MiR-106a-5p in extracellular vesicles derived from alveolar epithelial cells mediates cognitive dysfunction induced by chronic intermittent hypoxia in mice through MAPK signaling pathway

Background

Obstructive sleep apnea (OSA), characterized by chronic intermittent hypoxia (CIH), is frequently associated with cognitive dysfunction. However, the underlying peripheral-central interplay mechanism remains to be elucidated. Recent years have witnessed the proposal of the “lung-brain axis (LBA)” concept, suggesting that lung tissue can remotely regulate brain function via extracellular vesicles (EVs). This investigation aims to determine whether EVs derived from alveolar epithelial cells (AEC-EVs) mediate CIH-induced cognitive impairment and to delineate the associated molecular mechanisms.

Methods

Mice were exposed to CIH to model obstructive sleep apnea. EVs were isolated from brain tissue and MLE-12 cells via ultracentrifugation. CIH-AEC-EVs were administered to normal mice via tail vein injection; cognitive function was assessed using behavioral tests (Open Field, Y-Maze, Novel Object Recognition). In vitro, BV-2 cells were treated with CIH-AEC-EVs, and their polarization status was evaluated by Flow Cytometry (FCM), Quantitative Real-Time PCR (qPCR), Western Blotting (WB), and Immunofluorescence (IF). Key miRNAs and their target genes were screened and validated using miRNA sequencing, bioinformatics analysis, and dual-luciferase reporter assays. Finally, functional rescue experiments were performed using a miR-106a-5p inhibitor and a MAPK inhibitor to validate the functional outcomes both in vivo and in vitro.

Results

CIH-exposed mice exhibited cognitive impairment, hippocampal neuronal apoptosis, and increased M1 polarization of microglia. CIH markedly increased the abundance of alveolar-epithelial-cell-derived EVs (AEC-EVs) and microglial EVs in the brain, whereas neuron-derived EVs remained unchanged. CIH-AEC-EVs traversed the blood-brain barrier (BBB), were taken up by microglia, and induced M1 polarization while suppressing M2 polarization. Mechanistically, miR-106a-5p were enriched in CIH-AEC-EVs, which directly targeted DUSP2 mRNA, thereby relieving DUSP2-mediated suppression of ERK/MAPK signaling and facilitating M1 polarization. Administration of a miR-106a-5p antagonist or a MAPK inhibitor significantly reversed the aforementioned pathological alterations and ameliorated cognitive function.

Conclusion

Through the lung-brain axis, CIH enhances the transfer of miR-106a-5p-loaded AEC-EVs to the hippocampus, where they downregulate DUSP2 and activate the MAPK signaling pathway. This alteration results in an M1/M2 microglial imbalance, which contributes to cognitive dysfunction. Targeted suppression of AEC-EVs secretion or the miR-106a-5p/DUSP2 axis may provide a potential non-invasive therapeutic strategy for addressing cognitive impairments associated with OSA.

Graphical abstract

The alveolar epithelial cells of patients with chronic intermittent hypoxia secrete a large number of miR-106a-5p EVs through the lung brain axis, which act on brain tissue, leading to increased polarization of microglia M1 and neuronal cell apoptosis.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-025-03628-8.

Highlights

1. This study reveals that chronic intermittent hypoxia (CIH) mediates the lung-brain axis via alveolar epithelial cell-derived EVs (AEC-EVs), remotely regulating microglial polarization in the central nervous system and ultimately leading to cognitive dysfunction.

2. The study identifies that miR-106a-5p, highly enriched within AEC-EVs, drives M1 polarization and suppresses M2 polarization in microglia by targeting and inhibiting DUSP2, consequently activating the MAPK signaling pathway.

3. This research distinguishes, for the first time, the functional disparity between AEC-EVs and microglia-derived EVs under CIH conditions, demonstrating that the former predominantly drives neuroinflammation, while the latter may exert compensatory protective effects potentially via M2 polarization.

4.This study proposes non-invasive strategies targeting either AEC-EVs or the miR-106a-5p/DUSP2-MAPK axis, offering novel perspectives for the prevention and treatment of OSA-associated cognitive impairment.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-025-03628-8.

Introduction

Obstructive sleep apnea (OSA) is a prevalent sleep-related breathing disorder, pathophysiological characterized by chronic intermittent hypoxia (CIH) and sleep fragmentation. Globally, approximately 10–26% of adults are diagnosed with OSA [1–3]. In recent years, accumulating research has revealed a strong association between OSA and cognitive dysfunction, with pronounced impairments observed in attention, executive function, memory, and learning abilities [4–8].Structural alterations in the brain, particularly within the hippocampus, have been documented in OSA patients [9, 10].

The lung brain axis (LBA) is an important discovery in the field of neuroimmunology in recent years, referring to the bidirectional communication network formed between the lungs and the central nervous system (CNS) through mechanisms such as microbiome, immune regulation, and neural pathways. Previous studies have shown that lung microbiota, inhaled pollutants, and some chronic respiratory diseases can all cause neurological damage through the lung brain axis [11–15].

CIH is recognized as the core pathogenic factor leading to cognitive impairment. The repetitive hypoxia-reoxygenation cycles trigger neuroinflammation, oxidative stress, and mitochondrial dysfunction, thereby impairing brain regions critically involved in cognitive function, such as the hippocampus and prefrontal cortex. Extensive animal studies have corroborated that CIH induces neuronal apoptosis, synaptic impairment, and cognitive deficits [16]. Furthermore, microglial activation plays a pivotal role in CIH-induced neuroinflammation, exacerbating neural damage and cognitive dysfunction.

Microglia constitute the primary immune cells within the central nervous system. Under physiological conditions, microglia remain quiescent, continuously surveying the brain microenvironment to maintain neuronal homeostasis. However, when the cerebral milieu is perturbed—by hypoxia, infection, trauma, or neurodegenerative pathology—microglia rapidly become activated and release abundant pro-inflammatory mediators, including IL-1β, TNF-α, and reactive oxygen species (ROS), thereby fueling neuroinflammation, neuronal injury, and cognitive decline [17]. Microglial activation is broadly classified into pro-inflammatory (M1) and anti-inflammatory (M2) phenotypes. M1 microglia secrete pro-inflammatory cytokines that amplify neuroinflammation, whereas M2 microglia release anti-inflammatory factors that support neuroprotection and tissue repair. A delicate balance between these phenotypes is essential for microenvironmental homeostasis. Excessive M1 polarization or impaired M2 function disrupts this equilibrium, resulting in persistent neuroinflammation and accelerated cognitive decline [18].

Extracellular vesicles are secreted by diverse cell types [19]. They possess a lipid bilayer membrane and transport diverse bioactive cargos, including proteins, lipids, mRNAs, miRNAs and lncRNAs [20]. Accumulating evidence underscores their critical role in the pathogenesis and progression of cognitive dysfunction, and is especially relevant to Alzheimer’s disease (AD), diabetes-associated cognitive impairment and traumatic brain injury (TBI) [21].

