cysTMTRAQ—An Integrative Method for Unbiased Thiol-based Redox Proteomics*

Protein redox regulation plays important roles in many biological processes. Protein cysteine thiols are sensitive to redox changes and may function as redox switches, which turn signaling and metabolic pathways on or off to ensure speedy responses to environmental stimuli or stresses. Here we report a novel integrative proteomics method called cysTMTRAQ that combines two types of isobaric tags, cysteine tandem mass tags and isobaric tag for relative and absolute quantification, in one experiment. The method not only enables simultaneous analysis of cysteine redox changes and total protein level changes, but also allows the determination of bona fide redox modified cysteines in proteins through the correction of protein turnover. Overlooking the factor of protein-level changes in the course of protein posttranslational modification experiments could lead to misleading results. The capability to analyze protein posttranslational modification dynamics and protein-level changes in one experiment will advance proteomic studies in many areas of biology and medicine.

Protein redox regulation plays important roles in many biological processes. Protein cysteine thiols are sensitive to redox changes and may function as redox switches, which turn signaling and metabolic pathways on or off to ensure speedy responses to environmental stimuli or stresses. Here we report a novel integrative proteomics method called cysTMTRAQ that combines two types of isobaric tags, cysteine tandem mass tags and isobaric tag for relative and absolute quantification, in one experiment. The method not only enables simultaneous analysis of cysteine redox changes and total protein level changes, but also allows the determination of bona fide redox modified cysteines in proteins through the correction of protein turnover. Overlooking the factor of protein-level changes in the course of protein posttranslational modification experiments could lead to misleading results. The capability to analyze protein posttranslational modification dynamics and protein-level changes in one experiment will advance proteomic studies in many areas of biology and medicine. Changes in the redox states of protein cysteine thiols serve as regulatory switches in diverse biological processes (1). The redox cycle is regulated by well-known factors such as the ferredoxin-thioredoxin and glutathione-glutaredoxin systems, which reduce oxidized cysteines. Other oxidoreductases and oxidants such as reactive oxygen species act primarily to oxidize cysteine thiol groups (2,3). In order to map and quantify cysteine redox modifications on the proteome scale, several approaches and methods have been developed, mostly using thiol-specific reagents and isotope tags. Twodimensional gel electrophoresis technology combined with fluorescent dye labeling (e.g. monobromobimane (4,5) and cyanine dyes (6,7)) and gel-free technology with isotope tagging (e.g. isotope-coded affinity tagging (6 -8), cysTMT 1 (9), and iTRAQ labeling of enriched cysteine-containing peptides (10 -14)) are often used to identify potential redox-sensitive cysteine residues and quantify redox changes.
In addition to the well-known capabilities and limitations associated with two-dimensional gel electrophoresis-based and gel-free approaches (15), each method has its strengths and weaknesses in redox proteomics. For example, the twodimensional gel electrophoresis methods allow the inspection of spot patterns related to redox and protein-level changes. However, spot-volume-based quantification becomes problematic, as each spot often contains more than one protein species from complex samples. In addition, the limited number of fluorescent reagents compromises multiplexing capability, and the use of cyanine dyes does not allow mapping of the modified cysteines (6,7). Other thiol labeling approaches such as the use of N-ethylmaleimide, biotin-N-[6-(Biotinamido)hexyl]-3-(2-pyridyldithio) propionamide (16), and isotopecoded affinity tags allow specific enrichment of cysteinecontaining peptides, mapping of cysteine modification sites, and duplex experiments in the case of isotope-coded affinity tags (6,7). To enable multiplexing, 4-or 8-plex iTRAQ tags were recently used to label cysteine-containing peptides isolated from thiol-affinity chromatography (10,11,14,16). Another multiplexing technology, cysTMT, was developed to specifically label cysteines with free thiol groups of proteins from six different samples (9). Although these multiplexing technologies have found utility, they do not address the issue of protein turnover in the course of experiments, and many researchers have overlooked this important factor that could lead to misleading results (8,13,14,17). Only a small number of researchers have attempted to compare potential redox changes determined in proteomics experiments to total protein-level changes obtained from parallel or different studies (6,7,16). However, the success of this strategy is often low because many proteins quantified in redox experiments are either absent or not quantified with confidence in total proteomics experiments as a result of experimental variation and mass spectrometry stochastical sampling issues (18,19).
