 8 of 36 ACS Paragon Plus Environment

Page 8 of 41

Page 9 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(Gardner, MA) in powder form with >99% labeling efficiency verified by the manufacturer. Each cSIL was designed with 6 residues flanking both the N- and C-terminus of the signature peptide sequence, which were derived from the primary structure of human Tg in order to aid in emulating the digestion process of a full-length protein. The purity of each tSIL and cSIL was determine by HPLC and ranged from 59.7 to 99.5% at the time of use (Supporting Information, Table S-2). Amino acid analysis (AAA) traceable to NIST Standard Reference Material 2389a was performed on freshly reconstituted aliquots of each protein or peptide stock solution (Supporting Information, Table S-3). Measurements were performed in triplicate and the mean results utilized for concentration assignment of the stock solution.

The

imprecision was less than 5% (coefficient of variation) in all cases. NIST Bovine Serum Albumin Standard Reference Material 927e was processed in parallel with all stock solutions to qualify the accuracy of the AAA with respect to hydrolysis, which demonstrated recoveries between 99.7 and 104%. Following AAA of each stock solution, aliquots of each unlabeled Tg and SIL-rTg were gravimetrically diluted to nominal concentrations of 330 nmol/L (fmol/µL), not accounting for SDS-page purities. Likewise, all eight tSIL stock solutions and all eight cSIL stock solutions were gravimetrically combined and diluted in 0.001% Zwittergent 3-16 to create a tSIL sub-stock solution and cSIL sub-stock solution with each SIL peptide at a nominal concentration of 300 nmol/L (fmol/µL), not accounting for HPLC purities.

Working

samples/solutions were freshly diluted from sub-stocks by weight on the same day of their use to the indicated nominal concentration.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 41

Neat Tg Analysis Three test samples of human Tg (sTg, cTg, and rTg) prepared in 0.1% human serum albumin (HSA, MilliporeSigma Item #A1653) were quantified by bottom-up protein analysis. All samples were diluted by weight from their respective sub-stocks to maintain traceability of each component to their AAA value assignments. A generic description of the overall sample digestion procedure used for quantification of the human Tg samples is provided, depicting the point in time at which each SIL material was introduced into the workflow (Figure1C). SIL-rTg was gravimetrically introduced into each sample prior to denaturation, then the cSIL mixture was added just prior to digestion, and the tSIL mixture was added immediately following digestion. For a given denaturing condition, all three human Tg samples were digested in parallel as six replicates (Figure 1A). The first three replicates received SIL-rTg, but not the cSIL mixture. Conversely, the latter three replicates received the cSIL mixture, but not SIL-rTg. In this manner, the SIL-rTg and cSIL mixture were never present in the same sample to avoid isotopic interference of the VIL and LED signature peptides observed in blank control experiments (Supporting Information, Figure S-3). However, the tSIL mixture was added to all six replicates. Moreover, this design was repeated independently using 3 different denaturing conditions. Quantification of the three human Tg samples was carried out by IC using all 3 SIL materials (SIL-rTg, cSILs, or tSILs) as SIL CSs for a total of 27 determinations by IC (3 human Tg × 3 SIL CSs × 3 denaturants). Additionally, EC was carried out using the rTg sample as the external CS with each SIL material as SIL ISs for a total of 18 determinations by EC (2 human Tg × 3 SIL ISs × 3 denaturants).


Page 11 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Complex Matrix Analysis Two test samples of human-derived Tg (sTg and cTg) prepared in pooled human serum were quantified by bottom-up protein analysis using EC (Figure 1B). Three external calibrators were prepared using rTg as the CS spiked into three different surrogate matrices for human serum: chicken serum, 6% HSA in PBS, and 6% recombinant HSA (rHSA) in PBS. Both test samples and external calibration samples were prepared from their respective sub-stocks, diluting by weight to maintain traceability of each component to their AAA value assignments. The test samples and external calibrators were processed in parallel in triplicate, with all replicates receiving all 3 SIL materials to use as SIL ISs. Inclusion of all SIL materials in the same digestion was feasible given there was no isotopic interference observed in the blank controls for the two signature peptides considered in these studies (FSP and VIF). This experimental design was repeated independently using 3 different denaturing conditions for a total of 54 determinations by EC (2 human Tg × 3 surrogate matrices × 3 SIL ISs × 3 denaturants). Tryptic Digestion Digestion of the three Tg test samples in 0.1% HSA proceeded as follows. Fifty microliters of sample containing a nominal 3 pmol of unlabeled, human Tg and 3 pmol of SIL-rTg was mixed with an equal volume of denaturant solution (8 M Urea, 2% DOC, or 20% TFE) containing 10 mM DTT and incubated for 30 minutes at 56 °C. Denatured samples were further diluted with 100 µL of 125 mM Tris-HCl, 0.001% Zwittergent 3-16 (pH 8.0) containing a nominal 1.5 pmol of each cSIL. Subsequently, 5 µg TPCK-treated bovine trypsin in 25 µL of 50 mM acetic acid was added to produce an approximate enzyme:protein ratio of 1:10 (w/w) and digestions proceeded for 30 minutes at 37 °C after Page 11 of 36 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 41

