Phosphoproteomic response of cardiac endothelial cells to ischemia and ultrasound

Abstract

Myocardial infarction and subsequent therapeutic interventions activate numerous intracellular cascades in every constituent cell type of the heart. Endothelial cells produce several protective compounds in response to therapeutic ultrasound, under both normoxic and ischemic conditions. How endothelial cells sense ultrasound and convert it to a beneficial biological response is not known. We adopted a global, unbiased phosphoproteomics approach aimed at understanding how endothelial cells respond to ultrasound. Here, we use primary cardiac endothelial cells to explore the cellular signaling events underlying the response to ischemia-like cellular injury and ultrasound exposure in vitro. Enriched phosphopeptides were analyzed with a high mass accuracy liquid chromatrography (LC) - tandem mass spectrometry (MS/MS) proteomic platform, yielding multiple alterations in both total protein levels and phosphorylation events in response to ischemic injury and ultrasound. Application of pathway algorithms reveals numerous protein networks recruited in response to ultrasound including those regulating RNA splicing, cell-cell interactions and cytoskeletal organization. Our dataset also permits the informatic prediction of potential kinases responsible for the modifications detected. Taken together, our findings begin to reveal the endothelial proteomic response to ultrasound and suggest potential targets for future studies of the protective effects of ultrasound in the ischemic heart.

1. Introduction

Therapeutic ultrasound has been used as a tool for thrombolysis during myocardial infarction in multiple animal studies resulting in decreased infarct size, exhibiting potential for use as a clinical intervention [1–3]. However, beneficial effects of ultrasound (US) have also been demonstrated independently of vessel recanalization [4], indicating that it is able to elicit a biological response in heart tissue, in addition to that of physical clot lysis. Our previous studies have indeed confirmed direct tissue-salvaging effects of ultrasound [5].

Heart tissue is made up of many cell types including endothelial cells, which, forming the intimal lining of all blood vessels, are the closest cells to a vessel-obstructing thrombus and play a critical role in mediating ischemia-reperfusion induced injury [6–8]. We have demonstrated ultrasound-induced cellular responses of primary cultured cardiac endothelial cells, where production of multiple vasoactive compounds is altered upon ultrasound stimulation. Additionally, we found that ultrasound is able to increase endothelial cell viability following ischemia-like injury in vitro induced by oxygen glucose deprivation (OGD) [5]. The signaling events underlying the change in vasoactive compound production and cytoprotection have not been explored.

Studies of endothelial cell response to ultrasound have so far focused on individual signaling pathways or single proteins [5,9]. However, protein functions and signaling networks are regulated by multiple rapid phosphorylation events simultaneously. A reversible post-translational modification, protein phosphorylation allows signal transduction pathways to regulate several protein functions, including protein-protein interactions, enzyme activity and protein stability, that underly fundamental biological functions such as cell cycle regulation, survival, and energy metabolism [10,11]. As a potential therapeutic modality, it is important to define the nature and the mechanistic underpinning of the cellular response to ultrasound at this level.

This work represents a comprehensive approach to define how biochemical signaling is impacted by ultrasound. In this study, we have employed unbiased proteomics as an avenue to understand the cardiac endothelial response to ischemic insult as well as the response to ultrasound under normal and ischemic conditions. Global proteomic analysis allows us to produce a detailed analysis of acute changes in the expression levels and post-translational modifications of endothelial proteins in response to ischemic injury and low-intensity therapeutic ultrasound.

2. Materials and methods

Experiments were performed according to the National Institutes of Health Guidelines for the Use and Care of Laboratory Animals. Protocols were approved by the Institutional Animal Care and Use Committee of Oregon Health and Science University. Hearts from 5 mice were pooled to generate each culture; 15 mice in total were used to generate cultures for proteomic analysis (n = 3 independent cultures) and 35 mice were used to generate cultures for Western blotting (n = 7 independent cultures).

2.1. Endothelial cell isolation and culture

Mouse cardiac endothelial cells were isolated by selection using CD31- and CD102- conjugated Dynabeads. Hearts from 8-week old male C57BL6 mice (Charles River Laboratory, Wilmington, MA), were digested in collagenase type II (Worthington Biochemical) and triturated. Cells were pelleted, re-suspended in PBS containing 0.1% BSA and incubated with CD31 antibody (BD Pharmingen)-coated sheep anti-rat Dynabeads (Invitrogen) for 40 min at room temperature. The cell-Dynabead suspension was subsequently mounted on a magnetic separator; the CD31 Dynabead- bound cells were plated on collagen type IV (Sigma, St. Louis, MO)-coated flasks in high glucose DMEM supplemented with 20% fetal bovine serum, 50 μg/mL gentamycin, 2 mmol/L glutamine, 100 μg/mL heparin, 100 μg/mL endothelial cell growth supplement, endothelial mitogen (Biomedical Technologies Inc., Stoughton, MA), while non-bound cells were discarded. Once confluent, cells were detached (0.05% trypsin-EDTA, Sigma) and sorted, as above, using CD102 antibody (BD Pharmingen)-coated sheep anti-rat Dynabeads and cultured. Once confluent, cells with bound beads were removed; the remaining cells were plated in 12- well plates and grown to confluence.

2.2. Oxygen - Glucose Deprivation (OGD)

Cardiac endothelial cells were subjected to OGD for 2 h at 37 °C in an anaerobic chamber (Coy Laboratory Products, Grass Lake, MI) filled with an anoxic gas mixture (5% CO2, 5% H2, and 90% N2). The oxygen concentration was maintained at 0 ppm using a palladium catalyst. Anoxic conditions were monitored by an oxygen monitor (Oxygen-Hydrogen Gas Analyzer; Coy Laboratory Products) within the chamber. To initiate OGD, culture medium was removed, the cells were rinsed 3 times with DPBS, and replaced with glucose-free DMEM (Invitrogen) before placement in the anaerobic chamber. OGD was terminated by removing cells from the chamber, cells were rinsed 3 times with DPBS and the wells filled with degassed high glucose DMEM (Invitrogen) and sealed with MicroAmp optical adhesive film (Invitrogen), making sure no air bubbles were present and immediately exposed to ultrasound stimulation.

