Proteomics Analysis of Human Coronary Atherosclerotic Plaque

Cardiovascular disease presents significant variations in human populations with respect to the atherosclerotic plaque progression, inflammation, thrombosis, and rupture. To gain a more comprehensive picture of the pathogenic mechanism of atherosclerosis and the variations seen in patients, efficient methods to identify proteins from the normal and diseased arteries need to be developed. To accomplish this goal, we tested the feasibility and efficiency of protein identification by a recently developed method, termed direct tissue proteomics (DTP). We analyzed frozen and paraformaldehyde-fixed archival coronary arteries with the DTP method. We also validated the distinct expression of four proteins by immunohistochemistry. In addition, we demonstrated the compatibility of the DTP method with laser capture microdissection and the possibility of monitoring specific cytokines and growth factors by the absolute quantification of abundance method. Major findings from this feasibility study are that 1) DTP can be used to efficiently identify proteins from paraformaldehyde-fixed, paraffin-embedded, and frozen coronary arteries; 2) approximately twice the number of proteins were identified from the frozen sections when compared with the paraformaldehyde-fixed sections; 3) laser capture microdissection is compatible with DTP; and 4) detection of low abundance cytokines and growth factors in the coronary arteries required selective reaction monitoring experiments coupled to absolute quantification of abundance. The analysis of 35 human coronary atherosclerotic samples allowed identification of a total of 806 proteins. The present study provides the first large scale proteomics map of human coronary atherosclerotic plaques.

Cardiovascular disease presents significant variations in human populations with respect to the atherosclerotic plaque progression, inflammation, thrombosis, and rupture. To gain a more comprehensive picture of the pathogenic mechanism of atherosclerosis and the variations seen in patients, efficient methods to identify proteins from the normal and diseased arteries need to be developed. To accomplish this goal, we tested the feasibility and efficiency of protein identification by a recently developed method, termed direct tissue proteomics (DTP). We analyzed frozen and paraformaldehyde-fixed archival coronary arteries with the DTP method. We also validated the distinct expression of four proteins by immunohistochemistry. In addition, we demonstrated the compatibility of the DTP method with laser capture microdissection and the possibility of monitoring specific cytokines and growth factors by the absolute quantification of abundance method. Major findings from this feasibility study are that 1) DTP can be used to efficiently identify proteins from paraformaldehyde-fixed, paraffin-embedded, and frozen coronary arteries; 2) approximately twice the number of proteins were identified from the frozen sections when compared with the paraformaldehyde-fixed sections; 3) laser capture microdissection is compatible with DTP; and 4) detection of low abundance cytokines and growth factors in the coronary arteries required selective reaction monitoring experiments coupled to absolute quantification of abundance. The analysis of 35 human coronary atherosclerotic samples allowed identification of a total of 806 proteins. The present study provides the first large scale proteomics map of human coronary atherosclerotic plaques.

Molecular & Cellular Proteomics 6:1088 -1102, 2007.
Atherosclerosis is a disease of blood vessels presenting a wide spectrum of pathological manifestations ranging from simple thickening and narrowing of the blood vessels to catastrophic coronary arterial occlusion and myocardial infarction. Approximately 61 million Americans, almost one-quarter of the population, have some form of cardiovascular disease. Acute coronary syndromes and cerebrovascular stroke, the principal components of vascular diseases, are the first and third leading causes of death in the United States, accounting for more than 40% of all deaths (Centers for Disease Control statistics). To develop effective antiatherosclerotic interventions, it is critically important to understand the pathological mechanism of atherosclerotic vascular disease.
Multiple cell types are present in the atherosclerotic plaque, and the proteins from these cells are likely to contribute to the pathogenesis of atherosclerosis. Many proteins expressed in macrophages, vascular smooth muscle cells (SMCs), 1 and inflammatory cells, such as growth factors, connective tissue constituents, pro-and anticoagulants, lipid-associated proteins, metabolic regulators, and tissue enzymes have all been implicated in the complex process of atherogenesis. Therefore, careful analysis of proteins within the atherosclerotic vascular tissue will provide a repertoire of proteins that participate in vascular remodeling and atherogenesis. As a first step, the identification of the human atherosclerotic proteome will serve as a foundation for further studies to generate new hypotheses and to test these hypotheses.
The proteome is defined as all proteins expressed in a cell, tissue, or organism. According to this, proteomics is defined as the systematic analysis of proteins for identification, quan-tification, post-translational modifications, subcellular localization, protein-protein interactions, and enzymatic activities (1). Since the early 1990s, proteomics techniques using mass spectrometry coupled to chromatographic separation have been significantly improved and applied to the study of increasingly complex biological systems (2)(3)(4). The availability of human and other protein databases and bioinformatics tools now allows large scale studies where thousands of proteins can be analyzed simultaneously. In addition, laser capture microdissection (LCM) technology has emerged as a powerful technique to isolate specific areas or cell types within a tissue section, a critical method for refining tissue proteomics profiling (5)(6)(7). Although these proteomics approaches are becoming commonly available to many investigators, their application to human atherosclerosis has not been carried out to date.
In this study, we utilized the recently described paraffinembedded tissue (DTP) method (8) to test the feasibility of protein identification in human coronary vessels with various stages of atherosclerosis. Proteins identified by tandem mass spectrometry were carefully analyzed and classified. Many of these proteins have been implicated previously in atherosclerosis development; however, most of them have not been observed in human coronary atherosclerotic lesions. These proteins include some interesting factors and known regulatory proteins, such as pigment-epithelium derived factor (PEDF), periostin, milk fat globule-EGF-factor 8 (MFG-E8), and annexin I, all of which were validated by in situ immunohistochemistry (IHC). This study explored the most efficient way to analyze protein identification from human coronary arteries and atherosclerotic plaques. Our findings suggest that large scale protein profiling using frozen coronary artery samples represents the most efficient method for sample processing in terms of protein identification. Furthermore we demonstrated that LCM coupled to the shotgun proteomics technique is feasible. We also demonstrated that detection and quantification of low abundance growth factors and cytokines is possible by combining selective reaction monitoring mass spectrometry with AQUA.

