Complement protein levels in serum astrocyte-derived exosomes are associated with cognitive impairment in obstructive sleep apnea

Abstract

Study Objectives:

An association between neuroinflammation and cognitive decline has been established. The complement system regulates neuroinflammation. Dysregulation, impairment, or inadvertent activation of complement components contribute to preclinical Alzheimer’s disease. The astrocyte-derived exosome (ADE) complement proteins, including C3b and C5b-9, may be predictive biomarkers of mild cognitive impairment conversion to Alzheimer’s disease dementia. We hypothesized that complement proteins might be involved in cognitive impairment during obstructive sleep apnea (OSA). The aim of our study was to explore the correlation between the complement system and mild cognitive impairment (MCI) in patients with OSA.

Methods:

All participants with subjective snoring complaints from the Sleep Medicine Center underwent polysomnography. OSA was defined as apnea-hypopnea index ≥ 5 events/h. MCI was defined as the Montreal Cognitive Assessment < 26 and met the criteria: (1) a subjective cognitive impairment; (2) an objective impairment in 1 or more cognitive domains; (3) complex instrumental daily abilities can be slightly impaired but independent daily living abilities are maintained; and (4) no dementia. The ADEs were isolated immunochemically for enzyme-linked immunosorbent assay quantification of complement proteins, including C3b, C5b-9, and CD55. The participants who received continuous positive airway pressure were followed up and their complement protein levels were reassessed after 1 year of treatment.

Results:

A total of 212 participants (66.98% males; mean age of 56.71 ± 10.10 years) were divided into the OSA+MCI group (n = 90), OSA-MCI group (n = 79), and controls (normal cognitive state without OSA) (n = 43). The ADE levels of C3b and C5b-9 in the OSA+MCI group were higher than those in the OSA-MCI and control groups. The C3b and C5b-9 were independently associated with cognitive impairment in patients with OSA. The relationship between apnea-hypopnea index and Montreal Cognitive Assessment scores was mediated by C3b and C5b-9. We found no linear correlation between the complement proteins and the severity of OSA. The complement proteins were negatively correlated with global cognitive performance and cognitive subdomains. The complement protein levels significantly decreased after continuous positive airway pressure treatment.

Conclusions:

Complement proteins were implicated in cognitive impairment in patients with OSA and may be promising biomarkers for predicting cognitive impairment in patients with OSA.

Clinical Trial Registration: Registry: Chinese Clinical Trial Registry; Name: Study on early diagnostic markers in patients with dementia and mild cognitive impairment; URL: https://www.chictr.org.cn/; Identifier: ChiCTR1900021544.

Citation:

Li M, Sun C, Xue S, et al. Complement proteins levels in serum astrocyte-derived exosomes are associated with cognitive impairment in obstructive sleep apnea. J Clin Sleep Med. 2023;19(4):727–739.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Mounting evidence has verified the link between obstructive sleep apnea (OSA) and cognitive impairment, but very little information on the underlying mechanisms is available. The aim of our study was to investigate the correlation between the complement system and cognitive impairment in patients with OSA.

Study Impact: Our study suggests that complement protein contributes to cognitive impairment in patients with OSA. The relationship between the severity of OSA and cognitive performance was mediated by C3b and C5b-9. Complement proteins can be used as markers of cognitive impairment in patients with OSA.

INTRODUCTION

Obstructive sleep apnea (OSA), a common form of chronic sleep-disordered breathing, is characterized by recurrent, complete, or partial upper airway obstruction during sleep leading to intermittent hypoxia and sleep fragmentation,¹ which increases the risk of cognitive decline in older individuals.² Increased accumulation of β-amyloid (Aβ) and hyperphosphorylation of tau protein in the brain, episodic hypoxemia, oxidative stress, vascular inflammation, and systemic comorbidities may contribute to the pathogenesis.³

The complement system, as an effector of innate and adaptive humoral immunity, comprises soluble membrane-bound proteins produced by both neurons and glial cells. Complement pathway begins with the enzyme cascade, which proceeds via the classical, lectin, and alternative pathway. The soluble proteins in enzymatic cascade undergo conformational changes to generate proinflammatory anaphylatoxins (C3a and C5a), opsonins (C3b and C4b), and C5b (precursor to membrane attack complex formation), which initiates the second part of the system—the lytic pathway and osmolytic membrane attack complex (C5b-9, the terminal complement complex) formation.⁴ The complement system can enhance or assist in the role of antibodies or phagocytes to clear microbes or damaged cells, thus acting as a driver of immune-inflammatory processes.⁵ Overactivation or dysregulation of the complement system may promote excessive proinflammatory responses,⁶ mediating synapse loss and dysfunction early in Alzheimer’s disease (AD) and eventually leading to cognitive impairment.⁷ C3b is an important mediator in the process of antigen presentation and complement cascade amplification, and C5b-9 is implicated in the mechanisms of tissue injury during inflammation. CD55, as a membrane-bound complement inhibitor and a decay-accelerating factor, can facilitate the decay of C3 and C5 convertases to negate complement-mediated injury and maintain tissue homeostasis. Hypoxia can result in the expression of a proinflammatory state through the downregulation of CD55, and mediates microvascular injury.⁸ Astrocytes are capable of producing C3, C4, and the terminal pathway components and also synthesize regulatory proteins such as CD55, CD59, and membrane cofactor protein.⁶ Many studies indicate that endothelial cell inflammation, especially complement mediator levels in endothelial-derived plasma exosomes, are associated with white matter disease, brain degeneration, and cognitive changes.^8,9

Exosomes are membrane-derived microvesicles (30–150 nm) secreted by almost every cell type containing proteins, nucleic acids, and lipids, and are involved in the pathogenesis of many neuroinflammatory and neurodegenerative disorders.¹⁰ They play an essential role in brain homeostasis and the crosstalk between astrocytes with neurons, other glia, or brain endothelial cells. Under both physiological and pathological conditions, exosomes are released not only into the cerebrospinal fluid, but also into the blood.^11,12 Some studies have shown that complement protein levels in plasma astrocyte-derived exosomes (ADEs), as components of neurotoxic neuroinflammation, are closely related to cognition and brain damage and may be predictive biomarkers of mild cognitive impairment (MCI) conversion to AD.^13–15 Exosomal proteins released by activated astrocytes appear to reflect the neuroinflammatory processes in the brain.

