Multi-Omics Analysis of Circulating Exosomes in Adherent Long-Term Treated OSA Patients

Abstract

Obstructive sleep apnea (OSA) is a highly prevalent chronic disease affecting nearly a billion people globally and increasing the risk of multi-organ morbidity and overall mortality. However, the mechanisms underlying such adverse outcomes remain incompletely delineated. Extracellular vesicles (exosomes) are secreted by most cells, are involved in both proximal and long-distance intercellular communication, and contribute toward homeostasis under physiological conditions. A multi-omics integrative assessment of plasma-derived exosomes from adult OSA patients prior to and after 1-year adherent CPAP treatment is lacking. We conducted multi-omic integrative assessments of plasma-derived exosomes from adult OSA patients prior to and following 1-year adherent CPAP treatment to identify potential specific disease candidates. Fasting morning plasma exosomes isolated from 12 adult patients with polysomnographically-diagnosed OSA were analyzed before and after 12 months of adherent CPAP therapy (mean ≥ 6 h/night) (OSAT). Exosomes were characterized by flow cytometry, transmission electron microscopy, and nanoparticle tracking analysis. Endothelial cell barrier integrity, wound healing, and tube formation were also performed. Multi-omics analysis for exosome cargos was integrated. Exosomes derived from OSAT improved endothelial permeability and dysfunction as well as significant improvement in tube formation compared with OSA. Multi-omic approaches for OSA circulating exosomes included lipidomic, proteomic, and small RNA (miRNAs) assessments. We found 30 differentially expressed proteins (DEPs), 72 lipids (DELs), and 13 miRNAs (DEMs). We found that the cholesterol metabolism (has04979) pathway is associated with lipid classes in OSA patients. Among the 12 subjects of OSA and OSAT, seven subjects had complete comprehensive exosome cargo information including lipids, proteins, and miRNAs. Multi-omic approaches identify potential signature biomarkers in plasma exosomes that are responsive to adherent OSA treatment. These differentially expressed molecules may also play a mechanistic role in OSA-induced morbidities and their reversibility. Our data suggest that a multi-omic integrative approach might be useful in understanding how exosomes function, their origin, and their potential clinical relevance, all of which merit future exploration in the context of relevant phenotypic variance. Developing an integrated molecular classification should lead to improved diagnostic classification, risk stratification, and patient management of OSA by assigning molecular disease-specific therapies.

Keywords: OSA; exosomes; extracellular vesicles; lipids; miRNAs; multi-omics; omics; proteomics.

Figure 1

Exosomes derived from OSA subjects disrupt endothelial cell monolayer barrier integrity in vitro. (a) Ensemble-averaged curves of ECIS-measured endothelial cell barrier resistance changes over time after administration of exosomes from adult patients with OSA before treatment and after long-term-adherent CPAP (OSAT) therapy compared to endothelial cells incubated with plasma-free media and empty exosomes (control; black line). (b) Evaluation of ECIS-measured endothelial cell barrier resistance changes after exosome administration. Endothelial cells were grown on trans-well membranes to measure the integrity and permeability of the monolayer cells. (c) Continuous measurement of cell monolayer barrier function (TEER) using membrane inserts in multiple wells to measure the resistance across the trans-well membrane over time. TEER values for the average of each group: no exosomes, OSA, and OSAT. ** p < 0.001.

Figure 2

Plasma-exosome-induced wound healing in human endothelial cells. Comparative analysis of real-time endothelial barrier integrity following wounding on 8W10E arrays. Plasma-derived exosomes from OSA and OSAT patients applied to human microvascular endothelial cells (HMVEC-d) for wound healing using ECIS system in vitro. (a) Representative graphs of comparative analysis of hMVEC-d cells treated with and without exosomes following wound healing. HMVEC-d cells were seeded at 0 h at a density of 50,000 cells into ECIS system arrays (8W10E) for 24 h; cells were wounded, then exosomes derived from OSA (n = 12) and OSAT were added and monitored for another 48 h. (b) Histogram showing the effects of exosome cargos derived from OSA or OSAT, as well as unwounded cells and wounded cells with no exosomes, on endothelial cell wound healing and recovery of HMVEC-d cells. * Indicates p < 0.01, while ** p < 0.001.