To overcome this challenge, we have developed a doublelabeling strategy that uses iTRAQ and cysTMT in one experiment for the simultaneous determination of quantifiable cysteine redox changes and protein-level changes. This new strategy, named cysTMTRAQ, utilizes each of the tags for their specific chemical properties. cysTMT tags (m/z 126, 127, 128, 129, 130, and 131 for six samples) were used to label protein thiols responsive to a treatment, and iTRAQ tags (m/z 114, 115, 116, 117, 119, and 121 for six samples) were used to label the N termini of peptides for analysis of protein-level changes during the experiments. By taking advantage of the different mass tags and their labeling specificities, one can quantify changes in protein redox and total levels in the same experiment. As protein redox regulation is a ubiquitous process (1,2,12), the utility of this new integrative cysTMTRAQ method is expected to greatly advance redox proteomic studies in many fields of biology and medicine, and thus benefit a broad range of scientists.

EXPERIMENTAL PROCEDURES
Sample Preparation-A six-protein mixture containing bovine serum albumin, ␣-lactabumin, ␤-galactosidase, lysozyme, apotransferrin, and ␤-lactoglobulin from the AB Sciex iTRAQ labeling kit (AB Sciex Inc., Foster City, CA) was used for optimization of the doublelabeling procedure and oxidation experiments. Proteins were solubilized in a lysis buffer containing 6 M urea, 50 mM Tris, and 1 mM ethylenediaminetetraacetic acid, pH 7.5, reduced with 5 mM tris(2carboxyethyl)phosphine for 1 h at room temperature. Oxidation treatment was performed with 1 mM H 2 O 2 for 1 h at room temperature. Zeba columns (Thermo Scientific Inc.) were utilized after treatment to remove the reducing and oxidizing agents. Ellman's test for free thiols was performed to determine the free thiol concentration as described previously (20). Samples for optimization of the labeling procedure were prepared in 10-g, 20-g, and 40-g replicates in order to examine 1:2:4 fold change ratios. Two independent replicates of each concentration were prepared. Samples for the identification of redox proteins and cysteines were prepared at a fixed amount of 20 g and included reduced, four ratios of combined oxidized and reduced (1:1, 1:2, 1:4, and 1:8), and oxidized samples. Culture of Escherichia coli was grown aerobically in LB medium to an A 600 of 0.4 at 37°C and treated with 500 M NaOCl for 1 h. Protein extraction was conducted as previously described (21).
cysTMT and iTRAQ Labeling-Direct and reverse labeling of thiols were performed as suggested by the manufacturer (Thermo Scientific Inc., Bremen, Germany) and Parker et al. (22), respectively. Protein samples used for optimization of the labeling procedure were directly labeled after treatment with tris(2-carboxyethyl)phosphine and Zeba column clean-up. We labeled samples with a 1ϫ fold change with 126 m/z and 127 m/z, those with a 2ϫ fold change with 128 m/z and 129 m/z, and those with a 4ϫ fold change with 130 m/z and 131 m/z. Reverse labeling was performed for the redox experiments. The reduced and oxidized samples underwent alkylation with 10 mM 2-iodoacetamide for 1 h at 37°C. Reduced thiols for reverse labeling were produced through treatment with 5 mM tris(2-carboxyethyl)phosphine for 1 h at room temperature. A quick clean-up with a Zeba column was performed for both samples before cysTMT labeling. Samples were combined in ratios of oxidized to reduced sample, increasing the amount of reduced sample with each ratio while maintaining a constant total amount of protein. Ratios included 1:1 (127 m/z), 1:2 (128 m/z), 1:4 (129 m/z), and 1:8 (130 m/z). Fully reduced and oxidized samples were labeled with 126 m/z and 131 m/z, respectively. For each labeling method, a 3-fold excess of cysTMT tag to calculated free thiols was used, and labeling was performed at room temperature for 2 h. A 1:10-fold trypsin (Promega, Madison, WI) digestion was performed according to the iTRAQ manual (AB Sciex Inc.) at 37°C for 12 h (9). Peptides were subsequently cleaned up with C18 desalting columns (The Nest Group Inc., Southborough, MA) and then lyophilized to dryness.