which the digestion was terminated by the addition of 200 µL of 4% formic acid, 0.004% Zwittergent 3-16 containing nominally 3 pmol of each tSIL.

All incubations were

performed at 1500 rpm on a ThermoMixer® C equipped with a ThermoTop® (Eppendorf, Hauppauge, NY). Digestion of Tg-spiked human serum and its surrogate matrices proceeded as above, with two exceptions. Firstly, the nominal amounts of unlabeled Tg, SIL-rTg, cSILs, and tSILs were gravimetrically diluted to 100 fmol each and, secondly, 300 µg of trypsin was used to maintain an approximate enzyme:protein ratio of 1:10 (w/w). Additionally, following digestion in the serum/surrogate matrices, the FSP and VIF signature peptides were isolated using immunoaffinity enrichment essentially as previously described.20 LC-MS/MS Selected reaction monitoring (SRM) was utilized to monitor 2 transitions per signature peptide over the course of a 10.25 minute reverse phase HPLC separation. Two distinct MS acquisition methods were used: one for monitoring 8 signature peptides derived from the unlabeled Tg and the 3 SIL materials (Supporting Information, Table S4) and one for monitoring 39 signature peptides derived from only the unlabeled Tg and SIL-rTg (Supporting Information, Table S-5). The tSILs, cSILs, and SIL-rTg each had unique labeling of the respective signature peptides to provide unique SRM transitions (Supporting Information, Table S-1 and Table S-2). Matching transitions for SIL and unlabeled signature peptides were energy matched to best enable equivalent MS/MS response factors for IC.


Page 13 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Calculations All peak integration and review was performed in Skyline v3.5.0.9319,21 with subsequent export to Microsoft© Excel® for quantitative calculations.

Single-point

calibration was used in all cases for determination of the absolute amount of unlabeled Tg digested in each test sample according to the following equation:

‫ܥ‬஺ =

ோಲ ோ಴ೄ

× ‫ܥ‬஼ௌ

Equation (1)

where RA and RCS are the measured responses from the analyte (i.e., unlabeled, human Tg) and calibration standard (CS), respectively, CCS is the absolute value assigned to the CS, and CA is the absolute amount of analyte calculated in the test sample. The absolute value of the CS was corrected for analyte purity (as determined by SDS-PAGE or HPLC) as well as gravimetric dilutions. During IC, the CS was one of three SIL materials (e.g., SIL-rTg, cSIL, or tSIL) added to the test sample and the measured responses in Equation (1) were the absolute peak areas of the analyte and SIL CS in the test sample. This method of IC did not take into account or correct for potential differences in the LC-MS/MS response factors of the unlabeled and SIL signature peptides via so-called response factor normalization. This potential source of error, which could arise from differences in fragmentation efficiency or increased abundance of the monoisotopic precursor ion in SIL signature peptides,22 was expected to be negligible relative to the total error observed in these studies and is distinct from potential interference due to overlapping isotopic distributions of unlabeled and labeled peptides or conventional isotope effects altering the reaction rate/efficiency of enzymatic hydrolysis. Page 13 of 36 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 41

During EC, the CS was the rTg present in a separate calibration sample (i.e., external calibrator) processed in parallel with the test samples and the measured responses in Equation (1) were the analyte:SIL-IS peak area ratio (PARs) in the test sample (RA) and the CS:SIL-IS PAR in the external calibration sample (RCS). The definition used here for absolute accuracy was a measurement recovering between 80 and 120% of the target value, which allows for 95% confidence and 70% power with an expected coefficient of variance of 10% using triplicate measures. For digestions employing a nominal 3 pmol of each unlabeled Tg, absolute accuracy was thus defined as a measurement resulting in between 2.4 and 3.6 pmol. For digestions employing a nominal 100 fmol of each unlabeled Tg, absolute accuracy was thus defined as a measurement resulting in between 80 and 120 fmol.

Results & Discussion Neat Tg Quantification by Internal Calibration (IC) Eight signature peptides were considered when quantifying the 3 human sources of Tg (Supporting Information, Table S-2).

