Heat shock protein 90 is downregulated in calcific aortic valve disease

The aortic valves examined in this study were obtained from 50 patients at the time of aortic valve or aortic root surgery. All operations were made following normal surgical procedures. The study protocol was approved by the Research Ethics Committee of Oulu University Hospital and it conformed to the principles outlined in the Declaration of Helsinki. The aortic valve cusps were immersed immediately after removal into liquid nitrogen and stored at − 70 °C until analyzed.

For proteomics study, patients were divided into two groups: the control group (C, n = 5) consisted of patients with normal, non-calcified, smooth and pliable aortic valve cusps, operated due to ascending aortic pathology (aneurysm or dissection) or aortic regurgitation. The aortic stenosis group (AS, n = 7) consisted of patients who had non-rheumatic, severe aortic valve sclerosis with an increased degree of calcification. Patients who were identified as exhibiting macroscopic thickenings of aortic valve cusps, which were microscopically identified mainly as fibrotic and mild sclerotic lesions, were excluded from the study.

The patients’ demographics are presented in Table 1. There were no significant differences in gender, left ventricular ejection fraction or comorbidities between the study groups, and valve anatomy. However, the average age of the aortic stenosis (AS) patients was significantly higher than the patients in the control group. Histologically, the stenotic valves had a significantly elevated amount of calcium and more neovessels in comparison with control valves [4, 17, 18]. For validation of proteomics results, a separate matching group of patients (n = 39) was selected.

The proteins extracted from control (C, n = 5) and calcified (AS, n = 7) aortic valves were further purified by buffer exchange using an Amicon Ultra ultrafiltration unit with a 10 kDa cutoff (Millipore) and urea buffer (7 M urea, 2 M thiourea, 4% [w/v] CHAPS, 30 mM Tris, pH 8.5) and then the protein samples were sonicated and centrifuged. Protein amounts in the supernatants were determined with a Bradford-based assay according to the manufacturer’s instructions (Roti®-Nanoquant) and the aliquots were stored at − 70 °C. Protein labeling was performed with CyDye DIGE Fluor minimal dyes (GE Healthcare) according to the manufacturer’s protocol using 400 pmol Cy3 (pooled standard) and Cy5 (control, AS, respectively) for 50 μg protein. Proteins were separated as described earlier [19]. In brief, immobilized pH gradient (IPG) strips (pH 3–10 nonlinear, 24 cm, GE Healthcare) were incubated overnight in 650 μl rehydration buffer (7 M urea, 2 M thiourea, 4% [w/v] CHAPS, 130 mM [w/v] DTT, 2%[v/v] carrier ampholytes 3–10, Complete Mini protease inhibitor cocktail [Roche Life Science]). Isoelectric focusing (IEF) after anodic sample cup-loading was carried out with the Multiphor II system (GE Healthcare) under paraffin oil with 67 kVh. SDS-PAGE was performed overnight in polyacrylamide gels (12.5%) with the Ettan DALT II system (GE Healthcare) at 1–2 W per gel in 12 °C. Fluorescence signals were detected with a Typhoon 9400 (GE Healthcare) and 2-D gels analyzed with Delta2D 4.0 (Decodon). Theoretical spot positions were calculated with the Compute pI/Mw tool (http://​ca.​expasy.​org/​tools/​pi_​tool.​html). Principal Component Analysis was performed with the Delta2D v4.0 software (Decodon) according to the spot intensities on every gel image.