2.3. Ultrasound stimulation

Cells, in 12-well cell culture plates, were filled with degassed high glucose DMEM (Invitrogen), sealed with MicroAmp optical adhesive film (Invitrogen), and were submerged in an acrylic tank filled with 0.2 μm filtered and degassed water at 37.4 °C. A single-element high intensity focused ultrasound transducer operating at 250 kHz was used to insonify cells (H-171, Sonic Concepts, Bothell, WA). It was operated by a power amplifier (either RAM 5000, Ritec Inc., Warwick, RI or 600A225 Amplifier Research, Souderton, PA) in conjunction with an electrical impedance matching network. Before ultrasound stimulation, the transducer was calibrated for 1.2 MPa peak rarefactional acoustic pressure amplitude at the focus using a calibrated hydrophone (Reson TC4038, Teledyne Reson, Slangerup, Denmark). The transducer was mounted to a computer-controlled three-axis translation system (Velmex Inc., Bloomfield, NY) and positioned such that the transducer focus was at the location of the cultured cell surface. During US application, the transducer was moved, or scanned, over the central area (1 cm²) of each well. The scan consisted of ten back and forth 1-cm lines separated by 1 mm. Two such scans were performed per well and then the adjacent well was scanned and so on; each well was scanned for approximately 2 min. As the transducer was scanned, it was pulsed with 50 cycles at a pulse repetition frequency of 50 Hz which corresponds to a 1% duty cycle.

2.4. Sample collection

Following US stimulation, culture plates were extracted from the tank, the adhesive film removed, and left at room temperature for 15 min, as previously [5]. Since scanning of an entire 12-well plate takes approximately 24 min (2 min/well), 15 min were counted from the commencement of scanning of the 7th well, i.e. mid-way through the plate. Control plates were also filled with phenol red-free high glucose DMEM, sealed, and placed in the tank for the duration of US stimulation and left at RT until harvest to simulate identical experimental conditions as the US-exposed cultures. Two full 12 well plates were used per experimental condition, samples from 24 wells (50uL lysis buffer/well) were pooled for each sample. For proteomic and phosphoproteomic analysis, ECs were scraped in ice-cold lysis buffer (10 mM HEPES, pH 7.9, 10 mM Potassium Chloride, 0.1 mM Ethylenediaminetetraacetic Acid (EDTA), 0.1 mM Ethylene Glycol Tetraacetic Acid (EGTA), Sigma Phosphatase Inhibitor Cocktail 2, Sigma Phosphatase Inhibitor Cocktail 3 (both at10 uL/mL), cOmplete Mini EDTA-free Protease Inhibitor Cocktail (1 tablet/10 mL; Roche Diagnostics, Indianapolis, IN, USA), 0.5 mM Phenylmethanesulfonyl Fluoride (PMSF) and 0.5% NP40) and frozen on dry ice. For Western blot analysis, cells were lysed in ice-cold RIPA buffer containing cOmplete Mini EDTA-free Protease Inhibitor Cocktail and PhosSTOP Phosphate Inhibitor Cocktail (1 tablet/10 mL; both Roche Diagnostics).

2.5. Sample preparation

Mouse cardiac endothelial cell samples were analyzed at the Pacific Northwest National Laboratory (PNNL). Cell suspensions in the previous lysis buffer (10 mM HEPES, pH 7.9, 10 mM Potassium Chloride, 0.1 mM Ethylenediaminetetraacetic Acid (EDTA), 0.1 mM Ethylene Glycol Tetraacetic Acid (EGTA), Sigma Phosphatase Inhibitor Cocktail 2, Sigma Phosphatase Inhibitor Cocktail 3 (both at10 uL/mL), cOmplete Mini EDTA-free Protease Inhibitor Cocktail (1 tablet/10 mL; Roche Diagnostics, Indianapolis, IN, USA), 0.5 mM Phenylmethanesulfonyl Fluoride (PMSF) and 0.5% NP40) were centrifuged for 30 min at 1000 xg and then washed with 2 mL of NH₄HCO₃, pH 7.8 three times. Cell disruption/denaturing buffer (0.5 mL, 8 M Urea in NH₄HCO₃, pH 7.8) was added to the final cell pellets and they were homogenized for 5 min using a Qiagen TissueRuptor followed by a BCA assay (Pierce) to determine protein concentration.

2.5.1. Proteomics reagents

All chemicals and reagents were purchased from Sigma-Aldrich (St. Louis, MO) unless stated otherwise. Ammonium bicarbonate and acetonitrile were purchased from Fisher Scientific (Pittsburgh, PA), and Sequencing Grade Modified Trypsin was purchased from Promega (Madison, WI). Bicinchoninic acid (BCA) assay reagents and standards were obtained from Pierce (Pierce ThermoFisher Scientific, Waltham, MA); Ni-NTA-agarose beads were obtained from QIAGEN (Valencia, CA); purified, deionized water, >18 MW, (Nanopure Infinity ultrapure water system, Barnstead, Dubuque, IA) was used to make all aqueous buffers and solutions.

2.5.2. Protein extraction and trypsin digest

Proteins were denatured and reduced in this solution with 5 mM Dithiothreitol (DTT), followed by alkylation of free sulfhydryl groups with 10 mM Iodoacetamide in the dark with each reaction performed for 1 h at 37 °C with constant shaking at 1000 rpm a Thermomixer R (Eppendorf, Westbury, NY).

Samples were diluted 8-fold with 50 mM Ammonium Bicarbonate buffer pH 7.8 containing 1 mM CaCl2. Trypsin digestion was performed with Sequencing Grade Modified Trypsin prepared according to the manufacturer’s instructions. Trypsin was added to protein samples at a 1:50 (wt/wt) trypsin-to-protein ratio and samples were incubated for 3 h at 37 °C in Thermomixer R. After 3 h’s incubation, digestion reaction was stopped by acidifying samples to 0.1% Trifluoroacetic Acid (TFA) with 10% TFA stock solution. After 15 min 4000 xg centrifugation, samples were transferred to a fresh tube and stored at −80 °C until the next processing step.

