An Isobaric Protein-Level Labeling Strategy for Serum Glycoprotein Quantification Analysis by LC-MS/MS

Abstract

While peptide-level labeling using isobaric tag reagents has been widely applied for quantitative proteomics experiments, there are comparatively few reports of protein-level labeling. Intact protein labeling could be broadly applied to quantification experiments utilizing protein-level separations or enrichment schemes. Here, protein-level isobaric labeling was explored as an alternative strategy to peptide-level labeling for serum glycoprotein quantification. Labeling and digestion conditions were optimized by comparing different organic solvents and enzymes. Digestions with Asp-N and trypsin were found highly complementary; combining the results enabled quantification of 30% more proteins than either enzyme alone. Three commercial reagents were compared for protein-level labeling. Protein identification rates were highest with iTRAQ 4-plex when compared to TMT 6-plex and iTRAQ 8-plex using higher-energy collisional dissociation on an Orbitrap Elite mass spectrometer. The compatibility of isobaric protein-level labeling with lectin-based glycoprotein enrichment was also investigated. More than 74% of lectin-bound labeled proteins were known glycoproteins, which was similar to results from unlabeled and peptide-level labeled serum samples. Finally, protein-level and peptide-level labeling strategies were compared for serum glycoprotein quantification. Isobaric protein-level labeling gave comparable identification levels and quantitative precision to peptide-level labeling.

Introduction

Chemical labeling with isobaric tags is a common strategy in quantitative proteomics, since it is applicable to all kinds of samples (e.g., plasma and tissue) and can be multiplexed for up to eight samples.^1-3 While iTRAQ ⁴ and TMT ⁵ reagents have been extensively used, reports have almost exclusively applied isobaric labeling at peptide-level.^6-8 Preference for peptide-level labeling over protein-level labeling is the result of many factors. Organic solvents are required to solubilize the isobaric reagents and prevent hydrolysis of the amine reactive group. However, solvents often cause protein precipitation, which inhibits efficient labeling and digestion. Peptides generally have at least one primary amine, either at their N-termini or on a lysine residue, which allows almost all peptides to provide quantitative information. With protein-level labeling, tryptic cleavage at modified lysine residues is effectively blocked, leading to longer peptides that are difficult to sequence by MS/MS and sequencing of non-lysine containing peptides that cannot generate quantitative information.

Even with these disadvantages, protein-level labeling has potential benefits over peptide-level labeling. Protein-level labeling enables intact protein separation methods, such as SDS-PAGE, as the initial fractionation dimension. Another advantage, especially for quantitative applications, is that samples are mixed early in the sample preparation process. This eliminates variability from downstream preparation steps, such as protein-level fractionation and peptide digestion, as sources of quantification error. Protein-level labeling also leverages the multiplexing advantage of isobaric reagents by enabling concomitant sample preparation to minimize replication of time-consuming steps, which is beneficial for large sample cohorts. In 2007, Wiese et al. applied iTRAQ for intact protein labeling for the first time. ⁹ Standard proteins were tested and protein molecular weight was used to determine labeling efficiency using a MALDI-TOF. To address the issue of reduced trypsin cleavage, Engmann et al. used Asp-N to digest protein-level TMT labeled tissue samples to achieve similar quantitative proteomic coverage to trypsin in terms of peptides and proteins identified. ¹⁰ Chymotrypsin and Glu-C were evaluated in another study looking at TMT protein-level labeled serum samples.⁸ However, these enzymes have not been compared systematically.

We report an optimized protein-level isobaric tag labeling strategy for serum compatible with protein-level fractionation. Lectin affinity enrichment of glycoproteins for serum analysis was selected because glycosylation is a common protein post-translational modification¹¹ that plays important roles in adhesion, metastasis, signaling via cell-to-cell interactions, and disease progression. ^12-14 Optimal sample preparation conditions were determined by comparing organic solvents, followed by considering different enzymes as alternatives to trypsin. Three commercial isobaric tags (iTRAQ 4-plex, TMT 6-plex, and iTRAQ 8-plex) were tested to determine any influence on peptide identification resulting from the different tag structures. Lastly, isobaric protein-level and peptide-level labeling strategies were compared for serum glycoprotein quantification. It was demonstrated that isobaric protein-level labeling is a robust method that can be applied to glycoprotein quantification in serum that gives comparable performance to peptide-level labeling.

