Animal experiments were conducted in accordance with the guidelines of the Animal Ethic Committee of Medical Discovery Leader (MDL, Approval number: MDL2023-08-26-01). Eight-week-old male C57BL/6J mice (wild-type, METTL3^+⁣/+, and METTL3^+⁣/+GPX4^-⁣/-), weighing between 20 and 25 grams, were purchased from Cyagen Biotechnology Co., Ltd. (Suzhou, China). A total of 60 mice were used in the experiment. The mice were placed in a temperature-controlled room (23 $\pm$ 2 °C) with a controlled lighting schedule.

Before surgery, the mice were assigned codes and evenly distributed into the following groups: sham, TBI model, model+NC, model+METTL3, model+METTL3+siGPX4, model+METTL3+erastin groups (10 mice in each group). Mouse TBI models were created through a controlled cortical impact (CCI) procedure [26]. The mice were positioned face-down in a stereotaxic frame. A surgical incision was made to expose the skull, and a 3 mm craniotomy was performed on the left side, positioned roughly halfway between the bregma and lambda, lateral to the midline. The skull flap was meticulously removed without damaging the underlying dura mater. The CCI was inflicted perpendicularly to the brain surface using a pneumatic cortical impactor (AmScien Instruments, Richmond, VA, USA). The settings for the impactor were as follows: an impact pressure of 10 KPa, a depth of 1.0 mm, and a duration of 70 milliseconds. The craniotomy site was immediately sutured shut using standard surgical materials after TBI. Mice in the sham group underwent all surgical steps except for the actual CCI. For the model+METTL3+erastin groups, the mice were intraperitoneally administrated with erastin (329600, Sigma-Aldrich, St. Louis, MO, USA) at a dose of 20 mg/kg body weight, 1 h after TBI. To ensure consistency and reduce variability, the surgical procedures, drug administrations, and CCI operations were conducted by the same experienced personnel. All surgical interventions were performed under deep anesthesia induced by intraperitoneal injection of sodium pentobarbital (1% w/v in sterile saline, 50 mg/kg body weight). Anesthesia depth was confirmed by the absence of corneal reflex and negative response to the toe pinch test.

To assess the efficacy of the interventional treatments, the Morris Water Maze (MWM) test was performed 7 days post-TBI. Prior to the TBI, mice underwent a 3-day training period to acclimate to locating a submerged platform in the maze, with three attempts per day, each limited to 90 seconds. Mice unable to find the platform within this period were guided to it and remained there for 15 seconds to familiarize themselves with the location. A probe trial, conducted 24 hours after training, tracked the latency to first reach the platform and the frequency of traversing its former location. These metrics were recorded and analyzed using a computer-assisted video tracking system (EthoVision XT 16.0, Noldus Information Technology, Wageningen, Netherlands), providing precise records of the swimming paths and times.

Neurological severity score (NSS) was used to evaluate motor, balance, and reflex functions in mice post-TBI, offering a comprehensive assessment of neurological status [27]. Trained professionals conducted NSS assessments, ensuring accuracy and consistency through specialized training. The NSS ranges from 0 to 10, with increasing scores signifying more pronounced neurological deficits. The grading system is as follows: 0–3 points indicate nearly normal neurological function or mild deficits; 4–6 points suggest moderate deficits; and 7–10 points represent severe impairments.

After the completion of the experimental treatments, the mice were anesthetized using an intraperitoneal injection of sodium pentobarbital (1.0% w/v in normal saline, 50 mg/kg body weight) to ensure minimal distress. The anesthetized mice were then euthanized by cervical dislocation. Subsequently, the brain tissues were carefully excised and immediately placed in a chilled phosphate-buffered saline (PBS) solution to prevent degradation. Brain tissues were fixed in 4% paraformaldehyde for 24 hours at 4 °C, dehydrated through a 90% ethanol solution, and then cleared in xylene. The tissues were sectioned into 5-µm-thick slices using a microtome.

Brain tissue sections were stained with 0.1% crystal violet solution (C0121, Beyotime, Shanghai, China) for 30 minutes, then quickly rinsed with PBS to remove excess stain. Subsequent dehydration with 95% and 100% ethanol for 5 minutes each, followed by clearing with xylene for 5 minutes (repeated twice), preceded mounting with a neutral medium. The stained sections were examined under a Nikon TE300 microscope (Nikon, Tokyo, Japan).