Human Coronary Tissue
Archival samples of 35 coronary vessels in paraffin or frozen blocks, obtained from 28 patients undergoing cardiac transplant, were approved for the protein identification study by the Institutional Review Board as deidentified samples. Paraffin blocks (25 vessels) and optimal cutting temperature blocks (10 vessels) consisted of left anterior descending (25 samples), right coronary artery (six samples), and circumflex coronary artery (four samples). The tissue blocks were cut as 5-or 10-m sections and processed for either direct trypsin digestion, LCM analysis followed by direct trypsin digestion, or SDS-PAGE followed by in-gel trypsin digestion.

Histopathological Examination and Tissue Classification
Each paraffin and frozen block was cut in 5-m-thick serial sections. Samples were stained with hematoxylin and eosin (H&E) and for IHC using standard protocols. IHC with Ham-56 (macrophage marker) and ␣-actin (SMC marker) were used to characterize and classify each vessel according to American Heart Association guidelines for histopathological classification. Atherosclerotic lesions were classified into three histological categories: early, intermediate, and advanced.

Human Atherosclerotic Coronary Artery Mass Spectrometry Analysis
Sample Preparation from DTP-Serial tissue sections (5-m thickness) were scraped from slides, and the tissue was collected in microcentrifuge tubes. Samples were suspended and boiled in 30% ACN and 100 mM ammonium bicarbonate (NH 4 HCO 3 ) buffer for 10 min to reverse the cross-linking of proteins induced by formalin fixation (8). Proteins were digested with trypsin (12 ng/l) at 37°C overnight, and supernatant was collected after centrifugation for 10 min at 16,000 ϫ g acceleration. Peptides were dried in SpeedVac, resuspended in buffer A (5% ACN, 0.4% acetic acid, 0.005% heptafluorobutyric acid (HFBA) in water), and analyzed by LC-MS/MS. Sample Preparation from Frozen (Optimal Cutting Temperature) Embedded Tissue (In-gel Trypsin Digestion)-The tissue from serial sections (10-m thickness) was scraped from the slides and collected in microcentrifuge tubes. Samples were either digested and analyzed directly as described above or extracted in RIPA buffer (150 mM NaCl, 10 mM tris, 0.1% SDS, 1% Triton, 1% deoxycholate, 5 mM EDTA, protease inhibitor) and homogenized by ultrasonication. Protein gel electrophoresis (SDS-PAGE) was performed using 100 g of tissue lysate, and protein bands were visualized by staining gels with Coomassie Blue dye. After fixation and destaining, gel bands were cut, and in-gel trypsin digestion was performed as described previously (9). Samples were resuspended in buffer A and analyzed by LC-MS/MS.
Sample Preparation from Tissue Obtained by LCM-The LCM microscope, including the laser system, was purchased from Molecular Machines & Industries (Geneva, Switzerland). Histological information (H&E and IHC) from serial sections was used to identify different regions that were cut and pooled from multiple sections and analyzed to obtain area-specific proteomics information. Tissue sections from paraffin and frozen blocks were processed on special membrane slides with a metal rim. Specific areas were cut using an ultraviolet laser, and the tissue was collected on caps of specially designed microcentrifuge tubes. Samples were suspended in 30% ACN, 100 mM NH 4 HCO 3 buffer; boiled for 10 min; and subjected to in-solution trypsin digestion (12 ng/l at 37°C overnight). Peptides obtained from LCM experiments were resuspended in buffer A and analyzed by LC-MS/MS.

Liquid Chromatography and Tandem Mass Spectrometry
All the samples were analyzed on an LTQ (a two-dimensional ion trap) instrument equipped with a commercial nanospray source (Thermo Finnigan, San Jose, CA). Samples were loaded by a microautosampler (Famos, LC Packings, Sunnyvale, CA) onto an 11-cm ϫ 100-m fused silica capillary column packed with reverse C 18 material (5-m, 100-Å Magic beads, Michrom Bioresources, Auburn, CA). The solvent system was delivered by an HP1100 pump (Agilent Technologies, Palo Alto, CA). Peptides were eluted with a gradient from 100% buffer A to 80% buffer B (0.4% acetic acid, 0.005% HFBA in ACN) in 105 min. Each survey scan was followed by five MS/MS scans of the most intense ions. Dynamic exclusion features were enabled to maximize the fragmentation of less abundant peptide ions. Sample loading, solvent delivery, and scan function were controlled by the Xcalibur software (Thermo Finnigan).

Bioinformatics Data Processing and Analysis
A total of 327 LC-MS/MS runs were searched against a nonredundant human protein database concatenated to its reversed database (Advanced Biomedical Computing Center, December 1, 2004) using the SEQUEST algorithm (10). Searching parameters included mass tolerance of Ϯ1.5 Da, trypsin specificity, and differential methionine oxidation. SEQUEST results were automatically submitted to PeptideProphet for probability calculation. The generated data set was filtered using INTERACT based on the following criteria: ⌬Cn Ն 0.1, Xcorr: ϩ1, Ն1.9; ϩ2, Ն2.2; ϩ3, Ն3.7. Those proteins that were identified based on the presence of one peptide only were excluded during data filtering and analysis using the INTERACT software. These criteria were adjusted and selected based on reversed database search and analysis of false positive rate (the resultant false positive rate for proteins was 2.2%).