Previous studies indicate that CIH induces the release of EVs from pulmonary cells, including alveolar epithelial cells, endothelial cells, and fibroblasts. These EVs exhibit alterations in their quantity, size, protein markers, and miRNA cargo [22]. CIH-induced pulmonary EVs may carry pro-inflammatory miRNAs or proteins that activate alveolar macrophages and fibroblasts, promoting the release of inflammatory mediators such as IL-1β and TNF-α. This cascade thereby exacerbates pulmonary inflammation and fibrotic processes [23].

Given the differential impairment of pulmonary and brain tissues under CIH, we hypothesized that crosstalk exists between these organ systems. Herein, we aim to elucidate the molecular mechanisms by which alveolar epithelial cell-derived EVs (AEC-EVs) mediate CIH-induced cognitive deficits via the lung-brain axis.

Materials and methods

Animals

This study protocol followed the principle of “3Rs” and was approved by the ethics review committee of Shandong First Medical University. The C57BL/6 mice (Male, 6–8 weeks old) were obtained from Beijing Vital River Laboratory Animal Technology (Beijing, China). The mice were housed in an environment with a temperature of 25 ± 2℃ and a humidity of 50 ± 10% and a light/dark cycle for 12 h.

EVs were resuspended in 100 µL PBS (equivalent to 100 µg protein). The EVs-treated group received EVs via tail vein injection, while the control group received an equivalent volume of PBS. EVs were administered weekly for 10 consecutive weeks [24–26].

CIH model in vivo

The mice were exposed to CIH conditions from 8:00 AM-5:00 PM per day for about 10 weeks [27]. Firstly, the chamber was filled with N₂ for 85–95 s, and the oxygen level was reduced from 21% ± 1% (normal) to 7% ± 1% (hypoxia). The oxygen level was then maintained at 7% ± 1% for 15–20 s. Finally, the level was recovered to 21% ± 1% in 45–50 s and sustained for 10–20 s. A cycle of CIH lasted approximately 180 s, and there were approximately 20 cycles per hour [28]. The control group was placed in chambers with a constant 21% O₂, while all other experimental conditions were kept identical to those of the CIH group.

Grouping

Firstly, mice were allocated into two groups: the Normoxia group, which was housed continuously in chambers maintained at 21% O₂, and the CIH group, which underwent chronic intermittent hypoxia exposure for approximately 10 weeks (Fig. 1). In addition, mice were designated as Normoxia-EVs or CIH-EVs groups and were correspondingly infused with exosomes derived from MLE-12 cells cultured under normoxic or CIH conditions, respectively, through the tail vein for 10 weeks (Fig. 3). In a subsequent experiment, four cohorts were established: Normoxia-EVs, CIH-EVs, CIH-EVs + NC inhibitor, and CIH-EVs + miR-106a-5p inhibitor. Animals in the Normoxia-EVs cohort received vesicles from normoxic MLE-12 cultures, whereas the remaining three cohorts received vesicles from CIH-exposed MLE-12 cultures; the latter two groups were additionally administered either a negative-control inhibitor or a miR-106a-5p-specific inhibitor, respectively. All treatments were sustained for 10 weeks (Fig. 9).

Fig. 1.

CIH caused cognitive impairment and hippocampal tissue damage in mice. A Experimental flow chart. B Weight of mice (n=6). C Results of cognitive tests. OFT, Y-maze and NOR (n=6). D and E HE staining of mouse brain tissues. Scale bar: 0.2 mm. F and G Nissl staining of mouse brain tissues. Scale bar: 0.2 mm. H and I Immunofluorescence staining of mouse brain tissues. Blue: DAPI; red: NeuN; green: Tunel, Scale bar: 0.2 mm. J RT-qPCR analysis of iNOS and Arg-1 mRNA levels in mouse brain tissues (n=6). K and L Western blotting was used to detect the expression levels of iNOS and Arg-1 in mouse brain tissues (n=3). M-P Immunofluorescence staining of mouse brain tissues. Blue: DAPI; red: iNOS or Arg-1; green: IBA-1, Scale bar: 0.2 mm. The data are presented as the mean (SD), *p < 0.05, **p < 0.01, ***p < 0.005 and ****p < 0.001. Statistical significance was determined by two-sided Student’s t-test as appropriate.

Fig. 3.

AEC-EVs mediated cognitive impairment and microglial polarization in mice under CIH. A Confocal image of BV2 cells incubated with PKH67-labeled EVs. Blue: DAPI. Red: Phalloidin. Green: EVs; scale bar: 20 μm. B RT-qPCR analysis of iNOS and Arg-1 mRNA levels in BV2 cells (n=3). C and D Western blotting was used to detect the expression levels of iNOS and Arg-1 in BV2 cells (n=3). E-H Immunofluorescence staining of BV2 cells. Blue: DAPI. Red: iNOS or Arg-1. Green: IBA-1; scale bar: 20 μm. I Experimental flow chart. J Results of cognitive tests. OFT, Y-maze and NOR (n=6). K HE staining of mouse brain tissues. Scale bar: 0.2 mm. L and M Nissl staining of mouse brain tissues. Scale bar: 0.2 mm. N and O Immunofluorescence staining of mouse brain tissues. Blue: DAPI; red: NeuN; green: Tunel, Scale bar: 0.2 mm. P RT-qPCR analysis of iNOS and Arg-1 mRNA levels in mouse brain tissues (n=6). Q Western blotting was used to detect the expression levels of iNOS and Arg-1 in mouse brain tissues (n=3). R-U Immunofluorescence staining of mouse brain tissues. Blue: DAPI; Red: iNOS or Arg-1. Green: IBA-1, Scale bar: 0.2 mm. The data are presented as the mean (SD), *p < 0.05, **p < 0.01, ***p < 0.005 and ****p < 0.001. Statistical significance was determined by one-way ANOVA or two-sided Student’s t-test as appropriate.

Fig. 9.

The effect of ACE-EVs was mediated by miR‐106a-5p in vivo. A Experimental flow chart. B Results of cognitive tests. OFT, Y-maze and NOR (n=6). C and D HE staining of mouse brain tissues. Scale bar: 0.2 mm. E and F Nissl staining of mouse brain tissues. Scale bar: 0.2 mm. G and H Immunofluorescence staining of mouse brain tissues. Blue: DAPI; red: NeuN; green: Tunel, Scale bar: 0.2 mm. I-L Immunofluorescence staining of mouse brain tissues. Blue: DAPI; red: iNOS or Arg-1. green: IBA-1, Scale bar: 0.2 mm. M and N RT-qPCR analysis of iNOS, CD86, CD206 and Arg-1 mRNA level in mouse brain tissues (n=3). O Western blotting was used to detect the expression levels of p-ERK and ERK in mouse brain tissues (n=6). The data are presented as the mean (SD), *p< 0.05, **p < 0.01, ***p < 0.005 and ****p < 0.001. Statistical significance was determined by one-way ANOVA or two-way ANOVA as appropriate

Immunofluorescence (IF)

Following intracardial perfusion with cold PBS, mouse brains were collected and fixed in 4% paraformaldehyde overnight, followed by cryoprotection in 30% sucrose solution for two days. Coronal brain Sect. (20 μm) were sectioned at −18℃ using a cryostat. Sections were blocked with 1% BSA for 40 min at 37℃, then incubated with primary antibodies overnight at 4℃. After PBS washes, sections were incubated with secondary antibodies for 2 h at room temperature (RT). Finally, nuclei were counterstained with DAPI, and images were acquired using fluorescence microscopy. Quantification was performed using ImageJ software. All antibodies used in this study are listed in Supplementary Table S1.

