Proteomic analysis of tissue-derived extracellular vesicles shows region-specific molecular changes in a rat model of takotsubo syndrome

Abstract

Takotsubo syndrome (TS) is characterized by transient regional wall motion abnormalities (RWMA) of the heart following stress. Extracellular vesicles (EVs) play a significant role in cellular communication and disease pathophysiology, but remain unexplored in TS. Using a high-fidelity rat model of TS induced by isoprenaline infusion (n = 16), we isolated EVs from the tissue of affected apical and unaffected basal segments of the left ventricle at 24 h post-induction. The TS phenotype and cardiac function were assessed using high-resolution echocardiography. EVs were characterized by electron microscopy, western blot, and nanoparticle tracking analysis (NTA). Moreover, EV protein analysis was performed using tandem mass tag (TMT) proteomics. Pure, cup-shaped vesicles ranging from 50 to 500 nm were successfully isolated. NTA revealed lower particle concentrations in EVs isolated from the apex of TS24h hearts compared to their corresponding basal segments. Western blot experiments confirmed the presence of typical EV markers, including Flotillin 1, TSG101, and CD63. We identified 2093 proteins, with 238 differentially expressed (|FC| > 0.58, adj.P < 0.05) proteins between TS-apex and control-apex, and 562 between TS-apex and TS-base, indicating a unique molecular adaptation in the affected apex. Functional enrichment analysis showed increased abundance of proteins associated with immune response, tissue repair, and survival signalling pathways. Proteins related to mitochondrial function showed decreased abundance. Network analysis revealed an association between proteins involved in lipid processes and inflammation. Overall, this study presents the first proteomic characterization of EVs in TS hearts. Our results demonstrate a distinct EV protein abundance profile in the affected apical segments of TS hearts, with marked changes in proteins related to inflammatory responses, tissue repair mechanisms, energy metabolism, and cell survival pathways. This comprehensive proteomic profile of EVs in TS hearts provides potential candidates for therapeutic targets and diagnostic biomarkers, warranting further mechanistic and clinical validation studies.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-42812-5.

Introduction

Takotsubo syndrome (TS) resembles acute myocardial infarction in terms of mortality and clinical symptoms¹. Clinicians recognize the syndrome by its reversible regional wall motion abnormalities (RWMA), often showing “apical ballooning” on imaging. While the exact mechanisms of TS remain unclear, evidence supports a role for catecholamines². Human disease models are crucial for elucidating underlying mechanisms³. Our recently developed refined TS model replicates several key patient findings in both sexes, including apical ballooning, acute onset and recovery of RWMA within days to weeks, ECG changes, complication profile, recovery pattern, predictive factors, and clinically relevant preceding triggers^4,5.

Extracellular vesicles (EVs) are membrane-enclosed particles (30–1000 nm) released by cells into the extracellular space. These vesicles contain a complex cargo rich in proteins, lipids, and nucleic acids that reflects their cellular origin and have been shown to influence functions of recipient cells. This communication system has been shown to play an important role in various cardiovascular diseases^6,7. Increased shedding of EVs is observed in several cardiovascular conditions—including aortic stenosis⁸, cardiac hypertrophy⁹, and myocardial infarction¹⁰—and is believed to occur in response to cellular stress, notably hypoxia. Functionally, some examples include EVs released after myocardial infarction, which can stimulate monocytes and amplify acute inflammation in the heart¹⁰. In contrast, cardiomyocyte-derived EVs upon hypoxic stress have been shown to influence contractility and correlate with prognosis in aortic stenosis⁸. EVs have further been implicated in both disease progression⁶ and cardiac repair¹¹. Advances in proteomics have enabled detailed characterization of the diverse protein cargo within cardiac EVs¹², revealing how these molecular signatures can reflect underlying cellular states and disease processes, thereby allowing discovery of novel biomarkers and mechanistic pathways. For example, proteomic profiling has shown that loss of estrogen receptor alpha in mouse cardiomyocytes alters the protein cargo of cardiac EVs, mediating metabolic dysfunction beyond the heart¹³, while its analysis of regenerative neonatal cardiac EVs identified enrichment of Wdr75—a protein shown to promote cell proliferation, angiogenesis, and cardiac repair in recipient cells¹⁴. However, most prior work has focused on biofluids or isolated cells from whole organs, and region-specific EV dynamics remain largely unexplored in situ. To date, it remains unknown whether regions affected by distinctive conditions like TS release EVs with unique functional and molecular profiles, and whether different regions of the myocardium (e.g., transiently akinetic versus contracting zones) shed EVs with quantitatively and qualitatively distinct cargo. To our knowledge, no prior study has examined cardiac EVs in TS—whether in human or animal models, and regardless of material (biofluid or tissue). By characterizing region-specific EV profiles in TS, our study advances understanding of its pathophysiology and the diagnostic or therapeutic utility of cardiac EVs.

Given the central role of EVs in cell-to-cell communication, their capacity to reflect cellular states, and the region-specific nature of TS, we hypothesized that EVs from TS hearts may show region-specific differences in protein composition. We tested this hypothesis in our rat model of TS by analysing the proteome of EVs isolated from different regions in normal and TS hearts.