For each type of SIL CS, a total of 72

concentration assignments of Tg were possible by IC (3 Tg sources × 3 denaturing conditions × 8 signature peptides = 72 quantification measurements). Using tSIL CSs added post-digestion for IC, absolute accuracy was only achieved in 16 of 72 results (Figure 2) and only two signature peptides (VIF and SQA) achieved accurate protein value assignments in at least 2 of the 3 digestion conditions for each Tg. Indeed, there were obvious and gross differences between digestion conditions for nearly all signature peptides. For example, in the 72 IC measurements using tSIL CSs, the protein value


Page 15 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


assignments across the three denaturing conditions spanned nearly two orders of magnitude (0.04 to 3.44 pmol or 1.3 to 114.6% accuracy). Despite the inclusion of 6 residues flanking both termini of the signature peptides in each cSIL, only 14 of 72 protein value assignments using cSIL CSs for IC demonstrated absolute accuracy. In contrast to tSIL CSs, which produced predominantly negatively biased results (120% accuracy), 11 of 72 measurements with cSIL CSs resulted in a positive bias. Notably, the magnitude of the biases observed with the cSIL reference standards FYQ, VTW, and LHL (corresponding to the FSP, LED, and TFP signature peptides) were not explained by the magnitude of impurity observed in these standard preparations (Supporting Information, Figure S-2 and Table S-2). For example, large differences in results observed between denaturing conditions for same cSIL CS is not explained by an erroneous purity correction given the purity correction would systematically affect all value assignments for a given cSIL CS. In total, the 72 human Tg quantities obtained by IC with cSIL CSs spanned almost one order of magnitude (0.57 to 4.67 pmol or 19.1 to 155.6% accuracy). Although these results indicate some correction for digestion variance (as compared to tSILs), the cSILs proved largely ineffective as CSs, which is likely attributable to their inability to correct for denaturation and, thus, how digestion efficiency is impacted from higher order structure. With SIL-rTg as the CS, only 29 of 72 IC measurements produced absolute accuracy, while 42 of 72 results were positively biased (>120% accuracy). This apparent systematic positive bias may be explained by an error in the absolute value assignment of the SIL-rTg. Incorrect AAA or instability of the sub-stock solution was ruled out by re-analysis of the 4



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 41

protein sub-stocks by AAA following these studies (Supporting Information, Table S-3) and, further, the labeling efficiency of the SIL-rTg was demonstrated to be >99.8% (Supporting Information, Table S-1). Other pre-analytical artifacts occurring during manufacturing of the working solutions (e.g., adsorptive loss) were not characterized, but seem unlikely to uniquely affect SIL-rTg (and not rTg) and further unlikely to be reproducible across 3 separate experiments (one for each digestion condition). Alternatively, isotope effects reducing the tryptic digestion efficiency have been previously suggested for recombinant SIL proteins23 and cannot be ruled out in this study. Despite the apparent systematic bias, IC results observed with SIL-rTg as the CS were significantly less variable.

When considering the 24 Tg mean concentration

assignments obtained using SIL-rTg (3 denaturant conditions x 8 signature peptides), the imprecision in the mean quantity determined for sTg, cTg, and rTg was 13.5, 14.6, 6.5%, respectively. Notably, 6.5% is likely the optimal precision expected in the absence of systematic error between signature peptides given the level of replication and technical variance in this study.

By comparison, the imprecision for the same mean protein

quantities by IC was 51.5, 53.0, and 45.0%, respectively using cSIL CSs, and 88.8, 90.6, and, 87.7%, respectively using tSIL CSs. These results clearly demonstrate that control for digestion is key to reducing variance in protein concentration assignments by bottom-up analyses and, further, that optimal control for digestion and minimum variance is best obtained by controlling denaturation through use of SIL proteins. The poor performance of tSIL CSs and cSIL CSs for IC likely stems from the fact that neither accurately mirror the digestion (in)efficiency experienced by the Tg proteins being


