A NOTCH1/LSD1/BMP2 co-regulatory network mediated by miR-137 negatively regulates osteogenesis of human adipose-derived stem cells

MicroRNAs have been recognized as critical regulators for the osteoblastic lineage differentiation of human adipose-derived stem cells (hASCs). Previously, we have displayed that silencing of miR-137 enhances the osteoblastic differentiation potential of hASCs partly through the coordination of lysine-specific histone demethylase 1 (LSD1), bone morphogenetic protein 2 (BMP2), and mothers against decapentaplegic homolog 4 (SMAD4). However, still numerous molecules involved in the osteogenic regulation of miR-137 remain unknown. This study aimed to further elucidate the epigenetic mechanisms of miR-137 on the osteogenic differentiation of hASCs. Dual-luciferase reporter assay was performed to validate the binding to the 3′ untranslated region (3′ UTR) of NOTCH1 by miR-137. To further identify the role of NOTCH1 in miR-137-modulated osteogenesis, tangeretin (an inhibitor of NOTCH1) was applied to treat hASCs which were transfected with miR-137 knockdown lentiviruses, then together with negative control (NC), miR-137 overexpression and miR-137 knockdown groups, the osteogenic capacity and possible downstream signals were examined. Interrelationships between signaling pathways of NOTCH1-hairy and enhancer of split 1 (HES1), LSD1 and BMP2-SMADs were thoroughly investigated with separate knockdown of NOTCH1, LSD1, BMP2, and HES1. We confirmed that miR-137 directly targeted the 3′ UTR of NOTCH1 while positively regulated HES1. Tangeretin reversed the effects of miR-137 knockdown on osteogenic promotion and downstream genes expression. After knocking down NOTCH1 or BMP2 individually, we found that these two signals formed a positive feedback loop as well as activated LSD1 and HES1. In addition, LSD1 knockdown induced NOTCH1 expression while suppressed HES1. Collectively, we proposed a NOTCH1/LSD1/BMP2 co-regulatory signaling network to elucidate the modulation of miR-137 on the osteoblastic differentiation of hASCs, thus providing mechanism-based rationale for miRNA-targeted therapy of bone defect.

NOTCH signal is a fundamental pathway in bone remodeling and skeletal homeostasis [33][34][35]. Hairy and enhancer of split 1 (HES1), a downstream gene of NOTCH [36], is responsible for the actions of NOTCH in the skeleton, even though its osteogenic effects are cell type-specific and context-dependent. By interacting with runt-related transcription factor 2 (RUNX2), HES1 induces osteopontin promoter [37]. But HES1 also binds to the osteocalcin (OCN) promoter and suppresses its transcription in osteoblastic cells [38]. HES1 inactivation not only increases the femoral length and trabecular number in the limb bud of transgenic mice, but also enhances mineral apposition rate and suppresses bone resorption [39]. NOTCH1 has emerged as a target of miR-137 in human renal mesangial cells [40], retinal ganglion cells [41], neurons [42], non-small cell lung cancer cells [43], and breast cancer cells [44], but whether it is directly inhibited by miR-137 has not yet been identified in hASCs. In small cell lung cancer cells, NOTCH1 pathway is activated by LSD1 inhibitor and suppressed due to the binding of LSD1 [45]. Additionally, the induction of NOTCH signal impairs the activation of BMP pathway and the osteoblastic differentiation of dental follicle cells [46]. In contrast, NOTCH1 upregulates BMP2 expression in human aortic valve interstitial cells through the stimulation of NF-κB [47]. Our previous study confirmed that miR-137 knockdown induced BMP2-SMAD4 pathway through the downregulation of LSD1 dependently or independently [32], which coincides with the studies stating that LSD1 inhibition leads to increased BMP2 expression [48,49]. Accordingly, we postulate a signaling network entailing NOTCH1-HES1, LSD1, and BMP2-SMADs pathways to unveil the miR-137 modulation on the osteogenesis of hASCs.
This study identified the interactions of miR-137 and its downstream genes and revealed that the coregulatory signaling network of NOTCH1/LSD1/BMP2 mediated by miR-137 negatively modulated the osteogenesis of hASCs, suggesting that miR-137 might be applied as a promising therapeutic target for bone regeneration.