In this investigation, we enrolled participants with subjective snoring complaints, measured complement proteins (including C3b, C5b-9, and CD55) from ADEs, assayed plasma neuronal exosomal levels of Aβ and tau proteins, and quantified white matter hyperintensity volume, gray matter volume, hippocampal volume, and cortical thickness to explore the associations between complement components and cognitive impairment in patients with OSA without dementia. The ADE levels of complement proteins in patients treated with continuous positive airway pressure (CPAP) were remeasured 1 year later.

METHODS

Baseline recruitment and study procedures

For the cross-sectional study, we recruited participants with a chief complaint of snoring from the Sleep Medicine Center of Weihai Municipal Hospital from June 2019 to June 2021. Healthy controls were recruited from the community. The study was approved by the ethics committee of the same hospital, indicating compliance with all ethical regulations. Informed consent was obtained from all participants before study enrollment. Participants aged 35–80 years with snoring, not undergoing any neuropsychological assessment in the past year and without history of major surgery, trauma, or inflammatory disease (eg, pneumonia, arthritis, etc.) in the last 3 months, were included in the study. No participants had been diagnosed with OSA or received treatment for snoring before participating in the study.

The exclusion criteria were the following past or current medical history: (1) major neurological disorders, such as dementia, stroke, intracranial infection, traumatic brain injury, neurodegenerative disease, epilepsy, brain tumor, brain stem lesion affecting respiratory rhythm, restless legs syndrome; (2) major psychological disorders, such as major depression and schizophrenia; (3) the use of drugs that might confuse cognition or sleep (eg, benzodiazepines, antihistamines, tricyclic antidepressants, and donepezil); (4) systemic diseases, such as severe chronic heart or lung disease, liver cirrhosis, uremia, thyroid dysfunction, and autoimmune disease, and pregnancy. A total of 271 participants conforming to the inclusion criteria were recruited, of which 6 patients refused to participate in the study and 41 patients were excluded according to the exclusion criteria.

All participants subsequently underwent standardized clinical and laboratory evaluations, magnetic resonance imaging and polysomnography (PSG). OSA was defined as an apnea-hypopnea index (AHI) ≥ 5 events/h. They completed a detailed questionnaire, including data on age, sex, body mass index, education, hypertension, diabetes mellitus, hyperlipidemia, smoking (in the past 3 months), and alcohol consumption (≥ 300 g/week). Hypertension was defined as a blood pressure ≥ 140/90 mm Hg in 3 different measurements or current antihypertensive treatment. Diabetes was defined as repeated fasting glucose levels ≥ 126 mg/dL, glucose loads ≥ 200 mg/dL 2 hours after oral glucose administration or use of antidiabetic medications. Hyperlipidemia was defined as total cholesterol levels ≥ 200 mg/dL, triglyceride levels ≥ 150 mg/dL, low-density lipoprotein cholesterol levels ≥ 130 mg/dL, or lipid-lowering therapy.

For the longitudinal study, the participants with moderate-to-severe OSA who received CPAP therapy were followed and we reassessed them after 1 year of CPAP therapy to measure complement protein levels. Inclusion criteria were as follows: (1) CPAP therapy adherence was good, using the device for at least 4 hours per night and at least 5 nights per week¹⁶; (2) CPAP treatment was effective (AHI < 5 events/h); and (3) the patient could adhere to CPAP treatment for at least 1 year.

Sleep data acquisition

All participants underwent overnight in-laboratory PSG (Somté or Grael, Compumedics, 30–40 Abbotsford, Victoria, Australia) between 2200 and 0600 with the standard montage recommended by the American Academy of Sleep Medicine Version 2.5.¹⁷ The PSG protocol included a simultaneous electroencephalography recording (F3-M2, F4-M1, C3-M2, C4-M1, T3-M2, T4-M1, O1-M2, and O2-M1), right and left electrooculogram, chin electromyography, microphone, naso-oral thermistor, nasal pressure transducer, anterior tibial electromyography, electrocardiography, body position, thoracic and abdominal respiratory movement registration, and pulse oximetry sensor. According to the American Academy of Sleep Medicine criteria, all sleep staging, arousal, respiratory events, and movement event scoring were visually analyzed by a trained technologist blinded to the other information.¹⁷ Sleep parameters were recorded, including sleep latency, total sleep time, sleep efficiency, proportion of sleep stages, AHI, oxygen desaturation index (ODI), mean oxygen saturation during sleep (mean SaO₂), arousal index, and percentage of sleep time with oxygen saturation < 90% (T90).

ODI is defined as number of oxygen desaturations ≥ 3%; an obstructive apnea is defined as a ≥ 90% reduction in the thermal sensor signal for ≥ 10 seconds with continued or increased thoracoabdominal effort; an obstructive hypopnea is defined as a decrease in the nasal pressure signal by ≥ 30% for ≥ 10 seconds, terminating in a ≥ 3% oxygen desaturation from the pre-event baseline or arousal, along with snoring during the event, an associated thoracoabdominal paradox during the event but not prior to the event, or increased inspiratory flattening of the nasal (pressure.¹⁷ OSA severity was classified as mild for an AHI ≥ 5 events/h to < 15 events/h, moderate for an AHI ≥ 15 events/h to < 30, and severe for an AHI ≥ 30 events/h.¹⁸

The participants with any of the following conditions were also excluded: (1) total sleep time of PSG was less than 4 hours; (2) PSG showed rapid eye movement sleep behavior disorder or rapid eye movement sleep without atonia; and (3) PSG showed a large amount of periodic limb movements (total periodic limb movements index > 10/h).

Neuropsychological and psychopathological evaluation

All participants completed the following standardized questionnaires after eating in the morning: the Mini-Mental State Examination, Montreal Cognitive Assessment (MoCA) Beijing version, Epworth Sleepiness Scale, Hamilton Anxiety Scale, and Hamilton Depression Scale (HAMD)-17. All neuropsychological tests were administered by a trained physician in a quiet room. The MoCA scale, including visuoexecutive function, naming, attention, language, abstraction, delayed recall, and orientation, was scored from 0 to 30, with lower scores indicating poorer cognitive function. MCI was defined as MoCA < 26¹⁹ and meeting the revised Mayo Clinic criteria²⁰: (1) a subjective cognitive impairment; (2) an objective impairment in 1 or more cognitive domains; (3) slight impairment of complex instrumental daily abilities, but independent daily living abilities maintained; and (4) not enough to diagnose dementia.

