Quantification of Membrane and Membrane-Bound Proteins in Normal

By culturing normal and malignant breast cancer cells with light or heavy isotopes ... A number of proteins show increased expression levels in malign...
4 downloads 0 Views 474KB Size
Download PDF
Quantification of Membrane and Membrane-Bound Proteins in Normal and Malignant Breast Cancer Cells Isolated from the Same Patient with Primary Breast Carcinoma Xiquan Liang, Jenson Zhao, Mahbod Hajivandi, Rina Wu, Janet Tao, Joseph W. Amshey, and R. Marshall Pope* Life Sciences R&D, Protein Analysis, Invitrogen, Carlsbad, California 92008 Received March 28, 2006

More than 50% of all major drug targets are membrane proteins, and their role in cell-cell interaction and signal transduction is a vital concern. By culturing normal and malignant breast cancer cells with light or heavy isotopes of amino acids (SILAC), followed by cell fractionation, 1D gel separation of crude membrane proteins, and analysis of the digests using nanoelectrospray LC-MS/MS, we have quantified 1600 gene products that group into 997 protein families with approximately 830 membrane or membrane-associated proteins; 100 unknown, unnamed, or hypothetical proteins; and 65 protein families classified as ribosomal, heat shock, or histone proteins. A number of proteins show increased expression levels in malignant breast cancer cells, such as autoantigen p542, osteoblast-specific factor 2 (OSF-2), 4F2 heavy chain antigen, 34 kDa nucleolar scleroderma antigen, and apoptosis inhibitor 5. The expression of other proteins, such as membrane alanine aminopeptidase (CD13), epididymal protein, macroglobulin alpha2, PZP_HUMAN, and transglutaminase C, decreased in malignant breast cancer cells, whereas the majority of proteins remained unchanged when compared to the corresponding nonmalignant samples. Downregulation of CD13 and upregulation of OSF-2 were confirmed by immunohistochemistry using human tissue arrays with breast carcinomas. Furthermore, at least half the gene products displaying an expression change of 5-fold or higher have been described previously in the literature as having an association with cancerous malignancy. These results indicate that SILAC is a powerful technique that can be extended to the discovery of membrane-bound antigens that may be used to phenotype diseased cells. Keywords: Breast Cancer • Quantitative Proteomics • SILAC • Biomarkers • Membrane Proteins • Immunohistochemistry (IHC)

Introduction Stable isotope labeling with amino acids in cell culture (SILAC) is a simple and accurate approach to quantify differential protein expression and dynamic regulation of posttranslational modification.1-14 Two populations of cells are grown in identical media, except that one contains amino acids of natural atomic abundance (light amino acids) and the other contains amino acids that are comprised of the heavy nuclei of the corresponding atoms, such as 13C and 15N ( heavy amino acids). During cell culturing, heavy amino acids such as [U-13C6]lysine and [U-13C6]arginine or their light congeners are incorporated into proteins using the natural biosynthetic machinery of the cells. The resulting labeling efficiency is almost 100%, and the proteins or their component peptides are chemically identical, but isotopically distinct. Cells labeled with light or heavy amino acids are combined and treated as a single sample prior to any process or protein purification. This eliminates * Address correspondence to Marshall Pope, Department of Proteomics, 5781 Van Allen Way, Invitrogen Life Science, Carlsbad, CA 92008, USA. Tel, 760-476-7033; fax, 760-476-6867; e-mail: [email protected].

2632

Journal of Proteome Research 2006, 5, 2632-2641

Published on Web 09/21/2006

quantification error due to unequal sample preparation and increases the reproducibility of the results. When dual labels of heavy Lys and Arg are used, the specificity of trypsin digestion ensures that all protease fragments, other than C-terminal peptides, are separated by a mass tag. Because proteins and peptides labeled with light or heavy amino acids are chemically identical, they coelute in liquid chromatography or electrophoretic separations. Nevertheless, the mass difference between the light and heavy peptides is distinguishable by mass spectrometry. Once the isotopic peptide pairs have been correlated to their respective proteins of origin, the relative intensity of the peptide pairs can be used to quantify differential protein expression between normal and disease states. Membrane proteins play a pivotal role in regulating cellcell interaction, recognition, migration, adhesion, and signal transduction. Currently, more than 50% of all major drug targets for medicines are membrane proteins. In this report, we described a SILAC approach to the quantification of differential membrane expression between normal and malig10.1021/pr060125o CCC: $33.50

 2006 American Chemical Society

Discovery of Biomarkers of Malignant Breast Cells Using SILAC

nant breast cancer cells, isolated from a 74-year-old female with breast carcinoma. Approximately 40 proteins were identified with expression level changes of 3-fold or larger. A few differentially expressed proteins were validated by immunohistochemistry on human breast carcinoma tissue arrays.