Tryptic peptides were desalted, first, via reversed-phase SCX Solid-Phase Extraction (SPE) columns (Discovery-SCX, SUPELCO, Bellefonte, PA) using 80:20 ACN: 500 mM Triethylammonium Bicarbonate (TEAB) for peptides elution. Peptide samples were acidified to 0.1% TFA and sample volume was adjusted to 5 mL for each sample after concentration peptide samples in SpeedVac vacuum concentrator (Thermo Savant, Holbrook, NY). Next sequential desalting was performed via appropriately selected C18 Strata C18-E SPE columns (Phenomenex, Torrance, CA). Peptides were eluted from the SPE column with 80% MeCN: 20% water and concentrated in the SpeedVac vacuum concentrator.

A BCA Protein Assay was performed at this step to determine relative peptide concentration and 100 μg cleaned peptide aliquot from each sample was dried in SpeedVac and then stored at −80 °C until needed for further manipulations.

2.5.3. Phosphopeptide enrichment using IMAC

Magnetic Fe³⁺-NTA-agarose beads were freshly prepared using the Ni-NTA-agarose beads (QIAGEN) as previously described [12,13]. Dry peptides were reconstituted in IMAC binding/wash buffer (80% Acetonitrile, 0.1% TFA) and incubated with the 5% immobilized bead suspension for 30 min at RT. After incubation, the beads were washed 4 times with an IMAC wash buffer. After washing step, captured phosphopeptides were eluted from the beads using 1:1 (Acetonitrile:5% Ammonia) pH 8.0. Samples were eluted directly into LCMS vials, dried in SpeedVac and were reconstituted with 2% ACN, 0.1% TEA for LC-MS/MS analysis.

2.6. Instrument analysis

Global and phosphopeptide enriched samples were subjected to a custom high mass accuracy LC-MS/MS system as previously described [14], where the LC component consisted of automated reversed-phase columns prepared in-house by slurry packing 3-μm Jupiter C18(Phenomenex) into 35-cm x 360 μm o.d. x 75 μm i.d fused silica (Polymicro Technologies Inc.). The MS component consisted of a Q Exactive HF Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Scientific) outfitted with a custom electrospray ionization interface. Electrospray emitters were custom made using 360 μm o.d. x 20 μm i.d. chemically etched fused silica capillary. Analysis of the phosphoproteome samples applied similar conditions as used in the global proteome sample analysis, except that the spray voltage was 2.2 kV. All other instrument conditions were set as previously described [14].

2.7. Data analysis

Identification and quantification of the detected peptide peaks were performed using the Accurate Mass and Time (AMT) tag approach [15]. Briefly, the generated LC-MS/MS data was searched using MS-GF v6432 (09/08/2011); peptide search using MS-GF plus was performed with full tryptic digestion and allowed a maximum of two missed cleavages against a mouse UniProt database (version Oct 5, 2016, 26,460 entries) for both global and phosphoproteome datasets. The acetylation (protein N-term), oxidation (M) and phospho (STY) were set as variable modifications and the carbamidomethyl (C) was set as a fixed modification for phosphoproteome results. The minimal peptide length for identification was set as 5. Results were filtered with a score greater than or equal to 1E-10, resulting in a false discovery rate of ~1% at the level of peptides, which was then used to populate their respective AMT tag databases [16,17]. Multiple in-house developed informatics tools (publicly available http://ncrr.pnnl.gov/software) were used to process the LC-MS data and correlate the resulting LC-MS features to an AMT tag database that contained accurate mass and LC separation elution time information for peptide tags generated from serum proteins. Among the tools used were algorithms for peak-picking and for determining isotopic distributions and charge states [18]. Further downstream data analysis incorporated all possible detected peptides into a visualization program, VIPER, to correlate LC-MS features to the peptide identifications in the AMT tag database [19]. Peptide quantification was based upon the detected and matched peak intensity profile over elution time as calculated in VIPER. Protein level quantification was based upon the R-rollup approach as performed and described in Dante [20] (http://omics.pnl.gov). An overview of protein-protein interactions was generated by the Search Tool for the Retrieval of Interacting Genes (STRING) [21]. All raw data are deposited in MassIVE, accession number MSV000086652, and ProteomeXchange, accession number PXD023331.

2.8. Western blot analysis

Cells were lysed in ice-cold RIPA buffer containing protease (cOmplete tablets mini) and phosphate (PhosSTOP EASYpack) inhibitors (both Roche Diagnostics, Mannheim, Germany) and centrifuged, the supernatant was collected. Denatured protein samples (25μg) were separated by gel electrophoresis for 50 min at 200 V and transferred onto polyvinylidene difluoride membranes for 105 min at 30 V. Blots were blocked in 5% dry milk and incubated at 4 °C overnight with rabbit polyclonal antibody against NHP2L1 (1:500; #15802–1-AP), RABAC (PRAF1; 1:250; #10542–1-AP, both Proteintech, Rosemount, IL), BRIX1 (1:1000, NBP1–91708, Novus Biologicals, Centennial, CO) and rabbit monoclonal antibody against Sec61B (1:500; #14648, Cell Signaling, Danvers, MA) and PKR (EIF2AK2; 1:2000; ab184257,abcam). The signal was visualized using Amersham ECL anti-rabbit horseradish peroxidase-linked (1:1000; Cytiva, Marlborough, MA) supersignal chemiluminescent reagents (Thermo Fisher Scientific) and a Chemidoc Touch Imaging System (BIO-RAD, Hercules, CA). Blots were stripped using Restore Western Blot Stripping Buffer (Thermo Fisher Scientific) and reprobed using mouse monoclonal antibody against β actin (1:8000; NB600–501, Novus Biologicals) and reimaged. Densitometry was quantified using Image Studio Lite, version 5.2.5 (LI-COR Biosciences, Lincoln, NE); each protein was normalized relative to β actin.

