Protein Mobility Shifts Contribute to Gel Electrophoresis Liquid Chromatography Analysis

Abstract

Profiling of cellular and subcellular proteomes by liquid chromatography with tandem mass spectrometry (MS) after fractionation by SDS-PAGE is referred to as GeLC (gel electrophoresis liquid chromatography)-MS. The GeLC approach decreases complexity within individual MS analyses by size fractionation with SDS-PAGE. SDS-PAGE is considered an excellent fractionation technique for intact proteins because of good resolution for proteins of all sizes, isoelectric points, and hydrophobicities. Additional information derived from the mobility of the intact proteins is available after an SDS-PAGE fractionation, but that information is usually not incorporated into the proteomic analysis. Any chemical or proteolytic modification of a protein that changes the mobility of that protein in the gel can be detected. The ability of SDS-PAGE to resolve proteins with chemical modifications has not been widely utilized within profiling experiments. In this work, we examined the ability of the GeLC-MS approach to help identify proteins that were modified after a small hairpin RNA-dependent knockdown in an experiment using stable isotope labeling by amino acids in cell culture-based quantitation.

INTRODUCTION

Profiling of cellular and subcellular proteomes by liquid chromatography with tandem mass spectrometry (LC-MS/MS) after fractionation by SDS-PAGE has been in use since 2002.^1–3 The GeLC-MS (gel electrophoresis liquid chromatography-mass spectrometry) approach decreases complexity within individual MS analyses by size fractionation of the intact proteins resulting in greater sequence coverage for many of the identified proteins than is achieved using 2-dimensional chromatography of tryptic peptides. SDS-PAGE is considered an excellent fractionation technique for intact proteins with the properties of good resolution and applicability to proteins of all sizes, isoelectric points, and hydrophobicities. These properties have resulted in the GeLC-MS approach being widely adopted to increase depth of coverage for the proteome being examined and for individual targets within that proteome. Information beyond the protein identification can be derived from the mobility of the intact proteins after an SDS-PAGE fractionation. However, that information is not typically incorporated into profiling analysis. Because the first dimension fractionation in the GeLC-MS approach is based on mobility in the SDS-PAGE, information about chemical or proteolytic modification of the proteins that changes an individual protein’s mobility in the gel can be obtained. Mobility shifts due to proteolysis are well documented to occur for selected proteins during several physiologic and pathophysiologic processes.^4–7 For example, in a GeLC experiment, proteolysis-dependent mobility shifts have been documented for 261 of the 1648 identified proteins in cells undergoing autophagy.⁸ The study of autophagy is also notable as the mobility shift on SDS-PAGE of the LC3 protein caused by truncation, and lipidation is used as an assay for autophagy.⁹

Mobility shifts during SDS-PAGE fractionation occur with chemical modification of proteins, including phosphorylation,^10–13 glycosylation,^14,15 hydroxylation,^16–18 methylation,¹⁹ and ubiquitination.^20,21 The ability of SDS-PAGE to resolve and select proteins with chemical modifications has not been widely utilized within profiling experiments. In this work, we examined the ability of the GeLC-MS approach to help identify proteins that were modified in cells undergoing a small hairpin RNA (shRNA)-dependent knockdown in an experiment using stable isotope labeling by amino acids in cell culture (SILAC)-based quantitation.

Stem cell proteomes have been profiled by several groups, and SILAC has been employed to help in quantifying changes in stem cell proteins as the cells undergo differentiation.^22–30 One of the advantages of SILAC for profiling is that the labeled samples can be mixed and fractionated by SDS-PAGE without losing the trypsin cleavage site at Lys that is utilized by amine reactive reagents that block trypsin activity.^22,25,28,29 Technical variation in sample processing contributes to variance in quantitative proteomic analyses.^31–33 Mixing the samples before fractionation results in more precise protein quantitation and the ability to observe smaller changes in protein abundance.^34–36 For this work, we selected the mouse embryonic stem (mES) cell model with a constitutively expressed shRNA to decrease expression of the survival motor neuron (SMN) protein.³⁷ Cells were labeled with ¹³C₆-Arg and ¹⁵N₂,¹³C₆-Lys or were maintained in the same media with standard (i.e., light, Arg, and Lys). Total cell protein from light SMN-knockdown cells was mixed with an equal amount of protein from heavy-Arg plus heavy-Lys control cells, and the proteins were fractionated by 1-dimensional SDS-PAGE. Patterns of protein mobility on the SDS-PAGE and analysis of protein abundance differences based on gel mobility of the proteins were then evaluated.

