MiR-34a-5p promotes multi-chemoresistance of osteosarcoma through down-regulation of the DLL1 gene

MiR-34a-5p has been implicated in the tumorigenesis and progression of several types of cancer. However, the role of miR-34a-5p in osteosarcoma (OS) remains largely unknown. This study was performed in two multi-chemosensitive (G-292 and MG63.2) and two resistant (SJSA-1 and MNNG/HOS) OS cell lines. MiR-34a-5p promotes OS multi-chemoresistance via its repression of the Delta-like ligand 1 (DLL1) gene, the ligand of the Notch pathway, and thus negatively correlates with OS chemoresistance. The siRNA-mediated repression of the DLL1 gene suppressed cell apoptosis and de-sensitized G-292 and MG63.2 cells, while overexpression of DLL1 sensitized SJSA-1 and MNNG/HOS cells to drug-induced cell death. In agreement with the changes in the drug-induced cell death, the activity of the ATF2/ATF3/ATF4 signaling pathway was significantly altered by a forced reversal of miR-34a-5p or DLL1 levels in OS cells. DLL1 is a target of miR-34a-5p and negatively regulates the multi-chemoresistance of OS. This study suggested that miR-34a-5p, DLL1 and the ATF2/ATF3/ATF4 signaling pathway-associated genes are the potential diagnostic and/or therapeutic targets for an effective chemotherapy of OS. Our results also provide novel insights into the effective chemotherapy for OS patients.


DLL1 expression negatively correlates with the effect of miR-34a-5p on OS chemoresistance.
To explore the role of DLL1 in OS chemoresistance, we transfected G-292 cells with the miR-34a-5p mimic and the SJSA-1 cells with miR-34a-5p antagomiR and compared the cell death triggered by an IC 50 -dosed drug. The transfection of miR-34a-5p mimic in G-292 cells significantly increased drug resistance to some extent, with CDDP mostly to 1.66-fold, whereas the transfection of miR-34a-5p antagomiR in SJSA-1 cells decreased the chemoresistance for all four drugs (Fig. 3A). We further transfected si-DLL1 into G-292 cells to evaluate its effect on multi-chemoresistance. The transfection of si-DLL1 indeed decreased the level of DLL1 at both the mRNA (0.05:1.00) and protein level (0.46:1.00) compared with the corresponding control cells ( Fig. 3B and C). Following the decrease in the DLL1 level in G-292 cells upon transfection of miR-34a-5p mimic, the cell survival rate was increased byapproximately 5%-40% with the addition of the four drugs, excluding MTX (Fig. 3D). Furthermore, the transfection of si-DLL1 increased the cell survival rate with the addition of all four drugs (Fig. 3D). The results correlated well with the negative regulation of the multi-chemoresistance of OS cells. Conversely, the transfection of a GFP-tagged DLL1 expression construct raised the DLL1 protein level in SJSA-1 cells (Fig. 3E). The resultant chemoresistance was significantly decreased for all four drugs, excluding CDDP. As expected, the control cells with the transfection of only GFP protein showed a marginal effect on the chemoresistance of SJSA-1 (Fig. 3F).
Scientific RepoRts | 7:44218 | DOI: 10.1038/srep44218 Similar cases in which the chemoresistance was significantly decreased for all four drugs were also found with the transfection of miR-34a-5p antagomiR in SJSA-1 cells (Fig. 3F).
In agreement with its negative effect on chemoresistance ( Fig. 3D and F), a siRNA-mediated DLL1 repression decreased the apoptotic rate from 4.88% to 2.36%, and a similar effect was found in the miR-34a-5p-mimic-transfected G-292 cells (Fig. 4A,B and C). Taken together, both DLL1 and miR-34a-5p contribute substantially to the OS chemoresistance to Dox, Etop, MTX and CDDP.
