Skip to main content
NCBI home page
Molecular Medicine logoLink to Molecular Medicine
. 2026 Jul 10;32:173. doi: 10.1186/s10020-026-01566-6

Hydrogen alleviates sepsis-associated encephalopathy via ATG9B-dependent mitophagy

Xiaoli Yu 1,2,#, Xiaofan Huang 1,#, Yu Song 1,#, Jianfeng Liu 1,2, Fei Huang 2, Jiatian Cui 1,2, Keliang Xie 1,2,✉, Yan Cui 1,3,✉
PMCID: PMC13637375  PMID: 42432483

Abstract

Background

Sepsis-associated encephalopathy (SAE) is a severe complication with high mortality and limited therapeutic options. Molecular hydrogen (H₂) has shown neuroprotective potential, but its mechanism remains elusive.

Methods

SAE was induced in C57BL/6J mice by cecal ligation and puncture, with or without 2% H2 inhalation. HT22 hippocampal neurons were challenged with lipopolysaccharide in hydrogen-rich medium. H₂ intervention was administered via inhalation or hydrogen-rich medium. Cognitive function was evaluated by novel object recognition and Y-maze tests.

Transcriptome sequencing was used to identify key signaling pathways regulated by H₂. The neuroprotective mechanism of H₂ was explored by assessing mitophagy, apoptosis, reactive oxygen species, and mitochondrial membrane potential. ATG9B was knocked down using siRNA in vitro and AAV-shRNA in the hippocampal CA1 region in vivo to clarify its function.

Results

H₂ inhalation increased 7-day survival in CLP-induced septic mice from 40% to 75%, reduced systemic and hippocampal pro-inflammatory cytokines, and alleviated hippocampal neuronal damage and cognitive dysfunction. Transcriptomic profiling identified ATG9B as the most significantly upregulated mitophagy-related gene by H2. Mechanistically, H₂ upregulated ATG9B expression, which in turn enhanced PINK1-Parkin-mediated mitophagy flux, thereby coordinating mitophagosome formation and lysosomal fusion, leading to clearance of damaged mitochondria and reduced neuronal apoptosis. ATG9B knockdown completely abolished H₂-induced mitophagy flux, mitochondrial protection, and cognitive improvement.

Conclusions

H₂ alleviates SAE by upregulating ATG9B and restoring PINK1-Parkin-dependent mitophagy. The ATG9B-mitophagy axis represents a novel therapeutic target, and H₂ inhalation emerges as a potential strategy for sepsis-associated cognitive impairment.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s10020-026-01566-6.

Keywords: Hydrogen, Sepsis-associated encephalopathy, Mitophagy, ATG9B, Cognitive dysfunction

Introduction

Sepsis-associated encephalopathy (SAE) is a diffuse cerebral dysfunction secondary to systemic infection, presenting with symptoms ranging from delirium to long-term cognitive impairment (Sonneville et al. 2023, Li et al. 2025). Approximately half of ICU patients with sepsis develop SAE (Sonneville et al. 2017), which significantly increases mortality and reduces quality of life (Widmann and Heneka 2014). The hippocampus, a brain region critical for learning and memory (Knierim 2015), is particularly vulnerable to the insult of sepsis. Pathophysiologically, systemic inflammation disrupts the blood-brain barrier, activates microglia, and triggers neuronal injury and synaptic loss in the hippocampus (Kikuchi et al. 2019, Yin et al. 2023, Manabe and Heneka 2022). Despite its clinical importance, effective therapeutic strategies targeting the mechanisms underlying SAE remain lacking.

Accumulating evidence has implicated mitochondrial dysfunction as a central driver of neuronal injury in SAE (Yang et al. 2024, Gong et al. 2025). Damaged mitochondria generate excessive reactive oxygen species (ROS) and trigger apoptotic cascades (Hou et al. 2023). Mitophagy, the selective autophagic clearance of dysfunctional mitochondria, serves as a critical quality-control mechanism essential for neuronal homeostasis (Ding et al. 2022, Fang et al. 2019). This process involves the recognition of damaged mitochondria, encapsulation within autophagosomes, and subsequent lysosomal degradation (Picca et al. 2023). Impaired mitophagic flux is increasingly recognized as a contributing factor to the pathogenesis of various neurodegenerative diseases and cognitive disorders, underscoring its fundamental role in brain function (Yang et al. 2024, Liu et al. 2025). These observations position mitophagy as a potential therapeutic target in SAE.

Among the core autophagy-related proteins (ATGs), ATG9 is unique as the only transmembrane protein, and it plays a key role in efficient autophagosome biogenesis by providing and organizing the membrane source for phagophore expansion (Liang et al. 2020, Chiduza et al. 2024). Two homologous isoforms exist: ubiquitously expressed ATG9A and tissue-specific ATG9B. While ATG9A is ubiquitously expressed (Choi et al. 2024), ATG9B exhibits a tissue-specific expression pattern, with notable enrichment in certain immune cells and specific brain regions, suggesting unique functions beyond basal autophagy (Choi et al. 2024, Wang et al. 2022). Although ATG9 family proteins in basal autophagy have been well established, the precise regulatory functions of the tissue-specific ATG9B in neuron-selective mitophagy and its potential as a therapeutic target for SAE remain unexplored.

Molecular hydrogen (H₂) has emerged as a novel therapeutic agent with demonstrated efficacy in various models of organ injury, including cerebral ischemia and sepsis (Ohsawa et al. 2007, Yoshiura et al. 2025). Its pleiotropic effects include selective neutralization of cytotoxic hydroxyl radicals and modulation of inflammatory and anti-apoptotic properties (Qiu et al. 2019, Meng et al. 2025). While H₂ inhalation has been shown to improve survival and mitigate organ dysfunction in sepsis models (Wang et al. 2024, Cui et al. 2024), the precise neuroprotective mechanisms of H2 in SAE remain poorly defined. In particular, whether H2 protects against SAE by regulating ATG9B-related mitophagy remains unknown.

In this study, we tested the hypothesis that H₂ exerts neuroprotective effects in SAE by upregulating ATG9B to restore PINK1-Parkin-mediated mitophagic flux. We evaluated H2’s therapeutic effect in a CLP mouse model and LPS-stimulated HT22 neurons, and demonstrated that ATG9B is essential for H2-mediated activation of the PINK1-Parkin mitophagy pathway. Our findings show that the ATG9B-mitophagy axis is a novel and critical mechanism for H2-mediated neuroprotection, making it a promising therapeutic target in SAE. (Sch. 1)

Scheme 1.

Scheme 1

Open in a new tab

H₂ Upregulates ATG9B to Restore Mitophagic Flux and Cognitive Function in SAE

Materials and methods

Animals and ethics

Male C57BL/6J mice (6–8 weeks old, 20–25 g) were obtained from the Experimental Animal Center of the Chinese Academy of Medical Sciences, Beijing, China. Animals were housed under specific pathogen-free (SPF) conditions with a 12-h light/dark cycle, an ambient temperature of 22 ± 2 °C, a humidity of 50%-60%, and free access to food and water. All experimental procedures were approved by the Experimental Animal Management Committee of Tianjin Medical University [IRB2026-DW-31].

Sepsis model

Sepsis was induced by CLP (cecal ligation and puncture) as previously described (Dejager et al. 2011). Briefly, mice were anesthetized with 2% isoflurane. A 1 cm midline abdominal incision was made, and the cecum was exteriorized. Approximately 50% of the cecum was ligated with a 4‑0 silk suture just distal to the ileocecal valve and punctured once with a 21-gauge needle. A small amount of fecal content was extruded to ensure patency. The cecum was returned to the peritoneal cavity, and the abdomen was closed in two layers. Sham-operated mice underwent the same procedure without ligation and puncture. All animals received 1 mL of pre-warmed sterile saline subcutaneously immediately after surgery for fluid resuscitation. Seven days after surgery, the survival rate of mice in each group (20 mice in each group) was monitored, and their behavioral performance was evaluated. 24 h after surgery, the mouse sepsis score (MSS) was used for evaluation, and mouse serum and hippocampal tissue were collected for subsequent experiments.

