Programmed cell death in the cognitive impairment of obstructive sleep apnea

Abstract

Cognitive impairment (CI) is a significant and extraordinary complication of obstructive sleep apnea (OSA) patients. Programmed cell death (PCD) is an active and ordered process regulated by genes. A growing number of studies find that PCD is responsible for cognitive dysfunction and plays an important role in various neurological diseases, which involve apoptosis, necroptosis, pyroptosis, ferroptosis, and cell death associated with autophagy. However, the influence of PCD on OSA-CI remains unclear. We summarized the relevant studies that discussed the involvement of PCD in the CI of OSA and aimed to clarify the underlying mechanisms. Intermittent hypoxia (IH)-induced PCD had a critical effect on the mechanisms that produced the ultimate neurological deficit in OSA, and the PCD involved mainly included apoptosis, autophagy, ferroptosis, and pyroptosis. IH regulates PCD directly or through specific pathways, and drugs targeting related molecules have the potential to improve cognitive function. These findings enrich the pathogenesis of OSA-CI and provide new therapeutic insights.

Introduction

Obstructive sleep apnea (OSA) is the most common adult sleep disorder [1]. It is characterized by intermittent hypoxia (IH), sleep fragmentation (SF), and hypercapnia due to repeated upper airway collapse during sleep [2]. The prevalence of OSA continues to rise, and it is estimated that there are nearly 1 billion adults aged 30–69 years old with OSA worldwide [3, 4]. A recent large cohort study suggested that better sleep consolidation and absence of OSA were associated with better cognition [5]. Cognitive impairment (CI) is an important complication of OSA, causing serious distress to the lives of patients [6].

Cell death is a basic physiological process in all living organisms and plays a crucial role in maintaining tissue homeostasis and preventing disease [7]. Abnormal cell death is involved in the occurrence and progression of diseases. Insufficient cell death can lead to cancer and persistent infection, while excessive cell death is thought to be associated with degenerative and inflammatory diseases [8]. Traditional classifications of cell death include programmed cell death (PCD) and non-PCD. PCD is a regulated form of active cell death, including apoptosis, pyroptosis, necroptosis, ferroptosis, autophagy, cuproptosis, etc. [7]. Abnormal activation of PCD signaling cascades can be observed in various neurological diseases in response to different stress and inflammatory stimuli, such as apoptosis, necroptosis, pyroptosis, ferroptosis, and cell death associated with autophagy [9, 10]. A growing number of studies show that PCD is an important mechanism of CI in OSA patients [11, 12]. However, the processes leading to PCD and the type of PCD remain unclear.

In this review, we summarized and explored the recent advances in the roles of PCD in OSA-CI to figure out the underlying molecular mechanisms (Fig. 1) and provide new therapeutic targets.

Fig. 1.

Regulatory pathways or evidence of programmed cell death in OSA-CI

Clinical and imaging evidence of OSA-CI

OSA-CI involves a wide range of cognitive function areas, including memory, attention/alertness, executive function, and visuospatial deficits [13, 14]. In addition, some OSA patients suffer from emotional disorders, and severe cases can lead to permanent brain damage [6]. There is substantial clinical and imaging evidence supporting CI in OSA patients. Magnetic resonance imaging (MRI) of OSA patients showed decreased gray matter (GM) concentration or atrophy in multiple brain regions, including cortex (frontal, temporal, parietal), subcortical areas (thalamus, hippocampus, amygdala, cingulate gyrus), and cerebellum [15–17]. At the same time, the integrity of white matter (WM) in OSA patients is changed. The corpus callosum, cingulate cortex, pyramidal tract, insular cortex, basal ganglia, and limbic regions are commonly affected areas, which are mainly involved in the regulation of emotion, autonomic nervous system, and cardiovascular system [18]. Recent evidence suggests that there are biphasic changes in GM and WM in OSA. The first pattern is characterized by neuroimaging markers of GM loss and damage, suggesting chronic cell loss and accelerated aging. The other pattern presents with GM hypertrophy and reduced WM diffusivity, which may be related to acute and reactive changes [17]. In recent years, diffusion tensor imaging (DTI) has been widely used to assess brain microstructural alterations in OSA. Decreased fractional anisotropy and elevated mean and radial diffusivity in the anterior corpus callosum have been associated with impairments in prospective memory and sustained attention among OSA patients [19]. The disruption of white matter integrity and connectivity in regions such as the cingulate gyrus, accessory cingulate gyrus, and amygdala in OSA patients was reported when used DTI to construct brain structural network data, these abnormalities might be the basis for the reduced efficiency of inter-regional communication and cognitive information processing [20]. Diffusion kurtosis imaging (DKI) is another method based on a non-Gaussian diffusion model. Previous studies have reported an increase in brain mean kurtosis (MK) values in patients with OSA, which may reflect acute brain tissue injury. The mechanism may include impaired cell membrane permeability and energy pump dysfunction caused by ischemia and hypoxia, as well as changes in the volume fraction of neural tissue (cytotoxic edema or swelling of axons and neurons) [21]. Other studies have detected decreased brain MK values in patients with more severe OSA or CI, and this difference may be due to neuronal and oligodendrocyte degeneration or even cell death caused by hypoxia or ischemia [22].

Basic mechanisms of OSA-CI

There are multiple mechanisms of CI in OSA. IH and SF were seen as important initiating factors. IH can increase the level of oxidative stress in vivo, especially in the brain, which is one of the important pathological mechanisms of OSA-CI [23, 24]. The increase of reactive oxygen species and reactive nitrogen species under oxidative stress can further promote intracellular signaling cascades, leading to the increase of pro-inflammatory gene expression, and the increase of inflammatory response further aggravates oxidative stress [25]. Similarly, oxidation-antioxidant imbalance in OSA may increase inflammation, leading to neuronal apoptosis and microglial cell activation [24]. IH-induced oxidative stress and inflammation can also cause CI by damaging the blood–brain barrier (BBB), especially changes in microvascular permeability. When BBB is damaged, blood-derived glial activation inducers infiltrate the brain and cause glial activation, then induce neuronal death. On the other hand, glial activation further triggers BBB dysfunction by releasing inflammatory mediators [26]. Besides, previous studies have shown that OSA has pathological protein deposition similar to neurodegenerative diseases [27–30]. Abnormal protein deposition promotes damage or even death of nerve cells in the brain, mediating cognitive changes [31].

