The Notch Pathway Promotes Osteosarcoma Progression through Activation of Ephrin Reverse Signaling Free

Osteosarcoma is the most common primary malignant bone tumor, with an initial peak in diagnosis in the late adolescent and young adult period (1). Despite significant advances in the diagnosis and treatment of osteosarcoma, overall survival has remained relatively constant for the last 4 decades (2). In particular, patients with metastatic disease at diagnosis or those with recurrent disease have a poor prognosis, with only 20% surviving 5 years (3).

Notch signaling plays a key role in development in many different cell types and tissues (4). In mammalian cells, there are four Notch receptors (Notch1–4) and five Notch ligands (JAG1, 2, DLL1, 3, 4). Notch signaling is initiated by binding of Notch ligands to their receptors, followed by the release of the Notch intracellular domain (NICD) through sequential proteolytic events. NICD then translocates to the nucleus where it interacts with the transcriptional cofactor CBF1 and activates expression of target genes, such as the HES and HEY genes (5). Researchers have reported that the Notch signaling pathway participates in the progression of a variety of human cancers, including breast cancer (6, 7), prostate cancer (8, 9), and glioblastoma (10, 11). In many of these studies, it has been shown that increased Notch activity promotes tumor growth, whereas Notch pathway blockade inhibits it. A role for Notch signaling in osteosarcoma has also previously been identified (12). Conditional expression of NICD in immature osteoblasts is sufficient to drive the formation of bone tumors that displayed features of human osteosarcoma (13), and our previous studies have showed that Notch signaling participates in cisplatin-induced osteosarcoma stem cell enrichment (14). Nonetheless, the underlying mechanism of osteosarcoma progression warrants further investigation.

Eph receptor tyrosine kinases and their ephrin ligands are involved in a wide variety of physiologic and pathologic processes, including cell differentiation, proliferation, and apoptosis (15). Ephrins and Eph receptors are each divided into two classes based on sequence homology and binding specificity. EphA receptors bind glycosylphosphatidylinositol-anchored ephrin-A ligands, and EphB receptors bind transmembrane ephrin-B ligands (16). Interactions between Eph receptors and appropriate ephrin ligands result in bidirectional signaling (17). The deregulation of the Eph/ephrin signaling pathway has been implicated in tumorigenesis in a number of human cancers, such as breast, gastric, prostate, and osteosarcoma (18–21). Eph forward signaling is commonly tumor suppressive, whereas ephrin reverse signaling is often correlated with the promotion of malignancy (22). A previous study showed that the Grb4 transduces ephrin reverse signals, activation of ephrinB1 increases Fak activity, redistributes the Fak-binding protein paxillin, and leads to disassembly of focal adhesions (23). Recently, Cho and colleagues also found that ephrinB1 interacts with RhoGDI1 to promote cancer cell migration and invasion (24). Despite growing knowledge of ephrin-Eph signal transduction, detailed insights on its regulatory network are lacking.

In this study, we present both in vitro and in vivo evidence to clarify the role of Notch signaling in osteosarcoma progression. Moreover, we identify ephrinB1 as a transcriptional target of Notch pathway, which enhances ephrin reverse signaling and promotes tumor progression.

Twelve patients with osteosarcoma were enrolled in this study. This study was conducted in accordance with the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board (2015YW12) of Peking University Cancer Hospital and Institute (Beijing, China). The informed consent was written and obtained from all participants. Patients were diagnosed between January 2015 and December 2016 at the Department of Orthopedic Oncology, Peking University Cancer Hospital & Institute (Beijing, China). All patients were treatment-naïve prior to tumor biopsy. Patients received neoadjuvant chemotherapy prior to surgery, which includes cisplatin, methotrexate, and doxorubicin for 2 to 4 cycles according to the guideline of National Comprehensive Cancer Network. Patients whose tumors show more than 90% necrosis upon definitive surgery following neoadjuvant chemotherapy were defined as good responders, while the remaining were poor responders. Formalin-fixed paraffin-embedded (FFPE) osteosarcoma samples were obtained from biopsy and matched resected tumor. Clinical characteristics are summarized in Supplementary Table S1. Samples were conventionally sectioned and used for evaluating Notch activity by IHC.