All participants were divided into 3 groups based on AHI and MoCA scores: the OSA+MCI group (AHI ≥ 5 events/h and MoCA score < 26); the OSA-MCI group (AHI ≥ 5 events/h and MoCA ≥ 26), and the control group (AHI < 5 events/h and MoCA ≥ 26). Twelve of 224 enrolled participants for the cross-sectional study were excluded from the study because of missing oxygen saturation data caused by failure of the pulse oximeter sensor (n = 2), rapid eye movement sleep behavior disorder (n = 3), periodic limb movements (n = 2), and cognitive impairment without OSA (n = 5). In total, 212 patients were analyzed (Figure 1).

Figure 1. Description of the study population.

MRI = magnetic resonance imaging, OSA = obstructive sleep apnea, PLM = periodic leg movement, PSG = polysomnogram, RBD = rapid eye movement sleep behavior disorder.

Magnetic resonance imaging data acquisition and processing

Brain scans were acquired on the same 3.0 T Skyra magnetic resonance scanner (Siemens, Erlangen, Germany). We collected T1-weighted and T2-weighted fluid-attenuated inversion recovery imaging sequences using a magnetization prepared rapid acquisition gradient-echo pulse sequence with the following set of parameters: T1 (repetition time = 2,000 ms, echo time = 2.98 ms, inversion time = 900 ms, flip angle = 9°, matrix size = 256 × 256, field of view = 240 × 256 mm, slice thickness = 1.0 mm, slices = 176) and fluid-attenuated inversion recovery (repetition time = 9,000 ms, echo time = 96 ms, inversion time = 2,500 ms, flip angle = 150°, field of view = 240 × 240 mm, and slice thickness = 5.0 mm). An experienced neuroradiologist visually checked all the scans to ensure that the participants were free of stroke, traumatic brain injury, and brain tumor.

White matter hyperintensity (WMH) was defined as the presence of hyperintensity in the white matter area. The volume of WMHs was calculated using the semiautomated freeware 3-D slicer (http://www.slicer.org). The WMH regions were manually parcellated in template images layer by layer and were used in automated WMH extraction analysis.

The magnetic resonance images were preprocessed using FreeSurfer (version 6.0) (https://surfer.nmr.mgh.harvard.edu/). The total gray matter volume, total hippocampal volume, and average cortical thickness in AD signature regions (eg, entorhinal, inferior temporal, middle temporal, and fusiform regions)²¹ were calculated with default analysis settings.²² Simply put, intensity normalization, skull strip, volumetric registration, cortical surface reconstruction, spherical registration, and anatomical labeling were performed by the automatic pipeline. An experienced neuroradiologist who was blind to the clinical data manually assessed all image processing steps, including initial skull striping, spatial registration, and subcortical segmentation, according to FreeSurfer guidelines.

Extraction and quantification of complement proteins

The morning after PSG, peripheral blood samples were collected from all participants by venipuncture of forearm veins following an overnight fast. A technician collected and stored all blood samples using the same methods and procedures. Samples were centrifuged at 4,000 g for 10 minutes within 30 minutes of collection, and then the serum was promptly stored at −80°C until they were assayed.

A total of 550 µl of serum was centrifuged at 3,000 g for 20 minutes at 4°C to separate the cells from cell debris, and then 500 µl supernatant was transferred to a sterile vessel. The supernatants were incubated with 126 µl per tube of ExoQuick exosome precipitation solution (SystemBiosciences, EXOQ20A-1, Palo Alto, CA) and incubated at 4°C for 60 minutes. After centrifugation at 1,500 g for 30 minutes at 4°C, total exosomes were isolated from each sample. To enrich ADEs, total exosomes were resuspended in 250 µl of Dulbecco’s phosphate-buffered saline and incubated for 60 minutes at 20°C with 2.3 µl of glutamine aspartate transporter (ACSA-1) biotinylated antibody (Miltenyi Biotec, 130-118-984, Auburn, CA) in 50 μl of 3% bovine serum albumin. This was followed by the addition of 25 µl of streptavidin-agarose resin (Thermo Fisher Scientific, 53117, Waltham, MA) containing 50 μl of 3% bovine serum albumin and incubation for 30 minutes at 4°C. After centrifugation at 800 g for 10 minutes at 4°C and removal of the supernatant, each pellet was suspended in 100 μl of 0.05 M glycine-HCl (pH 3.0) with vortex for 10 minutes and centrifuged at 4,000 g for 10 minutes at 4°C. Supernatants were then transferred to prechilled Eppendorf tubes containing 100 µl of 10% bovine serum albumin and 20 µl of 1 M Tris-HCl (pH 8.6) before the addition of 550 µl of mammalian protein extraction reagent (Thermo Fisher Scientific, 78505). Resultant lysates of ADEs were stored at −80°C.

As shown in Figure 2, Western blotting and transmission electron microscopy were performed to confirm the success of exosome collection according to our previous protocols.²³

Figure 2. Western blotting and transmission electron microscopy identification of serum astrocyte-derived exosomes.

EXO = exosomes, CD81 = tetraspanin protein enriched in extracellular vesicles, HSP70 = heat shock protein 70, TSG101 = tumor susceptibility gene 101.

ADE proteins were quantified by enzyme-linked immunosorbent assay kits for the human tetraspanning exosome marker CD81 (RayBio, ELH-CD81, Norcross, GA), complement fragment C3b (Abbexa Ltd, abx252114, Cambridge, UK), terminal complement complex C5b-9 (Abbexa Ltd, abx054346), and decay-accelerating factor CD55 (RayBio, ELH-CD55). The mean value for all determinations of CD81 in each assay group was set at 1.00, and relative values of CD81 for each sample were used to normalize their recovery. The intra-assay coefficient of variation was less than 10%, and the interassay coefficient of variation was less than 10–12%. The enzyme-linked immunosorbent assays microplates were read using a Varioskan LUX 3020 instrument (Thermo Fisher Scientific Oy, Ratastie 2, Vantaa, Finland). Serum samples from participants were analyzed by the same technician blinded to the clinical data.