In OSA patients, SF is independently associated with CI including decreased sustained attention, impaired reaction time, and visuospatial deficits [32]. SF in OSA patients is associated with the unique patterns of insufficient cerebral perfusion, the regions with reduced perfusion overlap with the main areas of the default mode network and the attention network, indicating impaired attention and executive functions [33]. In addition, Alzheimer’s disease biomarkers and complement proteins derived from neural cell exosomes are regarded as mediating effects between sleep fragmentation and CI in patients with OSA [34]. In the mouse model, SF exposure led to increased oxidative stress [35, 36], neuroinflammation, microglia activation, and enhanced BBB permeability, resulting in decreased cognitive function [37].

PCD of OSA-CI

Apoptosis

Apoptosis was the earliest identified form of PCD, which is characterized by membrane blebbing, decreased cell size, nuclear fragmentation, chromatin condensation, exposure of phosphatidylserine on the cell surface, and apoptotic body formation [7, 38]. Apoptosis can be initiated through either an intrinsic (mitochondrial pathway of apoptosis) or an extrinsic pathway (death receptor-mediated apoptosis) [38]. Intrinsic apoptosis can be activated by intracellular signals such as DNA damage, Ca²⁺ overload, and elevated levels of reactive oxygen species (ROS), while extrinsic apoptosis is induced by external ligands that bind to cell surface death receptors [39]. OSA-CI is closely related to apoptosis (Table 1).

Table 1.

Apoptosis in the cognitive impairment of OSA model

Year	PMID	IH and SF model	Brain region/cell	Apoptosis Markers (methods)	Regulation mechanism	Behavior test
2012	22932184	SD rats: O2 level from 21 to 10% for 5 s every 90 s, 8 h/d, 4 weeks	Hippocampus/neuron	Apoptotic Index↑ (TUNEL)	Negatively correlated with thioredoxin	Morris water maze: escape latency↑, crossing number↓, dwell time↓
2015	26419512	SD rats: O2 decreased to a nadir level of 6.5–7% in 25–30 s and sustained for 30–35 s, then rose to 21% and lasted about 50 s. 2 min/cycle, 8 h/d, 30 days	Prefrontal cortex and hippocampus/neuron	Apoptotic cells↑, cleaved caspase-3↑, Bax/Bcl2↑ (TUNEL and WB)	–	Morris water maze: escape latency↑, crossing number↓, dwell time↓
2015	25843188	C57BL/6 J mice: IH consists of cycles of oxygen levels between 10 and 21% every 90 s, 40 cycles per hour, 8 h/d, 14 days	Hippocampus (CA1)/neuron	Percentage of apoptotic cells↑, Bcl2↓, cleaved caspase-3↑ (TUNEL and WB)	Endoplasmic reticulum stress↑	Open-field exploration: no significant difference Delay-dependent one-trial object recognition task: the percentage of preference for the novel object↓ Eight-arm maze: reference memory errors↑
2016	26996481	C57/BL mice: O2 level was reduced from 21 to 8% in 120 s, held at 8% for 120 s, returned to 21% for 50 s, and held at 21% for 300 s, 8 h/d, 5 weeks	Hippocampal dentate gyrus region/neuron	Apoptotic Index↑ (TUNEL)	Wnt Signaling Pathway (GSK-3β activity↑ and β-catenin↓)	Morris water maze: escape latency↑, crossing number↓, dwell time↓
2019	31678553	C57BL/6 J mice, KKAy type 2 diabetes model mice: (O2 concentration decreased from 21 to 5% over 30 s and then rebounded to 21% and was maintained 90 s) 5–21% O2, 30 cycles per hour, 8 h/d, 4 weeks HT22 cell: 1.5% O2 for 30 s and 21% O2 for 90 s, maintained 8 h	Hippocampus/neuron	Cleaved caspase-3↑, Bax/Bcl2↑ (WB and IHC)	HMGB1/TLR4 signaling pathway↑	–
2020	32209423	C57BL/6 mice: O2 decreased from 21 to 9% for 1.5 min and then gradually returned to 21% within 1.5 min. 20 times/h, 8 h/d, 3 weeks	Hippocampus (CA1, CA3, dentate gyrus)/neuron	Cleaved Caspase-3/pro- caspase-3↑, cleaved PARP/pro- PARP↑, Bcl2/Bax↓, apoptotic cells↑ (WB and TUNEL)	Iron overload	Morris water maze: escape latency↑, crossing number↓ Open field maze: time in open field↓, distance in center field↓
2020	32200526	Wistar rats: O2 concentration dropped from 21 to 8% in 1 min and remained at 8% for 2 min, then it went from 8 to 21% in 1 min and stayed at 21% for 2 min, 8 h/d, 4 weeks	Hippocampus/neuron	Apoptotic cells↑, Bax↑, Bcl2↓, caspase-3↑ (TUNEL and WB)	Nrdp1↑	Morris water maze: escape latency↑, crossing number↓, dwell time↓
2021	34471408	C57BL/6 J mice: concentration of O2 was maintained between 10 and 21%, cycling every 90 s for 8 h, 14 days PC12 cells: 8 episodic cycles of 21% O2 for 25 min and 0.1% O2 for 35 min	Hippocampus/neuron	Apoptotic cell ratio↑, cleaved caspase-3↑ in mitochondria: Bax↑, Bak↑, cytochrome c↓ in cytoplasm: Bax↓, Bak↓, cytochrome c↑ (flow cytometry analysis and WB)	PERK-ATF4-CHOP pathway↑	Open-field exploration: no significant difference Delay-dependent one-trial object recognition task: the percentage of preference for the novel object↓ Morris water maze: escape latency↑, number of attempts required to find the platform↑
2022	35030992	C57BL/6 mice: The O2 level was maintained at 7% for 15–20 s, recovered to 21% in 45–50 s, and sustained for 15–20 s. 20 cycles per hour, 8 h/d, 3 months	Hippocampal dentate gyrus region/neuroblast	Cleaved caspase-3↑ (IF)	Abnormal lipid droplet accumulation	Three-chamber social test: social novelty cognition↓ Morris water maze: escape latency↑, crossing number↓, dwell time↓, distances in the target quadrant↓ Fear conditioning test: fear memory impairment
2023	37553048	C57BL/6NJ mice: O2 ranged from 21.0% ± 0.5% to 9.0% ± 1.5% in 90 s every cycle, 7.5 h/d, 4 weeks PC12 cells: hypoxia (94%N2/1% O2/5%CO2) for 1 h and reoxygenation (74%N2/21%O2/5%CO2) for 30 min, a total of 6 cycles	Hippocampus (CA1)/neuron	Apoptotic index↑, cleaved caspase-3↑, Bcl2/Bax↓ (TUNEL and WB)	CaSR↑ induced PERK-ATF4-CHOP pathway↑	Eight-arm maze: working memory errors↑, reference memory errors↑, total errors↑
2023	36600080	C57BL/6 mice: O2 fluctuated from 9.0% ± 1.5% to 21% ± 0.5% in 90 s per cycle, 40 cycles/h, 7 h/d, 4 weeks PC12 cells: hypoxia (94%N2/1% O2/5%CO2) for 1 h and reoxygenation (74%N2/21%O2/5%CO2) for 30 min, a total of 6 cycles	Hippocampus (CA1)/neuron	Apoptotic index↑, cleaved caspase-3↑, Bcl2/Bax↓ (TUNEL and WB)	CaSR-PKC-ERK1/2 pathway↑	Eight-arm maze: working memory errors↑, reference memory errors↑, total errors↑
2023	37254290	C57BL/6 mice: alternating room air and 10% O2 every 90 s, and the replacement was realized within 10–30 s, 8 h/d, 4 weeks. Control the arousal frequency at 30 times/h	Hippocampus/neuron	Apoptotic cells↑ (TUNEL)	Nrf2 knockout increases apoptosis	Eight-arm maze: IH: working memory errors↑, reference memory errors↑, total errors↑ SF: reference memory errors↑, total errors↑
2023	37140776	SD rats: O2 fluctuated from 7.5 ± 0.5% to 21 ± 0.5%, 8 h/d, 4 weeks HT22 cell: 5% O2 for 30 min, and 21% O2 for 30 min, 5% CO2, 8 h in total	Hippocampus/neuron	Apoptotic rate↑, Bax↑, Bcl2↓, cleaved caspase-3/pro-caspase-3↑ (WB and flow cytometry analysis)	TGF-β3↓-Nrf-2/KEAP1/HO-1 pathway↑	Morris water maze: escape latency↑, crossing number↓, dwell time↓
2023	37268259	HT22 cell: 0.1% O2 3 min and 21% O2 7 min for 6 cycles/h, 48 h	–	Bcl2/Bax↓, cleaved caspase-3/caspase-3↑, fluorescence intensity of Annexin V-FITC (WB and apoptosis kit)	Autophagy↓	–
2023	37137262	C57BL/6J mice: The O2 level was maintained at 10% for 2 min, and recovered to 21% and sustained for 2 min. the replacement was realized within 2 min, 8 h/d, 4 weeks HT22 cell: O2 level alternate between 1 and 21% within 400 s, maintained 12 h	Hippocampus/neuron	Apoptotic cells and rate↑ (TUNEL, flow cytometry analysis)	–	Morris water maze: escape latency↑, crossing number↓, dwell time↓
2024	37977071	C57BL/6 mice: O2 levels within the cage were alternated between 6% for 30 s and 21% for 90 s, 8 h/d, 4 weeks HT22 cell: O2 in the incubator was alternated between 1% for 100 s and 21% for 100 s, maintained for 10 h	Hippocampus/neuron	Percentage of apoptotic cells↑ (TUNEL, flow cytometry analysis, CCK8)	–	Morris water maze: escape latency↑, crossing number↓, dwell time↓

