Nocturnal hypertension (NH) is a high-risk, underdiagnosed blood pressure (BP) phenotype strongly associated with cardiovascular morbidity. Obstructive sleep apnea (OSA), which is common in patients with NH, promotes vascular injury through sympathetic activation, inflammation, and endothelial dysfunction; however, the molecular mechanisms underlying this association in patients with NH remain poorly defined. This study explored these mechanisms using targeted proteomics and endothelial cell models.
MethodsAdults undergoing polysomnography (PSG) with NH, defined as nighttime BP ≥120/70mmHg at the time of 24-hour ambulatory BP monitoring, were included. Fasting blood samples were collected after PSG. Participants were classified as controls, defined as an apnea-hypopnea index (AHI) <15eventsh−1, or severe OSA, defined as AHI ≥30eventsh−1; patients with moderate OSA, defined as AHI between 15 and 30eventsh−1, were excluded to maximize contrast between groups. Participants were matched for age, sex, BMI, and nocturnal mean arterial pressure. The Olink® platform was used to quantify proteins. Differential abundance was assessed using linear models, and sparse partial least squares discriminant analysis was used to identify OSA-associated protein signatures. Endothelial integrity was assessed in cells exposed to extracellular vesicles derived from a subset of participants (n=8 controls; n=8 severe OSA).
ResultsA total of 58 matched participants were included, with 29 participants per group. Overall, 70.7% were men, the median age was 48 years, and the median BMI was 29.5kg/m2. Twenty-one proteins were differentially abundant and were enriched in pathways related to cell adhesion maintenance and extracellular vesicle composition. A 13-protein signature associated with OSA showed interconnectivity and enrichment for endothelial regulatory pathways. Extracellular vesicles from patients with OSA showed a trend toward increased endothelial barrier disruption compared with controls.
ConclusionsIn patients with NH, severe OSA appears to be associated with molecular alterations indicative of endothelial dysfunction, providing preliminary mechanistic insight into elevated cardiovascular risk that may help refine risk stratification.
Obstructive sleep apnea (OSA) is characterized by repeated episodes of upper airway obstruction during sleep. OSA leads to arousals, fragmented sleep, sympathetic overactivation, and oxidative stress [1]. Previous evidence indicates that OSA is associated with endothelial dysfunction, atherosclerosis, and blood pressure (BP) dysregulation [2], which increase the risk of cardiovascular disease (CVD) [3]. Consequently, OSA is considered an important condition in cardiovascular risk assessment, especially in high-risk individuals.
Patients with nocturnal hypertension (NH), defined as nighttime BP ≥120/70mmHg [4], constitute a population at high cardiovascular risk, with increased odds of developing target organ damage [5] and CVD [6]. The prevalence of isolated NH has been estimated to be approximately 20% [7], rising to nearly 80% overall in patients already affected by hypertension [8]. Nocturnal BP is the BP parameter most strongly correlated with cardiovascular events and mortality [9]. In line with this, the 2024 European Society of Cardiology clinical practice guidelines acknowledge the strong association between elevated nocturnal BP and cardiovascular risk [4]. The prevalence of OSA in patients with hypertension is high, ranging from 30% to 50% [10], and increases further among those with NH [11]. Given the high prevalence of OSA and its contribution to cardiovascular risk in this population [12,13], concomitant NH and OSA represent a clinically relevant condition. Consistently, patients with NH and coexisting OSA are at an increased risk of adverse cardiovascular outcomes [14].
Despite the existing evidence on the clinical consequences of OSA for cardiovascular risk, the specific molecular mechanisms by which OSA may promote cardiovascular damage in patients with NH remain largely undefined. Addressing this mechanistic gap is especially important in such a markedly high-risk population. Furthermore, because the clinical management of OSA still relies largely on symptoms such as excessive daytime sleepiness, studies evaluating the impact of OSA on cardiovascular injury may have important implications for treatment criteria, suggesting a therapeutic need independent of symptom burden.
