iTRAQ Is a Useful Method To Screen for Membrane-Bound Proteins

Pediatric Blood and Marrow Transplantation, University of Minnesota, 670 CCRB, 425 East River Road,. Minneapolis, Minnesota 55455. Received August 1 ...
0 downloads 0 Views 314KB Size
iTRAQ Is a Useful Method To Screen for Membrane-Bound Proteins Differentially Expressed in Human Natural Killer Cell Types Troy C. Lund,*,† Lorraine B. Anderson,‡ Valarie McCullar,† LeeAnn Higgins,‡ Gong H. Yun,† Bartek Grzywacz,§ Michael R. Verneris,§ and Jeffrey S. Miller† Department of Medicine, Department of Biochemistry, Molecular Biology, and Biophysics, Department of Pediatric Blood and Marrow Transplantation, University of Minnesota, 670 CCRB, 425 East River Road, Minneapolis, Minnesota 55455 Received August 1, 2006

We are interested in the biological as well as the molecular processes involved in natural killer (NK) cell development and function. Determining the proteomic complement could be a useful tool in predicting cellular function and fate. For the first time shown here, we have utilized iTRAQ, a new method that allows identification and quantification of proteins between multiple samples, to determine the expression of membrane-bound proteins in two previously characterized human NK cell populations. One population was derived from umbilical cord blood (UCB) stem cells (CD34+38-Lin-) and the other from expanded CD3-depleted adult peripheral blood. iTRAQ was employed for multiplex peptide labeling of proteins from fractionated membranes followed by two-dimensional high-performance liquid chromatography (2D-HPLC), and tandem mass spectrometry was used to identify protein signatures. We were able to identify and quantify differences in expression levels of 400-800 proteins in a typical experiment. Ontology analysis showed the majority of the proteins to be involved in cell signaling, nucleic acid binding, or mitochondrial function. Nearly all proteins were associated with the plasma membrane, membrane-bound organelle (lysosome or mitochondria), or nucleus. We found several novel proteins highly expressed in UCB stem cell derived NK cells compared to adult NK cells including CD9, alpha-2 macroglobulin, brain abundant signaling protein (BASP1), and allograft inflammatory factor-1 (AIF-1). In addition, we were able to confirm several of our iTRAQ results by RT-PCR, Western blot, and fluorescence-activated cell-sorting (FACS) analysis. This is the first demonstration and verification using iTRAQ to screen for membrane-bound protein differences in human NK cells and represents a powerful new tool in the field of proteomics. Keywords: Natural killer cell ‚ tandem mass spectrometry ‚ iTRAQ ‚ umbilical cord blood ‚ liquid chromatography ‚ membrane-bound

Introduction Our lab has a strong interest in natural killer (NK) cell biology especially with regard to in vitro derivation of NK cells from stem cells. We have previously described conditions to derive human NK cells from umbilical cord blood CD34+CD38-linstem cells as well as to expand NK cells from adult peripheral blood.1-3 We have recently begun to characterize the differences between the NK cell progeny of these two populations with the hypothesis that the stem cell derived NK cells may represent an intermediate cell type in the pathway to a mature NK cell. To determine possible differences in receptors, secreted proteins, and other intracellular differences, we utilized the method of iTRAQ labeling to screen for membrane-bound proteomic * Author to whom correspondence should be addressed at Department of Medicine, Division of Hematology, Oncology, and Transplantation, Mayo Mail Code 806, 420 Delaware St. SE, Minneapolis, MN 55455. Phone: 612626-4249; fax: 612-626-4074; e-mail: [email protected]. † Department of Medicine. ‡ Department of Biochemistry, Molecular Biology, and Biophysics. § Department of Pediatric Blood and Marrow Transplantation.