The ATF2/ATF3/ATF4 signaling pathway might be involved in OS multi-chemoresistance. To gain further mechanistic insights into the roles of miR-34a-5p/DLL1 axis, we used the Cignal reporter finder assay to compare the activities of 17 signaling pathways in G-292 versus SJSA-1cells 27 . The following six cancer-related signaling pathways: p53/DNA damage, NF-кB, MAPK/ERK, ATF2/ATF3/ATF4, cAMP/PKA and MEF2 were selected, whose activities differed by more than two-fold. Among them, two pathways, p53/DNA damage and NF-кB, showed higher activities in SJSA-1 cells, whereas the other four pathways, MAPK/ERK, ATF2/ATF3/ ATF4, cAMP/PKA and MEF2, showed higher activities in G-292 cells (Fig. 5A). These differences might imply a role in OS chemoresistance. Among them, two pathways, p53/DNA damage and NF-кB, showed higher activities in SJSA-1 cells, whereas the other four pathways, MAPK/ERK, ATF2/ATF3/ATF4, cAMP/PKA and MEF2, showed higher activities in G-292 cells (Fig. 5A). The expression of the corresponding transcription factors correlated well with the differential activity of these signaling pathways (Fig. 5B). We then compared the activities of all these pathways in miR-34a-5p mimic-transfected G-292 cells or miR-34a-5p antagomiR-transfected SJSA-1 cells. The activity of the p53/DNA damage pathway was elevated in the miR-34a-5p mimic-transfected G-292 cells, whereas it was reduced in the miR-34a-5p antagomiR-transfected SJSA-1 cells (Fig. 5C-F). By contrast, the activity of the ATF2/ATF3/ATF4 pathway was reduced in the miR-34a-5p mimic-transfected G-292 cells but increased in the miR-34a-5p antagomiR-transfected SJSA-1 cells (Fig. 5C-F). The activities of the above two pathways negatively correlated with miR-34a-5p's promoting effect on OS chemoresistance, indicating their association with OS chemoresistance. We further compared the pathway activities in DLL1 siRNA versus the mock siRNA-transfected G-292 cells. Although a slight decrease on the activities of the DNA damage and NF-кB pathways was observed, the other three pathways, including ATF2/ATF3/ATF4, were significantly repressed by DLL1 siRNA, indicating the correlation of these three pathways with OS chemoresistance. Taken together, only the ATF2/ATF3/ATF4 pathway was validated to be involved in the OS chemoresistance mediated by miR-34a-5p.
The DLL1 expression negatively correlates with the level of miR-34a-5p in the tumor xenografts of nude mice. Recently, miR-34a-5p was shown to promote Dox chemoresistance of OS in tumor xenografts of nude mice via repressing its target gene CD117 25 . In the present study, we semi-quantified the levels of DLL1 protein in the same set of the tumor tissues in mice via immuno-histological analysis. The intratumoral injection of miR-34a-5p's agomiR into G-292 decreased DLL1 expression. By contrast, the injection of miR-34a-5p's antagomiR into SJSA-1 increased DLL1 expression in Dox-or PBS-treated mice (Fig. 6). The results further confirmed that miR-34a-5p has a positive effect on both growth and chemoresistance of OS cell-derived tumor xenografts in nude mice.

Discussion
As a well-studied miRNA, miR-34a has been studied in several types of cancer, including Ewing's sarcoma 28 and colorectal cancer 27 . Several direct targets have been found for miR-34a, including Notch, c-Myc, c-Met, and c-Kit 29 . MiR-34a targets Notch1 and Notch2 in glioblastoma and medulloblastoma 19 . MiR-34a suppresses invasion via repression of Notch1 and Jagged1 in cervical carcinoma and choriocarcinomacells 30 . Moreover, much effort has been exerted to reveal the roles of miR-34a in cancer chemoresistance [31][32][33] . In this study, we showed that miR-34a-5p is also involved in the multi-chemoresistance of OS. MiR-34a-5p is up-regulated in multi-chemoresistant (SJSA-1 and MNNG/HOS) compared with multi-chemosensitive (G-292 and MG63.2) OS cell lines (Figs 1 and 2). Bisulfite Sequencing PCR (BSP) assay showed that the average methylation ratio of the miR-34a-5p promoter is comparable between SJSA-1 and G-292 cells (Additional File 2), which indicates that there are other factors that differentiate the expression of miR-34a-5p in SJSA-1 and G-292 cells. We next performed an RNA-seq analysis of G-292, MG63.2 and SJSA-1 cell lines and found a group of genes that were differentially expressed, including the DLL1 gene, which negatively correlates with chemoresistance (Additional File 1B). Both the role and mechanism of the DLL1 gene in the context of OS chemoresistance were systematically investigated in cultured cells and tumor xenografts in nude mice.