Materials and Methods

2.1 Materials

Reagents were purchased from Sigma-Aldrich (St. Louis, MO) unless otherwise specified. Tris (2-carboxyethyl) phosphine hydrochloride solution (TCEP, 0.5 M) and TMT 6-plex reagent was obtained from Thermo Scientific (Rockford, IL). iTRAQ 4-plex and 8-plex were purchased from AB Sciex (Foster City, CA). Sequencing grade trypsin, chymotrypsin, Glu-C, and Asp-N were purchased from Promega (Madison, WI). 3 kDa MWCO centrifugal filters and C₁₈ ZipTips were from Millipore (Billerica, MA). Agarose-bound Aleuria aurantia lectin (AAL) was purchased from Vector Laboratories (Burlingame, CA).

2.2 Serum depletion

Serum pooled from 30 healthy people was obtained from Bioreclamation LLC (Westbury, NY). Fourteen high-abundance proteins were depleted with an IgY-14 LC10 column (Sigma, St. Louis, MO). The depletion was performed with 250 μL serum according to the manufacturer’s instructions. The serum sample was diluted with 1× depletion buffer and loaded onto the IgY14 LC10 column. The flow-through fraction between 0 to 30 minutes was transferred into a 15 mL YM-3 centrifugal device (Millipore, Billerica, MA) and centrifuged at 4000 g, followed by buffer exchange three times with 5 mL deionized water. The final sample volume was 300 μL. The final protein concentration was measured using a Bradford assay kit (BioRad).

2.3 Isobaric labeling at the protein-level

A detailed protocol is provided in the Supporting Information. In brief, depleted serum was adjusted to at least 4 M urea using 8 M urea. Samples were reduced with 5 mM TCEP for 30 min at 37 °C and alkylated with 25 mM iodoacetamide for 1 h in the dark. The buffer was exchanged to 50 mM TEAB in 4 M urea with a final volume of 50 μL. Isobaric labeling reagents were dissolved in 60 μL DMSO, transferred to sample tubes, reacted for 2 h at room temperature, and incubated for 15 min with hydroxylamine (final concentration of 0.5%) to quench. Samples were combined, transferred to a YM-3 centrifugal filter, and exchanged to either 50 mM NH₄HCO₃ for digestion or lectin binding buffer (see below) for glycoprotein enrichment.

2.4 Glycoprotein enrichment

Glycoprotein enrichment was performed as described previously¹⁵ with some modifications. A column packed with 600 μL of agarose-bound AAL was washed with 3 mL binding buffer (20 mM Tris, 0.15 M NaCl, pH=7.5, protease inhibitor 1:100). Samples in 1 mL of binding buffer were loaded onto the column and incubated for 15 min twice. Five column volumes of binding buffer were used to wash away unbound proteins. Bound glycoproteins were eluted with four volumes of elution buffer (200 mM fucose in binding buffer). The elution buffer was exchanged using a YM-3 filter to 50 mM NH₄HCO₃ for digestion or 50 mM triethylammonium bicarbonate (TEAB) buffer for peptide-level labeling.

2.5 Enzymatic digestion

Glycoproteins were digested with trypsin, chymotrypsin, or Asp-N (1:30, enzyme to protein) at 37 °C overnight. For double digestion using trypsin and Glu-C, lyophilized tryptic digests were re-suspended in 50 mM NH₄HCO₃ solution and further digested by adding Glu-C (1:20, enzyme to protein) at 37 °C overnight. Glycopeptides were deglycosylated using PNGase F (New England Biolabs, Ipswich, MA) at 37 °C for 16 h and dried using a SpeedVac concentrator (Thermo Savant, Milford, MA). The samples were desalted using C₁₈ ZipTips before LC-MS/MS analysis.

2.6 Isobaric labeling at the peptide-level

Serum samples were subjected to lectin affinity enrichment. Eluted glycoproteins were reduced with 5 mM TCEP for 30 min at 37 °C, alkylated with 15 mM iodoacetamide in the dark at room temperature for 30 min, digested with trypsin (1:30, enzyme to protein) overnight, and dried down in a SpeedVac. Peptide-level isobaric labeling followed the manufacturers’ protocols. Digests were re-dissolved in 30 μL 0.5 M TEAB buffer. Approximately 70 μL of ethanol or 41 μL acetonitrile were added to one tube of isobaric labeling reagent (iTRAQ or TMT, respectively) to transfer the labeling reagent to the sample tube, followed by reaction for 1 h at room temperature. Two microliters of 5% hydroxylamine were added to quench the reaction for 15 min. Samples were combined, if required, and lyophilized using a SpeedVac concentrator.

