Abstract

Introduction

   Intervertebral disc degeneration (IVDD) is one of the main causes of
   low back pain. Existing literature has confirmed the occurrence of
   ferroptosis in IVDD. With the widespread application of artificial
   intelligence (AI), an increasing number of compounds have been screened
   for their potential to alleviate IVDD. BRD-K20733377 is one such
   compound with anti-aging properties. Preliminary experiments have shown
   that BRD-K20733377 can significantly inhibit cellular ferroptosis.
   However, research on the potential therapeutic targets and mechanisms
   of BRD-K20733377 in IVDD remains limited.

Objective

   This study aims to explore the main targets and potential mechanisms of
   BRD-K20733377 in the treatment of IVDD.

Methods

   Network pharmacology, bioinformatics, machine learning algorithms,
   molecular docking, molecular dynamics, and Mendelian randomization were
   used to comprehensively analyze the effects, potential targets, and
   mechanisms of BRD-K20733377 in IVDD. Rat nucleus pulposus-derived
   mesenchymal stem cells (NPMSCs) were selected for in vitro experiments.
   Cell viability was measured by CCK-8 and flow cytometry, while
   antioxidant defense, lipid peroxidation, and iron metabolism were
   explored using JC-1, Reactive Oxygen Species (ROS), FerroOrange dye,
   Lipid peroxides, Malondialdehyde (MDA), and Mitochondrial transmission
   electron microscopy. The expression levels of ferroptosis-related
   proteins were detected by Western blot and immunofluorescence. In the
   rat IVDD model, the effects of BRD-K20733377 on pain threshold, pain
   behavior, and its therapeutic efficacy were evaluated.

Results

   Predictive results identified 30 genes related to ferroptosis in
   BRD-K20733377 and IVDD, revealing that the potential mechanism of
   BRD-K20733377 in treating IVDD is closely associated with ferroptosis.
   In addition, functional enrichment analysis indicated that these genes
   are involved in multiple signaling pathways. Machine learning
   algorithms further identified the core targets STAT3 and NFKB1, and
   Mendelian randomization validated their direct causal relationship with
   IVDD. In vitro experiments confirmed that BRD-K20733377 inhibited IVDD
   by reducing intracellular Fe²⁺ levels and lipid peroxidation, thus
   regulating ferroptosis. Theoretically, BRD-K20733377 may inhibit NPMSCs
   ferroptosis via STAT3/NFKB1 axis. Ferroptosis-related proteins and
   immunofluorescence results further supported this mechanism. In vivo
   experiments showed that BRD-K20733377 significantly improved the
   behavior of SD rats, reduced pain scores, and alleviated IVDD.

Conclusion

   BRD-K20733377 inhibits ferroptosis through the STAT3/NFKB1 axis,
   thereby alleviating IVDD. This provides a new perspective for the study
   of IVDD and could serve as a potential therapeutictarget for IVDD.

Graphical abstract

   [42]graphic file with name 13287_2025_4662_Figa_HTML.jpg

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s13287-025-04662-2.

   Keywords: Intervertebral disc degeneration, Ferroptosis, Network
   pharmacology, Molecular dynamics, Mendelian randomization

Introduction

   In the past thirty years, low back pain (LBP) has become a leading
   cause of disability worldwide, with its incidence continuously rising
   [[43]1]. It is highly prevalent in the general population, affecting
   approximately 84% of individuals during their lifetime [[44]2], with up
   to 80% experiencing recurrent episodes. Intervertebral disc
   degeneration (IVDD) [[45]3], a common musculoskeletal disorder, is
   identified as a primary cause of LBP [[46]4], accounting for over 40%
   of cases globally and imposing significant socio-economic burdens
   [[47]5, [48]6].The intervertebral disc consists of three main
   components: the gelatinous nucleus pulposus (NP) at the core,
   surrounded by the annulus fibrosus (AF) and covered by endplates (EP)
   [[49]7, [50]8]. Various factors such as aging, unhealthy lifestyles,
   and biomechanical stress contribute to the transformation of NP tissue
   from a gelatinous to a fibrous state. This process involves reduced
   levels of type II collagen and aggrecan, crucial components that
   maintain disc integrity. Consequently, IVDD progresses as the disc
   structure deteriorates [[51]9, [52]10].Despite extensive research
   efforts investigating the mechanisms of IVDD, its precise pathogenesis
   remains incompletely understood [[53]11–[54]13]. Recent studies have
   highlighted the inflammatory microenvironment within the intervertebral
   disc as a crucial factor in the development of IVDD [[55]4,
   [56]14–[57]16]. Increased levels of pro-inflammatory mediators in the
   disc trigger a series of pathological events, including the production
   of reactive oxygen species (ROS), breakdown of the extracellular matrix
   (ECM), and different forms of programmed cell death such as apoptosis,
   necroptosis, and ferroptosis, with the latter being characterized by
   iron-dependent lipid peroxidation [[58]17, [59]18].

   Ferroptosis has been identified as a significant contributor to IVDD
   progression, as it accelerates IVDD by promoting immune cell
   infiltration. Disruption of iron metabolism and the dysfunction of
   glutathione peroxidase 4 (GPX4), often triggered by activation of the
   mechanosensitive Piezo1 channel, can further amplify ferroptosis in
   nucleus pulposus cells (NPCs), worsening the degeneration of the disc
   [[60]19]. Strategies aimed at stabilizing the interaction between
   ubiquitin-specific peptidase 11 (USP11) and sirtuin 3 (Sirt3) have
   shown potential in reducing oxidative stress-induced ferroptosis,
   offering a promising therapeutic approach for IVDD [[61]20].
   Furthermore, the use of polydopamine nanoparticles to target
   ferroptosis has shown positive results in preclinical models,
   suggesting their potential for treating IVDD [[62]21]. Together, these
   findings suggest that targeting ferroptosis could be an effective
   strategy to slow down or halt IVDD progression, highlighting the need
   for further research to fully understand the underlying molecular
   mechanisms.

   Additionally, it’s important to point out that current IVDD therapeutic
   drugs, like painkillers, primarily focus on symptom relief rather than
   addressing the underlying structural and functional issues of the disc
   [[63]22]. Given the high cost and extensive effort involved in
   developing new drugs, our approach is to evaluate existing drug
   compounds as potential treatments for IVDD. Natural products are highly
   effective with minimal side effects, providing a rich resource for
   exploration [[64]23]. Therefore, we recommend screening natural product
   libraries to identify potential therapeutic candidates for IVDD. In
   2023, Wong et al. [[65]24] used artificial intelligence (AI) to screen
   over 800,000 molecules, identifying three small molecule compounds:
   BRD-K20733377, BRD-K56819078, and BRD-K44839765. AI predicted that
   these compounds could precisely target anti-aging pathways, and
   subsequent cellular and animal tests validated these predictions. In
   this study, bioinformatics analysis revealed that ferroptosis plays a
   crucial role in IVDD [[66]25]. Additionally, we identified the compound
   BRD-K20733377 (BRD) as a potentially effective drug for modulating
   ferroptosis in IVDD.

   To investigate the role and mechanism of BRD in inhibiting ferroptosis
   and alleviating IVDD, this study employed strategies such as network
   pharmacology, bioinformatics, machine learning algorithms, molecular
   docking, molecular dynamics simulations, and Mendelian randomization.
   In addition, NPMSCs and Sprague-Dawley (SD) rats were selected for in
   vivo and in vitro experiments to further validate the research
   findings. This study reveals the mechanism by which BRD inhibits
   ferroptosis to alleviate IVDD, providing a structural foundation for
   the development of BRD-based drugs. The flow diagram is shown as the
   graphical abstract.