Detection and Quantification of Stromal Cell-derived Factor 1 ␣ (SDF1-␣) and Growth Factors in Human Atherosclerotic
Coronary Artery Samples SDF1-␣ and growth factors in the atherosclerotic human coronary arteries were detected and quantified by using a technique termed AQUA (11). The AQUA methodology consists of selection of a suitable tryptic peptide as a unique identifier of the protein of interest and the addition of the synthetic heavy isotope-labeled counterpart as an internal standard for quantification by mass spectrometry (see Fig.  3A). The methodology was implemented as described previously (12). The standard peptides and the product ions chosen for quantification of SDF1-␣ and the growth factors by liquid chromatography multiple reaction monitoring (MRM) are detailed in Supplemental Table 1. Amine-protected (Fmoc (N-(9-fluorenyl)methoxycarbonyl)) and heavy isotope ( 13 C 5 , 15 N)-labeled amino acids (valine and proline) were procured from Isotec (St. Louis, MO). The standard peptides were synthesized at the Rockefeller University Proteomics Resource Center. Isotope-labeled AQUA standard peptides were synthesized using an Intavis MultiPep parallel peptide synthesizer (Intavis, Koln, Germany) at the 5-mol scale. An optimum chromatographic method was developed for acquiring the data in a multiplexed manner. Proteins from frozen sections of atherosclerotic human coronary arteries were extracted using the RIPA buffer supplemented with 0.2% SDS. 100-g quantities of the frozen tissue protein extract were resolved by SDS-PAGE. Gel bands below 30 kDa were excised and subjected to in-gel trypsin digestion in the presence of ϳ1000 fmol of the AQUA standard peptides. The solvent used for dissolving the tryptic peptide mixtures was supplemented with 0.01% hydrogen peroxide and 5% formic acid to oxidize the methionine residues to their sulfoxide version.

Immunohistochemistry
For histological examination and classification of the tissue samples, H&E staining was performed according to standard procedures. The identification of specific cell populations within the plaque (i.e. macrophages and SMCs) was performed by antigen-specific immunostaining. IHC was performed with the use of the VECTASTAIN ABC system (Vector Laboratories). Sections were sequentially blocked with 3% H 2 O 2 , 5% normal serum (Vector Laboratories) matching the host of the secondary antibody, and avidin/biotin blocking solution (Vector Laboratories) for 30 min at room temperature. Annexin I staining was performed with antigen retrieval in 10 mM sodium citrate, 0.05% Tween 20, pH 6 buffer. Primary antibodies against human Ham-56 (1:50, mouse monoclonal) from DakoCytomation, ␣-actin (1:1000, mouse monoclonal) from Sigma, PEDF (1:200, mouse monoclonal) from Chemicon, periostin (1:100 rabbit polyclonal) from Abcam, MFG-E8 (1:50, monoclonal) from R&D Systems, and annexin I (1:100, rabbit polyclonal) from Santa Cruz Biotechnology were used for immunostaining. Secondary antibodies (horse anti-mouse and goat anti-rabbit) were used at 1:200 dilutions. Sections were developed by DAB substrate (Vector Laboratories) and counterstained with hematoxylin for microscopic visualization.

Apoptosis Assay
To evaluate the extent of apoptosis in the plaque, TUNEL assay was performed by using an in situ cell death detection kit (Roche Applied Science). We followed the manufacturer's protocol with minor modifications. Briefly tissue sections were deparaffinized and blocked for endogenous peroxidase activity (3% hydrogen peroxide in water) for 30 min at room temperature. Samples were treated with proteinase K (20 g/ml) for 30 min at room temperature. Positive controls were treated with DNase I (1 g/ml) for 10 min at room temperature. Terminal DNA labeling was performed at 37°C for 1 h with a reaction mixture containing fluorescein-conjugated nucleotides and terminal deoxynucleotidyltransferase enzyme. As a negative control, the enzyme was excluded during incubation. Samples were blocked with 2% BSA for 30 min at room temperature and treated with a horseradish peroxidase-conjugated antibody against fluorescein deoxynucleotides (30 min at 37°C). Finally the reaction was visualized with a DAB system.