2.7 LC-MS/MS analysis

Peptide mixtures were dissolved in 0.1% formic acid (FA) and loaded onto a Proxeon EASY nLC II System (Thermo) equipped with an in-house packed 25 cm column (75 μm i.d., Magic C₁₈AQ, 5 μm) coupled to an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific). Peptides were separated with 0.1% FA with 2% acetonitrile in water and 0.1% FA with 2% water in acetonitrile using a 90 min linear gradient from 2 to 32% solvent B at a flow rate of 400 nL/min. A Top 10 HCD method was used for MS/MS analysis; complete instrument settings are provided in Supporting Information.

2.8 Data analysis

Acquired MS/MS spectra were searched against a forward-reverse database generated from the UniProt human database (released Nov. 2010) using SEQUEST in Proteome Discoverer 1.1 (Thermo). Searches were performed using the following settings: precursor ion m/z tolerance: ± 10 ppm; fragment ion m/z tolerance: ± 0.03 Da; two missed cleavages allowed; static modification: carbamidomethylation (C); dynamic modifications: oxidation (M) and deamidation (N). For protein-level labeling, isobaric modification (iTRAQ 4-plex, TMT 6-plex, or iTRAQ 8-plex) of lysines and protein N-termini were considered; for peptide-level labeling, modification of lysines and peptide N-termini was considered. Identified peptides were filtered using a 1% peptide-level false discovery rate (FDR) and quantification was performed using reporter ions. The labeling efficiency was calculated based on the number of identified labeled spectra compared with the total identified spectra. All identified proteins were manually checked for the “glycoprotein” annotation by searching UniProt (www.uniprot.org).

Results and Discussion

3.1 Rationale for protein-level isobaric labeling for quantitative serum glycoproteomics

Most published quantitative proteomics methods for comparison of serum glycoproteins use label-free quantitative methods, such as spectral counting^{16, 17} and the peak area method¹⁸, or peptide-level isobaric labeling ^{19, 20} (Figure 1, Strategy A). Several sample preparation steps, including glycoprotein enrichment, digestion, and labeling, are completed in parallel until mixing, with each step adding to the overall method variance and sample preparation time. To minimize the influence of inconsistency during sample preparation, a quantitative proteomics method using isobaric labeling of intact proteins was developed (Figure 1, Strategy B). Since the samples are mixed prior to glycoprotein enrichment and digestion, variability from these steps is eliminated. Furthermore, the time required for protein-level sample preparation steps (e.g., glycoprotein enrichment/buffer exchange) is reduced by up to the multiplexing factor (currently 8-plex) of the isobaric reagent used. By overcoming many challenges associated with protein-level labeling, such as incomplete reaction efficiency and limited digestion, protein-level labeling should be broadly applicable to numerous proteomics workflows.

Figure 1. Workflow for peptide-level (A) and protein-level (B) isobaric labeling for relative serum glycoprotein quantitation.

For peptide-level labeling, glycoprotein enrichment is followed by digestion and reaction with isobaric tags. For protein-level labeling, labeling occurs prior to lectin affinity chromatography and digestion.

3.2 Organic solvent optimization

Organic solvents at >50% concentration are obligatory to dissolve isobaric tag reagents and to slow the hydrolysis rate of the amine reactive N-succinimidyl ester group. Solvents also influence reaction efficiency by altering reaction kinetics and protein higher-order structure. To improve protein-level labeling efficiency, four solvents were compared. Depleted serum samples were reduced, alkylated, reacted with isobaric reagents dissolved in ethanol, acetonitrile, methanol, or DMSO, and digested. DMSO yielded the greatest number of identified peptides, when compared to the other three solvents (Figure 2). With isobaric protein-level labeling, only peptides containing labeled lysines can generate quantitative information. Similar to the identification results, DMSO gave the highest percentage of lysine-containing (“quantifiable”) peptides and proteins (75 and 90%, respectively). Notably, all four solvents resulted in >95% labeling efficiency when used in 4 M urea. Even when the protein amount was doubled to 50 μg per tube of isobaric reagent, no significant decrease in labeling efficiency was observed. From a practical perspective, controlling the urea concentration during labeling is a critical factor. If the urea concentration was lower than 1.5 M, the protein-level labeling efficiency was significantly decreased (data not shown).

