Optimization of Protein-Level Tandem Mass Tag (TMT) Labeling Conditions in Complex Samples with Top-Down Proteomics

Abstract

Isobaric chemical tag labels (e.g., iTRAQ and TMT) have been extensively utilized as a standard quantification approach in bottom-up proteomics, which provides high multiplexing capacity and enables MS2-level quantification while not complicating the MS1 scans. We recently demonstrated the feasibility of intact protein TMT labeling for the identification and quantification with top-down proteomics of smaller intact proteoforms (< 35 kDa) in complex biological samples through the removal of large proteins prior to labeling. Still, the production of side products during TMT labeling (i.e., incomplete labeling or labeling of unintended residues) complicated the analysis of complex protein samples. In this study, we systematically evaluate the protein-level TMT labeling reaction parameters, including TMT-to-protein mass ratio, pH/concentration of quenching buffer, protein concentration, reaction time, and reaction buffer. Our results indicated that: (1) high TMT-to-protein mass ratio (e.g., 8:1, 4:1), (2) high pH/concentration of quenching buffer (pH > 9.1, final hydroxylamine concentration > 0.3%), and (3) high protein concentration (e.g., > 1.0 μg/μL) resulted in optimal labeling efficiency and minimized production of over/underlabeled side products. >90% labeling efficiency was achieved for E. coli cell lysate after optimization of protein-level TMT labeling conditions. In addition, a double labeling approach was developed for efficiently labeling limited biological samples with low concentrations. This research provides practical guidance for efficient TMT labeling of complex intact protein samples, which can be readily adopted in the high-throughput quantification top-down proteomics.

Introduction

The ability to identify and relatively quantify intact proteoforms using quantitative top-down proteomics techniques has enabled the study of diverse biological processes that are mediated by post-translational modifications (PTMs).^1,2 Various quantitative techniques have been adapted and applied to top-down proteomics, including label-free quantitation, metabolic labeling, and isobaric chemical labeling. However, isobaric chemical tag labeling methods including relative and absolute quantitation (iTRAQ)³, tandem mass tags (TMT)^4,5, and N, N-dimethylleucine (DiLeu) isobaric tags^6,7 have been widely applied to the quantification of peptides and proteins in bottom-up proteomics studies, the application to top-down proteomics has been limited.

Previously, intact protein-level isobaric tag labeling has been coupled with bottom-up proteomics methods to improve proteome coverage or characterize protein structures. iTRAQ was first applied to protein-level labeling using standard proteins by Wiese et al. to measure the labeling efficiency of intact proteins using matrix-assisted laser desorption ionization-time-of-flight (MALDI-TOF) mass spectrometry analysis.⁸ Sinclair et al. further applied protein-level TMT labeling to human serum samples; the TMT labeled proteins were separated using strong anion exchange (SAX) chromatography and digested to achieve a higher proteome coverage.⁹ Nie et al. applied isobaric chemical tag labeling for serum glycoprotein quantification and evaluated the effect of various digestion enzymes, isobaric chemical tags, and organic solvents on the identification and quantification of peptides.¹⁰ TMT was also applied to whole cells for protein quantification with combination of enrichment using anti-TMT antibody.¹¹ Kaiwen et al. also developed a native protein TMT method to profile Lys accessibility and applied it to investigate structural change in Alzheimer’s disease brain specimens.¹²

Top-down proteomics on protein-level isobaric chemical tag labeled standard proteins has been previously performed by Hung et al;¹³ however, the application of protein-level isobaric tag labeling to complex samples has been limited. One major challenge has been protein precipitation during labeling due to the presence of organic solvent which limits labeling efficiency. We previously demonstrated that the removal of large proteins by coupling molecular weight cutoff (MWCO) and size exclusion chromatography prevents intact proteins from precipitating under labeling conditions and allows for the identification and quantification of smaller proteoforms (≤ 35 kDa) in E. coli cell lysate.¹⁴ Moreover, the application of thiol-directed isobaric labeling for quantification and identification using top-down proteomics has also been demonstrated for E. coli and combined E. coli/yeast samples.¹⁵

Another challenge for protein-level TMT labeling is the production of side products (i.e., overlabeled and underlabeled species) during the labeling reaction. Overlabeling (labeling of residues other than lysine or the N-terminal) may occur on residues with side chains containing hydroxyl groups (Ser, Thr, Tyr). Underlabeling (missing labels) may occur on the N-terminal or lysine residues. Over and underlabeling does not have a significant impact on quantification results; however, the side products will decrease the signal-to-noise ratio and increase the sample complexity.^8,13

Recently, the TMT labeling condition for peptides was optimized by Zecha et al. and the impact of labeling reaction parameters was evaluated to achieve >99% labeling efficiency.⁸ Considering protein-level labeling is more complicated than peptide-level labeling, it is necessary to optimize protein-level labeling conditions to increase the labeling efficiency. In this study, we systematically optimized the parameters of the protein-level TMT labeling reaction using to maximize the production of the completely labeled species. Overall, we evaluated the effect of TMT-to-protein ratio (w/w), labeling reaction conditions (i.e., quenching buffer, reaction time, and reaction buffer), and protein concentration on protein-level TMT labeling efficiency. Our results demonstrate an optimized condition for intact protein-level TMT labeling of complex samples. We also innovated a double labeling strategy as an alternative approach to achieve sufficient labeling efficiency for mass limited samples that did not have adequate concentrations to implement the optimized labeling conditions (e.g., clinical samples).

EXPERIMENTAL

Chemicals and materials.

All reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA) unless noted otherwise. Acetonitrile (ACN), isopropanol (IPA), trifluoroacetic acid (TFA), and water used for LC mobile phases were LC-MS grade. Tris (2-carboxyethyl) phosphine hydrochloride (TCEP), Pierce^™ BCA Protein Assay Kit, molecular weight cutoff spin filters (10K and 100K MWCO), TMT 6-plex Isobaric Label Reagent set (90061) and TMT zero Isobaric Label Reagent set (90067) were obtained from Thermo Fisher (Waltham, MA, USA).