Exosome extraction and quantification of Aβ42, T-tau, and P-T181-tau

The fasting blood of all participants was drawn and stored in a polypropylene tube containing ethylenediaminetetraacetic acid. All the blood samples were centrifuged at 4,000 g for 10 minutes to obtain the plasma. Specific neural-derived exosomes (NDEs) were isolated according to our published protocol.²³ In short, by using ExoQuick exosome precipitation solution (EXOQ, EXOQ20A-1, System Biosciences), total exosomes were collected from plasma. NDEs were then isolated by coimmunoprecipitation using a rabbit anti-L1 cell adhesion molecule (L1CAM) antibody (eBiosciences, 13-1719-82, San Diego, CA) and labeled with biotin by the EZ-Link sulfo-NHS-biotin system (Thermo Fisher Scientific, 53117). Neuronal-derived exosomal proteins were measured by enzyme-linked immunosorbent assays kits for human Aβ42 (Invitrogen, KHB3544, Waltham, MA), total tau (T-tau) (Invitrogen, KHB0041), and tau phosphorylated at threonine 181 (P-T181-tau) (Invitrogen, KHO0631). As with the complement protein assay procedure, the neural-derived exosome levels of CD81 protein were measured to normalize the exosomal content. Exosomal protein assays were performed by the same investigator blinded to the clinical data.

Intervention

We recommended intervention for all patients diagnosed with moderate-to-severe OSA (AHI ≥ 15 events/h), but only 31 initiated CPAP treatment of 4–20 cm H₂O in the context of good self-efficacy and a supportive bed partner. Of these 31 participants with OSA, 10 patients who failed to adhere to CPAP treatment were excluded. In the end, 21 patients completed the longitudinal study, including 14 patients with MCI.

Statistical analysis

All analyses were performed using SPSS 22.0 software, R software version 4.1.1, and GraphPad Prism 8.3.0. Categorical variables are presented as frequencies (percentages). The normal distribution of continuous variables was tested using the Shapiro-Wilk test. Continuous variables conforming to a normal distribution are presented as the means ± SD, and data that did not conform to a normal distribution are reported in the form of medians (interquartile ranges). Categorical variables were compared using the χ² test to evaluate differences between groups, normally distributed variables were compared using an unpaired Student’s t test or analysis of variance, and variables with a non-normal distribution were compared using the nonparametric Mann-Whitney U test or Kruskal-Wallis test.

A univariate analysis was performed to investigate the demographic and clinical characteristics of the participants. A binary logistic regression analysis was performed to determine whether complement proteins were independent risk factors for MCI in patients with OSA by calculating the odds ratio and 95% confidence interval. We performed a mediation analysis with R to test whether the ADE levels of C3b and C5b-9 mediated the relationship between OSA severity and cognition after controlling for the influence of age, sex, and education. The AHI was considered the independent variable, and MoCA scores were the outcome, while the ADE levels of C3b and C5b-9 were the mediator. Pearson’s correlation coefficients were calculated to analyze the bivariate correlations between complement protein levels and classical biomarkers of AD in patients with OSA. A multivariate linear regression analysis was used to clarify the relationships between complement protein levels and global cognition, as well as its cognitive subdomain, in patients with OSA. A paired t test was used to explore the difference in complement protein levels of patients with OSA before and after CPAP treatment. All participants had complete data for primary outcomes (such as magnetic resonance imaging, PSG, complement proteins, Aβ42, T-tau, and P-T181-tau), and the extent of missing data was capped at < 5% for clinical measures. Statistical significance was defined as P < .05, and all reported P values 2 two-tailed.

RESULTS

Demographic and clinical characteristics of the study cohorts

All participants were an average of 56.71 ± 10.10 years old and 66.98% (142) were male. A total of 169 patients were diagnosed with OSA, including 90 (53.25%) in the OSA+MCI group and 79 (46.75%) in the OSA-MCI group. The remaining 43 participants belonged to the control group.

For the cross-sectional study, the demographic and clinical data of the participants are summarized in Table 1. There were significant differences between the 3 groups in age, body mass index, hypertension, hyperlipidemia, diastolic blood pressure, glucose, mean CD81-normalized neural-derived exosome (NDE) levels of Aβ42, NDE levels of T-tau, NDE levels of P-T181-tau, WMH volume, hippocampal volume, AHI, ODI, mean SaO₂, and T90. The glucose levels, Aβ42, T-tau, P-T181-tau, hippocampal volume, mean SaO₂, and T90 were all significantly higher in the OSA+MCI group than in both the OSA-MCI and control groups. However, no differences in sex, education, diabetes mellitus, drinking, smoking, Hamilton Anxiety Scare scores, HAMD scores, AD signature region cortical thickness, total gray matter volume, or total sleep time were manifested among the 3 groups.

Table 1.

Univariate analysis of demographic and clinical characteristics in three groups (n = 212).

