Quantitative proteomic and phosphoproteomic profiling of ischemic myocardial stunning in swine

Abstract

Despite decades of research on the pathophysiology of myocardial stunning, protein changes and/or phosphorylation status underlying alterations in cardiac function/structure remain inadequately understood. Here, we utilized comprehensive and quantitative proteomic and phosphoproteomic approaches to explore molecular mechanisms of myocardial stunning in swine. The closed-chest swine (n = 5 pigs) were subjected to a 10-min left anterior descending coronary artery (LAD) occlusion producing regional myocardial stunning. Tissues from the ischemic LAD region and a remote nonischemic area of the left ventricle were collected 1 h after reperfusion. Ion current-based proteomics (IonStar) and quantitative phosphoproteomics were employed in parallel to identify alterations in protein level and site-specific phosphorylation changes. A novel swine heart protein database exhibiting high accuracy and low redundancy was developed here to facilitate comprehensive study. Further informatic investigations identified potential protein-protein interactions in stunned myocardium. In total, we quantified 2,099 protein groups and 4,699 phosphorylation sites with only 0.4% missing values. Proteomic analyses revealed downregulation of contractile function and extracellular matrix remodeling. Meanwhile, alterations in phosphorylation linked with contractile dysfunction and apoptotic cell death were uncovered. NetworKIN/STRING analysis predicted regulatory kinases responsible for altered phosphosites, such as protein kinase C-mediated phosphorylation of cardiac troponin I-S199 and CaMKII-mediated phosphorylation of phospholamban-T17. In summary, the ion current-based proteomics and phosphoproteomics reliably identified novel alterations in protein content and phosphorylation contributing to contractile dysfunction, extracellular matrix (ECM) damage, and programmed cell death in stunned myocardium, which corroborate well with our physiological observations. Moreover, this work developed a comprehensive database of the swine heart proteome, a highly valuable resource for future translational research in porcine models with cardiovascular diseases.

NEW & NOTEWORTHY We first used ion current-based proteomics and phosphoproteomics to reliably identify novel alterations in protein expression and phosphorylation contributing to contractile dysfunction, extracellular matrix (ECM) damage, and programmed cell death in stunned myocardium and developed a comprehensive swine heart-specific proteome database, which provides a valuable resource for future research in porcine models of cardiovascular diseases.

INTRODUCTION

Transient contractile dysfunction following brief ischemia that completely normalizes within 24 h after reperfusion has been identified as “stunned myocardium” (7, 30, 31, 38). Stunned myocardium commonly occurs to various degrees after episodes of supply/demand imbalance that precipitate regional ischemia with or without symptomatic angina. Pathological features of acutely stunned myocardium include structural alterations of the collagen matrix that occur in the absence of myocyte sarcolemmal disruption on electron microscopy or pathological evidence of necrosis (75). Nevertheless, more recent studies have demonstrated that even brief ischemia can be accompanied by myocyte injury as reflected by an increase in myocyte apoptosis along with cardiac troponin I (cTnI) release from the heart (66). As a result, repetitive episodes of ischemia in chronic stenosis models can lead to regional myocyte loss and, over time, the development of chronic contractile dysfunction typical of hibernating myocardium (36) or even ischemic cardiomyopathy (14). Likewise, myocardial stunning contributes a partially reversible component to the contractile dysfunction that is present immediately following reperfusion of an ST segment elevation myocardial infarction (7).

Despite considerable investigation, the underlying causes of myocardial stunning remain unclear. There is general agreement that the generation of oxygen-derived free radicals and intracellular calcium overload are primary mediators of postischemic myocardial dysfunction (6). It has been suggested that the “oxyradical hypothesis” and “calcium hypothesis” are not mutually exclusive, but likely represent different components of the same pathophysiological process and may promote myocardial dysfunction by damaging proteins of the contractile apparatus or sarcoplasmic reticulum (12, 37). Nevertheless, in contrast to myocardial infarction, the extent to which proteins are altered via either degradation or posttranslational modifications (PTMs) in stunned myocardium remains incompletely understood. Most of our current understanding is based on in vitro observations in isolated myocytes during simulated ischemia and isolated buffer-perfused rodent hearts subjected to global ischemia. Unfortunately, these have not been consistently reproduced during regional ischemia in large animal models of myocardial stunning in vivo. For example, whereas cTnI proteolysis was proposed as a molecular mechanism of stunning following global ischemia in isolated rat hearts (19), this was not found during regional ischemia in swine (60, 66). Rather, elevated preload without ischemia could produce cTnI degradation in the absence of ischemia in isolated rat hearts as well as following preload elevation in swine (15, 65). These discordant observations underscore the need to examine molecular mechanisms of stunning in a large animal model.

During the past two decades, quantitative proteomic technologies have emerged as powerful tools to explore molecular mechanisms of disease, evaluate new therapeutics, and identify novel biomarkers in an unbiased fashion (56, 61). To date, proteomic analysis of stunned myocardium following ischemia-reperfusion injury has primarily been performed using two-dimensional gel electrophoresis (2DE) on heart tissue derived from rats (16) and rabbits (69), with identification of 39 and 53 altered protein spots, respectively. In addition to the aforementioned limitations of isolated hearts subjected to global ischemia, 2DE-based proteomic approaches are limited by a shallow depth of analysis, poor separation of highly hydrophobic proteins, and suboptimal quantitative sensitivity, accuracy, and precision (8, 49). Furthermore, the observation that protein differences between stunned and control myocardium were typically the result of isoelectric point shifts likely caused by chemical modifications (69) suggests that PTMs play a significant role in mediating alterations in protein function that underlie postischemic contractile dysfunction (53, 68).

Quantitative profiling of the phosphoproteome in stunned myocardium following transient regional ischemia holds the potential to identify novel, site-specific alterations in protein phosphorylation status that play a central role in the pathophysiology of postischemic myocardial stunning. However, this has been difficult to achieve previously, particularly in clinically relevant large animal models, because of limitations in technology and informatics resources including the lack of an appropriately annotated species-specific protein sequence database, challenges related to analysis of multiple biological samples, and the large protein concentration range of the myocardial proteome causing difficulty identifying lower-abundance proteins due to the presence of high-abundance muscle proteins (24, 56). To address these limitations, we examined myocardial proteins after 10 min of regional ischemia in closed-chest swine using a new proteomics pipeline. We enhanced myocardial protein identification by developing a porcine-specific cardiac protein sequence database based on RNA sequencing (RNA-seq) of normal swine myocardium. Identification of myocardial proteins altered by brief ischemia was enhanced by employing a cutting-edge liquid chromatography-mass spectrometry (LC-MS)-based proteomics strategy (IonStar). This method represents a significant improvement over 2DE methods, due to its increased depth of analysis, enhanced quantitative capacity, and ability to compare more biological replicates (54, 56, 57). In parallel, we employed quantitative phosphoproteomics to assess ~4,700 phosphorylation sites, thereby providing the most comprehensive cardiac phosphoproteomic analysis in any species to date. Taken together with assessment of phosphosite-kinase interactions via combined NetworKIN and STRING analysis, our results provide novel insight regarding the molecular mechanisms of three principal protein networks identified to be involved in myocardial stunning, namely, 1) contractile protein dysfunction, 2) extracellular matrix protein degradation, and 3) induction of apoptosis.

METHODS

All procedures and protocols conformed to institutional guidelines for the care and use of animals in research and were approved by the University at Buffalo Institutional Animal Care and Use Committee. Studies were conducted in male Yorkshire-crossbred pigs [n = 5 pigs; 48 ± 4 (SE) kg; 3–4 mo of age; WBB Farm, Alden, NY] that represent a subset of animals that were included in a previous report by our laboratory demonstrating cardiac troponin I (cTnI) release and apoptosis following brief ischemia in swine (66). The experimental design and workflow are shown in Fig. 1.

Fig. 1.

Experimental design and procedure of LC-MS-based proteomics and quantitative phosphoproteomics. Myocardial samples with or without 10-min left anterior descending coronary artery (LAD) occlusion followed by 1-h reperfusion (n = 10 samples, 5 from stunned LAD region vs. 5 from the remote normal region of the left ventricle) were collected and processed using SEPOD. The digested peptides then were separated into two fractions: 1) phosphoproteomic analysis including phosphopeptide enrichment, 10-plex tandem mass tag (TMT) labeling, high-pH fractionation, LC-MS/MS analysis, and phosphosite quantitation; and 2) label-free proteomics analysis with highly reproducible nano-LC separation followed by ultrahigh-field Orbitrap Fusion Lumos MS and ion current-based MS1 quantitation (IonStar). Protein identification was performed with a customized RNA-sequencing (RNA-seq)-derived swine heart protein database. The functional annotation, protein-protein interactions, and predicted upstream kinases were further analyzed in silico.