Methods

NP tissue collection

   This study collected human NP tissue samples from 9 patients who
   underwent posterior lumbar interbody fusion (PLIF) surgery for IVDD.
   Among these, 3 samples were classified as Pfirrmann grade III, 3 as
   grade IV, and 3 as grade V (The Pfirrmann grading is provided in
   Supplementary Table 1 (Table S1)). All tissue samples were stained with
   hematoxylin and eosin (HE) and scored according to a pathology scoring
   system (The pathological scoring is provided in Supplementary Table 2
   (Table S2)). Additionally, immunohistochemical analysis was performed
   to evaluate the expression levels of ferroptosis-related proteins
   acyl-coa synthetase long chain family member 4 (ACSL4), ferritin heavy
   chain 1 (FTH1), GPX4 and solute carrier family 7 member 11 (SLC7A11).
   In parallel, 6 three-month-old SD rats were selected, with 3 as control
   group and 3 induced to develop IVDD. After 2 weeks, disc tissue was
   harvested for HE staining and histological scoring (The histological
   scoring is provided in Supplementary Table 3 (Table S3)). Further,
   immunohistochemical analysis was conducted for ferroptosis-related
   proteins GPX4, ACSL4, FTH1 and SLC7A11, followed by quantitative
   analysis. All human tissue samples involved in this study were approved
   by the Ethics Committee of the Northern Jiangsu People’s Hospital
   Affiliated to Yangzhou University (Approval No. 2021ky050) on March 2,
   2021. All animal tissue samples involved in this study were approved by
   the Animal Ethics Committee of Yangzhou University (Approval No.
   202403237) on March 15, 2024. Detailed patient information is provided
   in Supplementary Table 4 (Table S4).

Network pharmacology

Potential target proteins of BRD, ferroptosis and IVDD

   The potential targets of BRD-K20733377 to Homo sapiens were searched
   from internet databases, including SwissTargetPrediction
   ([67]http://SwissTargetPrediction.ch/) [[68]26], SEA
   ([69]http://www.SEA Search Server) and SuperPred
   ([70]http://www.prediction.charite.de). After removing redundancies
   from the targets collected from the three databases, they were used as
   relevant targets for BRD.

   The potential targets of Ferroptosis were searched from Ferrdb
   ([71]http://www.zhounan.org/ferrdb/current/). Data from the ferroptosis
   database were categorized into four target types: driver, marker,
   suppressor, and unclassified. After removing redundancies, the
   remaining targets were identified as relevant to ferroptosis.

   Additionally, the disease target information related to IVDD was
   searched from web databases, including DisGenet
   ([72]https://www.disgenet.org/), Genecards
   ([73]https://www.genecards.org/) and OMIM ([74]https://omim.org/).
   Following the integration of these targets, redundancies were
   eliminated, and the remaining targets were considered relevant to IVDD.
   Through UniProt database ([75]https://www.uniprot.org/), the target was
   screened and limited to the species named “Homo sapiens”, which was
   standardized to the target genes.

Identification of common targets of BRD against IVDD ferroptosis and
construction of protein-protein interaction (PPI) networks

   Common targets among BRD, ferroptosis, and IVDD were extracted using
   Venny 2.1. Subsequently, data were exported, and co-expression data for
   the common targeted genes’ PPI networks were calculated using the
   STRING database ([76]https://string-db.org/). A co-expression network
   of the common targets was constructed using Cytoscape software (version
   3.9.1). Multiple attribute values were obtained through the CytoNCA
   plugin, and the nodes were ranked accordingly. The top 10 nodes were
   identified and visualized.

Gene ontology and Kyoto encyclopedia of genes and genomes pathway enrichment
analysis

   First, the DAVID 6.7 database is used ([77]https://david.ncifcrf.gov/)
   to perform enrichment analysis of Kyoto Encyclopedia of Genes and
   Genomes (KEGG) pathways and Gene Ontology (GO) biological processes.
   For KEGG pathway enrichment analysis, the top 15 pathways were selected
   based on Gene Ratio from highest to lowest. GO enrichment analysis
   includes three aspects: cellular components, molecular functions, and
   biological processes. Use the bioinformatics platform
   ([78]http://www.Bioinformatics.com.cn/) for visualization analysis to
   create bubble plots and histograms. Then, import the intersecting genes
   into MetaScape ([79]https://metascape.org/) to obtain enrichment
   analysis plots for functions and pathways.

Core target selection by two well-established machine learning algorithms

   The Least Absolute Shrinkage and Selection Operator (LASSO) regression
   is a machine learning regression algorithm based on linear
   relationships and regularization [[80]27]. Random Forest (RF) is a
   machine learning regression algorithm that uses bootstrap resampling
   and is based on decision trees [[81]28]. In order to identify candidate
   diagnostic biomarkers for ferroptosis in IVDD, this study combined
   LASSO and RF algorithms to identify potential biomarkers [[82]29,
   [83]30]. In brief, the top 10 genes obtained from the PPI network via
   the MCC algorithm were selected as input and analyzed the expression
   profiles of the [84]GSE266883 datasets. Subsequently, LASSO analysis
   was performed using the ‘glmnet’ package and RF analysis using the
   ‘randomForest’ package in R software. Finally, the overlapping genes
   identified by both machine learning algorithms were defined as
   candidate biomarkers for ferroptosis in IVDD.

Molecular docking

   Download the 3D structure of BRD from PubMed using its SMILES code, and
   convert the drug’s 3D structure to PDB format using PyMOL. Find the
   human source of the relevant protein from UniProt. Choose an
   appropriate protein structure based on Method, Resolution, Chain, and
   Positions, download the protein PDB structure, and import the prepared
   drug and protein structures into AutoDock Vina for molecular docking.
   Import the Complex file from the docking results into the
   Protein-Ligand Interaction Profiler (PLIP-Welcome (tu-dresden.de)) to
   obtain visualizations of various interactions between the drug and
   protein.

Molecular dynamics (MD) and simulation

   Molecular dynamics (MD) simulations were conducted on the active
   compounds and proteins identified through molecular docking using
   Desmond software [[85]31]. The topology files for the small molecules
   were created using the ATB server ([86]http://atb.uq.edu.au/index.py).
   For the NFKB1 and STAT3 proteins, their topologies were generated via
   the PDB2GMX tool, and the water model used was SPC/E, with sodium ions
   added to neutralize the system and maintain electrical neutrality.The
   system was simulated with periodic boundary conditions. Initially, the
   protein-ligand complexes underwent energy minimization for 1000 ps at
   300 K. Subsequently, the system was equilibrated under NVT and NPT
   ensembles for 100 ps each, with position restraints applied throughout.
   Finally, the position restraints were removed, and a 100 ns molecular
   dynamics simulation was performed at 300 K [[87]32].

The Mendelian randomization study identified the relationship between core
target and IVDD

   A Mendelian Randomization (MR) study employs genetic variants as
   instrumental variables to establish causal relationships between an
   exposure and its outcome [[88]33]. This method assists in evaluating
   whether an observed association reflects a causal link. In MR,
   confounding bias is minimized due to the random allocation of genetic
   variants at conception, while reverse causation is prevented because
   genetic variants are set before the disease develops. In this study, we
   conducted two-sample MR analyses to examine the causal relationship
   between Core target and IVDD. We utilized publicly available expression
   quantitative trait loci (eQTLs) and protein quantitative trait loci
   (pQTLs) that are associated with the susceptibility to IVDD. The
   summary-level eQTL data were retrieved from the IEU OpenGwas database
   ([89]https://gwas.mrcieu.ac.uk/) and the deCODE database
   ([90]https://www.decode.com/summarydata/) [[91]34]. The GWAS summary
   data for IVDD outcomes were sourced from the FinnGen database
   ([92]https://www.finngen.fi/). The effect estimates were derived using
   the inverse variance weighted (IVW) method.

