ROCK Inhibition Induces Terminal Adipocyte Differentiation and Suppresses Tumorigenesis in Chemoresistant Osteosarcoma Cells

Phenotypic and functional heterogeneity arise among cancer cells within the same tumor as a consequence of genetic alterations, environmental influences, and reversible changes in cellular properties (1). Tumors are composed of cancer stem cells (CSC) that manifest self-renewal capacity and pluripotency, progenitor cells, and more differentiated cells, with this hierarchy being similar to that of normal tissues (2, 3). Minor tumor cell subpopulations enriched in CSCs have been found to be more resistant to classical anticancer drugs compared with other tumor cells, with the tumor cells that remain after termination of treatment with such agents being a potential cause of relapse (4, 5). The identification of subpopulations of tumor cells that acquire chemoresistance and the development of novel therapeutic approaches that target these cells are important clinical challenges.

Osteosarcoma is the most common primary malignant bone cancer in childhood and adolescence (6). Although chemotherapy regimens including those based on doxorubicin or cisplatin have improved the survival rate of such patients, 20% to 30% of affected individuals are refractory to these conventional treatments, largely as a result of the unchecked survival of chemoresistant cells (7–9). Like other cancers, osteosarcoma comprises heterogeneous cell types, including CSC-like cells, progenitor cells, and differentiated cells (10), but it has been unclear which of these cell types is responsible for the resistance of osteosarcoma to chemotherapy.

We previously established osteosarcoma-initiating (OSi) cells by introducing the gene for c-Myc into bone marrow stromal cells of Ink4a/Arf knockout mice (11). Mice injected with these cells develop lethal osteosarcoma with pathological features similar to those of human osteosarcoma, including the production of osteoid. Moreover, we found that these OSi cells are composed of two distinct types: AO cells, which have a trilineage (adipogenic, osteogenic, and chondrogenic) differentiation potential similar to mesenchymal stem cells, and AX cells, which have a bilineage (osteogenic and chondrogenic) differentiation potential similar to osteochondro progenitor cells and are highly tumorigenic.

Differentiation therapy is an approach to suppress tumorigenesis through conversion of malignant undifferentiated cancer cells into differentiated cells of low tumorigenicity. All-trans retinoic acid is an effective therapeutic agent in patients with acute promyelocytic leukemia. It elicits terminal granulocytic differentiation in the leukemia cells by impairing transcriptional repression of genes necessary for differentiation (12). Differentiation therapy has the potential to be beneficial for cancer resistant to conventional agents, but limited evidence is available at present regarding its applicability to solid tumors (13, 14). We have therefore now pursued the identification of a chemoresistant subpopulation of OSi cells and found that, in contrast to AX cells, AO cells are highly resistant to doxorubicin. We also found that the Rho-associated coiled-coil containing protein kinase (ROCK) inhibitor fasudil induced terminal adipocyte differentiation of chemoresistant AO cells and suppressed their in vitro growth and in vivo tumorigenesis. Our findings suggest that ROCK inhibitors are potential novel therapeutic drugs for targeting chemoresistant stem-like osteosarcoma cells.

Mouse osteosarcoma OSi, AO, AX, and AOT cells were established as previously described (11). Human osteosarcoma U2OS, SAOS2, SJSA1, HOS, and 143B cell lines were purchased from ATCC in September 2016 (U2OS and SAOS2), January 2017 (SJSA1 and HOS), and February 2017 (143B). All cell lines were frozen down at two passages and used in the experiments within five passages after thawing. These cells were authenticated by examination of in vitro growth characteristics and morphological properties provided evidence of correct cell identity. All cells were regularly tested mycoplasma-free by PCR. Human and mouse osteosarcoma cells were cultured under 5% CO₂ at 37°C in Iscove's modified Dulbecco's medium (IMDM; Thermo Fisher Scientific) supplemented with 10% or 20% FBS, respectively, and penicillin–streptomycin (Nacalai Tesque). For induction of adipocyte differentiation, cells were exposed to DMEM supplemented with 10% FBS, 0.5 mmol/L 3-isobutyl-1-methylxanthine (Wako), 0.1 μmol/L dexamethasone (Wako), and insulin-transferrin-selenium-X supplement (final insulin concentration of 5 μg/mL; Thermo Fisher Scientific).