Large animal instrumentation.

Pigs were sedated with a Telazol (100 mg/mL)-xylazine (100 mg/mL) mixture (0.04 mL/kg im) and maintained on a continuous intravenous infusion of propofol (5–10 mg·kg⁻¹·h⁻¹) while being mechanically ventilated with supplemental oxygen. A 7-Fr sheath was placed into the right carotid artery; through this sheath a 5-Fr catheter (Millar) was positioned in the left ventricle for continuous pressure measurement. The side port of the introducer was used to measure arterial pressure. A second 7-Fr sheath was then placed into the left carotid artery to advance a balloon angioplasty catheter into the left anterior descending coronary artery (LAD) as described below. Animals were heparinized (100 U/kg iv), and hemodynamics were allowed to equilibrate (~20 min) before beginning the protocol. Hemodynamic parameters and systolic wall thickening using echocardiography (GE Vivid 7) were assessed as previously described (66).

After baseline measurements were completed, an appropriately sized balloon angioplasty catheter (Maverick, 3.0–4.0 mm; Boston Scientific) was advanced distal to the second diagonal branch of the LAD through a 6-Fr guiding catheter (Cordis Corporation, Miami Lakes, FL) under fluoroscopic guidance. To minimize the occurrence of lethal arrhythmias, all animals were pretreated with amiodarone (5 mg/kg iv) and lidocaine (1.5 mg/kg iv) boluses followed by continuous infusions (amiodarone, 0.04 mg·kg⁻¹·min⁻¹; lidocaine, 0.05 mg·kg⁻¹·min⁻¹) during coronary occlusion and 10 min into the reperfusion period. Balloon occlusion was documented with contrast angiography, and hemodynamic and functional measurements were repeated. After 10 min, the balloon was deflated, and reperfusion was confirmed angiographically. Blood sampling, hemodynamic measurements, and echocardiography were repeated at the end of ischemia and 10 min, 30 min, and 1 h after reperfusion (Table S1). Hearts were then arrested with intracardiac KCl under deep isoflurane anesthesia. Myocardial tissue from the ischemic LAD region (stunned myocardium) and normal remote area (posterior wall; “control myocardium”) of the left ventricle was excised for postmortem proteomic and phosphoproteomic analyses (Fig. 1 and described below), histopathology, and 2,3,5-triphenyltetrazolium chloride (TTC) staining (66).

Protein extraction, precipitation, and digestion.

The protein extraction, precipitation, and digestion follow our previous procedure, a surfactant cocktail-aided extraction/precipitation/on-pellet digestion (SEPOD; 2). Briefly, frozen heart samples (5 g) were quickly grinded in liquid nitrogen and homogenized with a Polytron homogenizer (Kinematica AG, Switzerland) in lysis buffer [50 mM Tris-formic acid (FA), 150 mM NaCl, 0.5% sodium deoxycholate, 2% SDS, and 2% Nonidet P-40, pH 8.0] supplemented with cOmplete protease inhibitor cocktail tablet and PhosSTOP phosphatase inhibitor cocktail tablet (Roche Applied Science, Indianapolis, IN). Adequate sonication (20 s, 3–4 cycles) with probe was performed until the solution became pellucid. Ultracentrifugation of samples was performed at 20,000 g for 30 min at 4°C. The supernatant was then transferred into 1.7-mL tubes on ice for measurement of protein yield via bicinchoninic acid (BCA) protein assay (Pierce Biotechnology, Inc., Rockford, IL) and subsequently stored at −80°C.

Protein lysates (5 mg) of each sample were reduced with 5 mM tris(2-carboxyethyl)phosphine (TCEP) and incubated at 37°C for 30 min in an Eppendorf Thermomixer (Eppendorf, Hauppauge, NY), followed by alkylation of cysteine residues with 14 mM iodoacetamide at 37°C for 30 min in the dark. Protein was subsequently extracted by methanol-chloroform precipitation and washed with ice-cold acetone. Air-dried pellets were then resuspended with 50 mM HEPES (pH 8.5). LysC (Wako, Japan; 1:100 enzyme-substrate ratio) was used for protein digestion. After incubation for 4 h, trypsin (Sigma; 1:20) was added, and samples were incubated overnight at 37°C. The digestion was terminated with 1% FA. For tandem mass tag (TMT)-labeling quantitative phosphoproteomics, acidified digested peptides were further subjected to C18 solid-phase extraction (SPE; Oasis PRiME HLB; Waters, Milford, MA).

Titanium dioxide bead-based phosphopeptide enrichment.

Because of the low stoichiometric nature of protein phosphorylation, phosphopeptides must be enriched before downstream mass spectrometry analysis. This was accomplished using methodology similar to our previously published protocol of TiO₂-based phosphopeptide enrichment (45) with minor modifications. Tryptic peptides (∼5 mg per channel) were resuspended in 500 µL (~1 mg/mL) 2 M lactic acid-1% trifluoroacetic acid (TFA)-50% acetonitrile (ACN; binding solution) and centrifuged at 15,000 g for 20 min. Supernatants were transferred to an Eppendorf tube containing 15 mg titanium dioxide beads (GL Sciences, Japan) and vortexed for 1 h at room temperature. Afterward, beads were washed with 1 mL 2 M lactic acid-50% ACN twice, 50% ACN-0.1% TFA twice, and 25% ACN-0.1% TFA twice. Phosphopeptides were eluted twice with 250 μL of 700 mM NH₃OH (pH 12), quenched with 5% formic acid, and desalted on Oasis HLB C18 columns.

Tandem mass tag labeling.

Isobaric labeling of the enriched phosphopeptides was performed using 10-plex TMT reagents (Thermo Fisher Scientific, Rockford, IL). TMT reagents were dissolved in 41 μL of anhydrous ACN, and 10 μL were used per 100 μg phosphopeptides in 100 μL HEPES (50 mM, pH 8.5). After incubation for 1 h at room temperature, the reaction was quenched by adding 8 μL of 5% hydroxylamine. Labeling efficiency was assessed and reached 97%.

High-pH reversed-phase peptide fractionation.

TMT-labeled phosphopeptides were dissolved with 300 µL 0.1% TFA and loaded on conditioned reversed-phase fractionation spin columns (Thermo Fisher Scientific, Rockford, IL). The column was washed with water followed by 5% ACN-0.1% triethylamine (TEA). Afterward, peptides were eluted sequentially by 10%, 12.5%, 15%, 17.5%, 20%, 25%, and 50% ACN with 0.1% TEA. The eluted fractions were dried with a SpeedVac concentrator (Thermo Fisher Scientific, Rockford, IL) and reconstituted in 50 µL 0.1% TFA for LC-MS analysis.

LC-MS/MS analysis for quantitative proteomics and phosphoproteomics.

LC-MS/MS analysis was performed on an Orbitrap Fusion Lumos Mass Spectrometer (Thermo Fisher Scientific, San Jose, CA) coupled to an ultrahigh-pressure Eksigent (Dublin, CA) Nano-2D Ultra capillary/nano-LC system. Mobile phase A and mobile phase B were 0.1% formic acid in 2% acetonitrile and 0.1% formic acid in 88% acetonitrile, respectively. Peptides were loaded to the trap (300-µm inner diameter × 1 cm) packed with Zorbax 5-µm C18 resins with 1% mobile phase B at 10 µL/min, and the trap was washed for 3 min. A gradient series (flow rate at 250 nL/min) was used to back-flush the trapped samples onto the long nano-LC column (75 µm × 100 cm) containing PepMap 3-μm C18 resins for further separation. The column was homogeneously heated to 52°C with heat-conductive silicone to improve the chromatographic resolution and reproducibility. The optimized gradients were as follows: 4% mobile phase B for 15 min, 13–28% mobile phase B for 110 min, 28–44% mobile phase B for 5 min, 44–60% mobile phase B for 5 min, 60–97% mobile phase B for 1 min, and 97% mobile phase B for 17 min.