2.9. Statistical analysis

For (phospho)proteomic data, generated peptide relative abundance values were Log₂ transformed and used for an ANOVA analysis and correlation plotting within Dante [20](http://omics.pnl.gov). Relative protein abundance was estimated by using median peptide intensity values [22], requiring a minimum of 2 unique peptides. Peptide level results for global analysis were first compiled into protein values utilizing the R-rollup approach as a feature in Dante [20] prior to ANOVA analysis and correlation plots. Statistical comparisons were based upon pval <0.05 criteria across three independent biological samples replicates per condition.

Kinase Set Enrichment Analysis (KSEA) was performed using known phosphorylation sites from Phosphosite Plus database (https://www.phosphosite.org/). Motif analysis was performed using WebLogo [23].

For analysis of Western blot data, values were analyzed by either paired t-test (2 conditions) or 1-way ANOVA with Tukey’s multiple comparisons test (3 conditions) using GraphPad Prism 9.1.0.

3. Results and discussion

We applied quantitative phosphoproteomics to determine specific proteins and residues of cardiac endothelial signaling pathways regulated by ischemia and ultrasound stimulation, under both basal and ischemic (OGD) conditions. Briefly, endothelial cells cultured from adult murine hearts underwent OGD and/or ultrasound stimulation and subsequent phosphoproteomic analysis, as depicted in Fig. 1A. Phosphopeptide enrichment coupled with global MS analysis allowed us to quantify changes in both total protein abundance and in the relative amount of phosphorylation of endothelial proteome. The experimental paradigm (Fig. 1B) was designed to mimic the metabolic state present following an ischemic event, such as myocardial infarction, in order to determine the effects of both ischemia and ultrasound on cardiac endothelial cells. To demonstrate safety of US stimulation, cells were labelled with propidium iodide (PI), uptake of which indicates cell damage due to loss of cell membrane integrity. As expected for our ultrasound parameters (250 kHz) and absence of either microbubbles or pharmacological agent, we observed no PI uptake upon exposure to US (Fig. S1).

Fig. 1.

Schematic of experimental workflow (A): Heart endothelial cells, magnetically selected and grown to confluence, undergo OGD and/or US stimulation. Cells are harvested, undergo tryptic digest and phosphoproteomic analysis is carried out. (B) Time course of sample treatments.

3.1. Quantitative analysis of cardiac endothelial proteome - regulation by ischemia and ultrasound

We investigated changes in cellular protein expression of cardiac endothelial cells induced by ischemia, ultrasound, and the combination of both ischemia and ultrasound. Overall, 3033 total proteins were detected with an average of 2630 proteins detected and quantified for each sample; of these, the abundance of 295 proteins were significantly altered (P < 0.05) when combined across all comparisons between conditions (see methods for details), as depicted by the heatmap in Fig. 2A. Interestingly of the 295 proteins changed across all conditions, little overlap between the conditions was observed, with no single protein being regulated in common among all conditions (Fig. 2B). The majority of proteins with altered expression levels were in response to ultrasound under normoxic conditions (baseline; BL), with 156 uniquely altered in this condition compared to unstimulated cells. Expression of 69 proteins were uniquely altered by US following OGD and only 46 proteins showed uniquely altered expression levels in response to OGD alone (Fig. 2B). These largely distinct proteomic responses are also depicted in PCA plots (Fig. S2), showing distinct clustering of data from each experimental condition.

Fig. 2.

Alterations in total proteins following US and OGD. (A) Heatmap representation of protein levels for all significantly altered proteins, p < 0.05. Each row represents log₂ ratios for each protein expressed in three independent samples of each experimental condition (BL, baseline; US, ultrasound: OGD, oxygen-glucose deprivation; OGD + US). (B) Venn diagram demonstrating the number of unique and shared proteins altered in each experimental condition.

3.2. OGD and US elicit distinct proteomic responses in cardiac endothelial cells

To better characterize these differences, we conducted pairwise comparisons of each experimental condition to its control group (normoxic baseline (BL) for US and OGD; OGD for OGD + US). These analyses allowed us to identify specific proteins that were either up- or down-regulated in response to each stimulus, as depicted in heatmaps in Fig. 3A-C. A broad range in magnitude of response was observed for each protein and condition (Fig. 3D-F). We found that ischemia (OGD) altered expression levels of 61 proteins, compared to baseline. Ultrasound stimulation led to regulation of 172 proteins under normoxic conditions, and 86 under ischemic conditions. Together with Fig. 2, these results indicate that the cellular response to ultrasound is more pronounced than it is to an ischemic event at the protein level, and that both of these stimuli elicit distinct protein expression profiles. Furthermore, the total number of proteins regulated by ultrasound is halved by ischemia. Our results demonstrate that the cellular response to ischemia and/or ultrasound appears to be a coordinated one, with upregulation of some proteins, and downregulation of others. This is an interesting observation as the vasoactive proteins and compounds we previously studied in cardiac endothelial cells were all upregulated by ultrasound [5]; this comprehensive study further delineates the cellular effect of ultrasound on these cells, demonstrating that ultrasound is also able to downregulate many proteins, both at baseline and following ischemia.

Fig. 3.

Pairwise comparisons of total protein levels. Heatmap representation of significant protein changes between BL and US (A), BL and OGD (B), and OGD and OGD + US; p < 0.05, red indicates increased, and green decreased, protein abundance. Volcano plots demonstrate the fold change of detected proteins altered in each comparison, log₂ FC was plotted on the x axis and -log₁₀P was plotted on the y-axis. Proteins with P < 0.05 were considered significant, those proteins are indicated as red dots (D-F; n = 3/condition).