Hydrogen gas (H₂) treatment

The administration of H₂ in vivo refers to the existing protocol (Qi et al. 2023). Briefly, the groups that need to inhale H₂ are placed in a sealed room at 1 h and 6 h after surgery, and 2% H₂ is inhaled for 60 min. A TF-1 flow meter (YUTAKA Engineering, Japan) was used to deliver H₂-air mixed gas (2%) at a rate of 4 L/min, and a HY-ALERTA 500 detector (H₂ Scan, USA) was used to continuously monitor the H₂ concentration. The control group was given air inhalation under the same conditions. For in vitro experiments, a hydrogen-rich medium was prepared as described below.

Preparation of hydrogen‑rich medium (HM)

A hydrogen-rich medium was prepared according to a previously described method with modifications (Yu et al. 2015). Briefly, H₂ gas was bubbled into Dulbecco’s modified Eagle’s medium (DMEM) under high pressure (0.4 MPa) for 4 h to reach a saturation concentration of approximately 0.6 mM. The saturated HM was then stored in a sealed aluminum bag or glass bottle under atmospheric conditions at 4 °C. HM was freshly prepared each week and sterilized by γ-irradiation (Co-60) before use. For cell experiments, HM was used within 30 min after opening the sealed container to avoid hydrogen loss.

Cell culture and treatment

Murine hippocampal neuronal HT22 cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 µg/mL streptomycin at 37 °C in 5% CO₂. To simulate neuroinflammation, cells were treated with 1 µg/mL lipopolysaccharide (LPS, Sigma Aldrich) for 24 h. For H₂ co-treatment, the culture medium was replaced with hydrogen-rich DMEM immediately before LPS challenge and refreshed every 6 h. In some experiments, cells were pretreated with 5 mM 3-methyladenine (3-MA, Sigma Aldrich) for 2 h prior to LPS/H₂ exposure.

siRNA knockdown of ATG9B

Small interfering RNA (siRNA) targeting mouse ATG9B (si-ATG9B) and negative control siRNA (si-NC) were designed and synthesized by OBiO (Shanghai, China). The sequences were: si ATG9B#1 sense 5’ CCUACAUCAGCUUCAUCAATT 3’. HT22 cells (50–60% confluent) were transfected with 50 nM siRNA using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol. After 48 h, cells were treated with LPS and/or H₂. Knockdown efficiency was verified by Western blotting.

Stereotaxic injection of adeno‑associated virus (AAV) into the hippocampus

To specifically regulate ATG9B expression in the hippocampus, an adeno-associated virus encoding shRNA targeting mouse ATG9B (AAV, sh-ATG9B) and the corresponding scrambled control virus (AAV, sh-NC) were constructed. Stereotaxic injection was performed according to the previous protocol (Keiser et al. 2018). The mice were anesthetized with isoflurane, secured in a stereotaxic apparatus, and the head was adjusted so that the front and rear fontanelles were at the same level. After disinfecting the skin, make a midline incision to expose the skull. The bilateral injection sites (AP -1.8 mm, ML ± 1.5 mm) were determined using bregma coordinates and then drilled. Using a microsyringe, inject 1 µL AAV (titer 1 × 1012 vg/mL) into the CA1 region of each hippocampus at a rate of 0.1 µL/min to a depth of 1.6 mm below the dura mater (DV -1.6 mm). After the injection is complete, leave the needle in place for 10 min, then slowly withdraw it. After suturing the incision, the mice were placed on a heating pad to recover. Three weeks after virus injection, follow-up experiments were conducted.

Enzyme‑linked immunosorbent assay (ELISA)

Twenty-four hours after surgery, the mice were euthanized, and orbital venous blood was collected and allowed to stand at room temperature for 1 h. The samples were centrifuged at 1000 g for 20 min to obtain serum. Use TNF-α (mlC50536-1, mlbio), IL-1β (mlC50300-1, mlbio), and IL-6 (mlc50325, mlbio) for detection, strictly following the instructions. Determine the levels of LDH, ALT, and BNP according to the kit instructions. The analysis was performed using an ELISA kit (Elabscience).

Behavioral tests

Open field test (OFT)

A square test chamber (50 × 50 × 40 cm) is used to measure spontaneous activity in mice. Each mouse is allowed to explore freely for 5 min. The total distance traveled during the test is recorded. After each mouse is tested, the test apparatus is cleaned with 75% ethanol to remove any residual odors.

New object recognition test (NOR)

The experimental setup uses a standard open-air plastic container consistent with the open-field experiment. On the first day of the experiment, no objects are placed inside the container; mice are placed inside to explore freely for 10 min to adapt to the environment. Conduct a familiarization test 24 h later. Place two familiar objects, A1 and A2, in the container and put the mice in to explore freely for 5 min. After each mouse is tested, wipe the container with 75% ethanol to remove the odor residue. After 24 h, the familiar object A2 was replaced with a new object B with a different shape and appearance, and the mice were allowed to explore freely for 5 min again. The preference index is calculated according to the following formula: time to explore new objects / (time to explore new objects + time to explore familiar objects).

Y-maze test

A Y-maze consisting of three arms of equal length (A, B, and C), arranged at 120° angles to each other, is used to test spatial working memory. During the experiment, mice undergo a 30-minute adaptation period, after which Arm C is blocked, allowing the mice to explore Arms A and B for 5 min. After 1 h, all three arms were opened, and the mouse was placed on arm A and allowed to explore freely for 5 min. Record the number of times you enter the new arm (C) and the time you stay in that arm.

Measurement of intracellular reactive oxygen species

Intracellular ROS levels were assessed using the Reactive Oxygen Species Assay Kit (Beyotime, S0034M) according to the manufacturer’s protocol. Briefly, HT22 cells were seeded in 12-well plates and treated as described. After the indicated treatments, cells were washed twice with PBS and incubated with 10 µM DCFH-DA diluted in probe diluent at 37 °C for 20 min. Following three washes with PBS to remove excess probe, fluorescence intensity was measured using a fluorescence microscope (Olympus IX73). For quantitative analysis, at least five random fields per well were captured, and the mean fluorescence intensity was analyzed using ImageJ software.

Measurement of mitochondrial membrane potential

Mitochondrial membrane potential (MMP) was evaluated using the Enhanced Mitochondrial Membrane Potential Assay Kit with JC-1 (Beyotime, C2003S). Following the designated treatments, HT22 cells were washed once with PBS and incubated with 1× JC-1 staining solution prepared by diluting JC-1 (200×) with JC-1 staining buffer at 37 °C for 20 min. Cells were then washed twice with JC-1 staining buffer to remove unbound probe. Fluorescence images were acquired using a fluorescence microscope (Olympus IX73). For quantitative analysis, the ratio of red to green fluorescence intensity was calculated using ImageJ software.

Propidium iodide (PI) staining for cell apoptosis

Cell death was assessed using the Propidium Iodide Apoptosis Detection Kit (Beyotime, C1008M). After the indicated treatments, HT22 cells cultured in 12‑well plates were gently washed twice with pre‑warmed PBS. Then, 300 µL of PI working solution was added to each well, and the plates were incubated at 37 °C in the dark for 15–20 min. Cells were then washed twice with PBS. For nuclear counterstaining, cells were incubated with Hoechst 33,342 (Beyotime, C1027) for 10 min at 37 °C. Fluorescence images were immediately captured using an inverted fluorescence microscope (Olympus IX73).

Transmission electron microscope observation

Hippocampal tissue was cut into small pieces of approximately 1 mm³ and fixed overnight at 4 °C in 2.5% glutaraldehyde. After post-fixation with 1% osmium tetroxide, dehydration with graded ethanol, and embedding in Epon-Araldite resin. Prepare 50 nm ultrathin sections, stain them with 2% uranyl acetate and lead citrate in sequence, and observe the ultrastructure of mitochondria under a transmission electron microscope (Hitachi, Japan). Ultrastructural analysis was performed by an observer who was unaware of the experimental group assignments.