Mice
The animal experiments were conducted in strict conformity with the guidelines of Animal Welfare Committee of Health Science Center in Peking University (LA2019019). Male, 5-week-old BALB/c-nu/nu nude mice (Charles River, Wilmington, MA, USA) were randomly assigned to 3 groups (n = 6 per group) and maintained with specific pathogen-free conditions.

Cell lines
The hASCs isolated from three separate donors were purchased in ScienCell Research Laboratories (Carlsbad, CA, USA). For each donor, the in vitro cell experiments were performed at least three times individually. For proliferation culture, cells were maintained in proliferation medium (PM), containing Dulbecco's modified Eagle's medium (Thermo Fisher Scientific, Rockford, IL, USA), 1% (v/v) penicillin/streptomycin (Thermo Fisher Scientific) and 10% (v/v) fetal bovine serum (ExCell Bio, Shanghai, China). When the cells reached 70-80% confluence, osteoinduction was performed by adding osteogenic medium (OM), which contained the above culture medium for promoting proliferation, 100 nM dexamethasone (Sigma-Aldrich), 0.2 mM L-ascorbic acid (Sigma-Aldrich), and 10 mM β-glycerophosphate (Sigma-Aldrich). The cell culture conditions were 37°C with 5% CO 2 and 100% relative humidity.

Alkaline phosphatase (ALP) staining and quantification
After 7 days of culture in PM or OM, the hASCs were used for ALP staining and activity test according to the published protocol [10]. ALP staining was operated following the BCIP/NBT staining kit (Beyotime, Shanghai, China) instructions. For the quantitative tests of ALP activities, cells were washed with phosphate buffer saline (PBS) and 1% Triton X-100 (Solarbio, Beijing, China), then scraped in milli-Q water and subjected to three cycles of freezing and thawing. By employing the BCA method and the pierce BCA protein assay kit (Thermo Fisher Scientific), total protein was read at 562 nm and computed with a bovine serum albumin standard curve according to the manufacturer's protocol. Afterwards, ALP activity was detected at 520 nm applying an alkaline phosphatase assay kit (Jiancheng, Nanjing, Jiangsu, China) and finally normalized to the total protein concentrations of cells.

Alizarin red S (ARS) staining and quantification
After 14 days of culture in PM or OM, the hASCs were applied to detect the matrix mineralization. Following being fixed with 95% ethanol for 30 min, cells were soaked in 1% ARS staining solution (pH 4.2; Sigma-Aldrich, St. Louis, MO, USA) for 20 min at room temperature. To assess the degree of mineralization, stained areas of each well were separately dissolved in 100 mM cetylpyridinium chloride (Sigma-Aldrich) for 1 h and the absorbances were detected at 562 nm. Finally, the relative ARS intensity was normalized to the total protein concentrations of cells.
RNA extraction, reverse transcription, and quantitative real-time polymerase chain reaction (qRT-PCR) After 3, 7, and 14 days of culture in PM or OM respectively, total RNA of cells was isolated with TRIzol (Invitrogen, Carlsbad, CA, USA) and synthesized into the first-strand cDNA using a reverse transcription system (Takara, Tokyo, Japan). All the transcripts were quantified using the FastStart universal SYBR green master (ROX) (Roche, Indianapolis, IN, USA) and a 7500 realtime PCR detection system (Applied Biosystems, Foster City, CA, USA). Relative expression levels of mRNA and miRNA were normalized to GAPDH mRNA and U6 snRNA, respectively. The sequences of the primers employed are listed in Additional file 6: Table S2.