For iTRAQ labeling, lyophilized samples were resuspended in 0.5 M triethylammonium bicarbonate. Samples were labeled according to the manufacturer's manual (AB Sciex Inc.), and a final concentration of 70% isopropanol was used for all the iTRAQ tags. After labeling, the samples were quenched with H 2 O, combined, and lyophilized. Strong Cation Exchange and LC-MS/MS-Peptides were dissolved in strong cation exchange solvent A (25% (v/v) acetonitrile, 10 mM ammonium formate, and 0.1% (v/v) formic acid, pH 2.8) and fractionated using an Agilent HPLC 1260 system with a polysulfoethyl A column (2.1 ϫ 100 mm, 5 m, 300 Å). Peptides were eluted with a linear gradient of 0% to 20% solvent B (25% (v/v) acetonitrile and 500 mM ammonium formate, pH 6.8) over 50 min followed by ramping up to 100% solvent B in 5 min. Six fractions were collected by monitoring the absorbance at 280 nm. The fractions were lyophilized and dissolved in 10 l of loading solvent (3% v/v acetonitrile, 0.1% v/v acetic acid) and loaded onto a C18 PepMap nanoflow column (75-m internal diameter, 3 m, 100 Å, Dionex Inc., Bannockburn, IL). The elution gradient started at 97% solvent A (0.1% v/v acetic acid, 3% v/v acetonitrile)/3% solvent B (0.1% v/v acetic acid, 96.9% v/v acetonitrile) and finished at 40% solvent A/60% solvent B within 90 min. After each run, the column was washed with 90% solvent B and re-equilibrated with solvent A. MS/MS analysis was carried out on an LTQ Orbitrap XL mass spectrometer in a positive mode applying data-dependent MS scanning and MS/MS acquisition (Thermo Scientific, Bremen, Germany). Survey full-scan MS spectra (m/z 400 -1800) were acquired in the Orbitrap with resolution r ϭ 60,000 at m/z 400. Data-dependent analysis of the 10 most intense ions with detection in MS/MS for acquiring HCD was set at 7500 at 400 m/z resolution with a 30,000 signal threshold, 5-ms activation time at 35 normalized collision energy, and dynamic exclusion enabled for 30 s with a repeat count of 1. Typical mass spectrometric conditions include a spray voltage of 2.2 kV, no sheath and auxiliary gas flow, a heated capillary temperature of 200°C, a capillary voltage of 44 V, a tube lens voltage of 165 V, and an ion isolation width of 1.0 m/z. The mass spectrometer calibration was performed according to the manufacturer's guidelines using a mixture of SDS, sodium taurocholate, the MRFA peptide, and Ultramark.
Database Searching and Data Analysis-Proteome Discoverer 1.4 with the SEQUEST algorithm (Thermo Scientific Inc., Bremen, Germany) was used to search a database composed of sequences of the six-protein mixture (AB Sciex Inc.) in the background of an IPI Homo sapiens sequence database (downloaded on July 12, 2012; 91,464 entries) for the six-protein samples and a bacterial database from NCBI (downloaded on April 3, 2013; 14,961,948 entries) for the E. coli samples. Proteome Discoverer nodes for spectrum grouper and spectrum selector were set to default with the spectrum properties filter set to a minimum and maximum precursor mass of 300 Da and 5 kDa, respectively. The SEQUEST algorithm was used for protein identification. Parameters were set to two maximum missed cleavage sites of trypsin digestion, absolute XCorr threshold of 0.4, and fragment ion cutoff percentage at 0.1. Tolerances were set to a 10-ppm precursor mass tolerance and a 0.01-Da fragment mass tolerance. Dynamic modifications included phosphorylation (ϩ79.966 (S,T,Y)), oxidation (ϩ15.995 Da (M)), carbamidomethyl (ϩ57.021 Da (C)), iTRAQ8plex (ϩ304.205 Da (K), and cysTMT6plex (ϩ304.177 Da (C)). The N-terminal modification was set for iTRAQ8plex (ϩ304.205 Da). Percolator was used for protein identification with parameters of a strict target false discovery rate of 0.01 and a relaxed target false discovery rate of 0.05. The reporter ion quantifier node was set with a peak integration tolerance of 20 ppm, and the event detector used mass precision set at 2 ppm with a signal-to-noise ratio threshold of 3. Quantification was performed for both cysTMT and iTRAQ reporter ion peak intensities. Peptide results were filtered to include only peptides with false discovery rate confidence and cysteine residues.