Data-dependent mode was employed for MS analysis. For label-free proteomics, MS1 survey scan [mass-to-charge ratio (m/z) 400–1,500] worked at a resolution of 120,000 with an automated gain control (AGC) target of 500,000 and maximum injection time of 50 ms. Precursors were then fragmentized in higher-energy collisional dissociation (HCD) activation mode at normalized collision energy of 35%. MS2 spectra were collected at a resolution of 15,000 with an AGC target of 50,000 and maximum injection time of 50 ms. For TMT-labeling identification and quantitation, Fourier transform mass spectrometry (FTMS1) spectra were collected at a resolution of 120,000 with an AGC target of 200,000 and a maximum injection time of 100 ms; Ion trap mass spectrometry (ITMS2) spectra were collected at an AGC of 4,000, maximum injection time of 120 ms, and Collision-induced dissociation (CID) collision energy of 35%; and FTMS3 spectra enabling more accurate reporter ion quantitation utilized an AGC target of 50,000 and a maximum injection time of 250 ms at 60,000 resolution, and HCD collision energy was increased to 55% to ensure maximal TMT reporter ion yield. Synchronous precursor selection (SPS) was enabled to include eight MS2 fragment ions in the FTMS3 scan (13).

Generation of a novel swine heart protein database via RNA sequencing.

Total RNA was extracted from myocardial samples as well as from brain tissue samples obtained from normal swine using the RNAeasy Mini Kit (Qiagen, Germany) For each RNA sample, cDNA libraries were prepared using the TrueSeq RNA Sample Preparation Kit (Illumina, San Diego, CA). The samples were then sequenced (rapid 100-cycle paired end at ~50 million reads per sample on an Illumina HiSeq 2500). Raw sequencing reads from the two experiments were mapped to the Ensembl Sus scrofa 10.2 reference genome using TopHat (v2.0.7) with default parameters and Illumina’s iGenomes transcript annotation file “genes.gtf” (Ensembl Sus scrofa 10.2) available at https://support.illumina.com/sequencing/sequencing_software/igenome.html. Transcript abundance data were obtained using R package “customProDB” (63). The resulting BAM file from TopHat was used as input for customProDB (63) to generate the customized protein database. Only transcripts with >1 read per kilobase per million mapped reads (RPKM) were considered.

Labeled protein identification and quantification.

The raw files (.raw) generated by LC-MS were matched to the RNA-seq swine heart database with 13,946 entries using the SEQUEST-HT search engine embedded in Proteome Discoverer (v2.1; Thermo Scientific). The search parameters were set as follows: Precursor Ion Tolerance, 20 ppm; Fragment Ion Tolerance, 0.6 Da; Maximum Missed Cleavages, 2; Static Modifications, carbamidomethylation/+57.021 Da (C); Dynamic Modifications, oxidation/+15.995 Da (M) and acetylation/+42.011Da (NH₂ terminus); and Decoy Database Search, target false discovery rate (FDR) 0.01. For quantitative phosphoproteomics, addition of several specific parameters was necessary: Static Modifications, peptide NH₂ terminus TMT 6-plex/+229.163 Da; Dynamic Modifications, phosphorylation/+79.966 Da (S,T); Site Probability Threshold, 50; Apply Quan Value Corrections, true; Co-Isolation Interference, 60%; Average Reporter S/N Threshold, 5; Tag Mass Tolerance, 15 ppm; Normalization Mode, none; and Scaling Mode, none.

Bioinformatics and statistics.

Gene Ontology (GO) annotation was performed using DAVID Bioinformatics Resources v6.8. Proteomic and phosphoproteomic networks were analyzed with STRING v10.0. Heat maps and hierarchical clustering analysis were achieved by gplots and the reshape2 package, respectively. The upstream kinases modulating a given phosphorylation site were predicted with NetworKIN v3.0. A paired Student’s t test was used to determine the statistical differences of average protein/phosphorylation level between stunned (LAD) and control (remote) myocardium. The box-and-whisker plot (10th to 90th percentile) and volcano plots were generated with GraphPad Prism 7.0. Data are expressed as means ± standard error (SE) in bar charts (Fig. 4A) and Supplemental Table S1 (all Supplemental Material is available at https://doi.org/10.6084/m9.figshare.11114396).

Fig. 4.

Quantification of significantly altered proteins and phosphosites in stunned swine myocardium. A: data are expressed as means ± SE. Echocardiographic assessment of regional wall thickening demonstrated a marked reduction in contractile function in the left anterior descending coronary artery (LAD) region during ischemia that improved but did not normalize 1 h after reperfusion, consistent with myocardial stunning (panel at left). Postmortem histopathological analysis of apoptosis by TUNEL staining showed a significant, ~5-fold increase in apoptotic myocytes in the ischemic LAD region compared with remote, nonischemic myocardium (panels at right). B: the expression of significantly altered phosphosites leading to contractile dysfunction (which includes proteins related to contractile fibers, Ca²⁺ overload, and cell junctions) and apoptosis across all samples is shown in a heat map. The protein intensity in an individual sample is normalized to the sum of protein intensities across all samples. In the color key, green indicates a normalized protein intensity of 0–7.0, black indicates a normalized protein intensity of 7.0–10.5, and red indicates a normalized protein intensity of 10.5 or higher. C: the expression of significantly altered proteins leading to cytoskeleton breakdown and extracellular matrix (ECM) damage across all samples is shown in a heat map with the same criteria as in B.

RESULTS

Swine subjected to a 10-min LAD occlusion exhibited severe regional contractile dysfunction during ischemia that persisted throughout the first hour of reperfusion (Supplemental Table S1). Although euthanasia and heart tissue collection 1 h postreperfusion were necessary for proteomic/phosphoproteomic analysis of stunned myocardium in the present study, we have previously shown that contractile dysfunction elicited by a 10-min coronary occlusion is reversible and normalizes by 3 h after reperfusion, consistent with myocardial stunning (66).

Development of a swine heart protein database.

Despite the well-recognized utility of large animal models for translation of scientific discoveries from experimental rodent models to humans (11), accurate and in-depth investigation of the myocardial proteome has been impeded by the lack of a nonredundant annotated swine heart protein database. To address this issue, we generated a customized RNA-seq-derived protein database for the porcine heart using customProDB, which revealed 13,947 transcripts with >1 RPKM. To determine heart-specific expression, we further compared the transcripts obtained from the heart versus those from the brain. Among the 14,768 transcripts from the swine brain, 10,385 were also identified in the swine heart. The histogram of the log10 ratio (heart vs. brain, Supplemental Fig. S1A) showed that 214 transcripts were at least 10-fold more abundant in the heart than in the brain, whereas 308 transcripts were at least 10-fold more abundant in the brain. Among the 214 abundant transcripts in the heart, 157 have known orthologs in the human genome. The top 10 highly expressed transcripts in the heart versus the brain (i.e., heart-specific transcripts) are listed in Supplemental Fig. S1B and include several well-known cardiac-specific proteins such as actin-α cardiac muscle 1 (ACTC1), troponin T (cardiac muscle, TNNT2), creatine kinase S-type (mitochondrial, CKMT2), and troponin I (cardiac muscle, TNNI3). All annotation information was acquired by searching TREMBL, Reference Sequence (RefSeq), and HUGO Gene Nomenclature Committee (HGNC) databases. Table 1 shows the performance of the RNA-seq-derived database by comparing it with other predicted protein databases based on genome sequences (RefSeq and TREMBL) using a single LC-MS run of a swine myocardium sample. We found that our RNA-seq database identified a similar number of unique proteins compared with others despite a smaller overall size, indicating that the database developed here showed less redundancy and higher accuracy while maintaining a similar level of completeness compared with current databases. This was also reflected in the fact that the RNA-seq database achieved the lowest number of misassigned proteins (86 vs. 164 in TREMBL and 373 in RefSeq), defined as two or more proteins with shared peptides leading to ambiguous identification. Moreover, 1,597 proteins identified by our customized database (~87% of all identified proteins) were given with unambiguous annotations, whereas the number of unambiguously annotated proteins from other databases was 671 (TREMBL) and 500 (RefSeq). Therefore, our RNA-seq-derived protein database allows identification of proteins with lower redundancy and less ambiguous annotation compared with other commercially available protein databases. In addition, the considerable reduction in database size substantially decreases redundant results and thereby provides more accurate identification, while also enabling more focused database searching to facilitate widespread application.

Table 1.

Comparisons of identified proteins across multiple databases

1% Peptide FDR	Total Entries	Identified Proteins	Potential Misassigned Proteins	Proteins with Annotations
RefSeq	32,222	1,858	373	500
TREMBL	26,053	1,894	164	671
RNA-seq	13,947	1,838	86	1,597

Values are numbers of proteins. FDR, false discovery rate; RefSeq, Reference Sequence; RNA-seq, RNA sequencing.