Profiling circulating plasma proteins can shed light on pathway-level alterations underlying OSA-related cardiovascular injury [15,16]. Complementarily, in vitro models provide a controlled environment to test OSA-mediated stress in specific tissues. In this regard, plasma-derived extracellular vesicles from patients with OSA applied to endothelial cell cultures have been shown to elicit measurable functional changes [17,18], supporting their use in studying cell-specific OSA-mediated effects. Here, by integrating advanced proteomic profiling with clinical data and mechanistic in vitro functional assays, we performed an exploratory study aimed at elucidating potential molecular mechanisms by which OSA may lead to elevated cardiovascular risk in patients with NH.
MethodsStudy designThis is an ancillary study of an observational, prospective, cross-sectional study designed to evaluate molecular profiles of cardiovascular risk in patients with OSA (ClinicalTrials.gov: NCT03513926). Briefly, the study included consecutive patients aged 18–60 years who were referred to the sleep unit for polysomnography. Key exclusion criteria included a previously diagnosed sleep disorder and a past medical history of continuous positive airway pressure treatment. Additional details on the initial study can be found in Supplementary Methods. For the present study, patients with a history of cardiac disease, listed in the Appendix, were excluded because of the potential of these conditions to alter several biomarkers [19].
Sleep evaluation and blood pressure monitoringOSA was diagnosed by in-laboratory polysomnography (PSG) using Philips Sleepware G3, Amsterdam, the Netherlands. Trained personnel analyzed the sleep studies according to standard criteria [20]. Apnea was defined as an interruption or reduction in oronasal airflow of ≥90% lasting at least 10s. Hypopnea was defined as a 30–90% reduction in oronasal airflow for at least 10s associated with oxygen desaturation ≥3% or evidence of arousal on electroencephalography. The apnea-hypopnea index (AHI) was calculated as the average number of apnea and hypopnea events per hour of sleep.
The morning after PSG, patients underwent 24-hour ambulatory blood pressure monitoring (ABPM) using the Mortara Ambulo 2400 (Milwaukee, WI, United States) according to internationally recommended procedures [21]. BP was measured every 15min during the day and every 30min at night, with sleep periods defined by patient reports. ABPM recordings were considered optimal when the percentage of valid measurements exceeded 70%, with at least 1 measurement every hour; otherwise, monitoring was repeated. Patients with NH, defined as nighttime BP ≥120/70mmHg [4], were included in the present study. A nondipping pattern was defined as a dipping ratio, calculated as the ratio of night/day BP, ≥0.9.
Patients were classified according to AHI into two extreme phenotypes: a control group, defined as AHI <15eventsh−1, and a severe OSA group, defined as AHI ≥30eventsh−1. Patients with moderate OSA, defined as AHI 15–29eventsh−1, were intentionally excluded to minimize phenotypic heterogeneity and specifically isolate the pathophysiological impact associated with severe disease. This extreme-group approach was chosen to enhance contrast and maximize the ability to detect OSA-related deleterious effects. A final subsample of patients was matched by age, sex, BMI, and nocturnal mean arterial pressure (MAP).
Proteomic analysisFasting venous blood samples were collected the morning after PSG in ethylenediaminetetraacetic acid anticoagulant tubes. Samples were centrifuged at 1500×g for 10min at 4°C to separate the plasma fraction, aliquoted, and stored in a dedicated −80°C freezer. Proteins were quantified using proximity extension assays (Olink Proteomics, Uppsala, Sweden). The standardized Cardiovascular II, Cardiovascular III, Cardiometabolic, and Organ Damage panels, which measured 366 unique proteins in total, were selected because of the relevance of the proteins included in mechanistic pathways involved in CVD. Details are available on the Olink website. All assays were performed in a blinded fashion.
Assay quality control was monitored using internal extension and interpolation controls to adjust for intra- and interrun variation. The data obtained are expressed as normalized protein expression (NPX) values. Only data from plates in which the internal control SD was within 0.2 median NPX for this control were reported, a criterion that all samples met. Sample-level quality control was evaluated using NPX distribution plots and multivariate visualization with principal component analysis (PCA) plots to detect outliers or global shifts in NPX distributions.
Duplicate proteins and proteins with more than 25% of values below the limit of detection were excluded from the analysis.
