Histone H3K9 lactylation activates the TXNIP/NLRP3 pathway to drive macrophage inflammation after spinal cord injury

Abstract

Spinal cord injury (SCI) induces metabolic and immune disruptions that impede tissue repair. However, the underlying mechanisms are poorly understood. In this study, we identified lactate accumulation as a critical driver of macrophage-mediated inflammation through histone H3K9 lactylation (H3K9la). Targeted metabolomics revealed elevated serum lactate levels in SCI patients, which were linked to increased glycolysis and lactate dehydrogenase activity. In mice, lactate accumulation after SCI was found to drive histone H3K9la in lesion-infiltrating macrophages and circulating monocytes. Integrated CUT&Tag and RNA-seq analysis revealed that thioredoxin-interacting protein (TXNIP) is a direct H3K9la target that activates the TXNIP-NLRP3 pathway, exacerbating inflammation and impairing mitochondrial function. In vitro, glycolytic inhibition reversed lactate-induced inflammation and mitochondrial dysfunction. In vivo, a hypoxia-responsive peptide inhibitor (H3K9la-pe) selectively reduced macrophage lactylation and inflammation, restored mitochondrial integrity, promoted axon regeneration, and significantly improved functional recovery in SCI mouse model. These findings elucidate a subacute metabolic-epigenetic-inflammatory axis in SCI and highlight that blocking macrophage H3K9la is a promising therapeutic strategy.

Graphical abstract

The schematic model illustrates that lactate accumulation in the injury microenvironment occurs not only locally within the spinal cord tissue but also exerts distal effects on peripheral blood, inducing H3K9la in macrophages. We further demonstrated that H3K9la transcriptionally activates the pro-inflammatory gene TXNIP, thereby triggering the TXNIP-NLRP3 inflammatory signaling pathway, which amplifies the inflammatory response and contributes to mitochondrial dysfunction.

1. Introduction

Spinal cord injury (SCI) is a severe traumatic disorder of the central nervous system characterized by irreversible sensory and motor deficits [1]. The global incidence and prevalence of SCI have steadily increased over recent decades [2,3]. The pathophysiology of SCI encompasses initial mechanical trauma and a cascade of secondary injury processes, prominently involving metabolic dysregulation. These secondary events include ischemia, edema, inflammation, rapid cell death, thrombosis, vasospasm-induced ischemia, cyst formation, and extensive infiltration of immune cells, all of which collectively contribute to progressive axon degeneration and impaired neural repair [1,[4], [5], [6]]. Persistent ischemia and hypoxia after SCI promote a shift toward aerobic glycolysis, significantly disrupting energy homeostasis, lipid metabolism, and redox balance [7,8]. However, the precise contribution of these metabolic alterations to SCI pathology remains poorly understood.

Metabolite-driven post-translational modifications (PTMs) are critical connections between altered cellular metabolism and transcriptional regulation. Histone lactylation is a novel epigenetic modification first described in 2019; it serves as an important regulator of chromatin remodeling, gene expression, and cellular phenotype transitions [9]. Under ischemic-hypoxic conditions, high lactate levels can drive histone lactylation, reprogramming gene networks involved in the resolution of inflammation, tissue repair, and immune cell polarization [10]. In contrast, in central nervous system disorders, increased histone lactylation is associated with an increase in glycolytic gene expression, glycolytic flux, and microglial activation, thereby amplifying inflammatory responses [11].

Neuroinflammation, driven primarily by the activation of innate immune cells in the central nervous system, is a key mediator of secondary injury after SCI. Macrophages (both monocyte-derived and resident) are essential immune effectors that orchestrate inflammatory responses during the acute and subacute phases following injury [12]. While early macrophage infiltration is crucial for the clearance of debris and initial tissue repair, sustained activation can prolong inflammation, exacerbate secondary damage, and hinder functional recovery [13,14]. Although macrophages are a prime cellular target for lactate-driven histone lactylation, the specific role of lactylation in modulating the macrophage phenotype and function in the SCI microenvironment remains largely undetermined.

In this study, we found that lactate accumulates not only in the local spinal cord microenvironment but also in the peripheral circulation, where it promotes H3K9la in macrophages. Through fluorescence--activated cell sorting combined with CUT&Tag and RNA sequencing, we found that H3K9la transcriptionally activates the key inflammatory gene TXNIP and thereby mediates activation of the TXNIP-NLRP3 inflammasome pathway, exacerbating inflammation and mitochondrial dysfunction. We engineered an H3K9la-targeted inhibitory peptide and accurately delivered it by coupling it to a 2-nitroimidazole derivative that exploits the hypoxic pathological microenvironment unique to SCI. Administering this peptide considerably suppressed macrophage lactylation and ultimately promoted neurological recovery after SCI. This is the first study to elucidate how aberrant lactate metabolism fine-tunes macrophage-mediated inflammation via histone lactylation. This study also revealed a unique metabolism-immune axis-based targeted therapeutic strategy for SCI.

2. Materials and methods

2.1. Mice SCI model

Eight-week-old female C57BL/6 mice (20–25 g) were obtained from Hunan SJA Laboratory Animal Company and maintained under specific-pathogen-free conditions at the Central South University Animal Facility. Animals were housed individually in standard cages under a 12 h light/12 h dark cycle, with ad libitum access to food and water, at 22–24 °C and 60–80% relative humidity. Manual bladder expression and general husbandry were performed two to four times daily. All procedures were approved by the Animal Care and Use Committee of Central South University (Changsha, China).

For SCI induction, T10 crush injury was performed at thoracic level 10 (T10) similar to that described before [15]. Briefly, mice were anesthetized, the dorsal surface prepared aseptically, and a midline incision was made over the T10 spinous process. Paraspinal muscles were retracted and a T10 laminectomy performed to expose the dorsal cord. Severe crush injury was delivered using No. 5 Dumont forceps (Fine Science Tools; tip width 0.5 mm) without spacers to bilaterally compress the cord for 5 s. The wound was irrigated and closed in anatomical layers before application of a sterile dressing. Postoperatively, mice received intramuscular penicillin (20,000 U) twice daily for three days for infection prophylaxis and were monitored with manual bladder voiding two to four times daily. Standard chow and water remained available at all times.

2.2. Cell culture and treatment

Primary bone marrow-derived macrophages (BMDMs) were obtained from eight-week-old C57BL/6 mice. Femurs and tibias were harvested aseptically after euthanasia and surface sterilization in 75% ethanol. Bone marrow was flushed with IMDM supplemented with 10% FBS, triturated, subjected to red blood cell lysis, and pelleted (1200 rpm, 5 min). Cells were resuspended in IMDM (12440053, Gibco) + 10% FBS (FBS-CP500, NEWZERUM) and plated in T75 flasks overnight to remove stromal adherents. Nonadherent cells were then cultured in IMDM +10% FBS with 20 ng/mL recombinant murine GM-CSF, with media changes every three days. Fully differentiated BMDMs were ready for experiments on day 7. For treatments, cells were exposed to l-lactate (20 mM) (L6402, sigma), sodium lactate (20 mM) (867-56-1, MCE), 2-deoxy-d-glucose (2-DG, 10 mM) (HY-13966, MCE), Mito-TEMPO (50 μM) (HY-112879, MCE) or PBS (control) for 24 h.

2.3. Measurement of lactate levels

Intracellular lactate level was measured by using d-lactate, l-lactate, LDH (Elabscience) Colorimetric/Fluorometric assay kit according to manufacturer's protocol.

2.4. Human peripheral blood plasma acquisition

Peripheral blood plasma samples were obtained from ten patients with spinal cord injury (SCI) who were admitted to the Department of Spine Surgery between May 2023 and May 2025. All participants underwent a comprehensive baseline neurological assessment in accordance with the International Standards for Neurological Classification of Spinal Cord Injury (ISNCSCI). From this cohort, 5 patients classified as American Spinal Injury Association (ASIA) Impairment Scale grade A were selected. Written informed consent and clinical data collection forms were completed prior to surgery. Preoperative peripheral blood was drawn, immediately centrifuged, and the plasma supernatant was flash-frozen on dry ice. Samples were transported on dry ice and stored at −80 °C until further analysis. All data involving human participants and/or human specimens were conducted under the approval of the Xiangya Hospital Ethics Committee (Ethics Approval No. 202307601). Detailed information for patients is shown in Supplementary Table S1 and Supplementary Table S2.

For this study, selection criteria included: 1. age 18-75 years old; 2. patients were diagnosed with SCI; 3. patients no prior surgical intervention.

Exclusion criteria included: 1. age <18 or >75 years old; 2. patients with concomitant severe injuries, particularly traumatic brain injury or pulmonary contusion; 3. patients with underlying comorbidities such as diabetes mellitus or hypertension; 4. clinical data unavailable.

2.5. LC–MS/MS analysis

Lactate- or vehicle-treated BMDMs were lysed in buffer containing 1% Triton X-100, 2 mM EDTA, protease inhibitors, 3 μM TSA, and 50 mM NAM, followed by incubation at 4 °C for 30 min. After centrifugation (20,000 g, 10 min), pellets were extracted with 0.2 M H₂SO₄, sonicated, and clarified by centrifugation. Protein concentration was determined by BCA assay.

Enriched lactylated peptides were dissolved in solvent A (0.1% formic acid, 2% acetonitrile) and separated on a custom C18 column (25 cm × 100 μm) using a nano Elute UHPLC system (Bruker) at 450 nL/min with a gradient of 7–24% solvent B over 42 min, 24–32% over 12 min, then to 80%. For global proteomics, tryptic peptides were separated using a 6–24% gradient over 70 min followed by 24–35%. Peptides were analyzed on a timsTOF Pro mass spectrometer (Bruker Daltonics) operating in parallel accumulation-serial fragmentation mode, with MS/MS scans acquired for precursors with charge states 2–5 and a dynamic exclusion of 30 s.

2.6. Peptide competition assay

Peptide competition assays were performed to validate the specificity of histone modification antibodies. Synthetic histone H3 peptides bearing H3K9 lactylation (H3K9la), H3K9 acetylation (H3K9ac), or no modification were used. Antibodies against H3K9la or H3K9ac were pre-incubated with the indicated peptides at a molar ratio of 1:20 (antibody:peptide) for 1 h at room temperature before being applied to WB. Antibodies incubated without peptide served as controls. Signal intensities were compared to assess modification-specific competition and cross-reactivity.

2.7. Fluorescent activated cell sorting (FACs)

Sham and 14 days post injury mice were anesthetized by intraperitoneal injection of 0.5 mL 0.3% pentobarbital sodium and confirmed to be in a deep, non-responsive state. Animals were placed supine on a heated pad to maintain body temperature, and limbs were secured to the operating table. Using sterile scissors, a midline thoracoabdominal incision was made: the skin and peritoneum were bluntly dissected to expose the diaphragm, which was punctured and then incised along its midline to open the thoracic cavity and reveal the heart. Pre-warmed heparin was drawn into a 1 mL syringe, which was then inserted into the left ventricle to withdraw 0.5 mL of peripheral blood; the sample was immediately transferred into a 15 mL tube containing anticoagulant.

