|Home | About | Journals | Submit | Contact Us | Français|
The vascular extracellular matrix (ECM) is essential for the structural integrity of the vessel wall and also serves as a substrate for the binding and retention of secreted products of vascular cells as well as molecules coming from the circulation. Although proteomics has been previously applied to vascular tissues, few studies have specifically targeted the vascular ECM and its associated proteins. Thus, its detailed composition remains to be characterized. In this study, we describe a methodology for the extraction of extracellular proteins from human aortas and their identification by proteomics. The approach is based on (a) effective decellularization to enrich for scarce extracellular proteins, (b) successful solubilization and deglycosylation of ECM proteins, and (c) relative estimation of protein abundance using spectral counting. Our three-step extraction approach resulted in the identification of 103 extracellular proteins of which one-third have never been reported in the proteomics literature of vascular tissues. In particular, three glycoproteins (podocan, sclerostin, and agrin) were identified for the first time in human aortas at the protein level. We also identified extracellular adipocyte enhancer-binding protein 1, the cartilage glycoprotein asporin, and a previously hypothetical protein, retinal pigment epithelium (RPE) spondin. Moreover, our methodology allowed us to screen for proteolysis in the aortic samples based on the identification of proteolytic enzymes and their corresponding degradation products. For instance, we were able to detect matrix metalloproteinase-9 by mass spectrometry and relate its presence to degradation of fibronectin in a clinical specimen. We expect this proteomics methodology to further our understanding of the composition of the vascular extracellular environment, shed light on ECM remodeling and degradation, and provide insights into important pathological processes, such as plaque rupture, aneurysm formation, and restenosis.
Vascular cells, in particular vascular smooth muscle cells, produce and maintain a complex meshwork of ECM.1 The ECM is not only the scaffold for the anchorage and mobility of residing cells but also absorbs and transduces the shear and strain forces of the blood flow. It is primarily composed of elastin, collagen, proteoglycans, and glycoproteins. The elastin fibers and type I and III fibrillar collagens form a rigid network of highly cross-linked interstitial matrix. They offer elasticity (elastin) and tensile strength (collagens). Proteoglycans, because of their negative charge, attract water and confer resistance to compression. Finally, glycoproteins participate in matrix organization and are essential for cell attachment.
The vascular ECM also serves as a substrate for the binding and retention of secreted, soluble proteins of vascular cells as well as molecules coming from the circulation, including lipoproteins, growth factors, cytokines, proteases, and protease inhibitors. These components are invariably associated with ECM proteins, especially proteoglycans. Together they comprise the vascular extracellular environment and are pivotal for disease processes, such as atherosclerosis and aneurysm formation (1).
Although proteomics has been previously applied to vascular tissues, only one study has specifically targeted the extracellular vascular environment (2). This study was focused on the isolation of intimal proteoglycans from human carotid arteries. Moreover, most proteomics studies use whole tissue lysates, which are rich in cellular proteins that inevitably mask the identification of the less abundant proteins of the vascular extracellular environment (3–5). Thus, the composition of the vascular ECM and its associated proteins remains poorly defined. In the present study, we used morphologically normal human aortic samples to develop a method for the extraction of proteins present in the extracellular environment, including ECM proteins and proteins attached to the ECM. We had three specific aims: first, to reduce the contamination with cellular proteins, thereby increasing the chance of identifying scarce extracellular proteins; second, to efficiently solubilize and deglycosylate ECM proteins to improve their analysis by liquid chromatography tandem mass spectrometry (LC-MS/MS); and third, to interface the nanoflow LC system to a recently developed injection device, which splits the flow from the analytical column, to allow the reanalysis of the same sample during a single LC-MS/MS run (RePlay, Advion).
Our methodology provides a detailed overview of the aortic ECM and its associated proteins, many reported for the first time in proteomics analysis of the vasculature. Most importantly, this method could be adapted for use with other tissues to further our understanding of the composition of extracellular environment and ECM turnover under various disease conditions.
The data presented were derived from three human aortic samples. The samples were obtained upon aortotomy performed during routine aortic valve replacement. They were collected from positions of the ascending aorta that were free of macroscopically evident vascular pathology, including atherosclerosis or aneurysm formation. The tissue was immediately snap frozen and subsequently kept in liquid nitrogen for later use. A fourth sample was also obtained upon aortotomy and processed as described below, but the proteomics analysis revealed proteolysis of ECM proteins. Therefore, we used it to examine the potential of our methodology to study proteolytic activity in the tissue. All procedures were approved by our local Regional Ethics Committee Board. Informed consent was obtained from all patients recruited into the study.