Statistical and bioinformatic analysesThe study population was characterized using descriptive statistics. Data are reported as frequencies and percentages for categorical variables and as median [Q1:Q3] for continuous variables. Clinical and sociodemographic characteristics were compared between study groups using Wilcoxon rank-sum tests for continuous variables and χ2 tests, or Fisher exact tests when expected cell counts were <5, for categorical variables.
To evaluate the relationship between OSA-related proteomic changes and nocturnal BP, a combination of linear models using empirical Bayes moderated t statistics, implemented with the R package limma [22], and Spearman rank correlations was used. Owing to the exploratory nature of the study, proteins were considered differentially abundant between the severe OSA and control groups if they met both of the following criteria: (1) statistical and fold-change criteria, defined as P≤.05 and fold change ≤0.87 or ≥1.15 and (2) correlation with at least 1 nocturnal BP parameter within the severe OSA group, defined as |ρ|>0.2. To study the biological relevance of these proteins, functional enrichment analysis of Gene Ontology (GO) terms was performed using the R package clusterProfiler, version 4.10.1 [23]. Only terms with a Benjamini–Hochberg-adjusted P≤.05 were reported.
To identify a protein signature associated with the presence of severe OSA, sparse partial least squares discriminant analysis (sPLS-DA) was performed using the mixOmics R package, version 2.4 [24]. A 2-component model was specified, and model parameters were tuned using 5-fold cross-validation. In-sample risk of severe OSA according to the presence of this protein signature was modeled using generalized additive models implemented with the mgcv R package [25]. Functional interactions between signature proteins were predicted using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) [26]. Only interactions whose evidence score reached the ≥0.4 threshold were reported. Data from the Genotype-Tissue Expression (GTEx) project [27] were queried to assess normal expression patterns of signature proteins in tissues and cell types of potential interest.
To evaluate the potential impact of antihypertensive medications on the circulating proteome [28], an identical analytical approach was applied after stratifying the study cohort by prescription of antihypertensive drugs.
Functional in vitro assaysOSA-mediated endothelial effects were explored using a battery of in vitro assays in human aortic endothelial cells (HAoECs) (Lonza, Basel, Switzerland) treated with plasma-derived extracellular vesicles from patients with and without OSA, using approaches similar to those reported elsewhere [29]. Briefly, extracellular vesicles were extracted from a subset of 16 patients from the full study cohort, comprising 8 controls and 8 patients with severe OSA, who were not receiving antihypertensive treatment to avoid the potential confounding effect of pharmacological treatment. They were matched by age, sex, BMI, and nocturnal MAP.
Extracellular vesicles were isolated using the Total Exosome Isolation Kit from plasma (Thermo Fisher Scientific, Waltham, MA, USA). Correct extracellular vesicle isolation was assessed structurally and molecularly using transmission electron microscopy and Western blotting of canonical extracellular vesicle markers.
The effects of extracellular vesicles on HAoECs were assessed by measuring expression of CDH5, the gene encoding vascular endothelial cadherin, using real-time quantitative polymerase chain reaction (RT-qPCR); measuring endothelial permeability through macromolecular permeability assays; and structurally visualizing CDH5 by immunofluorescence staining. Detailed protocols for in vitro-related procedures are provided in the Supplementary Data.
ResultsStudy populationThe selection process (Fig. 1) yielded 29 matched patients per group. Overall, the median age was 48 years, participants were predominantly male (70.7%), and median BMI was 29.5kg/m2 (Table 1). The most common reason for referral to the sleep unit was chronic snoring (79.3%), followed by observed apneas (44.8%) and somnolence (37.9%). A prior diagnosis of hypertension was present in 46.6% of patients, and 39.7% had been prescribed at least 1 antihypertensive drug. Median 24-hour systolic and diastolic BP values were 136 and 82mmHg, respectively, and a nondipping pattern was observed in 55.2% of the cohort.
Flowchart of the selection of patients for the protein quantification study. ABPM, ambulatory blood pressure monitoring; NH, nocturnal hypertension; PSG, polysomnography; OSA, obstructive sleep apnea; MAP, mean arterial pressure. Nocturnal hypertension was defined as systolic/diastolic blood pressure ≥120/70mmHg. Patients were controls if they experienced <15 apnea-hypopneas per hour, moderate if >15 but <30, and severe if >30.
