Article pubs.acs.org/ac
Covalent Labeling Denaturation Mass Spectrometry for Sensitive Localized Higher Order Structure Comparisons James A. Madsen, Yan Yin, Jing Qiao, Vanessa Gill, Kutralanathan Renganathan, Wing-Yee Fu, Stephen Smith, and James Anderson* Momenta Pharmaceuticals, 675 West Kendall Street, Cambridge, Massachusetts 02142, United States S Supporting Information *
ABSTRACT: Protein higher order structure (HOS) describes the three-dimensional folding arrangement of a given protein and plays critical roles in structure/function relationships. As such, it is a key product quality attribute that is monitored during biopharmaceutical development. Covalent labeling of surface residues, combined with mass spectrometry analysis, has increasingly played an important role in characterizing localized protein HOS. Since the label can potentially induce conformation changes, protocols generally use a small amount of label to ensure that the integrity of the protein HOS is not disturbed. The present study, however, describes a method that purposely uses high amounts of isobaric label (levels that induce denaturation) to enhance the sensitivity and resolution for detecting localized structural differences between two or more biological products. The method proved to be highly discriminative, detecting differences in HOS affecting as little as 2.5−5% of the molecular population, levels at which circular dichroism and nuclear magnetic resonance spectroscopy fingerprinting, both gold standard HOS techniques, were unable to adequately differentiate. The methodology was shown to have comparable sensitivity to differential scanning calorimetry for detecting HOS differences. In addition, the workflow presented herein can also quantify other product attributes such as post-translational modifications and site-specific glycosylation, using a single liquid chromatography−tandem mass spectrometry (LC−MS/MS) run with automated data analysis. We applied this technique to characterize a large (>90 kDa), multiply glycosylated therapeutic protein under different heat stress conditions and aggregation states.
M
raphy and conventional nuclear magnetic resonance (NMR) can be used, as long as the protein of interest can be crystallized or has a low molecular weight, respectively.5 Mass spectrometry (MS)-based analysis has recently gained significant popularity for protein higher order structure analysis and can provide valuable localized HOS information. Hydrogen−deuterium exchange coupled with mass spectrometry (HDX-MS), for example, measures the uptake of deuterium by amide hydrogens along the backbone of a protein in the presence of deuterium and can provide detailed information concerning protein structure, dynamics, and interactions.6−11 Hydroxyl radical footprinting12−17 and fast photochemical oxidation of proteins (FPOP),18−25 both of which cause oxidative modification to amino acid side chains, are two other MS-based methods that are complementary to HDX-MS and have recently shown promise for HOS characterization of therapeutic proteins.12,22,24 Covalent labeling using reagents that react with surface amino acid side chains (often amine or carboxylic acid containing residues) has also been used to
onoclonal antibodies (mAbs) and antibody-related proteins are the fastest growing group of therapeutic agents.1 These molecules are highly complex and require an array of analytical methodologies to thoroughly characterize all their physicochemical attributes during development. Protein higher order structure (HOS) is one of the key product quality attributes monitored during various development cycles as it describes the three-dimensional folding arrangement of a given protein and plays critical roles in structure/function relationships. Additionally, a protein therapeutic with a misfolded section of its structure can potentially cause aggregation, immunogenicity, and/or reduced drug efficacy.2−4 It is imperative, therefore, to have highly sensitive methods that are capable of detecting therapeutic proteins that have compromised higher order structure. There are a variety of analytical techniques for ensuring HOS integrity of protein therapeutics such as circular dichroism (CD), Fourier transform infrared (FT-IR) spectroscopy, differential scanning calorimetry (DSC), and intrinsic fluorescence.5 These methods yield valuable information on secondary and tertiary structures; however, they can only be utilized to probe global conformations. To characterize localized structural changes, methods such as X-ray crystallog© 2016 American Chemical Society
Received: December 14, 2015 Accepted: January 11, 2016 Published: January 11, 2016 2478
DOI: 10.1021/acs.analchem.5b04736 Anal. Chem. 2016, 88, 2478−2488
Article
Analytical Chemistry
by the addition of 50 μL of controls 1−3 and samples 4−6 (1 μg/μL) to the appropriate TMT aliquot. Reaction mixtures were immediately mixed by pipetting the solution up and down 10 times, and then incubated for 2 min at room temperature. After the incubation period, each reaction was simultaneously quenched and denatured by adding 50 μL of 6 M guanidine hydrochloride in 20 mM sodium phosphate/100 mM sodium chloride (pH 7.0) containing 5% hydroxylamine and mixed by pipetting the solution up and down 10 times. All six quenched reaction mixtures were equally mixed by adding 16 μL of each sample for a total volume of 96 μL. Samples were prepared for LC−MS/MS analysis per the section below. Protein Preparation and LC−MS/MS Analysis. Labeled samples were reduced and alkylated by adding 2 μL of 0.5 M tris(2-carboxyethyl)phosphine hydrochloride (TCEP) solution and incubating for 30 min at 37 °C, and then adding 2 μL of 1 M iodoacetamide and incubating for 1 h in the dark. Using Zeba spin columns, samples were then buffer-exchanged into 50 mM ammonium bicarbonate containing 12 mM methionine (final volume of 150 μL) and digested with either 2 μg of chymotrypsin or Asp-N (1:25 enzyme/substrate ratio) for 1 h in a Barocycler (Pressure Biosciences, South Easton, MA) operated at 37 °C and 20 000 psi. Enzymatic reactions were quenched with 2% formic acid. The resulting peptides were analyzed using LC−MS/MS by injecting 2 μg of sample onto a 2.1 mm × 50 mm (1.7 μm particle size) AQUITY BEH C18 column (Waters, Milford, MA) heated at 50 °C using a Dionex Ultimate 3000 RSLCnano (Santa Clara, CA) system. Peptide separation was performed with eluent A consisting of 0.1% formic acid in water and eluent B consisting of 0.1% formic acid in acetonitrile and a 95 min linear gradient from 4% to 35% eluent B at a flow rate of 50 μL/min. Data-dependent MS/MS was performed on a Thermo Scientific Q Exactive mass spectrometer (Bremen, Germany) as follows: the first event was the survey positive mass scan (m/ z range of 400−2000) followed by HCD events (30% NCE) on the 12 most abundant ions from the first event. Ions were generated using a spray voltage of 3.32 kV, a capillary temperature of 275 °C, and an S-Lens rf level of 85. Resolution and AGC were set at 35 000 and 1 × 106 for survey scans and 17 500 and 2 × 105 for MS/MS events. A maximum injection time of 250 ms was used for all scans. A dynamic exclusion duration of 20 s was used with a single repeat count. Both full MS and MS/MS spectra were produced from one microscan. Peptide Identification and Higher Order Structure Quantification. The entire peptide identification and quantitation data analysis workflow described below was fully automated using Proteome Discoverer, version 1.4 (Thermo Scientific, Bremen, Germany). Experimental MS/MS spectra were searched against a protein database composed of the sequence of M1 and common protein contaminants using both SEQUEST42 and Byonic43 within Proteome Discoverer. A mass tolerance of 10 ppm and 0.02 Da were used for precursor and fragment masses, respectively, and up to two missed cleavages were allowed. Carbamidomethyl of cysteine was used as a fixed modification. TMTduplex or TMTsixplex was set as a variable modification on lysine, serine, threonine, and the N-terminus; deamidation of asparagine and oxidation of methionine were also set as variable modifications. A database consisting of the top 13 glycans was used by Byonic for identifying glycopeptides.
