To the best of our knowledge, this is the first study to investigate the OSA severity-associated lipid profile in the CSF of patients with AD. We detected 11 CSF lipids that were differentially expressed between AD patients with and without severe OSA, regardless of the incorporation of confounding factors. This lipidomic signature was significantly correlated with different polysomnographic measures related to OSA. In addition, the discriminating power of this lipid signature in separating AD patients with severe OSA from AD patients with nonsevere OSA was much greater than the use of the STOP-Bang screening questionnaire alone. Importantly, the identified lipids possibly indicate the harmful effects produced by severe OSA at the brain level.
AD is a neurodegenerative disease usually accompanied by several comorbidities, including hypertension, diabetes, dyslipidaemia, depression, and sleep disorders. These comorbid conditions have been shown to be related to an increased risk of AD and its progression (5). OSA, which is the most common sleep disorder, has a high prevalence (45–90%) in patients with AD (6) and has been shown to be associated with increased AD pathology (8, 9) and disease incidence (33). In addition, OSA is a risk factor for several AD-related comorbidities, and its underdiagnosis may be associated with greater difficulties in the control of blood pressure, increased insulin resistance, metabolic syndrome, or obesity, all factors that worsen the cognitive evolution of patients (34, 35). The difficulties in performing PSG among patients with cognitive impairment, the high costs of the procedure, and the limited usefulness of the existing screening tests (17) in this population make it urgent to identify new biomarkers for the detection of severe forms of OSA.
The brain is one of the organs with a higher rate of oxygen consumption and is highly vulnerable to hypoxia. Acute hypoxia increases the formation of reactive oxygen species (ROS) in the brain (36). This organ is mainly composed of lipids, and hypoxia-induced ROS may increase lipoxidation and interfere with the proper functioning of these biomolecules (37). Lipoxidation is well documented in both the brain and blood of AD patients (19, 20). In addition, the existence of a link between OSA and lipid dyshomeostasis has been supported by several observational and meta-analyses studies (18, 38).
Lipidomics has been used to study blood (39) and CSF (40) lipid alterations as biomarkers for AD diagnosis and differential diagnosis from other types of dementia. Metabolomics, another discipline that detects all types of metabolites, including lipids, has revealed that lipids are the most dysregulated class of metabolites in OSA (38). Previous studies have detected phospholipids (41, 42), sphingolipids (43), and endocannabinoids (44) as biomarkers of OSA in cognitively unimpaired populations. These previous works have mainly used blood and urine samples for the detection of OSA biomarkers with the aim of detecting systemic lipid abnormalities caused by OSA. The search for OSA-related lipid alterations at the systemic level in patients with AD is scarce, and the only study conducted in this population showed lipid alterations in phospholipids and triglycerides with higher levels of oxidized lipid species in AD patients with severe OSA (45). However, AD principally affects brain physiology and function. Therefore, when searching for OSA-associated lipid alterations in AD patients, CSF would be a valuable sample because of its proximity to the brain tissue, and probably any OSA-provoked lipid alteration specific to AD would be captured in CSF better than any other biological fluid. In addition, CSF is collected in clinical settings for the measurement of AD core biomarkers needed for the diagnosis of AD, and the same sample can be used for lipidomics without increasing additional risk for the patients. As a result, we searched for a specific fingerprint of OSA in the CSF of AD patients. From 11 differentially expressed lipids between AD patients with and without severe OSA, we identified two: OxCer(40:6) and OxTG(57:2). The fact that both identified lipids were oxidized suggests that the increase in lipoxidation may be an important causative effect by which OSA exacerbates AD pathology or increases AD incidence, as reported previously (8, 46–48). In addition, having high discriminating power in separating AD patients with different OSA severities highlights the promising future of these biomolecules as biomarkers of OSA.
Ceramides are structural constituents of biological membranes and are also involved, as bioactive molecules, in a variety of biological events, including cell differentiation and proliferation, redox metabolism, inflammation, and apoptosis (49). Ceramide regulates the BACE1-mediated processing of amyloid precursor protein (APP), probably by the formation of ceramide-enriched platforms and the enhancement of BACE1 stability in cells. Inhibition of sphingomyelinase (SMase), the enzyme that mediates the conversion of sphingomyelin (SM) to ceramide, inhibits γ-secretase activity and leads to a reduction in Aβ secretion (50, 51). We found that higher levels of OxCer(40:6) in CSF were associated with the presence of severe OSA in AD patients. It has been suggested that the lipoxidation state is important for APP processing. When there is a high concentration of oxidized lipids, APP processing may shift from a nonamyloidogenic to an amyloidogenic pathway (50). Therefore, severe forms of OSA, by inducing lipoxidation of membrane-associated lipids, such as Cer(40:6), may increase AD pathology. In addition, oxidation may alter other functions of ceramides related to inflammation, redox homeostasis, and cell death that may play a role in AD pathogenesis.
Our results also revealed lower levels of TGs in CSF, specifically OxTG(57:2), in AD patients with OSA severity. Importantly, this observation remained significant after adjustment for medication use. Previous studies have reported significantly decreased serum TG levels in AD patients compared to a control group (52–54), which are, in addition, associated with early-AD biomarkers, including entorhinal cortical thickness and hippocampal volume measured by MRI scans, regions especially affected by AD. Our finding extends these previous observations from serum to CSF and suggests that impairments in TG metabolism among AD patients differ based on OSA severity. It remains unclear whether OSA severity also impairs the deterioration of the cortical parameters associated with AD. Inside cells, TGs are found in organelles called lipid droplets (LDs). In the brain, LDs have been found both in neurons and in glial cells (55). In the AD brain, it seems that the accumulation of LDs is more pronounced in the glia. Oxidative stress increases the number of LDs in glial cells, especially in astrocytes. Oxidized lipids produced in neurons are transported to surrounding astrocytes for detoxification and storage in LDs (56). TGs stored in LDs can serve as a source of energy and cell signalling molecules (57). In this context, it is proposed that the low TG levels in the CSF of AD patients with severe OSA express bioenergetic exhaustion of neural cells, with effects in oxidative stress conditions, which can favour cell death and the subsequent deterioration of brain functions and progression of AD. However, the specific mechanism(s) underlying the association of decreased TGs and AD and its impairment by OSA remain to be determined.
This untargeted approach is the initial step in this field of research and presents new knowledge from which new hypotheses can be generated. We believe that the validity of our findings should be determined by a targeted approach in an independent sample of AD subjects. From 11 differentially expressed lipids between groups, we identified two lipid species. Future advances and accessibility to finer equipment will surely help to identify more lipid species related to OSA severity in AD. We included patients with MMSE scores > 20, so the results should be interpreted with caution when extrapolating to patients at more advanced stages of the disease. The main strength of this study was the use of PSG for the diagnosis of OSA in our population. This allowed us to perform correlation analysis to examine the association between different PSG variables and the severity of OSA. In addition, our study population was well defined, and there was no significant difference regarding many sociodemographic characteristics and comorbidities that potentially would affect the results between groups. Moreover, we included a relatively large sample of consecutive AD subjects, which increased the generalizability of the data.