E. coli cell lysate preparation.

E. coli K12 cells were inoculated in 250 mL sterilized LB medium at 37 °C, 250 rpm for 8 hours. Then 250 mL E. coli was transferred into 2 L LB solution and incubated at 37 °C, 250 rpm for 12 hours. The cells were collected and pelleted through centrifugation at 10,000 × g, 4 °C for 20 minutes. Cell pellets were resuspended in 25 mM ammonium bicarbonate (ABC) buffer at the ratio of 5:1 (w/w) with addition of 0.1% (v/v) PMSF, as described previously.¹⁶ Cells were lysed using high pressure Avestin C3 EmulsiFlex homogenizer (ATA Scientific Instruments, Australia) with the operating pressure between 1500 psi and 2000 psi. The lysate was cooled on ice to offset the slow heat buildup during homogenization. After homogenization, the cell lysate was centrifuged at 10,000 × g, 4 °C for 40 minutes to remove the insoluble debris. Protein concentration was measured using Pierce^™ BCA Protein Assay Kit. The E. coli lysate was aliquoted and stored at −80 °C.

HeLa cell culture and cell lysate preparation.

HeLa cells were grown and lysed as previously described.¹⁷ Briefly, HeLa cells were cultured in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 2% penicillin-streptomycin under 5% CO₂ at 37 °C. Hela cells were washed with ice-cold PBS buffer and scratched off, followed by centrifugation at 2, 500×g, 4 °C for 20 minutes, resuspended into lysis buffer (1 μM PMSF and 20 mM NaF in PBS, pH = 7.4), and lysed using a sonicator. Cell lysate was centrifuged at 15,000 ×g, 4 °C for 30 min and the supernatant was collected. Protein concentration was measured by Pierce^™ BCA Protein Assay Kit. The HeLa cell lysate was then aliquoted and stored at −80 °C.

Protein-level TMT labeling.

Spin filters were washed using the labeling buffer (100 mM TEAB, pH=8.5 or 100 mM HEPES, pH=8.5) prior to protein filtration. Cell lysate were centrifuged at 12,000 × g and 4 °C for 20 minutes filtered in 100 kDa MWCO spin filters to remove high molecular weight proteins. The low MW filtrate was then concentrated using a pre-washed 10 kDa MWCO spin filter for the removal of small proteins and protein concentration by centrifugation at 12,000 × g and 4 °C for 15 minutes. Right before labeling, the protein sample was diluted to 1 μg/μL using 6 M fresh urea in the labeling buffer to the final urea concentration of 3 M. 800 μg proteins were reduced by adding 160 μL 0.5 M TCEP (incubating for 15 minutes at room temperature), and then alkylated by adding 215 μL of 375 mM IAA (incubating at room temperature at dark for 30 minutes). The reduced and alkylated proteins were then buffer-exchanged and concentrated with labeling buffer using a 10 kDa MWCO spin filter to the appropriate concentration for TMT labeling. TMTsixplex reagent or TMTzero reagent was used for the optimization of intact protein-level labeling. Reactant volumes, concentrations, and reaction conditions for each experimental condition are specified in Supplementary Table S1. The labeled sample was centrifuged at 12,000 × g and 4 °C for 30 minutes before LC-MS/MS analysis to remove any precipitation.

For TMT labeling of HeLa cell lysate, 100 kDa MWCO filtered HeLa cell lysate was reduced and alkylated as described above. This samples was then buffer-exchanged into 100 mM TEAB buffer, pH 8.5 using a 10 kDa MWCO filter. Protein concentration was diluted to 1.5 μg/μL before TMT labeling. HeLa cell lysate was aliquoted into 6 tubes and labeled with TMTsixplex individually at the optimal conditions: TMT reagent and proteins were incubated with a 4:1 TMT-to-protein mass ratio at room temperature for 1 hour, then the same amount of TMT reagent was added again, and the sample was incubated for an additional hour to complete double labeling. The TMT labeling reaction was quenched using hydroxylamine to a final concentration of 1.2% (pH 9.5). Six TMT-labeled samples were mixed with a mass ratio of 5:5:2:2:1:1 for LC-MS/MS analysis.

Top-down RPLC-MS/MS analysis.

10 μg of TMT-labeled protein sample was loaded onto a trapping column (150 μm i.d., 5 cm length, Jupiter particles, 5 μm diameter, 300 Å pore size) and separated using a home-packed C5 RPLC capillary column (75 μm i.d., 70 cm length, Jupiter particles, 5 μm diameter, 300 Å pore size) on a modified Thermo Scientific (Waltham, MA, USA.) Accela LC system.^18,19 Mobile phase A was 0.01% TFA, 0.585% acetic acid, 2.5% isopropanol, and 5% acetonitrile in water. Mobile phase B was 0.01% TFA, 0.585% acetic acid, 45% isopropanol, and 45% acetonitrile in water. A 200-min gradient from 10% to 70% of MPB was applied with a flow rate of 400 nL/min.^14,18–20 A customized nano-ESI interface¹⁴ was coupled to an LTQ Orbitrap Elite mass spectrometer using positive ion mode (Thermo Fisher Scientific, Hanover Park, IL, Bremen, Germany, USA). The temperature of the inlet capillary was set to 275 °C and the spray voltage was 2.6 kV. Full MS scans were collected with a resolving power setting of 240,000 (at m/z 400) with two micro scans. The top 6 most abundant precursor ions in the full MS scans were selected for MS/MS fragmentation. The MS/MS data were obtained using a resolving power setting of 60,000 (at m/z 400) with two microscans, an isolation window of 5.0 m/z and 60 s dynamic exclusion. Ions with charge state <4 were rejected for MS/MS fragmentation. Collision-induced dissociation (CID) with a normalized collision energy of 35 eV was applied. The maximum injection time was 1000 ms for full mass scans and 800 ms for MS/MS scans. The AGC target was 1×10⁶ for full mass scans and 6×10⁵ for MS/MS scans.