quantify solvent accessibility and specific surface deviations in proteins.26−40 Commercial isobaric reagents such as tandem mass tags (TMT)41 have recently been utilized as a HOS characterization tool for measuring labeling kinetics,33 probing lysine linkages for antibody−drug conjugates,34 and epitope mapping.35 Although many useful developmental innovations have been made in MS-based HOS techniques, many drawbacks still remainpoor sensitivity, resolution, reproducibility, and robustness are the most common limitations. Therefore, current methodologies are often not able to accurately and routinely detect low-abundant, localized structural changes for complex biotherapeutics.10 In this report, we describe an MS-based methodology that purposely uses high amounts of isobaric TMT labels (levels that induce significant protein denaturation) to maximize the sensitivity and resolution for detecting localized structural differences in a large (>90 kDa), multiply glycosylated therapeutic protein. In addition to HOS characterization, the methodology presented herein was also capable of quantifying product attributes such as post-translational modifications (PTMs) and site-specific glycosylation, using a single liquid chromatography−tandem mass spectrometry (LC−MS/MS) run with automated data analysis.
■
EXPERIMENTAL SECTION Materials and Reagents. All experiments were performed on M1, an antibody-like, multiply glycosylated therapeutic protein produced recombinantly in Chinese hamster ovary (CHO) cells with a molecular weight >90 kDa. Chymotrypsin and Asp-N were purchased from Promega (Madison, WI), and tandem mass tags were from Life Technologies (Carlsbad, CA). Solvents for liquid chromatography and all lab supplies were obtained from Fisher Scientific (Pittsburgh, PA). All other chemical reagents were purchased from Sigma-Aldrich (St. Louis, MO). Protein Stressing. M1 was diluted to 1 mg/mL in 1× PBS buffer (pH 7.4) prior to stressing. Samples were then properly sealed and heated for 18 h on a heat block at both 55 and 75 °C. After the incubations, samples were stored at room temperature. No obvious visible precipitation was observed after either heat-stressing condition. Protein Labeling. For the TMTduplex procedure, 126 and 127 TMT vials were thawed and allowed to reach room temperature for 30 min. Each 0.8 mg TMT vial was then reconstituted with 40 μL of acetonitrile and vortexed for at least 1 min. For each reaction, 5 μL of TMT 126 and 5 μL of TMT 127 were added to 50 μL of control (1 μg/μL) and 50 μL of sample (1 μg/μL), respectively. Reaction mixtures were immediately vortexed, centrifuged, and then incubated for 2 min. After the incubation period, each reaction mixture was quenched by adding 1 μL of 5% hydroxylamine and was immediately vortexed and centrifuged. The quenched samples were then denatured by adding 50 μL of 6 M guanidine hydrochloride in 20 mM sodium phosphate/100 mM sodium chloride (pH 7.0) and were mixed 1:1 by adding 50 μL of each reaction mixture to make a total volume of 100 μL. Samples were prepared for LC−MS/MS analysis per the section below. For the TMTsixplex procedure, 126, 127, 128, 129, 130, and 131 TMT vials were thawed and allowed to reach room temperature for 30 min. Each 0.8 mg TMT vial was reconstituted with 40 μL of acetonitrile and vortexed for at least 1 min. Using a multichannel pipet, 5 μL of each TMT solution was added to a separate microcentrifuge tube, followed 2479
DOI: 10.1021/acs.analchem.5b04736 Anal. Chem. 2016, 88, 2478−2488
Article
Analytical Chemistry
MS/MS run with automated data analysis. With these goals in mind, we developed an analytical workflow based on covalent labeling mass spectrometry that uses isobaric TMT labels to purposely cause protein denaturation. The schematic diagram of the overall experimental workflow can be seen in Figure 1. In
Fold changes were calculated by Proteome Discoverer using an integration tolerance of 20 ppm, isotopic quantitative value correction, and a maximum allowed fold change of 100. MS/ MS spectra that yielded fold changes greater than 2 were manually verified by assessing TMT channel interference and MS/MS patterns. To mimic the localized HOS difference plots that are standard for HDX-MS analysis of protein therapeutics,8,9 the reciprocal of decimal fold changes was used. Additionally, fold changes were made either positive or negative to show direction (i.e., an increase in positive fold changes indicates unfolding of the protein and an increase in negative fold changes indicates aggregation of the protein). Three replicates were prepared and analyzed at each condition (stressed or unstressed) for repeatability experiments, and error bars represent standard deviations of the fold changes. Size Exclusion Chromatography. An Agilent 1200 system (Santa Clara, CA) was utilized for all size exclusion chromatography (SEC) experiments. Heat-stressed and control M1 samples (1 mg/mL) were injected at 40 μL on a 7.8 mm × 300 mm (5 μm particle size) TOSOH TSKgel SEC column (Tokyo, Japan). A mobile phase consisting of 50 mM sodium phosphate, 300 mM NaCl (pH 7.0) was flowed at a rate of 1 mL/min for 20 min for separation. The detection wavelength was set to 280 nm. Data collection and analysis were performed with Empower software (Waters, Milford, MA). Circular Dichroism. All CD experiments were performed on a Jasco J-815 CD spectropolarimeter (Easton, MD). Samples were diluted to 0.2 mg/mL in 1× PBS buffer (pH 7.4) prior to analysis, and CD spectra were collected between 195 and 250 nm. Lower wavelength regions were not shown when the high-tension (HT) voltage exceeded ∼600 from detector saturation. In general, the TMT-labeled samples yielded higher HT voltages compared to unlabeled samples. Differential Scanning Calorimetry. All DSC experiments were performed on a TA Nano differential scanning calorimeter (New Castle, DE). Samples were diluted to 1.0 mg/mL in 1× PBS buffer (pH 7.4) prior to analysis, and DSC spectra were collected between 20 and 110 °C. Nuclear Magnetic Resonance Spectroscopy. One dimensional (1D) proton (1H) NMR spectroscopy experiments were performed on a Bruker 600 MHz NMR (Billerica, MA) equipped with a 5 mm cryoprobe. Measurements were made by taking 500 μL of a 1 mg/mL sample solution (in 1× PBS buffer, pH 7.4), adding 50 μL of 1 mM DSS-d6 D2O solution (Cambridge Isotope Laboratories, Inc., Tewksbury, MA) as the chemical shift reference standard, and then transferring the sample into a 5 mm NMR tube. The 1D 1H spectra were collected at 25 °C with 1024 scans, and a “zgespg” pulse program was utilized for peak suppression of the water signal. Raw data was Fourier transformed, phase corrected, and then calibrated to 0.00 ppm from the reference DSS-d6 singlet 1 H peak.