Dual-luciferase reporter assay
The 3′ untranslated region (3′ UTR) alignments of the target regions in NOTCH1 were predicted by TargetScan and RNA22. Reporter vectors were constructed based on the previous method [32]. The 3′ UTR sequences of NOTCH1, which contained the possible binding sites of miR-137, were PCR amplified and then inserted into pEZX-MT06 vectors (GeneCopoeia, Rockville, MD, USA) to create NOTCH1-WT (wild-type NOTCH1) luciferase reporter plasmids. Mutated forms were generated by site-directed mutagenesis (GeneCopoeia) and named NOTCH1-MT (mutant-type NOTCH1) luciferase reporter plasmids. For luciferase assay, the hASCs were planted on 24-well culture plates with a density of 5 × 10 4 /well and co-transfected with 1 μg NOTCH1-WT or NOTCH1-MT plasmids, 100 nM NC or miR-137 mimics, and lipofectamine 3000 (Invitrogen). The luciferase activities were examined by a dual-luciferase reporter assay system (Promega, Madison, WI, USA) 48 h later, and standardized to renilla luciferase activity for each transfected well.

Heterotopic osteogenesis examinations in vivo
After the transfection with NC, miR-137 and anti-miR-137 lentiviruses, hASCs of the third passage were maintained in PM for 1 week, collected, and incubated with auto-setting calcium phosphate cement (ACPC; Rebone, Shanghai, China) for 1 h at 37°C. Then the hASC-ACPC mixtures were transplanted subcutaneously to the dorsal regions of nude mice (n = 6 per group) for the analyses of heterotopic bone formation in vivo 8 weeks later. After being collected and fixed in 4% paraformaldehyde, the samples were photographed with soft X-ray. The radiograph was obtained by applying a Senograph 2000D molybdenum-rhodium twin target X-ray apparatus (GE, Fairfield, CT, USA). The radiation distance is 200 mm and the radiographing conditions were 22.0 kV, 35.0 mA. For histological evaluation, the specimens were decalcified in 10% ethylene diamine tetraacetic acid solution (pH 7.4) for 14 days, embedded into paraffin, then sliced into 5 μm-thick sections before the subsequent hematoxylin and eosin (HE) staining and Masson trichrome staining. The rabbit anti-OCN primary antibodies diluted to 1:100 (Servicebio, Wuhan, Hubei, China; GB11233) were used for immunohistochemical (IHC) staining.

Statistical analysis
Data and statistical analyses were conducted with SPSS Statistics 20.0 software (IBM, Armonk, NY, USA). All data were shown as mean ± standard deviation (SD) of three individual experiments. For the comparison of two independent or multiple groups, Mann-Whitney U test or Kruskal-Wallis test were applied, respectively. A twotailed test with p value < 0.05 was indicated as statistically significant.

MiR-137 reversely regulates hASC differentiation along osteoblastic lineage in vitro
Our previous study displayed an overall downward expression trend of miR-137 in hASCs during the osteoblastic induction and identified its negative role in this biological process [32]. Since lacking other studies on the osteogenic function of miR-137, we re-verified the reliability of our previous results. After transfecting hASCs with lentiviruses of NC, miR-137 overexpression, and miR-137 knockdown (Additional file 1: Fig. S1a), we evaluated the transfection rate was over 90% by computing the percentage of GFP-tagged cells (Additional file 1: Fig. S1b). Meanwhile, the transfection effects were quantitatively determined on 3 days, 7 days, and 14 days by qRT-PCR analysis (Additional file 1: Fig. S1c).
ALP staining and activity assays displayed that miR-137 overexpression reduced ALP activity of hASCs under proliferation condition or osteogenic induction, but miR-137 knockdown reversed the effects observed with miR-137 overexpression (Additional file 2: Fig. S2a,  b). ARS staining and quantification were applied to test the calcium deposits of extracellular matrix. More mineralized nodules were presented in miR-137 knockdown group while less in miR-137 overexpression group when both were compared with NC group (Additional file 2: Fig. S2c, d). Besides, the characteristic genes expressed in different stages of osteogenesis, including RUNX2, ALP, and OCN, were examined by qRT-PCR and presented significant decreases in miR-137 overexpression group but dramatically increased in miR-137 knockdown group (Additional file 2: Fig. S2e). According to these data, we substantiated that miR-137 inhibits in vitro osteoblastic activity of hASCs.