Brain tissue sections were immersed in Nissl staining solution (C0117, Beyotime, Shanghai, China) for 5 minutes at room temperature. After staining, the sections were rinsed with distilled water for 5 minutes. The stained sections were examined under a light microscope (Nikon, Tokyo, Japan) to visualize the morphology of neuron cells.

To identify degenerating neurons, FJB staining was performed on the brain sections according to the manufacturer’s protocol (AG310, Millipore, Darmstadt, Germany). The sections were first incubated in a 0.06% potassium permanganate solution for 10 minutes, washed with distilled water for 2 minutes, and stained with a 0.0004% solution of Fluoro-Jade B for 30 minutes. The stained sections were examined using a Nikon TE300 microscope to visualize degenerative neurons. The stained sections were examined under a Nikon TE300 fluorescence microscope equipped with filters to visualize degenerating neurons.

To identify iron deposition within cellular structures, Perls’ Prussian blue staining was performed using the Prussian Blue Iron Stain Kit (G1422, Solarbio, Beijing, China). Brain sections were immersed in the Perls Stain Solution for 30 minutes and rinsed with distilled water for 5 minutes. Subsequently, the sections were dehydrated with anhydrous ethanol for three times, each for 5 minutes, followed by clearing in dimethylbenzene for 5 minutes. The sections were then removed, and neutral resin was applied dropwise for mounting. Iron deposits were visualized as blue particles within cellular structures using a light microscope. These deposits were quantified with ImageJ software (version 1.53; National Institutes of Health, Bethesda, MD, USA).

Brain slices were rehydrated with PBS for 10 minutes and then blocked with a buffer consisting of 0.3% Triton X-100 and 10% BSA in PBS for 1 hour to reduce non-specific binding. The sections were incubated overnight at 4 °C with primary antibodies targeting GPX4 (ab125066, 1:250, Abcam, Cambridge, UK), SLC7A11 (ab307601, 1:500, Abcam), METTL3 (ab195352, Abcam), NeuN (ab104224, 1:3000, Abcam), GFAP (ab7260, 1:5000, Abcam), and Iba-1 (ab178846, 1:2000, Abcam). After washing with PBS, the sections were incubated with secondary antibodies for 1 hour at room temperature: Anti-rabbit IgG H&L (ab205718, 1:1000, Abcam) for GPX4, SLC7A11, METTL3, and GFAP; Anti-mouse IgG H&L (ab205719, 1:1000, Abcam) for NeuN; and Anti-goat IgG H&L (ab205720, 1:1000, Abcam) for Iba-1. For immunofluorescence staining, sections were mounted with a solution containing 4^′,6-diamidino-2-phenylindole (DAPI, 62248, Thermo Scientific, Rockford, IL, USA) to stain cell nuclei. The slides were examined under a fluorescence microscope (Leica, Wetzlar, Germany) to visualize cell density and specific protein expression within the brain tissue.

The levels of total iron and ferrous irons in the ipsilateral cortex were assessed using an iron assay kit (ab83366, Abcam, Cambridge, UK). Malondialdehyde (MDA), superoxide dismutase (SOD) and ROS levels in cortex tissues were measured using the lipid peroxidation MDA assay kit (S0131S, Beyotime, Shanghai, China), Total SOD detection kit (S0101M, Beyotime, Shanghai, China) and Tissue ROS test kit (BB-460512, Bestbio, Shanghai, China), respectively.

Mouse primary cortical neurons were extracted from E15.5 mouse embryos as previously described [28], and the mouse hippocampal neuronal cell line HT-22 was from the Wanwu Biotechnology Co., Ltd. (Hefei, China). Primary cortical neurons exhibited a clear cell body with multiple long, slender processes extending from the soma, while HT-22 cells displayed typical neuronal-like morphology with elongated cell bodies and processes. Both primary neurons and HT-22 cells were cultured in DMEM (Gibco, New York, NY, USA) supplemented with 10% fetal bovine serum (Gibco, USA), 1% streptomycin, and penicillin mix at 37 °C with 5% CO₂. To ensure cell purity and authenticity, mycoplasma detection and short tandem repeat (STR) profiling were conducted. No mycoplasma contamination was detected in any of the cell cultures.