Experimental Procedures Materials. NuPAGE gels, NuPAGE sample buffers, SimplyBlue SafeStain, Dulbecco’s modified Eagle’s medium (DMEM), Lys- and Arg-deprived DMEM, fetal bovine serum (FBS), dialyzed FBS, epidermal growth factor (EGF), SuperPicTure Polymer Detection Kit, and monoclonal anti-CD13 antibody were obtained from Invitrogen Life Sciences. [U-13C6]L-Lysine and [U-13C6]L-arginine were purchased from Cambridge Isotope Laboratories. Normal L-lysine, L-arginine, aprotinin, leupeptin hemisulfate, phenylmethanesulfonyl fluoride (PMSF), and insulin were purchased from Sigma. Rabbit anti-osteoblast specific factor 2 (OSF-2) antibody was ordered from Biovendor Laboratory Medicine, Inc. Trypsin was purchased from Promega. Benzonase was from Novagen. Normal (HTB-125) cells and malignant (HTB-126) breast cancer cells, isolated from a 74 year-old female with breast carcinoma, were purchased from the ATCC. Cell Labeling and Fractionation. Normal breast cells were maintained in DMEM containing 10% FBS and 30 ng/mL EGF. Malignant cells were maintained in DMEM containing 10% FBS and 30 ng/mL insulin. To initiate the incorporation of light or heavy labels, normal breast cells (105 cells each) were harvested and resuspended in 3 mL of modified DMEM supplemented with 10% dialyzed FBS, 30 ng/mL EGF, and 0.1 mg/mL light L-lysine and light L-arginine, and an equal quantity of malignant cells were resuspended in 3 mL of modified DMEM supplemented with 10% dialyzed FBS, 30 ng/mL insulin, and 0.1 mg/ mL heavy [U-13C6]L-lysine and heavy [U-13C6]L-arginine. Initially, normal and malignant cells were cultured in two separate 60 mm dishes. Every 3-4 days, the cells were split or the media was replaced with the corresponding light or heavy labeling medium. In approximately six doubling times, the cells should achieve almost 100% incorporation of heavy amino acids. Under our conditions, the total cell count would be 64 × 105 upon the completion of labeling. It is advisable to normalize the number of normal and malignant cells before they are mixed by cell counting. Normal cells (∼106 cells/100 mm dish) and malignant (∼106 cells/100 mm dish) breast cancer cells labeled with light or heavy amino acids were scraped off the dishes and mixed at 1:1 ratio based upon a predetermined cell density. The cell mixture was lysed on ice for 30 min in 1.6 mL of hypotonic buffer (10 mM Tris-HCl (pH 7.4), 1 mM MgCl2, 0.5 mM PMSF, 0.15 µM aprotinin, 1 µM leupeptin hemisulfate, and 10 U/ml benzonase), followed by 30 strokes of a Dounce homogenizer. To the cell suspension, we added 0.4 mL of 5× sucrose (1.25 M sucrose stock in H2O) and mixed the solution five times in the homogenizer. The homogenate was centrifuged at 500g for 10 min to remove the nuclei. The supernatant was then centrifuged at 100 000g for 1 h to obtain a crude membrane fraction. Membrane pellets were dissolved in 60 µL of 2× SDS sample buffer containing 50 mM DTT and heated at 95 °C for 5 min, and half of the sample was analyzed by SDS-PAGE. An entire gel lane was cut into as many as 45 fractions, and the gel pieces were subjected to in-gel tryptic digest.

research articles In-Gel Tryptic Digestion. Gel slices were chopped into small particles with a clean pipet tip and then destained three times with 25 mM NH4HCO3 (pH 8.0) in 40% acetonitrile, followed by dehydration with 100% acetonitrile. The gel pieces were dried for 5 min under Speed Vac, followed by rehydration with 10-20 µL of 10 ng/mL proteomics-grade trypsin in 25 mM NH4HCO3 (pH 8.0). The minced gel slices were kept on ice for 1-2 h and then incubated overnight at 37 °C. Approximately 50 µL of 1.5% TFA was added to the tryptic digestion mix, and the solution was incubated at room temperature for about 30 min. After vortexing the gel pieces for about 1 min, the peptide extract was transferred into a clean Axygen Eppendorf tube. The gel pieces were further extracted with 50 µL of 50% acetonitrile in 0.75% TFA for 30 min. Peptide extracts were combined and analyzed by matrix-assisted laser desorption ionization-time-of-flight (MALDI-TOF) mass spectrometry. For liquid chromatography-mass spectrometry (LCMS) analysis, the peptide extract was dried under vaccum and the peptides were resuspended in 10% acetonitrile in 0.1% formic acid. Mass Spectrometry and Data Analysis. Tryptic peptides labeled with light or heavy amino acids were analyzed with nanoelectrospray LC-MS/MS using a Q-TOF API-US instrument (Waters Corporation). An 0.1 × 100 mm Atlantis dC18 column containing 3 µm particle packing (Waters Corporation) was used for peptide separations. The protein abundance within a given gel slice was estimated based upon the total amount of Coomassie stain. A gradient of 5-45% (v/v) acetonitrile in 0.1% formic acid over 45 min, followed by 45-95% acetonitrile in 0.1% formic acid over 5 min, was used to resolve peptides extracted from bands 1-5 and 40-45. In contrast, a gradient of 5-45% (v/v) acetonitrile over 90 min, and then 4595% acetonitrile over 30 min, was used to fractionate bands 6-39. On the Q-TOF, the intensity criteria of signal-to-noise for MS to MS/MS switch is set to 10, whereas the threshold for MS/MS to MS switch is set to above 3500 counts/s. Four component triggers were used to acquire MS/MS data with 1.4 s scan time, followed cyclically by a survey scan of 1.1 s from 400 to 1800 m/z. Ions in charge states of 2+, 3+, and 4+ detected in the survey scan were selected for MS/MS. Raw data files from the Q-TOF instrument were processed with Mascot Distiller (Version 1.1.1.0, Matrix Science, London, U.K.) without smoothing, using charge states determined from the MS scans.15 The resulting centroid files were searched against the August 15, 2004 NCBInr database with the Mascot search algorithm (Version 2.0). With the use of the prescribed parser format, monoisotopic and average masses for light Lys (128.09497; 128.1741) and Arg (156.10112; 156.1875), as well as [U-13C6]Lys (134.09497; 134.1741) and [U-13C6]Arg (162.10112; 162.1875), were entered in Mascot modification file. Variable modifications considered include oxidation of Met, and pairs of light or heavy Lys and Arg with a mass shift of 6 Da. Searches were restricted to Homo sapiens species entries present in the NCBInr database from August 15, 2004, by allowing a maximum of one missed trypsin cleavage event. The mass tolerance of the precursor peptide ion was fixed at 200 ppm, whereas the mass tolerance for the MS/MS fragment ions was set to 0.5 Da. Quantification of peptide pairs was performed and validated manually by examining both MS and MS/MS spectrums in MassLynx 4.0 (Waters, Millford, MA) with no smoothing applied. To compile the summary of identified proteins (see Supporting Information), we employed the Protein Prophet and Journal of Proteome Research • Vol. 5, No. 10, 2006 2633

research articles

Liang et al.