Western blot assay

The total protein of mouse hippocampal tissue and HT22 cells was extracted using RIPA lysis buffer containing protease inhibitors, and the protein was quantified using the BCA method. Equal amounts of protein samples were separated by SDS-PAGE and transferred to a PVDF membrane. Blocked with 5% skim milk for 2 h at room temperature, and then incubated with the following primary antibodies: BCL-2 (Abcam, ab182858, RRID: AB_2715467, 1:1000), BAX (Abcam, ab32503, RRID: AB_725631, 1:2000), PINK1 (Santa Cruz, sc-517353, RRID: AB_2923157, 1:500), Parkin (Abcam, ab77924, RRID: AB_1566559, 1:2000), LC3B (Proteintech, 81004-1-RR, RRID: AB_2923695, 1:3000), SQSTM1 (Selleckchem, F0106, RRID: AB_3739912, 1:10000), ATG9B (Thermo Fisher, PA5-20998, RRID: AB_11153916, 1:1000) and GAPDH (Affinity, AF7021, T0004, RRID: AB_2839421, AB_2833041, 1:10000).

Primary antibodies were incubated overnight at 4 °C. The next day, the membrane was washed three times with TBST, and then incubated with the corresponding secondary antibodies (Proteintech, SA00001-1, SA00001-2, RRID: AB_2722565, AB_2722564, 1:6000) for 1 h at room temperature. Finally, images were developed and acquired using an enhanced chemiluminescence (ECL) system.

Immunofluorescence staining

The cells were fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100, and blocked with 3% bovine serum albumin for 2 h at room temperature. This was followed by incubation with the corresponding primary antibody overnight at 4 °C. The primary antibodies used included: LC3B (Proteintech, 18725-1-AP, RRID: AB_2137745, 1:200), SQSTM1 (Servicebio, GB121998, 1:500), and ATG9B (Thermo Fisher, PA5-20998, RRID: AB_11153916, 1:100). Then Alexa Fluor 488 or 594 labeled secondary antibodies were added, respectively. Lysosomes were stained with LysoTracker reagent kit (Beyotime, C1046), and mitochondria were stained with MitoTracker reagent (Beyotime, C1035). Cell nuclei were counterstained with DAPI or Hoechst 33,258 staining. Images were collected using an Olympus BX53 microscope.

Tissue immunofluorescence staining. Mouse brains were post-fixed in 4% paraformaldehyde (PFA) overnight. After dehydration through a sucrose gradient, the tissue was embedded in OCT and serially sectioned in the coronal plane on a freezing microtome at 10 μm thickness. The sections were permeabilized with 0.3% Triton X-100, blocked with 5% bovine serum albumin (BSA) at room temperature, and then incubated with primary antibody mixture at 4 °C overnight. Primary antibodies include: PSD95 (Abcam, ab238135, RRID: AB_2895158, 1:100), SYP (Abcam, ab32127, RRID: AB_2286949, 1:400) and ATG9B (Thermo Fisher, PA5-20998, RRID: AB_11153916, 1:100). Subsequently, sections were incubated with Alexa Fluor-labeled corresponding secondary antibodies (1:200) for 1 h at room temperature, and nuclei were counterstained with DAPI. Images were collected using a confocal microscope (Olympus BX53), and ImageJ was used to quantify fluorescence intensity in the CA1 region of the hippocampus.

Nissl staining

Mouse brain tissue was taken and fixed in 4% paraformaldehyde for 24 h. After fixation, the tissues were dehydrated through a graded ethanol series, cleared in xylene, and embedded in paraffin. Use a microtome to perform coronal serial sections with a thickness of 10 μm. The sections were deparaffinized in xylene, rehydrated in graded ethanol, and then stained with Nissl. The integrity of neurons in the hippocampus was observed under a light microscope (Olympus BX53). Neuron counting analysis was performed using ImageJ software.

Flow cytometry analysis of cell apoptosis

Flow cytometry analysis was performed using the BD CytExpert For DxFLEX flow cytometer (BD, USA) to evaluate neuronal apoptosis. HT22 cells were seeded in 6-well plates and cultured until reaching approximately 70% confluence, followed by the indicated treatments. After treatment, cells were harvested by trypsinization without EDTA, washed twice with cold PBS, and gently resuspended in 1× binding buffer. According to the manufacturer’s protocol for the Annexin V-FITC/PI Apoptosis Detection Kit (Beyotime, C1062), cells were incubated with 5 µL of Annexin V-FITC and 5 µL of propidium iodide (PI) in the dark at room temperature for 15 min. The samples were then immediately analyzed by flow cytometry. A gating strategy using FSC-A and SSC-A analysis was applied to exclude cellular debris and aggregates. The fluorescence signals were collected in the FITC (Annexin V) and PE (PI) channels. Flow cytometry data were analyzed using FlowJo v.10 software.

Statistical analysis

All data were analyzed using GraphPad Prism 10 (GraphPad Software, USA). The quantitative data were expressed as mean ± standard deviation (SD). The Shapiro-Wilk test was used to assess normality. For comparisons among multiple independent groups, one-way ANOVA followed by Tukey’s post-hoc test was performed. Survival rates were analyzed using Kaplan-Meier curves and log-rank tests. p-value < 0.05 was considered significant.

Results

Hydrogen ameliorates sepsis-induced neuroinflammation, cognitive impairment, and neuronal apoptosis

To evaluate the therapeutic potential of molecular hydrogen (H₂) in sepsis-associated encephalopathy (SAE), an in vivo sepsis model was established in mice. C57BL/6 mice underwent cecal ligation and puncture (CLP) to induce polymicrobial sepsis. As depicted in the experimental timeline (Fig. 1A), the experimental design included dedicated control groups treated with 2% H₂ alone to assess its biosafety profile.

Fig. 1.

Fig. 1

Open in a new tab

Hydrogen ameliorates sepsis-induced neuroinflammation, cognitive impairment, and neuronal apoptosis. A Flowchart of the in vivo animal experiment. B 7-day survival rate of mice (n = 20). C Serum levels of pro-inflammatory cytokines (TNF-α, IL-1β, IL-6) in mice, measured by ELISA (n = 6). D Total distance traveled (cm) by each group in the open field test(n = 10). E Novel object recognition test: preference index of mice in each group (n = 10). F-H Y-maze test: exploration time and entry frequency for the novel arm (n = 10). I Nissl staining of the hippocampus. Scale bars: 200 μm (overview), 50 μm (inset). J Number of abnormal neurons in the CA3 region (n = 6). K-L Expression levels of hippocampal apoptosis-related proteins Bcl-2 and Bax were determined by Western blotting, and the Bcl-2/Bax ratio was calculated. Statistics: Log‑rank test (B); one‑way ANOVA with Tukey’s post hoc test (C–L). Data are mean ± SD from independent experiments. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001

In CLP mice, H₂ treatment significantly improved survival (Fig. 1B). To assess the systemic inflammatory response, pro-inflammatory cytokine levels in the serum were measured. ELISA analysis revealed that CLP surgery led to significant elevations in circulating TNF-α, IL-1β, and IL-6. Remarkably, inhaling H₂ reduced this cytokine surge, resulting in lower levels (Fig. 1C). Notably, H₂ treatment alone did not alter basal cytokine levels, confirming its inert nature and the absence of pro-inflammatory side effects. Furthermore, to further evaluate sepsis-induced organ dysfunction, serum LDH, ALT, and BNP levels were measured. The results showed that these markers were significantly elevated in the CLP group, while H₂ treatment attenuated the extent of this elevation (Fig.S1A-C).