	OSA+MCI (n = 90)	OSA-MCI (n = 79)	Control (n = 43)	P
Demographics
Age, years (mean ± SD)	59.03 ± 10.27a	54.43 ± 10.13	56.05 ± 8.73	.011
Male, n (%)	60 (66.67)	59 (74.68)	23 (51.11)	.059
BMI, kg/m2 (mean ± SD)	27.97 ± 4.10b	28.32 ± 4.70b	25.38 ± 4.08	.001
Education, years, median (IQR)	10 (5)	12 (8)	12 (5)	.056
Hypertension, n (%)	55 (61.11)b	42 (53.16)	16 (37.21)	.035
Diabetes mellitus, n (%)	21 (23.33)	11 (13.92)	5 (11.63)	.145
Hyperlipidemia, n (%)	42 (46.67)b	28 (35.44)	10 (23.26)	.029
Drinking, n (%)	25 (27.78)	28 (35.44)	8 (18.60)	.140
Smoking, n (%)	20 (22.22)	23 (29.11)	6 (13.95)	.160
Systolic blood pressure, mm Hg, median (IQR)	139 (27)b	140 (24)	132 (25)	.069
Diastolic blood pressure, mm Hg, median (IQR)	86 (13)b	87 (12)b	80 (16)	.028
Laboratory Indicators
Total cholesterol, mg/dL (mean ± SD)	183.52 ± 46.09	188.23 ± 49.15	183.58 ± 33.51	.765
Triglycerides, mg/dL, median (IQR)	145.75 (96.13)	139.10 (65.56)	120.50 (79.74)	.227
LDL-C, mg/dL (mean ± SD)	112.36 ± 32.60	113.75 ± 32.24	108.74 ± 30.68	.710
Glucose, mg/dL, median (IQR)	97.20 (23.54)ab	94.86 (18.54)	93.60 (16.92)	.034
NDE levels of Aβ42, pg/mL (mean ± SD)	3.27 ± 0.64ab	3.00 ± 0.75b	2.56 ± 0.76	<.001
NDE levels of T-tau, pg/mL (mean ± SD)	164.52 ± 29.76ab	152.85 ± 28.66	142.47 ± 34.44	<.001
NDE levels of P-T181-tau, pg/mL (mean ± SD)	46.76 ± 10.17ab	41.30 ± 11.93	40.53 ± 11.28	.001
Neuropsychological Test
MMSE scores, median (IQR)	27 (3)ab	30 (1)	30 (1)	<.001
MoCA scores, median (IQR)	22 (4)ab	27 (2)	27 (2)	<.001
HAMA scores, median (IQR)	7 (7)	6 (7)	6 (5)	.137
HAMD scores, median (IQR)	7 (5)	5 (5)	5 (7)	.062
ESS scores, median (IQR)	9.5 (7)b	9 (8)b	5 (7)	<.001
Image Quantitative Measurement
WMH volume, mm3, median (IQR)	4842.67 (5311.97)b	4321.32 (3610.83)b	3673.27 (1798.16)	.006
AD-signature region cortical thickness, mm (mean ± SD)	2.43 ± 0.35	2.47 ± 0.32	2.54 ± 0.27	.251
Volume, mm3 (mean ± SD)
Total gray matter	450727.04 ± 50346.37	444414.42 ± 56600.04	461082.78 ± 59161.65	.275
Hippocampus	8584.52 ± 1240.06ab	9049.09 ± 1352.67	9193.05 ± 1078.11	.012
PSG Data
AHI, events/h, median (IQR)	33.90 (39.80)b	24.90 (34.20)b	2.10 (2.80)	<.001
5 events/h ≤ AHI <15 events/h, n (%)	19 (21.11)	21 (26.58)	0
15 events/h ≤ AHI <30 events/h, n (%)	23 (25.56)	22 (27.85)	0
30 events/h ≤ AHI events/h, n (%)	48 (53.33)	36 (45.57)	0	.570
ODI, events/h, median (IQR)	26.95 (35.80)b	20.20 (26.80)b	1.90 (1.80)	<.001
TST, minutes, median (IQR)	346.78 (100.20)	364.00 (87.4)	370.56 (103.00)	.620
Sleep latency, minutes, median (IQR)	21.50 (23.10)	24.50 (33.00)	29.00 (28.50)	.746
Sleep efficiency, %, median (IQR)	72.25 (20.87)	75.83 (18.20)	77.20 (21.46)	.620
Arousal index, events/h, median (IQR)	20.25 (25.30)b	16.80 (19.30)b	6.30 (9.70)	<.001
Mean SaO2, %, median (IQR)	94.00 (3.00)ab	95.00 (3.00)b	96.00 (1.00)	<.001
T90, %, median (IQR)	7.85 (18.30)ab	3.60 (9.60)b	0.00 (0.10)	<.001
NREM stage 1 of TST, % (mean ± SD)	17.46 ± 8.40b	15.13 ± 6.86b	9.76 ± 2.96	<.001
NREM stage 2 of TST, % (mean ± SD)	50.50 ± 8.45	57.71 ± 7.78	50.89 ± 7.06	.608
NREM stage 3 of TST, % (mean ± SD)	17.44 ± 6.50b	18.46 ± 6.96b	21.05 ± 6.24	.014
REM stage of TST, % (mean ± SD)	14.61 ± 4.22b	14.71 ± 4.56b	18.31 ± 5.16	<.001

^aP < .05 compared with OSA-MCI. ^bP < .05 compared with control. 1 mm Hg = 0.133kPa. AD = Alzheimer’s disease, AHI = apnea-hypopnea index, Aβ = amyloid-β, BMI = body mass index, ESS = Epworth Sleepiness Scale, HAMA = Hamilton Anxiety Scale, HAMD = Hamilton Depression Scale, IQR = interquartile range, LDL-C = low-density lipoprotein cholesterol, MCI = mild cognitive impairment, mean SaO₂ = mean oxygen saturation during sleep, MMSE = Mini-Mental State Examination, MoCA = Montreal Cognitive Assessment, NDE = neural-derived exosome, NREM = non-rapid eye movement, ODI = oxygen desaturation index, OSA = obstructive sleep apnea, P-T181-tau = tau phosphorylated at threonine 181, REM = rapid eye movement, SD = standard deviation, T90 = percentage of sleep time with oxygen saturation < 90%, TST = total sleep time, T-tau = total tau, WMH = white matter hyperintensity.

ADE levels of complement proteins in the OSA+MCI, OSA-MCI, and control groups

As shown in Figure 3, when the individual groups were analyzed in the cross-sectional study, the ADE levels of C3b and C5b-9 in the OSA+MCI group were higher than those in the OSA-MCI group (C3b: 145.72 ± 58.85 pg/mL vs 113.80 ± 33.22 pg/mL, P < .001, C5b-9: 1513.95 ± 756.48 pg/mL vs 502.80 ± 247.40 pg/mL, P < .001) and control group (C3b: 145.72 ± 58.85 pg/mL vs 98.88 ± 39.47 pg/mL, P < .001, C5b-9: 1513.95 ± 756.48 pg/mL vs 186.44 ± 88.67 pg/mL, P < .001). Moreover, the C3b and C5b-9 levels in the OSA-MCI group were higher than those in the control group (C3b: 113.80 ± 33.22 pg/mL vs 98.88 ± 39.47 pg/mL, P = .028, C5b-9: 502.80 ± 247.40 pg/mL vs 186.44 ± 88.67 pg/mL, P < .001). We also demonstrated that the CD55 levels of the OSA+MCI group were higher than those of the control group (841.97 ± 361.11 pg/mL vs 706.00 ± 239.02 pg/mL, P = .011). No significant difference was found between the OSA+MCI group and the OSA-MCI group (841.97 ± 361.11 pg/mL vs 784.60 ± 255.66 pg/mL, P = .231) or between the OSA-MCI group and the control group (784.60 ± 255.66 pg/mL vs 706.00 ± 239.02 pg/mL, P = .100).

Figure 3. The exosomal concentrations of complement proteins in cross-sectional OSA+MCI (n = 90), OSA-MCI (n = 79) and control (n = 43) groups.