Proteome Analysis of Human Atherosclerotic Coronary
Arteries-To gain insights into the biology of atherosclerosis, we applied a recently developed methodology, termed DTP, that allows protein identification directly from formalin-fixed, paraffin-embedded tissue samples (8). This method was developed and optimized using commercially available prostate cancer paraffin-embedded tissue arrays. A number of critical issues could not be addressed in the first study: 1) the efficiency of protein identification in paraformaldehyde-fixed, paraffin-embedded human coronary artery samples when compared with fresh frozen samples, 2) the efficiency of protein identification when the DTP method is compared with SDS-PAGE gel separation and in-gel digestion method, and 3) compatibility of LCM with the DTP method. Moreover given the limited availability of human coronary artery tissue samples in either paraformaldehyde-fixed or frozen preparations, we wanted to assess the best method to analyze these samples to maximize protein identification using LC-MS/MS tech- nologies. In addition, because every tissue type contains various quantities of highly abundant proteins, extracellular matrix material, and lipids, we tested the feasibility and efficiency of protein identification in human coronary atherosclerotic lesions.
The overall strategy of this study is outlined in Fig. 1. Tissues from two different sources (frozen or paraffin tissue blocks) were processed in three ways (Fig. 1, A-C) for protein extraction and trypsin proteolysis as described under "Experimental Procedures": A, total section deparaffinization and in-solution trypsin digestion; B, LCM of specific target areas in frozen or paraffin sections; and C, SDS-PAGE protein separation and in-gel trypsin digestion of frozen tissues. Tryptic peptides were separated by liquid chromatography and fragmented in the mass spectrometer by the process of CID (13). The resulting mass spectra were interpreted automatically by running the SEQUEST algorithm against the human protein database containing Ϸ56,709 entries. Combined data obtained from the three different sample treatments gave a total number of 2211 unique proteins including the identification of 1405 single peptide-matched proteins. After excluding the proteins identified based on single peptides we identified 806 unique proteins with high confidence.
We next compared the efficiency of protein identification from the three different strategies used (Fig. 1, A-C). The comparison of the number of identified proteins between similar quantities of starting coronary artery sections from frozen or paraformaldehyde-fixed tissues revealed that more than double the number of proteins was identified from fresh frozen samples (Table I). Although we utilized a total of 125-m thickness of the coronary artery sections with similar lesion types, we identified 225 proteins from the paraformaldehyde-fixed samples as multiple peptide hits, whereas 558 proteins were identified from the frozen sections. Supplemental Tables 2 and 3 list all of the identified proteins, the number of peptides and the independently identified peptides by tandem mass spectrometry, and other relevant information important for peptide and protein identification. From these results, we conclude that our strategy of proteomics profiling from frozen sections yielded significantly better proteome coverage because it identified at least twice the number of proteins identified from paraffin-embedded sections.
In addition, we tested the feasibility of protein identification from the LCM samples by mass spectrometry. In this procedure, tissue sections were first mounted on a polymer membrane followed by weak hematoxylin staining, laser dissection, and collection of the dissected tissue using commercially available microcentrifuge caps with a sticky surface. We were concerned about the efficiency of protein extraction from the polymer membranes as well as the extraction of undesired hydrophobic polymer ions that could potentially interfere with LC-MS/MS. Another reason for performing LCM was to possibly overcome the presence of high abundance extracellular matrix proteins that could overwhelm the low abundance proteins. As shown in Fig. 2A, using the Molecular Machines & Industries system, we were able to efficiently microdissect and capture the whole intima (top row), the remaining tunica media containing mainly vascular SMCs (middle row), and the tunica adventitia (bottom row). The flow diagram in Fig. 2B shows the general procedure by which 15 samples were analyzed. Using this procedure, we identified 495 multihit proteins, indicating that LCM is compatible with the DTP strategy using LC-MS/MS. Because the laser-dissected tissue sections were selected from regions where macrophages, endothelial cells, or vascular SMCs were enriched, we could not directly compare the protein identification obtained by this approach with the results obtained with the previously described strategies. The proteins identified by LCM-selected areas are available as supplemental data (Supplemental Table  4, intima hyperplasia; Supplemental Table 5, media; Supplemental Table 6, adventitia).

Feasibility of Detection of Selected Cytokines/Growth Factors in Human Coronary
Atherosclerotic Plaques-When we examined the total list of 806 identified proteins from human atherosclerotic plaques, we did not find known growth factors and cytokines attributed previously to the migration and proliferation of vascular and inflammatory cells. Thus, we wanted to test whether we could detect and quantify some of these important mediators by using more selective mass spectrometry and proteomics techniques. One such technique, termed A summary of the numbers of the proteins identified during the application of the three strategies used for tissue proteomics analysis is shown. The first and second columns indicate the tissue source (frozen or paraffin blocks) and the approach used for tissue processing. a Refers to the number of the unique identified proteins in each of the approaches followed (see Fig. 1, A, B, and C). b Indicates the number of proteins that has been determined based on the identification of only one peptide. c Represents the proteins identified with two or more peptides (c ϭ a Ϫ b). d Indicates the number of unique peptides identified from each of the different approaches.
AQUA, combined with MRM has been used in the laboratory previously (12). In this technique, a standard tryptic peptide representing the protein of interest is synthesized in the presence of heavy isotope-labeled amino acids, and the absolute quantity of the endogenous peptide is measured by direct comparison of the chromatographic peak area with that of the standard (Fig. 3B). We synthesized four heavy isotope-labeled standard peptides for the following cytokines/growth factors: transforming growth factor ␤ (TGF-␤), basic fibroblast growth factor, platelet-derived growth factor ␤ (PDGF-␤), and SDF1-␣ (Supplemental Table 1). As shown in Fig. 3C, we could successfully detect SDF1-␣ and TGF-␤ from whole coronary artery tissue lysate after 1D gel protein resolution.
These results indicate that the combination of efficient protein extraction (0.2% SDS buffer), reduction of sample complexity (1D gel electrophoresis), and a more selective and sensitive technique allows the detection of low abundance cytokines present in the coronary tissue. This strategy represents an additional tool for the specific qualitative and quantitative analysis of these types of factors that could not be identified by the shotgun proteomics approaches.
Analysis of Proteins Likely to Be Involved in Atherogenesis: Known Factors and Other Proteins Not Implicated Previously in the Context of the Disease-Proteomics analyses of human coronary atheroma resulted in the identification of many different types of proteins. Some of these are known to participate in atherosclerosis development, whereas others might be implicated in the progression of the disease but have not been identified in the atheroma by the currently available methods. Our goal was to validate a reasonable number of proteins identified by the proteomics analysis as well as to gain biological insight into the potential function of these proteins in atherogenesis. Toward this goal, we carefully analyzed our entire list of 806 proteins in the context of the published literature and selected four groups of proteins for further analysis: 1) extracellular matrix proteins, 2) lipid-binding and metabolism-associated proteins, 3) inflammation-related proteins, and 4) apoptotic-cell phagocytic ligands and receptors. Supplemental Table 7 shows the proteins selected to be included in these categories from the main list of 806 identified proteins. Although some of the factors are not directly related to atherosclerosis, evidence of their known function from the literature and their presence in the atherosclerotic lesions could be an indication that they may have a functional role in atherogenesis. To validate mass spectrometry-based protein identification, IHC was performed using 5-22 human coronary artery samples. The specimens were selected for the presence of regions that demonstrated histological hallmarks of human coronary atherosclerotic plaques, including macrophage foam cells, intima SMCs, fibrous caps, inflammatory angiogenic regions, cholesterol clefts with necrotic core, adventitial inflammation, and medial atrophy. Most of the coronary arteries selected contained several of these characteristics of atherosclerosis and, therefore, allowed us to validate protein expression from multiple coronary artery samples. Four proteins were selected from the main list of 806 proteins for further validation using immunohistochemical analysis.