MATERIALS AND METHODS

Materials

Acetic acid was from BDH (VWR International, Radnor, PA, USA), CaCl₂ was from Orion Calibration Standards (Thermo Fisher Scientific, Waltham, MA, USA), TFA was from Fisher Scientific (Thermo Fisher Scientific, Pittsburgh, PA, USA), and formic acid was from EMD Millipore (Billerica, MA, USA). Sequencing Grade Trypsin was from Promega (Madison, WI, USA). Cell culture media, dialyzed fetal bovine serum (FBS), and isotopically heavy and light amino acids were from Cambridge Isotope Laboratories (Tewksbury, MA, USA). All other reagents including solvents used for HPLC were the highest grade available from Sigma-Aldrich (St. Louis, MO, USA).

Cell Culture

FVB embryonic stem cells (ESCs) were maintained and expanded in complete ESC medium containing DMEM high glucose, 20% embryonic stem (ES)-qualified FBS, ES supplement [4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, l-glutamine, and 2-ME], 1× nonessential amino acids, and 1000 U/ml ESGRO leukemia inhibitory factor (EMD Millipore). Cells were passaged every 2 d at a 1:4 split in 0.1% gelatin-coated flasks. Cells were plated at 3 × 10⁶ in 10 cm gelatin-coated dishes in complete ESC medium and incubated overnight at 37°C 5% CO₂. On the day of transfection, cells were 60–70% confluent, and Lipofectamine 2000 (Invitrogen, Life Technologies, Carlsbad, CA, USA) and 6–8 μg plasmid DNA were added to each dish and incubated for 4 h. The medium was then removed and replaced with complete ESC media with 600–800 μg G418 to kill any cell that was not transfected. These colonies were removed from the 10 cm dish, and each colony was placed in 1 well of a 24-well plate with complete media and G418. As the colonies grew, they were expanded so individual cell lines could be frozen and stored. Each clonal cell line was then assayed by quantitative PCR and Western blot (protein) to determine the extent of SMN knockdown.³⁷

The selected control and knockdown cell lines were then maintained in RPMI 1640 medium for SILAC supplemented with 10% dialyzed FBS, 0.46 mM l-Lys-HCl or 0.46 mM ¹³C₆, ¹⁵N₂-Lys-HCl, 0.47 mM l-Arg-HCl, or 0.47 mM ¹³C₆-Arg-HCl, 200 mg/l l-proline,³⁸ 2 mm glutamine, 50 μM mercaptoethanol, 100 U/ml penicillin, and 100 µg/ml streptomycin in a humidified 5% CO₂ atmosphere. Cells were passaged 3 times a week and harvested for experiments after 6 passages in the SILAC media.

Cells were harvested for analysis by washing twice with ice-cold HBSS then scraping from the 10 cm culture dish. Cells were recovered in 1 ml ice-cold HBSS then pelleted by centrifugation and stored at −80°C as a cell pellet until analysis.

Sample Preparation

Cell pellets were resuspended in 100 μl water, then 100 μl 2% LiDS (Sigma-Aldrich) was added, and the mixture was immediately immersed in 95°C water for a 5-min incubation. Protein in lysates was determined using a bicinchoninic acid protein assay (Pierce, Rockford, IL, USA) and diluted to 2 mg/ml with 1% LiDS. Equal amounts of protein from SILAC heavy and light cell lysates were combined then treated with 10 mM DTT and alkylated with 30 mM iodoacetamide before adding 10 mM additional DTT. There were 3 sample pairs with heavy control and light shRNA knockdown from independent culture dishes fractionated by SDS-PAGE on 10% polyacrylamide gels and stained with Coomassie blue dye. Each of the 3 sample lanes was divided into 30 fractions with the edges of each lane removed prior to slicing for analysis. Proteins in the gel were digested overnight with 0.04 μg trypsin per slice in buffer containing 20 mM Tris (pH 8.0) and 10% acetonitrile. Eluted peptides solubilized in 0.1% formic acid were analyzed by LC-MS/MS without further purification.