Informatic analyses of these datasets yielded protein networks that provide some insights into the molecular pathways affected (Fig. 4). Ultrasound simulation alone produced a very complex interaction map (Fig. 4A). Several Kyoto Encyclopedia of Gene and Genomes (KEGG) terms were associated with the endothelial response to ultrasound at baseline including vesicular transport, focal adhesion, protein turnover and cellular metabolism (Fig. 4B; specific proteins associated with each KEGG pathway identified are listed in Supplemental Table 1). Surprisingly, this was not the case for the other conditions (OGD or OGD + US) which resulted in no associated KEGG terms, despite dense protein associations also including protein trafficking, cytoskeletal and transcriptional components (Fig. 4C,D).

Fig. 4.

Protein network analysis following US and OGD. Interactions of proteins regulated at baseline by US (A) and OGD (C), and by US after OGD (D). KEGG pathways identified for proteins regulated by US at baseline (B).

3.3. Differential response of normoxic and ischemic endothelial cells to ultrasound

Global proteomic analysis reveals that endothelial cell response to US is determined by cell state (Fig. 5). Of the 172 proteins significantly regulated by ultrasound under normoxic conditions, and the 86 after ischemia (Fig. 3A&C), only 9 proteins are common between the two conditions (Fig. 5A), revealing a differential endothelial proteomic response to ultrasound dependent on cell state. These 9 proteins regulated under both conditions may form the universal protein signature in response to ultrasound in cardiac endothelial cells. Plotting the linear intensity of these 9 proteins significantly altered by ultrasound under both conditions, normalized to their respective controls, reveals that 4 proteins are down-regulated under both normoxic and ischemic conditions, 4 up-regulated, and one, interferon-induced, double-stranded RNA-activated protein kinase (EIF2AK2), is upregulated under normoxic conditions and down-regulated following ischemia (Fig. 5B).

Fig. 5.

Differential protein expression in response to US at baseline and following OGD. (A) Venn diagram indicating number of proteins regulated by US at baseline and after OGD. (B) Magnitude and direction of change in expression of proteins regulated by US under both BL and OGD conditions relative to respective controls (linear intensity; n = 3/condition). Validation of proteomic data was carried out by Western blotting for EIF2AK2 (C) and PRAF1 (D; n = 4–5/condition).

The proteins upregulated under both normoxic and ischemic conditions are Cell Division Control Protein 41 Homolog (CDC42), Vesicle-Associated Membrane Protein 3 (VAMP3), Electron Transfer Flavoprotein β Subunit (ETFB) and Prenylated Rab acceptor protein 1 (PRAF1, alternate name Rabac1), the latter being the most robustly upregulated, with a 2.53 +/− 0.70 fold increase in normoxic cells and 1.31 +/− 0.30 in ischemic cells, compared to their normoxic/ischemic controls respectively. VAMP3, while upregulated under both conditions, is the protein with the largest upregulation observed under ischemic conditions, with a fold change of 1.51 +/− 0.59 compared to OGD alone. Both SNARE proteins, PRAF1 and VAMP3, are involved in Golgi sorting and vesicular trafficking [24,25]. While relatively little is known about PRAF1, with no studies describing its function in the heart or vasculature, VAMP3 is known to contribute to endothelial angiopoietin 2 (Angpt2) exocytosis, a vascular destabilizing factor, and biomarker of poor outcome in ischemic heart disease [26,27]. Additionally, VAMP3 is involved in trafficking of GLUT4 and CD36 in other cardiac cells types [28,29]. Whether upregulation in cardiac endothelial cells following MI is beneficial remains to be determined.

ETFB and CDC42 are both involved in mechanotransduction; ETFB in mechanoregulation of fibroblasts and CDC42 mediates endothelial response to sheer stress [30,31]. ETFB additionally is active in mitochondrial respiration and up-regulated in cardiac muscle of the diabetic rat [32]. Regulation of both these proteins by ultrasound may have implications following both an ischemic insult such as a heart attack but also other pathological conditions such as atherosclerosis [33–35].

The proteins downregulated under both normoxic and ischemic conditions are Proteasome 26S subunit, non-ATPase 1 (PSMD1), ADP-Ribosylation Factor 5 (ARF5), Eukaryotic Translation Initiation Factor 2 Subunit 3, X-linked (EIF2S3X) and Glutaredoxin-3 (Glrx3). While little is known about the functions of PSMD1 in the heart, it is a ubiquitin-proteasome system (UPS) protein involved in neutralization of damaged and misfolded proteins [36]. ARF5 is a member of the Arf family of GTP-binding proteins, which are key regulators of cell organization, and involved in both protein and lipid vesicular trafficking [37]. Consistent with this role in protein trafficking, Arf5 is implicated β1-integrin endocytosis and angiogenic sprouting [38]; down-regulation of Arf5 by ultrasound may therefore provide a novel avenue of limiting pathological angiogenesis in the heart.

EIF2S3X, the product of an X-linked gene, regulates the rate of protein translation. With higher expression in female than male mouse hearts, it is suggested to contribute to poorer cardiac functional recovery following I/R injury [39]. Down-regulation in cardiac endothelial cells by ultrasound may underlie the protection we have observed both in vitro and in vivo following ischemic insult [5], and also suggests that response to ultrasound may potentially to be sexually dimorphic.

The down-regulation we observe of Glrx3 (alternate name protein kinase C-interacting cousin of thioredoxin, PICOT) regardless of cell state is intriguing. Glrx3 is a negative regulator of cardiac hypertrophy and a positive inotropic regulator, with over-expression preventing further deterioration to the failing heart [40]. It is also implicated in the stress-induced DNA damage response, with Glrx3-deficient mice displaying a phenotype of an impaired anti-oxidant mechanism [41]. The functional effects of acutely decreased Glrx3, as observed in this study, have not been determined, nor have its effects in cardiac endothelial cells.