Methods

Animals and experimental set-up

All animal work adhered to national guidelines for experimental animal use, the study is reported in accordance with ARRIVE guidelines, and the study protocol received approval from the regional Animal Ethics Committee in Gothenburg (Göteborgs djurförsöksetiska nämnd, Dnr 5.8.18-02426/2023). At the end of the experiment, all animals were euthanized using isoflurane anaesthesia followed by heart excision, in accordance with approved ethical permit. No inotropic or vasoactive support was administered as part of the protocol. A total of 16 outbred male nRjHan rats (Sprague Dawley, Janvier Labs, France), aged 7 weeks, were used. Our established catecholamine-based model reproduces key TS features in both sexes, with comparable development of apical dysfunction in females and males⁴. Male rats were selected to reduce biological variance and preserve statistical power for discovery proteomics within a feasible sample size. For summary statistics, please refer to Table 1. The rats were housed 3 per open cage and acclimatized for a week in a temperature-controlled facility (19–21 °C) with a 12-h light/dark cycle and free access to food and water. Rats were anesthetized intraperitoneally with ketamine (70 mg/kg) and midazolam (3.5 mg/kg), and then randomized 8:8 to TS induction or control. For TS induction, 1 mg/kg of isoprenaline (Sigma Aldrich) was infused over 15 min via the tail vein using an infusion pump⁴, while controls received saline only. Using previously described echocardiographic methods, cardiac function and RWMA were evaluated at 6 and 24 h. Fractional area change (FAC) was used to assess cardiac function, while left ventricular akinesia index (LVAI) measured RWMA. Ejection fraction was not calculated, as FAC is a robust global index in this model and correlates strongly with ejection fraction and strain in our prior work⁵. TS was defined as any visible akinetic segment in the long-axis view extending circumferentially from the apex, with a narrow transition zone, to a contractile basal segment⁵.

Table 1.

Baseline characteristics and echocardiographic indices.

	Control (N = 8)	TS24h (N = 8)
Weight (kg)
Mean (SD)	0.339 (0.0339)	0.334 (0.0191)
Median [Min, Max]	0.327 [0.308, 0.399]	0.332 [0.310, 0.358]
Age (days)
Mean (SD)	48.8 (3.24)	47.9 (2.47)
Median [Min, Max]	47.0 [47.0, 54.0]	47.0 [47.0, 54.0]
Sex
Male	8 (100%)	8 (100%)
FAC at 6 h (%)
Mean (SD)	56.1 (7.28)	29.2 (7.11)
Median [Min, Max]	56.6 [44.6, 64.8]	31.1 [15.9, 38.4]
LVAI at 6 h (%)
Mean (SD)	0 (0)	26.7 (6.52)
Median [Min, Max]	0 [0, 0]	27.6 [13.4, 34.1]
Apical FS at 6 h (%)
Mean (SD)	44.0 (6.48)	8.81 (6.46)
Median [Min, Max]	44.0 [35.5, 54.7]	10.1 [-1.97, 15.7]
Basal FS at 6 h (%)
Mean (SD)	39.8 (7.01)	31.2 (7.81)
Median [Min, Max]	41.9 [28.9, 48.1]	30.7 [20.0, 44.0]
FAC at 24 h (%)
Mean (SD)	52.5 (7.80)	51.4 (5.20)
Median [Min, Max]	51.6 [44.5, 62.5]	53.4 [43.9, 58.0]
LVAI at 24 h (%)
Mean (SD)	0 (0)	11.6 (6.16)
Median [Min, Max]	0 [0, 0]	12.5 [0, 20.1]
Apical FS at 24 h (%)
Mean (SD)	51.5 (9.52)	36.1 (7.55)
Median [Min, Max]	53.6 [39.0, 60.0]	38.1 [20.4, 42.5]
Basal FS at 24 h (%)
Mean (SD)	42.2 (9.59)	45.3 (7.41)
Median [Min, Max]	39.4 [34.6, 55.6]	47.0 [32.1, 55.2]

FAC; Fractional Area Change, LVAI; Left Ventricle Akinesia Index, FS; Fractional Shortening.

EV isolation and characterization

Rats were euthanized 24 h after TS induction or saline infusion. The 24-h time point was selected to enable standardized region-resolved tissue-EV proteomics at a feasible, clinically relevant stage after TS induction, at which regional contractile dysfunction remains detectable while downstream responses are established beyond the immediate catecholamine surge. Hearts were removed, washed in ice-cold phosphate-buffered saline, without prior aortic perfusion, and tissue samples from the apex and base were collected and kept on ice. To increase EV yield, corresponding segments from two hearts were pooled per condition, creating 16 samples. These were then divided into four groups, with four samples each: control-apex, control-base, TS-apex, and TS-base. We adapted Crescitelli et al.’s protocol to isolate and characterize EVs from tissue (Fig. 1)¹⁵. Briefly, samples were sliced into small pieces in DMEM/high glucose medium and incubated with collagenase D (2 mg/mL) and DNase I (40 U/mL) at 37 °C for 30 min. After differential centrifugations, the EV pellets were re-suspended in PBS, followed by an iodixanol density cushion to separate EVs from protein contaminants. The combined pool of small and large EVs, representing a mixed EV population, were collected and stored at − 80 °C. EVs were characterized using western blot, nanoparticle tracking analysis, and transmission electron microscopy. Analysis of EV markers of the proteomic data was compared against the EVpedia database and to a curated list of common EV proteins obtained from the EVpedia and Vesiclepedia^16,17.