Subcellular protein fractionation.

Because highly abundant myofilament proteins reduce the dynamic range of protein identification in cardiac samples, subcellular fractionation was used to deplete myofilament proteins based on pH-dependent variability in protein solubility. The myofilament proteins were soluble in TFA buffer with extremely low pH (pH = 2.2), whereas the cytosolic fraction was extracted in mild HEPES buffer (pH = 7.8) and the hydrophobic membrane fraction was extracted using strong detergent cocktail. This enabled efficient depletion of myofilaments in both fractions (Supplemental Fig. S2A) and identification of unique proteins in both the cytosolic and membrane fractions (Supplemental Fig. S2B). To maximize the depth of protein identification, both trypsin and GluC were used for protein digestion, while strong cation-exchange liquid chromatography was employed for fractionation, collectively leading to improved analytical sensitivity and increased identification depth. As a result, a total of 7,943 proteins (FDR < 1%) were achieved from a total of 84 HPLC fractions (Supplemental Fig. S2C), which not only validated ~60% of our customized RNA-seq-derived protein database and but also supported our ability to comprehensively analyze the cardiac proteome in swine myocardium.

Proteomic and phosphoproteomic analysis of stunned myocardium.

Despite the presence of several highly abundant myocardial proteins (e.g., myosin), we were able to quantify 2,099 proteins, where only 0.4% of these proteins have missing data in any of the 10 samples. This represents a substantial improvement of depth compared with previous quantitative studies of the myocardial proteome regardless of species (<1,000 proteins; 16, 48, 69). The median intragroup coefficients of variation (CVs) were ~10% and 8.7% (Fig. 2A) for samples from remote normal regions and stunned regions, respectively, indicating relatively low biological variation among tissue samples. Furthermore, intragroup CV for technical replicates was <10%, demonstrating excellent quantitative reproducibility by the IonStar pipeline. Meanwhile, we used the experimental null method (55) to reliably determine the optimal cutoffs for altered proteins and false-positive alterations, which were fold change >1.3 to either direction, P < 0.1 (55). Here, 56 proteins were found to be significantly altered, and only 3 false positives were identified in the null comparison set (i.e., ~5% false-positive biomarker discovery, Fig. 2B). GO analysis of cellular locations revealed that these altered proteins were predominantly distributed in the extracellular matrix (ECM; e.g., extracellular region part and fibrillar collagen) and contractile fibers (e.g., contractile fiber part and cytoskeleton), suggesting dysregulation of ECM and contractile function in stunned myocardium (Fig. 3).

Fig. 2.

Reproducibility and statistical analysis of quantified proteins and phosphoproteins in stunned swine myocardium. A: the box-and-whisker (10th to 90th percentile) plot shows that the median intragroup coefficients of variation (CVs) of quantified proteins using IonStar pipeline were ~10% and 8.7% for samples from the normal remote region and stunned left anterior descending coronary artery (LAD) region, respectively. The median CVs of quantified phosphosites using the tandem mass tag (TMT)-labeling strategy were 24.39% in normal remote tissue and 17.35% in stunned tissue, indicating relatively low biological variation and good reproducibility of quantification among tissue samples. B: in total, 2,099 proteins were quantified in 10 samples with the IonStar pipeline, and 56 of them were significantly altered with an optimized cutoff employed (fold change >1.3, P < 0.1). The blue dots in the volcano plot indicate 56 altered proteins with statistical significance (left). Similarly, 4,699 phosphosites in 1,847 phosphoproteins were totally quantified in 10 samples with the TMT-labeling strategy, and 113 phosphosites were significantly changed under the optimal cutoff of fold change >1.5, P < 0.1. These altered phosphosites are represented by red dots in the volcano plot (right).

Fig. 3.

Gene Ontology (GO) annotations of significantly altered proteins and phosphoproteins in stunned swine myocardium. A: the cell component distribution of significantly altered proteins is presented in a bar graph. The x-axis indicates the number of proteins in each component. Different bar colors indicate different groups of cellular components: red bars, extracellular matrix; blue bars, contractile fiber and cytoskeleton, gray bars, other components. B: the distribution of cell components enriched by significantly altered phosphoproteins is shown. The x-axis indicates the number of proteins in each component. Different colors of bars indicate different groups of cellular components: blue bars, contractile fiber and cytoskeleton; orange bars, cell junction; gray bars, other components.

Meanwhile, 4,699 phosphosites in 4,332 phosphopeptides and 1,847 phosphoproteins were quantified within the 10 samples. The median intragroup CVs were 24.39% in control (remote) samples and 17.35% in stunned (LAD) myocardium (Fig. 2B). Compared with previous studies in rodents (51, 52, 58), the number of quantifiable phosphosites increased by >2-fold using our optimized, in-depth phosphoproteomic analysis procedure. Moreover, 113 phosphosites showed a significant difference between stunned LAD regions and normal remote regions of the porcine heart with optimized cutoffs of fold of changes and statistical scores (Fig. 2B). Cellular location analysis indicated that >75% phosphoproteins were distributed in contractile fibers (e.g., myofibril, contractile fiber part, and sarcomere), the cytoskeleton (e.g., microtubule cytoskeleton, myosin complex, striated muscle thick filament, thin filament, and stress fiber), and cell junctions (e.g., adherens junction, cell-substrate adherens junction, cell-cell junction, and anchoring junction), suggesting that alterations in phosphorylation status of proteins in these areas likely contribute to contractile dysfunction and structural remodeling in stunned myocardium.

Observed differences in protein level and phosphorylation status could be generally clustered into three groups based on related physiological processes: 1) contractile function, 2) ECM remodeling, and 3) apoptosis. Alterations in these processes were consistent with the observation of reduced wall thickening and an increased number of apoptotic myocytes in the stunned LAD region of the left ventricle compared with remote, nonischemic myocardium (Fig. 4A). Among the observed changes in protein level and phosphorylation, 49 significant changes related to contractile dysfunction were quantified (Fig. 4, B and C, and Supplemental Tables S2 and S3). At the protein level, the content of cytoskeletal proteins including myosin regulatory light polypeptide 9 (MYL9), palladin (PALLD), actin-related protein 2/3 complex subunit 3 (ARPC3), α-actinin-1 (ACTN1), PDZ and LIM domain protein 3 (PDLIM3), cysteine and glycine-rich protein 1 (CSRP1), calponin-1 (CNN1), and transgelin (TAGLN) was reduced in stunned myocardium, consistent with dysregulation of actin filament organization and contraction (Fig. 4C and Supplemental Table S3). Phosphoproteomic analysis revealed significant alterations of 43 phosphosites in stunned myocardium (Fig. 4B and Supplemental Table S2). For example, we observed increased phosphorylation of troponin I type 3 (TNNI3) at S202 and tropomyosin 1 (TPM1) at S283, along with decreased phosphorylation of phospholamban (PLN) at S16 and T17, alterations that may be linked to contractile dysfunction during ischemic stunning. Moreover, we identified 40 novel phosphorylation changes to contractile proteins, such as upregulated phosphorylation of calpastatin (CAST) at S374/S202/S334, synemin (SYNM) at S378, tight junction protein ZO-1 at S446, and cardiomyopathy-associated protein 5 (CMYA5) at S2269/S118/S123 and downregulated phosphorylation of xin actin-binding repeat-containing protein 1 (XIRP1) at S1662/S1664/1735, tensin-1 (TNS1) at S496, myosin-XVIIIb (MYO18B) at T67, leiomodin-2 (LMOD2) at T59, nebulette (NEBL) at S954/T797, and catenin δ-1 (CTNND1) at S861/S864. Importantly, the level of many of the proteins exhibiting altered phosphorylation status was not significantly affected, indicating that phosphorylation of contractile proteins, rather than protein degradation, may play a critical role in stunning-related contractile dysfunction. Examples of alterations in the level of proteins linked with ECM remodeling include degradation of collagen α-1(XIV) chain (COL14A) and collagen α-2(I) chain (COL1A2), as well as a decreased level of small leucine-rich proteoglycans (SLRPs) including decorin (DCN), lumican (LUM), proline/arginine-rich end leucine-rich repeat protein (PRELP), and osteoglycin (OGN), among others (Fig. 4C and Supplemental Table S4). Building on our recent observation that ischemic stunning is associated with induction of myocyte apoptosis, we observed a decreased level of defender against apoptotic cell death 1 (DAD1; Fig. 4C and Supplemental Table S5), along with 13 phosphorylation changes of apoptotic proteins including upregulated phosphorylation of E3 ubiquitin-protein ligase HUWE1 at S65 and reticulon-4 (RTN4) at S457, along with downregulated phosphorylation of AKT1 substrate 1 (AKT1S1) at 183, Tousled-like kinase 2 (TLK2) at S73, and MYC target 1 (MYCT1) at S49 (Fig. 4B and Supplemental Table S5). Many of the observed changes in phosphorylation status of proteins associated with programmed cell death have not been reported previously and therefore may represent novel mechanisms by which transient ischemia elicits irreversible myocyte injury via initiation of apoptotic signaling. A summary of the pathways and cellular localization that we have identified as dysregulated following brief ischemia is given in Fig. 5.