Figure 1. A schematic illustration of the SILAC approach for relative quantification of membrane proteins between normal and malignant breast cancer cells isolated from the same patient with breast carcinoma. Normal and cancer cells were grown in media containing either light Lys and Arg or heavy Lys and Arg for a minimun of six doubling times and then combined at a 1:1 ratio. The cell mixture was lysed in hypotonic buffer, and a crude membrane was obtained by centrifugation after removal of the nuclei. Membrane pellets were dissolved in SDS sample buffer and analyzed by SDS-PAGE. The entire gel lane was segmented into approximately 40 sections, followed by in-gel tryptic digestion. Peptide extracts were analyzed by nanoelectrospray LC-MS/MS, and the protein precursors were identified using Mascot Server. Finally relative quantification of protein expression was determined by the chromatographic response observed for each isotopic peptide pair in the MS.

Peptide Prophet algorithms, as implemented in version 1.05 of Scaffold (Proteome Software, Portland, OR).16,17 We required 95% confidence for individual peptides and a minimum protein confidence of 80%. A similar number of proteins were identified using a threshold model for Mascot scores. In brief, a difference of 10 between the peptide ion score and identity score was required using a significance threshold of p < 0.05. We also set minimum peptide threshold scores for +1, +2, +3, and +4 charge states at 20, 35, 40, and 40, respectively. Quantification of proteins with more than 3-fold change was achieved by averaging at least three peptide pairs with unique spectral sequences derived from the same protein (that is, matches to fragmentation patterns of 2+ and 3+ ions from the same peptide are considered independent). Immunohistochemistry. Nineteen formalin-fixed, paraffinembedded (FFPE) normal human breast tissue blocks and a FFPE MaxArray (Zymed-Invitrogen) human breast cancer array were used. Immunohistochemistry (IHC) was performed on 4 µm thick tissue sections. After deparaffinization and dehydration, antigens were heat-retrieved with 0.01 M citrate buffer (pH ) 6.0). After incubating primary rabbit anti-OSF-2 polyclonal antibody at a concentration of 0.125 µg/mL or mouse anti-CD13 monoclonal antibody at 1:50 dilution for 1 h, the goat anti-rabbit and goat anti-mouse HRP polymer detection system (SuperPicTure) was used. 2634

Journal of Proteome Research • Vol. 5, No. 10, 2006

Results Experimental Design and Sample Preparation. Figure 1 illustrates the strategy used to identify and quantify differential expression of membrane proteins between normal epithelial and malignant breast cancer cells isolated from a 74-year-old female with breast carcinoma. Normal and malignant breast cancer cells were grown in DMEM supplemented with either light lysine and arginine or heavy [U-13C6]lysine and [U-13C6]arginine, respectively. Because they are uniformly labeled with carbon 13, heavy isotopic forms of Lys and Arg are 6 Da heavier than their naturally abundant congeners. In a preliminary experiment, aliquots of either cell lineage were propagated a minimum of six doubling times in media supplemented with either light or heavy amino acids. The cells were lysed separately and analyzed simultaneously using SDSPAGE to fractionate crude membrane protein extracts (see Experimental Procedures). Protein bands with light or heavy labels were excised from adjacent gel lanes and subjected to tryptic digestion, and the extracted peptides were analyzed in parallel using MALDI-TOF. A shift of 6 Da for peptides containing heavy Lys or Arg was observed, indicating that the incorporation of heavy Lys and/or Arg was complete (data not shown). To characterize differential protein expression, normal and malignant cells were labeled with light or heavy amino acids

Discovery of Biomarkers of Malignant Breast Cells Using SILAC

research articles

Figure 2. Precision of quantification. Tryptic peptides extracted from light- and heavy-labeled cell states were analyzed by RPLC with Q-TOF detection. Thirty-seven isotopic peptide pairs derived from vimentin were recovered in a single LC-MS/MS run. The precision of quantification was determined by averaging the relative peak intensities across all peptide pairs, nine of which are shown. The relative ratio of isotopic peptide pairs is consistent regardless of the peptide charge state or the number of Lys and Arg residues within the peptide sequence (relative standard deviation of (8%)