Before extraction, the tissue pieces were partially thawed and weighed. Approximately 150 mg of tissue per aortic sample were immediately placed in ice-cold phosphate-buffered saline to remove plasma contaminants. Commercially available protease and phosphatase inhibitor mixtures (Sigma-Aldrich) were included according to the manufacturer's instructions to inhibit broad range proteinase activity, and 25 mm EDTA was included to ensure inhibition of metalloproteinases. While the tissue samples were immersed in the cold saline mixture, they were quickly diced with a scalpel into 8–10 smaller pieces. This would enhance (a) the initial removal of plasma contaminants and (b) the effective extraction of extracellular proteins, which is described below. The saline mixture was changed five times (~30 ml total per sample).
Then, the diced samples were incubated with 0.5 m NaCl, 10 mm Tris, pH 7.5 supplemented with proteinase/phosphatase inhibitor mixtures (Sigma-Aldrich) and 25 mm EDTA. The volume of the buffer was adjusted to 10:1 of the tissue weight (i.e. 100 mg in 1 ml), and the samples were mildly vortexed for 4 h at room temperature. The NaCl extracts were then desalted with centrifugation using desalting columns (Zeba Spin, Pierce). Following desalting, the extracts were mixed with 100% acetone (5:1 volume ratio) at −20 °C for 16 h. Proteins were precipitated with centrifugation (16,000 × g for 45 min), and the pellets were dried and redissolved in deglycosylation buffer (see below).
Subsequently, the aortic samples were incubated with 0.08% SDS (10:1 buffer volume to tissue weight) supplemented with proteinase/phosphatase inhibitor mixtures and 25 mm EDTA. The samples were mildly vortexed for 4 h at room temperature to achieve removal of cellular components. From our experience with tissue extractions, we took particular care to ensure a low vortexing speed to decrease the possibility of mechanical disruption of the ECM. The SDS solution was then removed and stored frozen for later use.
Finally, the samples were incubated in a 4 m guanidine HCl, 50 mm sodium acetate, pH 5.8 buffer (5:1 buffer volume to tissue weight) supplemented with proteinase/phosphatase inhibitor mixtures and 25 mm EDTA. Adjusting the pH to 5.8 is important; more than 30 years ago it was shown that at this pH guanidine is more effective in solubilizing vascular proteoglycans, whereas higher pH is required for the solubilization of the proteoglycans from other ECM-rich tissues, such as cartilage (6). The samples were then incubated for 48 h at room temperature and vortexed vigorously to enhance mechanical disruption of the ECM components. Subsequently, the guanidine extracts were mixed with 100% ethanol (5:1 volume ratio) at −20 °C for 16 h to ensure removal of guanidine (the presence of guanidine would hinder further biochemical processing). Proteins were precipitated with centrifugation (16,000 × g for 45 min), and the pellets were washed with 90% ethanol, dried, and redissolved in deglycosylation buffer (see below).
Deglycosylation (removal of glycosaminoglycan side chains) of the different extracts was achieved in a 150 mm NaCl, 50 mm sodium acetate, pH 6.8 buffer supplemented with proteinase/phosphatase inhibitors and 10 mm EDTA for 16 h at 37 °C. The deglycosylation enzymes (0.05 unit) used were chondroitinase ABC from Proteus vulgaris (It catalyzes the removal of polysaccharides containing 1→4-β-d-hexosaminyl and 1→3-β-d-glucuronosyl or 1→3-α-l-iduronosyl linkages to disaccharides containing 4-deoxy-β-d-gluc-4-enuronosyl groups. It acts on chondroitin 4-sulfate, chondroitin 6-sulfate, and dermatan sulfate glycosaminoglycan side chains.), keratanase from Bacteroides fragilis (cleaves internal 1→4-β-galactose linkages in unbranched, repeating poly-N-acetyl-lactosamine and keratan sulfate), and heparitinase II from Flavobacterium heparinum (cleaves heparan sulfate), all purchased from Sigma-Aldrich. Following deglycosylation, the solutions were clarified again with centrifugation (16,000 × g for 10 min) to ensure that the samples were free of turbidity, and protein concentration was estimated by UV absorbance at 280 nm using extinction coefficient of 1.1 of 0.1% mg/ml solution calculated on the basis of the frequency of tyrosine and tryptophan, which are the main UV light-absorbing amino acids at 280 nm in mammalian proteins (7).