Characteristics of the population selected for proteomic quantification.
| Characteristic | All participants(n=58) | Control(n=29) | Severe OSA(n=29) | P value | n |
|---|---|---|---|---|---|
| Sociodemographic and anthropometric characteristics | |||||
| Female sex | 17 (29.3%) | 9 (31.0%) | 8 (27.6%) | 1.00 | 58 |
| Age, y | 48.0 [44.2; 53.8] | 47.0 [45.0; 52.0] | 48.0 [44.0; 55.0] | .657 | 58 |
| BMI, kg/m2 | 29.5 [26.8; 33.3] | 29.7 [27.7; 32.7] | 29.4 [26.6; 33.3] | .981 | 58 |
| Smoking history | .098 | 58 | |||
| Non-smoker | 24 (41.4%) | 12 (41.4%) | 12 (41.4%) | ||
| Current smoker | 22 (37.9%) | 8 (27.6%) | 14 (48.3%) | ||
| Former smoker | 12 (20.7%) | 9 (31.0%) | 3 (10.3%) | ||
| Comorbidities | |||||
| Hypertension | 27 (46.6%) | 13 (46.4%) | 14 (50.0%) | 1.00 | 58 |
| Diabetes mellitus | 5 (8.6%) | 1 (3.4%) | 4 (13.8%) | .352 | 58 |
| Dyslipidemia | 14 (24.6%) | 6 (20.7%) | 8 (28.6%) | .701 | 57 |
| Sleep unit referral reasons | |||||
| Insomnia | 3 (5.2%) | 2 (6.9%) | 1 (3.4%) | 1.00 | 58 |
| Somnolence | 22 (37.9%) | 9 (31.0%) | 13 (44.8%) | .417 | 58 |
| Chronic snoring | 46 (79.3%) | 24 (82.8%) | 22 (75.9%) | .746 | 58 |
| Parasomnia | 1 (1.7%) | 1 (3.4%) | 0 (0.0%) | 1.00 | 58 |
| Observed apneas | 26 (44.8%) | 7 (24.1%) | 19 (65.5%) | .004 | 58 |
| Treatment | |||||
| Prescribed antihypertensive drugs, at least 1 | 23 (39.7%) | 12 (41.4%) | 11 (37.9%) | 1.00 | 58 |
| ACE inhibitors | 12 (20.7%) | 7 (24.1%) | 5 (17.2%) | .746 | 58 |
| β-Blockers | 6 (10.3%) | 2 (6.9%) | 4 (13.8%) | .670 | 58 |
| Diuretics | 8 (13.8%) | 5 (17.2%) | 3 (10.3%) | .706 | 58 |
| Calcium channel blockers | 8 (13.8%) | 5 (17.2%) | 3 (10.3%) | .706 | 58 |
| Angiotensin receptor blockers | 7 (12.1%) | 3 (10.3%) | 4 (13.8%) | 1.00 | 58 |
| Insulin | 1 (1.7%) | 0 (0.0%) | 1 (3.4%) | 1.00 | 58 |
| Lipid-lowering agents | 7 (12.1%) | 4 (13.8%) | 3 (10.3%) | 1.00 | 58 |
| Sleep parameters | |||||
| Apnea-hypopnea index, events/h | 23.0 [10.3; 50.7] | 10.1 [5.80; 13.1] | 50.8 [41.0; 60.6] | <.001 | 58 |
| Mean oxygen saturation during sleep, % | 93.0 [92.0; 94.0] | 93.0 [92.0; 95.0] | 93.0 [92.0; 94.0] | .179 | 58 |
| Minimum oxygen saturation during sleep, % | 85.0 [78.0; 88.0] | 88.0 [84.0; 90.0] | 80.0 [75.0; 85.0] | <.001 | 58 |
| Percentage of TST with oxygen saturation <90% | 2.30 [0.11; 7.71] | 0.12 [0.00; 1.35] | 4.10 [2.38; 14.1] | <.001 | 57 |
| Respiratory arousal index, events/h of sleep | 12.4 [6.32; 35.2] | 6.28 [3.41; 9.35] | 35.2 [23.8; 53.8] | <.001 | 55 |
| Epworth Sleepiness Scale score | 10.5 [8.00; 14.0] | 9.00 [7.50; 14.0] | 11.0 [8.00; 13.0] | .824 | 56 |
| ABPM parameters | |||||
| 24-h systolic blood pressure, mmHg | 136 [123; 144] | 138 [128; 147] | 133 [122; 144] | .455 | 58 |
| 24-h diastolic blood pressure, mmHg | 82.0 [79.3; 86.9] | 81.2 [79.5; 86.9] | 82.2 [77.4; 86.8] | .586 | 58 |