separately, harvested, counted, and mixed in a 1:1 ratio using the calculated cell density to adjust the combined volumes. We lysed the resulting cell mixture (ca. 2.5 × 106 cells) in 1.6 mL of hypotonic buffer. The crude membrane pellets were obtained by ultracentrifugation after removing nuclei. Membrane proteins are hydrophobic and difficult to dissolve in zwitterionic detergents. Hence, they are often intractable for 2D gel electrophoresis or most liquid chromatography methods. However, this method derives quantification based upon ratios of their constituent peptides. Therefore, we dissolved the crude membrane pellets in SDS sample buffer directly and fractionated the solutions using 1D SDS-PAGE. Following Coomassie staining, the entire gel lane was divided into approximately 40 sections, and the proteins were subjected to in-gel tryptic digestion. The resulting peptide extracts were analyzed by nESI LC-MS/MS using capillary HPLC coupled to a quadrupole time-of-flight instrument. The raw data was processed with Mascot Distiller 2.0 (Matrix Science), and protein assignments were made using an inhouse version of Mascot Server 2.0. Quantification was performed manually by direct analysis of the chromatographic response of the corresponding peptide pairs in the MS spectrum using MassLynx 4.0 acquisition software. LC-MS/MS Analysis and Standard Deviations. A maximun of 100 proteins could be assigned in a single LC-MS/MS run, and as expected, isotopic peptide pairs always coeluted. Most peptides appeared doubly and triply charged during nanocapillary LC-MS/MS analyses on the Q-TOF instrument. Only occasionally did some peptides appear with four net charges. In the MS spectrum, therefore, most isotopic peptide pairs are 3, 2, or 1.5 Da apart. Frequently, both light and heavy peptide congeners were isolated for MS/MS fragmentation. In this case, internal fragments and b daughter ions were identical, but singly charged y series fragment ions, which carry the C-

terminus of the precursor peptide, appeared as satellite pairs separated by 6 Da (data not shown). For abundant proteins, such as plectin 1 isoform 11, more than 65 isotopic peptide pairs were recovered in a single LC-MS/MS run. Thus, the precision of quantification could be improved by averaging the chromatographic response for multiple peptide pairs derived from the same protein. For example, nine of the 37 peptides recovered from vimentin are illustrated in Figure 2. The relative intensity ratio of isotopic peptide pairs was consistent, and the standard deviation within a single run was (8%. The intensity ratio of isotopic peptide pairs was also consistent regardless of the number of lysine or arginine residues or the charge state of the peptide. For example, the relative ratio of isotopic peptides of LQDEIQNMKEEMAR at charge state 2+ was almost identical to that at charge state 3+. Chemical labeling methods that target all peptides require that the two samples be handled separately until the labeling step that follows proteolytic digestion.18,19 When isotopic labels are incorporated metabolically, however, the two cell populations are mixed before any analytical work. Then, the cell mixture is treated as a single sample in all the subsequent steps. This makes the accuracy and reproducibility of SILAC methods easy to verify. We performed three individual biological replicates and found that, as long as cells from normal and malignant populations were normalized, the results were highly reproducible. Figure 3 illustrates the recovery of peptide LLQDSVDFSLADAINTEFK derived from vimentin and peptide VLQLINDNTATALSYGVFR derived from oxygen regulated protein precursor in three individual experiments. The intensity ratios for isotopic peptide pairs corresponding to either protein remain relatively constant with a standard deviation of (10%. Metabolic Conversion of Arginine to Proline. The relative quantification of peptides containing proline required special consideration. This is because amino acids provided in the Journal of Proteome Research • Vol. 5, No. 10, 2006 2635

research articles

Liang et al.

Figure 3. The reproducibility of SILAC quantification is demonstrated by three individual replications of the cell culture experiment. Peptides derived from vimentin (a) or oxygen regulated protein precursor (b) were recovered from separate experiments and used for quantification. The relative ratio of heavy-to-light peptides remained constant at 0.5 for vimentin and 2.0 for oxygen regulated protein precursor. Among runs, the relative standard deviation for peptides accurately assigned was (10%.

Figure 4. Satellite peaks are observed for proline-containing peptides because Arg can be converted into proline via R-ketoglutarate intermediate. Because proline retains five carbons from the backbone of heavy-labeled arginine, small satellite peaks appear at [M + 2.5] Da if the peptide is doubly charged and at [M + 1.6] Da if the peptide is triply charged. To correct the quantification of prolinecontaining peptides derived from the tumor cell state, we summed the monoistopic peaks corresponding to either isotopic form of the proline residue. Peptides containing multiple proline residues were omitted from the cumulative averages ascribed to their protein precursor.

culture medium are subject to cellular metabolism.3 Arginine can be converted to proline via a glutamic acid γ-semialdehyde intermediate. The percentage of conversion is likely to depend on the cell type as well as the amount of arginine and proline in the culture medium. The conversion of arginine to proline splits the mass signal stemming from proline-containing peptides of the heavy-labeled cell state into two channels. Because proline retains five 13C atoms, heavy isotopic forms of prolinecontaining peptides appear at 2.5 and 1.6 m/z units, respec2636

Journal of Proteome Research • Vol. 5, No. 10, 2006

tively, above the double- and triple-charge states of peptides. Accordingly, when proline-containing peptides are used for quantification, the correction factor in peak intensity for total heavy-labeled peptides should be the sum of peak intensities of these mass channels. For example, as depicted in Figure 4, the peak intensity for heavy-labeled peptide ETNLDSLPLVDTHSK would be the sum of peptides at m/z 837.93 and at m/z 840.45, although the peak intensity for the light peptide at m/z 834.91 remains unaltered. This correction is indepen-

Discovery of Biomarkers of Malignant Breast Cells Using SILAC

research articles

Figure 5. The relative expression of proteins having isobaric peptides that coelute can be disentangled by an examination of other peptides generated from the component proteins. (A) Peptides from an unnamed protein and coactivator overlap at 1156.6. (B) Quantification for the unnamed protein and coactivator are performed using different peptide pairs from their corresponding proteins.