MiR-137 reversely regulates hASC differentiation along osteoblastic lineage in vivo
In order to validate in vivo osteogenic effects of miR-137, hASCs were transfected with NC, miR-137, and anti-miR-137 lentiviruses and separately mixed with ACPC, and then the compounds were subcutaneously implanted into the dorsum of nude mice (Additional file 3: Fig. S3a). After 8 weeks, the total volume and mean density of the harvested samples were assessed and manifested an apparent enhancement in miR-137 knockdown group but remarkable reduction in miR-137 overexpression group (Additional file 3: Fig. S3b, c).
Histological analyses of bone formation were performed by staining of HE, Masson trichrome, and IHC staining for OCN. HE staining showed more new bone formation in miR-137 knockdown group when comparing with NC group, which displayed only a very small amount of osteoid, but we could hardly observe any new bone or osteoid in miR-137 overexpression group. Similarly, thicker and more compact blue-green-stained collagen fiber bundles were detected in miR-137 knockdown group than in another two groups, but overexpression of miR-137 led to the thinnest collagen deposition. Moreover, we found that dark-brown stained OCN granules were the most widespread in the cells of miR-137 knockdown group, fewer in NC group, and none could be discerned in miR-137 overexpression group (Additional file 3: Fig. S3d). Consequently, in accordance with the results of in vitro experiments, miR-137 subdues the osteoblastic activity of hASCs in vivo.
MiR-137 regulates NOTCH1-HES1 pathway by directly targeting NOTCH1 To ascertain the influences of miR-137 on NOTCH1 pathway, we first examined the expression of NOTCH1 and its downstream signal HES1 with miR-137 overexpression or knockdown. When compared with NC group, the mRNA and protein levels of NOTCH1 showed obvious increase in miR-137 knockdown group while marked reduction in miR-137 overexpression group. Contrary to NOTCH1, the expression tendency of HES1 accorded with the changes of miR-137 (Fig. 1a-c). These findings suggested that NOTCH1 is negatively regulated while HES1 is positively regulated by miR-137.
To further identify whether miR-137 could directly bind to NOTCH1 in hASCs as it does in other cell lines [40][41][42][43][44], dual-luciferase reporter assays were carried out. The presumed targeting sites of miR-137 in the 3′ UTR of NOTCH1 were forecasted by two prediction softwares (TargetScan and RNA22). Then the luciferase reporter vectors carrying the 3′ UTR of NOTCH1-WT or NOTCH1-MT were constructed (Fig. 1d, e) and the Data are shown as mean ± SD of three independent experiments performed in triplicate. *p < 0.05, **p < 0.01, ***p < 0.001 versus respective NC group relative luciferase activities were detected. MiR-137 mimics significantly repressed the luciferase activity in NOTCH1-WT group while they had no significant influences in NOTCH1-MT group when both groups were compared with their respective NC group (Fig. 1f). These results validated that miR-137 directly binds to the 3′ UTR of NOTCH1 and induces the expression of HES1 in hASCs.