To exclude post-modeling motor impairments as a potential confounding factor, an open-field test was conducted, revealing no significant differences in total locomotor activity among the groups (Fig. 1D, S1D). In the novel object recognition test, CLP mice exhibited a lower preference index compared with sham-operated mice, indicating impaired learning and recognition memory. Conversely, the preference index improved following hydrogen inhalation (Fig. 1E). Similarly, in the Y-maze test, the CLP group showed a significant reduction in both the time spent in and the number of entries into arm C. Hydrogen inhalation ameliorated these deficits (Fig. 1F-H). Nissl staining of the hippocampus was used to assess neuronal health and structural integrity. In the sham-surgery and H₂ alone group, hippocampal neurons exhibited a compact structure and distinct layering. However, marked pathological changes in the hippocampus (particularly in the CA3 region) were observed in the CLP group, characterized by neuronal atrophy and nuclear condensation. H₂ treatment effectively preserved neuronal morphology and Nissl body density (Fig. 1I-J).

Given established links between neuroinflammation and neuronal apoptosis, hippocampal cell death was assessed. Western blot analysis of apoptosis regulators revealed that LPS stimulation shifted the Bcl-2/Bax balance toward apoptosis, resulting in decreased Bcl-2 and increased Bax levels. H₂ co-treatment reversed this imbalance, increasing Bcl-2 and suppressing Bax expression (Fig. 1K-L). This modulation provides a molecular basis for the observed decrease in neuronal death.

In conclusion, these data confirm that H₂ not only reduces the overall mortality rate in CLP mice but also alleviates cytokine storms, reduces neuroinflammation, inhibits neuronal apoptosis, and protects hippocampal neuronal structure. Most importantly, hydrogen can reverse sepsis-induced cognitive dysfunction. These fundamental findings provide a solid theoretical foundation for subsequent research into mechanisms.

Transcriptomic analysis reveals that H2 enriches mitophagy-related pathways and upregulates ATG9B in the hippocampus of septic mice

Based on the established neuroprotective effects of H₂, we further explored the underlying cellular mechanisms. Since mitochondrial dysfunction plays a crucial role in sepsis-induced neuronal injury (Yin et al. 2026), the damaged mitochondria act as a persistent source of ROS and pro-inflammatory signals, thereby exacerbating neuroinflammation (Wang et al. 2024). Therefore, we focused on mitochondrial quality control. To gain an unbiased insight into the molecular pathways modulated by H₂, RNA sequencing of hippocampal tissue treated with CLP or CLP + H₂ was performed. The volcano plot highlighted a distinct transcriptional profile in the H₂-treated group. Differential gene expression analysis identified 1588 significantly upregulated and 1704 downregulated genes in the CLP group compared to the sham, confirming widespread transcriptional dysregulation (Fig. 2A). Figure 2B presents a volcano plot illustrating the distribution of differentially expressed genes (DEGs) between the CLP+H2 and CLP groups. The analysis identified 3735 genes with significant expression changes: 1200 up-regulated and 886 down-regulated, based on |log2FC| ≥ 2 and p-value ≤ 0.05.

Fig. 2.

Fig. 2

Open in a new tab

Transcriptomic analysis reveals that H2 enriches mitophagy-related pathways and upregulates ATG9B in the hippocampus of septic mice. A Volcano plot depicting DEGs between CLP and the sham group. B Volcano plot showing DEGs between the CLP+H2 and CLP groups. C KEGG pathway enrichment analysis of genes upregulated in the CLP compared to the sham group. D KEGG pathway enrichment analysis of genes upregulated in the CLP+H2 compared to the CLP group. E-G Heat map of DEGs involved in autophagy, mitophagy, and apoptosis modulated by hydrogen therapy. Threshold for differential expression: |log2FC| ≥ 2 and p-value ≤ 0.05

The DEGs were functionally annotated according to the Kyoto Encyclopedia of Genes and Genomes (KEGG). Compared to the Sham group, the DEGs in the CLP group are mainly associated with the TNF signaling pathway, NF-Kappa B signaling pathway, and cytokine-cytokine receptor interaction pathways. (Fig. 2C). This transcriptomic signature corroborates our earlier findings that CLP induces an inflammatory signaling cascade. Compared to the CLP group, the CLP+H2 group is mainly associated with autophagy, apoptosis, mitophagy, lysosomes, and AMPK signaling pathways (Fig. 2D). The enrichment of autophagy and mitophagy confirms that H2 treatment potently activates these processes in the hippocampus.

Subsequently, the main DEGs were displayed in heat maps illustrating autophagy, mitophagy, and apoptosis. In the autophagy and mitophagy analyses, ATG9B was identified. It was downregulated in the CLP group yet significantly upregulated following hydrogen intervention (Fig. 2E-G).

ATG9B, the sole multi-pass transmembrane protein in the core autophagy-related (ATG) machinery, plays a non-redundant role in autophagosome biogenesis (Chiduza et al. 2024). Unlike other ATG proteins that function in the cytoplasm, ATG9B is essential for supplying and organizing membrane sources for the growing phagophore (Wang et al. 2022). Given that ATG9B plays a crucial role in the early stages of autophagosome formation, it may act as a critical downstream effector in the H₂ therapy, coordinating the membrane dynamics necessary for efficient mitophagy.

Hydrogen activates PINK1-Parkin-dependent mitophagy to alleviate neuronal injury in vitro and in vivo

Transcriptomic analysis revealed that H₂ treatment enriched pathways related to autophagy, mitophagy, and apoptosis in the hippocampus of septic mice. Therefore, it is necessary to elucidate whether H₂ alleviates sepsis-associated neuronal injury. Firstly, the key components of classical mitophagy pathways in hippocampal tissue were analyzed. Western blot analysis showed that, compared with the sham group, the hippocampal expression levels of PINK1 and Parkin in the CLP group were downregulated, accompanied by accumulation of p62 and decreased LC3-II, suggesting that mitophagy was inhibited. In contrast, after hydrogen treatment, the trend in the above protein was reversed, indicating improved mitophagic function. The autophagy inhibitor 3-MA (20 mg/kg) can reverse these effects caused by hydrogen (Fig. 3A-B).

Fig. 3.

Fig. 3

Open in a new tab

Hydrogen activates PINK1-Parkin-dependent mitophagy to alleviate neuronal injury in vitro and in vivo. A-D Western blot analysis and quantification of mitophagy-related and apoptosis-related proteins (n = 6). E Total distance traveled (cm) in the open field test (n = 10). F Preference index in the novel object recognition test (n = 10). G-I Y-maze test: time spent and entries into the novel arm (n = 10). J TEM images of hippocampal mitochondrial morphology. Scale bar = 1 μm (main), 500 nm (enlarged view) (n = 3; Each group consisted of three biologically independent mice). K-L ROS representative images showing HT22 treated with LPS or co-cultured with H2 and H2 + 3-MA groups, Scale bar = 50 μm (n = 3). M-N Representative images of JC 1 staining, Scale bar = 50 μm (n = 3). O-P Assessment of cell death by PI staining in HT22 cells, Scale bar = 50 μm (n = 3). Statistics: one‑way ANOVA with Tukey’s post hoc test. Data are mean ± SD from independent experiments. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001

Given that dysfunctional mitochondria that escape mitophagic clearance can trigger the intrinsic apoptosis pathway by disrupting the Bcl-2/Bax balance (Baechler et al. 2019). Under SAE conditions, H2 treatment shifted the Bcl-2/Bax balance toward an anti-apoptotic state compared to the CLP group, whereas the addition of 3-MA reversed the protective effect of hydrogen. (Fig. 3C-D). These results indicate that H₂ not only enhances the clearance of damaged mitochondria via mitophagy but also restores mitochondrial anti-apoptotic capacity. In contrast, the autophagy inhibitor 3-MA attenuated the therapeutic effects of hydrogen, confirming that hydrogen mediates enhanced mitophagy in the context of autophagy. Behavioral testing revealed that 3-MA impaired improved working memory and recognition memory, weakened the cognitive protective effects of H₂ in SAE mice (Fig. 3F-I), and did not affect their overall motor activity (Fig. 3E, S1E). Consistent with these findings, TEM revealed that mitochondria in the CLP group were swollen, with fragmented cristae and vacuolation, whereas H2 treatment promoted encapsulation of damaged mitochondria within double-membrane autophagosomes (Fig. 3J).