For TMT-labeled HeLa samples, a 60-min gradient from 10% to 70% MPB was applied and the other LC parameters were consistent with the previously stated parameters. MS detection was performed using and Orbitrap Exploris 240 mass spectrometer with positive mode (Thermo Fisher Scientific, Hanover Park, IL, Bremen, Germany, USA). Full MS resolution was set at 120,000 with three microscans. The resolution of MS/MS was 120,000 with 2 microscans. The AGC target was set to 3×10⁶ for MS and 1×10⁶ for MS/MS. Maximum injection time was 1000 ms for MS and 500 ms for MS/MS. Top 6 most abundant precursor ion peaks were selected for MS/MS fragmentation, and 35% of normalized HCD collision energy was used for MS/MS. The isolation window was set to 2 m/z. Dynamic exclusion was 20 s. Charge 4–50 ions will be included for MS/MS fragmentation.

Data analysis.

All top-down proteomics datasets were deconvoluted using Biopharma Finder (Thermo Fisher Scientific) and were searched against the annotated E. coli protein database (UniProt 2019-10-13, 4519 species) or Human protein database (UniProt 2020-07-11, 20368 species) using TopPIC Suite.^21,22 The TopPIC parameters were MS1 signal-to-noise ratio was 3, the precursor window size was 5.0 m/z and the maximum mass was 50,000 Daltons. Decoy database searching was used and the FDR cutoff was set to 0.01 for the spectrum and proteoform levels. The maximum number of mass shifts was 2 and the mass shift range was ± 500 Daltons. TMT labeling at the N-terminal and lysine residues was set as a fixed PTM and the variable PTM list is given in Supplementary Table S2. All other parameters were set as default values. MASH Suite²³ and ProSight Lite²⁴ were used for manual interpretation and spectrum presentation.

The labeling efficiency was evaluated and calculated using the identification heatmap as well as the average TMT labeling status (weighted average) under each condition (example in Figure 1, data analysis details in Supplementary Data Analysis (Labeling efficiency evaluation) section in Supplementary document).

Figure 1. Optimization of TMT-to-protein mass ratio.

(A) Quantities and concentrations of TMT reagent (blue) and intact E. coli protein lysate (grey). The reaction volume and protein quantity were kept constant at 55.5 μL and 45 μg for all 4 conditions. The mass of the TMT reagent was increased from 45 to 360 μg in a constant volume of ACN yielding an increasing TMT-to-protein mass ratio from 1:1 to 8:1. (B) MS of 2 identified proteins labeled using the different TMT-to-protein ratios: Phosphocarrier protein HPr, P0AA04 and Glutaredoxin 3, P0AC63 (asterisk (*) represents coeluted species). (C) Identification heatmap demonstrating all labeled species observed under each TMT-to-protein ratio condition. Y-axis shows the number of “protein groups” ranked from high to low MW of the completely labeled proteoform. The labeling status is shown for each intact proteoform as x-axis with under-/over-labeled status indicated by the brackets. Grey scale (0–1) represents the relative intensity of each proteoform. (D) Violin plots of average labeling status to show overall TMT labeling efficiency.

RESULTS AND DISCUSSION

Removal of High Molecular Weight Proteoforms

We previously found that removal of large molecular weight proteins (>100 kDa) decreased sample precipitation and increased labeling efficiency for protein-level TMT labeling,¹⁴ so we evaluated and adapted this sample preparation process for the removal of large proteins prior to TMT labeling (e.g., molecular weight cutoff (MWCO) filtration step to enrich proteins <100 kDa). We examined the reproducibility of protein-level TMT labeling of complex samples using E. coli cell lysate. We labeled these samples using TMTzero reagent as described before with minor modifications.¹⁴ In short, we reduced, alkylated, and buffer-exchanged the lysate and labeled with TMTzero reagent with a TMT-to-protein mass ratio of 1:4 at room temperature for 1 hour. The labeling reaction was quenched with 5% hydroxylamine to a final concentration of 0.6% and incubated for 15 minutes. Lysate was labeled in duplicate to evaluate the labeling reproducibility. No significant precipitation was observed during the labeling process, indicating that the 100 kDa molecular weight cutoff approach can efficiently remove large proteins for protein-level TMT labeling. In total, 352 unique proteoforms were identified from these two samples (Supplementary Table S3). Base peak chromatogram (BPC) shown in Figure S1A demonstrated high reproducibility between the two replicate samples. An identification heatmap (Figure S1B) was utilized to compare the overall labeling efficiency between two replicates. Each row in heatmap represents a “protein group” which contains all TMT labeled products from one proteoform, including completed labeled, underlabeled, and overlabeled products. The greyscale indicates the relative intensity of each labeled product. The heatmap demonstrated highly similarity of labeling efficiency between two replicates.

Overall, we found that the protein-level TMT labeling is reproducible and 100 kDa MWCO filtration removes large proteins to avoid protein precipitation during the TMT labeling reaction. Using this MWCO filtering method to prepare intact protein samples for TMT labeling, we systematically evaluated protein-level TMT labeling reaction parameters, including TMT-to-protein ratios, pH/concentration of the quenching buffer, initial protein concentration, reaction time, and reaction buffer to minimize the under-/over- labeling commonly observed in protein-level TMT labeling.^10,13

Optimization of TMT-to-protein mass ratios in the labeling reaction

The TMT-to-protein mass ratio recommended by the manufacturer is 8:1, which is an approximately 20-fold molar excess of the labeling reagent.⁸ However, a recently published bottom-up study determined that a TMT-to-peptide mass ratio of 1:1 demonstrated a high labeling efficiency with high reproducibility. Here, to minimize under-labeling and over-labeling products in the protein-level TMT labeling process, TMT-to-protein mass ratio (1:1, 2:1, 4:1 to 8:1) was evaluated (Figure 1A). Briefly, 45 μg of E. coli proteins were labeled with TMTsixplex reagent at different ratios. The protein reaction concentration after mixing was 0.8 μg/μL (e.g., the final TMT concentration varied from 2.3 mM to 18.2 mM).