Fig. 1.

Regional EV isolation from the left ventricle. Rats underwent Takotsubo syndrome (TS24h) induction or control treatment, after which heart tissue was collected from apical and basal segments. Segments from two hearts were pooled per condition, resulting in n = 16 samples (4 per group: control-apex, control-base, TS-apex, TS-base). The illustrated workflow details sequential steps: tissue mincing and enzymatic digestion (collagenase D and DNase I), filtration, stepwise differential centrifugation, and iodixanol cushion separation to isolate large and small EVs. The isolated mixed EV population were analysed by tandem mass tag-based proteomics. Protocol adapted from Crescitelli R. et al. Nature Protocols 2021¹⁵.

Immunoblotting

Frozen heart tissue was homogenized, and proteins were extracted with the T-PER™ Tissue Protein Extraction Reagent (Thermo Fisher Scientific). EVs were lysed with RIPA Cell Lysis Buffer (10X) (Cell Signaling) with protease/phosphatase inhibitors (Thermo Scientific REF89901) and protein was quantified by Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific). Equal amounts of total protein lysate were loaded and separated on a NuPAGE 4–12% Bis–Tris gel (Invitrogen) and transferred to a PVDF membrane (Bio-Rad). Blots were probed with specific antibodies for Calnexin (ab13504), Flotillin 1 (ab133497), Rab4 (ab109009), TSG101 (ab125011) from Abcam; CD63 (PA5-92370) from Invitrogen and Laminin A/C (4777) from Cell Signaling Technologies and then with the corresponding horseradish peroxidase-conjugated secondary antibody. Immunoblots were visualized with Immobilon Western Chemiluminescent Horseradish Peroxidase Substrate (Millipore) and a ChemiDoc Touch Imaging System (BioRad). Bands were quantified with Image Lab Software (Bio-Rad).

Transmission electron microscopy

EVs were analyzed by negative staining as previously described¹⁵. Briefly, 5 µg of EVs was applied onto glow-discharged 200-mesh formvar/carbon-coated copper grids (Electron Microscopy Sciences, Hatfield Township, PA). After two washes with H₂O, EVs were fixed in 2.5% glutaraldehyde, followed by two additional washes. The samples were then stained with 2% uranyl acetate for 1.5 min. Grids were examined using a digitized Talos L120C transmission electron microscope (Thermo Fisher Scientific) operating at 120 kV and equipped with a CCD camera.

Nanoparticle tracking analysis

EV samples were analyzed for particle concentration and size distribution using the ZetaView® PMX120 system (Particle Metrix, Germany). The instrument was calibrated with 100-nm polystyrene beads (Sigma) prior to measurement. For each sample, a dilution in PBS was prepared, with typical dilution factors ranging from 1:1000 to 1:5000 depending on EV concentration. The diluted sample was introduced into the instrument via a syringe, and measurements were acquired at eleven positions, with three cycles recorded per position. Data were processed using ZetaView® software (version 8.05.11, SP1). The camera sensitivity was set to 80, shutter speed to 100, The brightness threshold was set to 20, and the particle size detection range was defined between 10 and 1000 nm.

Global relative quantification

Details for the experiment are provided in Supporting Information. Relative quantification was performed to compare protein expression in extra vesicular isolations from rat heart tissue. The samples were processed using modified filter-aided sample preparation (FASP) method¹⁸. Samples (40 µg) were reduced, transferred to filters, washed several times with 8 M urea and once with digestion buffer prior to alkylation and digestion overnight. Peptides were labelled using TMTpro 18-plex isobaric mass tagging reagents (Thermo Fisher Scientific) and pooled into one TMT-set, purified by HiPPR Detergent Removal Resin and Pierce™ and Peptide Desalting Spin Columns (both Thermo Fischer Scientific). The TMT-set was basic reversed-phase chromatography (bRP-LC, pH10) fractionated into 30 fractions over 70 min. Each fraction was analysed on Orbitrap Eclipse Tribrid mass spectrometer equipped with the FAIMS Pro ion mobility system interfaced with nLC 1200 liquid chromatography system (all Thermo Fisher Scientific). Peptides were separated on a C18 35 cm column over 90 min and data were acquired with SPS MS3 method. Raw files were processed and analyzed with Proteome Discoverer (Ver 2.4, Thermo Scientific) against UniProt Swiss-Prot Rattus norvegicus using Sequest as a search engine.

The MS proteomics data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository¹⁹ with the data set identifier PXD069879.