For protein identification, additional 2-D gels were run with a higher amount of unlabelled protein (400–600 μg) combined with 50 μg Cy3-labelled internal standard. After detection of the fluorescence signals (see above) and silver staining, labelled and unlabelled protein patterns were matched with the 2-D PAGE image analysis software Melanie 3.0 (GeneBio). Spots with correctly matched centers were excised, digested with trypsin (recombinant; Roche) and prepared for MALDI-TOF mass spectrometry as described previously [19]. The extracted and dried peptides were dissolved in 5 μl alpha-Cyano-3-hydroxycinnamic acid (98%, recrystallized from ethanol-water, 5 mg/ ml in 50% acetonitrile and 0.1% TFA) and 0.5 μl applied onto the sample plate using the dried-droplet method. Proteins were identified from PMF obtained with a VOYAGER-DE™ STR (Applied Biosystems) as described earlier [19]. In general, the clearest peaks (up to 50) visible in the mass spectrum were used to identify proteins with Mascot (http://​www.​matrixscience.​com/​) using Swiss-Prot as the corresponding protein database. Search parameters were enzyme: trypsin; modifications: oxidation of Met; missed cleavage: 1; resolution: monoisotopic; ion mode: [M + H]; threshold: 50 ppm. The protein identification was accepted if at least 4 major peaks matched to the protein with the highest Mascot score. In addition, the identification was confirmed by analyzing the induced spot from different gels. During later stages of the project, mass spectra of the tryptic digests were obtained with a UltrafleXtreme MALDI TOF/TOF instrument (Bruker Daltonics) where up to 10 ions from each peptide fingerprint were subjected to the MS/MS measurement. Data were processed with Flexanalyis and Biotools (Bruker) and combined PMF/MS/MS spectra were searched against the NCBI or Swiss-Prot non-redundant protein database using Mascot (Matrix science) with standard search parameters (MS tolerance: 30 ppm, MS/MS tolerance: 0.7 Da, modifications: Carbamidomethyl (Cys) and optional oxidation of Met, up to 1 missed cleavage).

In the western blot experiments, aortic valve samples were obtained from a separate matching cohort (C, n = 19 AS, n = 20). The samples were ground in liquid nitrogen, and then homogenized for 10 min in a lysis buffer containing inhibitors. The lysis buffer itself contained 1 M Tris (pH 7.5), 3 M NaCl, 0.25 M EDTA (pH 8.0), 0.1 M EGTA (pH 7.9), 1 mmol/l β-glycerophosphate, 1 mmol/l Na3VO4, 2 mmol/l benzamidine, 1 mmol/l phenylmethylsulfoxide, 50 mmol/l NaF, 1 mmol/l dithiothreitol and 10 μg/ml each of leupeptin, pepstatin, aprotininand and distilled water. The valve tissue samples were homogenized using a MagnaLyser instrument (Roche). After homogenization, the samples were centrifuged for 20 min in 12,500 rpm and + 4 °C and then the supernatant was collected for protein isolation. 5x NEB lysis buffer (100 mM Tris-HCl [pH 7.5], 750 mM.

NaCl, 5 mM EDTA, 5 mM EGTA, 5% Triton X 100, 12 mM sodium pyrophosphate, 5 mM β-glycerophosphate, 5 mM Na₃VO₄) was added and mixed following centrifugation for 20 min in 12,500 rpm in + 4 °C. Supernatant containing the total fraction was collected. Western blot was performed using a 1.0 mm, 12% gel with 40 μg of protein/well. The following primary antibodies were used: HSP90α (ADI-SPS-771) and HSP90β (ADI-SPA-844) from Enzo Life Sciences, Protein kinase B (Akt) (#9272), Phospho- Akt (#4056), p38 mitogen activated protein kinase (MAPK) (#9212), phospho-p38 MAPK (#9211), extracellular signal regulated kinase p44/42 MAPK (Erk1/2) (#9102) and Phospho-p44/42 MAPK (pErk1/2) (#9106) from Cell Signaling Technology, Inc., Anti-Annexin II (610,068, BD Transduction Laboratories), and Anti-Galectin 1 (ab25138, Abcam). Anti-mouse-IgG HRP-labeled (GE Healthcare), Anti-rabbit-IgG Peroxidase conjugate (Calbiochem), anti-IgG HRP-linked rabbit (#7074, Cell Signaling Technology, Inc.), and anti-IgG HRP-linked mouse (#7076, Cell Signaling Technology, Inc.) secondary antibodies were used. Data was quantified using the QuantityOne Software (Bio-Rad).

The localization of HSP90α and HSP90β in the aortic valve cusps was studied by using immunohistochemical stainings. The aortic valve samples, sent for routine diagnistics, were fixed in buffered formalin solution and embedded in paraffin. Decalcification with EDTA was done if needed. For total valve area and calcified valve area slides were photographed with a Leica DFC420 camera (Wetzlar) and areas were quantified with Image J analysis software. Calcified area to total area was calculated with the following formula: (calcified valve area/ total valve area)*100. Before application of the primary antibodies, the 5-μm-thick sections of valves samples were heated in a microwave oven in citrate buffer, pH 6.0, for 30 min. Rabbit monoclonal antibodies ab133492 at a dilution of 1:2000 (Abcam) for HSP90α and ab32568 at a dilution of 1:300 (Abcam) for HSP90β were used to stain. 3,3′Diaminobenzidine (DAP) was used as the chromogen in immunostaining process. Negative control stainings were carried out by substituting nonimmune rabbit serum for the primary antibodies.