dent of the charge state of the peptide as well as the number of lysine or arginine residues in the peptide (compare panels C and D with panels A and B of Figure 4). The relative abundance of proline residues in membrane proteins, especially matrix proteins, is often high. Hence, the conversion of arginine to proline should be taken into consideration when the relative ratio of isotopic peptide pair is calculated. The appearance of multiple proline residues in a single peptide complicates the spectrum further. For example, a peptide of ISLPLPNFSSLNLR derived from vimentin is depicted in Figure 4E. In this case, an accurate measure for the intensity of heavylabeled peptides involves the sum of peptides at m/z 788.96, 791.47, and 793.96. Overlapping Peptides. In cases where peptides or peptide pairs overlapped, the overlapping peptides were not used for quantification. For example, the heavy form of peptide NLPGLVQEGEPFSEEATLFTK derived from 100 kDa coactivator (gi|7512259) overlaps with the light form of peptide VEGTEPTTAFNLFVGNLNFNK derived from an unnamed protein product (gi|271750187) (see Figure 5A). It is not practical to disentangle the respective envelopes of these isobaric peptides, nor is it necessary for quantification. Instead, the expression ratio for the individual proteins from normal and malignant cells may be recovered by examination of other peptide pairs from the corresponding proteins. For example, by examining the m/z 1251.14 peptide from gi|271750187 and the m/z 733.38 peptide from the 100 kDa coactivator (Figure 5B), we determined that gi|271750187 was upregulated about 2.4-fold in breast cancer cells and that the 100 kDa coactivator remained relatively unchanged between normal and cancer cells. Analysis of Differential Expression and Validation. Using the SILAC approach, we have identified 997 protein families with approximately 830 membrane or membrane-associated proteins; 100 unknown, unnamed, or hypothetical proteins; and 65 protein families classified as ribosomal, heat shock, or histone proteins. The majority of proteins remained unchanged

when compared to the corresponding nonmalignant samples. Table 1 displays proteins with differential expression larger than 3-fold. Certain cell adhesion or matrix proteins, such as epican, vimentin, integrin beta 1 isoform 1A precursor, annexin 1, A2, VI, and especially annexin V, show decreased expression levels in malignant breast cancer cells. Disregulation and loss of annexin have been related to prostate carcinogenesis and progression.20-22 Proteins regulating the function of ion channels and transporters, such as stomatin, plasma membrane calcium-transporting ATPase, and vesicle amine transport protein 1, showed decreased expression in cancer cells. Alpha 2 macroglobulin (A2M), a large glycosylated protease inhibitor, showed a significant decrease in malignant breast cancer cells. This is consistent with previous reports of a marked decrease of serum alpha 2 macroglobulin in prostate cancer.23 Pregnancy zone protein precursor (PZP), a near homologue, was also decreased in tumor cells. PZP is known to inhibit all four forms of serine proteinases by a unique trapping mechanism in which cleavage of the “bait” region causes a conformational change along the PZP periphery to trap the proteinase.24 Suppression of this class of protease inhibitors would seem consistent with accelerated invasiveness of the tumor. A relatively large number of expression differences may be catalogued in the initial, discovery-oriented phase of a search for biomarker candidates. We have noted, in particular, that the primary normal and malignant cells were supplemented with two different growth factors, and this could certainly contribute to expression artifacts. It is essential, therefore, to substantiate the protocol by repeating the study with a wider variety of matched cells, or preferably, by direct validation in human tissue arrays. The deregulation of membrane alanine aminopeptidase precursor (CD13), a zinc-dependent metallopeptidase, is illustrative. CD13 has been shown to be differentially expressed in normal prostatic stromal and epithelial cells, with increased Journal of Proteome Research • Vol. 5, No. 10, 2006 2637

research articles

Liang et al.

Table 1. Proteins Displaying a 3-fold or Higher Change in Differential Expression between Normal and Malignant Human Breast Cells Downregulated Proteins protein name

NCBI GI number

Membrane or Membrane-Bound Proteins Chain C, Annexin V 809190 CD81 antigen 12804239 1-8D 23396 epididymal protein 23092553 glutaminase 12044394 hexokinase 1 isoform HKI 4504391 inter-alpha-trypsin inhibitor heavy chain 3 precursor 422961 lysosome-associated membrane protein-3 variant 21070332 membrane-type matrix metalloproteinase 793763 macroglobulin alpha2 224053 membrane alanine aminopeptidase precursor 4502095 type-2 phosphatidic acid phosphohydrolase 3015569 plasminogen activator inhibitor type 1, member 2 24307907 transglutaminase 2 isoform a; transglutaminase C 39777597 Unknown (Downregulated Protein) unnamed protein product 7023123

SILAC ratio

5.5 ( 0.5 3.7 ( 0.4 3.4 ( 0.3 8.6 ( 0.9 5.5 ( 0.6 3.5 ( 0.4 5.3 ( 0.5 3 ( 0.5 4.3 ( 0.4 7.5 ( 0.8 10 ( 0.9 5.1 ( 0.5 3.8 ( 0.4 6.5 ( 0.6 7.5 ( 0.8