Whole blood (0.5 mL) was mixed with 5 mL of Lysing Buffer (pH 7.1–7.4) (555899, BD Pharmingen), inverted to ensure uniform suspension, and incubated in the dark at room temperature for 15 min. Samples were centrifuged at 200 g for 5 min, and the supernatant discarded. If a red pellet remained visible, lysis and centrifugation were repeated until the pellet was colorless. Cells were washed with 2 mL PBS containing 1% FBS and pelleted again (200 g, 5 min). The resulting cell pellet was resuspended in 150 μL Stain Buffer and blocked with anti-mouse CD16/CD32 (553141, BD Fc Block; 50:1 dilution) for 15 min at 4 °C, then washed (200 g, 5 min). Pellets were stained with 100 μL Zombie Aqua™ (423101, Biolegend) viability dye for 15 min at room temperature in the dark, washed twice (200 g, 5 min, 4 °C, and resuspended in Stain Buffer. Cells were then incubated on ice for 60 min with anti-Ly-6C-APC (128015, Biolegend), anti-CD45-APC/Cy7 (157617, Biolegend), anti-Ly-6G-PE (127607, Biolegend), anti-CD11b-FITC (101205, Biolegend), washed twice in Stain Buffer, and finally resuspended in 1 mL Stain Buffer for flow cytometric analysis.

2.8. Western blotting

Cell lysates were prepared by incubating samples in RIPA buffer supplemented with protease inhibitors on ice for 30 min, followed by centrifugation at 12,000×g for 15 min to remove debris. Protein concentration was determined using a BCA assay. Samples were mixed 4:1 with 5 × Laemmli buffer, denatured at 95 °C for 10 min, and stored at −20 °C. Equal amounts of protein were separated by SDS–PAGE (90 V for 30 min, then 120 V until resolved), and transferred onto PVDF membranes (preactivated in ethanol) at 300 mA for 90 min in an ice-water bath. Membranes were blocked in 5% nonfat milk/TBST for 1.5 h, washed, and incubated overnight at 4 °C with primary antibodies in TBST: anti- Pankla (PTM-1401, PTM Bio, 1:1000), anti- Histone H3(HL102-01, Transgen, 1:1000), anti-β-actin (66009-1-Ig, Proteintech, 1:5000), anti-H3K9la (PTM-1421RM, PTM Bio, 1:1000), anti-H3K14la (PTM-1414, PTM Bio, 1:1000), anti-H3K18la (PTM-1406RM, PTM Bio, 1:1000), anti-H3K9ac (PTM-112RM, PTM Bio, 1:5000), anti- H3K9me (PTM-643, PTM Bio, 1:1000), anti-H3K27ac (PTM-116 RM, PTM Bio, 1:1000), anti- Caspase 1(81482-1-RR, Proteintech, 1:10000), anti- NLRP3(MA5-32255, Invitrogen, 1:1000), anti-TXNIP (14715, CST, 1:1000), anti-IL-1β(26048-1-AP, Proteintech, 1:3000), anti-TNF-α(17590-1-AP, Proteintech, 1:5000), anti-iNOS(22226-1-AP, Proteintech, 1:500). After washing, membranes were incubated 1 h at room temperature with HRP-conjugated secondary antibodies, washed again, and developed with chemiluminescent substrate. Bands were imaged on a ChemiDoc XRS Plus and quantified with ImageJ, normalizing target signal to β-actin or Histone H3.

2.9. RNA extraction, reverse transcription and quantitative RT-PCR

Total RNA was isolated with TRIzol™ reagent and reverse-transcribed using the Transcriptor First Strand cDNA Synthesis Kit according to the manufacturers’ instructions. Quantitative real-time PCR was carried out with the FastStart Universal SYBR® Green Master Mixon a real-time PCR system. Transcript levels were normalized to β-actin, and primer sequences are provided in Supplementary Table S3.

2.10. Immunofluorescence

Spinal cord cryosections were equilibrated to room temperature for 15 min and rinsed twice in PBS. Sections were outlined with a hydrophobic barrier pen and permeabilized in 0.5% Triton X-100/PBS for 30 min, followed by two PBS washes. Samples were blocked in 5% BSA in PBS for 30 min at room temperature, then incubated overnight at 4 °C with primary antibodies diluted in PBS: anti- F4/80 (ab6640, Abcam, 1:400), anti-IBA1 (011-27991, WAKO, 1:400), anti- Pankla (PTM-1401, PTM Bio, 1:50), anti- NLRP3 (MA5-32255, Invitrogen, 1:200), anti- TXNIP (MA5-32771, Invitrogen, 1:100), anti-β3-Tubulin (D71G9, CST, 1:400), anti-Neurofilament-L (C28E10, CST, 1:400), anti- GAP43 (D9C8, CST, 1:400) and anti- H3K9la (PTM-1421RM, PTM Bio, 1:50). After removing primary antibody, sections were washed twice in 0.3% PBST (10 min each) and twice in PBS (10 min each). Fluorophore-conjugated secondary antibodies were applied for 1 h at room temperature in the dark, followed by two washes each in 0.3% PBST and PBS (10 min each). Sections were then counterstained with DAPI (GeneTex), mounted, and imaged by widefield or confocal microscopy. Secondary-only controls were included to assess background. Fluorescence quantification was performed using ImageJ (Fiji v1.53), and three-dimensional reconstructions of z-stacks were generated in Imaris 10.1.

2.11. Mitochondrial oxygen consumption

Oxygen consumption rates (OCR) were measured using a Seahorse XF96 Extracellular Flux Analyzer (Seahorse Biosciences) following the manufacturer's protocol. BMDMs were plated at 4 × 10⁴ cells per well and cultured for the indicated durations in assay medium. Basal mitochondrial respiration was recorded before sequential injections of oligomycin (1 μM), FCCP (1 μM), and rotenone/antimycin A (0.5 μM) to determine ATP-linked respiration, maximal respiratory capacity, and non-mitochondrial respiration.

2.11.1. RNA-seq

Total RNA was extracted from BMDMs using TRIzol™ reagent (Invitrogen, Cat. No. 15596026) following Chomczynski and Sacchi's protocol. Genomic DNA was removed by DNase I treatment. RNA purity (A₆₂₀/₂₈₀) was assessed on a NanoDrop™ One spectrophotometer (Thermo Fisher), and integrity was verified by 1.5% agarose gel electrophoresis. Concentrations were determined with a Qubit™ 3.0 fluorometer using the RNA Broad-Range Assay Kit (Life Technologies, Q10210).For transcriptome analysis, 2 μg of total RNA was used to construct stranded libraries with the KC-Digital™ Stranded mRNA Library Prep Kit (Wuhan Seqhealth; Cat. No. DR08502), which incorporates unique molecular identifiers (8 nt UMIs) to eliminate PCR and sequencing bias. Libraries (200–500 bp) were size-selected, quantified, and sequenced on a DNBSEQ-T7 platform (MGI Tech) in paired-end 150 bp mode, with an average sequencing depth of approximately 20 million reads per sample.

2.11.2. CUT&Tag

CUT&Tag assays were performed using a commercial kit (HD102-01, Vazyme) according to the manufacturer's instructions with minor modifications. Briefly, cells were immobilized on Concanavalin A-coated magnetic beads and gently permeabilized using digitonin. Samples were incubated with a primary antibody specific for the target protein, followed by incubation with Protein G-fused micrococcal nuclease (pG-MNase). Upon calcium activation, MNase selectively cleaved genomic DNA in close proximity to antibody-bound chromatin. Released DNA fragments were purified and used for library construction and sequenced on an Illumina platform. Raw sequencing reads were aligned to the mouse reference genome using standard pipelines. Peak calling was performed using MACS2 v2.2.7.1 with the following parameters: f BAM --call-summits --nomodel --extsize 147 -B -p 0.05. The sequencing depth reached approximately 2.0 Gb per sample, corresponding to ∼0.73×genome coverage based on the mouse genome size (∼2.7 Gb). Peak annotation and downstream analyses were conducted using standard bioinformatics tools.

2.11.3. Chromatin immunoprecipitation (ChIP)-qPCR

BMDMs were seeded at 1 × 10⁵ cells/mL in 10 cm dishes and grown to ∼80% confluence. Cells were crosslinked in 1% formaldehyde (in PBS with 100 mM PMSF and SDS lysis buffer), quenched, washed twice in ice-cold PBS + 1 mM PMSF, and harvested by scraping. Pellets were resuspended in SDS lysis buffer (1 mM PMSF) and sonicated on ice (15 cycles of 10s on/20 s off at 20 W) to shear chromatin to 200–1000 bp. Lysates were cleared (12,000×g, 5 min, 4 °C) and diluted tenfold in ChIP buffer (with 1 mM PMSF). An aliquot (20 μL) was reserved as input. Samples were pre-cleared with 70 μL Protein A/G–agarose–salmon sperm DNA (30 min, 4 °C), then incubated overnight at 4 °C with 1 μg anti-H3K9la (negative controls omitting antibody). Immune complexes were captured with an additional 60 μL Protein A/G–agarose (1 h, 4 °C), washed sequentially with low-salt, high-salt, LiCl, and TE buffers, and eluted twice with 250 μL elution buffer (1% SDS, 0.1 M NaHCO₃). Crosslinks were reversed with 5 M NaCl (65 °C, 4 h), and samples (including input) were treated with EDTA, Tris-HCl, and Proteinase K (45 °C, 1 h), then purified using a commercial DNA cleanup kit. DNA was eluted in 20 μL RNase-free water and analyzed by qPCR.

2.12. Dual-luciferase reporter assay

A dual-luciferase reporter assay was performed in HEK293T cells seeded in 12-well plates. Cells were co-transfected in triplicate with pGL3-Basic or Txnip promoter–pGL3-Basic together with Hist1h3a WT or K9Q mutant expression plasmids, and pRL-TK was included as an internal control at a 1:10 ratio to the firefly reporter. At 24–48 h post-transfection, cells were lysed in 1 × Passive Lysis Buffer, and firefly and Renilla luciferase activities were sequentially measured using a dual-luciferase reporter system. Firefly luciferase activity was normalized to Renilla activity, and the relative luciferase signals were used to assess Txnip promoter activity.

2.13. Locus targeting CRISPR epigenetic editing

Short guide RNAs (sgRNAs) used in CRISPRi were designed using CRISPOR [16]. sgRNAs were cloned into a U6-driven vector using BbsI, and co-transfected with an EF1α–dCas9-KRAB–2A-Blasticidin effector plasmid using Lipofectamine™ 3000 (L3000015, Invitrogen). After antibiotic selection, Txnip expression was assessed by WB, confirming effective promoter repression.

2.13.1. Co-immunoprecipitations assay

Primary BMDMs were seeded at 1 × 10⁵ cells/mL in T75 flasks and grown to ∼80% confluence, detached with Accutase, and pelleted. Lysates were prepared by incubating 10⁶ cells in 100–200 μL ice-cold lysis buffer containing protease inhibitors for 15 min, then cleared by centrifugation (14,000 g, 5 min, 4 °C). Supernatants (500 μL) were incubated overnight at 4 °C with 1 μg anti-TXNIP or control rabbit IgG (1:50 in TBS) bound to 20 μL Protein A/G magnetic beads. Beads were washed three times with 500 μL TBS (magnetic separation, gentle pipetting), then eluted by resuspending in 20 μL lysis buffer. Eluates (20 μL) were mixed with 100 μL SDS–PAGE loading buffer, denatured at 95 °C for 5 min, and analyzed by WB.