Cellular experiments verified that BRD inhibited Erastin-induced ferroptosis
in NPMSCs

Isolation and culture of NPMSCs

   SD rats (weighing 250–350 g, aged 2.5–3.5 months) were purchased from
   the Animal Experimental Center of Yangzhou University (License No.:
   SYXK (SU) 2021–0027). The experimental study was approved by the Ethics
   Committee of Northern Jiangsu People’s Hospital Affiliated to Yangzhou
   University. The separation of NP tissues and the isolation of NPMSCs
   were performed as previously described [[93]35]. Briefly, the SD rats
   were euthanized, and the NP tissue from the caudal vertebrae was
   carefully separated under sterile conditions, followed by washing with
   phosphate-buffered saline (PBS)(Biosharp, China) containing 1%
   penicillin-streptomycin (Gibco, USA). The NP tissue was cut into 1 mm³
   pieces and digested at 37 °C for 12 h using collagenase type II (Gibco,
   USA). After washing with regular saline and centrifuging at 1000 rpm
   for 4 min, the cells were resuspended in MSC complete medium (Cyagen,
   USA) supplemented with 10% fetal bovine serum (FBS)(Clark, USA) and 1%
   penicillin-streptomycin solution (Gibco, USA), and cultured in a
   sterile incubator at 37 °C with 5% CO2. The culture medium was changed
   every three days. When the adherent cells reached 80% confluence, they
   were passaged at a 1:3 ratio. The third-generation cells obtained by
   this method were used for subsequent experiments.

Cell viability and cell proliferation assay

   NPMSCs were plated in a 96-well plate at a density of 1500 cells per
   well, with varying concentrations of BRD (MCE, China) at 0 µM, 3 µM, 6
   µM, 9 µM, 12 µM, and 15 µM. The cells were cultured for 24, 48, and
   72 h. Afterward, a fresh culture medium containing 10% Cell Counting
   Kit-8 (CCK-8)(Beyotime, China) was added to each well, and the plates
   were incubated at 37 °C for 2 h. The optical density (OD) of each well
   was then measured at 450 nm using a microplate reader (Bio-Rad, USA).
   To establish a ferroptosis model of NPMSCs, the cells were cultured in
   complete medium with varying concentrations of Erastin (MCE, China) (0,
   2.5, 5, 7.5, 10, and 12.5 µM) for 24 h. Prior to Erastin treatment, the
   cells were pretreated with 6 µM of BRD for 24 h. Cell viability was
   assessed using a Cell Counting Kit-8 according to the manufacturer’s
   protocol. The cell cycle phases of NPMSCs were determined using a cell
   cycledetection kit (KeyGEN, China) and analyzed by flow cytometry,
   further comparing the proportion of cells in the G2 phase during
   different proliferative stages. In brief, NPMSCs were harvested from a
   six-well plate, washed twice with cold PBS, fixed with 75% ethanol at
   4 °C for 2 h to overnight, and incubated in the dark for 30 min with a
   mixture of propidium iodide (PI) and RNase A. The cell cycle phases
   were analyzed by flow cytometry (BD Company, USA).

JC-1 assay

   Mitochondrial membrane potential (MMP) was measured using a JC-1
   Detection Kit (KeyGEN, China). Briefly, NPMSCs were incubated with 2 µM
   JC-1 dye, a cationic dye, for 20 min. After two washes with incubation
   buffer, the cells were cultured in complete medium, observed under a
   fluorescence microscope, and analyzed using ImageJ software.

ROS assay

   In accordance with the manufacturer’s protocol for the ROS detection
   fluorescent probe-DHE kit (KeyGEN, China), NPMSCs were incubated with
   20 µM DHE at 37 °C for 30 min. After incubation, the cells were
   observed under a fluorescence microscope and analyzed using ImageJ
   software.

FerroOrange assay

   According to the manufacturer’s instructions provided by the
   FerroOrange detection kit (Dojindo, Japan), NPMSCs were incubated with
   a 1 µmol/L FerroOrange working solution at 37 °C in a 5% CO2 incubator
   for 30 min. The cells were then observed under a fluorescence
   microscope and analyzed using ImageJ software.

Liperfluo assay

   Following the manufacturer’s guidelines for the Liperfluo detection kit
   (DOJINDO, Japan), NPMSCs were treated with a 1 mM Liperfluo working
   solution and incubated at 37 °C with 5% CO2 for 30 min. After washing,
   the cells were examined under a fluorescence microscope and analyzed
   using ImageJ software.

Transmission electron microscopy

   On transmission electron microscopy (TEM) sections, mitochondrial
   function was assessed by observing whether the mitochondria were
   swollen and whether the mitochondrial cristae density was increased.

Malondialdehyde assay

   According to the manufacturer’s protocol for the Malondialdehyde (MDA)
   Assay Kit (Solarbio, China), the supernatant was obtained after NPMSCs
   were lysed. The MDA detection working solution was added as required,
   followed by heating, cooling, and centrifugation. The absorbance was
   measured at 532 nm.

Extraction of total cellular protein and Western blot analysis

   Total protein from NPMSCs was extracted using the Whole Cell Lysis
   assay (KeyGEN, China), and protein concentration was quantified using a
   BCA protein assay kit (Beyotime, China). For each sample, 30 µg of
   protein was loaded onto a 10%-12% SDS-PAGE gel for separation, followed
   by transfer to a polyvinylidene difluoride (PVDF) membrane (Millipore,
   USA). The membranes were blocked with 5% skim milk for 2 h at room
   temperature, followed by overnight incubation at 4 °C with the primary
   antibodies: GPX4 (1:3000; Proteintech, China), ACSL4 (1:20000;
   Proteintech, China), FTH1 (1:1500; Proteintech, China), SLC7A11
   (1:3000; Proteintech, China), and GADPH (1:1000; Proteintech, China).
   Afterward, the membranes were incubated with HRP-conjugated secondary
   antibodies (1:500; Proteintech, China). Protein bands were visualized
   using an enhanced chemiluminescence system, and the protein expression
   was quantified using ImageJ software.

Immunofluorescent staining

   NPMSCs treated with different conditions were subjected to
   immunofluorescence staining for GPX4, ACSL4, FTH1, and SLC7A11. In
   brief, the NPMSCs were fixed with 4% paraformaldehyde (PFA) for
   25–30 min, followed by permeabilization with 0.1% Triton X-100 for
   10 min. After blocking with 5% BSA at room temperature for 60 min, the
   NPMSCs were incubated overnight with the corresponding primary
   antibodies at 4 °C. The cells were then washed three times with cold
   PBS and incubated with a 1:500 dilution of secondary antibody
   (Proteintech, China) at room temperature for 60 min. Nuclei were
   stained with 2-(4-amidinophenyl)-1 H-indole-6-carboxamidine (DAPI).
   Finally, the NPMSCs were mounted with an anti-fluorescence quenching
   agent (Servicebio, China) and observed under a fluorescence microscope.

Establishment of rat IVDD model

   A total of 30 male SD rats, aged 2 to 4 months and weighing between 200
   and 300 g, were randomly assigned to one of three groups: Control group
   (n = 10), IVDD group (n = 10), and BRD group (n = 10). The IVDD model
   in SD rats was created following the method outlined previously
   [[94]36]. In brief, the rats were anesthetized with pentobarbital, and
   the tail skin was disinfected with povidone iodine. A 21 G needle was
   inserted into the coccygeal intervertebral discs (Co 4–5, Co 6–7, Co
   8–9), rotated 180°, and left in place for 5 s at a depth of 5 mm. Two
   weeks after the surgery, the BRD group received treatment every 4
   weeks, with BRD (5 µl, suspended in 0.9% PBS at a concentration of 6 µM
   [[95]37]) injected into the disc using a 31 G needle. The control group
   received no special treatment, while the IVDD group underwent disc
   injections of 0.9% PBS.

Behavioral assessment

   The pain behavior of rats was assessed through a field test at 0 weeks,
   2 weeks, 4 weeks, 6 weeks, 8 weeks, and 10 weeks [[96]38, [97]39]. The
   animals were placed in a 60 × 60 cm open-field box, and their behavior
   was recorded for 10 min from a bird’s-eye view. The animals were
   habituated for 10 min in a pen the day before the test. The room
   lighting and time were kept consistent. A gait refinement analysis
   system(SANS SA114)(SCIMON Science Intelligent Technology, Chinese) was
   used to analyze 11 parameters related to LBP, as in previous studies
   [[98]39]. The Von Frey filament pain meter was used to perform the Von
   Frey test. Three animals were allowed to adapt for 15 min in a
   transparent acrylic chamber placed on a metal grid. A maximum force of
   30 g was applied, gradually increasing over 10 s, and held for up to
   40 s. The withdrawal threshold and time were sequentially and
   alternately assessed by applying the probe to the tail base and
   surgical site. This process was repeated at least three times, and the
   average value was calculated. During the behavioral tests, the
   researchers were blinded to the groupings.