Validation of a Proinflammatory Factor and an Extracellular Matrix Protein Expression in Atherosclerotic Coronary
Arteries-Macrophage infiltration and inflammation are character-istic features of atherosclerotic lesion progression. To validate protein detection and identification by mass spectrometry, we selected one of the proteins involved in inflammation/angiogenesis and explored its expression by IHC. PEDF is a member of the serine/protease inhibitor family that has a known neurotrophic activity (14). We selected PEDF as a target for validation and examined the expression of this protein in the context of the atherosclerotic plaque. To define the cells in the coronary arteries, we utilized an anti-␣-smooth muscle actin antibody for vascular SMCs and a Ham-56 antibody for detecting macrophage foam cells and endothelial cells. As shown in Fig. 4, A-C, low power magnification of a human Fig. 3. Detection and quantification of the cytokines/growth factors in the atherosclerotic human coronary arteries. Schematic diagram of the AQUA methodology used for the detection of selected cytokines/growth factors is shown in A and B. Synthesized stable isotope-labeled peptides act as internal standards for the quantification of endogenous peptides derived from the tissue sections. 100-g quantities of frozen tissue protein extract were used for the detection of the cytokines and growth factors. Multiple reaction monitoring chromatograms corresponding to the standard peptides (red) and the endogenous peptides (blue) selected for quantifying SDF1-␣, unprocessed TGF-␤1, basic fibroblast growth factor (FGF), and PDGF-␤ are shown in C. Absolute quantities of SDF1-␣ and unprocessed TGF-␤1 were found to be in low femtomole amounts, and fibroblast growth factor and PDGF-␤ were below the detection limit. Quantities of SDF1-␣ and TGF-␤1 were found to be 27.2 and 167.6 pg, respectively, in 100 g of coronary artery lysate. coronary atherosclerotic plaque stained for H&E (A) showed strong positive staining for PEDF in macrophage-derived foam cells (C). Higher power magnification of adjacent sections stained for PEDF and macrophages revealed that lipidloaded macrophage foam cells are strongly positive for PEDF (Fig. 4, AЈ-CЈ). PEDF antibody also stained the inflammatory angiogenic areas present in the intima (Fig. 5, A and B, and Supplemental Fig. S1). Interestingly staining for PEDF was very weak or absent in coronary arteries with diffused intima thickening, indicating that medial SMCs from normal coronary arteries have very little or no PEDF expression (Supplemental Fig. S2, A-D) and that the expression of the factor is associated with the infiltration of the intima with inflammatory cells and the development of intima hyperplasia. In support of this idea, anti-PEDF antibody stained strongly the intima of preatheroma lesions that were also positive for Ham-56 staining (Supplemental Fig. S3). Positive staining was also observed in highly inflammatory areas within the adventitia that were also positive for Ham-56 (Fig. 5, C, D, E, and F). The specificity of anti-PEDF antibody was tested by omitting the primary antibody and by including the isotype-matched non-immune IgG as controls that do not show staining (Supplemental Figs. S1, C and D, and S2, E and F). These IHC images are representative of 22 stained atheroma lesions for PEDF expression. Of the total 22 cases studied by IHC, 15 (68%) showed variable levels of macrophage and inflammatory infiltration coincident with Ham-56 and PEDF staining. In contrast, 7 of 22 (32%), most demonstrating histological features of diffuse, non-inflammatory intima thickening, were all negative for PEDF. These results indicate that proteomics identification of PEDF in coronary arteries can be validated by using IHC.
Extracellular matrix proteins, including proteoglycans as well as glycoproteins, are known to be expressed at increased levels during the progression of atherosclerosis. Proteomics profiling with mass spectrometry allowed the identification of many of these factors (listed in Supplemental Table 7), including a protein identified as periostin. Although periostin has been shown to be induced in a rat carotid balloon injury and restenosis model (15), the expression of periostin in human atherosclerosis has not been examined to date. This prompted us to analyze periostin expression together with cell identification markers in human coronary atherosclerotic plaques. As shown in Fig. 6, periostin is distributed mainly in the intima as it can be confirmed based on Ham-56 and ␣-actin staining. As shown in Supplemental Fig. S4, we were able to confirm this pattern of staining; anti-periostin antibody stains the intima of an intermediate lesion with the atheroma in progress (foam cells, large amount of extracellular matrix deposition, and an incipient necrotic core). This sample shows how periostin expression decreased toward the area of intima-media transition (Supplemental Fig. S4, B and C, label decreased in the direction of the dashed arrows). On the other hand, advanced lesions presenting calcification show that anti-periostin antibody stains areas adjacent to the calcified core and the atrophied media (Supplemental Fig. S5). Negative controls for periostin staining where the primary antibody was excluded are shown in Supplemental Figs. S4D and S5D. Periostin expression was tested in 11 samples, many of them with advanced complicated lesions with different levels of calcifications. Eight of the samples (73%) were positive for periostin staining. The three negative samples (27%) presented diffuse intima thickening without excess core calcifications and advanced inflammatory areas.
Engulfment Ligands Are Present in Human Atherosclerotic Plaques-Apoptosis of vascular and inflammatory cells has been implicated in the pathophysiology of atherosclerosis (16) particularly during the rupture of the atherosclerotic plaque (17,18). It has been proposed that during atheroma progression endothelial cells, SMCs, and macrophages are exposed to factors that induce apoptosis, turning the plaque from a stable lesion to a vulnerable, unstable, and inflammatory tissue. As part of the apoptosis process, phagocytosis of apoptotic cells is essential to clear cell debris and avoid excess production and secretion of proinflammatory cytokines (19,