After peptide and protein identification, cysTMT and iTRAQ datasets were analyzed, and fold changes were compared among the respective reporter ions. The reporter ion peak intensities of identical peptides were summed, and those values were log 2 transformed. Outliers were removed as described previously (23). Ratios generated during the optimization experiment were set relative to 1ϫ, and means were normalized to their respective replicates (24). Ratios generated during the application of redox treatment were set relative to 1:1, and background peak intensity values generated for the reduced sample were subtracted from all sample peak intensities. Means were also normalized to their respective replicates (24). For protein quantification, at least three unique peptides were required. Protein quantification was performed by summing the reporter ion peak intensities of all corresponding peptides (25), and ratios were generated as described above. To analyze the data from biological samples, we calculated the ratios accordingly from the median normalized peak intensity values, as reported above. Student's t test (two-tail) on the log2-transformed treated/control ratio was performed, and the p values were further corrected via the q-value false discovery rate method (26). A peptide with a q-value less than 0.05 and a fold change greater than 1.2 or less than 0.8 was considered statistically significant. The significant peptides labeled with cysTMT were confronted with the significant proteins quantified via iTRAQ. Student's t test was conducted between the fold change of cysTMTlabeled peptides and the fold change of the corresponding proteins quantified via iTRAQ. A correction factor was applied to the fold change of cysTMT-labeled peptides, taking into account the fold change of the protein quantified via iTRAQ (supplemental Fig. S1).
The list of all peptides and proteins identified can be found in the supplemental material (supplemental Tables S2, S3, and S4).

Optimization of CysTMT and iTRAQ Double-labeling CysT-MTRAQ Procedures-
To test the feasibility of double-labeling of peptides derived from a six-protein mixture with cysTMT and iTRAQ tags, a direct labeling method was implemented, and the amounts of different samples were controlled in order to observe the experimental fold changes between samples (supplemental Fig. S2). Proteins were aliquoted into six samples, two at 10 g, two at 20 g, and two at 40 g. After reduction, the cysTMT tags were used to label the six sam-ples. The samples then underwent trypsin digestion and iTRAQ labeling followed by mixing to create theoretical fold changes of 1:1:2:2:4:4. Reporter ion ratios were exported from Proteome Discoverer, and tagging efficiency was determined by examining the deviation of the median values of the reporter ion ratios from the expected values for both peptides and proteins. Using 1ϫ, 2ϫ, and 4ϫ fold changes to examine the double-labeling efficiency, we found that cysteine-containing peptides labeled with cysTMT followed the expected fold changes (Fig. 1A, supplemental Fig. S3). Peptides with amine-labeling iTRAQ tags also showed the expected fold changes (Fig. 1A, supplemental Fig. S3). The greater variability observed in the iTRAQ quantification relative to the cysTMT quantification might be due to the difference in the number of peptides analyzed. When only the cysteine-containing peptides were analyzed, the variations in both cysTMT and iTRAQ quantification were similar, (supplemental Fig. S4). Interestingly, the peptide-level quantification can be extended to the protein level (i.e. the protein-level changes also followed the expected ratios) (supplemental Fig. S5). MS/MS spectral analysis of cysteine peptides revealed that the fold changes of iTRAQ tags in the low-mass region were similar to the fold changes of cysTMT tags (Fig. 1B), suggesting both iTRAQ and cysTMT tags labeled the peptides at different functional groups with the same efficiency. Correlation analysis of the iTRAQ quantitative data and the cysTMT quantitative data showed that the two datasets were highly correlated (R 2 ϭ 0.94) (Fig. 1C). These results demonstrate that doublelabeling of peptides is feasible using the different isobaric tags (iTRAQ and cysTMT). The expected fold changes of the samples can be determined at both peptide and protein levels.