TUNEL TdT-mediated dUTP Nick-End Labeling, WB western blot, IF immunofluorescence; IHC immunocytochemistry

IH-induced neuronal apoptosis and CI seem to increase with the duration of hypoxia within a certain range. A novel object recognition (NOR) test was performed to evaluate the changes in the cognitive function of mice, significant reductions in the recognition index were observed that progressed over the first 45 days and stabilized thereafter [40]. In the IH rat model, it was found that neuronal apoptosis in the hippocampus and cortex was significantly increased, and the results of the water maze suggested that cognitive decline and damage were more severe after 4 weeks of IH treatment than after 2 weeks [41]. Si et al. [42] treated HT22 (mouse hippocampal neurons) cells with IH for 48 h and found that cell activity gradually decreased from 12 h of intervention.

It seems that most IH-induced neuronal apoptosis belongs to intrinsic apoptosis, which can be regulated by multiple upstream pathways, and occurs mostly in the hippocampus and cortex. IH-induced neuronal apoptosis is closely related to oxidative stress. IH-treated rats could detect increased neuronal apoptosis in the hippocampus, which is negatively correlated with the expression of antioxidant thioredoxin [43]. Transforming growth factor-β3 (TGF-β3) exerts a protective effect on hippocampal neurons by binding to the TGF-β type receptor I (TGF-βRI), this neuroprotective effect is mediated by activation of the nuclear transcription factor 2 (Nrf-2)/Kelch-like ECH associated protein 1 (KEAP1)/hemo oxygenase 1 (HO-1) pathway, which strengthens the antioxidant defense and supports neuronal survival [44].

Endoplasmic reticulum stress (ERS) signaling can promote apoptosis [45], and acts as the linkage between IH and neuronal cell apoptosis. Increased ERS was found in the hippocampus of IH-treated mice, inducing neuronal apoptosis by upregulation of C/EBP homologous proteins (CHOP) and caspase-12, oxidative stress, and mitochondrial dysfunction [46]. After IH treatment, Bcl2 associated X protein (Bax) and Bcl2 antagonist 1 (Bak) were accumulated in mitochondria, which induced the release of cytochrome c and initiated cell apoptosis in the hippocampus of mice, and this process was regulated by the protein kinase-like endoplasmic reticulum kinase (PERK)- activating transcription factor 4 (ATF4)-CHOP pathway [47]. Another study also focused on ERS and apoptosis, they found calcium sensitive receptor (CaSR) increased, and mediated neuronal apoptosis through the PERK-ATF4-CHOP pathway under IH condition [48]. On the other hand, CaSR accelerated apoptosis and synaptic plasticity deficit by upregulating protein kinase C (PKC) and extracellular signal-regulated kinases 1/2 (ERK1/2) [49].