To analyze the effect of Notch genes expression on prognosis in patients with osteosarcoma, several publicly available datasets (GSE21257, GSE39058, GEO2R) were analyzed. In GSE21257, genome-wide gene expression profiling of prechemotherapy biopsies of patients with osteosarcoma who developed metastases within 5 years (n = 34) were compared with those who did not develop metastases (n = 19). GSE39058 contains miRNA and mRNA expression data in paraffin-archived human osteosarcoma specimens that aims to reveal profiles with reproducible and independent prognostic value. GEO2R was used to extract the expression value and clinical characteristics, and Kaplan–Meier survival curve of patients with osteosarcoma with low or high expression of Notch genes were generated. The log-rank test were used to evaluate the outcomes of patients with different expression profiles. Spearman correlation analysis was applied for correlation analysis.

The human osteosarcoma cell lines U2OS, MG63, and 143B and were obtained from ATCC. The cells were cultured in DMEM containing 10% FBS and 1% antibiotics. Cells were propagated at 37°C with 5% CO₂ and 100% humidity. Cell viability was determined using Trypan blue staining. Culture medium was replaced every three days. Authentication of these cell lines was performed by short tandem repeat (STR) profiling. All experiments performed on cells that were passaged less than 20 times. Cells were routinely monitored for Mycoplasma contamination, and cell cultures were free of Mycoplasma. Cell line identification was performed at the end of experiments and showed 100% STR profiles matching to corresponding cell lines reported in ATCC.

Lentivirus/GV308-NICD1 (TetIIP-NICD1-3FLAG-Ubi-TetR-IRES-Puromycin) and Lentivirus/GV329-shRBPJ (TetIIP-shRBPJ-Ubi-TetR-IRES-Puromycin) and their corresponding control lentiviruses, lentivirus/GV308, and lentivirus/GV329, respectively, were obtained from Genechem. To generate stably transduced cell lines, osteosarcoma cells were infected with the viral supernatant according to the manufacturer's protocol. Infected cells (therefore named NICD1-OE, RPBJ-shRNA and CON) were incubated for 48 hours prior to selection of stable transfectants by addition of puromycin (5 μg/mL) for two weeks in selective medium. To induce the expression of NICD1 or RBPJ-shRNA, the cells were incubated with 2 μg/mL of doxycycline for 48 hours before subsequent experiments. For in vivo xenograft assay, the animals were administered 2 mg/mL of doxycycline in drinking water. EphrinB1 (eph) cDNA and shRNAs plasmids were purchased from Sino Biological and Santa Cruz Biotechnology, respectively, and they were transfected into osteosarcoma cells using FuGENE6 Transfection Reagent (Promega).

Cell viability was measured with the Cell Counting Kit-8 (CCK8, Dojindo) assay according to manufacturer's recommendations. A total of 5 × 10³ cells were seeded in a 96-well plate. The culture medium was replaced by 90 μL fresh medium mixed with 10 μL CCK-8 solution. Then the cells were incubated at 37°C for 2 hours. The OD450 value was measured and was used to calculate cell viability.

Five hundred cells per well were seeded in a 6-well plate. The culture medium was replaced every other day. After being cultured for 2 weeks, the cells were fixed with 4% paraformaldehyde for 30 minutes and stained with 1% crystal violet for 20 minutes. Clones with a diameter ≥ 50 μm were calculated and counted using microscope.

To determine the apoptosis rate, cells were harvested using 0.25% trypsin without EDTA. Cells were then stained with PE-labeled Annexin-V and PI and PBS was added as a negative control. To analyze the Stro-1/CD117 double positive cells, cells were digested and resuspended with PBS. According to the manufacturer's protocol, in brief, antibody (STRO-1 and CD117; BioLegend) or isotype control IgG treated with each test at 4°C for 30 minutes. Subsequently, the cells washed with cold PBS and flow cytometry analysis was performed with BD FACSCalibur (BD Biosciences) and the fluorescent intensities were measured by Cell Quest Software (BD Biosciences).