To evaluate the mitochondrial protective effects via H2, additional cellular-level experiments were conducted. In LPS-treated (1 µg/mL) HT22 cells, hydrogen significantly reduced intracellular ROS levels (Fig. 3K). JC-1 staining confirmed that hydrogen restored LPS-reduced mitochondrial membrane potential (MMP). However, this mitochondrial protection was antagonized by 3-MA (Fig. 3M). In addition, the number of neuronal deaths increased after LPS treatment, and 3-MA partially abrogated H2’s protective effect against LPS-induced cell death (Fig. 3O).

In summary, hydrogen activates the PINK1-Parkin pathway to enhance mitophagic flux in septic conditions. These findings position mitochondrial quality control via mitophagy as a central mechanism underlying the H₂ therapy in sepsis-associated neuronal injury.

ATG9B coordinates autophagosomal membrane dynamics and lysosomal fusion to sustain mitophagic flux

To further elucidate the functional role of ATG9B in mitophagy, loss-of-function experiments on HT22 hippocampal neurons under normal and inflammatory conditions were performed. Cells were transfected with control siRNA (si-NC) and siRNA targeting ATG9B (si-ATG9B), respectively. The RNA sequence with the highest transfection efficiency (S2A-B) was selected, and its knockdown efficiency was verified by Western blot. (Fig. 4A-B). Western blot analysis revealed that ATG9B knockdown alone under basal conditions impaired the mitophagy machinery, as evidenced by alterations in mitophagy-related proteins compared to the control group (Fig. 4C-D). In LPS-induced cells, ATG9B knockdown exacerbated this impairment, resulting in a more pronounced blockade of mitophagic flux than the LPS group (Fig. 4C-D).

Fig. 4.

Fig. 4

Open in a new tab

ATG9B coordinates autophagosomal membrane dynamics and lysosomal fusion to sustain mitophagic flux. A-B Western blot validation and quantification of ATG9B knockdown efficiency in HT22 cells (n = 3). C-D Western blot analysis and quantification of ATG9B, PINK1, Parkin, LC3, and p62 expression in HT22 cells under the indicated conditions (n = 3). E Representative immunofluorescence images showing LC3-mitochondria colocalization and colocalization coefficients. Scale bar = 10 μm (n = 3). F Representative immunofluorescence images showing LC3–p62 colocalization and the corresponding Pearson’s correlation coefficients in HT22 cells. Scale bar = 20 μm (n = 3). G Representative immunofluorescence images showing mitochondria-lysosome colocalization and the corresponding Pearson’s correlation coefficients in HT22 cells. Scale bar = 10 μm (n = 3). Statistics: one‑way ANOVA with Tukey’s post hoc test (A-D); Pearson’s correlation coefficients (E-G). Data are mean ± SD from independent experiments. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001

Given ATG9B’s established role in supplying membranes for autophagosome formation, the envelopment of mitochondria by autophagosomal membranes was examined next. Immunofluorescence (IF) staining for LC3 and MitoTracker revealed that LPS stimulation impaired the co-localization of LC3 with mitochondria, indicating that mitophagosome formation was inhibited. However, ATG9B knockdown similarly diminished this co-localization, suggesting that ATG9B is essential for the damaged mitochondria envelopment (Fig. 4E). Furthermore, the co-localization of LC3 with p62, an autophagy adaptor that bridges ubiquitinated cargo to the autophagosomal membrane, was assessed. Consistent with the above findings, ATG9B knockdown reduced LC3-p62 co-localization, indicating that ATG9B facilitates the autophagic cargo to autophagosome (Fig. 4F). These results demonstrate that ATG9B functions as a critical membrane supplier, enabling the efficient encapsulation of mitochondria by autophagosomes.

Subsequently, whether ATG9B also influences the later stages of mitophagy was further investigated. Co-localization analysis of MitoTracker with LysoTracker showed that LPS treatment diminished mitochondria-lysosome fusion. Importantly, ATG9B knockdown reduced this colocalization, indicating that ATG9B is required not only for autophagosome formation but also for subsequent fusion with lysosomes (Fig. 4G).

Collectively, these data establish that ATG9B orchestrates multiple stages of the mitophagic process, from membrane nucleation and cargo encapsulation to lysosomal fusion, and that its function is impaired under inflammatory conditions.

ATG9B is essential for hydrogen-induced mitophagic flux and neuroprotection against LPS-stimulated neuronal damage

To further verify whether ATG9B is a key downstream molecule in hydrogen-induced mitophagy flux, hydrogen treatment was administered after knockdown of ATG9B expression in HT22 hippocampal neurons. First, the ATG9B protein levels were confirmed to be consistent with the transcriptome data (Fig. S2C-D). Western blot analysis revealed that hydrogen treatment upregulated PINK1 and Parkin expression, enhanced LC3-II conjugation, and promoted p62 degradation following LPS stimulation; these effects were also observed in cells transfected with control siRNA. In contrast, ATG9B knockdown abolished these effects (Fig. 5A-B). These findings indicate that ATG9B is essential for H₂ activation of the PINK1-Parkin mitochondrial autophagy pathway.

Fig. 5.

Fig. 5

Open in a new tab

ATG9B is essential for hydrogen-induced mitophagic flux and neuroprotection against LPS-stimulated neuronal damage. A-B Western blot analysis and quantification of ATG9B, PINK1, Parkin, p62, and LC3 in HT22 cells transfected with si-ATG9B and treated as indicated (n = 3). C Representative IF images of LC3 colocalization with mitochondria in HT22 cells. Scale bar = 10 μm (n = 3). D Representative IF images of LC3 colocalization with p62. Scale bar = 20 μm (n = 3). E Representative IF images of mitochondria colocalizing with lysosomes. Scale bar = 10 μm (n = 3). F-G Flow cytometry assessment of neuronal apoptosis under different treatment conditions (n = 3). Statistics: one‑way ANOVA with Tukey’s post hoc test (A, B, F, G); Pearson’s correlation coefficients (C-E). Data are mean ± SD from independent experiments. ns, not significant, *p < 0.05, **p < 0.01, ***p < 0.001

Considering ATG9B’s established role in providing membranes for autophagosome formation, whether hydrogen promotes ATG9B-mediated physical envelopment for damaged mitochondria was assessed. IF staining with LC3 and MitoTracker revealed that H₂ treatment promoted the colocalization of LC3 with mitochondria. In contrast, ATG9B knockdown reduced the efficient encapsulation of mitochondria into autophagosomes via hydrogen (Fig. 5C). Next, the colocalization of LC3 with p62 was assessed. Consistent with the above findings, hydrogen treatment promoted the colocalization of LC3 and p62, and ATG9B knockdown reversed this effect (Fig. 5D). These data establish that H₂ coordinately regulates membrane supply and substrate recognition steps by upregulating ATG9B, thereby ensuring efficient initial isolation of damaged mitochondria.

Subsequently, whether ATG9B’s involvement in the later stages of mitophagy was also associated with H2 therapy was investigated. Through dual staining with MitoTracker and LysoTracker, the images showed that H₂ enhanced the fusion of mitochondria and lysosomes. However, in ATG9B-deficient cells, this effect was attenuated (Fig. 5E). To verify whether knocking down ATG9B would reverse the protective effect of hydrogen on neurons, this study used flow cytometry to detect the apoptosis level of HT22 cells under different treatment conditions. The results showed that hydrogen treatment effectively attenuated LPS-induced neuronal apoptosis. However, this protective effect was reversed by ATG9B knockdown (Fig. 5F-G).

