Extracellular vesicles derived from GMSCs stimulated with TNF-α and IFN-α promote M2 macrophage polarization via enhanced CD73 and CD5L expression

Immunoregulatory properties of mesenchymal stem cell (MSC)-derived extracellular vesicles (EVs) are promising. Gingival tissue-derived MSCs (GMSCs) have unique immunoregulatory capacity and secrete large amounts of EVs. Recent findings suggest that priming MSCs with inflammatory stimuli is an effective strategy for cell-free therapy. However, the precise mechanism by which the contents of EVs are customized has not been fully elucidated. Here, we show that EVs derived from GMSCs primed with a combination of two pro-inflammatory cytokines, tumor necrosis factor-α (TNF-α) and interferon-α (IFN-α), synergistically promote anti-inflammatory M2 macrophage polarization by increasing the expression of cluster of differentiation 73 (CD73) and CD5 molecule-like (CD5L). Expression of CD73 by TNF-α/IFN-α stimulation was transcriptionally upregulated by the activation of mammalian target of rapamycin signaling and nuclear translocation of hypoxia-inducible factor 1α in GMSCs. TNF-α/IFN-α treatment also significantly increased the expression of CD5L mRNA via the transcription factor DNA-binding protein inhibitor ID3 and liver X receptor. Interestingly, exosomal CD5L is a prerequisite for the synergistic effect of EVs-mediated M2 macrophage polarization. These results indicate that combined pre-licensing with TNF-α and IFN-α in GMSCs is ideal for enhancing the anti-inflammatory function of EVs, which contributes to the establishment of a therapeutic tool.


TNF-α and INF-α-induced CD73 expression in GMSCs were mediated by mammalian target of rapamycin (mTOR)-hypoxia-inducible factor-1α (HIF-1α) axis.
Further, we explored the universal molecular mechanisms of TNF-α and IFN-α-mediated expression of CD73 in GMSCs. Considering the recent findings of the induction of HIF-1α by inflammatory cytokines 27 and the regulation of CD73 expression by HIF-1α, we hypothesized that the TNF-α/IFN-α-mediated upregulation of CD73 could be induced by HIF-1α.
To test this hypothesis, we first confirmed that dual TNF-α and IFN-α stimulation synergistically enhanced the expression of HIF-1α mRNA in GMSCs (Fig. 2a). Western blot analysis further confirmed the dual stimulationinduced upregulation of HIF-1α protein expression for 24 h (Fig. 2b). Immunofluorescent staining captured by confocal microscopy demonstrated that both TNF-α and IFN-α-induced the expression of HIF-1α within 6 h (Fig. 2c). Furthermore, significant nuclear accumulation of HIF-1α, which indicates the activation of HIF-1α, was observed after GMSCs were treated with TNF-α and IFN-α for 24 h (Fig. 2d). To examine the direct involvement of HIF-1α in TNF-α and IFN-α-mediated upregulation of CD73 expression, GMSCs were transfected with a HIF-1α overexpression plasmid (pHIF1) or siRNA for HIF-1α (siHIF1). Twelve hours after transfection, pHIF1 enhanced HIF-1α mRNA expression in GMSCs in a dose-dependent manner. However, the expression levels of CD73 mRNA were unchanged (Fig. 3a). However, increased expression of CD73 and HIF-1α was observed 24 h after transfection with pHIF1 ( Fig. 3b), suggesting that the expression of CD73 was induced after enhanced expression of HIF-1α in GMSCs. In accordance with our established protocol, EVs from pHIF1 transfected GMSCs were isolated from cell culture supernatants. Forty-eight hours after transfection, the expression levels of both HIF-1α and CD73 mRNA were still elevated (Fig. 3c). Similarly, a significant increase in the exosomal CD73 expression was observed in pHIF1-transfected GMSCs compared to that in control transfectants (Fig. 3d).
In contrast, transfection with siHIF1 attenuated both the TNF-α and IFN-α-induced expression of CD73 mRNA (Fig. 4a) and endogenous expression levels of CD73 protein in GMSCs (Fig. 4b). Consequently, the knockdown of HIF-1α suppressed the expression of CD73 in EVs from GMSCs (Fig. 4c). We further examined the pathway involved in HIF-1α-mediated CD73 expression. As HIF-1α expression is known to be regulated by the mTOR pathway 27 , rapamycin was used to inhibit mTOR. mTOR inhibition decreased both HIF-1a and CD73 expression in response to TNF-α/IFN-α stimuli (Fig. 4d). These results suggest that TNF-α and INF-α-induced CD73 expression is regulated by the activation of mTOR signaling and the subsequent nuclear translocation of HIF-1α in GMSCs.
EVs from TNF-α and INF-α-stimulated GMSCs synergistically reprogram macrophage from M1 to M2 phenotype. Our previous study confirmed that EVs from TNF-α-stimulated GMSCs enhanced the ability to convert the macrophage phenotype from pro-inflammatory M1 to anti-inflammatory M2 and further demonstrated that the expression level of CD73 in EVs was associated with the ability of EVs to switch M1 macrophages into M2 phenotype 16 . To compare and validate the modulatory effect of TNF-α and IFN-α preconditioning on macrophage reprogramming from the M1 to M2 phenotype, GMSCs were stimulated with TNF-α and INF-α individually and in combination before isolation of EVs from the cell culture supernatant. To achieve this, M1-activated macrophages were stimulated with or without EVs, and the expression of M1 (CD86) and M2 markers (CD206 and CD163) were examined by flow cytometry (Fig. 5a). Compared with GMSC-derived EVs without preconditioning (EVs-Ctrl), stimulation with TNF-α-preconditioned GMSC-derived EVs (EVs-TNF) showed a decreased expression of CD86 (26.6% vs. 20.6%) (Fig. 5b) and an enhanced expression of CD206 and After stimulation by TNF-α/IFN-α for 6 h and 24 h, whole-cell lysates prepared from GMSCs were subjected to immunoblot analysis with an anti-HIF-1α antibody. β-actin was used as a control. Representative western blot analysis of total Grp78 is shown in the upper panel. Quantification of total HIF-1α protein levels relative to β-actin was performed using MultiGauge software (lower panel). (c,d) Time course confocal images of HIF-1α internalization. GMSCs were incubated with TNF-α/IFN-α for 6 h (c) and 24 h (d), respectively. The cells were stained with anti-HIF-1α antibody followed by an Alexa 488 secondary antibody (green). Nuclei were stained with DAPI dye (blue). The intensities of HIF-1α fluorescence (c; right) and nuclear translocated HIF-1α (d; right) for each sample were measured from all the cells in each group, which were counted in three randomly chosen fields. Scale bars: 20 μm. *P < 0.05; **P < 0.01; ***P < 0.001. Error bars represent mean ± SD, n = 3. The significance of differences between groups was determined using one-way Tukey's test. Full blots are presented in Supplementary materials.   (Fig. 5c,d). These results indicated that the combined preconditioning of GMSCs with TNF-α and IFN-α synergistically augmented the effect of EVs-mediated M2 macrophage reprogramming.

TNF-α and INF-α stimulation induced CD5L expression in GMSCs-derived EVs contributed M2 macrophage polarization.
To gain further insights into the EVs-TNF/IFN-mediated M2 macrophage polarization, we hypothesized that the exosomal proteome of immunomodulatory proteins could contribute to the synergistic effect of TNF-α and INF-α stimulation in GMSCs. Since recent studies have demonstrated that CD5 molecule-like (CD5L)/apoptosis inhibitor of macrophages is expressed in human bone marrow MSCs (hBM-MSCs) 28 and is involved in M2 macrophage polarization 29 , we examined the effects of TNF-α/IFN-α stimulation on the expression of CD5L mRNA in GMSCs (Fig. 6a). Combined stimulation with TNF-α and INF-α dramatically upregulated the expression of CD5L mRNA in comparison with TNF-α or IFN-α stimulation after 24 h. To validate if this effect was transcriptionally controlled by the DNA-binding protein inhibitor ID3 (ID3) 29 and liver X receptor (LXR) 30 , time-course expression was monitored. Dual stimulation of TNF-α and INF-α significantly upregulated ID3 mRNA within 6 h and gradually decreased within 24 h. However, induction of  Figure 3. HIF-α regulates the expression of CD73 in GMSCs. (a,b) Effect of HIF-1α-overexpression on cellular CD73 expression in GMSCs. GMSCs were seeded on 6-well plates (5 × 10 5 cells/well) a day before transfection. At approximately 60-70% confluence, the cells were transfected with indicated amounts of HIF-1α-overexpressed plasmid (pHIF-WT) (0.125, 0.25, and 0.5 μg/2 mL) for 24 h. The mRNA and protein expression of HIF-1a and CD73 in GMSC was analyzed by quantitative real-time PCR (a) and western blot analysis (b). Relative HIF-1α and CD73 protein expressions were measured by quantifying the density of bands and normalized against β-actin using the MultiGauge software (right). After transfection of pHIF-WT (0.25 µg/2 mL) for 48 h (c), GMSC-derived EVs were isolated from culture media, and the expression of CD73 in EVs was evaluated by western blotting (d). Exosome-associated CD63 and CD81 proteins were used as loading controls. Relative CD73 protein expressions were measured by quantifying the density of bands and normalized against CD81 using the MultiGauge software (d; right). *P < 0.05; **P < 0.01; ***P < 0.001. Error bars are means ± SD., n = 3. Data represent mean ± SD. The significance of differences between groups was determined by one-way Tukey's test. Full blots are presented in Supplementary materials. www.nature.com/scientificreports/ LXR mRNA peaked at 24 h. Although the amount of secreted CD5L by TNF-α/IFN-α stimulation was less than 10 pg/mL and the stimulation had little effect on secretion for 48 h (Fig. 6b). TNF-α/IFN-α stimulation clearly enhanced CD5L expression in GMSC-derived EVs (Fig. 6c). To examine whether CD5L could induce M2 macrophage polarization, the macrophages were stimulated with recombinant CD5L (Fig. 7a). CD5L strongly decreased M1 marker expression lower than the IL-4/13 stimulation (37.6% vs. 11.0%) (Fig. 7b) and enhanced M2 marker expression relative to the unstimulated control macrophages (20.3% vs. 32.7%) (Fig. 7b). To further explore the role of exosomal CD5L in M2 macrophage polarization, GMSCs were transfected with CD5L siRNA before preconditioning and M1 polarized macrophages were treated with EVs (Fig. 8a). The efficiency of the CD5L knockdown (si-CD5L) and the upregulation of CD5L by TNF-α/INF-α stimulation in control siRNA (si-Ctrl) transfected GMSCs were confirmed by western blot analysis (Fig. 8b). Compared to unstimulated control (Ctrl), control siRNA-transfected GMSC-derived EVs (EVs-siCtrl) upregulated CD206 expression (  (c) Exosomal CD73 protein expression in equal amounts (5 μg) was compared by immunoblot analysis with an anti-CD73 antibody. Exosome-associated CD81 proteins were used as loading controls. Relative CD73 protein expressions were measured by quantifying the density of bands and normalized against CD81 using the MultiGauge software (lower). (d) Effect of rapamycin treatment on the expression of HIF-1α and CD73 in TNF-α, INF-α, and TNF-α/ INF-α-stimulated GMSCs. Inhibition of mammalian target of rapamycin (mTOR) activity was confirmed using a phosphorylated-mTOR antibody (p-mTOR, Se2448). Relative HIF-1α and CD73 protein expressions were measured by quantifying the density of bands and normalized against β-actin, and relative phosphorylation of mTOR were measured by quantifying the density of bands and normalized against mTOR using the MultiGauge software (right). using the MultiGauge software (b; right). *P < 0.05; **P < 0.01; ***P < 0.001. Error bars represent mean ± SD, n = 3. The significance of differences between groups was determined using one-way Tukey's test. Full blots are presented in Supplementary materials.  www.nature.com/scientificreports/ (Fig. 8c). These results suggested that TNF-α and INF-α stimulation-induced CD5L in GMSC-derived EVs also contributed to M2 macrophage activation.

Discussion
This study aimed to establish an effective and reproducible protocol for GMSC-derived EVs-based therapy. In our study, we used serum-free medium, since our preliminary experiments demonstrated that the quantity of GMSC-derived EVs increased under serum deprivation compared with the exosome-depleted fetal bovine serum (FBS)-containing culture conditions, which presents a clinical advantage. In general, starvation stress can cause cells to activate survival pathways and secrete factors that counteract severe conditions. While serum starvation has the potential to influence exosomal biology 31 , previous studies have demonstrated that serum-free conditions can enhance the therapeutic efficiency of MSCs 32,33 . In this context, in addition to reducing the possibility of introducing bovine-derived artifacts 34 , approximately 45% of the published papers used serum-free medium for isolation of MSC-EVs 35 . Considering these evidences, we used serum-free medium as the optimal culture condition in this study.
In the present study, we reported that dual TNF-α/IFN-α priming in GMSCs synergistically augmented EVs-induced M2 macrophage polarization through the upregulation of CD73 and CD5L. Previous studies have shown that the inflammatory environment modulates the immunoregulatory capacity of MSCs 36 , and the effects of priming MSCs with pro-inflammatory cytokines, such as TNF-α, IFN-γ, IL-1β, and LPS have been reported 37 . Among the few studies that validated the effect of combined cytokine priming on the immunoregulatory properties of MSCs, representative priming with TNF-α and IFN-γ improved their immunomodulatory functions 38 . Meanwhile, studies on other cytokine combinations have recently reported a synergistic effect of TNF-α/IL-10 on the secretion of PGE2 39 and IFN-γ/IL-1β on the expression of PGE2 and IDO 40 , which promotes the immunosuppressive capacity of MSCs. However, owing to the complex cross-talk between cytokine signaling, detailed downstream molecular signaling has not been elucidated. In our previous study, we demonstrated that TNF-α stimulation enhanced exosomal CD73 expression in GMSCs, which is essential for M2 macrophage induction 16 . In this study, we observed that the dual stimulation of GMSCs with TNF-α and IFN-α synergistically increased exosomal CD73 expression in EVs (Fig. 1). The expression of CD73 is upregulated at the site of inflammation and is increased by INF-α in endothelial cells 24 . However, it is also simultaneously activated by TNF-α 41 . Both of these are consistent with our current data. We also observed that enhanced CD73 expression was caused by the upregulation and nuclear translocation of HIF-1α (Fig. 2). Recent studies have revealed the transcription factors regulating CD73 expression 42 , and CD73 has a HIF-1 binding DNA-consensus motif in the promoter region. Importantly, HIF-1α binds directly to the CD73 promoter, thereby, activating CD73 expression 43 and is also induced by both TNF-α 44 and IFN-α 45 . Consistent with these reports, our current study demonstrated enhanced expression of CD73 by HIF-1α overexpression plasmid (Fig. 3) and inhibition of CD73 expression by HIF-1α siRNA under TNF-α/IFN-α stimulation (Fig. 4) in both GMSCs and GMSC-derived EVs. Furthermore, signal transduction between mTOR-dependent induction of HIF-1 activity 46 and TNF-α/IFN-α-mediated mTOR activation 47,48 was confirmed by rapamycin treatment of GMSCs (Fig. 4d), indicating that the mTOR-HIF-1α-CD73 axis is essential for TNF-α/IFN-α-induced CD73 expression in GMSCs. In the current study, we examined the effect of EVs on macrophage polarization after M1 macrophage induction (Fig. 5a). Accordingly, we further confirmed that EVs-TNF/IFN exhibited superior ability to switch M1 macrophages into the M2 phenotype (Fig. 5). It is widely accepted that HIF-1α is a central transcription factor in the hypoxia response, and hypoxic preconditioning of MSCs is one of the main MSC priming strategies 49 . Therefore, the involvement of HIF-1α in TNF-α/IFN-α priming is reasonable, and further application of GMSC-derived EVs should be validated in future studies.
We next investigated the possibility of another molecular mechanism that reinforces EVs-TNF/IFN-mediated M2 macrophage polarization. While attempting to identify any exosomal protein that modulates inflammatory responses independently of CD73, we found interesting studies reporting a novel role of CD5L as a driver of M2 macrophage polarization 29 , and CD5L was identified as a new cell surface marker in hBM-MSCs 28 . CD5L is a soluble protein belonging to the scavenger receptor cysteine-rich superfamily, and is mainly secreted from macrophages 50 . While CD5L is involved in the pathogenesis of the inflammatory processes to that control lipid metabolism by inhibiting the apoptosis of macrophages, it also promotes an anti-inflammatory cytokine profile in response to TLR activation 51 . Recent studies have revealed that CD5L is secreted from the olfactory mucosa MSC 52 . It is also compiled in the EV-specific database, ExoCarta 53 . However, the effect of exosomal CD5L on Figure 5. EVs from TNF-α/IFN-α-stimulated GMSCs enhance M2 macrophage polarization. (a) Strategy diagram for the validation of phenotypic switching capacity of EVs from M1 to M2 macrophages. After M1 macrophages were generated by LPS (10 ng/mL) and IFN-γ (25 ng/mL) stimulation from resting macrophages derived from CD14 + peripheral blood-derived monocytes (PBMCs) for 24 h, EVs (2.5 µg/mL) from GMSC with or without (Exo-Ctrl) priming of TNF-α (100 ng/mL; Exo-TNF), IFN-α (100 ng/mL; Exo-IFN), and TNF-α/ IFN-α (50 ng/mL, each; Exo-TNF/IFN) were incubated with M1 macrophages for 48 h. (b) Effects of cytokine primed GMSCs-derived EVs on CD86 expression in M1 macrophages. (c,d) Effects of TNF-α-treated GMSCderived EVs on CD206 (c) or CD163 (d) expression in M1 macrophages. Percentage of double positive cells (CD11b + CD206 + ) (c) and (CD11b + CD163 + ) (d) was analyzed to measure the ratio of macrophages polarized to M2 phenotype. Percentage of double positive cells (CD11b + CD86 + ) representing M1 macrophages (b), and M2/ M1 balance was determined by using the ratio of (CD11b + CD206 + )/(CD11b + CD86 + ) (c) or (CD11b + CD163 + )/ (CD11b + CD86 + ) (d) macrophage population. *P < 0.05; **P < 0.01; ***P < 0.001. Error bars represent mean ± SD, n = 3. The significance of differences between groups was determined using one-way Tukey's test. www.nature.com/scientificreports/ macrophage phenotype remains unknown. These reports prompted us to investigate the possible involvement of CD5L in EVs-TNF/IFN-mediated M2 macrophage polarization. In the current study, we observed a significant upregulation of CD5L mRNA expression by the combined stimulation of TNF-α and IFN-α in GMSCs (Fig. 4). Our data showed that TNF-α/IFN-α stimulation dramatically enhanced exosomal CD5L expression with a slight increase in the secreted CD5L protein (< 10 pg/mL), indicating that TNF-α and IFNα-induced expression of CD5L proteins was mostly released in the form of CD5L-overexpressed EVs-TNF/IFN of GMSCs. The expression of CD5L is positively regulated by two transcription factors, ID3 29 and LXR 54 , which regulate the inflammatory responses by inhibiting inflammatory cytokines 55 . Accordingly, qRT-PCR analysis revealed that combined stimulation with TNF-α and INF-α initially upregulated ID3 mRNA expression within 6 h and then enhanced LXR mRNA expression at 24 h (Fig. 6). Altogether, our data suggest that this time lag between the induction of these transcription factors contributes to the continuous expression of CD5L in GMSCs. After confirmation that CD5L has the ability to induce M2 macrophage polarization (Fig. 7), knockdown of CD5L in GMSCs was performed. This inhibited EVs-mediated M2 macrophage marker expression (Fig. 8), demonstrating that exosomal CD5L was essential for TNF-α/IFN-α induced synergistic effects on M2 macrophage induction.
In conclusion, our study demonstrated that EVs from GMSCs stimulated with the combination of TNF-α and IFN-α promote M2 macrophage polarization by the upregulation of exosomal CD73 and CD5L. Compared with stimulation with a single cytokine, the combination of TNF-α and IFN-α synergistically enhanced the expression of CD73 through the activation of the HIF-1α-mTOR pathway and the upregulation of CD5L was transcriptionally mediated by LXR and ID3 in GMSCs. These findings based on the molecular mechanisms are summarized in Fig. 9. Considering the impact of macrophage plasticity on disease 56 , GMSCs-derived EVs preconditioned by TNF-α and IFN-α could be a promising therapeutic tool for the treatment of inflammatory diseases. (c) Exosomal CD5L protein expression in equal amounts of the protein (5 μg) was compared by immunoblot analysis with an anti-CD5L antibody. Exosome-associated CD63, CD9, and CD81 proteins were used as loading controls. Relative CD5L protein expressions were measured by quantifying the density of bands and normalized against CD81 using the MultiGauge software (right). **P < 0.01; ***P < 0.001. Error bars represent mean ± SD, n = 3. The significance of differences between groups was determined using one-way Tukey's test. Full blots are presented in Supplementary materials.  (c) Percentage of double positive cells (CD11b + CD206 + ) was analyzed to measure the ratio of macrophages polarized to M2 phenotype and M2/M1 balance was determined by using the ratio of (CD11b + CD206 + )/(CD11b + CD86 + ) macrophage population. *P < 0.05; **P < 0.01; ***P < 0.001. Error bars represent mean ± SD, n = 3. The significance of differences between groups was determined using one-way Tukey's test.

Materials and methods
Ethics statement and preparation of human tissue samples. All human gingival tissues were obtained as discarded clinical samples using the approved institutional review board (IRB) protocol at Kyushu University Hospital. The procedures for using human samples were conducted in accordance with the Declaration of Helsinki and were approved by the Kyushu University Institutional Review Board for Human Genome/ Gene Research (Protocol Number: 2019-374). Written informed consent was obtained from all the subjects. Human gingival tissues were obtained from three 27-32 years old adult donors who were early stages of periodontitis. After initial treatment for periodontitis, gingivectomy was performed to eliminate the distal pockets adjacent to the terminal molars. The inclusion and exclusion criteria of subjects were listed in Supplementary  Table 1.  Percentage of double positive cells (CD11b + CD206 + ) was analyzed to measure the ratio of macrophages polarized to M2 phenotype. *P < 0.05; **P < 0.01; ***P < 0.001. Error bars represent mean ± SD, n = 3. The significance of differences between groups was determined using one-way Tukey's test. Full blots are presented in Supplementary materials.  16,20,57,58 . Briefly, gingival tissues were gently separated, minced, and digested with phosphate-buffered saline (PBS) solution containing collagenase type I (2 mg/mL) (Worthington Biochemicals, Lakewood, NJ, USA) and dispase II (4 mg/ mL) (Sanko Junyaku, Tokyo, Japan) for 1 h at 37 °C. Single-cell suspensions from the gingiva were obtained by passing the culture through a 70-μm strainer (Falcon, BD Biosciences Discovery Labware, Bedford, MA, USA). All nucleated cells were seeded in 100-mm culture dishes with complete media containing alpha modification of Eagle's medium (Invitrogen, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS, Hyclone Laboratories, USA), 2 mM l-glutamine (Invitrogen, Carlsbad, CA, USA), 10 mM l-ascorbic acid phosphate (FUJIFILM Wako, Japan), and 100 U/mL penicillin/streptomycin (Gibco Life Technologies, country), followed by an initial incubation for 48 h at 37 °C in 5% CO 2 and 95% humidity. The cultures were subsequently washed twice with PBS to eliminate non-adherent cells. The attached cells were further cultured for another 12 days under the same conditions in the complete medium mentioned above, and cells at the 3rd-5th passages were used for subsequent experiments. Human CD14 + peripheral blood-derived monocytes (PBMCs; lot no. 2050813521) were purchased from Lonza (Basel, Switzerland). Cells were seeded in 6-well plates (2 × 10 5 cells/well) and cultured in RPMI 1640 medium (Nacalai Tesque, Kyoto, Japan) supplemented with 10% heatinactivated FBS, 2 mM glutamine (Nacalai Tesque), 1% sodium pyruvate (Nacalai Tesque), 1% nonessential amino acids (Nacalai Tesque), and 25 ng/mL of M-CSF. To generate macrophages, monocytes were cultured for 7 days with a single change of fresh medium on day 3-4 after the initiation of the culture. To polarize M1 or M2 macrophages, macrophages were further treated with 10 ng/mL LPS, 20 ng/mL IFN-γ or 20 ng/mL IL-4, and 20 ng/mL IL-13 for 24 h. Isolation and characterization of EVs. EVs were prepared according to the protocol reported in our previous study 16 and the recommendations of the International Society of Extracellular Vesicles 59 . For preconditioning of GMSCs, cells in 100-mm culture dishes were grown to achieve a confluence of 70-80%. After the medium was aspirated, cells were rinsed three times with PBS and treated with 100 ng/mL of recombinant human TNF-α and IFN-α, and 50 ng/mL of TNF-α and IFN-α in a serum-free medium and incubated for 48 h prior to supernatant harvesting. Controls are maintained in a serum-free medium without any additions. GMSC-derived EVs were isolated from serum-free conditioned media using the MagCapture™ Exosome Isolation Kit PS (FUJIFILM Wako) 60 . Cells were preconditioned in a serum-free medium for 48 h. Medium was then collected and centrifuged at 10,000×g for 30 min to eliminate other large extracellular vesicles. The cleared supernatants were passed through 0.22 mm filter membranes and concentrated using a Vivaspin-20 concentrator (Sartorius, Göttingen, Germany). The amount of protein in EVs was quantified using a BCA Protein Assay Kit (Takara Bio, Otsu, Japan), and 2.5 mg/mL of EVs were used for the stimulation of macrophage. Transmission electron microscopy (TEM) analyses were performed by the Hanaichi Ultrastructure Research Institute  Quantitative real-time PCR (qRT-PCR) analysis. qRT-PCR was performed as previously described 16 .

Cytokines and reagents.
Total RNA was isolated from cells using ISOGEN II (Nippon Gene, Tokyo, Japan), and first-strand cDNA was synthesized using PrimeScript RT Master Mix (Takara Bio, Otsu, Japan Western blot analysis. Western blot was performed as previously described 16  Enzyme-linked immunosorbent assay. Cell supernatants were collected, and the expression levels of CD5L were measured using a Human CD5L ELISA Kit (Thermo Fisher Scientific), according to the manufacturer's instructions. The plates were read using a microplate reader at 450 nm with a background correction at 630 nm.
Cell transfection. GMSCs were seeded in 6-well plates (5 × 10 5 cells/well) a day before transfection. At approximately 60-70% confluence, the cells were transfected with indicated amounts of the expression plasmids using 2 µL of Lipofectamine 3000 (Thermo Fisher Scientific), according to manufacturer's instructions. For the control, cells were transfected with empty vectors under the same conditions. Human HIF-1α expression plasmid (pCAG-HIF1 alpha) and control plasmid (pCAGEN) were obtained from Addgene (http:// www. addge ne. org/). Transfection of siRNAs was carried out using Lipofectamine™ RNAiMAX Transfection Reagent (Thermo Fisher Scientific) for 24 h according to our reverse transfection protocol 61 . Stealth™ RNAi duplexes, comprising mixtures of three different siRNAs against human CD73 (HSS107326, HSS107328, and HSS181585), and HIF-1α (HSS179231, HSS104774, and HSS104775) were obtained from the Invitrogen Corporation (Invitrogen Life Technologies, Carlsbad, CA, USA). Stealth™ RNAi negative control duplex (Medium GC Duplex, Invitrogen Life Technologies, Carlsbad, CA, USA) was used as the control.
Statistical analysis. Data were analyzed using the JMP Pro 16 software (SAS Institute Inc., Cary NC, USA).
Comparisons between two groups were performed using an independent unpaired two-tailed Student's t-test, and comparisons between more than two groups were performed using one-way analysis of variance (ANOVA) with Bonferroni correction. Statistical significance was set at P < 0.05.

Data availability
The data used to support the findings of this study are available from the corresponding author on request.