circ_0072464 Shuttled by Bone Mesenchymal Stem Cell-Secreted Extracellular Vesicles Inhibits Nucleus Pulposus Cell Ferroptosis to Relieve Intervertebral Disc Degeneration

Ferroptosis, as an iron-dependent form of necrotic cell death, has been reported to affect activities of nucleus pulposus cells (NPCs). However, its role in the pathogenesis of intervertebral disc degeneration (IDD) is largely unknown. Notably, our bioinformatics analysis predicted that circ_0072464 was downregulated in nucleus pulposus of IDD mice. Therefore, this study is aimed at clarifying the mechanisms of extracellular vesicle- (EV-) encapsulated circ_0072464 derived from bone marrow mesenchymal stem cells (BMSCs) in NPC ferroptosis in IDD. EVs were extracted from mouse BMSCs (BMSC-EVs) and then cocultured with IL-1β-induced NPCs, followed by detection of matrix synthesis, proliferation, and ferroptosis of NPCs based on gain- or loss-of-function experiments. It was found that the uptake of BMSC-EVs by NPCs alleviated IDD. circ_0072464 and NRF2 were downregulated, and miR-431 was upregulated in IDD. Mechanistically, circ_0072464 competitively bound to miR-431, which targeted and inhibited NRF2 expression. BMSC-derived EVs carrying circ_0072464 inhibited NPC ferroptosis to promote matrix synthesis and proliferation of NPCs by inhibiting miR-431 and upregulating NRF2. Besides, in vivo experiments also confirmed that BMSC-EVs alleviated intervertebral disc lesions in mice with IDD through the circ_0072464/miR-431/NRF2 axis. Collectively, BMSC-EV-loaded circ_0072464 inhibited NPC ferroptosis to relieve IDD via upregulation of miR-431-mediated NRF2, therefore providing a potential therapeutic target against IDD.


Introduction
Intervertebral disc degeneration (IDD), as a multifactormediated and age-related disease, is a universal chronic process that can lead to serious spinal symptoms, including low back pain [1]. The cellular and biochemical changes in the microenvironment of intervertebral disc (IVD) occur in the development of IDD, resulting in progressive functional and structural damage [2]. During IDD, mechanical loading may cause extracellular matrix damage, leading to abnormity in microenvironment in nucleus pulposus cells (NPCs) [3]. It has been reported that ferroptosis of NPCs is implicated during the process of IDD [4]. Ferroptosis is a newly studied mechanism of cell death with the essence presented as the disorder of cell lipid metabolism catalyzed by iron ions [5].
New data are accumulating on the beneficial effects of mesenchymal stem cell-(MSC-) derived extracellular vesicles (EVs) in various human diseases, including IDD [6]. EVs (exosomes and microvesicles), as membranous vesicles that are secreted by most cell types, are increasingly considered as important mediators of intercellular communication and biomarkers of diseases [7]. The pathophysiological role of EVs in IDD has been increasingly recognized [8]. Moreover, exosome-transported circular RNAs (circRNAs) have been demonstrated to exert effects on IDD [9]. circRNAs are a large class of endogenous noncoding RNAs, whose 5 ′ and 3 ′ ends are connected together to form a closed-loop structure, and play a vital role in various pivotal biological processes, including cell proliferation, apoptosis, and metabolism [10].
In addition, circRNAs, as a competitive endogenous RNA (ceRNA), competitively bind to microRNAs (miRNAs) to regulate the expression of miRNA target messenger RNA (mRNA), which serves as a potential biomarker for IDD treatment [11]. miRNAs, as small noncoding RNAs, affect some biological processes, such as apoptosis, by regulating the target gene expression [12]. Existing evidence has revealed that dysregulated miRNA expression is involved in the pathogenesis of IDD [12,13]. Our bioinformatics analysis prior to investigations revealed putative binding relation between circ_0072464 and miR-431. It is interesting to note that miR-431, negatively regulated by circSEMA4B, modulates the progression of IDD [14]. However, the mechanism by which BMSC-derived EV communication affects NPC ferroptosis in IDD involving the interplay between circ_0072464 and miR-431 is still poorly understood.
Hence, we hypothesized that the transfer of circ_ 0072464 via BMSC-derived EVs (BMSC-EVs) might affect NPC ferroptosis in IDD.

Ethics Statement.
The study protocols were approved by the Ethics Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology. All animal procedures were in line with the Guide for the Care and Use of Laboratory animals published by National Institutes of Health.
In addition, BMSCs were cultured in OriCell™ MSC osteogenic, adipogenic, or chondrogenic differentiation medium (Cyagen Biosciences Inc., Guangzhou, China), respectively. Next, Alizarin red staining was used to observe calcium deposition on the 14 th day of induction, oil red O staining to observe lipid droplets on the 21 st day of induction, and Alcian blue staining to observe extracellular matrix on the 28 th day of induction.

Isolation and Characterization of BMSC-EVs.
BMSCs were ultracentrifuged at 100000 × g at 4°C in the medium containing serum overnight to remove EVs. Next, BMSCs were cultured in conditioned medium (DMEM/F12+10% EV-free serum) for 48 h. The culture supernatant of BMSCs in logarithmic growth phase was collected to extract EVs. BMSCs were centrifuged (500 × g) for 10 min at 4°C to collect supernatant, which was repeated one time. BMSCs were then centrifuged (2000 × g, 30 min) to remove large vesicle, centrifuged (10000 × g, 20 min, 4°C) to remove cellular debris, filtered using a 0.22 μm filter, and centrifuged (110000 × g, 70 min) again. At last, precipitate was resuspended by PBS, centrifuged, resuspended again with 100 μL sterile PBS, and stored at -20°C.

Surgically Induced IDD Mouse
Model. Fifty C57BL/6 male mice (aged 8 weeks) provided by Beijing Vital River Laboratory Animal Technology Co., Ltd., (Beijing, China) were raised under specific pathogen-free conditions (22°C, 50%-55% humidity, and 12 h light/dark cycle). The specific procedures were performed as previously reported method [18]. The Co6-Co7 discs were collected at week 6 and 12 after surgery from the mice for subsequent experiments.
2.5. Safranin O Staining. The discs from mice were fixed in 10% neutral-buffered formalin for 1 week, decalcified in EDTA for 2 weeks, embedded by paraffin, and sliced into sections (5 μm). The sections were stained with eosin and safranin O-fast green. Then, the Olympus BX51 microscope (Olympus Center Valley, PA) was employed to analyze histological images. The cellularity and morphology of the annulus fibrosus (AF), nucleus pulposus (NP), and the border between the two structures were examined. Three pathologists scored independently according to the scoring criteria of histological sections [18].
2.6. X-Ray Examination. At weeks 6 and 12 after puncture, radiographs were conducted. The disc height index (DHI) was applied to evaluate IVD height. The control discs and the corresponding punctured discs were measured. The heights of disc and the adjacent vertebral body were measured on the midline, and 25% of the disc width on both sides was measured from the midline. DHI was presented as the mean value of the three measurements from the midline to the 50% of disc width boundary at the center divided by the average of the two adjacent vertebral body heights. CapitalBio Technology (Beijing, China) was adopted to generate and sequence the sequencing library with 5 μg RNA per sample. The detailed methods were in line with previously published study [19]. Finally, the paired-end sequencing was performed on an Illumina NextSeqCN500 sequencer.
2.9. Quality Control and Comparison. The quality of pairedend reads of the original sequencing data was checked using FastQC software v0.11.8. The Cutadapt software 1.18 was utilized to pretreat original data to remove Illumina sequencing connector and poly (A) tail sequence. Perl script was used to remove reads with N more than 5%. FASTX Toolkit software 0.0.13 was employed to extract reads with 70% base mass above 20. Double-ended sequence was repaired by the BBMap software. Finally, the filtered highquality read fragments were compared with the human reference genome by the bwa software (0.7.12).
2.10. Bioinformatics Data. The CIRI software [20] was adopted to identify reverse splicing events, and sequencing data that cannot be directly compared with the reference genome were used for subsequent circRNA analysis. The screened circRNA was annotated using circBase website. The differentially expressed circRNAs were analyzed by limma (v3.32.10) package of R software [21] with jlogFCj > 1 and p value < 0.05.
The Circinteractome database was utilized to predict the key circRNAs that regulate miRNA. The downstream target genes of candidate miRNA were predicted by RNA22 database and miRDB database. Next, IDD-related mRNA data GSE56081 was retrieved from GEO database. This microarray dataset included 5 IDD samples and 5 control samples. The differentially expressed mRNAs were screened with jlogFCj > 0:2 and p value < 0.01. The "limma" package of R language was utilized for differential analysis. The predicted results of RNA22 and miRDB database were intersected with downregulated mRNA in the IDD sample of GSE56081 to screen the target genes regulated by the downstream of candidate miRNA.
2.13. Uptake of EVs by NPCs. EVs were labeled with green fluorescent dye (PKH67; Sigma-Aldrich) and then cocultured with Hoechst-labeled NPCs at 37°C for 2 h and fixed 3 Oxidative Medicine and Cellular Longevity in 4% paraformaldehyde for 15 min. Observation of nitracellular green fluorescence was performed with the help of a fluorescence microscope (CX23, Olympus, Tokyo, Japan). Then, EV-stimulated receptor cells were collected, and circ_0072464 expression was determined by reverse transcription quantitative polymerase chain reaction (RT-qPCR).
2.14. Cell Counting Kit-8 (CCK8) and EdU Assays. Cell viability of NPCs was detected according to the instructions of CCK8 (Dojindo Molecular Technologies, Inc., Kumamoto, Japan). The OD values were measured at the wavelength of 450 nm.
EdU assay was carried out based on previous study [22]. The EdU positive rate was calculated: EdU positive rate = EdU − positive cells ðred cellsÞ/total Hoechst 33258 − positive cells ðblue cellsÞ.

RNA Fluorescent In Situ Hybridization (FISH).
After transduction with circ_0072464 and miR-431 expressing vectors, NPCs or NP tissue sections were subjected to FISH assay [23]. The IX-53 microscope (Olympus, Tokyo, Japan) was employed to obtain images.
2.17. BODIPY Staining. Lipid reactive oxygen species (ROS) levels were measured by C11-BODIPY 581/591 (Thermo Fisher, D3861). Specifically, the cells were incubated with 10 μM C11-BODIPY 581/591 at 37°C for 30 min. After washing with PBS three times, cells were detached with trypsin and resuspended in fresh PBS. A flow cytometer (BD Biosciences) was utilized to analyze fluorescence emission peaks. The peak from 590 nm to 510 nm is inversely proportional to the production of lipid ROS.

Measurement of Labile Iron
Contents. Based on the instructions, an iron assay kit (ab83366, Abcam) was applied for iron concentration detection. Next, cells were treated with various compounds, collected, and homogenized by 5 times the volume of iron assay solution on ice, followed by centrifugation (13,000 × g, 10 min) at 4°C to obtain the supernatant. Cells were incubated with iron reducer (5 μL) for 30 min at 37°C and then with iron probe (100 μL) for 60 min at 37°C in the dark. Absorbance value was measured using a microplate reader (Thermo Fisher) at the wavelength of 593 nm.
2.20. RT-qPCR. Total RNAs were extracted using the miR-Neasy Mini Kit (Qiagen, Venlo, The Netherlands). RNA was reversely transcribed using PrimeScipt RT Kit (Takara). miR-431 was reversely transcribed using specific tailing reaction kit (B532451, Sangon, Shanghai, China). RT-qPCR was carried out on the 7500 Sequence Detection System (ABI, Foster City, CA) using SYBR Premix Ex Taq (Takara). The 2 -ΔΔCt method was used to quantify relative expression levels of target genes. For circ_0072464, total RNAs were incubated with or without 3 U/μg of RNase R (Epicentre, San Diego, CA) at 37°C for 20 min, and the resultant RNA was subsequently purified using the RNeasy MinElute Cleanup Kit (Qiagen). The circRNA was amplified using the specific divergent primers for the back-splice junction of circ_ 0072464. The agarose gel electrophoresis and sequencing were conducted to detect the amplified products. U6 and β-actin served as internal reference for miRNA and mRNA, respectively. All primers are shown in Table S1.
2.21. Western Blot Analysis. Total protein was extracted using radioimmunoprecipitation assay (RIPA) lysis buffer (R0010, Solarbio Science & Technology Co., Ltd., Beijing, China) containing PMSF, and protein contents were measured using the Micro BCA Protein Assay Kit (Thermo Fisher). The lysates were centrifuged and separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis gels, electrotransferred onto a polyvinylidene fluoride membrane. The membranes were blocked and incubated with the specific primary antibody overnight at 4°C. After washing with TBST, the membrane was subjected to incubation with horseradish peroxidase-(HRP-) labeled anti-rabbit IgG 2.25. Statistical Analysis. All data were presented as mean ± standard deviation and analyzed by the SPSS 21.0 statistical software (IBM Corp., Armonk, NY). The normality and homogeneity of variance test were conducted. Comparisons of data between two groups were analyzed by independent sample t -test, and those among multiple groups were tested by one-way analysis of variance (ANOVA) or repeated measures ANOVA, followed by Tukey's post hoc test. p < 0:05 was regarded as statistically significant.

Results
3.1. BMSC-EVs Alleviated IDD. BMSCs were isolated from bone marrow, and EVs were extracted from BMSCs, both of which were well characterized. Isolated BMSCs under ordinary microscope showed fibrous aggregation growth, and Alizarin red staining, oil red O staining, and Alcian blue staining further confirmed that the isolated BMSCs had the potential to differentiate into osteoblasts, adipoblasts, and chondroblasts ( Figure S2A-B). These findings suggested that BMSCs were successfully isolated.
Next, TEM showed that EVs were irregularly circular (Figure 1(a)). In addition, NTA presented that the average diameter of EVs was 110 nm (Figure 1(b)). Western blot analysis exhibited that the levels of CD63, TSG101, and CD81 in EVs were higher than those in cell lysate, and there was no calnexin expression ( Figure 1(c)). It can be concluded that EVs were successfully extracted from BMSCs.
Subsequently, Alizarin red staining displayed that as mice grew older, the IVD of mice showed morphological disorder of NP and AF, which indicated that mice had IDD (Figure 1(d)). After successful establishment of IDD mouse models, safranin O staining displayed that compared with IDD mice at the 6 th week, the degree of IDD in mice aggravated significantly at the 12 th week, and compared with IDD mice, the degree of IDD in IDD mice injected with EVs decreased significantly at the 6 th or 12 th week (Figure 1(e)). MMP13 and Col II are critical factors reflecting matrix synthesis in the disc during IDD [24,25]. Immunofluorescence staining revealed that MMP13 level decreased and Col II level increased in IDD mice injected with EVs at the 12 th week (Figure 1(f)).
The obtained data suggested that BMSCs and BMSCs-EVs were successfully isolated, and BMSCs-EVs effectively relieved IVD lesions in IDD mice.

circ_0072464 Shuttled by BMSC-EVs Promoted Matrix
Synthesis and Proliferation of NPCs. Through sequencing (GSE189551) the NP tissues of IDD mice and control mice, 5 Oxidative Medicine and Cellular Longevity 233 upregulated and 317 downregulated circRNAs were obtained, and circ_0072464 expression decreased in the NP tissues from IDD mice (Figure 2(a)), suggesting that circ_ 0072464 may play an important role in the development of IDD.
RT-qPCR also exhibited that circ_0072464 was poorly expressed in NP tissues from IDD mice (Figure 2(b)), which was further verified in NPCs that circ_0072464 was also poorly expressed in IL-1β-treated NPCs (Figure 2(c)).
Next, the effects of BMSC-EVs carrying circ_0072464 on IDD were explored. Confocal microscope exhibited that BMSC-EVs were internalized by NPCs (Figure 2(d)). Next, BMSCs were transduced with sh-circ_0072464, from which EVs were extracted. RT-qPCR revealed that circ_0072464 expression reduced in BMSCs and EVs after circ_0072464 knockdown (Figure 2(e)). Subsequently, EVs were cocultured with NPCs. RT-qPCR displayed that circ_0072464 expression increased in NPCs cocultured with EVs, and after silencing of circ_0072464, circ_0072464 expression decreased in NPCs (Figure 2(f)). These results confirmed that BMSC-EVs delivered circ_0072464 into NPCs.
Moreover, the roles of BMSC-EVs carrying circ_0072464 in cellular function of NPCs were investigated. CCK8 and EdU assay revealed that cell proliferation of NPCs cocultured with EVs was promoted, yet inhibited after silencing of circ_ 0072464 (Figures 2(g) and 2(h)). Western blot analysis and immunofluorescence staining showed that levels of Col II and Aggrecan increased, and levels of MMP13 and ADAMTS-5   Oxidative Medicine and Cellular Longevity reduced in NPCs cocultured with EVs, while circ_0072464 knockdown exerted opposite effects (Figures 2(i) and 2(j)).
It can be concluded that BMSC-EVs delivered circ_ 0072464 into NPCs to promote matrix synthesis and proliferation of NPCs.

circ_0072464
Competitively Bound to miR-431. The downstream mechanism of circ_0072464 was further studied, and the downstream regulatory miRNAs of circ_ 0072464 were predicted by Circinteractome database, results of which presented that circ_0072464 targeted miR-431. In 9 Oxidative Medicine and Cellular Longevity addition, IDD-related mRNA expression profile GSE63492 was retrieved from GEO database for differential analysis. circ_0072464 and miR-431 are highly conservative in both humans and mice. The NCBI-BLAST comparative analysis indicated that circ_0072464, also known as circARL15, is encoded by ARL15, sequences of which had strong homology in both humans and mice ( Figure S3A). Similar results were also detected regarding miR-431 ( Figure S3B). The results showed that miR-431 was overexpressed in patients with IDD (Figures 3(a) and 3(b)). RT-qPCR also exhibited that miR-431 expression increased in NP tissues of IDD mice (Figure 3(c)), and in IL-1β-treated NPCs (Figure 3(d)).
Next, we moved to illuminate the roles of circ_0072464 in the regulation of miR-431. Bioinformatics website predicted that there was a binding site between circ_0072464 and miR-431 (Figure 3(e)), which was validated by dualluciferase reporter gene assay (Figure 3(f)). It was found that the luciferase activity of circ_0072464-WT was suppressed by restoring miR-431, while no evident difference was found regarding luciferase activity of circ_0072464-MUT. RNA pull-down and RIP assays confirmed the interaction between circ_0072464 and miR-431 (Figures 3(g) and  3(h)). In addition, RNA FISH assay showed that there was colocalization between circ_0072464 and miR-431 in NPCs (Figure 3(i)).

miR-431
Targeted and Inhibited NRF2. In order to further study the downstream mechanism, the target genes of miR-431 were predicted through RNA22 and miRDB databases. The predicted results of RNA22 and miRDB databases were intersected with the significantly downregulated mRNA in the IDD sample of GSE56081 (Figure 4(a)   13 Oxidative Medicine and Cellular Longevity NRF2 in the intersection. A recent study has also demonstrated that NRF2 expression gradually decreased with the occurrence of IDD [26]. It was noted from PCR and Western blot analysis that NRF2 was poorly expressed in NP tissues of IDD mice (Figures 4(b) and 4(c)) and IL-1β-treated NPCs (Figures 4(d) and 4(e)).
Whether miR-431 regulated NRF2 was further explored. There was a binding site between miR-431 and NRF2 through miRDB database (Figure 4(f)), which was verified by dual-luciferase reporter gene assay. The luciferase activity of NRF2-3′UTR-WT was inhibited by overexpressed miR-431, while no evident difference was found with regard to NRF2-3′UTR-MUT (Figure 4(g)). RNA pull-down and RIP assay confirmed that there was interaction between miR-431 and NRF2 (Figures 4(h) and 4(i)). In addition, miR-431 expression was elevated, and NRF2 expression was reduced in NPCs transfected with miR-431 mimic, while the results were opposite in NPCs transfected with miR-431 inhibitor (Figures 4(j) and 4(k)).
These findings concluded that miR-431 targeted and suppressed NRF2 expression.
3.5. circ_0072464 Upregulated NRF2 by Sponging miR-431 to Inhibit NPC Ferroptosis. It has been demonstrated that ferroptosis can promote the death of AF cells and NPCs to mediate oxidative stress response, thereby promoting IDD [27]. Therefore, we further studied whether circ_0072464 regulated ferroptosis by upregulating the miR-431/NRF2 axis. The expression of circ_0072464 and NRF2 was increased, and miR-431 expression was deceased in NPCs transduced with oe-circ_0072464, while further treatment of miR-431 mimic resulted in no obvious difference in circ_0072464 expression but elevated miR-431 and diminished NRF2 (Figures 5(a) and 5(b)).
As shown in Figures 5(c) and 5(d), cell proliferation was augmented upon circ_0072464 overexpression while it was reduced in the presence of further miR-431 overexpression. In addition, the results of Western blot analysis and immunofluorescence staining suggested an increase in the expression of Col II and Aggrecan yet a reduction in that of MMP13 and ADAMTS-5 in NPCs transduced with oe-circ_0072464 while opposite results were found in response to oe-circ_0072464+miR-431 mimic treatment (Figures 5(e) and 5(f)).
The effects of circ_0072464 regulating miR-431/NRF2 on NPC ferroptosis were explored. ACSL4 expression was elevated, and GPX4 expression was decreased in NP tissues of IDD mice ( Figure 5(g)) and IL-1β-treated NPCs ( Figure 5(h)). Next, NPCs were overexpressing circ_ 0072464 and/or sh-NRF2. It was observed that ACSL4 expression was reduced, and GPX4 expression was increased in NPCs overexpressing circ_0072464, while the results were opposite in NPCs transduced with sh-NRF2. In NPCs overexpressing circ_0072464, sh-NRF2 led to elevated ACSL4 expression and diminished GPX4 expression ( Figure 5(i)).
In addition, flow cytometry, TEM, Western blot analysis, and labile iron detection results (Figures 5(j)-5(m)) presented that the content of lipid ROS was decreased, the morphology of mitochondria was intact, and the levels of FTL and intracellular free iron were decreased in NPCs overexpressing circ_0072464. After silencing of NRF2, the content of lipid ROS in NPCs was increased, the mitochondria were damaged, and the levels of FTL and intracellular free iron were increased. Moreover, the content of lipid ROS in NPCs was increased, the mitochondria were seriously damaged, and the levels of FTL and intracellular free iron were obviously increased in NPCs in response to overexpression of circ_0072464 and silencing of NRF2.
These findings demonstrated that circ_0072464 promoted NPC proliferation and matrix synthesis while inhibiting NPC ferroptosis by sponging miR-431 and upregulating NRF2.
3.6. circ_0072464 Inhibited NPC Ferroptosis to Promote Matrix Synthesis and Proliferation of NPCs via miR-431/ NRF2. Furthermore, NPCs were treated with oe-circ_ 0072464 and/or Erastin to examine the effects of circ_ 0072464/miR-431/NRF2 axis on cell activities of NPCs. RT-qPCR and Western blot analysis showed that expression of circ_0072464, NRF2, and GPX4 was elevated, and expression of miR-431 and ACSL4 was reduced after overexpression of circ_0072464. Additional treatment of Erastin led to diminished GPX4 expression and elevated ACSL4 expression, but expression of circ_0072464, miR-431, and NRF2 displayed no evident difference. Relative to oe-circ_ 0072464 alone, GPX4 expression was decreased, and ACSL4 expression was increased upon cotreatment of oe-circ_ 0072464 and Erastin (Figures 6(a) and 6(b)).
Subsequently, CCK8 assay and EdU assay exhibited that overexpression of circ_0072464 enhanced NPC proliferation, and Erastin suppressed NPC proliferation. Besides, proliferation was repressed by Erastin in NPCs overexpressing circ_0072464 (Figures 6(c) and 6(d)). Western blot analysis and immunofluorescence staining showed that circ_0072464 overexpression elevated levels of Col II and Aggrecan and reduced levels of MMP13 and ADAMTS-5 in NPCs, while Erastin exerted the opposite effects. Simultaneous treatment with oe-circ_0072464 and Erastin decreased levels of Col II and Aggrecan and increased levels of MMP13 and ADAMTS-5 in NPCs in comparison to oe-circ_0072464 alone (Figures 6(e) and 6(f)). Thus, Erastininduced ferroptosis inhibited the facilitating role of circ_ 0072464 in matrix synthesis and proliferation of NPCs.
The obtained data confirmed that circ_0072464 repressed NPC ferroptosis to enhance matrix synthesis and proliferation of NPCs by regulating the miR-431/NRF2 axis.

circ_0072464 Delivered by BMSC-EVs Alleviated IVD
Lesions in IDD Mice through miR-431/NRF2. At last, whether circ_0072464-containing BMSC-EVs inhibit NPC ferroptosis to relieve IDD via the miR-431/NRF2 was vali-dated in vivo. IDD mice were injected with EVs and/or Erastin, and NP tissues were isolated for further experiments.
We found elevations in expression of circ_0072464, NRF2, and GPX4 and reductions in expression of miR-431 and ACSL4 in NP tissues of IDD mice injected with EVs, while GPX4 expression reduced and ACSL4 expression elevated in NP tissues of IDD mice injected with Erastin. Compared with IDD mice injected with EVs, GPX4 expression decreased and ACSL4 expression increased in NP tissues of IDD mice injected with EVs+Erastin (Figures 7(a)-7(c)).
Safranin O staining and X-ray exhibited that the degree of morphological disorder of NP and AF and the degeneration of IVD were reduced in IDD mice injected with EVs, while the results were opposite in IDD mice injected with Erastin. Compared with IDD mice injected with EVs, the degree of morphological disorder and degeneration of NP and AF were also aggravated in IDD mice injected with EVs+Erastin (Figures 7(d) and 7(e)). In addition, immunofluorescence staining presented that MMP13 expression decreased and Col II expression increased in NP tissues of IDD mice injected with EVs, while Erastin exerted the opposite effects. Compared with IDD mice injected with EVs, MMP13 expression elevated and Col II expression decreased in NP tissues of IDD mice injected with EVs+Erastin (Figure 7(f)).
These findings demonstrated that the transfer of circ_ 0072464 by BMSC-EVs regulated the miR-431/NRF2 axis to repress NPC ferroptosis, thereby alleviating IVD lesions in IDD mice.

Discussion
In recent years, MSC-secreted EVs have emerged as a potential therapeutic strategy for IDD [28]. The data obtained in our study demonstrated that circ_0072464 shuttled by BMSC-derived EVs was transferred into NPCs, where circ_ 0072464 suppressed NPC ferroptosis to relieve IDD by regulating miR-431/NRF2 both in vitro and in vivo.
Evidence accumulates on the biological intervention of BMSCs in IDD, suggesting a potential for identifying suitable biotherapeutic targets [29]. EVs have been documented to be secreted by BMSCs, which provide therapeutic benefits for IDD [30]. Our findings confirmed that the transfer of BMSCs-EVs to NPCs alleviated IDD in mice, as reflected by decreased MMP13 level and increased Col II level. MMP13 and Col II are known to reflect the synthesis of IVD matrix during IDD [24,25,31]. Consistently, a recent study has also demonstrated that BMSC-derived EVs protect IVD against degeneration by reducing levels of expression of Col II and Aggrecan while inhibiting NPC apoptosis [15].
Moreover, we reported that circ_0072464 shuttled by BMSC-EVs promoted matrix synthesis and proliferation of NPCs through inhibiting NPC ferroptosis. NPCs are capable of inducing the generation of extracellular matrix for maintaining the hydration state of IVD [32]. The typical feature of IDD includes the degradation of NP and AF matrix components, which impairs the bearing performance of IVD [33]. Therefore, the alleviatory effect of circ_0072464 could be confirmed. In addition, our data also demonstrated that

18
Oxidative Medicine and Cellular Longevity to be upregulated in IDD and circARL15 acts as a ceRNA to reduce miR-431-5p expression, thus inhibiting NPC apoptosis but inducing NPC proliferation to alleviate IDD [35], providing validation of our finding.
Further mechanistic explorations in this study confirmed that miR-431 targeted and repressed NRF2, which was poorly expressed in IDD. Evidence exists reporting that NRF2 expression is decreased in NP tissues of patients IDD [26]. The important roles of NRF2 in alleviating lipid peroxidation and suppressing ferroptosis have also been verified [27]. Available evidence has indicated that ferroptosis of NPCs is implicated in pathogenesis of IDD, suggesting that targeting NPC ferroptosis may be a novel target for treating IDD [36]. Both in vitro and in vivo experiments in the present study confirmed that circ_0072464 shuttled by BMSC-secreted EVs inhibits NPC ferroptosis by inducing matrix synthesis and proliferation of NPCs to relieve IDD via downregulation of miR-431 and upregulation of NRF2.

19
Oxidative Medicine and Cellular Longevity To sum up, our study demonstrated that circ_0072464 transferred by BMSC-derived EVs upregulated NRF2 expression through competitive binding to miR-431, the mechanism of which led to inhibition of NPC ferroptosis and alleviation of IDD (Figure 8). Our findings pave way for the development of potential therapeutic strategies for inhibiting NPC ferroptosis to alleviate IDD. Due to the limited known researches on the functional role of circ_ 0072464, the underlying mechanism concerning circ_ 0072464 shuttled by BMSC-derived EVs, miR-431, and NRF2 should be more clearly investigated in the future. More importantly, some other differentially expressed cicrR-NAs in IDD were revealed from our RNA-seq results in addition to circ_0072464, highly suggestive of the need to identify the potential therapeutic implication of novel cicrR-NAs in IDD. Meanwhile, the administration and dosage of agents to inhibit ferroptosis process are urgently required to be determined to improve treatment efficacy against IDD.

Data Availability
The datasets generated for this study are available on request to the corresponding author.

Conflicts of Interest
The authors have declared that no competing interest exists.

Authors' Contributions
Conceptualization was contributed by XJY, HRX, and HK. Data curation was done by XJY and QKL. Formal analysis was performed by XJY, SXW, and YQJ. Investigation was carried out by QKL and YGW. Methodology was contributed by HRX and RL. Project administration was done by HK. Resources were contributed by HK. Software was contributed by HRX and SXW. Supervision was done by YQJ. Validation was carried out by YGW. Visualization was performed by XJY and RL. Writing-original draft was done by XJY, QKL, YGW, YGW, SXW, and RL. Writing-review and editing was done by HRX, YQJ, HK, and WHH. Revision was carried out by WHH, XJY, HRX, HK, SXW, YQJ, QKL, and YGW. Figure S1: isolation of NP tissues and identification of NPCs. Figure S2: identification of mouse BMSCs. Figure S3: circ_ 0072464 and miR-431 have strong homology in humans and mice. Table S1: primer sequences of RT-qPCR.