MS

All analyses were performed on a Thermo Scientific Q Exactive Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Fisher Scientific, Waltham, MA, USA). Peptides were separated by reversed-phase chromatography using an EASY-nLC 1000 Liquid Chromatograph system (Thermo Fisher Scientific, Waltham, MA, USA) and Acclaim PepMap 100 (75 μm × 2 cm trap) with Acclaim PepMap RSLC (75 μm × 15 cm column; Dionex, Sunnyvale, CA, USA). Peptides were eluted with a 2-h gradient from 5–30% acetonitrile with pH maintained by 0.1% formic acid. Column effluent was analyzed directly by MS/MS using high-energy collisional dissociation fragmentation.

Database Searching

Mass spectra were extracted from raw files and analyzed using MaxQuant version 1.4.1.2 (Max Planck Institute of Biochemistry, Martinsried, Germany). They were searched against the Universal Protein Resource Mouse Proteome database (downloaded August 2013, containing ∼43,539 entries; European Molecular Biology Laboratory-European Bioinformatics Institute, Cambridge, United Kingdom) along with the MaxQuant contaminants database assuming the digestion enzyme trypsin. The mass tolerances for parent ions were 20 ppm for the first search and 4.5 ppm for the second search. All fragment mass tolerances were 20 ppm. The iodoacetamide derivative of cysteine was specified as a fixed modification. Oxidation of methionine, acetylation of the N terminus, and phosphorylation of serine, threonine, and tyrosine were specified as variable modifications.

Criteria for Protein Identification

False discovery rates were calculated by searching a reversed database and were set to 0.01 for peptide-spectra matches and 0.05 for protein identification. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony.

Protein Quantification

To capture quantitative data on protein migration, it was necessary to calculate quantitative data for each individual slice rather than the whole lane. Therefore, 2 separate searches of the data were conducted using the 2 different quantification approaches. The first was an individual slice-by-slice quantification, where each slice was submitted as its own experiment, and so each protein was quantified up to 30 times per lane. The second was a MuDPIT (multidimensional protein identification technology)-style search in which the spectra from all 30 slices from a lane were submitted to MaxQuant as a single experiment. For the MuDPIT-style quantification, MaxQuant calculated a single quantification value for each protein in each lane. Unless otherwise stated, the data presented are from the individual slice quantification.

Data Analysis

Data analysis was carried out in R version 3.0.0 (R Foundation for Statistical Computing, Vienna, Austria). Proportional Venn diagrams were made with the eulerAPE tool,³⁹ available at http://www.eulerdiagrams.org/eulerAPE/. Values are presented as means ± SEM.

RESULTS

Control mES cells and SMN shRNA knockdown cells were grown in media with isotopically heavy Lys and Arg or isotopically normal Lys and Arg, respectively. A total of 50 μg protein from each of 3 independently cultured heavy and light cell lysates was pooled to obtain 3 independent samples then separated on a 10% polyacrylamide gel. Each lane was divided into 30 slices, and those were digested with trypsin and analyzed by LC-MS/MS (Fig. 1). Resolution of proteins was good, and distribution was even over the length of the lane (Fig. 2). Each of the 3 lanes was cut into approximately equal size slices. The pattern of the dye indicates that the majority of discernable bands on the gel are localized within only 1 slice. As indicated by the grid, slices were taken from the interior of each lane so that the streaked areas on the outside of the lanes were avoided where possible.

Figure 1.