The DLL1 gene encodes one of the ligands in the Notch signaling pathway, which is highly conserved to regulate cell fate determination, stem cell self-renewal, proliferation and apoptosis 34 . The Notch pathway is composed of four receptors (Notch1-4) and five ligands (Jag1-2, DLL1, DLL3-4). Upon ligand binding, the activation of Notch is governed by a two-step proteolytic cleavage by ADAMS family protease and then γ -secretase to release the intracellular domain of Notch (ICN), which is subject to forming a complex with Mastermind-like 1 (MAM1) 35 . Recent studies have demonstrated the oncogenic potential of Notch as well as its constitutive activation in different type of cancers, such as human T-cell acute lymphoblastic leukemia 36 , non-small cell lung cancer 37 , ovarian carcinomas 38 , and pancreatic cancer 39 . Meanwhile, Notch1 signaling was suggested to be activated in human OS and may be associated with tumor invasion and metastasis [40][41][42] . The metastatic OS cell lines have higher levels of Notch 1, Notch 2, DLL1 and the Notch-induced gene Hes1 43 . Notch2, Jagged1, HEY1, and HEY2 were over-expressed in OS biopsy specimens whereas Notch1 and DLL1 were down-regulated 44 . In addition, the miR-34 cluster was inversely correlated with invasiveness in a small panel of OS tumors, suggesting that the miR-34 family members may be responsible for regulating Notch expression 42 . Here, we showed that DLL1 expression is elevated in multi-chemoresistant OS cell lines. DLL1's involvement in OS appears more complicated, as it can be up-or down-regulated in different OS cell lines.
For further mechanistic understanding, we compared the activity of cancer-related signaling pathways, including Notch, in G-292 and SJSA-1 cells. No significant difference in the Notch pathway was found, even though DLL1 is a Notch ligand. Alternatively, we found that the ATF2/ATF3/ATF4 signaling pathway is involved in the DLL1-mediated repression of OS chemoresistance. We therefore further searched for the interactions between DLL1 and the master transcription factor gene for the ATF2/ATF3/ATF4 signaling pathway (Fig. 7) using GeneMANIA 45 . In total, 16 interacting genes of DLL1 were found in both GeneMANIA and the UniHIdatabase 5 . Then, these 16 genes were subjected to search for the relationships with the ATF2/ATF3/ATF4 transcription factors. Based on literature mining, a sophisticated connection network concerning a series of interactions was generated (Fig. 7). The DLL1 gene showed a direct genetic interaction with ATF2. In addition, DLL1 regulates ATF4 via a physical interaction with PSEN1, whereas DLL1 affects ATF3 via the interaction network of DLL1-MFNG-NOTCH1-ATF3. Notably, the PSEN1 gene regulates miR-193a-3p's promoting effect on multi-chemoresistance in bladder cancer 46 . Notch1 plays a central role in the interplay of DLL1 with ATF2/ATF3/ ATF4 and MEF signaling pathways. Thus, Notch1 may be the key player for the regulation of DLL1 in the ATF2/ ATF3/ATF4 pathway. Further studies are needed for elucidating the fine mechanisms of the DLL1-regulated OS multi-chemoresistance. In addition, the links between the observations found in this study and the clinical practice for better OS chemotherapy warrant further exploration. Notably, a previous report suggested that the expression of miR-34a reduces drug resistance by targeting CD117 in colorectal cancer 26 , which is contradictory to the promoting effect of miR-34a-5p in OS drug resistance. This conflict might reflect the complicated regulatory mechanism of the miR-34a-mediated cancer drug resistance in different cells.

Conclusion
In summary, we demonstrated that DLL1 is a target of miR-34a-5p and negatively regulates the OS multi-chemoresistance. The ATF2/ATF3/ATF4 signaling pathway might be involved in the miR-34a-5p's promoting effect on OS chemoresistance. Targeting miR-34a-5p as well as its target gene DLL1 through novel therapeutics may provide a valuable strategy to overcome OS chemoresistance. RNA-seq analysis. RNA-seq analysis was performed by BGI-Tech (Shenzhen, China), the method has been described in our previous report 48 . The raw data were deposited to the GEO dataset with the accession code of GSE89930. RNA analysis. Total RNA was extracted using TRIzol (Vazyme). For the mRNA analysis, the cDNA primed by oligo-dT was made with RT reagent kit (Tiangen, China), and the mRNA level of DLL1 was quantified by a duplex-qRT-PCR analysis where the TaqMan probes with a different fluorescence for β -actin (Shing Gene, China) were used in the FTC-3000P PCR instrument (Funglyn, Canada). The miRNA expression level was normalized using U6 small nuclear RNA (HmiRQP9001) as an internal control, as previously described 51 . Using the 2− ΔΔ Ct method, the β -actin level was normalized before comparing the relative level of the target genes. The sequences of primers and probes used for the qRT-PCR analysis are as follows: Protein analysis. Total proteins were extracted from cultured cells with cell lysis buffer (60 mMTris-HCl, pH 6.8, 2% SDS, 20% glycerol, 0.25% bromophenol blue, and 1.25% 2-mercaptoethanol) and heated at 95 °C for 10 min. Anti-DLL1 (20230-1-AP) was purchased from San Ying Biotechnology, China (Proteintech). The target proteins were then probed with anti-rabbit IgG peroxidase-conjugated antibody. The target bands were revealed by an enhanced chemiluminescence reaction (Pierce), and the relative density (level) of proteins over the GAPDH band was quantified with the Gel-Pro Analyzer (Media Cybernetics). Cell apoptosis analysis. Apoptosis was analyzed using Annexin V/PI double staining. 48 hr after transfection, the cells in the logarithmic growth phase were harvested and rinsed twice with ice-bathed PBS, then 3 μ l FITC-labeled enhanced annexin V and 3 μ l propidium iodide were added to the 150 μ l cell suspension. After incubation for 30 min, flow cytometry was performed on a FACSCalibur instrument. The number of apoptotic and necrotic cells were calculated by flow cytometry (Becton-Dickinson, USA) and analyzed by Flowjo Software. The ratio of early apoptosis was used for the test results. The experiments were performed three times independently, and a representative is shown.