Figure 1. General experimental workflow of the methodology described herein. Isobaric mass tags are used to covalently modify protein samples with high label amounts to induce protein denaturation. Reactions are subsequently quenched with excess amine and mixed. The protein mixture is then reduced, alkylated, enzymatically digested, and analyzed by LC−MS/MS. After dissociation by MS/MS, peptides are identified by database searching, and reporter ion ratios are used to calculate fold changes (i.e., localized structural deviations) for each labeled peptide. The workflow only illustrates the TMTduplex procedure; however, TMTsixplex was also used to increase multiplexing capabilities.
RESULTS AND DISCUSSION The motivation for the work presented herein was to create high-resolution analytical methodology capable of detecting and quantifying low-level HOS differences in localized regions of large, multiglycosylated therapeutic proteins. In addition to maximizing sensitivity, our goal was to develop a straightforward workflow that was reproducible, robust, and not dependent on overly specialized equipment. Lastly, we wanted the methodology to simultaneously quantify (dis)similarities in HOS, PTMs, and site-specific glycosylation from a single LC−
short, therapeutic protein samples are diluted to 1 mg/mL in 1× PBS, pH 7.4, and either TMTduplex or TMTsixplex is added at amounts capable of denaturing the protein. The samples are then quenched after a specified time, equally mixed, reduced, alkylated, enzymatically digested, and analyzed by data-dependent LC−MS/MS. Unmodified, TMT-labeled, glycosylated, and PTM peptides are identified by database searching, and reporter ion ratios are used to calculate fold changes (i.e., localized structural deviations) for each labeled peptide. The performance of the methodology and its
■
2480
DOI: 10.1021/acs.analchem.5b04736 Anal. Chem. 2016, 88, 2478−2488
Article
Analytical Chemistry
Figure 2. SEC (A) and CD (B) of M1 unstressed, heat-stressed at 55 °C for 18 h, and heat-stressed at 75 °C for 18 h.
Figure 2B. When M1 was heat-stressed at 75 °C for 18 h, the monomer was almost completely eliminated (Figure 2A), and dimer, trimer, and soluble higher order aggregation species were present as confirmed by SEC-MALS (data not shown). These aggregates accounted for 95% of the total area in the SEC chromatogram. The 75 °C heat-stressed sample was also analyzed by CD (Figure 2B) and showed drastic HOS differences compared to the unstressed and 55 °C stressed M1 with a significant portion of the structure likely in a denatured, random coil formation. These samples with different stress-induced HOS changes and aggregation states serve as useful positive controls with known HOS changes and were further analyzed by the covalent labeling denaturation MS workflow as described below. Maximizing Sensitivity and Resolution for Localized HOS Comparisons. Covalent labeling mass spectrometry has been used widely as a complementary tool to NMR and X-ray crystallography for characterizing native protein structures.26−30,36,38 Since covalent labeling can potentially perturb native conformations, it is important to carefully ensure no label-induced protein structural changes have occurred during the reaction.26−30,33,36,38 However, product quality attributes
application to localized HOS characterization will be assessed in the following sections. Generation of Aggregated, Heat-Stressed Samples for Assessing Method Performance. To test the performance of the covalent labeling methodology, biotherapeutic samples were stressed to generate HOS differences. Specifically, M1 was analyzed under unstressed and heat-stressed conditions (55 °C for 18 h and 75 °C for 18 h) to forcibly induce HOS changes. These samples were analyzed by CD and SEC to assess the degree of aggregation and HOS change. As seen in Figure 2A, very little aggregation was seen in the unstressed “control” sample; however, 36% aggregation was observed after M1 was stressed at 55 °C for 18 h. The peak coming off the column before the monomer was determined to be a dimer species, as confirmed by SEC with multiangle light scattering (MALS) detection (data not shown). The aggregate at the 55 °C heat-stressing condition does not necessarily ensure a significant HOS change, as the monomer species that make up the dimer commonly have a native-like structural form.11,44 The sample was therefore analyzed further by CD, and a significant secondary structural change was observed for the 55 °C sample as compared to the unstressed “control” sample as seen in 2481
DOI: 10.1021/acs.analchem.5b04736 Anal. Chem. 2016, 88, 2478−2488
Article
Analytical Chemistry
Figure 3. Increasing the amount of label increases sensitivity (higher fold changes for more peptides across the protein) and resolution (the number of TMT-labeled amino acid residues and peptides). TMTduplex was used for each labeling condition as follows: TMT 127 was used to label a stressed sample (75 °C for 18 h) and TMT 126 was used to label an unstressed (control) sample. Unique TMT-labeled peptides are arranged from lowest to highest fold change. A fold change near 1 indicates equivalence between the two samples for a given TMT-labeled peptide.