Hematoxylin and Eosin (HE) and Nissl staining

Following pretreatment, fresh brain tissues were collected and fixed in 4% paraformaldehyde. After dehydration, samples were paraffin-embedded, sectioned coronally, dewaxed, stained, and dehydrated. Images were captured under a microscope (Nikon, Tokyo, Japan) for subsequent analysis. Damaged neuronal cells undergo cell body contraction/swelling, nuclear condensation, disappearance of Nissl bodies, and appearance of deep red cytoplasm. As necrosis progresses, the nucleus dissolves and disappears, leaving only a faint outline called a ghost cell. Determine the degree of hippocampal tissue damage by calculating the proportion of normal morphology of neuronal cells [29].

Behavioral testing

Cognitive function was assessed using three established paradigms:

Open field test (OFT)

The mice were transported to the test room and acclimatized for 1 h before the experiment. Each mouse was then placed in an open field box and left to explore freely for 10 min. Finally, the movement of mice was automatically recorded by software for further analysis. The time spent on the central stage indicated the anxiety level of mice. After each mouse completed the experiment, the experimental site was washed with 75% alcohol to avoid the influence of odor [30].

Y Maze Test

The Y-maze consisted of three arms (A, B, and C) with an angle of 120° to each other. In the training phase, mice freely explored in the A and B arms. In the test phase, a new arm C was added, and the mice were free to explore for 5 min in three arms A, B, and C. Novel arm preference was then calculated based on the exploration time (E), using the formula: Novel arm preference = EC/(EC + EB + EA).

Novel Object Recognition Test (NOR)

The test has two phases. In the first stage, objects A1 and A2 (the same material and size) were placed in an open box in which the mouse moved freely for 10 min and video recording was performed. After 1 h of rest, the second stage was started. Object A2 is replaced by Object B of different shape. The discrimination index (DI) was then calculated based on the exploration time (E), using the formula: DI = (EB − EA1)/(EA1 + EB) [31].

All mice were euthanized after behavioral experiments and their brains were collected for subsequent biological analysis.

EVs isolation and characterization

The mouse brain tissue cells suspension was prepared using a fully automatic tissue processor (gentleMACS, Germany). Extracellular vesicles were extracted from the cell suspension collected through ultracentrifugation. The extracted exocrine body weight was suspended in PBS and stored at −80℃ for further use. The morphology of the isolated EVs was observed by transmission electron microscopy (TEM). The particle size distribution and concentration of extracellular vesicles were analyzed using nanoparticle tracking analysis (NTA). The protein biomarkers ALIX and TSG101 were extracted from EVs through western blotting detection.

The culture medium was collected from MLE-12 cells maintained under either normoxic (room air) or intermittent hypoxia (IH) conditions. Sequential centrifugation was performed at 500 × g for 30 min, 2000 × g for 30 min, and 10,000 × g for 30 min at 4 °C to remove cell debris and large particles. The resulting supernatant was transferred to an ultracentrifuge tube and subjected to centrifugation at 120,000 × g for 70 min at 4 °C. After discarding the supernatant, the pellet was gently resuspended in an appropriate volume of PBS. A final centrifugation step at 120,000 × g for 70 min yielded a pellet containing purified extracellular vesicles.

Cells

MLE-12 and BV-2 cells were purchased from Shanghai Zhong Qiao Xin Zhou Biotechnology (Shanghai, China). MLE-12 cells were cultured in DMEM/F-12 medium supplemented with 10% fetal bovine serum. BV-2 cells were cultured in DMEM high glucose medium supplemented with 10% fetal bovine serum. Suspended cell culture consumables were purchased from Bioland.

CIH model in Vitor

During each period of CIH, the oxygen fraction in the chamber was uniformly reduced from 21% to 1% over a 15 min period, and maintained at 1% for 10 min, then uniformly increased to 21% over a 15 min period, and maintained at 21% for 1 min. The whole process was cyclical, with a constant balance of 5% CO₂ and nitrogen [32].

Transfection

When the cell density reached 70%, BV2 cells were transfected with si-Dicer, miR-106a-5p mimics or miR-106a-5p inhibitor using Lipofectamine 3000. After 6 h, the medium was replaced with original medium and the transfected cells were used for further experiments after 48 h. The transfection efficiency was verified by western blotting analysis.

Immunofluorescence and confocal microscopy

The BV2 cells were fixed with 4% paraformaldehyde. After blocking with 5% BSA, the cells were incubated overnight at 4℃ with different primary antibodies. Then, the cells were washed three times in PBS and incubated with the corresponding secondary antibodies for 1 h. The cell nuclei were stained with DAPI. Finally, the cells were photographed using a laser confocal microscope (A1R HD25, Nikon, Japan) and semi-quantitative analysis was performed using ImageJ. The antibodies used in this study are listed in Supplementary Table S1.

Western blotting

After in vivo and in vitro experimental treatments, proteins were extracted from brain tissues and cells. The protein concentration was determined by using the BCA protein assay kit (Solarbio, Beijing, China, Cat: PC0020). Equal amounts of proteins were separated using 10% SDS polyacrylamide gels and then transferred to polyvinylidene difluoride membranes. Thereafter, the membrane was blocked with 5% BSA and then incubated with primary antibodies overnight at 4℃. The membrane was then incubated with secondary antibodies. A protein imaging system (ImageQuant800, cytiva, USA) was used to detect protein imaging, and ImageJ software (National Institutes of health, National Institutes of health, National Airlines) was used for quantitative analysis [33]. All antibodies used in this study are listed in Supplementary Table S1.

RNA isolation and RT‒PCR

Total RNA was extracted from brain tissue or cultured cells using the Fastagen RNA extraction kit (Shanghai, China), and cDNA was synthesized using the PrimeScript™ RT reagent Kit (TaKaRa, Tokyo, Japan), according to the manufacturer’s instructions. Subsequent RT-qPCR analysis was performed on a Light Cycler instrument (Bio-Rad, California, USA), employing the FastStart Essential DNA Green Master Kit (Roche, Basel, Switzerland). The expression data were normalized relative to the mRNA levels of GAPDH or the miRNA U6. The results were evaluated using the 2^−ΔΔCt method [33].

Dual luciferase reporter assay

293 T cells were seeded into 96 well plates at 70% confluence in triplicate, and the cells were transfected with dual luciferase reporter plasmids containing wild-type or mutated dusp2 sequences, as well as NC mimic or miR-106a-5p using Lipofectamine 2000 according to the manufacturer’s instructions. 48 h after transfection, firefly luciferase and Renilla luciferase signals were detected using a dual luciferase reporter kit (Solarbio, Beijing, China, Cat: D0011). Relative promoter activity was expressed as the ratio of firefly to Renilla luciferase activity.