Methods
Luciferase reporter assay. Cells were seeded in 24-well plate at a concentration of 2 × 10 5 cells/per well and co-transfected 24 hr later with pGL3-luc-Rab27B UTR WT and miR-20a-5p mimic/antagomiR or NC. 48 hr after transfection, cells were collected, and the relative luciferase activity was performed using Dula-Luciferase Reporter Assay Kit (Promega). The relative firefly luciferase activities of the UTR construct was analyzed as previously reported 11 .
Signaling pathway analysis. The reporter construct encodes the firefly luciferase reporter gene under the control of a basal promoter element (TATA box) joined to tandem repeats of a specific transcriptional response element. The NC construct encodes firefly luciferase under the control of a basal promoter element (TATA box) without any additional transcriptional response elements, is critical for identifying specific effects and determining background reporter activity. The positive control construct is a mix of constructs that constitutively express GFP, firefly luciferase and Renilla luciferase. The cells were transfected in triplicate with each firefly luciferase reporter construct in combination with the Renilla luciferase-based control construct using the riboFECT CP transfection reagent, and both the luciferase activities were measured in the cell extracts 24 hr after transfection. The luciferase activities (luciferase unit) of the pathway reporter relative to those of the negative control in the transfected cells were calculated as a measurement of the pathway activity.
In vivo studies. Four-week BALB/C nude mice were purchased from Silaike Experimental Animal China.
All of the animal experiments were carried out in strict adherence with the regulations for the Administration of Affairs Concerning Experimental Animals approved by the State Council of People's Republic of China. All Figure 7. A simplified interaction map was analyzed between the target gene (DLL1) with the TGFβ , Myc/ Max and ATF2/ATF3/ATF4 pathways by GeneMANIA (http://genemania.org/). Orange nodes represent the purposed genes. Gray nodes represent the genes that related with purposed genes through GeneMANIA method. The web-based interface searches gave a large set of functional association data to return related genes based on available genomic and proteomic data. The association data include protein, DNA and genetic interactions, pathways, gene and protein expression data, phenotypic screens and shared protein domains.
procedures involving animals and their care in this study were approved and performed by the Institutional Animal Care and Use Committee (IACUC) of the University of Science and Technology of China. The analyzed as previously reported 25 . Immunohistochemistry. The expression of DLL1 protein was measured using immunochemical analysis of 5-mm slices of formalin-fixed paraffin-embedded tumor xenografts from the nude mice. The tissue slides from all six groups were placed on a single slide and simultaneously subjected to the same immunestaining. Antigens were retrieved by pretreating dewaxed sections in a microwave oven at 750 watts for 5 min in citrate buffer (pH 6) and processed with the Super Sensitive Link-Labeled Detection System (Biogenex, Menarini, Florence, Italy). The enzymatic activities were determined using 3-amino-9-ethylcarbazole (Dako, Milan, Italy) as a chromogenic substrate. Following counterstaining with Mayer hematoxylin (Invitrogen), the slides were mounted in aqueous mounting medium (glycergel, Dako). Pictures were taken usinga LEICA DM 4000B microscope. Accordingto the rates of positive cell in each field, we marked the chipsas 0 (pigment free), + (light yellow),+ + (yellow),+ + + (51∼ 75%), and + + + + (brownish yellow) by their color intensity 52 .
Statistical analyses. The data are presented as the mean, and the error bars indicate the S.D. All statistical analyses were performed with Excel (Microsoft, Redmond, WA, USA). Two-tailed Student's t-test, a one-way analysis of variance was used to calculate statistical significance. A P-value of < 0.05 was considered significant.