such as HOS are most often assessed as a comparison during the development of therapeutic proteins (whether that be a biosimilar compared to an innovator product, scale-up material compared to an established reference, or a protein stressed in two different formulations, among others); therefore, the exact native conformation is often not as critical. What is important, however, is the ability of a given analytical methodology to detect change, even if that change (e.g., HOS conformational change) comes from a portion of the protein molecules that is stoichiometrically low. With this in mind, we assessed whether adding a substantially higher amount of label compared to the protein (up to a label/protein molar ratio of approximately 1100:1), and thus purposely denaturing the protein, would result in an increase in sensitivity for detecting localized differences in protein therapeutic HOS. Conversely, when all localized conformations were the same between two biologic products, we assessed whether the high degree of label incorporation and any induced structural changes would be equivalent and sufficiently reproducible. To test for sensitivity improvements at different label amounts, we utilized TMTduplex to compare unstressed M1 (labeled with the TMT 126 reagent) with the 75 °C, 18 h stressed M1 sample (labeled with the TMT 127 reagent) using 2 min reaction times. These two samples were compared since they had known and fairly dramatic HOS differences as shown in the previous section. The labeling reactions were performed separately using 0.3, 0.5, 1.1, 5.3, or 9.8 mM concentrations of both TMT reagents, and LC−MS/MS and data analysis was
performed per the Experimental Section. Figure 3 shows a plot of fold change (i.e., the ratio of the 127 and 126 mass tag ion intensities) versus all identified unique TMT-labeled peptides for each reaction condition identified from the LC−MS/MS data. The TMT peptides on the x-axis were ordered from lowest to highest fold change; a fold change near 1 indicates equivalence between the two samples for a given TMT-labeled peptide. As expected, significantly more unique TMT-labeled peptides were identified with increasing TMT label amount. However, the fold changes of many of these peptides were also dramatically higher compared to those using less label amount. Adding high amounts of label can potentially influence enzymatic digestion patterns and thus the resulting fold changes; therefore, we also performed these TMT experiments using Asp-N instead of chymotrypsin. As seen in Figure S1, the same trends were observed for both chymotrypsin and Asp-N digested TMT samplesincreasing fold changes (higher sensitivity) were observed with increasing TMT concentration. While the high-label conditions cannot be used to probe native conformations of the proteins, these results demonstrated that the most sensitive detection of HOS deviations between biotherapeutic samples was achieved by using high levels of label (levels which will cause protein denaturation). To gain more insight into how the label induces protein conformational change, we performed CD, 1D 1H NMR fingerprinting, and dose−response experiments. First, the two labeling extremes (0.3 and 9.8 mM) for the unstressed sample were analyzed by CD to assess the degree of label-induced 2482
DOI: 10.1021/acs.analchem.5b04736 Anal. Chem. 2016, 88, 2478−2488
Article
Analytical Chemistry
showed the highest fold changes between the stressed and unstressed M1 samples were ones that often contained either multiple TMT modifications or modifications to other amino acids besides lysine (e.g., serine, threonine, tyrosine, etc.). These peptides were usually only observed when high amounts of label were used and thus were likely generated during labelinduced denaturation. We hypothesize, therefore, that the highest fold changes are most likely a result of protein unfolding at the site of TMT modification (or the microenvironment), which in turn causes further labeling to occur at a rate that is faster than linear. For example, a biotherapeutic sample with an HOS section that is more exposed and modifiable would have a faster nonlinear rate of label incorporation versus a sample with the same section, but less exposed. These conditions would lead to higher TMT fold changes compared to if the labeling conditions did not induce conformation changes and label incorporation scaled linearly (see Figure S5). The resolution, as expected, was also improved by using high label amounts. The 5.3 and 9.8 mM TMT concentrations have the ability to label lysine residues that are not only on the surface but those located more in the interior of the protein. This increase in labeling is likely aided by the label-induced denaturation of the protein as described previously. Hydroxylcontaining amino acids are also readily modified at high label concentrations, in addition to lysine residues. For instance, 100% of the lysine residues, 85% of the serines, and 69% of the threonines were modified with the highest TMT label amount. Other amino acids were also intermittently modified such as tyrosine and histidine; these residues were not specifically selected as dynamic modifications for TMT labels during database searches (to minimize search space); however, the Nterminus of all peptides was allowed to have TMT labels in order to expand the detection of unforeseen TMT modifiable residues. The high concentrations of label also increases the level of TMT modification at each site, thus increasing the number of detectable and quantifiable unique TMT-labeled peptides. As seen in the legend for Figure 3, the 5.3 and 9.8 mM TMT concentrations yielded the highest resolution with a TMT-labeled residue every 3.6 and 3.7 amino acids, which is nearly 2-fold better resolution compared to the 0.3 mM TMT reactions. Since the 5.3 and 9.8 mM TMT reactions yielded similar sensitivity and resolution (degree of fold changes, number of unique TMT-labeled peptides and amino acids), the protein is likely near saturation with label at the 5.3 mM TMT conditions, and thus little analytical value was added by using the 9.8 mM TMT load. Therefore, the 5.3 mM TMT concentration was used for localized HOS comparisons in the remainder of this study (except where noted) to limit label consumption and cost. Given that protein therapeutic products would unlikely be subjected to excessive denaturants (e.g., high heat, extreme pH differences, oxidizing reagents, etc.) during their lifetime of manufacturing, shipping, and storage, any potential changes in HOS would only be expected for a small population of the total protein molecules. Thus, we assessed the sensitivity of the methodology described herein for detecting low-level HOS differences by spiking in low amounts of stressed M1 (75 °C for 18 h) into unstressed M1. Figure 4A shows a plot of fold change versus all identified unique TMT-labeled peptides for three separate control runs (unstressed M1 labeled with TMT 127 versus the same unstressed sample labeled with TMT 126), stressed M1 spiked into unstressed M1 at 5% (labeled with
HOS change. As seen in Figure S2, the 0.3 mM labeling condition induced no significant or very little HOS change, while the 9.8 mM reaction imparted substantial denaturation of the native conformation. Next, 1D 1H NMR fingerprinting was used to assess the global conformational change upon TMT labeling as it is arguably one of the most sensitive techniques for assessing HOS45 and is a complementary method to CD. The full fingerprinting spectra for the hydroxylamine reaction quench, 9.8 mM TMT reagent, M1 reacted with 9.8 mM TMT for 2 min and quenched with hydroxylamine, and M1 (no reaction) are shown in Figure S3A. A zoomed in version of each of these spectra highlighting the region consisting of proton signals from backbone amines, aromatics, and side-chain amines that is especially sensitive to changes in HOS can be seen in Figure S3B. Striking profile differences were observed between unreacted M1 and the TMT-reacted M1 sample, and proton signals from the hydroxylamine quench and TMT reagent did not cause significant interference in the highlighted fingerprinting region. These results, therefore, indicate denaturation of the M1 conformation upon TMT labeling, similar to what was observed for CD. Lastly, we monitored the labeling kinetics of the TMT reaction by generating dose− response plots. The details of this procedure have been reported previously.33 The amount of covalent attachment between TMT reagent and protein will scale linearly with reagent concentration; however, a deviation in linearity indicates a change in protein conformation. The benefit of this type of analysis is that structural perturbations can be sensitively detected at localized sites of the protein. We generated dose−response plots by first reacting unstressed M1 with TMT 126 using the same reagent concentrations as was used for Figure 3 (0.3, 0.5, 1.1, 5.3, and 9.8 mM). LC−MS/MS analysis was then performed as previously described. Precursor areas were obtained from the resulting TMT-modified peptide identifications, and plots were produced using the following equation: ⎛ ⎞ area of unmodified ⎟ vs [TMT] ln⎜ ⎝ area of unmodified + area of modified ⎠