(A) The C3b levels in astrocyte-derived exosomes from the OSA+MCI and OSA-MCI groups were higher than those in the control group, and the C3b levels in the OSA+MCI group were higher than those in the OSA-MCI group. (B) The C5b-9 levels in astrocyte-derived exosomes from the OSA+MCI and OSA-MCI groups were higher than those in the control group, and the C5b-9 levels in the OSA+MCI group were higher than those in the OSA-MCI group. (C) There was no significant difference in exosomal concentrations of CD55 among the 3 groups, but the exosomal concentrations of CD55 in the OSA+MCI group were higher than those in the control group. MCI = mild cognitive impairment, OSA = obstructive sleep apnea.

Pearson’s correlation analysis showed that the severity of OSA (including AHI, ODI, mean SaO₂, and T90) was not associated with complement protein levels (C3b: AHI r = .055, P = .607; ODI r = .071, P = .505; mean SaO₂ r = .016, P = .881; T90 r = −.066, P = .536. C5b-9: AHI r = .055, P = .609; ODI r = .053, P = .617; mean SaO₂ r = .189, P = .075; T90 r = −.116, P = .276. CD55: AHI r = .089, P = .356; ODI r = .098, P = .356; mean SaO₂ r = −.041, P = .700; T90 r = −.057, P = .594.).

Correlation between ADE levels of complement proteins and MCI in patients with OSA

In the binary logistic regression analysis, the ADE levels of C3b and C5b-9 were significant predictors of cognitive impairment in patients with OSA, except CD55, after adjusting for age, sex, education, vascular risk factors (eg, hypertension, diabetes, hyperlipidemia), HAMD scores, and AHI. For every 100 pg/mL increase in C3b, there was a corresponding greater risk of MCI (odds ratio: 3.619, 95% confidence intervals: 1.666–7.858). Similarly, for every 100 pg/mL increase in C5b-9, there was a corresponding increase in the risk of MCI (odds ratio: 1.590, 95% confidence intervals: 1.357–1.863) (Table 2).

Table 2.

Logistic regression analysis between the ADE levels of complement proteins and MCI in patients with OSA (n = 169).

	Model 0		Model 1		Model 2
	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
C3b, per 100 pg/mL	4.268 (2.038–8.935)	<.001	3.720 (1.759–7.868)	.001	3.619 (1.666–7.858)	.001
C5b-9, per 100 pg/mL	1.552 (1.349–1.785)	<.001	1.544 (1.340–1.779)	<.001	1.590 (1.357–1.863)	<.001
CD55, per 100 pg/mL	1.060 (0.962–1.168)	.241	—	—	—	—

Model 0: unadjusted. Model 1: adjusted for age, sex, and education. Model 2: adjusted for age, sex, education, vascular risk factors (eg, hypertension, diabetes mellitus, hyperlipidemia), HAMD scores, and AHI. ADE = astrocyte-derived exosome, AHI = apnea-hypopnea index, CI = confidence interval, OR = odds ratio, OSA = obstructive sleep apnea, MCI = mild cognitive impairment.

We used a mediation analysis to determine the association between the continuous AHI and MoCA scores mediated by complement proteins. The analyses showed that the relationship between AHI and MoCA scores was mediated by complement proteins, including C3b (Figure 4A) and C5b-9 (Figure 4B), controlling for age, sex, and education. The effect was considered partial mediation with the proportion of mediation varying from 36.5% to 102.26%. There is no evidence that AHI contributes to cognitive impairments via modulating CD55.

Figure 4. Mediation analyses with Montreal Cognitive Assessment (MoCA) as cognitive outcome.

The relationship between AHI and MoCA scores was mediated by C3b (A) and C5b-9 (B). AHI = apnea-hypopnea index, DE = direct effect, IE = indirect effect.

Adjusted relationship between imaging parameters and complement proteins in patients with OSA

In addition, multivariate linear regression analysis was performed with the imaging parameters (WMH volume, AD signature region cortical thickness, total gray matter volume, and hippocampal volume) as dependent variables and age, sex, vascular risk factors (eg, hypertension, diabetes, hyperlipidemia), AHI, and the levels of complement proteins (including C3b, C5b-9, and CD55) as independent variables in the entire OSA group. We found that the ADE levels of C3b (β: 0.244, SE: 6.187, P = .002) and C5b-9 (β: 0.231, SE: 0.422, P = .004) had a significant linear relationship with WMH volume. Unexpectedly, we did not find any correlation between the complement proteins and total gray matter volume, AD signature region cortical thickness, or hippocampal volume (Table 3).

Table 3.

Multivariable linear regression analysis between the ADE levels of complement proteins and imaging parameters in patients with OSA (n = 169).

	C3b		C5b-9		CD55
	β (SE)	P	β (SE)	P	β (SE)	P
WMH volume	0.244 (6.187)	.002	0.231 (0.422)	.004	0.023 (1.028)	.767
AD-signature region cortical thickness	−0.044 (0.001)	.582	−0.029 (0.000)	.725	−0.013 (0.000)	.874
Total gray matter volume	−0.134 (81.666)	.088	−0.116 (5.564)	.149	0.045 (13.267)	.566
Hippocampal volume	−0.139 (1.976)	.073	−0.142 (0.134)	.073	−0.091 (0.320)	.241

AD = Alzheimer’s disease, ADE = astrocyte-derived exosome, OSA = obstructive sleep apnea, WMH = white matter hyperintensity.

Association between complement proteins and AD classical biomarkers in exosomes in patients with OSA

Pearson’s correlation coefficient was used to verify the correlation between the complement proteins and AD classical biomarkers in patients with OSA. As shown in Figure 5, there were positive correlations between the ADE levels of C3b and the NDE levels of Aβ42, T-tau, and P-T181-tau (r = .237, P = .002; r = .421, P < .001; r = .209, P = .006, respectively). Similar to C3b, there were positive correlations between the ADE levels of C5b-9 and the NDE levels of Aβ42, T-tau, and P-T181-tau (r = .260, P = .001; r = .342, P < .001; r = .289, P < .001, respectively). We also found that CD55 was positively correlated with Aβ42, T-tau, and P-T181-tau (r = .158, P = .0411; r = .250, P = .001; r = .207, P = .007, respectively).

Figure 5. Correlation analysis between the complement protein levels and AD classical biomarkers levels from exosomes of patients with OSA (n = 169).