Specific ubiquitin ligases, glycogen synthase kinases, and inflammatory pathways, as well as autophagy and iron overload, have been reported to be involved in IH-induced neuronal apoptosis. IH could promote the expression of neuregulin receptor degradation protein-1 (NRDP1), which leads to neuronal apoptosis [50]. Similar neuronal apoptosis was also found in the hippocampus of the IH-exposed mice, and this might be regulated by increased activity of glycogen synthase kinase-3β (GSK-3β) and decreased β-catenin expression [51]. IH may induce apoptosis in hippocampal neurons by brain iron overload [52]. A recent study has shown that IH could promote neuronal apoptosis and oxidative stress by reducing levels of autophagy [42]. Besides, IH could promote neuronal apoptosis in normal mice and diabetes mice models, the latter had higher apoptosis levels and could be modulated by high mobility group box-1 protein (HMGB1)/toll-like receptor 4 (TLR4) signal [53]. IH-induced apoptosis of neuroblasts has also received attention apart from neurons. Li et al. detected abnormal lipid droplet accumulation in the hippocampus of IH-treated mice, and this change promoted neuroblast apoptosis [54].

In addition, it should be noted that SF is also an important pathogenic mechanism of OSA-CI [55], but there is little literature focusing on the effect of SF on neuronal apoptosis. A recent study has found that IH and SF treatment in mice can regulate hippocampal neuronal apoptosis through Nrf2 [35].

Autophagy

Autophagy encompasses different pathways that route cytoplasmic material to lysosomes for degradation, including 3 main types: macroautophagy, microautophagy, and chaperone-mediated autophagy [56]. It helps maintain cell homeostasis and recycle nutrients while removing toxic cellular components. However, under abnormal conditions (such as nutrient deprivation, oxidative stress, or exposure to cytotoxic agents), dysregulated autophagy can lead to cell death [7]. The normal function of the nervous system is thought to be highly dependent on autophagy because post-mitotic neurons are unable to dilute abnormal protein and organelle accumulation through cell division [56].

At present, studies on autophagy in OSA patients are pretty insufficient, and there is a lack of relevant clinical evidence to clarify the relationship between cognitive function and the level of autophagy in OSA patients. Abnormal expression of autophagy markers was observed in OSA patients. Previous studies have reported that Beclin-1 levels in the peripheral circulation of OSA patients increased with disease severity [57]. Peripheral blood cells of OSA patients showed impaired autophagy activity (decreased expression of LC3B/ATG5/BECN1/ULK1 and increased accumulation of p62) and increased DNA methylation in the promoter region of the LC3B gene, meanwhile corresponding in vitro studies showed similar autophagy trends, suggesting that IH-induced epigenetic regulation of autophagy damage [58].

Different IH conditions result in different autophagy patterns in hippocampal neurons, the duration, mode, and object of hypoxia all have effects (Table 2). Song et al. [59] used 3 IH neuron models (hypoxia duration: 2, 5, 10 min) to determine the appropriate hypoxia time to study autophagy, their results indicated that the hypoxia phase (1.5%O₂) for 5 min and reoxygenation phase (21%O₂) for 10 min led to significant difference on the markers of apoptosis and autophagy. When SD rats were treated with IH for 2, 4, and 6 weeks, the level of autophagy was increased in the hippocampus as the treatment time increased, but there were no significant differences between rats treated for 4 and 6 weeks [60].

Table 2.

Autophagy in the cognitive impairment of OSA model

Year	PMID	IH model	Brain region	Autophagy markers (methods)	Regulation mechanism	Behavior test
2019	31678553	C57BL/6 J mice, KKAy type 2 diabetes model mice: (O2 decreased from 21 to 5% over 30 s and then rebounded to 21% and was maintained 90 s) 5–21% O2, 30 cycles per hour, 8 h/d, 4 weeks HT22 cell: 1.5% O2 for 30 s and 21% O2 for 90 s, maintained 8 h	Hippocampus/neuron	LC3II/LC3I↓, P62↑, Beclin1↓ (WB and IF)	HMGB1/TLR4 signaling pathway↑	–
2020	33176803	C57BL/6 J mice: O2 fluctuated from 21 to 5% with a period of 60 s, 8 h/d, 3 weeks BV2 cell: O2 fluctuated from 21 to 5%, maintained 24 h	Hippocampus/microglia	BNIP3/NIX, BNIP3 ATG-7 and LC3II↑; P62, TOM20 ↓, MMP↓, mitochondrial fragmentation and swelling, and loss of cristae (WB, IF, flow cytometry analysis, TEM)	JNK-ERK signaling pathway↑ (p-JNK/JNK↑, p-ERK/ERK↑)	Morris water maze: escape latency↑, crossing number↓, dwell time↓
2020	31549780	C57BL/6 J mice: an orbital rotor with a speed of 55 Hz and a repeated cycle of 10 s on, 110 s off, during the light‐on phase, continuously for 2 months	Cortex	LC3B↑, Beclin1↑ and UVRAG↑ (WB and IF)	–	Morris water maze: escape latency↑, crossing number↓, dwell time↓; Novel object recognition test: discrimination index↓; Open field test: time in the central zone↓, total distance↑
2021	34469698	SD rats: The compressed air and nitrogen were filled into the chamber alternately at 30 s intervals. O2 fluctuated from 21% to 7.5%, 8 h/d, 4 weeks	Hippocampus (CA1)/neuron	Autophagic vacuole↑, LC3I↓, LC3II↑, P62↓ (TEM, WB and IF)	p38MAPK signaling pathway↑(p-p38MAPK/MAPK↑)	Morris water maze: escape latency↑, crossing number↓, dwell time↓
2021	33717152	C57BL/6 mice: O2 fluctuated from 24 to 7% with a period of 60 s, 10 h/d, 5 weeks BV2 cell: O2 fluctuated from 21 to 5%, maintained 24 h	Hippocampus/microglia	ATG-5, ATG-7, PINK1, Parkin, Beclin1 and LC3 II↑; P62, TOM20↓, MMP↓, mitochondrial swelling and loss of mitochondrial cristae (WB, IF, flow cytometry analysis, TEM)	NLRP3	Contextual fear conditioning test: freezing times in the contextual and tone conditional↓
2021	34685469	C57BL/6 mice: sleep was interrupted at 2 min intervals during the 12 h light period, maintained 5d	Striatum and hippocampus	striatum: LC3II↑, Beclin1↑, P62↑ (WB); hippocampus: LC3II↑, Beclin1↓, P62↓ (WB)	–	–
2022	35401114	C57BL/6 mice: O2 fluctuated from 21.0 ± 0.5% to 9 ± 1.5%, 90 s/cycle, 7 h/d, 4 weeks	Hippocampus (CA1)/neuron	Autophagosomes↑, LC3II/LC3I↑, P62↑, Beclin1↑, ATG5↑ (TEM, WB and IF)	–	Eight-arm maze: working memory errors↑, reference memory errors↑, total errors↑
2022	36611953	C57BL/6 mice: sleep was interrupted at 2 min intervals during the 12 h light period, maintained for 3 weeks	Hippocampus	LC3II↑ (WB)	–	–
2023	37268259	HT22 cell: 0.1% O2 3 min and 21% O2 7 min for 6 cycles/h, 48 h	–	Fluorescence intensity of autophagy↓, autophagosomes↓, LC3II/LC3I↓, P62↑, Beclin1↓ (autophagy staining test kit, TEM, WB)	AMPK-mTOR signaling pathway (P-AMPK/AMPK↓, P-mTOR/mTOR↑)	–