In total, 747 unique proteoforms were identified from these 4 samples using top-down LC-MS/MS analysis (Supplementary Table S4). The labeling efficiency for identified proteoforms in each sample is demonstrated in an identification heatmap (Figure 1C). We found that 31.00% (31 out of 100), 34.76% (65 out of 187), 52.69% (98 out of 186), and 67.39% (124 out of 184) of the protein groups from the samples labeled using 1:1, 2:1, 4:1 and 8:1 TMT-to-protein ratios, respectively, were found to only include the completely labeled products; improperly labeled side products were not detected. 67.00%, 64.71%, 36.02% and 12.50% of protein groups, respectively, were found to include underlabeled products; 4.00%, 6.42%, 23.66%, and 25.54% of protein groups were found to include overlabeled products. The percentage of protein groups that contained a completely labeled product with >80% intensity (e.g., completely labeled product was the dominant protein peak) was found to be 35.00%, 36.36%, 59.68%, and 80.98%, respectively.

We also evaluated the average labeling status using a violin plot (Figure 1D). The mean of the weighted averages for the 1:1, 2:1, 4:1 and 8:1 TMT-to-protein labeled samples were −1.19 ± 1.22, −0.56 ± 0.53, −0.09 ± 0.36 and 0.07 ± 0.27. This suggests that as TMT-to-protein mass ratio increases, the amount of completely labeled protein groups increases as demonstrated by the mean value approaching 0. Additionally, at labeling ratios of 4:1 and 8:1, the dominant distributions are localized around 0, indicating a majority of protein groups primarily contain completely labeled proteins. However, in the sample labeled at 1:1 and 2:1 ratios, a large portion of protein groups were underlabeled. Also, in the samples labeled with a 1:1 TMT-to-protein ratio, multiple underlabeled products were observed. In the sample labeled with a 2:1 ratio, the dominant product is underlabeled by 1 TMT tag (e.g., a dominant distribution at −1 on the violin plot).

Two examples of identified proteoforms are shown in Figure 1B, Phosphocarrier protein HPr (Uniprot: P0AA04) and Glutaredoxin 3 (Uniprot: P0AC62), which showed multiple labeled products in the MS spectra. For Phosphocarrier protein HPr protein, the dominant peak that represented the completely labeled proteoform was observed at 8:1 and 4:1 TMT-to-protein labeling ratios. However, in the sample with a 4:1 ratio, a small peak that represented the underlabeled product (underlabeled by 1 tag) was also observed. In the samples labeled with 1:1 and 2:1 labeling ratios, there were significant underlabeled products detected. Interestingly, the sample with a 1:1 ratio showed a normal distribution of underlabeled peaks with multiple peaks representing underlabeled products. However, the sample with a 2:1 ratio only contains one major underlabeled product (underlabeled by 1 tag) which is consistent with the violin plot (Figure 1D). The MS/MS fragmentation confirmed that the missing label was located on the N-terminus (Figure S2A–C), which is consistent with our previously reported results that the decreased reactivity of the N-terminus compared with lysine (pKa of N-terminal is 7.7 ± 0.5 and lysine is 10.5 ± 0.1) causes underlabeling to occur more frequently on the N-terminus.^14,25 Similar results were observed for P0AC62 with the exception that the sample labeling with a 2:1 TMT-to-protein ratio had no observable underlabeling. Overall, analysis of MS/MS spectra for identified proteoforms confirmed that the missing labels in the 2:1 TMT-to-protein ratio sample was primarily localized to the N-terminus (Figure S2D–E).¹⁴

Overall, higher TMT-to-protein ratio (w/w: 8:1, 4:1) reaction conditions resulted in higher ratios of completely labeled products compared with underlabeled products. However, some protein groups labeled using 8:1 and 4:1 TMT-to-protein ratios showed overlabeled products, which suggested that overlabeling also needs to be minimized by optimization.

Optimization of quenching buffers in the labeling reaction

For bottom-up TMT labeling, 5% hydroxylamine (0.3% final concentration) was recommended as the quenching buffer by the manufacturer. However, Tris buffer (50 mM at pH 8.0) has also been utilized as the quenching buffer in previous literature during peptide-level TMT labeling.⁸ In this study, we evaluated different quenching buffers for intact protein-level TMT labeling (e.g., hydroxylamine buffer with a final concentration of 1.2%, 0.6%, or 0.3%, and Tris buffer with a final concentration of 50 mM Figure 2A). We also measured final pH values of the reaction solution after quenching: 9.1 (0.3% hydroxylamine), 9.3 (0.6% hydroxylamine), 9.5 (1.2% hydroxylamine), and 8.0 (50 mM Tris).

Figure 2. Optimization of quenching buffer in protein-level TMT labeling.

(A) Quantities and concentrations of TMT reagent (blue) and intact E. coli protein (grey). The reaction parameters were kept the same for all four experiments but the final concentration of the quenching buffer was varied: 1.2%, 0.6%, and 0.3% hydroxylamine or 50 mM Tris which resulted in final pH of 9.5, 9.3, 9.1, and 8.0, respectively. (B) MS of 2 identified proteins under different quenching conditions: Phosphocarrier protein HPr, P0AA04 and Glutaredoxin 3, P0AC63 (asterisk (*) represents coeluted species). (C) Identification heatmap of quenching buffer optimization group. Y-axis shows the number of “protein groups” ranked from high to low MW of the completely labeled proteoform. The labeling status is shown for each intact proteoform as x-axis with under-/over-labeled status indicated by the brackets. Grey scale (0–1) represents the relative intensity of each proteoform. (D) Violin plots demonstrating the average labeling status to show overall TMT labeling efficiency.