The workflow for the GeLC analysis. SILAC heavy Arg and Lys control and normal Arg and Lys shRNA knockdown mES cells were mixed then separated into 30 fractions by SDS-PAGE. Proteins in gel slices were digested then analyzed by LC-MS/MS. Differences in protein abundance were evaluated with the migration distance considered in the analysis.

Figure 2.

Separation of proteins by SDS-PAGE. SILAC-labeled cell proteins were separated on a 10% polyacrylamide gel. The grid overlaid on replicate 1 demonstrates the area taken for individual slices. Streaking was more prevalent on the lane edges, and those were avoided in sampling.

With the high sample load on the SDS-PAGE and allowing 60 h of MS time for analysis of each lane, a depth of coverage of 7050 proteins was achieved. Of the identified proteins, 7013 (99.5%) were quantifiable in at least 1 lane, with 93.0% of all proteins quantifiable in at least 2 of the lanes (Fig. 3). The reproducibility of the analysis was excellent for moderate-to-low abundance proteins, as demonstrated by the consistent detection of spectra of peptides from proteins that are markers of pluripotency (Fig. 4). Six proteins that are associated with pluripotency were detected in all 3 samples with a low SEM for sequence coverage and the number of unique peptides identified.

Figure 3.

LC-MS/MS analysis for SILAC-based protein abundance quantitation. The 3 GeLC-30 samples produced a total of 7050 protein identifications. Of all identified proteins, 82% were quantified in all 3 samples and 93% quantified in at least 2 of the 3 samples.

Figure 4.

Stem cell markers are confidently identified. The percent sequence coverage for the 3 samples is shown for the proteins in (A). The number of unique peptides for each protein is shown in (B). The proteins selected are POU domain, class 5, transcription factor 1 (Pou5f1; Oct-3 and Oct-4), Sox2, Nanog, c-Kit, Klf4, and alkaline phosphatase (ALP). Values are mean ± sem, n = 3.

Figure 5 shows the distribution in lane 3 of 4181 proteins selected as having limited distribution in the lane according to their MW rank based on their primary sequence, as a function of the gel slices in which they were found. Proteins were selected if they had a total intensity determined by MaxQuant of >10E⁷ and were identified in a single slice or in adjacent slices. It is apparent that proteins with a wide range of MWs migrate the same distance in the gel and are observed in the same gel slice. It is also apparent that the majority of proteins were identified in ≤2 contiguous slices and follow a distribution that roughly matches their MW. Considering only values with at least 5% of the total signal for each protein, 3529.7 ± 148.7 or 54.9% of all proteins quantified were located in a single slice, with an additional 2233.0 ± 226.3 or 35.0% found in only 2 slices (Fig. 6). In order to visualize the deviations in mobility from the expected mobility of the proteins on SDS-PAGE, the slice containing the largest signal for each protein was determined. The protein mobilities were then plotted as a function of the expected migration distance, which was defined as the moving average (median) of the nearest 100 neighboring proteins (Fig. 7).

Figure 5.

Protein migration pattern for abundant proteins with a narrow distribution in the gel. Proteins in lane number 3 that had a total signal intensity >10E⁷ and had a distribution limited to 1 slice or to adjacent slices were plotted by their MW (Mol. Wt.) rank and the slice number(s) where they were detected. The total protein signal intensity including both the heavy and light forms is represented by the polygon height. Zero intensity points adjacent to measured signals were used as coordinates to complete the polygons. Therefore, proteins that were quantified in only 1 slice appear as triangles.

Figure 6.

Distribution of proteins within the SDS-PAGE lanes. The number of slices required to account for 95% of the MS signal of each quantifiable protein is shown. On average, 89.6% (±0.2%) of quantified proteins were localized to 1 or 2 slices. Values are mean ± sem, n = 3.

Figure 7.

Protein migration distance can deviate from that predicted using the primary sequence. Each protein was assigned to a single slice where it had the greatest intensity. The migration distance in slice number of each quantified protein in lane 3 was predicted using a rolling median of the migration distances for 100 proteins with the closest MWs calculated from their primary sequences. (A) The difference between the actual and the predicted position for each protein is shown as a function of that protein’s MW rank. (B) The same data are plotted as a histogram. Arrows indicate that the shift in mobility was a reduction in migration (Towards Gel Top) or an increase in migration (Towards Gel Bottom).