Flow cytometry analysis

The cells were scraped from the culture dish and subsequently washed three times with cold PBS. According to the manufacturer’s recommended protocol, APC anti mouse CD86 and PE anti mouse CD206 antibodies were used for staining. Use CytExpert software (version 1.0.7; Tree Star, Ashland, OR, USA) for data collection and further analysis.

EVs uptake assay and co-culture

The extracted EVs were labeled with PKH67 (Beijing fluorescence Biotechnology Co., Ltd., PKH67), added to BV2 cells and cultured for 4 h, and observed by laser confocal microscope. BV2 cells were treated with EVs (20 µg/mL) for 24 h, followed by CIH exposure [33, 34].

Bioinformatics analysis

Expression profiles of EVs secreted by A549 cells under room air (RA) or intermittent hypoxia (IH) conditions were retrieved from the GEO dataset GSE181067. Differentially expressed genes (DEGs) were identified using the thresholds |fold change (FC)| >1.0 and p < 0.05.

Statistical analysis

All experiments were conducted as biological replicates, with sample size determined by statistical power requirements. Data are expressed as mean ± standard deviation (SD). All data passed the Shapiro-Wilk normality test (n < 50). Comparisons between groups were made using a t-test (two-sided). One-way and two-way analysis of variance (ANOVA) followed by Tukey’s post hoc tests were used for comparison of data from multiple groups, respectively. p < 0.05 was considered significant.

Results

CIH leads to cognitive dysfunction and microglial M1 polarization in mice

Previous studies reported significantly reduced food intake and body weight in CIH-exposed mice, suggesting that CIH may affect the CNS through altered nutrition [35]. To eliminate confounding effects from differential food consumption, we implemented a pair-feeding protocol in which both Normoxia and CIH groups received equal daily food/water per mouse throughout the study. Consequently, no significant weight difference was observed between groups during the modeling phase (Fig. 1 A and B). Behavioral analyses revealed that, relative to normoxic controls, mice exposed to CIH exhibited a significantly shorter duration in the central zone, reduced exploration of the novel arm, and diminished exploration of the novel object (Fig. 1 C).H&E staining revealed marked hippocampal injury in CIH-exposed mice (Fig. 1 D and E). Nissl staining and IF demonstrated extensive neuronal apoptosis in the hippocampus of the CIH group (Fig. 1 F-I). Compared with normoxic controls, CIH mice exhibited pronounced microglial M1 polarization, evidenced by increased M1 marker iNOS and decreased M2 marker Arg-1 on RT-qPCR, WB and IF (Fig. 1 J-P). Collectively, paired analysis of CIH-exposed mice indicates that CIH elicits cognitive impairment, hippocampal damage and microglial polarization.

CIH elevates the abundance of both AEC-EVs and microglia-derived EVs (MG-EVs) in murine brain tissue

To investigate the impact of CIH on brain-derived EVs, we isolated EVs from cerebral tissues of both Normoxia and CIH-exposed mice (Fig. 2 A). Subsequent characterization revealed that extracellular vesicles purified from murine brain exhibited characteristic cup-shaped morphology under transmission electron microscopy (TEM) (Fig. 2B). Nanoparticle tracking analysis (NTA) confirmed vesicle diameters within the 100–150 nm range across experimental groups (Fig. 2 C). Western blotting analysis validated EVs’ identity through positive detection of markers ALIX and TSG101 with concurrent absence of the negative control calnexin (Fig. 2D). Quantitative immunoblotting further showed that levels of neuron-derived EVs remained unchanged, whereas microglia-derived EVs (MG-EVs) and alveolar epithelial cell–derived EVs (AEC-EVs) were significantly elevated in CIH-exposed mice compared with normoxia controls (Fig. 2D and E).

Fig. 2.

CIH led to an increase in EVs derived from alveolar epithelial cells and microglia in mouse brain tissue. A Schematic diagram of EVs extraction process. B Representative TEM images of the EVs. C Measurement of the particle size of the EVs via NTA. D and E Western blotting detection of EVs markers ALIX, TSG101, and Calnexin, as well as microglial marker protein CD11b, alveolar epithelial marker protein SP-C, and neuronal marker protein Synaptophysin (n=3). The data are presented as the mean (SD), **p< 0.01. Statistical significance was determined by two-sided Student’s t-test as appropriate

CIH-MG-EVs can increase the M2 polarization level of BV2 cells

To investigate the impact of MG-EVs on microglial function, BV2 cells were cultured under Normoxia or CIH conditions followed by EV isolation from supernatants. Isolated EVs were then co-cultured with naïve BV2 cells for 24 h. Both qPCR and Western blotting analyses demonstrated that EVs from CIH-exposed BV2 cells significantly upregulated Arg-1 expression and enhanced M2 polarization in recipient microglia (Figure S1 A-C). Flow cytometry confirmed increased F4/80 + CD206 + double-positive cells indicating augmented M2 polarization following co-culture with CIH-derived EVs (Figure S1 D and S1E), with concordant findings observed via immunofluorescence (Figure S1 F-I).

Then, BV2 cells were divided into three experimental groups: Normoxia, CIH-exposed, and CIH-exposed with EVS secretion inhibitor GW4869 (MedChemExpress, Cat#HY-19363). Comparative analysis revealed that CIH-exposed microglia exhibited increased M1 polarization and reduced M2 polarization relative to Normoxia controls. Notably, GW4869 treatment further suppressed M2 polarization without altering M1 polarization (Figure S2). We co cultured MG-EVs with neuronal cells and astrocytes and found that CIH-MG-EVs did not directly cause neuronal cell apoptosis (Figure S3). These results collectively indicate that CIH-primed BV2 cells’ EVs autonomously enhance M2 polarization while minimally influencing M1 activation states.

CIH-AEC-EVs modulate microglial polarization

Subsequent investigation examined the impact of AEC-EVs on BV2 microglial polarization in vitro. MLE-12 cells were cultured under Normoxia or CIH conditions, with EVs isolated from supernatants and co-cultured with CIH-exposed BV2 cells. Both EVS groups were internalized by BV2 microglia (Fig. 3 A). BV2 cells were treated with PBS, Normoxic MLE-12 EVs, or CIH-exposed MLE-12 EVs during CIH exposure for 24 h. To eliminate the influence of EVs secreted by BV2 cells themselves, we added GW4869 in advance to BV2 cell culture medium. Molecular analyses revealed that while Normoxic EVs failed to alter iNOS or Arg-1 expression, CIH-exposed MLE-12 EVs significantly upregulated iNOS and suppressed Arg-1 in BV2 cells via RT-qPCR and Western blotting (Fig. 3 B-D), corroborated by immunofluorescence (Fig. 3 E-H). In addition, we co cultured AEC-EVs with neuronal cells and astrocytes and found that CIH-AEC-EVs did not directly cause neuronal cell apoptosis (Figure S3).