To overexpress METTL3 and YTHDF2, mouse METTL3 (NM_019721.2) and YTHDF2 (NM_145393.4) were synthesized and cloned into the pLVX plasmid vector (General Biosystems, Anhui, China) and the complete plasmid sequence have been provided in Supplementary Fig. 1. The siRNA targeting METTL3 (5^′-UCUAUCUCCAGAUCAACAUCG-3^′; 5^′-GCUACAAUCACAUCGCAGUCC-3^′; 5^′-GAUAUCACAACAGAUCCACUG-3^′), siRNA targeting YTHDF2 (5^′-UAGUAACUGGGUAAGUAGGAG-3^′; 5^′-UGGUUUAUUCUCGUUGUUCUC-3^′; 5^′-UUUGAAAUCAAAUUAAUCCUG-3^′), siRNA targeting GPX4 (5^′-AGUUUACGUCAGUUUUGCCUC-3^′) and nonsense siRNA (5^′-UUCUCCGAACGUGUCACGU-3^′) were constructed by RiboBio (Guangzhou, China). Cells were transfected for 24 hours, after which a mechanical scratch injury was created using a 200 µL pipette tip to establish the cell injury model. Transfections were performed using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA).

Cell viability was assessed by the Cell Counting Kit-8 (C0038, Beyotime, Shanghai, China). Briefly, cells were harvested and resuspended in complete culture medium to prepare a cell suspension with a density of 2 $\times$ 10⁴ cells/mL. A volume of 100 µL of the cell suspension, containing 2000 cells, was added to each well of a 96-well plate. After the incubation period, 10 µL of Cell Counting Kit 8 (CCK-8) solution was added to each well, and the plate was gently mixed to ensure even distribution of the reagent. The cells were then incubated for 1 hour at 37 °C in a humidified atmosphere with 5% CO₂. The absorbance of each well was measured at a wavelength of 450 nm using a Microplate Reader (800TSUV, BioTek, Winooski, VT, USA). The cell viability rate was calculated by comparing the absorbance values of the experimental groups with those of the control group.

The samples were homogenized in lysis buffer (P0023, Beyotime, Shanghai, China) and sonicated. The homogenate was then centrifuged at 4 °C for 20 minutes at 12,000 g to separate soluble proteins. Protein concentrations were measured using a BCA kit (P0009, Beyotime, Shanghai, China). Proteins were resolved on SDS-PAGE gels ranging from 7.5% to 12.5% (Bio-Rad, Hercules, CA, USA) and transferred onto polyvinylidene fluoride membranes (Millipore, Darmstadt, Germany), and incubated overnight at 4 °C with primary antibodies against SLC7A11 (ab307601, 1:1000, Abcam, Cambridge, UK), IL-6 (ab290735, 1:1000, Abcam), TNF-$\alpha$ (ab183218, 1:2000, Abcam), IL-1$\beta$ (ab315084, 1:1000, Abcam), GPX4 (ab125066, 1:2000, Abcam), METTL3 (ab195352, 1:1000, Abcam), YTHDF2 (ab220163, 1:1000, Abcam), and $\beta$-actin (AF7018, 1:3000, Affinity, Cincinnati, OH, USA). The next day, membranes were incubated with secondary antibodies (ab97051, 1:2000, Abcam) for 1.5 hours at room temperature. Immunoreactive protein bands were visualized using an enhanced chemiluminescence system (Thermo, USA), and band intensities were quantified using ImageJ software (version 1.53, National Institutes of Health, USA).

Total RNA was extracted from the samples using TRIzol reagent (15596018CN, Invitrogen, CA, USA), and quantified using a NanoDrop 2000C spectrophotometer (Thermo, USA). Reverse transcription was performed using PrimeScript™ RT Master Mix (RR036A, Takara, Japan) to generate cDNA. qPCR reactions were conducted using a SYBR Green master mix (04913850001, Roche, Basel, Switzerland) to ensure accuracy. mRNA levels were normalized to the housekeeping gene $\beta$-actin as an endogenous control, and relative quantification was calculated using the comparative cycle threshold (Ct) method. Primer sequences used are as follows:

GPX4-F: 5^′-TGGTTTACGAATCCTGGCCT-3^′, GPX4-R: 5^′-GGCATCGTCCCCATTTACAC-3^′;

SLC7A11-F: 5^′-GTCTGCCTGTGGAGTACTGT-3^′, SLC7A11-R: 5^′-ATTACGAGCAGTTCCACCCA-3^′;

METTL3-F: 5^′-TGGCCTCTTCAGCATCAGAA-3^′, METTL3-R: 5^′-ACTGACCTTCTTGCTCTGCT-3^′;

YTHDF2-F: 5^′-TGCTTGGTCTACTGGAGGTG-3^′, YTHDF2-R: 5^′-AATGGAGTGCTACCTAGGGC-3^′;

IL-6-F: 5^′-ACTTCCATCCAGTTGCC-3^′, IL-6-R: 5^′-ATGTGTAATTAAGCCTCCGAC-3^′;

TNF-$\alpha$-F: 5^′-ACGGCATGGATCTCAAAGACAAC-3^′, TNF-$\alpha$-R: 5^′-AGATAGCAAATCGGCTGACGG-3^′;

IL-1$\beta$-F: 5^′-TTGAAGTTGACGGACCCCAA-3^′, IL-1$\beta$-R: 5^′-CCACAGCCACAATGAGTGA-3^′;

$\beta$-actin-F: 5^′-CTCCTGAGCGCAAGTACTCT-3^′, $\beta$-actin-R: 5^′-TACTCCTGCTTGCTGATCCAC-3^′.

The production of inflammatory factors IL-6 (ab222503, Abcam, Cambridge, UK), TNF-$\alpha$ (ab208348, Abcam), and IL-1$\beta$ (PI301, Beyotime, Shanghai, China) was measured by using ELISA kits. Absorbance was measured at 450 nm using a microplate reader (Varioskan LUX, Thermo, USA). The concentrations of the inflammatory factors were determined by comparing the absorbance values to a standard curve generated using known concentrations of recombinant proteins.

The level of m6A RNA methylation in cells was quantified using an m6A RNA Methylation Quantification Kit (ab185912, Abcam, Cambridge, UK). In brief, mRNA was extracted and bound to a strip well for 90 minutes. After washing, captured, detection, and enhancer antibodies were added sequentially. The color-developing solution was then added, and the absorbance was measured at 450 nm using a microplate reader.

The MeRIP assay was performed using the Magna MeRIP m6A Kit (17-10499, Millipore, Darmstadt, Germany) to analyze the m6A modification in RNA samples. The enrichment of GPX4 mRNA was measured by qPCR.

The interaction between YTHDF2 and GPX4 mRNA was measured by Pierce™ Magnetic RNA-Protein Pull-Down Kit (PI20164, Thermo, USA), which contains streptavidin-coated magnetic beads (1 µm diameter, silica core, binding capacity: 1 µg biotinylated probe per 1 mg beads). In brief, the biotin-labeled GPX4-specific probe was reacted with magnetic beads, which were then mixed with protein lysate samples. After washing and elution, YTHDF pulled down by GPX4 mRNA was detected by western blot.

The RIP assay was performed using antibodies against YTHDF2 or a negative IgG control with Magna RIP™ RNA-Binding Protein Immunoprecipitation Kit (17-700, Millipore, Darmstadt, Germany). Protein A/G-coated magnetic beads (1 µm diameter, agarose matrix, provided in the kit) were pre-washed with RIP wash buffer and incubated with 5 µg of anti-YTHDF2 antibody or negative control IgG (I9145, Sigma-Aldrich, St. Louis, USA) for 1 hour at 4 °C. Cells were lysed in RIP lysis buffer, and the lysates were incubated with RIP immunoprecipitation buffer containing immunoprecipitated with A/G magnetic beads. The magnetic frame was used to fix bead-bound complexes and wash away unbound material. RNA was extracted from the bead-bound complexes, and GPX4 mRNA levels were analyzed by qPCR.

Cells transfected with siMETTL3, siYTHDF2, or siNC were treated with actinomycin D for 0, 3, and 6 h. RNA was extracted, and GPX4 mRNA levels were measured by qPCR.