The NaCl, guanidine, and SDS extracts were denatured and reduced in sample buffer containing 100 mm Tris, pH 6.8, 40% glycerol, 0.2% SDS, 2% β-mercaptoethanol, and 0.02% bromphenol blue and boiled at 96 °C for 10 min. 35 μg of protein per sample were loaded and separated on Bis-Tris discontinuous 4–12% polyacrylamide gradient gels (NuPAGE, Invitrogen). Prestained protein standards were run alongside the samples to allow molecular mass estimation of proteins (All Blue, Precision Plus, Bio-Rad).
After electrophoresis, gels were stained using the PlusOne Silver staining kit (GE Healthcare). Silver staining was used for band excision to avoid cross-contamination with fainter gel bands. (Coomassie staining will predominantly stain the abundant proteins; hence, fainter gel bands could be missed.). The gel bands were excised in identical parallel positions across lanes, and no “empty” gel pieces were left behind. Subsequently, all gel bands were subjected to in-gel digestion with trypsin using an Investigator ProGest (Genomic Solutions) robotic digestion system. Tryptic peptides from the NaCl and guanidine extracts were separated on a nanoflow LC system (Dionex UltiMate 3000) and eluted with a 40-min gradient (10–25% B in 35 min, 25–40% B in 5 min, 90% B in 10 min, and 2% B in 30min where A is 2% ACN, 0.1% formic acid in HPLC H2O and B is 90% ACN, 0.1% formic acid in HPLC H2O). The column (Dionex PepMap C18, 25-cm length, 75-μm internal diameter, 3-μm particle size) was coupled to a nanospray source (Picoview) using RePlay (Advion). After the direct LC-MS run, the flow was switched, and the portion stored in the capillary of the RePlay device was reanalyzed (“replay run”) (8). Spectra were collected from an ion trap mass analyzer (LTQ-Orbitrap XL, Thermo Fisher Scientific) using full ion scan mode over the mass-to-charge (m/z) range 450–1600. MS/MS was performed on the top six ions in each MS scan using the data-dependent acquisition mode with dynamic exclusion enabled. MS/MS peak lists were generated by extract_msn.exe and matched to a human database (UniProtKB/Swiss-Prot Release 14.6, 20,333 protein entries) using SEQUEST version 28 (revision 13) (Bioworks Browser 3.3.1 SP1, Thermo Fisher Scientific) and X! Tandem (version 2007.01.01.2). Carboxyamidomethylation of cysteine was chosen as a fixed modification, and oxidation of methionine was chosen as a variable modification. The mass tolerance was set at 1.5 amu for the precursor ions and at 1.0 amu for fragment ions. Two missed cleavages were allowed. Scaffold (version 2.0.5, Proteome Software Inc., Portland, OR) was used to calculate the spectral counts and to validate MS/MS-based peptide and protein identifications (9, 10). According to the default values in the Scaffold software, the following peptide thresholds were applied: X! Tandem: −log(Expect scores) > 2.0; SEQUEST: ΔCn > 0.10 and XCorr > 2.5 (2+), 3.5 (3+), and 3.5 (4+). Peptide identifications were accepted if they could be established at greater than 95.0% probability as specified by the Peptide Prophet algorithm (9). Protein identifications were accepted if they could be established at greater than 99.0% probability (10) with at least two independent peptides and a precursor ion mass accuracy ≤10 ppm.