TEM transmission electron microscopy, WB western blot, IF immunofluorescence, MMP mitochondrial membrane potential

To date, the effects of IH on autophagy levels remain inconclusive, with inconsistent findings reported across different studies. IH could reduce the level of autophagy by altering the expression of adenosine 5′-monophosphate-activated protein kinase (AMPK)-mammalian target of rapamycin (mTOR) signaling pathway [42]. A study found autophagy was reduced both in normal mice and diabetes mice after IH treatment, the autophagy level was lower and could be regulated by HMGB1/TLR4 signaling in the diabetes model [53]. Moreover, Li et al. [61] detected that the expressions of LC3II/LC3I, P62, Beclin1, and autophagy-related genes (ATG) 5 were increased, and autophagosomes increased, which indicated autophagy flux inhibition in the hippocampus of IH mice. However, other studies show that IH could promote autophagy. In IH-treated SD rats, the level of autophagy activated, inhibiting the p38 mitogen-activated protein kinase (MAPK) signaling pathway could further activate autophagy in hippocampal nerve cells, thus reducing nerve cell injury [60]. In addition, IH was shown to aggravate CI in brain ischemia/reperfusion rat model, the underlying mechanism was associated with phosphatidylinositol 3‑kinase (PI3K)‑mTOR‑autophagy pathway activation and nerve cell damage [62]. The reason for the inconsistent results may be the different disease courses caused by the differences in IH conditions.

IH initiated time-dependent mitophagy to remove damaged mitochondria in microglia, which changed levels of inflammation and oxidative stress [63, 64]. Activated BCL2-interacting protein 3 (BNIP3)-dependent and PTEN-induced putative kinase1 (PINK1)-Parkin pathway-mediated mitophagy were detected in microglia of IH model, the former was activated by c-Jun N-terminal kinase (JNK)-extracellular regulated kinase (ERK) signaling pathway and the latter was regulated by nucleotide‐binding domain like receptor protein 3 (NLRP3) expression [63, 64].

SF is also involved in the CI of OSA through autophagy dysregulation. The unique autophagy rhythm of the hippocampus is altered by SF [65]. Short-term SF leads to dysregulated autophagy in the striatum and hippocampus but not in the frontal cortex, indicating that SF-related autophagy may tend to occur in specific brain regions [66]. Endosome-autophagosome-lysosome pathway dysfunction and microglia-mediated neuroinflammation were considered to be similar mechanisms for chronic SF and neurodegenerative diseases [67]. Short-term SF-activated microglia are in the striatum, while long-term SF-activated microglia are in the hippocampus [66, 68]. SF-induced microglial activation in the hippocampus was related to dysregulation of autophagy [68].

Ferroptosis

Ferroptosis is defined as an iron-dependent regulated necrosis that is caused by massive lipid peroxidation-mediated membrane damage, and mainly occurs through the extrinsic or transporter-dependent pathway and the intrinsic or enzyme-regulated pathway [69, 70]. It is implicated in the pathological cell death associated with many diseases including neurodegenerative diseases and ischemia–reperfusion injury [69]. However, the role of ferroptosis in OSA-CI is still in the preliminary stage of exploration.

The expression of ferroptosis-related genes and iron metabolism are altered in OSA patients. Serum iron levels in obese patients with OSA were significantly higher than those without OSA, and transferrin saturation correlated with OSA severity and duration of hypoxia, suggesting an association between OSA-induced hypoxia and iron metabolism in obese patients [71]. Huang et.al identified 13 ferroptosis-related differentially expressed genes in OSA patients with CPAP treatment as potential targets [72]. Recently, a large cohort study demonstrated that iron homeostasis was altered in the brain of OSA patients, increased iron accumulation was observed in specific cortical regions by quantitative MRI analysis, and nocturnal hypoxemia was closely related to higher iron levels [73]. On the other hand, malondialdehyde (MDA), 8-iso-prostaglandin F2α (8-iso-PGF2α), oxidized low-density lipoprotein (oxLDL), and thiobarbituric acid-reactive substances (TBARS) are common markers of lipid peroxidation in OSA [24, 74]. The central nervous system is rich in lipids, ROS can oxidize polyunsaturated fatty acids easily, and IH induces the increase of ROS and the imbalance of the oxidation-antioxidant system [24, 35, 75]. Therefore, lipid peroxidation may affect the cognition of OSA.

