Mouse osteosarcoma AXT cells, which were previously established from Ink4a/Arf-null mice by overexpression of human c-MYC (5), or human osteosarcoma cells (U2OS (#HTB-96), MG63 (#CRL-1427), Saos2 (#HTB-85) purchased from ATCC (Manassas, VA, USA) were cultured in IMDM (Nacalai Tesque, Kyoto, Japan) or McCoy's 5A medium (Thermo-Fischer Scientific, Carlsbad, CA, USA), respectively, containing 10% FBS under 5% CO₂ at 37°C (27,28). Mouse bone marrow stromal cells (wild-type or Ink4a/Arf null BMSCs) were established as previously described (5). Briefly, bone marrow cells from C57BL/6J or Ink4a/Arf null mice were collected from femurs and tibias and hematopoietic cells were depleted with antibodies to CD45 and lineage-specific antigens. Adherent cells were cultured in IMDM supplemented with 20% FBS and used as BMSCs within ten passages.

Nintedanib (#HY-50904 MedChemExpress, Monmouth Junction, NJ, USA) was dissolved in dimethyl sulfoxide (DMSO) at a stock concentration of 20 mM.

AXT cells or human osteosarcoma MG63, Saos2, and U2OS cells (5×10² in 50 µl of IMDM for AXT cells or McCoy's 5A medium for MG63, Saos2, and U2OS cells, containing 10% FBS) were cultured in a 96-well cell culture plate. Then, 50 µl of the corresponding medium containing test reagents at twice the desired concentrations was added to each well. Cell viability was evaluated using a Cell Titer Glo assay kit (Promega, Madison, WI, USA). Assays were performed at least in triplicate and data are shown as the means ± SD relative (fold) against the corresponding control value for cells incubated in the absence of test reagents.

A 96-well cell culture plate was coated with 30 µl of Matrigel (Corning, Corning, NY, USA) and Human Umbilical Vein Endothelial Cells (HUVEC (#KE-4109) purchased from CRABO, Osaka, Japan) cultured with Endothelial Cell Growth Medium (Takara, Shiga, Japan) were seeded after passing through the 40 mm-cell strainer (2×10⁴ in 45 µl of the culture medium per each well). Then, 25 µl of the culture medium with or without 400 nM or 4 µM of nintedanib was added and the final total volume was 100 µl. Tube formation was evaluated after 14 h.

All mouse experiments were performed in accordance with the guidelines of Hoshi University, and the present study was approved by the Committee on Animal Research of Hoshi University (approval number: 20-071). Mice were housed in ventilated cages (five mice/cage; floor area 501 cm²) under specific pathogen-free conditions. Mice were fed a standard chow diet and water ad libitum. Rooms were maintained at 22°C and kept on a 12 h light and dark cycle with inspection every weekday to ensure that they were not under distress throughout the experiments.

The detailed duration of the experiments and the number of animals used was shown in main figures. Briefly, 1–2×10⁶ AXT cells were suspended in 100 µl of IMDM and injected subcutaneously into the bilateral flanks (two injections in total) of 7-week-old female C57BL/6J mice (SLC, Shizuoka, Japan) or 20-week-old female C57BL/6 SCID mice (The Jackson Laboratory, Bar Harbor, ME, USA) under inhalation anesthesia with 4% isoflurane (Wako, Tokyo, Japan) for the induction and 2% for maintenance.

Nintedanib at 50 mg/kg was orally administered once a day. The endpoint criteria were as follows: i) The mean tumor diameter exceeds 20 mm. ii) The combined tumor burden is more than 15% of body weight (~20 g of 7-week-old mice). iii) Occurrence of ulceration, infection, or necrosis of tumor. iv) Body weight loss is more than 20% of weight. The results did not reach the endpoint criteria in this study. Bilateral two tumors developed in each mouse in this study. The major and minor axes of the bilateral tumors were measured, and the estimated tumor volume was calculated according to the guideline of Washington State University (https://iacuc.wsu.edu/documents/2017/12/tumor-burden-guidelines.pdf/).

Estimated tumor volume (mm³)=d² × D/2. D and d were the major and minor diameter in mm, respectively. The maximum diameter and the volume of bilateral tumors observed in each mouse was listed in Table SI. When the mice were euthanized, a lethal dose (200 mg/kg) of pentobarbital sodium (Tokyo Kasei Kogyo, Tokyo, Japan) was intraperitoneally injected. Death was confirmed by cardiac arrest. None of the mice was unexpectedly dead or found dead during the study.

2X Laemmli sample buffer (Bio-Rad, Hercules, CA, USA) supplemented with β-mercaptoethanol was used for the preparation of cell lysate. Immunoblot analyses were performed according to standard semi-dry transfer methods using 5–20% gradient precast polyacrylamide gels (ePAGEL, ATTO, Tokyo, Japan). The primary antibodies were listed in Table SII. Signal intensities were quantitated using the ImageJ software (29), and the relative (fold) values against the corresponding control value normalized against the intensities of the α-Tubulin bands are shown.

Activated tyrosine kinase in vivo was screened using the mouse RTK phosphorylation antibody array C1 kit (RayBiotech, Norcross, GA, USA). The array map is shown in Table SIII. AXT cells were inoculated into C57BL/6J mice and nintedanib was administrated. The tumor lysate from two tumors of two mice was extracted by disruption using a BioMasher homogenizer (Nippi, Tokyo, Japan). The total cell lysate used was 500 µg.

Trypsinized cells were washed with PBS, fixed with 70% ethanol for ≥48 h at −20°C, washed twice with ice-cold PBS, and stained with PBS supplemented with 10 µg/ml propidium iodide and 20 µg/ml RNase. The DNA content of 20,000 singlet cells was measured by FACSVerse (BD Biosciences, San Jose, CA, USA).

Trypsinized cells were washed with ice-cold PBS, and stained with allophycocyanin-conjugated annexin V (eBioscience, Carlsbad, CA, USA) and propidium iodide. Stained cells were analyzed by FACSVerse. FlowJo software (Tree Star, Ashland, OR, USA) was used for the data management.

Total RNA was extracted and RT and qPCR analyses were performed using the NucleoSpin RNA kit and PrimeScript (Takara). To evaluate circulating tumor cells (CTCs), whole blood was collected from the right ventricle of euthanized mice and total RNA was extracted from 200 µl blood with a NucleoSpin RNA blood kit (Takara, Shiga, Japan). Since GFP is constitutively expressed in AXT cells, CTCs were quantitated based on the expression level of Gfp relative to Actb mRNA. The sequences of PCR primers are as follows: GFP forward: 5′-GACGTAAACGGCCACAAGTT-3′, reverse: 5′-TTGCCGGTGGTGCAGATGAA-3′, Ccnd1 forward: 5′-CAACAACTTCCTCTCCTGCT-3′, reverse: 5′-ACTCCAGAAGGGCTTCAATC-3′, Actb forward: 5′-CAACCGTGAAAAGATGACCC-3′, reverse: 5′-TACGACCAGAGGCATACAG-3′. qPCR analysis was performed with StepOne thermal cycler (Thermo Fisher Scientific) with the 2-step protocol; 60 sec at 95°C, then 40 cycles of 15 sec at 95°C and 60 sec at 60°C followed by the melting and dissociation curve analysis.

Immunohistochemical analysis was performed according to standard procedures (5,6,8). Deparaffinized sections were stained with primary antibodies listed in Table SII. Hematoxylin was used for nuclear staining.

Deparaffinized tumor sections were stained with primary antibodies to aSMA (Abcam) derived from rabbit and to CD31 (eBiosience) derived from rat. Alexa Fluor 555-conjugated anti-rabbit and Alexa Fluor 647-conjugated anti-rat secondary antibody (both from Abcam) were used. DAPI solution (Dojindo, Kumamoto, Japan) was used for nuclear staining. Samples were observed with FV3000 confocal microscopy (Olympus, Tokyo, Japan).

Unless indicated otherwise, quantitative data are expressed as the means ± SD relative to the control value. All assays were performed at least in triplicate. Data were analyzed with a Student's t-test or one-way analysis of variance (ANOVA) with the Dunnet post hoc test, using Graph Pad Prism 9 (GraphPad Software, San Diego, CA, USA). P<0.05 was considered to indicate a statistically significant difference.

First, the anti-proliferative effect of nintedanib on osteosarcoma cells in vitro was evaluated. Nintedanib decreased the viability of mouse osteosarcoma AXT cells in a concentration-dependent manner (Fig. 1A). To examine the effect of nintedanib on the non-tumor counterparts of AXT cells, Ink4a/Arf null BMSC and wild-type BMSC derived from the C57BL/6 background were used (5). The growth of both cell lines was also inhibited by nintedanib in a concentration-dependent manner and these cell lines were more sensitive to nintedanib than AXT cells (Fig. 1A). Next, the anti-proliferative effect of nintedanib on human osteosarcoma cells was evaluated. Similar to findings with AXT cells, nintedanib had a direct anti-proliferative effect on human U2OS and Saos2 osteosarcoma cells, but the effect of the drug on MG63 cells was less pronounced (Fig. 1B) and concentrations of nintedanib ≤3 µM did not suppress growth. These results indicate that the effect of nintedanib in osteosarcoma is dependent on cell type in vitro.

To elucidate the mechanisms underlying the suppression of osteosarcoma cell growth by nintedanib, the cell-cycle status was examined. A 20 h exposure of AXT cells to nintedanib decreased the fraction in S-phase (Fig. 2A and B). Treatment with nintedanib increased the sub-G1 fraction (Fig. 2C), indicating the emergence of apoptotic cells. To further evaluate the induction of apoptosis, AXT cells were treated with nintedanib for 23 h. Although the level of annexin V-positive apoptotic cells was not high, nintedanib concentration-dependently increased the percentage of this cell type (Fig. 2D and E). Together, these findings suggest that nintedanib inhibited cell-cycle progression and increased apoptotic cells in AXT cells.

To clarify the molecular events induced by nintedanib treatment, we evaluated activated kinases by immunoblotting. Phosphorylation of AKT was decreased by 21.5 h treatment with nintedanib or doxorubicin as a positive control (Fig. 3A). The phosphorylation level of ERK1/2 was also attenuated by nintedanib treatment. However, this suppression was not concentration-dependent and ERK1/2 activation was restored by 5 µM nintedanib treatment (Fig. 3B). The expression level of cyclin D1 was downregulated by nintedanib treatment. Notably, similar to the ERK1/2 phosphorylation result, 5 µM nintedanib restored cyclin D1 protein levels to control values (Fig. 3C). Cells treated with 5 µM nintedanib had higher gene expression levels of cyclin D1 than controls, but the level was not statistically significant (Fig. 3D). Thus, the high expression level of cyclin D1 induced by 5 µM nintedanib could be mainly attributable to the accumulation of cyclin D1 protein.

We further examined the activation of signaling molecules downstream of growth factor or cytokine receptors such as IL-6 and CXCL8 which are suggested to be important to promote osteosarcoma metastasis (30–32). Phosphorylation level of STAT3 was not high under the culture with 10% FBS containing medium and tended to be suppressed only at low concentration of nintedanib like p-ERK (Fig. 3E). p38MAPK and SAPK/JNK were activated by the supplement of nintedanib (Fig. 3E and F). These molecules might not be activated as downstream of cytokine or growth factor receptors. On the other hand, Src was inactivated by the supplement of high concentration of nintedanib, consistent with a previous report that nintedanib suppresses many kinases related to Src activation at high concentrations (10).

Collectively, these results showed that nintedanib altered the expression levels of intracellular signaling molecules, but the effect of nintedanib on these molecules was not necessarily concentration-dependent.

We examined whether nintedanib might exert anti-tumor activity against AXT cells in vivo. AXT cells were inoculated into syngeneic C57BL/6 mice and then the mice were treated with a single daily dose of nintedanib at 50 mg/kg (Fig. 4A). Previous toxicity tests using mice reported that no major adverse events occurred when daily 100 mg/kg nintedanib was orally administered for 14 days, and the approximate lethal single dose is over 2,000 mg/kg. The in vivo studies using mouse IPF models demonstrated that a daily dose of 24.9–83 mg/kg nintedanib in mice was efficacious and tolerated when administered for 10 to 30 days (URL: [Ofev, INN-nintedanib (europa.eu)], [https://www.ema.europa.eu/en/documents/assessment-report/ofev-epar-public-assessment-report_en.pdf]). Therefore, we set 50 mg/kg of nintedanib to evaluate the anti-tumorigenic effect.

Administration of nintedanib significantly decreased the primary tumor size and weight (Fig. 4B and C). Treatment with nintedanib did not affect the mouse body weight (Fig. 4D) and all mice were alive, well, and not exhausted during the treatment. Since AXT cells were labeled with GFP, circulating tumor cells could be quantitated (27,28). Treatment with nintedanib decreased the number of circulating tumor cells, although the effect was not statistically significant (Fig. 4E). Consistent with those findings, nintedanib also significantly reduced the levels of lung metastatic lesions (Fig. 4F and G). Notably, large metastatic lesions over 100 µm diameter were not detected in nintedanib-treated mice.

Therefore, these findings show that nintedanib as a single agent exerted an anti-tumor effect on primary lesions and metastatic growth of osteosarcoma.

To elucidate the molecular mechanisms underlying the anti-osteosarcoma effect in vivo, tumor lysate was prepared from tumors developed in mice inoculated with AXT cells as shown in Fig. 5A, and then the phosphorylation levels of receptor tyrosine kinases per unit of protein were compared. Phosphorylation was still detected in 31 of 71 kinases in the lysate from nintedanib-treated mice (Fig. 5B). FGFR, PDGFR, and VEGFR are the molecular targets for nintedanib (10–12), but phosphorylation levels of FGFR1 and FGFR2 were very weak and not affected by nintedanib treatment. Activation of PDGFRα/β, VEGFR2, and VEGFR3 could not be detected using this phosphorylation array system.

To further evaluate the activation levels of PDGFRs and VEGFR2, the same lysate was subjected to immunoblotting. The activation of PDGFRα/β was attenuated by nintedanib administration (Fig. 5C), suggesting that nintedanib inhibited its molecular targets in vivo. Phosphorylation of Tyr951 or Tyr996 of VEGFR2 was not detected (data not shown). We examined the activation status of downstream molecules. Consistent with our in vitro results at low concentrations, nintedanib decreased ERK1/2 phosphorylation levels (Figs. 3B and 5D). Cyclin D1 was slightly downregulated (Fig. 5E). However, phosphorylation levels of mTOR and AKT were not affected by nintedanib (Fig. 5F).

We evaluated the activation of PDGFR in AXT cells in vitro and compared to the result from the in vivo samples (Fig. S1A). Our previous studies using immunoprecipitation western blotting showed that PDGFR activation was difficult to detect under the culture with FBS containing medium without supplement of PDGF ligands (7). In addition, under the presence of serum in vitro, survival of AXT cells was not dependent on the PDGFR signaling. Consistent with our previous findings, the activation of PDGFR in vitro was considerably weaker than that in vivo. Interestingly, the expression level of PDGFRα increased with nintedanib treatment, but the associated receptor activation was not observed (Fig. S1B). In in vivo tumor samples, phosphorylation status of Src and STAT3 was not attenuated by nintedanib administration (Fig. 5G and H). On the other hand, like in vitro results, p38MAPK and SAPK/JNK were activated by nintedanib administration (Figs. 3E, F, 5H and I). These results suggest that nintedanib does not block STAT3 signaling under 50 mg/kg administration, and p38MAPK and SAPK/JNK might not be activated as downstream of cytokine or growth factor receptors.

To clarify nintedanib-mediated histological changes and the mechanisms underlying anti-osteosarcoma effect in vivo, immunohistochemical analyses were performed.

Previous reports suggest that nintedanib exerts anti-tumor activity by inhibiting the functions of CAFs or fibrogenic reprogramming that confers to metastatic ability to osteosarcoma cells (21,22,26). α-smooth muscle actin (αSMA) is a useful marker for smooth muscle cells and a part of fibroblasts including myofibroblasts or CAFs (33,34). In osteosarcoma developed according to Fig. 4A, αSMA positive cells consisted of a variety of cell types; a part of osteosarcoma cells, fibroblastic cells surrounding osteosarcoma (possibly including CAFs), CD31-negative cells localized around blood vessels, and CD31/αSMA double-positive cells (Figs. 6A and S2). However, these cells were also abundant in the tumors of nintedanib-treated mice, indicating that nintedanib treatment did not result in quantitative changes in αSMA-positive cells.

A previous report suggested that nintedanib promoted anti-tumor immunity and potentiated the effects of immune checkpoint inhibitors (23). However, tumor-infiltrating CD8-positive T cells were not numerous in either control or nintedanib-treated mice (Fig. 6B). In addition, nintedanib exhibited anti-osteosarcoma activity in C57BL/6 SCID mice, in which T- and B-cell function is obsolete (Fig. 6C and D).

Next, we evaluated the effect of nintedanib on vasculature formation in osteosarcoma. Immunohistochemical staining revealed that nintedanib reduced the formation of CD31-positive ducts, despite the variable abundance of tumor vasculatures (Fig. 7A and B). To evaluate the effect of nintedanib on vascular formations more directly, endothelial cell tube formation assay was performed. HUVEC formed tubes in 14 h in Matrigel and they became thinner and interrupted by the supplement of nintedanib compared to the control (Fig. 7C and D). Nintedanib treatment also attenuated the viability of HUVEC (Fig. 7E).

These findings indicated that nintedanib exerted anti-tumor activity in our osteosarcoma model mainly by inhibiting tumor vascular formation.