Fig. 5.

Summary of the pathways and cellular distributions of quantified proteins and phosphosites modulating contractile dysfunction, extracellular matrix damage, and apoptosis in stunned swine myocardium. Upregulated molecules are colored in red, whereas downregulated molecules are colored in green. Well-confirmed changes such as downregulated ATP1A1-S650, TPM1-S283, and PLN-T17 and upregulated TNNI3-S202, as well as novel changes are included. FAK, focal adhesion kinase.

Upstream kinases mapped to observed phosphorylation sites.

After identifying ischemic stunning-induced alterations in phosphorylation status at specific phosphosites, we subsequently predicted the upstream kinases responsible for regulating these phosphorylation changes. Herein, two essential and complementary parameters were set: 1) NetworKIN score, which interprets kinase-substrate correlation based on motif analysis; and 2) STRING score, which shows the protein-protein interaction based on functional association analysis. Cutoffs of NetworKIN score >1.0 and STRING score >0.4 were used for this analysis. In total, 27 contractile proteins corresponding to 12 kinase groups were identified (Fig. 5), and biological interactions are shown in Supplemental Fig. S3A. Similarly, 26 apoptotic proteins corresponding to 13 kinase groups were identified (Fig. 5), and their biological interactions are shown in Supplemental Fig. S3B. The strongest correlations of 18 phosphosites of contractile proteins and 12 phosphosites of apoptotic proteins with predicted kinases are shown in Tables 2 and 3, respectively. Examples of specific kinases that are predicted to mediate observed changes in phosphorylation status during myocardial stunning include protein kinase C (PKC; increased phosphorylation of cTnI at S199), calmodulin-dependent kinase II (CaMKII; decreased phosphorylation of PLN at T17), and protein kinase A (PKA; decreased phosphorylation of PLN at S16). Furthermore, phosphorylation of CDK18, a critical factor regulating transcription and cell proliferation, was found to be downregulated at S12 via PKA in stunned myocardium. Other than these well-known kinase-substrate interactions, a series of novel interactions were also identified by our analysis. These include PAK1 or CaMKIIα-mediated downregulation of CTNND1 phosphorylation at S861, CK2 or GSK3β-mediated downregulation of CTNND1 phosphorylation at S864, PKC or CK1-mediated upregulation of ACTB phosphorylation at S62, PKC or GSK-mediated downregulation of AKT1S1 phosphorylation at S183, PAK-mediated upregulation of vesicle-associated membrane protein-associated protein B/C (VAPB) phosphorylation at S158, MAPK-mediated downregulation of KIF1C phosphorylation at S914, and MAPK or JNK-mediated downregulation of MAVS phosphorylation at S220 (Tables 2 and 3 and Fig. 6). Although the biological significance of these interactions will require validation in subsequent experiments, these results offer novel, hypothesis-generating insight that may facilitate efforts to further elucidate molecular mechanisms underlying altered intracellular signaling during ischemic myocardial stunning.

Table 2.

Interaction scores between predicted kinases and contractile dysfunction-related phosphosites in stunned swine myocardium

Gene Name	Protein Name	Phosphosites on Sus scrofa	Phosphosites on Homo sapiens	Phosphopeptides	NetworKIN Score	STRING Score	Kinase	Group	P Value	Ratio
ACTB	Actin, cytoplasmic 1	S62	S60	DSYVGDEAQsK	8.761	0.8	PKCθ	PKC	0.098	1.535*
TNNI3	Troponin I type 3 (cardiac)	S202	S199	NIDALsGMEGR	1.444	0.895	PKCα	PKC	0.039	1.729*
SORBS1	Sorbin and SH3 domain containing 1	S256	S465	YsFSEDTK	4.806	0.601	PAK2	PAK	0.001	1.592*
CTNND1	Catenin (cadherin-associated protein), delta 1	S861	S861	SQsSHsYDDSTLPLIDR	1.595	0.509	PAK1	PAK	0.047	0.655†
CTNND1	Catenin (cadherin-associated protein), delta 1	S864	S864	SQsSHsYDDSTLPLIDR	2.414	0.523	CK2α	CK2	0.047	0.655†
ACTB	Actin, cytoplasmic 1	S62	S60	DSYVGDEAQsK	9.531	0.577	CK1α	CK1	0.098	1.535*
CTNND1	Catenin (cadherin-associated protein), delta 1	S864	S864	SQsSHsYDDSTLPLIDR	5.883	0.65	GSK3β	GSK3	0.047	0.655†
CTNND1	Catenin (cadherin-associated protein), delta 1	S861	S861	SQsSHsYDDSTLPLIDR	1.962	0.558	CaMKIIα	CaMKII	0.047	0.655†
PLN	Sus scrofa phospholamban	T17	T17	AStIEMPQQAR	11.356	0.963	CaMKIIβ	CaMKII	0.017	0.425†
PLN	Sus scrofa phospholamban	S16	S16	RAsTIEMPQQAR	8.740	0.855	PKAα	PKA	0.038	0.644†
KIF1C	Kinesin family member 1C	S914	S915	APPARPSsPPLSSWER	1.678	0.293	MAPK3	MAPK	0.073	0.635†
VAPB	Sus scrofa VAMP (vesicle-associated membrane protein)-associated protein B and C	S158	S156	ALsSALDDTEVKK	1.294	0.197	PAK2	PAK	0.003	1.782*
NEBL	Nebulette	S387	S388	GRSsLDLDK	1.362	0.177	PAK1	PAK	0.007	1.718*
MTUS2	Microtubule-associated tumor suppressor candidate 2	T284	T1345	LQtGDPTSPVKLSPTSPIYR	1	0.129	DAPK3	DAPK	0.096	0.655†
LMOD2	Leiomodin 2 (cardiac)	T59	T59	tPTGTFSR	1.356	0.11	CK1α	CK1	0.097	0.559†
DYNC1LI1	Dynein, cytoplasmic 1, light intermediate chain 1	S162	S207	DFQEYVEPGEDFPAsPQRR	2.761	0.232	CDK1	CDK	0.013	0.64†
NEBL	Nebulette	S387	S388	GRSsLDLDK	1.319	0.225	PKAα	PKA	0.007	1.718*
MTUS2	Microtubule-associated tumor suppressor candidate 2	T284	T1345	LQtGDPTSPVKLSPTSPIYR	1.412	0.266	TTK	TTK	0.096	0.655†

^*Upregulated ratios (>1.5);

^†downregulated ratios (<0.667).

Table 3.