Figure S4 illustrates dose−response plots for all unique TMTlabeled peptides that had unmodified and modified area counts that were detectable for at least four of the five concentrations. A straight line was drawn to highlight any deviations from linearity. For the lowest concentrations (0.3, 0.5, and 1.1 mM), the points scaled linearly for the majority of the TMT peptides, indicating that no significant structural perturbations were detected. However, substantial deviations from linearity were observed for the two highest concentrations (5.3 and 9.8 mM) for essentially all TMT peptides. These results, as well as those obtained from CD and NMR fingerprinting, exemplify that significant protein denaturation is present after the highconcentration labeling conditions that are utilized in this study. Covalent labeling denaturation depends on a variety of characteristics of the label (e.g., size, bulkiness, charge density, hydrophobicity, among others).26 The TMT labels used in this study are quite large and contain a bulky mass tag, which are probable attributes that readily lead to conformational changes during the attachment of TMT labels to a protein. Labelinduced structural deviations likely lead to perturbations in the local amino acid environment and permit the modification of residues that were not easily modifiable under low-label conditions, thus furthering the cascade of denaturation. In the data presented here, we found the TMT-labeled peptides that 2483
DOI: 10.1021/acs.analchem.5b04736 Anal. Chem. 2016, 88, 2478−2488
Article
Analytical Chemistry
and human error due to manual execution of the labeling reactions. Specific areas of the protein will also be more flexible than other regions, which may increase the variability of the labeling reaction and contribute to fold changes that deviate from 1. Therefore, we always compare sample runs to control runs for localized HOS assessments. As shown in Figure 4A, significant differences (i.e., high fold changes) were detected by the covalent labeling denaturation methodology in samples containing as little as 2.5−5% degraded protein. For the 5% spiked sample, seven unique TMT-labeled peptides were detected with fold changes greater than 2. Of those seven, four had no PTMs and three were deamidated (asparagine converted to aspartic acid). For TMT-labeled peptides that also contain PTMs (such as deamidation), assessing the degree of HOS change is difficult to separate from changes in PTM amounts (i.e., fold changes will represent the degree of PTM difference plus any HOS differences). Consequently, the assessment of HOS changes is more straightforward from TMT-labeled peptides that do not contain PTMs. Since the 5% spiked sample had several TMT peptides without oxidation, deamidation, and glycosylation that yielded high fold changes, it can be concluded that HOS changes were indeed detected at this low level. The 2.5% spiked sample resulted in three TMT peptides with fold changes greater than 2; however, all three peptides contained PTMs, and thus determining the degree of HOS change for these peptides is less straightforward. The two TMT peptides with the highest fold changes both had deamidation modifications, which have been shown to cause structural changes due to alterations in local negative charge.48−55 On the basis of these results, we conclude that changes can be detected when between 2.5−5% of the molecular population has a localized HOS deviation, a value range that is expected to be underestimated since 100% of the molecular population is unlikely denatured for the 75 °C, 18 h heat-stressed sample (see Figure 2). Interestingly, as shown in Figure 4B, when the same spiked samples were analyzed by CD (a technique commonly utilized for characterizing the HOS of protein therapeutics), no significant differences between spiked samples and controls were detected. 1D 1H NMR fingerprinting was also used to analyze these spiked samples as well as several samples with higher percentages of stressed M1. Figure S6A shows the full fingerprinting spectra for several spiked percentages, and Figure S6B presents a zoomed in version of the spectra highlighting the region consisting of proton signals from backbone amines, aromatics, and side-chain amines that is especially sensitive to changes in HOS. While a significant profile change can be observed for the 100% stressed M1 (75 °C for 18 h) and the 50% spiked sample, the samples spiked with lower percentages of stressed material are difficult to differentiate from the control (unstressed M1). To more quantitatively assess the NMR fingerprinting data, we utilized an automated program to measure the peak intensities of the 23 most resolved peaks from the 1D 1H spectra. Each peak intensity was the average of 10 points (five points below and five points above each specific peak chemical shift) to minimize the variation caused by minor peak shifting. The peak intensities were then correlated between samples, and coefficient of determination (R2) values were calculated. In general, the R2 cutoff value for assessing similarity is approximately 0.980 based on replicate analysis of 28 different lots of M1 (data not shown). The spike experiments show that for the 75 °C heat-stressed sample, the detection limit was ∼25% for the 1D 1H NMR method as seen in Table S1. Lastly,
Figure 4. (A) Assessing method sensitivity by spiking stressed (75 °C for 18 h) into unstressed (control) material and comparing fold changes. Three replicate runs of unstressed (control) vs the same unstressed (control) were also analyzed. (B) The same spiked samples analyzed by CD were indistinguishable from the unstressed controls.
TMT 127) versus unstressed M1 (labeled with TMT 126), and stressed M1 spiked into unstressed M1 at 2.5% (labeled with TMT 127) versus unstressed M1 (labeled with TMT 126). The results were plotted in a similar fashion as described for Figure 3. The control runs illustrate that fold changes of around 1 were dominant when comparing two of the same unstressed samples and thus exemplifies that the rate of label incorporation and any label-induced structural changes are sufficiently reproducible. A few of the TMT peptides exhibited fold changes that increased to approximately 2 due to experimental error. This error often arises from coisolation of interfering species,46,47 MS2 product ions that are isobaric with the TMT reporter ions,33 peptides at or near the quantification limits of the mass spectrometer,46 2484
DOI: 10.1021/acs.analchem.5b04736 Anal. Chem. 2016, 88, 2478−2488
Article
Analytical Chemistry
Figure 5. (A) TMTsixplex labeling of unstressed (control) vs the same unstressed (control). (B) TMTsixplex labeling of different heat-stressing conditions and aggregation states (55 °C for 18 h or 75 °C for 18 h) vs unstressed (control) material. Each condition was performed in triplicate (e.g., TMT 126, 127, and 128 were reacted with three replicates of control aliquots and TMT 129, 130, and 131 were reacted with three replicates of heat-stressing aliquots).
HOS differences (comparable to the covalent labeling technique). These spike experiments serve as a means of benchmarking the sensitivity of the proposed MS methodology against several gold standard HOS methodologies that are routinely employed in biologics development. Assessing Local HOS Changes and Repeatability. We used TMTsixplex to characterize the protein locations in M1 where HOS changes were occurring for both heat-stressing conditions (75 and 55 °C for 18 h) and to more thoroughly assess the repeatability of the methodology. The main benefit of TMTsixplex is that three replicates of controls (unstressed M1) and three replicates of stressed conditions (either 75 or 55 °C for 18 h) can be analyzed using a single sample preparation and one LC−MS/MS run. Because of this multiplexing capability, variability from differences in MS ionization, data-dependent peak picking, and sample preparation are less problematic. Thus, repeatability and robustness are improved. Figure 5A depicts a plot of fold change versus unique TMT peptides (arranged from N- to C-terminus) from TMTsixplex labeling of a control run (all six TMT channels were used to label unstressed M1). The fold changes are from the average of the 129/126, 130/127, and 131/128 reporter ions (decimal fold
these spiked samples were analyzed by DSC, which specifically assesses conformational stability of separate domains of the protein and thus provides a less global measurement of HOS. M1 has two melting transitions, and both had essentially the same melting temperature (Tm) values for all samples. The 100% 75 °C heat-stressed sample was the only exception as it had no signal for either melting transition (see Table S2). A slight decrease in signal was observed in the DSC profile for the 5% spiked sample as compared to the unstressed M1 control, and a substantially larger decrease was seen for the 50% spiked sample as illustrated in Figure S7. This decrease in signal is likely the result of the almost completely unfolded M1 conformation present in the 75 °C heat-stressed sample. The profile of the 2.5% spiked sample, however, could not be adequately differentiated from the unstressed M1 control. The culmination of all the experiments above shows that, by using the covalent labeling denaturation methodology presented herein, minor localized HOS deviations can be readily differentiated and detected. Minor structural perturbations, however, can get buried in the global signal for CD and NMR fingerprinting. Compared to these two other gold standard techniques, DSC was shown to be more sensitive in detecting 2485
DOI: 10.1021/acs.analchem.5b04736 Anal. Chem. 2016, 88, 2478−2488
Article
Analytical Chemistry