20)
. Although the contribution of macrophages seems to be critical in animal models of atheroma progression (21), the protective or contributory role for phagocytosis of apoptotic cells during atherosclerosis progression has not been well defined. Supporting data suggest that impaired phagocytosis of apoptotic cells promotes disease development (22,23). In this context we decided to analyze the expression of proteins specifically associated with the phagocytosis of apoptotic cells, including phagocytic receptors as well as bridging molecules required for engulfment. A list of these proteins is available in Supplemental Table 7. MFG-E8 and annexin I were selected to validate their mass spectrometry identification by using IHC. In Fig. 7, vessel staining by MFG-E8 antibody demonstrates that the protein is mainly expressed in the intima, consistent with areas that also stain for the macrophage marker Ham-56 (Fig. 7, A and B). This staining pattern is better appreciated at higher magnification in Fig. 7, D and E. The MFG-E8 antibody also stains areas that implicate migratory SMCs in the tunica intima. Supporting this, Fig. 7 (F and G) shows an area in the intima clearly positive for both MFG-E8 and ␣-actin staining, denoting the presence of SMCs in the MFG-E8-positive area. Overall a subset of macro-phages that are positive for Ham-56 are also positive for MFG-E8, whereas there is a macrophage population that is clearly negative for MFG-E8 (Fig. 7, D and E). Similar results were also found for vascular SMCs when serial sections were analyzed for MFG-E8 and ␣-actin (Fig. 7, F and G). The MFG-E8 expression profile was confirmed by staining five different coronary arteries. One representative example of MFG-E8 staining is shown in Supplemental Fig. S6. We found that this particular sample represents an advanced and complex lesion where the lumen has been occluded and later reorganized. Therefore, it was possible to test multiple areas rich in SMCs as well as foam cells especially in the disrupted intima. MFG-E8 antibody stains foam cells in the area surrounding the remaining lumen that are clearly negative for ␣-actin (Supplemental Fig. S7, C and D). Among the five different coronary arteries that were stained for MFG-E8, four of them (80%) are clearly positive, whereas one sample presenting diffuse intima thickening showed no staining.
In addition to MFG-E8, we also validated annexin I expression in the atherosclerotic plaque by IHC. Fig. 8, A and  shows a representative sample with anti-annexin I antibody staining lipid-loaded macrophages (foam cells) that are also positive for Ham-56 staining. The red arrows indicate the same cell in adjacent sections positive for both Ham-56 and annexin I. To further characterize the role of annexin I, we measured apoptotic nuclei by TUNEL staining in the plaque and evaluated whether the positive areas coincide with annexin I expression. In this analysis, we observed that apoptosis (Fig. 8, G and H) is occurring in annexin I-positive areas (Fig. 8, D and F) coinciding with foam cell-enriched areas within hyperplastic intima (Ham-56-positive areas; Fig. 8, C  and E). In Fig. 8H shows how apoptotic cells (black arrows, brown nuclei) and normal foam cells (red arrows, blue nuclei) are distributed within the same area. These results indicate that annexin I is expressed in macrophage foam cells in the human atherosclerotic plaques. Furthermore we found that annexin I-positive areas also contained TUNEL-positive cells (Fig. 8, G and H). Annexin I expression was evaluated in 12 different samples, many of them showing foam cells and inflammation in the tunica adventitia. Seven of the 12 (58%) samples showed positive staining in the tunica intima and adventitia for annexin I. DISCUSSION Despite significant advancement in the treatment of coronary and cerebrovascular diseases, comprehensive understanding of the pathologic mechanisms is still lacking. To gain more comprehensive understanding of atherosclerosis progression, it is critical to better define the expressed proteins in the vessel wall. Proteins are the major effectors of most biological processes and are the most suitable molecules for use as biomarkers and risk factors as well as targets for disease treatment. Profiling of protein expression from pathological tissues provides a rough survey of the pathological state. Because proteins are the key effectors of function, the ability to efficiently utilize mass spectrometry to identify thousands of proteins in a single tissue and/or condition is extremely valuable. The major focus of this study was to identify proteins expressed within atherosclerotic lesions of coronary arteries that can be used as a platform for a better understanding of human atherosclerotic plaque development and progression. In the present study, we present a large scale protein profile of human atherosclerotic coronary arteries. The feasibility of DTP analysis was shown recently for prostate cancer (8). Despite the challenges of the multiple cell types and components of human atheroma lesions, we were able to show that DTP analysis is applicable to atherosclerotic coronary arteries even from archived paraffin blocks. Extracting proteins with a buffer containing SDS and separating the extracted proteins by 1D gel electrophoresis improved the number of identified proteins. In addition, when we utilized atherosclerotic coronary arteries preserved as frozen blocks instead of paraffin blocks, we observed that approximately twice the number of proteins can be identified by applying the DTP method (data not shown). Furthermore using the frozen samples in combination with SDS-PAGE, we were able to identify 558 individual proteins versus 225 proteins from the DTP paraffin block analysis, concluding that the analysis of frozen samples by the proposed methodology significantly increased the number of identified proteins. The absence of highly cross-linked proteins that limit extraction is a likely possibility for the improvement in protein identification seen in the frozen sections. In addition, resolving proteins by gel electrophoresis reduced the complexity of the samples, thus increasing protein identification by mass spectrometry.