| Daytime systolic blood pressure, mmHg | 136 [126; 148] | 141 [130; 149] | 133 [126; 146] | .446 | 58 |
| Daytime diastolic blood pressure, mmHg | 84.8 [80.7; 88.4] | 84.8 [81.7; 89.5] | 84.6 [79.0; 88.0] | .323 | 58 |
| Nighttime systolic blood pressure, mmHg | 122 [114; 139] | 122 [117; 142] | 122 [112; 138] | .363 | 58 |
| Nighttime diastolic blood pressure, mmHg | 76.0 [73.0; 82.2] | 75.6 [73.5; 83.8] | 76.1 [73.0; 78.8] | .576 | 58 |
| Dipping ratio | 0.91 [0.87; 0.95] | 0.89 [0.85; 0.96] | 0.91 [0.89; 0.94] | .479 | 58 |
| Dipping pattern | .428 | 58 | |||
| Dipper | 26 (44.8%) | 15 (51.7%) | 11 (37.9%) | ||
| Nondipper | 32 (55.2%) | 14 (48.3%) | 18 (62.1%) | ||
Variables are expressed as number (percentage) for categorical variables or as median [Q1; Q3] for continuous variables.
ABPM, ambulatory blood pressure monitoring; BMI, body mass index; OSA, obstructive sleep apnea; TST, total sleep time.
No differences between groups were observed in sociodemographic characteristics, comorbidities, or BP parameters. Patients with OSA were more often referred to the sleep unit because of observed apneas and were characterized by severe AHI, with a median of 50.8eventsh−1.
Proteomic profilingQuality controlAll samples passed quality control based on intersample relationships in global protein expression (Supplementary Fig. 1A) and protein-level distribution across samples (Supplementary Fig. 1B). Eighteen proteins were discarded, as listed in Supplementary Table 1, because of low expression in the study population, with >25% of their quantification values below the limit of detection. A total of 348 unique proteins remained in the analysis.
Differentially abundant proteins between controls and patients with severe OSAUnivariate protein analysis identified 26 proteins with significant and relevant differences between the severe OSA and control groups, of which 17 were upregulated and 9 were downregulated in patients with OSA (Fig. 2A and Supplementary Table 2). The relationships between these proteins and different nocturnal BP parameters are shown in Fig. 2B. The strongest observed correlations were between FGF23 and diastolic dipping ratio (ρ=0.47) and between CDH5, a cadherin specifically found in vascular endothelium, and nocturnal systolic BP (ρ=−0.4). Proteins that met the significance and correlation criteria described in the “Methods” section were considered differentially abundant proteins (DAPs) (n=21). Of these proteins, 43% were associated with cardiovascular functions, 38% with organ damage, and 19% with cardiometabolic processes (Fig. 2C). DAPs were significantly associated with several GO terms (Fig. 2D), with the strongest enrichments observed in proteins involved in anatomical structure morphogenesis and extracellular vesicles.