If SMN shRNA knockdown caused posttranscriptional modifications that altered protein mobility in SDS-PAGE, those modifications should be apparent in the analysis of individual gel slices but less so in the MuDPIT-style analysis where changes in protein abundance in one slice could be nullified by averaging with the value from another slice. The protein could either be changed in the opposite direction in the second slice or unchanged so that the weighted average fold change of the 2 slices is <2. To determine if the analysis of individual gel slices might identify more proteins with abundance differences in the shRNA knockdown cells, the data from the same LC-MS/MS analyses were submitted to MaxQuant as individual samples or pooled as a single sample for a MuDPIT-style analysis. For this analysis, quantification was done using protein SILAC ratios. In the slice-by-slice analysis, if a protein was found in several gel slices and, therefore, assigned >1-fold change value, the maximum fold change for the protein from any slice was taken as the fold change for that protein. The slice-by-slice analysis resulted in 6414.3 ± 60.9 proteins quantified in the 3 samples, with 804.3 ± 32.8 having a ≥2-fold difference in abundance in any slice. The pooled MuDPIT-style analysis resulted in 6332.0 ± 59.3 proteins quantified, with 257.7 ± 10.9 having a ≥2-fold difference in abundance. The slice-by-slice analysis of abundance differences identified on average 546.7 more proteins that were at least 2-fold different in abundance in any one lane as compared to the MuDPIT-style analysis (P < 0.003, t test). Although many of these proteins are grouped close to the 2-fold change criterion, many have much greater differences with SEM limits for the 3 samples that do not overlap with the values from the MuDPIT-style analysis (Fig. 8). Of the proteins quantified in the MuDPIT-style analysis, 96.5% were also quantified in the individual slice analysis. Of the proteins with a SILAC ratio >2-fold in the MuDPIT analysis, 80.0% also had at least one 2-fold change in an individual slice. The mean and SEM were calculated for each protein, and the median SEM was 0.095 (log2). The median SEM for protein fold change determined by the pooled slice analysis was 0.068 (log2).

Figure 8.

Proteins that have a 2-fold change in abundance only when slice-by-slice GeLC analysis is used. Open circles represent data from the slice-by-slice analysis and closed circles the MuDPIT-style analysis of the same spectra. For the three samples an average of 546.7 ± 25.2 proteins have a >2-fold change in abundance in slice-by-slice analysis that do not have a 2-fold change when the pooled spectra are quantified. Values are mean ± sem, n = 3.

One explanation for proteins having retarded mobility on SDS-PAGE is that they are highly phosphorylated. Retarded mobility is a well-established phenomenon for phosphoproteins undergoing SDS-PAGE.^10,12,13 In order to determine if we could detect this phenomenon in this data set, proteins with a ratio of identified phosphorylation events-to-protein sequence length >1–100 were designated as highly phosphorylated and plotted as the gel slice they were identified in vs. MW. From this data set, 320 proteins were identified with at least 1 phosphorylation, and 37 of those were highly phosphorylated. Each phosphoprotein was assigned to the single gel slice in which it was found with the highest signal intensity. To establish the expected migration range for proteins, a cutoff of total signal intensity of at least 10E⁷ was set, and only proteins found in a single gel slice or in adjacent gel slices were included. Approximately 90% of the proteins in each lane fit these criteria and were considered to have had a discrete migration pattern. The typical mass range for proteins in each slice was determined by ranking the masses of proteins with discrete migration patterns in that slice. The highest 10% and lowest 10% of proteins as ranked by MW were removed, and the remaining range of masses was considered typical for that gel slice. In comparison to the proteins with discrete migration, the highly phosphorylated proteins have slower migration (open circles in Fig. 9). Highly phosphorylated proteins were defined as proteins having >1 identified phosphorylation site per 100 amino acids. The number of phosphorylation sites per highly phosphorylated protein varied from 1 to 15, with most proteins having 2 identified phosphorylation sites. The median MW for highly phosphorylated proteins was smaller than the median for all proteins identified (15.5 vs. 50.7 kDa). This may reflect the requirement we established to have at least 1 phosphorylation event for every 100 amino acids in the protein. We plotted randomly chosen, size-matched proteins with no detectable phosphorylation selected from the same analysis (filled circles in Fig. 9) to demonstrate that slow migration for highly phosphorylated proteins was not a function of their small size. The gray areas in Fig. 9 indicate the mass range containing 80% of proteins with discrete gel migration in each slice. The same information is presented in Supplemental Fig. S1 for phosphoproteins in lanes 1 and 3.