To corroborate our in-vitro observations, we conducted a complementary in-vivo experiment. We found that injection of PBS and Normoxia-EVs into the tail vein of mice did not affect their cognitive function (Figure S4A- S4C). Subsequently, mice received intravenous injections of Normoxia-EVs or CIH-EVs derived from MLE-12 cells (Fig. 3 I). Behavioral assessments demonstrated that, relative to Normoxia-EVs-treated controls, CIH- EVs -treated mice exhibited a shorter duration in the central zone, diminished exploration of the novel arm, and reduced investigation of the novel object (Fig. 3 J). Histopathological evaluation by H&E staining disclosed pronounced hippocampal damage in the CIH- EVs group (Fig. 3 K), while Nissl staining and immunofluorescence confirmed marked neuronal apoptosis (Fig. 3 L-O). The distribution of AEC-EVs in mouse brain tissue was tracked using Dir dye, and confocal microscopy revealed significant colocalization between AEC-EVs and microglia (Figure S4D). RT-qPCR and western blotting analyses of hippocampal tissue revealed a significant up-regulation of iNOS and a concurrent down-regulation of Arg-1 in the CIH-EVs cohort (Fig. 3 P and Q), findings that were corroborated by immunofluorescence (Fig. 3 R-U). Collectively, these data indicate that AEC-derived EVs generated under CIH are sufficient to elicit cognitive dysfunction, hippocampal injury, and a shift in microglial polarization in vivo.

CIH-AEC-EVs drive BV2 microglial polarization via MAPK signaling

Given the established role of the MAPK pathway in microglial polarization, we hypothesized that AEC-EVs-mediated effects on BV2 cells involve MAPK activation. As anticipated, both Western blotting and immunofluorescence analyses demonstrated that EVs derived from CIH-exposed MLE-12 cells—but not normoxic controls—elevated phosphorylated ERK (p-ERK) levels in BV2 cells (Fig. 4 A-D). However, CIH-AEC-EVs did not lead to an increase in phosphorylation levels of p38 and JNK (Figure S4F and S4G). Pharmacological inhibition of ERK phosphorylation using U0126(MedChemExpress, Cat#HY-19363) effectively abrogated this response (Fig. 4 E-H). Subsequent molecular profiling revealed that U0126 pretreatment suppressed the CIH-EVs-induced upregulation of iNOS and reversed Arg-1 downregulation (Fig. 4 I-K), with concordant findings observed via immunofluorescence (Fig. 4 L-O). These data collectively establish that CIH-primed alveolar epithelial EVs modulate BV2 polarization primarily through ERK/MAPK pathway activation.

Fig. 4.

AEC-EVs regulated microglial polarization by modulating ERK phosphorylation under CIH conditions. A and B Western blotting was used to detect the expression levels of ERK and p-ERK in BV2 cells (n=3). C and D Immunofluorescence detection of p-ERK expression in BV2 cells from different groups (n=3). Blue: DAPI; red: p-ERK, Scale bar: 20 μm. E and F Western blotting was used to detect the expression levels of ERK and p-ERK in BV2 cells (n=3). G and H Immunofluorescence detection of p-ERK expression in BV2 cells from different groups (n=3). Blue: DAPI; red: p-ERK, Scale bar: 20 μm. I and J Western blotting was used to detect the expression levels of iNOS and Arg-1 in BV2 cells (n=3). K RT-qPCR analysis of iNOS, CD86, CD206 and Arg-1 mRNA levels in BV2 cells (n=3). L-O Immunofluorescence detection of iNOS or Arg-1 expression in BV2 cells from different groups (n=3). Blue: DAPI; red: iNOS or Arg-1; green: IBA-1, Scale bar: 20 μm. The data are presented as the mean (SD), *p < 0.05, **p < 0.01 and ***p < 0.005. Statistical significance was determined by one-way ANOVA as appropriate

miR-106a-5p as a candidate MiRNA responsible for the effect of CIH-AEC-EVs

To determine whether miRNAs serve as critical mediators in CIH-AEC-EVs, miRNA-depleted MLE-12 -derived EVs were generated by means of siRNA-mediated knockdown of Dicer, an endoribonuclease essential for miRNA maturation (Fig. 5 A-C). BV2 cells were co-cultured with four EVs groups: Normoxia-EVs, CIH-EVs, CIH-EVs + NC siRNA, and CIH-EVs + Dicer siRNA. miRNA depletion significantly attenuated CIH-EVs-induced MAPK pathway activation (Fig. 5 D-G) and concurrently reversed the CIH-EVs-driven upregulation of iNOS mRNA and protein with concomitant suppression of Arg-1 expression (Fig. 5 H-J). Flow cytometry further demonstrated diminished CD86⁺F4/80⁺ double-positive cells and restored CD206⁺F4/80⁺ populations (Fig. 5 K and L), with immunofluorescence corroborating these polarization shifts (Fig. 5 M-P). These results establish that miRNAs within alveolar epithelial CIH-EVs govern MAPK activation and drive microglial M1/M2 imbalance.

Fig. 5.

The biological effect of AEC-EVs was partly mediated by miRNAs. A and B Western blotting was used to detect the expression level of Dicer in MLE-12 cells (n=3). C RT-qPCR analysis of Dicer mRNA level in MLE-12 cells (n=3). D and E Western blotting was used to detect the expression levels of ERK and p-ERK in BV2 cells (n=3). F and G Immunofluorescence detection of p-ERK expression in BV2 cells from different groups (n=3). Blue: DAPI; red: p-ERK, Scale bar: 20 μm. H and I Western blotting was used to detect the expression levels of iNOS and Arg-1 in BV2 cells (n=3). J RT-qPCR analysis of iNOS, CD86, CD206 and Arg-1 mRNA levels in BV2 cells (n=3). K and L Flow cytometry analysis of the proportion of M1 macrophages (CD86⁺F4/80⁺) and M2 macrophages (CD206⁺ F4/80⁺) in BV2 cells (n=3). M-P Immunofluorescence detection of iNOS or Arg-1 expression in BV2 cells from different groups (n=3). Blue: DAPI; red: iNOS or Arg-1; green: IBA-1, Scale bar: 20 μm. The data are presented as the mean (SD), *p < 0.05, **p < 0.01, ***p < 0.005 and ****p < 0.001. Statistical significance was determined by one-way ANOVA or two-sided Student’s t-test as appropriate

We next identified specific miRNAs responsible for CIH-EVs bioactivity. Analysis of the GEO dataset GSE181067—profiling EVs’ miRNAs from A549 cells under Normoxia versus intermittent hypoxia (IH)—revealed 87 upregulated and 23 downregulated miRNAs in CIH-EVs (Fig. 6 A). Functional enrichment analysis linked these differentially expressed miRNAs to neuroregulatory processes, gliogenesis, and hippocampal development (Fig. 6 B). KEGG pathway analysis identified MAPK signaling as the most significantly enriched pathway (Fig. 6 C), consistent with our mechanistic findings. Protein-protein interaction (PPI) network mapping further delineated miRNA-target gene relationships within the MAPK pathway (Fig. 6 D), indicating that CIH-AEC-EVs likely modulated MAPK signaling through these candidate miRNAs.