644

Journal of Proteome Research 2007, 6, 644-653

Published on Web 12/22/2006

differences between these two populations and also to quantify such differences. In the past few years, there has been a rapid development of equipment and methods in the area of proteomics exampled by automated two-dimensional gel analysis, tandem mass spectrometry (MS/MS), and chip-based antibody arrays. We have chosen to utilize iTRAQ, a method of protein/peptide tagging that uses a multiplexed set of four isobaric reagents. iTRAQ chemistry allows for labeling of primary amine groups of peptides generated by tryptic digests of two to four different protein samples. After labeling, the samples are combined prior to downstream fractionation and identification. Tagging of primary amine groups allows one to label virtually all peptides in a mixture unless both lysine and a reactive N-terminus are lacking. This offers an advantage over the isotope-coded affinity tag (ICAT) method which relies only on cysteine labeling subsequently imposing bias toward cysteine-rich peptides.4 iTRAQ tagging allows for the effective quantification of peptides and therefore relative protein abundance on the basis of the 10.1021/pr0603912 CCC: $37.00

 2007 American Chemical Society

Screening NK Cell Membrane-Bound Proteins by iTRAQ

intensity of reporter ions in product ion spectra resulting from simultaneous tandem mass spectrometric analysis of the isobarically tagged peptides. Additionally, we performed twodimensional liquid chromatography (2D LC) of labeled peptides prior to MS/MS analysis to allow for extensive separation of the peptide fragments, now a standard method in proteomics. Although iTRAQ can be used to evaluate total protein expression,4-6 in our experiments we sought to focus our attention on a subset of the total proteome: the membranebound proteins. We expected that the reduced sample complexity would result in more protein identifications as compared to an unfractionated sample partly on the basis of having a smaller dynamic range of proteins. Analysis of membranebound proteins would also allow for the potential differential determination of cell receptors that could aid in our understanding of the relationship of stem cell derived NK cells to mature NK cells. Using iTRAQ, we were able to identify over 400 unique proteins (in summation from three data sets). A significant proportion of the protein set was shown to be membranebound (plasma membrane, mitochondria, or lysosomal) but also included some structural and cytoplasmic proteins. We were able to detect proteins that showed significant differences in expression between stem cell derived and mature NK cells. In addition, we undertook verification of our results by performing immunoblot or fluorescence-activated cell-sorting (FACS) analysis for the identified proteins. The iTRAQ and noniTRAQ methods for quantitation did not agree precisely in terms of the measured fold-change in proteins, however, the results from the multiple techniques were in qualitative agreement. Several potentially interesting proteins were found to be highly expressed in stem cell derived NK cells compared to adult NK cells and included brain abundant signaling protein (BASP1), alpha-2 macroglobulin (A2M), and allograft inflammatory factor-1 (AIF-1). These proteins will be used as a starting point in elucidating the pathway of differentiation from stem cell to mature NK cell. Thus, we show that iTRAQ followed by 2D LC-MS/MS can be used as a thorough, highly specific initial screening tool to determine proteins of interest that are differentially expressed between two cell types.

Materials and Methods Culture of Umbilical Cord Blood Stem Cell Derived Natural Killer Cells. The isolation and derivation of NK cells from umbilical cord blood (UCB) stem cells was carried out as previously described.1,3,7 In brief, UCB mononuclear cells were enriched for CD34+ cells by using the magnetic cell system (MACS) as recommended by the manufacturer (Miltenyi Biotec, Oberlin, CA). Resultant cells were stained with CD34 allophycocyanin (APC; BD Biosciences, San Diego, CA), and fluorescein isothiocyanate (FITC)-conjugated antibodies against CD2, CD3, CD4, CD5, CD7, CD8, CD10, and CD19 were used for the lineage (Lin) cocktail (all from BD Biosciences). Phycoerythrin (PE)-conjugated anti-CD38 (BD Biosciences) was used as the third fluorescent color for multicolor cell separation and sorting using the fluorescence-activated cell sorter (FACS) Star Plus (Becton Dickinson, San Jose, CA). CD34+/Lin-/38- cells were plated at 50 cells per well into 96-well plates pre-established with AFT024 stroma as previously described.3 Cells were cultured in a previously defined media developed by our lab for the derivation of NK cells from UCB.1 Cultures were maintained for 46 days after which aliquots were withdrawn to determine the NK population percentage. This was done by

