Optimization of the SDS-PAGE gel slicing approach for identification of human liver microsomal proteins via MALDI-TOF mass spectrometry

Copyright: © 2012 Petushkova NA, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.


Introduction
Although shotgun proteomics commonly employs the method based on multidimensional chromatographic separation of proteins (nLC-MS/MS), the SDS-PAGE has not as yet lost its popularity [1]. One important aspect of SDS-PAGE application is connected with its ability to analyse membrane proteins with enriched fractions -such as microsomes [2,3]; dorsal root ganglion plasma membrane [4]; enriched plasma membranes from embryos [5]; membrane proteins of yeast mitochondria [6] and human colon carcinoma [7]; and also, platelet from human blood [8].
In rather simple cases -e.g. with plasma membrane proteome of germinating barley embryos [5] or rat muscle [9] -after the separation procedure the stained bands were cut out and the identification of proteins occurring in these bands is performed. In complex mixtures the boundaries between bands are not distinct; in view of this, the sliceby-slice cutting is employed, i.e. the SDS-PAGE lane is cut into slices whereupon the protein content of each slice is analyzed by MS method [2,3,7]. Analysis of obtained slices by peptide mass fingerprinting (PMF) has shown that each slice often appears to be a mixture of 2 to 5 proteins [2,10]. Mixed spectra impeded identification, which necessitated the development of additional functions for the software; e.g. such an option as a signal-processing algorithm based on neural network improves the results of identification by Profound [10].
Ambiguity of results obtained by PMF compelled the researches to turn to LC-MS/MS for analyzing protein content in SDS-PAGE slices. Thus, Simpson et al. [7], while investigating membrane proteins of human colon carcinoma, cut the SDS-PAGE lane into 16 equal slices with the width ~3 mm each. Thiede et al. [11], in analyzing the proteins of Jurkat T-cells, cut sequentially the 150-mm gel lane into 100 slices, i.e. the width of each slice was about 1.5 mm. To identify the cytochrome P450 isoforms from rat [12] or human [13] liver microsomes, authors chose a definite interval of the gel lane (~0.8 cm) corresponding, in their case, to the appropriate molecular weight of cytochromes P450 (48 to 62 kDa) and cut this region into 5 or 6 slices, respectively. Thus, the width of each such slice varied between 1.3 -1.6 mm.
Recently, we have demonstrated the advantage of cutting the gel into much thinner slices (about 0.2 mm thick) [14]. The mass spectrometric information from neighboring slices was found to contain different peptides belonging to one protein, which enabled to enhance the integral degree of sequence coverage by 12 ± 5% [2].
Although the sequential slicing of SDS-PAGE is used in many experiments, the choice of optimal slice width is yet an unsolved problem and requires additional systematic investigation. Using human liver microsomes (HLM) as an example, we have compared the results of PMF-based protein identifications that were obtained after cutting the fixed gel region into 40 and 20 slices and demonstrated that the lesser is the cutting step the greater number of proteins was identified. Besides, it was shown that the decrease of slice thickness leads to the 2 to 6-fold increase in the peptide peak's intensity in the mass spectrum.

Sample preparation and SDS-PAGE
Purified preparations of microsomal membranes were obtained as described previously [2]. In brief, a sample of human liver tissue (8-12 g) was obtained during surgical interventions conducted at the National Research Centre of Surgery as nontumorous tissue surrounding surgically removed liver metastases [14]. Informed consent was obtained from all patients. The protocol for this study was approved by the National Research Centre of Surgery, Russian Academy of Medical Sciences. The resulting pellet of microsomal membranes was suspended in 100 mM Na-phosphate buffer (pH 7.4) that contained 100 mM KCl, 1mM EDTA, 1mM dithiotreitol, and 20% glycerol (v/v) and used for SDS-PAGE separation. The protein concentration was determined by the Bradford assay using bovine serum albumin (BSA) as a standard [15]. Protein samples (110µg each) were mixed (ratio, 1:10) with a sample buffer (i.e. 0.06M Tris-HCl, pH 6.8, 100mM dithiothreitol, 2% β-mercaptoethanol, 2% SDS, 10% glycerol, 0.01% bromophenol blue), then boiled for 5 min in water bath at 95°C and loaded onto a 12% separation gel (size, 70 mm x 100 mm x 1 mm). Composition of the electrode buffer was 26mM Tris base, 195mM glycine, 3.5mM SDS in water. Electrophoresis was carried out at 150V until the bromophenol blue front reached the bottom of the gel. The proteins were stained with Coomassie Brilliant Blue G-250 as described previously [16]. Quantity One (version 4.6.1) was used to examine the relative quantity of the 37 to 75 kDa range (the quantity if this region was measured by its intensity and expressed as a percentage of the total intensity of the lane).