in Figure 2A, we also expected to see significantly high negative fold changes due to the 36% aggregated dimer species present in the 55 °C heat-stressed sample. However, no such peptides were observed. We hypothesized, therefore, that at the high levels of TMT reagent used in these experiments, the label may interfere with the protein−protein contact sites and break apart the aggregate. We would expect this to be particularly true for aggregates formed from weak noncovalent interactions, such as those formed commonly in antibody dimer species. To investigate this hypothesis, the 55 °C heat-stressed sample was subjected to labeling conditions using the lowest TMT reagent concentration from Figure 3 (0.3 mM). Additionally, both chymotrypsin and Asp-N digestions were utilized to improve sequence identification overlap. Figure S9 depicts a plot of fold change versus unique TMT peptides for these results. Interestingly, several peptides yielded high negative fold changes in the binding domain for both digestions, an area that did not show increased fold changes for the high-concentration labeling reactions. This domain as well as an area in the Fc CH2 were shown to be protected in the 75 °C heat-stressed sample even after high-labeling conditions; however, the harsher stressing generated higher amounts of aggregated species (see Figure 2A) and likely stronger aggregate contacts. As shown in Figure S9, high levels of error accompanied the high negative fold changes for the 55 °C heat-stressed sample labeled with the low TMT concentration. Due to the small amount of label, many of the TMT-modified peptides had low ion abundances, which often resulted in poorer MS2 spectra and only a single count for the fold change calculation. Therefore, to increase the confidence in the aggregation site determination, we performed targeted MS2 experiments and exclusively focused on the TMT peptides that yielded high negative fold changes from the datadependent analyses. We specifically targeted the cluster of TMT peptides circled in red in Figure S9. The resolution was increased from 17 500 to 70 000, and the isolation width was narrowed from 2 to 1 m/z to further decrease potential contaminating ion species and maximize MS2 spectral quality. The targeted MS2 peak profile of one of these peptides is shown in Figure S10A and exemplifies the data output generated using this approach. Figure S10B illustrates the TMT region of the MS2 spectrum after averaging across the peak profile in Figure S10A. This TMT profile shows a clear decrease in labeling after heat-stressing M1 at 55 °C (i.e., TMT 126, 127, and 128 were reacted with three replicates of unstressed aliquots and TMT 129, 130, and 131 were reacted with three replicates of 55 °C heat-stressing aliquots). Furthermore, the fold change results for all four targeted peptides can be seen in Figure S10C. Since every replicate of each targeted peptide yielded a high negative fold change, the region of the binding domain from where these peptides originated is likely a site of aggregation. While the targeted MS2 experiments significantly increased the confidence of the TMTbased quantification over using data-dependent analysis alone, we anticipate further improvements in accuracy and precision by incorporating recent MS advances such as synchronous precursor selection (SPS)60 as well as robotic liquid handling systems for the labeling reactions, both of which we plan to utilize in future studies. On the basis of the results herein, the low-label conditions (no label-induced denaturation) were best for the detection of aggregation sites (especially for aggregates that have weak contact sites), while the high-label conditions that cause label-induced denaturation yielded the most sensitive
changes were converted to negative reciprocals before averaging as discussed in the Experimental Section). Once again, the control run showed fold changes of 1 or lower to be dominant, but a few peptides had fold changes of around 2 due to experimental error (as discussed earlier). The replicate measurements were also reproducible as shown by the error bars. Figure 5B depicts the same control analysis along with TMTsixplex labeling of the two different heat-stressing conditions (55 °C for 18 h or 75 °C for 18 h vs unstressed M1). Three replicates were performed for each condition (e.g., TMT 126, 127, and 128 were reacted with three separate unstressed aliquots and TMT 129, 130, and 131 were reacted with three separate aliquots from the same heat-stressing condition). As stated previously, the assessment of HOS changes is more straightforward from TMT-labeled peptides that do not contain PTMs because HOS change is difficult to separate from the degree of PTM change. Therefore, the data from Figure 5B, with all oxidized, deamidated, and glycosylated peptides filtered out, can also be seen in Figure S8. M1 has two melting transitions per DSC, one near 55 °C and one slightly higher than 75 °C, and was therefore the main reason we picked these temperatures for the heat-stressing samples in this study. On the basis of previous reports, the first transition (∼55 °C) is likely the unfolding of both the protein binding and Fc CH2 domains, and the second transition (∼75 °C) is likely the unfolding of the Fc CH3 domain.56−59 Interestingly, while the 75 °C heat-stressed sample yielded regions of high fold change throughout the protein (due to the high level of stress), two general regions in the Fc CH3 domain yielded the most dramatic differences. Many of these peptides with very high fold changes also had a considerable amount of associated error, usually because the TMT reporter ions from the unstressed M1 sample were near the MS detection limit, and the TMT channels that were not detected were replaced with minimum spectral intensities. Since this sample was highly stressed, many conformations will be present and certain unfolded sections will likely be highly flexible, both of which should increase error compared to what was observed for the control (Figure 5A). There are also areas of the protein in the 75 °C heat-stressed sample that have substantial negative fold changes (i.e., regions that have become more protected in the heat-stressed vs the unstressed sample). The protein section that yielded the most peptides with the highest negative change was near the Cterminal end of the Fc CH2 domain; one TMT peptide in the protein binding domain also had a significantly negative fold change. Although it is possible that these regions are simply areas where sections of amino acids have become buried in the structure, the more likely cause of the negative fold changes are from sites of aggregation, since the areas of the protein affected by the heat-stressing should be more unfolded in general (less buried). The largest differences for the 55 °C heat-stressed sample were located in the Fc CH2 domain. These fold changes, however, were from labeled peptides that were also deamidated. As stated previously, while deamidation can cause HOS changes via the addition of negative charge,48−55 it is difficult to separate the fold change contribution of HOS versus PTM change for the results herein. The 55 °C heat-stressed did yield a TMT peptide with a significantly elevated positive fold change that did not contain a PTM, thus highlighting the sensitivity of detecting HOS differences in lesser stressed samples. This peptide was located in the binding domain and had a fold change of 7.3 ± 3.4. On the basis of the SEC profile 2486
DOI: 10.1021/acs.analchem.5b04736 Anal. Chem. 2016, 88, 2478−2488
Article
Analytical Chemistry
other experiment common to therapeutic protein development that requires thorough product attribute comparisons.
detection of unfolded areas of the protein, or any other potential structural deviations. Quantitation of PTMs and Site-Specific Glycosylation. As mentioned previously, our analytical workflow can also be used to automatically quantify other product attributes such as PTMs and site-specific glycosylation, in addition to localized HOS. Deamidation, glycosylation, and oxidation were specifically searched for and quantified using our automated data analysis workflow, as these PTMs are often the most common and important modifications to therapeutic proteins61 and are routinely monitored during biologics development. We detected 25, 21, and 7 quantifiable unique TMT-labeled peptides (i.e., ones that had fold changes) that were deamidated, glycosylated, and oxidized, respectively, from the combined data of the control and two stressed conditions presented in Figure 5. These peptides covered 16 unique PTM sites. As expected, the heat-stressed samples showed substantially higher degrees of deamidation compared to the control samples (up to 62-fold change)especially for the “hot spot” PTMs that tend to increase at the highest rates. Oxidation was often elevated as well for the stressed samples, but more moderately (up to 4-fold change). Interestingly, many of the areas in the protein with high fold changes for deamidation also had high fold changes associated with HOS unfolding. This denaturation of the HOS could be a direct result from an increase in PTM amount; however, unfolding of the protein due to heating could also cause an increase in solvent exposure for residues such as asparagine and methionine (ones that are highly susceptible to modification). An increase in solvent exposure could then lead to increases in PTM quantities. For the results presented here, it is difficult to decipher whether PTMs induced conformation denaturation or conformational denaturation induced higher levels of PTMs.