research articles FACS analysis using an allophycocyanin (APC) conjugated antibody directed at CD56 (BD Pharmingen, San Diego, CA). Expanded cells were >98% CD56 positive, which indicates specificity for NK cells. All tissues were approved by the Committee on the Use of Human Subjects in Research at the University of Minnesota. Culture of Adult NK Cells. Adult peripheral blood mononuclear cells (PBMC) were CD3 depleted and expanded in NK cell-promoting conditions as previously described.2 Briefly, PBMC were labeled with anti-hCD3 magnetic beads (Miltenyi Biotec, Oberlin, CA) for 30 min at 4 °C, and immunodepletion was done using the MACS system as recommended by the manufacturer (Miltenyi Biotec, Oberlin, CA). Resultant CD3 depleted cells were cultured for 14 days in previously described media which specifically promotes NK cell derivation and expansion.2 Aliquots were withdrawn to determine the NK population percentage. This was done as described for stem cell derived NK cells. Expanded cells were >98% CD56 positive. All tissues were approved by the Committee on the Use of Human Subjects in Research at the University of Minnesota. Membrane-Bound Protein Preparation. Cells were harvested and washed twice in PBS followed by suspension in hypotonic lysis buffer at 107 cells per mL: 10 mM Tris (pH 7.5), 1.5 mM KCl, and proteinase inhibitors were supplied by a Complete tablet (Calbiochem, San Diego, CA). Cells were placed on ice for 10 min followed by 30 s of high-speed vortexing and 20 strokes in a glass homogenizer. The unbroken cells and nuclear debris were pelleted by centrifugation at 800g for 10 min at 4 °C. The supernatant was then ultracentrifuged at 110 000g for 45 min at 4 °C. The resulting pellet was considered the membrane component (also containing membrane-bound organelles). The pellet was washed once in hypotonic lysis buffer sans proteinase inhibitors. The insoluble product was snap frozen in aliquots and was stored at -80 °C. iTRAQ Labeling. Membrane-bound proteins from the derived NK cells (5.0 × 107 cells) were labeled with iTRAQ reagents using the protocol provided by the manufacturer (Applied Biosystems (ABI), Foster City, CA). Briefly, the membrane pellet was resuspended in dissolution buffer (0.5 M triethylammonium bicarbonate pH 8.5) supplied in the iTRAQ kit and was incubated on ice until the suspension clarified. A small amount of undissolvable material was usually present and was removed by centrifugation. Protein concentration was determined by Bradford method. Each sample was denatured and reduced, and cysteine residues were blocked with reagents supplied by the manufacturer. Peptides generated by trypsin digestion at 37 °C overnight were labeled with iTRAQ reagents at lysines, terminal amine groups, and tyrosines. The four isobaric labels have a nominal mass of 145 Da and consist of a “reporter” functional group (114, 115, 116, or 117 m/z), a “balance” group (31-28 amu), and a peptide reactive group. Each of the four chemical labels dissociates in the mass spectrometer to produce one of the discrete reporter ions, which is measured in an MS/ MS scan and provides the peaks used for peptide quantitation. Protein digests (20 µg for each sample) were labeled in parallel with a specific reporter and were combined into one tube. UCB protein was labeled with reagent 114 and adult protein was labeled with reagent 117. The combined peptide mixture was dried in vacuo, was resuspended in 0.1% trifluroacetic acid (TFA), and was applied to a Sep-Pak C18 cartridge (Waters Corporation, Milford, MA) to remove buffer, trypsin, and salts. The eluted peptides were dried in vacuo and were resuspended in water before online 1D reversed-phase high-performance Journal of Proteome Research • Vol. 6, No. 2, 2007 645