NOTCH1 inhibitor reverses the effects of miR-137 knockdown on osteogenesis and downstream genes expression
To confirm whether miR-137 regulated osteogenesis through NOTCH1, we employed NOTCH1 inhibitor (tangeretin) in hASCs transfected with miR-137 knockdown lentiviruses. In vitro osteogenic stainings and quantification manifested that when compared with NC  = 100 μm). c, d ARS staining (c) and quantification (d) of transfected hASCs after a 14-day culture in PM or OM. e Relative expression analyses of RUNX2, ALP, and OCN by qRT-PCR in transfected hASCs on 3 days, 7 days, and 14 days. f Relative expression analyses of NOTCH1, HES1, LSD1, BMP2, and SMAD4 by qRT-PCR in transfected hASCs on 3 days, 7 days, and 14 days. g, h Western blotting (g) and band intensity analyses (h) of NOTCH1, HES1, LSD1, BMP2, SMAD4, and p-SMAD1/5 in transfected hASCs on 7 days. Data are shown as mean ± SD of three independent experiments performed in triplicate. *p < 0.05, **p < 0.01, ***p < 0.001 versus respective NC group group, miR-137 knockdown evidently strengthened ALP activity and extracellular matrix calcification, whereas the promoted osteogenic ability caused by miR-137 knockdown were completely reversed by tangeretin treatment, and miR-137 overexpression group displayed impaired osteogenic differentiation potential as previously mentioned (Fig. 2a-d). In accord with the results of osteogenic stainings, qRT-PCR detection showed that the expression of RUNX2, ALP, and OCN were enhanced in miR-137 knockdown group while reduced in miR-137 overexpression group, but the addition of tangeretin abrogated the induction of these genes by miR-137 knockdown and even suppressed their expression (Fig. 2e).
To further authenticate whether NOTCH1 mediated the mechanisms of the osteogenic regulation by miR- Fig. 3 NOTCH1 knockdown impairs osteogenesis by inducing HES1. a, b ALP staining (a) and quantification (b) of transfected hASCs after a 7-day culture in PM or OM (scale bar = 100 μm). c, d ARS staining (c) and quantification (d) of transfected hASCs after a 14-day culture in PM or OM. e Relative expression analyses of NOTCH1, HES1, and RUNX2 by qRT-PCR in transfected hASCs on 3 days, 7 days, and 14 days. f, g Western blotting (f) and band intensity analyses (g) of NOTCH1, HES1, and RUNX2 in transfected hASCs on 7 days. Data are shown as mean ± SD of three independent experiments performed in triplicate. *p < 0.05, **p < 0.01, ***p < 0.001 versus respective NC group 137, we examined the expression of miR-137 downstream signals in the above four groups. Both in miR-137 overexpression and tangeretin-treated groups, NOTCH1 was significantly repressed, though miR-137 knockdown induced its expression. The mRNA and protein expression showed that HES1 and LSD1 changed synchronously with the alterations of miR-137, but tangeretin-treated group inverted the downregulation of these two genes by miR-137 knockdown when compared with NC group. In addition, the mRNA or protein expression levels of BMP2, SMAD4, and p-SMAD1/5 showed that these genes were induced by miR-137 knockdown while inhibited by miR-137 overexpression. Similarly, tangeretin inhibited the activation of BMP2-SMADs pathway initiated by miR-137 knockdown (Fig.  2f-h). Combined with the above results, we affirmed that NOTCH1 inhibitor reverses the osteogenic modulation effects of miR-137 knockdown.

NOTCH1 knockdown impairs osteogenesis by inducing HES1
To determine the influences of NOTCH1 knockdown on the osteoblastic potential of hASCs, we applied ALP and ARS stainings combined with quantitative analysis and found that NOTCH1 knockdown attenuated ALP activity and extracellular mineralization (Fig. 3a-d). As the downstream molecules of NOTCH1, HES1, and RUNX2 were further detected at  = 100 μm). c, d ARS staining (c) and quantification (d) of transfected hASCs after a 14-day culture in PM or OM. e Relative expression analyses of HES1, RUNX2, ALP, and OCN by qRT-PCR in transfected hASCs on 3 days, 7 days, and 14 days. Data are shown as mean ± SD of three independent experiments performed in triplicate. *p < 0.05, **p < 0.01, ***p < 0.001 versus respective NC group mRNA and protein levels in hASCs transfected with NOTCH1 knockdown lentiviruses. Coincident with the impacts of miR-137 on NOTCH1-HES1 pathway, NOTCH1 knockdown induced the expression of HES1 while repressed RUNX2 (Fig. 3e-g). Our results indicated that NOTCH1 knockdown impedes the osteogenic potential of hASCs by the stimulation of HES1, corroborating the former conclusions that miR-137 inhibits osteogenesis by the downregulation of NOTCH1 and upregulation of HES1.
Then we checked the impacts of HES1 on the osteogenic differentiation of hASCs. After knocking down HES1, we examined the osteogenic ability by the application of osteogenic stainings and quantification as well as qRT-PCR detection of RUNX2, ALP, and OCN. All these in vitro osteogenic tests showed promoted osteogenic potential of hASCs with HES1 knockdown (Fig. 4). Therefore, we demonstrated that HES1 plays a negative role in the osteoblastic differentiation of hASCs and intervenes in NOTCH1induced osteogenesis.