In-gel digestion
The procedure developed by Shevchenko et al. [17] was used for in-gel digestion. The gel was first pre-frozen, and then cut with a manual microtome into thin slices: for the chosen mass range (length, 10 mm) 20 or 40 slices were obtained. After that each slice was divided into three equal parts, and the middle piece of each slice (volume, 0.8 mm 3 or 0.4 mm 3 ) was sampled for in-gel trypsinolysis. For this purpose, the Coomassie-stained slices were destained with water and washed with a mixture that contained 50% acetonitrile in 100mM ammonium bicarbonate (v/v), pH 8.9, for 20 min at 50°C. They were then incubated for 20 min in 100% acetonitrile. After the acetonitrile had been removed and the gel pieces dried, 6.3 ± 2.0 µL of trypsin solution (25ng/µL modified trypsin in 50mM bicarbonate ammonium) were added and the mixture was incubated at 37°C overnight. Then 15µl of 0.7% trifluoroacetic acid were added to each gel piece and the samples were incubated for 2 h at room temperature. The extracted tryptic peptides were used for mass spectrometric analysis.

MALDI-TOF MS
Each mixture of proteolytic peptides (1µL) was spotted on a MALDI target (600/384 Anchor chip: Bruker Daltonics, Germany) in three replicates and dried in air. For ionization, a solution of 2.5-dihydrobenzoic acid (3mg/mL) in acetonitrile: 0.7% trifluoroacetic acid (1:1 v/v) was used. Mass spectra in the m/z range 600 to 4000 were manually acquired using FlexControl software (Bruker Daltonics) in the reflection/delayed extraction mode with an accelerating voltage of 25kV and a 135ns delay using Ultraflex II (Bruker Daltonics). All mass spectra represented signal averaging of 1000 laser shots from one location on sample spot. From each sample spot 4-6 mass spectra were acquired. Laser fluency was adjusted above the desorption threshold of the matrix to obtained the best resolution and the highest mass measurement accuracy. Signals with a S/N ratio > 6, a maximum of 100 peaks per spectrum, were considered to build peak lists using the SNAP algorithm (FlexAnalysis 2.0, Bruker Daltonics) and internally calibrated with trypsin autolysis products (m/z 842.5094 Da and 2211.1046 Da). Resulting peak lists were used to search against NCBI non-redundant human sequences database (NCBInr) (ftp://ftp.ncbi.nih.gov/blast/ db/FASTA/nr.gz), UniProtKB/Swiss-Prot (ftp://ftp.ebi.ac.uk/pub/ databases/uniprot/current_release/knowledgebase/complete/, April 2010) or IPI_HUMAN (http://www.ebi.ac.uk/IPI/IPIhelp.html). Proteins were identified by their PMF using Mascot (Matrix Science, USA). During the database search, a maximum of one missed cleavage was allowed, considered trypsin cleavage; a mass tolerance of 0.15 Da was used, and variable modifications, such as methionine oxidation and cysteine modification with acrylamide, were taken into account.

Results and Discussion
The aim of the present paper was to investigate usefulness of SDS-PAGE thin slicing for enrichment of proteome coverage of the human liver microsomal fraction. Human liver microsomes (HLM) represent the most popular in vitro model for studying hydrophobic membrane proteins which display a large group of physiologically important receptors, transporters, and enzymes as well as key pharmacological targets. There are now multiple methods for analyzing of membrane proteomes. Fractionation of proteins by SDS-PAGE, followed by in-gel digestion of gel slices prior to MS of peptides has been widely used for the analysis of HLM proteome [13].