■
ASSOCIATED CONTENT
S Supporting Information *
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.5b04736. Fold changes vs amounts of label when using Asp-N as the digestion enzyme (Figure S1), circular dichroism spectra (Figure S2), 1H NMR spectra (Figure S3), dose− response plots for unstressed M1 (Figure S4), hypothesis as to why TMT reactions that utilize high label amounts yield higher fold changes between proteins with HOS differences (Figure S5), 1H NMR spectra of samples of stressed M1 spiked into an M1 control (no stress) at various percentages (Figure S6), coefficient of determination values from the correlation between 23 representative peaks from 1H NMR spectra (Table S1), DSC melting temperature values of samples of stressed M1 spiked into unstressed M1 (Table S2), DSC profiles of spiked samples of stressed M1 spiked into unstressed M1 (Figure S7), localized HOS assessment of heatstressed M1 vs unstressed M1 with all oxidized, deamidated, and glycosylated peptides filtered out (Figure S8), localized HOS assessment of heat-stressed M1 vs unstressed M1 using the lowest TMT label concentration and both chymotrypsin and Asp-N (Figure S9), localized HOS assessment of heat-stressed M1 vs unstressed M1 using the lowest TMT label concentration, multiple enzymes (chymotrypsin and Asp-N), and targeted MS2 (Figure S10) (PDF)
■
■
CONCLUSION We demonstrated herein the utility of using isobaric TMT labels combined with mass spectrometry for localized higher order structure characterization of complex protein biotherapeutics. Specifically, we showed that higher fold changes (sensitivity) and resolution are the result of using label amounts high enough to induce conformational denaturation of the protein sample. This methodology was shown to have similar or better sensitivity compared to several gold standard HOS techniques such as CD, DSC, and NMR fingerprinting, yet provided in-depth analysis of localized structural deviations instead of solely a global HOS measurement. Even with purposeful, label-induced structural changes to the biotherapeutic, we were still able to generate useful information pertaining to likely sites of aggregation for heat-stressed samples that contained high amounts of aggregated species and strong protein−protein contact sites. For aggregates that have limited abundance and weak contact sites, we demonstrated that the lowest label condition (no label-induced denaturation) was best for aggregate site detection, and thus the combination of low and high reagent concentrations may be best for the most thorough HOS characterization of certain samples. Furthermore, the methodology was shown to simultaneously quantify common post-translational modifications such as deamidation, oxidation, and glycosylation. The combination of all these analyses can also be performed on data collected from as little as one LC−MS/MS experiment. We anticipate this methodology to be particularly beneficial for the development of biosimilars, forced degradation studies, and any
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare the following competing financial interest(s): All authors are employees of Momenta Pharmaceuticals with stock compensation.
■
ACKNOWLEDGMENTS We thank Sunyoung Bang and Jim Prescott (Momenta Pharmaceuticals) for providing SEC-MALS analysis of the heat-stressed samples.
■
REFERENCES
(1) Beck, A.; Wurch, T.; Bailly, C.; Corvaia, N. Nat. Rev. Immunol. 2010, 10, 345−352. (2) Hermeling, S.; Crommelin, D. J.; Schellekens, H.; Jiskoot, W. Pharm. Res. 2004, 21, 897−903. (3) Wang, W.; Singh, S. K.; Li, N.; Toler, M. R.; King, K. R.; Nema, S. Int. J. Pharm. 2012, 431, 1−11. (4) Ratanji, K. D.; Derrick, J. P.; Dearman, R. J.; Kimber, I. J. Immunotoxicol. 2014, 11, 99−109. (5) Berkowitz, S. A.; Engen, J. R.; Mazzeo, J. R.; Jones, G. B. Nat. Rev. Drug Discovery 2012, 11, 527−540. (6) Houde, D.; Arndt, J.; Domeier, W.; Berkowitz, S.; Engen, J. R. Anal. Chem. 2009, 81, 2644−2651. (7) Houde, D.; Peng, Y.; Berkowitz, S. A.; Engen, J. R. Mol. Cell. Proteomics 2010, 9, 1716−1728. (8) Houde, D.; Berkowitz, S. A.; Engen, J. R. J. Pharm. Sci. 2011, 100, 2071−2086. 2487
DOI: 10.1021/acs.analchem.5b04736 Anal. Chem. 2016, 88, 2478−2488
Article
Analytical Chemistry
(43) Bern, M.; Cai, Y.; Goldberg, D. Anal. Chem. 2007, 79, 1393− 1400. (44) Roberts, C. J. Trends Biotechnol. 2014, 32, 372−380. (45) Poppe, L.; Jordan, J. B.; Rogers, G.; Schnier, P. D. Anal. Chem. 2015, 87, 5539−5545. (46) Rauniyar, N.; Yates, J. R., 3rd J. Proteome Res. 2014, 13, 5293− 5309. (47) Maes, E.; Valkenborg, D.; Baggerman, G.; Willems, H.; Landuyt, B.; Schoofs, L.; Mertens, I. PLoS One 2015, 10, e0120115. (48) Wright, H. T.; Urry, D. W. Crit. Rev. Biochem. Mol. Biol. 1991, 26, 1−52. (49) Takata, T.; Oxford, J. T.; Demeler, B.; Lampi, K. J. Protein Sci. 2008, 17, 1565−1575. (50) Huang, L.; Lu, J.; Wroblewski, V. J.; Beals, J. M.; Riggin, R. M. Anal. Chem. 2005, 77, 1432−1439. (51) Luo, Q.; Joubert, M. K.; Stevenson, R.; Ketchem, R. R.; Narhi, L. O.; Wypych, J. J. Biol. Chem. 2011, 286, 25134−25144. (52) Robinson, N. E.; Robinson, A. B. Proc. Natl. Acad. Sci. U. S. A. 2001, 98, 12409−12413. (53) Kosky, A. A.; Razzaq, U. O.; Treuheit, M. J.; Brems, D. N. Protein Sci. 1999, 8, 2519−2523. (54) Geiger, T.; Clarke, S. J. Biol. Chem. 1987, 262, 785−794. (55) Liu, Y. D.; van Enk, J. Z.; Flynn, G. C. Biologicals 2009, 37, 313− 322. (56) Liu, H.; Bulseco, G. G.; Sun, J. Immunol. Lett. 2006, 106, 144− 153. (57) Demarest, S. J.; Glaser, S. M. Curr. Opin. Drug Discovery Dev. 2008, 11, 675−687. (58) Ghirlando, R.; Lund, J.; Goodall, M.; Jefferis, R. Immunol. Lett. 1999, 68, 47−52. (59) Fast, J. L.; Cordes, A. A.; Carpenter, J. F.; Randolph, T. W. Biochemistry 2009, 48, 11724−11736. (60) McAlister, G. C.; Nusinow, D. P.; Jedrychowski, M. P.; Wuhr, M.; Huttlin, E. L.; Erickson, B. K.; Rad, R.; Haas, W.; Gygi, S. P. Anal. Chem. 2014, 86, 7150−7158. (61) Liu, H.; Ponniah, G.; Zhang, H. M.; Nowak, C.; Neill, A.; Gonzalez-Lopez, N.; Patel, R.; Cheng, G.; Kita, A. Z.; Andrien, B. MAbs 2014, 6, 1145−1154.
