Introduction

Diabetes mellitus (DM) is a very common metabolic disorder characterized by hyperglycemia due to insulin deficiency, resistance or increased glucagon secretion1. It was estimated that DM affected approximately 463 million patients in 2019, and the number was predicted to rise to 578 million by 2030 and 700 million by 20452. Besides hyperglycemia, many complications of DM gradually develop, such as heart diseases, neuropathy, nephropathy and foot damage, causing disability and even death3. Diabetic foot ulcer (DFU) represents one of severe complications of DM and causes lower extremity amputation as infection progresses4. Almost 15% of diabetic patients are expected to develop DFU in their lifetime, and 5-24% of them will eventually suffer from amputation5. Therefore, promoting the wound healing of DFU is quite important for reducing the risk of severe infection and DFU management.

Exosomes are a class of nanosized extracellular vesicles with a diameter of 30-150 nm and released by most cells6, which can be found in body fluids including urine, saliva and plasma7. Exosomes contain cell constituents such as protein, lipid, DNA and RNA and deliver them into other cells by which exosomes play key roles in regulating the function and behavior of recipient cells8. Importantly, exosome-mediated intercellular communication is associated with wound-healing process in DFU. Adipose-derived stem cell-derived exosomes carrying Nrf2 facilitate the angiogenesis of endothelial progenitor cells and accelerate the wound healing of DFU via enhancing vascularization9. Exosomal lncRNA H19 derived from mesenchymal stem cells promote wound healing in DFU through reducing miR-152-3p and upregulating PTEN10. Xiong et al. reported that suppression of exosomal miR-15a-3p promoted the wound healing of DFU through activating the NOX5/ROS signaling11. However, studies on exosome function in DFU are still limited.

Circular RNA (circRNA) is a type of non-coding RNA forming a closed loop through covalent ligation of the 5’ and 3’ termini and exerts vital functions in various physiological processes and pathological conditions such as cancers, heart diseases and diabetes12,13. In recent years, emerging evidence has demonstrated the implication of circRNAs in DFU. Upregulation of hsa_circ_0084443 indicates that it is associated with DFU14. Adipose-derived stem cell-derived exosomes overexpressing mmu_circ_0000250 improve wound healing via inhibiting miR-128-3p and upregulating SIRT115. A recent study analyzed the GSE114248 dataset from GEO database for exploring differentially expressed circRNAs between DFU and normal wounds. Twenty-eight upregulated circRNAs and thirty-nine downregulated circRNAs were identified. Serum and exosomal hsa_circ_0000907, also known as circMYO9B, is a promising biomarker for early diagnosis of DFU16. However, the role and underlying mechanism of circMYO9B in DFU have not been reported.

In summary, the purpose of this study is to explore the roles of mesenchymal stem cell (MSC)-derived exosomes carrying circMYO9B and molecular mechanism. Here, we demonstrate that MSC-derived exosomes promote endothelial cell proliferation, migration and angiogenesis and accelerate the wound healing process of DFU via delivering circMYO9B and consequently targeting the hnRNPU/CBL/KDM1A/VEGFA axis. Our study elucidates exosomal circMYO9B-mediated regulation of wound healing of DFU and provides potential exosome-based therapies for DFU management.

Results

Characterization of MSCs and MSC-derived exosomes

Human bone marrow MSCs (BMSCs) were isolated from healthy donors and cultured for morphological observation. Cells began to adhere to the bottom of flasks and exhibit a fusiform form at day 3 (Supplementary Fig. 1a). Moreover, cells grew in a swirly or clustered manner with distinct nucleus, exhibiting typical MSC characteristics (Supplementary Fig. 1a). We evaluated the differentiation potential of BMSCs into osteoblasts, adipocytes and chondrocytes. BMSCs overlapped and showed calcified nodules with calcium deposition after osteogenic induction (Supplementary Fig. 1b). After adipogenic induction, lipid was accumulated, and beaded lipid droplets were observed (Supplementary Fig. 1b). Glycosaminoglycan synthesis was confirmed by Alcian blue staining after chondrogenic differentiation (Supplementary Fig. 1b). Moreover, cells were almost 100% positive for CD105, CD90 and CD73 but negative for HLA-DR, CD45, CD34 and CD19 (Supplementary Fig. 1c), presenting typical surface antigen expression pattern of MSCs. These observations validated that BMSCs possessed typical characteristics and pluripotency of MSCs.

Exosomes derived from BMSCs (BMSC-exos) were mainly round vesicles and 30–120 nm in diameter (Supplementary Fig. 2a). Dynamic light scattering showed that the predominant diameter of exosomes was approximately 120 nm (Supplementary Fig. 2b). More than 90% of exosomes were positive for CD63 (Supplementary Fig. 2c), an exosome surface marker17. Besides, compared to BMSCs, BMSC-derived exosomes showed increased expression of exosome markers including Alix, TSG101, CD81 and CD9 and decreased Calnexin expression (Supplementary Fig. 2d), showing typical exosome characteristics. In addition, we observed that PKH-26-labeled exosomes could enter HUVECs (Supplementary Fig. 2e), indicating that exosomes could be internalized by endothelial cells. Besides, we checked the purity of exosomes via examining the abundance of non-exosomal biomarkers including TFIIB, Lamin A/C, GM130 and Cytochrome c. We observed that all of them could be detected in MSCs but not in exosomes (Supplementary Fig. 2f), suggesting high purity of isolated exosomes. The concentration of BMSC-exos was approximately 6.2 × 109 exosomes/mL determined with dynamic light scattering (Supplementary Fig. 2g). Sucrose density gradient fractionation was applied for exosome purification from BMSCs, and the vesicles were isolated in a subfraction at a density of 1.12 g/mL (Supplementary Fig. 2h).

MSC-derived exosomes affect the viability, proliferation, apoptosis, migration and angiogenesis of HG-induced endothelial cells

To investigate the biological activity of MSC-derived exosomes in diabetic wound healing, HUVECs were co-cultured with BMSC-derived exosomes and treated with normal glucose (NG, 5.5 mM) or high glucose (HG, 15 and 30 mM). As shown in Fig. 1a, high glucose (30 mM) severely impaired HUVEC viability with increased treatment time compared to NG treatment, and moderate decrease of cell viability was also observed after HG (15 mM) treatment. Decreased HUVEC viability was largely rescued by BMSC-derived exosomes (Fig. 1b). Moreover, HUVECs exhibited reduced proliferation and enhanced apoptosis after HG treatment, but these effects were abrogated by exosomes (Fig. 1c, d). Transwell assays showed that HG suppressed the migratory capacity of HUVECs, but MSC-exos promoted the motility of HUVECs (Fig. 1e). More tubes were formed in HG-induced HUVECs co-cultured with MSC-derived exosomes (Fig. 1f), suggesting that MSC-exos promoted the angiogenesis of HUVECs. As VEGFA is a master regulator in angiogenesis18, we found that HG inhibited VEGFA secretion of HUVECs, but MSC-exos enhanced VEGFA secretion (Fig. 1g). HG-mediated suppression of VEGFA and other angiogenesis-related markers including VEGFR219, PDGF20 and EGF21 were largely reversed by MSC-exos (Fig. 1h). These data indicated that MSC-derived exosomes regulated the viability, proliferation, apoptosis, migration and angiogenesis of HG-treated HUVECs.

Fig. 1: MSC-derived exosomes affect the viability, proliferation, apoptosis, migration and angiogenesis of HG-induced endothelial cells.
figure 1

a HUVECs were treated with normal glucose (5.5 mM, NG) or high glucose (15 or 30 mM, HG) for 0, 24, 48 or 72 h, and cell viability was examined with MTT (n = 3). HG-treated HUVECs were co-cultured with BMSC-exos or equal volume of PBS. NG and HG-treated cells were used as controls. b HUVEC viability was examined with MTT (n = 3). c EdU incorporation analysis of HUVECs (n = 3, scale bar, 100 µm). d Cell apoptosis analysis by TUNEL (n = 3, scale bar, 100 µm). e Transwell assays for HUVEC migration analysis (n = 3, scale bar, 100 µm). f Tube formation analysis of HUVECs (n = 3, scale bar, 100 µm). g The concentration of VEGFA in the supernatants of HUVECs (n = 3). h The expression of VEGFA, VEGFR2, PDGF and EGF was assessed by western blot (n = 3). GAPDH was used a normalization control. n = 3 represents biologically independent experiments. Data were presented as mean ± SD. *P < 0.05, **P < 0.01 and ***P < 0.001.

Full size image

MSC-exos carrying circMYO9B promotes the proliferation, migration and angiogenesis of HG-induced endothelial cells

As circMYO9B is implicated in DFU, we examined and found high abundance of circMYO9B in BMSC-derived exosomes (Fig. 2a), suggesting the possibility that MSC-exos exerted function through delivering circMYO9B. Subsequently, circMYO9B was overexpressed in cells transfected with oe-circMYO9B but silenced through sh-circMYO9B transfection (Fig. 2b). Moreover, exosomes were extracted, and the expression of circMYO9B was increased in exosomes derived from circMYO9B-overexpressing MSCs but decreased in exosomes derived from circMYO9B-silencing MSCs (Fig. 2b). Exosomes were co-cultured with HUVECs in the presence of HG. Compared to HG-treated control cells, the addition of Exo-oe-NC or Exo-sh-NC increased EdU incorporation (Fig. 2c), reduced apoptotic cell rate (Fig. 2d) and promoted the migration (Fig. 2e) and tube formation (Fig. 2f) of HG-treated HUVECs. Intriguingly, Exo-oe-circMYO9B exhibited strongest pro-proliferative, anti-apoptotic, pro-migratory and pro-angiogenic effects (Fig. 2c–f). In contrast, Exo-sh-circMYO9B dramatically suppressed HUVEC proliferation, migration and angiogenesis but enhanced its apoptosis (Fig. 2c–f). These data implied that BMSC-exos carrying circMYO9B promoted the proliferation, migration and angiogenesis of HG-induced endothelial cells and inhibited cell apoptosis. Additionally, the secretion and expression of VEGFA were enhanced by Exo-oe-NC, Exo-sh-NC or Exo-oe-circMYO9B. and Exo-oe-circMYO9B showed strongest effect, but Exo-sh-circMYO9B reduced its secretion and expression (Fig. 2g, h). The expression of angiogenesis-related markers was upregulated by Exo-oe-circMYO9B but downregulated by Exo-sh-circMYO9B (Fig. 2h), suggesting that VEGFA might be the potential target of MSC-exos carrying circMYO9B.

Fig. 2: MSC-exos carrying circMYO9B promotes the proliferation, migration and angiogenesis of HG-induced endothelial cells.
figure 2

a RT-qPCR analysis of circMYO9B in BMSCs and BMSC-exos (n = 3). b RT-qPCR analysis of circMYO9B in BMSCs (control), BMSCs transfected with oe-NC, oe-circMYO9B, sh-NC or sh-circMYO9B (n = 3) and BMSC-Exos. HG-treated HUVECs were co-cultured with Exo-oe-NC, Exo-oe-circMYO9B, Exo-sh-NC or Exo-sh-circMYO9B. HG-treated HUVECs were used as control cells. c EdU incorporation analysis of HUVECs (n = 3, scale bar, 100 µm). d Cell apoptosis analysis by TUNEL (n = 3, (scale bar, 100 µm). e Transwell assays for HUVEC migration analysis (n = 3, scale bar, 100 µm). f Tube formation analysis of HUVECs (n = 3, scale bar, 100 µm). g The concentration of VEGFA in the supernatants of HUVECs (n = 3). h The expression of VEGFA, VEGFR2, PDGF and EGF was assessed by western blot (n = 3). GAPDH was used a normalization control. n = 3 represents biologically independent experiments. Data were presented as mean ± SD. *P < 0.05, **P < 0.01 and ***P < 0.001.

Full size image

Circular characterization of circMYO9

As illustrated in Supplementary Fig. 3a, circMYO9B is generated from the exon 2 of MYO9B transcript on chromosome 19 through back-splicing and has a length of 890 nucleotides (nt), and the junction site was identified by Sanger sequencing. Linear MYO9B and GAPDH were successfully amplified from both cDNA and gDNA from MSCs, but circMYO9B was only amplified from cDNA, not from gDNA (Supplementary Fig. 3b). Moreover, we found that circMYO9B was highly resistant to RNase R digestion or actinomycin D, while linear MYO9B and GAPDH were efficiently digested, and circMYO9B exhibited longer half-life than linear MYO9B and GAPDH mRNAs (Supplementary Fig. 3c, d), demonstrating that circMYO9B had very stable circular structure. There observations confirmed the circular characterization of circMYO9B.

MSC-exos carrying circMYO9B regulates the proliferation, migration and angiogenesis of HG-induced endothelial cells through KDM1A

As KDM1A regulates VEGFA expression22, we knocked down KDM1A in HUVECs, and VEGFA expression was reduced in KDM1A-knockdown HUVECs (Fig. 3a). Three potential binding sites (BS1 -1485/-1470, BS2 -289/-286 and BS3 -226/-205) for KDM1A in the promoter of VEGFA were predicted via AnimalTFDB (Fig. 3b). Further ChIP assays showed that KDM1A could bind to BS3, not BS1 and BS2 (Fig. 3c). Wildtype and BS3-mutated promoter were inserted into the pGL3 luciferase reporter vector (VEGFA-WT and VEGFA-MUT), and we found that knockdown of KDM1A suppressed the luciferase activity of VEGFA-WT but did not affect the luciferase activity of VEGFA-MUT (Fig. 3d). These data suggested that KDM1A directly bound to the promoter of VEGFA and promoted its expression. Subsequently, we examined whether MSC-exos carrying circMYO9B-mediated regulation was dependent on KDM1A. Exos-oe-circMYO9B-induced upregulation of KDM1A was obviously reduced by sh-KDM1A in HUVECs, and Exo-sh-circMYO9B-mediated suppression of KDM1A was reversed by KDM1A overexpression (Fig. 3e, f). Exo-oe-circMYO9B-mediated accelerative effects on HUVEC proliferation, migration and tube formation and inhibition of cell apoptosis were reversed by KDM1A knockdown (Fig. 3g–j). Consistently, Exo-sh-circMYO9B-mediated inhibition of cell proliferation, migration and tube formation and promotion of cell apoptosis were also rescued by KDM1A overexpression (Fig. 3g–j). Moreover, Exo-oe-circMYO9B-induced secretion of VEGFA and Exo-sh-circMYO9B-mediated reduced VEGFA secretion were abolished by knockdown and overexpression of KDM1A, respectively (Fig. 3k). These observations indicated that MSC-derived exosomal circMYO9B regulated the proliferation, migration and angiogenesis of HG-treated HUVECs via targeting KDM1A.

Fig. 3: MSC-exos carrying circMYO9B regulates the proliferation, migration and angiogenesis of HG-induced endothelial cells via upregulating KDM1A.
figure 3

a KDM1A and VEGFA were detected in HUVECs transfected with sh-NC or sh-KDM1A via western blot (n = 3). GAPDH was used a normalization control. b Three potential binding sites (BS1: −1485/−1470, BS2: −289/−268 and BS3: −226/−205) for KDM1A in VEGFA. c ChIP assays for evaluating the interaction between KDM1A and BS1, BS2 or BS3 (n = 3). d Luciferase activity of VEGFA-WT and VEGFA-MUT reporters in HUVECs co-transfected with sh-NC or sh-KDM1A (n = 3). HUVECs transfected with sh-NC, sh-KDM1A, vector or KDM1A were co-cultured with Exo-oe-circMYO9B or Exo-sh-circMYO9B. HG-treated HUVECs were used as control cells. The expression of KDM1A was assessed by RT-qPCR (e, n = 3) and western blot (f, n = 3). GAPDH was used a normalization control. g EdU incorporation analysis of HUVECs (n = 3, scale bar, 100 µm). h Cell apoptosis analysis by TUNEL (n = 3, scale bar, 100 µm). i Transwell assays for HUVEC migration analysis (n = 3, scale bar, 100 µm). j Tube formation analysis of HUVECs (n = 3, scale bar, 100 µm). k The concentration of VEGFA in the supernatants of HUVECs (n = 3). n = 3 represents biologically independent experiments. Data were presented as mean ± SD. *P < 0.05, **P < 0.01 and ***P < 0.001.

Full size image

CircMYO9B binds to hnRNPU to enhance its cytoplasmic translocation in endothelial cells

RNA pull-down assays were performed to explore targets of circMYO9B, and hnRNPU could be pulled down by circMYO9B, but the antisense circMYO9B (As-circMYO9B) did not pulled hnRNPU down (Fig. 4a). Moreover, RNA immunoprecipitation (RIP) assays showed increased enrichment of circMYO9B in the complex immunoprecipitated by the hnRNPU antibody, confirming the direct interaction between circMYO9B and hnRNPU (Fig. 4b). The cytoplasmic and nuclear fractions of HUVECs were separated, and circMYO9B was localized in both cytoplasm and nucleus but mainly distributed in the cytoplasm (Fig. 4c). To investigate whether circMYO9B regulates hnRNPU expression, circMYO9B was overexpressed in HUVECs. No change of protein levels of hnRNPU was observed after circMYO9B overexpression (Fig. 4d). However, we found that overexpression of circMYO9B enhanced the cytoplasmic distribution of hnRNPU but reduced its distribution in the nucleus (Fig. 4e). IF assays showed increased cytoplasmic abundance of hnRNPU in HUVECs with circMYO9B overexpression (Fig. 4f). To conclude, circMYO9B bound to hnRNPU and facilitated its cytoplasmic translocation in HUVECs.

Fig. 4: CircMYO9B binds to hnRNPU to enhance its cytoplasmic translocation in endothelial cells.
figure 4

a Western blot analysis of hnRNPU in samples pulled down by circMYO9B or AS-circMYO9B and input (n = 3). b The enrichment of circMYO9B by an hnRNPU antibody or a normal IgG (n = 3). c RT-qPCR analysis of the abundance of circMYO9B, U6 and GAPDH in cytoplasmic and nuclear fractions (n = 3). d Western blot analysis of hnRNPU in HUVECs transfected with oe-NC or oe-circMYO9B (n = 3). Tubulin was used as a control. e Western blot analysis of hnRNPU in cytoplasmic and nuclear fractions (n = 3). Tubulin and Lamin B were used as cytoplastic and nuclear controls. f IF staining of hnRNPU (red) in HUVECs transfected with oe-NC or oe-circMYO9B (n = 3, scale bar, 100 µm). Blue, DAPI, nucleus. n = 3 represents biologically independent experiments. Data were presented as mean ± SD. ***P < 0.001.

Full size image

CircMYO9B inhibits CBL expression by forming the circMYO9B-hnRNPU-CBL mRNA complex

As a RNA binding protein, hnRNPU binds to target mRNAs and regulates their stability23,24. To explore the mechanism by which circMYO9B upregulates VEGFA expression, we examined the interaction between hnRNPU and VEGFA and found that hnRNPU did not directly interacted with VEGFA (Fig. 5a). Overexpression of circMYO9B did not affect ZNF24 expression, a transcriptional regulator of VEGFA25, and no change of the mRNA level of KDM1A was observed (Fig. 5b, c). However, circMYO9B overexpression reduced the protein level of KDM1A (Fig. 5b, c), suggesting that circMYO9B might regulate KDM1A expression at a post-translational level. IP assays showed that hnRNPU did not directly interacted with KDM1A (Fig. 5d). E3 ubiquitin ligases which potential modify KDM1A were predicted through UbiBrowser, and CBL was selected due to its highest reliability (Fig. 5e). We found that overexpression of circMYO9B downregulated CBL in HUVECs (Fig. 5f, g). Furthermore, we observed increased enrichment of circMYO9B and CBL by the hnRNPU antibody, additionally, overexpression of circMYO9B promoted the interaction between hnRNPU and CBL mRNA (Fig. 5h), and overexpression of hnRNPU reduced CBL expression in HUVECs (Fig. 5i). Overexpression of circMYO9B enhanced CBL mRNA decay in HUVECs in response to actinomycin D (Fig. 5j). Taken together, these data demonstrated that circMYO9B destabilized CBL mRNA and reduced its expression through forming the circMYO9B-hnRNPU-CBL mRNA complex.

Fig. 5: CircMYO9B inhibits CBL expression by forming the circMYO9B-hnRNPU-CBL mRNA complex.
figure 5

a The interaction between hnRNPU and VEGFA mRNA was evaluated by RIP assays (n = 3). b, c RT-qPCR and western blot analysis of KDM1A and ZNF24 in HUVECs transfected with oe-NC or oe-circMYO9B (n = 3). GAPDH was used a normalization control. d The interaction between hnRNPU and KDM1A was analyzed by IP assays (n = 3). e Potential E3 ubiquitin ligases which interacted with KDM1A. f, g RT-qPCR and western blot analysis of CBL in HUVECs transfected with oe-NC or oe-circMYO9B (n = 3). GAPDH was used a normalization control. h Enrichment of circMYO9B, CBL and U6 in the hnRNPU antibody or normal IgG-immunoprecipitated fractions was analyzed by RIP assays (n = 3). i Western blot analysis of hnRNPU and CBL in HUVECs transfected with oe-NC or oe-hnRNPU (n = 3). GAPDH was used a normalization control. j The stability of CBL mRNA in HUVECs transfected with oe-NC or oe-circMYO9B after Act D treatment for 0, 2, 4, 6 or 8 h. n = 3 represents biologically independent experiments. Data were presented as mean ± SD. **P < 0.01 and ***P < 0.001.

Full size image

Exosomal circMYO9B promotes the proliferation, migration and angiogenesis of HG-induced endothelial cells via suppressing CBL

To investigate whether CBL regulates KDM1A ubiquitination, CBL was knocked down in HUVECs. We observed that knockdown of CBL markedly reduced the ubiquitination of KDM1A and increased its half-time in HUVECs (Fig. 6a, b). Furthermore, overexpression of circMYO9B reduced the ubiquitination of KDM1A in HUVECs (Fig. 6c). Overexpression of circMYO9B promoted the expression of KDM1A and VEGFA but suppressed CBL expression in HUVECs, and their expression were restored by simultaneous overexpression of CBL (Fig. 6d). These data demonstrated that circMYO9B upregulated VEGFA expression through regulating CBL expression and subsequent KDM1A ubiquitination in endothelial cells. Exo-oe-circMYO9B-mediated CBL downregulation and upregulation of KDM1A and VEGFA were abrogated by overexpression of CBL in HG-treated HUVECs (Fig. 6e). Besides, overexpression of CBL reversed Exo-oe-circMYO9B-mediated promotion of HUVEC proliferation, migration and tube formation and suppression of cell apoptosis (Fig. 6f-i). Our findings demonstrated that MSC-derived exosomal circMYO9B regulated HUVEC proliferation, migration and angiogenesis dependent on CBL.

Fig. 6: Exosomal circMYO9B promotes the proliferation, migration and angiogenesis of HG-induced endothelial cells via suppressing CBL.
figure 6

a Western blot analysis of the ubiquitination of immunoprecipitated KDM1A in HUVECs transfected with si-NC or si-CBL (n = 3). b The stability of KDM1A in HUVECs transfected with si-NC or si-CBL after CHX treatment for 0, 2, 4, 6 or 8 h (n = 3). c Western blot analysis of the ubiquitination of immunoprecipitated KDM1A in HUVECs transfected with oe-NC or oe-circMYO9B (n = 3). d Western blot analysis of the expression of CBL, VEGFA and KDM1A in HUVECs transfected with oe-NC, oe-circMYO9B, oe-circMYO9B+oe-NC, oe-circMYO9B+oe-CBL (n = 3). GAPDH was used as a normalization control. HUVECs were transfected with oe-NC or oe-CBL and treated with HG, and Exo-oe-circMYO9B or Exo-sh-circMYO9B was added for coculture. HG-treated HUVECs were used as control cells. e Western blot analysis of CBL, VEGFA and KDM1A. GAPDH was used a normalization control. f EdU incorporation analysis of HUVECs (n = 3, scale bar, 100 µm). g Cell apoptosis analysis by TUNEL (n = 3, scale bar, 100 µm). h Transwell assays for HUVEC migration analysis (n = 3, scale bar, 100 µm). i Tube formation analysis of HUVECs (n = 3, scale bar, 100 µm). n = 3 represents biologically independent experiments. Data were presented as mean ± SD. *P < 0.05, **P < 0.01 and ***P < 0.001.

Full size image

MSC-exos carrying circMYO9B accelerates wound healing in mice with DFU

To evaluate the role of MSC-exos in regulating diabetic wound healing in vivo, we established a DFU model in db/db mice. Mice were divided into PBS, MSC and MSC-GW4869 groups. In the MSC-GW4869 group, MSCs were pre-treated with GW4869, a widely used inhibitor of exosome generation. Compared to PBS, MSC reduced wound area, but MSC-GW4869 decelerated wound healing (Supplementary Fig. 4a). H&E and Masson’s Trichrome staining showed that MSC-GW4869 mice exhibited a suppressive effect on would healing compared to MSC mice, including increased inflammatory cells around the wound, thinned granulation tissue and increased collagen deposition (Supplementary Fig. 4b). Thus, MSC-mediated protective effects were obviously inhibited by GW4869 treatment (Supplementary Fig. 4a, b). Compared to MSC, MSC-GW4869 obviously increased apoptotic cells and reduced CD31 expression (Supplementary Fig. 4c, d). In addition, MSC-GW4869 enhanced CBL expression and reduced the expression of KDM1A and VEGFA in vivo, whereas MSC showed an opposite effect, decreasing CBL expression and promoting the expression of KDM1A and VEGFA (Supplementary Fig. 4e). Therefore, MSC-derived exosomes greatly promoted wound healing in DFU.

Besides, compared to PBS, MSC-Exo-oe-NC and MSC-Exo-oe-circMYO9B reduced wound area (Fig. 7a). MSC-Exo-oe-circMYO9B accelerated wound healing, including decreased inflammatory cells around the wound, thickened granulation tissue and reduced collagen deposition (Fig. 7a, b). MSC-Exos-mediated protective effects were obviously inhibited by circMYO9B knockdown (Fig. 7a, b). Compared to MSC-Exo-oe-NC, MSC-Exo-oe-circMYO9B obviously reduced apoptotic cells and enhanced CD31 expression, whereas MSC-Exo-sh-circMYO9B enhanced cell apoptosis and reduced CD31 expression in wound tissues (Fig. 7c, d). In addition, MSC-Exo-oe-circMYO9B reduced CBL expression and enhanced the expression of KDM1A and VEGFA in vivo, while MSC-Exo-sh-circMYO9B promoted CBL expression but downregulated KDM1A and VEGFA (Fig. 7e). Therefore, MSC-derived exosomal circMYO9B greatly promoted wound healing in DFU possibly via circMYO9B secretion.

Fig. 7: MSC-exos carrying circMYO9B greatly accelerates wound healing in mice with DFU.
figure 7

A mouse model of DFU was established, and PBS (control), MSC-Exo-oe-NC, MSC-Exo-oe-circMYO9B or MSC-Exo-sh-circMYO9B was injected into surrounding areas of wound (n = 8). a Representative images and calculation of wound healing rates at day 0, 3, 7 and 13. b H&E and Masson’s Trichrome staining of wound tissues (scale bar, 100 µm). c Cell apoptosis analysis in wound tissues by TUNEL (scale bar, 100 µm). d IHC staining of CD31 in wound tissues (scale bar, 100 µm). e Western blot analysis of the expression of CBL, KDM1A and VEGFA in wound tissues (n = 8). GAPDH was used as a normalization control. n = 8 represents biologically independent animals. Data were presented as mean ± SD. *P < 0.05, **P < 0.01 and ***P < 0.001.

Full size image

Discussion

As a devastating complication of DM, DFU is No.1 cause of hospitalization in diabetic patients and non-traumatic amputation globally26,27. High blood glucose causes abnormal angiogenesis and slows down wound healing of DFU28. In this study, we reported that MSC-derived exosomal circMYO9B promoted the proliferation, migration and angiogenesis of HG-treated HUVECs and reduced cell apoptosis through translocating hnRNPU and subsequently targeting the CBL/KDM1A/VEGFA axis. In vivo assays showed that exosomes carrying circMYO9B promoted the wound healing process of DFU in diabetic mice. Collectively, MSC-derived exosomal circMYO9B accelerated wound healing of DFU by enhancing angiogenesis through the hnRNPU/CBL/KDM1A/VEGFA axis.

Exosomes carrying ncRNAs, especially miRNAs and lncRNAs, are emerging as pivotal regulators in the pathogenesis of DM29. Intriguingly, exosomes regulate wound healing through various mechanisms. Adipose-derived stem cell-derived exosomes facilitate wound healing via promoting vascularization9. MSC-derived exosomes contribute to diabetic wound healing via promoting angiogenesis30 or adjusting macrophage polarization31. In consistence, we found that MSC-derived exosomes improved wound healing of DFU via enhancing angiogenesis. Because of excellent biocompatibility and low immunogenicity, exosomes secreted by MSCs have potency to be engineered as a drug cargo for DFU management. As exosomes were pelleted after centrifugation, we examined whether the supernatant improved cell viability, proliferation, and migration. As shown in Supplementary Fig. 5, HG-induced impaired viability, proliferation and migration, and increased apoptosis could be improved by MSC-derived exosomes but not by the supernatant (Supplementary Fig. 5a–d), showing that the supernatant has a similar effect as that of PBS.

Recent studies have shown that circRNAs are implicated in angiogenesis and neovascularization32. Circ-Amotl1 enhances wound healing via increasing the nuclear translocation of STAT3 and regulating the expression of Dnmt3a and miR-17-5p33. Dysregulated circRNAs are closely associated with the development of diabetic complications including DFU34,35. Differentially expressed circRNAs between DFU and normal wounds were identified via analyzing the GSE114248 dataset, and further assays indicated serum and exosomal circMYO9B as potential biomarkers for early diagnosis of DFU16. However, the implication of circMYO9B in regulating wound repair of DFU is still unknown. We firstly reported that exosomal circMYO9B promoted the proliferation, migration and angiogenesis of endothelial cells and improved diabetic wound repair. Furthermore, knockdown of circMYO9B inhibited cell proliferation, migration and angiogenesis. Our study identifies an important circRNA regulator in diabetic wound healing.

The competitive endogenous RNA (ceRNA) hypothesis suggests that circRNAs exert their activities via functioning as miRNA sponges to competitively bind to miRNAs, and thereby relieving miRNA-mediated suppression of downstream mRNA targets, known as a circRNA-miRNA-mRNA regulatory network36. Intriguingly, circRNAs also exert their function through interacting with proteins37. Wu and colleagues found that circFOXO3 suppressed the progression of cell cycle via interacting with p21 and CDK238. CircAmotl1 was reported to promote tumorigenesis via enhancing the nuclear translocation of c-myc39. More investigations are required to elucidate the regulatory network of circRNAs with proteins in wound healing of DFU. Here, we demonstrated that circMYO9B interacted with hnRNPU to promote its cytoplastic translocation in HUVECs.

As VEGFA is a pivotal regulator in angiogenesis40 and increased VEGFA expression was observed, we hypothesized that hnRNPU might regulate VEGFA expression. Previous studies have reported that KDM1A is required to maintain VEGFA expression22 and HG treatment reduced KDM1A expression in HUVECs (Supplementary Fig. 5e), suggesting the possibility that hnRNPU might regulate VEGFA expression through modulating KDM1A activity. After excluding direct interaction between hnRNPU and VEGFA, we found that hnRNPU interacted with CBL and reduce its expression via promoting CBL mRNA decay. In this study, we screened RNA binding proteins (RBPs) that bind to both circMYO9B and CBL through StarBase. Ten potential RBPs with high reliability including ELAVAL, G3BP1, SCAF8, PCBP2, CTCF, LIN28B, HNRNPA2B1, FUS, MBNL1, HNRNPU were identified. These RBPs were knocked down, and we observed that knockdown of ELAVAL, HNRNPA2B1 or HNRNPU could enhance CBL expression. Knockdown of HNRNPU showed the most obvious effect (Supplementary Fig. 6a). Subsequently, we examined the interaction of these three RBPs and circMYO9B through RIP. CircMYO9B was highly enriched by a hnRNPU antibody (Supplementary Fig. 6b). Based on these observations, we chose to study hnRNPU as the primary RBP for circMYO9B and CBL. CBL is an E3 ubiquitin ligase and has been reported to facilitate KDM1A ubiquitination and reduce its stability41. Consistently, we demonstrated that CBL promoted the ubiquitination and degradation of KDM1A, thus inhibiting VEGFA expression in endothelial cells. Therefore, circMYO9B promoted the translocation of hnRNPU from nucleus to cytoplasm where hnRNPU interacted with CBL and reduced its expression, and suppression of CBL stabilized KDM1A and promoted VEGFA expression.

To summarize, we firstly demonstrate that MSC-derived exosomal circMYO9B accelerates wound healing of DFU by translocating hnRNPU and enhancing VEGFA expression through the CBL/KDM1A axis in endothelial cells (Supplementary Fig. 7). Our study not only highlights exosomal circMYO9B-mediated facilitation of diabetic wound healing, but also provides potential therapeutic strategies for DFU management. To achieve this, further investigations are required to elucidate molecular mechanisms in detail. In addition, samples from DFU patients are required to clarify the pathological mechanism for clinical applications in future.

Methods

Cell isolation, culture and differentiation

Human bone marrow mesenchymal stem cells (BMSCs) were isolated from bone marrow aspirates from healthy donors42 and maintained in Dulbecco’s Modified Eagle Medium (DMEM, Gibco, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS, HyClone, Logan, UT, USA). BMSCs were characterized by flow cytometry analysis of HLA-DR, CD105, CD90, CD45, CD34, CD19 and CD73. BMSCs were incubated with the following fluorescent monoclonal mouse anti-human antibodies (Biolegend (San Diego, CA, USA): CD105 PE; CD90 PC5.5; CD45 PE; CD34 PC5.5; CD19 APC; CD73 APC; HLA-DR FITC. BMSCs were trypsinized and resuspended in MSC osteogenic (PromoCell, Heidelberg, Germany), adipogenic (STEMCELL, Vancouver, Canada) or chondrogenic (Cyagen, Suzhou, Jiangsu, China) differentiation medium at 1 × 105 cells/mL in 6-well plates. Alizarin red, oil red O and Alcian blue staining were performed for identifying osteogenic, adipogenic and chondrogenic differentiation. Human umbilical vein endothelial cells (HUVECs) obtained from American Type Culture Collection (RRID: PCS-100-013) were authenticated by STR profiling and tested for mycoplasma contamination, and then cultured in Medium 200 (Gibco), Three doses of glucose were used to treat HUVECs including 5.5 mM (normal glucose, NG), 15 and 30 mM (high glucose, HG). HUVECs were cocultured with BMSC-derived exosomes (100 μg/mL) in some assays, which were divided into MSC-exos, Exos-oe-NC, Exos-oe-circMYO9B, Exos-sh-NC or Exos-sh-circMYO9B. For cycloheximide (CHX) treatment, cells were treated CHX (Sigma, St. Louis, MO, USA) at 50 mg/mL for 0, 2, 4, 6 or 8 h. Donors in this study were informed and provided written consent, and our study was approved by the Ethics Committee of the Third Xiangya Hospital Central South University.

Cell transfection

For overexpression of circMYO9B, KDM1A, CBL and hnRNPU, coding regions for circMYO9B, KDM1A, CBL and hnRNPU were inserted into pcDNA3.1 vector (oe-circMYO9B, oe-KDM1A, oe-CBL and oe-hnRNPU, ThermoFisher Scientific, Waltham, MA, USA). Empty pcDNA3.1 vector (oe-NC) was used as a negative control. The shRNAs against circMYO9B and KDM1A (sh-circMYO9B/sh-KDM1A), siRNA against CBL (si-CBL) and negative controls (sh-NC and si-NC) were provided by RiboBio (Guangzhou, China), and the target sequences of shcircMYO9B (5’-CTGTGCTGGAGCTCCAGGACA-3’) and shKDM1A(5’-GCCTAGACATTAAACTGAATA-3’), and si-CBL(5’-GCCTAGACATTAAACTGAATA-3’) were provided. BMSCs were transfected with oe-NC, oe-circMYO9B, sh-NC or sh-circMYO9B using Lipo 3000 (ThermoFisher Scientific). HUVECs were transfected with shNC, shKDM1A, oe-NC, oe-circMYO9B, si-NC, si-CBL, oe-circMYO9B + oe-NC or oe-circMYO9B + oe-CBL using Lipo 3000 (ThermoFisher Scientific). After 72 h, cells were harvested for subsequent assays.

Sucrose density gradient centrifugation

Exosomes were purified through sucrose density gradient fractionation. Briefly, exosomes were isolated, resuspended in sucrose (0.25 M) and loaded onto a gradient with layers of various concentrations of sucrose (2, 1.3, 1.16, 0.8, and 0.5 M sucrose). Subsequently, samples were ultra-centrifugated at 100,000 × g for 2.5 h. Subfractions containing different densities were extracted, and the subfraction containing exosomes was washed and ultra-centrifuged at 100,000 × g for 2 h. The concentrated exosomes were stored used for the subsequent assays.

Exosome isolation, characterization and internalization

Exosomes were isolated using the Total Exosome Isolation reagent (ThermoFisher Scientific) following the manual. Exosome morphology was examined by transmission electron microscopy (TEM, ThermoFisher Scientific). Dynamic light scattering was used to analyze the size distribution of exosomes. Exosomes were absorbed on 4 µm latex beads and stained with the PE-labeled CD63 antibody from BioLegend for flow cytometry analysis. Alix, TSG101, CD81, CD9, Calnexin, TFIIB, Lamin A/C, GM130 and Cytochrome c were examined using western blot.

Exosomes were incubated with PKH26 dye (Sigma) at 1 µM for 20 min and cultured with HUVECs for 24 h. HUVECs were fixed in 3.7% formaldehyde solution and permeabilized in 0.2% Triton-X100. HUVECs were stained with Alexa Fluor 488-conjugated phalloidin (Cell Signaling Technology, Beverly, MA, USA) at 0.5 µM for 30 minutes. Cells were mounted in anti-fade mounting medium with DAPI (Beyotime, Shanghai, China) and imaged.

Real-time quantitative reverse transcription-PCR (RT-qPCR)

Total RNA was extracted from HUVECs and BMSCs with TRIzol reagent from ThermoFisher Scientific and quantified using Nanodrop 2000. Isolation of RNA from exosomes was performed using Total Exosome RNA and Protein Isolation Kit (ThermoFisher Scientific). Nuclear and cytoplasmic fractions of HUVECs were separated with NE-PER™ Nuclear and Cytoplasmic Extraction Reagents (ThermoFisher Scientific), and total RNA in nuclear and cytoplasmic fractions were extracted using TRIzol reagent. RNA was reversely transcribed into cDNA using iScript cDNA synthesis kit (Bio-Rad, Hercules, CA, USA). VEGFAKDM1A, ZNF24, CBL, circMYO9B, GAPDH and U6 snRNA were examined by quantitative real-time PCR. The relative expression was calculated using the 2−∆∆Ct method. Primers used in RT-qPCR were shown in Supplementary Table 1.

Protein extraction and western blot

Total Exosome RNA and Protein Isolation Kit (ThermoFisher Scientific) was used to extract protein from exosomes. HUVECs and wound tissue homogenates were lysed in RIPA lysis buffer (Yeasen, Shanghai, China). Protein was quantified with BCA kit (Beyotime), loaded for electrophoresis and transferred to polyvinylidene fluoride (PVDF) membranes (GE, Chicago, IL, USA). After block, membranes were incubated with rabbit antibodies against Alix (1:2000; ab275377, Abcam), TSG101 (1:1000; ab125011, Abcam), CD81 (1:2000; ab109201, Abcam), CD9(1:500; ab236630, Abcam), Calnexin (1:800; ab22595, Abcam), VEGFA (1:1000; ab46154, Abcam), VEGFR2 (1:1000; ab315238, Abcam), PDGF (1:500; ab234666, Abcam), EGF (1:1000; ab206106, Abcam), hnRNPU (1:10000;ab180952, Abcam), Tubulin (1:500; ab6046), Lamin B (1:500; ab32535, Abcam), CBL (1:1000; ab32027, Abcam), KDM1A (1:1000; ab17721; Abcam), TFIIB (1:1000; ab109106, Abcam), Lamin A/C (1:1000; ab108595, Abcam), GM130 (1:1000; ab52649, Abcam) and Cytochrome c (1:1000; ab133504, Abcam), ZNF24 (1:1000; ab176589, Abcam) and GAPDH (1:5000; ab8245, Abcam). Primary antibodies were purchased from Abcam (Cambridge, UK). Membranes were washed and incubated an HRP-labeled goat-anti rabbit IgG antibody (ThermoFisher Scientific) for 1 hour. DAB substrate (BOSTER, Wuhan, Hubei, China) was used to visualize bands, and band intensity was analyzed using ImageJ software.

MTT assay

HUVECs were seeded in 96-well plates and treated with NG, HG, HG + PBS or HG + MSC-exos for 0, 24, 48 or 72 h. Subsequently, the medium was replaced with fresh medium (100 µL). 10 µL of MTT (R&D Systems, Minneapolis, MN, USA) was added and incubated for 5 h. The insoluble formazan was reconstituted by adding 50 µL of DMSO, and the absorbance (490 nm) was recorded.

Tube formation assay

Matrigel (Corning, NY, USA) was added into 24-well plates (200 µL each well) for gelation at 37 °C. HUVECs were seeded on Matrigel (1 × 105 cells each well), treated and incubated for 10 h. Capillary-like structures were imaged using a BX51 microscope (Olympus, Tokyo, Japan).

EdU incorporation assay

HUVECs were seeded in 96-well plates at 1 × 105 cells per well. Five hours later, the medium was replaced with fresh medium supplemented with 10 µM EdU, and cells were incubated for 4 h. Cells were fixed in 4% formaldehyde solution and permeabilized in 0.3% Triton-X100. Incorporated EdU was detected using Click-iT reaction cocktail (ThermoFisher scientific). Cells were stained with Hoechst 33342.

$${{\rm{The}}}\; {{\rm{EdU}}}\; {{\rm{positive}}}\; {{\rm{rate}}}( \% )=\frac{{{\rm{the}}}\; {{\rm{number}}}\; {{\rm{of}}}\; {{\rm{positive}}}\; {{\rm{cells}}}}{{{\rm{the}}}\; {{\rm{number}}}\; {{\rm{of}}}\; {{\rm{total}}}\; {{\rm{cells}}}}\times 100 \%$$
(1)

Transwell migration assay

Transwell chambers with 8 µm pore membranes were purchased from Corning. Cells (1 × 106 cells/mL) were seeded on the upper chamber and incubated for 24 hours. Cells which migrated to the lower chamber were fixed in 4% formaldehyde solution and stained with 1% crystal violet solution (Sigma) for imaging.

Terminal deoxynucleotidyl transferase dUTP nick end labeling assay

HUVECs were fixed in 4% formaldehyde solution and permeabilized in 0.3% Triton-X100 solution. For wound tissues, samples were embedded in paraffin and cut into 5-µm sections. Sections were deparaffinized, fixed and permeabilized. Subsequently, cells or sections were performed with TdT rection and Click-iT rection consecutively. Samples were mounted in anti-fade mounting medium with DAPI (Beyotime)andimaged.Click-iTPlusTUNELAssaykitwasprovided by ThermoFisher Scientific.

Enzyme-linked immunosorbent assay

The secretion of VEGFA in the supernatants of HUVECs was examined with the human VEGFA ELISA kit (Sino Biological, Beijing, China). The supernatants were collected and stored at -80°C until use. The absorbance (450 nm) was recorded.

Immunofluorescence staining

HUVECs were fixed in 3.7% formaldehyde solution and permeabilized in 0.3% Triton-X100 solution. Cells were blocked in 5% normal goat serum solution for 1 hour and incubated with a rabbit polyclonal antibody against hnRNPU (1:500, ab264142, Abcam) overnight. Next day, cells were incubated with an Alexa Fluor 555-conjugated goat anti-rabbit secondary antibody (1:1000, ThermoFisher Scientific) for 1 hour. After wash, cells were mounted in anti-fade mounting medium with DAPI (Beyotime) and imaged.

RNA pull-down assay

For RNA pull-down, circMYO9B and its antisense RNA were labeled with biotin using the Biotin RNA Labeling Mix (Sigma). Total protein was extracted from HUVECs using the Protein Extraction Kit (Bio-Rad). Biotin-conjugated circMYO9B and its antisense RNA (50 pM) were mixed with total protein (1 mg) from HUVECs and incubated for 4 h. Streptavidin beads were added into the mixture and incubated for 1 h. Protein was recovered and subjected to electrophoresis. The enrichment of hnRNPU pulled down by circMYO9B was validated by western blot.

RIP

Protein A/G magnetic beads were pre-coated with a rabbit hnRNPU antibody (1:50; ab264142, Abcam) or a normal rabbit IgG (1:50; ab172730, Abcam). Cells were lysed in RIP lysis buffer, and cell lysates and magnetic bead-antibody complex were mixed and incubated overnight. Subsequently, RNA was extracted and subjected to RT-qPCR to evaluate the binding of RNAs (circMYO9B, KDM1A and CBL) to hnRNPU.

RNA pull-down assay

For RNA pull-down, circMYO9B and its antisense RNA were labeled with biotin using the Biotin RNA Labeling Mix (Sigma). Total protein was extracted from HUVECs using the Protein Extraction Kit (Bio-Rad). Biotin-conjugated circMYO9B and its antisense RNA (50 pM) were mixed with total protein (1 mg) from HUVECs and incubated for 4 hours. Streptavidin beads were added into the mixture and incubated for 1 h. Protein was recovered and subjected to electrophoresis. The enrichment of hnRNPU pulled down by circMYO9B was validated by western blot.

A mouse model of DFU

The mouse model of DFU was established as previously described in ref. 10. Five-to six-week-old male C57BLKS/J db/db diabetic mice were obtained from Hunan SJA Laboratory Animal Co., Ltd (Changsha, China). Mice were fed with high fat and glucose diet for six weeks and administrated with intraperitoneal injection of 0.45% streptozotocin (Sigma). After three days, blood glucose was measured, and twenty-four mice with high blood glucose (≥ 16.7 mM for consecutive days) were selected. Two weeks later, these mice were intraperitoneally injected with pentobarbital sodium (50 mg/kg, Sigma) for anesthetization, and the back of foot was exposed by shaving. Full-thickness wounds of 1 cm were generated by skin excisions. MSCs were pre-treated with GW4869 (Selleck, Houston, TX, USA) at 10 µM for 24 h (MSC-GW4869). MSCs and MSC-GW4869 (1 × 106 cells in 100 µL PBS per mouse) or 100 μL of PBS (n = 8 per group) were injected around the wounds. In addition, MSC-derived exosomes (MSC-Exos-oe-NC, MSC-Exos-oe-circMYO9B and MSC-Exos-sh-circMYO9B, 100 µg exosomes in 100 μL PBS per mouse) or 100 μL of PBS (n = 8 per group) were injected around the wounds. At day 0, 3, 7 and 13, wounds were imaged for monitoring the wound-healing process and determining wound area. Wound tissues were harvested for histological examination and TUNEL staining. Animal procedures were approved by the Animal Care and Use Committee of the Third Xiangya Hospital Central South University.

Histological examination

Wound tissues were fixed in 4% formaldehyde and embedded in paraffin. Samples were cut into 5 µm sections and stained with Hematoxylin and eosin (H&E, Solarbio, Beijing, China) and Masson’s Trichrome (Solarbio). For IHC staining, sections were antigen-retrieved and blocked with H2O2 and 8% BSA solution. Subsequently, sections were stained with a rabbit CD31 antibody (1:50, ab28364, Abcam) and incubated with a goat anti-rabbit HRP-conjugated secondary antibody. DAB substrate (Beyotime) was added to visualize the signal. Sections were stained with hematoxylin and imaged.

Statistics and reproducibility

All experiments were performed in at least three biological replicates, and each biological replicate contained three technical replicates. Results were expressed as mean ± standard deviation (SD). The Student’s t-test and one-way analysis of variance (ANOVA) were used for comparing the variance of two and multiple groups, respectively. GraphPad Prism 8.0 (GraphPad Software, San Diego, CA, USA) was used for statistical analysis. P < 0.05 was statistically significant. *P < 0.05, **P < 0.01 and ***P < 0.001.