Similar to the MRI data of OSA patients, iron overload was found in the hippocampus of IH murine models, and neurological damage was attenuated when IH-induced ferritin and transferrin receptor 1 were inhibited [52]. Neuronal ferroptosis was also found in the prefrontal cortex of IH mice, which led to neuronal loss and cognitive decline [76]. One study examined ferroptosis-related markers in hippocampus of IH rat model and found mitochondrial damage, increased Fe²⁺, Malondialdehyde (MDA) levels, and Acyl-CoA synthetase long-chain family member 4 (ACSL4) protein expression were observed in the IH group, while superoxide dismutase (SOD), glutathione (GSH), and glutathione peroxidase 4 (GPX4) protein expression were decreased [12].

Pyroptosis

Pyroptosis is a PCD mediated by gasdermin protein, which is triggered by caspases activated by some inflammasomes. It is manifested as cell swelling, plasma membrane cleavage, chromatin fragmentation, and the release of proinflammatory substances in cells [77, 78]. Cytokine release in pyroptosis induces cell injury, leads to cell damage and dysfunction of neurons and blood vessel cells, and results in loss of memory and executive function [79]. Ischemia/hypoxia can lead to pyroptosis of brain neurons, microglia, and microvascular endothelial cells [80–83]. However, the relationship between pyroptosis and IH in CI of OSA remains unclear.

A previous study showed that IH aggravated neuroinflammation and pyroptosis in early brain injury after subarachnoid hemorrhage through the Apoptosis-associated speck-like protein containing a CARD (ASC)/hypoxia-inducible factor-1α (HIF-1α) pathway [84]. Chen et al. [61] isolated exosomes from the plasma of severe OSA patients, coincubated mouse hippocampal neurons with exosomes, or injected exosomes into mice via caudal veins, their results suggested that OSA plasma-derived exosomes promoted neuronal pyroptosis and increased expression of inflammatory factors and led to cognitive dysfunction in mice [85]. The inflammasome, also known as the pyroptosome, is a supramolecular entity that initiates the pyroptotic cell death process [79]. NLRP3 is a member of the inflammasome family, and NLRP3 signaling is increased in a HIF-1α-dependent manner in severe OSA patients and IH monocytes, with upregulated inflammatory cytokines (interleukin-1β and interleukin-18) and gasdermin D [86]. Expression levels of pyroptosis protein including NLRP3, cleaved caspase-1, and ASC are also increased in the hippocampus of IH mice [64]. These studies indicate that pyroptosis may be involved in the progression of CI in OSA.

Interaction between PCDs

Different types of PCD can be considered as a unified cell death program, in which the individual pathways are highly linked and can be flexibly coordinated and compensated for each other [87]. Ferroptosis has been found to interact with apoptosis and autophagy in neurodegenerative diseases, the balance of PCDs can be controlled by regulating their same protein target [88]. In OSA, oxidative stress, ERS, and mitochondrial damage may be the common pathways regulating PCDs. Under the condition of oxidative stress, intricate interactions occur among apoptosis, autophagy, and ferroptosis. Cytoprotective autophagy contributes to cellular homeostasis by eliminating damaged organelles and misfolded protein aggregates, thereby reducing reactive oxygen species and mitigating apoptosis. Nevertheless, under specific pathological conditions, autophagy may transition into an apoptotic response. Ferroptosis inducers have been shown to activate apoptotic pathways via endoplasmic reticulum and mitochondrial stress and can also initiate non-selective autophagy. Moreover, excessive autophagy, in the presence of oxidative stress and lipid peroxidation, may facilitate ferroptosis. Selective and chaperone-mediated autophagy further potentiates ferroptosis by degrading key inhibitory proteins, leading to increased accumulation of intracellular free iron and lipid peroxides [89]. A recent study suggested that IH-induced neuronal ferroptosis mediates the occurrence of ERS [76]. Meanwhile, ERS can promote neuronal apoptosis under IH treatment [47, 48]. Furthermore, mitochondrial dysfunction or damage triggered by IH has been identified as an upstream pathway of apoptosis and ferroptosis [90, 91]. Few studies have investigated the interaction between PCDs under IH conditions. Activating the Nrf2 pathway and autophagy can protect the cardiac function of IH mice and reduce cardiomyocyte apoptosis [92, 93]. While inhibition of autophagy in liver kupffer cells could aggravate IH-induced apoptosis [94]. Ferrostatin-1, one of the ferroptosis inhibitors, was reported to be able to ameliorate apoptosis and injury in aortic endothelial cells [90] and lung cells [95] under IH condition. However, crosstalk between PCDs has not been fully discussed yet in OSA-CI. In IH treated neuron model, Tanshinone IIA could inhibit oxidative stress and neuronal apoptosis by activating autophagy [42]. In IH-treated mice, Sulforaphane alleviates hippocampal neuronal apoptosis by enhancing Nrf2 nuclear translocation and autophagy [61]. Notably, simultaneous inhibition of multiple PCDs is considered to have more potential to improve ischemia–reperfusion injury than a single type of PCD [96]. Therefore, it is worth exploring the links between different PCDs, which may have potential clinical value.

PCD induced by intercellular interactions

IH can directly activate microglia and promote the release of inflammatory cytokines in the central nervous system. Excessive nerve inflammation further promotes the activation of glial cells, causes synaptic loss, and neuron damage, and eventually leads to neurocognitive defects [11]. The majority of studies on PCD induced by intercellular interactions in OSA focused on apoptosis, neuronal apoptosis can be indirectly mediated by IH-induced microglial activation. Wang et al. found SUMO-specific proteases 1 (SENP1) expression was decreased in IH-treated microglia, which reduced the level of Target of Myb 1 (TOM1) and peroxisome proliferator-activated receptor γ (PPARγ) by promoting their SUMOylation, and aggravated neuroinflammation and neuronal apoptosis [97, 98]. Another study also focused on the effects of microglia on neurons and found that IH induced microglial activation and release of pro-inflammatory cytokines through the HMGB1/TLR4/NF-κB signaling pathway, leading to neuronal apoptosis [99]. Besides, the combined intervention of high glucose and IH can activate microglia, lead to the release of neuroinflammatory factors (ROS, TNF-α, IL-1β), and mediate the apoptosis of HT22 cells in co-culture through the PI3K/Akt/GSK-3β signaling pathway [100].