These data demonstrate that hydrogen’s therapeutic effect in enhancing mitophagy is associated with upregulation of ATG9B. It both coordinates the initial phagocytosis of damaged mitochondria and promotes the subsequent fusion of mitophagosomes with lysosomes, thereby restoring mitophagic flux to inhibit neuronal apoptosis. These data position ATG9B as an upstream regulator of the PINK1-Parkin pathway, linking membrane dynamics to hydrogen-mediated neuroprotection.

Hippocampus-specific ATG9B knockdown abolishes hydrogen-mediated neuroprotection in SAE mice

The mechanistic insights from cell-based studies prompted us to validate the functional significance of ATG9B in vivo. ATG9B expression was knocked down by three-dimensional injection of adenovirus (AAV-shATG9B) into the hippocampus. First, the knockdown efficiency was verified. Western blotting results showed that AAV-shATG9B significantly reduced ATG9B protein levels, indicating a clear knockdown effect (Fig. 6A-B). Subsequently, based on this, hydrogen-based treatment is carried out. IF showed that ATG9B fluorescence intensity in the hippocampus of the CLP group decreased and then increased after hydrogen treatment. In contrast, its expression was inhibited upon ATG9B knockdown (Fig. 6C-D). Western blot analysis showed that the H2-mediated upregulation of mitophagy flux observed in the AAV-shNC group was suppressed in the KD group. (Fig. 6E-F). Behavioral results showed that H2 inhalation improved performance in the novel object recognition and Y-maze tests compared with the CLP group. Crucially, this therapeutic effect was completely lost in CLP mice with hippocampal ATG9B knockdown, despite receiving H₂ therapy (Fig. 6H-K). Locomotor activity remained unchanged across all groups, confirming that the observed differences were specific to cognitive function rather than general activity (Fig. 6G, S3). This result provides direct evidence that ATG9B is required for H₂ to ameliorate sepsis-associated cognitive impairment.

Fig. 6.

Fig. 6

Open in a new tab

Hippocampus-specific ATG9B knockdown abolishes hydrogen-mediated neuroprotection in SAE mice. A-B Representative Western blot and quantification of ATG9B in hippocampal lysates from Sham, Sham + AAV-shNC, and Sham+H2+AAV-shATG9B groups (n = 6). C-D IF staining and quantification of ATG9B in mouse hippocampal neurons. Scale bar = 100 μm (n = 6). E-F Western blot analysis and quantification of LC3, p62, and ATG9B in four groups (n = 6). G Total distance traveled (cm) by each group in the open field test (n = 10). H Novel object recognition test: preference index of mice in each group (n = 10). I-K Y-maze test: exploration time and entry frequency for the novel arm (n = 10). L Representative IF images showing colocalization of PSD95 and SYP in hippocampal frozen sections, with quantification of the colocalization coefficient. Scale bar = 100 μm (n = 6). Statistics: one‑way ANOVA with Tukey’s post hoc test. Data are mean ± SD from independent experiments. ns, not significant, *p < 0.05, **p < 0.01, ***p < 0.001

In addition, the colocalization analysis of the presynaptic marker SYP and the PSD95 showed that the colocalization area was significantly reduced in the CLP group, indicating that the synaptic structure was damaged. Hydrogen treatment restores synaptic integrity. However, in mice with ATG9B knockdown, this protective effect on synaptic structures was abolished. (Fig. 6L). These data solidify the link between the ATG9B-dependent mitophagy and the preservation of synaptic connectivity.

Collectively, these in vivo studies provide robust evidence that ATG9B is a key downstream signaling pathway for hydrogen therapy in SAE. This confirms that targeting and activating this specific molecular axis is the basis for hydrogen’s neuroprotective and cognitive-protective effects in SAE.

Discussion

SAE is a prevalent and devastating complication characterized by acute brain dysfunction and long-term cognitive impairment, yet effective targeted therapies remain lacking (Sonneville et al. 2023, Catarina et al. 2021). Molecular hydrogen is a safe and penetrative therapeutic gas with documented anti-oxidative, anti-inflammatory, and organ-protective properties (Nogueira et al. 2018), but its precise mechanism against SAE, especially regarding mitochondrial quality control, remains poorly understood. This study demonstrates that H₂ inhalation rescues cognitive dysfunction, reduces neuroinflammation, and preserves hippocampal neuronal integrity in a CLP-induced SAE mouse model. Mechanistically, we identify ATG9B as a novel and indispensable downstream effector of H2 that is essential for activating PINK1-Parkin-dependent mitophagy. Unlike prior studies focusing on H2’s antioxidant properties, our work reveals that H2 upregulated ATG9B expression to restore PINK1-Parkin-mediated mitophagic flux, promoting autophagosome formation and mitochondria-lysosome fusion to clear damaged mitochondria. Both in vitro knockdown and in vivo hippocampus-specific silencing of ATG9B abolished the beneficial effects of H2, confirming that ATG9B serves as a key molecular hub. These findings not only elucidate a novel mechanism for hydrogen therapy in SAE but also pinpoint the ATG9B-mitophagy axis as a promising therapeutic target for cognitive dysfunction in sepsis.

Mitochondrial dysfunction is a central event in SAE pathogenesis. Damaged mitochondria release excessive ROS, trigger inflammatory cascades, and activate intrinsic apoptotic pathways, ultimately leading to neuronal loss and synaptic breakdown. Mounting evidence has linked impaired mitophagy to cognitive dysfunction in multiple neurodegenerative diseases and acute brain injuries (Lu et al. 2023, Antico et al. 2025). Consistent with these studies, our results showed that mitophagic flux was severely blocked in septic mice and LPS‑stimulated HT22 cells, as evidenced by decreased PINK1-Parkin expression, accumulation of p62, and reduced LC3-II conversion. H2 treatment effectively reversed these changes and restored mitophagy. Furthermore, the autophagy inhibitor 3-MA largely abolished H2-mediated mitochondrial protection, anti-apoptotic effects, and cognitive improvement. These results firmly establish that enhancing mitophagy is a key mechanism by which hydrogen exerts neuroprotection in SAE.

A central finding of this study is the elucidation of ATG9B’s pivotal role in H₂-mediated enhancement of mitophagy. Previous studies have indicated that ATG9B possesses lipid scramblase activity and is essential for supplying membrane sources for efficient autophagosome biogenesis (Wang et al. 2017). However, compared with ATG9A, the specialized functions of ATG9B in neuron-selective autophagy and in disease models remain limited. The present work addresses this gap. Under basal conditions, knockdown of ATG9B alone is sufficient to impair mitophagic flux. In the context of sepsis, the H₂-induced activation of the PINK1-Parkin pathway, the enhancement of autophagosome-lysosome fusion, and the ultimate neuroprotective effects are abrogated by ATG9B deficiency. This indicates that H₂ does not activate all autophagic pathways but instead relies specifically on ATG9B to coordinate the precise clearance of damaged mitochondria. This finding suggests that targeting ATG9B may offer a more cell-type-specific intervention strategy than directly targeting PINK1 or Parkin.

Notably, beyond its classical role in autophagosome formation, the result also implies that ATG9B is involved in the later stages of autophagosome-lysosome fusion. A significant reduction in the co-localization of mitochondria and lysosomes was observed in ATG9B-knockdown cells. This phenomenon suggests that ATG9B may indirectly regulate fusion events through interactions with SNARE complexes or other lysosomal tethering proteins (Jahn et al. 2024), or that proper membrane curvature and closure mediated by ATG9B are prerequisites for subsequent successful fusion. This mechanistic detail warrants further investigation using high-resolution live-cell imaging and protein interaction proteomics in future studies.