Radiographic evaluation and histological analysis

   After 6 weeks of puncture, 5 SD rats from the control group, IVDD
   group, and BRD group were selected, totaling 15 rats. After 10 weeks of
   puncture, the remaining 15 SD rats from these groups were also
   selected. All 30 SD rats were anesthetized with pentobarbital and
   placed in a prone position. X-ray and MRI images were collected, and
   the disc height index (DHI) [[99]40] was calculated using ImageJ
   software. The degree of IVDD was assessed using the Pfirrmann grading
   system [[100]41]. After euthanizing the rats with an overdose of
   pentobarbital, tail specimens were harvested for NP area analysis and
   gross specimen histological analysis [[101]42]. The specimens were
   fixed in 4% paraformaldehyde for 48 h, decalcified in EDTA for 1 month,
   dehydrated in graded ethanol, and embedded in paraffin. Safranin O-fast
   green (SF), hematoxylin and eosin staining (HE) were used on paraffin
   tissue sections.

Immunohistochemistry

   IVD paraffin sections underwent antigen retrieval by incubation in
   citrate buffer (pH 6.0) (Servicebio, China) and microwaving on high
   heat for 6–8 min. To prevent nonspecific protein binding, sections were
   blocked with 5% bovine serum albumin (BSA) at room temperature for
   30 min. Subsequently, sections were incubated overnight at 4 °C with
   primary antibodies (anti-ACSL4, anti-FTH1, anti-GPX4, anti-SLC7A11,
   anti-COL2, anti-MMP13, 1:1000, Proteintech, China). Following this,
   sections were incubated with appropriate secondary antibodies for 1 h
   at ambient temperature. Finally, sections were dehydrated, sealed, and
   digitally scanned using a slide scanner.

Data statistical analysis

   All experiments were performed in biological triplicates, and the
   results are presented as mean ± standard deviation (Mean ± SD).
   Statistical analysis was conducted using one-way analysis of variance
   (one-way ANOVA). Data were analyzed with GraphPad Prism 9.0. The
   significance levels are indicated as follows: ns denotes no statistical
   significance, * indicates P < 0.05, ** indicates P < 0.01, ***
   indicates P < 0.001, **** indicates P < 0.0001.

   The work has been reported in line with the ARRIVE guidelines 2.0.

Results

Ferroptosis in NP tissue

   To evaluate the involvement of ferroptosis in promoting IVDD, a total
   of 9 human NP tissues with varying degrees of degeneration, as
   classified by the Pfirrmann gradingsystem, were collected and MRI
   (Fig. [102]1A). In this study, H&E staining was used to perform
   pathological scoring on NP tissue samples of different Pfirrmann grades
   (Fig. [103]1B, H). Additionally, IHC analysis was conducted to measure
   the expression of ACSL4, FTH1, GPX4, and SLC7A11 in degenerative NP
   tissues (Fig. [104]1C–G). The results showed that with increasing
   degeneration grade, the expression of FTH1, GPX4, and SLC7A11
   significantly decreased, while ACSL4 expression increased. To further
   explore whether this phenomenon also exists in SD rats, 3 cases of
   normal and degenerated caudal intervertebral disc tissues from SD rats
   were collected, and MRI was performed (Fig. [105]1I). H&E staining was
   used to perform pathological scoring on normal and degenerated disc
   tissue samples from SD rats (Fig. [106]1J, P). Furthermore, IHC
   analysis was conducted to measure the expression of ACSL4, FTH1, GPX4,
   and SLC7A11 in normal and degenerated intervertebral disc tissues from
   SD rats (Fig. [107]1K–O). The results showed that in the degenerated
   group, the expression of ACSL4 significantly increased, while the
   expression of FTH1, GPX4, and SLC7A11 significantly decreased.

Fig. 1.

   [108]Fig. 1
   [109]Open in a new tab

   Ferroptosis immunohistochemistry in human and rat nucleus pulposus (NP)
   tissues. A Magnetic resonance imaging (MRI) scans showing different
   degrees of human intervertebral disc (IVD). B Hematoxylin and eosin
   (H&E) staining showing degenerative human NP tissues from patients
   (upper panel, scale bar = 200 μm; lower panel, scale bar = 50 μm). C–F
   Representative acyl-coa synthetase long chain family member 4 (ACSL4),
   ferritin heavy chain 1 (FTH1), glutathione peroxidase 4 (GPX4), and
   solute carrier family 7 member 11 (SLC7A11) immunostaining images in
   degenerative human NP tissues (upper panel, scale bar = 200 μm; lower
   panel, scale bar = 50 μm). G Quantification of ACSL4, FTH1, GPX4, and
   SLC7A11 in degenerated NP tissues (n = 3 per group). H Pathological
   scoring of human degenerated NP tissues. I MRI scans showing normal and
   degenerated rat IVD. J H&E staining showing rat NP tissues (upper
   panel, scale bar = 1000 μm; lower panel, scale bar = 250 μm). K–N
   Representative ACSL4, FTH1, GPX4, and SLC7A11 immunostaining images in
   rat NP tissues (upper panel, scale bar = 200 μm; lower panel, scale bar
   = 50 μm). O Quantification of ACSL4, FTH1, GPX4, and SLC7A11 in rat NP
   tissues (n = 3 per group). P Histological scoring of rat NP tissues.
   Data are presented as mean ± standard deviation (SD). * indicates
   P < 0.05, ** indicates P < 0.01, *** indicates P < 0.001, ****
   indicates P < 0.0001

Network pharmacological analysis

   In this study, we used database mining to identify all relevant targets
   of BRD, Ferroptosis, and IVDD. The network pharmacology was employed to
   predict the potential mechanism of BRD in combating IVDD through
   ferroptosis. By utilizing target prediction websites, we obtained 465
   targets related to BRD, 1225 targets related to ferroptosis, and 1823
   targets related to IVDD. Subsequently, we used the Venny 2.1.0 software
   to generate the Venn diagram, which shows the overlap of 30 targets
   between BRD, Ferroptosis, and IVDD (Fig. [110]2C). A PPI interaction
   network was constructed based on the STRING database, consisting of 30
   nodes and 256 edges (Fig. [111]2D). Subsequently, we used Cytoscape
   software (version 3.9.1) to visualize these shared targets based on the
   following criteria: Closeness ≥ 0.025136333, Betweenness ≥ 12,

Fig. 2.

   [112]Fig. 2
   [113]Open in a new tab

   Identification of common targets and analysis of protein-protein
   interaction (PPI) network. A The 2D structure of BRD-K20733377. B The
   3D structure of BRD-K20733377. C Venn diagram to acquire the
   intersection of genes as potential therapeutic targets. D PPI network
   based on the STRING database. E High-ranking targets selection and PPI
   network visualization analysis

   and Degree ≥ 17.0666666. Therefore, PPI network of the top ten
   high-ranking targets was obtained. These high-ranking targets were:
   NFE2L2, TNF, PARP1, STAT3, NFKB1, SIRT1, MTOR, IL6, CTNNB1, and AKT1
   (Fig. [114]2E). GO and KEGG pathway enrichment analyses were conducted
   on these 10 high-ranking targets genes. The results of the GO analysis
   were shown in Fig. [115]3A. The potential target genes are primarily
   enriched in the following Biological Process (BP) categories:
   Regulation of oxidative stress-induced intrinsic apoptotic signaling
   pathway, Cellular response to peptide hormone stimulus, Regulation of
   small molecule metabolic process, Intrinsic apoptotic signaling pathway
   in response to oxidative stress and Response to peptide hormone. The
   enriched Molecular Function (MF) terms include RNA polymerase
   II-specific DNA-bindingtranscription factor binding, DNA-binding
   transcription factor binding, Nuclear receptor binding, Nuclear
   estrogen receptor binding and NAD binding.Furthermore, according to
   KEGG pathway enrichment analysis, these high-ranking targets genes were
   primarily enriched in signaling pathways such as Pathways in Human
   cytomegalovirus infection, Adipocytokine signaling pathway, Insulin
   resistance, Fluid shear stress and atherosclerosis, Alcoholic liver
   disease, Hepatitis C, Kaposi sarcoma − associated herpesvirus infection
   and Acute myeloid leukemia (Fig. [116]3B). Additionally, we performed
   Metascape functional and pathway enrichment analyses on the 10 high
   ranking genes. The main enrichments were observed in the following
   functions and pathways: Molecular pathway for oxidative stress,
   Cellular response to chemical stress, Regulation of apoptotic signaling
   pathway, Cellular senescence, Positive regulation of programmed cell
   death, Cellular responses to stimuli (Fig. [117]3C, D).