Statistical analyses were conducted using GraphPad Prism 9 (San Diego, CA, USA). Comparisons between two groups were analyzed using Student’s t-test, while multiple groups were analyzed by one-way ANOVA followed by Tukey’s post hoc test. Statistical significance was defined as p 0.05. Data are presented as mean $\pm$ standard deviation (SD) unless otherwise noted.

We established a mouse model of TBI to investigate the ferroptosis. MWM test was employed to assess cognitive function in mice. Compared to the sham group, TBI resulted in increased path lengths (Fig. 1A) and a decreased number of dentate gyrus neurons (Fig. 1B). Brain tissue images and the NSS confirmed the successful creation of the TBI model (Fig. 1C,D). Histological analyses, including HE, Nissl, Perls, FJB staining, and immunofluorescence, revealed significant brain injury (Fig. 1E), cytoplasmic shrinkage or nuclear pyknosis (Fig. 1F), iron-positive cells (Fig. 1G, Supplementary Fig. 2), and degeneration of cortical neurons (Fig. 1H, Supplementary Fig. 2) in TBI tissues, along with reduced expression of the neuronal marker NeuN (Fig. 1I, Supplementary Fig. 2). Additionally, TBI tissues exhibited significantly elevated levels of MDA, Fe²⁺, total iron, and lipid ROS, along with significantly decreased SOD levels (Fig. 1J, p 0.05), indicative of ferroptosis. Correspondingly, protein and mRNA levels of the ferroptosis-suppressing factors GPX4 and SLC7A11 were reduced in TBI mice brain tissues (Fig. 1K,L, p 0.05). Immunofluorescence staining further revealed significantly diminished fluorescence intensity for GPX4 and SLC7A11 in the TBI group compared to the sham group (Fig. 1M, Supplementary Fig. 3, p 0.05). Screening of m6A modulators in TBI tissues identified METTL3, an m6A writer, as significantly downregulated at the mRNA level (Fig. 1N, p 0.05). This decrease in METTL3 expression was further confirmed by western blot and immunofluorescence (Fig. 1O,P, Supplementary Fig. 3, p 0.05). These findings suggest that METTL3 may play a role in regulating ferroptosis during TBI.

Fig. 1.

Downregulation of METTL3 in TBI-induced brain injury. (A) The Morris Water Maze (MWM) test was performed on day 7 post-TBI. (B) Crystal violet staining of brain tissues. Scale bar = 40 µm. (C) Brain representative images. Scale bar = 1 mm. (D) Neurological severity score (NSS) scores indicating brain tissue damage. (E–I) Representative images of (E) HE, scale bar = 2 mm, (F) Nissl, scale bar = 100 µm, (G) Perls (arrows indicate iron-positive cells), Scale bar = 100 µm, (H) Fluoro-Jade B (FJB), scale bar = 100 µm, and (I) immunofluorescence staining of markers of neuron (NeuN) in brain sections, scale bar = 200 µm or 50 µm; cortical contusion areas are represented by rectangles. (J) The cortical levels of MDA, SOD, Fe²⁺, total iron, and lipid ROS were measured. (K–M) The mRNA and protein levels of GPX4 and SLC7A11 in cortex tissues of TBI mice were measured by (K) qPCR, (L) western blot, and (M) immunofluorescence staining, scale bar = 100 µm. (N) The mRNA levels of m6A modification regulators in cortex tissues were measured by qPCR assay. (O,P) The protein level of METTL3 in cortex tissues was detected by (O) western blot and (P) immunofluorescence staining. Scale bar = 100 µm. **p 0.01. n = 3 for each group. TBI, traumatic brain injury; MDA, malondialdehyde; SOD, superoxide dismutase; ROS, reactive oxygen species; m6A, N6-methyladenosine; qPCR, Quantitative PCR.

