In this investigation, we demonstrated plasma metabolic profiling for young and middle-aged male patients with OSA using an integrated method of GC–MS and LC–MS untargeted metabolomics, as well as PSG and ROS detection. The findings revealed that 59 metabolites from GC–MS and 27 metabolites from LC–MS were significantly altered in sOSA patients when compared to mOSA. Extensive pathways were disrupted distinctly, with the only one in common being retrograde endocannabinoid signaling from both forms of metabolomics. Other enriched pathways included the FoxO signaling pathway and its upstream PI3K-Akt signaling pathway, different amino acid and nitrogen metabolism, lipid metabolism and TCA, particular synaptic function and neurodegeneration, renal tubular reabsorption, pyrimidine metabolism, and so on. Although there is no significant difference in ROS between mOSA and sOSA, the statistical significance of ROS between high and low fragmentation subgroups (N3 threshold: 16 min, arousal index: 30/h) was clearly established.

Various fatty acids and amino acids, glycosides or nucleosides, as well as small molecules of phosphates, intermediary metabolites in energy metabolism, and arachidonic acids, catecholamines, were revealed, as shown in previous OSA studies compared with the healthy controls [7,8,9,10,11], but fewer in sOSA patients comparing with the mOSA. In a mice model, H2S alleviates CIH-induced myocardial damage through PI3K/AKT/mTOR pathway, inhibiting oxidative stress and enhancing autophagy [13]. In another IH model in vitro and in vivo, GLP-1 analogue liraglutide’s protective mechanisms in combating cognitive deficits associated with CIH was via the Nrf2/HO-1 and MAPK/NF-κB pathways [14].The PI3K/Akt signaling system is essential for cell growth, apoptosis, translation, and metabolism. Activated Akt can directly phosphorylate a number of downstream effectors, inhibiting or activating their actions; one significant target is the FoxO family [15, 16]. The FoxO pathway regulates the cell cycle, apoptosis, oxidative stress resistance, and metabolism, and it is thought to play a significant role, primarily in response to stress circumstances rather than as a critical mediator of normal physiology [17]. An interesting finding in basic sleep research is that FOXO and the target of rapamycin (TOR) occupy opposing and agonistic positions within PI3K/Akt pathway. During the early stages of sleep, the insulin-like growth factor-1 (IGF-1) signaling pathway is activated. This activation occurs due to the secretory peaks of growth hormone (GH) and the downregulation of IGF-1 binding protein-1 (IGFBP-1), which in turn triggers the mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/ERK) and PI3K/Akt signaling cascades. Phosphorylated Akt activates TOR, which promotes protein synthesis and growth processes, acting as a catalyst for growth and aging [18]. In the later stages of sleep, the IGF-1 signaling axis is downregulated. This process is mediated by declining GH levels, the somatostatin—and corticosteroid-induced upregulation of IGFBP-1. The absence of IGF-1 signaling inhibits PI3K/Akt activity, leading to the activation and nuclear translocation of FoxO. Once in the nucleus, FoxO regulates various cellular functions, including stress resistance and energy metabolism[T18]. Based on our study findings, the FoxO and PI3K/Akt pathways likely play a role in the pathophysiology of OSA.

The endocannabinoid system, which includes cannabinoid receptors, endogenous ligands, and enzymes that synthesize, degrade, and transport endocannabinoids, is found throughout the central and peripheral nervous systems, the endocrine system, the gastrointestinal tract, and inflammatory cells [19, 20]. It acts as a key modulator, regulating neurochemical systems, autonomic function, the circadian sleep–wake cycle, anxiety, and mood [19, 21, 22]. Preliminary research revealed its significance in OSA and insomina patients, as well as related models [21,22,23]. The best-studied endocannabinoids, 2-arachidonoyl-glycerol (2-AG) and anandamide, have a major impact on a variety of biochemical and physiological processes through cannabinoid receptors CB1R and CB2R. These mechanisms include thermoregulation, HPA axis activation, metabolism, memory consolidation, inflammation, and reward [21].