Page 17 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


quantified. The premise of calibration by Equation (1) assumes equivalent digestion efficiency between the analyte protein and the CS, which these results would indicate is most often a fallacy when using tryptic or cleavable peptides as CSs. Given absolute accuracy requires equivalent digestion of the analyte protein and CS, accurate protein value assignments are only obtained with tryptic peptide CSs when “complete digestion” of the protein is observed – that is, when stoichiometric conversion of the protein into the signature peptide is observed. Given the variety of signature peptides and denaturing conditions considered in this study, the results would indicate that “complete digestion” is the exception rather than the rule and, thus, that absolute quantification with tryptic peptide CSs is rarely achieved. Any claims of absolute protein quantification (i.e., absolute accuracy) when using tryptic peptide reference standards should consequently be supported by demonstration of “complete digestion” (i.e., as performed in these studies). Nonetheless, this data indicates that tSIL CSs are suitable for accurately quantifying the amount of peptide product derived from the digestion process by IC. By comparing the amount of peptide product quantified by this strategy to the absolute amount of substrate (Tg, SIL-rTg, cSIL), it is straight forward to derive the true conversion efficiency of substrate to product peptide, i.e., the “absolute digestion efficiency” (Supporting Information, Figure S-4). Comparing the absolute digestion efficiency of the human Tg and SIL CSs can facilitate identifying the source of inaccuracy during quantification. When the SIL CS exhibited greater digestion efficiency than the analyte Tg, a negatively biased protein value assignment was observed. Conversely, a positively biased protein value assignment was observed when the SIL CS exhibited lower digestion efficiency than the analyte Tg, while



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 41

accuracy was only observed when equivalent digestion efficiency (not necessarily complete digestion) of the Tg and SIL CS was obtained. This relationship explains why a negative bias was the most common outcome for IC using tSIL CSs, which by definition represent 100% digestion efficiency as they are not subjected to the digestion procedure. Additionally, this was the more frequent outcome for cSIL CSs, which lack the higher order structure of the analyte protein that may otherwise reduce the digestion efficiency. On the other hand, there were also multiple examples of positive bias by IC using cSIL CSs. In lieu of cSIL peptide impurity (which we have accounted for), reduced digestion efficiency of smaller substrates seems counter intuitive given their lack of higher order structure; however, such an occurrence may arise when the lack of higher order structure favors the formation of dead-end, missed-cleavage peptides.12 It should also be noted that digestion efficiency may be the net product of both peptide formation and degradation. Consequently, if cSIL CSs experience the same amount of formation, but a greater amount of degradation as the analyte protein, the net “digestion efficiency” of the CS could potentially be lower than that of the analyte protein, thereby resulting in a positive bias in protein value assignment by IC. Indeed, this has been proposed for tSIL CSs which are added pre-digestion and, thus, experience a greater amount of degradation due to the fact that they are capable of degrading throughout the digestion step, while the peptide derived from the analyte protein is differentially degraded after it has the chance to form over time.24,25 Peptide degradation was ruled out as a likely culprit of bias in the present studies by performing a time-course analysis of the digestion of all 3 trypsin substrates (unlabeled Tg, SIL-rTg, and cSIL peptides, Figure S-6), but could


Page 19 of 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


have also been ruled out through a comparative study adding the tSIL CSs pre- and postdigestion.26 Neat Tg Quantification by External Calibration (EC) To exemplify the difference in the use of SIL materials as CSs for IC versus ISs for EC, quantification of the 2 human-derived Tg proteins (sTg and cTg) was performed via EC using the unlabeled, rTg as the CS in combination with each of the SIL materials (tSIL, cSIL, and SIL-rTg) as ISs (i.e., SIL ISs). A total of 48 value assignments of human-derived Tg were obtained with each SIL IS (2 human-derived Tg × 3 digestion conditions × 8 signature peptides = 48 measurements). Perhaps surprisingly, tSIL ISs produced accurate protein quantification in 43 of 48 measurements, while cSIL and SIL-rTg produced accurate results in only 37 and 36 cases, respectively, when used as ISs (Figure 3).

This marked

improvement in the quantitative accuracy relative to the use of the SIL materials for IC is the result of the accuracy being primarily dependent upon the consistent digestion efficiency of the human-derived Tg and the rTg CS, rather than the relative digestion efficiency of the human derived Tg and the SIL ISs. The performance of the tSIL ISs in this assessment was not likely due to a fundamental property of tSILs, but more likely the result of having reduced imprecision through having double the number of replicates in the experimental design (Figure 1A) – indeed, if the results are bifurcated into two independent experiments of triplicate measures with respect to the tSIL ISs, the tSIL ISs produce accurate results in only 35 of 48 cases (i.e., the first three replicates) and 39 of 48 cases (i.e., from the second three replicates). Consequently, these results suggest all 3 forms of ISs may be equally effective in achieving absolute quantification, at least in the



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 41

absence of matrix effects (i.e., both test samples and the external reference samples were prepared in 0.1% HSA matrix). Perhaps the most interesting observation from this analysis of EC was that one of eight signature peptides, TFP, produced negatively biased measurement of sTg and cTg (

Absolute Protein Quantification by Mass ... - ACS Publications

Recommend Documents