Fig. 3.

   [118]Fig. 3
   [119]Open in a new tab

   Gene ontology (GO) and kyoto encyclopedia of genes and genomes (KEGG)
   enrichment analysis. A Enriched GO terms for biological processes (BP),
   molecular functions (MF): A combined bar chart (Top 5). B KEGG enriched
   pathways sankey bubble plot (Top 8). C, D Metascape functional and
   pathway enrichment analysis plot (Top 20)

Machine learning algorithms

   This study combined two machine learning algorithms to identify
   potential relevant biomarkers from the 10 central genes screened from
   the PPI network. First, we used the CytoNCA plugin to calculate the top
   10 high-ranking genes, which include: NFE2L2, TNF, PARP1, STAT3, NFKB1,
   SIRT1, MTOR, IL6, CTNNB1, and AKT1. Subsequently, four biomarkers were
   identified using the LASSO algorithm, including PARP1, STAT3, NFKB1,
   and AKT1 (Fig. [120]4A-B). Meanwhile, the RF algorithm determined three
   potential biomarkers based on their importance, including NFKB1,
   cross-referenced to identify genes for candidate biomarkers,
   specifically signal transducer and activator of transcription 3 (STAT3)
   and nuclear factor kappa B subunit 1 (NFKB1) (Fig. [121]4E).

Fig. 4.

   [122]Fig. 4
   [123]Open in a new tab

   Determining target hub genes through machine learningmethods. A The
   least absolute shrinkage and selection operator (LASSO) regression of
   the 10 high ranking targets calculated by CytoNCA. B LASSO regression
   parameter selection using crossvalidation. C, D Random forest (RF) and
   importance analysis. E Venn diagram of hub target

Molecular docking, molecular dynamics (MD) and simulation

   Docking simulation is a convenient and effective method used to explore
   the interactions between small molecules and their target proteins. In
   this study, AutoDock Vina software was used to simulate the docking of
   BRD with the NFKB1 and STAT3 proteins, and their binding energy scores
   were obtained. A negative binding energy indicates the potential for
   binding. Typically, a binding energy value lower than-5 kcal/mol is
   considered more likely to result in binding [[124]43, [125]44]. In this
   study, the binding energies of BRD with NFKB1 and STAT3 proteins were
   − 7.0 kcal/mol and − 7.4 kcal/mol, respectively, indicating that BRD
   has ideal binding affinity for both NFKB1.

   and STAT3. BRD binds to the surface pocket of the NFKB1 protein, which
   is formed by the amino acids LEU202, PHE151, VAL150, THR205, YLS147,
   and LEU210. Among these amino acids, the main interactions between BRD
   and NFKB1 are hydrogen bonds (blue solid lines) and hydrophobic
   interactions (gray dashed lines) (Fig. [126]5A). Additionally, BRD
   binds to the surface pocket of the STAT3 protein, which is formed by
   the amino acids LYS591, ARG609, TRP623, SER636, PRO639, GLN635, and
   ILE659. The primary interactions between BRD-K20733377 and STAT3 among
   these amino acids were hydrogen bonds (blue solid lines), hydrophobic
   interactions (gray dashed lines), and π-cation interactions (yellow
   dashed lines) (Fig. [127]5B). These docking results suggested that BRD
   may interact directly with NFKB1 and STAT3, and potentially exert
   biological effects through these proteins in the process of ferroptosis
   in IVDD.During the simulation of the STAT3 protein, no significant RMSD
   fluctuations occurred, indicating that the protein structure did not
   undergo any disruption. The RMSD of BRD stabilized early in the
   simulation and fluctuated within a very small range of 0-1.5 Å.
   Moreover, during the later stages of binding, the fluctuations of the
   complex became even more stable. This indicates that BRD binds very
   tightly with STAT3, and the stability of the BRD-STAT3 binding is high
   (Fig. [128]5C). Except for some local regions, the RMSF of the protein
   remained below2.5 Å, suggesting that the main structure of the protein
   is highly rigid, possibly due to the binding of BRD (Fig. [129]5D). The
   radius of gyration (RoG) of the BRD-STAT3 complex fluctuated within a
   very small range of 0-0.5 Å, indicating that the system exhibits good
   compactness (Fig. [130]5E). For the BRD-STAT3 complex, the number of
   hydrogen bonds was 1–2 in the early stages of the simulation, and 1–3
   in the later stages, indicating that hydrogen bonding contributes
   moderately to the stability of the complex (Fig. [131]5F). After NFKB1
   binds to BRD, the RMSD reached stability early in the simulation and
   fluctuated within a very small range of 0–1 Å. Furthermore, in the
   later stages of binding, the fluctuations of the complex became even
   more stable. This suggests that BRD binds very tightly with NFKB1, and
   the binding stability of BRD-NFKB1 is high (Fig. [132]5G). After
   binding BRD, the RMSF fluctuations of the BRD-NFKB1 complex
   significantly decreased, indicating that the main structure of the
   protein became highly rigid, likely due to the binding of BRD
   (Fig. [133]5H). The radius of gyration (RoG) of the BRD-NFKB1 complex
   fluctuated within a very small range of 0-0.5 Å, indicating that the
   system has good compactness (Fig. [134]5I). For the BRD-NFKB1 complex,
   the number of hydrogen bonds was 1–2 in the early stages of the
   simulation and 2–3 in the later stages, indicating that hydrogen bonds
   contribute moderately to the stability of the complex (Fig. [135]5J).

Fig. 5.

   [136]Fig. 5

   [137]Fig. 5
   [138]Open in a new tab

   Molecular docking and molecular dynamics simulation of BRD-K20733377
   with key targets. A The results show the overall binding view, local
   binding view, and local 2D view of the nuclear factor kappa B subunit 1
   (NFKB1)- BRD-K20733377 complex obtained from docking. B The results
   show the overall binding view, local binding view, and local 2D view of
   the signal transducer and activator of transcription 3
   (STAT3)-BRD-K20733377 complex obtained from docking. C–F Results of
   molecular dynamics (MD) simulation analysis illustrating root mean
   square deviation (RMSD), root mean square fluctuation (RMSF), radius of
   gyration (Rog), and number of Hbonds for STAT3 protein and
   BRD-K20733377 ligand complexes. G–H Results of molecular dynamics (MD)
   simulation analysis illustrating root mean square deviation (RMSD),
   root mean square fluctuation (RMSF), radius of gyration (Rog), and
   number of Hbonds for NFKB1 protein and BRD-K20733377 ligand complexes

The MR identified the relationship between STAT3, NFKB1 and IVDD

   Through MR analysis using pQTL data, we identified suggestive evidence
   of an association between STAT3, NFKB1, and IVDD risk (IVW, OR = 1.02,
   95% CI = 1.00-1.04, p = 0.025; IVW, OR = 1.09, 95% CI = 1.00-1.19,
   p = 0.043). The Cochran Q test indicated no evidence of heterogeneity
   in the MR results (Q_p > 0.05, Q_p > 0.05). Both the MR-Egger
   regression (p > 0.05, p > 0.05) and MR-PRESSO analysis (p > 0.05,
   p > 0.05) indicated no significant pleiotropy at the overall level.
   These findings were illustrated using scatter plots. Furthermore, the
   leave-one-out analysis revealed that the overall estimate remained
   unaffected by the removal of any individual SNP (Fig. [139]6A–H).