(9) Iacob, R. E.; Bou-Assaf, G. M.; Makowski, L.; Engen, J. R.; Berkowitz, S. A.; Houde, D. J. Pharm. Sci. 2013, 102, 4315−4329. (10) Majumdar, R.; Middaugh, C. R.; Weis, D. D.; Volkin, D. B. J. Pharm. Sci. 2015, 104, 327−345. (11) Zhang, A.; Singh, S. K.; Shirts, M. R.; Kumar, S.; Fernandez, E. J. Pharm. Res. 2012, 29, 236−250. (12) Deperalta, G.; Alvarez, M.; Bechtel, C.; Dong, K.; McDonald, R.; Ling, V. MAbs 2013, 5, 86−101. (13) Wang, L.; Chance, M. R. Anal. Chem. 2011, 83, 7234−7241. (14) Maleknia, S. D.; Ralston, C. Y.; Brenowitz, M. D.; Downard, K. M.; Chance, M. R. Anal. Biochem. 2001, 289, 103−115. (15) Xu, G.; Chance, M. R. Anal. Chem. 2005, 77, 4549−4555. (16) Guan, J. Q.; Chance, M. R. Trends Biochem. Sci. 2005, 30, 583− 592. (17) Takamoto, K.; Chance, M. R. Annu. Rev. Biophys. Biomol. Struct. 2006, 35, 251−276. (18) Hambly, D. M.; Gross, M. L. J. Am. Soc. Mass Spectrom. 2005, 16, 2057−2063. (19) Hambly, D.; Gross, M. Int. J. Mass Spectrom. 2007, 259, 124− 129. (20) Gau, B. C.; Sharp, J. S.; Rempel, D. L.; Gross, M. L. Anal. Chem. 2009, 81, 6563−6571. (21) Chen, J.; Rempel, D. L.; Gross, M. L. J. Am. Chem. Soc. 2010, 132, 15502−15504. (22) Jones, L. M.; Sperry, J. B.; Carroll, J. A.; Gross, M. L. Anal. Chem. 2011, 83, 7657−7661. (23) Chen, J.; Rempel, D. L.; Gau, B. C.; Gross, M. L. J. Am. Chem. Soc. 2012, 134, 18724−18731. (24) Yan, Y.; Chen, G.; Wei, H.; Huang, R. Y.; Mo, J.; Rempel, D. L.; Tymiak, A. A.; Gross, M. L. J. Am. Soc. Mass Spectrom. 2014, 25, 2084− 2092. (25) Vahidi, S.; Stocks, B. B.; Liaghati-Mobarhan, Y.; Konermann, L. Anal. Chem. 2012, 84, 9124−9130. (26) Mendoza, V. L.; Vachet, R. W. Mass Spectrom. Rev. 2009, 28, 785−815. (27) Mendoza, V. L.; Vachet, R. W. Anal. Chem. 2008, 80, 2895− 2904. (28) Mendoza, V. L.; Antwi, K.; Baron-Rodriguez, M. A.; Blanco, C.; Vachet, R. W. Biochemistry 2010, 49, 1522−1532. (29) Mendoza, V. L.; Baron-Rodriguez, M. A.; Blanco, C.; Vachet, R. W. Biochemistry 2011, 50, 6711−6722. (30) Vasicek, L.; O’Brien, J. P.; Browning, K. S.; Tao, Z.; Liu, H. W.; Brodbelt, J. S. Mol. Cell. Proteomics 2012, 11, O111.015826. (31) Cammarata, M.; Lin, K. Y.; Pruet, J.; Liu, H. W.; Brodbelt, J. Anal. Chem. 2014, 86, 2534−2542. (32) Zhou, X.; Lu, Y.; Wang, W.; Borhan, B.; Reid, G. E. J. Am. Soc. Mass Spectrom. 2010, 21, 1339−1351. (33) Zhou, Y.; Vachet, R. W. Anal. Chem. 2013, 85, 9664−9670. (34) Gautier, V.; Boumeester, A. J.; Lossl, P.; Heck, A. J. Proteomics 2015, 15 (16), 2756−2765. (35) Venable, J. D.; Steckler, C.; Ou, W.; Grunewald, J.; Agarwalla, S.; Brock, A. Anal. Chem. 2015, 87, 7540−7544. (36) Kaur, P.; Tomechko, S. E.; Kiselar, J.; Shi, W.; Deperalta, G.; Wecksler, A. T.; Gokulrangan, G.; Ling, V.; Chance, M. R. MAbs 2015, 7, 540−552. (37) Zhang, L.; Lilyestrom, W.; Li, C.; Scherer, T.; van Reis, R.; Zhang, B. Anal. Chem. 2011, 83, 8501−8508. (38) Zhou, Y.; Vachet, R. W. J. Am. Soc. Mass Spectrom. 2012, 23, 708−717. (39) Zhang, H.; Liu, H.; Blankenship, R. E.; Gross, M. L. J. Am. Soc. Mass Spectrom. 2016, 27, 178−181. (40) Borotto, N. B.; Zhou, Y.; Hollingsworth, S. R.; Hale, J. E.; Graban, E. M.; Vaughan, R. C.; Vachet, R. W. Anal. Chem. 2015, 87, 10627−10634. (41) Thompson, A.; Schafer, J.; Kuhn, K.; Kienle, S.; Schwarz, J.; Schmidt, G.; Neumann, T.; Hamon, C. Anal. Chem. 2003, 75, 1895− 1904. (42) Eng, J. K.; McCormack, A. L.; Yates, J. R., III J. Am. Soc. Mass Spectrom. 1994, 5, 976−989. 2488
DOI: 10.1021/acs.analchem.5b04736 Anal. Chem. 2016, 88, 2478−2488