(A, D, G) The C3b levels were positively correlated with Aβ42, T-tau, and P-T181-tau, respectively. (B, E, H) The C5b-9 levels were positively correlated with Aβ42, T-tau, and P-T181-tau, respectively. (C, F, I) The CD55 levels were positively correlated with Aβ42, T-tau, and P-T181-tau, respectively. Aβ = β-amyloid, P-T181-tau = tau phosphorylated at threonine 181, T-tau = total tau.

Correlation between complement proteins and cognitive performance in patients with OSA

We observed a negative correlation between the ADE levels of complement proteins (C3b and C5b-9) and cognitive performance. After adjusting for age, sex, education, hypertension, diabetes mellitus, hyperlipidemia, HAMD scores, and AHI, significant negative associations were still observed between C3b levels and total MoCA scores (β: −0.334, SE: 0.005, P < .001), attention (β: −0.320, SE: 0.001, P < .001), language (β: −0.203, SE: 0.001, P = .009), abstraction (β: −0.307, SE: 0.001, P < .001), delayed recall (β: −0.211, SE: 0.002, P = .004), and orientation (β: −0.211, SE: 0.001, P = .008). The C5b-9 level was negatively correlated with total MoCA scores (β: −0.559, SE: 0.000, P < .001), visuoexecutive function (β: −0.317, SE: 0.000, P < .001), attention (β: −0.297, SE: 0.000, P < .001), language (β: −0.297, SE: 0.000, P < .001), abstraction (β: −0.422, SE: 0.000, P < .001), delayed recall (β: −0.482, SE: 0.000, P < .001), and orientation (β: −0.192, SE: 0.000, P = .018). The details are summarized in Table 4.

Table 4.

Multivariable linear regression analysis between the ADE levels of complement proteins and cognitive performance in patients with OSA (n = 169).

	Model 0		Model 1		Model 2
	β (SE)	P value	β (SE)	P value	β (SE)	P value
C3b
Total MoCA scores	−0.407 (0.005)	<.001	−0.350 (0.005)	<.001	−0.334 (0.005)	<.001
Visuoexecutive	−0.113 (0.002)	.144	—	—	—	—
Naming	−0.157 (0.001)	.041	−0.143 (0.001)	.064	—	—
Attention	−0.368 (0.001)	<.001	−0.341 (0.001)	<.001	−0.320 (0.001)	<.001
Language	−0.235 (0.001)	.002	−0.222 (0.001)	.005	−0.203 (0.001)	.009
Abstraction	−0.349 (0.001)	<.001	−0.309 (0.001)	<.001	−0.307 (0.001)	<.001
Delayed recall	−0.304 (0.002)	<.001	−0.230 (0.002)	.001	−0.211 (0.002)	.004
Orientation	−0.227 (0.001)	.003	−0.206 (0.001)	.009	−0.211 (0.001)	.008
C5b-9
Total MoCA scores	−0.622 (0.000)	<.001	−0.573 (0.000)	<.001	−0.559 (0.000)	<.001
Visuoexecutive	−0.383 (0.000)	<.001	−0.320 (0.000)	<.001	−0.317 (0.000)	<.001
Naming	−0.122 (0.000)	.114	—	—	—	—
Attention	−0.351 (0.000)	<.001	−0.320 (0.000)	<.001	−0.297 (0.000)	<.001
Language	−0.331 (0.000)	<.001	−0.327 (0.000)	<.001	−0.297 (0.000)	<.001
Abstraction	−0.451 (0.000)	<.001	−0.413 (0.000)	<.001	−0.422 (0.000)	<.001
Delayed recall	−0.565 (0.000)	<.001	−0.494 (0.000)	<.001	−0.482 (0.000)	<.001
Orientation	−0.223 (0.000)	.004	−0.198 (0.000)	.013	−0.192 (0.000)	.018

Model 0: unadjusted. Model 1: adjusted for age, sex, and education. Model 2: adjusted for age, sex, education, vascular risk factors (eg, hypertension, diabetes mellitus, hyperlipidemia), HAMD scores, and AHI.

AHI = apnea-hypopnea index, ADE = astrocyte-derived exosome, HAMD = Hamilton Depression Scale, MoCA, Montreal Cognitive Assessment, OSA = obstructive sleep apnea.

Alterations in the levels of complement proteins after CPAP intervention

For the longitudinal study, 21 patients with moderate-to-severe OSA were treated by CPAP, of which 14 patients had a clinical diagnosis of MCI. After 1 year of CPAP treatment, we observed that the ADE levels of C3b (109.884 ± 49.444 pg/mL vs 140.000 ± 59.277 pg/mL, P < .001), C5b-9 (918.618 ± 615.718 pg/mL vs 1,153.788 ± 747.746 pg/mL, P < .001), and CD55 (638.616 ± 344.820 pg/mL vs 777.531 ± 470.714 pg/mL, P = .008) decreased significantly compared with baseline (Table 5), but we did not observe significant improvement in MoCA scores compared with baseline (24.43 ± 3.27 vs 24.14 ± 3.51, P = .083).

Table 5.

Changes of the ADE levels of complement proteins in patients with OSA before and after CPAP treatment (n = 21).

	Baseline	Follow-up	P
C3b, pg/mL (mean± SD)	140.000 ± 59.277	109.884 ± 49.444	<.001
C5b-9, pg/mL (mean± SD)	1153.788 ± 747.746	918.618 ± 615.718	<.001
CD55, pg/mL (mean± SD)	777.531 ± 470.714	638.616 ± 344.820	.008

ADE = astrocyte-derived exosome, CPAP = continuous positive airway pressure, OSA = obstructive sleep apnea.

DISCUSSION

In this study, we obtained novel results: (1) the ADE levels of C3b and C5b-9 in the OSA+MCI group were higher than those in the OSA-MCI and control groups; (2) the ADE levels of C3b and C5b-9 were independently associated with MCI in patients with OSA, were negatively correlated with global cognitive performance and cognitive subdomains, and were positively correlated with NDE levels of Aβ42, T-tau, and P-T181-tau; (3) the relationship between AHI and MoCA scores was mediated by C3b and C5b-9; and (4) the ADE levels of C3b, C5b-9, and CD55 after 1 year of CPAP intervention were significantly lower than at baseline.