Differential plasma protein presence between controls and severe OSA patients in nocturnal hypertension patients. (A) Volcano plot showing labeled differential proteins between groups that met statistical and fold-change criteria (P value ≤0.05 and FC ≤0.87 or FC ≥1.15). (B) Spearman correlation matrices of levels of differentially present proteins in the severe OSA subgroup with nocturnal blood pressure (BP) parameters for each theme. (C) Voronoi diagram showing proteins that met statistical and fold-change criteria, as well as correlation (|ρ|>0.2) to nocturnal BP colored according to associated thematic. (D) Dotplots representing the top gene ontologies terms that differential proteins correlated nocturnal BP were enriched for. GeneRatio represents the ratio of the number of proteins annotated to the specific pathway over the total number of proteins comprising the input protein list, Count represents the natural number of proteins annotated to the term and p.adjust represents a Benjamini–Hochberg adjusted P value. DBP, diastolic blood pressure; SBP, systolic blood pressure; DR, dipping ratio.
sPLS-DA identified a protein signature comprising 13 proteins, of which LILRB5, EPHB4, and CDH5 emerged as key contributors, with absolute correlation coefficients exceeding 0.3 (Fig. 3A, left). Lower levels of CDH5 and EPHB4 and higher levels of LILRB5 were associated with severe OSA. Moreover, the first component showed a quasilinear relationship with the in-sample risk of log odds of severe OSA (Fig. 3A, right). Functional interaction analysis predicted 7 interactions among these proteins (Fig. 3B), linking 6 proteins within a network consisting of HSPG2, ANG, CDH5, EPHB4, LYVE1, and LILRB5. Notably, two of the predicted interactions, EPHB4:CDH5 and EPHB4:LYVE1, are supported by experimental evidence, and at least five others have been shown to coexpress. The proteins comprising this network were all shown to be regularly expressed by potential tissues of interest, such as cardiac and arterial tissue (Fig. 3C). Single-cell data for the signature proteins in cardiac tissues showed expression across a variety of cell types, with higher and more homogeneous expression in endothelial cell types (Fig. 3D).
Severe OSA-associated protein signature identified in nocturnal hypertension patients. (A) Protein signature selected by sparse partial least squares discriminant analysis (sPLS-DA). The left figure shows a forest plot listing all the proteins composing the signature with their individual contributions. The right figure shows the modeled in-sample log odds risk of severe OSA according to the presence of the identified protein signature (first component). (B) Interaction graph representing results of functional interaction predicted with the STRING database (protein–protein interaction enrichment score P-value: 1.48e−05). Colored halo gradient represents base 2 logarithm of the fold change (red more negative, blue more positive) of the protein in the severe OSA group. (C) Heatmap showing the tissue expression patterns of signature proteins in potential tissues of interest (inferred through an external database with expression data of healthy adult subjects). The color gradient represents relative expression of the genes encoding those proteins in transcripts per million (TPM) and proteins are clustered based on similarities in their expression patterns. (D) Ballon plot representing single cell expression data of our protein signature in heart tissue. Size represents the percentage of expression in the specific cell type whereas color represents a measure of the homogeneity in the expression of that protein in the specific cell type through base 10 logarithm of count per 10,000 (log CP10K).
Proteomic profiling analysis was reassessed after stratifying the population by antihypertensive prescription to evaluate the potential confounding effect of pharmacological antihypertensive treatment on the circulating proteome. Briefly, the untreated and treated groups included 35 and 23 individuals, respectively. Their characteristics are summarized in Supplementary Tables 3 and 4. A smaller set of DAPs (n=6) was identified between the severe OSA and control groups in the treated population (Supplementary Fig. 2 and Supplementary Table 5), with no retrievable signature associated with OSA. In contrast, the DAP set identified in the antihypertensive-naive population comprised 47 proteins, enriched in cell adhesion-related processes and extracellular vesicle components, which had also emerged in the global analysis, as well as immune system processes (Supplementary Fig. 3 and Supplementary Table 6). In addition, a signature associated with severe OSA consisting of 26 proteins was identified in this population. This signature showed a linear relationship with the in-sample log odds of OSA risk and contained several proteins predicted to interact with each other (Supplementary Fig. 4).
OSA-related extracellular vesicle effects on endothelial cellsThe characteristics of patients selected for extracellular vesicle isolation are shown in Supplementary Table 7. This population consisted entirely of men, and no significant differences were observed in sociodemographic characteristics or BP parameters between the control and severe OSA groups. Structural and molecular characterization (Supplementary Fig. 5) corroborated proper isolation technique.