Statistical analysis

Randomization to induction was blocked by day, blinded to the researchers, and uploaded to the REDCap electronic data capture tools²⁰, where all study data was collected. Data analyses were conducted using R 4.3.3 with R Studio 2023.12.1 + 402. For EV yield endpoints, protein content (µg/g tissue), particle concentration (particles/g tissue), and particle-to-protein ratio (particles/µg protein) were log10-transformed prior to analysis and analyzed using a repeated-measures mixed-model two-way ANOVA with fixed effects of condition (control vs TS24h), region (apex vs base), and their interaction (condition × region). Animal ID was specified as the subject (random) factor and region as the within-subject repeated factor (compound symmetry covariance). Planned contrasts (base vs apex within each condition; TS24h vs control within each region) are reported with 95% confidence intervals, and p-values were adjusted using the Benjamini–Hochberg procedure (Supplementary Table 3).

All identified proteins were selected for principal component analysis (PCA). Raw protein-level abundance values were log2-transformed for variance stabilization, followed by differential protein expression analysis with the limma package (version 3.58.1)²¹. Proteins with absolute log2 fold changes of > 0.58 and benjamini–hochberg adjusted p-value of < 0.05 were, if not otherwise stated, considered differentially expressed. The visualisation of differentially expressed proteins (DEPs) was carried out with the EnhancedVolcano package (1.20.0). Heatmaps of protein expression data were generated using Z-score normalization. Hierarchical clustering was performed with the average linkage method, utilizing euclidean distance for row clustering and correlation distance for column clustering. P-values for the hierarchical clustering were calculated using the pvclust package with multiscale bootstrap resampling (2.2.0)²². Over representation analysis (ORA) was performed on DEPs with enrichment analysis using the clusterPorfiler (4.10.1) by querying the Gene Ontology database (GO, BP; Biological Processes, MF; Molecular Functions, CC; Cellular Component). Top 10 significantly enriched GO terms were visualised on dot plots. Gene set enrichment analysis (GSEA) was performed by quering the GO-BP and Reactome Pathway data base²³. The results from GSEA of all significant GO-BP were further analysed using an enrichment map approach²⁴. Semantic similarity among enriched biological processes was computed for Rattus norvegicus using the Wang method through the GOSemSim (2.28.1) package and visualized through Cytoscape (3.10.1)^25,26. GO term descriptions were mapped, and edges with a similarity score greater than 0.425 were retained to highlight closely related processes. An MCL cluster algorithm was later applied within the Cytoscape for further clustering and labelling using AutoAnnotate (v1.4.1). Gene set variation analysis (GSVA) was performed on enriched Reactome pathways, to quantify the differences between conditions and segments, using the GSVA package (1.50.5)²⁷. Protein–protein interaction analysis was conducted in Cytoscape, with interactions enriched through the STRING database. Protein function was characterized by querying the Pharos database²⁸. Weighted correlation network analysis (WGCNA) was performed through the WGCNA package (1.73)²⁹.

Results

EVs successfully isolated from cardiac tissue

After isolating EVs from the cardiac tissue of both apical and basal segments from TS and normal hearts, we observed pure, cup-shaped, vesicles ranging from 50 to 400 nm in size by electron microscopy (Fig. 2C). The proteomic analysis of EVs identified 85% of the EV markers listed in EVpedia and confirmed the presence of key EV-associated proteins. These included proteins categorized under Rabs, Annexins, Tetraspanins, Common EV markers, Heat shock proteins, and ESCRT complexes (Supplementary File Table 1). Western blot experiments confirmed the presence of typical EV markers, including Flotillin 1, TSG101, and CD63 (Fig. 2A). A proteomics-based marker assessment did not suggest disproportionate non-vesicular co-isolate markers in the EV preparations (Supplementary Fig. 1; Supplementary Table 2). No significant differences in EV protein or particle concentrations were observed between the apex and base regions in control hearts. However, TS hearts exhibited a notable reduction in both EV protein content and particle number specifically in the apex region, while the base remained relatively unchanged. Importantly, the average ratio of protein content per EV particle was comparable across all groups, indicating that the decrease in total EV protein in the TS apex is primarily attributable to fewer EVs recovered. Nanoparticle tracking analysis revealed similar size distribution profiles across all groups, characterized by a peak around 150 nm; however, the TS apex group displayed a markedly lower EV concentration throughout this size range (Fig. 2B, D; Supplementary Table 3). Collectively, the data illustrate that EVs isolated from cardiac tissue showed typical EV morphology and protein marker profiles, validating successful isolation, and while numbers of EVs recovered was uniform across regions in normal hearts, TS hearts displayed a region-specific reduction in EVs from the affected apex.

Fig. 2.

Characterization of isolated EVs. (A) Representative western blot confirming canonical EV marker presence and minimal contamination in preparations from apical and basal regions of control and TS24h hearts. The bands shown were cropped for clarity; original, full-length uncropped blots are available in Supplementary File. Similar purity was observed in all isolations. (B) EV protein content and particle concentration (per gram tissue) and particle-to-protein ratio for apex and base regions in control and TS24h groups. TS24h hearts show a pronounced reduction in recovered EV protein and particle number in the apex, but not the base. Results shown as mixed-model ANOVA with Benjamini–Hochberg correction; *P < 0.05, **P < 0.01, ns = not significant. (C) Transmission electron microscopy images illustrating the morphology of EVs, revealing typical cup-shaped vesicles ranging from 50 to 400 nm in both control and TS24h samples. (D) Nanoparticle tracking analysis shows similar size profiles (~ 150 nm peak), with reduced EV concentration in the TS apex group.