When placed in the context of sepsis treatment, hydrogen therapy exhibits significant advantages over traditional broad-spectrum antioxidants (Peng et al. 2024). The antioxidant effect of H₂ is selective and can neutralize only highly cytotoxic free radicals, without interfering with reactive oxygen species that have physiological signaling functions. Secondly, H₂ can not only alleviate oxidative damage, but also actively restore cellular homeostasis by enhancing the intrinsic mitophagy mechanism. This mechanism of action may lead to longer-lasting protective effects. In addition, H₂’s extremely high permeability and good safety profile make it an ideal drug candidate for clinical translation (Ahmad et al. 2022). Preliminary clinical trials have explored H₂ inhalation therapy in patients with post-cardiac arrest syndrome and cerebral injury (Li et al. 2020, Otsuka et al. 2023), confirmed its safety, and observed a trend toward improved neurological outcomes. However, it must be acknowledged that the pathogenesis of SAE is highly complex and involves multiple pathophysiological processes, such as blood-brain barrier disruption, microglial polarization, and peripheral immune suppression (Gofton and Young 2012). Although hydrogen has been shown to directly activate neuronal mitophagy, the regulatory effects of hydrogen on microglial and astrocyte mitophagy, and its contribution to the overall neuroinflammatory network, remain to be further elucidated.

Several limitations of the present study should be considered. First, while the focus was placed on the PINK1-Parkin pathway, other mitophagy receptors (e.g., BNIP3, FUNDC1) warrant further investigation. Second, the precise upstream sensor that H₂ upregulates ATG9B remains to be fully characterized. Whether H₂ activates ATG9B transcription through modulation of specific transcription factors or via epigenetic modifications is an important direction for future research. Third, while the in vivo model validates this axis within the hippocampus, its role in other brain regions affected by SAE requires further exploration. Furthermore, the young male mouse model used in this study does not fully recapitulate the complex clinical scenario of patients with sepsis, as such patients are often older and have underlying comorbidities. Future verification of H₂ efficacy and ATG9B-targeting strategies in mouse models of both genders, in elderly or diabetic mouse models will have greater clinical relevance.

From a translational perspective, identifying ATG9B as the rate-limiting mediator of H₂-induced mitophagy paves the way for targeted therapeutic interventions. Although small-molecule agonists that directly target ATG9B have not yet been reported, structure-based drug screening or gene therapy strategies represent potential avenues for intervention. In addition, monitoring ATG9B levels in peripheral blood or cerebrospinal fluid, or assessing its potential as a biomarker to predict SAE severity and the efficacy of hydrogen therapy, is a direction for future research.

In summary, this study elucidates a novel mechanism whereby molecular hydrogen attenuates sepsis-associated cognitive impairment by upregulating ATG9B expression and activating PINK1-Parkin-dependent mitophagy. ATG9B has been shown to be a key hub molecule that coordinates autophagosomal membrane dynamics, sustains mitophagic flux, and mediates the neuroprotective effects of H₂. This discovery enhances the mechanistic understanding of H₂ therapy and identifies the ATG9B-mitophagy axis as a promising new target for intervention in SAE.

Conclusion

In summary, this study presents that molecular hydrogen effectively attenuates SAE by protecting hippocampal neurons, preserving synaptic integrity, and improving cognitive function. Mechanistically, H₂ upregulates ATG9B expression in hippocampal neurons, thereby activating PINK1-Parkin-dependent mitophagy, which in turn enhances autophagosomal membrane dynamics and promotes efficient clearance of damaged mitochondria, thereby inhibiting neuronal apoptosis. Genetic knockdown of ATG9B abolishes H₂-induced mitophagic flux and neuroprotection, establishing ATG9B as a critical downstream effector. These findings provide mechanistic insights into H2 therapy and position the ATG9B-mitophagy axis as a novel therapeutic target for the prevention and treatment of SAE.

Supplementary Information

Supplementary Material 1. (2.8MB, pdf)
Supplementary Material 2. (1.5MB, docx)

Acknowledgements

We thank all partners in this study for their efforts and assistance.

Authors’ contributions

Xiaoli Yu: Methodology, Validation, Writing – original draft, Data curation, Software, Formal analysis. Xiaofan Huang, Yu Song and Jianfeng Liu: Project administration, Methodology, Visualization, Validation, Software. Fei Huang and Jiatian Cui: Data curation, Software, Formal analysis. Yan Cui and Keliang Xie: Resources, Funding acquisition. All authors approved the final version of the manuscript.

Funding

This work was supported by a grant from the research project of the National Natural Science Foundation of China (No. 82572501 to Yan Cui), Joint Funds of the Natural Science Foundation of Tianjin (No. 25JCLMJC00230 to Yan Cui), and Tianjin “the Belt and Road” Joint Laboratory Project (No. 24PTLYHZ00200 to Keliang Xie).

Data availability

Data can be provided by the corresponding author upon request.

Declarations

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Xiaoli Yu, Xiaofan Huang and Yu Song contributed equally to this work.

Contributor Information

Keliang Xie, Email: mzk2011@126.com.

Yan Cui, Email: cuiyanbio45@163.com.