EIF2AK2 (alternate name Protein kinase R1, PKR), which is differentially regulated by ultrasound depending on cell state, up-regulated under normoxic and down-regulated under ischemic conditions, is a ubiquitously expressed interferon-inducible serine/threonine protein kinase involved in multiple stress-induced cellular responses. The kinase is a negative regulator of translation factor eukaryotic initiation factor 2α (eIF2α), leading to inhibition of general translation, thus regulating cellular apoptosis, proliferation and differentiation. Increased levels of EIF2AK2 are suggested to be involved in the aging process [42]. It has multiple roles in various endothelial cell beds, of relevance to the heart it mediates endothelial senescence, a phenotype characteristic of endothelial dysfunction and cardiovascular disease. Activation of EIF1AK2 contributes to endothelial cell senescence, and conversely experimentally-induced senescence can be reversed by its inhibition in vitro [42,43]. In cardiomyocytes, EIF2AK2 inhibition is also protective against reactive oxygen species (ROS)-induced injury, reducing apoptosis [44]. The decreased abundance of this kinase that we observe upon ultrasound stimulation under ischemic conditions may contribute to the endothelial- and cardio-protection that we have demonstrated both in vivo and in vitro following ischemic insult [5]. However, the upregulation we observe to normoxic endothelial cells by ultrasound may not be beneficial to the healthy heart, indicating that the use of ultrasound therapeutically is context dependent, and should be studied in that manner.

Interestingly, of the 9 proteins we suggest form the universal protein signature of cardiac endothelial response to ultrasound, at least 4 are involved in insulin regulation and/or lipid synthesis: PSMD1, CDC42, VAMP3 and ETFB, with the latter being up-regulated in cardiac tissue of the diabetic heart [28,29,32,36,45–47]. Our data therefore indicate that studies into the use of therapeutic ultrasound in diabetic-induced cardiovascular pathology warrants consideration.

Our data demonstrating largely distinct endothelial cell proteomic responses to ultrasound depending on cell state highlight the importance of studying therapeutic ultrasound in disease-specific models. This is also an important consideration for clinical ultrasound applications, with potentially very different outcomes depending on disease state, with different proteins regulated following a myocardial infarction or other ischemic event, than in healthy tissue. While ultrasound is protective to the ischemic heart and endothelial cells [5], this may not be the case in a healthy state due to a differing proteomic response as we demonstrate here. For example, it should not be assumed that ultrasound would be protective as a potential preconditioning intervention prior to high-risk heart surgery, as the protein signature of the ultrasound response in that setting may be very different than during or after MI, potentially leading to very different consequences.

3.4. Ultrasound normalization of protein levels Dysregulated by ischemia

Having determined that the endothelial proteomic response to ultrasound is largely distinct in normoxic versus ischemic conditions, we next compared the nature of the proteomic response to ischemia with that of ischemic cells stimulated with ultrasound. We found that again, the proteins regulated in both conditions were largely distinct (Fig. 6A). More proteins were regulated by ultrasound in ischemic cells than by ischemia alone, with 78 and 53 proteins altered in abundance respectively, with only 8 proteins common to both conditions. Of the 8 proteins significantly regulated both by ischemia (BL versus OGD) and by ultrasound under ischemic conditions (OGD versus OGD + US), we asked whether ultrasound was able to return protein levels back to those observed in normoxic conditions (BL).

Fig. 6.

Normalization of protein levels by US following OGD. (A) Venn diagram indicating number of proteins regulated by OGD and by OGD + US. Magnitude and direction of change in expression of proteins in response to US of proteins (B) upregulated and (C) downregulated by OGD (linear intensity; n = 3). Validation of proteomic data was carried out by Western blotting for BRX1 (D), NHP2L1 (E) and SEC61B (F; n = 5–7/condition).

Plotting the linear intensity of the 8 proteins identified as significantly altered by ischemia and responsive to ultrasound following ischemia, reveals that three proteins are increased by OGD and all reduced back toward baseline levels by ultrasound (Fig. 6B). These proteins are RNA-Binding Protein 39 (RBM39), Ribosome Biogenesis Protein BRIX1 homolog (BRX1) and EIF2S3X. All 3 proteins are involved in gene expression: splicing, ribosomal biogenesis and protein synthesis respectively, and are all linked to cardiovascular disease [39,48–52]. Preventing the ischemia-induced increase in one, or all, of these proteins may underlie the increase in cell survival of endothelial cells stimulated with ultrasound following OGD [5].

Four proteins were identified that showed the opposite trend – they were down-regulated by OGD, and returned back toward baseline levels upon treatment with ultrasound, these are Receptor-type Tyrosine-protein Phosphatase Beta (PTPRB), Prohibitin (PHB), Protein Transport Protein Sec61 Subunit Beta (SEC61B) and NHP2-like protein 1 (NHP2L1, alternate name: small nuclear ribonucleoprotein, SNU13) (Fig. 6C). No studies describe the regulation of NHP2L1 in endothelial cells or the heart, however the modulation of NHP2L1 levels by ultrasound we observe has the potential to influence multiple biological processes due its function in RNA processing and splicing [53].

Interestingly, PHB has been identified as a therapeutic target in cardiovascular disease [54]. PHB protein levels are reduced in the mouse heart by MI in vivo [55], ischemic regulation that we also observe here in cardiac endothelial cells following ischemia in vitro. Functionally, cardiac-specific over-expression of PHB leads to a reduction in MI-induced mitochondrial fission and apoptosis [55], and also protects cardiomyocytes against hypoxia-induced death by inhibiting cytochrome c release, in line with its role in mitochondrial homeostasis [56]. The role of PHB in endothelial cells has not yet been determined, however the decreased protein levels we observe in response to OGD, and increase by ultrasound following OGD may underlie ultrasound-mediated cell protection of ischemic cells [5].

PTRPB (alternate name vascular endothelial protein tyrosine phosphatase, VE-PTP) is another protein involved in the endothelial response to sheer stress and angiogenesis [57,58]. It is an endothelial-specific phosphatase which stabilizes VE-cadherins junctions by reducing the rate of VE-cadherin internalization [59]. Interestingly, PTRPB is mechanistically linked to CDC42 [58], which makes up our suggested proteomic signature in response to ultrasound, discussed above.