NOTCH1 knockdown induces LSD1 while inhibits BMP2-SMADs pathway
Since our previous study has demonstrated that miR-137 upregulated LSD1 while downregulated BMP2 and SMAD4 in hASCs [32], we further investigated the influences of NOTCH1 knockdown on LSD1 and BMP2-SMADs pathway. As predicted, the expression of LSD1 significantly increased after knocking down NOTCH1 while BMP2, SMAD4, and p-SMAD1/5 decreased apparently (Fig. 5). Therefore, we deduced that the osteogenic inhibition of NOTCH1 knockdown is also dependent on the activation of LSD1 and suppression of BMP2-SMADs pathway.
LSD1 knockdown regulates NOTCH1-HES1 pathway As above, we have affirmed that NOTCH1 acted as a negative regulator in LSD1 expression. But considering the  complex interplay of signaling molecules, we tried to clarify whether LSD1 had feedback effects on NOTCH1-HES1 pathway. Notably, we found that LSD1 knockdown led to a higher level of NOTCH1 while lowered the expression of HES1 when compared with NC group (Fig. 6a-c), thus prompting a reciprocal negative relationship between NOTCH1 and LSD1.
BMP2 knockdown inhibits HOTCH1 while induces HES1 and LSD1 To gain further insights into the relationships between NOTCH1, LSD1, and BMP2 signals, we then examined the expression of NOTCH1, HES1, and LSD1 at mRNA and protein levels after knocking down BMP2. Importantly, the expression of NOTCH1 dramatically decreased while HES1 and LSD1 increased with BMP2 knockdown (Fig. 6d-f). Combining with the above results that NOTCH1 knockdown inhibited BMP2-SMADs pathway, we verified a positive feedback loop between NOTCH1 and BMP2. Furthermore, our previous study revealed that silencing of LSD1 promoted the osteoblastic potential of hASCs by stimulating BMP2-SMAD4 signaling pathway [32]. Here again, we observed upward tendencies in the expression of BMP2, SMAD4, RUNX2, and ALP with LSD1 knockdown (Additional file 4: Fig.  S4). Therefore, a negative interplay between LSD1 and BMP2 was also confirmed by us.