Figure 3

Plasma exosomes altered human endothelial cell tube formation (angiogenesis) in vitro. Plasma-derived exosomes from OSA or OSAT were applied on a 3-D matrix endothelial cell culture system to assess angiogenesis, and tube lengths were quantified using ImageJ software 2.9.0. The formation of tube-like structures was observed under bright field. (a) Representative of phase contrast micrographs of the capillary-like tubular structures of OSA and OSAT and compared with no exosomes for 24 h. Arrow indicates tube formation. (b) Line graphs showing total tube length, (c) longest tube length, and (d) shortest tube length. Tube formation was quantified by counting the number of branching points in the total photographed area. n = 12. Scale bar 100 μm.

Figure 4

Lipidomic analysis of OSA exosomes using LC-MS/MS. (a) Supervised OPLS-DA model showing separation of OSA and OSAT groups. (b) Heatmap showing the top differentially expressed lipids. (c) Volcano plot depicting the total differentially expressed lipids. The red color indicates up-regulation, and the blue color indicates down-regulation. n = 12/group.

Figure 5

Proteomic analysis of OSA and OSAT using in situ digestion on LC-MS. (a) Orthogonal partial least square discriminant analysis (OPLS-DA) for the separation of two groups, (b) heatmap analysis, (c) volcano plots and (d) protein–protein network for differentially expressed proteins. The red color indicates up-regulation, and the blue color indicates down-regulation. n = 12/group. Temporal changes of gene ontology and KEGG pathways for exosome proteomics analysis for OSA and OSAT subjects. The differentially expressed proteins (DEPs) were subjected to KEGG, GO, and disease description. (e) Top terms from GO functional enrichment analyses based on DEPs of biological processes (BP), cellular components (CC) and molecular functions. (f) Top KEGG pathways based on pathway enrichment analysis KEGG) of DEPs pathways, and (g) disease description. All proteins associated with disease tissue networks were identified according to the DISEASES database.

Figure 6

Plasma-derived exosome miRNA profiling for OSA and OSAT. (a) Orthogonal partial least square discriminant analysis (OPLS-DA); (b) Heatmap illustrating miRNA expression patterns in exosomes (dark red: increased miRNA expression; light blue: reduced miRNA expression). The dendrograms show hierarchical clustering representing the similarities and dissimilarities in expression profiles among individuals and miRNAs; and (c) volcano plots. Gene Ontology (GO) for the target prediction of differentially expressed miRNAs (miRNAs) in exosomes derived from OSA and OSAT subjects. GO analysis for (d) cellular components, (e) biological processes, and (f) molecular functions, and (g) KEGG pathways identified in target predication genes found in differentially expressed miRNAs in OSAT vs. OSA. Network visualization of the differentially expressed exosomal miRNAs derived from OSA and OSAT. (h) Network for 13 miRNAs, and (i) list of the miRNAs and their associated target predication genes. n = 12.

Figure 7

Graphical representation of a multi-block analysis performed on OSA and OSAT. (a) Variable plot highlighting the contributions for OSA clinical data, exosome cargo including lipids, proteins, and miRNAs. Markers in the outer circle are more significant and contribute more to separating the two conditions. Those in the inner circle are less significant than ones in outer circle. (b) Circos plot depicting the strongest correlation biomarkers in the multi-omic biomarker panel in OSA and OSAT. Circos plot of clinical, lipids, proteins, and miRNAs showing Spearman’s correlation analysis (p < 0.05) between OSA and OSAT multi-omics, classified based on the superclass. (c) Heatmap clustering of the variables (clinical data, lipids, proteins, and miRNAs) to represent the muti-omic profiles for the 7 samples. Red indicates a positive correlation, and blue indicates a negative correlation. The variables were selected by applying DIABLO to cellular frequency, gene, and metabolite module datasets depicted using a Circos plot. The variables indicated in the ideogram are connected with either red or blue to other variables if the correlation is either positive or negative. Only correlation above a certain threshold is depicted (r = 0.8). The lines around the ideogram are drawn by connecting the average expression value of a given variable for a certain phenotypic group.