Interaction scores between predicted kinases and apoptosis-related phosphosites in stunned swine myocardium

Gene Name	Protein Name	Phosphosites on Sus scrofa	Phosphosites on Homo sapiens	Phosphopeptides	NetworKIN Score	String Score	Kinase	Group	P Value	Ratio
CDK18	Cyclin-dependent kinase 18	S12	S14	RFsLSVPR	1.319	0.545	PKAα	PKA	0.037	0.612†
AKT1S1	AKT1 substrate 1 (proline-rich)	S183	S183	sLPVSVPVWAFK	1.555	0.538	GSK3β	GSK3	0.069	0.664†
AKT1S1	AKT1 substrate 1 (proline-rich)	S183	S183	sLPVSVPVWAFK	1.693	0.557	PKCα	PKC	0.069	0.664†
SLK	STE20-like kinase	S340	S336	sKRASSDLSIASSEEDK	2.553	0.538	PKCθ	PKC	0.012	1.524*
MAVS	Mitochondrial antiviral signaling protein	S220	S222	QDTELGSAHTAGTVSSPTsPR	2.022	0.589	MAPK3	MAPK	0.006	0.558†
MAVS	Mitochondrial antiviral signaling protein	S220	S222	QDTELGSAHTAGTVSSPTsPR	1.159	0.548	MAPK8	JNK	0.006	0.558†
AK1	Adenylate kinase isoenzyme 1	S38	S38	YGYTHLsTGDLLR	1.007	0.230	PAK1	PAK	0.074	0.642†
CDC42EP1	CDC42 effector protein (Rho GTPase binding) 1	S192	S192	RSDsLLSFR	3.958	0.113	PAK1	PAK	0.072	1.575*
MYCT1	Myc target 1	S49	S115	SsYSHGLNR	1.362	0.15	PAK1	PAK	0.050	0.629†
MYCT1	Myc target 1	S49	S115	SsYSHGLNR	1.434	0.193	PKCβ	PKC	0.050	0.629†
OPTN	Sus scrofa optineurin	S336	S346	KNSATPsELNEK	1.298	0.189	PKCθ	PKC	0.002	1.563*
TLK2	Tousled-like kinase 2	S73	S99	ISDYFEFAGGSGPGTsPGR	1.006	0.227	MAPK3	MAPK	0.050	0.633†

^*Upregulated ratios (>1.5);

^†downregulated ratios (<0.667).

Fig. 6.

Summary of proposed molecular mechanisms involved in myocardial stunning of swine. The molecular mechanisms of myocardial stunning in swine are proposed at the level of protein expression and protein phosphorylation. On one hand, observed changes in protein expression demonstrate 1) contractile dysfunction, which is caused by degradation of cytoskeletal molecules via Ca²⁺-dependent calpain activation; 2) extracellular matrix (ECM) damage reflected by the degradation of collagen, another mechanism of calpain activation; and 3) apoptosis, which is represented by repressed DAD1 and confirmed for the first time at molecular level in the present analysis of stunned swine myocardium. On the other hand, phosphorylation changes support contractile dysfunction through PKC-mediated TNNI-S202 phosphorylation in a Ca²⁺/calpain-dependent manner and repressed phosphorylation of PLN-T17 mediated by CaMKII, whereas myocyte apoptosis during ischemic myocardial stunning is further supported by inhibition of CDK18-S12 phosphorylation dependent on PKA. Besides well-confirmed interactions represented by solid arrows, multiple novel interactions are also shown using dashed arrows. Red squares represent upregulated proteins/phosphosites, green squares represent downregulated proteins/phosphosites, and blue ellipses represent predicted kinases.

DISCUSSION

The present study provides novel insight regarding the molecular mechanisms involved in postischemic myocardial stunning using two complementary analytic strategies: global, ion current-based quantitative proteomics (IonStar) and accurate, reproducible quantitative phosphoproteomics. The results represent the most in-depth investigation of the swine cardiac proteome and phosphoproteome published to date, in large part through the use of a novel, less redundant RNA-seq-derived protein database with complete annotation that should offer a strong platform for biological researchers working on similar studies in the future. Correspondingly, the proposed molecular mechanisms underlying myocardial changes during ischemic stunning were elucidated in terms of protein content and phosphorylation status (Fig. 6). Observed changes in protein levels reveal 1) contractile dysfunction caused by degradation of cytoskeletal molecules via Ca²⁺-dependent protease (e.g., calpain) activation, 2) ECM damage supported by collagen degradation, and 3) apoptotic cell death, represented by repressed DAD1, all of which could contribute to the pathogenesis of myocardial stunning. Phosphorylation changes linked to contractile dysfunction included PKC-mediated TNNI-S202 phosphorylation via Ca²⁺-calpain activation, repressed CaMKII-mediated PLN-T17 phosphorylation, as well as PKA-induced PLN-S16 phosphorylation (Fig. 6). Although a large number of the observed changes in phosphorylation status occurred at conserved phosphosites, we discovered many novel alterations with as-yet-unknown biological relevance. Collectively, this work offers new insight into the molecular mechanisms underlying several pathophysiological features of ischemic myocardial stunning, while also providing new targets for future studies aimed at advancing the understanding of how the heart responds to brief ischemia and the development of therapeutic strategies to ameliorate postischemic stunning and chronic adverse left ventricular (LV) remodeling associated with repetitive exposure to this phenomenon.

Contractile dysfunction and ECM remodeling in stunned myocardium.

Contractile dysfunction is the principal functional abnormality associated with myocardial stunning (6). We quantified a total of 51 significantly altered molecules conferring this dysfunction, either via alterations in protein abundance (Fig. 4C and Supplemental Table S3) or via changes in phosphorylation status (Fig. 4B and Supplemental Table S2). Interestingly, though a number of phosphosites changed significantly in stunned myocardium, the abundance of proteins on which these sites were found remained constant (Fig. 4B), indicating that global proteomic and phosphoproteomic analyses provided complementary insights. Because prior studies of phosphosite-specific function are limited, we matched the phosphosites discovered in swine with conserved phosphosites in humans via peptide alignment. As a result, we were able to investigate how observed alterations in phosphorylation status may be linked with potential molecular mechanisms of myocardial stunning.

Currently, Ca²⁺ overload is one plausible mechanism for contractile dysfunction in stunned myocardium (23). It has been postulated that inactivation of the Na⁺-K⁺ ATPase on the cell membrane is an important component of this process since it leads to the loss of the electrochemical gradient that facilitates active Ca²⁺ efflux. As a result, intracellular Ca²⁺ accumulates. We found that phosphorylation of Na⁺-K⁺ ATPase ATP1A1 was depressed at phosphosite S650, without any changes in total protein abundance. The conserved human phosphosite, S653, is known to be involved in ATP hydrolysis, suggesting that this site-specific alteration in phosphorylation and the associated loss of ATP1A1 activity may be an important mechanism by which Na⁺-K⁺ ATPase dysfunction leads to Ca²⁺ overload and contractile dysfunction in stunned myocardium (28, 46).

Intracellular Ca²⁺ accumulation also leads to the activation of Ca²⁺-dependent proteases (e.g., calpain I), which can promote myofibrillar degradation and impair contractile function (23). We found several cytoskeletal proteins including MYL9, PALLD, ARPC3, ACTN1, PDLIM3, CSRP1, CNN1, and TAGLN to be downregulated in stunned myocardium (Fig. 4C), which may contribute to abnormal actin filament organization and contractile dysfunction. We also observed increased phosphorylation of troponin I type 3 (TNNI3) at S202 in stunned myocardium, a conserved site at S199 in humans (70, 74), without a significant change in protein abundance. Prior studies in human cardiomyocytes have revealed that phosphorylation status at S199 regulates calcium sensitivity, cTnI’s binding affinity to actin-tropomyosin, and calpain-induced proteolysis (70), supporting the notion that this site-specific alteration in phosphorylation status may be involved in several pathophysiological processes associated with ischemic myocardial stunning. We also observed decreased phosphorylation of tropomyosin 1 (TPM1) at S283 in stunned myocardium. This may underlie the observation that cooperative binding of myosin is impaired at this phosphosite in stunned human myocytes (50). Finally, we observed decreased phosphorylation of phospholamban at both T17 and S16, without a change in phospholamban abundance. These changes are consistent with previous reports in dogs and immature swine and can contribute to contractile dysfunction by limiting sarcoplasmic reticulum (SR) Ca²⁺ reuptake (9, 26, 35).

In addition to confirming changes in phosphorylation that have been observed in previous reports, we identified several novel site-specific changes in phosphorylation status in stunned myocardium (e.g., CAST-S374/S202/S334, SYNM-S378, ZO1-S446, XIRP1-S1662/S1664/1735, TNS1-S496, CMYA5-S2269/S118/S123, and MYO18B-T67). The biological significance of these changes is difficult to predict because of the lack of conserved phosphosites in humans. For example, the phosphorylation of calpastatin (CAST) was upregulated in stunned swine myocardium at sites S374, S202, and S334, which are located at the four repetitive inhibitory units of calpastatin (40). We speculate that this upregulation may reduce calpastatin’s inhibitory efficiency against calpain based on known (but not conserved) phosphosites in human myocardium, resulting in Ca²⁺-dependent calpain activation and the proteolysis of cytoskeletal proteins (3). Meanwhile, several phosphosites with conserved sites in humans were altered in stunned myocardium (e.g., LMOD2-T59, NEBL-S954/T797, and CTNND1-S861/S864), but there remains a paucity of knowledge about their biological significance (Supplemental Table S2). Thus, future studies investigating the role of these alterations in phosphorylation status in the pathophysiology of myocardial stunning are necessary.