Upregulated Proteins protein name

NCBI GI number

Membrane or Membrane-Bound Proteins alpha 1 type III collagen; 4502951 alpha 3 type VI collagen isoform 3 precursor; 17149807 autoantigen p542 3334899 apoptosis inhibitor 5 30583025 Chain A, Synthetic Ubiquitin With Fluoro-Leu At 50 And 67 31615803 Chain D, Tsg101(Uev) Domain In Complex With Ubiquitin 48425523 Carbamoyl phosphate synthetase 2/aspartate transcarbamylase 41351087 cathepsin B 3929733 coproporphyrinogen oxidase 433888 cysteine protease 1890050 4F2 heavy chain antigen 177216 farnesyl-diphosphate famesyltransferase 435677 FBRL_HUMAN; 34 kDa Nucleolar Scleroderma antigen 3399667 putative G-binding protein 3153873 integrin beta 4 binding protein isoform a 4504771 lysosomal proteinase cathepsin B 181178 matrin 3 6563246 methylene tetrahydrofolate dehydrogenase 2 precursor 5729935 Osteoblast-specific factor 2 480007 Opa-interacting protein OIP2 2815604 RAD50 homologue isoform 1 19924129 SP-H antigen 743447 SWI/SNF-related matrix-associated actin-dependent regulator 18606276 scaffold attachment factor B 1213639 Solute carrier family 7 (cationic amino acid transporter) 27503713 thyroid-lupus autoantigen p70 4503841 Unknown (Upregulated Proteins) unnamed protein product 21750187 unnamed protein product 32097 unnamed protein product 7022744 unnamed protein product 21749696 unnamed protein product 10434070 unknown 9789023 hypothetical protein 21739574 hypothetical protein 13276691 hypothetical protein 21739884 Hypothetical protein DKFZp564H2171.1 7512749 Hypothetical protein FLJ10292 15012020 Hypothetical protein FLJ12525 15778927 Human mRNA, complete cds 348239 FTSJ3 protein [Homo sapiens] 12652761 PREDICTED: similar to RIKEN cDNA 0610009D07 50745107 MGC2477 protein 12655159 Others (Upregulated Proteins)a B23 nucleophosmin 825671 chromosome segregation protein smc1 2135244 CPSF6 protein 12653847 DDX17 protein 12653635

2638

Journal of Proteome Research • Vol. 5, No. 10, 2006

SILAC Ratio

4 ( 0.5 3.2 ( 0.3 7.5 ( 0.7 4.9 ( 0.5 3.7 ( 0.4 3.6 ( 0.3 3 ( 0.3 3 ( 0.4 3.4 ( 0.3 3.3 ( 0.3 4 ( 0.4 3.3 ( 0.3 8 ( 0.7 3.1 ( 0.3 3.4 ( 0.3 5 ( 0.5 4.5 ( 0.4 4.3 ( 0.4 12 ( 1.5 4.5 ( 0.5 3.3 ( 0.3 5 ( 0.5 4.5 ( 0.4 3.2 ( 0.3 3.3 ( 0.3 5 ( 0.5 4 ( 0.4 5.8 ( 0.6 6.1 ( 0.6 7.5 ( 0.8 4.4 ( 0.4 3.9 ( 0.4 4.2 ( 0.5 4 ( 0.4 4.1 ( 0.4 7.5 ( 0.8 6.3 ( 0.6 6.4 ( 0.6 7.5 ( 0.8 5 ( 0.5 8 ( 0.9 4.4 ( 0.4 4.5 ( 0.4 6.1 ( 0.6 6.2 ( 0.6 4.2 ( 0.4

research articles

Discovery of Biomarkers of Malignant Breast Cells Using SILAC Table 1 (Continued) Upregulated Proteins protein name

Others (Upregulated DNA-activated protein kinase, catalytic subunit - human U5 snRNP-specific protein, 200-KD hnRNA-binding protein M4 similar to snRNP Sm D1 (Sm-D autoantigen) histone deacetylase 2 Nucleolar protein NOP5/NOP58 protein kinase, DNA-activated PTB-associated splicing factor SmB/B′ autoimmune antigene a

NCBI GI number

SILAC Ratio

1362789 40217847 479852 34877889 4557641 21595782 32140473 38458 36495

6.8 ( 0.7 6.2 ( 0.6 10 ( 1.1 7 ( 0.7 5.2 ( 0.5 7.7 ( 0.8 7 ( 0.8 8 ( 0.9 9 ( 0.8

Proteins)a

Proteins that are not classified as membrane or membrane-bound proteins.

Figure 6. Immunohistochemical staining was used to validate the results of SILAC differential expression measurements. (A) Immunocytochemistry of human normal breast tissues and breast cancer samples using antibodies against membrane alanine aminopeptidase (CD13) and OSF-2. (B) Statistics correlating expression levels of CD13 and OSF-2 with breast cancer. The presence and degree of staining with anti-CD13 antibody was scored for 46 independent breast cancer tissue samples and 20 independent normal breast cancer tissue samples. The presence and degree of staining with anti-OSF-2 antibody was scored for 58 independent breast cancer tissue samples and 19 independent normal breast cancer tissue samples. (C) Normal and breast cancer cells were cultured in light (L) and heavy (H) amino acids, respectively. Membrane alanine aminopeptidase (CD13) was identified as downregulated 10fold, and osteoblast specific factor 2 (OSF2) was identified as upregulated 12-fold by mass spectrometry. Journal of Proteome Research • Vol. 5, No. 10, 2006 2639

research articles expression in the stroma surrounding prostate cancer cells and undetectable expression in most tumor cells.25 Expression of the gene encoding CD13 is controlled by two separate promoters, and the transcription rate varies with cell type.26 Both CD13 transcripts encode the same polypeptide, but the low-abundance CD13 transcripts found in myeloid cells and fibroblasts derive from promoters located 8 kb upstream from the ATG codon. We observed a downregulation of 10-fold in the cancer cells used for this SILAC study. Although we identified CD13 as differentially expressed from a single cell type, it was validated with a total of 46 human breast cancer tissues and 20 normal tissues, correlating the in vitro and in vivo studies. As depicted in Figure 6, panels A and B, the immunoreactivity patterns and statistics obtained with tissue sections from breast carcinomas suggest that the expression of CD13 is significantly decreased in breast carcinomas, which is in agreement with the result obtained with mass spectrometry (Figure 6C). Many upregulated proteins we identified have been reported as having increased expression in tumors. For example, oxygen regulated protein precursor (150 kDa)27,28 and cathepsin B29-31 are proposed to serve as prognostic markers for breast cancer. Expression of 4F2 heavy chain antigen is associated with the progression of oral carcinomas.32,33 Interestingly, a few other upregulated antigens or autoantigens are related to autoimmune disease or Epstein Barr viral infections.34,35 Expression of osteoblast specific factor 2 (OSF-2, periostin), a secreted matrix protein, increased 12-fold in breast carcinomas in this SILAC study (Table 1). OSF-2 is reported to be upregulated in epithelial ovarian tumors and may enhance cell motility and invasiveness.36 Differential expression of OSF-2 was further validated using normal or tumor tissue sections consisting of 59 breast cancer tissues and 19 normal tissues. As depicted in Figure 6, panels A and B, our immunohistochemical results on human tissue arrays suggest that OSF-2 is overexpressed in patients with breast carcinomas, which is in agreement with the result obtained with mass spectrometry (Figure 6C).