The complement system, an effector arm of the innate immune system, is also an important driver of age-related synapse loss and cognitive decline and plays an essential role in neurodevelopment, homeostasis, neuroplasticity, and neurodegeneration processes.^24–27 Complement proteins occur at low concentrations under physiological conditions and are involved in diverse processes, including neural progenitor proliferation and differentiation, neuronal migration, and synaptic pruning, which increase in levels following the inflammatory response and contribute to the pathogenesis of several inflammatory diseases and cognitive impairment.²⁸ Astrocytes are the most numerous cell type in the brain and are involved in various functions, such as metabolic support, synaptic regulation, and homeostatic system control,²⁹ and can also contribute to neuroinflammation and neurodegeneration.³⁰ Astrocytes locally synthesize most of the complement components of both the classical and alternative pathways, including proteins involved in complement activation, complement receptors, and regulators.²⁷ Recently, a study confirmed that the ADE complement protein isolated from patients with AD, such as C1q, C4b, C3d, C3b, and C5b-C9, was significantly increased and may potentially damage neurons in the late inflammatory phase of AD.¹³ A study suggested sleep loss promotes memory impairment and neurogenic decline through complement activation.³¹ Another study determined that intermittent hypoxia in OSA impaired endothelial protection against complement activity and promoted internalization of the major complement inhibitor from the endothelial cell surface, resulting in increased complement activity and inflammation in patients with OSA.³² Consistent with previous research, we determined that the patients with OSA had higher complement protein levels, and the ADE levels of C3b and C5b-9 were significantly increased in patients with OSA with MCI compared with patients with OSA without cognitive impairment. Additionally, we found that the relationship between AHI and MoCA scores in patients with OSA was mediated by C3b and C5b-9. C3b and C5b-9, but not CD55, and were negatively correlated with cognitive subdomains, including attention, language, abstraction, delayed recall, and orientation. These findings confirm that OSA significantly increases cognitive impairment progression risk, potentially through the complement pathway. Unexpectedly, there were no significant differences in the ADE levels of CD55 between the OSA-MCI group and the control group or between the OSA+MCI group and the OSA-MCI group. If the sample size of the study was increased, a significant difference might be obtained. Moreover, after CPAP treatment, the complement protein levels, including C3b, C5b-9, and CD55, all decreased significantly, confirming that CPAP therapy has considerable therapeutic value for complement dysfunction in patients with OSA.

Numerous lines of evidence suggest an association between inflammatory pathways, amyloid, and tau pathology. Inflammatory processes are required for amyloid-related synaptotoxicity, and Aβ can incite inflammation and specifically activate the complement cascade.^7,33 Complement proteins can increase Aβ production and decrease Aβ clearance. Aβ accumulation, in turn, promotes a significant oxidative and inflammatory mechanism.³⁴ Some studies have also shown that complement proteins are closely associated with tau.³⁵ Previous studies have demonstrated that OSA could impair Aβ clearance. Hypoxia results in an increased production and decreased clearance of Aβ. As OSA severity increases, the rate of Aβ burden increases.^36–38 OSA acts in synergy with Aβ and tau together, resulting in synergistic neurodegenerative outcomes.³⁹ Similar to prior studies, our study demonstrated that the mean CD81-normalized NDE levels of Aβ42, T-tau, and P-T181-tau were all significantly higher in the OSA+MCI group than in both the OSA-MCI and control groups. Additionally, we reported a positive correlation between complement proteins and Aβ and tau protein in patients with OSA, which also confirmed that the inflammatory process is closely related to classical biomarkers of AD and interacts with each other.

In the context of the previous studies in OSA, researchers found that OSA was associated with changes in white matter⁴⁰ and cortical thickness over different regions, which may indicate a distinct time course within which OSA exerts detrimental effects on brain integrity, reflective of either reactive or inflammatory processes.⁴¹ Inflammation is associated with brain structure, increased inflammation, and poor white matter integrity.⁴² Complement system activation plays a critical role in mediating neuroinflammation and white matter lesions.⁴³ In this study, our findings revealed that there were significant differences in WMH volume and hippocampal volume among the OSA+MCI, OSA-MCI, and control groups. Furthermore, we reported that the complement proteins had a significant linear effect on WMH volume, indicating that OSA may cause white matter lesions through complement pathways in addition to vascular factors. However, we found no significant difference in cortical thickness and total gray matter volume between the 3 groups.

The limitations of this study are as follows. (1) Since only snoring patients were included in our study, there was a selection bias in this study. A larger sample size including controls without OSA with or without cognitive impairment will be needed to delineate the relative roles of the complement pathways in patients with OSA. (2) The MoCA is an informative tool for a brief cognitive screening, however, a comprehensive neuropsychological evaluation is lacking in our study. (3) Due to insufficient patient adherence, only a few patients received and adhered to CPAP treatment, which may interfere with the outcome evaluation.

CONCLUSIONS

In summary, the ADE levels of complement proteins in patients with OSA, especially those with cognitive impairment, were higher. The complement proteins, which affected white matter changes and were associated with Aβ and tau, may play an important role in cognitive impairment in patients with OSA. OSA acts in synergy with Aβ and tau together, resulting in cognitive impairment, possibly through the complement pathway. The ADE levels of complement proteins decreased significantly after CPAP treatment, confirming that the complement proteins in serum ADEs can be used as a marker of cognitive impairment in patients with OSA. In the future, various follow-up studies on the cytotoxic mechanisms of complement proteins in ADEs will be necessary to validate our initial findings.

DISCLOSURE STATEMENT

All authors have seen and approved this manuscript. Work for this study was performed at Weihai Municipal Hospital. This study was supported by Medical and Health Technology Development Program in Shandong Province (2019WS226). The authors report no conflicts of interest.

ABBREVIATIONS

Aβ

β-amyloid

AD

Alzheimer’s disease

ADE

astrocyte-derived exosome

AHI

apnea-hypopnea index

CPAP

continuous positive airway pressure

HAMD

Hamilton Depression Scale

MCI

mild cognitive impairment

mean SaO₂

mean oxygen saturation during sleep

MoCA

Montreal Cognitive Assessment

NDE

neural-derived exosome

NREM

non-rapid eye movement

ODI

oxygen desaturation index

OSA

obstructive sleep apnea

PSG

polysomnography

P-T181-tau

tau phosphorylated at threonine 181

SaO2

mean oxygen saturation

T-tau

total tau

T90

percentage of sleep time with oxygen saturation < 90%

WMH

white matter hyperintensity