Fig. 6.

The identification of potentially effective candidate miRNAs in AEC-EVs. A EVs miRNA sequencing results from the GSE181067. B GO enrichment analysis. C KEGG enrichment analysis. D Protein-protein interaction (PPI) network mapping. E RT-qPCR analysis of miRNA levels in EVs (n=3). F and G RT-qPCR analysis of miRNA levels in BV2 cells (n=3). H The uptake of the Cy5‐labelled miR‐106a-5p EVs or miR‐106a-5p EVs were visualized by immunofluorescence analysis in BV2 cells after incubation for 24h. Blue: DAPI; red: CY-5; green: PKH67, Scale bar: 20 μm. The data are presented as the mean (SD), *p< 0.05 and **p < 0.01. Statistical significance was determined by one-way ANOVA, two-way ANOVA or two-sided Student’s t-test as appropriate

To pinpoint the active miRNAs, we identified six upregulated DE-miRNAs (miR-3620-5p, miR-2861, miR-762, miR-1231-5p, miR-1231-3p, and miR-106a-5p) that were highly expressed, had large-fold changes, were statistically significant, or closely related to the MAPK signaling pathway. RT-qPCR showed that compared to Normoxia-EVs, these six miRNAs were upregulated in CIH-EVs (Fig. 6E).

In BV2 cells co-cultured with CIH-EVs, only miR-106a-5p was significantly upregulated (Fig. 6 F) in a dose-dependent manner (Fig. 6G). In addition, the content of miR-106a-5p in mouse plasma EVs also increased under CIH environment (Figure S4E). To confirm whether the increased miR-106a-5p levels in BV2 cells were directly due to CIH-EVs transfer, we transfected MLE-12 cells with miR-106a-5p or Cy5-labeled miR-106a-5p. We then collected conditioned media from these cells to isolate two EVs types. After adding these EVs to BV2 cell cultures, we detected Cy5 signals only in BV2 cells incubated with Cy5-labeled miR-106a-5p-EVs using confocal microscopy (Fig. 6H). This indicates that AEC-EVs mediate miR-106a-5p transfer. In summary, these data suggest that AEC-EVs mediated miR-106a-5p shuttling.

MiR-106a-5p targets DUSP2 to regulate the MAPK signaling pathway

DUSP2, which controlled ERK dephosphorylation and was implicated in multiple disorders [36–39], was predicted by database analysis to be a target of miR-106a-5p, with the binding site identified (Fig. 7 A). The findings demonstrated that miR-106a-5p significantly reduced the luciferase activity of the wild-type 3ʹ UTR of DUSP2, while it had no inhibitory effect on the mutant 3ʹ UTR of DUSP2 (Fig. 7B). Transfection of BV2 cells with miR-106a-5p mimics reduced DUSP2 mRNA expression, as detected by RT-qPCR (Fig. 7 C). Further validation using Western blotting and immunofluorescence showed that miR-106a-5p mimics also decreased DUSP2 protein levels (Fig. 7D-G) and increased ERK phosphorylation (Fig. 7H-K) in BV2 cells treated with LPS. These findings indicate that miR-106a-5p inhibits DUSP2 expression by directly targeting the 3ʹ UTR of DUSP2 mRNA.

Fig. 7.

MiR-106a-5p regulated the ERK MAPK signaling pathway by targeting DUSP2. A Prediction of DUSP2 regulated targets by miR-106a-5p. B Dual luciferase reporter assay (n=3). C RT-qPCR analysis of DUSP2 mRNA level in BV2 cells (n=3). D and E Western blotting was used to detect the expression level of DUSP2 in BV2 cells (n=3). F and G Immunofluorescence detection of DUSP2 expression in BV2 cells from different groups (n=3). Blue: DAPI; red: DUSP2, Scale bar: 20 μm. H and I Western blotting was used to detect the expression levels of ERK and p-ERK in BV2 cells (n=3). J and K Immunofluorescence detection of p-ERK expression in BV2 cells from different groups (n=3). Blue: DAPI; red: p-ERK, Scale bar: 20 μm. The data are presented as the mean (SD), *p< 0.05, **p < 0.01, ***p < 0.005 and ****p < 0.001. Statistical significance was determined by one-way ANOVA or two-sided Student’s t-test as appropriate

CIH-AEC-EVs mediates the polarization of microglia by miR-106a-5p in vivo and in vitro

MLE-12 cells under CIH were transfected with a miR-106a-5p inhibitor or NC inhibitor (Fig. 8 A) to generate miR-106a-5p inhibitor-CIH-EVs and NC inhibitor-CIH-EVs for in vitro validation (Fig. 8 B). As expected, compared to CIH-EVs or NC inhibitor-CIH-EVs treatment, miR-106a-5p inhibitor-CIH-EVs treatment partially abrogated the effects of CIH-EVs, as evidenced by increased DUSP2 protein levels (Fig. 8C-F) and reduced ERK phosphorylation (Fig. 8 C, D and G, and 8H). Additionally, the miR-106a-5p inhibitor-CIH-EVs treatment partially inhibited the increase in iNOS protein and mRNA expression and the decrease in Arg-1 protein and mRNA expression (Fig. 8I-O).

Fig. 8.

ACE-EVs exerted microglial polarization through miR‐106a-5p in vitro. A RT-qPCR analysis of miR-106a-5p levels in BV2 cells (n=3). B RT-qPCR analysis of miR-106a-5p levels in EVs (n=3). C and D Western blotting was used to detect the expression levels of DUSP2, p-ERK and ERK in BV2 cells (n=3). E and F Immunofluorescence detection of DUSP2 expression in BV2 cells from different groups (n=3). Blue: DAPI; red: DUSP2, Scale bar: 20 μm. G and H Immunofluorescence detection of p-ERK expression in BV2 cells from different groups (n=3). Blue: DAPI; red: p-ERK, Scale bar: 20 μm. I and J Western blotting was used to detect the expression levels of iNOS and Arg-1 in BV2 cells (n=3). K-N Immunofluorescence detection of iNOS or Arg-1 expression in BV2 cells from different groups (n=3). Blue: DAPI; red: iNOS or Arg-1; green: IBA-1, Scale bar: 20 μm. O RT-qPCR analysis of iNOS, CD86, CD206 and Arg-1 mRNA level in BV2 cells (n=3). The data are presented as the mean (SD), *p< 0.05, **p < 0.01, ***p < 0.005 and ****p < 0.001. Statistical significance was determined by one-way ANOVA or two-sided Student’s t-test as appropriate

To verify these in vitro findings, in vivo experiments were conducted to examine whether introducing a miR-106a-5p inhibitor could partially abrogate the effects of CIH-AEC-EVs on cognitive dysfunction in mice (Fig. 9 A). Behavioral analyses showed that, compared to mice in the CIH-EVs + NC inhibitor group, those in the CIH-EVs + miR-106a-5p inhibitor group spent more time in the central zone, exhibited greater exploration of the novel arm and showed more investigation of the novel object (Fig. 9 B). H&E staining revealed attenuated hippocampal injury in the CIH-EVs + miR-106a-5p inhibitor group (Fig. 9 C and D). Nissl staining and immunofluorescence also demonstrated reduced neuronal apoptosis in the hippocampus of this group (Fig. 9 E-H). Moreover, mice in the CIH-EVs + miR-106a-5p inhibitor group displayed diminished microglial M1 polarization. As shown by RT-qPCR and immunofluorescence, iNOS expression decreased while Arg-1 expression increased (Fig. 9 I-N). Western blotting results further indicated upregulated DUSP2 protein expression and downregulated ERK phosphorylation in the CIH-EVs + miR-106a-5p inhibitor group (Fig. 9 O). These data suggest that EVs secreted by alveolar epithelial cells in CIH mice induce polarization of microglia both in vivo and in vitro, partially through miR-106a-5p.

