The effect of gap junction-mediated transfer of miR-200b on osteogenesis and angiogenesis in a co-culture of MSCs and HUVECs

Currently, cell-based strategies to vascularize engineered bone tissue in vitro are arousing great interest for bone regeneration. It is anticipated that the vascular network in grafts could be improved through co-culture of osteogenic cells and vasculogenic cells, and that this network would amalgamate with the host vasculature and promote survival and regeneration of bone tissue (Tsigkou et al., 2010). However, the problem of insufficient vascularization in newly formed bone tissue still exists. In vivo, bone development and remodeling depend on complex interactions between bone-forming cells and other cells present within the bone microenvironment, particularly endothelial cells (ECs), which might be one of the pivotal cell types involved in a complex interactive communication network in bone. For successful engineering of pre-vascularized bone tissue in vitro, understanding the interactions between vasculogenic cells (endothelial cells or endothelial progenitors) and osteogenic cells [osteoblasts or mesenchymal stem cells (MSCs)] is a prerequisite. However, the cellular and molecular bases of osteogenic–endothelial cell interactions, as well as their impact on endothelial vascularization and osteogenic differentiation, are not fully understood.

The communication between vasculogenic cells and osteogenic cells may not only deploy diffusible soluble paracrine factors, but also involve the junctional communication through the formation of a multicellular network (Villars et al., 2000, 2002; Guillotin et al., 2004). Moreover, the secretion of soluble factors, such as VEGF, has been reported to rely on cell–cell contact via gap junctions (Grellier et al., 2009). Gap junctions are specialized intercellular channels formed by membrane proteins called connexins, which can mediate intracellular exchange of regulatory ions, small molecules (maximum molecular mass of 1.5 kDa) (Moorer and Stains, 2017), and siRNAs and microRNAs (miRNAs) (minor diameters of 1.0 nm) between adjacent cells (Valiunas et al., 2005; Inose et al., 2009). Connexin 43 (Cx43; also known as GJA1), a predominant gap junction protein in osteoblasts, has been demonstrated to form gap junctions between ECs and osteogenic cells in co-culture in several studies. The functional inhibition of gap junctions by means of treatment with 18α-glycyrrhetinic acid (18GA) or through inhibiting Cx43 synthesis with antisense oligodeoxyribonucleotide can attenuate the effect of human umbilical vein endothelial cells (HUVECs) on osteogenic differentiation in co-culture (Villars et al., 2002). In another study, gap junctions between apposed HUVECs and human osteoprogenitor cells were shown to form from Cx43 after a short-term co-culture (48 h), and synthesis of Cx43 was required for the stimulated alkaline phosphatase (ALPase) activity in this co-culture system (Guillotin et al., 2008). Moreover, Herzog et al. revealed that in co-culture of outgrowth endothelial cells (OECs) and osteoblasts with direct cell–cell contacts, both cellular organization of OECs into vascular structures and osteogenic differentiation of osteoblasts were promoted via Cx43-based gap junctions (Herzog et al., 2014).

miRNAs are a group of non-coding endogenous RNAs that are ∼21–25 nucleotides long, and usually function as negative regulators of gene expression through complementary binding to their target mRNAs. Numerous miRNAs have been identified to play critical regulatory roles in biological and pathological processes, such as cell differentiation, angiogenesis and cancer. For instance, miRNAs are critical to stem cell function, particularly in the maintenance of stemness or the lineage-specific differentiation (Kim, 2005; Lu et al., 2005). In particular, miRNAs are able to regulate the function of ECs during angiogenesis (Akhtar et al., 2015; Climent et al., 2015). Furthermore, several miRNAs have already been confirmed to regulate osteogenesis and angiogenesis simultaneously. Inhibition of miR-222 expression was shown to accelerate bone healing by stimulating both osteogenesis and angiogenesis (Yoshizuka et al., 2016). Bone regeneration has been to shown to be promoted through miR-26a mediating positive regulation of angiogenic-osteogenic coupling (Li et al., 2013). Yang et al. reported that the miR-497–195 cluster regulates the angiogenesis coupled with osteogenesis through maintaining endothelial Notch activity and HIF-1α stability (Yang et al., 2017).

miR-200b, which belongs to the miR-200 family, is expressed in a variety of cells and regulates key cellular functions, such as cell proliferation, motility, apoptosis, stem cell properties, angiogenesis and epithelial–mesenchymal transition (EMT) (Brabletz and Brabletz, 2010). As an anti-angiogenic miRNA, miR-200b targets v-Ets erythroblastosis virus E26 oncogene homolog 1 (ETS1), thereby inhibiting the angiogenic response of human microvascular endothelial cells (HMECs). When transfected with miR-200b, the ability of HMECs to migrate and to form tubes on Matrigel is inhibited (Chan et al., 2011). As an angiogenic inhibitor, by negatively regulating VEGF signaling, via targeting VEGF and its receptor KDR, its therapeutic potential in cancer has also been implicated (Choi et al., 2011). It has been reported that the TGF-β/ZEB/miR-200b signaling network creates a double-negative feedback loop that plays an essential role in the initiation of EMT and cancer metastasis (Xiong et al., 2012; Shen et al., 2015). Moreover, miR-141 and miR-200b could inhibit BMP-2-induced pre-osteoblast differentiation through translational repression of Dlx5 (Vimalraj and Selvamurugan, 2013).

Mounting evidence indicates that miRNAs can be considered as intercellular signals that allow cells to communicate. Cells can release miRNAs to specifically modulate physiological processes in recipient cells via exosomes (Montecalvo et al., 2012; Baglio et al., 2015) or gap junctions (Calderón and Retamal, 2016; Thuringer et al., 2016; Zong et al., 2016). In the present study, the objective was to gain an insight into the molecular mechanisms in cell coupling and reciprocal interactions between ECs and MSCs. In a direct co-culture between bone marrow-derived MSCs (BMSCs) and HUVECs under osteogenic induction conditions, we found that osteogenesis was significantly stimulated while angiogenesis was compromised. It was therefore hypothesized that miR-200b might function as a regulator in this process. Results demonstrated that BMSCs communicated with HUVECs via miR-200b, which could transfer from BMSCs to HUVECs through gap junctions formed of Cx43. In particular, TGF-β was found to trigger the transfer of miR-200b to HUVECs. As a negative regulator, the decrease of miR-200b in BMSCs induced upregulation of VEGF-A, which was partly responsible for the stimulatory effects on osteogenic differentiation. Once inside HUVECs, miR-200b reduced the angiogenic potential of HUVECs by downregulating of ZEB2, ETS1, KDR and GATA2. Treatment with the TGF-β inhibitor SB431542 or TGF-β-neutralizing antibody in co-culture caused the formation of capillary-like structures to be partially rescued. These findings could be fundamental to the engineering of pre-vascularized bone tissue in vitro.

Rat BMSCs at passage 3 were tagged with conjugated antibodies against CD34, CD45 (also known as PTPRC), CD44, CD73 (NT5E), CD105 (ENG), CD106 (VCAM1), CD29 (integrin β1, encoded by ITGB1) and CD90 (THY1). Flow cytometric analysis indicated that cells were strongly positive for the MSC markers CD29 and CD90 (≥99%), positive for CD44, weakly positive for CD106 and CD73, and negative for the hematopoietic stem cell markers CD45, CD34 and CD105 (Fig. S1A). Results showed that after subculture, hematopoietic cells were eliminated from bone marrow-derived cells, and BMSCs used in this study were of high purity.

Rat BMSCs at passage 3 were induced for osteo-, adipo- and chondro-genesis for 21 days to assess their capacity for multi-lineage differentiation [with osteogenic induction medium (OIM), adipogenic induction medium (AIM) or chondrogenic medium; see Materials and Methods]. After induction, the differentiated cells were stained with the indicated dyes (Fig. S1B). After osteogenic induction, BMSCs were able to produce mineral nodules as evidenced by strong intensities of Alizarin Red S staining. After adipogenic induction, lipid vacuoles were seen within the cells upon Oil Red O staining. For analyzing chondrogenic differentiation, we used Safranin O staining, which detects glycosaminoglycan (GAG), a characteristic extracellular matrix synthesized by chondrocytes. These cell differentiation assays demonstrated that BMSCs had multi-lineage differentiation potential.

As shown in Fig. 1A, co-culture of HUVECs and BMSCs with direct cell–cell contact in osteogenic induction conditions induced a significant amount of Ca²⁺ accumulation within 14 days of co-culture as compared to what was seen with BMSC monoculture, as shown by the stronger intensities of Alizarin Red S staining and greater number of mineralized nodules. The activity of ALPase within 14 days of osteogenic induction, as detected by Nitro Blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate (NBT/BCIP) staining showed a noticeable difference between co-culture and monoculture (Fig. 1B). More intensive NBT/BCIP staining and higher ALPase activity was observed in co-culture.

qRT-RCR analysis confirmed that the osteoblastic phenotype markers genes, such as Alp, Runx2, Ocn (also known as Bglap) and collagen 1 (also known as Col1a1), were greatly upregulated in direct co-culture compared with the level in BMSC monoculture (Fig. 1C). Importantly, the level of Cx43, a predominant gap junctional molecule, was significantly increased by 6.82-fold at the early time point in co-culture (1 day). Chemokine (C-X-C motif) ligand 9 (Cxcl9) was also enhanced by 9.23-fold on the first day of co-culture. The level of another critical chemokine for cell migration and cell differentiation, Cxcl12, in co-culture was lower than that in BMSC monoculture within 14 days. Direct cell contacts between BMSCs and HUVECs stimulated the expression of Vegf-a by 4.43-fold on day 1 in co-cultured BMSCs compared with monoculture.

The amount of cytokine secreted in culture medium was also quantified by using ELISA. As shown in Fig. 1D, more Cxcl9, VEGF-A and Cxcl12 was secreted by BMSCs, and more TGF-β was produced by HUVECs. Within 7 days, the secretion of Cxcl9 in co-culture was almost equivalent to the amount in BMSC monoculture. The content of VEGF-A was significantly enhanced in co-culture compared with that in monoculture. Co-culture led to a higher level of Cxcl12 production on day 1 and 3, but a lower level on day 5 and 7 compared with monoculture. TGF-β reached the peak value at the early time point (1 day) in co-culture, which was significantly higher than that in HUVEC monoculture.

To observe the formation of capillary-like structures, immunofluorescence staining was performed to detect the expression of CD31 in HUVECs in monolayers, co-culture in Transwell inserts (hereafter called Transwell co-culture) and direct co-culture in osteogenic induction conditions after 24 h. As shown in Fig. 2A, HUVECs in monoculture or Transwell co-culture failed to display capillary-like structure. Surprisingly, the formation of capillary-like network was also compromised in direct co-culture. However, co-cultured BMSCs and HUVECs in growth medium (GM) arranged into clusters, which were bridged with the interconnected network comprised of elongated CD31-positive cells (Fig. S2).

qRT-PCR analysis (Fig. 2B) demonstrated a significant downregulation of angiogenic genes, ZEB2, ETS1, KDR and GATA2, in direct co-culture compared with what was seen in monoculture of HUVECs. However, the levels of these genes in Transwell co-culture were similar to those in HUVEC monoculture. The expression of CXCR3, the receptor for Cxcl9, in monoculture, direct co-culture and Transwell co-culture remained unchanged. Interestingly, as seen in Fig. 2C, after direct co-culture, the gene expression level of Cx43 in HUVECs had a 5.93-fold increase over that seen in monoculture, whereas no obvious upregulation was found when they were co-cultured in Transwell.

Western blotting was further applied to determine the protein levels. As shown in Fig. 2D, KDR, the receptor for VEGF, was significantly decreased in direct co-culture compared with monoculture and Transwell co-culture. Additionally, the amount of phosphorylated KDR (P-KDR) in direct co-culture was decreased. Direct contact induced significant downregulation of ETS1 expression in comparison with monoculture and Transwell co-culture. There was no significant difference in the levels of CXCR3 among HUVEC monoculture, direct co-culture and Transwell co-culture.

As demonstrated above, direct co-culture induced upregulation in gene expression of Cx43 in both HUVECs and BMSCs. Hence, the formation of gap junctions between HUVECs and BMSCs was investigated. Immunofluorescence staining of Cx43 was performed after 24 h in co-culture, and HUVECs were identified with PE-conjugated human (h)CD31. As shown in Fig. 3A, Cx43 appeared mainly as bright, punctate staining, and was located at sites where BMSCs and HUVECs were close to each other, indicating the formation of gap junctions. A dye-coupling assay further confirmed that BMSCs and HUVECs were able to form functional gap junctions, as Calcein dye transfer took place within 3 h of cell ‘parachuting’ (see Materials and Methods) (Fig. 3B).

To elucidate the role of gap junctions in co-culture, the gap junction inhibitor 18GA (100 μM) was applied to block intercellular communication via gap junctions.

The osteogenesis observed after osteogenic induction for 14 days in cultures supplemented with 18GA was evaluated. Cells were stained with NBT/BCIP and Alizarin Red S to detect ALPase activity and Ca²⁺ accumulation, respectively. As shown in Fig. 4A, cells treated with 18GA showed obviously weaker staining compared to controls without 18GA. Quantification of Ca²⁺ accumulation and ALPase activity further confirmed this observation (Fig. 4B). In monoculture, ALPase activity was not affected by 18GA, except on day 7 and 14, and in the co-culture, ALPase activity was greatly decreased after 18GA treatment; only the 14-day treatment with 18GA induced lower Ca²⁺ accumulation in monoculture, while co-culture with 18GA gave a significant decrease of Ca²⁺ deposition after 7 and 14 days. qRT-PCR analysis demonstrated a significant decrease in mRNA level of Alp (also known as Alpl) in the presence of 18GA compared to co-culture without the inhibitor (Fig. 4C). The level of Runx2 was mainly affected in monoculture on day 3 and 7 after treatment with 18GA, while it remained unchanged in co-culture. 18GA had no obvious influence on the levels of Ocn and collagen 1 in both monoculture and co-culture. After treatment with 18GA in co-culture, the expression of Cx43, Cxcl9 and Vegf-a decreased significantly on day 1. The expression of Cxcl12 was not affected by 18GA in monoculture and co-culture. Western blotting results further showed that the levels of Runx2 and collagen 1 were hardly affected by 18GA, but the expression of Cxcl9 and VEGF-A was significantly impaired by 18GA, especially on day 1, showing synchronous changes with Cx43. Moreover, the amount of phosphorylated ERK1 and ERK2 (ERK1/2, also known as MAPK3 and MAPK1, respectively) was enhanced in co-culture in comparison to in monoculture, with total ERK1/2 unchanged (Fig. 4D).

After blocking gap junction function with 18GA for 24 h, the formation of capillary-like structures in co-culture was partially recovered, as seen in Fig. 4E. Furthermore, upon 18GA treatment, along with downregulation of Cx43 in co-culture, the decrease in expression of angiogenic genes in co-culture was attenuated; the levels of ZEB2, ETS1, KDR and GATA2 were significantly increased to different extents (Fig. 4F). In contrast, although 18GA induced a lower expression of Cx43 in HUVEC monoculture, the levels of angiogenic genes were unchanged.

Parathyroid hormone (PTH) was used to stimulate gap junction formation by increasing the expression level of Cx43. A BMSC monoculture and co-culture were treated with PTH (10 nM), with no PTH treatment serving as control. During a 14-day induction, cells were stained with NBT/BCIP and Alizarin Red S to detect ALPase activity and Ca²⁺ accumulation, respectively (Fig. S3A). Cells treated with PTH showed obviously stronger Alizarin Red S staining compared to controls. As shown in Fig. S3B, a quantitative measurement indicated that ALPase activity was not affected in monoculture by PTH, but was significantly increased in co-culture on day 1. Treatment with PTH induced greater Ca²⁺ accumulation in monoculture after 14 days and in co-culture after treatment for both 7 and 14 days. qRT-PCR analysis demonstrated that a significant increase in Alp mRNA level was observed in the presence of PTH in co-culture compared to control on day 1, that the level of Runx2 was altered in monoculture after treatment with PTH, but not in co-culture, and the levels of Ocn and collagen 1 were hardly affected by PTH in both monoculture and co-culture (Fig. S3C). After treatment with PTH in co-culture, the expression of Cx43 increased significantly on day 1, and Cxcl9 and Vegf-a were changed accordingly, showing the close correlation between gap junction formation and the levels of these mRNAs. By contrast, the level of Cxcl12 was hardly altered by PTH in monoculture and co-culture.

To further investigate the role of gap junction in regulating Cxcl9 and VEGF, siRNAs, named 3646 and 3710, respectively (see Materials and Methods), were used to specifically knockdown Cx43 in BMSCs. As shown in Fig. S4A, after siRNA treatment for 36 h, BMSCs showed a significant reduction in the Cx43 mRNA, level by 80% and 70% with siRNA 3646 and 3710, respectively, compared with control. When siRNA-transfected BMSCs were directly co-cultured with HUVECs in osteogenic induction conditions for 24 h, the level of Cx43 was decreased by 50%, whereas the level of Cx43 in non-transfected BMSCs and HUVECs co-culture was increased by 2.8-fold. Semi-quantitative PCR further demonstrated that siRNAs 3646 and 3710 could effectively knockdown Cx43 in BMSCs. However, the decrease of Cx43 in BMSCs did not directly affect the levels of Cxcl9 and Vegf-a. After co-culturing siRNA-transfected BMSCs with HUVECs, the levels of Cxcl9 and Vegf-a were obviously lower than that in non-transfected co-cultures (Fig. S4B). The protein levels of Cx43, Cxcl9 and VEGF-A were also measured through western blotting (Fig. S4C). After siRNA transfection for 48 h, a significant reduction of Cx43 was noticed in BMSCs, whereas the levels of VEGF-A and Cxcl9 remained unchanged. For co-culture between siRNA-transfected BMSCs and HUVECs, the levels of Cxcl9 and VEGF-A were slightly lower than those in co-culture with non-transfected cells. These data indicate that Cxcl9 and VEGF are not directly regulated by Cx43 in BMSCs. Instead, molecules communicated through Cx43-based gap junctions between BMSCs and HUVECs might be responsible for regulating Cxcl9 and VEGF.

miRNAs can regulate both osteogenesis and angiogenesis, and miR-200b is an anti-angiogenic regulator (Pecot et al., 2013). We therefore hypothesized that miR-200b could move through gap junctions between BMSCs and HUVECs to function as a regulator in co-culture.

As shown in Fig. 5A, the level of miR-200b in BMSC monoculture was significantly higher than that in HUVEC monoculture. After BMSCs and HUVECs were co-cultured for 24 h, cells were sorted by flow cytometry after staining with phycoerythrin (PE)-conjugated anti-hCD31 antibody, with PE-positive cells being HUVECs, and PE-negative cells being BMSCs. The ratio between BMSCs and HUVECs after sorting was almost the same as the original seeding ratio (2:1) (Fig. S5A). After co-culture, the level of miR-200b was significantly decreased by 88% in co-cultured BMSCs, whereas miR-200b had increased by 30-fold in co-cultured HUVECs (Fig. 5B).

To ascertain whether the transfer of miR-200b between cells actually happened, BMSCs transfected with FAM-labeled miR-200b were cocultured with Dil-stained HUVECs. After co-culture for 2 h, the green fluorescence in BMSCs was scattered outside HUVECs, which indicated miR-200b was still inside BMSCs (Fig. 5C). However, after 24 h, most of the green spots began to enter HUVECs, overlapping with the red cell membrane (Fig. 5C). and appearing yellow in HUVECs.

To further determine whether the transfer of miR-200b was mediated by gap junctions or secreted exosomes, the gap junction inhibitor 18GA and activator PTH, and the exosome inhibitor GW4869 were applied to the direct co-culture for 24 h. Co-cultured cells were then subjected to flow cytometry sorting, and the level of miR-200b was measured by qRT-PCR. As shown in Fig. 5D, when co-culture was carried out in the presence of 18GA, the decrease of miR-200b initially observed in co-cultured BMSCs was reversed to some extent, and the increase in co-cultured HUVECs was diminished. In contrast, PTH aggravated the decrease of miR-200b in co-cultured BMSCs and enhanced the increase in co-cultured HUVECs. However, the exosome inhibitor GW4869 did not significantly affect the levels of miR-200b in BMSCs and HUVECs in co-culture. Transwell co-culture was also performed for 24 h with or without GW4869. However, the level of miR-200b was not affected by the presence of GW4869 (Fig. S5B). Moreover, with GW4869 treatment, the ALPase activity in BMSCs monoculture and co-culture was unaffected, and GW4869 only slightly affected Ca²⁺ accumulation in the later stage (14 days) (Fig. S5C and D). These results indicate that the transfer of miR-200b from BMSCs to HUVECs takes place via gap junctions in direct co-culture.

To confirm the role of miR-200b on osteogenesis, BMSCs were treated with either miR-200b mimics or inhibitor. As shown in Fig. 6A, a far higher level of miR-200b, an ∼889-fold increase, was present in BMSCs transfected with miR-200b mimics compared to non-transfected cells, whereas a reduction by 85% in miR-200b was observed in BMSCs treated with the inhibitor. Furthermore, transfection with miR-200b mimics induced an obvious decrease in the expression of Cxcl9 and Vegf-a, while cells treated with the inhibitor showed the opposite effect (Fig. 6B). As shown in Fig. 6C, cells transfected with miR-200b inhibitor showed greater Ca²⁺ accumulation on day 7 and 14, and enhanced ALPase activity on day 7 compared with non-transfected cells.

Subsequently, the effects of VEGF on osteogenesis of BMSCs in co-culture were investigated. Upon supplementing the OIM with rat VEGF-A (10 ng/ml), osteogenesis was stimulated, as evidenced by enhanced ALPase activity and Ca²⁺ accumulation compared with that in the untreated group after 14 days of osteogenic induction (Fig. 6D). A blocking antibody against VEGF was also employed in co-culture. As shown in Fig. S6, co-culture with the antibody present led to a weaker NBT/BCIP and Alizarin Red S staining compared with co-culture without the antibody, and quantitative measurement of ALPase activity and Ca²⁺ accumulation further confirmed these results.

To investigate the effects of miR-200b on angiogenesis, miR-200b mimics and inhibitor were transfected into HUVECs. After transfection with the mimics, the level of miR-200b in HUVECs had a 4128-fold increase compared with non-transfected cells, while it was reduced by 60% in cells transfected with the inhibitor (Fig. 7A). Additionally, as shown in Fig. 7B, HUVECs transfected with the mimics showed a significant decrease in the expression of ZEB2, ETS1, KDR and GATA2, whereas cells treated with the inhibitor showed the opposite.

Cell migration was also evaluated. A higher level of miR-200b in HUVECs transfected with the mimics induced a lower cell migration rate than what was seen in control, while the migration rate of cells treated with the inhibitor was greatly enhanced (Fig. 7C). Then, an in vitro Matrigel tube formation assay was exploited. It was found that HUVECs transfected with miR-200b mimics remained spherical and isolated, and no small cellular clusters or tubes appeared (Fig. 7D). In contrast, cells treated with the inhibitor formed branching and anastomosing tubes, resulting in a network with a capillary-like structure. These data suggest that miR-200b has an angiostatic effect on HUVECs.

TGF-β signaling is critical to cell differentiation and vessel stabilization, and it had been reported that TGF-β can regulate the transfer of miRNAs (Climent et al., 2015). Since it was shown above that direct contact co-culture between BMSCs and HUVECs on day 1 induced a 50% increase of TGF-β compared to that in monoculture, it was hypothesized that TGF-β could be the critical factor regulating the transfer of miR-200b to HUVECs. A specific TGF-β pathway inhibitor, SB431542 (10 μM), and TGF-β-neutralizing antibody (1.0 μg/ml) were employed to block the function of TGF-β during co-culture. As seen in Fig. S7A, with neutralizing antibody supplemented, the level of TGF-β in the supernatant was significantly lower than that in untreated co-culture. In addition, co-cultures treated with SB431642 secreted significantly lower TGF-β on day 1, whereas there was no significant difference on day 3, 5 and 7. Importantly, when co-cultures were treated with SB431542 or neutralizing antibody for 24 h, a significant reduction in miR-200b transfer to HUVECs was observed, and the decrease of miR-200b in BMSCs was not seen (Fig. S7B). However, when co-cultures were treated with TGF-β (10 ng/ml), the level of miR-200b in BMSCs and HUVECs was hardly affected. Through qRT-PCR analysis, it was further demonstrated that after blocking TGF-β in co-culture, ZEB2, ETS1, KDR and GATA2 were all significantly upregulated, although their expression levels were still lower than those in HUVEC monoculture (Fig. S7C).

A vessel formation assay on Matrigel was further undertaken to evaluate the angiogenesis in co-culture after supplementing TGF-β, SB431542 or neutralizing antibody. As shown in Fig. S7D, HUVECs alone on Matrigel could form intact capillary-like structures, whereas cells in co-culture without any treatment remained spherical and dispersed. Co-cultures treated with TGF-β exhibited similarly dispersed, heterogeneous cell clusters on Matrigel. In contrast, angiogenesis was partly recovered in co-cultures when they were treated with SB431542 or neutralizing antibody, as evidenced by formation of small cellular nests and short tubes.

Bone regeneration occurs in close spatial and temporal association with angiogenesis in vivo. Therefore, to engineer a vascularized bone tissue in vitro, it is necessary to take both osteogenesis and angiogenesis into consideration. By using co-culture between BMSCs and HUVECs, miR-200b was found to be an important regulator for both osteogenesis and angiogenesis, in a manner that depended on the formation of functional gap junctions between the two cell types. A proposed mechanism of action is illustrated in Fig. 8. BMSCs and HUVECs in direct co-culture could form functional gap junctions via Cx43, through which miR-200b could transfer from BMSCs to HUVECs. TGF-β secreted by HUVECs was the critical factor triggering the transfer of miR-200b. As a negative regulator, the reduction of miR-200b in BMSCs induced upregulation of VEGF, thus promoting the osteogenic differentiation. Once inside HUVECs, miR-200b could diminish the angiogenesis of HUVECs by reducing the expression of ETS1, GATA2, ZEB2 and KDR, thus inhibiting cell migration and the formation of a vascular-like network. Although VEGF could promote angiogenesis, the downregulation of its receptor KDR in HUVECs limited its function.

Previous studies have concentrated on assessing the angiogenesis in co-culture between ECs and MSCs in proliferative medium. To address the impact of ECs on osteogenesis of MSCs, cells were also co-cultured in either osteogenic or osteogenic factor-free medium. However, few studies have addressed osteogenesis and angiogenesis simultaneously in co-culture in osteogenic induction conditions. In the present work, rat BMSCs and HUVECs were directly co-cultured in osteogenic medium containing endothelial cell growth supplement (ECGS) to keep endothelial cells alive. It was found that the osteogenic differentiation of BMSCs was significantly stimulated in co-culture compared to what was seen in monoculture within 14 days. However, the formation of capillary-like structures in co-culture was compromised. Previous studies have shown that when MSCs and ECs are co-cultured in proliferative medium, MSCs can support EC forming the capillary-like structures (Li et al., 2011, 2015; Carrion et al., 2013). Consistent with these findings, in co-culture between BMSCs and HUVECs in growth medium, vascularized structures were seen in the present study. It is possible that the osteogenic state of BMSCs may not support the formation of vascularized structure by ECs. In fact, the capillary-like structures failed to form in co-culture between the osteogenic-induced BMSCs and HUVECs (data not shown). Scherzed et al. have also reported that differentiated human MSCs showed restricted migration ability, which meant they were no longer able to support tube formation by ECs (Scherzed et al., 2016). On the other hand, the impact of ECs on osteogenesis of MSCs is controversial. Most studies have found that ECs had positive effects on osteogenesis of MSCs, with enhanced osteogenic gene expression, Ca²⁺ accumulation and ALPase activity (Gershovich et al., 2013; Kim et al., 2013; Gurel Pekozer et al., 2016; Wu et al., 2016). Other studies have indicated that ECs may negatively affect osteogenesis of MSCs. Duttenhoefer et al. found that endothelial progenitor cells impaired the osteogenic differentiation of BMSCs (Duttenhoefer et al., 2015). Co-culture between adipose tissue-derived MSCs and HUVECs led to less-efficient osteogenesis and repair of rat cranial defects than in the MSC monoculture group (Ma et al., 2014). Meury et al. indicated that ECs inhibited osteogenesis of BMSCs through interfering with Osterix (also known as SP7) expression (Meury et al., 2006).

Communication, regulation of cellular processes and development in cell culture are impossible without connexins. They are important and obligatory mediators that are involved throughout angiogenesis and osteogenesis. Cx43 is the most abundant connexin in the osteoblast lineage and plays critical roles in bone formation (Watkins et al., 2011). In addition, downregulation of endothelial Cx43 causes endothelial dysfunction and impaired angiogenesis (Wang et al., 2013). In vivo, osteogenic cells residing in a perivascular niche lie in close proximity to ECs, and therefore gap junctional communication between heterotypic cells is inevitable (Crisan et al., 2008). Several studies have revealed that by directly co-culturing osteogenic cells and ECs, gap junctions were formed via Cx43, and only with direct contact, could osteogenesis be enhanced (Villars et al., 2000, 2002; Guillotin et al., 2008; Herzog et al., 2014). In addition, various pathways acting downstream of Cx43 in bone, including those involving ERK, PKCδ, β-catenin and protein kinase A, have been reported (Stains and Civitelli, 2005; Hebert and Stains, 2013; Gupta et al., 2016). Cx43-dependent ERK and PKCδ pathways further regulate the transcriptional activity of Runx2, with increased Cx43 level promoting Runx2 activity, which drives the expression of several osteoblast genes and vice versa (Niger et al., 2013). In the present study, when BMSCs and HUVECs were in direct co-culture, osteogenic differentiation of BMSCs was significantly enhanced. In addition, inhibition of Cx43 in BMSC monoculture could significantly decrease the expression of Runx2. However, blocking the gap junctional intercellular communication between the two cell types could not induce the downregulation of Runx2. In addition, 18GA, an inhibitor of gap junction function, had no obvious negative impact on the expression of ERK. In one report, inhibition of gap junctions was shown to block the expression of Ocn and Bsp in osteoblasts, and its effect on the expression of collagen type I was low (Schiller et al., 2001). However, the effects of 18GA on the expression of Ocn and collagen I in BMSC monoculture and co-culture were very minimal in the present study. Importantly, it was found that the expression level of Cxcl9 and Vegf-a was regulated by gap junctions formed between BMSCs and HUVECs. Moreover, by exploiting siRNAs, a correlation between Cxcl9 and Vegf-a levels and gap junction function was established. The elevated level of VEGF-A could partly be responsible for the enhancement of osteogenesis in BMSCs in co-culture. The secreted Cxcl9 had almost no influence on ALPase activity, and slightly inhibited Ca²⁺ accumulation in the late phase of osteogenesis (data not shown). Cxcl9, as an angiostatic factor secreted by osteoblasts, could interact with VEGF and prevent its binding to ECs and osteoblasts, thus abrogating angiogenesis and osteogenesis both in vivo and in vitro (Huang et al., 2016). In co-culture as described in the present study, Cxcl9 slightly inhibited cell proliferation and migration and prevented HUVECs from forming capillary-like structures. However, Cxcl9 was not responsible for downregulation of angiogenic genes, including ETS1, GATA2, ZEB2 and KDR (data not shown).

The miR-200 family comprises five members, miR-200a, -200b, -200c, miR-429 and miR-141, which regulate the EMT, and all appear to regulate tumor angiogenesis by targeting and repressing the expression of several key mRNAs. Several reports have linked low miR-200 levels with stimulation of angiogenesis through upregulation of ETS1, TUBB3, VEGF, KDR, SUZ12, ERRFI-1 and ZEB1/2; however, the clinical relevance of miR-200 expression appears to vary for different cancers (Pecot et al., 2013). ETS1, originally detected in lymphoid cells of adult tissues, has been identified as being the first transcription factor expressed during the angiogenesis of ECs in the embryo (Lelièvre et al., 2001). It has been reported that miR-200b targets ETS1 and induces inhibition of angiogenic response in ECs (Chan et al., 2011). GATA2 is critical for proliferation and differentiation of early hematopoietic progenitors (Tsai et al., 1994). GATA2 regulates the expression of vascular endothelial growth factor receptor-2 (VEGFR-2) during both vascular development and angiogenesis, the process by which ECs form new blood vessels from an existing vascular network (Mammoto et al., 2009). Transcription factors of the ZEB protein family (ZEB1 and ZEB2) and miR-200 family members can form a double-negative feedback loop, which controls EMT and mesenchymal–epithelial transition (MET) programs in both development and tumorigenesis (Hill et al., 2013). It has also been demonstrated that KDR, as an important VEGFR, is negatively regulated by miR-200b (Choi et al., 2011). In the present study, after co-culture, the level of miR-200b in HUVECs was significantly enhanced, whereas a decreased level was noticed in BMSCs. The critical role of miR-200b in co-culture was further confirmed upon the transfection of BMSCs with its mimics and inhibitor. As an angiostatic factor, a higher level of miR-200b in HUVECs reduced the expression of ETS1, GATA2, ZEB2 and KDR, thus inhibiting cell migration and the formation of a vascular-like network. Moreover, inhibition of miR-200b in BMSCs derepressed the expression of VEGF and thus promoted osteogenesis.

Mounting evidence implies that miRNAs could be considered as a language that allows cells to communicate with each other, stressing their important roles in the intercellular signal transduction. For example, smooth muscle cells (SMCs) can deliver miR-143 and/or -145 (miR-143/145) to ECs via intercellular tubes. Once in ECs, miR-143/145 repressed hexokinase II and integrin β8 and thereby reduces the angiogenic potential of the recipient cell (Climent et al., 2015). Zhou et al. found that endothelial miR-126 acts as a key intercellular mediator for an increase in SMC turnover by repressing the expression of FOXO3, BCL2 and IRS1 (Zhou et al., 2013). Sun et al. demonstrated that osteoclasts secreted miRNA-enriched exosomes, which mediated a transfer of miR-214 into osteoblasts to inhibit their function (Sun et al., 2016). MSC-derived exosomes might serve as an important mediator of cell-to-cell communication within the tumor microenvironment that would suppress angiogenesis by transferring anti-angiogenic molecules, such as miR-16 (Lee et al., 2013). Additionally, miR-145 has been shown to transfer from microvascular ECs to colon cancer cells, thus inhibiting angiogenesis (Thuringer et al., 2016). However, the roles of miRNAs in the crosstalk between MSCs and ECs remain largely elusive. Liang et al. found human adipose-derived MSCs secreted exosomes enriched with miR-125a, which could regulate angiogenesis of ECs in vitro and in vivo (Liang et al., 2016). Here, we show that miR-200b is transferred from cell-to-cell via gap junctions, as well as demonstrating its role in intercellular communication between MSCs and ECs. Therefore, it would be possible to modulate the expression as well as transfer of miR-200b for better formation of vascularized bone tissues. Taken together, our findings may be fundamental to the development of cell-based bone regeneration strategies by providing new insight into how an efficient strategy for engineering pre-vascularized bone tissue in vitro could be developed.

All procedures on rats were performed in full compliance with the guidelines of the ethics committee at East China University of Science and Technology. We used 4-week-old Sprague Dawley rats (SD rats), purchased from Shanghai SLAC Laboratory Animal Co., Ltd, that were SPF grade, male and weighed ∼80–120 g. Bone marrow mesenchymal stem cells (BMSCs) were isolated using a bone marrow adherence method as previously described (Jin et al., 2018). Cells were cultured in minimum essential medium α medium (α-MEM) (Gibco) supplemented with 10% fetal bovine serum (FBS) (Hyclone) at 37°C in a 5% CO₂ humidified atmosphere (proliferative medium). Cells at passage 3 to 6 were used in our experiment. Cell surface antigens were analyzed on a flow cytometer (BD FACS Calibur). The antibodies used are shown in Table S1.

Human umbilical vein endothelial cells (HUVECs) were purchased from ScienCell and cultured in endothelial cell growth medium (ECM) (ScienCell), containing 5% FBS, 100× penicillin-streptomycin solution and supplemented with 100× endothelial cell growth supplement (ECGS). Cells at passage 3 to 8 were used in our experiment.

BMSCs at passage 3 were seeded in 24-well plates at a density of 1×10⁴ cells/well and grown in α-MEM. After cell adhesion, the proliferative medium was discarded, and a specialized medium, according to the desired differentiation, was directly added to the cell culture. The assays were performed in triplicate for each group, with three independent experiments.

The cultured cells were treated with osteogenic induction medium (OIM), comprising Dulbecco's modified Eagle's medium (DMEM) (Gibco) supplemented with 10% FBS, 10⁻⁷ M dexamethasone (Sigma), 10 mM β-glycerol phosphate (Sigma) and 50 μg/ml L-ascorbic acid (Sigma). The medium was replaced twice a week. According to different experiments, factors such as rat VEGF-A (10 ng/ml) (Peprotech) were added to OIM as needed. After induction, mineralized nodules were detected by Alizarin Red S staining. ALP enzymatic activity was analyzed by using the NBT/BCIP alkaline phosphatase color development kit (Beyotime). The quantification of Ca²⁺ accumulation and ALPase activity was performed as previously described (Fan et al., 2017).

The cells were cultured in adipogenic induction medium consisting of DMEM, 10% FBS, 1 μM dexamethasone, 0.1 mM indomethacin (Sigma), 0.1 mM 3-isobutyl-1-methyl-xanthine (IBMX) (Sigma) and 10 mg/l insulin (Sigma). The medium was changed twice a week. After induction, cells were stained with 0.2% Oil Red O solution (Sigma) to detect lipid droplets within the cells.

1×10⁶ cells were suspended in 20 μl α-MEM to form a high-density cell pellet. The cell pellet was induced using chondrogenic medium, comprising DMEM, 1% ITS+ (BD Biosciences), 50 μg/ml L-ascorbic acid, 0.1 μM dexamethasone, 40 μg/ml L-proline (Sigma), 1 mM sodium pyruvate (Sigma) and supplemented with 10 ng/ml TGF-β1 (Peprotech). The medium was changed every 3 days. After induction, cells were stained with Safranin O (Sigma).

To carry out direct co-culture, both cell types were expanded separately in the media described above and then co-cultured in 24-well plates or cell culture flasks. For co-culture, ECGS was used to maintain the survival of ECs. BMSCs and HUVECs were mixed and seeded at a ratio of 1:2 in growth medium (1:1 α-MEM:ECM supplemented with 100× ECGS), at the density of 5000 cells/cm² and 10,000 cells/cm², respectively. Cells were seeded in the growth medium. After cell adhesion, the growth medium was replaced with OIM, and according to the different experiments, 100 μM 18GA (Sigma), 10 nM PTH (Sigma), 10 μM GW4869 (Sigma), 10 μM SB431542 (Sigma), 10 ng/ml TGF-β (Peprotech), 1.0 μg/ml neutralizing antibody of TGF-β (R&D) or 0.5 μg/ml VEGF blocking antibody (R&D) was added into the OIM.

Indirect co-culture was established using Transwell inserts with 0.4-μm filters (Corning). Depending on different experiments, either BMSCs or HUVECs were plated in the Transwell inserts.

Total RNA was extracted using Trizol reagent (Invitrogen) and cDNA was obtained using MLV reverse transcriptase (Promega) following the manufacturer's instructions. Quantitative real-time PCR (qRT-PCR) was performed on a RT-PCR system (Bio-Rad CFX96) using a SYBR mix (Roche) according to the manufacturer's instructions. PCR conditions were 95°C for 10 min, followed by 40 cycles of amplification at 95°C for 10 s and 57°C–63°C (depending on the primers used) for 20 s. GAPDH was used as a house keeping gene.

For semi-quantitative PCR analysis, PCR conditions were 94°C for 3 min, followed by 30 cycles of amplification at 94°C for 30 s, and 60°C for 30 s. PCR products in a 25 μl aliquot were size separated by electrophoresis in 2% agarose gels.

For miRNA isolation, total RNA was extracted with a miRNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. Reverse transcription of 1 μg RNA was carried out with a miScript II RT Kit (Qiagen). A MiScript SYBR Green PCR kit was used to quantitatively analyze the level of mature miRNAs. U6 snRNA was used as an internal control. The relative quantitative expression of miR-200b was normalized to the level of U6 snRNA. The specific primers for miR-200b and U6 were purchased from Qiagen.

Gene expression was analyzed according to the ΔΔCT method. The specifics of primers are listed in Table S2.

Cells were harvested and lysed in protein extraction reagent RIPA lysis buffer (Beyotime) containing 10 mM phenylmethylsulphonyl fluoride (PMSF) (Beyotime) on ice for 30 min. After centrifugation for 1 h at 12,000 g, the supernatant was retained. Protein concentrations were determined by using the BCA assay (Beyotime) following the manufacturer's instruction. After being heated for 5 min at 95°C in the sample buffer (Beyotime), equal aliquots of cell lysates were run on a 10% SDS-polyacrylamide gel. The separated proteins were electrically transferred onto PVDF membrane (Millipore). The membrane was blocked with 5% (w/v) non-fat dry milk in TBST buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.05% Tween-20) for 1 h at room temperature, and then incubated overnight at 4°C with the primary antibodies (Table S3). The bound primary antibodies were detected with secondary antibodies conjugated to alkaline phosphatase (ALPase) or horseradish peroxidase (HRP), and visualized through NBT/BCIP staining (Beyotime) or enhanced chemiluminescence (Millipore). β-actin was used as an internal control, and its expression was used to standardize input protein to analyze relative protein expression.

Cells in 24-well plates were fixed in 4% (w/v) paraformaldehyde for 10 min at room temperature. After washing in PBS, cells were permeabilized with 0.25% Triton X-100 for 10 min and blocked with 1% (w/v) bovine serum albumin (BSA) for 30 min. Cells were incubated for 60 min at room temperature with primary antibodies (Table S3), each diluted at 1:100. As secondary antibodies, Alexa Fluor 488 goat anti-rabbit-IgG (Santa Cruz Biotechnology) at a dilution of 1:200 were used. The cell nucleus was visualized via DAPI (Invitrogen) staining. Fluorescence staining was visualized in a fluorescence microscope (Nikon Eclipse Ti).

The function of the gap junctions was assessed using a ‘parachute’ dye-coupling assay. Donor cells (BMSCs or HUVECs) were double-labeled with 1 μM Calcein AM (Sigma), which is intracellularly converted into the gap junction-permeable dye Calcein, and 5 μM Dil (Beyotime), which is a membrane dye that cannot spread to coupled cells, for 30 min. Unincorporated dye was removed by three consecutive washes with culture medium. Donor cells were then trypsinized and seeded onto the receiver cells (HUVECs or BMSCs) at a 1:200 ratio, and the two cell types were allowed to attach to the monolayer and form gap junctions for 3 h at 37°C. Fluorescence microscope was used for imaging.

Cells were transfected by means of Lipofectamine RNAiMAX according to the manufacturer's protocol (Invitrogen).

siRNAs targeting the rat Cx43 gene (3646, 5′-GCUGGUUACUGGUGACAGA-3′; 3710, 5′-GAACUACAGCGCAGAGCAA-3′) (RiboBio) were used to specifically downregulate the level of Cx43 in BMSCs, and nonspecific sequence siRNAs were used as a negative controls (NC) (RiboBio). Cells without transfection served as control, and cells treated with transfection reagent were considered mock. After 24 to 48 h of siRNA transfection, the knockdown efficiency was confirmed by qRT-PCR or western blot analysis.

miR-200b mimics (FAM labeled or not) or inhibitor (RiboBio) were transfected into BMSCs and HUVECs to evaluate its function, and negative control miRNAs (mimics NC and inhibitor NC) (RiboBio) were also used. MiR-200b mimics are synthetic double stranded RNAs, same as endogenous miRNA-200b. Transfection with miR-200b mimics can lead to a high level of miRNA-200b, and their regulatory role would be boosted, meaning that there would be gain-of-function effects. miR-200b inhibitor is a synthetic single-stranded RNA, and functions as antisense oligonucleotides that can prevent miR-200b from binding to its target mRNAs, thereby weakening the gene silencing effects. Their negative control miRNAs are the oligonucleotides that do not react with any mRNAs in cells. After 24 h of transfection, the level of miR-200b was measured by qRT-PCR or visualized with a fluorescence microscope. Sequences were as follows: miR-200b mimics, sense, 5′-UAAUACUGCCUGGUAAUGAUGA-3′ and antisense, 5′-UCAUCAUUACCAGGCAGUAUUA-3′; miR-200b mimics NC, sense, 5‘-UUUGUACUACACAAAAGUACUG-3′ and antisense, 5‘-CAGUACUUUUGUGUAGUACAAA-3′; miR-200b inhibitor, sense, 5′-UCAUCAUUACCAGGCAGUAUUA-3′; miR-200b inhibitor NC, sense, 5‘-CAGUACUUUUGUGUAGUACAAA-3′.

ELISA was used to quantify the content of rat vascular endothelial growth factor (VEGF) (E-EL-R2603c, Elabscience), rat chemokine (C-X-C motif) ligand 9 (Cxcl9) (E-EL-R0854c, Elabscience), rat chemokine (C-X-C motif) ligand 12 (Cxcl12) (E-EL-R0922c, Elabscience) and human TGF-β1 (E-EL-H0110c, Elabscience) expressed by cells. Culture media was collected and analyzed at the specific time points. Assays were carried out according to the manufacturer's protocol and analyzed on a microplate reader at a wavelength of 450 nm (Biotek ELX800). The assays were performed in triplicates for each group, with three independent experiments.

Co-cultures were incubated with 18GA (100 μM), PTH (10 nM), GW4869 (10 μM), SB431542 (10 μM), TGF-β1 (10 ng/ml) or TGF-β-neutralizing antibody (1.0 μg/ml) in 75 cm² culture flasks. After 24 h, cells were trypsinized and incubated with PE-conjugated anti-hCD31 antibody (R&D; Table S1). Co-cultured BMDCs and HUVECs can be separated by flow cytometry based on the fluorescence dye.

HUVECs were seeded at the density of 5000 cells/cm² in 24-well plates. Confluent cells (or transfected HUVECs) were serum-deprived for 16 h, and a linear wound was created in monolayer by scratching with a sterile pipette tip (200 μl yellow tip). Cells were washed with PBS to remove floating cells. After additional 6 h, cell migration into the wound was assessed by using a digital inverted microscope. The percentage of wound closure for a given time was calculated from at least five images.

HUVEC monoculture (or transfected HUVECs) or co-cultures were seeded on growth factor-depleted Matrigel (BD). The Matrigel was assessed for capillary-like structures 12 h later. Microscopic fields containing the tube structure formed on the gel were photographed at low magnification (×40). Five fields per test condition were examined.

All the experiments were repeated at least three times using three rats to provide BMSCs and from pooled HUVECs. All the data are presented as means±s.d. Statistical significance was evaluated using one-way ANOVA with the Student–Newman–Keuls post-hoc test when data could be assessed as normally distributed. Non-normally distributed data were statistically analyzed using the nonparametric Kruskal–Wallis test. A value of P<0.05 was considered statistically significant.