research articles liquid chromatography (HPLC) separation of 1% of the total mixture directly into the mass spectrometer. The preliminary 1D LCMS allowed us to ensure proper labeling had occurred. The remaining amount of the complex peptide mixture was separated by 2D LC to increase detection of peptides from low abundance proteins. Peptides were separated in the first dimension by strong cation exchange (SCX) chromatography on a Magic 2002 HPLC system (Michrom BioResources, Inc., Auburn, CA) using a Polysulfoethyl A column (150 mm length × 1.0 mm ID, 5 µm particles, 300 Å pore size) (PolyLC Inc., Columbia, MD). The sample was dissolved in 350 µL of SCX buffer A (20% v/v acetonitrile (ACN), 5 mM KH2PO4 pH 3.2 with phosphoric acid) and was loaded onto the column. Peptides were eluted with a gradient of 0-35% buffer B (20% v/v ACN, 5 mM KH2PO4 pH 3.2, 500 mM KCl) over 12 min and from 35% to 100% buffer B for 38 min at 35 µL/min column flow rate. During elution, the absorbance at 280 nm was monitored, fractions were collected at 1.5-min intervals, and each fraction was speed-vacuumed to dryness. Fractions 13-23, which showed mAU280 > 2, were analyzed in the second dimension, C18 reversed-phase separation. The entire contents of each fraction was reconstituted in reversed-phase load buffer and then was injected onto a Dionex/LC Packings (Sunnyvale, CA) capillary LC system, online with a QSTAR Pulsar i mass spectrometer (ABI), as described previously.8 Briefly, peptides were loaded onto an LC Packings C18 Pepmap trap cartridge and were washed for 17 min at 35 uL/min with load buffer (98:2, water:acetonitrile, 0.1% formic acid). Peptides were eluted during an increasing gradient of acetonitrile from 5 to 35% acetonitrile over 47 min. Tandem mass spectral data was acquired in 3-s intervals using the information-dependent acquisition (IDA) mode. Collision energy for fragmentation of iTRAQ peptides was increased 20% above the standard level used for non-iTRAQ petides. Data Analysis and Interpretation. Relative abundance quantitation and peptide and protein identification were performed using ProQUANT software (ABI). For protein identification using peptide tandem MS data, Interrogator Algorithm9 was used to search the MS/MS data set against a human protein database from March 2, 2005, containing 187 752 proteins (Celera Discovery System, ABI) ProQUANT search parameters were minimum peptide confidence level 90%; peptide mass and fragment ion masses tolerances 0.15 and 0.1 Da, respectively; 1 trypsin missed cleave site, MMTS fixed modification of cysteines; and variable oxidation of methinonines. The peak areas of the reporter ions (114, 115, 116, and 117 m/z) were used by ProQUANT for relative protein quantitation (see ABI’s ProQUANT manual). The threshold for protein identification was set at a confidence of >95%. ProGroup Viewer Software (ABI) was used to compile the results from the database searches into protein groups, to remove protein redundancy, to provide protein-based ratios of relative abundance, and to export data and statistical parameters for further analysis and calculations. ProQUANT provides an estimate of the 95% confidence p-value for each protein hit. Peptides found in more than one protein are not used in quantitation. For our final reports, quantitation results were reported only when the standard deviation of the quantitation was below 20%, which is indicated as an “error factor” (EF) of 90% confidence per peptide (actual numbers were 412, 632, and 809 proteins). We compared iTRAQ ratios for UCB and adult proteins in three separate experiments, that is, three separate membrane-bound protein preparations of NK cells derived from three adult donors and three UCB donors. For each experiment, we examined the original protein list generated by ProQUANT and filtered it according to the following criteria: (1) two or more peptide hits for protein identification (minimum 90% confidence per peptide), (2) a minimum of three peptides used for calculation of iTRAQ ratios, and (3) ratios with EF’s < 2.0 (EF defined in Materials and Methods). Among the three experiments, the range of proteins in the filtered list was 129-325. Next, this subset was divided into two groups on the basis of p-value: (1) p-values < 0.05, (Figure 1a) and (2) p-values > 0.05 (Figure 1b). For iTRAQ ratios, the p-value reported by ProQUANT is a measure of the probability that the calculated ratio differs from unity by chance. When the p-value is 3 for each ion, and peptide fragment ion patterns were inspected to ensure the presence of at least four consecutive b- or y-type fragment ions and proper assignment of all product ion peaks in the MS/MS spectrum. The ratios from the filtered data of grouped proteins for one experiment were plotted in Figure 1a which shows log ratio ( log EF for 137 proteins from UCB stem cell derived NK cell and adult NK cell membrane-bound protein preparations where the p-value < 0.05. The curve shows a fairly even distribution of proteins with regard to increased and decreased expression in either cell type. This same data distribution was seen for each experiment performed. Log ratios for 149 proteins from UCB stem cell derived NK and adult NK membrane-bound protein preparations where the p-value > 0.05 are shown in Figure 1b. These protein ratios are not statistically different from 1 according to the data acquired. Manual inspection of precursor ion signal in the TOF