Discussion

This study explores how chronic intermittent hypoxia (CIH) induces cognitive impairment through the lung - brain axis. We innovatively propose that AEC-EVs may be a key mediator linking peripheral CIH stimulation with central microglial polarization imbalance. The miR-106a-5p in AEC-EVs targets DUSP2, releasing MAPK pathway inhibition. This drives microglial M1 polarization and suppresses M2 polarization, causing hippocampal neuron apoptosis and cognitive decline. Using CIH mouse models, EVs tracing, miRNA depletion/replenishment, and dual-luciferase reporter assays, we identified the molecular axis: AEC-EVs - miR-106a-5p - DUSP2 - MAPK - microglial polarization.

The lung-brain axis, a recently proposed interorgan communication network, refers to the bidirectional information exchange between the lung and the central nervous system (CNS) through multiple pathways, including neural, immune, endocrine, and microbiota-metabolite routes, thereby jointly regulating physiological homeostasis and disease progression [40]. Similar to the traditional “gut-brain axis,” the LBA emphasizes that peripheral lung tissue is not only regulated by the brain but can also actively influence brain function through soluble mediators, EVs, the vagus nerve, and microbiota metabolites. For instance, acute lung injury (ALI) or mechanical ventilation can increase norepinephrine in the brain via the sympathetic-adrenomedullary axis, triggering neurogenic pulmonary edema or cognitive impairment [41].Pulmonary infections or chronic obstructive pulmonary disease release pro-inflammatory cytokines such as IL-1β, TNF-α, and IL-6. These cytokines can enter the brain parenchyma through weak areas of the blood-brain barrier (e.g., the circumventricular organs), activate microglia, and lead to depression and memory decline [42].

A 2022 Nature study first confirmed that pulmonary commensal bacteria can reprogram microglia to a high interferon type I - expressing state via LPS, thereby suppressing the autoimmune response in experimental autoimmune encephalomyelitis (EAE) [43].Dysbiosis of the lung microbiota, caused by factors such as smoking or antibiotic use, reduced short - chain fatty acids (SCFAs), decreased tight junction proteins of the blood - brain barrier (e.g., ZO-1 and Claudin-5), and exacerbated neuroinflammation [44]. Additionally, studies have found that intravenously administered mesenchymal stem cells (MSCs) engraft in the lung and activate 5 - HT neurons via the vagus nerve - nucleus tractus solitarius - dorsal raphe nucleus pathway, thereby improving depressive-like behaviors in mice [45].

Most research on CIH-induced cognitive impairment focuses directly on brain tissue. For example, prolonged intermittent hypoxia alters orexin and its receptor expression in mice brains, inducing anxiety-like behavior [46]. CIH and amyloid proteins can impact hippocampal activity, cholinergic regulation, hippocampal lipid metabolism, and memory [47, 48]. CIH also facilitates organ crosstalk. For example, CIH induces pulmonary vascular endothelial EVs to promote atherosclerosis. CIH-induced epididymal remodeling is associated with early myocardial interstitial fibrosis [49]. In our study, CIH causes cognitive dysfunction in mice as AECs secrete EVs to the hippocampus, driving microglial polarization. This shows LBA crosstalk, with EVs playing a key role. Many prior studies also prove that EVs are crucial in lung - brain communication [21, 50, 51].

In this study, treating BV2 cells in a CIH-exposed environment with the EVs inhibitor GW4869 reduced M2 polarization (Figure S2). This implies that BV2 cell-derived EVs in CIH enhance M2 polarization, possibly as a form of cellular self-protection. This is consistent with prior research showing that hypoxic microglial EVs promote M2 polarization [52]. Thus, the increased M1 and decreased M2 polarization in hippocampal microglia of CIH-exposed mice is not due to autocrine EVs, but likely results from exogenous EVs (e.g., AEC-EVs).

Previous studies have indicated that miR-106a-5p is linked to neuroinflammation. In the serum and cerebrospinal fluid of patients with acute cerebral infarction, miR-106a-5p exhibits significant upregulation, and its expression levels in these fluids are positively correlated with pro-inflammatory cytokines in such patients. A reduction in miR-106a-5p inhibits the inflammation of BV-2 cells induced by oxygen-glucose deprivation [53]. This finding aligns with the results of the present study, which demonstrated that CIH-AEC-EVs carrying miR-106a-5p to the hippocampus resulted in increased M1 polarization of microglia in hippocampal tissue and the onset of neuroinflammation. Furthermore, research has revealed that miR-106a-5p can activate the MAPK signaling pathway by mediating the downregulation of RPS6KA2 and upregulating the expression of MAPK14 in glioblastoma and ovarian cancer [54, 55]. Unlike this study, previous research has highlighted miR-106a-5p’s role in the P38 MAPK signaling pathway, whereas this study emphasizes the impact of miR-106a-5p on ERK phosphorylation following its interaction with DUSP2. This expands our comprehension of the mechanisms through which miR-106a-5p influences the MAPK signaling pathway.

Previous studies have shown that DUSP2 has the ability to simultaneously regulate the phosphorylation of ERK, p38, and JNK [39, 56, 57]. The strength of its effect depends on cell type, subcellular localization, and stimulation conditions. In our study, we found that CIH-EVs showed an increase in ERK phosphorylation levels, but had no significant effect on p38 and JNK phosphorylation levels. This indicates that the regulation of microglia by CIH-AEC-EVs not only depends on DUSP2 but also on other mechanisms, which requires further in-depth research in the future.

In summary, our study establishes that alveolar epithelial cells remotely regulate CNS immune homeostasis via EVs’ miRNAs within the LBA. AEC-EVs and miR-106a-5p/DUSP2 axis play important roles in it.

Moreover, this study offers fresh insights into the “periphery-to-center” mechanisms underlying OSA-related cognitive impairment and suggests a potential non-invasive therapeutic avenue for patients with OSA-induced cognitive dysfunction, especially for those who cannot tolerate continuous positive airway pressure.