References

  1. Sonneville R, Benghanem S, Jeantin L, et al. The spectrum of sepsis-associated encephalopathy: a clinical perspective. Crit Care (London England). 2023;27(1):386. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Li J, Jia Q, Yang L, et al. Sepsis-associated encephalopathy: Mechanisms, Diagnosis, and Treatments update. Int J Biol Sci. 2025;21(7):3214–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Sonneville R, de Montmollin E, Poujade J, et al. Potentially modifiable factors contributing to sepsis-associated encephalopathy. Intensive Care Med. 2017;43(8):1075–84. [DOI] [PubMed] [Google Scholar]
  4. Widmann CN, Heneka MT. Long-term cerebral consequences of sepsis. Lancet Neurol. 2014;13(6):630–6. [DOI] [PubMed] [Google Scholar]
  5. Knierim JJ. The hippocampus. Curr biology: CB. 2015;25(23):R1116–1121. [DOI] [PubMed] [Google Scholar]
  6. Kikuchi DS, Campos ACP, Qu H, et al. Poldip2 mediates blood-brain barrier disruption in a model of sepsis-associated encephalopathy. J Neuroinflamm. 2019;16(1):241. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Yin XY, Tang XH, Wang SX, et al. HMGB1 mediates synaptic loss and cognitive impairment in an animal model of sepsis-associated encephalopathy. J Neuroinflamm. 2023;20(1):69. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Manabe T, Heneka MT. Cerebral dysfunctions caused by sepsis during ageing. Nat Rev Immunol. 2022;22(7):444–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Yang X, Duan H, Li S, et al. Yap1 alleviates sepsis associated encephalopathy by inhibiting hippocampus ferroptosis via maintaining mitochondrial dynamic homeostasis. J Cell Mol Med. 2024;28(19):e70156. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Gong H, Liu H, Yuan M, et al. Dihydroartemisinin alleviates sepsis-associated encephalopathy by reducing microglial iron accumulation and mitochondrial dysfunction via HIF1A/HMOX1 pathway. Phytomedicine: Int J phytotherapy phytopharmacology. 2025;148:157413. [DOI] [PubMed] [Google Scholar]
  11. Hou Y, Zhang Y, Jiang S, et al. Salidroside intensifies mitochondrial function of CoCl(2)-damaged HT22 cells by stimulating PI3K-AKT-MAPK signaling pathway. Phytomedicine: Int J phytotherapy phytopharmacology. 2023;109:154568. [DOI] [PubMed] [Google Scholar]
  12. Ding H, Li Y, Chen S, et al. Fisetin ameliorates cognitive impairment by activating mitophagy and suppressing neuroinflammation in rats with sepsis-associated encephalopathy. CNS Neurosci Ther. 2022;28(2):247–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Fang EF, Hou Y, Palikaras K, et al. Mitophagy inhibits amyloid-β and tau pathology and reverses cognitive deficits in models of Alzheimer’s disease. Nat Neurosci. 2019;22(3):401–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Picca A, Faitg J, Auwerx J, Ferrucci L, D’Amico D. Mitophagy in human health, ageing and disease. Nat metabolism. 2023;5(12):2047–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Yang Y, Ke J, Cao Y, Gao Y, Lin C. Melatonin regulates microglial M1/M2 polarization via AMPKα2-mediated mitophagy in attenuating sepsis-associated encephalopathy. Biomed pharmacotherapy = Biomedecine pharmacotherapie. 2024;177:117092. [DOI] [PubMed] [Google Scholar]
  16. Liu J, Meng S, Wang Z, et al. Molecular hydrogen-mediated SIRT1 activation alleviates sepsis-associated encephalopathy by promoting mitophagy. Eur J Med Res. 2025;31(1):113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Liang W, Moyzis AG, Lampert MA, Diao RY, Najor RH, Gustafsson ÅB. Aging is associated with a decline in Atg9b-mediated autophagosome formation and appearance of enlarged mitochondria in the heart. Aging Cell. 2020;19(8):e13187. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Chiduza GN, Garza-Garcia A, Almacellas E, et al. ATG9B is a tissue-specific homotrimeric lipid scramblase that can compensate for ATG9A. Autophagy. 2024;20(3):557–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Choi J, Jang H, Xuan Z, Park D. Emerging roles of ATG9/ATG9A in autophagy: implications for cell and neurobiology. Autophagy. 2024;20(11):2373–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Wang LH, Yuan Y, Wang J et al. ASCL2 Maintains Stemness Phenotype through ATG9B and Sensitizes Gliomas to Autophagy Inhibitor. Advanced science (Weinheim, Baden-Wurttemberg, Germany). 2022;9(27):e2105938. [DOI] [PMC free article] [PubMed]
  21. Ohsawa I, Ishikawa M, Takahashi K, et al. Hydrogen acts as a therapeutic antioxidant by selectively reducing cytotoxic oxygen radicals. Nat Med. 2007;13(6):688–94. [DOI] [PubMed] [Google Scholar]
  22. Yoshiura T, Kawauchi S, Nishida S, et al. Hydrogen inhalation and intrathecal magnesium sulfate ameliorate ischemia by suppressing cortical spreading depolarization in a rat subarachnoid hemorrhage model. Neurotherapeutics: J Am Soc Experimental Neurother. 2025;22(5):e00617. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Qiu P, Liu Y, Zhang J. Recent Advances in Studies of Molecular Hydrogen against Sepsis. Int J Biol Sci. 2019;15(6):1261–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Meng S, Liu J, Luo Y, et al. Molecular hydrogen protects against sepsis-induced cardiomyopathy through improving Golgi stress-mediated autophagy, inflammation and apoptosis. Br J Pharmacol. 2025;182(22):5627–46. [DOI] [PubMed] [Google Scholar]
  25. Wang Y, Han Q, Liu L, et al. Natural hydrogen gas and engineered microalgae prevent acute lung injury in sepsis. Mater today Bio. 2024;28:101247. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Cui Y, Li Y, Meng S, Song Y, Xie K. Molecular hydrogen attenuates sepsis-induced cardiomyopathy in mice by promoting autophagy. BMC Anesthesiol. 2024;24(1):72. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Dejager L, Pinheiro I, Dejonckheere E, Libert C. Cecal ligation and puncture: the gold standard model for polymicrobial sepsis? Trends Microbiol. 2011;19(4):198–208. [DOI] [PubMed] [Google Scholar]
  28. Qi B, Song Y, Chen C, et al. Molecular hydrogen attenuates sepsis-induced cognitive dysfunction through regulation of tau phosphorylation. Int Immunopharmacol. 2023;114:109603. [DOI] [PubMed] [Google Scholar]
  29. Yu Y, Ma X, Yang T, et al. Protective effect of hydrogen-rich medium against high glucose-induced apoptosis of Schwann cells in vitro. Mol Med Rep. 2015;12(3):3986–92. [DOI] [PubMed] [Google Scholar]
  30. Keiser MS, Chen YH, Davidson BL. Techniques for Intracranial Stereotaxic Injections of Adeno-Associated Viral Vectors in Adult Mice. Curr protocols mouse biology. 2018;8(4):e57. [DOI] [PubMed] [Google Scholar]
  31. Yin H, Wei Y, Han Z, Li Z, Liu K, Zhu L. BNIP3-mediated mitophagy alleviates traumatic brain injury by inhibiting neuronal apoptosis and oxidative stress. Int Immunopharmacol. 2026;168(Pt 1):115793. [DOI] [PubMed] [Google Scholar]
  32. Wang H, Ye J, Peng Y, et al. CKLF induces microglial activation via triggering defective mitophagy and mitochondrial dysfunction. Autophagy. 2024;20(3):590–613. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Baechler BL, Bloemberg D, Quadrilatero J. Mitophagy regulates mitochondrial network signaling, oxidative stress, and apoptosis during myoblast differentiation. Autophagy. 2019;15(9):1606–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Catarina AV, Branchini G, Bettoni L, De Oliveira JR, Nunes FB. Sepsis-Associated Encephalopathy: from Pathophysiology to Progress in Experimental Studies. Mol Neurobiol. 2021;58(6):2770–9. [DOI] [PubMed] [Google Scholar]
  35. Nogueira JE, Passaglia P, Mota CMD, et al. Molecular hydrogen reduces acute exercise-induced inflammatory and oxidative stress status. Free Radic Biol Med. 2018;129:186–93. [DOI] [PubMed] [Google Scholar]
  36. Lu Y, Li Z, Zhang S, Zhang T, Liu Y, Zhang L. Cellular mitophagy: Mechanism, roles in diseases and small molecule pharmacological regulation. Theranostics. 2023;13(2):736–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Antico O, Thompson PW, Hertz NT, Muqit MMK, Parton LE. Targeting mitophagy in neurodegenerative diseases. Nat Rev Drug Discovery. 2025;24(4):276–99. [DOI] [PubMed] [Google Scholar]
  38. Wang N, Tan HY, Li S, Feng Y. Atg9b Deficiency Suppresses Autophagy and Potentiates Endoplasmic Reticulum Stress-Associated Hepatocyte Apoptosis in Hepatocarcinogenesis. Theranostics. 2017;7(8):2325–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Jahn R, Cafiso DC, Tamm LK. Mechanisms of SNARE proteins in membrane fusion. Nat Rev Mol Cell Biol. 2024;25(2):101–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Peng Z, Li XJ, Zhou Y, et al. Hydrogen exerts neuroprotective effects after subarachnoid hemorrhage by attenuating neuronal ferroptosis and inhibiting neuroinflammation. Free Radic Biol Med. 2024;215:79–93. [DOI] [PubMed] [Google Scholar]
  41. Ahmad A, Baig AA, Hussain M, et al. Narrative on Hydrogen Therapy and its Clinical Applications: Safety and Efficacy. Curr Pharm Design. 2022;28(31):2519–37. [DOI] [PubMed] [Google Scholar]
  42. Li L, Li X, Zhang Z, Liu L, Zhou Y, Liu F. Protective Mechanism and Clinical Application of Hydrogen in Myocardial Ischemia-reperfusion Injury. Pakistan J Biol sciences: PJBS. 2020;23(2):103–12. [DOI] [PubMed] [Google Scholar]
  43. Otsuka Y, Tomura S, Toyooka T, et al. Hyperbaric hydrogen therapy improves secondary brain injury after head trauma. Undersea hyperbaric medicine: J Undersea Hyperbaric Med Soc Inc. 2023;50(4):403–11. [PubMed] [Google Scholar]
  44. Gofton TE, Young GB. Sepsis-associated encephalopathy. Nat reviews Neurol. 2012;8(10):557–66. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1. (2.8MB, pdf)
Supplementary Material 2. (1.5MB, docx)

Data Availability Statement

Data can be provided by the corresponding author upon request.

Articles from Molecular Medicine are provided here courtesy of The Feinstein Institute for Medical Research at North Shore LIJ

RESOURCES