Bone Marrow Mesenchymal Stem Cell-Derived Extracellular Vesicles Carrying circ_0050205 Attenuate Intervertebral Disc Degeneration

Objective It has been reported that bone marrow mesenchymal stem cells (BMSCs) are a potential source of autologous stem cells to support the nucleus pulposus (NP) regeneration in intervertebral disc degeneration (IDD). Herein, we aim to study the mechanism underlying the effects of BMSC-derived extracellular vesicles (BMSC-EVs) on nucleus pulposus cells (NPCs) in IDD. Methods EVs were isolated from BMSCs. An IDD model was surgically established in C57BL/6J mice. NPCs were exposed to tBHP to establish an IDD cell model. RNA sequencing was performed to identify differentially expressed circRNAs in NP tissues harvested from mice with IDD. Interactions among circ_0050205, miR-665, and GPX4 were validated, and different interventions were used to study the roles of these molecules in NPC biological functions. Results BMSC-EVs promoted NPC survival and inhibited NPC apoptosis and extracellular matrix (ECM) degradation. circ_0050205 expression was downregulated in the NP tissues of IDD mice, and BMSC-EVs facilitated NPC survival and suppressed ECM degradation in NPCs by transferring circ_0050205. circ_0050205 sponged miR-665 and upregulated GPX4 expression. BMSC-EVs expressing circ_0050205 promoted NPC survival-inhibited ECM degradation in NPCs and alleviated IDD in mice via the miR-665/GPX4 axis. Conclusion In conclusion, BMSC-EVs promoted NPC survival-inhibited ECM degradation in NPCs and attenuated IDD progression via the circ_0050205/miR-665/GPX4 axis.


Introduction
Intervertebral disc degeneration (IDD) is considered a major contributor to lower back pain, and even disability, worldwide [1]. IDD results from excessive destruction of the extracellular matrix (ECM) due to an imbalance between catabolism and anabolism in discs, and IDD represents an abnormal and cellmediated response to progressive structural failure [2,3]. Nucleus pulposus cells (NPCs), the primary cell type constituting the nucleus pulposus (NP), bear great responsibility in the generation of extracellular matrix (ECM) molecules [4]. Importantly, NPC dysfunction and ECM degradation are the major causes of IDD [5,6]. Furthermore, oxidative stress and ferroptosis in the NP play crucial roles in the pathogenesis of IDD [7,8].
Recently, cell-based strategies have been considered in attempts to regenerate NP tissues to affect IDD [9]. Specifically, bone marrow mesenchymal stem cells (BMSCs) have been applied in rabbit and pig models of IDD as BMSC transplantation has the ability to facilitate IVD regeneration [10,11]. It has been reported that MSCs possess their therapeutic potentials by paracrine signaling through releasing bioactive factors and soluble peptides [12]. Extracellular vesicles (EVs), nanosized vesicles (30 to 100 nm in diameter), are produced by the MSCs and act importantly in cell-cell communication by carrying various biomolecules, including DNAs, lipids, proteins, and noncoding RNAs; EVs have the ability to drive regenerative processes in many diseases [13].
Circular RNAs (circRNAs), a newly identified group of endogenous noncoding RNAs, are involved in NPC functions and ECM synthesis or degradation [14]. Interestingly, cir-cRNAs have been shown to be enriched and expressed at stable levels in EVs [15], and thus, circRNAs mediate the effects of EVs on IDD [16]. There is increasing interest in investigations of a new paradigm for understanding the regulatory mechanisms of ncRNAs, including circRNAs, microRNAs (miRNAs), and lncRNAs, termed the competitive endogenous RNA (ceRNA) hypothesis. The circRNA acts as ceRNAs or natural miRNA sponges and therefore results in increased expression of the corresponding miR target genes [14,17]. Based on the aforementioned evidence, we are interested in further understanding the circRNA-based mechanism underlying the paracrine effects of BMSCs on NPCs. In this study, we first performed RNA sequencing to identify differentially expressed cir-cRNAs in the NP tissues harvested from the established IDD mouse models. EVs were then isolated from mouse BMSCs (mBMSCs) and cocultured with the NPCs exposed to tBHP to evaluate the effect of BMSC-derived EVs (BMSC-EVs) on the NPC biological functions. Next, we investigated the transfer of a candidate circRNA from BMSCs to NPCs via EVs and further constructed a ceRNA regulatory network related to the function of this candidate circRNA in IDD. This study will reveal the molecular mechanism of IDD and offers a new target for the treatment of IDD.

Ethics Statement.
Animal experiments were ratified by the animal ethics committee of Tongji Hospital, Tongji Medical College of Huazhong University of Science and Technology.

IDD Mouse Model Establishment.
A total of 40 C57BL/6J male mice (age: 12 weeks, weight: 18-22 g) were procured from the Beijing Institute of Pharmacology and Toxicology, Chinese Academy of Medical Sciences (Beijing, China). All mice were housed in a specific pathogen-free-grade environment with conditions of 60-65% humidity, at 22-25°C under a12 h light/dark cycle (eat and drink freely). After 7 days of acclimation, the IDD model was prepared in C57BL/6J mice by annulus fibrosis (AF) needle puncture as previously reported [18]. First, the mice were administered with ketamine (100 mg/kg) for general anesthesia. A sagittal small skin incision was made from the Co6 to Co8 coccygeal discs to locate the disc position to insert the needles in the tail. Next, a puncture was made vertically with a syringe needle at the Co6-Co7 disc, followed by 180°axial rotation; then, the needle position was maintained for 10 s. A puncture was made parallel to the endplates with a 31-G needle through the AF into the nucleus pulposus, and the needle was inserted 1.5 mm into the disc to depressurize the nucleus. Other undisturbed segments were considered controls. Six weeks after surgery, the Co6-Co7 disc was harvested for further investigation.
The mice were randomized into a sham group (mice received sham operation), aIDD group (mice received IDD model establishment), a BMSC-EV + vector group (IDD mice injected with EVs from vector-transduced BMSCs), a BMSC-EV + LV-circ_0050205 + sh-NC group (IDD mice injected with EVs from BMSCs transduced with adenovirus carrying LV-circ_0050205 + sh-NC), and a BMSC-EV + LV-circ_0050205 + sh-GPX4 group (IDD mice injected with EVs from BMSCs transduced with adenovirus carrying LV-circ_ 0050205 + sh-GPX4) (8 mice in each group). Adenovirus (10 μL, 5 × 10 9 PFU, Sangon Biotech, Shanghai, China) was immediately injected into the puncture site. A week later, the mice (except the sham-operated mice) were injected with EVs (3 μg/μL in 100 μL) via the caudal vein once a week for 6 consecutive weeks. Finally, the mice were euthanized and the NP tissues were harvested and isolated for further experiments.
2.3. Radiographic Analysis. Sedated mice underwent X-ray after IVD puncture and treatment. X-ray was started in a 21 lp/mm detector which provided over 5x geometric magnification (Faxitron VersaVision; Faxitron Bioptics LLC, Tucson, AZ. USA). The disc height indices (DHIs) were calculated by three spinal surgeons who were blinded to the treatment utilizing the following formula: DHI = IVD height/adjacent IVD body height [19].

Histological
Examinations. NP tissues were rinsed in normal saline, fixed with 4% paraformaldehyde (30-50 min), dehydrated, cleared, paraffin embedded, and sliced. The slices were mounted onto slides and then heated at 45°C. For hematoxylin-eosin (HE) staining, after deparaffinization and dehydration with gradient alcohol, the slides were stained with hematoxylin for 5 min. Then, 1% hydrochloric acid ethanol was added and incubated to differentiate the slices for 3 s. The slices were then incubated in 5% eosin for 3 min, dehydrated with gradient ethanol, and cleared in xylene. For oil red O staining, after deparaffinization and dehydration with gradient alcohol, the slides were stained with hematoxylin for 5 min and hydrolyzed in 1% hydrochloric acid ethanol for 15 s. The slices were stained with aqueous solutions of 0.02% malachite green for 3 min and rapidly washed with 1% ethylic acid, followed by staining with 0.1% oil red O for 3 min. The sections were dehydrated with gradient alcohol and permeabilized with xylene. Finally, the slides were observed under an optical microscope and 10 random views were selected per slide.

Quality Control and
Comparison. The quality of the paired-end reads of the raw sequencing data was examined using FastQC software. The raw sequencing data was preprocessed by Cutadapt software 1.18 to remove the Illumina sequencing adaptors and poly (A) tailing sequences. Reads with N concentration > 5% were removed by the Perl Script and those with 70% base mass > 20% were identified using the FASTX Toolkit software (v0.0.13). Both ends were repaired using the BBMap software. The high-quality fragments of the reads were mapped onto the human reference genome for comparison.
2.13. Cell Viability Assay. NPCs were seeded at a density of 5 × 10 3 cells/well in a 96-well plate. Upon reaching 60% confluence, the cells were exposed to different doses of tBHP and cultured for 6 or 12 h; alternatively, after 12 h of transfection, the cells were exposed to different doses of tBHP for 6, 12, 24, 48, and 72 h. Then, 10 μL CCK-8 (K1018, APExBIO) was added to each well, followed by 2 h of incubation at 37°C. The absorbance (at 450 nm) was determined by a Multiskan FC reader (51119080, Thermo Fisher Scientific) to quantify cell viability.
2.14. Cell Proliferation Assay. NPCs were seeded at a density of 5 × 10 3 cells/well in a 96-well plate and treated with 10 μM EdU (COO81L, Beyotime) for 24 h. After fixation with paraformaldehyde, the cells were stained with Hoechst 33342. Images were captured using a fluorescence microscope (Thermo Fisher Scientific) at excitation wavelengths of 350 nm and 550 nm. The numbers of EdU-positive NPCs were counted to evaluate proliferation. The EdU staining positive rate = EdU − stained positive cells/total cells × 100%: 2.15. Cell Apoptosis Assay. Transfected and tBHP-exposed NPCs were incubated with a mixture (100 μg/mL) of 5 μL Annexin V-APC and 1 μL PI for 15 min according to the kit instructions (Keygen Biotech, China). Apoptotic cells were examined using the FACScan flow system (Becton Dickinson, San Diego, CA) and analyzed employing FlowJo software.

ROS Production Measurement. ROS production in cells was detected by the ROS Detection Kit (CA1410, Beijing
Solarbio Science & Technology Co. Ltd., Beijing, China). NPCs were incubated with 1 mL diluted dichlorodihydrofluorescein diacetate (DCF-DA) for 20 min and rinsed with serum-free medium three times to remove the excess DCF-DA that failed to enter the cells. A confocal microscope (Olympus FV10C-W3) was used to observe the cells. The fluorescence intensity was quantified at 485 nm (excitation wavelength) and 535 nm (emission wavelength) using a SpectraMax M5 reader.
2.18. Measurement of Labile Iron Levels. The iron levels were determined using a commercial kit (ab83366, Abcam). Briefly, NPCs were homogenized in a 5× volume of iron assay buffer on ice, followed by centrifugation at 13000 × g and 4°C for 10 min. The supernatant was incubated with 5 μL of an iron reducer for 30 min at 37°C, followed by incubation with 100 μL of an iron probe for 60 min at 37°C in the dark. The absorbance value was quantified at a wavelength of 593 nm utilizing a microplate reader (Thermo Fisher Scientific).
2.20. TEM. NPCs were centrifuged at 1000 × g for 15 min. The precipitates were fixed with PBS supplemented with 2.5% glutaraldehyde for 2 h and incubated with 1% osmium tetroxide for 2 h. Then, the precipitates were dehydrated with gradient alcohol and embedded with Epon 812. After ultrathin sectioning, the sections were stained with uranyl acetate and lead citrate and examined using a TEM (Tecnai G2 12, FEI Company, Netherlands).

2.22
. RNA Immunoprecipitation (RIP) Assay. The binding of miR-665 and circ_0050205 to the Ago2 protein was examined using the RIP kit (Millipore, Billerica, MA). Briefly, mononuclear cells were washed, lysed by incubation with RIPA buffer (P0013B, Beyotime) for 5 min, and centrifuged (7000 × g, 10 min) at 4°C. One part of the cell extract was used as input, 4 Oxidative Medicine and Cellular Longevity and another part was used for immunoprecipitation. Briefly, 50 μL magnetic beads were resuspended in 100 μL RIP wash buffer and incubated with 5 μg antibody (rabbit anti-human Ago2, ab186733, 1 : 50, Abcam, or rabbit anti-human IgG, ab109489, 1 : 100, Abcam) to allow binding. Following washing, the bead-antibody complex was resuspended in 900 μL RIP wash buffer and incubated with 100 μL cell extract overnight. Then, the bead-protein complex was collected. The immunoprecipitate and input were digested with proteinase K and subjected to RNA extraction and PCR analysis.
2.23. RNA Pulldown Assay. Bio-miR-665-WT and Bio-miR-665-MUT (GeneCreate, Wuhan, China) labeled with 50 nM biotin were used to transfect mononuclear cells. After 48 h, the mononuclear cells were collected and incubated with specific lysis buffer (Ambion, Austin, Texas) for 10 min. The cell lysates were incubated with M-280 streptavidin magnetic beads (S3762, Sigma-Aldrich) precoated with RNase-free BSA and yeast tRNA (TRNABAK-RO, Sigma-Aldrich). After washing with prechilled lysis buffer twice, low-salt buffer three times, and high-salt buffer once, the RNA was purified with TRIzol and analyzed by PCRto examine the enrichment of circ_ 0050205 and GPX4.

Real-Time Quantitative Polymerase Chain Reaction
(RT-qPCR). Following total RNA extraction using TRIzol reagent (16096020, Thermo Fisher Scientific), circRNA and mRNA were used to synthesize cDNA with an RT kit (RR047A, Takara, Japan). A PolyA Tail Detection Kit (B532451, Sangon) was used to generate cDNA from miRNA. RT-qPCR was performed using the SYBR® Premix Ex Taq™ II kit (DRR081, TaKaRa) and the PCR instrument (ABI 7500, ABI, Foster City, CA). U6 was used to normalize miRNA expression, and GAPDH was used to normalize circRNA and mRNA expression. The primer sequences are listed in Table S1. The results were calculated using the 2 −△△CT method.

Statistical
Analysis. Measurement data represented the results of at least three independent experiments and were described as mean ± standard deviation. Unpaired t-test was adopted for comparison between the two groups; oneway or two-way ANOVA was employed for data comparison among multiple groups and multiple time points, with the Tukey's test used for the post hoc test. A P value < 0.05 indicated significant difference. All statistical analyses were started utilizing SPSS 21.0 statistical software (IBM, Armonk, NY).

Ferroptosis Was Associated with NPC Viability Reduction and Extracellular Matrix (ECM) Degradation
Caused by Oxidative Stress (OS). Previous evidence has shown that superoxide-triggered cell death is related to ferroptosis and OS and ECM degradation are involved in the pathogenesis of IDD [29][30][31]. Here, we were interested in whether ferroptosis participated in the NPC death and ECM degradation observed in IDD. For this purpose, we used tBHP to treat NPCs and performed CCK-8 and EdU assays to examine NPC proliferation. As shown in Supplementary  Figures 1(a) and 1(b), NPC proliferation decreased in a concentration-dependent manner when the NPCs were exposed to 25, 50, and 100 μM of tBHP for 6 h or 12 h. A concentration-dependent increase was also noted in NPC apoptosis after exposure to different concentrations of tBHP (Supplementary Figure 1c). In addition, increased expression of two catabolic markers MMP13 and ADAMTS 5 and decreased expression of two anabolic markers COL II and aggrecan were observed in NPCs exposed to tBHP stimulation (Supplementary Figure 1d). These results suggested that superoxide could lead to decreased NPC viability, NPC death, and ECM degradation. Ferroptosis is characterized by ROS production and free Fe 2+ augmentation [32]. As shown in Supplementary  Figure 1e-g, tBHP stimulation resulted in increased ROS production, lipid peroxidation, and total free iron in NPCs in a concentration-dependent manner. Subsequently, we treated NPCs with the ferroptosis-specific inhibitors Fer-1, Lip-1, and DFO. As shown by CCK-8 and EdU assays, Fer-1, Lip-1, or DFO treatment rescued the tBHP-induced decrease in NPC viability (Supplementary Figure 1h Figure 1k) demonstrated that treatment of ferroptosis-specific inhibitors weakened the effects of tBHP stimulation on NPC apoptosis and expression of catabolic and anabolic markers. Similarly, tBHP-induced increase in the ROS production, lipid peroxidation, and total free iron was abrogated in NPCs upon treatment of ferroptosis-specific inhibitors ( Supplementary Figure 1l-n). TEM observation showed that tBHP stimulation led to mitochondrial damage, such as mitochondrial shrinkage, mitochondrial crista disintegration, increased mitochondrial membrane density, and ferroptotic characteristics (Supplementary Figure 1o). These data showed that ferroptosis was involved in OS and ECM degradation, which was associated with the pathogenesis of IDD. 5 Oxidative Medicine and Cellular Longevity

BMSC-EVs Promoted the NPC Survival and Suppressed ECM Degradation in NPCs.
BMSCs have been reported to prevent IDD progression via the transfer of EVs [33,34]. In addition, BMSCs have been found to inhibit ferroptosis [35]. Therefore, we hypothesized that BMSC-EVs could inhibit IDD progression by augmenting survival and reducing ECM degradation in NPCs. Observation results of the morphology of BMSCs under a light microscope showed that BMSCs have a long-shuttle shape or are spindle shaped (Supplementary Figure 2A). After differentiation induction under different conditions, red calcium nodules, red lipids contained in cells, and blue collagen staining appeared (Supplementary Figure 2B), indicating that BMSCs had theosteogenic, adipogenic, and chondrogenic differentiation potentials. In addition, Figure 1(a) reflects the characterization of BMSC-EVs by TEM observation and Figure 1(b) shows the size of the BMSC-EVs, with diameters ranging from 30 to 150 nm. As shown in Figure 1(c), CD63 and TSG101 were expressed in the BMSC-EVs while calnexin was not expressed. The abovementioned results indicated the successful EV extraction. Microscopic observation demonstrated NPC uptake of PKH67-labeled BMSC-EVs (Figure 1(d)).
Next, BMSC-EVs were cocultured with the 50 μM tBHPexposed NPCs for 12 h. The results of CCK-8 assay (Figure 1(e)), EdU staining (Figure 1(f)), and flow cytometric analysis (Figure 1(g)) showed that BMSC-EVs enhanced NPC viability and proliferation but reduced apoptosis under tBHP stimulation. Immunofluorescence detection demonstrated decreased expression of MMP13 and ADAMTS 5 and elevated expression of COL II and aggrecan in tBHPexposed NPCs following the addition of BMSC-EVs (Figure 1(h)). As expected, a significant decrease was noted in ROS production, lipid peroxidation, and total free iron in tBHP-exposed NPCs following the addition of BMSC-EVs (Figures 1(i)-1(k)). These data proved that BMSC-EVs favored NPC survival and repressed ECM degradation in NPCs.

BMSC-EVs Carried circ_0050205 to Enhance NPC Survival and Inhibit ECM Degradation in NPCs.
To study the mechanism underlying the effects of BMSC-EVs on NPC survival and ECM degradation, we performed RNA sequencing, with screened 337 upregulated circRNAs and 283 downregulated circRNAs in the NP tissues collected from mice with IDD ( Supplementary Figure 3a-b). Among these circRNAs, circ_0050205 exhibited the largest fold change in expression. The results of RT-qPCR showed that circ_0050205 expression was downregulated in the NP tissues of IDD mice compared to that of sham-operated mice (Figure 2(a)). Similarly, we observed a declined expression of circ_0050205 in NPCs after tBHP stimulation (Figure 2(b)).
Next, BMSC-EVs expressing Cy3-circ_0050205 were allowed to coculture with NPCs. As depicted in Figure 2(c), Cy3-circ_0050205 was observed in the NPCs, suggesting that circ_0050205 was transferred from BMSCs into NPCs via EVs. RT-qPCR revealed an elevated expression of circ_0050205 in tBHP-exposed NPCs after coculture with BMSC-EVs (Figure 2(d)). These data suggested that circ_0050205 can be transferred from BMSCs into NPCs via EVs.
Further investigation was performed to investigate whether circ_0050205 mediated the effects of BMSC-EVs on NPC survival and ECM degradation in NPCs. We generated circ_0050205-overexpressing BMSCs by the delivery of the circ_0050205 expression vector (Figure 2(e)). Then, we isolated EVs from circ_0050205-overexpressing BMSCs and cocultured them with NPCs in the presence of tBHP. RT-qPCR results demonstrated elevated expression of circ_ 0050205 in tBHP-exposed NPCs cocultured with EVs from circ_0050205-overexpressing BMSCs (Figure 2(f)). Results of CCK-8 assay (Figure 2(g)), EdU staining (Figure 2(h)), and flow cytometric analysis (Figure 2(i)) showed that EVs derived from circ_0050205-overexpressing BMSCs could promote the viability and proliferation and inhibit the apoptosis of tBHPexposed NPCs. Immunofluorescence detection demonstrated decreased expression of MMP13 and ADAMTS 5 yet elevated expression of COL II and aggrecan in tBHP-exposed NPCs cocultured with EVs derived from circ_0050205-overexpressing BMSCs (Figure 2(j)). We also found a significant decrease in ROS production, lipid peroxidation production, and total free iron in tBHP-exposed NPCs cocultured with EVs derived from circ_0050205-overexpressing BMSCs (Figures 2(k)-2(m)). These results collectively indicated that BMSC-EVs facilitated NPC survival and inhibited ECM degradation in NPCs by transferring circ_0050205.

circ_0050205 Sponged miR-665 to Upregulate GPX4
Expression. Subsequently, we performed a computer-based prediction of interactions between circ_0050205 and its downstream miRNAs in the Circbank and CircInteractome databases. These two databases yielded 4 common downstream miRNAs (miR-665, miR-1248, miR-1272, and miR-1265) of circ_0050205 (Figures 3(a) and 3(b)). Considering the upregulated miR-665 expression observed in the context of IDD from previous evidence [36], miR-665 was selected for further investigation. We also observed increased miR-665 expression in the NP tissues of IDD mice compared to sham-operated mice (Figure 3(c)). Similarly, upregulated miR-665 expression was observed in tBHP-exposed NPCs (Figure 3(d)). Luciferase assay demonstrated reduced luciferase activity of circ_0050205-WT after miR-665 mimic transfection, while the luciferase activity of circ_0050205-Mut did not change (Figure 3(e)). In addition, RIP assay demonstrated significant enrichments of circ_0050205 and miR-665 in the immunoprecipitates using anti-Ago2 compared with IgG (Figure 3(f)). Meanwhile, RNA pulldown assay demonstrated that more miR-665 was pulled down utilizing Bio-circ_0050205-WT than Bio-circ_0050205-Mut (Figure 3(g)). The interaction between circ_0050205 and miR-665 was further confirmed by RT-qPCR. We delivered the circ_0050205 expression vector or sh-circ_0050205 into NPCs and constructed circ_0050205-overexpressing or circ_0050205-knockdown NPCs (Figure 3(h)). Decreased expression of miR-665 was observed in tBHP-treated NPCs after circ_0050205 overexpression, while the opposite trend was observed aftercirc_0050205 knockdown (Figure 3(i)). 6 Oxidative Medicine and Cellular Longevity   Oxidative Medicine and Cellular Longevity Next, computer-based miRNA-mRNA prediction was performed with the RNA22 database and the miR-665 binding sites in the GPX4 3'UTR were found (Supplementary Figure 3c). The KEGG pathway analysis found that GPX4 was part of the ferroptosis regulatory network (Supplementary Figure 3d). As shown in Figures 4(a) and 4(b), GPX4 was poorly expressed in the NP tissues of IDD mice and tBHP-exposed NPCs. Pearson correlation analysis demonstrated a negative correlation between miR-665 expression and GPX4 expression in the NP tissues of IDD mice (Figure 4(c)). Results of luciferase assay found that the luciferase activity of GPX4-WT was reduced after miR-665 mimic transfection, while there was no alteration in the luciferase activity of GPX4-Mut (Figure 4(d)). RNA pulldown assay data demonstrated that Bio-miR-665-WT bound more GPX4 than Bio-miR-665-Mut (Figure 4(e)). To further study whether GPX4 was a target gene of miR-665, we treated NPCs with the miR-665 mimic, inhibitor, and their corresponding NCs, followed by tBHP stimulation (Figure 4(f)). Elevated expression of miR-665 and declined expression of GPX4 were noted in tBHP-exposed NPCs upon miR-665 mimic transfection while opposing tendency was found in tBHP-exposed NPCs upon miR-665 inhibitor transfection (Figure 4(g)). These results suggested that GPX4 was a target gene of miR-665.
Finally, we aimed to study the regulatory network that includes circ_0050205, miR-665, and GPX4. We observed increased GPX4 expression in tBHP-exposed NPCs upon circ_0050205 overexpression. When the circ_0050205 expression vector and miR-665 mimic were concomitantly transfected into NPCs in the presence of tBHP, we observed a decrease in GPX4 expression compared with LV-circ_ 0050205 transfection alone (Figure 4(h)). These results collectively revealed that circ_0050205 functioned as a ceRNA to sponge miR-665 and then upregulated GPX4 expression.

circ_0050205
In BMSC-EVs Attenuated IDD Progression via the miR-665/GPX4 Axis. To confirm the therapeutic role of BMSC-EVs in IDD, IDD was surgically established in C57BL/6J mice. Then, IDD mice were treated with EVs derived from circ_0050205-overexpressing BMSCs and/or adenovirus containing sh-GPX4. X-ray examination (Figure 7(a)) 8 weeks after surgery showed that compared with sham-operated mice, IDD mice presented degenerative characteristics. When IDD mice were injected with BMSC-EVs or EVs derived from circ_0050205-overexpressing BMSCs, the degeneration was attenuated. When IDD mice were injected with EVs derived from circ_0050205-overexpressing BMSCs and adenovirus containing sh-GPX4, GPX4 knockdown negated the inhibitory effects of circ_0050205 overexpression on IDD.   16 Oxidative Medicine and Cellular Longevity Similarly, histological examinations by HE staining and oil red O staining confirmed the X-ray examination (Figures 7(b) and 7(c)). As shown in Figures 7(d) and 7(e), compared with sham-operated mice, IDD mice exhibited reduced immunohistochemical staining of COL II, aggrecan, and GPX4 concomitant with increased immunohistochemical staining of MMP13 and ADAMTS 5. After injection of BMSC-derived EVs or EVs derived from circ_0050205-overexpressing BMSCs, IDD mice showed increased immunohistochemical staining of COL II, aggrecan, and GPX4 with reduced immunohistochemical staining of MMP13 and ADAMTS 5. Similarly, injection of adenovirus containing sh-GPX4 negated the effects of EVs derived from circ_ 0050205-overexpressing BMSCs on the expression of these markers.
It was found that IDD mice exhibited increased total free iron compared with sham-operated mice. Injection of BMSC-derived EVs or EVs derived from circ_0050205-overexpressing BMSCs reduced total free iron in IDD mice, while injection of adenovirus containing sh-GPX4 negated the inhibitory effects of EVs derived from circ_0050205overexpressing BMSCs on total free iron (Figure 7(f)).

Discussion
MSC-based cell strategies have been developed to promote NPC regeneration in IDD. BMSCs secrete EVs that can shuttle their cargos, such as circRNAs, between cells. Many circRNAs act importantly in the pathogenesis of IDD through acting as sponges for miRNAs [37]. As the obtained findings demonstrated, circ_0050205, a circRNA of interest, could be transferred from BMSCs into NPCs via EVs and then increase NPC survival and inhibit ECM degradation by binding to miR-665 and upregulating GPX4 expression. EVs were isolated from BMSCs and added to NPC culture medium in the presence of tBHP. The promoting effects of BMSC-EVs on NPC survival were subsequently observed. Accumulating evidence shows that OS contributes to the pathogenesis of IDD [38] and ferroptosis is a form of cell death that is associated with OS and inflammation [39]. NPC ferroptosis is attracting more attention because of its role in contributing to the pathogenesis of disc degeneration. Consistent with our results, Sun et al. demonstrated that BMSCs exert protective effects on liver grafts donated after circulatory death by inhibiting hepatocyte ferroptosis [35]. In addition, Yang and his team provided evidence that EVs secreted by vascular endothelial cells inhibit glucocorticoidinduced osteoporosis by reducing ferritinophagy and inhibiting the ferroptosis of osteoblasts [40]. EVs derived from adipose-derived stem cells inhibit ferroptosis in mice with intracerebral hemorrhage [41]. In addition, we also found that BMSC-EVs could prevent ECM degradation in NPCs, as evidenced by decreased expression of two catabolic markers MMP13 and ADAMTS 5 concomitant with increased expression of two anabolic markers COL II and aggrecan; these results were consistent with another study [42]. Disc degeneration occurs due to an imbalance between catabolism and anabolism in discs, leading to excessive ECM degradation. EVs can affect the ECM, including matrix degradation, matrix protein crosslinking, and matrix calcification [43]. Zhang et al. found that NPCs with activated autophagy release EVs and prevent ECM destruction [44].
In the following experiments, we studied the mechanism underlying the promoting effects of BMSC-EVs on NPC survival and inhibitory effects on the ECM degradation in NPCs. Our RNA sequencing results identified circ_ 0050205, as it exhibited the largest fold change in expression between IDD tissues and normal tissues. We found that circ_0050205 mediated the promoting effects of BMSC-EVs on NPC survival and inhibitory effects on the ECM degradation in NPCs. In contrast to this favorable circRNA, Song et al. identified an unfavorable circRNA, circ_ 0000253, that was carried by EVs derived from NPCs of different degenerative grades and whose expression was upregulated in degenerative NPCs [16]. Other favorable circRNAs that protect the intervertebral disc from degeneration were identified. For example, circARL15 expression was downregulated in IDD and its overexpression prevented NPC apoptosis but promoted NPC proliferation [45]. Knockdown of circSNHG5 expression in chondrocytes reduced cell proliferation, which was followed by degradation of ECM [46]. A circRNA-related ceRNA network has been studied in IDD. In the present study, we also demonstrated that circ_ 0050205 functioned as a ceRNA to sponge miR-665 and then upregulated GPX4 expression. We found an increased  (e) Immunofluorescence detection of MMP13, ADAMTS 5, COL II, and aggrecan proteins in tBHP-exposed NPCs treated with LV-GPX4 (25 μm). (f) The ROS production was examined using the DCF-DA kit in tBHP-exposed NPCs treated with LV-GPX4. (g) Lipid peroxidation determined by C11-BODIPY581/591 in tBHP-exposed NPCs treated with LV-GPX4. (h) The total free iron in tBHPexposed NPCs treated with LV-GPX4. * P < 0:05 vs. tBHP + vector. Cell experiments were conducted three times independently. 19 Oxidative Medicine and Cellular Longevity  Figure 6: circ_0050205 in BMSC-EVs contributes to the survival of NPCs and inhibits ECM degradation in NPCs via the miR-665/GPX4 axis. (a) GPX4 expression in tBHP-exposed NPCs cocultured with BMSC-EVs-LV-circ_0050205 + sh-GPX4. (b) CCK-8 assay for viability of tBHP-exposed NPCs cocultured with BMSC-EVs-LV-circ_0050205 + sh-GPX4. (c) EdU staining for proliferation of tBHP-exposed NPCs cocultured with BMSC-EVs-LV-circ_0050205 + sh-GPX4. (d) Flow cytometric analysis of apoptosis of tBHP-exposed NPCs cocultured with BMSC-EVs-LV-circ_0050205 + sh-GPX4. (e) Immunofluorescence detection of MMP13, ADAMTS 5, COL II, and aggrecan proteins in tBHP-exposed NPCs cocultured with BMSC-EVs-LV-circ_0050205 + sh-GPX4 (25 μm). (f) The ROS production was examined using DCF-DA kit in tBHP-exposed NPCs cocultured with BMSC-EVs-LV-circ_0050205 + sh-GPX4. (g) Lipid peroxidation determined by C11-BODIPY581/591 in tBHP-exposed NPCs cocultured with BMSC-EVs-LV-circ_0050205 + sh-GPX4. (h) The total free iron in tBHP-exposed NPCs cocultured with BMSC-EVs-LV-circ_0050205 + sh-GPX4. * P < 0:05 vs. tBHP + BMSC-EVs + LV-circ_ 0050205 + sh-NC. Cell experiments were conducted three times independently. 21 Oxidative Medicine and Cellular Longevity  23 Oxidative Medicine and Cellular Longevity miR-665 expression in the NP tissues of IDD mice compared to sham-operated mice, which was consistent with other evidence [36]. It was reported that miR-665 silencing attenuated ischemia/reperfusion injury-induced ROS production and inhibited cardiomyocyte apoptosis during myocardial infarction [47]. GPX4 was first demonstrated to be a target gene of miR-665 in this study. The GPX4 is an antioxidant defense enzyme and belongs to the selenoprotein family, which could limit the activity of ferroptosis by reducing lipid peroxidation [48]. GPX4 knockdown in mice may contribute to the ferroptosis in kidney tubular cells, inducing acute renal failure, which was efficiently suppressed

24
Oxidative Medicine and Cellular Longevity by Fer-1 and DFO [49]. In our study, GPX4 was poorly expressed in the NP tissues of IDD mice and the IDD cell model, revealing that NPC ferroptosis possibly results from the lack of GPX4 function. More importantly, when we overexpressed GPX4 in NPCs, inhibited ferroptosis concomitant with reduced ECM degradation was noted. Similarly, a recent study has also verified the involvement of GPX4 in OS-induced NPC ferroptosis in IDD [8]. Furthermore, the involvement of GPX4 in the circRNA-related ceRNA network has been revealed in previous studies. Previous data demonstrated that silencing of circFNDC3B limited GPX4 expression and enhanced ROS, total iron level, and Fe 2+ in oral squamous cell carcinoma cells via the miR-520d-5p/ SLC7A11 axis [50]. In addition, Xu et al. showed that cir-cIL4R inhibited ferroptosis through the miR-541-3p/GPX4 axis [51]. For confirming the in vitro data, we surgically established IDD in mice and demonstrated the therapeutic effects of BMSC-EVs on IDD; additionally, we showed that the mechanism by which the circ_0050205/miR-665/GPX4 axis mediated the inhibitory effects of BMSC-EVs in vitro was also relevant for the effects BMSC-EVs in vivo.

Conclusions
Overall, these experimental data revealed that circ_0050205 was transferred via EVs from BMSCs into NPCs. BMSC-EVs carrying circ_0050205 promote NPC survival and attenuate IDD progression by regulating miR-665 and GPX4 expression (Figure 8). These data reveal a unique regulatory network associated with NPC survival, namely, the circ_0050205/miR-665/GPX4 network. circ_0050205 might be an advantageous target for improving MSC-based cell strategies for treating IDD.

Data Availability
The datasets generated and/or analyzed during the current study are available in the manuscript and supplementary materials.

Ethical Approval
All procedures and protocols done to the animals were approved by the committee of Tongji Hospital, Tongji Medical College of Huazhong University of Science and Technology.

Conflicts of Interest
The authors declare that they have no competing interests.