Although we identified 806 proteins with high confidence, we did not identify cytokines and growth factors, which are expected to be present in lesions, even when we considered the single peptide hit proteins. Vascular endothelial growth factor (VEGF) as well as the VEGF receptors, fibroblast growth factor 2, vascular cell adhesion molecule, intercellular adhesion molecule, tumor necrosis factor ␣, TGF-␤, SDF1-␣, interleukin 6, and other growth factors and cytokines could not be identified directly by this method. The absence of the above mentioned factors could be explained by one of the following reasons. First, abundance of these key molecules is likely to be much lower than the connective tissue proteins. Thus, there may be improvements that can be made for the identification of molecules present in low abundance. Second, many of these cytokines and receptors are post-translationally modified, most notably with advanced glycosylation, which may impact the peptide pool available for identification especially if a high threshold is set for multiple peptide identification. Third, most of these inflammatory cytokines are involved in plaque inflammation, instability, and rupture, whereas the coronary arteries in the present study were typically pathologically early, advanced stable, or healed plaques with organized thrombus. As none of our tissues apparently represent unstable plaques or acute rupture, the possibility of finding these cytokines at very high levels may be low. Finally evidence from our laboratory has shown that repeated analysis of the same sample significantly increased the total number of identified proteins. 2 Based on this evidence, it is clear that the repeated analysis of each of the samples, by running them several times, may significantly improve total proteomics coverage for each sample and may lead to cytokine and growth factor detection. To test whether these cytokines and growth factors were present in the sample, we decided to apply MRM coupled to AQUA. By applying this selective methodology, we were able to detect and quantify TGF-␤ and SDF1-␣, confirming that these factors are present in the vessel wall but are undetectable by the less sensitive DTP method. These results suggest that MRM in association with AQUA is more sensitive for the targeted identification and quantification of low abundance proteins in histological tissue sections.
As we expected based on the literature, by mass spectrometry-based sequencing, we were able to confirm the presence of proteins such as extracellular matrix proteins that are involved in atherosclerosis progression. The increased expression of these proteins is associated with the proliferation of SMCs, a phenotypic change, and the migration from the tunica media to the intima (24 -27). On the other hand, the overexpression of this group of proteins represents a positive feedback on the proliferation, migration, and changes in SMC phenotype that occurs during the atherosclerosis progression (27). The other group of proteins that we were able to confirm in the atherosclerotic vessel wall is the one represented by lipoproteins (Supplemental Table 7). This was also expected because modified lipoproteins accumulate in atherosclerotic lesions, a phenomenon partially due to the increased levels of extracellular matrix proteins (28,29). The identification of proteins that are known to have an active role during atherosclerosis development validates our method and supports the utilization of this type of approach for the identification of novel proteins involved in the development of the disease.
Among the selected proteins from the general list (Supplemental Table 7), PEDF, periostin, MFG-E8, and annexin I were validated by IHC. While this manuscript was in preparation PEDF expression was reported in the atherosclerotic plaque in the aorta and coronary arteries (30). Mass spectrometry detection and IHC staining observed in our study further confirmed the presence of PEDF in the human plaque, suggesting that the factor might play a role in the evolution of the disease. Supporting this idea, it was reported that PEDF is expressed in vitro in a model of differentiated macrophages when these cells are exposed to oxidized low density lipoprotein (31). It has also been shown that PEDF is expressed during embryogenesis particularly in association with the developing bone matrix (32). In the same study the authors showed that osteoblasts and to a lesser extent osteoclasts express and secret PEDF (32). This evidence suggests that PEDF could be participating in extracellular matrix remodeling, and by doing this, it might be involved in controlling calcification, cell migration, and/or angiogenesis. On the other hand we found PEDF expression predominantly in two areas: macrophage-derived foam cell-rich areas and inflammatory areas in the intima and in the adventitia. This indicates a probable role of PEDF in inflammatory cells migration and proliferation. In support of this idea, exposure to PEDF induces the release of proinflammatory factors such as interleukin 1␤; interleukin 6; macrophage inflammatory proteins 1, 2, and 3 ␣; and tumor necrosis factor ␣ in isolated microglia, astrocytes, and cerebellar granule neurons (33)(34)(35)(36). Similarly the presence of PEDF in the inflammatory areas of human coronary atherosclerotic plaques suggests that this factor may have an autocrine and/or paracrine effect on inflammatory cells of the plaques. Although all this evidence strongly suggests that PEDF is involved in the atherosclerotic process, further experiments in animal models are needed to unravel the specific role of this factor in atherogenesis.