Another molecular mechanism underlying the functional abnormalities that characterize myocardial stunning is ECM damage (75). For example, activation of endogenous procollagenase has been observed in a canine model of stunned myocardium, resulting in a significant loss of collagen after 90 min of reperfusion (10). Our proteomic data support collagen degradation as a key characteristic of ischemic myocardial stunning. We observed decreased abundance of type XIV, α₁-collagen (COL14A) and type I, α₂-collagen (COL1A2) 1 h after reperfusion (Fig. 4C and Supplemental Table S4). Moreover, we are the first to demonstrate decreased abundance of small leucine-rich proteoglycans (SLRPs) such as decorin (DCN), lumican (LUM), proline/arginine-rich end leucine-rich repeat protein (PRELP), and osteoglycin (OGN) in stunned myocardium (Fig. 4C and Supplemental Table S4). DCN can stimulate collagen fibril assembly in myocardial infarction (MI) to alleviate infarct dilatation, an important process of healing (67). Similarly, LUM can modulate matrix arrangement and collagen assembly by interacting with collagen, aggrecan, and integrins (34). Therefore, the downregulation of SLRPs observed with proteomic profiling suggests that interruption of the collagen fibril assembly is an important pathophysiological molecular mechanism involved in ischemia-induced myocardial stunning.

Apoptotic cell death in stunned myocardium.

Our recent study demonstrated a transient increase in myocyte apoptosis in stunned myocardium following brief ischemia (66), but the detailed mechanisms underlying this phenomenon remain unknown. Based on GO and a search of the existing literature, the changes in abundance and phosphorylation status of 14 apoptosis-related proteins are listed in Supplemental Table S5. Among them, we identified significant site-specific alterations in phosphorylation status of 13 proteins, with only 1 protein exhibiting a difference in abundance (Fig. 4, B and C).

The defender against apoptotic cell death 1 (DAD1) is a mammalian cell death suppressor that acts via interactions with the downstream bcl-2 protein (25). We observed a decreased abundance of DAD1 in stunned myocardium, which may be involved in promoting myocyte apoptosis. The biological relevance of observed changes in phosphorylation status was assessed by evaluating conserved phosphosites in humans. For example, AKT1 substrate 1 (AKT1S1) can dissociate from Raptor, allowing the rapamycin complex 1 (mTORC1) access to downstream effectors and activation of mTORC1-mediated cell survival through phosphorylation (59). Indeed, the loss of phosphorylation at S183 was previously reported to suppress mTORC1 activity in HeLa cells (21). Here, we found decreased phosphorylation of AKT1S1 at the same phosphosite (S183) in swine stunned myocardium, suggesting that suppression of mTORC1-mediated cell survival may be involved in myocyte apoptosis following brief ischemia (Supplemental Table S5). Similar to contractile dysfunction-related proteins, we observed several novel phosphorylation changes in apoptosis-related proteins (e.g., CCAR2-S124, HUWE1-S65, TLK2-S73, MYCT1-S49, OPTN-S336, and SLK-S340). For example, HUWE1, an E3 ubiquitin ligase activated by phosphorylation, mediates ubiquitination and proteasomal degradation of targets such as MCL-1, Mfn2, RASSF1C, and p53, which is required for DNA damage-induced apoptosis (18, 73). Here, we observed increased phosphorylation of HUWE1 at S65 (conserved as site S3560 in humans) and speculate that it may contribute to the activation of HUWE1-mediated apoptosis of stunned myocardium, even in the absence of alterations in HUWE1 protein abundance (Supplemental Table S5). Similarly, Tousled-like kinase 2 (TLK2) initiates DNA replication during the S phase (33), but reports on the effects of TLK2 phosphorylation are lacking. In the present study, we found decreased phosphorylation of TLK2 at a novel phosphosite S73 (conserved phosphosite S99 in humans) in stunned myocardium. Such a change could suppress the activity of this kinase, potentially interrupting cell cycle progression during postischemic stress. Furthermore, MYC target 1 (MYCT1) is the direct target gene of c-MYC, and its overexpression promotes apoptosis (71). Here, decreased phosphorylation of MYCT1 at S49 (conserved phosphosite S115 in humans) was identified in stunned swine myocardium, which may be a mechanism involved in stunning-related myocyte apoptosis. Finally, several other novel phosphorylation changes (e.g., RTN4-S457 and MAVS-T208) were uniquely identified in the present study, although determination of their biological significance will require further investigation.

Investigation of predicted regulatory kinases and networks.

Our observation that brief ischemia elicits a greater number of changes in phosphorylation status than changes in protein abundance highlights the role of PTMs in myocardial stunning. Because our in-depth phosphoproteomic analysis allows identification of exact phosphopeptide sequences and their corresponding conserved human ortholog, we were able to investigate potential correlations between regulatory kinases and substrate networks. To this end, NetworKIN, a package combining linear motif analysis with network information on cellular context to accurately and sensitively link kinases to their corresponding modification sites (5, 22), was used to predict the potential kinases responsible for observed changes in phosphorylation status. Because swine are not included in the current NetworKIN database (humans and yeast only), conserved phosphosites aligned with humans are prioritized in this analysis. Briefly, with the criterion of NetworKIN score >1, 27 substrates corresponding to 12 kinase groups were identified in contractile dysfunction-related molecules and 26 substrates corresponding to 13 kinase groups were identified in apoptotic cell death-related molecules. The distribution of kinase groups and substrate numbers are shown in Supplemental Fig. S4. As the prediction of kinase groups based solely on theoretically linear motif analysis with the NetworKIN score is not sufficient to interpret biologically functional correlations between kinase groups and substrates, we further utilized STRING analysis, which provides specific and meaningful protein associations with support from biological data. With the additional stringent criterion of STRING score >0.4, the correlation confidence of filtered substrates and corresponding kinases conferring contractile dysfunction and apoptosis is shown with STRING (Supplemental Fig. S3), and the quantitative ratio of each substrate is shown in Tables 2 and 3.

On the basis of this analysis, we were able to identify several substrates and kinases with correlations supported by prior experimental data. For example, we observed increased phosphorylation of cTnI at S199 in stunned myocardium, which is defined as a target of PKC from both reported data in a conserved human phosphosite (70) and our own analysis (Supplemental Fig. S3). The increased PKC-mediated phosphorylation of cTnI at S199 will make this protein more susceptible to proteolysis (70), which may contribute to contractile dysfunction during myocardial stunning. Furthermore, Ca²⁺ overload-induced calpain activation can induce irreversible activation of PKC by cleaving off its NH₂ terminus regulatory subunit (27), which may result in PKC-mediated phosphorylation of cTnI at S199 during myocardial stunning.

Our analysis also predicts that the observed reduction in phosphorylation of PLN at T17 is mediated by calmodulin-dependent kinase II (CaMKII), whereas the decrease in phosphorylation of PLN at S16 is likely mediated by PKA. Both predictions are supported by reported data (64) and may offer insight into the mechanisms underlying alterations in PLN phosphorylation that influence Ca²⁺ handling and regulation of contractility in myocardial stunning. Decreased PKA activity is also predicted to mediate the reduction in phosphorylation of CDK18 at S12 (conserved phosphosite S14 in humans), which may have important biological effects based on CDK18’s role in the regulation of cell proliferation and gene transcription (39). Additionally, we found several novel phosphosites and kinases that have not been previously investigated in the context of myocardial ischemic injury. For example, the phosphorylation of p120 catenin (CTNND1) plays a role in regulating cadherin stability, adhesion-induced signaling, and cancer progression and can be phosphorylated by Src and EGFR at tyrosines (e.g., Y96, Y112, Y228, Y291, and Y302) or PKC at serine sites (e.g., S268, S269, and S288; 32). Increased phosphorylation at S268 causes unstable cadherin and promotes metastasis of cancer cells (72). Here, we observed downregulation of CTNND1 phosphorylation in stunned swine myocardium at two phosphosites: S861, which is mediated by PAK1 or CaMKIIα, and S864, which is mediated by CK2α or GSK3β, based on NetworKIN and STRING prediction. Similar to PKC, GSK3β could also be activated by calpain through cleavage of its regulatory subunit, which results from Ca²⁺ overloading in stunned myocardium (47). β-Actin (ACTB) is a major component of the contractile and cytoskeletal apparatus, and its phosphorylation by PKC has been shown to be involved in leukemic cell apoptosis (62). Other proteomic analyses have revealed that multiple kinases (e.g., PAK1, Src, and PKA) phosphorylate actin without identifying direct functional consequences or specific phosphosites (17). In the present study, we identified increased phosphorylation of ACTB at S62 in stunned myocardium, which is likely mediated by PKC or CK1. Finally, alterations in phosphorylation status of some protein phosphosites and activity of associated kinases that have not been previously identified were revealed in the present study, including PAK-mediated VAPB phosphorylation, MAPK-mediated KIF1C phosphorylation, and MAVS phosphorylation. However, further studies are necessary to determine the biological significance of the observed alterations of these signaling pathways in myocardial stunning.