research articles

Figure 1. iTRAQ technology identified and quantified membranebound proteins from UCB stem cell derived and adult NK cells. The log ratios ( log error factors were plotted for 137 individual proteins identified by MS and were quantified using iTRAQ technology. The error factor sets the 95% confidence interval for an average ratio. The results plotted in top panel (a) are a subset of proteins with EF < 2 and p-value < 0.05 which indicate that the observed ratios are significantly different from 1:1 (or 0 on log scale). The minimum number of peptides considered for protein identification and quantitation was two and three, respectively. The bottom panel (b) shows a subset of 123 proteins from the same experiment as above with the following limits: two peptide minimum for identification; three peptide minimum for quantitation; EF < 2.0; p-value > 0.05. Protein ratios in this subgroup do not differ significantly from 1:1 (or 0 on log scale) according to the data sets obtained.

MS scans and reporter ion signal intensity and quality could reveal outlier peptides whose ratio contributes to the average ratio for a protein. We did not inspect the data sets for possible outliers, therefore, further inspection of data and recalculation of statistics could possibly reveal ratios in this protein group that are statistically different from unity. Next, we looked at the ontology of the generated data set. Figure 2a displays the PANTHER biological process ontology inclusion criteria as for Figure 1a (two or more peptides for ID and three or more peptides for ratio, p-value < 0.05). A large percentage (33%) of the proteins was involved in cell adhesion, trafficking, and signaling as would be expected from membranebound proteins. Another large portion of the proteins, 27%, was involved in cell metabolism as the result of mitochondria copurification. Nuclear proteins were also present indicating that even after centrifugation to remove nuclei, some proteins remained in the preparation as cross-contaminants. Figure 2b shows the biologic process ontology for protein with p-value > 0.05. The graph shows that 39% of the proteins were involved in cellular metabolism again signifying mitochondrial presence. Again, a large portion (26%) was involved in adhesion and signaling. The comparison of Figure 2a and 2b shows a very similar breakdown indicating a uniformity in the protein sets even though one set indicated a change in expression and one Journal of Proteome Research • Vol. 6, No. 2, 2007 647

research articles

Lund et al.

Western blots were performed on membrane preparations to verify upregulated proteins found by iTRAQ labeling. Figure 4 shows clearly that BASP1, A2M, and MPO had higher expression in UCB stem cell derived NK cells versus that of adult. In fact, very little to no protein could be detected in the membranes of adult derived NK cells. These data confirm that these proteins identified by iTRAQ are, in fact, expressed, and they are also present in significantly higher levels in the membrane fraction from UCB derived NK cells as was determined by iTRAQ quantification. Immunoblot of whole cell lysates confirmed the higher overall expression of BASP1 and A2M in stem cell derived NK cells. This is different from the finding for MPO that showed a similar whole cell lysate expression level (but higher expression at the membrane) suggesting a disparity in subcellular localization between the two cell populations. Figure 5 shows RT-PCR data for A2M, MPO, and AIF-1. As mentioned previously, AIF-1 is a cell surface protein we identified by a single peptide from the tandem MS data that was later verified by FACS analysis (Figure 6). RT-PCR shows there is more message for A2M and AIF-1 in UCB stem cell derived NK cells and supports the iTRAQ data. Message for MPO was similar for both cell types (consistent with the whole-cell lysate immunoblot), which indicates that expression differences may be occurring because of posttranslational processing resulting in intracellular trafficking differences. These data indicate that the increase in protein abundance determined by iTRAQ could be validated by Western blot analysis and RT-PCR.