While little is known about the functions of SEC61B in the heart, or in endothelial cells, monocyte SEC61B has been identified as a candidate gene linking systemic inflammation to atherosclerosis [60]. A subunit of the Sec61 translocon complex, SEC61B interacts with microtubules, facilitating the maintenance of endoplasmic reticulum (ER) homeostasis [61]. ER-stress occurs in endothelial cells, as well as other cardiac cell types, upon hypoxia/reoxygenation injury [57,62,63], and has been proposed as a therapeutic target in heart disease [63,64]. The mechanism of protection observed in response to many interventions in MI studies has been attributed to attenuation of the ER-stress response [65–67]; our results suggest that ultrasound-mediated protection following ischemia both in vivo and in vitro [5] may be mediated by normalizing SEC61B levels, and thus attenuating ER stress.

Dynein light chain 1 (DYNLL1; also known as Protein Inhibitor of Neuronal Nitric Oxide Synthase, PIN), was the only protein regulated in cardiac endothelial cells both by ischemia and by ultrasound following ischemia that is not normalized by ultrasound; it is down-regulated by OGD, and further down-regulated with ultrasound (Fig. 6C). DYNLL1 inhibits nitric oxide production by neuronal Nitric Oxide Synthase (nNOS), which contrary to its name is expressed in endothelium, and contributes to basal NO production [68,69]. nNOS is known to protect against MI-induced cardiac remodeling and may play an antiinflammatory role in cardiac vessels [69,70]. Our results suggest that both ischemia and ultrasound may remove basal nNOS inhibition, thus producing more nitric oxide, potentially protecting the heart against ischemic insult; a mechanism warranting further study. Interestingly, we have previously demonstrated increased phospho-eNOS phosphorylation (S1177) upon ultrasound stimulation of cardiac endotheilial cells 15 min post-exposure [5]. Increased eNOS phsophorylaton implies increased NO production downstream. Together these data suggest that ultrasound may increase NO production by endothelial cells by multiple mechanisms.

3.5. Ultrasound stimulated endothelial Phosphoproteome dynamics

In addition to exploring the impact of ischemia and ultrasound on endothelial protein levels, we also analyzed our cell lysates for phosphorylation changes. Our quantitative mass spectrometry approach allowed us to ascertain significant changes in relative phosphorylation for sites throughout the endothelial proteome in response to OGD (192 phospho-peptides), ultrasound (375 phospho-peptides), or ultrasound after OGD (164 phospho-peptides)(Fig. 7A). Similar to our findings for the total proteome in each setting, proteins targeted for post-translational modification can be shared between two conditions, yet the majority of phosphorylation events are unique to each condition (Fig. 7B). Interestingly, a Venn diagram of modified proteins indicates that several proteins produce multiple differentially-modified peptides in each condition (Fig. 7B). Pairwise heatmap and volcano plot comparisons further highlight the distinct response profiles observed in each condition (Fig. 8A-F). To better understand which pathways are engaged by OGD and ultrasound, we employed STRING network analysis to generate large relational maps (Supplemental Fig. 3). Mapping of strongly represented KEGG pathways provides key insights into potential endothelial phospho-signaling alterations to cellular homeostasis (Fig. 9A-C, Supplemental Tables 2–4). Some KEGG pathways appear in all conditions including RNA splicing, proteoglycan metabolism, tight junctions, and actin cytoskeleton regulation. While these pathways are observed in all cases, the specific phospho-proteins identified for each category and condition are sometimes shared and others are unique. Closer examination of specific residues reveals that in a particular protein identical sites can be altered (see Figs. 10B and Fig. 11) or, alternatively, the residues modified in the same protein can vary by condition. For example, we detect multiple phosphorylation events (T1564, S315, and S326) in the tight junction protein Tjp1/ZO-1 (Zona Occludens-1) following ultrasound under normoxic conditions. Only a single site is observed in OGD (S117), but this modification is accompanied by S131 only in OGD plus ultrasound (Supplemental Fig. 4). Importantly, ZO-1 is known to underlie critical endothelial endothelial functions [71] and alterations in its expression and phosphorylation are linked to challenges such as hemodynamic stress [72]. It is also known to be affected by pathologic conditions such as cerebral ischemia [73]. Our findings indicate that while specific pathways within endothelial cells can experience post-translational remodeling due to ischemia or ultrasound, the specific cellular response can also depending on physiologic state. We also observe specific pathways affected uniquely in each setting such as Apelin signaling (US), focal adhesions (OGD) or adherens junctions (OGD and US) (Fig. 9A-C).

Fig. 7.

Alterations in phosphopeptides following US and OGD. (A) Heatmap representation of phosphopeptides for all significantly altered peptides, p < 0.05. Each row represents log₂ ratios for each phosphopeptide regulated in three independent samples of each experimental condition. (B) Venn diagram demonstrating the number of unique and shared peptides altered in each experimental condition.

Fig. 8.

Pairwise comparisons of phosphopeptides. Heatmap representation of significant phosphopeptide changes between BL and US (A), BL and OGD (B), and OGD and OGD + US (C); p < 0.05. Volcano plots demonstrate the fold change of detected peptides altered in each comparison, log₂ FC was plotted on the x axis and -log₁₀P was plotted on the y-axis. Proteins with P < 0.05 were considered significant, those proteins are indicated as red dots (D–F).

Fig. 9.

Phosphopeptide network analysis following US and OGD. KEGG interactions of peptides regulated at baseline by US (A) and OGD (B), and by US after OGD (C). KEGG descriptions listed in supplemental tables 12–14.

Fig. 10.

Signature of phosphopeptide changes in response to US unique to OGD. (A) Venn diagram indicating number of phosphopeptides altered by US at BL and following OGD. (B) Relative fold-change in protein modifications by US at BL and after OGD (linear intensity). (C) KEGG interations of phosphoproteins regulated by US only following OGD.

Fig. 11.

Phosphomotif response to each condition. Relative fold-change (linear intensity) in phosphorylation of a specific phospho-peptide normalized to baseline levels. (A) Residues elevated during OGD and returned toward baseline after ultrasound stimulation. (B) Phosphopeptides which are dephosphorylated during OGD and recover toward baseline levels of phosphorylation with ultrasound.