A total of 610 unique proteoforms were identified from 4 samples with different quenching buffers using LC-MS/MS analysis (Supplementary Table S5). An identification heatmap was generated to evaluate the labeling efficiency for each identified protein (Figure 2C). In the solution quenched by Tris buffer (pH 8.0), many overlabeled products were observed with relatively high intensity. With the increased pH/final concentration of hydroxylamine, the relative intensity of overlabeled products decreased and more completely labeled product were detected. The percentage of protein groups that only included the completely labeled product (no improperly labeled products) were 89.24% (141 out of 158), 77.86% (109 out of 140), 80.16% (101 out of 126), and 60.13% (92 out of 153) for the samples quenched at pH 9.5, pH 9.3, pH 9.1 and pH 8.0, respectively. 3.16%, 15.00%, 15.08%, and 35.29% of protein groups, respectively, included overlabeled products, which suggested that higher pH after quenching led to a decrease in overlabeled products. Additionally, 7.59%, 7.86%, 4.76%, and 8.50% of the protein groups were found to include underlabeled products. The percentage of protein groups that contained a completely labeled product with >80% intensity (e.g., completely labeled product was the dominant peak) was found to be 92.41%, 85.00%, 84.92%, and 67.97%, respectively.

Mean values of the weighted averages for the four samples were calculated to be −0.03 ± 0.20, 0.02 ± 0.30, 0.06 ± 0.33 and 0.26 ± 0.54 for pH 9.5, pH 9.3, pH 9.1, and pH 8.0, respectively (Figure 2D), which indicated that quenching at pH > 9.1 decreases overlabeling.

As shown in Figure 2B, for Phosphocarrier protein HPr (P0AA04), only a very small portion of the overlabeled product was observed in the reaction quenched at high pH (hydroxylamine 1.2%, pH 9.5). Generally, as the quenching pH increased, overlabeling decreased. In the solution quenched with Tris buffer (pH=8.0), the dominant peak represented the product overlabeled by one TMT label. Similar results were observed for Glutaredoxin 3 (P0AC62). MS/MS fragmentation (Figure S3A) for Phosphocarrier protein HPr protein localized the overlabeling site at T₃₀; Figure S3B for Glutaredoxin 3 indicated the overlabeling site was within a range with several STY residues, which is consistent with previous reported results that TMT overlabeling tends to occur on STY amino acids.^8,10

The side chains of lysine and tyrosine have similar pKa values (Lys: 10.5 ± 1.1; Tyr: 10.3 ± 1.2);²⁵ therefore, it is expected that tyrosine would also be labeled by TMT at basic pH. It has also been reported that O-acylation can occur on serine and threonine because of the nucleophilicity of the hydroxyl oxygen due to the hydrogen bonding of hydroxyl hydrogen to proton acceptors such as histidyl imidazole groups.^26,27 Hydroxyl-containing amino acids also show reactivity towards NHS esters when separated by one or two residues from a histidyl residue to form an H-X-[STY] or [STY]-X-H motif.^8,27 O-acylation was found to hydrolyze within a few minutes at basic pH values,^8,28,29 which explains why hydroxylamine quenching buffer can be used to reverse the overlabeling by increasing the pH to ≥ 9.0.^26,28,28,30

Overall, our results demonstrate that a hydroxylamine quenching buffer (final concentration of 0.3% or higher) resulting in >pH 9.1 after quenching can reverse unwanted O-acylation reactions and reduce the production of overlabeled products.

Optimization of the initial protein sample concentration in the labeling reaction

In the TMT and protein labeling reaction, higher molarity of reactants (TMT and protein) increases the rate of labeling and reduces the hydrolysis of the TMT reagent which may reduce the production of underlabeled species.⁸ Hence, to fully understand the effect of reactant concentration on intact protein-level TMT labeling, the initial concentration of protein was varied (2 μg/μL, 1.5 μg/μL, 1.0 μg/μL, and 0.5 μg/μL) while maintaining the same TMT-to-protein mass ratio of 4:1, Figure 3A. Specifically, 200 μg TMT in 10 μL of ACN solvent was reacted with 50 μg protein in different buffer volumes from 25 μL to 100 μL, which varied the final reactant concentrations of TMT and proteins from 5.4 to 16.9 mM and 0.5 to 1.4 μg/μL, respectively.

Figure 3. Optimization of initial protein concentration in the labeling reaction.

(A) Quantities and concentrations of TMT reagent (blue) and intact E. coli protein (grey). The reaction TMT and protein quantity were kept constant at 200 μg and 50 μg for all 4 conditions and the total reaction volume was varied from 35 μL to 110 μL by changing initial protein concentration from 2.0 μg/μL to 0.5 μg/μL. (B) MS of 2 identified protein examples under different reaction conditions: Phosphocarrier protein HPr, P0AA04 and Glutaredoxin 3, P0AC63 (asterisk (*) represents coeluted species). (C) Identification heatmap of all four samples under initial protein concentration optimization group. Y-axis shows the number of “protein groups” ranked from high to low MW of the completely labeled proteoform. The labeling status is shown for each intact proteoform as x-axis with under-/over-labeled status indicated by the brackets. Grey scale (0–1) represents the relative intensity of each proteoform. (D) Violin plots demonstrating the average labeling status to show overall TMT labeling efficiency.