Figure 2. Identified and quantified peptides using different organic solvents for protein-level isobaric labeling.

Depleted serum was labeled with iTRAQ 4-plex reagent dissolved in ethanol, acetonitrile, methanol, or DMSO. Column heights reflect the average of duplicate experiments with error bars indicating one standard deviation.

3.3 Enzymes for digestion of protein-level isobaric labeled serum samples

Protein-level isobaric labeling modifies lysine residues, which blocks tryptic cleavage. Trypsin then cleaves only after arginine, producing longer peptides that are less efficiently identified by MS/MS. To address this issue, Asp-N, chymotrypsin, and sequential digestion with trypsin/Glu-C were explored as alternatives to trypsin. With respect to total peptide or protein identifications, similar results were obtained with Asp-N and trypsin, which both outperformed chymotrypsin and sequential digestion with trypsin/Glu-C (Table 1). When considering quantifiable proteins, Asp-N performed the best among the four digestion strategies, followed by trypsin and chymotrypsin. When combining results from replicate runs of Asp-N or trypsin digests, only a 10% increase in quantified proteins was observed over the results from a single run. However, a 30% increase in quantified proteins was obtained by combining one run of Asp-N with one run of trypsin, when compared to a single run of either. These results support the recommendation that isobaric protein-level samples be digested with trypsin and Asp-N in parallel to improve protein identification and quantification numbers.

Table 1. Comparison of different digestion conditions for protein-level labeled samples.

Trypsin, Asp-N, chymotrypsin, and trypsin followed by Glu-C were used to digest proteins labeled at protein-level with isobaric tags. Asp-N + trypsin results were obtained by combining the results from separate Asp-N and trypsin results together.

Enzyme	Peptides	Quantifiable Peptides	% of Quantifiable Peptides	Identified Proteins	Quantifiable Proteins	% of Quantified Proteins
Asp-N	368	256	88%	87	77	69%
Trypsin	333	177	78%	91	71	54%
Chymotrypsin	283	216	85%	71	61	76%
Trypsin/Glu-C	100	27	40%	49	20	27%
Asp-N + Trypsin	738	480	82%	122	100	65%

3.4 Comparison of isobaric labeling reagents for protein level labeling

There are currently three commercially available isobaric tags: iTRAQ 4-plex, TMT 6-plex, and iTRAQ 8-plex. Differences in their chemical structures have lead to reports of variable performance for peptide-level labeling experiments. Pottiez et al. observed that iTRAQ 8-plex provided more consistent ratios than iTRAQ 4-plex by MALDI-TOF/TOF.²¹ However, Pichler et al. demonstrated that iTRAQ 4-plex outperforms TMT 6-plex and iTRAQ 8-plex for peptide identification by CID and quantification by HCD using an LTQ Orbitrap.⁴ It is noted that different ionization mechanisms (MALDI vs. ESI) and fragmentation methods (high energy CID vs. trap-type CID and HCD) were used, which may explain the observed differences. Motivated by these seemingly contradictory reports, the performance of the isobaric tags for protein-level labeling was investigated. The greatest number of identified and quantified peptides was obtained with iTRAQ 4-plex, followed by iTRAQ 8-plex and TMT 6-plex (Figure 3A). In comparison to iTRAQ 4-plex, 20% fewer proteins were identified and quantified using TMT 6-plex and iTRAQ 8-plex (Figure 3B). For these conditions, using electrospray for ionization and HCD for fragmentation, iTRAQ 4-plex protein-level labeling yielded the highest identification rates with TMT 6-plex and iTRAQ 8-plex found to be similar. While TMT 6-plex did not outperform iTRAQ 8-plex, as reported by Picher et al., the general trend is in agreement. Differences in the instrumentation, fragmentation modes used, and physicochemical properties of the peptides studied may be responsible for the observed difference.

Figure 3. Comparison of three isobaric tags using protein-level labeling.

The total number of unique identified and quantified peptides (A) and proteins (B) using three isobaric labeling reagents with protein-level labeling. Column heights reflect the average of duplicate experiments with error bars indicating one standard deviation.