Figure 9.

Highly phosphorylated proteins have slower migration through the gel than expected based on MW. Open circles represent proteins with ≥1 identified phosphorylation events detected per 100 amino acids for proteins in lane 1. Filled circles represent randomly chosen, size-matched proteins with no detectable phosphorylation selected from lane 1. The gray areas indicate the mass range containing the median 80% of all proteins.

The final point that is evident and is unique to the GeLC analysis is the migration of a subset of the proteins to noncontiguous slices in the gel. Although the majority of proteins migrated in the gel as single species with distribution in 1 or 2 adjacent slices, there are distinct sets of proteins that have either a multimodal distribution or are very poorly localized. A multimodal distribution was observed for 640 proteins, defined as having at least 1 slice with lower intensity for that protein than both of the adjacent slices. (Fig. 10).

Figure 10.

Multimodal distribution of the nuclear autoantigenic sperm protein (Nasp; Q99MD9) after SDS-PAGE. The Nasp protein was detected in at least 3 noncontiguous slices in each of the 3 independent analyses.

DISCUSSION

SDS-PAGE is the most versatile option available for fractionating submilligram quantities of intact proteins. The advantages of SDS-PAGE include the ability to fractionate proteins regardless of size, isoelectric point, or hydrophobicity. In addition, although every fractionation technique has the potential for nonuniform loss of analyte, SDS-PAGE minimizes nonuniform protein loss. Essentially any protein that is soluble in sodium dodecyl sulfate (SDS) or a saturated urea solution will be retained in the gel and be available for the next step in the workflow. Although it has only modest resolving power in comparison to HPLC or capillary electrophoresis, the resolution obtainable with SDS-PAGE is adequate for subsequent MS-based proteomic analysis. This is supported by the data in Fig. 6 that indicate that 54.9% of all proteins were located in only 1 gel slice that represents only 3.3% of the resolving gel and that 89.6% of all proteins were identified in no more than 2 slices. With LC-MS/MS analysis typically using second-dimension liquid chromatography gradient times of ≥2 h, it is preferable to limit the number of first-dimension fractions to minimize the time required for the complete analysis. The time required for multiple LC-MS/MS analyses and the potential to lose analytes during the fractionation procedure combine to make it counterproductive to fractionate beyond the ability of the MS system to evaluate all the proteins in each fraction. SDS-PAGE is versatile enough to allow the number of fractions taken to be adjusted for the depth of coverage desired and the MS system being used.

In this analysis, we took care to avoid areas of the gel where streaking or nonuniform protein mobility was apparent. Removing the edges of each lane prior to slicing the lane into fractions appears to be necessary more frequently when using heavily loaded gels. For this work, the samples were labeled with isotopically distinct Arg and Lys so they could be mixed prior to the electrophoresis and the potential nonuniform loss of protein that could occur by trimming the edges of replicate samples did not affect the quantitation. We estimate that with 30 slices, the loss of the edges of each lane would result in ∼2 μg protein being available in each fraction of the 100 μg total protein loaded.