We have reported previously that combination of sequential thin slicing of SDS-PAGE gel lanes with consequent slice-by-slice massspectral protein identification can be used to differentiate between polymorphic variants of cytochromes P450 [2]. In the current report, the working hypothesis has been proposed that the decrease of gel slices' thickness enhances the number of identified membrane proteins under the same experimental conditions. To verify this hypothesis, the following experimental steps were performed. The proteins comprising human liver microsomal ghosts [14] were separated by SDS-PAGE. Equal amounts of protein (0.11mg) were deposited on two parallel gel lanes. After the separation, the 10-mm region of the gel lane, corresponding to molecular weights of 37 to 75 kDa (a prestained molecular mass marker was used to estimate the position of this region) was cut sequentially into slices: one lane was cut into 20 slices and another (parallel) lane, into 40 slices. Thus, the slices obtained differed in thickness by twofold, constituting 0.5 mm in case of 20 slices and 0.25 mm in case of 40 slices ( Figure 1). Thus, slice no.1 in the 20-slice series approximately corresponds to slices nos. 1 and 2 in the 40-slice series. Accordingly, slice no. 2 (20-slice series) corresponds to slices nos. 3 and 4 (40-slice series) and so on. Each slice was divided into three fragments and the middle piece of each slice (piece sizes, 0.5 х 0.8 х 2 mm and 0.25 х 0.8 х 2 mm for 20-and 40-sliced gel lanes, respectively) was subjected to in-gel proteolysis by trypsin.
A series of pilot experiments was conducted with the aim to select the optimal amount of trypsin solution. It was established that the amount of required trypsin depends both on the amount of protein loaded on the SDS-PAGE gel lane and on the volume of the gel slice taken for trypsinolysis. In our experimental conditions, the amount of trypsin solution varied within 6.3 ± 0.5μL. The average number of peptide peaks in the mass spectrum was calculated and found to be 96 ± 23 for 0.25-mm slices and 90 ± 26 for 0.5-mm slices. Difference in the peaks' number has no statistical significance, i.e. the decrease of the cutting step does not lead to the increase in the number of observed peptides; thus, irrespective of slice thickness, the susceptibility of proteins to hydrolysis was nearly the same.
Mass spectra for each slice were obtained in three replicates. From two to four MALDI-TOF mass spectra, taken from each of the three replicates (positions on the MALDI target) per one slice, the best one (i.e. the one with the greatest score for identified protein and the highest percentage of sequence coverage) was selected.
Protein identification was performed by PMF. Among several important parameters of PMF search engines (pI, taxonomy, cleavage enzyme, the number of peptides matched, the percentage of sequence coverage, etc.) peptide mass tolerance should receive a special attention [18][19][20][21]. For automated protein identification for a set of PMF-spectra the proprietary program PMFScan was developed to dispatch a batch of spectra with specific parameters to the local Mascot v.2.1 server. Using PMFScan, the tolerance between the theoretical and experimental peptide masses would screened over the range of values [20]. Figure 2 shows how the number of identified HLM proteins varied depending on the peptide mass tolerance in the range from 0.02 to 0.4 Da with a step of 0.01 Da. The asymmetric bell-shaped dependency with a widened maximum at 0.15 Da was observed for both 20 and 40 slices with two protein databases -UniProtKB/SwissProt and IPI_human. Upon cutting the gel lane into 20 slices, the number of identified proteins did not significantly differregardless of the database used. With 40 slices, the use of the IPI_human database yielded a 10% higher number of identifications than the use of UniProtKB/SwissProt. Such mass tolerance dependencies of database searching results are in qualitative agreement with the earlier reported data [18,21]. For example, Ossipova et al. [18] reported that human plasma protein identification results vary with tolerance value and achieved the maximum at 0.1 -0.15 Da.
Similar dependencies to those presented in Figure 2 were obtained when PMF was searched against NCBInr (data not shown). Overall, upon searching against NCBInr at the mass tolerance of 0.15 Da, 351 and 195 proteins were identified in 40 and 20 slices, respectively. Changing the value of mass tolerance from 0.15 Da led to reduction of the total number of identified proteins in both 40 and 20 slices. For example, at the mass tolerance 0.1 Da we could identify 241 proteins in 40 slices and only 119 proteins in 20 slices. It is to be noted that the total number of HLM proteins identified by use of the NCBInr database significantly surpasses the one obtained by UniProtKB/SwissProt and IPI_human. Such a difference is observed since there are over 10 mln entries in NCBInr whereas the other two databases, UniProtKB/ SwissProt and IPI_human, contain a much lesser number of entries: 500 thousand and 250 thousand entries, respectively.