Figure 2. Ontology of cumulative identified NK cell membrane proteins determined by iTRAQ and 2D-LC MS/MS. The PANTHER classification scheme (http://www.pantherdb.org) provided protein source and function. The data shown are grouped by biologic process and show data with p-value < 0.05 (a) or > 0.05 (b). The graphs represent the data shown in Figure 1a and 1b.

did not. The higher metabolic component of the P-value > 0.05 data is not unexpected as often this contains “housekeeping” type proteins which are not expected to vary too much between closely related cell types. In addition, since our membrane preparation was an “enrichment” method, and therefore somewhat crude, cytoskeletal elements were present and expected. Verification of iTRAQ by Western Blot and RT-PCR. Several proteins with a 2-fold or greater change in abundance were chosen for further verification. These included alpha-2 macroglobulin (A2M), myeloperoxidase (MPO), brain abundant signal protein (BASP1), allograft inflammatory factor-1 (AIF1), CD9, and granzyme. Incidentally, CD9 and AIF-1 were identified having only single peptide “hits” in the original data sets. The aforementioned proteins were chosen on the basis of their novelty and repetitive appearance in sequential experiments (except AIF-1 was found in a single experiment). Figure 3 summarizes the ratios of these “singled-out” proteins as well as CD44, CD59, CD54, and CD56, proteins that were identified and did not show 2-fold changes (actual values available in supplemental data.) Of the latter four proteins, CD54 and CD 59 were identified with only one peptide. The iTRAQ ratios for each of the proteins present in two or more experiments are consistent from experiment to experiment (AIF-1 is an exception). 648

Journal of Proteome Research • Vol. 6, No. 2, 2007

Verification of iTRAQ by FACS Analysis. Using iTRAQ technology, we were able to identify several cell surface markers present on both UCB stem cell derived and mature adult NK cells including CD9, CD44, CD54, CD56, and CD59. We therefore sought to confirm the iTRAQ results by FACS analysis which would allow us to verify the presence of the CD markers and to compare differences between the two populations using another method (Figure 6). Each FACS plot represents cells that were gated for CD56+ which ensures that the antibodies were binding to cell surface markers present on NK cells (over 98% of cells derived in our culture system were NK cells). The difference in CD9 expression was quite striking. Eleven percent of UCB stem cell derived NK cells showed expression, while little to no expression was seen on adult NK cells. Evaluation of AIF-1 previously determined by a single peptide hit by iTRAQ (Figure 3) showed higher expression in stem cell derived NK cells on FACS analysis when compared to adult cells: 5.7% (stem cell) versus 1.5% (adult) as shown in Figure 6. RT-PCR of AIF-1 confirmed that the corresponding mRNA was produced in both cell populations, but slightly higher expression was found in UCB stem cell derived NK cells (Figure 5). Because we measured marked expression of A2M on UCB stem cell derived NK cells, we also looked for the A2M receptor, CD91, and found it to be exclusively expressed on 10% of the UCB stem cell derived NK cells (data not shown). Comparison of expression levels of stem cell derived NKs to adult NKs with regard to CD44, CD54, CD56, and CD59 proteins using iTRAQ technology showed ratios very near unity (Figure 3). The FACS results (Figure 6) show that nearly all cells, regardless of derivation, expressed CD56 or CD59. CD44 and CD54 were also 100% present on both populations similarly (data not shown). These results verify the observation that iTRAQ ratios near 1.0 indicate similar expression levels between the two populations being studied.

Screening NK Cell Membrane-Bound Proteins by iTRAQ

research articles

Figure 3. Proteins from NK cell membranes determined by iTRAQ and 2D-LC MS/MS chosen for further investigation. UCB stem cell derived NK membrane-bound peptides were labeled with isobaric tag 117 and those from adult NK with 114. The Y-axis shows the log 117:114 ratio of each protein as it was detected in one or more experiments (experiments #1-#3 as indicated). The error bars show the log EF at 95% confidence if the EF was