The protein concentration in the SDS extracts was estimated using the Groves equation, (A280 × 1.55) − (A260 × 0.76) (11), which takes into account nucleic acid interference. 35 μg of protein per sample were loaded and separated on Bis-Tris discontinuous 4–12% polyacrylamide gradient gels as above. The gel was silver-stained, and protein bands were digested by trypsin as described above. Tryptic peptides were separated on a split-free, nanoflow LC system (EASY-nLC, Proxeon) on a C18 column (Easy-nLC C18, Proxeon, 10-cm length, 75-μm internal diameter, 3-μm particle size) and eluted with an 80-min gradient (10–25% B in 35 min, 25–40% B in 5 min, 90% B in 10 min, and 2% B in 30min where A is 2% ACN, 0.1% formic acid in HPLC H2O and B is 90% ACN, 0.1% formic acid in HPLC H2O). The nanoflow LC system was not coupled to a RePlay device. Spectra were collected from a dual pressure, linear ion trap (LTQ Velos, Thermo Fisher Scientific) using full ion scan mode over the m/z range 300–1800. MS/MS was performed on the top 20 ions in each MS scan using the data-dependent acquisition mode with dynamic exclusion enabled. MS/MS peak lists were generated by extract_msn.exe and matched to a human database (UniProtKB/Swiss-Prot Release 15.13, 20,276 protein entries) using SEQUEST version 28 (revision 13) (Proteome Discoverer 1.2 beta, Thermo Fisher Scientific). Carboxyamidomethylation of cysteine was chosen as a fixed modification, and oxidation of methionine was chosen as a variable modification. The mass tolerance was set at 0.5 amu for the precursor ions and at 0.5 amu for fragment ions. Two missed cleavages were allowed. Scaffold (version 3, Proteome Software Inc.) was used to calculate the spectral counts and to validate MS/MS-based peptide and protein identifications as before. According to the default values in the Scaffold software, the following peptide thresholds were applied: SEQUEST: ΔCn > 0.10 and XCorr > 2.5 (2+), 3.5 (3+), and 3.5 (4+). Peptide identifications were accepted if they could be established at greater than 95.0% probability as specified by the Peptide Prophet algorithm (9). Protein identifications were accepted if they could be established at greater than 99.0% probability (10) with at least two independent peptides.
Aliquots of the NaCl, guanidine, and SDS extracts were mixed with denaturing sample buffer and boiled. 20 μg of protein per sample were loaded and separated on 4–12% gradient gels as above. Proteins were then transferred on nitrocellulose membranes. Membranes were blocked in 5% fat-free milk powder in PBS and then probed for 16 h at 4 °C with primary antibodies to β-actin (sc-130656), periostin (sc-67233), and fibronectin (sc-56391). The antibodies were purchased from Santa Cruz Biotechnology and were used at a 1:500 dilution in 5% BSA. The membranes were then treated with the appropriate secondary, peroxidase-conjugated antibodies (Dako) at a 1:2000 dilution. Finally, the blots were imaged using enhanced chemiluminescence (ECL; GE Healthcare), and films were developed on a Xograph processor. The densitometry for the lanes from developed blots for periostin and asporin was performed using ImageJ software.
Gelatinolytic zymograms were used to detect the presence of MMP-2 and -9 in the 0.5 m NaCl extracts. Extract aliquots (20 μg of protein) were mixed with non-reducing sample buffer containing 100 mm Tris, pH 6.8, 40% glycerol, 0.2% SDS, and 0.02% bromphenol blue and separated by 10% SDS-PAGE in gels containing 0.5 mg/ml gelatin (Sigma-Aldrich). Prestained protein standards were run alongside the samples as before. To allow in-gel MMP activity, SDS was removed from the gels with three washes in a buffer containing 2% Triton X-100, 50 mm Tris, pH 7.4, and 200 mm NaCl for 45 min. Gels were then incubated for 16 h to detect MMP-9 or 36 h to detect MMP-2 at 37 °C in a buffer containing 50 mm Tris, pH 7.4, 200 mm NaCl, and 10 mm CaCl2. Finally, gels were stained for 60 min with Coomassie Brilliant Blue (Sigma-Aldrich) and destained for 2 h to visualize the MMPs.
Small representative pieces from each aortic sample were placed in optimal cutting temperature (OCT) compound (BDH), which was subsequently frozen. The frozen tissue blocks were used to generate consecutive 7-μm-thick cross-sections. The sections were air-dried (30 min) and fixed in 100% acetone for 20 min. Subsequently, sections were used either to generate hematoxylin- and eosin-stained histological slides or for immunohistochemical analysis using antibodies to podocan used at 2 μg/ml (MAB4220) (R&D Systems), sclerostin (sc-130258) used at 1 μg/ml, and agrin (sc-25528) used at used at 2 μg/ml (Santa Cruz Biotechnology). Negative controls were generated with either mouse (podocan) or rabbit (sclerostin and agrin) isotype IgG antibodies (Sigma-Aldrich) used at 2 μg/ml. The immunostaining was generated using the DAB+ (diaminobenzidine) peroxidase-based chromogen system (Dako) according to the manufacturer's instructions. Images were taken by a Zeiss Axioplan 2ie microscope interfaced to AxioVision software (version 3.0.6).