From the group of extracellular matrix proteins, periostin was selected for further characterization. The connection between periostin and atherosclerosis is supported by in vivo experiments. Lindner et al. (15) showed that in the rat carotid balloon injury periostin mRNA is one of the most strongly up-regulated factors in SMCs during neointima formation. In addition, it was shown that overexpression of this protein induces cell migration, and this effect is blocked by periostin antibody pretreatment. On the other hand it has been shown that periostin is up-regulated in rat lung and in isolated pulmonary arterial SMCs in response to hypoxic stress (37), a situation likely to occur during atherosclerosis progression. Finally periostin expression has been associated with angiogenesis, and its expression is increased in human breast cancer (38). This study shows that different cell lines overexpressing periostin induce tumor growth and angiogenesis of xenografts in mice (38). In addition, in a co-culture system of tumor cells overexpressing periostin and human microvascular endothelial cells, it was shown that the protein induces cell migration and proliferation. Supporting its role in angiogenesis, the authors showed that exposure to periostin increases VEGF receptor expression in human microvascular endothelial cells (38). Although these reports implicate periostin in pathological conditions in vivo, no information is available regarding periostin expression in human coronary atherosclerotic plaque. In the present study, we showed periostin expression in several different types of lesions. Periostin has been identified extensively by mass spectrometry in about 90% of the analyzed samples with 40% coverage based on the percentage of identified amino acids. From our IHC data, we have confirmed the presence of periostin in human atherosclerotic plaques specifically in the intima, preatheroma regions, and extracellular matrix of more advanced lesions presenting calcification. However, published data (15) suggest that periostin may act as a potent migratory stimulus in the neointima; our analysis of human lesions validates the presence and association of periostin in human atherosclerosis development.
Finally we also analyzed two proteins related to the phagocytosis of apoptotic cells present in the plaque. It is known that apoptosis occurs during atherosclerosis development (16,39,40), and it has been associated with the instability of the plaque (18,41,42). On the other hand, phagocytosis of apoptotic cells is a critical aspect of controlling inflammation that greatly contributes to tissue homeostasis. By mass spectrometry-based protein identification we found several proteins that have been shown to be involved in the phagocytosis of apoptotic cells. We chose to validate two of the phagocytosis ligands, MFG-E8 and annexin I, by IHC. We found MFG-E8 expression by IHC in the atherosclerotic plaque. An anti-MFG-E8 antibody stained strongly the tunica intima. Interestingly MFG-E8 is also expressed strongly in the subset of SMCs in the intima, indicating that not only macrophages express this molecule ( Fig. 7 and Supplemental Fig. S6). In regard to this, Geng et al. (39) have shown that SMCs in the intima of atherosclerotic samples express Fas and die by apoptosis. During apoptotic cell engulfment, MFG-E8 is secreted by macrophages and binds to the phosphatidylserine at the surface of apoptotic cells. Assuming this is the ongoing sequence of events, we think that the MFG-E8 seen in the plaque is attached to the surface of cells that have died by apoptosis and need to be removed by phagocytosis. Although the role of MFG-E8 in the phagocytosis of apoptotic cells has been well documented (44), whether this protein is involved in atherosclerosis needs to be further investigated.
Our results suggest that annexin I is expressed in macrophages present in the tunica intima. Furthermore these macrophages present a foam cell phenotype, and they are posi-tive for Ham-56 antibody staining. There is strong evidence demonstrating the role of annexin I in the phagocytosis of apoptotic cells (45). During apoptosis, cells expose annexin I at the surface, and the molecule acts as an "eat me signal" that contributes toward recognition and subsequent phagocytosis of the apoptotic cells (45). On the other hand, it has been shown that annexin I can be expressed by macrophages during apoptotic T lymphocyte clearance (46). Although these studies might explain the presence of annexin I in the plaque, we cannot rule out the possibility that the cross-reaction of the foam cell with anti-annexin I antibody is due to the contribution from the phagocytosis of apoptotic cells. In an attempt to clarify this, we analyzed apoptosis in the plaque. We found that Ham-56-positive areas (macrophages) presenting foam cells are also positive for annexin I staining (Fig. 8, C, D,  E, and F). In addition, the TUNEL assay showed high specificity for overlapping areas in serial sections. These results support the idea that annexin I participates in the engulfment of apoptotic bodies present in the plaque. The presence of two known engulfment ligands in the plaque suggests that engulfment of apoptotic cells is likely to be a factor involved in the development and progression of atherosclerosis. This is the first study providing large scale protein identification of human coronary arteries and coronary atherosclerotic plaques that is freely available in a searchable format for the larger scientific community. We performed the present feasibility study to assess the practical limitations associated with analysis of coronary arteries and atherosclerotic plaques. Our study demonstrates that we were able to identify a large number of high abundance proteins expressed in the human coronary arteries. Four of these biologically interesting proteins were validated by IHC. Furthermore we were able to combine the LCM method with the DTP method, paving the way for future investigations where the proteome from a specific cell type can be enriched and identified. Moreover using the AQUA method, we were able to demonstrate the presence of low abundance cytokine/growth factor in the human coronary arteries. We anticipate that the proteomics methodologies we have utilized will be embraced in the future for detection of key protein molecules in clinically relevant human tissues. This information will be useful for basic understanding of disease processes as well as for clinical applications such as diagnosis and early detection of pathological conditions. Biology, Center for Vascular Biology, University of Connecticut School of Medicine, 263 Farmington Ave., Farmington, CT 06030. Tel.: 860-679-2444; Fax: 860-679-1201; E-mail: han@nso.uchc.edu.