The TS phenotype is coupled with distinct regional protein alterations in EVs

A total of 2093 proteins were quantified. In control hearts, apical and basal EVs clustered close to the lower-central region of PC1 and PC2 (explaining 32% and 18% of the variation, respectively), thus indicating global proteomic similarity within the normal heart (Fig. 3A). Following isoprenaline challenge and the development of the TS phenotype with apical contractile dysfunction, a notable shift along both PC1 and PC2 towards the top-left quadrant was seen in apical EVs, suggesting TS-induced proteomic changes. However, the basal segment, exposed to similar catecholaminergic surge, did not show any obvious changes in PC1 or PC2 in their EV proteome, indicating a localized proteomic response in the affected apical segment and disrupting the previously uniform proteomic state. In support of these observations, differential protein expression analysis between EVs of TS-base vs control-base, and control-apex vs control-base showed few DEPs (Fig. 3B). We further investigated the EV samples similarities across the dataset and applied hierarchical clustering using a multiscale bootstrap of all proteins across all samples. By doing so, the distinct separation of EVs from TS-apex was apparent (Fig. 3C; Supplementary Fig. 2). The remaining EV groups did not cluster in a significant way. The two suspected outliers observed in the PCA plot (Fig. 2), integrated well with the main clusters in the heatmap and was not removed. In the differential protein expression analysis between TS-apex vs control-apex and TS-apex vs TS-base, a total of 238 and 562 DEPs, respectively, were observed (Fig. 3D). The top ten up- and downregulated proteins for these two key contrasts (TS-apex vs control-apex and TS-apex vs TS-base) are provided in Supplementary Fig. 3. The greater number of DEPs in the TS-apex vs TS-base comparison likely reflects the region-specific nature of protein expression changes in response to stress, with apex being more affected, coupled with differences in intra- and inter-group variability that influence the detection of significant differences. Among the most highly upregulated proteins in both comparisons were the proteins typical of immune activation and regulation, such as Ficolin-2; known to activate the complement system via the lectin pathway. A list of the up- and downregulated proteins can be found in the supplementary dataset file 2. ORA for the GO-BP of the 196 DEPs intersected between TS-apex vs control-apex and TS-apex vs TS-base, revealed 394 biological processes (Supplementary dataset file 3). The most significantly differentially regulated process among the proteins was the immune system process (GO:0002376). Similar results were observed when analysing them separately (Fig. 3E). Together, this indicates that proteomics analysis of EVs from heart tissue of an animal TS-model undergo distinct regional alterations, with inflammatory and immunologic processes playing an important role.

Fig. 3.

Differentially expressed proteins from different segments (apex, base) and cardiac states (TS, control). (A) Schematic overview of study design and principal component analysis (PCA) of control and 24 h after TS-induction. The 24-h phase is recognized by its recovery phase in the TS model, this induced a shift along both principal components (PC1 and PC2). Colours indicate cardiac segment and condition, and PC1 and PC2 explain 32% and 18% variance, respectively. (B) Volcano plots. Dotted lines represent the threshold for DEPs (horizontal, adjusted p-value < 0.05; vertical, absolute Log2FC > 0.58 or 1.5-fold change). (C) Heat map of all identified proteins, DEPs in TS-apex vs Control-apex, and TS-apex vs TS-base respectively, with z-score normalization values. (D) Venn diagram of shared and unique DEPs between TS-apex vs Control-apex (red) and TS-apex vs TS-base (green), number of up- and downregulated proteins are annotated. (E) Overrepresented GO “Biological Process”, “Cellular Component”, and “Molecular Function” terms. The dotplot visualizes the top 10 significantly enriched terms, ranked by their adjusted p-values (Benjamini–Hochberg correction). Each dot represents a GO term, where the color intensity indicates the enrichment significance (adjusted p-value), and the dot size reflects the number of DEPs associated with each term. The terms are ordered by their relative representation (GeneRatio). This analysis highlights the GO terms most significantly affected in the apex 24 h after TS-induction.

EV proteome of TS-apex express changes related to immune response, tissue repair, survival, and mitochondrial function