Previous research has found that endocannabinoids are increased in OSA patients compared to healthy persons [7]. The expression of enzymes involved in endocannabinoid production and breakdown, such as diacylglycerol lipase (DAGL) and monoacylglycerol lipase (MAGL), may change or enhance endogenous cannabinoid levels [24]. Furthermore, MAGL levels showed a moderate correlation with ArousaI, reflecting the severity of sleep fragmentation [25]. Wang and colleagues demonstrated that AEA and 2-AG levels increased in sOSA patients compared to the mild or moderate OSA category, with AEA linked with AHI and the homeostasis model of assessment for insulin resistance index (HOMA-IR) [26]. Furthermore, the principal conclusion of our investigation confirmed the participation of endocannabinoids in signaling pathways, while more widespread impacts on brain function could not be ruled out. Similarly, certain synaptic activities and neurodegenerative pathways, as shown in pathway enrichment by GC–MS, may have a latent influence on sleep-related neurologic disorders. A few studies concentrating on the administration of exogenous cannabinoids had ever demonstrated certain effects on sleep characteristics or AHI. However, due to the limitations in stability and reliability, AASM does not currently recommend cannabinoids as a treatment for OSA [21, 23].

Patients with OSA commonly undergo periods of intermittent hypoxia, similar to ischemia/reperfusion [27, 28]. This situation, characterized by repetitive cycles of oxygen deprivation and restoration, has a major impact on the development and progression of OSA and associated comorbidities, particularly cardiovascular illnesses and metabolic syndrome, owing to increased ROS production [27,28,29,30,31]. ROS, the most common result of oxidative stress, are transient intermediates in both normal and abnormal metabolism. And its production is a complicated and multifaceted process. ROS generation is influenced by a variety of parameters, including oxidative stress, mitochondrial function, nutritional and antioxidant status, age or life stage, and epigenetic changes [32]. Intrinsic changes to the electron transport chain (ETC), such as deficiencies in component assembly, mutations in ETC subunits, and ETC inhibition, can also increase ROS production [33]. Differences in redox disruption appeared to be related to both aging and physiologic exercise [34]. Mitochondrial uncoupling of oxidative phosphorylation, or proton leak, reduces the electrochemical proton gradient (Δp) and accounts for up to 25% of basal metabolism in the presence of oxidative stress, such as diabetes, drug resistance in tumor cells, ischemia–reperfusion injury, or aging. Variations in mitochondrial uncoupling proteins (UCP) expression may influence the effectiveness of ROS generation in the cardiovascular system [35]. Notably, OSA may drive ROS production through multiple distinct mechanisms beyond classical oxidative stress pathways. First, recurrent oxygen desaturation events induce sleep fragmentation and chronic sleep deprivation, which disrupt redox homeostasis via circadian rhythm dysregulation. Second, OSA-associated systemic inflammation promotes ROS generation through proinflammatory cytokine-mediated activation of NADPH oxidases (NOX), particularly in vascular endothelium. Furthermore, metabolic perturbations secondary to IH—including altered mitochondrial respiration and sleep deprivation-induced lipolysis—create a pro-oxidant microenvironment that perpetuates ROS overproduction [36]. Our findings show that sleep fragmentation can cause noticeable changes in ROS levels between subgroups of patients, regardless of AHI or ODI. Young and middle-aged flies have been shown to have preserved neuroprotective mechanisms in response to sleep fragmentation. In cases of poor sleep quality, elevated neuronal ROS levels may promote neuronal insulin signaling by blocking cytochrome C release, a mechanism that protects neurons from death caused by unfolded protein responses (UPR) [37]. This finding shows that, even in the absence of severe AHI or ODI, sleep fragmentation may play an unignorable role in ROS formation.

In our study sOSA group showed a significantly higher BMI compared to the mOSA, which displayed the robust link between obesity and OSA. Epidemiologically, 70% of OSA patients are obese, and higher BMI correlates with more severe OSA, particularly in males and younger populations [3, 36], consistent with our study. Cervical fat accumulation narrows the oropharynx and diminishes upper airway traction, while obese men with OSA exhibit less weight loss after a one-year dietary and exercise intervention compared to similarly obese men without OSA. Emerging evidence highlights intermediate mechanisms—including oxidative stress, endothelial dysfunction, insulin resistance, and inflammation—as key contributors to the concurrence of both [3, 36]. In a longitudinal cohort study of 544 participants, OSA was identified as an independent risk factor for developing obesity-associated T2DM, with a significant positive correlation between AHI and BMI. This association remained statistically significant after adjusting for confounding variables (age, sex, BMI, race, etc.) [36]. Notably, comorbidities such as heart failure, stroke, and COPD could not be fully attributed to BMI or obesity [3], which further highlights OSA's role as an independent risk factor for systemic disorders. However, the bidirectional and synergistic interplay between obesity and OSA renders the efforts to differentiate their independent impacts on complications, particularly in cohort with limited sample sizes. We must acknowledge our current study's limitations. The sample size is small, and all of the cases are male and of Chinese Han ethnicity. The possibility of sample bias and individual differences may have an impact on its generalizability.