To assess the impact of METTL3 on TBI progression, we administered METTL3 overexpression in the TBI model.

qPCR and western blot analyses confirmed that this treatment successfully restored the mRNA and protein levels of METTL3 in TBI-damaged brain tissues (Fig. 2A–C, Supplementary Fig. 4, p 0.05). Notably, METTL3 overexpression enhanced cognitive function (Fig. 2D) and reduced brain damage induced by TBI (Fig. 2E–G, p 0.05). Histological staining of cortex tissues revealed that METTL3 overexpression alleviated the brain injury (Fig. 2H), nuclear pyknosis (Fig. 2I), presence of iron-positive cells (Fig. 2J, Supplementary Fig. 5), and degeneration of cortical neurons (Fig. 2K, Supplementary Fig. 5), while also increasing NeuN-positive cells (Fig. 2L, Supplementary Fig. 5). Moreover, METTL3 overexpression reduced the production and secretion of inflammatory factors IL-6, TNF-$\alpha$, and IL-1$\beta$ (Fig. 2M–O, Supplementary Fig. 6, p 0.05) and inhibited the activation of Iba-1+ microglia and GFAP+ astrocytes during TBI (Fig. 2P, Supplementary Fig. 7). Furthermore, METTL3 treatment reduced ferroptosis biomarkers, including the MDA, Fe²⁺, total iron, and lipid ROS, upregulated SOD levels (Fig. 2Q, p 0.05), and concurrently promoted the transcription and expression of GPX4 and SLC7A11 (Fig. 2R–T, Supplementary Figs. 7,8, p 0.05).

Fig. 2.

Overexpression of METTL3 alleviates the TBI damages and ferroptosis. (A–C) The mRNA and protein levels of METTL3 in mouse brain tissues were measured by (A) qPCR, (B) western blot, and (C) immunofluorescence staining. Scale bar = 100 µm. (D) The Morris Water Maze (MWM) test was performed on day 7 post-TBI. (E) Crystal violet staining of brain tissues. Scale bar = 40 µm. (F) Representative images of the brain. (G) NSS scoring of brain tissue damage. (H–L) Representative images of (H) HE, scale bar = 2 mm, (I) Nissl, scale bar = 100 µm. (J) Perls, scale bar = 100 µm. (K) Fluoro-Jade B (FJB), scale bar = 100 µm, and (L) immunofluorescence staining of markers of neuron (NeuN) in brain sections, scale bar = 100 µm; Cortical contusion areas are represented by rectangles. (M–O) The mRNA and protein levels of IL-6, TNF-$\alpha$, and IL-1$\beta$ in cortex tissues of TBI mice were measured by (M) qPCR, (N) western blot, and (O) ELISA experiment. (P) Immunofluorescence staining of GFAP and Iba-1. Scale bar = 100 µm. (Q) The cortical levels of MDA, SOD, Fe²⁺, total iron, and lipid ROS were measured. (R–T) The mRNA and protein levels of GPX4 and SLC7A11 in cortex tissues of TBI mice were measured by (R) qPCR, (S) western blot, and (T) immunofluorescence staining. Scale bar = 100 µm. *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001. n = 3 for each group.

To investigate the protective role of METTL3 on neurons in vitro, we first isolated mouse primary cortical neurons, which expressed the neuronal marker NeuN (Supplementary Fig. 9) and subsequently established a cellular model to simulate mechanical injury akin to that experienced during TBI. We treated these cells with METTL3 overexpression. The downregulation of METTL3 observed in damaged neurons was effectively reversed by its overexpression (Fig. 3A,B, p 0.05). Notably, METTL3 overexpression significantly enhanced neuronal viability (Fig. 3C, p 0.05) and mitigated ferroptosis triggered by mechanical stress (Fig. 3D,E, p 0.05). Furthermore, while the transcription and expression levels of GPX4 and SLC7A11 were diminished in mechanically damaged neurons, METTL3 overexpression restored these levels (Fig. 3F–K, Supplementary Fig. 10, p 0.05), indicating a suppressive effect on ferroptosis.

Fig. 3.

Overexpression of METTL3 protects neuron viability and alleviates ferroptosis in vitro. Primary neurons and HT-22 cells were processed by mechanical injury and treated with METTL3 overexpression. The mRNA and protein levels of METTL3 were measured by (A) qPCR and (B) western blot. (C) Cell viability was detected by CCK-8 assay. (D,E) The levels of MDA, SOD, Fe²⁺, total iron, and lipid ROS were measured. (F,G) The mRNA levels of GPX-4 and SLC7A11 in cells were detected by qPCR. (H–K) The protein expression of GPX4 and SLC7A11 in cells was determined by (H,I) immunofluorescence staining, scale bar = 100 µm, and (J,K) western blot assay. *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001. n = 3 for each group.