Possible treatment for OSA-CI

Most studies have been performed in animal or cellular models of IH because of limited access to brain specimens from OSA patients. In addition to traditional CPAP treatment, previous studies have figured out various drugs that protect brain tissues from IH-induced CI by targeting apoptosis and autophagy (Fig. 2).

Fig. 2.

Possible treatments for OSA-CI

CPAP treatment

The first-line treatment for OSA is continuous positive airway pressure (CPAP) [101]. Most current research on the treatment of OSA-CI has focused on CPAP, and the data for other treatments is pretty limited. Adherence to CPAP therapy seems to be able to slow and reverse the cognitive decline in OSA patients and reduce the risk of progression to dementia [102]. The duration of CPAP therapy is critical to the effect of cognitive improvement in OSA patients. Previous studies showed that one month of CPAP treatment could reduce the sleepiness of OSA patients and improve verbal episodic memory [103], 12 weeks of treatment also showed a positive effect on verbal memory [104], and three months of treatment led to significant improvements in episodic memory, short-term memory, and executive function [105]. A systematic review suggested that CPAP therapy should be persisted for at least 4 weeks to improve OSA-CI [106]. Scholars have reported the time trajectory of NOR test performance in IH mouse models. Switching to normoxia in the early stage of intermittent hypoxia intervention and over a specific period, the CI of mice is partially reversed instead of full recovery [40]. However, there is little evidence of the pathological improvement of CI in OSA by CPAP or normoxia restoration. Therefore, we cannot summarize the effect of similar treatments on the different PCDs of hippocampal nerve cells.

Anti-apoptosis treatments to manage OSA-CI

Anti-apoptosis is considered to be an effective way to improve CI in neurodegenerative diseases and brain injury [107–109]. At present, most of the neuroprotective substances found by the existing evidence act on hippocampal neuron apoptosis in IH models.

Antioxidants

IH induces an oxidation-antioxidant imbalance in OSA [24]. OSA-CI is strongly associated with neuronal damage in brain regions most sensitive to hypoxia and oxidative stress, such as the hippocampus and cerebral cortex regions [2, 42, 110]. Imaging data indicating changes in brain function were correlated with the apnea–hypopnea index and oxygen desaturation index in OSA patients [111]. The level of oxidative stress in the peripheral blood of OSA patients is closely related to hypoxic sleep parameters and cognitive scores [112, 113]. Increased oxidative stress in the brain has also been demonstrated in rodent models with IH treatment [52].

Improving oxidative stress can alleviate CI by reducing neuron loss [114, 115]. Many plant extracts and active ingredients of herbal medicines and novel compounds play an important role in anti-apoptosis and anti-oxidative stress, thereby reducing the nerve damage of IH. A plant-derived drug named apocynin was reported to inhibit NADPH oxidase activity and attenuate neuronal apoptosis in the hippocampus of IH rats [116]. Protocatechuic acid is a simple phenolic compound with antioxidant and anti-inflammatory effects. It was found that protocatechuic acid can lessen oxidative stress and apoptosis in the hippocampus of IH rats [117]. Astragalus was reported to protect against IH-induced hippocampal neuron apoptosis in rats with decreased HIF-1α expression [118]. Sulforaphane (Nrf2 agonist) was found to have a neuroprotective effect in the IH and SF mice model, which could reduce the apoptosis of hippocampal neurons by increasing the level of Nrf2 [35]. Sulforaphane could also alleviate neuronal apoptosis by activating autophagy [61]. Moreover, N-acetylcysteine can increase the level of thioredoxin, reduce neuronal apoptosis, and improve cognitive function in the hippocampus of IH rats [43]. Thus, Antioxidants have been shown to have neuroprotective effects in IH models.

Anti-inflammatory agents

The systemic inflammation of OSA is mediated by IH [119]. Peripheral inflammatory signals pass through the damaged blood–brain barrier and vagus nerve into the central nervous system, causing neuroinflammation [120]. Many studies have found that peripheral inflammation in OSA patients is associated with cognitive decline [112, 121, 122]. On the other hand, IH can directly activate microglia and astrocytes and stimulate excessive neuroinflammation, leading to synaptic damage and loss, and neuronal apoptosis [11].

Neuroprotective therapy based on anti-neuroinflammation is effective in many neurodegenerative diseases with cognitive decline [123–125]. It seems to be effective for OSA as well. Toll-like receptors (TLRs) family are the initiating molecules of inflammation, which are widely expressed in microglia, astrocytes, and neurons. The activation of glial cells in neuroinflammation is closely related to TLRs [126, 127]. Increased expression of TLR2 can stimulate glial cells to secrete inflammatory factors, ultimately leading to TLR2-induced neuronal apoptosis [128, 129]. TLR2 gene knockout could reduce neuron loss and abnormally activated glia in the hippocampus of IH mice [130]. Additionally, pseudo ginsenoside GQ was found to have anti-inflammatory effects, significantly ameliorated spatial learning deficits, and inhibited microglial activation, pro-inflammatory cytokine release, and neuronal apoptosis in the hippocampus of IH mice [99]. Recently, the application of Banxia-Houpu decoction showed an increase in the Bcl-2/Bax ratio, thereby mitigating neuronal damage in IH mice [131]. Moreover, SENP1 promotes microglia migration by alleviating SUMOylation of protein TOM1, reducing neuroinflammation, neuronal pathological deposition, and neuronal apoptosis subsequently [97].

ERS inhibitors

Misfolded/unfolded proteins within the endoplasmic reticulum are a common feature of nervous system diseases, inducting endoplasmic reticulum stress. It could initiate an unfolded protein response to maintain protein homeostasis, but cell death and inflammation will be activated if the damage is irreversible [132]. Previous studies have proposed that hypoxia elicits ERS [133]. ERS-mediated apoptosis is one of the important mechanisms of cognitive dysfunction in OSA [41].