Fig. 6.

   [140]Fig. 6
   [141]Open in a new tab

   Results of mendelian randomization (MR) analysis. A–D The MR analysis
   investigating the association between signal transducer and activator
   of transcription 3 (STAT3) and intervertebral disc degeneration(IVDD).
   The x-axis represents the impact of the single nucleotide polymorphism
   (SNP) on the exposure, while the y-axis shows its effect on the
   outcome. A positive slope (> 0) indicates that elevated STAT3 levels
   are linked to an increased risk of IVDD. Funnel plots show no evidence
   of heterogeneity among the SNPs. In the forest plot of SNPs, the
   combined results (depicted by the red line) illustrate that higher
   STAT3 levels are associated with an increased risk of IVDD. E, F The MR
   analysis investigating the association between NFKB1 and IVDD. The
   x-axis represents the impact of the single nucleotide polymorphism
   (SNP) on the exposure, while the y-axis shows its effect on the
   outcome. A positive slope (> 0) indicates that elevated NFKB1 levels
   are linked to an increased risk of IVDD. Funnel plots show no evidence
   of heterogeneity among the SNPs. In the forest plot of SNPs, the
   combined results (depicted by the red line) illustrate that higher
   NFKB1 levels are associated with an increased risk of IVDD

BRD inhibits Erastin-induced ferroptosis in NPMSCs

   The CCK-8 assay was used to evaluate the effects of BRD and Erastin on
   NPMSCs viability. As shown in Fig. [142]7A–C, BRD at a concentration of
   6 µM for 72 h did not show significant cytotoxic effects. Therefore, a
   concentration of 6 µM BRD was chosen for subsequent experiments. After
   Erastin treatment, the viability of NPMSCs decreased in a
   dose-dependent manner (Fig. [143]7D), and 5 µM Erastin was selected for
   furtherexperiments. Cell cycle analysis was performed to assess the
   effect of the drugs on the G2/M phase. Compared to the control group,
   the percentage of NPMSCs in the G2/M phase was higher in the Erastin
   group, indicating cell cycle arrest, which was reduced by BRD
   treatment, lowering the percentage of NPMSCs arrested in the G2/M phase
   (Fig. [144]7E–H). JC-1 was used to detect mitochondrial fluorescence
   intensity during ferroptosis. When the MMP is high, JC-1 exists as
   J-aggregates in the mitochondria, producing red fluorescence. When the
   MMP is low, JC-1 is released from the mitochondrial matrix as monomers,
   producing green fluorescence in the cytoplasm. As shown in
   Fig. [145]7I, J, MMP decreased after Erastin treatment, and this effect
   was partially reversed by BRD treatment. ROS levels, which reflect
   cellular reactive oxygen species during ferroptosis, were measured in
   Fig. [146]7K–L. Compared to the control group, NPMSCs in the Erastin
   group had higher ROS levels, which were partially reversed by BRD
   treatment. We explored the potential of BRD to inhibit Erastin-induced
   ferroptosis. After Erastin stimulation, NPMSCs exhibited an increase in
   Fe²⁺ content and accumulation of lipid peroxides. However, 6 µM BRD was
   able to suppress the accumulation of Fe²⁺ and lipid peroxides
   (Fig. [147]7M-P), and the fluorescence of FerroOrange and Liperfluo was
   partially reduced. In transmission electron microscopy of mitochondria,
   Erastin-stimulated NPMSCs showed mitochondrial shrinkage, increased
   mitochondrial cristae density, BRD treatment changes in mitochondrial
   morphology were rescued (Fig. [148]7Q), with mitochondrial morphology
   restored and mitochondrial cristae density reduced. MDA is a metabolic
   product of lipid peroxidation and is one of the markers of ferroptosis.
   After Erastin stimulation, MDA levels increased, whereas BRD treatment
   significantly reduced MDA levels (Fig. [149]7R). These results
   confirmed that BRD can inhibit Erastin-induced ferroptosis. GPX4,
   ACSL4, FTH1, and SLC7A11 are key proteins involved in ferroptosis. In
   Erastin-treated NPMSCs, the expression levels of GPX4, FTH1, and
   SLC7A11 in total cell protein decreased, while the expression of ACSL4
   in total protein increased (Fig. [150]8A–E). Furthermore,
   immunofluorescence staining of NPMSCs revealed that the fluorescence
   intensity of GPX4, FTH1, and SLC7A11 was reduced in the Erastin-treated
   group, whereas the fluorescence intensity of ACSL4 was elevated. In
   contrast, these fluorescence intensities were reversed in the
   BRD-treated group (Fig. [151]8F–M). The immunofluorescence staining
   results were consistent with Western blot analyses.

Fig. 7.

   [152]Fig. 7

   [153]Fig. 7
   [154]Open in a new tab

   BRD-K20733377 inhibits Erastin-induced ferroptosis in nucleus pulposus
   mesenchymal stem cells (NPMSCs). A-C) CCK-8 assay results of NPMSCs
   treated with different concentrations of BRD-K20733377 for 24, 48, and
   72 h. D CCK-8 assay results of NPMSCs treated with different
   concentrations of Erastin for 24 h. E–G Cell cycle analysis of NPMSCs
   in different treatment groups. H Quantitative analysis of the cell
   cycle results. I Fluorescence-based detection of mitochondrial membrane
   potential (MMP) in different groups. Red fluorescence indicates
   aggregated JC-1 in mitochondria, while green fluorescence indicates the
   JC-1 monomer. J Quantitative analysis of MMP results. K
   Fluorescence-based detection of reactive oxygen species (ROS) in
   different groups. Green fluorescence represents high levels of ROS. L
   Quantitative analysis of ROS results. M, N Fluorescence detection of
   intracellular ferrous ions (FerroOrange) in different groups. O, P
   Fluorescence detection of lipid peroxides (Liperfluo) in different
   groups. Q Transmission electron microscopy (TEM) of mitochondrial in
   different groups. (R) Malondialdehyde (MDA) in different groups. Data
   are expressed as the mean ± standard deviation. n = 3, * indicates
   P < 0.05, ** indicates P < 0.01, *** indicates P < 0.001

Fig. 8.

   [155]Fig. 8
   [156]Open in a new tab

   Western blot and immunofluorescence of ferroptosis-related proteins.
   A–E Western blot was conducted to examine the protein levels of
   acyl-coa synthetase long chain family member 4 (ACSL4), ferritin heavy
   chain 1 (FTH1), glutathione peroxidase 4 (GPX4), solute carrier family
   7 member 11 (SLC7A11). Full-length blots/gels are presented in
   supporting materials. F–L Immunofluorescence staining was conducted to
   examine the expression and localization of ACSL4, FTH1, GPX4, and
   SLC7A11 (green). n = 3, * indicates P < 0.05, ** indicates P < 0.01,
   *** indicates P < 0.001, **** indicates P < 0.0001

BRD modulated painlike behaviors in vivo

   The IVDD induction and drug injection surgeries were successfully
   performed with no complications. Figure [157]9A showed the experimental
   procedure diagram. The Von Frey test indicated that, compared to the
   PBS group, the BRD group significantly reduced the pain response
   threshold at the tail base surgery site (Fig. [158]9B) and the surgical
   site (Fig. [159]9C) (p < 0.05). The open field test, aimed at assessing
   spontaneous pain-like behavior changes, showed no significant
   differences between the groups (p > 0.05). Although not reaching
   statistical significance, rats in the BRD group showed higher average
   speed, total distance, and duration of movement (Fig. [160]9D–N).

Fig. 9.