Discussion Using a SILAC approach, we have established a general, highthroughput method to characterize changes within a membrane proteome. As demonstrated in Figures 1-5, the protocol is a simple and accurate approach to identify cell surface markers that differentiate tumor and normal cells. In considering the overall sensitivity of the method, we note it employs a crude membrane preparation lacking sophisticated enrichment steps. Among the ca. 1000 protein families identified, however, we found G proteins, G binding proteins, farnesyltransferases, ephrins, integrins, porins, alpha catenin, EGFR precursor, CD44, and solute carrier proteins among others. In short, we identified many of the same low-abundance proteins reported by Celis and co-workers in their recent characterization of the interstitial fluid perfusing the breast tumor microenvironment.37 However, we did not identify 14-3-3 σ, an epithelial-specific marker that is downregulated in breast and many other cancers. Because we were able to identify other isoforms of 14-3-3 with fair coverage (eta, tau, zeta, gamma, and epsilon), we presume this omission stems from the complexity of the sample rather than the sensitivity of the method. Further improvements may therefore be expected if we were to specifically enrich for integral plasma membrane proteins. Zhao and co-workers recently demonstrated a biotinylation procedure that should be entirely compatible with SILAC labeling.38 Moreover, with advanced technology for selection of homogeneous cell popu2640

Journal of Proteome Research • Vol. 5, No. 10, 2006

Liang et al.

lations such as laser capture microscopy, one could precisely isolate 500-800 normal and tumor cells from patient biopsies. It is conceivable then, that such cells may be cultured in either light- or heavy-labeled media for 6-8 doublings, and researchers may use subtractive proteomics to create a robust and accurate analysis of antigenic factors unique to the cell surface of an individual. The further improvements that would be required to make this a reality are not extraordinary. Multidimensional LC separations coupled to MS/MS techniques can increase the number of components identified by 3- to 4-fold.39 Alternatively, preliminary studies applying the same protocol to prepare samples quantified by a combination of off-line capillary HPLC and MALDI-TOF/TOF suggest that one could identify more proteins and enhance the dynamic range of the method using a Nano LC-Probot system (LC Packings, Dionex) coupled to a 4700 proteomics analyzer (Applied Biosystems). Clearly, the complexity and heterogeneity of breast cancer mandates the analysis of more patient samples to pinpoint common biomarkers. Even as we seek to improve our fractionation and software analysis tools, we anticipate translating this powerful technique from the arena of primary cell lines to the characterization of individual patient proteomes, because it holds great promise in biomarker discovery and the validation of drug targets.

Acknowledgment. We thank Dr. Akhilesh Pandey from Johns Hopkins University and Dr. Matthias Mann from the University of Southern Denmark for helpful advice on SILAC. We also thank Dr. John Cottrell from Matrix Science, Ltd. and Dr. Brian Searle from Proteome Software for helpful advice on data analysis. Supporting Information Available: Summary of proteins identified. This material is available free of charge via the Internet at http://pubs.acs.org. References (1) Ong, S. E.; Blagoev, B.; Kratchmarova, I.; Kristensen, D. B.; Steen, H.; Pandey, A.; Mann, M. Mol. Cell. Proteomics 2002, 1, 376386. (2) Sechi, S.; Oda, Y. Curr. Opin. Chem. Biol. 2003, 7, 70-77. (3) Blagoev, B.; Kratchmarova, I.; Ong, S. E.; Nielsen, M.; Foster, L. J.; Mann, M. Nat. Biotechnol. 2003, 21, 315-318. (4) Ong, S. E.; Kratchmarova, I.; Mann, M. J. Proteome Res. 2003, 2, 173-181. (5) Ibarrola, N.; Kalume, D. E.; Gronborg, M.; Iwahori, A.; Pandey, A. Anal. Chem. 2003, 75, 6043-6049. (6) Ibarrola, N.; Molina, H.; Iwahori, A.; Pandey, A. J. Biol. Chem. 2004, 279, 15805-15813. (7) Everley, P. A.; Krijgsveld, J.; Zetter, B. R.; Gygi, S. P. Mol. Cell. Proteomics 2004, 3, 729-735. (8) Ong, S. E.; Mittler, G.; Mann, M. Nat. Methods 2004, 1, 119-126. (9) Gruhler, A.; Olsen, J. V.; Mohammed, S.; Mortensen, P.; Faergeman, N. J.; Mann, M.; Jensen, O. N. Mol. Cell. Proteomics 2005, 4, 310-327. (10) Anderson, J. S.; Lam, Y. W.; Leung, A. K. L.; Ong, S. E.; Lyon, C. E.; Lamond, A. I.; Mann, M. Nature 2005, 433, 77-83. (11) Shui, W.; Liu, Y.; Fan, H.; Bao, H.; Liang, S.; Yang, P.; Chen, X. J. Proteome Res. 2005, 4, 83-90. (12) Zhu, H.; Hunter, T. C.; Pan, S.; Yau, P. M.; Bradbury, E. M.; Chen, X. Anal. Chem. 2002, 74, 1687-1694. (13) Chen, X.; Smith, L. M.; Bradbury, E. M. Anal. Chem. 2000, 72, 1134-1143. (14) Gu, S.; Liu, Z.; Pan, S.; Jiang, Z.; Lu, H.; Amit, O.; Bradbury, E. M.; Hu, C. A.; Chen, X. Mol. Cell. Proteomics 2004, 3, 998-1008. (15) Perkins, D. N.; Pappin, D. J.; Creasy, D. M.; Cottrell, J. S. Electrophoresis 1999, 20, 3551-3567. (16) Nesvizhskii, A. I.; Keller, A.; Kolker, E.; Aebersold, R. Anal. Chem. 2003, 75, 4646-4658. (17) Keller, A., Nesvizhskii, A. I., Kolker, E., and Aebersold R. Anal. Chem. 2002, 74, 5383-5392.