Our study has several limitations. First, we did not differentiate the roles of various AEC-EVs subpopulations, like micro vesicles and apoptotic bodies. Second, while we highlighted miR-106a-5p’s role in CIH-AEC-EVs-induced cognitive dysfunction, our study did not assess the potential contributions of non-miRNA cargo, such as proteins, mRNAs, lipids, or DNA fragments, which are also functionally active components of EVs. These molecules may independently or synergistically influence microglial polarization, neuroinflammation, or synaptic function. For instance, exosomal proteins like cytokines or enzymes could directly modulate signaling pathways in recipient cells, while mRNA transcripts might be translated into biologically active proteins post-uptake. Future studies could employ multi-omics approaches (e.g., proteomics, transcriptomics, lipidomics) to comprehensively characterize the full cargo spectrum of CIH-AEC-EVs and elucidate the integrated role of both miRNA and non-miRNA components in mediating neuroimmune responses. Prior studies indicated that CIH affected brain regions differently in male and female rats [58, 59], but we only used male mice. Although our focus was on AEC-EVs’ impact on brain tissue, sex-based differences might exist and should be explored in future studies to determine if they influence CIH-AEC-EVs effect. In addition, the results of this study showed that CIH-MG-EVs can induce M2 polarization in BV2 cells in vitro (Figures S1 and S2). However, the overall animal level showed CIH driven polarization of mouse microglia towards M1 type. This phenomenon suggests that under CIH stress, microglia may attempt to limit inflammatory damage and protect themselves by releasing protective EVs, but their effect is much weaker than the comprehensive damage caused by AEC-EVs and other pro-inflammatory signals, ultimately leading to M1 polarization dominance. The exact molecular mechanism needs further clarification through subsequent research. Furthermore, the M1/M2 classification of microglia is relatively clear in in vitro stimulation models; however, in vivo, there is rarely a pure M1 or M2 state. Instead, microglia exhibit mixed or transitional phenotypes, such as M1, M2a, M2b, and M2c. With the advancement of technologies like single-cell sequencing and spatial transcriptomics, research has progressively unveiled the high degree of heterogeneity and dynamic plasticity of microglia under both physiological and pathological conditions, along with the identification of novel microglia phenotypes, including Disease-Associated Microglia, Interferon-Response Microglia, Proliferative-Region-Associated Microglia, and Neurodegenerative Microglial Signature. The primary focus of this study is to explore the M1/M2 polarization types of microglia, and subsequent research should delve into whether different types of microglia exert distinct functions. Furthermore, the lungs are composed of various types of cells, such as alveolar epithelial cells, pulmonary vascular endothelial cells, alveolar macrophages, and other types of cells. This study mainly explores the effect of extracellular vesicles secreted by alveolar epithelial cells on cognitive dysfunction caused by CIH. Further investigation should be conducted to determine whether other types of cells in lung tissue also play a role. And this study only examined type 2 alveolar epithelial cells (AT2) without considering type 1 alveolar epithelial cells (AT1). Current research shows that type 2 alveolar epithelial cells have a stronger ability to secrete extracellular vesicles under physiological and pathological conditions. Multiple studies have clearly indicated that AT2 cells are the main source of extracellular vesicle secretion in alveolar epithelium, especially in disease models such as acute lung injury [60, 61], sepsis [62], and hypoxia [63]. Extracellular vesicles secreted by AT2 cells have been widely isolated and studied. In contrast, there is very little research on the secretion of extracellular vesicles by type 1 alveolar epithelial cells, and there is currently no clear evidence to suggest that AT1 cells secrete large amounts of extracellular vesicles under physiological or pathological conditions, or that their extracellular vesicles play an important role in diseases. Therefore, this study mainly focuses on type II alveolar epithelial cells. In the future, we will conduct further exploration to distinguish the effects of type I and type II alveolar epithelial cells on CIH mediated cognitive dysfunction. In addition, previous studies have shown that behavioral experiments such as open field, Y-maze, and novel object tests can reflect cognitive dysfunction in mice [64, 65], but subsequent studies have shown that they can also reflect anxiety and motor changes in mice. In subsequent studies, the anxiety level of mice can be evaluated by adding elevated plus maze or light dark box experiments to eliminate anxiety interference, and the interference of motor ability can be eliminated by recording the total distance traveled by mice. Moreover, mouse models don’t fully replicate human OSA hypoxia patterns, so clinical sample validation is necessary. For instance, we could collect BALF, plasma, or CSF from OSA patients to extract EVs and examine the link between EVs’ miR-106a-5p and cognitive issues [26]. Lastly, our exploration of the synergistic effects of the MAPK pathway (e.g., p38/JNK) downstream of DUSP2 and neuroglial interactions was superficial. Future studies should integrate single-cell sequencing with EVs analysis of CSF from OSA patients to refine treatment targets.

Conclusion

In this study, we demonstrated that CIH-induced cognitive impairment in mice was mediated by the lung-brain axis. The specific mechanism involved miR-106a-5p, which suppressed DUSP2 expression, thereby reducing DUSP2’s inhibitory effect on the MAPK signaling pathway. CIH increased the levels of miR-106a-5p in AEC-EVs. These EVs entered the systemic circulation and released miR-106a-5p in the hippocampus. This reduced DUSP2 expression in microglia, enhanced MAPK signaling, and skewed microglial polarization toward M1 and away from M2, leading to hippocampal injury and cognitive dysfunction. Our findings suggest a potential pathway for treating or preventing cognitive impairment in patients with OSA.

Abbreviations

OSA

Obstructive sleep apnea

CIH

Chronic intermittent hypoxia

AEC-EVs

EVs produced by alveolar epithelial cells

CIH-AEC-EVs

EVs produced by alveolar epithelial cells in chronic intermittent hypoxia environment

MG-EVs

Microglia-derived EVs

qPCR

Quantitative Real-Time PCR

WB

Western Blotting

IF

Immunofluorescence

BBB

The blood-brain barrier

CNS

Central nervous system

AD

Alzheimer’s disease

TBI

Traumatic brain injury

RT

Room temperature

HE Staining

Hematoxylin and Eosin Staining

OFT

Open field test

NOR

Novel Object Recognition Test

TEM

Transmission electron microscopy

NTA

Nanoparticle tracking analysis

RA

Room air

IH

Intermittent hypoxia

DEGs

Differentially expressed genes

SD

Standard deviation

LBA

Lung-brain axis

ALI

Acute lung injury

Authors’ contributions

**B.W.** Conceptualization, Data curation, Methodology, Writing - original draft, Writing - review & editing, Visualization. **C.S.** Conceptualization, Data curation, Methodology, Writing - original draft, Writing - review & editing. **R.Z.** Investigation, Software, Methodology. **A.G.** Investigation, Validation. **M.Z.** Investigation, Funding acquisition. **X.Z.** Supervision, Funding acquisition. **C.G.** Conceptualization, Supervision, Writing - review& editing. All authors reviewed the manuscript.

Funding

This work was supported by the State Key Laboratory of Drug Regulatory Science (2025SKLDRS0326).

Data availability

Data is provided within the manuscript or supplementary information files.

Declarations

Ethics approval and consent to participate

All the animal experiments were approved by the ethics review committee of Shandong First Medical University (W202507040811).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.