   [161]Fig. 9
   [162]Open in a new tab

   Behavioral assessment. A Rat experimental procedure diagram. B, C Von
   Frey test using a tactile meter to assess pain threshold. D–N Rat pain
   behavior assessed through the open field test. The representative
   behavioral parameters evaluated include Mean velocity (D), Total
   distance (E), Movement (F), Not Movement (G), Probability of waking
   (H), Unsupported rearing (I), Duration of unsupported rearing (J),
   Walking duration (K), Walking frequency (L), Supported rearing (M),
   Duration of supported rearing (N). n = 5, * indicates P < 0.05, **
   indicates P < 0.01, *** indicates P < 0.001

BRD effectively prevents IVDD in rat model

   To investigate the effects of BRD on in vivo IVDD, we established an
   IVDD rat model and performed radiological assessments at different
   postoperative time points (6 weeks and 10 weeks) to observe the
   progression of IVDD, including changes in the DHI and Pfirrmann
   grading. Post-surgery, PBS or BRD was injected into the rat discs.
   X-ray and MRI imaging were used to evaluate changes in DHI and NP
   signal intensity. In the PBS injection group, DHI significantly
   decreased, while BRD treatment reversed this reduction
   (Fig. [163]10A-C). In the PBS group, NP signal intensity significantly
   decreased, and the Pfirrmann grade significantly increased, whereas BRD
   treatment reversed these changes (Fig. [164]10D-E). Additionally, we
   performed measurements of the NP area and histological scoring on gross
   specimens of discs from differentgroups. In the PBS group, the relative
   area of the NP progressively decreased over time, while histological
   scores progressively increased; however, BRD treatment reversed these
   changes (Fig. [165]10F–K). In addition to radiographic assessments,
   histological analysis was conducted to evaluate the efficacy of the
   different treatments. Specifically, rat disc sections were stained with
   HE and SF staining techniques. Four weeks after PBS treatment,
   compression-induced deformation of the NP was observed, leading to a
   reduction of approximately 20–25% in its area. The NP in the PBS group
   had almost completely disappeared, with the intervertebral space being
   replaced by chondrocytes extending inward. In contrast, the progression
   of IVDD was significantly alleviated in the BRD group (Fig. [166]11A).
   After SF staining, NP tissue appeared red due to collagen accumulation,
   while the AF tissue appeared blue due to fiber accumulation. In the PBS
   group, the NP tissue area ratio was the lowest compared to the control
   group, and the defective NP tissue was replaced by blue fibrous rings.
   In contrast, a larger red region remained in the BRD group, indicating
   collagen enrichment and suggesting a better therapeutic effect
   (Fig. [167]11B). The MMP13 immunohistochemistry results were as follows
   (Fig. [168]11C). In the control group, the NP exhibited a light brown
   color. Compared to the control group, the NP area was reduced in the
   PBS group, with increased histone staining intensity. In contrast to
   the PBS group, the BRD group showed reduced brown staining, indicating
   downregulation of MMP13. The COL-II immunohistochemistry results showed
   the strongest brown staining in the control group (Fig. [169]11D). In
   contrast, the PBS group displayed the lightest brown color. In the BRD
   group, the staining intensity was higher than that of the PBS group.
   The results of HE and SF staining, as well as the immunohistochemical
   scoring, were shown in the heatmap in Fig. [170]11E. Meanwhile, the
   expression level of ACSL4 in the BRD group was lower than that in the
   PBS group, while the expression levels of FTH1, GPX4, and SLC7A11 were
   higher than those in the PBS group, indicating that BRD can inhibit
   ferroptosis at the tissue level (Fig. [171]12A-D). The results of the
   immunohistochemistry scoring are presented as heat maps in
   Fig. [172]12E. In summary, these results indicated that BRD can
   alleviate the IVDD process in the puncture-induced rat model.

Fig. 10.

   [173]Fig. 10
   [174]Open in a new tab

   Imaging assessment and histologic evaluation after BRD-K20733377
   treatment in vivo. A X-ray images at 0 weeks, 6 weeks, and 10 weeks
   post-puncture in different groups. B Measurement of disc height index
   (DHI). C Quantitative analysis of DHI. D Magnetic resonance imaging
   (MRI) images at 0 weeks, 6 weeks, and 10 weeks post-puncture in
   different groups. E Heatmap of Pfirrmann grading in different groups.
   F–H Different groups of 6-week-old nucleus pulposus gross specimens,
   nucleus pulposus area, and histological score. (I-K) Different groups
   of 10-week-old nucleus pulposus gross specimens, nucleus pulposus area,
   and histological score. n = 5, ** indicates P < 0.01, *** indicates
   P < 0.001

Fig. 11.

   [175]Fig. 11
   [176]Open in a new tab

   Imaging assessment and histologic evaluation after BRD-K20733377
   treatment in vivo. A Hematoxylin and eosin (HE) staining images of rat
   caudal spines at 6 and 10 weeks post-surgery (n = 5). Scale bar: 1 mm.
   B Safranin-O/Fast Green (SF) staining images at 6 and 10 weeks
   post-surgery (n = 5). Scale bar: 1 mm. C Immunohistochemical analysis
   of matrix metallopeptidase 13 (MMP13) at 6 and 10 weeks post-surgery
   (n = 5). Scale bar: 1 mm. D Immunohistochemical analysis of collagen II
   (COL-II) at 6 and 10 weeks post-surgery (n = 5). Scale bar: 1 mm. E
   Quantitative analysis of the data presented as heat maps, based on
   parameters such as the morphology and cellularity of the nucleus
   pulposus, the border between the anulus fibrosus and nucleus pulposus,
   the morphology of the anulus fibrosus, and the cellularity of the
   anulus fibrosus. Data are shown as mean ± standard deviation

Fig. 12.

   [177]Fig. 12
   [178]Open in a new tab

   Immunohistochemical analysis of ferroptosis-related proteins. A
   Immunohistochemical analysis of acyl-coa synthetase long chain family
   member 4 (ACSL4) at 6 and 10 weeks post-surgery (n = 5). Scale bar:
   1 mm. B Immunohistochemical analysis of ferritin heavy chain 1 (FTH1)
   at 6 and 10 weeks post-surgery (n = 5). Scale bar: 1 mm. C
   Immunohistochemical analysis of glutathione peroxidase 4 (GPX4) at 6
   and 10 weeks post-surgery (n = 5). Scale bar: 1 mm. D
   Immunohistochemical analysis of solute carrier family 7 member 11
   (SLC7A11) at 6 and 10 weeks post-surgery (n = 5). Scale bar: 1 mm. E
   Quantitative analysis of the data presented as heat maps, based on
   parameters such as the morphology and cellularity of the nucleus
   pulposus, the border between the anulus fibrosus and nucleus pulposus,
   the morphology of the anulus fibrosus, and the cellularity of the
   anulus fibrosus. Data are shown as mean ± standard deviation

