miR-450a-5p within rat adipose tissue exosome-like vesicles promotes adipogenic differentiation by targeting WISP2

In adipose tissue, adipocytes are constantly being created and destroyed throughout life (Rosen and Spiegelman, 2014). In humans, ∼10% of adipocytes are turned over every year, while nearly 0.6% of adipocytes are renewed every day in mice (Rigamonti et al., 2011, Spalding et al., 2008). Adipose tissue-derived stem cells (ADSCs) serve as a reservoir and allow the continued renewal of precursor cells that can differentiate into adipocytes (Bowers et al., 2006). However, differentiation of adipocytes is a complex and multi-step process with many factors and signaling pathways involved (Gustafson et al., 2015b). The terminal differentiation of adipocytes has been extensively characterized with the use of hormonal cocktails that typically contains dexamethasone (DEX), insulin, indomethacin and isobutylmethylxanthin (IBMX) (Rosen and Spiegelman, 2014, Wu et al., 2010). The cocktails act as direct inducers of genes that regulate adipogenesis such as peroxisome proliferator-activated receptor γ (PPARγ), and CCAAT-enhancer-binding protein β (C/EBPβ) and δ (C/EBPδ) (Moseti et al., 2016; Wu et al., 2010). However, the differentiation cocktails cannot represent the natural adipogenic process in vivo (Sarkanen et al., 2012). We know little about the adipogenic process in vivo, as the means of studying this are most indirectly (Rosen and Spiegelman, 2014). Understanding adipogenesis in adipose tissue may provide means to develop strategies for the treatment and prevention of obesity and the application of adipose tissue regeneration.

Previous studies have demonstrated that adipose tissue extract promotes adipogenesis and angiogenesis, indicating that adipose tissue can produce bioactive factors to regulate adipogenesis via paracrine signals (Sarkanen et al., 2012; Li et al., 2014). For example, BMP4, secreted by adipocytes, can induce ADSCs to undergo adipogenic commitment (Gustafson et al., 2015a). In addition to secreting soluble proteins, adipose tissue also releases extracellular vesicles (EVs) [including exosomes, small membrane vesicles (30–100 nm) that are released by multivesicular bodies fusing with the cell membrane and microvesicles, 100–1000 nm vesicles originated by shedding of the plasma membrane] (Deng et al., 2009; Connolly et al., 2015; Muller et al., 2011; Fleury et al., 2016). However, the size range of exosomes overlaps partially with that of microvesicles, which hinders a complete size-discrimination between these two populations of EVs (Campoy et al., 2016). Therefore, enriched vesicles having similar morphology and size consistent with exosomes are defined as exosome-like vesicles (Altadill et al., 2016). Exosomes can facilitate intercellular communication by transferring proteins and non-coding RNAs (such as microRNAs; miRNAs) (Kahlert and Kalluri, 2013; Braicu et al., 2014). Although some functions or mechanisms remain elusive, it has been demonstrated that exosomes can modulate their neighboring cells by influencing major cellular process including apoptosis, cell differentiation, proliferation and metabolism (Qin et al., 2016; Braicu et al., 2014). One obvious point about the content of exosomes is that they contain miRNAs, which can inhibit translation of mRNAs by targeting 3′ untranslated regions (UTRs) (Batiz et al., 2015). Transported miRNAs are capable of targeting mRNAs in recipient cells and, hence, are associated with cell-to-cell signaling and communication (He et al., 2014; Valadi et al., 2007).

Emerging evidence shows that inhibition of miRNAs biogenesis impairs differentiation of stem cells into adipocytes, which supports the idea that miRNAs may directly modulate adipocyte differentiation (Xie et al., 2009; Zhang et al., 2016; Arner and Kulyté, 2015). For example, miR-30a is a positive regulator of adipocyte lineage commitment that acts through modulation of Runx2 while miR-143 is required for terminal differentiation by modulating MAP2K5–ERK5 (ERK5 is also known as MAPK7) signaling (Chen et al., 2016; Chen et al., 2014; Hamam et al., 2014). In addition, exosomal miRNAs also contribute to the process of cell differentiation (Forterre et al., 2014; Xu et al., 2014). Previous studies have shown that exosomes can promote differentiation of mesenchymal stem cells (MSCs) (such as osteogenesis, neurogenesis and angiogenesis) through exosomal miRNAs (Narayanan et al., 2016; Takeda and Xu, 2015; Cui et al., 2016; Lopatina et al., 2014). In adipose tissue, adipocyte-derived exosomes deliver anti-osteoblastic miRNAs (miR-30c, miR-125a, miR-125b and miR-31) to osteoblasts, which reduces the levels of osteoblastic differentiation markers (osteocalcin and osteopontin) (Martin et al., 2015).

Although adipose tissue extract induces adipogenic differentiation of MSCs, the role that exosomes play in this process remains unknown. In this study, we hypothesized that exosome-like vesicles derived from adipose tissue (Exo-AT) could be enriched with miRNA that would induce adipogenic differentiation of MSCs. We demonstrated that Exo-AT contained adipogenesis-associated miRNAs that could promote adipogenic differentiation of ADSCs. In particular, miR-450a-5p, which was enriched in Exo-AT, could promote adipogenesis through targeting WISP2, a negative regulator of adipogenesis (Hammarstedt et al., 2013). This study expands our understanding of the adipogenic process in the adipose tissue microenvironment.

Exosome-like vesicles derived from adipose tissue (Exo-AT) or from ADSCs (Exo-ADSC) were extracted using Total Exosome Isolation™ reagent. The morphology of Exo-AT and Exo-ADSC were observed directly through transmission electron microscopy (TEM). The results revealed that Exo-ADSC and Exo-AT were rounded and double-membrane structures with a size ranging from 50–100 nm in diameter (Fig. 1A). The size distribution as measured by the Zetasizer Nano ZS analysis system indicated that Exo-AT and Exo-ADSC were homogeneous, with a peak at 85 nm and 73 nm respectively, which is consistent with them being exosomes (Fig. 1B). Finally, the characteristics of Exo-AT and Exo-ADSC were further validated by western blotting for the expression of CD9, CD63 (a tetraspanin enriched in late multivesicular bodies, and, hence, in exosomes) and TSG101 (a protein related to exosomes biogenesis) (Romancino et al., 2013; Kowal et al., 2014). The results indicated that CD9, CD63 and TSG101 were present in vesicles. In contrast, the cellular protein β-actin was detected exclusively in cell lysates (Fig. 1C). To explore whether Exo-AT could be internalized by ADSCs, ADSCs were co-cultured with the DiO-labeled exosome-like vesicles for 6 h then visualized with confocal microscopy. The DiO-labeled (green) Exo-ADSCs or Exo-AT were seen to surround the nuclei after entering cells, which implies that endocytosis might be the main mechanism through which ADSCs internalized exosome-like vesicles (Fig. 1D).

To examine the effects of Exo-AT on adipogenesis, 10⁵ ADSCs were continuously co-cultured with 80 μg of exosome-like vesicles for 10 days. The culture medium was changed every 3 days to maintain the treatment effect. Adipogenic induction medium was used as a positive control. Lipid droplets could be observed after 10 days induction in Exo-AT-treated cells (Fig. 2A). Adipogenesis was further determined by Oil Red O staining. The results showed that optical density (OD) value of extracted Oil Red O was also higher in Exo-AT group than in the Exo-ADSC group (Fig. 2B). Similarly, the mRNA levels for the adipogenic genes encoding PPARγ2, aP2 (also known as FABP4) and adiponectin, as analyzed by real-time quantitative PCR (qPCR), were slightly increased expression in Exo-AT group on day 5 and remarkably elevated on day 10 compared to in the Exo-ADSC group (Fig. 2C). However, the expression of mRNA encoding C/EBPδ, an early adipogenic gene, remained at a similar level (Fig. 2C). Western blotting revealed that Exo-AT had an increased amount of PPARγ2, lipoprotein lipase (LPL) and adiponectin, while non-treated ADSCs and the Exo-ADSC group had a lower amount of these proteins (Fig. 2D; Fig. S1A).

The emerging evidence supports the theory that exosome cargo miRNAs function as important regulators in cell differentiation. Therefore, in order to determine the mechanism by which Exo-AT mediated adipogenesis, we profiled the miRNAs in Exo-AT in two replicates by using high-throughput sequencing (miRNA-seq); Exo-ADSC was also profiled as a comparative sample. After trimming low-quality reads, contaminants, adaptors and reads smaller than 15 nt, the remaining reads were mapped to miRBase v21 (Table S2). A total of 148 and 154 types of known miRNAs were identified in Exo-ADSC and Exo-AT, respectively (Fig. 3A). Among these miRNAs, 103 miRNAs were detected in both Exo-ADSC and Exo-AT. Compared to Exo-ADSC, 45 conserved miRNAs were enriched (expressed ≥2-fold, FDR<0.05) in Exo-AT (Fig. 3B). A total of 14 of the 45 miRNAs that were enriched in Exo-AT (31.11%, such as miR-30a-5p and miR-148a-3p) are reported to participate in regulation of adipogenesis, whereas eight miRNAs (17.78%, such as miR-93-5p and miR-150-3p) are known to negatively control osteoblastic differentiation of MSCs, nine miRNAs (20%, such as miR-150-3p, miR-126a-3p) are associated with angiogenesis, four miRNAs (8.89%, such as miR-224) are related to metabolism and 15 miRNAs have not been reported to regulate adipogenesis (Table S3). To validate these miRNA-seq results, the expression of three most abundant miRNAs in Exo-AT (miR-450a-5p, miR-99a-5p and miR-30a-5p) were measured by qPCR. Consistent with the miRNA-seq results, qPCR showed that Exo-AT were enriched in miR-450a-5p, miR-99a-5p and miR-30a-5p compared with Exo-ADSC (Fig. 3C).

In order to further investigate the function of the miRNAs that were enriched in Exo-AT, miR-450a-5p was selected as candidate since it is one of abundant miRNAs (reads count>1000) and also the most highly enriched miRNA in Exo-AT (56.69-fold) compared with Exo-ADSC, and it had also not been investigated in the context of adipogenesis, in contrast to other miRNAs (Table S3). We first determined the expression profile of miR-450a-5p in ADSCs during adipogenesis induced by adipogenic medium at different time points (0, 1, 2 or 3 days). miR-450a-5p was statistically upregulated during adipogenesis, which suggests that it has a contribution to adipogenesis (Fig. 4A). To further investigate the function of miR-450a-5p in adipogenesis, miR-450a-5p mimic or inhibitor was transfected into ADSCs. The efficiency of transfection was validated by qPCR (Fig. 4B). Morphological observation of lipid droplets showed that overexpression of miR-450a-5p (mimics) improved adipogenesis, while inactivation of the miR-450a-5p (inhibitors) impaired adipogenesis (Fig. 4C). Analysis of the amount of Oil Red O staining, showed that there was an increased accumulation of lipid droplets when miR-450a-5p (mimics) was overexpressed, while inactivation of the miR-450a-5p (inhibitors) decreased the accumulation of lipid droplets (Fig. 4D). Finally, we checked the expression of adipogenic genes, including those encoding PPARγ2, aP2 and adiponectin, by qPCR and western blotting. These results showed that the cells with overexpressed or inactivated miR-450a-5p had a similar adipogenesis characterization, as determined by Oil Red O staining (Fig.4E,F; Fig. S1B).

To identify the molecular mechanism through which miR-450a-5p is involved in adipogenesis, Targetscan, miRanda and miRbase were used to predict target genes. The genes predicted by all these algorithms were chosen as targets for further analysis. Among 110 potential targets identified for miR-450a-5p, only FZD1 and WISP2 have been described in the context of adipogenesis (Jeon et al., 2016). A previous study has reported that WISP2 plays a central role in regulation of adipogenic commitment and that inhibition of WISP2 could induce spontaneous adipogenesis in stem cells (Hammarstedt et al., 2013). FZD1 is a transmembrane receptor that mediates Wnt signaling. In the prediction results, FZD1 would be downregulated by miR-450a-5p, which means that FZD1 and miR-450a-5p should show a reverse expression profile during adipogenesis. However, Park et al. showed that FZD1 was upregulated during the early stage of adipogenesis (48 h) (Park et al., 2008), which is similar to miR-450a-5p expression profile (Fig. 4A). The controversy between the prediction and expression profile indicated that FZD1 might not be the target gene of miR-450a-5p. Moreover, there is no evidence that inhibition of FZD1 promotes spontaneous adipogenesis of MSCs (Yu et al., 2013). Therefore, we assessed the potential for WISP2 to be a target gene of miR-450a-5p for further research (Fig. 5A). We found that miR-450a-5p was upregulated (Fig. 4A) while WISP2 was downregulated during adipogenic differentiation of ADSCs (Fig. 5B). These results are consistent with the prediction. To further test the effects of miR-450a-5p on WISP2, ADSCs were treated with miR-450a-5p mimic at different concentrations (0 nM, 10 nM, 50 nM and 250 nM) to compare the expression level of WISP2. The results demonstrated that WISP2 was repressed by miR-450a-5p at both the mRNA and protein levels in a dose-dependent manner (Fig. 5C,D). Finally, to further validate whether WISP2 was a target gene of miR-450a-5p, we performed a dual-luciferase reporter assay to demonstrate that WISP2 could be directly inhibited by miR-450a-5p. Luciferase reporters that contained either a wild-type (pMIR-WISP2-WT) or mutated (pMIR-WISP2-MUT) WISP2 3′UTR were transfected into HEK-293 cells, along with either miR-450a-5p mimic or mimic control. When cells were transfected with miR-450a-5p mimic, luciferase reporter activities reduced. In contrast, when the WISP2 3′UTR mutant was transfected into cells, there was no effect on luciferase activity (Fig. 5E). These data indicate that miR-450a-5p could directly target the 3′UTR of WISP2 to regulate adipogenesis.

To identify whether Exo-AT could influence the expression of WISP2, ADSCs were co-cultured with exosome-like vesicles for 5 days. The expression level of miR-450a-5p in cells co-cultured with Exo-AT was upregulated compared with cells incubated with Exo-ADSC (Fig. 6A). Meanwhile, Exo-AT reduced the expression of WISP2 on day 5, indicating that Exo-AT could also regulate expression of WISP2 (Fig. 6B). To validate a crucial role of miR-450a-5p in the functions of Exo-AT, ADSCs were transfected with miR-450a-5p inhibitor and then treated with Exo-AT for 5 days. ADSCs treated with miR-450a-5p inhibitor and Exo-ADSC were used as a negative control. qPCR analysis showed that the level of miR-450a-5p was greatly elevated by Exo-AT while this effect was reversed by miR-450a-5p inhibitor (Fig. 6A). ADSCs treated with miR-450a-5p inhibitor failed to suppress the expression of WISP2 at both the RNA and protein levels (Fig. 6B,C; Fig. S1C). Moreover, the expression of adipogenic markers including PPARγ2, aP2 and adiponectin were also decreased at the mRNA (Fig. 6D) and protein level (Fig. 6E; Fig. S1D), indicating that inactivation of the miR-450a-5p impaired Exo-AT-mediated adipogenesis.

Adipose tissue secretes considerable amounts of multiple adipogenic factors in appropriate combinations to contribute to new adipose tissue development (Sarkanen et al., 2012). However, only some of the adipogenesis inducers in adipose tissue have been discovered because of the complexity of adipose tissue extract (Huttala et al., 2016). As a novel type of cell communication, exosomes have recently drawn much attention in cell differentiation (Choi et al., 2016). For example, it has been demonstrated that conditioned medium from neurons induces differentiation of MSCs into neural cells while exosomes derived from the conditioned medium can also account for differentiation (Krampera et al., 2007; Takeda and Xu, 2015). Therefore, in addition to growth factors and cytokines, exosomes from adipose tissue may also contribute to pro-adipogenesis effects.

In this study, we showed that ADSCs could uptake both Exo-AT but also Exo-ADSC; however, only Exo-AT were able to induce adipogenesis while Exo-ADSC had no corresponding effect. These results indicate that Exo-AT contains adipogenic factors that could trigger adipogenic signaling in ADSCs, resulting in adipogenic differentiation. In fact, previous studies have proved that the exosomes derived from osteoblasts, neurons and muscle cells can mediate differentiation of MSCs, although no work had been undertaken on Exo-AT (Narayanan et al., 2016; Cui et al., 2016; Batiz et al., 2015; Choi et al., 2016). Therefore, we conclude that exosomes may serve as a mediator of communication between stem cells and mature tissue cells. Once the stem cells differentiate into mature cells, they, in turn, induce stem cells to differentiate through exosomes, thus a positive-feedback loop mechanism is formed during differentiation (Fig. 7).

Exosomes contain a wide contingent of functional proteins, mRNAs and miRNAs (Sun et al., 2013). It has been indicated that proteins and miRNAs are involved in exosome-mediated differentiation. For example, TGF-β1 in cancer cell exosomes induce the differentiation of MSCs into cancer-associated myofibroblasts via the TGF-β1/SMAD signaling pathway (Chowdhury et al., 2015). Several studies have shown that exosomal miRNAs participate in multiple differentiation processes, such as osteogenesis, neurogenesis, angiogenesis and myogenesis (Lee et al., 2014; Qi et al., 2016), which indicates that miRNAs could be the main mechanism for exosome-mediated differentiation. However, whether miRNAs in Exo-AT acted as a significant component in exosomes to promote adipogenesis had not been elucidated. Therefore, we focused on the role of exosomal miRNAs in cell communication during adipogenesis. Interestingly, 14 miRNAs enriched in Exo-AT had been previously reported to promote adipogenic differentiation of MSCs. Several of these miRNAs (fold change ≥20 compared to Exo-ADSC) had been described as regulating adipogenesis (e.g. miR-125b-5p, miR-30e-3p and miR-204-5p), whereas several of them (e.g. miR-93-5p and miR-203-3p) had been reported to negatively regulate osteogenesis (Table S3) (Kennell et al., 2008; Cho et al., 2016; Hamam et al., 2014; Yang et al., 2012; An et al., 2014). Eguchi et al. also analyzed the miRNAs in adipocytes derived from EVs and detected several adipogenic miRNAs (including miR-148a-3p, miR-181a-5p, miR-30a-5p, miR-125b-5p) (Eguchi et al., 2016), which is consistent with our results (Table S3).

Here, we identified that miR-450a-5p, one of the most abundant miRNAs in Exo-AT, was a regulator of ADSC differentiation into adipocytes. Normally, adipogenesis-related miRNAs are downregulated in the obese subjects (Williams and Mitchell 2012). Taken together with the fact that decreased expression of miR-450a-5p has been reported in obesity (Delic et al., 2016), miR-450a-5p might be involved in adipogenesis. Moreover, miR-450a-5p is downregulated when Wnt signaling, a negative signal regulating adipogenesis, is activated by Li⁺ during adipogenesis of preadipocytes (Qin et al., 2010). Li⁺ can activate Wnt signaling through the inhibition of GSK-3β and further blocks the phosphorylation of β-catenin (Ross et al., 2000). Therefore, the downregulation of miR-450a-5p in Li⁺-treated pre-adipocytes indicates that miR-450a-5p might act as a positive regulator of adipogenesis by suppressing Wnt signaling. In our results, inactivation of miR-450a-5p partially diminished the adipogenic activity of Exo-AT. This suggests miR-450a-5p might be involved in Exo-AT-mediated adipogenic differentiation.

To determine the molecule mechanism for the action of exosomal miR-450a-5p during adipogenesis, we performed a bioinformatics analysis for potential targets of miR-450a-5p and found WISP2. WISP2, also known as CCN5, is a novel adipokine secreted by adipose tissue (Lehr et al., 2012) that is highly expressed in MSCs and preadipocytes (Hammarstedt et al., 2013). A previous study has suggested that WISP2 promoted proliferation and inhibited adipogenesis, while inhibition of WISP2 induced spontaneous adipogenesis (Hammarstedt et al., 2013). WISP2 is also known to regulate BMP4-dependent adipogenesis by forming a complex with ZNF423, a transcriptional activator of PPARγ2, or by directly activating Wnt signaling (Gustafson et al., 2013). Hence, we hypothesize that miR-450a-5p transferred from exosome-like vesicles might promote adipogenesis through downregulation of WISP2. As expected, ADSCs treated with Exo-AT or overexpressing miR-450a-5p had a substantial decrease in the expression of WISP2. These findings indicate a new role of miR-450a-5p transferred from exosome-like vesicles in regulating adipogenesis by modulating its target WISP2.

Although our data suggest that miR-450a-5p had an important role in the Exo-AT-mediated adipogenesis, we cannot rule out the contribution of other exosomal cargoes. Inhibition of miR-450a-5p does not arrest Exo-AT-mediated adipogenesis completely. Eguchi et al. performed a comprehensive protein content characterization of EVs derived from stressed adipocytes using tandem mass spectroscopy (LC-MS/MS) and found that the functions of proteins in EVs were mainly involved in translation, carbohydrate and lipid metabolism, cytoskeleton, extracellular matrix, ribonucleoproteins and nucleosomes (Eguchi et al., 2016). Connolly et al. have reported that EVs from adipocytes are enriched in markers of adipogenesis (such as PREF-1, also known as DLK1, and PPARγ). The amount of PREF-1, a negative regulator of adipogenesis, decreased in EVs derived from cells during the later stages of differentiation (Connolly et al., 2015). Other adipogenesis-related proteins were also detected in adipocyte-derived exosomes including adenylate kinase 2 (AK2), and matrix metalloproteinase-2 and -9 (MMP-2 and MMP-9) (Lee et al., 2015; Aoki et al., 2007). Furthermore, mRNAs encoding the adipogenesis-related proteins PPARγ, leptin, C/EBP-α and C/EBP-δ have also been detected in adipocyte-derived EVs and as being delivered to osteoblasts (Martin et al., 2015). Therefore, components other than miRNAs in exosome-like vesicles may also participate in regulating adipogenesis.

Our results provide novel insights into the intercommunication between ADSCs and adipose tissue cells in the adipose tissue microenvironment. Considering that exosome-like vesicles are completely cell-free and not highly immunogenic, exosome-like vesicles could be an important tool for use during adipose tissue engineering (Konala et al., 2016). Therefore, further research is needed to study the possibility that exosome-like vesicles act as a proadipogenic factor to enable use during adipose tissue engineering.

The study was approved by the Ethical Committees of the State Key Laboratory of Oral Diseases, West China School of Stomatology, Sichuan University, China. ADSCs were prepared and cultured as described in our previous studies (Jing et al., 2011; Zuk et al., 2002). Briefly, inguinal fat pads were collected from 4-week-old Sprague-Dawley (SD) rats and washed extensively with sterile phosphate-buffered saline (PBS) to remove the debris and red blood cells. Then the samples were cut into small pieces (1–2 mm³) and treated with 0.075% collagenase (type I) for 30 min at 37°C. The cells were maintained at 37°C in a CO₂ incubator. When the cells reached 80% confluence, the cells were trypsinized and passaged at a 1:3 ratio.

To collect the culture medium, cells were cultured to 90% confluence, washed twice with PBS, switched to serum-free medium (α-MEM, 100 U/ml penicillin and 100 μg/ml streptomycin) and cultured for 3 days. The serum-free culture medium was centrifuged at 150 g for 10 min to remove cell debris and kept at 4°C for further experiments.

Exo-AT or Exo-ADSC were obtained from adipose tissue extract (ATE) or ADSC culture medium, respectively.

To obtain adipose tissue extract, 5 g adipose tissue explants, achieved by mincing adipose tissue from 4-week-old SD rats, was added into the Celstir Spinner Flask (Wheaton, IL) supplemented with 75 ml α-minimal essential medium (α-MEM, Hyclone), 100 U/ml penicillin and 100 μg/ml streptomycin, and cultured with shaking at 100 rpm for 3 days. ATE was centrifuged at 150 g for 10 min to remove the cell debris and stored at −80°C, after determining its protein concentration by the BCA method.

ATE or ADSC culture medium was first introduced into Amicon® Ultra-15 Centrifugal Filter Units with an Ultracel-3 membrane (3 kDa molecular mass cut off membrane, Millipore) and centrifuged at 5000 g for 30 min. The concentrate was then mixed with 0.5 volume of Total Exosome Isolation™ reagent (Life Technologies), incubated overnight at 4°C and spun down for 1 h at 10,000 g at 4°C. The pellet was re-suspended in 100 μl PBS and stored at −80°C, after determining its protein concentration by the BCA method.

Exosome-like vesicles were fixed with 1% glutaraldehyde at 4°C overnight. After washing, vesicles were loaded onto formvarcarbon-coated grids, negatively stained with aqueous phosphotungstic acid for 60 s and imaged with a transmission electron microscope (Hitachi H7500 TEM, Japan).

The size of vesicles was determined by a dynamic light scattering technique using a ZetasizerNano ZS analysis system (Zetasizer version 6.12; Malvern Instruments, UK). A size distribution plot, with the x-axis showing the distribution of estimated particle radius (nm) and the y-axis showing the relative percentage, was made.

Exosome-like vesicles were labeled with a membrane-labeling dye DiO (Invitrogen) as previously described (Choi et al., 2016), and were then washed and resuspended in serum free α-MEM. Next, ADSCs were co-cultured with DiO-labeled vesicles for 6 h, washed with PBS three times, fixed in 4% paraformaldehyde, stained with phallotoxins (Invitrogen), washed with PBS and imaged by confocal microscopy (Olympus FV1000, Japan).

ADSCs at passage 3 were plated in six-well plates at a density of 10⁵ cells/well, cultured for 24 h, then rinsed with PBS and incubated with 2 ml of one of four different culture media for up to 10 days. The media used were: (1) basal medium [α-MEM supplemented with 10% fetal bovine serum (FBS)], as a negative control; (2) basal medium supplemented with Exo-ADSC (the concentration of Exo-ADSC was 40 µg/ml); (3) basal medium supplemented with Exo-AT (the concentration of Exo-AT was 40 µg/ml); and (4) adipogenic medium (α-MEM supplemented with 10% FBS, 1 mM DEX, 10 mM insulin, 200 mM indomethacin and 0.5 mM IBMX), as a positive control (PC) (Zuk et al., 2002). The medium was changed every 3 days. The cells were collected at day 5 or 10 for qPCR analysis or western blotting. After 10 days in culture, adipogenic differentiation was determined by Oil Red O (Sigma-Aldrich) staining. Then, Oil Red O in cells was extracted with 100% isopropanol for 15 min. The absorbance was measured at 510 nm with a spectrophotometer (MultiskanGO, Thermo Scientific).

Total RNA from Exo-AT and Exo-AT was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. Total RNA was fractionated on a 15% Tris-borate-EDTA (TBE) polyacrylamide gel (Invitrogen) and a band corresponding to small RNAs (18–30 nt) was excised. Isolated small RNAs were terminal repaired by adding 5′- and 3′-adapter, reverse transcribed into cDNA and amplified. Raw small RNA sequence data were obtained by using a Illumina HiSeq™2000 machine. After sequencing, the Solexa CHASTITY quality filtered reads were harvested as Clean Reads. The trimmed reads were aligned to miRBase v21 using Novoalign software (v2.07.11) allowing, at most, one mismatch. miRNAs in Exo-AT and Exo-ADSCs were profiled in two biological replicates. Our miRNA profiling data have been deposited at the Gene Expression Omnibus (GEO) repository, with accession numbers from GSM2425598–GSM2425601 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE92313). We tested for a difference between the Exo-AT and Exo-ADSC with a paired two-sided t-test and corrected for multiple testing by using the Bonferonni method. Fold change and P-values were calculated for each miRNA. These P-values were used in testing the false discovery rate (FDR) values for each miRNA, which was further used as a filter to get significant miRNA hits with a fold change ≥2 and FDR<0.05. We analyzed expression data with Multi Experiment View (MEV) cluster software. To identify putative targets of different expressed miRNAs, TargetScan (http://www.targetscan.org/), miRanda (http://www.microrna.org/microrna/) and miRbase (http://www.mirbase.org/) were applied. The genes predicted by all these algorithms were chosen as target genes.

ADSCs (5×10⁴ per well) were seeded in 12-well plates and cultured overnight. 500 μl transfection mix (Lipofectamine 3000; Invitrogen) with miRNA mimic or inhibitor (Ruibo, China) was added into cells and incubated at 37°C for 6 h. The transfection efficiency of mimic (50 nM) and inhibitor (200 nM) was confirmed by qPCR. To explore functions of miR-450a-5p during adipogenesis, ADSCs were transfected with mimic or inhibitor then cultured with 1 ml adipogenic medium (α-MEM supplemented with 10% FBS, 1 mM DEX, 10 mM insulin, 200 mM indomethacin and 0.5 mM IBMX) for 5 days (fresh adipogenic medium were changed every 3 days). RNA and protein were extracted at day 5 after induction. To explore the role of miR-450a-5p in Exo-AT-mediated adipogenesis, ADSCs (5×10⁴ per well, 12-well plates) were transfected with miR-450a-5p inhibitor (200 nM), then cultured in 1 ml medium supplemented with exosome-like vesicles (40 µg/ml) for 5 days. The medium was changed every 3 days.

756 bp of the WISP2 3′UTR region was amplified and cloned into a firefly luciferase reporter vector (Obio Technology, China), named as pMIR-WISP2-WT. A mutant version of the WISP2 3′UTR reporter plasmid (pMIR-WISP2-MUT) was also generated by mutating the seed region for miR-450a-5p. To evaluate the direct interaction between miR-450a-5p and the 3′UTR from WISP2, HEK-293 cells were co-transfected with 50 nM miRNA mimic, 1 μg of firefly luciferase reporter vector (pMIR-WISP2-WTor pMIR-WISP2-MUT) and with 100 ng of Renilla luciferase pRL-CMV vector (Obio Technology, China). Luciferase activity was measured using the Dual-Glo luciferase assay system (Promega) at 48 h post transfection. Normalized firefly luciferase activity was compared between different groups.

To identify exosome-like vesicles, 30 μg vesicles, as determined by the BCA method, was dissolved in RIPA Lysis Buffer (KeyGEN, China), resolved on a 10% polyacrylamide gel and blotted on to a nitrocellulose membrane. The membranes were blocked and then incubated with primary antibodies against CD9 (1:1000, Zen Bioscience, China, 220642), TSG101 (1:1000, Zen Bioscience, China, 341000), CD63 (1:1000, Zen Bioscience, China, 615509) and β-actin (1:1000, Abcam, ab3280) at 4°C overnight, followed by horseradish peroxidase (HRP)-conjugated secondary antibodies for 1 h at room temperature. Immobilon Western Chemiluminescent HRP Substrate (Millipore) was used for the detection following the manufacturer's instructions. Signals were visualized with a ImageQuant LAS 4000 mini machine (GE Healthcare). To detect adipogenic markers during adipogenesis, total proteins were extracted from treated cells at various times by using a Total Protein Extraction Kit (KeyGEN, China). 30 µg cellular protein from each sample was resolved by SDS-PGAE followed by immunoblotting as above. Primary antibodies against PPARγ2 (Abcam, ab45036), adiponectin (Abcam, ab62551), WISP2 (Zen Bioscience, China, 507456) (all used at a dilution of 1:1000) and GAPDH (1:5000, Zen Bioscience, China, 200306-7E4) were used in this study. Band intensities were determined using ImageJ software and normalized to internal control GAPDH.

Total cellular RNAs were extracted using RNAiso Plus (TaKaRa Biotechnology) according to the manufacturer's instruction. The RNAs were transcribed into cDNAs with a First Strand cDNA Synthesis Kit (Thermo Scientific). Specific cDNAs were amplified with SYBR Premix ExTaq (TaKaRa Biotechnology, Japan) utilizing an Eco Real-time PCR System (Illumina). Reaction conditions were: 95°C for 2 min; followed by 40 cycles of 95°C for 5 s, 60°C for 30 s. The results were analyzed by using the 2^−ΔΔCT relative quantitative method, with GAPDH as an internal control. miRNAs from 1 mg vesicles were extracted with an miRNA Isolation kit (OMEGA) according to the manufacturer's instructions, then transcribed into cDNA by using an miRcute Plus miRNA First-Strand cDNA Synthesis Kit (TIANGEN, China). cDNAs were amplified with an miRcute miRNA qPCR Detection Kit (SYBR Green) (TIANGEN, China) by utilizing an Eco real-time PCR system (Illumina). U6 was used as an internal control. Primer sequences are listed in Table S1.

Each experiment was repeated at least three times. Data are expressed as mean±s.d. Statistical analysis was performed with a paired Student's t-test. P<0.05 was considered statistically significant.