A trend toward lower CDH5 expression was observed in HAoECs treated with extracellular vesicles from the OSA group compared with the control group (Fig. 4A). A similar trend suggestive of increased endothelial permeability was observed in HAoECs treated with OSA extracellular vesicles (Fig. 4B). Consistent with these findings, qualitative (Fig. 4C, left) and quantitative (Fig. 4C, right) assessments of endothelial monolayer integrity using VE-cadherin visualization suggested greater barrier disruption in HAoECs treated with OSA extracellular vesicles.
In vitro assays results. The panel represents results obtained from a battery of in vitro assays performed in human aortic endothelial cells treated with extracellular vesicles extracted from patients of untreated, uncontrolled nocturnal hypertension with severe OSA or controls. The number displayed between boxplots represents P value for Wilcoxon rank-sum tests. (A) Boxplot representing negative delta cycle threshold (Ct) distributions for the gene encoding VE-cadherin in cells treated with extracellular vesicles (n=16). (B) Boxplot representation of an indirect measure of endothelial permeability through quantification of fluorescent dextran passage (n=16). (C) On the left, a staining of VE-cadherin (green) and cell nuclei (blue) in cells treated with extracellular vesicles where the red arrows point towards discontinuities observed in the endothelial monolayer (representative of a total of 40 images). On the right, a boxplot representing the mean area left exposed by the discontinuities in the endothelial monolayer of cells treated with extracellular vesicles (n=8) calculated from immunofluorescence images. Each dot represents the mean area left by the discontinuities across the endothelial monolayers of cells treated with extracellular vesicles from individual patients of both groups. Effect sizes reported in each boxplot display as a Cohen d coefficient.
This study describes an initial exploration of potential cardiovascular-altering mechanisms associated with OSA in patients with NH via a combined approach of targeted proteomics and in vitro mechanistic studies. In a cohort of 58 individuals matched for age, sex, BMI and nocturnal MAP, 21 proteins related to nocturnal BP were present at different levels in the severe OSA population compared to the control group. In addition, a molecular signature of OSA comprising 13 proteins showing high interconnectivity was identified. Both sets contained proteins present in cardiac and arterial tissues that are regularly expressed in immune and endothelial cell types.
Several identified DAPs and signature proteins have been linked to cell adhesion, a crucial process for maintaining vascular integrity. Notably, CDH5, essential for endothelial junctions [30], was downregulated and negatively correlated with nocturnal systolic BP in OSA patients. This, along with the upregulation of YES1, which phosphorylates CDH5 to promote vascular permeability [31], supports the hypothesis that OSA may increase vascular permeability. Moreover, the interaction of CDH5 with EPHB4 highlights a potentially synergistic role in vascular remodeling or permeability (see potential relationships discussed in Supplementary Table 8). Additionally, ICAM2, a protein whose loss leads to increased vascular permeability and abnormal blood vessel formation [32,33], was also downregulated. These changes suggest that OSA may be associated with molecular alterations potentially leading to endothelial dysfunction, a precursor to hypertension and CVD.
Consistent with the literature regarding OSA and systemic inflammation [34], enrichment of immune and inflammatory proteins was observed. Higher levels of CCL3, a chemokine associated with atherosclerosis and increased cardiovascular risk [35,36], were observed in the OSA group. A heightened inflammatory profile and an increase in vascular permeability could possibly explain the increased levels of subclinical atherosclerosis markers observed in OSA patients [37]. These differences were more pronounced in OSA individuals with no antihypertensive prescription, suggesting a heightened cardiovascular inflammatory burden associated with greater cardiovascular risk, particularly relevant in the context of masked hypertension [38].
Subgroup analysis revealed distinctions based on antihypertensive treatment status. The enriched pathways in the antihypertensive-naïve population mirrored those observed in the full cohort, encompassing cardiovascular function, immune response, and tissue damage. In contrast, the lack of identification of a proteomic signature in treated hypertensive patients may reflect a medication-driven modulation of key biological pathways typically altered by OSA. Antihypertensive agents may attenuate OSA-induced molecular changes, thereby masking differential protein expression. This observation is reinforced by the identification of a broad and rich DAP signature in the untreated patient group. It suggests that antihypertensive treatments, beyond their hemodynamic effects, may attenuate the biological impact of OSA on vascular and inflammatory tissues.