2.14. Flow cytometry and ROS detection

BMDMs cultured in six-well plates were harvested and fixed in Fixation Buffer (Biosciences, 554655) for 20 min at 4 °C. Cells were then permeabilized with 0.1% Triton X-100 for 10 min and blocked in 3% bovine serum albumin for 20 min. Intracellular reactive oxygen species were detected by staining with 2′, 7′-dichlorofluorescein diacetate (DCFH-DA; Beyotime Biotechnology, S0033S) following the manufacturer's protocol. Samples were analyzed on a flow cytometer, and data were quantified using FlowJo V10.

2.15. Mitochondrial imaging by transmission electron microscopy (TEM)

Primary BMDMs (1 × 10⁵ cells/mL, 15 mL) were plated in T75 flasks and grown to >80% confluence, then treated for 24 h with PBS, 20 mM lactate, or 10 mM 2-DG. Cells were washed twice with PBS, detached with Accutase, and pelleted (1000 rpm, 5 min). Pellets were fixed in 1 mL cold 2.5% glutaraldehyde at 4 °C for 4–6 h, washed three times in 0.1 M phosphate buffer (15 min each), post-fixed in 1% osmium tetroxide for 2 h, and washed again (3 × 15 min). Samples were dehydrated in graded acetone (50%, 70%, 90% for 15 min each; 100% twice for 20 min), then infiltrated in 1:1 acetone: resin at 37 °C for 12 h, followed by pure resin for 12 h. Blocks were polymerized at 37 °C overnight and at 60 °C for 24 h. Ultrathin sections (100 nm) were cut on a Leica UC7, double-stained with 3% uranyl acetate and lead citrate, and imaged on a Hitachi HT7700 TEM. Three magnifications of each field were captured.

2.16. Peptide synthesis

All peptides were synthesized by Guoping Pharmaceutical Inc. (Hefei, China) using standard Fmoc solid-phase peptide synthesis on Rink amide resin (0.15 mmol scale). Amino acids were sequentially coupled from the C-terminus to the N-terminus using DIC/HOBt or HBTU/DIEA as coupling reagents. Fmoc deprotection was performed with 20% piperidine in DMF. Site-specific lysine lactylation was introduced on-resin after selective removal of the MMT protecting group, followed by coupling of the lactyl moiety using DIC/HOBt chemistry. After chain assembly, peptides were cleaved from the resin using a TFA-based cleavage cocktail (95% TFA, 1% H₂O, 2% EDT, 2% TIS) for 2 h at room temperature. Crude peptides were precipitated in cold diethyl ether, collected by centrifugation, and lyophilized. Purification was performed by preparative reverse-phase HPLC (Shimadzu HPLC-20AD/AT system) using a C18 column (20 × 250 mm, 8 μm; Daiso), with a linear gradient of 0.1% TFA in water (solvent A) and 0.1% TFA in acetonitrile (solvent B). Purified peptides exhibited a purity of >95%, as confirmed by analytical HPLC, and molecular weight was verified by LC-MS (Shimadzu LCMS-2020).

For in vivo experiments, peptides composed of D-amino acids were used to enhance proteolytic stability. Peptides were dissolved in sterile PBS, stored on ice, protected from light, and equilibrated to room temperature immediately prior to use. Based on preliminary tissue distribution and pharmacokinetic analyses, a dose of 5 mg/kg was selected to achieve sufficient spinal cord exposure. Mice were randomly assigned to experimental groups and received once-daily intraperitoneal injections of H3K9la-pe or vehicle beginning immediately after spinal cord injury and continuing throughout the subacute phase.

2.17. Hematoxylin and Eosin (H&E) staining

On 28 days post-injury, mice were euthanized and bladders excised for gross imaging. Tissues were fixed in formalin, dehydrated through an ethanol series, embedded in paraffin, and sectioned. Sections were stained with a H&E Staining Kit (Beyotime, C0105S) according to the manufacturer's instructions. Stained sections were imaged on a Nikon inverted microscope and analyzed using ImageJ.

2.18. In vivo imaging

Vivo 3D Multimodal Imaging System (Ex/Em: 754/778 nm). Images were analyzed with Milabs-Ol-PP v2.4.13 software, and mean fluorescence intensity was quantified over an identical region of interest encompassing the injury. All injections and imaging procedures were conducted under sterile conditions and in accordance with institutional animal care and use guidelines.

2.19. Axon tracing

Axon regeneration in SCI mice was evaluated using anterograde viral tracing. Under anesthesia (0.3% pentobarbital sodium, 0.5 mL i.p.), the T8 vertebral level was exposed via midline skin and muscle incisions, and a partial laminectomy was performed to expose the dura. Mice were secured in a stereotaxic frame, and the dorsal spinal artery at the center of the exposed segment was designated as the origin (X/Y = 0). A 3 × 2 injection grid was defined at X = −0.3, 0, and 0.3 mm and Y = 0 and 0.4 mm. At each site, 2 μL of AAV2/9-Syn-EGFP (Genechem) was injected in three depths (Z = 0.9, 0.6, and 0.4 mm) using a microinjection system (100 nL per depth at 80 nL/min), with the needle left in place for 2 min post-injection to facilitate dispersion. Muscle and skin were sutured in layers, the wound disinfected, and animals were recovered on a 37 °C warming pad. Postoperatively, penicillin (50,000 U/kg, i.m.) was administered twice daily for three days to prevent infection. Two weeks later (4 weeks post-SCI), mice were perfused, and spinal cords were harvested, cryosectioned at 20 μm, and counterstained with DAPI. EGFP-labeled axons were imaged by confocal microscopy, and regeneration distance and fiber density were quantified.

2.20. Corticospinal tract tracing

The procedure of tracing the corticospinal tract (CST) follows the protocol described in this published paper [17]. Briefly, a vertical midline incision was made from the interocular region to the posterior skull. Using bregma as the reference point for both mediolateral and anteroposterior coordinates, bilateral cranial windows (5 mm × 2 mm) were created, with the medial edges positioned 0.5 mm lateral to bregma. Using a digital stereotaxic injector (Item No. 51709, Stoelting, USA), 0.5 μL of biotin-dextran amine (BDA; MW 10,000; 10% in PBS; Molecular Probes) was injected at each of 10 cortical sites (5 per hemisphere). Injection coordinates were as follows: mediolateral, 1.5 mm lateral to bregma; anteroposterior, −1.0, −0.75, −0.5, 0, and −1.5 mm from bregma; and dorsoventral, 0.5 mm below the cortical surface. BDA was delivered at a rate of 0.1 μL/min, and the injection needle was left in place for 5 min after each injection to ensure adequate tracer diffusion. At 28 dpi mice were anesthetized and perfused with 4% paraformaldehyde, and spinal cord tissues were collected for CST distribution analysis.

2.21. Electrophysiology

Capture the mice to be tested and anesthetize with 0.3% sodium pentobarbital. Shave, disinfect the animal, and securely fixate it in a stereotaxic apparatus, maintaining body temperature with a heating pad. Perform a craniotomy to expose the M1 area of the motor cortex. Insert stimulation electrodes, guided by the stereotaxic device, to a depth of 700-1000 μm from the brain surface, targeting corticospinal neurons in the sensorimotor cortex. Position recording electrodes to penetrate the contralateral thigh's sciatic nerve distal end to record muscle action potentials induced by electrical stimulation. Amplify and record bioelectrical signals using the BL-420F Biological Function Experiment System. Wave width was defined as the time interval between the positive peak and the subsequent negative peak of the evoked waveform. For each animal, wave width was averaged from 3 to 5 consecutive traces, and the mean value was used for statistical analysis. Stimulation parameters are as follows: stimulation type: micro-voltage, mode: single pulse; delay 100 ms; frequency 100 Hz, stimulation intensity: 14 V.

2.22. BMS score evaluation and LSS swimming

Hindlimb motor recovery was assessed using the Basso Mouse Scale (BMS; 0–9) pre-injury and at 1, 3, 7, 14, and 28 dpi. Prior to each session, mice were acclimated on the testing platform and observed for 4 min.

For the swim test, mice received three days of pre-operative training in a 5 × 15 cm water tank. On 28 dpi, swimming performance was recorded for 30 s and scored using the Louisville Swim Scale (LSS), which evaluates hindlimb movement, alternation, forelimb dependence, trunk stability, and body angle.

All mice were randomly assigned, and two trained, blinded observers independently scored each animal. Final BMS and LSS values represent the mean of both observers’ scores.

2.23. Hindlimb kinematic analysis

For detailed hindlimb kinematic analysis, mice were placed on a transparent walkway (80 cm in length and 4 cm in width) to assess overground locomotion. Anatomical landmarks for the hindlimb included the iliac crest, hip joint, knee joint, ankle joint, and distal toe. For each mouse, 4 to 6 consecutive gait cycles were recorded in the central 20 cm region of the walkway at 4K resolution and 120 frames per second, ensuring that the head remained stable and without rotation. Videos were analyzed offline using DeepLabCut to automatically track bony landmarks. Blinded analyses were performed to quantify the maximum iliac crest height, toe height, stride length, and knee–ankle angle oscillation for each gait cycle. Stick diagrams of hindlimb movement were subsequently generated in R studio.

2.24. Statistical analysis

Statistical analyses were performed using GraphPad Prism (version 9.0; GraphPad Software). Data are presented as mean ± standard deviation (SD). Normality of data distribution was assessed using the Shapiro–Wilk test.

For comparisons between two groups, variance homogeneity was evaluated using an F test. When assumptions of normality and equal variance were met, an unpaired two-tailed Student's t-test was applied. For comparisons involving more than two groups, variance homogeneity was assessed using the Brown–Forsythe test. When appropriate, one-way analysis of variance (ANOVA) followed by Tukey's multiple-comparisons test was used. For experiments involving two independent variables (e.g., treatment and time), two-way ANOVA with post hoc multiple-comparisons testing was performed as indicated.

All statistical tests were two-sided. A P value < 0.05 was considered statistically significant. Exact P values and sample sizes (n) are provided in the figure legends. Statistical significance is denoted in the figures as follows: ns denotes p ≥ 0.05, ∗ denotes p < 0.05, ∗∗ denotes p < 0.01, ∗∗∗ denotes p < 0.001, and ∗∗∗∗ denotes p < 0.0001.

3. Results

3.1. Aberrant elevation of lactate levels after SCI

By comparatively analyzing 110 metabolites in the cerebrospinal fluid and serum of SCI patients, Wu et al. identified lactate, as potential biomarkers for assessing the severity of SCI, and subsequent experimental studies have demonstrated that lactate can also distinguish injury severity in animal models of SCI [18,19]. Though lactate level increases in various neurological disorders, including brain injury, neurodegenerative diseases, and neuropathic pain [20,21], the systemic alterations in lactate metabolism following SCI remain to be investigated. To determine the systemic metabolic changes, we collected peripheral blood serum samples from clinically diagnosed SCI patients within seven days of injury and performed targeted energy metabolomics profiling. To compare the metabolic alterations in the serum of SCI patients, we constructed a heatmap after hierarchical clustering analysis, which revealed 50 metabolites detected in the lesion center of the spinal cord (Fig. 1A). Volcano plot analysis of 32 energy metabolism–related metabolites in serum samples from SCI patients revealed a significant post-injury increase in glycolysis-associated metabolites, including lactate, d-glucose-6-phosphate, and β-d-fructose-6-phosphate (Fig. 1B). Among the six most significantly altered metabolites, the lactate level exhibited pronounced and consistent increase following injury (Fig. 1C).

Fig. 1.

Aberrant Elevation of Lactate Levels in Peripheral Blood and Lesion Site After SCI. (A) The heatmap illustrates the top 50 significantly altered energy metabolism–related metabolites in patients with spinal cord injury (SCI). (B) Comparative analysis of these 50 metabolites using a volcano plot identified the most significantly dysregulated functional metabolites (highlighted in red), which exhibited marked changes in the serum following SCI. (C) Box plots show the expression levels of six representative energy metabolism–related metabolites that were significantly altered after SCI (n = 5). (D) Quantification of l-lactate and d-lactate concentrations in spinal cord tissue from sham-operated mice and at days 1, 3, 7, and 14 post-SCI (n = 3, one-way ANOVA, Tukey's post hoc test). ns: not significant; ∗∗p < 0.01; ∗∗∗∗p < 0.0001. (E) Quantitative analysis of relative LDH activity in spinal cord tissue at corresponding time points (n = 3, one-way ANOVA, Tukey's post hoc test). ns: not significant; ∗p < 0.05; ∗∗∗p < 0.001. (F) Quantification of l-lactate concentration and LDH activity in peripheral blood at days 1, 3, 7, and 14 post-SCI, compared to sham controls (n = 3, one-way ANOVA, Tukey's post hoc test). ns: not significant; ∗∗p < 0.01. (G) Quantitative analysis of peripheral blood l-lactate levels in clinical SCI patients versus healthy individuals (n = 6, unpaired t-test). ∗∗∗∗p < 0.0001.

To further determine whether lactate significantly accumulates in the injured region after SCI in a time-dependent manner, we collected spinal cord samples from SCI mice at different time points. Colorimetric assay revealed that the levels of l-lactate and d-lactate remained low in the sham group but progressively increased after injury, peaking on 7 days post-injury (dpi). By 14 dpi, however, no significant difference was found in lactate levels compared to their levels on 7 dpi (Fig. 1D). l-lactate is the predominant isoform in the human body and plays essential roles in energy metabolism and acid-base homeostasis, whereas d-lactate is typically produced by bacteria through a distinct metabolic pathway [22]. We further showed that d-lactate exhibited a more rapid increase, with significant changes observed as early as 3 dpi, which may be associated with the acute inflammatory response following SCI. In contrast, l-lactate, an endogenous metabolite, showed more pronounced changes over time, with a significant increase detected on 7 dpi. Therefore, we selected l-lactate for further characterization in subsequent experiments.

Lactate dehydrogenase (LDH) catalyzes the conversion of pyruvate to lactate under hypoxic or low-oxygen conditions, thereby supporting the energy production. We assessed LDH activity in tissue samples from the injury epicenter at various time points post-SCI. The results revealed a significant increase in LDH activity on 7 dpi compared to that in the sham group, whereas no significant difference was found between 7 and 14 dpi (Fig. 1E). These findings agree with previous results and suggest that excessive lactate accumulation is closely associated with high LDH activity. To assess whether SCI alters systemic lactate metabolism, we collected peripheral blood at multiple post-injury time points and measured serum l-lactate and LDH activity. LDH activity increased over time, peaked at 7 dpi, and remained comparably elevated at 14 dpi (Fig. 1F). This finding supports the hypothesis that the observed increase in lactate levels may result from an increase in systemic glycolytic activity.

We confirmed these findings in clinical samples by analyzing serum from SCI patients. Consistent with our targeted metabolomic analysis, serum lactate levels were significantly higher in SCI patients than in healthy controls (Fig. 1G). In vitro, we extracted bone marrow-derived macrophages (BMDMs) from mice and exposed them to various durations of hypoxia to mimic the ischemic and hypoxic microenvironments post-SCI. Compared to normoxic conditions, BMDMs showed a time-dependent increase in lactate concentration with prolonged hypoxia (Fig. S1A–B).

These results demonstrate that both SCI mice and patients have high systemic lactate levels, which may be driven by an increase in the glycolytic activity.

3.2. Subacute-phase induction of histone lactylation in macrophages after SCI

To determine the temporal and spatial dynamics of the ability of macrophage lactylation to drive reparative gene activation after SCI, we tracked lactylation changes over time in circulating monocytes and lesion-infiltrating macrophages in the SCI microenvironment during the acute phase. We first isolated primary BMDMs from wild-type mice. To mimic the ischemic-hypoxic microenvironment present after SCI, BMDMs were cultured under hypoxic conditions for various durations, and corresponding protein extracts were collected. The results of Western blot (WB) analysis demonstrated that histone lactylation in BMDMs increased markedly with prolonged hypoxia (Fig. 2A and B). We also treated BMDMs with increasing concentrations of lactate to replicate the acidified environment following SCI and identify the optimal in vitro conditions for inducing histone lactylation. The results of the WB analysis revealed a dose-dependent increase in lactylation, with 20 mM lactate producing a significantly greater effect than lower concentrations. Therefore, subsequent experiments were conducted using 20 mM lactate-treated BMDMs as an in vitro model of macrophage histone lactylation (Fig. 2C and D). These findings indicate that the metabolic alterations simulated by hypoxia and lactate treatment recapitulate the post-SCI microenvironment and strongly induce lactylation in BMDMs.

Fig. 2.

Subacute-phase Induction of Histone Lactylation in Macrophages After SCI. (A) Representative WB of Pankla and histone H3 in BMDMs treated with increasing concentrations of l-lactate. (B) Quantification of Pankla band intensity normalized to histone H3 in (A) (n = 6, one-way ANOVA. ns, not significant; ∗∗∗∗p < 0.0001.) (C) Representative WB of Pankla and histone H3 in BMDMs cultured under hypoxic conditions for the indicated durations. (D) Quantification of Pankla band intensity normalized to histone H3 in (C) (n = 6, one-way ANOVA. ns, not significant; ∗∗∗∗p < 0.0001.) (E) Representative immunofluorescence of spinal cord sections from sham and at 3, 7, 14, and 28dpi mice showing F4/80 (green), Pankla (red) and DAPI (blue). Scale bar, 200 μm; dashed outline denotes lesion epicenter. (F) Quantification of F4/80 and Pankla double-positive cells as a percentage of F4/80-positive macrophages in (E) (n = 3; one-way ANOVA. ∗p < 0.05; ∗∗∗∗p < 0.0001.) (G) Comparison of the Pankla-positive area normalized to DAPI across lesion core and uninjured margin at 14dpi in (E) (n = 3, unpaired t-test. ∗∗p < 0.01.) (H) Histogram of Pankla fluorescence intensity (gray value) at different locations in the 14dpi in (E). (I) Confocal images at 14 dpi mice showing colocalization of F4/80 and Pankla; main panels scale bar, 5 μm; insets scale bar, 3 μm. White boxes indicate inset regions. (J) Relative fluorescence intensity profiles for Pankla and DAPI along identical pixel coordinates in (I). (K) Representative WB of Pankla, histone H3 and β-actin in spinal cord lysates from sham and 14 dpi mice. (L) Quantification of target protein band intensities normalized to histone H3 in (K) (n = 8, unpaired t-test. ∗∗p < 0.01.).

To determine whether macrophage lactylation occurs after SCI in vivo, we collected spinal cord tissue from mice at different dpi. Immunofluorescence analysis revealed that macrophage lactylation first became detectable at 3 dpi. By 7 dpi, the extent of lactylation was greater, although the relative number of lactylated macrophages did not differ significantly from that observed at 3 dpi (Fig. 2E). At 14 dpi, lactylation was predominantly localized to macrophages, whereas by 28 dpi, the number of lactylated macrophages decreased (Fig. 2E). Quantitative analysis of the proportion of macrophages subjected to global pan-lactylation (Pankla) staining confirmed these findings, revealing a marked increase in lactylation at 14 dpi (Fig. 2F). Quantitative comparison of the Pankla‐positive area versus the DAPI‐positive area at the epicenter of the injury and adjacent margins at 14 dpi revealed that lactylation was concentrated in the lesion core and colocalized with the distribution of macrophages (Fig. 2G). A grayscale histogram of Pankla fluorescence intensity across the entire spinal cord also revealed significant upregulation of lactylation within the lesion at 14 dpi (Fig. 2H). To examine the subcellular localization of lactylation more precisely, confocal imaging was performed on spinal cord sections at 14 dpi. Fluorescence colocalization analysis showed that lactylation signals were primarily nuclear and closely associated with chromatin signal in macrophages (Fig. 2I). Further pixel‐by‐pixel analysis of Pankla and DAPI fluorescence intensity confirmed a high degree of colocalization, which indicated that lactylation of histones in macrophages occurred after SCI (Fig. 2J). Finally, to determine whether circulating monocyte‐derived macrophages also underwent lactylation, we isolated peripheral blood mononuclear cell proteins from both sham and 14 dpi mice and performed quantitative analysis of Pankla. Compared to those from the sham controls, mononuclear cells from 14 dpi mice also exhibited an increase in lactylation (Fig. 2K and L).

These results indicate that during the subacute phase of SCI (14 dpi), there is robust histone lactylation in resident macrophages within the lesion and in peripheral monocyte‐derived macrophages.

3.3. Significant elevation of H3K9la in macrophages and circulating monocytes 14 days after SCI

To identify the predominant histone lactylation site in macrophages following SCI, we first selected three lysine residues on histone H3 that are known to participate in macrophage lactylation: H3K9, H3K14, and H3K18. Compared to those of the sham controls, all three H3 lactylation marks were elevated in the injured spinal cord at 14 dpi (Fig. 3A and B), but H3K9la exhibited the most pronounced increase. After each band intensity was normalized to that of the loading control, only H3K9la was significantly upregulated, whereas H3K14la and H3K18la presented more modest, nonsignificant changes (Fig. 3C). Lactylation-focused proteomic analysis further independently identified H3K9 as the lactylation site, providing antibody-independent validation of H3K9la in our system (Fig. 3D). To further confirm the specificity of H3K9la detection, peptide competition assays were performed. Pre-incubation of the H3K9la antibody with an H3K9la-modified peptide markedly abolished the detected signal, whereas pre-incubation with H3K9ac-modified or unmodified H3 peptides had minimal effects. Conversely, an H3K9ac antibody was selectively competed by the H3K9ac peptide but not by the H3K9la peptide, confirming modification-specific antibody recognition (Fig. S1C–D). These results indicate that H3K9 is the principal site of histone lactylation in the injured spinal cord after SCI.

Fig. 3.

Significant Elevation of H3K9la in Macrophages and Circulating Monocytes 14 Days After SCI. (A) Representative WB of H3K9la, H3K14la, H3K18la and total H3 in spinal cord lysates from uninjured mice and at 14 days post‐injury (14 dpi). (B) Quantification of H3K9la, H3K14la and H3K18la band intensities normalized to total H3 in (A) (n = 6, unpaired t-test. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗∗p < 0.0001). (C) Quantification of target protein band intensities normalized to histone H3 in (A) (n = 6, one‐way ANOVA. ∗∗∗∗p < 0.0001). (D) Representative tandem mass spectra of histone H3-derived peptides containing lysine lactylation detected in control and lactate-treated samples. (E) Representative immunofluorescence of spinal cord sections from sham or at 7 and 14 dpi mice, stained for F4/80 (green), H3K9la (red) and DAPI (blue). Scale bar = 100 μm; dashed outline marks the lesion core; white boxes indicate regions shown at higher magnification. (F) High‐magnification view of the 14 dpi lesion core showing colocalization of F4/80 and H3K9la. Scale bar = 10 μm. (G) Fluorescence intensity profiles for H3K9la and F4/80 along the line indicated in (E) within the lesion core. (H) Quantification of F4/80⁺/H3K9la⁺ double‐positive cells as a percentage of total F4/80⁺ macrophages at 7 and 14 dpi in (E) (n = 3; unpaired t-test. ∗∗∗∗p < 0.0001). (I) Representative WB of H3K9la and total histone H3 in peripheral blood samples from sham and SCI mice at 14 dpi. (J) Quantification of H3K9la band intensity normalized to histone H3 in (I) (n = 8; unpaired t-test. ∗∗∗∗p < 0.0001). (K) Representative WB of H3K9la and total histone H3 in BMDMs treated with increasing concentrations of l-lactate. (L) Quantification of H3K9la band intensity normalized to histone H3 in (K) (n = 6, one-way ANOVA. ns, not significant; ∗∗p < 0.01; ∗∗∗∗p < 0.0001).

Immunofluorescence imaging was subsequently performed on spinal cord sections from sham‐operated mice and from mice at 7 and 14 dpi to determine the cellular localization of H3K9la (Fig. 3E). At 14 dpi, higher‐magnification images revealed that H3K9la signals were predominantly present within macrophages and extensively colocalized with nuclear, highly consistent with those of prior Pankla and F4/80 co-staining (Fig. 3F). Fluorescence colocalization analysis demonstrated a significant increase in H3K9la expression in macrophages at 14 dpi compared to that at 7 dpi; pixel‐by‐pixel analysis of the relative fluorescence intensity for H3K9la and F4/80 in the 14 dpi images confirmed a high degree of overlap. In addition, Immunofluorescence staining revealed relatively low H3K9la signals in Iba-1–positive microglia compared with F4/80-positive macrophages in the injured site, indicating preferential enrichment of H3K9la in infiltrating macrophages after SCI (Fig. S1E–G).These findings confirmed that H3K9la occurs primarily in macrophages within the lesion area at 14 dpi (Fig. 3G and H).

Previous studies have shown that macrophages within the spinal lesion are predominantly derived from the recruitment and local differentiation of circulating monocytes by 14 dpi [23]. Moreover, we have revealed a temporal increase in peripheral blood lactate levels following SCI. To determine whether monocyte H3K9la is similarly activated, we isolated peripheral blood mononuclear cells (PBMCs) from 14 dpi and sham‐operated mice. Compared to that in control PBMCs, H3K9la expression in PBMCs was significantly elevated at 14 dpi (Fig. 3I and J), indicating that circulating monocytes also undergo H3K9la during the subacute phase of SCI.

Next, we prepared primary BMDMs and cultured them with a lactate gradient treatment to assess H3K9la levels. BMDMs showed dose‐dependent increase in H3K9la expression level after treatment (Fig. 3K and L). To further exclude a pH‐dependent effect, BMDMs were treated with graded concentrations of sodium lactate, which likewise produced a progressive upregulation of H3K9la (Fig. S1H–I). Under 5-10 mM sodium lactate, H3K9la levels remained unchanged, whereas 20 mM sodium lactate elicited a significant increase. These findings suggest that extracellular lactate accumulation, rather than acidification, primarily drives the increase in this post-translational modification. To more closely simulate the hypoxic environment of the injured spinal cord, BMDMs were cultured under hypoxia for various durations, and H3K9la levels were systematically evaluated. No noticeable increase in H3K9la was observed after 6-12 h of hypoxia, potentially reflecting an initial cellular compensatory response. However, after 24 h of hypoxia, H3K9la was significantly elevated (Fig. S1J–K), confirming that prolonged oxygen deprivation can also potentiate histone lactylation in macrophages.

These findings indicate that during the subacute phase of SCI, the high levels of lactate in the peripheral blood induce remote activation of H3K9la in circulating monocytes. After monocytes migrate to and differentiate at the site of the injury, sustained metabolic dysregulation of the local microenvironment further perpetuates H3K9la in macrophages.

3.4. Genome‐wide analysis of H3K9la‐mediated transcriptional changes after SCI

Histone modification alterations can affect both the transcriptional activation and inhibition of target genes [24]. To screen candidate target genes of histone lactylation in SCI, we isolated circulating monocyte‐derived macrophages from sham‐operated and 14 dpi SCI mice and performed CUT&Tag using an anti‐H3K9la antibody (Fig. S2A). The results showed that H3K9la was markedly enriched at gene promoters and upstream regions following SCI (Fig. 4A). In circulating monocyte‐derived macrophages from SCI mice, 18,797 differential H3K9la peaks were identified, 8.37% of which were localized to promoter regions (±3 kb). In contrast, only 4740 differential peaks were detected in the sham controls, with 3.98% residing in the promoter region (Fig. 4B). These findings indicate that H3K9la modification is heavily concentrated at promoter sites after SCI, thereby activating the transcription and translation of associated genes. Gene Ontology enrichment analysis revealed that genes upregulated simultaneously with increased H3K9la occupancy after SCI were predominantly involved in pathways regulating cell migration, adhesion, inflammation, and oxidative stress (Fig. 4C). The heatmap revealed that H3K9la‐marked upregulated genes included key inflammatory and immune response mediators such as TXNIP, IL‐1β, and CD14 (Fig. 4D). GSEA revealed that these differentially modified genes were enriched predominantly in the SCI-related and pro-inflammatory pathways (Fig. 4E).

Fig. 4.

Genome‐Wide Analysis of H3K9la‐Mediated Transcriptional Changes After SCI. (A) Heatmap of H3K9la binding peaks in peripheral blood monocyte–derived macrophages from sham and 14 dpi mice. Color intensity corresponds to normalized read counts; hierarchical clustering groups genes with similar binding profiles to illustrate global lactylation occupancy. (B) Genome-wide distribution of significantly upregulated H3K9la peaks in peripheral blood monocyte–derived macrophages at 14dpi mice compared with sham mice. (C) Bubble plot of the top 10 Gene Ontology terms enriched among genes associated with upregulated H3K9la peaks. Bubble size reflects gene count and color intensity indicates enrichment significance. (D) Heatmap of representative inflammation-related genes identified from CUT&Tag analysis, showing relative H3K9la enrichment. (E) Gene Set Enrichment Analysis of H3K9la-bound genes, highlighting pathways significantly enriched in the 14dpi macrophage population. (F) Venn diagram showing the overlap between differentially bound genes identified by CUT&Tag and differentially expressed genes from RNA sequencing. (G) Diagonal volcano plot of RNA-seq results, highlighting the 28 genes that overlap with the CUT&Tag differential binding set. (H) Genome browser tracks of representative target loci, showing CUT&Tag signal intensity in sham (red) and 14 dpi (blue) macrophages. (I) Quantitative RT-PCR analysis of indicated gene expression levels (n = 3, unpaired t-test. ∗∗p < 0.01; ∗∗∗p < 0.001). (J) Chip-qPCR quantification of H3K9la enrichment at target promoter regions (n = 3; unpaired t-test. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001). (K) Dual-luciferase reporter assay quantification of TXNIP promoter activity (n = 3; one-way ANOVA. ns, not significant; ∗∗∗∗p < 0.0001).

Additionally, BMDMs were treated with 20 mM lactate or PBS and subjected to mRNA sequencing, which identified 591 DEGs. These genes intersected with the 988 differential peaks from CUT&Tag sequencing; the resulting Venn diagram revealed 28 genes commonly regulated by both assays (Fig. 4F). We then generated a diagonal volcano plot of the RNA‐seq data from lactate‐treated BMDMs, highlighting those genes overlapping with CUT&Tag. Cross‐analysis revealed that thioredoxin‐interacting protein (TXNIP) and Hmox1 were the genes whose expression was most significantly upregulated and downregulated, respectively (Fig. 4G). Finally, IGV‐based peak analysis revealed that H3K9la at the TXNIP promoter was significantly elevated after SCI, whereas H3K9la at the Hmox1 promoter was markedly reduced compared to that in the sham controls (Fig. 4H). To validate the circulating monocyte CUT&Tag and RNA‐seq findings at 14 dpi, we performed RT‐qPCR and ChIP‐qPCR to confirm H3K9la enrichment at the promoters of the most significantly upregulated genes (TXNIP and Notch1) and the downregulated gene (Hmox1). RT‐qPCR analysis revealed that in lactate‐treated BMDMs, TXNIP expression was significantly increased, Notch1 expression was modestly elevated, and Hmox1 expression was markedly reduced (Fig. 4I). Consistent with these results, the findings of ChIP‐qPCR assays revealed significant enrichment of H3K9la at the TXNIP promoter in lactate‐treated BMDMs, whereas H3K9la occupancy at the Hmox1 promoter decreased considerably (Fig. 4J). To determine whether H3K9la directly enhances TXNIP transcriptional activity, a dual‐luciferase reporter assay was performed. The TXNIP promoter region containing the H3K9la‐enriched site was cloned upstream of a luciferase reporter and co‐transfected with an H3K9Q construct to mimic lactylation. H3K9Q expression significantly increased TXNIP promoter–driven luciferase activity compared with control constructs, providing functional evidence supporting H3K9la-mediated regulation of TXNIP transcription. (Fig. 4K).

To summarize, we identified H3K9la target genes in monocyte-derived macrophages after SCI and showed that H3K9la is associated with selective activation of inflammatory gene programs, among which TXNIP represents a key downstream target.

3.5. H3K9la-mediated regulation of macrophage TXNIP transcription and activation of the NLRP3/caspase-1 signaling pathway

TXNIP binds to thioredoxin (Trx) and inhibits its reductase activity, thereby playing a key role in oxidative stress signaling and the pathogenesis of multiple diseases [25]. Recent studies have revealed a central regulatory function for TXNIP in metabolic disorders: as a key modulator of lipid and glucose metabolism, TXNIP participates in diabetic complications via the ROS-TXNIP-NLRP3 axis, and inhibition of this pathway substantially ameliorates metabolic dysregulation [26]. At the protein level, TXNIP has been shown to directly interact with NLRP3, facilitating inflammasome assembly and activation under conditions of oxidative or metabolic stress [27]. Moreover, in CNS diseases, TXNIP plays a role in redox imbalance and inflammasome activation, and knocking out the TXNIP gene can restore redox homeostasis in the brain and reduce ischemia‐induced neuronal injury [28].

We further determined the spatiotemporal dynamics of TXNIP expression following SCI. In the sham group, neither F4/80 nor TXNIP signals were detected. TXNIP expression in macrophages was significantly higher at 14 dpi than at 7 dpi, consistent with the changes in H3K9la expression in macrophages observed in our previous findings (Fig. 5A and C). Furthermore, higher‐magnification images of the 14 dpi samples revealed colocalization of TXNIP with F4/80, confirming that TXNIP was abundant in macrophages (Fig. 5B). Spatial transcriptomics dataset revealed that TXNIP expression was elevated at 14 dpi and predominantly localized to the lesion core (Fig. 5D). Additionally, single‐cell RNA‐seq analysis of a mouse SCI model revealed that TXNIP expression in macrophages was almost undetectable at 0.5 dpi, started increasing at 1 dpi, but did not increase significantly at 3 or 7 dpi. Substantial upregulation of TXNIP was observed only at 14 dpi, and this increase in TXNIP expression persisted at 28 dpi and 35 dpi (Fig. 5E).

Fig. 5.

Molecular Mechanism of Macrophage H3K9la-Mediated Activation of TXNIP Transcription. (A) Representative immunofluorescence images of spinal cord sections from sham mice and SCI mice at 7 and 14dpi, stained for F4/80 (green), TXNIP (red) and DAPI (blue). Scale bar, 100 μm. White boxes denote regions shown at higher magnification. (B) Higher magnification of the 14dpi section in (A), Scale bar, 10 μm. (C) Line-scan analysis of relative fluorescence intensity for TXNIP and F4/80 along the same pixel coordinates indicated in (A). (D) Spatial transcriptomic feature plot illustrating TXNIP expression in spinal cord sections from sham and 14 dpi mice. The color bar denotes relative expression levels. (E) Temporal expression trajectory of TXNIP in spinal cord macrophages, as inferred from the single-cell RNA-seq dataset GSE162610. (F) Representative immunofluorescence images of spinal cord sections from sham mice or from mice at 7 and 14dpi, stained for F4/80 (green), NLRP3 (red) and DAPI (blue). Scale bar, 100 μm; white boxes indicate regions shown at higher magnification. (G) Line-scan analysis of relative fluorescence intensity for NLRP3 and F4/80 along identical pixel coordinates indicated in (F). (H) WB analysis of Co-IP samples, demonstrating the presence of TXNIP and its interaction partner NLRP3. (I) Representative WB of NLRP3, pro‐Caspase-1, cleaved Caspase-1, TXNIP, and β-actin. (J) Quantification of NLRP3, cleaved Caspase-1 (normalized to total Caspase-1), and TXNIP band intensities relative to β-actin in (I) (n = 3; one-way ANOVA). (K) Representative WB of TNF-α, iNOS, IL-1β, and β-actin in BMDMs treated with lactate and/or 2-DG. (L) Quantification of TNF-α, iNOS, and IL-1β band intensities normalized to β-actin in (K) (n = 3; one-way ANOVA. ns, not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.) (M) Intracellular ROS levels in BMDMs assessed by DCFH-DA staining and flow cytometry. (N) Oxygen consumption rate (OCR) profiles measuring mitochondrial respiration in BMDMs (n = 5; mean ± SD). O, oligomycin; F, FCCP; R/A, rotenone/antimycin A. (O) Quantification of OCR parameters from panel N, including basal respiration, ATP‐linked respiration, maximal respiration, and spare respiratory capacity (n = 5; one-way ANOVA. ns, not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001).

Macrophages can secrete TXNIP via both autocrine and paracrine mechanisms, thereby activating NLRP3 and caspase‐1 to promote the maturation and release of IL-1β, which exacerbates inflammation [29]. To determine whether TXNIP activates NLRP3 in lesion‐infiltrating macrophages at 14 dpi, we performed immunofluorescence staining on spinal cord sections from sham‐operated, 7 dpi, and 14 dpi. In the sham controls, neither F4/80 nor NLRP3 signals were detected, which indicated the absence of macrophage infiltration and inflammasome activation. In contrast, at both 7 dpi and 14 dpi, many macrophages accumulated in the lesion. NLRP3 expression in macrophages was significantly higher at 14 dpi than at 7 dpi (Fig. 5F and G).

Lactate treatment significantly increased H3K9la fluorescence intensity compared with PBS, whereas, 2-Deoxy-d-glucose (2-DG), a glucose analog that inhibits glycolysis, alone had no effect. Co-treatment with lactate and 2-DG markedly reduced lactate-induced H3K9la signals (Fig. S3A–B). These findings were confirmed by WB, which showed increased H3K9la protein levels after lactate treatment and suppression upon lactate plus 2-DG co-treatment (Fig. 5I). TXNIP expression was then examined by immunofluorescence. Lactate markedly increased TXNIP fluorescence compared with PBS, while 2-DG alone had no significant effect; co-treatment with lactate and 2-DG reduced TXNIP levels relative to lactate alone (Fig. S4A–B). Similarly, lactate increased NLRP3 fluorescence intensity, which was reduced by lactate plus 2-DG co-treatment, whereas 2-DG alone had no significant effect (Fig. S5A–B). Co-immunoprecipitation confirmed a direct interaction between TXNIP and NLRP3 in lactate-treated BMDMs (Fig. 5H). WB analysis further showed that lactate increased TXNIP and caspase-1 protein levels, both of which were suppressed by lactate plus 2-DG treatment; NLRP3 protein levels were also reduced by co-treatment compared with lactate alone (Fig. 5I and J). To provide locus-specific validation, promoter-targeted CRISPR interference (CRISPRi) using dCas9-KRAB was applied to repress the H3K9la-enriched Txnip promoter region identified by CUT&Tag. This intervention significantly reduced TXNIP and NLRP3 protein expression, accompanied by decreased levels of inflammation-related proteins assessed by WB (Fig. S5C–E).

3.6. H3K9la-mediated transcriptional regulation of TXNIP triggers inflammatory responses in BMDMs

The TXNIP/NLRP3 signaling axis plays a key role in initiating inflammatory responses [30]. Accordingly, we measured the levels of pro-inflammatory markers in BMDMs treated with lactate and/or 2-DG. After 24 h of exposure to lactate, the TNF-α and iNOS levels were markedly increased, whereas the IL-1β level remained unchanged, mirroring the pattern observed for NLRP3. Treatment with 2-DG alone significantly decreased the levels of all the measured inflammatory markers. In cells cotreated with lactate and 2-DG, TNF-α and IL-1β levels were modestly elevated compared to those in cells treated with 2-DG alone, but they remained below the levels observed in cells treated with lactate alone, while iNOS expression was significantly suppressed (Fig. 5K and L). To further define the position of TXNIP within this signaling cascade, siRNA-mediated knockdown of TXNIP was performed, which markedly attenuated NLRP3 inflammasome activation and downstream effector protein expression, supporting TXNIP as a key downstream mediator linking H3K9la to NLRP3 activation (Fig. S6A–B).

Excess TXNIP not only sequesters Trx and impairs its antioxidant function, thereby increasing the level of intracellular reactive oxygen species (ROS), but also promotes mitochondrial ROS production, leading to mitochondrial DNA damage and inflammation [31,32]. To assess the effect of TXNIP on mitochondrial function, we measured intracellular ROS levels in BMDMs treated with lactate and/or 2‐DG by flow cytometry. Lactate treatment increased the ROS fluorescence intensity relative to that of the PBS controls, whereas combined lactate + 2‐DG treatment attenuated ROS accumulation (Fig. 5M). We further evaluated mitochondrial respiratory function by measuring the oxygen consumption rate (OCR). Lactate‐treated macrophages presented modest increases in basal respiration, maximal respiration, and ATP production, which is consistent with lactate serving as both a metabolic end‐product and a substrate for mitochondrial respiration. However, spare respiratory capacity decreased after exposure to lactate, which indicated that the ability to meet additional energy demands decreased despite elevated baseline respiration. In contrast, 2‐DG treatment improved spare respiratory capacity (Fig. 5N and O). To determine the contribution of mitochondrial oxidative stress, cells were treated with the mitochondria-targeted antioxidant Mito-TEMPO. Mito-TEMPO markedly reduced intracellular ROS levels and suppressed NLRP3 and downstream inflammasome-related protein expression, while TXNIP protein levels remained largely unchanged (Fig. S6C–E).

To summarize, H3K9la in macrophages is functionally linked to inflammatory responses and mitochondrial dysfunction, with TXNIP and mitochondrial ROS acting as critical intermediates connecting H3K9la to NLRP3 signaling.

3.7. Targeted peptide inhibitors suppress H3K9la to attenuate inflammatory responses in BMDMs

Macrophage H3K9la drives both inflammatory activation and mitochondrial dysfunction, highlighting that a strategy that specifically targets H3K9la in lesion‐infiltrating macrophages needs to be developed [33]. Synthetic inhibitory peptides based on PTM motifs represent a promising approach. By analyzing the sequence surrounding histone H3 lysine 9, we designed four candidate peptides, each comprising a CPP linked to an oxygen‐sensitive nitroaromatic self‐immolative linker (Fig. 6A). These conjugates remain stable under normoxic conditions, but within the hypoxic lesion core, intracellular nitroreductases convert the nitro group to an amino group, triggering linker cleavage and release of the active inhibitory peptide. All peptides were purified by high‐performance liquid chromatography (Fig. S7A) and shown by high‐resolution mass spectrometry to exceed 95% purity (Fig. S7B). Functional screening of lactate‐pretreated BMDMs revealed that hypoxia‐responsive H3K9la‐peptide‐3# was the most potent inhibitor of H3K9la, as determined by WB analysis (Fig. 6B and C). Therefore, we selected this construct, denoted as H3K9la-pe, as our tool for targeted inhibition of H3K9 la in subsequent experiments. In addition, we examined the effects of H3K9la-pe on other histone modifications at related lysine residues. WB analyses showed that H3K9la-pe selectively reduced H3K9la, with minimal effects on H3K9 acetylation, H3K9 methylation, H3K27 acetylation, and Pankla levels (Fig. S7C–D), indicating limited perturbation of other histone modifications.

Fig. 6.

Targeted Peptide Inhibitors Suppress H3K9la to Attenuate Inflammatory Responses in BMDMs. (A) Schematic representation of peptide design. Green denotes the cell‐penetrating peptide (CPP), blue indicates the hypoxia‐sensitive linker (2-nitroimidazole derivative, Nitro-Im), and red highlights the H3K9la recognition motif. (B) Representative WB of nuclear H3K9la and total histone H3 in BMDMs co‐treated with lactate and each of the four candidate peptide inhibitors. (C) Quantification of H3K9la band intensities normalized to histone H3 from panel B (n = 3, one‐way ANOVA. ns, not significant; ∗p < 0.05; ∗∗∗∗p < 0.0001). (D) Volcano plot of RNA-seq data comparing H3K9la-pe–treated and control groups. (E) Gene set variation analysis (GSVA) plot depicting pathway activity scores following H3K9la-pe treatment. (F) Heatmap showing expression patterns of inflammation-associated genes across experimental groups. (G) Representative WB of nuclear H3K9la and total histone H3 in BMDMs treated for 24 h with lactate, 2-DG and H3K9la-pe. (H) Quantification of H3K9la band intensities normalized to histone H3 in (G) (n = 3, one-way ANOVA. ns, not significant, ∗p < 0.05; ∗∗∗∗p < 0.0001.) (I) Representative TEM images showing mitochondrial morphology in BMDMs under different treatment conditions. The enlarged view (right) corresponds to the dashed box in the left image; red arrows indicate mitochondria. (J) Intracellular ROS levels were assessed using DCFH-DA staining followed by flow cytometry analysis. (K) OCR measurements evaluating mitochondrial respiratory function in BMDMs (n = 5, mean ± SD; OCR: oxygen consumption rate; O: oligomycin, F: FCCP, R/A: rotenone and antimycin A). (L) Quantification of mitochondrial respiration parameters shown in (J), including basal respiration, ATP production, maximal respiration and spare respiratory capacity (n = 5, one-way ANOVA, Tukey's post hoc test. ns: not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001).

To assess potential off-target transcriptional effects, RNA-seq analysis following H3K9la-pe treatment showed a significant reduction in Txnip expression, whereas other inflammation-related genes displayed modest downward trends without statistical significance (Fig. 6D). Gene set variation analysis indicated that E2F and p53 signaling pathways were not significantly activated after peptide treatment (Fig. 6E). Consistently, heatmap analysis revealed marked suppression of inflammation-associated gene signatures, without broad activation of unrelated transcriptional programs, supporting limited off-target transcriptional effects (Fig. 6F).

To validate the in vitro efficacy of H3K9la-pe on macrophage lactylation, BMDMs were first treated with lactate and then cotreated for 24 h with either the glycolysis inhibitor 2-DG or H3K9la-pe. WB analysis revealed that H3K9la-pe inhibited H3K9la more potently than 2-DG. We next evaluated the effect of H3K9la-pe on the TXNIP-NLRP3 signaling axis: lactate markedly upregulated TXNIP, NLRP3, and caspase-1, whereas H3K9la-pe treatment suppressed their expression more effectively than 2-DG (Fig. 6G and H). To assess downstream inflammatory responses, we measured pro-inflammatory marker levels via WB analysis. Co-treatment with lactate and 2-DG reduced iNOS and TNF-α but had a negligible effect on IL-1β; in contrast, co-treatment with lactate and H3K9la-pe considerably decreased the levels of all three markers (Fig. S7E–F).

Transmission electron microscopy examinations revealed that lactate treatment induced widespread mitochondrial vacuolization and even membrane rupture in BMDMs, while 2-DG partially ameliorated vacuolization. Some cells still exhibited severe mitochondrial damage, whereas H3K9la-pe treatment greatly restored mitochondrial integrity (Fig. 6I). Flow cytometry analysis of intracellular ROS revealed that lactate greatly increased ROS fluorescence, which was significantly reduced by both 2-DG and H3K9la-pe, with H3K9la-pe resulting in a greater decrease than 2-DG alone (Fig. 6J). Assessment of mitochondrial function by the OCR demonstrated that lactate modestly increased basal respiration, maximal respiration, and ATP production. H3K9la-pe not only mitigated lactate-induced hyperactivation of mitochondrial respiration but also significantly increased spare respiratory capacity, indicating that it had a restorative effect on mitochondrial function (Fig. 6K and L). In the in vitro BMDM H3K9la model, H3K9la-pe more effectively ameliorated inflammation and restored mitochondrial function than 2-DG.

3.8. H3K9la-pe inhibits TXNIP downstream gene expression and promotes axon regeneration after SCI

To evaluate the in vivo distribution of the hypoxia-responsive H3K9la peptide, lyophilized H3K9la-pe was reconstituted in PBS containing the near-infrared dye DiR and administered intraperitoneally on day 1 post-SCI. Live fluorescence imaging was performed at 12, 24, and 48 h post-injection, followed by ex vivo imaging of the spinal cord lesion and major organs at 48 h (Fig. 7A and B). No signal was detected in PBS-treated controls. DiR alone showed transient whole-body distribution with peak signal at 24 h and marked decline by 48 h, whereas DiR-H3K9la-pe exhibited higher fluorescence intensity at 24 h and sustained signal at the lesion site (Fig. 7C). Ex vivo analysis confirmed greater accumulation of DiR-H3K9la-pe in the injured spinal cord compared with DiR alone (Fig. 7D). Lower signals were detected in peripheral organs including the spleen, stomach, and ileum (Fig. 7E). To assess peptide uptake, H3K9la-pe was conjugated with Rhodamine B. H3K9la-pe fluorescence predominantly co-localized with macrophage markers at the lesion site, with minimal overlap observed in neurons, astrocytes, or microglia (Fig. S8A–B), and in vitro uptake assays further confirmed preferential internalization by macrophages (Fig. S8C). Although ex vivo imaging revealed relatively strong fluorescence in the liver, in vivo signals from deep visceral organs were substantially attenuated by tissue absorption and scattering, resulting in limited detectability, representing a known technical limitation of optical imaging [34]. Based on the observed peak tissue signal at 24 h, daily intraperitoneal administration of H3K9la-pe or PBS was applied for 14 days to maintain exposure during the subacute phase after SCI. To assess systemic safety, H&E staining was performed on major organs, including the heart, liver, spleen, lung, and kidney, after administration of the experimental dose of H3K9la-pe (Fig. S8D). In parallel, CCK-8 assays were conducted in vitro to evaluate the effect of H3K9la-pe on primary BMDMs, showing no significant differences compared with control groups (Fig. S8E). Histological examination further revealed no detectable pathological alterations in major organs, indicating that H3K9la-pe did not produce overt toxicity under the dosing conditions used in this study in either in vivo or in vitro settings.

Fig. 7.

H3K9la-pe Inhibits TXNIP Downstream Gene Expression and Promotes Axon Regeneration After SCI. (A) Representative in vivo imaging of SCI mice at 12, 24, and 48 h following intraperitoneal injection of H3K9la-pe, demonstrating targeted accumulation at the lesion site. Postmortem images of dissected organs were also captured, showing fluorescence intensity in mice injected with saline, DiR, or DiR-H3K9la-pe. (B) Schematic diagram illustrating the anatomical locations of major dissected organs. (C) Quantitative analysis of fluorescence signal intensity at the injury site at 12, 24, and 48 h post-injection in the saline, DiR, and DiR-H3K9la-pe groups (n = 3, one-way ANOVA). (D) Quantification of fluorescence efficiency (log-transformed) from dissected organs shown in (A) (n = 3, mean ± SD). (E) Quantitative analysis of fluorescence intensity specifically in spinal cord tissue in (A) (n = 3, mean ± SD, one-way ANOVA, ns: not significant, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.) (F) Representative WB images showing the expression levels of nuclear H3K9la, NLRP3, Caspase 1, Cleaved Caspase 1, TXNIP, β-actin, and H3 in the Sham, Control, and H3K9la-pe treatment groups. (G) Quantitative analysis of the relative protein levels in (F), including H3K9la normalized to H3, NLRP3 and TXNIP normalized to β-actin, and Cleaved Caspase 1 normalized to total Caspase 1 (n = 3, one-way ANOVA. ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). (H) Quantitative analysis of the percentage of BDA- and NF-positive areas within the lesion site shown in (I-J) (n = 6, one-way ANOVA. ∗∗p < 0.01; ∗∗∗∗p < 0.0001.) (I) Representative immunofluorescence images of corticospinal tract tracing at 28dpi in Sham, Control, and H3K9la-pe-treated SCI mice (BDA: white, DAPI: blue; scale bar = 200 μm). (J) Representative immunofluorescence images of NF staining at day 28 post-injury in Sham, Control, and H3K9la-pe-treated SCI mice, including magnified views of the lesion area (NF: white, DAPI: blue; scale bar = 200 μm, magnified view = 50 μm).

Immunofluorescence staining of spinal cord sections (Fig. S9A–B) revealed that compared to PBS, H3K9la-pe treatment significantly reduced the proportion of macrophages positive for H3K9la and the overall fluorescence intensity. To assess its effects on the TXNIP-NLRP3 signaling axis, we performed immunofluorescence assays for TXNIP and NLRP3 (Fig. S9C–F). We found a marked decrease in the percentage of positive macrophages and signal intensity in the H3K9la-pe group versus the control group. WB analysis of spinal cord lysates from the sham, PBS-treated, and H3K9la-pe-treated groups confirmed that H3K9la-pe strongly suppressed the expression of H3K9la, TXNIP, NLRP3, and caspase-1 (Fig. 7F and G). H3K9la-pe also significantly decreased the levels of the pro-inflammatory markers TNF-α, iNOS, and IL-1β compared to their corresponding levels in vehicle-treated SCI mice (Fig. S10A–B).

We further examined the effect of H3K9la-pe treatment on axon regeneration in SCI mice. Corticospinal tract (CST) tracing via motor cortex injection demonstrated that some CST axons could regenerate into the injured area in the treatment group (Fig. 7I). Immunofluorescence examinations of the axon markers neurofilament (NF), β3-tubulin, GAP-43 showed a substantial increase in fiber density and more coherent axon morphology at the injury center in H3K9la-pe-treated mice than in control mice (Fig. 7H and J; Fig. S11A–B). Additionally, at 14 dpi, AAV2/9-Syn-EGFP was injected into the T8 spinal segment, and the spinal cords were collected at 28 dpi for analysis. Syn-EGFP tracing revealed that in H3K9la-pe-treated mice, a subset of labeled axons penetrated the lesion site, with distinct fibers visible at the epicenter of the injury (Fig. S11C–D). High-magnification images revealed continuous, fiber-like axon projections in the lesion cores of treated animals, whereas control mice predominantly exhibited dystrophic retraction bulbs.

3.9. H3K9la-pe treatment promotes neurological functional recovery in SCI mice

Finally, to evaluate the effect of H3K9la-pe on functional recovery after SCI, we performed electrophysiological, behavioral, and histological assessments in sham-operated, vehicle-treated, and H3K9la-pe-treated mice. At 28 dpi, motor‐evoked potential (MEP) amplitudes in vehicle‐treated SCI mice were significantly lower than those in sham control mice, whereas H3K9la-pe treatment significantly restored hindlimb MEP amplitudes (Fig. 8A and B). Peak-to-peak duration was significantly prolonged after SCI and was partially reduced following H3K9la-pe treatment (Fig. S11E). Basso Mouse Scale (BMS) scores indicated that from day 5 onward, H3K9la-pe-treated mice exhibited progressive locomotor improvements, achieving scores of 4–6 between 14 and 28 dpi (Fig. 8C). In the Louisville Swim Scale (LSS) test, vehicle‐treated mice showed high trunk instability and increased body angles, whereas H3K9la-pe-treated mice showed high trunk stability, a decrease in forelimb reliance, and improvements in hindlimb propulsion, resulting in significantly greater LSS scores (Fig. 8D and E). At 28 dpi, H3K9la-pe-treated SCI mice showed substantial gains in hindlimb weight support and stepping ability, as confirmed by the results of the detailed kinematic analysis (Fig. 8F and G). Histological examination of neurogenic bladder function via H&E staining revealed that vehicle‐treated SCI mice exhibited greater muscle thinning than sham control mice, whereas H3K9la-pe treatment preserved the detrusor thickness, indicating that bladder function improved substantially (Fig. 8H and I). These findings together demonstrate that administering H3K9la-pe significantly enhances both motor and autonomic recovery after SCI.

Fig. 8.

H3K9la-pe Treatment Promotes Neurological Functional Recovery in SCI Mice. (A) Representative MEP waveforms recorded on day 28dpi in Sham, Control, and H3K9la-pe treatment groups. (B) Quantitative analysis of MEP amplitudes shown in (A) (n = 6, one-way ANOVA.) (C) BMS scores plotted over time from pre-injury to day 28dpi in Sham, Control, and treatment groups (n = 6, two-way ANOVA). (D) Representative images of trunk inclination during swimming tests on day 28 post-injury in Sham, Control, and treatment groups. (E) Quantitative analysis of LSS scores in (D) (n = 6, one-way ANOVA.) (F) Representative stick diagram decomposition of leg movements, oscillation of the whole limb (virtual limb linking the hip to the toe) on day 28dpi in Sham, Control, and H3K9la-pe treatment groups. (G) Quantitative analysis of maximum iliac crest height, stride length, and toe maximum height in (F) (n = 6, one-way ANOVA.) (H) Representative images showing detrusor muscle thickness of the bladder on day 28 post-injury in Sham, Control, and H3K9la-pe treatment groups (scale bar: 200 μm). (I) Quantitative analysis of detrusor muscle thickness in (H) (n = 6, one-way ANOVA. ns: not significant; ∗p < 0.05, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.).