For each clinical sample (n = 4) and biochemical fraction (n = 2) analyzed, we obtained a technical replicate by using the RePlay device. The reproducibility of the RePlay analysis was assessed by calculating the Pearson correlation coefficient for the spectral counts. All spectra from proteins identified with at least two unique peptides were included in the analysis, and no outliers were removed. The p value for the linear dependence between the RePlay runs was obtained using the Fisher transformation. To assess biological variability among the four clinical specimens, we computed the coefficient of variation for all identified proteins using Scaffold software (version 2.0.5, Proteome Software Inc.). A coefficient of variation exceeding 100% was applied as a cutoff. Statistical significance was determined by using analysis of variance (ANOVA). A p value <0.05 was considered significant.
Our extraction methodology was developed on human aortic samples obtained during cardiac surgery for aortic valve replacement of degenerative aortic valve stenosis. Fig. 1A is a schematic summary of the methodology used in this study. The samples (n = 3) were macroscopically classified as normal during surgery, and upon histological examination, they showed no pathological signs. Extraction of the proteins of the vascular extracellular environment and reduction of the contamination with cellular proteins were achieved by a three-step extraction procedure. First, the tissue samples were treated with 0.5 m NaCl. The salt ions induce displacement of polyionic interactions between proteins, thus enabling the extraction of loosely bound extracellular proteins, including newly synthesized ECM proteins and degradation products (12), providing an overview of tissue turnover. The tissue was then decellularized using SDS. In contrast to other decellularization reagents, such as non-ionic detergents (Triton X-100) (13), 0.1% SDS has been previously shown to be very effective in solubilizing cytoplasmic and nuclear membranes, thereby removing cellular components from tissues (14). However, the ionic detergent above its critical micelle concentration can also cause protein denaturation and disrupt native ECM proteins (15).
To substantially remove the cellular material while preserving the ECM and its associated proteins as much as possible, we used 2.8 mm (0.08%) SDS in distilled water. This concentration is well below the critical micelle concentration of SDS in distilled water, which is 8 mm (16). In addition, at this concentration, SDS is effective in solubilizing cell membrane lipid bilayers (17). Treatment with 0.08% SDS was sufficient to remove most cellular components as shown by the marked reduction of nuclear staining (hematoxylin; Fig. 1B, blue). The cytoplasmic staining (eosin; pink) was also reduced, but other eosinophilic structures, such as collagen and elastin, were preserved in the tissue. Finally, the specimens were subjected to denaturing extraction in a 4 m guanidine HCl buffer. Guanidine HCl extractions were developed more than 40 years ago as a very effective way to solubilize most of the strongly bound ECM components, including large aggregating proteoglycans (versican, aggrecan, etc.), small proteoglycans (decorin, biglycan, etc.), cell attachment matrix glycoproteins (such as type VI collagen, fibronectins, and laminins), and basement membrane components (perlecan, type IV collagen, etc.). Guanidine induces disaggregation of ECM components by destabilizing the ionic, disulfide-dependent protein conformations (18). Following dissociative extraction, only the insoluble, heavily cross-linked interstitial matrix, namely type I and III collagens and elastin, remains in the tissue. Western blotting confirmed that β-actin immunoreactivity is predominantly found in the SDS extracts (Fig 1C, middle panel), which also had the highest protein concentration per milliliter (supplemental Fig. 1). On the other hand, the extracellular AEBP1 isoform, an ECM-associated glycoprotein, was detected in the NaCl and guanidine extracts by Western blotting (Fig 1C, upper panel), indicating that these extracts are enriched in proteins of the extracellular space. Silver staining revealed a distinct but reproducible protein pattern in the three different extraction steps (Fig. 1C, bottom panel).
The different extracts were subjected to enzymatic deglycosylation to remove chondroitin, keratan, and heparan sulfate glycosaminoglycan (GAG) chains from the matrix glycoproteins. Deglycosylation is a critical step in our methodology. First, GAGs increase the molecular weight of glycoproteins and affect their separation and resolution by one-dimensional electrophoresis. Second, the presence of GAGs greatly reduces the chance for the identification of proteoglycans by mass spectrometry.