Total RNA was isolated from cells with the use of the TRIzol reagent (Thermo Fisher Scientific), and portions (500 ng) of the RNA were subjected to reverse transcription (RT) with the use of ReverTra Ace qPCR RT Master Mix with gDNA Remover (Toyobo). Real-time PCR analysis was performed in triplicate with the use of TaqMan Fast Universal PCR Master Mix (Applied Biosystems) and a StepOne Plus thermocycler (Applied Biosystems). The probes for mouse PPARγ (Pparg; GenBank accession no. NM_011146.3; Mm00440940_m1), mouse GLUT4 (Slc2a4; NM_009204.2; Mm01245502_m1), mouse perilipin 1 (Plin1; NM_175640.2; Mm00558672_m1), mouse ROCK1 (Rock1; NM_009071.2; Mm00485745_m1), mouse ROCK2 (Rock2; NM_009072.2; Mm01270843_m1), mouse MKL1 (Mkl1; NM_153049.2; Mm00461840_m1), mouse ACTA2 (Acta2; NM_007392.3; Mm01546133_m1), mouse MYL9 (Myl9; NM_172118.1; Mm01251442_m1), mouse MMP9 (Mmp9; NM_013599.3; Mm00442991_m1), mouse tenascin C (Tnc; NM_011607.3; Mm00495662_m1), mouse Sp7 (Sp7; AK032521.1; Mm00504574_m1), human PPARγ (PPARG; NM_015869.4; Hs01115513_m1), human SLC2A4 (SLC2A4; NM_001042.2; Hs00168966_m1), human PLIN1 (PLIN1; NM_001145311.1; Hs0016017_m1), and human MKL1 (MKL1; NM_001269589.1; Hs00252979_m1) genes were obtained as TaqMan Pre-Developed Assay Reagents (Applied Biosystems). A TaqMan probe (4352339E; Applied Biosystems) for the mouse glyceraldehyde-3-phosphate dehydrogenase gene (Gapdh; NM_008084.2) and a TaqMan probe (4319413E; Applied Biosystems) for the eukaryotic 18S ribosomal RNA gene (X03205.1) were included as endogenous controls for mouse and human, respectively.

Nuclear and cytoplasmic fractions of cells were prepared with the use of a NE-PER Nuclear and Cytoplasmic Extraction Reagent Kit (Pierce). Immunoblot analysis was performed according to standard procedures (15) with primary antibodies to PPARγ (1:250 dilution; #2435; Cell Signaling Technology), to FABP4 (1:500 dilution; ab13979; Abcam), to perilipin A/B (1:500 dilution; P1873; Sigma-Aldrich), to FLAG (1:1000 dilution; F1804; Sigma-Aldrich), to lamin A/C (1:500 dilution; #2032; Cell Signaling Technology), and to GAPDH (1:1000 dilution; G8795; Sigma-Aldrich).

Cells were washed three times with PBS and then fixed for 15 minutes at room temperature with 4% paraformaldehyde. Immunofluorescence staining was performed according to standard procedures (15) with primary antibodies to PPARγ (1:250 dilution; #2435; Cell Signaling Technology), to FABP4 (1:250 dilution; ab13979; Abcam), to perilipin A/B (1:500 dilution; P1873; Sigma-Aldrich), and to MKL1 (1:50 dilution; sc21558; Santa Cruz Biotechnology). Secondary antibodies conjugated with Alexa Fluor 594 or Alexa Fluor 647 were obtained from Molecular Probes. Filamentous actin (F-actin) was stained with Alexa Fluor 488–labeled phalloidin (200 U/mL; Molecular Probes) in PBS. Cells were counterstained with Hoechst 33342 (Sigma-Aldrich) at 5 μg/mL diluted in water and were observed with an FV10i confocal laser-scanning microscope (Olympus) or a BIOREVO BZ-9000 fluorescence microscope (Keyence).

Cells were counted with a TC20 automated cell counter (Bio-Rad). Live and dead cells were identified by staining with trypan blue (Sigma-Aldrich). Cell proliferation was measured with the use of a CellTiter-Glo Cell Proliferation Assay Kit (Promega). All assays were performed in triplicate.

Cells were exposed to the following agents: doxorubicin (Aspen), cisplatin (Yakult), fasudil (AdooQ Bioscience), Y-27632 (Nacalai Tesque), CT-04 (Rho Inhibitor I; Cytoskeleton), latrunculin A (Calbiochem), or doxycycline (Sigma-Aldrich).

Cells were fixed with 70% ethanol for 2 hours at −20°C, washed with PBS containing 2% FBS and 0.01% NaN₃, and stained with PBS containing propidium iodide (PI; Sigma-Aldrich) at 2 μg/mL. The cell-cycle profile of 20,000 single cells was determined by flow cytometry with an Attune instrument (Thermo Fisher Scientific) and Flowjo software (TreeStar).

All experiments using mice were approved by the Institutional Animal Care and Use Committee (IACUC; approval number 12066) of Keio University and performed according to IACUC guidelines. AO cells (4 × 10⁶) or OSi cells (8 × 10⁶) were injected subcutaneously and bilaterally into the flank of syngeneic 6-week-old male C57BL/6 mice (Sankyo Labo). Drugs were administered according to the schedules described in Supplementary Materials and Methods or figure legends. Tumor volume was calculated from caliper measurements as 0.5 × length × width², as previously described (16). Tumors were isolated for analysis on day 25 or 35.