LC-MS/MS analysis identified 559 unique proteoforms among 4 samples (Supplementary Table S6). In the identification heatmap (Figure 3C), we found that 61.22% (90 out of 147), 75.76% (125 out of 165), 79.61% (121 out of 152), and 75.89% (107 out of 141) of the protein groups for samples labeled with protein concentration from 0.5 μg/μL to 2 μg/μL, respectively, which suggested underlabeled products could be reduced by increasing protein initial concentration. The percentage of protein groups that contained underlabeled products were 37.41%, 20.00%, 8.55% and 10.64% as protein concentration increased; the percentage of protein groups that include overlabeled products decreased slightly as concentration increased (3.40%, 5.45%, 13.82% and 15.60%). In addition, the means of weighted averages (Figure 3D) from were calculated to be −0.16 ± 0.38, −0.08 ± 0.27, 0.02 ± 0.25 and 0.03 ± 0.28 from 0.5 μg/μL to 2 μg/μL. This trend further demonstrates that under labeling was reduced as reactant concentration increased. The percentage of protein groups that contained a completely labeled product peak with >80% intensity (e.g., completely labeled product was dominant) was found to be 69.39%, 86.06%, 88.16%, and 83.69% for samples with protein concentrations from 0.5 μg/μL to 2 μg/μL, respectively.

To further demonstrate the influence of reactant concentration on labeling efficiency, MS data for two identified proteins are shown in Figure 3B. The MS spectra of Phosphocarrier protein HPr (P0AA04) showed that the dominant peak that represented the completely labeled product was observed in all samples. However, the sample with an initial protein concentration of 0.5 μg/μL showed 2 underlabeled peaks with similar intensities as the completely labeled peak (Figure 3B). Another example, Glutaredoxin 3 (P0AC62), also showed the completely labeled product peak as the highest peak in all samples; however, a small peak that represents the underlabeled product was detected in the samples with initial protein concentration of 1.0 μg/μL and 0.5μg/μL (Figure 3B). Fragmentation and MS/MS spectra for the two examples in Figure S4 suggested that proteoforms that were under labeled by one TMT tag were missing labels on the N-terminal which is consistent with previous observations.¹⁴

Overall, our results indicated that initial protein concentration >1.0 μg/μL with a TMT-to-protein ratio of 4:1 (wt/wt) or higher was best for intact protein TMT labeling. These conditions minimized underlabeling and resulted in high labeling efficiency to produce primarily completely labeled species.

Investigation of TMT labeling reaction conditions influence in labeling efficiency

Here, we evaluated other reaction parameters that might affect the efficiency of protein-level TMT labeling, including (1) reaction time (60 minutes and 30 minutes); (2) pH 8.5 reaction buffer (100 mM TEAB and 100 mM HEPES) (Figure 4A). LC-MS/MS analysis identified 367 proteoforms among 3 samples (Supplementary Table S7).

Figure 4. Investigation of TMT labeling reaction conditions influence in labeling efficiency.

(A) Quantities and concentrations of TMT reagent (blue) and intact E. coli protein (grey). The reaction TMT and protein quantity were kept constant at 200 μg and 50 μg for all 4 conditions. Reaction buffer and time are indicated in the figure. (B) MS of 2 identified protein examples with labeling status at different reaction conditions: Phosphocarrier protein HPr, P0AA04 and Glutaredoxin 3, P0AC63 (asterisk (*) represents coeluted species). (C) Identification heatmap for reaction optimization group. Y-axis shows the number of “protein groups” ranked from high to low MW of the completely labeled proteoform. The labeling status is shown for each intact proteoform as x-axis with under-/over-labeled status indicated by the brackets. Grey scale (0–1) represents the relative intensity of each proteoform. (D) Violin plots of average labeling status to show overall TMT labeling efficiency.

Reaction time evaluation:

The manufacturer protocol recommended a 60-min reaction time for TMT labeling. To evaluate the effect of reduced labeling time on labeling efficiency, we changed the labeling time to 30 minutes and kept the other parameters the same (Supplementary Table S1). Two protein examples: Phosphocarrier protein HPr (P0AA04) and Glutaredoxin 3 (P0AC63), Figure 4B, demonstrated similar labeling for the samples incubated for 30 and 60 minutes with the exception that the sample incubated for 60 minutes generated a higher intensity of over labeled peak in P0AA04. In the identification heatmaps (Figure 4C), 77.08% (148 out of 192) of protein groups from the 60-min incubation sample and 75.47% (160 out of 212) of protein groups from the 30-min incubation sample demonstrated only the completely labeled products. 11.46% and 12.26% of protein groups, respectively, were found to include under labeled products and 12.50% and 15.09% of protein groups were found to include over labeled products. The percentage of protein groups that contained a completely labeled product with >80% intensity (e.g., completely labeled product was dominant) was 80.21% and 80.19% for 60-min and 30-min incubations, respectively. There is no significant difference between these two samples with respect to labeling efficiency. Additionally, the weighted averages shown in the violin plot (Figure 4D) indicate similar overall TMT labeling with mean values of 0.02 ± 0.36 for 60-min incubation and 0.05 ± 0.39 for 30-min incubation. These results indicate that a 30-minute incubation time is sufficient for high protein-level TMT labeling efficiency.

Reaction buffer evaluation:

To determine if different reaction buffers will affect the labeling efficiency, we tested the protein-level TMT labeling in 100 mM TEAB (pH=8.5) and 100 mM HEPES (pH=8.5), the other parameters were shown in Supplementary Table S1. The examples (Phosphocarrier protein HPr (P0AA04) and Glutaredoxin 3 (P0AC63)) shown in Figure 4B suggest a similar labeling efficiency between two samples except that the HEPES sample had a higher ratio of over labeled product peaks compared to TEAB buffer for P0AA04. According to the identification heatmap in Figure 4C, the labeling efficiency was similar between the HEPES and TEAB buffers. 77.08% (148 out of 192) and 78.95% (165 out of 209) of the protein groups in TEAB and HEPES, respectively, were found to only produce the completely labeled products. 11.46% and 9.09% of protein groups were found to contain under labeled products; 12.50% and 12.44% of protein groups included over labeled products. The percentage of protein groups that contained a completely labeled product with >80% intensity (e.g., completely labeled product was dominant in majority of protein groups) was 80.21% and 82.78% for TEAB and HEPES, respectively. The weighted average distribution (Figure 4D) for the two samples was also quite similar with mean values of 0.02 ± 0.36 and 0.03 ± 0.33, respectively, which was consistent with the heatmap, suggesting similar labeling efficiency between labeling buffers.