3.5 Isobaric protein-level vs. peptide-labeling for lectin-based glycoprotein enrichment

Lectin affinity chromatography is a common approach for glycoprotein enrichment from biological samples to study differential glycoprotein abundance. ^{16, 18} However, lectin affinity for glycoproteins is easily influenced by organic solvents and other solutes. When subjecting protein-level isobaric labeled or unlabeled serum samples to AAL affinity enrichment, glycoprotein recoveries of ~5% were obtained for both. To confirm glycosylation, all identified proteins were manually checked for the “glycoprotein” annotation in the UniProt database. The ratio of known glycoproteins to identified proteins for the protein-level labeled serum sample was 75%, which is in agreement with the peptide-level labeled and label-free serum samples, both at 71%. Approximately 89% of known glycoproteins were quantified by either isobaric protein-level or peptide-level labeling, suggesting equivalent performance. The similarity in both the recovered protein amounts and enrichment factor for glycoproteins shows that isobaric labeling at the protein-level does not have a significant influence on lectin affinity enrichment.

Since the main impetus for this work was to use protein-level fractionation strategies with isobaric labeling, glycoprotein enrichment with peptide-level and protein-level labeling strategies were compared. For protein-level labeling, two 100 Fg depleted serum samples were labeled with iTRAQ 4-plex, combined, subjected to glycoprotein enrichment, and divided into two equal fractions for parallel digestion with Asp-N and trypsin. For peptide-level labeling, two 100 μg depleted serum samples were each was passed through separate AAL columns. The lectin bound fractions were digested using trypsin, labeled with iTRAQ 4-plex, and combined. All samples were analyzed in triplicate; detailed results are summarized in Supplementary Table S-1. While more peptides were identified with the combined protein-level runs than the peptide-level runs (912 vs. 805, respectively), the number of quantifiable peptides was comparable (600 vs. 593, respectively). Protein-level labeling slightly outperformed peptide-level labeling with respect to both identified (169 vs. 140, respectively) and quantified proteins (135 vs. 125, respectively). Although protein-level labeling gave better absolute numbers than peptide-level labeling, the ratios of quantifiable peptides and proteins to the total identifications were lower with protein-level labeling. Compared to the predicted percentage, lysine-containing peptides were slightly over-represented (+11%) with protein-level labeling (Supplementary Figure S-1). These results show that two different digestion methods (trypsin and Asp-N), in conjunction with additional LC-MS/MS runs, can be used with protein-level labeling to obtain comparable identification numbers to peptide-level labeling.

One theoretical advantage of isobaric protein-level labeling is that labeling occurs before glycoprotein enrichment and digestion, which eliminates the contribution of these steps to the overall method variance. The distribution from protein-level labeling (RSD = 11%) was narrower than that from peptide-level labeling (RSD = 15%), indicating higher overall precision (Figure 4). It is noted that the reduction in variance could be partially due to the additional instrumental runs performed for the protein-level samples (3 runs/enzyme, 6 total) compared to peptide-level samples (3 total). Although these were not exactly replicates, it is anticipated that some reduction in variance should be observed due to multiple measurements. Regardless, the strategy of using two enzymes with additional instrument runs is a workable strategy to increase the number of quantifiable proteins for protein-level labeled samples.

Figure 4. Quantitative precision for proteins identified from isobaric peptide-level or protein-level labeling.

It is noted that three LC-MS/MS runs were performed for peptide-level labeling and six LC-MS/MS runs for protein-level labeling. The box represents one standard deviation and the whiskers represent the entire range.

Conclusion

An intact protein isobaric tag labeling strategy was developed by comparing various solvents, enzymes, and commercially available isobaric tags. The optimum labeling conditions used DMSO for labeling, trypsin or Asp-N for digestion, and iTRAQ 4-plex as the isobaric tag. Trypsin and Asp-N digests generated complementary peptide identifications that increased overall protein coverage by ~30%. No observable effect on glycoprotein enrichment using AAL was observed, both in terms of the relative and absolute number of glycoproteins. Compared to peptide-level labeling, increased quantitative precision was obtained from isobaric protein level labeling using two enzymes and additional LC-MS/MS runs. Overall, the isobaric protein-level labeling strategy described herein provides a novel quantitative strategy for serum glycoproteomics.