Discussion
Dysfunction of miR-137 has been implicated in several cancer types such as glioblastoma [19], melanoma [20], multiple myeloma [21], non-small cell lung cancer [22], and endometrial cancer [23]. In addition, miR-137 has been found to play an essential role in neural development and maturation, with several studies displaying an association with cell proliferation and neurogenic differentiation [24][25][26][27][28][29][30][31]. However, little is known concerning its functions and regulatory mechanisms on the osteoblastic differentiation, especially in mesenchymal stem cells. Silencing of miR-137-3p is found to facilitate the osteogenesis of bone marrow-derived mesenchymal stem cells by targeting RUNX2 [50]. Previously, we demonstrated that silencing of miR-137 promoted the osteoblastic activity in hASCs and revealed its modulation on the signaling network of LSD1/BMP2/SMAD4 as part of the mechanisms [32]. Interestingly, our former research confirmed a positive role of miR-137 in LSD1 expression, which was contrary to several studies reporting that miR-137 directly binds to LSD1 [23,26,51,52]. The contradicted outcomes might be associated with the various biological features of different cell types, and we deduced that there probably exist intermediary regulators working between miR-137 and LSD1 during the osteogenesis of hASCs. Notably, this study identified that NOTCH1 was a direct target of miR-137 in hASCs, and NOTCH1-HES1 pathway was engaged in the crosstalk between LSD1 and BMP2-SMADs pathway. In this way, NOTCH1 signal mediated the control of miR-137 on LSD1/BMP2/SMAD4 network. Moreover, the interrelations of the above signals were validated comprehensively and a NOTCH1/LSD1/BMP2 co-regulatory network was established, further elucidating the epigenetic mechanisms of miR-137 during the process of hASCs differentiating into osteoblastic lineage. After reconfirming the inhibitory impacts of miR-137 on the osteoblastic activity of hASCs both in vitro and in vivo, we demonstrated that miR-137 negatively regulated the expression of NOTCH1 while positively regulated HES1. NOTCH signaling pathway influences tumorigenesis as well as embryonic development [53] because of its crucial role in cell fate determination, proliferation, differentiation, and apoptosis [54]. Though it is still debatable whether NOTCH signal serves as a positive or negative regulator in osteogenesis [55], our data displayed impaired osteogenic capacity of hASCs after knocking down NOTCH1. Most noteworthy, NOTCH1 was validated as a direct target gene of miR-137 in hASCs, the same as in other cell lines [40][41][42][43][44]. HES1 is known as a potential downstream gene of NOTCH1 in many studies, but it is not affected in NOTCH1 knockout mice while the expression of HES5 and HES-related repressor protein (HERP)1, -2, and -3 are greatly diminished [56][57][58]. As a transcriptional regulator in the NOTCH signaling pathway, recombination signal binding protein (RBPJ) gene disruption in homozygous mice exhibits reduced HES5 expression, but not for HES1 [56]. Given the various effects on OCN, osteopontin, and RUNX2 [37,38], HES1 influences the osteogenesis inconsistently depending on the different cellular environments. This study exhibited enhanced osteogenic differentiation of hASCs with HES1 knockdown. Moreover, NOTCH1 knockdown upregulated HES1, indicating that NOTCH1 acts as a negative regulator in the expression of HES1. To further identify whether NOTCH1 mediated the osteogenic differentiation of hASCs modulated by miR-137, we applied NOTCH1 inhibitor (tangeretin) in hASCs transfected with miR-137 knockdown lentiviruses and found that tangeretin reversed the impacts of miR-137 knockdown on osteogenesis and downstream genes expression, thus verifying the vital role of NOTCH1 in the osteogenic regulation of miR-137. Collectively, we brought insight into how the NOTCH1-HES1 pathway was influenced and involved in the osteogenic modulation of miR-137.
LSD1 has been linked to the repression of NOTCH1 pathway in various cell types [45,[59][60][61][62][63], though one study states that it functions as a corepressor when associated with RBPJ-repressor complex and as a NOTCH1 coactivator upon NOTCH activation [64]. Nevertheless, few studies have reported the interplay between NOTCH1 and LSD1 during the osteogenesis of hASCs and whether this interaction contributes to the osteogenic control of miR-137 is still unknown. Coincident with the influences of miR-137 on NOTCH1 and LSD1, we uncovered a negative interaction between NOTCH1 and LSD1 with separate knockdown of them. More intriguingly, despite inducing the expression of NOTCH1, LSD1 knockdown repressed HES1. Thus, the opposite expression trends of NOTCH1 and HES1 caused by LSD1 knockdown might reinforce the downregulation of HES1 by NOTCH1 alone. These results authenticated the crosstalk between NOTCH1-HES1 pathway and LSD1, through which miR-137 regulates the osteogenic differentiation of hASCs.
BMP signal is a canonical pathway in skeleton and BMP2-SMAD4 pathway has been shown to participate in the osteogenic regulation of miR-137 by us [32]. After knocking down NOTCH1 or BMP2 individually, we observed suppressed expression of NOTCH1, BMP2, and SMADs simultaneously, indicating a positive interrelationship between NOTCH1 and BMP2-SMADs pathway. Besides, increased HES1 expression with BMP2 knockdown proved the negative effects of NOTCH1 on HES1 from another aspect. Considering our previous results that LSD1 knockdown activated the BMP2-SMAD4 pathway [32], in turn, we investigated the impacts of BMP2 on LSD1. As expected, LSD1 was significantly upregulated with BMP2 knockdown. Thus, these findings suggested a negative feedback loop between LSD1 and BMP2-SMADs pathway.
In conclusion, our results revealed that depending on the reciprocal negative regulation between NOTCH1 and LSD1, LSD1 and BMP2, as well as the synergistic function between NOTCH1 and BMP2, miR-137 negatively regulates osteogenesis of hASCs through the NOTCH1/ LSD1/BMP2 co-regulatory signaling network (Fig. 7).

Conclusions
In summary, our study provided a relatively comprehensive rationale for the negative modulation of miR-137 during the hASC differentiation towards osteoblastic lineage and established a co-regulatory network of NOTCH1/LSD1/BMP2 to elucidate the underlying mechanisms, which is of substantial importance for potential targeted therapy of bone-related diseases.