The majority of proteome studies performed via PMF do not involve detailed analysis of the number of identified proteins based on the parameters specified for the PMF search engine. For instance, tolerance values are taken according to the manufacturing characteristics assuming the quality of internal calibration for several spectra. These values are usually decelerated, which in turn may result in the rapid decrease in the number of identified proteins [18,20]. However, our results have shown that peptide mass tolerance values significantly influence the number of identified proteins and, hence, the tolerance value for the every protein search requires careful selection.    more than 40% of identified proteins could be attributed to membrane proteins according to the UniProtKB/Swiss-Prot annotations ( Table  1). Replicate PMF analyses of the gel slices showed that 68% membrane proteins from human liver microsomal ghosts occurred in each of the repeats. Table 1 presents a list of 62 identified membrane proteins with molecular weights in the 35 to 90 kDa range. Twenty-five proteins were identified in 20 SDS-PAGE gel slices, while in 40 SDS-PAGE gel slices 53 membrane proteins were identified; seventeen proteins were common to both slice series. Half of the proteins listed in Table  1 were located in the endoplasmic reticulum, a major constituent of microsomes; the rest were the proteins from mitochondria, as well as from the inner and outer cellular membranes but not from ribosome. The procedure for obtaining the human liver microsomal fraction employed in our study presupposed the presence in this fraction of all above-mentioned proteins. The data obtained provide evidence for the possible presence within the microsomal fraction of elsewhere-localized proteins. Sutton et al. [3] have shown that 82% of mouse liver microsomal proteins identified in the 42-72 kDa bands on SDS-PAGE gel lane were associated with endoplasmic reticulum, mitochondria, membranes and cytosol; while nuclear proteins were under-represented. The twofold increase in the number of slices (from 20 to 40) leads not only to the increase of the total number of identified proteins but also to the proportional (double) increase in the number of membrane proteins, which commonly reside in liver microsomes. As regards the core microsomal inhabitants, cytochromes P450 (CYPs), the experiment with 40 slices allowed us to identify 10 different CYP forms while with 20 slices only 5 CYPs were identified.
Twofold divergence in the total number of proteins identified in the 40 versus 20 gel slices occurred due to detection of additional proteotypic peptides in the thinner slice (Figure 3 and Figure 4). Figure 3 reports the mass spectra with nos. 18 (0.5-mm thickness) and 34 (0.25-mm thickness) obtained for arylacetamide deacetylase (AAAD), a specific protein from the microsomal fraction of liver cells that is involved in activation of arylamine substrates to ultimate carcinogens [22]. On cutting the gel into 20 slices, this protein was not identified by the search engine: overall, only nine matching m/z values were registered in slice 18 (Figure 3a) with low sequence coverage of 21% and Mascot score of 41, which was below significant threshold of 56 (Figure 4a). However, with cutting into 40 slices, the enzyme was identified in four slices (nos. 33-36). The PMF spectra of AAAD (Figure 3b) in slice no. 34 contained 13 masses and sequence coverage of 34% was obtained under these conditions (Mascot score: 84, Figure  4b). Among peaks in both spectra, eight masses were shared between slices of 0.25-mm and 0.5-mm thickness (Figure 3 and  Figure 4d) were observed in more than 50% mass spectra in which the AAAD was identified, which allowed declaring these peptides as proteotypic [23].