research articles

Discovery of Biomarkers of Malignant Breast Cells Using SILAC (18) Zappacosta, F.; Annan, R. S. Anal. Chem. 2004, 76, 6618-6627. (19) DeSouza, L.; Diehl, G.; Rodrigues, M. J.; Guo, J.; Romaschin, A. D.; Colgan, T. J.; Siu, K. W. J. Proteome Res. 2005, 4, 377-386. (20) Smitherman, A. B.; Mohler, J. L.; Maygarden, S. J.; Ornstein, D. K. J. Urol. 2004, 171, 916-920. (21) Zhang, X.; Zhi, H. Y.; Zhang, J.; Wang, X. O.; Zhou, C. N.; Wu, M.; Sun, Y. T.; Liu, Z. H. Zhonghua Zhongliu Zazhi 2003, 25, 353355. (22) Chetcuti, A.; Margan, S. H.; Russell, P.; Mann, S.; Millar, D. S.; Clark, S. J.; Rogers, J.; Handelsman, D. J.; Dong, Q. Cancer Res. 2001, 61, 6331-6334. (23) Kanoh, Y.; Ohtani, N.; Mashiko, T.; Ohtani, S.; Nishikawa, T.; Egawa, S.; Baba, S.; Ohtani, H. Anticancer Res. 2001, 21, 551556. (24) Christensen, U.; Simonsen, M.; Harrit, N.; Sottrup-Jensen, L. Biochemistry 1989, 28, 9324-9331. (25) Bogenrieder, T.; Finstad, C. L.; Freeman, R. H.; Papandreou, C. N.; Scher, H. I.; Albino, A. P.; Reuter, V. E.; Nanus, D. M. Prostate 1997, 33, 225-232 (26) Shapiro, L. H.; Ashmun, R. A.; Roberts, M. W.; Look, A. T. J. Biol. Chem. 1991, 266, 11999-12007. (27) Tsukamoto, Y.; Kuwabara, K.; Hirota, S., Kawano; K. Yoshikawa, K.; Ozawa, K.; Kobayashi, T.; Yanagi, H.; Stern, D. M.; Tohyama, M.; Kitamura, Y.; Ogawa, S. Lab. Invest. 1998, 78, 699-706. (28) Asahi, H.; Koshida, K.; Hori, O.; Ogawa, S.; Namiki, M. BJU Int. 2002, 90, 462-466.

(29) Castiglioni, T.; Merino, M. J.; Elsner, B.; Lah, T. T.; Sloane, B. F.; Emmert-Buck, M. R. Hum. Pathol. 1994, 25, 857-862. (30) Krepela, E.; Vicar, J.; Cernoch, M. Neoplasma 1989, 36, 41-52. (31) Thomssen, C.; Schmitt, M.; Goretzki, L.; Oppelt, P.; Pache, L.; Dettmar, P.; Janicke, F.; Graeff, H. Clin. Cancer Res. 1995, 1, 741746. (32) Kim do, K.; Ahn, S. G.; Park, J. C.; Kanai, Y.; Endou, H.; Yoon, J. H. Anticancer Res. 2004, 24, 1671-1675. (33) Yoon, J. H.; Kim, Y. B.; Kanai, Y.; Endou, H.; Kim do, K. Anticancer Res. 2003, 23, 3877-3881. (34) Rhodes, G. H.; Valbracht, J. R.; Nguyen, M. D.; Vaughan, J. H. J. Autoimmun. 1997, 10, 447-454. (35) Lapeyre, B.; Mariottini, P.; Mathieu, C.; Ferrer, P.; Amaldi, F.; Amalric, F.; Caizergues-Ferrer, M. Mol. Cell. Biol. 1990, 10, 430434. (36) Gillan, L.; Matei, D.; Fishman, D. A.; Gerbin, C. S.; Karlan, B. Y.; Chang, D. D. Cancer Res. 2002, 62, 5358-5364. (37) Celis, J. E.; Gromoy, P.; Cabezon, T.; Moreira, J. M.; Ambartsumian, N.; Sandelin, K.; Rank, F.; Growmova, I. J. Proteome Res. 2004, 3, 327-344. (38) Zhao, Y.; Zhang, W.; Kho, Y.; Zhao, Y. Anal. Chem. 2004, 76, 18171823. (39) Gu, S.; Du, Y.; Chen, J.; Liu, Z.; Bradbury, E. M.; Hu, C. A.; Chen, X. J. Proteome Res. 2004, 3, 1191-1200.

PR060125O

Journal of Proteome Research • Vol. 5, No. 10, 2006 2641