The GeLC-30 analysis over a total of 60-h MS time for the 30 fractions produced on average 6414 protein group identifications for the 3 individual samples and a total of 7050 for the sample set. Because these cells are a mouse stem cell model, markers associated with pluripotency were used as an indication of the depth of coverage. The ability to better estimate the abundances of such proteins in small samples is considered a highly desirable goal. Oct4, Sox2, Nanog, c-Kit, and Krüppel-like factor 4 (Klf4) were all quantified with 3–15 unique peptides in each of the 3 samples using MuDPIT-style analysis. It is clear that the depth of coverage achieved in the GeLC-30 analysis was sufficient to allow observations of low abundance proteins. The reproducible detection of such proteins with small variance establishes the sensitivity of the GeLC technique to identify and quantify targets that are important in the biology of the cells as well as better characterize the identity of the cell population of interest.

Protein electrophoresis in 1 dimension has sufficient resolving power for proteomic analysis of complex samples such as cell lysates. The migration distance for our set of identified proteins correlates well with the log of the protein MW: R² = 0.79 for the correlation between the slice of maximum intensity for each protein and log₁₀(MW). However, any particular protein will have a wide range into which it can fall. For a typical set of 10 proteins with adjacent MWs as calculated by the protein primary sequences if each is assigned only to the slice where they are at their maximum intensity, the 10 proteins will be found over a range of 5 gel slices (Fig. 5). Reasons for the variability include splice variants that have different MWs, posttranslational modifications (PTMs) such as phosphorylation, and partial folding in SDS and complex formation. In an extreme case of protein modification, a single protein can migrate to ≥2 distinct places in the gel. This splitting of the protein, as shown in Fig. 10, can provide information about the state of the protein that is being observed. High incidence of PTMs and/or isoform/variant proteins may indicate a cellular process that is tightly regulated.

Intentionally exaggerating a shift in mobility is also possible and can potentially extend the use of GeLC analysis even further. This type of shift in mobility has been demonstrated for individual proteins⁴⁰ and for groups of proteins⁴¹ that are phosphorylated. In this case, a Zn²⁺-coupled reagent retards the mobility of phosphorylated proteins in SDS-PAGE. The combination of a mobility-shifting reagent and GeLC analysis is yet to be tested for a highly complex sample such as a cell lysate.

Analysis of individual slices has the potential to provide additional information about the proteins that have observed differences between samples. Searching for abundance differences slice by slice identified 3.1 times as many proteins with a 2-fold change as the MuDPIT-style analysis of the same experimental data (Fig. 8). The additional proteins with a 2-fold change that are observed by using slice-by-slice analysis cannot be attributed to search space differences because the 2 searches both identify approximately the same number of proteins (6414.3 ± 60.9 and 6332.0 ± 59.3, respectively). The increased number of changing proteins in the slice-by-slice analysis demonstrates the potential of SDS-PAGE migration data to highlight protein modifications. The use of SILAC labeling allows us to determine the relative abundance between the 2 samples for a protein in any gel slice. Estimating changes in any protein abundance between slices, as would be expected for a shift due to phosphorylation, may also be possible and could provide information about the extent of the response being observed. Those estimates would need to be made on the basis of signal intensity instead of SILAC ratios.

The rate of migration on SDS-PAGE for proteins we could identify as being highly phosphorylated was reduced such that they were found in different gel slices than proteins of the same MW that had not been determined as being highly phosphorylated (Fig. 9 and Supplemental Fig. S1). Phosphoproteins that we identify by sequence matching consistently migrate slower than expected and are localized outside of the mass range for all proteins. Arbitrarily selected size-matched proteins migrate within the expected range for all proteins. Proteins that are separated by a physical selection such as mobility through an SDS-PAGE can still be identified by the peptides with nonmodified residues. This shift in mobility from that that would be expected might, therefore, be a useful indicator that a PTM is present.

The GeLC workflow has been widely adopted for proteomic profiling due to advantages in ease of the procedure, broad applicability to all classes of proteins, and low percentage of protein loss during the separation. By incorporating SILAC-based quantitation, nonuniform loss of proteins between samples is further reduced while retaining all the advantages of the GeLC workflow. The behavior of individual proteins during the SDS-PAGE in regard to migration to a single position or to multiple positions provides additional information about the proteins that is frequently not reported. This additional information is dependent on chemical differences in the individual proteins and, therefore, is expected to provide important data about cellular processes and regulatory events.