Establishment of CysTMTRAQ in Identification of Redox Proteins and Cysteines-To implement the double-labeling method in a redox experiment, two sets of proteins were reduced and oxidized respectively, and then combined in different ratios while maintaining a constant total protein amount in each sample (supplemental Fig. S6). An indirect labeling procedure was used in which the six protein mixtures underwent thiol alkylation, reduction of oxidized thiols, and subsequent labeling with the thiol-specific cysTMT tags and amine-labeling iTRAQ tags (supplemental Fig. S6). CysTMT ratios of the treated samples reflected the relative amounts of oxidized proteins in the samples, and the experimental ratios turned out to fit the expected values for all the samples ( Fig.  2A, supplemental Fig. S7A). Analysis of iTRAQ tags revealed that the peptide amounts did not change among the samples (Fig. 2B, supplemental Fig. S7B), and the results at the protein level were consistent with those at the peptide level (supplemental Fig. S8). Spectral analysis showed a similar trend in cysTMT and iTRAQ labeling of cysteine-containing peptides (Fig. 2C), and no change in iTRAQ labeling in non-cysteinecontaining peptides (Fig. 2D). These results clearly demonstrate that the double isobaric tag labeling procedure allows one to examine cysteine redox changes while monitoring protein level changes, if any.
Application of cysTMTRAQ in Redox Proteomics of E. coli Cells under Oxidative Stress-We further tested the application of cysTMTRAQ in redox proteomics of complex samples. To expose the cells to oxidative stress, we treated exponentially growing E. coli cells with 500 M NaOCl for 1 h to induce both cellular redox and protein-level changes (supplemental Fig. S9). Cell culture was harvested directly into ice-cold 100% (w/v) TCA. This "acid trap" step is extremely important because the addition of TCA not only lyses the cells but also protonates all thiolates, which significantly slows thiol-disulfite exchange reactions (27). In total, we identified 912 cysteine-containing peptides, and 96.5% of them had both iTRAQ and cysTMT labels. Out of the 912 cysteine-containing peptides, 33 were significantly redox-regulated, showing a q-value of Յ0.05 and either a Ն1.2-fold increase or a Յ0.8fold decrease (supplemental Table S1). Among the redox- regulated proteins, 10 (e.g. outer membrane protein A, glycerol-3-phosphate dehydrogenase, and malate dehydrogenase) were previously reported to have redox-sensitive cysteines (supplemental Table S1). When the protein-level change was taken into account, 5 of the 33 peptides containing redox-sensitive cysteines also showed significant protein-amount changes (Fig. 3A). Colicin receptor I and fructose-bisphosphate aldolase were assigned as redox regulated when only cysTMT quantification was analyzed. However, when the protein-turnover correction was applied (based on iTRAQ quantification), these proteins did not show significant redox regulation (Fig. 3B). DISCUSSION The results presented here have demonstrated that doublelabeling of proteins with two different series of isobaric tags, cysTMT and iTRAQ, is feasible and can be used to examine protein redox and total amount changes simultaneously in one experiment. Because no enrichment step of cysteine peptides is involved, the method has the potential to reveal the stoichiometry of redox modifications. However, an enrichment step can be included to capture low-abundance redox proteins after the double-labeling procedure. In this case, both the enriched fraction and the flow-through will need to be analyzed via LC-MS/MS. Compared with existing redox proteomics approaches (e.g. 6,7,16), this newly developed cysTMTRAQ method is advantageous because it allows researchers to accurately identify and quantify redox cysteines, peptides, and proteins and obtain high efficiency in the correlation between redox changes and total protein changes. Because cysTMT and iTRAQ are employed in the same experiment with the same samples, the sample variation and complication of proteinlevel changes in redox peptide/protein determination are resolved. High confidence and accuracy in the determination of the redox sensitivity of cysteine peptides can be achieved with high efficiency. Because this method acquires both redox and total protein-level information, samples in small quantities such as stomata preparations and patient biopsies can be preserved. In contrast to the use of two different experiments that lead to low correlation and high variation of identified peptides and proteins, this integrated cysTMTRAQ method is efficient and effective. However, it should be noted that the inherent problem of precursor interference associated with MS2 tandem mass tag and iTRAQ technologies has caused ratio compression (28,29). The MS3 function on the new generation of mass spectrometers (e.g. the Thermo Scientific Orbitrap Fusion) can overcome this problem. Nevertheless, this cysTMTRAQ technological breakthrough will stimulate progression in redox proteomics, as well as other posttranslational modification proteomics, in many areas of biology and medicine.