The up- or down-regulated proteins with their respective expression values were uploaded for processing by the Ingenuity Pathway Analysis (IPA) software (Qiagen). A core analysis was performed with the following parameters: core analysis, reference set user-defined (i.e., only the set of differentially expressed genes by GeneSpring-software mapped to the IPA database), direct and indirect relationships included, confidence = experimentally observed. Then, the IPA-software was used to generate a molecular network showing the interrelationships between up- or down-regulated proteins as previously described [20], based on the information contained in the Ingenuity Pathways Knowledge database.

The proteomic analysis, based on the minimal DIGE, identified a total of 15 differentially abundant protein spots in stenotic valves as compared to control valves according to the selection criteria (fold change ≥1.5 and P ≤ 0.05). A typical 2D gel representing calcified aortic valve proteins is shown in Fig. 1. Further, mass spectrometry analyses identified seven proteins within 12 spots (Table 2) which had undergone significant upregulation i.e. complement 9, serum amyloid P-component (APCS) and transgelin (1.7-, 2.3- and 3.5-fold, respectively, P < 0.05) as well as downregulation of heat shock protein HSP90 (α/β; genes HSP90AA1/AB1), protein disulfide isomerase A3 (PDIA3), annexin A2 (ANXA2) and galectin-1 (2.1-, 3.5-, 2.2- and 2.2-fold, respectively, P < 0.05) in stenotic valves (Fig. 1, Table 2).

Western blot was used to validate the the proteomic results of HSP90, ANXA2 and galectin-1. The candidate proteins were selected based on their unknown role in CAVD. A marked downregulation of HSP90β protein levels was detected in stenotic valves compared to controls (Fig. 2a-b), whereas no change in HSP90α protein levels was observed (data not shown). Considerable interindividual variability was seen in the levels of ANXA2 protein, since it was highly expressed in only two out of three control samples (Fig. 3c). There was no significant difference in galectin-1 protein levels between stenotic and control levels (Fig. 3a-b).

To test if valvular anatomy impact on our data, we performed an unsupervised clustering analysis (principal component analysis, PCA) on the raw proteomic data (Additional file 1: Figure S1). We hypothesized that if valvular anatomy would affect protein expression profile, bicuspid valves should cluster together and distinctly from tricuspid valves. However, we find no evidence of such clustering and in fact, the clearest distinction emerges between control, and AS.

We correlated HSP90 expression levels both with age and calcification of the valves (expressed as proportion of calcified area in aortic valve cusps to total aortic valve). As shown in Additional file 2: Figure S2A-B, the HSP90β protein levels correlated with age of the patients (P < 0.01) and calcification of the valves (P < 0.05). In addition, the valvular calcification correlated with the age (P < 0.01) (Additional file 2: Figure S2C).

In the immunohistochemical stainings, the localization of HSP90α and HSP90β was virtually identical (Fig. 2b-e). VICs in both normal and calcified valves were positive. In addition, the endothelium of neovasculature was widely positively stained, whereas in the surface endothelium, the positive reaction was more patchy. Furthermore, most of the inflammatory cells, mainly lymphocytes, were also positively stained.

An IPA-analysis was used to determine the biological relationships among the differentially expressed proteins. The main molecular network exhibiting expression changes based on Fisher’s exact test is shown in Fig. 4.

Among the novel putative interactions suggested by IPA, HSP90 was linked to Akt and ERK, and further to p38 MAPK (Fig. 4). Therefore, we conducted Western blot analyses to evaluate the activation of Akt, ERK and p38 MAPK kinases in control and stenotic valves. The ratio of phosphorylated ERK to total ERK was increased (1.5-fold, P < 0.05) whereas the ratio of phosphorylated Akt to total Akt was reduced (0.7-fold, P < 0.05) suggesting that the Akt and ERK pathways were disturbed in the stenotic valves (Fig. 5a,c). Instead, there was no change in the phosphorylation pattern of p38 MAPK in stenotic valves (Fig. 5b).