The overlap of phospho-peptides affected by ultrasound under both normoxic and ischemic conditions can be represented by Venn diagram (Fig. 10A). It is clear that, in both cases, the majority of peptides were unique to ultrasound response under each condition, but a number were shared. In cases where identical peptides and residues could be assessed for each case, we found similar direction and magnitude of modifications (Fig. 10B, Supplemental Table 5). We also conducted STRING analysis similar to Fig. 9 using the proteins represented by the altered phospho-peptides (Fig. 10C) to highlight proteins for which phosphorylation changes occurred uniquely when ultrasound is applied to ischemic cells. While no KEGG pathway terms emerged from this analysis, interestingly numerous protein kinases exist as nodes including: Slk, Lats1, Bmp2k, Ptk2b (FAK2), Mapk3 (ERK) and ARaf, Clk3, Map4k4 and Peak1 (Fig. 10C, blue). Similar to our observations in total proteome dynamics, these findings that normoxic cells exposed to ultrasound lack these modifications only found in ischemic cells indicates that OGD shapes endothelial responses to ultrasound.

Consistent with the response of ischemic cells to therapeutic ultrasound and paralleling our findings for total protein levels, we identified a set of alterations in protein phosphorylation that are OGD-driven and reversed by ultrasound (Fig. 11A,B). In some cases, protein modifications increased by OGD were restored to control levels by ultrasound (Fig. 11A, Supplemental Table 6). The proteins responding in this manner affect multiple aspects of cellular physiology including autophagy (GAPR1), cytoskeletal dynamics (MAP4), protein dephosphorylation (SHIP1), RNA splicing (SRRM2), and epigenetic modification (TASOR/FAM208A/RAP140). Similarly, proteins were found that undergo dephosphorylation in response to OGD and this is reversed by the addition of ultrasound (Fig. 11B, Supplemental Table 6). Here again, we observed a diverse set of cellular functions potentially impacted by these modifications: cytoskeletal dynamics (ARHGC/ARHGEF12, RHG21/ARHGAP21, BNIP2, CLAP1/CLASP1, MAP4), intracellular transport (DYHC1), exocytosis (EXOC2), RNA metabolism and translation (FXR1, IF4B), lysosome function/autophagy (MCLN1/MCOLN1) and intracellular trafficking (MYH9). As with many global phospho-proteomic studies, most of the modifications in our dataset do not yet have known physiologic influence, but some sites have been previously observed in ischemic tumor tissue (SRRM2-S2388, TASOR-S903, and ARHGC/ATHGEG12-T736) [13]. However, within our set of reversed modifications, serine 1943 in the non-muscle myosin heavy chain (MYH9) has been shown by others to inhibit dimerization [74], binding of S100A/Mts1 [75] and very recently was reported to play a role in the nuclear function of MYH9 [76]. Future studies will be required to determine how each of these alterations may contribute to the protective effects of ultrasound during ischemia.

3.6. Motif analysis reveals consensus signature of endothelial response to ultrasound

We employed bioinformatics to determine which kinases or phosphatases may be responsible for the observed changes in the observed post-translational modification observed in each condition. We first used Kinase Set Enrichment Analysis (KSEA) which tests for statistical enrichment of up- or down-regulated phosphopeptides that are known to be substrates of particular kinases [23]. This predicted the activity of multiple kinases (Fig. 12A) including multiple members of the CDK family and protein kinase A, yet this analysis derived from only a limited number of peptides in our data. We performed a more comprehensive analysis by selecting phospho-peptides that are significantly different between conditions and aligning them centered on their phosphorylation sites (Fig. 12B). Unfortunately these alignment comparisons failed to reveal consistent motifs, but it is likely that these sites will prove more informative as additional target motifs are mapped to specific kinases. The candidate kinases revealed here will allow future studies to apply specific inhibitors to determine how each contributes to the post-translational modifications described in our dataset and may serve as potential sites of therapeutic intervention to enhance any protective role from a particular kinase’s activation during myocardial infarction.

Fig. 12.

Phosphopeptide motif analysis. (A) KSEA analysis using substrate consensus to predict which kinases are targeting the peptides identified in each condition. (B) Pileup graphs of consensus sequences from each dataset centered on the phosphorylation site.

4. Conclusions

In this study, we applied an unbiased global proteomic approach to identify the endothelial protein response to ischemia, ultrasound and ischemia followed by ultrasound. A subset of proteins were observed to change their overall level in each condition, reflecting a proteo-dynamic contribution to cellular adaption over short periods of time. Signal transduction events act on more rapid time scales and we also observed up- and down-regulation of phosphorylation events in our proteomic dataset. Informatic analysis revealed specific proteins, peptides, and pathways affected in each condition, demonstrating that ultrasound is able to elicit a more pronounced proteomic reponse than ischemia. Importantly our work highlights that the endothelial response to ultrasound is largely determined by cell state, with a differential (phospho) proteomic response in normoxic and ischemic cells. We also identify 9 proteins suggestive of a universal endothelial protein signature in response to ultrasound, including proteins involved in mechanotransduction and trafficking. Studies demonstrate that therapeutic ultrasound is beneficial to the ischemic heart [4,5], however the cellular pathways mediating this response have not been elucidated. Our work identifies candidate proteins and modifications in endothelial cells that may underlie the therapeutic effects of ultrasound, demonstrating ultrasound-induced normalization of some proteins dysregulated by ischemia, as well as proteins regulated by ultrasound only following ischemia. Future studies focused on these proteins and specific modifications will determine their contributions to improved outcome in the ischemic heart upon delivery of therapeutic ultrasound. Studies aimed at delineating underlying mechanisms of the protective effects of ultrasound will also require similar analyses of other constituent cell types in the heart, revealing common and unique response profiles between cell types. Our work provides a framework for such studies and contributes knowledge of the endothelial (phospho)proteomic response to both ischemia and therapeutic ultrasound.