4. Discussion

Repetitive ischemia and inflammation drive secondary injury after SCI, evolving from acute hemorrhage to subacute thrombosis and vasospasm, causing persistent hypoxia and metabolic disruption. This induces a Warburg-like shift—enhanced glycolysis and lactate accumulation [35]. Our previous research demonstrated that SCI induces a localized pseudo-Warburg effect due to reduced lactate clearance by spinal endothelial cells [36]. The current study extend these findings by demonstrating systemic lactate elevation in SCI patients, likely resulting from early systemic inflammation. Elevated lactate correlates with increased H3K9la in lesion macrophages and circulating monocytes, activating the TXNIP–NLRP3 inflammatory pathway, exacerbating inflammation, and impairing mitochondrial function. Targeted inhibition of H3K9la reduces inflammation, improves mitochondrial health, and enhances neurological recovery, underscoring its therapeutic potential for SCI [37].

Lactate acts not only as a metabolic byproduct but also as a signaling molecule that modulates immune responses. Lactate can exert both pro-inflammatory and anti-inflammatory effects. For example, lactate signaling inhibits the lipopolysaccharide (LPS)-induced production of numerous cytokines and chemokines [38], and in macrophages and monocytes, it decreases glycolytic flux and downregulates several receptor-mediated signaling pathways [39]. In contrast, lactate can also increase the secretion of pro-inflammatory mediators, including interleukin-6 (IL-6), matrix metalloproteinase-1 (MMP1), and IL-1β, by macrophages via monocarboxylate transporters (MCTs) and activate NF-κB [40]. To determine the dual roles of lactate in immunity, the molecular mechanisms by which lactate influences macrophage function need to be elucidated.

Lysine lactylation is a novel post-translational epigenetic mark derived from lactate [9]. In immune cells stimulated by hypoxia, LPS, or bacterial challenge, histone lactylation accumulates at gene promoters to drive further lactate production and directly regulate transcription [41]. For example, lactate-induced lactylation of the RNA methyltransferase METTL3 enhances N6-methyladenosine (m6A) modification in tumor-infiltrating myeloid cells (TIMs), promoting their immunosuppressive phenotype and facilitating tumor immune escape [42]. In the context of CNS diseases, the pathophysiological importance of lactylation is greatly pronounced, given the chronic interplay among metabolic dysregulation, inflammation, and oxidative stress characteristic of these disorders. Pan et al. reported that in Alzheimer's disease, elevated lactylation in pro-inflammatory microglia upregulates the transcription of glycolytic enzyme-encoding genes, thereby exacerbating neuroinflammation. Similarly, lactylation can induce the apoptosis of hippocampal neurons in patients with schizophrenia [11].

Macrophages also play key roles following SCI. After SCI, circulating monocytes infiltrate the lesion and differentiate through multiple stages into macrophages, which then orchestrate inflammation, cell proliferation, and differentiation, and tissue remodeling [43,44]. In this study, we found that macrophage lactylation peaks at 14 dpi and that these lactylated macrophages are most likely derived from the bone marrow. These findings provide a theoretical basis for the use of BMDMs as our in vitro model and lay the foundation for elucidating the precise mechanisms underlying macrophage lactylation after SCI.

By integrating lactylome analysis with CUT&Tag and RNA sequencing, we found that spinal cord injury induces histone H3 lysine 9 lactylation in macrophages, which is associated with transcriptional activation of the TXNIP gene. TXNIP binds to thioredoxin and inhibits its reductase activity, playing a key role in immune-inflammatory responses. First described by Zhou et al., in 2010 in a diabetes model, TXNIP links inflammasome activation to oxidative stress [27]. Park et al. later found that macrophages release TXNIP via autocrine and paracrine mechanisms; free TXNIP directly binds to NLRP3, triggering a conformational change that recruits ASC and caspase 1, assembles a functional inflammasome, and promotes the maturation and secretion of IL-1β, thereby amplifying inflammation [29]. In the CNS, TXNIP also disrupts redox homeostasis and activates the inflammasome, whereas genetic deletion of TXNIP restores redox balance and decreases ischemia-induced neuronal damage [28]. Moreover, the TXNIP-NLRP3 pathway is associated with the release of ROS from mitochondria and loss of membrane potential [45]. Although the key role of TXNIP is established in many CNS disorders, the mechanism by which it contributes to the progressive hypoxia and inflammation characteristic of SCI remains undefined. TXNIP is a master regulator of energy metabolism, and its expression is modulated by glucose, amino acids, and insulin [46,47]; however, no study has determined whether lactate, the end product of glycolysis, can induce the expression of TXNIP. Using complementary in vitro and in vivo models, we showed that macrophage H3K9la epigenetically controls immune-inflammatory pathways and oxidative stress responses, and we determined the molecular mechanism by which this lactylation process, acting through the TXNIP and NLRP3 signaling axes, orchestrates the inflammatory cascade following SCI.

The available strategies aimed at mitigating inflammation after SCI have focused predominantly on surgical decompression, cell-based therapy, biomaterial-mediated interventions, and the administration of corticosteroids or growth factors to reduce activation of the immune system. However, SCI involves an intricate array of pathological processes. While existing strategies have shown promise in preclinical studies, their clinical translation has encountered significant obstacles due to inherent limitations in animal models and the restricted scope of therapeutic effects [48]. Interventions targeting protein lactylation, such as histone deacetylase inhibitors or physical modulation techniques, have demonstrated potential regulatory efficacy in vitro, but their clinical application is hindered by the engagement of nonselective targets and pleiotropic biological effects. For example, conventional epigenetic modulators often disrupt multiple lysine modifications, including acetylation and methylation, with unpredictable consequences for genomic stability. Similarly, metabolic approaches designed to lower lactate concentrations may effectively reduce lactylation but disrupt the delicate metabolic equilibrium between glycolysis and oxidative phosphorylation. Targeted inhibitory peptides are an innovative class of synthetic mini-proteins for therapeutic intervention in lactylation pathways. By incorporating specific post-translational modification motifs into chemically stabilized structures, these peptides selectively disrupt intracellular protein-protein interactions that are otherwise inaccessible to traditional small-molecule drugs or biologics [49]. In this study, we identified the consensus motif critical for H3K9la, and guided by the hypoxic microenvironment characteristic of injured spinal cord tissue, we rationally designed hypoxia-responsive peptide inhibitors. These peptides selectively disrupt the enzyme-complex assembly required for H3K9la by releasing steric hindrance blocks following enzymatic reduction. We subsequently validated their therapeutic potential through preliminary efficacy evaluations in SCI animal models. In the future, studies should investigate the combined use of multiple targeted peptides for synergistic effects along with advanced delivery systems to optimize therapeutic potency and safety for clinical application.

This study, however, had several limitations. The human serum analysis was performed in a relatively small cohort of carefully matched SCI patients. Cohort expansion was constrained by strict inclusion and exclusion criteria—including injury severity, time window after injury, and availability of paired clinical data—as well as limited sample availability. As a result, the human data presented in this study should be regarded as exploratory. Future multi-center studies with larger, well-characterized cohorts will be required to validate lactate as a reliable metabolic biomarker for SCI. In addition, while the present work focused on the H3K9la–TXNIP axis, our CUT&Tag analyses revealed that H3K9la is associated with additional target genes, including Notch1 and Hmox1, suggesting that TXNIP is not the sole downstream effector of H3K9la signaling. Elucidating the functional contributions of these pathways and their interplay with chromatin remodeling mechanisms will be an important direction for future studies. Finally, our evaluation of the hypoxia-responsive peptide H3K9la-pe was confined to the acute and subacute phases of SCI; comprehensive assessments of long-term safety, epigenomic specificity, and sustained therapeutic efficacy will be required before considering clinical translation.

To summarize, this is the first study to systematically integrate post-injury metabolic dysregulation with immune regulatory mechanisms following SCI. We extensively characterized subacute lactate accumulation, examined its role in driving H3K9la in infiltrating macrophages, and identified TXNIP as a direct transcriptional target of this lactylation event. Based on these mechanistic insights, we rationally developed and validated a hypoxia-responsive peptide inhibitor targeting H3K9la. We also found that it can efficaciously promote functional recovery in an SCI mouse model. Our findings offered a targeted epigenetic-metabolic reprogramming strategy, laying a framework for future therapeutic interventions aimed at improving the outcomes of SCI.

Ethics approval and consent to participate

This study was reviewed and approved by the Clinical Ethics Committee of Xiangya Hospital, Central South University (Approval No. 202307601). All procedures involving human participants were conducted in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

All animals research protocols were approved by the Animal Ethics Committee of Central South University. Animal care and use during our experiment were conducted under the guidelines of the Administration Committee of Affairs Concerning Experimental Animals in Hunan Province, China.

Consent for publication

Not applicable.

Fundings

This work was supported by the National Natural Science Foundation of China (No. 82572949; No. 82501664), the Natural Science Foundation of Hunan Province (No. 2024JJ6636, 2025JJ40098, and 2025JJ60482), the Natural Science Foundation of Changsha (kq2502038), the Science Foundation of Xiangya Hospital for Young Scholar (Grant No. 2024q15), the Special Postdoctoral Funding Program of China (Grant No. 2024T171059), and Graduate students of Central South University independently explore innovative projects (2022ZZTS094).

CRediT authorship contribution statement

Chaoran Shi: Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing. Feifei Yuan: Investigation, Validation. Xingyi Chen: Validation, Visualization. Sheng Xu: Investigation. Yinghe Ding: Investigation. Haicheng Wen: Investigation. Chunyue Duan: Writing – review & editing. Tianding Wu: Writing – review & editing. Hongbin Lu: Conceptualization, Methodology, Project administration, Resources, Writing – review & editing. Jianzhong Hu: Conceptualization, Data curation, Investigation, Project administration, Resources, Supervision, Visualization. Jiaqi Xu: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Software, Validation, Visualization, Writing – review & editing. Liyuan Jiang: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

List of abbreviations

2-DG	2 -Deoxy-d-glucose
ASIA	American Spinal Injury Association
BMDMs	bone marrow-derived macrophages
BMS	Basso Mouse Scale
ChIP	Chromatin immunoprecipitation
CNS	Central nervous system
Co-IP	Co-immunoprecipitations
CRISPRi	CRISPR interference
CUT&Tag	Cleavage under targets and tagmentation
dpi	days post-injury
H&E	Hematoxylin and Eosin
H3K9la	Histone H3 Lysine 9 Lactylation
HPLC	High performance liquid chromatography
IL-6	interleukin-6
ISNCSCI	International Standards for Neurological Classification of Spinal Cord Injury
LDH	Lactate dehydrogenase
LSS	Louisville Swim Scale
m6A	N6-methyladenosine
MCTs	monocarboxylate transporters
MEP	motor‐evoked potential
MMP1	matrix metalloproteinase-1
MNase	micrococcal nuclease
NLRP3	Nod-like receptor family pyrin domain containing 3
OCR	oxygen consumption rate
Pankla	global pan-lactylation
PBMCs	peripheral blood mononuclear cells
PTMs	post-translational modifications
ROS	reactive oxygen species
SCI	Spinal cord injury
sgRNAs	Short guide RNAs
TIMs	tumor-infiltrating myeloid cells
TXNIP	thioredoxin-interacting protein
WB	Western Blot

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

CUT&Tag data have been deposited in the NCBI SRA under the accession number PRJNA1313616, RNA-seq data have been deposited in the NCBI SRA under the accession number PRJNA1395828. Additional data are available from the corresponding authors upon request.