This study examined extracellular vesicles as potential mediators of OSA-mediated endothelial injury, which several DAPs were associated with. The functional assays showed trends suggestive of increased endothelial permeability and barrier disruption in cells exposed to severe OSA vesicles, reinforcing the notion that they may drive OSA-related endothelial injury. An illustration depicting this hypothetical mechanistic pathway is shown in Fig. 5.
This study has several strengths: (1) a well-characterized cohort using PSG and 24-hour ABPM, the gold standards for OSA and NH diagnosis; (2) a robust proteomic analysis methodology from fasting plasma samples obtained immediately after the sleep study via the Olink Proteomics platform [39]; and (3) the integration of an in vitro approach to explore the effects of OSA on endothelial cell types. However, several limitations also exist in this study: (1) corrections for multiple comparisons in our statistical testing for protein level comparison analysis were not applied. However, this study was intended as an initial exploration of plausible OSA mechanisms associated with vascular damage, mechanisms which we explored in depth in an in vitro endothelial model; (2) a limited sample size for the in vitro studies which limited statistical power and the inclusion of females. These findings should therefore be interpreted with caution, and require validation in larger, more controlled samples; (3) while the exclusion of moderate OSA from the study was intended to enhance contrast and likelihood of detecting effects, it limits generalizability of the findings and precludes dose–response analysis. Future studies with more diverse samples are warranted to determine whether these effects extend to moderate OSA; (4) there is a risk the control group did not represent a NH population completely free of sleep disorders. Given that all patients had been referred to the sleep unit and that somnolence was equally present in both groups, the potential for partial phenotype overlap cannot be excluded. This may have hindered our ability to detect sleep disturbance-related proteomic differences, potentially accounting for the modest differences observed between groups; and (5) although the study population represents a clinically relevant sample of the NH population, in whom pharmacological burden is prominent, treatment may have significantly altered proteomic profiles. This possibility is supported by the secondary analyses, in which antihypertensive prescription appeared to attenuate OSA-related signal.
Taken together, the available evidence provides context for interpreting the relationship between OSA and cardiovascular risk. Although the three major clinical trials showed a neutral effect of continuous positive airway pressure (CPAP) on cardiovascular risk reduction [40–42], it is important to highlight that treatment adherence was low across all of them. Notably, an on-treatment analysis based on a meta-analysis of these trials demonstrated that adequate adherence to CPAP was associated with a reduction in cardiovascular event recurrence [43]. These findings are consistent with our results, which may aid ongoing efforts to clarify the mechanisms through which OSA contributes to elevated cardiovascular risk.
Furthermore, a post hoc analysis of the ISAACC trial showed that, within a specific patient phenotype without prior cardiovascular events, individuals with OSA had a significantly higher cardiovascular risk compared to non-OSA controls [44]. In our analysis, the mechanisms underlying the increased cardiovascular risk associated with OSA appeared to be more pronounced when comparing patients at lower baseline cardiovascular risk. Although these findings may suggest that the effect of OSA is more evident in patients at earlier stages of cardiovascular disease, the results should be interpreted with caution given the study's multiple limitations.
Our findings are consistent with a possible association between OSA and endothelial dysfunction and systemic inflammation in NH patients, particularly in patients whose hypertension is untreated.
Declaration of generative AI and AI-assisted technologies in the writing processNo material for this article was generated using artificial intelligence.
FundingThis research was funded by the Instituto de Salud Carlos III (ISCIII), grant number PI21/00337, co-funded by the European Union; FUCAP Grant Call 2022, Maria Grant, sponsored by the Naccari Ravà Foundation; and the Sociedad Española de Neumología y Cirugía Torácica (SEPAR), grant number 1209/2022.
ADST received financial support from the Instituto de Salud Carlos III through Miguel Servet 2023, grant number CP23/00095, co-funded by the European Union.
IJ-G was supported by the Instituto de Salud Carlos III (ISCIII), PFIS 2024, grant number FI24/00084, co-funded by the European Union.
Conflicts of interestNone declared.
The following are the supplementary data to this article:
