Esketamine alleviates COPD-depression comorbidity in rats via MAPK/NF-κB inhibition and gut-lung-brain axis modulation

Abstract

Background

Chronic obstructive pulmonary disease (COPD) and depression frequently co-occur, yet the biological basis of this comorbidity and effective therapeutic strategies remain poorly defined.

Methods

We established a rat model of COPD-depression comorbidity through sequential cigarette-smoke exposure and chronic unpredictable mild stress. Pulmonary function, depression-like behaviors, histopathology, and MAPK/NF-κB signaling in lung and hippocampus were assessed. Esketamine or esketamine plus the TLR1/2 agonist Diprovocim was administered for 14 days. Cytokines, oxidative-stress markers, neuronal apoptosis, and microglial activation were evaluated. Complementary in-vitro studies used NR8383 alveolar macrophages (CSE model) and HAPI microglia (LPS + CSE). Gut and lung microbiota were profiled by 16 S rRNA sequencing and correlated with physiological and inflammatory indices.

Results

Comorbid rats displayed airflow limitation, depression-like behaviors, systemic inflammation, oxidative stress, and MAPK/NF-κB activation. Esketamine improved pulmonary function and behavior, reduced neuronal apoptosis and microglial activation, and suppressed MAPK/NF-κB signaling; these effects were partly reversed by Diprovocim. In vitro, esketamine increased macrophage and microglial viability, lowered proinflammatory cytokines and oxidative markers, and inhibited pathway activation. Microbiota profiling showed dysbiosis of gut and lung communities, with loss of beneficial taxa and expansion of conditional pathogens, whereas esketamine partially restored balance by promoting commensals and reducing potential pathogens.

Conclusions

These findings delineate a gut-lung-brain inflammatory-microbial network in COPD-depression comorbidity and identify esketamine as a multi-target intervention capable of modulating signaling pathways, inflammation and oxidative stress, and microbial homeostasis.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-026-03699-1.

Introduction

Chronic obstructive pulmonary disease (COPD) is a progressive respiratory disorder defined by persistent airflow limitation and is associated with high prevalence, disability, and mortality worldwide [1]. Depression, a common psychiatric illness characterized by persistent low mood, cognitive impairment, and somatic symptoms, markedly reduces quality of life and social functioning [2]. According to the Global Burden of Disease (GBD) study, both COPD and depression are major contributors to morbidity and disability globally, with disease burden continuing to rise [3, 4]. Epidemiologic studies highlight a strong comorbidity between the two conditions. Meta-analyses estimate that depression affects approximately 34.5% of patients with COPD—3.5 times the prevalence in those without COPD [5]. In China, data from the Health and Retirement Longitudinal Study reported depressive symptoms in nearly half of COPD patients [6]. Importantly, comorbid depression worsens respiratory symptoms, increases the risk of acute exacerbations, prolongs hospitalization, reduces quality of life, and elevates mortality [7], posing a critical challenge to prognosis and health care delivery.

Although COPD and depression are traditionally classified as respiratory and neuropsychiatric disorders, respectively, growing evidence points to shared pathophysiological mechanisms. Chronic systemic inflammation, oxidative stress, and immune dysregulation may represent a common foundation [8, 9]. In COPD, chronic hypoxia and inflammation can disrupt central monoamine metabolism and neuroendocrine function, promoting depressive symptoms [6, 10]. Conversely, depression is often accompanied by neurotransmitter imbalance, hyperactivation of the hypothalamic-pituitary-adrenal (HPA) axis, and impaired neuroimmune regulation [11], which in turn can enhance pulmonary inflammation and barrier dysfunction, accelerating COPD progression [12, 13]. These findings suggest a bidirectional pathological loop between COPD and depression, mediated by inflammatory, endocrine, and neuroimmune pathways. Yet the precise mechanisms and therapeutic targets remain unclear.

Beyond these established mechanisms, the concept of the gut-lung-brain axis offers a new perspective on cross-organ interactions in chronic disease [14]. The gut microbiota is essential for immune homeostasis, short-chain fatty acid metabolism, neurotransmitter synthesis, and inflammatory regulation [15–17]. Dysbiosis has been linked to both depression [18] and pulmonary disease [19]. Similarly, the lung microbiota, now recognized as a distinct ecosystem, contributes to chronic airway disorders, where pathogenic overgrowth and depletion of protective species shape local and systemic inflammation [20]. Evidence suggests that lung and gut microbiota interact through vagal, inflammatory, and metabolic pathways, forming a cross-organ microbial network [21, 22]. Disruption of this network may impair barrier integrity, activate systemic inflammation, and disturb neural function, thereby aggravating COPD-depression comorbidity. However, systematic studies of lung-gut microbial alterations in this context are lacking.

Given the multifactorial nature of COPD-depression comorbidity, therapies targeting single systems are insufficient. Interventions capable of modulating inflammation, neural activity, and microbiota simultaneously are urgently needed. The N-methyl-D-aspartate receptor (NMDAR), expressed in both the central nervous system and pulmonary immune cells, has emerged as a potential cross-system target [23–25]. In the brain, NMDAR overactivation contributes to excitotoxicity, impaired synaptic plasticity, and depression [26]. In the lung, it promotes cytokine release and tissue injury [27]. Thus, modulating NMDAR activity may offer a unified strategy to alleviate both respiratory and neuropsychiatric pathology [28].

Esketamine, the S-enantiomer of ketamine [29], is a noncompetitive NMDAR antagonist widely used for anesthesia and analgesia [30]. It has recently gained attention for its rapid and robust antidepressant effects [31, 32], mediated by enhanced hippocampal synaptic plasticity, restoration of brain-derived neurotrophic factor (BDNF), and regulation of mTOR and AKT signaling [33]. Preclinical evidence also indicates that esketamine reduces inflammation and oxidative stress. In models of lung injury, it attenuated inflammatory signaling, mitigated epithelial damage, and improved pulmonary function [34, 35]. Moreover, esketamine has been shown to alter gut microbiota composition and function, potentially alleviating neuropsychiatric dysfunction through microbial-neuroimmune pathways [36]. However, its effects in COPD-depression comorbidity, particularly on lung-gut microbial interactions, have not been systematically explored.

To address this gap, we combined animal modeling, cellular experiments, and microbiota sequencing to investigate the mechanistic interplay between COPD and depression and to evaluate the therapeutic potential of esketamine. We established and validated a comorbid COPD-depression model, characterized inflammatory signaling pathways, assessed esketamine’s effects on pulmonary function, behavioral phenotypes, and systemic inflammation, and further examined its actions in alveolar macrophages and microglia. Finally, we profiled lung and gut microbiota alterations and esketamine’s modulatory effects. Together, these investigations provide mechanistic insights into COPD-depression comorbidity and identify esketamine as a candidate therapy targeting inflammation, neuroregulation, and the gut-lung-brain axis.

Methods

Animals and ethics

Male specific-pathogen-free Sprague-Dawley rats (8 weeks; 200–250 g; SPF [Beijing] Biotechnology, China) were housed in the Fujian Medical University animal facility under controlled temperature (20–22 °C), relative humidity (≈ 50%), and a 12-h light/dark cycle, with autoclaved chow and water ad libitum. All procedures conformed to institutional guidelines and were approved by the Ethics Committee of The Second Affiliated Hospital of Fujian Medical University (Approval No. 2025 − 198).

Experimental design, modeling, and treatment

Rats were randomized (n = 7 per group) to control (Con), COPD only (COPD), depression only (YY), or comorbidity (CY). COPD was induced by whole-body cigarette-smoke exposure twice daily (30 min per session) for 16 weeks. A sealed, custom chamber (transparent polyethylene storage box with lid) delivered smoke from lit cigarettes via a syringe connected to a three-way valve and tubing through an inlet port; a second vent port permitted controlled aeration. The system provided stable sealing and real-time observation; after smoke delivery, the lid was opened and the chamber was rapidly ventilated. The target chamber concentration was approximately 5% by volume (smoke volume/box volume ≈ 1/20) for each exposure (Patent No. CN210784835U). Commercial cigarettes (Taishan, “White General,” China) were used for all exposures.

YY rats underwent chronic unpredictable mild stress (CUMS) during weeks 13–16. CY rats received 12 weeks of smoke exposure followed by concurrent smoke plus CUMS during weeks 13–16 to model respiratory-psychiatric comorbidity. CUMS consisted of one stressor per day, applied in an unpredictable order to sustain low-intensity chronic stress: 24-h food deprivation, 24-h water deprivation, 45° cage tilt (24 h), ice-water swim (5 min), tail pinch (2 min), wet bedding (24 h), or reversed light/dark cycle. Noninvasive pulmonary function was assessed every 4 weeks during modeling; at week 16, behavioral testing was performed, followed immediately by invasive pulmonary mechanics and tissue harvest for downstream assays.

For pharmacologic studies, successfully modeled CY rats were re-randomized (n = 7 per group) to vehicle (CY), esketamine (CYK), or esketamine plus Diprovocim (CYKD). These animals belonged to one original cohort that completed the 16-week modeling concurrently; no separate animal cohorts were used. Esketamine (2 mL/50 mg; Batch No. H20213735; Jiangsu Hengrui Pharmaceuticals, China) was administered intraperitoneally at 10 mg/kg once daily. Diprovocim (5 mg; Cat. No. HY-123942; MedChemExpress, USA) was administered intraperitoneally at 1 mg/kg every 3 days. Interventions lasted for 14 days. The 14-day treatment duration was selected based on prior evidence showing that two weeks of daily esketamine produces stable anti-inflammatory and antidepressant-like effects in rodent models of neuroinflammation and stress-induced depression [37]. Age-matched Con rats (n = 7) received neither modeling nor drug. Behavioral tests and noninvasive pulmonary function were performed at week 16 (pre-treatment) and week 18 (post-treatment); invasive mechanics and terminal sampling followed the final assessments. For microbiota profiling, fecal pellets and lung tissue samples for rRNA sequencing were collected at the terminal sampling point (week 18), immediately after completion of post-treatment behavioral and pulmonary function assessments.

Pulmonary function testing

Conscious rats were evaluated by whole-body plethysmography (WBP; Buxco). After acclimation to stable breathing, 5-min recordings were obtained and artifact-free segments were analyzed for expiratory flow at 50% of tidal volume (EF50) and peak expiratory flow (PEF). For invasive mechanics, rats were anesthetized with pentobarbital (0.3% w/v; 1 mL/100 g intraperitoneally), tracheostomized, cannulated with a dual-lumen tube, and connected to a Buxco invasive system. After stabilization of pressure and flow traces, forced vital capacity (FVC) and the FEV_0.1/FVC ratio were recorded according to manufacturer’s recommendations.

Behavioral assays

For the sucrose preference test (SPT), rats underwent 48 h of adaptation (24 h with two bottles of 1% sucrose; then 24 h with one bottle of water and one of 1% sucrose). After 24 h of water deprivation, a 12-h test was conducted with two identical bottles (positions swapped at 6 h). Sucrose preference (%) was calculated as sucrose intake/(sucrose + water) × 100. The open-field test was conducted in a square arena (100 × 100 × 40 cm) under uniform illumination (~ 120 lx). The central zone was defined as the inner 50 × 50 cm area. Rats were gently placed in a corner and allowed to explore for 5 min. Total distance traveled, immobility time, and center-zone time and entries were quantified using automated tracking. The arena was cleaned with 70% ethanol between trials to eliminate residual olfactory cues. For the forced-swim test (FST), rats were pre-exposed for 15 min the day prior; on test day they swam for 6 min in 25 ± 1 °C water (≈ 30 cm depth), and immobility during the final 4 min was scored by a blinded observer. Behavioral assays were administered in a fixed sequence to reduce cumulative stress effects: the sucrose preference test was followed by the open-field test, and the forced-swim test was conducted last under thermoneutral conditions.

Histology and immunohistochemistry

Lungs and brains were fixed in 4% paraformaldehyde (≥ 24 h), processed, paraffin-embedded, and sectioned at 4 μm. For hematoxylin-eosin staining, sections were deparaffinized in xylene, rehydrated through graded ethanol to water, stained with hematoxylin (with acid alcohol differentiation and bluing as required), counterstained with eosin Y, dehydrated, cleared, and mounted with neutral resin. Lung morphometry was performed on randomly selected, nonoverlapping parenchymal fields (avoiding bronchi and vessels > 100 μm and artifacts). After pixel calibration in ImageJ, mean linear intercept (MLI) and mean alveolar number (MAN) were computed using standard line-intersection methods; the mean of three fields per animal served as the biological replicate. Hippocampal CA1 injury was graded on a 0–3 scale by two independent, blinded pathologists (0, intact cytoarchitecture; 1, mild disarray with occasional pyknotic neurons; 2, moderate disorganization with scattered karyopyknosis/vacuolation and neuron loss; 3, severe disruption with widespread neuron loss, edema, and vascular dilation). Discrepancies were resolved by consensus.

For immunohistochemistry, antigen retrieval was performed in sodium citrate buffer (microwave heating), endogenous peroxidase was quenched with 3% H₂O₂ (25–30 min), and nonspecific binding was blocked with 5% BSA (30 min at 37 °C). Sections were incubated overnight at 4 °C with phospho-p38 (Bioss, bs-0636R, 1:200 for lung / 1:100 for hippocampus), phospho-JNK (Proteintech, 80024-1-RR, 1:200), or phospho-p65 (Affinity, AF2006, 1:200 for lung / 1:150 for hippocampus) at empirically optimized dilutions for lung and brain, followed by HRP-conjugated secondary antibodies (30 min, 37 °C). DAB chromogen was used for development under microscopic monitoring; nuclei were counterstained with hematoxylin, blued, dehydrated, cleared, and mounted. Whole slides were digitized with an automated scanner. Within predefined regions of interest (ROIs), integrated optical density (IOD) and area were measured in ImageJ and average optical density (AOD = IOD/area) was used as a semiquantitative index of phospho-protein expression.

TUNEL and microglial Immunofluorescence

Apoptosis in hippocampal CA1 was assessed with a fluorescein TUNEL kit. After deparaffinization and rehydration, sections underwent proteinase-K permeabilization (15–30 min at 37 °C), PBS washes, and incubation with TUNEL reaction mixture (60 min at 37 °C in the dark). Autofluorescence quenching was applied per kit instructions; sections were counterstained with DAPI and coverslipped with antifade medium. Slides were scanned on a fluorescence slide scanner. In randomly selected high-power fields within CA1, TUNEL-positive nuclei were counted and expressed as a percentage of total DAPI-positive nuclei.

Microglial activation was evaluated by IBA-1 immunofluorescence. After antigen retrieval and blocking, sections were incubated with anti-IBA-1 (1:250, overnight at 4 °C), washed in PBST, and incubated with a 594-nm fluorescent secondary antibody (1:500, 30 min at 37 °C). DAPI counterstaining and antifade mounting were then applied. All images were acquired from the left hippocampal hemisphere to maintain anatomical consistency. Quantification was performed in the CA1 region, a brain area strongly implicated in depressive-like behavior and stress-related neuroinflammation. For each animal, 5 non-overlapping high-power fields within anatomically matched CA1 ROIs were selected. Group size was n = 3 biologically independent animals per group, which is commonly used for fluorescence-based histological quantification, while multiple fields per animal were analyzed to ensure adequate sampling depth. Image analysis was performed in ImageJ by a blinded evaluator. A uniform intensity threshold was applied across all images to identify IBA-1–positive cells. Positive-cell percentages were determined by counting IBA-1–labeled cells relative to the total number of DAPI-stained nuclei within each ROI. Fluorescence intensity was quantified as mean fluorescence intensity (MFI), calculated by normalizing the integrated fluorescence signal of IBA-1 labeling to the analyzed ROI area.

Western blot

Lung and hippocampal tissues were homogenized in ice-cold RIPA buffer containing protease and phosphatase inhibitors. Lysates were clarified (12,000 × g, 20 min, 4 °C), and protein concentrations were determined by BCA assay. Equal protein (≈ 40 µg per lane) was denatured with 5× loading buffer (95–100 °C, 10 min), resolved by SDS-PAGE, and transferred to PVDF membranes (methanol-activated). After blocking with 5% non-fat milk in TBS-T, membranes were incubated overnight at 4 °C with the following primary antibodies at validated dilutions: phospho-p65 (Affinity, AF2006, 1:1,000), total p65 (Proteintech, 10745-1-AP, 1:2,000), phospho-p38 (Bioss, bs-0636R, 1:1,000), total p38 (Proteintech, 14064-1-AP, 1:2,000), phospho-JNK (Proteintech, 80024-1-RR, 1:1,000), total JNK (Proteintech, 24164-1-AP, 1:2,000), and GAPDH (Proteintech, 60004-1-Ig, 1:5,000). Membranes were then washed and incubated with HRP-conjugated secondary antibodies for 1 h at room temperature. Immunoreactive bands were visualized using enhanced chemiluminescence and imaged under identical exposure settings. Band intensities were quantified in ImageJ with background subtraction, and phospho/total ratios (p-p65/p65, p-p38/p38, p-JNK/JNK) were calculated for each sample.

Serum cytokines and oxidative-stress markers

Serum (and cell-culture supernatants where indicated) was analyzed by ELISA for IL-6, IL-8, IL-10, and TNF-α according to manufacturers’ instructions; standards and samples were run in duplicate or triplicate and read at 450 nm. Malondialdehyde (MDA) was determined by thiobarbituric-acid reaction (absorbance at 532 nm), myeloperoxidase (MPO) activity colorimetrically (460 nm), and superoxide dismutase (SOD) activity by the WST-1 method (550 nm). To minimize variability, all assays for a given analyte were performed on the same plate when feasible.

Cell culture and in-vitro interventions

Rat alveolar macrophages (NR8383; F-12 K supplemented with 15% FBS and 1% penicillin/streptomycin) and rat microglia (HAPI; DMEM/F-12 with 10% FBS and 1% penicillin/streptomycin) were maintained at 37 °C in 5% CO₂. Pilot CCK-8 assays established working concentrations of cigarette smoke extract (CSE; 0–500 µg/mL) and esketamine (0–20 µg/mL) that induced inflammatory stress without excessive cytotoxicity. CSE was obtained from Python Co. (catalog AAPR551-B2). Mainstream smoke from 3R4F Kentucky reference cigarettes was collected by a cold-trap method and solubilized in PBS to 5 mg/mL (nicotine 120 µg/mL; pH 7.2), sterile-filtered (0.22 μm), and supplied as a standardized preparation with batch-to-batch consistency. Immediately before use, CSE was diluted aseptically to target concentrations; all working dilutions were freshly prepared to minimize loss of volatile constituents.

NR8383 experiments (24 h unless noted) comprised: control; CSE; CSE plus esketamine; CSE plus Diprovocim (500 nM, final 2 h); and CSE plus esketamine plus Diprovocim. HAPI experiments comprised: control; LPS (1 µg/mL) plus CSE; LPS + CSE plus esketamine; LPS + CSE plus Diprovocim (500 nM, final 2 h); and LPS + CSE plus esketamine plus Diprovocim. Readouts included CCK-8 viability, ELISA cytokines/oxidative indices in supernatants, and Western blot analysis of p-p38, p-JNK, and p-p65. For Diprovocim co-stimulation, the compound was applied during the final 2 h based on prior evidence that it induces rapid TLR1/2 pathway activation [38]. All cell experiments were performed with n = 3 biological replicates and three technical replicates per assay to ensure reproducibility.

Serum exosomal miRNA sequencing

Serum was clarified by sequential low-speed centrifugation and ultracentrifugation to isolate exosomes. Vesicle size and morphology were quality-controlled as needed (e.g., NTA, TEM). Total RNA was extracted from exosomes, and miRNA libraries (e.g., QIAseq) were constructed, quality-checked (Bioanalyzer), and sequenced on an Illumina platform. After adapter and quality trimming, reads were mapped to the rat genome and miRNA references; counts were generated and differential expression was analyzed with DESeq2/edgeR (P < 0.05 and |log₂ fold change| ≥ 1). Predicted target genes of differentially expressed miRNAs were subjected to KEGG pathway enrichment with Benjamini-Hochberg FDR control (P < 0.05).

Gut microbiota 16S rRNA sequencing

Colonic contents were collected at necropsy into sterile tubes and stored at − 80 °C. Microbial DNA (FastPure Stool DNA kit) served as template for V3-V4 amplification using primers 338 F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). PCR reactions (20 µL) contained 2× Pro Taq (10 µL), 5 µM primers (0.8 µL each), ≈ 10 ng DNA, and nuclease-free water; cycling was 95 °C for 3 min; 27 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 45 s; and 72 °C for 10 min. Triplicates were pooled, purified from 2% agarose, quantified (Synergy HTX), prepared into libraries with NEXTFLEX^® Rapid DNA-Seq, and sequenced on the Illumina NextSeq 2000 (PE300) platform according to standard protocols by Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China). Reads were filtered with fastp (v0.19.6), merged with FLASH (v1.2.11), clustered into OTUs at 97% identity with chimera removal (USEARCH v11), and rarefied to 20,000 reads per sample (Good’s coverage ≈ 99%). Taxonomy was assigned with RDP Classifier (v11.5) against Silva v138.2 (70% confidence). Predicted functions were inferred by PICRUSt2 (v2.2.0) and mapped to KEGG (Level-3). Raw data are available in SRA (SRP599604).

Lung microbiota 16S rRNA five-region sequencing

Aseptic lung tissue was snap-frozen at − 80 °C. Total DNA (FastPure Stool DNA kit) was amplified across five hypervariable regions (V2, V3, V5, V6, V8) using barcoded primers: V2 (F1 TGGCGAACGGGTGAGTAA / R1 CCGTGTCTCAGTCCCARTG), V3 (F2 ACTCCTACGGGAGGCAGC / R2 GTATTACCGCGGCTGCTG), V5 (F3 GTGTAGCGGTGRAATGCG / R3 CCCGTCAATTCMTTTGAGTT), V6 (F4 GGAGCATGTGGWTTAATTCGA / R4 CGTTGCGGGACTTAACCC), V8 (F5 GGAGGAAGGTGGGGATGAC / R5 AAGGCCCGGGAACGTATT). PCR reactions (20 µL) contained 5× TransStart FastPfu buffer (4 µL), 2.5 mM dNTPs (2 µL), 5 µM primers (0.8 µL each), TransStart FastPfu (0.4 µL), ≈ 10 ng DNA, and water; cycling was 95 °C for 3 min; 29 cycles of 95 °C for 30 s, 53 °C for 30 s, and 72 °C for 45 s; and 72 °C for 10 min. Each sample had three technical replicates and process blanks spanning sampling, extraction, and PCR. Amplicons were pooled, gel-purified, quantified (Synergy HTX), converted to libraries as above, and sequenced on the Illumina NextSeq 2000 platform according to standard protocols by Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China). Taxonomic assignment used the Short Multiple Regions Framework (SMURF) to integrate multi-region reads against Greengenes (May 2013). Given the low biomass of lung tissue, stringent background filtering was applied: samples with < 1,000 reads and taxa with relative abundance < 10⁻⁴ were removed; taxa present in extraction/PCR blanks or sampling blanks at > 30% frequency were excluded. Raw data are available in SRA (SRP599622).

Statistical analysis

Analyses were performed in SPSS 29.0 and GraphPad Prism 10. Normality was assessed by the Shapiro-Wilk test and homoscedasticity by Bartlett’s test. Parametric data are presented as mean ± SD; non-parametric data as median (interquartile range). For multi-group comparisons, one-way ANOVA with Bonferroni or LSD post-hoc testing was used when assumptions were met; Welch’s ANOVA with Dunnett T3 was used for unequal variances; otherwise, Kruskal-Wallis with Dunn’s correction was applied. Two-sided P < 0.05 denoted statistical significance.

For microbiota analyses, alpha diversity was computed with mothur (v1.30.2) or the vegan package (R); beta diversity used Bray-Curtis distances with principal-coordinates analysis (PCoA) and ANOSIM for between-group separation. Differential taxa were evaluated by Wilcoxon rank-sum tests (P < 0.05) and LEfSe (LDA ≥ 2, P < 0.05). Functional pathway differences were tested by Kruskal-Wallis with Benjamini-Hochberg false-discovery-rate correction. Spearman correlations were computed between microbial features and physiological, behavioral, and biochemical indices and visualized as heatmaps. All tests were two-tailed.

Result

Pulmonary impairment and emphysematous changes in the comorbid model

At the end of the modeling period, clear differences in general condition were observed across groups. Control rats remained active, with smooth fur and steady weight gain. In contrast, rats exposed to cigarette smoke (COPD group) and those subjected to both cigarette smoke and CUMS (CY group) appeared lethargic, with coarse fur, cough, and tachypnea.

At week 16, pulmonary function testing revealed significant group differences (P < 0.05). Compared with controls, both COPD and CY rats exhibited marked reductions in EF50, PEF, FVC, and FEV_0.1/FVC (Fig. 1a-d), indicating airflow limitation and impaired ventilatory capacity.

Fig. 1.

Establishment and validation of the comorbid COPD-depression rat model. a-d, Pulmonary function at week 16, including EF50, PEF, FVC, and FEV0.1/FVC (n = 7). e-h, Behavioral assessments showing sucrose preference, forced-swim immobility, open-field locomotor distance, and immobility time (n = 7). i, Representative open-field trajectories. j, Representative lung sections stained with hematoxylin-eosin (HE) and quantification of mean linear intercept (MLI) and mean alveolar number (MAN) (n = 3). k, Representative hippocampal sections showing HE staining, TUNEL assay, and IBA-1 immunofluorescence, with quantification of pathological scores, apoptotic index (TUNEL-positive cells), and microglial activation (IBA-1-positive cells and mean fluorescence intensity) (n = 3). Con, control; COPD, COPD model; YY, depression model; CY, comorbid model. Data are expressed as mean ± SD or median (IQR) as appropriate. a-h: one-way ANOVA with Bonferroni post-hoc correction (*P < 0.05, **P < 0.01 vs. Con); Welch test with Dunnett’s T3 correction (△P < 0.05 vs. Con). j, k: one-way ANOVA with LSD post-hoc correction (*P < 0.05 vs. Con); for non-parametric variables, Kruskal-Wallis test with Dunn’s correction was applied (*P < 0.05 vs. Con)

Histopathological examination further confirmed the presence of emphysematous changes. Hematoxylin-eosin staining showed intact alveolar structures in control and YY rats, whereas COPD and CY rats displayed septal thickening, disrupted alveolar architecture, epithelial shedding, and inflammatory infiltration. Quantitative morphometry demonstrated significantly increased mean linear intercept (MLI) and reduced mean alveolar number (MAN) in COPD and CY rats (P < 0.05; Fig. 1j).

Neurobehavioral deficits and hippocampal pathology in comorbid COPD-depression

Behavioral assessments demonstrated pronounced neurobehavioral disturbances in YY and CY rats. Compared with controls, both groups showed reduced sucrose preference, prolonged immobility in the forced-swim test, decreased locomotor distance, and increased immobility time in the open-field test (all P < 0.05; Fig. 1e-i). These findings indicate anhedonia, behavioral despair, and reduced exploratory activity.

Histological evaluation of the hippocampal CA1 region revealed preserved neuronal morphology in control and COPD rats. In contrast, YY and CY rats exhibited neuronal loss, shrinkage, and disorganized cellular arrangement, accompanied by significantly elevated pathological scores (P < 0.05). TUNEL staining confirmed increased neuronal apoptosis in these groups, and IBA-1 immunofluorescence demonstrated marked microglial alterations, with higher numbers of IBA-1 positive cells and increased fluorescence intensity (P < 0.05; Fig. 1k).

Serum exosomal miRNA profiling suggests involvement of MAPK/NF-κB related pathways

Serum exosomal miRNA sequencing was performed in control, comorbid, and esketamine-treated rats. Distinct sets of differentially expressed miRNAs were identified when comparing comorbid rats with controls, as well as when comparing esketamine-treated and untreated comorbid rats. KEGG pathway enrichment analysis of the predicted target genes revealed significant over-representation of several signaling cascades, with MAPK- and NF-κB–related pathways appearing consistently across comparisons (Figure S1). These findings suggested that dysregulated upstream signaling may contribute to the systemic and neural alterations observed in the comorbid model. Accordingly, subsequent analyses evaluated whether key components of MAPK- and NF-κB–associated pathways were altered in vivo and whether these molecular changes were modulated by esketamine.

Systemic inflammation, oxidative stress, and MAPK/NF-κB activation in lung and hippocampus

Serum inflammatory cytokines and oxidative stress markers differed markedly across groups. Levels of IL-8, TNF-α, and IL-10 were elevated in COPD, YY, and CY rats compared with controls, whereas IL-6 was selectively increased in COPD and CY rats. Oxidative stress indices showed a parallel pattern: MPO was increased and SOD was reduced in all three disease groups, and MDA was significantly elevated in COPD and CY rats (Fig. 2a).

Fig. 2.

Activation of MAPK/NF-κB signaling accompanied by systemic inflammation and oxidative stress in the COPD-depression comorbid rat model. a, Serum cytokines and oxidative stress markers, including IL-6, IL-8, TNF-α, IL-10, MPO, SOD, and MDA (n = 7). b, Representative western blots of lung tissue showing p-p38, p38, p-JNK, JNK, p-p65, p65, and GAPDH, with quantification of phosphorylation ratios (n = 3). c, Representative western blots of hippocampal tissue showing p-p38, p38, p-JNK, JNK, p-p65, p65, and GAPDH, with quantification of phosphorylation ratios (n = 3). d, Representative immunohistochemical staining of lung sections for p-p38, p-JNK, p-p65, with quantitative analysis of mean optical density (n = 3). e, Representative immunohistochemical staining of hippocampal sections for p-p38, p-JNK, and p-p65, with quantitative analysis of mean optical density (n = 3). Con, control; COPD, COPD model; YY, depression model; CY, comorbid model. Data are shown as mean ± SD. a: one-way ANOVA with Bonferroni post-hoc correction was used (*P < 0.05, **P < 0.01 vs. Con); Welch test with Dunnett’s T3 correction was used where appropriate (^△P < 0.05 vs. Con). b-e: one-way ANOVA with LSD post-hoc correction (*P < 0.05 vs. Con)

Western blot analyses demonstrated robust activation of MAPK/NF-κB–related signaling in both lung and hippocampal tissue. In the lung, phosphorylation of JNK was increased in COPD, YY, and CY rats, while phosphorylation of p38 and p65 was elevated in COPD and CY rats (Fig. 2b). In the hippocampus, phosphorylation of p38 and p65 was increased in COPD, YY, and CY rats, whereas JNK phosphorylation was selectively elevated in YY and CY rats (Fig. 2c).

Immunohistochemistry corroborated these biochemical findings. Expression of p-p38, p-JNK, and p-p65 was markedly enhanced in the lungs of COPD and CY rats, with broader distribution and stronger staining intensity. In the hippocampus, p-p38 and p-p65 expression was increased in YY and CY rats, whereas p-JNK levels did not show significant differences across groups (Fig. 2d-e).

These results together indicate that the comorbid COPD-depression model is characterized by systemic inflammation, oxidative stress, and coordinated activation of MAPK/NF-κB signaling in both lung and brain.

Esketamine ameliorates pulmonary dysfunction and lung pathology

Following the intervention period, the general condition of animals diverged markedly across groups. Control rats remained active with smooth fur and stable weight gain. CY rats appeared lethargic, with coarse fur, decreased activity, and labored respiration. Esketamine-treated CYK rats showed partial recovery, including improved activity, cleaner fur, and more regular breathing, whereas CYKD rats resembled the CY group and exhibited deterioration relative to CYK rats.

Pulmonary function testing demonstrated significant between-group differences before and after treatment. At week 16, EF50 and PEF were reduced in CY, CYK, and CYKD rats compared with controls (P < 0.05). By week 18, untreated CY rats remained impaired, exhibiting lower EF50, PEF, FVC, and FEV_0.1/FVC than controls (P < 0.05). Esketamine administration improved airflow limitation in CYK rats, which displayed significantly higher EF50, PEF, FVC, and FEV_0.1/FVC compared with CY rats (P < 0.05). These improvements were attenuated in CYKD rats, which displayed lower values than CYK rats but remained higher than those of untreated CY rats (P < 0.05; Fig. 3a).

Fig. 3.

Esketamine improves pulmonary, behavioral, and neuropathological outcomes in comorbid COPD-depression rats. a, Pulmonary function at week 16 and week 18, including EF50, PEF, FVC, and FEV0.1/FVC (n = 7). b-c, Behavioral performance, including sucrose preference, forced-swim immobility, open-field locomotor distance, and open-field immobility time (n = 7). d, Representative lung sections stained with hematoxylin-eosin (HE) and quantification of mean linear intercept (MLI) and mean alveolar number (MAN) (n = 3). e, Representative hippocampal sections showing HE staining, TUNEL assay, and IBA-1 immunofluorescence, with quantification of pathological scores, apoptotic index (TUNEL-positive cells), and microglial activation (IBA-1-positive cells and mean fluorescence intensity) (n = 3). Con, control; CY, comorbid model; CYK, CY rats treated with esketamine; CYKD, CY rats treated with esketamine and Diprovocim. Data are expressed as mean ± SD or median (IQR) as appropriate. a, b: one-way ANOVA with Bonferroni post-hoc correction was used (*P < 0.05, **P < 0.01 vs. Con; #P < 0.05, ##P < 0.01 vs. CY; ^▲P < 0.05, ^▲▲P < 0.01 vs. CYK); Welch test with Dunnett’s T3 correction was used (^△P < 0.05, ^△△P < 0.01 vs. Con; ^◎P < 0.05 vs. CY). d, e: one-way ANOVA with LSD post-hoc correction was used (*P < 0.05 vs. Con; #P < 0.05 vs. CY; ^▲P < 0.05 vs. CYK); for non-parametric variables, Kruskal-Wallis test with Dunn’s correction was applied (*P < 0.05 vs. Con)

Histological evaluation further supported pulmonary improvement with esketamine. Control lungs exhibited intact alveolar architecture, whereas CY rats showed hallmark COPD-like changes—including septal thickening, alveolar disruption, epithelial shedding, and inflammatory infiltration. CYK rats demonstrated clearer alveolar contours and reduced inflammatory changes, whereas CYKD rats again resembled CY rats. Quantitative morphometry confirmed increased MLI and decreased MAN in CY rats, improvement in CYK rats, and deterioration in CYKD rats (P < 0.05; Fig. 3d).

Esketamine reverses depression-like behaviours and hippocampal injury

At week 16, CY, CYK, and CYKD rats exhibited reduced sucrose preference, prolonged immobility in the forced-swim test, and decreased locomotor activity in the open-field test compared with controls (P < 0.05). Open-field immobility time was similarly increased (P < 0.05). By week 18, CY rats continued to display pronounced depression-like behaviors. In contrast, CYK rats showed behavioral improvement, with increased sucrose preference, shorter immobility times, and greater locomotor activity relative to CY rats (P < 0.05). CYKD rats showed diminished improvement compared with CYK rats, although their depressive-like behavioral measures remained directionally better than those of CY rats (P < 0.05; Fig. 3b-c).

Histopathological analysis of the hippocampal CA1 region revealed preserved neuronal structure in controls. CY rats displayed neuronal loss, cellular shrinkage, and disorganized arrangement. Esketamine-treated CYK rats showed partial restoration of neuronal morphology, whereas CYKD rats resembled the CY group. Pathological scores were significantly increased in CY rats (P < 0.05). TUNEL staining confirmed enhanced neuronal apoptosis in CY rats, attenuation in CYK rats, and renewed elevation in CYKD rats. Immunofluorescence staining similarly demonstrated pronounced microglial alterations in CY rats, attenuation of these alterations in CYK rats, and partial recurrence in CYKD rats (P < 0.05; Fig. 3e).

Esketamine suppresses pulmonary MAPK/NF-κB signaling

In the lung, CY rats showed marked activation of MAPK/NF-κB signaling, with increased phosphorylation of p38, JNK, and p65 compared with controls. Esketamine treatment substantially attenuated these phosphorylation levels in CYK rats, whereas co-administration of Diprovocim reversed this suppression, restoring p-p38, p-JNK, and p-p65 to levels comparable to the untreated comorbid group (Fig. 4b). Immunohistochemistry corroborated these findings: CY rats exhibited stronger and more widespread staining for p-p38, p-JNK, and p-p65, reductions were evident in CYK rats, and Diprovocim again reinstated the heightened expression pattern (Fig. 4d).

Fig. 4.

Esketamine suppresses MAPK/NF-κB signaling and restores inflammatory and oxidative balance in comorbid COPD-depression rats. a, Serum cytokines and oxidative stress markers, including IL-6, IL-8, TNF-α, IL-10, MDA, MPO, and SOD (n = 7). b, Representative western blots of lung tissue showing p-p38, p38, p-JNK, JNK, p-p65, p65, and GAPDH, with quantification of phosphorylation ratios (n = 3). c, Representative western blots of hippocampal tissue showing p-p38, p38, p-JNK, JNK, p-p65, p65, and GAPDH, with quantification of phosphorylation ratios (n = 3). d, Representative lung sections stained for p-p38, p-JNK, and p-p65, with quantification of mean optical density (AOD) (n = 3). e, Representative hippocampal sections stained for p-p38, p-JNK, and p-p65, with quantification of mean optical density (AOD) (n = 3).Con, control; CY, comorbid model; CYK, CY rats treated with esketamine; CYKD, CY rats treated with esketamine and Diprovocim. Data are expressed as mean ± SD. a: one-way ANOVA with Bonferroni post-hoc correction was used (**P < 0.01 vs. Con; ##P < 0.01 vs. CY; ^▲P < 0.05, ^▲▲P < 0.01 vs. CYK). b-e: one-way ANOVA with LSD post-hoc correction was used (*P < 0.05 vs. Con; #P < 0.05 vs. CY; ^▲P < 0.05 vs. CYK)

Esketamine modulates hippocampal MAPK/NF-κB signaling

A similar pattern was observed in the hippocampus. Phosphorylation of p38, JNK, and p65 was elevated in CY rats, reduced following esketamine treatment, and re-elevated with the addition of Diprovocim (Fig. 4c). Immunohistochemical analysis confirmed attenuation of p-p38 and p-p65 expression in CYK rats and renewed upregulation in CYKD rats, whereas p-JNK showed no significant group differences in staining intensity (Fig. 4e). These results indicate that esketamine partially normalizes hippocampal MAPK/NF-κB signaling disturbances in comorbid COPD-depression, and that Diprovocim antagonizes these effects.

Esketamine restores systemic inflammatory and oxidative balance

Serum inflammatory and oxidative stress markers showed marked dysregulation in comorbid rats and a clear therapeutic response to esketamine. CY rats exhibited significantly elevated concentrations of IL-6, IL-8, TNF-α, and IL-10 compared with controls, together with increased MDA and MPO and reduced SOD activity. Esketamine treatment partially normalized this inflammatory profile, lowering IL-6, IL-8, and TNF-α while further enhancing IL-10 in CYK rats. A parallel improvement was observed in oxidative stress indices, with reductions in MDA and MPO and restoration of SOD activity. These beneficial effects were attenuated in CYKD rats, which again showed heightened pro-inflammatory cytokines, increased oxidative burden, and diminished antioxidant capacity (Fig. 4a).

Cellular validation of pulmonary MAPK/NF-κB modulation in alveolar macrophages (NR8383 cells)

In NR8383 cells, CCK-8 assays demonstrated a dose-dependent reduction in viability with increasing concentrations of CSE (P < 0.05). A concentration of 200 µg/mL reduced viability to approximately 74% and was therefore selected for subsequent modeling. Esketamine improved cell viability at 0–15 µg/mL, with maximal enhancement observed at 15 µg/mL, whereas 20 µg/mL decreased viability (P < 0.05). Accordingly, 200 µg/mL CSE and 15 µg/mL esketamine were used as the modeling and intervention conditions. In grouped assays, CSE decreased viability, esketamine restored it, Diprovocim further suppressed it, and combined treatment diminished the protective effect of esketamine (P < 0.05; Fig. 5a, b, d).

Fig. 5.

Cellular validation of MAPK/NF-κB signaling in alveolar macrophages and microglia. a-b, Dose-dependent effects of CSE and esketamine on NR8383 cell viability (CCK-8 assay). c, Western blot analysis of p-p38, p38, p-JNK, JNK, p-p65, p65, and GAPDH in NR8383 cells, with quantification of phosphorylation ratios (n = 3). d-f, Grouped NR8383 assays showing cell viability (CCK-8) and release of inflammatory cytokines (IL-6, IL-8, TNF-α, IL-10) and oxidative stress markers (MDA, MPO, SOD) (n = 3). g-h, Dose-dependent effects of LPS + CSE and esketamine on HAPI cell viability (CCK-8 assay). i, Western blot analysis of p-p38, p38, p-JNK, JNK, p-p65, p65, and GAPDH in HAPI cells, with quantification of phosphorylation ratios (n = 3). j-l, Grouped HAPI assays showing cell viability (CCK-8) and release of inflammatory cytokines and oxidative stress markers (n = 3). Con, control; CSE, cigarette smoke extract; LPS, lipopolysaccharide. Data are expressed as mean ± SD; *P < 0.05 vs. Con; #P < 0.05 vs. CSE or LPS + CSE; ^▲P < 0.05 vs. CSE + Esketamine or LPS + CSE + Esketamine (LSD correction)

Inflammatory and oxidative indices showed parallel trends. CSE increased IL-6, IL-8, and TNF-α and decreased IL-10, while esketamine reversed these changes; Diprovocim produced opposing effects, and co-treatment attenuated esketamine’s benefit (P < 0.05). Oxidative stress markers followed a similar pattern: MDA and MPO were elevated and SOD decreased after CSE exposure; esketamine normalized these indices; Diprovocim exacerbated them; and combined treatment again blunted esketamine’s effects (P < 0.05; Fig. 5e, f).

Western blotting confirmed modulation of MAPK/NF-κB signaling. CSE increased phosphorylation of p38, JNK, and p65; esketamine reduced phosphorylation of JNK and p65; Diprovocim further enhanced these signals; and esketamine plus Diprovocim re-elevated phosphorylation compared with esketamine alone (P < 0.05; Fig. 5c).

Cellular validation of hippocampal MAPK/NF-κB modulation in microglia (HAPI cells)

In HAPI microglial cells, viability declined under combined LPS (1 µg/mL) and CSE stimulation, with a marked reduction observed at 250 µg/mL CSE (approximately 72%), which was chosen for modeling (P < 0.05). Esketamine improved viability at 10–15 µg/mL but decreased viability at 20 µg/mL (P < 0.05). Thus, 250 µg/mL CSE with 1 µg/mL LPS and 15 µg/mL esketamine were used in subsequent experiments. In grouped assays, LPS + CSE reduced viability, esketamine restored it, Diprovocim further suppressed it, and co-treatment again attenuated the beneficial effect of esketamine (P < 0.05; Fig. 5g, h, j).

Inflammatory cytokines and oxidative stress markers showed comparable trends to those in NR8383 cells. LPS + CSE elevated IL-6, IL-8, and TNF-α and decreased IL-10 relative to controls (P < 0.05). Esketamine reversed these alterations, whereas Diprovocim produced the opposite pattern, and combined treatment reduced the degree of esketamine-related improvement (P < 0.05). Oxidative stress exhibited the same sequence: LPS + CSE increased MDA and MPO and decreased SOD; esketamine normalized these changes; Diprovocim aggravated them; and esketamine plus Diprovocim re-elevated MDA and MPO and reduced SOD compared with esketamine alone (P < 0.05; Fig. 5k, l).

Western blotting corroborated these effects on MAPK/NF-κB signaling. LPS + CSE increased phosphorylation of p38, JNK, and p65; esketamine attenuated these increases; Diprovocim enhanced them; and combined treatment again re-elevated phosphorylation relative to esketamine alone (P < 0.05; Fig. 5i).

Gut microbiota diversity, composition, and correlations

Alpha diversity analysis showed that richness indices (Sobs, Ace, Chao1) were lower in CY rats than in controls, although differences were not significant (P > 0.05). After esketamine treatment, these indices increased significantly in CYK rats compared with CY rats (P < 0.05) but decreased again in CYKD rats (P < 0.05). Community diversity, assessed by Shannon and Simpson indices, was reduced in CY rats relative to controls (P < 0.05). Esketamine reversed these changes in CYK rats, whereas CYKD rats again showed lower Shannon and higher Simpson indices (P < 0.05). Rarefaction curves reached a plateau with increasing sequencing depth, indicating sufficient coverage and reliable data quality (Fig. 6a, c).

Fig. 6.

Gut microbiota diversity, composition, and correlations in comorbid COPD-depression rats. a, Alpha diversity indices (Sobs, Ace, Chao1, Shannon, Simpson) across groups (n=7). b, Principal coordinates analysis (PCoA) based on Bray-Curtis distances and ANOSIM results. c, Rarefaction curves showing sequencing depth and OTU coverage. d, Differential genera between Con and CY rats identified by Wilcoxon test and LEfSe. e, Differential genera between CY and CYK rats identified by Wilcoxon test and LEfSe. f, Spearman correlations between the top 50 genera and host phenotypes (lung function, behavior, inflammatory cytokines, oxidative stress indices). g, Predicted KEGG level 3 pathways differing among groups based on PICRUSt2 analysis. Con, control; CY, comorbid model; CYK, CY rats treated with esketamine; CYKD, CY rats treated with esketamine and Diprovocim. Data are expressed as mean ± SD or median (IQR) as appropriate; a: *P < 0.05, **P < 0.01 vs. Con; #P < 0.05, ##P < 0.01 vs. CY; ^▲P < 0.05, ^▲▲P < 0.01 vs. CYK (Bonferroni corrected). ††P < 0.01 indicates significance vs. CY; ‡‡P < 0.01 indicates significance vs. CYK (Dunn’s pairwise test). d, e: *P < 0.05 , **P < 0.01 compared with the indicated group. f: *P < 0.05, **P < 0.01, ***P < 0.001 indicate significant correlations. g: *P < 0.05, **P < 0.01, ***P < 0.001 between two groups

Beta diversity assessed by PCoA based on Bray-Curtis distances revealed distinct clustering among groups. ANOSIM confirmed significant differences in community structure (R = 0.4231, P = 0.001). Pairwise comparisons showed clear separation between Con and CY (R = 0.6744, P = 0.001), CY and CYK (R = 0.4616, P = 0.001), and CYK and CYKD (R = 0.4937, P = 0.001) (Fig. 6b).

Community composition analysis identified 1009 OTUs shared by all groups, representing 48.6% of the total. OTU numbers decreased in CY rats, recovered in CYK rats after esketamine treatment, and declined slightly again in CYKD rats. Dominant genera across groups included Ligilactobacillus, Muribaculaceae, Bacteroides, Prevotellaceae_UCG-003, Clostridia_UCG-014, Dubosiella, Lachnospiraceae_NK4A136_group, and Allobaculum (Supplementary Fig. S2).

Differential abundance analysis showed that, compared with controls, CY rats were enriched in Ligilactobacillus and depleted in Dubosiella, Bifidobacterium, Romboutsia, Faecalibaculum, norank_f__Erysipelotrichaceae, Clostridium, Turicibacter, norank_f_UCG-010, and Anaerobiospirillum (P < 0.05). LEfSe confirmed enrichment of Ligilactobacillus, Xylanibacter, Streptococcus, Eubacterium ventriosum group, Eubacterium ruminantium group, norank Rs-E47 termite group, and the Rs-E47 termite group family in CY rats. In contrast, Dubosiella, Staphylococcus, Bifidobacterium, and Faecalibaculum were enriched in controls, together with members of the Erysipelotrichaceae, Ruminococcaceae, and Bifidobacteriaceae families and their higher-level taxa including Bifidobacteriales, Erysipelotrichales, and the phylum Actinobacteria (Fig. 6d).

Compared with CY rats, CYK rats had reduced abundance of Ligilactobacillus and enrichment of Dubosiella, norank_f_[Eubacterium]_coprostanoligenes_group, Colidextribacter, NK4A214_group, norank_f_Lachnospiraceae, norank_f_Oscillospiraceae, Intestinimonas, norank_f_Desulfovibrionaceae, and Monoglobus (P < 0.05). LEfSe further confirmed enrichment of Ligilactobacillus and related taxa (Lactobacillaceae, Lactobacillales, Bacilli) and the Eubacterium ventriosum group in CY rats. By contrast, CYK rats were enriched in members of the class Clostridia and the orders Lachnospirales, Oscillospirales, and Erysipelotrichales, along with the families Lachnospiraceae, Oscillospiraceae, and Erysipelotrichaceae. Additional enriched taxa in CYK rats included Agathobaculum, Dubosiella, and members of the Ruminococcaceae family (Fig. 6e).

Functional prediction with PICRUSt2 identified 30 KEGG level 3 pathways that differed significantly among groups, suggesting distinct metabolic shifts associated with esketamine treatment (Fig. 6g).

Correlation analysis further demonstrated associations between gut microbiota and host phenotypes. Ligilactobacillus negatively correlated with EF50, FEV_0.1/FVC, and sucrose preference, and positively with forced-swim immobility, IL-6, IL-8, TNF-α, MDA, and MPO, while negatively with SOD (P < 0.05). Conversely, Dubosiella and Faecalibaculum positively correlated with EF50, FEV_0.1/FVC, and sucrose preference, and negatively with immobility and inflammatory/oxidative markers, while positively with SOD (P < 0.05). Additional genera, including Ruminococcus, norank_f_[Eubacterium]_coprostanoligenes_group, norank_f__Erysipelotrichaceae, Intestinimonas, and Clostridium, also showed significant associations with behavioral indices. In addition, Ruminococcus negatively correlated with IL-8 and MDA and positively with SOD, whereas Clostridium negatively correlated with IL-8, MDA, and MPO (Fig. 6f).

Lung microbiota diversity, composition, and correlations

Analysis of alpha diversity showed no significant differences in richness (Sobs) across groups (P > 0.05). However, CY rats exhibited higher Shannon and lower Simpson indices than controls, indicating increased diversity and reduced evenness (P < 0.05). Esketamine treatment increased Simpson index in CYK rats, whereas CYKD rats reverted toward the CY phenotype (P < 0.05). Rarefaction curves plateaued with sequencing depth, confirming adequate coverage and data reliability (Fig. 7a, b).

Fig. 7.

Lung microbiota diversity, composition, and correlations in comorbid COPD-depression rats. a-b, Alpha diversity indices (Sobs, Shannon, Simpson) and rarefaction curves. c, Principal coordinates analysis (PCoA) based on Bray-Curtis distances and ANOSIM results. d-e, Differential species identified by Wilcoxon test and LEfSe analysis. f, Heatmap of Spearman correlations between the top 50 species and host phenotypes (lung function, behavior, serum cytokines, oxidative stress). g, Cross-organ correlations between gut and lung microbiota at the genus level. Con, control; CY, comorbid model; CYK, CY rats treated with esketamine; CYKD, CY rats treated with esketamine and Diprovocim. Data are expressed as mean ± SD or median (IQR) as appropriate; a: ^§P < 0.05, ^§§P < 0.01 indicate significant differences vs. Con; †P < 0.05 indicates significance vs. CY (Dunn’s test). d, e: *P < 0.05, **P < 0.01 compared with the indicated group. f, g: *P < 0.05, **P < 0.01, ***P < 0.001 indicate significant correlations

Beta diversity assessed by PCoA based on Bray-Curtis distances revealed distinct clustering among groups. ANOSIM confirmed significant differences (R = 0.4654, P = 0.001), with clear separation between Con and CY (R = 0.8785, P = 0.001), CY and CYK (R = 0.7736, P = 0.001), and CYK and CYKD (R = 0.4179, P = 0.001) (Fig. 7c).

Community composition analysis identified 403 shared species across groups. Dominant taxa included Lactobacillus murinus, Proteus penneri, CF231_Unknown_g_CF231, Klebsiella pneumoniae, Elizabethkingia meningoseptica, Providencia rettgeri, Bacillus circulans, and Deinococcus aquaticus (Supplementary Fig. S3).

Differential abundance analysis revealed that, compared with controls, CY rats had lower Lactobacillus murinus but higher Proteus penneri, Elizabethkingia meningoseptica, Providencia rettgeri, Bacillus circulans, Klebsiella pneumoniae, Deinococcus aquaticus, Blautia producta, Lactobacillus helveticus, and Micrococcus luteus (P < 0.05). LEfSe confirmed enrichment of Lactobacillus murinus and related higher-level taxa in controls, together with Eubacteriales and Lachnospiraceae. In contrast, CY rats were enriched in Pseudomonadota, Gammaproteobacteria, Enterobacterales, Actinomycetota, Flavobacteriia, and Morganellaceae (Fig. 7d).

Compared with CY rats, CYK rats showed increased Lactobacillus murinus and decreased Proteus penneri, Elizabethkingia meningoseptica, Bacillus circulans, Providencia rettgeri, Klebsiella pneumoniae, Acinetobacter lwoffii, Lactobacillus helveticus, Microbacterium laevaniformans, and Brevundimonas vesicularis (P < 0.05). LEfSe confirmed enrichment of Lactobacillus murinus and related taxa (Lactobacillus, Lactobacillaceae, Lactobacillales, Bacilli, Bacillota), together with Clostridia, Eubacteriales, Bacteroidia, and Bacteroidales in CYK rats. CY rats, by contrast, were enriched in Pseudomonadota, Gammaproteobacteria, Enterobacterales, Bacillales, Actinomycetota, Actinomycetes, Morganellaceae, and Flavobacteriales (Fig. 7e).

Correlation analysis demonstrated significant associations between lung microbiota and host phenotypes. Lactobacillus murinus and CF231_Unknown_g_CF231 positively correlated with EF50, FEV_0.1/FVC, and sucrose preference, and negatively with immobility and pro-inflammatory/oxidative markers, while positively with SOD (P < 0.05). Conversely, Proteus penneri, Elizabethkingia meningoseptica, Providencia rettgeri, Deinococcus aquaticus, Lactobacillus helveticus, and Chryseobacterium anthropi negatively correlated with lung function and sucrose preference, and positively with immobility and inflammatory/oxidative markers (P < 0.05). Additional correlations were observed: Klebsiella pneumoniae, Bacillus circulans, Acinetobacter guillouiae, and Corynebacterium tuberculostearicum negatively correlated with lung function; Streptococcus cristatus positively correlated with EF50 and FEV_0.1/FVC, and negatively with inflammatory markers; Microbacterium laevaniformans negatively correlated with sucrose preference and positively with immobility; while Bifidobacterium animalis, Ruminococcus gnavus, and Helicobacter rodentium positively correlated with sucrose preference and negatively with immobility, accompanied by favorable correlations with SOD (P < 0.05; Fig. 7f).

Cross-organ correlations between gut and lung microbiota

Cross-organ analysis at the genus level further revealed correlations between lung and gut microbiota. Among the top 30 genera, five were shared between lung and gut (Ruminococcus, Bacteroides, Streptococcus, Blautia, Allobaculum). Ruminococcus abundance was positively correlated between lung and gut, while Bacteroides was negatively correlated. Strong correlations (|r|>0.6, P < 0.05) were identified for 12 lung-gut genus pairs, including positive associations between gut Ligilactobacillus and lung Proteus, Providencia, Bacillus, Microbacterium, Elizabethkingia, Allobaculum, and Chryseobacterium, as well as between gut Lachnospiraceae_NK4A136_group and lung Bacteroides. Negative associations were observed between gut Ligilactobacillus and lung Lactobacillus, and between gut Xylanibacter and lung Bacteroides (Fig. 7g).

Discussion

We established a comorbid COPD-depression rat model by sequential cigarette-smoke exposure followed by chronic unpredictable mild stress (CUMS). The model reproduced airflow limitation and lung injury together with depression-like behaviors and hippocampal pathology. Building the organ disease first and then layering the neurobehavioral stressor accords with common practice for chronic medical-psychiatric comorbidity models [39, 40] and mirrors clinical chronology, in which persistent somatic disease, therapeutic burden, and low-grade inflammation often precede mood symptoms [41]. Smoke alone produced COPD-like defects after 8–16 weeks, consistent with prior work [42, 43], whereas robust depression-like phenotypes appeared only after CUMS, as expected from the reliability of CUMS to induce anhedonia and behavioral despair [44–46] and its advantages over alternative stress paradigms in multisystem models [47, 48]. Avoiding intratracheal LPS minimized confounding systemic inflammation, procedure-related mortality, and interference with stress induction [49, 50], improving stability and interpretability.

Although COPD is clinically associated with an increased susceptibility to depression, accumulating preclinical evidence indicates that cigarette-smoke exposure alone rarely induces stable or statistically robust depressive-like behaviors in rodent models. Consistent with prior reports, our preliminary CS-only cohort exhibited mild reductions in locomotor activity and sucrose preference, but these changes did not reach statistical significance, suggesting that smoke exposure by itself is insufficient to reproduce the full behavioral spectrum of COPD-related affective disturbances. For this reason, CUMS was incorporated to enhance the reliability and reproducibility of the neurobehavioral phenotype. This sequential strategy—beginning with chronic pulmonary injury and followed by stress sensitization—aligns with established approaches for modeling chronic medical–psychiatric comorbidity and reflects the clinical trajectory in which prolonged somatic illness, inflammatory burden, and physiological stress often precede the onset of mood symptoms. We acknowledge, however, that the addition of CUMS introduces an exogenous stressor that may partially intersect with endogenous COPD-related neuropsychiatric mechanisms. This constitutes a limitation of the model and should be considered when interpreting disease-specific versus stress-related contributions to the observed behavioral and molecular alterations.

Our focus on the MAPK/NF-κB pathway was informed by both exploratory and established evidence. Pilot exosomal miRNA sequencing revealed consistent enrichment of MAPK pathway targets in comorbidity and after esketamine treatment, nominating this pathway as a regulatory hub. The MAPK and NF-κB cascades are functionally intertwined, converging on transcriptional programs that drive cytokine release, oxidative stress, and apoptosis [51]. Both have been implicated in COPD-related airway inflammation and in stress-induced neuroinflammation [52]. Thus, concurrent activation of this axis in lung and hippocampus is mechanistically plausible and clinically relevant. Targeting this pathway allowed us to validate its role in comorbidity and to determine whether esketamine exerts therapeutic benefit through its suppression.

In the comorbid model, MAPK/NF-κB signaling was synchronously activated in lung and hippocampus—elevated p-p38, p-JNK, and p-p65—accompanied by higher circulating cytokines and oxidative stress. In the lung, the strongest activation occurred in the COPD and CY groups, whereas the YY group exhibited only minimal changes, indicating that cigarette-smoke exposure was the primary driver of pulmonary signaling abnormalities. This pattern matches smoke-triggered macrophage and epithelial activation with ROS and cytokine amplification that sustain remodeling and airflow limitation [53–55]. Slightly greater activation in CY than in COPD suggests stress-related neuroimmune inputs that potentiate peripheral inflammation [56]. In the hippocampus, neuronal apoptosis, microglial activation, and increased p-p38/p-p65 in CY and YY are consistent with neuroinflammatory mechanisms of depression—glial priming, cytokine induction, impaired neurogenesis, and synaptic dysregulation [57]. CUMS activates TLR-dependent cascades, sustaining microglial pro-inflammatory states and reducing plasticity [58], and diverts tryptophan metabolism through IDO, lowering 5-HT bioavailability [59]. Notably, hippocampal p-JNK was elevated in YY and CY rats but not in COPD rats, indicating that JNK activation in the hippocampus is more strongly associated with chronic stress than with smoke exposure. Hippocampal p-JNK differences were detected by Western blot but not by immunohistochemistry, likely reflecting subregion heterogeneity, temporal dynamics, and sampling limits; broader regional and temporal mapping may resolve this.

The bidirectional “lung-brain” coupling is mechanistically plausible. Chronic pulmonary inflammation can disrupt the blood-brain barrier and propagate danger signals to the CNS, whereas HPA-axis activation can amplify peripheral inflammation, creating a feed-forward loop that converges on MAPK/NF-κB nodes [60]. Emerging clinical evidence further supports this bidirectional lung–brain coupling. In COPD, systemic inflammatory mediators such as GDF-15 have been shown to correlate with reduced physical activity and cognitive vulnerability, indicating that peripheral inflammatory and metabolic stress can directly map onto central functional decline [61]. Under our experimental conditions, hippocampal activation in COPD alone was minimal, suggesting that the central consequences of peripheral inflammation may become more apparent when cigarette-smoke exposure is accompanied by chronic stress, rather than occurring with cigarette-smoke exposure alone.

Against this background, esketamine produced coherent, cross-organ benefits. In comorbid rats, it improved ventilatory performance (EF50, PEF, FVC, FEV_0.1/FVC), alleviated depression-like behaviors (higher sucrose preference, reduced immobility, greater locomotion), and mitigated lung and hippocampal pathology. Systemically, it lowered pro-inflammatory cytokines and oxidative stress and increased IL-10 and SOD. At the signaling level, it suppressed phosphorylation of p38, JNK, and p65 in both tissues. These findings align with reports of rapid central antidepressant actions together with anti-inflammatory effects in peripheral organs [62]. Diprovocim, a potent TLR1/2 agonist that activates MyD88-dependent MAPK/NF-κB cascades [63], partially reversed these benefits—attenuating lung protection and histologic improvement and re-inducing microglial activation, neuronal apoptosis, and depression-like behaviors—consistent with central roles for MAPK/NF-κB in COPD and stress-related neuroinflammation [64]. Importantly, Diprovocim’s ability to attenuate—but not completely abolish—the protective effects of esketamine lends further support to the notion that modulation of MAPK/NF-κB signaling represents a key mechanism underlying esketamine’s cross-organ therapeutic actions.

In NR8383 alveolar macrophages (CSE) and HAPI microglia (LPS + CSE), we modeled peripheral and central inflammation under comorbid conditions. In depression research, LPS is commonly used to activate microglia and mimic neuroinflammatory states associated with depressive-like behaviors [65], whereas CSE has been applied to assess the direct impact of smoke exposure on central inflammation [66]. Because our animal model combined cigarette-smoke exposure with CUMS to induce COPD-depression comorbidity, we accordingly used LPS together with CSE in vitro to more accurately simulate neuroinflammatory features under comorbid conditions. Although a CSE-only microglia group was not included, which limits our ability to fully distinguish LPS-dependent from CSE-specific effects, we acknowledge this as a study limitation and will address it in future experiments. In these models, esketamine increased viability, reduced IL-6/IL-8/TNF-α and MDA/MPO, increased IL-10/SOD, and downregulated p-p38/p-JNK/p-p65 (with cell-type-specific patterns). Diprovocim counteracted these effects. Using standardized CSE improved reproducibility over manual preparations and supports translational relevance [67]. The relatively modest p38 suppression in NR8383 compared with in vivo lung tissue likely reflects cell-type and microenvironment dependencies; p38’s sustained role in particulate-triggered macrophage responses may blunt ex vivo modulation. Co-culture systems and time-course studies could clarify these contingencies.

Microbiota profiling showed organ-specific dysbiosis in comorbidity with partial restoration after esketamine. In the gut, CY rats had reduced evenness (lower Shannon, higher Simpson), enrichment of Ligilactobacillus, and depletion of taxa linked to epithelial integrity and anti-inflammatory tone (e.g., Dubosiella, Bifidobacterium, Faecalibaculum, Romboutsia)—a signature compatible with barrier impairment and systemic inflammation in COPD and depression. Ligilactobacillus may be protective at baseline yet expand reactively under inflammatory stress, potentially displacing SCFA producers [68, 69]. Depletion of Bifidobacterium may perturb tryptophan-AhR/HCA3 signaling and Treg-mediated tolerance [70], while loss of SCFA-linked taxa may further erode epithelial and neuroimmune resilience [71]. In the lung, Lactobacillus murinus—a candidate immunoregulatory species—was depleted, whereas Proteus penneri, Elizabethkingia meningoseptica, Providencia rettgeri, and Klebsiella pneumoniae were enriched, indicating a shift toward conditional pathogens [72–75]. Correlations linked these microbial changes to lung function, behavior, and inflammatory/oxidative indices, supporting a pathophysiologic role for lung-gut dysbiosis.

Esketamine partly normalized both ecosystems: gut richness and evenness improved with enrichment of Dubosiella, Intestinimonas, and Colidextribacter and other SCFA/anti-inflammatory taxa; aberrant Ligilactobacillus expansion receded. In the lung, Lactobacillus murinus increased and conditional pathogens declined. PICRUSt2 suggested parallel shifts in microbial functions. Although classically an NMDAR antagonist, esketamine may restore mucosal homeostasis through systemic anti-inflammatory actions, with feedback along the gut-lung-brain axis. The blunted microbiota recovery with Diprovocim underscores the need to suppress inflammatory signaling to enable ecological re-equilibration.

Limitations of this study should be acknowledged. First, microbiota profiling was performed only at a single terminal time point, providing a cross-sectional rather than temporal view of gut and lung microbial dynamics. Because samples were collected immediately after post-treatment behavioral and pulmonary assessments, it remains unclear whether microbial shifts preceded, accompanied, or followed the physiological and inflammatory improvements. Second, although esketamine-associated microbial changes correlated with lung function, behavioral phenotypes, and systemic inflammatory markers, the present data cannot establish causality or differentiate direct drug-microbiota effects from secondary consequences of reduced inflammation. Third, the study did not include esketamine-only or Diprovocim-only control groups, limiting the ability to fully distinguish intrinsic drug effects from their modulation of comorbid pathology. Future work incorporating longitudinal sampling, microbiota manipulation approaches such as fecal transfer, metabolomic profiling, and expanded control groups will be essential to define causal pathways and drug-specific mechanisms.

These findings identify MAPK/NF-κB as a shared, tractable node connecting pulmonary and central inflammation in COPD-depression comorbidity and nominate esketamine as a systemic modulator that improves organ phenotypes while fostering microbiota realignment. Clinically, they support testing anti-inflammatory-neuroregulatory strategies in patients with COPD and mood disorders, with consideration of microbiota-directed adjuncts. Future priorities include temporal mapping across lung-gut-brain compartments, dose-response and durability studies, strain-resolved meta-omics, and microbiota manipulation to establish mechanism.

In summary, we developed a stable comorbid COPD-depression model characterized by systemic inflammation, oxidative stress, and MAPK/NF-κB activation in lung and hippocampus. Esketamine improved pulmonary and behavioral phenotypes, attenuated inflammation and oxidative stress, and suppressed pathway activation, whereas Diprovocim partially reversed these effects. Gut and lung dysbiosis tracked with physiologic and inflammatory readouts and was partially corrected by esketamine. Together, these data delineate an inflammatory-microbial network along the lung-brain axis and support esketamine as a multi-target intervention capable of modulating both signaling and microecology in COPD-depression comorbidity. The proposed mechanistic framework is summarized in Fig. 8.

Fig. 8.

Mechanistic model of esketamine in COPD-depression comorbidity. Chronic cigarette-smoke exposure combined with CUMS induced airflow limitation, depressive behaviors, hippocampal injury, systemic inflammation, and oxidative stress via MAPK/NF-κB activation and microbiota imbalance. Esketamine improved pulmonary and behavioral outcomes, inhibited apoptosis and microglial activation, restored cytokine and oxidative balance, and partially corrected gut-lung dysbiosis. Diprovocim attenuated these effects

Authors’ contributions

Conception and design: A Liu, X-Q Zhang, C-Y Li and Z-Y Chen. Collection and assembly of data: A Liu, X-Q Zhang, J-X Guo, Q-Y Wen, K Dai, W-J Zheng and J-H Wu. Data analysis and interpretation: A Liu, X-Q Zhang, J-X Guo and Z-Y Chen. Manuscript writing: All authors. Final approval of manuscript: All authors.

Funding

This study was supported by the Horizontal Project of the Second Affiliated Hospital of Fujian Medical University (Grant Number: 2025004HX).

Data availability

The gut microbiota sequencing data have been deposited in the NCBI BioProject database under accession number PRJNA1289055 (SRA accession: SRP599604), available at: [https://dataview.ncbi.nlm.nih.gov/object/PRJNA1289055] (https:/dataview.ncbi.nlm.nih.gov/object/PRJNA1289055) . The lung microbiota sequencing data have been deposited in the NCBI BioProject database under accession number PRJNA1289098 (SRA accession: SRP599622), available at: [https://dataview.ncbi.nlm.nih.gov/object/PRJNA1289098] (https:/dataview.ncbi.nlm.nih.gov/object/PRJNA1289098) . All other data supporting the findings of this study are available within the article and its Supplementary Information files, or from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study adhered to ethical guidelines approved by the Ethics Committee of The Second Affiliated Hospital of Fujian Medical University (Approval No. 2025 − 198).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.