After separation by SDS-PAGE, protein bands were excised from silver-stained gels, subjected to in-gel tryptic digestion, and analyzed by LC-MS/MS. The peptides were injected into a high mass accuracy ion trap (LTQ-Orbitrap XL, Thermo Fisher Scientific) using RePlay (Advion). This injection system splits the gradient from the analytical column and allows the reanalysis of the same sample in a single LC-MS/MS run (8). 272 and 321 proteins were identified in the NaCl and guanidine extracts, respectively. All the peptides identified by LC-MS/MS in the NaCl and guanidine extracts are shown in supplemental Tables 1 and 2. According to their gene ontology annotation (European Molecular Biology Laboratory-European Bioinformatics Institute database, www.uniprot.org), a quarter of these proteins were located in the extracellular space: among the 61 extracellular proteins in the NaCl extracts (Table I) and the 84 extracellular proteins in the guanidine extracts (Table II), proteins highlighted with Footnote b were, to the best of our knowledge, identified for the first time in the vasculature by proteomics. Lipoproteins were included (even though they are plasma-derived) because of their intimate relationship with proteoglycans and their relevance in cardiovascular pathology. The classification of the identified extracellular space proteins in the NaCl and the guanidine extracts is shown in Fig. 2, A and B. As expected, the denaturing guanidine extracts were particularly enriched in ECM proteoglycans and glycoproteins (50%), whereas the mildly dissociative NaCl extracts were enriched in proteases and protease inhibitors (21%), indicating the importance of this subproteome for studying proteolytic processes in tissues.
To identify proteins of the extracellular space extracted by 0.08% SDS, we performed proteomics analysis of the SDS extracts. Because of the increased complexity of these samples, tryptic peptides were injected into a dual pressure, linear ion trap (LTQ Velos, Thermo Fisher Scientific) with faster scan speed. As expected, the majority of the identified proteins were cellular (data not shown), and only 28 extracellular space proteins were identified (Table III). The classification of the identified extracellular space proteins in the SDS extracts is shown in Fig. 2C. Interestingly, apart from type XIV collagen, all other extracellular space proteins have been previously identified by proteomics in the vasculature. Also, of note is the absence of proteases and protease inhibitors in these extracts either because they are obscured by abundant cellular proteins or because those that remain in the tissue following 0.5 m NaCl extraction are tightly bound to the ECM and require dissociative conditions to be extracted (i.e. guanidine). Peptides of extracellular space proteins identified by LC-MS/MS in the SDS extracts are shown in supplemental Table 3.
Among the identified proteins in the NaCl and guanidine extracts were three recently discovered glycoproteins that have never been detected in the aortic ECM at the protein level. Podocan, an atypical proteoglycan, which has so far been found in the basement membrane of the glomeruli as a product of kidney podocytes (19), was detected in the NaCl extracts. Immunohistological staining confirmed the presence of podocan in the aortic intima (Fig. 3A; the isotype negative control is shown in Fig. 3B). Sclerostin, a potent bone morphogenetic protein antagonist secreted by osteoblasts and osteocytes (20), was identified in the guanidine extract and was localized in the subendothelial layer of the aortic intima (Fig. 3C; the isotype negative control is shown in Fig. 3D). The function of these glycoproteins in the vasculature is to date unknown. Finally, agrin is a large heparan sulfate proteoglycan. It has been found in the basement membrane of various tissues and is known to participate in the organization of the neuromuscular junction (21). Immunohistochemical staining confirmed the presence of agrin on the aortic endothelium (Fig. 3E; the isotype negative control is shown in Fig. 3F), which is also the site for the accumulation of other heparan sulfate proteoglycans, such as perlecan (22). It is therefore possible that agrin participates in the formation of the vascular endothelial barrier. Representative MS/MS spectra for the three glycoproteins are shown in supplemental Fig. 2.
For the wider application of proteomics in clinical research, it is essential to maximize the information obtained from valuable clinical samples. Given the scarcity of human tissue available, we used the RePlay device (Advion), which allowed us to obtain a technical replicate from each clinical specimen in a single LC-MS/MS run. Representative examples of chromatograms using RePlay are shown in Fig. 4, A and B. Correlation analyses of the spectral counts in the three NaCl extracts and the three guanidine extracts are presented in Fig. 4, C and D. The mean spectral counts for the NaCl and the guanidine extracts were 32.7 ± 9.9 (median, 9.5) and 126.5 ± 41 (median, 40), respectively. The R2 values (Pearson correlation coefficient) from all six extracts exceeded 0.94 (p values <0.001), demonstrating good reproducibility in the duplicate analyses (supplemental Fig. 3).
