Abstract

   Intervertebral disc degeneration (IVDD) is a chronic degenerative
   disease involving the aging and loss of proliferative capacity of
   nucleus pulposus cells (NPCs), processes heavily dependent on
   mitochondrial dynamics and autophagic flux. This study finds that the
   absence of BCL2/adenovirus E1B 19 kDa interacting protein 3 (BNIP3) is
   associated with senescence‐related NPC degeneration, disrupting
   mitochondrial quality control. Bone marrow mesenchymal stem cells
   (BMSCs) have multidirectional differentiation potential and produce
   extracellular vesicles containing cellular activators. Therefore, in
   this study, BMSCs are induced under hypoxic stimulation to deliver
   BNIP3‐rich extracellular vesicles to NPCs, thereby alleviating
   aging‐associated mitochondrial autophagic flux, promoting damaged
   mitochondrial clearance, and restoring mitochondrial quality control.
   Mechanistically, BNIP3 is shown to interact with the membrane‐bound
   protein annexin A2 (ANXA2), enabling the liberation of the
   transcription factor EB (TFEB) from the ANXA2‐TFEB complex, promoting
   TFEB nuclear translocation, and regulating autophagy and lysosomal gene
   activation. Furthermore, a rat model of IVDD is established and
   verified the in vivo efficacy of the exosomes in repairing disc
   injuries, delaying NPC aging, and promoting extracellular matrix (ECM)
   synthesis. In summary, hypoxia‐induced BMSC exosomes deliver BNIP3‐rich
   vesicles to alleviate disc degeneration by activating the mitochondrial
   BNIP3/ANXA2/TFEB axis, providing a new target for IVDD treatment.

   Keywords: bone marrow mesenchymal stem cells, exosomes,
   hypoxia‐preconditioned mesenchymal stem cells, Intervertebral disc
   degeneration, matrix reconstruction, mitophagy
     __________________________________________________________________

   This study delves into the therapeutic efficacy of hypoxia‐induced bone
   marrow mesenchymal stem cell exosomes enriched with BCL2/adenovirus E1B
   19 kDa interacting protein 3 (BNIP3) in mitigating intervertebral disc
   degeneration (IVDD). Through mechanistic insights, this study explores
   the restoration of mitochondrial quality control and delays in nucleus
   pulposus cell aging, offering promising avenues for IVDD treatment.

   graphic file with name ADVS-11-2404275-g006.jpg

1. Introduction

   Lower back pain is a prevalent global health issue that significantly
   impacts individuals worldwide, leading to substantial healthcare costs
   and disability.^[ [62]^1 ^] From the pathophysiological perspective,
   intervertebral disc degeneration (IVDD) is a significant factor
   contributing to lower back pain.^[ [63]^2 ^] Anatomically, the
   intervertebral disc is composed of a central nucleus pulposus (NP), an
   outer annulus fibrosus (AF), and upper and lower cartilaginous
   endplates (CEP).^[ [64]^3 ^] Disc degeneration is attributed to a
   multitude of factors, including diminished water content within the NP,
   reduced elasticity of the aging AF, and ossification of the CEP.^[
   [65]^4 ^] Cellular senescence, inflammation, and damage to the
   mitochondria, essential organelles for cellular energy production and
   reactive oxygen species (ROS) generation, collectively contribute to
   reduced cell density and compromised extracellular matrix (ECM)
   degradation and synthesis in the NP.^[ [66]^5 ^] The current study
   aimed to investigate the pathological mechanisms underlying IVDD at the
   molecular level to identify novel targets for enhancing the treatment
   of this condition.

   Bone marrow‐derived mesenchymal stem cells (BMSCs) possess remarkable
   multilineage differentiation and cellular regeneration modulation
   abilities, as well as low immunogenicity. These qualities have made
   BMSCs a highly sought‐after cellular system for treating orthopedic
   disorders.^[ [67]^6 ^] Typically, stem cells possess the capacity to
   migrate to damaged tissues and repair impaired cells through various
   mechanisms such as direct differentiation, cell–cell interactions, and
   paracrine release of active cytokines.^[ [68]^7 ^] Theoretically, stem
   cell transplantation holds promise as an excellent option for cell
   repair and regeneration. However, the intercellular microenvironment,
   ischemia, inflammation, oxidative stress, limited homing ability, lack
   of targeted migration, and low post‐transplantation survival rates
   present significant challenges for stem cell therapy.^[ [69]^8 ^]
   Therefore, researchers are currently focusing on exploratory studies of
   stem cells.

   Extracellular vesicles, specifically exosomes, serve as vital modes of
   paracrine secretion from stem cells, enabling the delivery of parent
   cell characteristics and corresponding genes to recipient cells through
   noncellular contact mechanisms.^[ [70]^9 ^] Typically, in vitro
   experiments involve exposing BMSCs to normoxic conditions (21% O^2),
   which significantly deviate from the low‐oxygen microenvironment (2–8%)
   within the human body. Such culture conditions may be critical factors
   that influence the potential of mesenchymal stem cells.^[ [71]^10 ^]
   Hypoxia enhances the intrinsic activity of stem cells by reducing
   glucose consumption, lactate levels, and the release of cytochrome c.
   Additionally, extracellular vesicles derived from low‐oxygen‐treated
   cells exhibit promising therapeutic effects in various disease
   models.^[ [72]^11 ^] Therefore, we used hypoxic preconditioning to
   induce related changes in intracellular metabolism, enhancing the
   therapeutic efficacy of the resulting exosomes.^[ [73]^12 ^]

   BCL2/adenovirus E1B 19 kDa interacting protein 3 (BNIP3) is a
   mitochondria‐targeted protein that shares homology with BCL2
   interacting protein 3 like (BNIP3L) and potentially possesses a BH3
   domain. It is localized to the outer mitochondrial membrane.^[ [74]^13
   ^] The phosphorylation of serine residues in the LC3‐interacting region
   (LIR) of the N‐terminus of BNIP3 facilitates its interaction with
   LC3B.^[ [75]^14 ^] The C‐terminus of BNIP3 contains multiple
   phosphorylation sites, including TM, S/T, and RRLT sequences. Under
   oxidative stress or hypoxic conditions, complete phosphorylation of the
   BNIP3 C‐terminus attenuates its cytotoxic effects.^[ [76]^15 ^] BNIP3
   functions as a mitochondrial autophagy receptor activated under hypoxia
   and nutrient deprivation to maintain mitochondrial quality control and
   safeguard cell survival.^[ [77]^16 ^] Studies have demonstrated that
   BNIP3 plays a role in regulating mitochondrial autophagic flux and
   mitochondrial–lysosomal crosstalk.^[ [78]^4 , [79]^17 ^] However, the
   exact role and specific molecular mechanisms by which BNIP3 maintains
   intervertebral disc integrity and mitochondrial quality control in
   nucleus pulposus cells (NPCs) remain unclear. Therefore, in this study,
   we examined the regulatory network of BNIP3 in mitochondrial autophagy,
   which coordinates mitochondrial quality control and promotes the
   formation of autolysosomes to eliminate damaged mitochondria, thereby
   delaying senescence in NPCs (Scheme [80] 1 ).

Scheme 1.

   Scheme 1
   [81]Open in a new tab

   Revealing the exosome therapy strategy of hypoxia preconditioning and
   its specific molecular mechanisms in IVDD.

2. Results

2.1. Differential Expression of BNIP3 and Assessment of Mitochondrial Damage
in Degenerated NP Tissue

   The NP is centrally located within intervertebral disc tissue, and any
   damage to NPCs inevitably contributes to the development of IVDD. In
   addition, age‐associated mitochondrial dysfunction in NP tissues has
   been identified as a key factor in the irreversible degeneration caused
   by IVDD. In this study, we assessed the severity of IVDD using the
   Pfirrmann grading system based on magnetic resonance imaging (MRI;
   Table [82]S1, Supporting Information). We collected 16 NP samples from
   different patients with IVDD, categorized them based on the degree of
   degeneration observed in imaging (Figure [83] 1A), and examined the
   protein expression of the senescence‐associated factors tumor protein
   53 (TP53)/p53, cyclin‐dependent kinase inhibitor 1A (CDKN1A)/p21, and
   cyclin‐dependent kinase inhibitor 2A (CDKN2A)/p16 within the samples
   (Figure [84]1B). Our findings revealed elevated expression levels of
   TP53/p53, CDKN1A/p21, and CDKN2A/p16 in the severe IVDD group (Grade
   III/IV) compared to those in the mild IVDD group (Grade II/III). To
   investigate BNIP3 expression in degenerating NP tissues, we performed
   immunohistochemical staining for CDKN2A and BNIP3. Our results revealed
   decreased expression of BNIP3 and upregulation of CDKN2A in the grade V
   group compared to levels in the grade II group. (Figure [85]1C).
   Notably, immunofluorescence analysis of rat IVDD and degenerating human
   NP tissues also indicated downregulation of BNIP3 (Figure [86]1D) in
   both models. To simulate an in vitro model of IVDD, we treated NPCs
   with tert‐butyl peroxide (TBHP) (Figure [87]1E). This treatment
   resulted in an increased number of β‐galactosidase‐positive cells and
   reduced the proliferative efficiency of EdU‐labeled cells, suggesting
   that NPCs are prone to senescence upon exposure to TBHP. Similarly,
   after TBHP treatment of NPCs, we observed a significant accumulation of
   ROS in cells upon staining with 2′−7′ dichlorofluorescein (DCFH‐DA).
   Furthermore, the mitochondrial superoxide indicator MitoSox
   demonstrated an accumulation of ROS within the mitochondria.
   Concurrently, MitoTracker staining revealed a shift in mitochondrial
   morphology from elongated strips to fragmented, dot‐like structures or
   a non‐fluorescent appearance (Figure [88]1F). These results suggest
   that disrupted mitochondrial homeostasis in senescent NPCs contributes
   to the persistent and irreversible degeneration of these cells.

Figure 1.

   Figure 1
   [89]Open in a new tab

   Mitochondrial damage and BNIP3 downregulation in senescent degenerating
   nucleus pulposus tissues. A) T2‐weighted magnetic resonance imaging
   (MRI) was utilized to procure human nucleus pulposus tissues exhibiting
   varying degrees of degeneration, as per the Pfirrmann grading system.
   B) Senescence‐associated proteins (tumor protein 53 [TP53],
   cyclin‐dependent kinase inhibitor 1A [CDKN1A], and cyclin‐dependent
   kinase inhibitor 2A [CDKN2A]) in degenerated tissues were
   quantitatively evaluated via western blotting. n = 3, *p < 0.05;
   **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant. C)
   Immunohistochemical staining was performed to examine the expression of
   CDKN2A and BNIP3 in different degenerating tissues. n = 3, scale bar:
   500 µm. *p < 0.05. D) Tissue fluorescence assessment of BNIP3
   expression in degenerating human and rat nucleus pulposus tissue. Human
   tissue: n = 3, scale bar: 100 µm. Rat tissue: n = 3, scale bar: 200 µm.
   **p < 0.01; ***p < 0.001. E) Nucleus pulposus cells were stimulated
   with tert‐butyl peroxide (TBHP), and differences between the
   TBHP‐treated and control groups were assessed by β‐galactosidase, EdU,
   and immunofluorescence staining. β‐gal and EdU: n = 3, scale bar:
   100 µm. Immunofluorescence staining for BNIP3; scale bar: 50 µm.
   **p < 0.01; ***p < 0.001. F) Mitochondrial function in isolated human
   nucleus pulposus cells was evaluated by 2′−7′ dichlorofluorescein
   (DCFH‐DA), MitoSox, and MitoTracker assays following exposure to TBHP
   at concentrations of 10, 20, 30, and 40 µmol/mL. DCFH‐DA: n = 3, scale
   bar: 100 µm. MitoSox: n = 3, scale bar: 100 µm. MitoTracker: n = 3,
   scale bar: 20 µm. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001;
   ns, not significant. Data are expressed as mean ± standard deviation.

   In this regard, we concluded that differences existed between the
   degenerated NP tissue and undegenerated groups, as evidenced by the
   upregulation of senescence‐related genes, restriction of cell
   proliferation, and downregulation of the mitochondrial autophagy gene
   BNIP3. We successfully constructed a cellular model of senescence and
   mitochondrial damage in NPCs using TBHP, which lays the foundation for
   subsequent studies.

2.2. Characterization and Identification of Exosomes for Restoration of NPC
Extracellular Matrix Homeostasis

   Hypoxia, a condition characterized by a limited oxygen supply, can
   activate cellular responses in the form of stress‐induced release or
   adaptive changes. These cellular adaptations to hypoxic environments
   may be beneficial.^[ [90]^18 ^] Therefore, we conducted experiments on
   human bone marrow‐derived mesenchymal stem cells (HBMSCs) to
   investigate the effects of hypoxic chambers on the production of
   hypoxia‐inducible factor 1 (HIF‐1). We subjected the HBMSCs to
   different durations of hypoxia (4, 8, 12, and 24 h) in chambers
   maintained at 37 °C under 5% CO[2] and 2% O[2]. Our results showed that
   the intracellular production of HIF‐1 peaked after 24 h of hypoxia
   treatment (Figure [91] 2A,B). Western blot analysis of exosome surface
   markers (tumor susceptibility 101 [TSG101], heat shock protein 70
   [HSP70], and CD63) revealed slightly elevated expression of exosomes in
   hypoxia‐treated cells (hypoxia‐exo) compared to that in the exosomes of
   the normoxia‐treated group (normoxia‐exo) (Figure [92]2C). Nanoparticle
   tracking analysis results demonstrated that both hypoxia‐exo and
   normoxia‐exo had particle diameters of ≈100 nm (Figure [93]2D,E).
   Hypoxia pretreatment weakly altered the exosome particle size, while
   the appearance, size, and basic characteristics of both exosome groups
   remained consistent. To further assess the endocytic effect of exosomes
   on NPCs, we labeled the exosomes with PKH26 and co‐incubated them with
   NPCs for 24 h (Figure [94]2F,G). Using high‐magnification fluorescence
   microscopy, we observed that the hypoxia‐exo group exhibited superior
   endocytosis compared with that of the normoxia‐exo group.

Figure 2.

   Figure 2
   [95]Open in a new tab

   Characteristics and functional analysis of exosomes in human bone
   marrow‐derived mesenchymal stem cells (HBMSCs) subjected to hypoxic
   conditions. A) Hypoxia‐inducible factor 1 (HIF‐1) expression in HBMSCs
   was detected by western blotting after subjecting them to hypoxia for
   4, 8, 12, or 24 h. B) The quantitative analysis of western blotting:
   n = 3, ****p < 0.0001; ns, not significant. C) Surface marker proteins
   (tumor susceptibility 101 [TSG101], heat shock protein 70 [HSP70], CD63
   and endoplasmic reticulum protein [Calnexin]) of exosomes in
   hypoxia‐treated cells (hypoxia‐exo) and those of the normoxia‐treated
   group (normoxia‐exo) were quantified using western blotting. D)
   Nanoparticle tracking analysis revealed that both hypoxia‐exo and
   normoxia‐exo had a size distribution ranging from 40–150 nm. E)
   Differences in exosome morphology were observed under transmission
   electron microscopy. Scale bar: 5 µm. F) Exosomes were labeled with
   PKH26 and co‐incubated with nucleus pulposus cells for 24 h. Laser
   confocal microscopy demonstrated that hypoxia‐exo had stronger
   endocytosis efficiency than did normoxia‐exo. Scale bar: 20 µm. G) A
   schematic diagram was created to illustrate the co‐incubation process.
   H) Western blot analysis of catabolism‐ (matrix metallopeptidase 3
   [MMP3], matrix metallopeptidase 13 [MMP13], and ADAM metallopeptidase
   with thrombospondin type 1 motif 5 [ADAMTS5]) and anabolism‐related
   markers (collagen II [COL‐II] and aggrecan) after different treatments.
   I) Immunofluorescence staining of biomarkers related to matrix
   catabolism (MMP13; red) and anabolism (COL‐II; green). Scale bar:
   100 µm. J) Quantitative analysis of MMP13 and COL‐II fluorescence
   intensity. n = 3, **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not
   significant. K) Quantitative analysis of MMP3, MMP13, ADAMTS5, COL‐II,
   and aggrecan expression. n = 3, *p < 0.05; **p < 0.01; ***p < 0.001;
   ****p < 0.0001; ns, not significant. L) Quantitative analysis of the
   senescence‐related markers tumor protein 53 (TP53)/p53, cyclin
   dependent kinase inhibitor 1A (CDKN1A)/p21, and cyclin dependent kinase
   inhibitor 2A (CDKN2A)/p16 expression. n = 3, **p < 0.01,
   ****p < 0.0001. M) Western blot analysis of senescence‐related markers
   (TP53/p53, CDKN1A/p21, and CDKN2A/p16). N) Quantitative analysis of
   exosome function after proteinase K pretreatment. n = 3, *p < 0.05;
   **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant. O)
   Western blot analysis of exosome function after proteinase K
   pretreatment. Data are expressed as mean ± standard deviation. BMSC,
   bone mesenchymal stem cell.

   We then analyzed expression of catabolism (matrix metallopeptidase
   [MMP13]) and anabolism (collagen II [COL‐II]) markers in NPCs by
   western blotting, and after TBHP treatment, we exogenously added
   normoxia‐treated (normoxia‐exo) or hypoxia‐treated BMSC exosomes
   (hypoxia‐exo) to these cells while treating the control group with
   phosphate‐buffered saline (PBS) only (Figure [96]2H,K). The results
   showed that treatment with hypoxia‐exo was superior to that with
   normoxia‐exo in terms of ECM rescue. The western blotting and
   immunofluorescence results were consistent, and treatment with
   hypoxia‐exo downregulated the expression of the catabolism markers
   matrix metallopeptidase 3 (MMP3), MMP13, and ADAM metallopeptidase with
   thrombospondin type 1 motif 5 (ADAMTS5) and upregulated the expression
   of the anabolism markers COL‐II and aggrecan (Figure [97]2I,J). We
   further validated the efficacy of hypoxia‐exo at the mRNA level with
   quantitative PCR experiments; the results were congruent with the
   western blot findings (Figure [98]S1, Supporting Information). To
   confirm the critical role of HIF‐1α in HBMSCs, we used siRNA targeting
   HIF‐1α to knock down its expression in BMSCs, which resulted in partial
   dysfunction of exosomal components. In the knockdown of HIF‐1α
   (si‐Hif‐1α) treatment group, we found that the improvements provided by
   treatment with hypoxia‐exo were partially reversed (Figure [99]2H–K).
   In addition, we evaluated the effects of hypoxia on senescence‐related
   markers in NPCs. Western blot analysis of TP53/p53, CDKN1A/p21, and
   CDKN2A/p16 showed that hypoxia‐exo treatment effectively reversed the
   TBHP‐induced senescence phenotype of the NPCs (Figure [100]2M,L).
   Similarly, in si‐Hif‐1α‐treated BMSCs, a partial loss of function in
   exosomes secreted under hypoxic conditions was observed
   (Figure [101]2M,L). In addition, the protective effect of hypoxia‐exo
   was partially attenuated by pretreatment with proteinase K, suggesting
   that the therapeutic effect of exosomes depends on intravesicular cargo
   delivery (Figure [102]2N,O).

2.3. Hypoxia‐Exo Ameliorates IVDD in Rats In Vivo

   To investigate the influence of exosomes on IVDD in vivo, we
   established a rat model of IVDD and radiographically evaluated the rats
   according to different postoperative periods (four and eight weeks) to
   observe IVDD progression and changes in disc height index (DHI). After
   surgery, PBS, normoxia‐exo, or hypoxia‐exo was injected into the
   intervertebral discs of the rats. Using X‐ray imaging and MRI, we
   assessed the changes in the intensity of the DHI (Figure [103]S2,
   Supporting Information) and the high signal intensity of the NP in the
   T2‐weighted phase (Table [104]S1, Supporting Information). In the
   PBS‐injected group, the DHI was significantly decreased, characterized
   by cone collapse, intervertebral space wear, and osteophyte formation
   (Figure [105] 3A). This finding was confirmed by MRI (Figure [106]3B).
   While all experimental groups showed a loss of NP signal, this was most
   evident in the PBS‐injected group, followed by the normoxia‐exo‐treated
   group, whereas the hypoxia‐exo‐treated group showed the smallest
   change. The differences in the degree of degeneration between the
   groups were small at four weeks and significant at eight weeks
   (Figure [107]3C).

Figure 3.

   Figure 3
   [108]Open in a new tab

   Imaging assessment and histologic evaluation after exosomal treatment
   in vivo. A) Postoperative X‐ray assessment of rats with intervertebral
   disc degeneration (IVDD) at 4 and 8 w (n = 4). B) Postoperative
   magnetic resonance imaging (MRI) assessment of rats at 4 and 8 w
   (n = 4). C) Quantitative analysis of rat disc height index scores and
   Pfirrmann grades. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001;
   ns, not significant. D) Hematoxylin and eosin (HE) staining images of
   rat caudal spines 4 and 8 w after surgery (n = 4). Scale bar: 1 mm. E)
   Safranin‐O/Fast Green (SF) staining images 4 and 8 w after surgery
   (n = 4). Scale bar: 1 mm. F) Immunohistochemistry of matrix
   metallopeptidase 13 (MMP13) and collagen II (COL‐II) 4 and 8 w after
   surgery (n = 4). Scale bar: 500 µm. G) Quantitative analysis of the
   data was plotted as heat maps based on the morphology of the nucleus
   pulposus, cellularity of the nucleus pulposus, border between the
   anulus fibrosus and nucleus pulposus, morphology of the anulus
   fibrosus, and cellularity of the anulus fibrosus. Data are expressed as
   mean ± standard deviation.

   In addition to radiographic evidence, histology was used to evaluate
   the efficacy of the different treatments. Specifically, we utilized
   hematoxylin and eosin (HE) and Safranin‐O/Fast Green (SF) staining of
   rat disc sections (Table [109]S2, Supporting Information). Four weeks
   after treatment with PBS, we observed compression‐induced deformation
   of the NP, leading to a reduction of ≈20–25% in its area. The
   normoxia‐exo‐treated group showed partial amelioration, whereas the
   hypoxia‐exo‐treated group demonstrated the most pronounced therapeutic
   effect. After eight weeks, the PBS‐treated group exhibited a nearly
   complete disappearance of NPCs and replacement of the intervertebral
   space with inwardly extending chondrocytes compared with features in
   the control group. In the normoxia‐exo‐treated group, IVDD progression
   was mitigated, and the hypoxia‐exo‐treated group displayed the most
   significant amelioration (Figure [110]3D). After SF staining, the
   medullary tissue appeared red owing to collagen enrichment, whereas the
   NP tissue appeared blue owing to fiber enrichment. In the PBS‐treated
   group, we observed the lowest ratio of medullary tissue area compared
   to that in the control group, and the deficient medullary tissue was
   replaced by a blue fibrous ring, features which were second‐most common
   in the normoxia‐exo‐treated group. In contrast, the red color,
   indicative of collagen enrichment, was still present in a larger area
   in the hypoxia‐exo‐treated group eight weeks after the operation,
   indicating superior treatment efficacy (Figure [111]3E).

   MMP13 immunohistochemistry demonstrated the following (Figure [112]3F):
   in the control group, NPCs exhibited a lighter brown color; in the
   PBS‐treated group, the NP area decreased, and the intensity of histone
   staining increased compared to those in the control group; in the
   normoxia‐exo‐treated group, the degree of brown coloration was slightly
   decreased relative to that in the PBS‐treated group; and the
   hypoxia‐exo‐treated group exhibited a further decrease in brown
   coloration, indicating downregulation of MMP13.

   The COL‐II immunohistochemistry results revealed that the control group
   exhibited the strongest brown coloration (Figure [113]3F). Conversely,
   the PBS‐treated group displayed the lightest brown coloration, with
   most of the NP retaining its original color. In the
   normoxia‐exo‐treated group, the staining intensity was slightly higher
   than that in the PBS‐treated group. Notably, the hypoxia‐exo‐treated
   group displayed a superior staining intensity compared to both the
   normoxia‐exo‐treated and PBS‐treated groups, confirming the efficacy of
   hypoxia‐pretreated BMSC exosomes both in vivo and in vitro in rats. The
   results of the HE and SF staining and immunohistochemistry scoring are
   presented as heat maps in Figure [114]3G.

2.4. Genetic Differences in Hypoxia‐Preconditioned Exosomes

   TBHP triggers mitochondrial dysfunction and senescence in NPCs, and
   treatment with hypoxia‐exo significantly reversed this outcome. To
   quantify differential gene expression between hypoxia‐ and
   normoxia‐preconditioned BMSC exosomes, we used the NCBI database
   ([115]https://www.ncbi.nlm.nih.gov/geo/) for data retrieval. The
   specific dataset employed in this study was extracted from sample
   [116]GSE229825, and the processing conditions closely followed those
   provided by the data submitters. Subsequently, the obtained data were
   subjected to heat map analysis for visualization (Figure [117] 4A).
   Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathway analysis demonstrated that genes within the hypoxia‐induced
   exosomes were primarily associated with “reactive oxygen species,”
   “mTOR signaling pathway,” and “autophagy pathway” (Figure [118]4B),
   signifying their important role in the hypoxia‐mediated intracellular
   signaling pathway. Notably, the molecular function enrichment map
   (Figure [119]4C) reveals that several hypoxia‐inducible genes were
   significantly upregulated in the biological process (BP) category.
   Correspondingly, some mitochondrial membrane, inner and outer
   mitochondrial membrane, and organelle membrane genes were significantly
   upregulated in the cell component (CC) category. Furthermore, several
   oxidoreductase and kinase activity‐related genes were significantly
   upregulated at the molecular function (MF) level. Thus,
   hypoxia‐preconditioned extracellular vesicles contain abundant
   hypoxia‐responsive factors that are closely linked to the mitochondria.
   To further investigate this attribute, we generated volcano plots
   (Figure [120]4D) based on the following: false discovery rate < 0.001
   and log FC > 1.5. Among the exocytosis‐related genes upregulated by
   hypoxia, BNIP3 expression was significantly increased.

Figure 4.

   Figure 4
   [121]Open in a new tab

   Hypoxic exosomes protect nucleus pulposus cells (NPCs) from oxidative
   stress‐induced mitochondrial dysfunction. A) Differential gene
   enrichment maps in hypoxic and normoxic exosomes. B) Differential gene
   pathway enrichment analysis (Kyoto Encyclopedia of Genes and Genomes
   [KEGG] analysis). C) Gene ontology (GO) analysis of differential genes.
   D) Volcano plot showing differential gene expression in hypoxic and
   normoxic groups. False discovery rate < 0.001, logFC > 1.5. E) Western
   blot analysis of BCL2/adenovirus E1B 19 kDa interacting protein 3
   (BNIP3) levels in tert‐butyl peroxide (TBHP)‐treated NPCs. n = 3,
   **p < 0.01. F) Western blot analysis of BNIP3 levels in NPCs treated
   with hypoxic versus normoxic exosomes. n = 3, **p < 0.01. G)
   Mitochondrial length measurements of NPCs under TBHP stimulation.
   n = 3, scale bar: 20 µm. ****p < 0.0001. H,I) Co‐localization of
   PKH67‐labeled exosomes with MitoTracker‐labeled mitochondria. n = 3,
   scale bar: 20 µm. J) Fluorescence analysis of hypoxic and normoxic
   exosomes for mitochondrial membrane potential (JC‐1; scale bar: 100 µm)
   and mitochondrial morphology (MitoTracker; scale bar: 50 µm).
   Mitochondrial reactive oxygen species (MitoSox) in NPCs. Scale bar:
   100 µm. K) Quantitative analysis of relative fluorescence intensity of
   JC‐1 and MitoSox, as well as quantification of mitochondrial
   fragmentation ratio. n = 3, *p < 0.05; **p < 0.01; ***p < 0.001;
   ****p < 0.0001; ns, not significant. Data are expressed as
   mean ± standard deviation.

   To validate the differential expression of BNIP3 among degenerating
   NPCs, an exogenous degenerating NPC model was induced using TBHP;
   subsequently, BNIP3 protein expression was verified by western blot
   analysis (Figure [122]4E). The results demonstrated significant
   downregulation of BNIP3 in degenerating NPCs. Additionally, we
   conducted immunohistochemical and tissue fluorescence analyses of rat
   tail intervertebral disc slices. Consistent with previous results, we
   observed higher BNIP3 expression in the hypoxia‐exo‐treated group
   (Figure [123]S3, Supporting Information). To validate this finding, we
   simultaneously introduced exogenous hypoxia‐exo and normoxia‐exo into
   the NPCs. Western blot analysis confirmed that BNIP3 expression
   increased in NPCs following the addition of hypoxia‐exo compared to
   BNIP3 levels in the normoxia‐exo‐treated group, thus confirming that
   hypoxia‐exo upregulated BNIP3 gene expression after endocytosis into
   NPCs (Figure [124]4F). Assessment of mitochondrial morphology in
   TBHP‐treated NPCs using MitoTracker revealed a shortened, flexed
   phenotype along with a blurred structure and weakened fluorescence
   intensity, indicating impaired mitochondrial integrity
   (Figure [125]4G).

   To visualize and confirm the endocytosis of exosomes into NPCs and
   their interactions, we stained exosome membranes with PKH67 and
   simultaneously labeled mitochondria in NPCs with MitoTracker (both
   groups of cells were treated with TBHP). Fluorescence microscopy
   revealed the co‐localization of exosomes with mitochondria, suggesting
   their interaction (Figure [126]4H). Furthermore, we observed that
   hypoxia‐exos had a higher likelihood of penetrating the NPC membrane
   than did normoxia‐exos. Moreover, the hypoxia‐exo‐treated group
   demonstrated superior recovery of mitochondrial morphology and
   fluorescence intensity compared to that in the normoxia‐exo‐treated
   group (Figure [127]4I). Mitochondrial membrane potential was then
   evaluated using JC‐1, MitoTracker and MitoSox probes in TBHP‐treated
   NPCs (Figure [128]4J). Our results revealed a decrease in mitochondrial
   membrane potential (ΔΨm), with healthy mitochondria showing JC‐1
   aggregation in the mitochondrial matrix that formed polymers that
   emitted strong red fluorescence, while unhealthy mitochondria
   (PBS‐treated) displayed JC‐1 as a cytosolic monomer producing green
   fluorescence. MitoTracker detects increased fragmentation of
   intracellular mitochondria after TBHP treatment, which improves upon
   addition of Hypoxia‐exo. Additionally, the MitoSox probe showed
   oxidation by ROS in mitochondria, resulting in intense red
   fluorescence. However, the addition of hypoxia‐exo resulted in partial
   restoration of mitochondrial morphology, with a corresponding reversal
   of the JC‐1 red/green ratio and reduction of mitochondrial ROS. These
   findings demonstrate the partial rescue of the mitochondrial phenotype
   after the endocytosis of exosomes by NPCs, leading to reduced
   mitochondrial fragmentation and oxidative stress damage. Remarkably,
   the rescue effect was more pronounced in the hypoxia‐exo‐treated group
   than in the normoxia‐exo‐treated group. Furthermore, we silenced
   hypoxia‐inducible factor HIF‐1α to illustrate its important role and
   found that the effect of hypoxia‐exo treatment was partially
   attenuated, indicating that genes in the exosomes of hypoxia‐pretreated
   cells are under upstream transcription factor regulation
   (Figure [129]4K).

2.5. Hypoxia‐Induced Exosome‐Delivered BNIP3 Activates Mitochondrial
Autophagy, Alleviating Aging in NPCs

   We quantified autophagic flux in NPCs treated with TBHP using western
   blot analysis. Moreover, we generated NPCs with reduced BNIP3 levels
   using a targeted BNIP3 siRNA (si‐BNIP3) approach. Subsequently, we
   supplemented wild‐type NPCs with hypoxia‐exo and observed a remarkable
   increase in BNIP3 protein expression concomitant with an increase in
   the LC3II/I ratio, a crucial marker of autophagy (Figure [130] 5A).
   Subsequently, BNIP3 knockdown led to a notable decline in autophagic
   flux and a concurrent increase in substrate release resulting from
   accumulated P62. Following this, we introduced hypoxia‐exo into BNIP3
   knockdown NPCs, resulting in the restoration of BNIP3 expression and
   subsequent recovery of autophagic flux (Figure [131]5B). We conducted a
   comprehensive assessment of the impact of BNIP3 using TEM. The results
   revealed that in the si‐Control+TBHP group, mitochondria displayed
   swelling, loss of cristae, and vacuole‐like degeneration
   (Figure [132]5C). This phenomenon was ameliorated upon addition of
   hypoxia‐exo. Subsequently, upon BNIP3 knockdown and exposure to TBHP,
   mitochondrial damage was notably more pronounced compared to that in
   the si‐control group. Additionally, supplementation of hypoxia‐exo to
   the BNIP3 knockout group resulted in restoration of impaired
   mitochondrial morphology. To further evaluate the effect of this
   process on NPCs, we performed double‐immunofluorescence staining using
   MitoTracker and BNIP3 (Figure [133]5D). In wild‐type NPCs, hypoxia‐exo
   treatment significantly ameliorated the aberrant mitochondrial
   morphology induced by TBHP, restoring the rod‐like elongated
   mitochondrial structure. Furthermore, the fluorescence intensity of
   MitoTracker staining was enhanced, and a significant increase in BNIP3
   fluorescence intensity was observed compared to baseline levels. To
   ascertain the involvement of BNIP3 in mitochondrial autophagy, we
   induced mitochondrial damage by adding TBHP subsequent to BNIP3
   knockdown. The results demonstrated further deterioration of
   mitochondrial morphology in BNIP3‐knockdown NPCs compared to that in
   wild‐type NPCs. However, upon administration of hypoxia‐exo, successful
   delivery of BNIP3 within the exosomes to the NPCs was observed, leading
   to the upregulation of BNIP3 fluorescence intensity and partial
   restoration of mitochondrial morphology (Figure [134]5D,E). To further
   investigate this phenomenon, immunofluorescence was used to label the
   translocase of outer mitochondrial membrane 20 (TOMM20) along with LC3
   in NPCs. Upon TBHP‐induced mitochondrial damage, TOMM20 was recruited
   to the outer mitochondrial membrane and subsequently activated.
   Research indicates that activation of TOMM20 recruits parkin, leading
   to the ubiquitination of outer membrane proteins. The ubiquitinated
   products bind to LC3 and are enclosed in autophagic vesicles.^[
   [135]^19 ^] By incorporating hypoxia‐exo, the fluorescence intensity
   and co‐localization of TOMM20 with LC3 were further increased,
   indicating the successful activation of mitochondrial autophagy
   (Figure [136]5F).

Figure 5.

   Figure 5
   [137]Open in a new tab

   BCL2/adenovirus E1B 19 kDa interacting protein 3 (BNIP3) maintains
   mitochondrial homeostasis through autophagy activation. A) Western blot
   analysis of BNIP3, LC3, and P62 expression in wild‐type and
   BNIP3‐knockout NPCs treated with tert‐butyl peroxide (TBHP) with or
   without exosomes from hypoxia‐treated cells (hypoxia‐exo). B)
   Quantification of BNIP3, LC3, and P62 expression. n = 3, *p < 0.05;
   ***p < 0.001; ****p < 0.0001; ns, not significant. C) Mitochondrial
   morphology changes as observed by transmission electron microscopy. D)
   Double‐immunofluorescence staining of MitoTracker and BNIP3. Scale bar:
   20 µm. E) Quantitative analysis of fluorescence intensity of BNIP3 and
   mitochondrial length measurements in different treatment groups. n = 3,
   **p < 0.01; ****p < 0.0001; ns, not significant. F) Immunofluorescence
   detection of changes in the fluorescence intensity of LC3 and
   translocase of outer mitochondrial membrane 20 (TOMM20) in nucleus
   pulposus cells (NPCs) after different treatments. Scale bar: 20 µm. G)
   X‐ray images of BNIP3 wild type (BNIP3 WT) and BNIP3‐/‐ mice after
   different treatments. H,I) Hematoxylin and eosin (HE) and Safranin‐O
   (SO) staining of BNIP3 WT and BNIP3‐/‐ mice. Scale bar: 1 mm. J,K)
   Immunohistochemical analysis of cyclin dependent kinase inhibitor 2A
   (CDKN2A) expression in BNIP3 WT and BNIP3‐/‐ mice. Scale bar: 200 µm.
   Data are expressed as mean ± standard deviation.

   To investigate the role of BNIP3 in IVDD, we compared the imaging and
   histological scores of BNIP3 wild‐type (BNIP3 WT) and BNIP3 knockout
   (BNIP3‐/‐) mice in both the control and puncture groups at four weeks
   post‐surgery. The results showed that the NP tissue integrity and
   intervertebral space height were both lower in the BNIP3‐/‐ group
   compared to that in the BNIP3 WT group (Figure [138]5G). Quantitative
   analysis results are provided in the supplementary materials
   (Figure [139]S5, Supporting Information). Additionally, following IVDD
   surgery, the postoperative repair capacity of the BNIP3‐/‐ group was
   significantly reduced. However, the integrity of intervertebral disc
   tissue was improved after exogenous addition of hypoxia‐exo
   (Figure [140]5H,I). Afterward, we conducted immunohistochemistry to
   assess the levels of age‐related proteins (CDKN1A, CDKN2A, and TP53).
   The results revealed mild signs of aging in the NP of BNIP3‐/‐ mice,
   which became more pronounced following IVDD surgery. Addition of
   exogenous hypoxia‐exo subsequently partially mitigated the elevation of
   aging marker proteins (Figure [141]5J,K). The immunohistochemistry
   results for TP53 and CDKN1A were consistent with the previous findings
   (Figure [142]S5, Supporting Information).

   To further elucidate the relationship between BNIP3 and mitochondrial
   autophagy activation in the regulation of human NPC aging, we induced
   the aging phenotype of NPCs using TBHP and then applied 3‐methyladenine
   (3‐MA) to inhibit autophagosome formation. We found that 3‐MA treatment
   exerts beneficial effects on TBHP‐induced aging of NPCs (wild‐type), as
   evidenced by western blotting analysis of TP53, CDKN1A, and CDKN2A
   expression and cell staining with β‐gal and EdU (Figures [143]S6
   and [144]S7, Supporting Information). Concurrently, MitoTracker
   analysis revealed an increase in TBHP‐induced mitochondrial
   fragmentation, which was alleviated by 3‐MA treatment (Figure [145]S7,
   Supporting Information). In contrast, 3‐MA treatment had deleterious
   effects on NPCs overexpressing BNIP3 (oe‐BNIP3), exacerbating the aging
   phenotype compared with that in wild‐type cells. Furthermore, cell
   proliferation decreased and mitochondrial fragmentation increased in
   oe‐BNIP3 NPCs treated with 3‐MA (Figure [146]S7, Supporting
   Information). This suggests that mitochondrial autophagy can be induced
   by TBHP in both wild‐type and oe‐BNIP3 NPCs and that BNIP3 plays a
   beneficial role in promoting mitochondrial autophagy, which is crucial
   for alleviating mitochondrial dysfunction and cellular aging.

2.6. The Mitophagy Receptor BNIP3 Interacts with ANXA2 in NPCs, Facilitating
the Nuclear Translocation of TFEB

   We investigated the potential mechanisms underlying the involvement of
   BNIP3 in mitophagy. To validate the potential downstream targets of
   BNIP3, co‐immunoprecipitation (CO‐IP) and mass spectrometry (MS) were
   employed to identify the downstream effector proteins associated with
   BNIP3‐mediated autophagy activation (Figure [147] 6A). We performed a
   range of analyses, including GO, KOGs, KEGG, IPR, and subcellular
   localization analyses, on the coprecipitated material to obtain
   additional information on these downstream effectors (Figure [148]S8,
   Supporting Information). In summary, annexin A2 (ANXA2) was chosen as a
   protein that strongly interacts with BNIP3 from a pool of ten potential
   candidates based on protein abundance analysis and was therefore used
   for further validation (Table [149]S3, Supporting Information). We
   validated the protein–protein interactions between BNIP3 and ANXA2
   using CO‐IP analysis (Figure [150]6B). ANXA2 is a calcium‐dependent
   membrane‐binding protein that plays a crucial role in membrane
   trafficking and fusion.^[ [151]^20 ^] In nonalcoholic fatty liver
   disease, ANXA2 obstructs autophagic flux mediated by the AMP‐activated
   protein kinase (AMPK)/mechanistic target of rapamycin (mTOR) pathway.^[
   [152]^21 ^] However, we observed no significant change in the
   fluorescence intensity of ANXA2 when BNIP3 expression was upregulated
   in NPCs (Figure [153]6C). Hence, we hypothesized that BNIP3 regulates
   autophagy via an indirect interaction with ANXA2.

Figure 6.

   Figure 6
   [154]Open in a new tab

   The BCL2/adenovirus E1B 19 kDa interacting protein 3 (BNIP3)–annexin A2
   (ANXA2) interaction influences transcription factor EB (TFEB)
   localization. A) Schematic representation of co‐immunoprecipitation
   (CO‐IP) and mass spectrometry. B) BNIP3 interaction with ANXA2
   demonstrated by CO‐IP. n = 3. C) Differential expression of BNIP3 and
   ANXA2 following overexpression or empty vector expression assessed by
   immunofluorescence staining analysis. n = 3, scale bar: 50 µm. D)
   Nucleus pulposus cells (NPCs) were transfected with either an BNIP3
   plasmid to overexpress BNIP3 or an empty plasmid, followed by CO‐IP to
   validate the reciprocal binding between BNIP3, ANXA2, and TFEB. n = 3.
   E) CO‐IP experiment confirming TFEB binding to ANXA2. n = 3. F) NPCs
   were subjected to ANXA2 knockdown, and subsequent western blot analysis
   was employed to examine the cytoplasmic and nuclear expression levels
   of ANXA2 and TFEB proteins. n = 3. G) Quantitative analysis of
   cytoplasmic and cytosolic TFEB expression was performed. n = 3,
   *p < 0.05. H) In NPCs, ANXA2 was either silenced or BNIP3 was
   overexpressed, and the spatial localization of TFEB was assessed using
   immunofluorescence. n = 3, scale bar: 50 µm, **p < 0.01. I) Spatial
   changes in TFEB expression were analyzed by western blot analysis
   following treatment with tert‐butyl peroxide (TBHP) or an
   overexpression control plasmid (oe‐control) or overexpression of BNIP3
   (oe‐BNIP3) in NPCs. J) Quantification of cytoplasmic and nuclear TFEB
   expression. n = 3, **p < 0.01; ****p < 0.0001; ns, not significant.
   Data are expressed as mean ± standard deviation.

   Previous studies have indicated that ANXA2 serves as a direct target of
   bleomycin in interstitial pulmonary fibrosis, forming a complex with
   transcription factor EB (TFEB) and tyrosine 3‐monooxygenase/tryptophan
   5‐monooxygenase activation protein (YWHA) in the cytoplasm, resulting
   in the cytoplasmic retention of TFEB.^[ [155]^20 ^] TFEB has been
   reported to be an autophagy gene involved in lysosome biogenesis and
   autophagolysosome regulation. Inhibition of the nuclear translocation
   of TFEB suppresses lysosomal gene activation and initiation of
   mitochondrial autophagic flux. Therefore, we conducted
   immunoprecipitation experiments to determine whether ANXA2 is involved
   in inhibiting the nuclear translocation of TFEB. The results revealed
   an interaction between ANXA2 and TFEB (Figure [156]6E). To verify this
   interaction, we overexpressed BNIP3 in NPCs and preserved the
   interacting proteins. Subsequently, we validated this interaction using
   western blot analysis. The results demonstrated that the interaction
   between BNIP3 and ANXA2 was enhanced by BNIP3 overexpression, whereas
   the interaction between ANXA2 and TFEB was weakened (Figure [157]6D).
   Next, siRNA (si‐ANXA2) was used to suppress intracellular ANXA2
   expression. Western blot analysis of nuclear and cytoplasmic proteins
   revealed that silencing ANXA2 downregulated the cytoplasmic expression
   of TFEB and upregulated its nuclear expression. (Figure [158]6F,G).
   Similarly, we treated NPCs with si‐ANXA2 and performed
   immunofluorescence analysis. The results revealed that, compared to the
   negative control cells (si‐control), cells treated with si‐ANXA2
   displayed enhanced nuclear expression of TFEB (Figure [159]6H). These
   results indicate that ANXA2 forms a complex with TFEB in the cytoplasm,
   preventing TFEB from entering the nucleus and becoming deactivated.
   Subsequently, we overexpressed BNIP3 in NPCs, and immunofluorescence
   analysis revealed that increased levels of BNIP3 promoted the nuclear
   translocation of TFEB. Similarly, we performed western blot analysis
   after separating nuclear and cytoplasmic proteins and observed that
   oe‐BNIP3 in NPCs did not result in a significant difference in
   cytoplasmic TFEB expression compared to that in the control group.
   However, an upward trend in nuclear TFEB protein levels compared to
   those in the control group was observed (Figure [160]6I,J). In
   conclusion, we established an association between BNIP3 and ANXA2 in
   NPCs, which reduced the interaction between ANXA2 and TFEB and promoted
   the nuclear translocation of TFEB.

2.7. TFEB is a Pivotal Regulator of Lysosomal Biogenesis and Function

   During mitophagy, the degradation and recycling of mitochondria depend
   on gene regulation within the lysosome. We further speculated that TFEB
   expression may enhance autophagic flux in NPCs, thereby activating the
   expression of lysosome‐associated genes. Through western blot analysis
   of the lysosomal markers lysosomal associated membrane protein (LAMP)−1
   and LAMP2 and lysosomal protease cathepsin D (CTSD), we observed a
   reduction in lysosomal protein expression in NPCs under TBHP treatment
   (Figure [161] 7A). In NPCs overexpressing TFEB, TFEB upregulation
   promoted the expression of the lysosome‐associated proteins LAMP1,
   LAMP2, and CTSD (Figure [162]7B). This finding was further confirmed by
   immunofluorescence analysis (Figure [163]7C,D). Subsequently, we
   evaluated the autophagic function mediated by TFEB and found that TFEB
   overexpression increased autophagy related 5 (ATG5) and LC3B‐II
   expression, while reducing the levels of the substrate P62
   (Figure [164]7E,F).

Figure 7.

   Figure 7
   [165]Open in a new tab

   The influence of TFEB abundance and subcellular localization on
   lysosomal function. A) Expression analysis of lysosomal associated
   membrane protein (LAMP)−1, LAMP2, and cathepsin D (CTSD) using western
   blot analysis. B) Quantitative assessment of relative protein levels of
   LAMP1, LAMP2, and CTSD. n = 3, *p < 0.05, ****p < 0.0001. C)
   Fluorescence intensity detection of LAMP1, LAMP2, and CTSD by
   immunofluorescence assay. Scale bar: 50 µm. D) Quantitative analysis of
   LAMP1, LAMP2, and CTSD using immunofluorescence. n = 3, *p < 0.05,
   **p < 0.01, ***p < 0.001. E) Expression analysis of LC3, autophagy
   related 5 (ATG5), and P62 using western blot analysis. F) Protein
   quantification analysis of LC3, ATG5, and P62. n = 3, *p < 0.05;
   **p < 0.01; ****p < 0.0001; ns, not significant. G) Spatial
   distribution analysis of TFEB after different treatments using
   immunofluorescence. Scale bar: 100 µm. H) Quantitative analysis of
   nuclear/cytoplasmic protein ratio of TFEB. n = 3, ****p < 0.0001; ns,
   not significant. I) Immunofluorescence staining for LMAP1 and
   galectin‐3 GAL3 after different treatments. Scale bar: 50 µm. J)
   Quantitative analysis of GAL3 fluorescence intensity. n = 3, *p < 0.05,
   **p < 0.01. K) Treatment with 20‐deoxyingenol enhances TFEB nuclear
   translocation to augment Lyso Tracker Red staining. L) Quantitative
   analysis of the relative changes in Lyso Tracker Red. **p < 0.01;
   ***p < 0.001. Data are expressed as mean ± standard deviation.

   To further explore the influence of TFEB and its subcellular
   localization on gene function, we conducted experiments using
   20‐deoxyingenol (20‐Deo), which induces the translocation of TFEB to
   the nucleus, thereby enhancing autophagy and lysosomal biogenesis. NPCs
   were subjected to different treatments, including PBS, 20‐Deo,
   overexpression of a control plasmid (oe‐control), overexpression of
   ANXA2 (oe‐ANXA2), and oe‐ANXA2+20‐Deo, to assess the changes in
   autophagic flux within the NPCs. Immunofluorescence results revealed
   that compared to levels in the control group, treatment with 20‐Deo
   significantly facilitated the nuclear expression of TFEB, whereas no
   significant difference was observed between the oe‐control‐treated and
   control groups. Furthermore, oe‐ANXA2 reduced the nuclear translocation
   of TFEB, which was partially restored by exogenous addition of 20‐Deo
   (Figure [166]7G,H).

   Galectin‐3 (GAL3) is a specific marker of lysosomal and endolysosomal
   membrane damage and has been extensively employed in the investigation
   of lysosomal membrane injury.^[ [167]^22 ^] We conducted a
   co‐localization immunofluorescence analysis of GAL3 and LAMP1 (a
   lysosomal membrane protein) to examine these phenomena. We found that
   the addition of TBHP induced lysosomal damage, leading to an increase
   in GAL3 levels (Figure [168]7I,J). Additionally, when ANXA2 was
   overexpressed in NPCs and TBHP was introduced, a further increase in
   the expression of GAL3 was observed. The subsequent administration of
   20‐Deo to ANXA2‐overexpressing NPCs partially reduced the fluorescence
   intensity of GAL3.

   Next, we utilized flow cytometry to assess the lysosomal acidification
   fluorescent probe, Lyso Tracker. The results revealed that 20‐Deo
   facilitates TFEB nuclear translocation, thereby enhancing the
   biological occurrence of Lysotracker (Figure [169]7K,L). Conversely,
   overexpression of ANXA2 led to a reduction in Lysotracker intensity by
   inhibiting TFEB nuclear translocation. These results indicate that
   nuclear translocation of TFEB effectively increases the number of
   acidified lysosomes, promoting lysosomal biogenesis.

2.8. The BNIP3/ANXA2 Axis Regulates IVDD Progression in Rats

   To validate the efficacy of ANXA2 and BNIP3 in vivo in rats, we
   established a rat IVDD model and introduced ANXA2 and BNIP3
   adenoviral‐associated viruses (AAVs) into NPCs through disc injection.
   Four weeks post‐surgery, X‐ray imaging and DHI analysis revealed a
   significant degenerative trend in the PBS+AAV‐control (CTL) group
   compared to the Sham+AAV‐CTL group. However, the addition of exogenous
   hypoxia‐exo showed certain improvements in the intervertebral space.
   Notably, co‐administration of hypoxia‐exo and AAV‐ANXA2 partially
   counteracted the efficacy of the exosomes (Figure [170] 8A,B). MRI
   T2‐weighted images of the rat intervertebral discs revealed a reduction
   in intervertebral space and decreased NPC hydration in the PBS+AAV‐CTL
   group compared to the Sham group. However, treatment with
   hypoxia‐exo+AAV‐CTL showed the most favorable therapeutic effects,
   superior to treatment with hypoxia‐exo+AAV‐ANXA2. Nevertheless,
   exogenous supplementation of AAV‐BNIP3 rescued the negative effects
   induced by AAV‐ANXA2 (Figure [171]8C,D). Furthermore, HE and SO
   staining revealed a substantial disappearance of NPCs in the PBS+AAV
   CTL group, with a more pronounced degeneration observed at eight weeks
   compared to four weeks post‐treatment (Figure [172]8E,F). Conversely,
   the hypoxia‐exo+AAV‐CTL group exhibited the most favorable therapeutic
   efficacy, while the addition of AAV‐ANXA2 attenuated the NP repair
   capability. Addition of AAV‐BNIP3 promoted ECM synthesis in the NPCs
   while inhibiting its degradation. We then conduced histological
   evaluations of the morphology of the NP, cellularity of the NP, border
   between the AF and NP, morphology of the AF, and cellularity of the AF
   and presented the results in heat maps (Figure [173]8I). Tissue
   fluorescence analysis of CDKN2A content revealed significant aging
   phenotypes in the PBS+AAV CTL group compared to the Sham+AAV‐CTL group.
   Treatment with hypoxia‐exo demonstrated the most effective rescue of
   aging in NPCs. In the hypoxia‐exo+AAV‐ANXA2 group, partial defects were
   observed in the NP tissue, accompanied by increased expression of
   CDKN2A (Figure [174]8G,H). Fluorescence results of CDKN1A and TP53
   after different treatments are provided in the supplementary materials
   (Figure [175]S9, Supporting Information).

Figure 8.

   Figure 8
   [176]Open in a new tab

   The BCL2/adenovirus E1B 19 kDa interacting protein 3 (BNIP3)/annexin A2
   (ANXA2) axis regulates the progression of intervertebral disc
   degeneration (IVDD) in vivo. A) X‐ray images of rat tail vertebrae were
   taken at 4 and 8 w post‐injection with adenoviral‐associated viruses
   (AAVs). B) Quantitative analysis of rat disc height index scores.
   n = 4, **p < 0.01; ***p < 0.001; ****p < 0.0001; C) Magnetic resonance
   imaging (MRI) with T2‐weighted sequences was conducted at 4 and 8 w
   post‐injection with AAVs. D) Quantitative analysis of Pfirrmann grades.
   n = 4, *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; E)
   Hematoxylin and eosin (HE) staining of rat tail vertebrae slices after
   different treatments. Scale bar: 1 mm. F) Safranin‐O/Fast Green (SO)
   staining of rat tail vertebrae slices after different treatments. Scale
   bar: 1 mm. G) Fluorescent staining images of cyclin dependent kinase
   inhibitor 2A (CDKN2A) in rat tail vertebrae slices from each
   experimental group. Scale bar: 200 µm. H) Quantitative analysis of the
   relative fluorescence intensity of CDKN2A in each experimental group.
   n = 4, ***p < 0.001; ****p < 0.0001; I) Plotting histological scores of
   each experimental group in heatmap format. n = 4. Data are expressed as
   mean ± standard deviation.

3. Discussion

   Treatment of IVDD, an irreversible form of cellular aging, is primarily
   focused on alleviating the associated symptoms of lower back and leg
   pain but remains a challenge. Current therapeutic approaches mainly
   involve nonpharmacological interventions, pharmacotherapy, and surgical
   procedures.^[ [177]^23 ^] However, relying solely on nonsteroidal
   anti‐inflammatory drugs or total disc excision surgery is insufficient
   to effectively reverse the degenerative trend and may even lead to more
   severe consequences.^[ [178]^24 ^] Therefore, cellular and molecular
   approaches that target the pathogenesis and underlying mechanisms of
   this condition are highly desirable. Moreover, the inevitable
   senescence and limited proliferative capacity of NPCs are significant
   factors contributing to the compromised integrity of NP tissues.^[
   [179]^25 ^] Cellular senescence, along with the inflammatory response
   mediated by the senescence‐associated secretory phenotype, is a
   critical factor contributing to the progressive deterioration of the
   intervertebral disc. In this study, we initially observed a significant
   upregulation of aging‐related protein markers in degenerated NP tissues
   accompanied by impaired mitochondrial function. To investigate this
   further, we established an in vitro model of NP degeneration using
   TBHP.^[ [180]^26 ^]

   Hypoxia restricts the cellular susceptibility to oxidative DNA damage
   and genetic mutations.^[ [181]^27 ^] Hence, hypoxia is considered a
   crucial factor for enhancing cellular function, and we therefore aimed
   to optimize the state of stem cells through hypoxic preconditioning.
   Research suggests that hypoxia‐preconditioned exosomes derived from
   BMSCs exhibit superior therapeutic efficacy in osteoarthritis compared
   to the those in an untreated group.^[ [182]^6 ^] The gain‐of‐function
   of paracrine secretion induced by stem cell preconditioning has
   positioned exosome preconditioning as a promising therapeutic approach.
   Following a simple, effective, and easily implementable method, we
   cultured BMSCs in a 2% hypoxic incubator, extracted the extracellular
   vesicles, and revealed a novel mechanism of mitochondrial autophagy
   mediated by the HIF‐1α‐BNIP3‐ANXA2 axis through differential gene
   enrichment analysis. As a key regulator of the hypoxic response, HIF‐1α
   undergoes proteasomal degradation via prolyl hydroxylation under
   physiological conditions.^[ [183]^28 ^] Additionally, HIF‐1α is
   subjected to multiple post‐translational modifications, including
   acetylation and phosphorylation.^[ [184]^29 ^] Under conditions of
   oxygen deprivation, the activity of hydroxylases is suppressed, leading
   to the accumulation of HIF‐1α. Subsequently, HIF‐1α binds to hypoxia
   response elements within the cells, thereby controlling the
   transcription and regulation of downstream genes.^[ [185]^30 ^]
   Similarly, the decrease in mitochondrial membrane potential (ΔΨm),
   which is negatively regulated under hypoxia, inhibits the accumulation
   of cytochrome c and the activation of caspase‐3 and 9, thereby
   effectively restoring a stable mitochondrial membrane potential
   state.^[ [186]^31 ^]

   Four outer mitochondrial membrane proteins, FUN14 domain containing 1
   (FUNDC1), BCL2 like 13 (BCL2L13), NIX, and BNIP3, are direct receptors
   for mitophagy.^[ [187]^32 ^] As a member of the BCL‐2 family, BNIP3 is
   widely localized in the outer mitochondrial membrane and exhibits high
   sensitivity to HIF‐1α.^[ [188]^29 ^] In this study, we demonstrated for
   the first time that hypoxia triggers the upregulation of HIF‐1α in
   mesenchymal stem cells, resulting in increased BNIP3 expression and its
   packaging into secreted exosomes. Exogenously delivered BNIP3 rescued
   the downregulation of BNIP3 caused by IVDD. In aging NPCs, the lack of
   BNIP3 limited the activation of mitochondrial autophagic flux, which
   crucially regulates the initiation of mitochondrial autophagy through
   independent or tandem receptor activation pathways.

   ANXA2 is a calcium‐dependent membrane‐binding protein that exists in
   cells as a monomer or heterotetramer. Monomers are present in the
   cytoplasm, nucleus, and early nucleoli, whereas heterotetramers are
   present on the cell membrane.^[ [189]^33 ^] Recent studies have
   reported that ANXA2 aggravates hepatic lipid metabolic injury by
   blocking AMPK/mTOR signaling‐mediated autophagic flux in the liver.^[
   [190]^21 ^] In addition, ANXA2 can bind to FAM3 metabolism regulation
   signaling molecule A (FAM3A) and activate Rho GTPase to regulate the
   AKT serine/threonine kinase 1 (AKT1)‐mTOR‐unc‐51 like autophagy
   activating kinase (ULK)−1/2 signaling axis.^[ [191]^34 ^] TFEB, a
   member of the MiT‐TEF family of transcription factors, is a critical
   regulator of lysosomal transcription. It is activated by protein
   phosphatase 3 (PPP3)/calcineurin‐mediated dephosphorylation during
   ROS‐mediated lysosomal Ca^2+ release and regulates the expression of
   receptor genes optineurin (OPTN) and calcium binding and coiled‐coil
   domain 2 (CALCOCO2) through nuclear translocation.^[ [192]^35 ^] In
   addition, studies have demonstrated a significant correlation among
   mitochondrial ROS‐mediated lysosomal Ca^2+ release, endoplasmic
   reticulum lysosomal Ca^2+ filling, and TFEB activation.^[ [193]^36 ^]
   In this study, we conducted the first screen to identify ANXA2 as an
   interacting partner of BNIP3. The binding of BNIP3 to ANXA2 facilitated
   the release of TFEB and reduced its cytoplasmic retention.
   Additionally, this release promoted the direct interaction of TFEB with
   the ATG family members LC3, p62/SQSTM1.

   In summary, we developed a hypoxia treatment strategy for optimizing
   the bioavailability of stem cell‐derived exosomes in NP tissues. By
   enriching exosomes with BNIP3, we introduced a novel approach for
   exogenous autophagy receptor gene delivery. The impact of hypoxia
   pretreatment on the physicochemical properties of exosomes was
   negligible. However, further investigation is needed to understand the
   enhanced characteristics of recipient NPC uptake compared to that of
   untreated mesenchymal stem cell‐derived exosomes. Furthermore, we
   validated the biocompatibility of the exosomes by in vivo injection
   into Sprague‐Dawley rats. These findings demonstrate the significant
   potential of this novel hypoxia‐induced BMSC‐derived exosome strategy
   for the future clinical treatment of IVDD.

4. Experimental Section

Human BMSCs

   The human mesenchymal stem cell samples used in this study were
   obtained from patients at the Second Affiliated Hospital of Wenzhou
   Medical University. This study was approved by the Medical Ethics
   Committee of the Second Affiliated Hospital of the Wenzhou Medical
   University (Wenzhou, China; approval number: LCKY2020‐157). The cells
   were seeded into 25 cm^2 flat‐bottomed cell culture flasks with BMSC
   complete medium (basal medium; OriCell, USA) supplemented with 10%
   fetal bovine serum (FBS; OriCell). The cells were subsequently cultured
   in an incubator at 37 °C under 5% CO[2]. Passaging was carried out when
   the cells reached 70–80% confluence.

Collection of Human Intervertebral Disc Samples

   The human NP tissue samples used in this study were procured from
   patients at the Second Affiliated Hospital of Wenzhou Medical
   University with the approval of the Medical Ethics Committee of the
   Second Affiliated Hospital of Wenzhou Medical University (Wenzhou,
   China; approval number: LCKY2020‐157). All patients provided written,
   informed consent prior to the procedure. The control group consisted of
   NP tissue specimens obtained from patients with trauma‐induced lumbar
   spine fractures (n = 4; mean age, 27 years; age range: 24–30 years).
   The degenerative disc group was further divided into groups II, III,
   IV, and V based on T2‐weighted images obtained via MRI. The study
   population comprised patients aged 45–65 years (n = 4; mean age, 54
   years).

Animal Model

   Sprague‐Dawley rats were selected as experimental animals in this
   study. All rats were obtained from the Zhejiang Provincial Laboratory
   Animal Center (Hangzhou, China), and their use was approved by the
   Laboratory Animal Ethics Committee of Wenzhou Medical University (No.
   WIUCAS23020208). Rats weighing 250–300 g were selected for needle
   puncture surgery. The needle gauge 22G was used. Needle punctures were
   performed at the 7th/8th caudal vertebrae (Co7/8) under radiographic
   guidance. A needle was inserted into the intervertebral disc space at a
   fixed depth of 3 mm. After insertion, the needle was rotated 360° and
   left in place for 30 s. Following the needle puncture procedure,
   exosomes from the different treatment groups were injected into the
   central intervertebral disc space using a microinjector. Every 7 days,
   injections were administered at a dosage of 5 µl, with a concentration
   of 1 µg µl^−1. The schematic diagram of the rat acupuncture model and
   exosome injection can be found in Figure [194]S10 (Supporting
   Information). Experimental procedures involving animal models and
   exosome injection can be found in Figure [195]S11 (Supporting
   Information). Tissue degeneration levels were assessed using the DHI
   method and graded according to the Pfirrmann grading system, aided by
   radiography and MRI. The grading was evaluated by three experts
   specializing in orthopedics.

Hypoxia Treatment and Exosome Extraction

   Cultured BMSCs that reached 70–80% confluence were subjected to a
   series of procedures. First, the cells were rinsed three times with
   PBS, and the medium was replaced with basal medium (OriCell)
   supplemented with 10% exosome‐free serum (OriCell). The cells were then
   cultured in a low‐oxygen incubator at 37 °C under a gas mixture of 2%
   O[2] and 5% CO[2] for 24 h. After incubation, the cell culture medium
   was collected. Subsequently, exosomes were extracted using differential
   ultracentrifugation and purification. The extraction process involved
   centrifugation at 300×g for 10 min to separate the cells into groups.
   Dead cells were then removed by centrifugation at 2000×g for 10 min,
   followed by the removal of cell debris at 10000×g for 30 min. The
   exosome precipitate was separated by ultracentrifugation to obtain
   exosome samples. Purification was then performed by washing the
   exosomes with PBS and subjecting them to ultracentrifugation at
   100000×g to remove residual proteins. The exosome surface markers
   TSG101, HSP70, CD63 and Calnexin were detected using western blot
   analysis. The morphological characteristics of the exosomes were
   observed by using TEM (Hitachi, Tokyo, Japan). The sizes of the
   exosomes were determined by nanoparticle tracking analysis. To verify
   endocytosis efficiency, exosomes were labeled with PKH26 and PKH67 dyes
   (Solarbio, Beijing, China) and then internalized into NPCs.

ROS Measurements

   Dihydroethidium (DHE; Beyotime) and DCFH‐DA (Beyotime) staining were
   used for ROS detection. The final concentration of the working solution
   was diluted to 10 µmol L^−1, according to the manufacturer's
   instructions, and was subsequently added to the NPCs, which were
   incubated for 20 min at 37 °C. The cells were then removed, washed
   three times with PBS, and observed and photographed using a confocal
   microscope (Olympus, Tokyo, Japan).

Mitochondrial Analysis

   NPCs were seeded onto confocal dishes and co‐incubated with MitoTracker
   Red (Invitrogen) for 30 min. The cells were then observed and imaged
   using a confocal microscope. JC‐1 (C2006; Beyotime) was used to analyze
   the membrane potential of living cells. NPCs were labeled with JC‐1 for
   30 min, followed by observation and imaging. MitoSox reagent
   (40778ES50; Yeasen Biotechnology, Shanghai, China) was used to measure
   superoxide production, and the powder was dissolved into dimethyl
   sulfoxide to prepare a 5 mm stock solution, which was then diluted to a
   5 µm working solution. The resulting working solution was added to a
   12‐well plate containing the cells. In addition to MitoSox staining,
   all cell nuclei were stained with 4′,6‐diamidino‐2‐phenylindole (DAPI;
   Solarbio, C0065) and visualized using a laser confocal microscope
   (Olympus).

β‐Gal Cell Senescence Staining

   A Senescence β‐Galactosidase Staining Kit (C0602; Beyotime) was used,
   according to the manufacturer's instructions, to assess cellular
   senescence in the fixed NPCs. Specifically, 1 mL of the staining
   fixative was added to the NPCs, which were then treated and fixed for
   15 min at room temperature. Following this, the staining working
   solution was prepared according to the recommended ratio; NPCS were
   incubated with this solution overnight at 37 °C. Subsequently, stained
   cells were observed and counted under a light microscope (Olympus,
   Tokyo, Japan).

EdU Cell Proliferation Staining

   Cell proliferation was assessed using the Cell‐Light EdU Apollo In
   Vitro Kit ([196]C10310; Ribobio, Guangzhou, China), which employs EdU
   labeling. EdU, an analog of thymine deoxyribonucleotide that can
   substitute thymine deoxyribonucleoside (T) in the cell nucleus during
   the proliferative phase, was introduced into the treated NPCs.
   Co‐incubated cells were fixed with 4% paraformaldehyde, washed, and
   permeabilized. Subsequently, the Apollo staining reaction solution was
   used for staining. The cells were then incubated at room temperature,
   shielded from light, and decolorized on a shaker for 30 min. Later, the
   NPCs were labeled with Hoechst33342 for 30 min, and the resulting
   images were observed under a fluorescence microscope.

Western Blotting

   Cellular protein extraction was performed by placing cells in a RIPA
   lysis and extraction buffer supplemented with protease and phosphatase
   inhibitors (Thermo Fisher Scientific). The protein concentration was
   normalized using a BCA protein assay kit (Beyotime). Subsequently,
   8–12% Sodium dodecyl‐sulfate polyacrylamide gel electrophoresis gels
   were used for electrophoresis, and the separated proteins were
   transferred onto polyvinylidene fluoride membranes. The membranes were
   then sealed using a rapid‐closure solution (Beyotime) at room
   temperature for 20 min before being incubated with the primary antibody
   overnight at 4 °C. After incubation with the appropriate secondary
   antibody at room temperature for 1 h, the membranes were treated with
   an Enhanced ECL Chemiluminescence Detection Kit (Vazyme, Nanjing,
   China). For detailed information on the antibodies used, please refer
   to the Table [197]S4 (Supporting Information).

Quantitative Real‐Time PCR

   NPCs were grown in 6‐well plates, and after different treatments, total
   RNA was extracted using a Total RNA Extraction Kit (Beyotime). cDNA was
   synthesized using a PrimeScript RT kit (Takara, Kyoto, Japan) according
   to the manufacturer's instructions. An RT‐PCR system (LightCycler 96;
   Roche, Basel, Switzerland) was used for amplification. The RT‐PCR
   parameters were as follows: 40 cycles of 30 s at 95 °C, 5 s at 95 °C,
   and 30 s at 60 °C. Primer sequences are available in the Table [198]S5
   (Supporting Information).^[ [199]^38 ^]

Cell Transfection

   To silence HIF‐1α, BNIP3, ANXA2, and TFEB, targeted mRNA and their
   negative control small interfering RNAs (siRNAs) (Ribbio, China) were
   transiently transfected, according to the manufacturer's protocol. All
   the sequences are listed in Table [200]S6 (Supporting Information).

Immunofluorescence

   NPCs were seeded in confocal dishes and subjected to various
   treatments. Subsequently, the cells were washed three times with PBS
   and fixed with 4% paraformaldehyde. Cell membranes were permeabilized
   by treatment with 0.1% Triton X‐100, followed by blocking with 5% goat
   serum albumin. The primary antibodies used were against MMP13
   (#18165‐1‐AP; Proteintech, Rosemont, IL, USA), COL‐II (#ab307674;
   Abcam, Cambridge, UK), BNIP3 (#sc56167; Santa Cruz Biotechnology,
   Dallas, TX, USA), TOMM20 (#42 406; Cell Signaling Technology, Danvers,
   MA, USA), ANXA2 (#11256‐1‐AP; Proteintech), and TFEB (#13372‐1‐AP;
   Proteintech). The cells were incubated with primary antibodies and
   subsequently treated with FITC‐labeled goat anti‐rabbit or anti‐mouse
   IgG (Beyotime) or Cy3‐labeled goat anti‐rabbit or anti‐mouse IgG
   (Beyotime). DAPI (Solarbio, C0065) was used to stain cell nuclei.

Co‐IP Followed by MS

   A co‐immunoprecipitation (Co‐IP) kit (Thermo Fisher Scientific) was
   used to isolate protein complexes. The lysate was subjected to gentle
   centrifugation and incubated overnight at 4 °C with the primary
   antibody targeting the protein of interest, BNIP3, following the
   manufacturer's instructions. The next day, antigen/antibody complexes
   were captured using Pierce protein A/G beads (Thermo Fisher Scientific)
   and incubated for 1 h at room temperature. The beads were then washed
   twice with IP lysis/washing buffer and once with purified water. The
   antigen/antibody complexes were eluted and subjected to western blot
   analysis. The eluted products were further analyzed using an EASY‐nLC
   liquid chromatography system with an HFX mass spectrometer (Thermo
   Fisher Scientific), and the resulting raw files were directly imported
   into the Proteome Discoverer 2.5 software (Thermo Fisher Scientific)
   for database searching, peptide identification, and protein
   quantification.

X‐Ray and MRI Examinations

   X‐ray images of rat tails were obtained before surgery and 4 and 8
   weeks after surgery using a XPERT.8 X‐ray machine (Kubtec, Stratford,
   CT, USA). DHI was determined from the captured images according to the
   methodology outlined in the Appendix. Additionally, weighted MRI of the
   rat tails was conducted 4 and 8 w after surgery using an Achieva 3.0 T
   MRI system (Philips, Amsterdam, The Netherlands). The evaluation and
   analysis of the images focused on the high‐signal‐intensity region
   corresponding to the NP on T2‐weighted images in the sagittal plane.
   The degree of IVDD was assessed by a panel of more than three
   orthopedic fellows. They measured and evaluated the IVDD based on
   predetermined criteria.

Histological Analysis

   Human‐derived NP tissues were obtained and fixed in 4% paraformaldehyde
   (PFA). Following fixation, the tissue was dehydrated, embedded in
   paraffin, and sectioned into blocks. Immunofluorescence staining was
   performed on sections for further analysis. Rat intervertebral disc
   tissue was harvested from Co7/8 of the rat tails. The collected caudal
   vertebral tissues were fixed with PFA, followed by decalcification
   using 10% Ethylenediamine tetraacetic acid for 8 w. After
   decalcification, the tissues were dehydrated, paraffin‐embedded, and
   sectioned to a thickness of 5 µm. To observe morphological changes in
   the medulla, surrounding fibrous ring, and cartilage endplates, the
   sections were subjected to hematoxylin and eosin (HE), Safranin‐O/Fast
   Green (SF), and immunohistochemical (IHC) staining procedures.

Statistical Analysis

   All data were presented as mean ± standard deviation. Statistical
   analyses were performed using the Prism 9 software (GraphPad Software,
   La Jolla, CA, USA). To assess group differences, the data were
   subjected to statistical analysis using a two‐tailed Student's t‐test
   or one‐way analysis of variance, followed by a post‐hoc Tukey's
   multiple comparison test. Statistical significance was set at
   *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. “ns” denotes
   non‐significant differences.

Ethics Approval Statement

   The human experiments in this study were approved by the Medical Ethics
   Committee of the Second Affiliated Hospital of the Wenzhou Medical
   University (Wenzhou, China; approval number: LCKY2020‐157). The animal
   experiments in this study were approved by the Laboratory Animal Ethics
   Committee of Wenzhou Medical University (No. WIUCAS23020208).

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   Y.J., O.W., and Q.C. contributed equally to this work. Y.J., O.W, and
   Q. C. performed investigation, data curation, methodology, and wrote
   the original draft. L.C. and Z.Z. performed methodology and software.
   H.T., H.Z., and K.Z. performed conceptualization and investigation.
   J.G., X.W., Z.G., and J.S. performed formal analysis and validation.
   K.K., M.J., Y.L., and X. W. performed supervision and visualization.
   E.Z. and P.M. performed wrote the original draft, and wrote, reviewed,
   and edited the draft. S.S. performed methodology and supervision. A.W.
   performed project administration, supervision, and provided resources.

Supporting information

   Supporting Information
   [201]ADVS-11-2404275-s001.docx^ (31.8MB, docx)

Acknowledgements