Discussion

   IVDD is a major cause of low back pain, often leading to disability and
   imposing a significant economic burden on society today [[179]45].
   Despite progress in the treatment of IVDD in recent years, current
   therapies primarily involve nonsteroidal anti-inflammatory drugs
   (NSAIDs) and surgical treatments [[180]46]. Therefore, there is an
   urgent need for new therapeutic approaches. Increasing evidence
   suggests that ferroptosis plays a crucial role in the pathogenesis of
   IVDD [[181]47–[182]49]. In the present study, immunohistochemical
   analysis of ferroptosis-related proteins in intervertebral discs at
   various stages of degeneration demonstrated that it is a contributing
   factor to the degeneration process (Fig. [183]1), indicating the
   necessity of targeting ferroptosis as a treatment for IVDD.
   Furthermore, AI screening identified the compound BRD, which
   selectively targets senescent cells [[184]24]. However, no studies have
   yet explored the role of BRD in IVDD, and its molecular mechanism
   remains unclear. This study aims to investigate the effects of BRD on
   ferroptosis in IVDD and its potential mechanisms through bioinformatics
   and in vivo and in vitro experiments. In terms of research methods,
   this study utilizes systems pharmacology and bioinformatics analysis.
   It identifies the potential key biological targets and possible
   molecular pathways through which BRD regulates ferroptosis to
   counteract IVDD, and these results are validated through in vitro
   experiments. First, we predict the targets related to BRD, ferroptosis,
   and IVDD, and take the intersection. Using PPI networks and machine
   learning, we discover that STAT3 and NFKB1 are two biomarkers with
   diagnostic value in IVDD and are key regulators of BRD-K20733377’s
   inhibition of ferroptosis. Signal Transducer and Activator of
   Transcription 3 (STAT3) is an important transcription factor, belonging
   to the STAT (Signal Transducer and Activator ofTranscription) protein
   family. STAT3 plays a crucial role in various physiological and
   pathological processes by mediating signaling pathways triggered by
   cytokines and growth factors in cells [[185]50]. Studies have shown
   that inhibiting STAT3 activity induces ferroptosis in gastric cancer
   cells through lipid peroxidation and Fe^2+ accumulation [[186]51].
   Another study integrated differentially expressed genes in ulcerative
   colitis from the GEO database with ferroptosis-related genes in FerrDb,
   conducted bioinformatics analysis, and successfully screened out STAT3
   as a core gene related to ferroptosis. This suggests that STAT3 might
   serve as a potential biomarker for diagnosing and treating ulcerative
   colitis [[187]52]. Nuclear Factor Kappa B Subunit 1 (NFKB1) is a gene
   encoding a protein known as p50, a subunit of the NF-κB (nuclear factor
   kappa-light-chain-enhancer of activated B cells) family of
   transcription factors. NF-κB plays a critical role in regulating immune
   responses, inflammation, cell survival, and stress responses [[188]53].
   Studies have found that NFKB1 may regulate the expression of CYBB and
   HMOX1 in M1 microglia in spinal cord injury, thus establishing the link
   between microglial ferroptosis and neuroinflammation [[189]54]. Other
   studies suggest that ferroptosis plays an important role in the
   development of diabetic cardiomyopathy (DCM) and demonstrate that
   ubiquitin-specific protease (USP) activates NF-κB in DCM to promote
   ferroptosis [[190]55]. Luo et al. [[191]56] theoretically and
   experimentally verified that Lipocalin 2 (LCN2) promotes the molecular
   mechanism of ferroptosis in hypoxic-ischemic brain injury (HIBD) by
   activating the NF-κB/STAT3 signaling pathway.

   Molecular docking and molecular dynamics simulations were used to
   validate the molecular mechanism by which BRD inhibits ferroptosis to
   alleviate IVDD. Molecular docking studies indicated that BRD has a high
   affinity with the core targets. The molecular dynamics simulation
   results showed that the binding of BRD with the core targets is more
   stable, suggesting that STAT3 and NFKB1 proteins are indeed key players
   in the inhibition of ferroptosis and alleviation of IVDD by BRD.
   Therefore, we hypothesize that BRD may alleviate IVDD by inhibiting
   ferroptosis through the STAT3/NFKB1 axis, and further analysis will be
   conducted to determine whether there is a direct causal relationship
   with IVDD.

   MR is an epidemiological method based on genetic variation used to
   study causal relationships. It uses genes as instrumental variables to
   help infer causal relationships between exposure factors (such as diet,
   lifestyle, or biomarkers) and disease outcomes [[192]57]. Through
   Mendelian randomization analysis, we found a significant causal
   relationship between STAT3/NFKB1 and IVDD. This further validates our
   hypothesis that BRD influences IVDD through the STAT3/NFKB1 axis.

   To verify the effect of BRD in inhibiting ferroptosis and alleviating
   IVDD, we conducted in vitro cell experiments to study the inhibitory
   effect of BRD on ferroptosis. At the same time, we established a rat
   model of intervertebral disc degeneration and investigated the effects
   of BRD on rat behavior and degenerated intervertebral discs. Our
   results showed that BRD treatment significantly improved
   Erastin-induced ferroptosis in NPMSCs, as evidenced by an increase in
   mitochondrial fluorescence intensity, a decrease in cellular ROS
   levels, reduced FerroOrange and Liperfluo staining fluorescence,
   restoration of mitochondrial morphology, and a decrease in MDA content.
   These cellular ferroptosis indicators are consistent with existing
   studies, indicating that BRD inhibits ferroptosis at the cellular level
   [[193]58–[194]62]. Furthermore, we observed that at the protein level,
   Erastin induced an increase in ACSL4 and a decrease in GPX4, SLC7A11,
   and FTH1. After BRD treatment, the expression of ACSL4 was reduced,
   while the expression of GPX4, SLC7A11, and FTH1 was increased.
   Immunofluorescence results corroborated the Western blot results. These
   findings are consistent with the ferroptosis detection results in the
   literature [[195]56, [196]63–[197]65].

   Additionally, our study provides interesting insights into pain-related
   behaviors. Although the Von Frey test successfully demonstrated that
   BRD could increase the pain threshold at the surgical site, no
   significant differences were observed in other behavioral outcomes,
   consistent with previous studies [[198]38, [199]39, [200]66]. This may
   be attributed to the specific nature of tail IVD pain, which may not
   have a substantial impact on overall behavior, particularly considering
   the limited weight-bearing role of the tail. Furthermore, the acute
   degeneration model used may not fully replicate the complex environment
   required to simulate discogenic pain [[201]67]. At the same time, we
   found that BRD showed significant benefits in vivo. BRD was able to
   reduce the loss of disc cells and collagen, as well as inhibit
   extracellular matrix degradation. Although tail intervertebral discs
   provide feasibility, assessability, and repeatability, it is important
   to note that the tail puncture model used in this study has certain
   limitations. In fact, this model cannot fully replicate the complexity
   of chronic human IVDD, especially the differences in cell types (such
   as rat NPMSCs [[202]45]) and biomechanical factors. These differences
   may limit our comprehensive understanding of BRD.

   Recent studies have shown that ferroptosis is closely associated with
   IVDD, and inhibiting ferroptosis can effectively improve IVDD. To
   explore the potential clinical applications and translational
   development of BRD as a ferroptosis inhibitor, we analyzed its effects
   on NPMSCs. Notably, our results confirmed that BRD can inhibit
   ferroptosis in NPMSCs. In subsequent experiments, we will further
   investigate the specific mechanisms by which BRD regulates ferroptosis
   in NPMSCs. Additionally, the therapeutic effects of BRD in other
   ferroptosis-mediated diseases and its anti-ferroptosis activity also
   warrant further study.

   Although we have explored the potential role of BRD in IVDD, there are
   still some limitations. For example, the acute tail intervertebral disc
   degeneration rat model used in this study, while feasible, evaluable,
   and reproducible, cannot fully replicate the complexity of human
   chronic IVDD. The differences in cell types and biomechanical factors
   between humans and rats may limit our comprehensive understanding of
   BRD’s clinical application. Furthermore, the study was conducted with
   short-term experiments, which may not fully reveal the long-term
   effects and potential side effects of BRD. Future research should focus
   on the long-term impact of BRD on intervertebral disc degeneration and
   its potential resistance or side effects. Additionally, this study
   identified STAT3/NFKB1 as the potential targets of BRD through
   bioinformatics, but the detailed molecular mechanisms by which BRD
   specifically regulates these signaling pathways remain unclear. Future
   research could explore these mechanisms further using more refined
   molecular biology techniques. Finally, although this study provides
   preliminary evidence for the potential of BRD in IVDD, the translation
   of its application into preclinical research and human treatment still
   faces many challenges. These limitations suggest that future research
   should not only validate the current findings but also employ more
   comprehensive experimental designs and more complex models to better
   understand the potential of BRD in IVDD and its clinical application
   prospects.

Conclusion

   Our study for the first time reveals that BRD-K20733377 inhibits
   ferroptosis and alleviates IVDD through the STAT3/NFKB1 signaling axis.
   Furthermore, the anti-ferroptosis effect of BRD-K20733377 in
   Erastin-induced NPMSCs has been elucidated, and its protective effect
   in a rat IVDD model has been validated. Overall, our data suggest that
   BRD-K20733377 can serve as a new ferroptosis inhibitor with promising
   progress and translational value, providing a potential therapeutic
   approach for IVDD and other ferroptosis-mediated diseases.

Supplementary Information

   Below is the link to the electronic supplementary material.
   [203]Supplementary Material 1.^ (22.9KB, docx)

Acknowledgements