Densitometry analysis of the protein region between 37 and 75 kDa demonstrated that this area contain 46.3 ± 3.2% of the total HLM protein loaded on the SDS-PAGE gel lane. As about 2 nmol of HLM protein were deposited on each parallel gel lanes, so one slice in 20-slices series contained in the region of 9.3 pmol of HLM protein, whereas one slice in 40-slices series contained approximately 4.6 pmol of HLM protein. After the in-gel trypsin digestion the concentration of HLM protein in the final solution volume was nearly equal in both slice series, and the concentration of matrix solution was 20 mg/mL in both cases. The final volume of mixture, consisting of proteolytic peptides and a matrix, spotted on each position on the MALDI target was 2.0µL. As a result a matrix/analyte molar ratio was nearly the same (around of 250:1) as in the case of 20-slice series and 40-slice series (with the assumption that the trypsin digestion was complete, all the peptides were extracted from the gel, and the concentration of the resultant peptide mixture was the same as the protein content per slice in both cases). Therefore our sample preparation conditions for both types of slices were close to optimal recommendations, according to which a final matrix : analyte molar ration must be at least 100:1 and a final volume of 0.5-2.0µL [24].     In analyzing peptide mixtures derived from gel slices with different thicknesses, we, in fact, examined the effect of proteomic sample dilution (1:2) on hydrolysis peaks intensity upon identification of HLM membrane proteins. It was established that in thinner slices the intensity of theses peaks is 2 to 6 times higher ( Figure 5). The intensities of 13 coincident hydrolysis peaks corresponding to the peptide fragments' masses of CYP2A6 (MW 56481) for 0.5-mm slice no. 15 and for 0.25-mm slice no. 30 are presented in Table 2. The intensity of peaks was enhanced; e.g. for the peak with m/z 2408 the accretion was about 600%. Only for two peaks (with m/z 1450 and 1474) the intensities from thinner slices were somewhat lower. Analogous results were obtained for liver carboxylesterase 1 and CYP 2C18 (data not shown). The hydrolysis peak's intensity for liver carboxylesterase 1 (MW 62481) in 0.25-mm slice no. 14 was 2 to 5.5 times higher than in 0.5-mm slice no. 7. For example, the accretion for the coincident peaks with m/z 1639 and m/z 949 was, respectively, 360% and 540%. Besides, in 0.25-mm slices the overall intensity of all peaks was, on average, 4.2 times higher than it was in 0.5-mm slices. Of note, the intensities of CYP 2C18 (MW 55711) peaks in one of 0.5-mm slices reached the maximum value of 3.7 x 10 5 while in 0.25-mm slice it was 1.5 x 10 6 .
Regular search engines do not use peak intensities as parameter for protein identification [25] however, peak intensity is an inherent attribute of the experimental spectrum useful for protein identification [19]. Peak intensity is related to matrix : analyte ratio as well to the ingel-concentration of the individual protein and to the complexity of the sample [26,27].
We assume that the influence of the sample complexity was the main reason for the increased MALDI peaks intensity for the thinner slices. For the soft ionization there exist selective effects that may lead to spectral suppression of distinct peptide ions. MALDI-TOF-MS reveals strong differences in signal intensity among peptide peaks [27], including the complete absence in spectra of the peaks for the relatively abundant peptide ions due to competitions for protons [28]. Due to cutting the protein bands on SDS-PAGE gel lane into the thinner gel slices leads to diminishing the complexity of the protein mixture in the piece, thus to declining the complexity of resultant peptide mixture from the thinner gel piece compared with the resultant peptide mixture from the thicker gel piece. Simplification of the peptide mixture created the preferential conditions for the matrix-assisted ionization, so we observed higher intensities of peaks and also the appearance of the proteotypic peaks, which trigger the identification of more proteins and enhance the sequence coverage.

Conclusions
Separation of liver microsomal membrane proteins can be achieved by SDS-PAGE, resulting in more than one protein per gel band. Several proteins are assumed to occur within each protein band, so the direct analysis of a peptide mixture resulting from proteolytic digestion often becomes a problem for peptide detection and protein identification. The simplified mixture can be obtained by cutting a protein band in to several slices. Our approach involved obtaining of the most possible number of thin slices in the selected region on the SDS-PAGE gel lane with subsequent MS identification of protein in each particular slice.
It was shown, that the number of identified proteins increased twofold when the selected gel region was cut into the two times thinner slices. There were 53 membrane proteins positively identified when   1-cm long fragment of SDS-PAGE gel lane that was cut into 40 slices, comparing to 25 proteins identified in 20 slices, with 17 proteins overlapped. Analysis of three to six replicates of MALDI-TOF mass spectra revealed that although the total number of peaks was not changed, the mass fingerprints were different for thin and thick slices, the former demonstrating increased peak intensity. PMF spectra for thin slices were enriched by few interpretable peaks, picked up by the search engine for identifying, additional proteins and for increasing the sequence coverage. We conclude, that step of SDS-PAGE gel slicing sufficiently influences on the protein identification by peptide mass fingerprinting.