Cells were seeded until 90% to 95% confluence and then wounded using a yellow pipette tip. Three wounds were made for each sample and migration distance was photographed and measured at zero time, 12 hours, and 24 hours, respectively.

Cells were seeded on Matrigel (Sigma-Aldrich) coated polycarbonate membrane insert (6.5 mm in diameter with 8.0-μm pores) in a Transwell apparatus (Corning) and maintained in DMEM without serum. DMEM containing 10% FBS was added to the bottom chamber. After incubation for 12 hours at 37°C, cells on the top surface of the insert were removed and cells that migrated to the bottom surface of the insert were fixed, stained with 0.1% crystal violet, and then subjected to microscopic inspection. Cells were counted on the basis of five field digital images taken randomly at × 200 magnification.

Tumor cells were dispersed into single cells and then seeded into 6-well ultra-low attachment plates at a density of 5 × 10³ cells per well. The culture medium was used RPMI1640 supplemented with 20 ng/mL EGF (Sigma-Aldrich), 20 ng/mL bFGF (Sigma-Aldrich), and 1 × B27 (Invitrogen). After being cultured for 14 days, each well was taken with five randomly selected regions by microscope, tumor colonies with a diameter ≥ 50 μm were deemed as spheres, and the number of spheres were calculated.

FFPE sections (5 mm) were routinely dewaxed and rehydrated. Antigen retrieval was performed by incubating slides in 10 mmol/L citric buffer (pH 6.0) and microwaving for 15 minutes. After blocking, they were incubated with primary antibody anti-Hes1 (R&D Systems, # MAB3317, 5 μg/mL), anti-Notch1 (Abcam, #ab27526, 1:200), and anti-Ki67 (Abcam, #ab92742, 1:1,000) at 4°C overnight. After washing, they were incubated with biotinylated secondary antibody at 1/100 for 30 minutes at room temperature, then with Strep–ABC complex at 1/100 for 30 minutes at room temperature. Sections were finally developed with DAB substrate kit, counterstained with hematoxylin, and mounted. Sections were analyzed on high-power fields (×400) using a Leica digital microscope (Leica Microsystems). The positive ratio was used to evaluate the expression, since NICD is translocated to the nucleus following proteolytical processing, we regarded nucleus staining of Notch1 as NICD1 positive. Cells with distinctive brown particles in nucleus were considered positive. The slides were evaluated by two of the authors independently, and they were blind to the clinical data before scoring.

Total RNA was isolated and cDNA was synthesized according to the manufacturer's protocol. Amplifications were performed in an ABI PRISM 7000 Sequence Detection System (Applied Biosystems). Thermal cycler conditions were 50°C for 2 minutes and 95°C for 10 minutes, then 40 cycles of 15 seconds at 95°C followed by 1 minute at 60°C. β-Actin was used to normalize gene expression. Gene expression levels were calculated using the 2^−ΔΔC_t method and normalized to the β-actin. The gene-specific primers used are listed in Supplementary Table S2.

Proteins were extracted with Protein Lysis Buffer (Beyotime Biotechnology). Lysates were centrifuged at 10,000 × g at 4°C for 10 minutes, and supernatants collected. Cell lysates containing 40 μg protein were separated by SDS-PAGE and transferred to a polyvinylidene difluoride membrane using a transfer apparatus according to the manufacturer's protocols (Bio-Rad). After incubation with 5% nonfat milk in TBST (10 mmol/L Tris, pH 8.0, 150 mmol/L NaCl, 0.5% Tween 20) for 60 minutes, the membrane was washed once with TBST and incubated at 4°C, overnight with the indicated antibodies. The antibody we used to detect endogenous Notch1-NICD was from Abcam (#ab27526). Specific bands equal to 120 kDa were regarded as cleaved portion of NICD1. The following antibodies were obtained from Abcam: anti-Ephrin B1 (#ab99029, 1:200) and anti-phosphor-ephrinB1 (Y344; #ab30563, 1:500). Membranes were washed three times for 10 minutes and incubated with a 1:3,000 dilution of horseradish peroxidase–conjugated anti-mouse or anti-rabbit antibodies for 2 hours. Blots were washed with TBST three times and developed with the ECL system (Amersham Biosciences) according to the manufacturer's protocol. The films were scanned and the optical density of the target band was analyzed using Alpha processing system. All the experiments were repeated at least three times. The densitometry ratios shown on figures are mean value from multiple blots.

EphrinB1 promoter with/without the wild-type Notch-binding site and with a mutated Notch-binding site were cloned into pGL4 Luciferase Reporter Vectors (Promega) according to the manufacturer's protocol. Promoter activities were evaluated using a dual-luciferase reporter assay (Promega). The luciferase reporters were transfected using Lipofectamine 2000 (Qiagen). Forty-eight hours posttransfection, cells were lysed and luciferase assays were performed on a luminometer following the manufacturer's protocol. Luminescence of the firefly luciferase was normalized to that of the Renilla luciferase for each experiment and triplicates were averaged.

Chromatin immunoprecipitation assay (ChIP) assays were performed via a commercially purchased chromatin immunoprecipitation kit (Millipore) according to manufacturer's instructions. Cells were routinely cross-linked and sonicated. The samples were then centrifuged and the supernatants were sent for preclear. Cross-linked chromatin was incubated with 5 μL anti-Notch1 (Abcam, Immunogen from C-terminal of human Notch-1, ChIP Grade #ab27526) or control IgG in a total volume of 1 mL at 4°C overnight. Antibody–protein–DNA complexes were isolated by immunoprecipitation with salmon sperm DNA/protein A. After extensive washing, pellets were eluted by elution buffer (1% SDS, 0.1 mol/L NaHCO₃). Formaldehyde cross-linking was reversed. Samples were purified through PCR purification kit columns (Qiagen). The immunoprecipitated DNA was amplified by with primer pairs flanking the predicted Notch-binding sites in the human ephrinB1 promoter. An intronic sequence of the human β-actin gene served as negative control. The primer list is available in Supplementary Table S3.

A total of 3 μg RNA per sample was used as input material for the RNA sample preparations. Sequencing libraries were generated using the TruSeq RNA Sample Preparation Kit (Illumina). DNA fragments with ligated adaptor molecules on both ends were selectively enriched using Illumina PCR Primer Cocktail in a 15 cycle PCR reaction. All of the DNA fragments were processed by 2 × 150 base, paired-end sequencing on a Hiseq platform (Illumina). DESeq2 with negative binomial distribution was adopted to analyze differentially expression of mRNAs. Reads abundance was evaluated and RPKM (Reads Per Kilo bases of Per Million Reads Mapped) were obtained. Gene Ontology (GO) annotations and KEGG pathway analysis were applied to investigate the roles of all DE mRNAs. The RNA-sequencing data have been deposited in Sequence Read Archive (accession number SRP151852).

Male athymic nude mice (nu/nu, 4–5 weeks old) were purchased and maintained at the Center for Animal Experiment, Renmin Hospital of Wuhan University (Wuhan, China). All animal experiments were approved by the Ethics Committee of Renmin Hospital of Wuhan University (approval number 20150419). Osteosarcoma cells transfected with the NICD1-OE, RBPJ-shRNA, or control construct were counted by Trypan blue staining, and suspended in 100 μL of 50% Matrigel/PBS.

For extreme limiting dilution assay (ELDA), tumors were grown following subcutaneous inoculation with 1 × 10⁴ to 1 × 10⁶ cells, with 5 mice in each group. Tumor growth was defined at >2 mm diameter and mice were monitored for up to 6 months.

To determine in vivo tumorigenicity, mice were randomly divided into three groups (NICD1-OE, RBPJ-shRNA, and CON). A total of 5 × 10⁶ cells were subcutaneously injected, with 6 mice in each group. Tumor volume was monitored weekly, and was calculated using the following formula: V = 1/2 × width² × length. Data were accumulated from at least three independent experiments. After 8 weeks, or when the tumor volume was more than 2,000 mm³, mice were sacrificed and the tumor tissues and corresponding lungs were removed and sent for further analysis.

Values are presented as means ± SEM and error bars represent the SEM of at least three independent experiments. If not stated otherwise, two-tailed Student t test (paired and unpaired) and ANOVA with Tukey post hoc test were performed for statistical analysis. P values less than 0.05 were considered statistically significant.

The mean expression of Hes1 was 15.49% in primary osteosarcoma biopsy samples, in contrast, only low levels of Hes1⁺ cells (approximately 1.4%) were detected in the adjacent normal bone marrow and muscular cells (Fig. 1A). We then analyzed paired diagnostic biopsy and surgical sample and found the resected postchemotherapeutic tumors had greater levels of NICD1 and Hes1 compared with their biopsy counterparts (Fig. 1B). Furthermore, tumors that responded poorly to chemotherapy had significantly greater levels of NICD1 and Hes1 expression in their chemo-naïve biopsy samples (Fig. 1C). Therefore, activated Notch signaling correlates with a worse response to chemotherapy. In addition, we analyzed datasets from the Gene Expression Omnibus (GEO) to determine the clinical significance. Using GSE39058, we found that the recurrence-free survival was poorer in patients with high Notch1 and Hes1 expression compared with low ones (Fig. 1D). Using GSE21257, the metastasis-free survival and overall survival were poorer in patients with high Hey1 and HeyL expression compared with low ones (Fig. 1E and F). However, not all of the comparisons discussed are statistically significant, and it would be due to small sample size. Taken together, activation of Notch signaling is associated with poor prognosis in osteosarcoma.

We found that NICD1 and Hes1 were expressed in all three osteosarcoma cells tested (Supplementary Fig. S1A). Osteosarcoma cells were then divided into NICD1-overexpressing (NICD1-OE), RBPJ-shRNA, and mock-transduced (CON) groups (Supplementary Fig. S1B). Colony-forming assays revealed that NICD1-OE cells showed a moderate to significant growth advantage, while Notch inhibition by RBPJ-shRNA strongly inhibited growth in all three osteosarcoma cell lines (Fig. 2A). Cell proliferation assays revealed that NICD1-OE cells showed a moderate to significant growth advantage, while Notch inhibition by RBPJ-shRNA strongly inhibited growth in all three osteosarcoma cell lines (Fig. 2B). Supporting this, we found the two cell-cycle kinase complexes CDK4/6-CyclinD and CDK2-CyclinE were upregulated in NICD1-OE cells while downregulated in RBPJ-shRNA cells (Supplementary Fig. S1C).

By treating with cisplatin at different concentrations for 48 hours, NICD1-OE cells showed increased chemoresistance compared with the vehicle control. Conversely, Notch inhibition by RBPJ-shRNA promoted chemosensitivity of osteosarcoma to chemotherapeutic reagents (Fig. 2C; Supplementary Fig. S1D and S1E). Accordingly, an increased Bcl2/Bax ratio and decreased expression of caspase-3 protein were observed in NICD1-OE 143B cells, implying that activated Notch signaling promotes chemoresistance, while Notch inhibition contributes to enhanced cisplatin chemosensitivity (Fig. 2D).

Wound-healing assays showed that the motility of RBPJ-shRNA 143B cells was dramatically compromised compared with vehicle control, whereas NICD1-OE cells demonstrated accelerated cell migration (Supplementary Fig. S1F). Similarly, using Transwell Matrigel assays, we observed enhanced invasive capabilities in NICD1-OE osteosarcoma cells, and impaired invasiveness in RBPJ-shRNA cells (Fig. 2E). These data imply that Notch activity promotes osteosarcoma proliferation, chemoresistance, and invasion.

We then tested the in vivo effect of Notch signaling. Tumor growth was significantly inhibited in the RBPJ-shRNA group compared with the control group, whereas NICD1-OE cells showed increased tumorigenicity (Fig. 2F and G). Compared with control, NICD1-OE treatment induced the expression of the Notch target gene Hes1, together with an increase in proliferation as measured via Ki67 staining. In contrast, RBPJ-shRNA reduced Hes1 and Ki67 expression (Fig. 2H). These data indicate that Notch activity is essential for tumor establishment and maintenance. Furthermore, we examined metastatic potential of the different cells. The number of metastatic foci/lung was significantly higher in NICD1-OE mice than control mice, whereas RBPJ-shRNA mice exhibited significantly less foci (Fig. 2I). Taken together, our data suggest Notch signaling promotes osteosarcoma progression both in vitro and in vivo.

We have previously demonstrated that γ-secretase inhibition selectively eliminates osteosarcoma stem cells (14). Therefore, we sought to examine the role of Notch within the cancer stem cell (CSC) population in osteosarcoma. To validate the enrichment of CSCs, we analyzed cell surface markers commonly used for the identification of cancer stem cell subpopulations (Stro-1, CD117; ref. 25). NICD1-OE showed a significant enrichment of Stro-1/CD117 double positive (DP) cells, with these making up about 10% of the total cell population (Fig. 3A). We also analyzed the expression of a set of genes that have been shown to maintain the stem cell phenotype (Sox2, Oct4, TERT). All these genes were upregulated when NICD1 was overexpressed, whereas RBPJ-shRNA inhibited their expression (Fig. 3B and C; Supplementary Fig. S2A). Several publications have shown that osteosarcoma cells grown under serum-free, low attachment conditions are enriched for specific stem-like characteristics (26). We were able to grow cell spheres from all cell lines tested. As predicted, NICD1-OE cells were more efficient in forming spheres than control cells (Fig. 3D). To test the role of Notch signaling on tumorigenicity in vivo, we performed the ELDA assay. NICD1-OE cells were much more efficient in forming tumors than RBPJ-shRNA cells. Injecting 1 × 10⁵ NICD1-OE 143B cells resulted in 100% tumor formation, while 1 × 10⁵ RBPJ-shRNA 143B cells presented with no tumor occurrence. Furthermore, as few as 1 × 10⁴ NICD1-OE 143B cells were able to initiate tumor formation (Fig. 3E) and these cells were successfully capable to form new tumors by serial transplantation. When ELDA analysis was performed to calculate stem cell frequency, the NICD1-OE 143B cells were approximately 50 times more efficient in forming xenograft tumors than RBPJ-shRNA 143B cells (Supplementary Table S4). These data support the hypothesis that Notch activity promotes stem cell–like properties in osteosarcoma cells.

To investigate the underlying mechanism of Notch-induced osteosarcoma progression, we applied sequencing analysis to compare differentially expressed genes between 143B RBPJ-shRNA and CON cells. A total of 285 genes were found to be downregulated and 556 genes were found to be upregulated in RBPJ-shRNA cells compared with CON cells (Fig. 4A). These genes were enriched in cytokine–cytokine receptor interaction, signaling pathways regulating pluripotency of stem cells, and chemokine signaling, among others. Eph/ephrin signaling participates in a wide spectrum of developmental processes and tumorigenicity, and we found ephrinB1 was downregulated in RBPJ-shRNA cells. We measured mRNA expression levels of ephrinB1 and Hes1 in 143B cells, and confirmed the findings of the RNA sequencing (Fig. 4B). In addition, using publicly available dataset (GSE21257), we found that ephrinB1 expression was positively correlated with Hes1 (Fig. 4C). On the basis of these results, we sought to explore whether the expression of ephrinB1 is directly promoted by the activation of NICD1. We performed in silico analysis to find the putative Notch-binding site in the ephrinB1 promoter region, and three were identified (-1438 to -1431, -2430 to -2423, and -2911 to -2904; Fig. 4D). Using ChIP assay to enrich the NICD1-bound DNA fragments, we detected that the strongest binding site of NICD1 to ephrinB1 was -1438 to -1431 (Fig. 4E). Furthermore, we manifested the regulatory activity of NICD1 on the ephrinB1 promoter via dual-luciferase reporter assays, we found that ephrinB1 promoter activity was decreased in the 143B cells transfected with no binding site and with a mutated CBF1-binding motif in -1438 to -1431 compared with the control cells transfected with the wild-type CBF1-binding motif in -1438 to -1431 (Fig. 4F). To further confirm the regulatory effect of NICD1 on ephrinB1, the protein expression levels of ephrinB1 in NICD1-OE, RBPJ-shRNA, and their corresponding control cells were measured by Western blot assays. Consistent with previous results, we confirmed that expression of ephrinB1 was correlated with Notch activity (Fig. 4G; Supplementary Fig. S2B). These results imply that ephrinB1 is a direct NICD1 target.

Once ephrinB1 was confirmed as a Notch target gene, we wonder whether Ephrin-reverse signaling was activated. The phosphorylation of ephrinB1 was detected, and we found that phosphorylation of ephrinB1 was enhanced along with upregulated ephrinB1 expression after Notch activation (Fig. 4G). We further investigated whether ephrinB1 mediates the Notch-driven osteosarcoma proliferation and chemoresistance, the efficacy of ephrinB1 cDNA and shRNAs were confirmed using Western blot analysis (Supplementary Fig. S2C). Cell proliferation assays showed that NICD1 overexpression significantly increased the proliferation of osteosarcoma cells, and knockdown of ephrinB1 in NICD1-OE cells reversed Notch-induced proliferation. We also observed that the proliferation of osteosarcoma cells was significantly reduced by RBPJ-shRNA, and was rescued by the accompanying ephrinB1 overexpression (Fig. 5A). In addition, cell toxicity assays showed that NICD1 overexpression induced chemoresistance of osteosarcoma cells, and ephrinB1 depletion reversed this phenotype. Meanwhile, ephrinB1 overexpression compromised the chemosensitivity induced by RBPJ-shRNA (Fig. 5B).

Several factors contribute to poor survival of osteosarcoma, including late diagnosis, resistance to chemotherapy, and metastatic disease (27). In this study, we measured Notch pathway proteins using IHC. We found that human osteosarcoma samples overexpress Notch pathway proteins, and Hes1 expression was enriched after neoadjuvant chemotherapy. These data are consistent with similar reports from a range of other human cancers (6, 8, 10). Moreover, we provide evidence that upregulated Hes1 expression correlates with decreased chemotherapy response, suggesting that Notch gene expression in a patient's biopsy sample might predict outcome to chemotherapy. Our data suggest NICD1 or Hes1 are potential biomarkers to stratify patients and direct individualized treatment. In addition, through data mining from publicly available dataset, we found that Notch pathway genes correlate strongly with poorer prognosis, including recurrence, metastasis, and overall survival. The drawback in using these data is their relatively small sample size (n = 53 in GSE21257, and n = 42 in GSE39058), which makes statistics for Hes1 insignificant.

The Notch signaling pathway is well known for its complexity. Its role in cancer has been extensively studied in the context of oncogenic Notch mutations (28, 29). A previous study established a role for Notch in inhibiting osteoblast differentiation, and showed that its activation in committed osteosarcoma cells induces a highly tumorigenic phenotype (13). Consistent with our study, Tanaka and colleagues also reported that Notch pathway inhibition prevents osteosarcoma growth by cell-cycle regulation (30). Wang and colleagues demonstrated that Notch1 is critical for cisplatin sensitivity in osteosarcoma (31). Our previous observations showed that Notch signaling is activated in chemoresistant osteosarcoma cells (14). In this study, we further confirmed that Notch signaling strongly facilitates the acquisition of tumor cell progression including proliferation and chemoresistance.

Outcome for patients with metastatic disease is much poorer than for those with localized osteosarcoma (32). Tsuru and colleagues found that Notch target gene Hey1 could promote osteosarcoma metastasis via MMP9 expression (33), and Mu and colleagues reported that Notch signaling is associated with ALDH activity and increased metastatic behavior in osteosarcoma cells (34). Accumulating evidence suggests that many sarcomas can undergo EMT-related processes, which may be associated with aggressive clinical behavior (35). Our previous study has provided evidence that there is heterogeneity among osteosarcoma cells with regard to their EMT status (36). Consistently, we have also found that Notch signaling promotes invasiveness in vitro, therefore Notch inhibition may provide opportunities to improve clinical outcomes.

Using xenograft models, we clearly demonstrated that inhibition of Notch signaling is efficacious in reducing tumor growth and metastasis. Although Notch-silenced cells exhibited pulmonary metastasis, we found these lesions to be Notch positive. This could be due to these cells either escaping from shRNA treatment, or contamination of uninfected cells, further supporting the idea that Notch is indispensable for pulmonary metastasis of osteosarcoma.

Cancer stem cells are proposed to play major roles in drug resistance, tumor recurrence, and metastasis (37). Recent studies have shown evidence that osteosarcoma also possesses cells with stem cell–like properties (38). Our previous study showed that Notch activity is activated in osteosarcoma stem cells, but its stem cell–promoting effect in osteosarcoma remains unclear (14). We demonstrated here that Notch signaling is critical for these osteosarcoma stem cells and that Notch regulates genes that establish stemness. At the functional level, we showed that Notch signaling is required for self-renewal in osteosarcoma cells, and that raising Notch levels promotes these CSC-related traits. Using ELDA xenograft models, we clearly demonstrated that osteosarcoma stem cells are highly enriched in Notch-active cells, while Notch inhibition reduced the stem cell pool.

The downstream signals and molecular mechanisms mediating Notch activity in osteosarcoma are just unfolding. It has been previously shown that ephrinB1 is expressed in osteosarcoma and correlates with poorer prognosis (21). In our study, we found ephrinB1 to be upregulated and act as a direct target gene of Notch pathway. Ephrin reversing signal was generally correlated with tumor progression and initiated through tyrosine kinase–dependent and -independent manner (22). The former one is initiated by the recruitment of Src-family kinases and subsequent phosphorylation of ephrinB proteins, while kinase-independent signal occurs via interaction of PDZ domain proteins with ephrinB proteins, which modulated chemokine receptor function. We confirmed that phosphorylation of ephrinB1 was enhanced after Notch activation, suggesting that ephrin reversing signal is initiated after Notch activation via a tyrosine kinase–dependent manner. Therefore, we presumed that ephrinB1 is induced by Notch activation, and ephrin reverse signal is then enhanced, followed by tumor progression (Fig. 5C).

In conclusion, we provide preclinical evidence for the Notch pathway as both a molecular marker to predict osteosarcoma prognosis, and as a therapeutic target against osteosarcoma. In addition, we are the first to identify ephrin reverse signaling as a key mediator of Notch pathway. Therefore, using Notch signaling inhibitors could be a useful therapeutic strategy to sensitize tumors to neoadjuvant chemotherapy, and ultimately improve patient outcome.

No potential conflicts of interest were disclosed.

Conception and design: L. Yu, K. Xia, Z. Fan, W. Guo

Development of methodology: L. Yu, K. Xia, T. Gao, W. Guo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Yu, K. Xia, J. Chen, X. Sun, Z. Fan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Yu, K. Xia, T. Gao, J. Chen, X. Sun, B.M. Simoes, R. Eyre

Writing, review, and/or revision of the manuscript: L. Yu, K. Xia, J. Chen, X. Sun, R. Eyre, R.B. Clarke

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z. Zhang

Study supervision: W. Guo

The study was supported in part by grants from the Natural Science Foundation of China (81802689 and 81502575), Guidance Fund from Renmin Hospital of Wuhan University (RMYD2018M68), and Fundamental Research Funds for the Central Universities (2042015kf0069). B.M. Simoes and R. Eyre were supported by EU FP7 and Breast Cancer Now research funding, respectively. R.B. Clarke and B.M. Simoes are supported by the NIHR Manchester Biomedical Research Centre (IS-BRC-1215-20007).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.