To explore the underlying mechanisms of METTL3-mediated neuronal functions, we conducted a knockdown of METTL3 in primary neurons and HT-22 cells, selecting siMETTL3-1 for further analysis (Fig. 4A,B, p 0.05). The knockdown of METTL3 led to a decrease in total m6A levels on RNA (Fig. 4C, p 0.05) and reduced m6A enrichment on GPX4 mRNA (Fig. 4D, p 0.05). Additionally, METTL3 depletion resulted in enhanced GPX4 RNA degradation in neurons and HT-22 cells (Fig. 4E, p 0.05). YTHDF2 was found to bind to GPX4 mRNA as demonstrated by RNA pulldown assays (Fig. 4F), suggesting its role in METTL3-mediated regulation of GPX4 m6A modification and stability. We further depleted YTHDF2 using siRNAs to elucidate the METTL3/YTHDF2 axis in neurons. siYTHDF2-1 effectively downregulated YTHDF2 mRNA and protein expression (Fig. 4G,H, Supplementary Figs. 11,12, p 0.05). RIP assays showed that METTL3 knockdown decreased YTHDF2 binding to GPX4 mRNA (Fig. 4I, p 0.05). Overexpression of YTHDF2 rescued GPX4 mRNA and protein levels and stability in the absence of METTL3 (Fig. 4J, p 0.05), consequently increasing the GPX4 protein expression (Fig. 4K, Supplementary Fig. 13, p 0.05). Neurons were treated with actinomycin D, and the mRNA level of GPX4 was detected by qPCR assay.

Fig. 4.

METTL3 epigenetically regulates GPX4 expression in neurons. (A,B) Neurons were transfected with siMETTL3, and the (A) mRNA and (B) protein levels of METTL3 were measured by qPCR and western blot assay. (C) The total m6A level was measured by m6A-ELISA assay. (D) The m6A enrichment on GPX4 mRNA was detected by MeRIP assay. (E) Neurons were treated with actinomycin D, and the mRNA level of GPX4 was detected by qPCR assay. (F) The interaction between YTHDF2 and GPX4 mRNA was measured by RNA pulldown assay. (G,H) Neurons were transfected with siYTHDF2, and the (G) mRNA and (H) protein levels of YTHDF2were measured by qPCR and western blot assay. (I) RIP-qPCR assay was performed to measure the binding of YTHDF with GPX4 mRNA. (J) Neurons were transfected with siMETTL3 and YTHDF2 overexpression vectors, and the mRNA level of GPX4 was measured by qPCR assay. (K) The protein level of GPX4 was measured by western blot assay. **p 0.01, ***p 0.001, ****p 0.0001. n = 3 for each group.

To determine if METTL3 influences ferroptosis in TBI-induced neuronal damage through GPX4 regulation, we conducted further analyses. qPCR and western blot assays revealed that GPX4 knockdown or erastin treatment did not affect METTL3 expression in neurons (Fig. 5A,B, Supplementary Fig. 14). Notably, the neuronal proliferation enhanced by METTL3 in the damage model was counteracted by GPX4 knockdown and erastin treatment (Fig. 5C, p 0.05). This effect was characterized by significantly increased levels of MDA, Fe²⁺, total iron, and ROS, coupled with decreased SOD levels in the GPX4 knockdown and erastin groups (Fig. 5D,E).

Fig. 5.

METTL3 modulates ferroptosis in neurons via regulating GPX4 in vitro. Primary neurons and HT22 cells were induced with mechanical damage and treated with METTL3 overexpression, siGPX4, and erastin. (A,B) The mRNA and protein levels of METTL3 were measured by (A) qPCR and (B) western blot. (C) Cell viability was detected by CCK-8 assay. (D,E) The levels of MDA, SOD, Fe²⁺, total iron, and lipid ROS were measured. (F,G) The mRNA levels of GPX4 and SLC7A11 in cells were detected by qPCR. (H,I) The protein expression of GPX4 and SLC7A11 in cells was determined by (H) western blot and (I) immunofluorescence staining assay. Scale bar = 100 µm, ns: p $>$ 0.05, *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001. n = 3 for each group.