In conclusion, we evaluated the reaction conditions for protein-level TMT labeling and concluded that the reaction buffer and time did not significantly affect labeling efficiency. The result indicated that the labeling reaction time can be reduced and HEPES may be used as alternative reaction buffer.

Double labeling: to achieve complete labeling on low-concentration samples

Our results indicate that high protein and TMT reagent concentrations are optimum for intact protein labeling. However, some samples (e.g., clinical samples) may be limited and may not be sufficient to produce high concentration protein solutions (e.g., >1.0 μg/μL). For efficient labeling of low concentration protein samples, we designed and optimized a double labeling technique using various initial protein concentrations (2 μg/μL, 1.5 μg/μL, 1.0 μg/μL, and 0.5 μg/μL), Figure 5A. Briefly, TMT labeling was conducted twice before quenching; 200 ug of TMT in 10 μL ACN was incubated with the protein solutions in 100 mM TEAB buffer (pH 8.5) for 1 hour followed by a second addition of TMT reagent and incubation for an additional hour. TMT-to-protein ratio remained constant and the final reactant concentrations were varied from 10.8 to 33.8 mM (TMT) and 0.5 to 1.4 μg/μL (protein).

Figure 5. Double labeling: to achieve complete labeling on low-concentration samples.

(A) Quantities and concentrations of TMT reagent (blue) and intact E. coli protein (grey). The reaction TMT and protein quantity were kept constant at 200 μg and 50 μg for all 4 experiments but the final reaction volume after double labeling was varied from 45 μL to 120 μL by changing initial protein concentration from 2.0 μg/μL to 0.5 μg/μL. (B) MS of 2 identified protein examples under different reaction conditions: Phosphocarrier protein HPr, P0AA04 and Glutaredoxin 3, P0AC63 (asterisk (*) represents coeluted species). (C) Identification heatmap of all four samples under initial protein concentration optimization group. Y-axis shows the number of “protein groups” ranked from high to low MW of the completely labeled proteoform. The labeling status is shown for each intact proteoform as x-axis with under-/over-labeled status indicated by the brackets. Grey scale (0–1) represents the relative intensity of each proteoform. (D) Violin plots demonstrating the average labeling status to show overall TMT labeling efficiency.

The LC-MS/MS analysis identified 619 unique proteoforms among the 4 samples (Supplementary Table S8). The heatmap in Figure 5C suggested a similar labeling efficiency for all 4 conditions. The percentage of protein groups that only included completely labeled products was found to be 87.73% (193 out of 220), 88.54% (170 out of 192), 90.78% (187 out of 206), and 89.15% (189 out of 212) from 0.5 μg/μL to 2 μg/μL, respectively. 10.45%, 7.81%, 5.83% and 6.60% of protein groups, respectively, were found to include under labeled products and 2.27%, 4.17%, 3.88% and 4.72% of protein groups were found to include over labeled products. The percentage of protein groups that contained a completely labeled product with >80% intensity (e.g., majority of protein product was completely labeled) was found to be 93.64%, 92.71%, 93.69%, and 93.87% for initial protein concentration samples from 0.5 μg/μL to 2 μg/μL, respectively. As shown in the violin plot (Figure 5D), there is no significant difference among the mean values of the weighted averages (−0.04 ± 0.20, −0.03 ± 0.22, −0.03 ± 0.20 and −0.02 ± 0.18). Overall, we found that there was no significant difference in TMT labeling efficiency among 4 samples with different initial protein concentrations when the double labeling technique was used. Furthermore, the MS spectra of both Phosphocarrier protein HPr (P0AA04) and Glutaredoxin 3 (P0AC62) proteins showed that the completely labeled product was the only observable proteoform peak in all samples after double labeling (Figure 5B).

Our results indicated that the double labeling can be an alternative approach to achieve high labeling efficiency for limited samples with low protein concentration (i.e., ≥0.5 μg/μL initial protein concentration).

Application of optimized protein-level TMT labeling conditions on HeLa cell lysate

We applied the optimized conditions to the TMT labeling of HeLa cell lysate to evaluate the TMT labeling efficiency and quantification accuracy. Briefly, HeLa proteins (1.5 μg/μL) were aliquoted into 6 tubes and labeled with TMTsixplex under the optimized conditions: TMT-to-protein mass ratio of 8:1 (double labeling) and hydroxylamine quench (final concentration of 1.2%, pH 9.5). Then, the six TMT-labeled HeLa protein samples were mixed to a mass ratio of 5:5:2:2:1:1 and analyzed using LC-MS/MS.

179 proteoforms were identified from the LC-MS/MS analysis (Supplementary Table S9). MS spectra of four protein examples are given in Figure 6A. All four proteins were found to only show the completely labeled proteoforms (indicated by *) and the dashed lines labeled with −1 and +1 represent the theoretical location of proteoforms under labeled and over labeled by one TMT tag, respectively. An identification heatmap with 133 clustered protein groups (Figure 6B) was generated to illustrate the overall labeling efficiency. 86.47% (115 out of 133) protein groups were found to only include completely labeled products without any improperly labeled side products, 10.53% protein groups contained under labeled products and 6.77% protein groups included over labeled products. 91.73% of protein groups that contained a completely labeled product with >80% intensity (e.g., completely labeled product was dominant in majority of protein groups). The mean value of the weighted average was −0.05 ± 0.21, which confirmed that most of the protein groups showed primarily completely labeled products. In Figure 6C, the Jupiter microtubule associated homolog 1 protein (Q9UK76) was found to include only the completely labeled proteoform. Additionally, the intensities of all reporter ions were extracted and the ratios were calculated as previously described.¹⁴ The average reporter ion ratios were 5.20 ± 1.50, 5.00 ± 1.37, 1.97 ± 0.76, 1.98 ± 0.67, 1.05 ± 0.34 (theoretical ratios as 5:5:2:2:1), which represents a sizeable improvement compared with our previously published data before reaction optimization.¹⁴

Figure 6. Application of optimized protein-level TMT labeling conditions on HeLa cell lysate.