Some drugs that can directly or indirectly inhibit ERS have attracted people’s attention. Tauroursodeoxycholate (TUDCA), which is a member of the ERS inhibitor, can reduce IH-induced neuronal apoptosis in the hippocampal CA1 region by upregulating the anti-apoptotic protein Bcl-2 in mice [46]. Treatment with CaSR inhibitors alleviates the apoptosis of hippocampal neurons in IH models by inhibiting the ERS pathway indirectly [48].

Endogenous factors

Some endogenous factors were proven to be involved in the apoptosis of hippocampal neurons under IH conditions. Orexin, a neuropeptide produced in the lateral hypothalamus, has attracted much attention in sleep disorders and the treatment of neurological diseases [134, 135]. Orexin A was reported to improve IH-induced hippocampal apoptosis and oxidative stress [136]. Hypoxia facilitates the accumulation of the presynaptic neuromodulator adenosine, which modulates synaptic plasticity through its interaction with the inhibitory adenosine 1 receptor (A1R) and facilitatory adenosine _2A receptor (A_2AR). When mice were treated with IH, activation of adenosine A1R and blockade of adenosine A_2AR reduced apoptosis of hippocampal neurons, attenuated long-term potentiation, and alleviated memory impairment [137, 138]. Melatonin was also found to have a protective effect against apoptosis of hippocampal neurons in IH rats [139].

Metal element regulators

Abnormal deposition/distribution of metal ions in different brain regions induces oxidative stress, endoplasmic reticulum stress, mitochondrial and autophagy dysfunction, and is involved in CI in a variety of neurodegenerative diseases [140, 141]. Emerging evidence also suggests that CI in OSA is related to the imbalance of metal homeostasis. Some scholars have detected abnormal iron accumulation in the brain of OSA patients through imaging [73]. Similarly, IH mice showed increased brain iron levels [131]. Iron may specifically contribute to nerve cell death when the level of glutathione is reduced [142]. In addition, treating mice with IH for a long time could increase brain cobalt, predominantly in the white matter [143].

Some chelating agents and metal-based drugs that regulate metal ion homeostasis have been proposed as one of the alternative treatment options for neurodegenerative diseases [144]. Several studies have set their sights on these unique therapeutic targets in OSA. Mood stabilizing agent LiCl decreased the activity of GSK-3β and increased the expression of β-catenin, and partially reversed neuronal apoptosis and spatial memory deficits in IH mice [51]. Huperzine A, which acts as an effective iron chelator, could attenuate apoptosis, oxidative stress, and synaptic plasticity mediated by IH in mice [52].

Inhibition of lipid accumulation

Lipid dysregulation can stimulate pathological protein deposition, and lead to mitochondrial and endoplasmic reticulum dysfunction and even nerve cell death. It can also increase the burden on the cerebrovascular system, induce insulin resistance, and thus affect the structure of neurons indirectly [145]. A recent study examined the changes in lipid metabolism in the cerebrospinal fluid and found that the primary mechanism for the association between OSA and Alzheimer’s disease may be an increase in lipid oxidation in the central nervous system.

It has been reported that cognition can be improved by regulating the lipids of nerve cells. Gedam et al. found that hypoxia in microglia can induce lipid droplet accumulation, while depletion of the C3a receptor complement saves the dysregulated lipid profile and improves the phagocytosis and aggregation ability of microglia [146]. An active component of Salvia miltiorrhiza, SMND-309, dramatically alleviated CI in mice by decreasing lipid droplet accumulation in the hippocampus, and this further reduced neuronal injury, neuroblast apoptosis, and glial activation [54].

Active-autophagy treatments to manage OSA-CI

Autophagy has been considered as a potential therapeutic target for neurodegenerative diseases according to available evidence [147, 148]. Autophagy and mitophagy are important potential targets for OSA neurocognitive impairment as well. Tanshinone IIA is extracted from salvia miltiorrhiza, which is a traditional Chinese medicine, it can inhibit oxidative stress and neuronal apoptosis by activating the AMPK/mTOR autophagy pathway under IH conditions [42]. Pinocembrin is a natural flavonoid drug, that has antimicrobial, anti-inflammatory, and antioxidant properties [149]. Pinocembrin can activate BNIP3-dependent mitophagy through the JNK-ERK pathway to exert neuroprotective function in IH models, inhibiting the formation of NLRP3 inflammasome [64], and NLRP3 deficiency was reported to protect against IH-mediated neuroinflammation and mitochondrial ROS [63].

Conclusion and prospects

CI is an important complication of OSA, which brings great inconvenience to the lives of patients [6]. CPAP can partially improve CI in OSA [106], but poor adherence drives us to explore new effective treatments. In this review, we briefly summarized the relevant studies that discussed the involvement of PCD in the CI of OSA. We found that IH-induced PCD had a critical effect on the mechanisms that produced the ultimate neurological deficit, and the PCD involved mainly included apoptosis, autophagy, ferroptosis, and pyroptosis. IH regulates PCD directly or through specific pathways, and drugs targeting related molecules have the potential to improve cognitive function.

In the future, the following contents are what we need to pay attention to in this field: (a) Current IH disease models (oxygen concentration, hypoxia duration, and hypoxia frequency and extent) need to be unified as much as possible, select the one that is more relevant to the clinical reality, and fully consider the role of SF, then describe PCD changes in different disease phases. (b) New and comprehensive PCDs deserve to be investigated to find out whether there is a dominant type of PCD in OSA-CI and to explore the deeper molecular mechanisms. (C) Reciprocal regulation between PCDs involved. (d) PCD-based therapy must be developed and evaluated in clinical trials.

Author contributions

D.Z., R.O., and Y.O. made the conceptualization. Y.O. conducted manuscript writing. X.W. helped visualization. D.Z. and R.O. took charge of funding acquisition and project administration. All authors reviewed and edited the manuscript.

Funding

This work was supported by grants from the Natural Science Foundation of Hunan Province (Grant No. 2024JJ5488), the Health Commission of Hunan Province (Grant No. W20243062), and the National Key Clinical Specialty Construction Projects of China.

Availability of data and materials

Not applicable.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no conflict of interest.