Additionally, both mRNA and protein levels of GPX4 and SLC7A11 were elevated with METTL3 overexpression, but this increase was attenuated by GPX4 knockdown and erastin treatment (Fig. 5F–I, Supplementary Figs. 14,15, p 0.05).

We investigated the role of METTL3 in regulating GPX4 during TBI using a mouse model with METTL3 overexpression and treatments with siGPX4 or erastin. MWM test results showed that METTL3 enhanced cognitive function in TBI mice (Fig. 6A), reduced NSS scores (Fig. 6B, p 0.05), and promoted brain recovery (Fig. 6C). These beneficial effects were counteracted by siGPX4 and erastin treatments. Crystal violet and HE staining revealed fewer dentate gyrus neurons (Fig. 6D) and smaller damaged brain areas (Fig. 6E) in METTL3-treated mice, which were negated by GPX4 depletion and erastin-induced ferroptosis. Additionally, METTL3 ameliorated nuclear pyknosis (Fig. 6F), iron-positive cells (Fig. 6G), cortical neuron degeneration (Fig. 6H), and loss of NeuN-positive neurons (Fig. 6H, Supplementary Fig. 16), effects reversed by siGPX4 and erastin. Furthermore, GPX4 knockdown and erastin increased the production and secretion of inflammatory factors IL-6, TNF-$\alpha$, and IL-1$\beta$ (Fig. 7A–C, Supplementary Fig. 17, p 0.05). The METTL3-mediated decrease in Iba-1+ microglia and GFAP+ astrocytes during TBI was also reversed by siGPX4 and erastin (Fig. 7D).

Fig. 6.

METTL3 alleviates brain damage during TBI via regulating GPX4 in vivo. (A) The Morris Water Maze (MWM) test was performed on day 7 post-TBI. (B) NSS score of brain damage. (C) Representative images of the brain. (D) Crystal violet staining of brain tissues. Scale bar = 100 µm. (E–H) Representative images of (E) HE, scale bar = 2 mm, (F) Nissl,scale bar = 100 µm, (G) Perls,scale bar = 100 µm, (H) Fluoro-Jade B (FJB) and immunofluorescence staining of markers of neuron (NeuN) in brain sections, scale bar = 100 µm; Cortical contusion areas are represented by rectangles. ***p 0.001, ****p 0.0001. A1, sham; A2, model + NC; A3, model + METTL3; A4, model + METTL3 + siGPX4; A5, model + METTL3 + erastin. n = 3 for each group.

Fig. 7.

METTL3 alleviates brain inflammation response during TBI via regulating GPX4 in vivo. (A–C) The mRNA and protein levels of IL-6, TNF-$\alpha$, and IL-1$\beta$ in cortex tissues of TBI mice were measured by (A) qPCR, (B) western blot, and (C) ELISA experiment. (D) Immunofluorescence staining of GFAP and Iba-1. Scale bar = 100 µm, **p 0.01, ***p 0.001, ****p 0.0001. A1, sham; A2, model + NC; A3, model + METTL3; A4, model + METTL3 + siGPX4; A5, model + METTL3 + erastin. n = 3 for each group.

Assessments of ferroptosis biomarkers revealed that METTL3 mitigated TBI-induced ferroptosis, as evidenced by decreased levels of MDA, Fe²⁺, total iron, and lipid ROS, coupled with increased SOD levels (Fig. 8A, p 0.05).

Fig. 8.

METTL3 alleviates ferroptosis during TBI via regulating GPX4 in vivo. (A) The cortical levels of MDA, SOD, Fe²⁺, total iron, and lipid ROS were measured. (B–D) The mRNA and protein levels of GPX4 and SLC7A11 in cortex tissues of TBI mice were measured by (B) qPCR, (C) western blot, and (D) immunofluorescence staining. Scale bar = 100 µm, ns: p $>$ 0.05, *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001. n = 3 for each group.

Conversely, siGPX4 silencing and erastin treatment negated these protective effects. Additionally, siGPX4 and erastin treatment downregulated the METTL3-induced upregulation of GPX4 and SLC7A11 at mRNA and protein levels (Fig. 8B–D, Supplementary Figs. 18,19, p 0.05). Collectively, these findings indicate that METTL3 exerts neuroprotective effects against TBI by modulating GPX4 expression.