Relationship to previous studies.

Although proteomic and phosphoproteomic analysis approaches have previously been utilized in swine to characterize changes in protein content and/or phosphorylation status after prolonged myocardial ischemia leading to myocardial infarction (4, 20) and chronic hibernating myocardium (42–44), evaluation of brief ischemia-induced myocardial stunning has been limited. However, Ledee et al. (35) recently performed quantitative cardiac phosphoproteomic profiling of myocardium collected 2 h after brief (10 min) ischemia in open-chest immature swine. Myocardial function was not assessed. Using Isobaric Tags for Relative and Absolute Quantification (iTRAQ), the authors identified 1,896 phosphopeptides in LV tissue samples collected from control animals (n = 6) and animals subjected to brief ischemia (n = 6). Of the identified peptide phosphorylation sites, 111 sites on 86 proteins met statistical requirements for differential phosphorylation, many of which were related to calcium handling and contractile function, similar to the present study. Furthermore, Ledee et al. observed decreased phosphorylation of pyruvate dehydrogenase (PDH; at S293) and phospholamban (PLN; at S16) following brief ischemia and validated these findings with semiquantitative immunoblots. Reduced phosphorylation of PLN and PDH was also observed in our analysis and has been previously reported in mice and humans (1). Thus, alterations in phosphorylation status of these proteins may therefore be regarded as a quality control marker in future experiments employing ischemia-reperfusion protocols. Importantly, the present study offers several advantages that allow more extensive characterization of the myocardial proteome and phosphoproteome in ischemic myocardial stunning. For example, our study was performed at an earlier time point in closed-chest mature swine. Our detailed assessment of serial hemodynamics and echocardiographic indexes of global and regional LV function documented regional contractile dysfunction. Furthermore, our approach involved parallel as well as complementary investigation at both the proteomic and phosphoproteomic level to provide a superior depth of profiling during brief ischemia-induced stunning. Combined with the use of our novel RNA-seq-derived swine heart protein database (as opposed to more traditional databases such as UniProt or TREMBL) with more complete annotations and lower redundancy, we achieved a more comprehensive cardiac phosphoproteomic data set (>4,500 phosphosites) with a very low amount of missing data (0.4%). To our knowledge, this is the most comprehensive data set published to date and undoubtedly improves our ability to more thoroughly analyze molecular interactions that occur during ischemic myocardial stunning in a clinically relevant large animal model.

Methodological considerations.

There are several methodological considerations that merit discussion. First, we utilized healthy swine without preexisting coronary artery disease or other cardiovascular comorbidities. Although this was critical in enabling evaluation of the effects of brief ischemia in the absence of chronic disease, this may influence extrapolation of our results to humans, many of whom have chronic ischemic heart disease and/or comorbidities that may affect the response to brief ischemia. In addition, although regional ischemia following a brief coronary occlusion is the most common animal model that has been employed to study myocardial stunning, reversible dyssynergy may also arise following prolonged myocardial ischemia or “short-term hibernation,” reperfused myocardial infarction, chronic hibernating myocardium, and global cardiac ischemia associated with cardioplegia and cardiac arrest (29). The proteomic and phosphorylation changes we report here may not be responsible for delayed recovery of function after perfusion is restored in all of these pathophysiological states. Second, our study used whole myocardial tissue to assess proteomic and phosphorylation changes after reperfusion. As a result, protein changes will reflect alterations in myocytes as well as endothelial cells and the extracellular matrix. Although myocytes contribute less than half of the cells in the heart, they contribute to the majority of LV mass. Thus, whole myocardial tissue proteomics will be biased toward myocyte proteins rather than endothelial cells and fibroblasts. Since we are comparing relative protein changes in stunned myocardium with normally functioning tissue and do not have confounding changes in myocyte number or mass as there is no irreversible injury, relative changes in myocardial proteins and phosphoproteins should be unaffected by the “dilution” of nonmyocyte components provided that the protein of interest is myocyte specific. For proteins present in all cell types, the relative changes in myocytes could be attenuated. Third, temporal variations in phosphorylation in myocardial tissue are likely to be complex during recovery from brief ischemia. We chose to study myocardial tissue 1 h after reperfusion when contractile dysfunction was changing slowly and in a “quasistatic state.” This avoids early transients that would be expected to resolve quickly after reperfusion and provide a more restricted candidate list that would more likely be related to the mechanism of persistent contractile dysfunction. How these individual candidate mechanisms impact the recovery of contractile function in stunned myocardium will require further study. Fourth, the experimental design is influenced by the TMT-labeling quantification strategy, which is feasible on ≤10 samples and therefore limits the sample size of the study. Future experimentation with IonStar-based, label-free quantification strategies may enable proteomic/phosphoproteomic studies on larger cohorts. Finally, we chose the remote zone as the normal comparator to provide an internal control for multiple variables that could affect global protein phosphorylation (e.g., neurohormonal background, anesthesia, etc.). This internal reference also permitted a more robust paired statistical analysis that reduced the sample size required. We have previously demonstrated that comparison of ischemic myocardium with remote myocardium can minimize some protein changes when structural remodeling is present as in chronic hibernating myocardium (41). Although we do not believe that this type of remote zone remodeling would be present 60 min after a brief coronary occlusion, the use of remote myocardium could have resulted in underestimating some protein changes if normal sham animals were used as a control.

Conclusions and perspectives.

In summary, the results of the present study provide a comprehensive profile of the multiple alterations in myocardial protein levels and phosphorylation status that develop in response to brief ischemia-induced myocardial stunning. Rather than a single molecular target, our results support the notion that reversible dyssynergy arises from multiple protein alterations including contractile proteins as well as the extracellular matrix. They further support the notion that even brief ischemia stimulates myocyte injury from apoptosis. Our innovative analytical approach to identify protein alterations, changes in phosphorylation status, and phosphosite-kinase interactions provides a variety of new molecular targets for therapeutic strategies aiming to ameliorate acute and chronic ischemia-induced left ventricular dysfunction.

On the one hand, the present study provides a comprehensive assessment of alterations in protein content and phosphorylation status in a clinically relevant porcine model of brief ischemia-induced myocardial stunning. Through the use of parallel quantitative proteomic and phosphoproteomic analyses, a variety of novel molecular mechanisms that may contribute to contractile dysfunction, extracellular matrix remodeling, and myocyte apoptosis were identified. Investigation of possible phosphosite-kinase interactions revealed several regulatory kinases that may be responsible for a large number of the observed changes in protein phosphorylation, including protein kinase C, calmodulin-dependent kinase II, and protein kinase A.

On the other hand, by significantly extending our current understanding of the molecular signature associated with ischemic myocardial stunning, the present study identifies a variety of new mechanisms that could be targeted therapeutically to ameliorate the adverse consequences of repetitive stunning in patients with ischemic heart disease. In addition to enhancing our mechanistic understanding of brief ischemia-induced myocardial stunning in a clinically relevant large animal model, the development of a nonredundant and well-annotated database of the swine heart proteome provides a valuable resource for future translational studies aiming to profile the cardiac proteome and phosphoproteome in porcine models of cardiovascular disease.

GRANTS

This work was supported by National Heart Lung and Blood Institute Grants HL-055324, HL-061610, and F32HL-114335, American Heart Association Grant 17SDG33660200, National Center for Advancing Translational Sciences Grant UL1-TR-001412, Department of Veterans Affairs Grant 1IO1BX002659, and the Albert and Elizabeth Rekate Fund in Cardiovascular Medicine.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

X.W. and X.S. conceived and designed research; X.W. and X.S. performed experiments; X.W. and X.S. analyzed data; X.W. and X.S. interpreted results of experiments; X.W. and B.R.W. prepared figures; X.W. drafted manuscript; X.W., B.R.W., R.F.Y., J.M.C., and J.Q. edited and revised manuscript; J.M.C. and J.Q. approved final version of manuscript.