Because of their scarcity, MMPs are rarely detected by tissue proteomics (23). Our subfractionation, however, enabled us to obtain spectral evidence for the presence of MMP-2 in one of the NaCl extracts (aortic sample 1). Consistent with our proteomics data, gelatin zymography confirmed the presence of MMP-2 in this sample (Fig. 5A). A representative MS/MS spectrum for MMP-2 is shown in supplemental Fig. 4A. Importantly, the non-denaturing composition of the NaCl buffer allowed the detection of MMP-2 directly without further processing of the extracts.
In the guanidine extracts, we performed Western immunoblotting for the ECM protein periostin, which was unambiguously identified (Fig. 5B, upper panel). Periostin mRNA was recently shown to be expressed in rat carotid arteries following balloon injury (24). We also confirmed the presence of the scarce glycoprotein asporin in the vascular ECM (Fig. 5B, bottom panel). Its biological role is to date unclear, but it was recently shown to inhibit TGFβ-induced chondrogenesis in articular cartilage (25) and is a potential regulator of collagen mineralization (26). Finally, the Western blots for periostin and asporin were analyzed by densitometry, and the densitometry values were correlated with the spectral counts obtained from the RePlay analysis (n = 2, mean ± S.D.) (Fig. 5, C and D). Although the technical replicates obtained using the RePlay device were accurate, there is expected variability in the different clinical samples. Representative MS/MS spectra for periostin and asporin are shown in supplemental Fig. 4, B and C.
Finally, we investigated whether the NaCl extract represents a subproteome enriched in degradation products. Currently, our understanding of ECM proteolysis is based on in vitro experiments, which do not necessarily relate to in vivo processes. In our methodology, the original molecular mass distribution of the identified proteins is preserved by the gel LC-MS/MS approach. Thus, our method can be important for advancing our understanding of protein degradation in vivo. For instance, we noted that in a fourth aortic specimen, which was classified as normal upon aortotomy, the coefficient of variation for the number of fibronectin spectra in the NaCl extract exceeded 100%, implying a significant difference in the dispersion of the data among the analyzed samples (p ANOVA <0.05). On the other hand, the coefficient of variation for samples 1–3 is not significant (58%; p ANOVA >0.05). We also noted that there was an increase in fibronectin sequence coverage (Fig. 6, A and B) and total fibronectin spectra (~4-fold) in the NaCl extract of the fourth specimen. Closer inspection of the molecular mass distribution of fibronectin spectra indicated the breakdown of fibronectin in this particular sample with more than 80% of fibronectin spectra identified in gel bands below 150 kDa. Western blotting using an antibody raised against full-length human fibronectin confirmed extensive laddering of the glycoprotein in both the NaCl and guanidine extracts (Fig. 6, C and D). Fibronectin degradation products have been implicated in the induction of proinflammatory responses in the vasculature and other tissues (1). Interestingly, mass spectrometry provided spectral evidence for MMP-9 only in this particular sample (Fig. 6E), and the presence of MMP-9 was confirmed by gelatin zymography (Fig. 6C, lower panel). Fibronectin is a known substrate for MMP-9 (27), but obviously, we cannot exclude that other proteinases contribute to the breakdown of fibronectin in this specimen in vivo. The example, however, demonstrates the potential of our methodology to analyze proteolytic activity based on the presence of proteolytic enzymes and corresponding degradation products in tissues. Importantly, this key information on protein breakdown within clinical samples that were considered normal macroscopically could be lost by using a conventional shotgun proteomics approach.
In this study, we combined biochemical fractionation with proteomics to comprehensively characterize the protein composition of the vascular extracellular environment. The main advantage of our methodology is the stepwise approach, which results in the isolation of proteins and protein degradation products loosely bound to the ECM in the NaCl fraction, the depletion of cellular proteins in the SDS fraction, and the effective solubilization of ECM proteins in the guanidine fraction.