When we analysed the protein–protein interactions of all the significant proteins observed in TS-apex vs control-apex and TS-apex vs TS-base, subnetworks involved in specific biological processes emerged (Fig. 4A). Some of the biological processes included an increased abundance linked to lipids response, collagen biosynthesis, blood coagulation, and MAPK-signalling. We also observed a decreased abundance of key proteins related to the mitochondrial respiration, such as NADH dehydrogenase [ubiquinone] flavoprotein 3, mitochondrial; accessory subunit of the mitochondrial membrane respiratory chain NADH dehydrogenase (Complex I), Cytochrome c oxidase subunit 4 isoform 1, mitochondrial; Component of the cytochrome c oxidase (Complex IV), and Solute carrier family 25 member 51; mediator of mitochondrial uptake and maintenance of NAD⁺ levels (Fig. 4B,C). This echoed our findings from the enrichment analysis conducted via STRING on all the up- and downregulated proteins, respectively (Supplementary Fig. 4 & 5). To overcome the limitations of predefined thresholds and enable the identification of biologically meaningful pathway-level changes that may have been missed by ORA, we applied GSEA. Although GSEA is traditionally applied to transcriptomic data, here it was used to analyse proteomics data, identifying gene sets showing coordinated changes in protein abundance (Supplementary dataset file 4). Through the enrichment map analysis, we were able to categorize the biological processes into 6 themes based on their biological relevance, functional connectivity and network structure (Fig. 5A). We found gene sets related to mitochondrial function and energy production showing coordinated downregulation in the TS-apex. In contrast, the remaining five themes were predominantly upregulated, indicating an active response in TS-apex. Immune defence and inflammatory responses were prominent, alongside increased signalling related to cell motility, adhesion, and immune activation. Pathways involved in tissue repair and vascular homeostasis also showed enrichment, reflecting ongoing processes in the aftermath of stress. Additionally, processes linked to regulation of gene expression and protein activity, as well as signalling and structural differentiation, were enriched, consistent with a dynamic cellular remodelling environment in the affected apex. Similar themes were observed from the GSEA Reactome Pathway analysis (Supplementary dataset file 5 & Supplementary File). GSVA corroborated these findings, which further confirmed the upregulation of immune response and cellular organization pathways, alongside the downregulation of mitochondrial function and energy production pathways in the TS-apex (Fig. 5B). To further explore the coordinated expression of proteins in our dataset and identify key drivers of biological processes, we employed WGCNA. Unlike the enrichment-based methods previously described, WGCNA enables the detection of modules of co-expressed proteins, providing a systems-level perspective on how proteins may function together within the same regulatory networks. By correlating these modules with our phenotypic groups, we were able to identify clusters of proteins that show statistically significant co-expression patterns in the TS-apex, particularly highlighting proteins involved in mitochondrial dysfunction and immune activation (tan, and red and yellow respectively, Supplementary Fig. 6 & 7 and supplementary dataset file 6). Notably, the “midnightblue” module, which displayed distinct expression patterns with phenotypic groups, was enriched in KEGG pathways associated with RAS and chemokine signalling as well as cancer-related processes, with MAPK1 emerging as a top hub protein (Supplementary Fig. 7). Given MAPK’s key role in promoting cell proliferation, survival, and growth, these findings suggest that pathways related to cell survival and proliferation may be central to the heart’s response to stress in the TS-apex and are associated with EV-isolated samples. In summary, these analyses reveal that the TS-apex EV proteome displays a decreased abundance of proteins associated with mitochondrial function and energy production alongside increased abundance of proteins related to immune defence, inflammation, tissue repair, and signalling pathways including MAPK.

Fig. 4.

Protein–Protein Interaction Network Analysis of Differentially Expressed Proteins. (A) Network visualization of protein interactions showing DEPs and their functional relationships. Interactions were filtered using a confidence score cut-off of 0.4. Node colours indicate protein function. Each node is surrounded by a circle divided into two semicircular sections: the first half (12 o’clock clockwise) represents TS-apex vs Control-apex comparison, while the second half represents TS-apex vs TS-base comparison. Red slices indicate upregulated proteins while blue slices represent downregulated proteins. Network clustering was performed using the MCL algorithm through AutoAnnotate, with edge weights derived from STRING confidence scores. (B) Heat maps of biological processes and the identified proteins with z-score normalization values. Hierarchical clustering of columns was performed using Canberra distance metric. (C) Focused subnetwork highlighting the downregulated proteins specifically involved in mitochondrial respiration pathways.

Fig. 5.

Enrichment Map Analysis of Biological Processes from GSEA. (A) Network visualization of enriched biological processes. Node represents significant GO-BP, and edges indicate the semantic similarity score (> 0.425) between processes, calculated using the Wang method. Inner rings compare TS-apex vs Control-apex, while outer ring compares TS-apex vs TS-base. Ring colour indicates enrichment direction (red: positive; blue: negative). Clusters are labelled using AutoAnnotate. An additional higher-level clustering revealed 6 broader clusters/themes (highlighted with coloured boundaries) representing major biological functions affected in the system. (B) Radar chart depicting the results of GSVA of selected Reactome pathways for the respective groups. The distance from the centre indicates the GSVA score (scale -0.6—0.6).

Discussion

This work presents the first comparative proteome analysis of tissue-derived EVs isolated from the apical and basal regions of the LV in both normal and TS phenotype hearts. We found that TS induces marked changes in the EV protein composition, shifting from a homogeneous apex/base profile in controls to a distinct signature in the TS apex that differs significantly from both the base and normal hearts. EVs from the transiently non-contracting apex thus represent a unique molecular state, further supporting the regional specificity of TS, even under uniform systemic catecholamine exposure. We also observed a lower EV yield from the TS apex compared with the TS base and controls, in contrast to the increased EV levels often reported in other cardiovascular conditions such as myocardial infarction. This divergence may reflect differences in myocardial response and pathophysiological mechanisms between TS and ischemic injury. Importantly, the particle-to-protein ratio and vesicle size distribution were similar across groups, suggesting broadly preserved vesicle properties and cargo loading. However, because tissue-derived EV isolation quantifies EVs recovered from the tissue rather than directly measuring EV biogenesis, release, or clearance in vivo, the lower EV yield from the TS apex could reflect several non-mutually exclusive mechanisms, including reduced local formation or release, increased export from the tissue, altered retention or uptake by recipient cells, enhanced clearance, or regional structural changes (e.g., edema or injury-related tissue remodelling) that affect digestion and recovery efficiency. In addition, region- or condition-dependent co-isolation of non-vesicular proteins could influence protein-based normalization. Accordingly, the biological interpretation of the lower TS-apex EV yield—whether it reflects true differences in EV biogenesis/trafficking in vivo or region—cannot be determined from the present data and will require targeted studies of EV production, trafficking, and fate in TS.

The EV proteome from each cardiac region may reflect the local cellular state or act as a vehicle for altering recipient cell function. Both perspectives should be considered when interpreting the regional protein profiles presented here. Because PPI networks, ORA/GSEA, enrichment maps and WGCNA are all inference-based frameworks, the pathway-level conclusions presented here should be considered hypothesis-generating, aimed at prioritizing candidate processes and proteins for future validation rather than proving specific mechanistic pathways.

We found a significant downregulation of proteins related to mitochondrial function and energy metabolism in the TS-apex. Energetic alterations have been reported in TS patients during the acute phase, characterized by reduced PCr/ATP ratio³⁰. Our data are consistent with the idea of multi-level energetic changes, including alterations in electron transport chain function, NAD + homeostasis, and oxidative phosphorylation. Although GSEA indicates coordinated changes in protein abundance, further functional studies are needed to confirm the impact of these changes on mitochondrial activity and metabolic remodelling. Moreover, it remains to be determined whether these changes are a primary driver or a secondary consequence of the condition.

Immune defence and inflammatory responses were dominant features of the EV proteome. We identified biological processes related to both immune activation and regulation, including processes involved in cell motility, adhesion, and immune cell activation. This dual profile indicates that the EV proteome reflects both activation and regulatory components of the inflammatory response, consistent with a complex and regulated process rather than a purely pro-inflammatory state. This interpretation aligns with clinical and experimental evidence that TTS is accompanied by myocardial and systemic immune activation, including macrophage-associated myocardial inflammation and altered circulating cytokines/monocyte subsets^31,32. Consistent with the clinical relevance of inflammatory burden in TS, higher inflammatory biomarkers are independently associated with delayed recovery in left ventricular systolic function beyond 10 days (~ 25% of patients) and worse long-term outcomes³³. In this context, the prominent immune and inflammatory signatures in the TS-apex EV proteome at 24 h may reflect regional inflammatory signalling during the subacute phase and support the concept that modulating inflammatory pathways may be therapeutically relevant in selected TS phenotypes³⁴. EV profiling may therefore complement future efforts to define inflammatory TS phenotypes, but our data remain hypothesis-generating. In parallel, our finding of coordinated immune pathway enrichment alongside downregulation of mitochondrial/energy-related proteins in TS-apex EVs may reflect a coupled inflammatory–metabolic stress response during early recovery. Whether these EV-associated signatures reflect protective versus adverse processes cannot be determined from the present data and will require mechanistic studies.

Our network analysis further revealed a notable relationship between lipid metabolism and inflammation in TS. Histologically, oedema and lipid accumulation appear as early as 6 h post-stress, followed by inflammatory cell infiltration at 24 h, and ultimately fibrosis at 30 days in our model⁴. Comparable lipid accumulation and inflammatory response have been reported in patients^31,35,36. The strong functional connectivity between lipid response and inflammatory processes observed in our network analysis suggests that early lipid accumulation may reflect and potentially contribute to the initiation of the subsequent inflammatory cascade. This hypothesis is supported by our GSEA and enrichment map analysis, which showed coordinated upregulation of both lipid response and immune defence processes. The EVs appear to reflect and potentially contribute to this pathological sequence, carrying proteins involved in lipid handling and immune activation. This link between lipid metabolism and inflammation may represent an important downstream cascade, potentially contributing to both acute manifestations and chronic tissue changes in TS patients. Alongside the inflammatory response, we observed significant upregulation of biological processes related to coordinated tissue repair and vascular adaptation processes in the TS-apex. These included pathways involved in wound healing, blood coagulation, angiogenesis, and collagen biosynthesis. This active repair signature suggests a complex adaptive response involving both extracellular matrix remodelling and vascular changes, characterized by a procoagulative state and promotion of angiogenesis. Such repair processes may reflect and potentially participate in the development of thrombus formation and microscopic fibrosis observed in TS patients. They could be associated with the long-term sequela³⁶, while also possibly representing compensatory responses to restore microvascular integrity and improve perfusion in regions affected by stress-induced damage.

A hallmark of TS is recovery of systolic function despite a severe initial presentation. In our tissue-EV proteomics, TS-apex EV cargo was enriched for proteins linked to cell survival and regulation of cell death, including MAPK pathway components. This aligns with prior reports implicating PI3K/AKT-related survival signalling in TS patients and with experimental data showing that EVs can activate MAPK-dependent protective responses in cardiomyocytes^37,38. Together, these findings suggest that the TS-apex EV proteome reflects activation of pro-survival programs that may relate to recovery and cardioprotection during resolution of RWMA⁵. Enrichment of MAPK-related EV proteins is also consistent with catecholamine-triggered β-adrenergic stress signalling. Beyond canonical G-protein signalling, β-arrestin–dependent pathways at β1-adrenergic and AT1 receptors can engage ERK and PI3K/AKT signalling and have been linked to cardiomyocyte survival under stress^39,40. In this framework, β-blockers could reduce catecholamine-G protein-mediated injury while potentially preserving signalling through β-arrestin-mediated pro-survival pathways. Carvedilol provides a precedent for such pathway selective modulation⁴¹. Registry data in a propensity score–matched cohort further report an association between beta-blocker therapy at discharge and lower follow-up mortality in TS, with early separation of event curves, although causality cannot be inferred⁴². Given the mitochondrial/energy-related changes in our TS-apex EV proteome, future studies could also examine whether β-blockade modulates EV-associated energetic pathways. EV-associated stress/survival programs may additionally be of cardio-oncology interest, considering the bidirectional relationship between TS and cancer and evidence that cardiac EVs can influence tumour growth after cardiac injury^43,44.

Limitations

Tissue-derived EV preparations have inherent constraints. Our approach profiles a mixed EV population recovered from fresh myocardial tissue and therefore cannot capture circulating EVs, resolve EV trafficking (release, uptake, clearance, biogenesis), or assign cellular origin. Hearts were rinsed in ice-cold PBS, and tissue processing can introduce contributions from residual intravascular/extravasated blood-associated material, digestion-related intracellular carryover, and non-vesicular co-isolates. Although all groups were processed identically with paired apex/base sampling within each animal and EV enrichment included differential centrifugation followed by an iodixanol (OptiPrep) density cushion, we cannot fully exclude region- or condition-specific effects or co-isolates influencing the tissue-EV proteome. Western blots were used qualitatively, additional negative markers would further strengthen purity assessment. The study design is limited to adult male rats and a single post-induction time point (24 h), which captures a subacute/early recovery phase rather than peak dysfunction (~ 6 h) or later complete recovery; generalizability to human TS, female rats, and other TS phases is therefore limited. Finally, bioinformatic pathway/network analyses are inference-based and should be interpreted as hypothesis-generating; independent mechanistic validation and temporal profiling are needed to establish causality and define EV functional roles.

Future directions

Our work presents the first isolation and characterization of EVs in the setting of the TS phenotype and stress-induced cardiomyopathy. Future studies should focus on several key aspects to further understand the role of EVs and mechanisms in TS pathophysiology. First, temporal profiling of EV content at multiple time points would provide valuable insights into the dynamic changes during progression and recovery. Second, functional studies investigating EV trafficking and cellular uptake could help determine how EVs mediate cell-to-cell communication between different cardiac regions and cell types. Third, the relationship between TS development and recovery with the activation of cardioprotective pathways and energetic impairments. Finally, investigation of EVs in TS patients could provide potential diagnostic biomarkers and therapeutic targets.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

TS

Takotsubo syndrome

FAC

Fractional area change

LVAI

Left ventricular akinesia index

RWMA

Regional wall motion abnormality

EVs

Extracellular vesicles

Author contributions

E.Z.: Performed animal experiments, bioinformatics and data analysis, drafted the manuscript, contributed to study design and interpretation of data. A.N., M.K.: Contributed to data collection and interpretation, and critically reviewed the manuscript. A.M.: Isolated extracellular vesicles, performed protein quantitation and immunoblotting, and critically reviewed the manuscript. K.E.: Performed nanoparticle tracking analysis, contributed to study design and data interpretation, and critically reviewed the manuscript. R.C.: Performed transmission electron microscopy, contributed to study design and data interpretation, and critically reviewed the manuscript. A.E.: Contributed to data collection and interpretation, and critically reviewed the manuscript. A.K.: Performed global relative protein quantification, contributed to study design and data interpretation, and critically reviewed the manuscript. R.O.B.: Contributed to data interpretation and study design, and critically reviewed the manuscript. B.R.: Contributed to data interpretation and critically reviewed the manuscript. E.O.: Conceived the study, supervised the project, contributed to study design and data interpretation, ensured ethical oversight, and critically reviewed the manuscript.

Funding

Open access funding provided by University of Gothenburg. This study was funded by the Swedish Research Council (2020-02592), the Swedish Heart and Lung Foundation (20200826), and the Swedish state under the agreement between the Swedish government and the country councils (ALF-agreement, ALFGBG-966521). RC has received a research grant from by Vetenskapsrådet Etableringsbidrag (Starting Grant from the Swedish Research Council) (Grant # 2023-02239), the Assar Gabrielsson’s Foundation (Grant # FB23-01), the Serena Ehrenström foundation, the Ann-Lisa och Bror Björnssons Foundation, Wilhelm och Martina Lundgrens Vetenskapsfond (Grant # 2025-SA-4821 and 2023-SA-4142) and Magnus Bergvalls Foundation (2024-1009).

Data availability

The data underlying this article are available in the article and its online supplementary material.

Declarations

Competing interests

The authors declare no competing interests.