(A) MS of 4 identified proteoforms. The completely labeled proteoform is indicated by an asterisk and the −1 and +1 represent the theorical location of proteoform peaks under labeled and over labeled by one TMT tag. (B) Identification heatmap of TMT-labeled HeLa proteoforms. Y-axis shows the number of “protein groups” ranked from high to low MW of the completely labeled proteoform. The labeling status is shown for each intact proteoform as x-axis with under-/over-labeled status indicated by the brackets. Grey scale (0–1) represents the relative intensity of each proteoform. (C) MS/MS spectra of the completely labeled Jupiter microtubule associated homolog 1 protein. (D) Normalized intensity ratios (126/131, 127/131, 128/131, 129/131, and 130/131) of TMT-labeled HeLa proteins. The error bars represent the the standard deviation and the red dashed lines represent the theoretical ratio (5:5:2:2:1).

Overall, the intact protein-level TMT labeling under these optimized conditions not only increased the labeling efficiency, but also improved the quantification accuracy.

CONCLUSION

We have determined that the optimized conditions for protein-level TMT labeling in complex samples, such as E. coli cell lysate and HeLa cell lysate. A protein solution with initial concentration > 1.0 μg/μL labeled with TMT as 4:1 or higher TMT-to-protein mass ratio, then quenched by hydroxylamine to a final concentration as > 0.3% (final pH > 9.1) resulted in optimal labeling efficiency and minimized production of over/underlabeled side products). Double labeling can also be applied for low concentration samples for a better labeling efficiency. With optimal protein-level TMT labeling conditions, >90% labeling efficiency (percentage of protein groups that only contained completely labeled product) was achieved. When the optimal conditions were applied to HeLa cell lysate, 86.47% labeling efficiency was achieved and 91.73% of protein groups contained a completely labeled product with >80% intensity (e.g., completely labeled product was dominant). These optimized conditions can potentially be adapted to the other isobaric chemical tag labeling approaches (e.g., iTRAQ, DiLeu).

In previous work, optimization of peptide-level TMT labeling suggested that a 1:1 TMT-to-peptide mass ratio was sufficient for labeling;⁸ however, our results indicated a higher TMT-to-protein mass ratio (4:1 or 8:1) is required for efficient top-down TMT labeling. The higher TMT-to-protein mass ratio required for sufficient labeling may be due to structural hindrance caused by the secondary structure that may be present in intact proteins even after urea denaturation, reduction, and alkylation.³¹ Higher organic concentration and/or higher labeling temperature may improve intact protein denaturation by reducing the remaining secondary structure, which can potentially enhance labeling efficiency at a lower TMT-to-protein mass ratio.³² This may be determined by further optimization of organic content and temperature during TMT labeling.

We also demonstrated that the enrichment of low molecular weight proteins using 100 kDa MWCO only was sufficient for protein-level TMT labeling of complex samples to limit precipitation under the labeling conditions. Simplified sample preparation reduced the sample processing time and may prevent sampling bias. Other molecular weight strategies such as Gel-Eluted Liquid Fraction Entrapment Electrophoresis (GELFrEE)³³ can also be utilized in the future for enhanced enrichment of low molecular weight proteoforms. The presented protocol was optimized for low molecular weight proteoforms quantification using protein-level TMT labeling, which may not be readily applicable to larger proteins such as IgG (~150 kDa). In the future, MS compatible detergent such as 4-hexylphenylazosulfonate (Azo)^1,34, will be evaluated to optimize TMT labeling for high molecular weight proteins.

In addition, we will optimize the fragmentation parameters in the future to increase the number of proteoform identifications and increase proteome coverage. Also, fractionation methods such as high-pH and low-pH reversed phase chromatography^17,35,36 can also be applied to improve intact proteoform identification and characterization.

Supplementary Material

Supplementary Material

Supplementary Data Analysis (Labeling efficiency evaluation)

Supplementary Table S1. Experimental design of protein-level TMT labeling optimization

Supplementary Table S2. Post-translational modifications (PTMs) in TopPIC Suite searching

Supplementary Figure S1. Simple sample preparation and reproducibility evaluation.

Supplementary Figure S2. Fragmentation of examples from optimization of TMT-to-protein mass ratio group.

Supplementary Figure S4. Fragmentation examples of protein-level TMT labeling quenching buffer optimization group.

Supplementary Figure S4. Fragmentation examples of initial protein concentration with single labeling optimization group.

Supplementary Tables

Supplementary Table S3. Supplementary Table S3 TopPIC Suite Searching results for group 0: reproducibility test

Supplementary Table S4. Supplementary Table S4 TopPIC Suite Searching results for optimization group 1: TMT-to-protein mass ratio

Supplementary Table S5. Supplementary Table S5 TopPIC Suite Searching results for optimization group 2: Quenching buffer

Supplementary Table S6. Supplementary Table S6 TopPIC Suite Searching results for optimization group 3 protein initial concentration with single labeling

Supplementary Table S7. Supplementary Table S8 TopPIC Suite Searching results for optimization group 5: Protein-level TMT reaction conditions

Supplementary Table S8. Supplementary Table S7 TopPIC Suite Searching results for optimization group 4 protein initial concentration with double labeling

Supplementary Table S9. Supplementary Table S9 TopPIC Suite Searching results for group 5: Hela cell lysate with optimized conditions