Until now, a comprehensive analysis of extracellular space components comprising the vascular proteome in humans has not been performed. Previous attempts to utilize the analytical power of proteomics to study ECM components and associated proteins have been limited by the fact that without prefractionation cellular and plasma proteins in mammalian tissues overwhelm the analytical capacities of the instruments and obscure the identification of scarce proteins in the extracellular environment. Therefore, it is not surprising that all extracellular space proteins found in the SDS extracts (except from type XIV collagen) have been reported before by proteomics studies of different vascular tissues.
On the other hand, the tissue subfractionation resulted in the identification of numerous proteins in the NaCl and guanidine extracts, which have not been investigated in the vasculature until now. The relative insolubility and the extensive post-translational modifications, especially glycosylation, render ECM proteins difficult to analyze. For example, we identified and validated the presence of three glycoproteins: podocan, originally described as a glycoprotein secreted by kidney podocytes; sclerostin, another small glycoprotein, which has recently received attention because of its ability to act as a potent negative regulator of bone morphogenetic protein signaling and bone mineralization; and agrin, a large basement membrane proteoglycan, which participates in the organization of the synapses in the neuromuscular junction. We also identified and validated the presence of the atypical proteoglycan asporin in our aortic extracts.
Moreover, the NaCl and guanidine extracts contained two ECM glycoproteins, tenascin-X and AEBP1. The first is believed to be a modulator of collagen fibrillogenesis (28). Interestingly, genetic deletion of tenascin-X in mice resulted in increased activity of MMP-2 and -9 (29). Less is known about AEBP1; the extracellular isoform was originally found to play a role in wound healing (30), and more recently, Majdalawieh et al. (31) showed that intracellular AEBP1 represses macrophage cholesterol efflux.
Among collagens, collagen VI appears to be very abundant in our extracts. It serves mainly as a substrate for the attachment of cells and matrix molecules and also anchors basement membranes to underlying connective tissues (32). More interesting, however, is the identification of type XIV collagen (33). Previous proteomics studies probably failed to identify this collagen because it is heavily modified by chondroitin sulfate saccharides in vivo (34). These modifications greatly increase its size and alter its electrophoretic properties but are removed by the deglycosylation step in our extraction methodology.
Finally, many scarce proteins, including LIM and cysteine-rich domains protein 1 (LMCD1), target of Nesh-SH3 (TARSH), and RPE-spondin (a previously hypothetical protein (35)), were all identified for the first time by proteomics in the vasculature, and representative MS/MS spectra are shown in supplemental Fig. 5. Their function in the vasculature is unknown.
The subfractionation presented here could be an important tool for studying catabolic processes in tissues. In the NaCl fraction, we found fibronectin degradation products and the presence of MMP-2 and -9 as verified by gelatinolytic zymography. Mass spectrometry also identified two mast cell proteases, namely chymase and tryptase β-1. The first was recently found to participate in the development of mouse abdominal aneurysm (36), whereas the second, an activator of MMP-3 (37), has been shown to degrade the antiatherogenic high density lipoprotein (HDL) (38). Their presence in the aortic extracts might suggest an early involvement of mast cells in aortic pathology (39).
As far as vascular research is concerned, the methodical and accurate characterization of the vascular extracellular environment is essential to advance our understanding of the ECM composition in different vascular territories and their predisposition to disease processes. For instance, atherosclerosis, the major cause for morbidity and mortality in the developed world, is intimately related to dynamic changes in the ECM and its associated proteins: the retention of lipoproteins by proteoglycans leads to the development of atherosclerotic lesions, but ultimately, it is the breakdown of ECM that facilitates plaque ruptures, triggering the onset of clinical symptoms. These changes, if comprehensively characterized, could lead to new therapeutic options for cardiovascular disease. Finally, because metabolic labeling cannot be used with clinical samples, spectral counting is an alternative method to assess relative protein abundance (40), but its accuracy is poor for proteins with low spectral counts (41).
In conclusion, the proteomics method described allowed (a) the identification of scarce proteins present in the vascular extracellular space, (b) the identification of novel glycoproteins, and (c) the interrogation of proteolytic activity within tissues based on the identification of proteolytic enzymes and corresponding degradation products. Because ECM proteins accumulate in tissues, mRNA transcripts do not correspond to actual protein levels. In contrast, a proteomics-led approach has the potential to provide a more comprehensive analysis of the ECM composition and shed new light on pathological disease processes.
* This work was supported in part by grants from the Oak Foundation and the British Heart Foundation.
This article contains supplemental Figs. 1–5 and Tables 1–3.
1 The abbreviations used are: