Baculovirus-Mediated miR-214 Knockdown Shifts Osteoporotic ASCs Differentiation and Improves Osteoporotic Bone Defects Repair

Osteoporotic patients often suffer from bone fracture but its healing is compromised due to impaired osteogenesis potential of bone marrow-derived mesenchymal stem cells (BMSCs). Here we aimed to exploit adipose-derived stem cells from ovariectomized rats (OVX-ASCs) for bone healing. We unraveled that OVX-ASCs highly expressed miR-214 and identified 2 miR-214 targets: CTNNB1 (β-catenin) and TAB2. We demonstrated that miR-214 targeting of these two genes blocked the Wnt pathway, led to preferable adipogenesis and hindered osteogenesis. As a result, OVX-ASCs implantation into OVX rats failed to heal critical-size metaphyseal bone defects. We further engineered the OVX-ASCs with a novel Cre/loxP-based hybrid baculovirus vector that conferred prolonged expression of miR-214 sponge. Gene delivery for miR-214 sponge expression successfully downregulated miR-214 levels, activated the Wnt pathway, upregulated osteogenic factors β-catenin/Runx2, downregulated adipogenic factors PPAR-γ and C/EBP-α, shifted the differentiation propensity towards osteogenic lineage, enhanced the osteogenesis of co-cultured OVX-BMSCs, elevated BMP7/osteoprotegerin secretion and hindered exosomal miR-214/osteopontin release. Consequently, implanting the miR-214 sponge-expressing OVX-ASCs tremendously improved bone healing in OVX rats. Co-expression of miR-214 sponge and BMP2 further synergized the OVX-ASCs-mediated bone regeneration in OVX rats. This study implicates the potential of suppressing miR-214 by baculovirus-mediated gene delivery in osteoporotic ASCs for regenerative medicine.

binding sites to miR-214) promotes OVX-BMSCs osteogenesis and augments the ability of OVX-BMSCs to heal critical-size bone defects in osteoporotic rats 12 . Although OVX-BMSCs can be engineered to heal osteoporotic bone defects, adipose-derived stem cells (ASCs) have become attractive for bone regeneration because ASCs can be isolated in large quantities by liposuction [13][14][15] . However, ASCs are inferior to BMSCs in osteogenic differentiation capability, often resulting in delayed or incomplete repair of large bone defects 16,17 . To overcome this problem, we have developed Cre/loxP-based hybrid baculovirus (BV) vector system comprising two BV: one expressing Cre recombinase and the other substrate BV harboring the transgene cassette flanked by loxP sites 18 . After co-transduction of ASCs with the two BV, the expressed Cre recognizes the loxP sequences, excises the transgene cassette off the substrate BV genome, thereby leading to the formation of episomal DNA minicircle encompassing the transgene within the cells. Such hybrid BV system enables sustained transgene expression 19 and improves ASCs differentiation and bone healing in vivo 12 .
Despite the promise of ASCs for bone healing, whether osteoporotic ASCs exhibit more favorable adipogenic differentiation vs. osteogenic differentiation and the underlying mechanism remain unknown. Also, whether osteoporotic ASCs can be used to heal osteoporotic bone defects has yet to be explored. Using ASCs isolated from OVX rats (OVX-ASCs), here we investigated the miR-214 expression level, adipogenesis/osteogenesis preference, molecular pathway and how miR-214 affects the differentiation of surrounding OVX-BMSCs in vitro and bone healing in OVX rats. We further exploited the Cre/loxP-based BV persistently expressing miR-214 sponge to engineer OVX-ASCs, so as to knockdown intracellular and exosomal miR-214 levels, reverse the differentiation preference, substantiate osteogenesis and ameliorate bone healing in OVX rats.
To explore whether OVX-ASCs favored adipogenic differentiation and the role of miR-214 overexpression, OVX-ASCs (Mock and 214 S groups) and Sham-ASCs (Sham group) were cultured in osteoinduction or adipoinduction medium and analyzed by qRT-PCR. In comparison with the Sham group, the Mock group expressed lower (p < 0.05) levels of osteogenic (runx2 and ocn) genes upon osteoinduction ( Fig. 1b) but higher (p < 0.05) levels of adipogenic (ppar-γ and c/ebp-α) genes upon adipoinduction (Fig. 1c). Alizarin red and AdipoRed staining performed at day 14 ( Fig. 1d) further revealed that Mock group exhibited poorer mineralization (upper panel) and more triglycerides accumulation (lower panel) than Sham group. Conversely, 214 S group significantly (p < 0.05) elevated the runx2 and ocn expression (Fig. 1b), attenuated the ppar-γ and c/ebp-α expression (Fig. 1c), triggered more evident mineralization and dampened the accumulation of intracellular triglycerides at 14 dpt (Fig. 1d). These data collectively confirmed that OVX-ASCs aberrantly overexpressed miR-214 and favorably committed to adipogenic rather than osteogenic lineage, but alleviating miR-214 level switched the differentiation from adipogenic towards osteogenic lineage.
Sham-ASCs and OVX-ASCs were mock-transduced (Sham-Mock and OVX-Mock) or co-transduced with BacECre/Bac214S (Sham-214S and OVX-214S), followed by transfection with one of these 4 plasmids and measurement of Fluc and Gluc activities 3 days later. Transfection efficiency was calibrated by Gluc activity and the relative luciferase activities were obtained by normalizing Fluc/Gluc to those in the Sham-Mock or Sham-214S groups.
Furthermore, we analyzed 10 bone-associated cytokines in the supernatant at day 15 using a protein array. The qualitative (Fig. 4b) and quantitative (Fig. 4c) array data revealed that 214 S group secreted significantly higher levels of BMP7 and osteoprotegerin (OPG) and lower levels of osteopontin (OPN) than the Mock group.
The μCT imaging at 2W and 5W illustrated poor bone healing and shattered bone structure near the implantation site in the Mock and LEBW groups (Fig. 5a), indicating that implanting OVX-ASCs and even OVX-ASCs expressing potent osteogenic factor BMP2 failed to heal the defects. Intriguingly, the cortical bone in the Mock and LEBW groups at 5W were also compromised and the bone mineral densities (BMD) were evidently lower than those of the intact cortical bone. Conversely, both the 214 S and LEBW/214S groups completely filled the defects with osseous tissues at 5W (Fig. 5a). Of note, at 5W the LEBW/214S group not only increased the BMD at the defect (transverse view, Fig. 5a), but also elevated the cortical bone BMD along the shaft (sagittal view,   5a). Quantitative μCT analysis (Fig. 5b) attested that the BMD from defect center to the mid-shaft in the LEBW/214S group significantly (p < 0.05) exceeded those in the Mock and LEBW groups, and was superior to those in the 214S group and non-operated bone although the difference was statistically insignificant.
The front and rear views of 3D images (Fig. 6a) further confirmed that the Mock and LEBW groups failed to completely fill the defect at 2W and 5W, whereas the 214S and LEBW/214S groups filled the entire defects in both exterior and interior sides at 5W. Of note, the LEBW/214S group also improved trabecular bone-like structure formation within the metaphysis at 5W. After μCT scanning at 5W, the left femora were removed for analyses. The Mock and LEBW groups were filled with fibrous tissues within the defect areas as judged from the H&E staining (Fig. 6b). The 214S group improved the bone formation while the LEBW/214S group yielded even better new bone formation with abundant matrix and more compact bone structure, which histologically resembled the non-operated bone (Fig. 6b).
Furthermore, we calculated bone formation and microarchitecture parameters in the operated region using the μCT data. Compared with the Mock and LEBW groups, the 214S group gave rise to higher bone volume to total volume ratio (BV/TV, Fig. 7a), BMD (Fig. 7b), trabecular thickness (Tb. Th, Fig. 7c) and trabecular number (Tb.N, Fig. 7d) as well as lower distance between trabeculae (Tb.Sp, Fig. 7e) at 5W, although the difference was not statistically significant in all parameters. Conversely, at 5W the LEBW/214S group conferred significantly

Discussion
The overriding objective of this study was exploiting osteoporotic ASCs to heal bone defects in osteoporotic rats. Using OVX rats as the osteoporotic animal model, we found that OVX-ASCs highly expressed miR-214 and exhibited impaired osteogenesis capability but favored adipogenic differentiation (Fig. 1). Such finding agreed with the observations in osteoporotic BMSCs whose lineage commitment favorably shift to adipogenesis [23][24][25] . Nonetheless, by using the hybrid BV persistently expressing miR-214 sponge (214S group), we were able to knockdown miR-214 in OVX-ASCs to a level similar to that in Sham-ASCs, hence reversed the adipogenic/osteogenic differentiation bias and enhanced OVX-ASCs osteogenesis (Fig. 1). miR-214 was recently found to target baculoviral IAP repeat-containing 7 in human osteoblasts 26 , Osx in C2C2 cells 27 and ATF4 (another TF important for osteogenic differentiation) in mouse OVX-BMSCs 28 . Here we unraveled two new miR-214 targets: CTNNB1 (β-catenin) and TAB2 (Fig. 2). β-catenin and TAB2 are key mediators in the Wnt pathway 21,29 which stimulates Runx2 expression and directs mesenchymal precursor commitment to osteogenic lineage 30,31 while simultaneously represses chondrogenesis 32 and adipogenesis 33,34 . In the canonical Wnt pathway, β-catenin promotes BMSCs osteogenesis through its interaction with Runx2 to augment bone formation 35,36 while prevents BMSCs adipogenesis by downregulating PPAR-γ and C/EBP-α 7,37 . In the noncanonical Wnt pathway, TAB2 is a scaffold protein responsible for transmitting the signal 37 . Consequently, miR-214 overexpression in OVX-ASCs suppressed β-catenin and TAB2, resulting in the blockade of Wnt signaling and impaired osteogenesis potential. However, suppressing miR-214 by the BV-expressed miR-214 sponge reprogrammed the differentiation preference. These data shed light on how miR-214 functions as a "molecular switch" controlling the OVX-ASCs differentiation, and provided new insights into why repressing miR-214 levels can shift the OVX-ASCs differentiation from adipogenic to osteogenic.
Additionally, we uncovered that knocking down miR-214 levels in OVX-ASCs (214S group) stimulated the osteogenesis of co-cultured OVX-BMSCs via a paracrine manner (Fig. 3), which was accompanied by reduced release of exosomal miR-214 and OPN as well as increased secretion of BMP7 and OPG (Fig. 4). Very recently it was reported that exosomal miR-214 secreted by osteoclast can be transferred to osteoblasts to inhibit osteoblastic activity 38 . Conversely, BMP7 can promote osteogenic differentiation of osteoporotic BMSCs 25 . The alleviated exosomal miR-214-triggered inhibition and enhanced BMP7-mediated stimulation contributed to the enhanced osteogenesis of co-cultured OVX-BMSCs. The OPN and OPG might be implicated in the in vivo repair (see below).
Owing to the compromised osteogenesis capability of OVX-ASCs, implantation of the mock-transduced OVX-ASCs (Mock group) failed to heal the bone defects in the OVX rats (Figs 5-7). Even engineered OVX-ASCs persistently expressing the potent osteoinductive BMP-2 (LEBW group) failed to heal the defects (Figs 5-7), which underlined the difficulty to heal the bone defects in osteoporotic patients/animals using autologous cells. This also suggested that in OVX-ASCs either BMP2 is insufficient to activate the downstream BMP and MAPK/ ERK pathways, or these pathways require additional cues to orchestrate osteogenic differentiation. The latter conjecture is supported by our recent finding that implantation of rat ASCs co-expressing BMP2 and SDF1 (which potentiates MAPK/ERK pathway and stabilizes Runx2) repairs difficult-to-heal calvarial bone defects more efficiently than ASCs expressing BMP2 alone 39 .
Despite the difficulty to repair osteoporotic bone defects, implanting OVX-ASCs whose intracellular miR-214 level was knocked down by BV-expressed miR-214 sponge (214S group) remarkably repaired the bone defects in 5 weeks (Figs 5-7). Such striking healing may be ascribed to several factors. First, suppressing miR-214 reprogramed the propensity of OVX-ASCs differentiation from adipogenic to osteogenic by reviving the Wnt signaling mediator β-actin/TAB2 (Figs 1 and 2). Second, implantation surgery itself provokes host BMSCs infiltration to the injury site 40 and the implanted OVX-ASCs stimulated the osteogenesis of host OVX-BMSCs (Fig. 3) by secreting osteoinductive BMP7 (Fig. 4). Third, exosomal miR-214 undermines in vivo bone formation 38 . Suppressing intracellular miR-214 in OVX-ASCs concomitantly decreased the exosomal miR-214 secretion, thereby alleviating the inhibitory effects and promoting bone repair. Fourth, OPG is an antagonist of RANKL that inhibits osteoclast maturation. OVX-ASCs transduced with BacECre/Bac214S secreted more OPG (Fig. 4), which can repress the maturation of osteoclast that impedes bone formation early in the regeneration process. In the later stage of repair process, the implanted cells should have been eradicated 41 , so that at 5W no OPG was secreted from the implanted cells to inhibit osteoclast.
By co-expressing BMP2 and miR214 sponge in the OVX-ASCs, the LEBW/214S group further substantiated the bone regeneration, as evidenced by the complete defect filling, improved cortical bone structure (Figs 4 and 5), highest BV/TV, BMD, Tb.Th, Tb.n and lowest Tb.Sp among all groups (Fig. 7). These data demonstrated that co-expressing BMP2/miR-214 sponge in OVX-ASCs exerted synergistic bone healing effects when compared with expressing miR-214 sponge alone. Since β-catenin pathway enhances mesenchymal cell responsiveness to BMP2 35 and β-catenin requires interactions with such stimulatory signals as BMP2 to induce osteogenesis of stem cells 36 , it is likely that BMP2 not only coordinated with BMP7 to trigger BMP and MAPK/ERK signaling cascades but also potentiated the Wnt/β-catenin pathway, which acted in concert to provoke the osteogenesis and suppress adipogenesis of OVX-ASCs, thereby further ameliorating the bone repair. Furthermore, miR-214 and miR-148b antagonize the effects of each other in cancer cells 42 . miR-148b can effectively enhance ASCs osteogenesis 43 by targeting noggin 44 , a BMP2 antagonist that negatively regulates BMP2-induced osteoblast differentiation and bone formation 3 . It is possible that suppressing miR-214 concurrently elevates miR-148b and decreases noggin expression, thus enhancing bone repair by alleviating the negative regulation. It should also be noted that implantation surgery induces host immune responses that trigger death of implanted cells. We previously found that implantation of rabbit ASCs into the femoral defects in rabbits healed the critical-size defects but the implanted cells were eradicated in 4 weeks 40 . Since the cells were implanted into immunocompetent rats, it was likely that the cells were eradicated after a period of time. However, the fate of implemented cells awaits further investigation.
In light of our data, we propose a model (Fig. 8) that OVX-ASCs express aberrantly high level of miR-214, which targets TAB2 and CTNNB1 and hence blocks the Wnt pathway, leading to the overexpression of PPAR-γ and C/EBP-α and preferential differentiation into adipocytes. miR-214 also targets osterix 27 and ATF4 28 that dictate the differentiation into mature osteoblasts and to osteocytes, respectively, thereby further repressing osteogenesis. High miR-214 levels in OVX-ASCs also increases exosomal miR-214/OPN secretion but decreases BMP7/OPG secretion, which concurs with compromised osteogenesis of co-cultured OVX-BMSC. Consequently, suppressing miR-214 activates Wnt pathway, which not only blocks PPAR-γ/CEBP-α to mitigate adipogenesis but also enhances Runx2 levels, de-represses osterix and ATF4, hence additively promoting the osteogenic differentiation of OVX-ASCs. Meanwhile, knocking down miR-214 levels in OVX-ASCs enhances BMP7/OPG secretion and lowers exosomal miR-214/OPN secretion, which simultaneously enhances the osteogenesis of host OVX-BMSCs and attenuates the host osteoclast activity. These mechanisms converge to improve the commitment of miR-214 sponge-expressing OVX-ASCs towards osteogenic lineage and augment bone healing in OVX rats via autocrine and paracrine effects. BMP2/miR-214 sponge co-expression further synergizes the OVX-ASCs osteogenesis and bone healing in OVX rats. In conclusion, this study implicates the potential of engineering osteoporotic ASCs by repressing miR-214 as a means to treat osteoporotic fractures.

Methods
Osteoporotic rat model, ASCs and BMSCs isolation and expansion. All animal experiments were performed in compliance with the Guide for the Care and Use of Laboratory Animals (Ministry of Science and Technology, Taiwan) and experimental protocols were approved by the Institutional Animal Care and Use Committee of National Tsing Hua University. To generate osteoporotic rat models, Sprague-Dawley female rats (8-weeks old, BioLASCO, Taiwan) were subjected to bilateral ovariectomy (OVX) or sham operation (Sham) as described 12 . After ovariectomy, the rats were injected with methylpredinisolone hemisuccinate (1 mg/kg body weight/day, Sigma) every day for 4 weeks. Osteoporosis induced by this method was confirmed previously 12 .
ASCs were harvested subcutaneously from the inguinal fat pads of OVX rats (OVX-ASCs) or Sham rats (Sham-ASCs) and isolated the same way as porcine ASCs isolation 45 . BMSCs were isolated from the bone marrow of hind limb of OVX rats (OVX-BMSCs) or Sham rats (Sham-BMSCs) as described 12 . All isolated cells were cultured in DMEM medium containing 10% fetal bovine serum (FBS, Hyclone), 100 IU/ml penicillin and 100 IU/ml streptomycin, incubated at 37 °C (5% CO 2 ) and were passaged 3-5 times for experiments.
Quantitative real-time reverse-transcription PCR (qRT-PCR) and miR-214 analysis. Total cellular RNA was isolated using the NucleoSpin RNA II kit (Machereye-Nagel) and reverse transcribed to cDNA using the Omniscript RT Kit (Qiagen). The osteogenic and adipogenic genes were analyzed using StepOnePlus Real-Time PCR Systems (Applied Biosystems) and gene-specific primers (Table S1). Mature miRNAs in the cells were isolated using Trizol (Invitrogen) while miR-214 in the exosomes were isolated using miRCURY ™ Exosome Isolation Kit (Exiqon), followed by miRNA extraction using Trizol. The cellular and exosomal miR-214 levels were analyzed using the TaqMan MicroRNA Assays kit (Applied Biosystems). The gene expression levels were normalized against that of U6 (for miRNAs) or gapdh (for osteogenic and adipogenic genes) and referenced to those of selected cells (see Results).
Alizarin red staining and AdipoRed TM staining. Cells were stained by Alizarin red to assess mineralization as described 12 . To assess adipogenesis, cells were washed with PBS twice, stained 15 min using AdipoRed TM (Lonza) which stains intracellular lipid droplets, and observed under a confocal microscope (TE2000-E, Nikon). The stained triglyceride emitted green fluorescence at 572 nm.
For luciferase reporter assays, the 4 plasmids (2.5 μg) were separately transfected into cells using Lipofectamine 3000 (Invitrogen) and cells were collected 2 days later. The Fluc and Gluc activities in the lysates were measured with the Pierce ™ Gaussia-Firefly Luciferase Dual Assay Kit (Thermo Fisher) and read using a liquid scintillation microplate reader (Spectra Max M2, Molecular Devices). Transfection efficiency was calibrated by Gluc activity and the relative luciferase activities were obtained by normalizing the ratio of Fluc to Gluc (Fluc/Gluc) to those in reference cells (see Results).