IHC analysis was performed as previously described (15). Frozen human osteosarcoma sections were obtained from OriGene. Frozen or paraffin-depleted sections were stained with primary antibodies to GFP (1:100 dilution; 598; Medical and Biological Laboratories, Nagoya, Japan), to PPARγ (1:100 dilution; #2435; Cell Signaling Technology), to perilipin A/B (1:500 dilution; P1873; Sigma-Aldrich), and to MKL1 (1:100 dilution; HPA030782; Sigma-Aldrich), and they were counterstained with hematoxylin (Wako). For immunohistofluorescence analysis, paraffin-depleted sections were stained with primary antibodies to perilipin A/B (1:500 dilution; P1873; Sigma-Aldrich) and then with Alexa Fluor 594–conjugated secondary antibodies (Molecular Probes). Sections were observed with a BIOREVO BZ-9000 fluorescence microscope (Keyence).

Expression of shRNAs was achieved with the retroviral expression vector pRePS (kindly provided by T. Hara, Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan), which also contains a puromycin resistance gene. The sequences of the sense oligonucleotides were 5′-CGAGGACTATTTGAAACGGAA-3′ for MKL1 shRNA #1, 5′-CCCACTCAGGTTCTTTCTCAA-3′ for MKL1 shRNA #2, and 5′-CGTACGCGGAATACTTCGA-3′ for luciferase shRNA (nonspecific control). The pRePS vectors were introduced into Plat-E packaging cells (17) by transfection for 24 hours with the FugeneHD reagent (Roche). The culture supernatants collected after a second incubation for 24 hours were passed through a 0.45-μm cellulose acetate filter (Iwaki) to obtain the generated retroviruses. AO cells were infected with the retroviruses for 48 hours in six-well plates, and the infected cells were then subjected to selection in the presence of puromycin (10 μg/mL).

The coding region for mCherry was amplified from the vector pMXs-3 × FLAG-mCherry-IP (15) by PCR with the primers 5′-CCGCTCGAGATGGTGAGCAAGGGCGAGGAGGATAACATG-3′ (XhoI_mCherry forward primer) and 5′-TTTTCCTTTTGCGGCCGCTTACTTGTACAGCTCGTCCATG-3′ (NotII_mCherry reverse primer), and the coding region for GFP was amplified from pMXs-IG (kindly provided by T. Kitamura, The University of Tokyo) by PCR with the primers 5′-CCGCTCGAGATGGTGAGCAAGGGCGAGGAGCTGTTCACC-3′ (XhoI_GFP forward primer) and 5′-TTTTCCTTTTGCGGCCGCTTACTTGTACAGCTCGTCCATG-3′ (NotII_GFP reverse primer). The PCR products were digested with XhoI and NotI and then ligated into the XhoI and NotI sites of pMXs-IP (kindly provided by T. Kitamura) to yield the vectors pMXs-mCherry-IP and pMXs-GFP-IP, respectively. The cDNA for a constitutively active mutant (MKL1-N100) of human MKL1 was obtained by digestion of Addgene plasmid #27176 (18) with EcoRI and BamHI, and the released fragment was cloned into the EcoRI and BamHI sites of the custom pRetroX-3 × FLAG-Tight-Pur vector (19) to yield pRetroX-3 × FLAG-MKL1-N100-Tight-Pur). Retroviral vectors were introduced into Plat-E packaging cells (17), recombinant retroviruses were collected, and cells were infected with the retroviruses and subjected to selection as described above for RNAi. For generation of AO-TetOn-FLAG-MKL1-N100 cells, AO cells were infected with the pRetroX-Tet-on Advanced-hygro response virus for 48 hours (19), and the infected cells were subjected to selection in the presence of hygromycin (200 μg/mL). Cells resistant to hygromycin were then infected with the pRetroX-3 × FLAG-MKL1-N100-Tight-Pur virus for 48 hours and subjected to selection with puromycin (10 μg/mL).

Quantitative data are presented as means ± SD and were analyzed with Student t test for comparisons between two groups. Mann–Whitney U test was applied for analysis of differences in tumor size. A P value of <0.05 was considered statistically significant.

Additional and more detailed methods are described in Supplementary Materials and Methods.

We first attempted to identify a subpopulation of OSi cells highly resistant to chemotherapeutic agents. The effects of doxorubicin and cisplatin, both of which are administered clinically for the treatment of osteosarcoma, on the viability of AO cells, which possess adipogenic potential (Supplementary Fig. S1A and S1B), and of AX cells, which have lost this potential, were thus examined. The IC₅₀ values of doxorubicin (2,881 vs. 36.3 ng/mL) and cisplatin (3,584 vs. 383 ng/mL) were both markedly higher for AO cells than for AX cells (Fig. 1A). AO cells also showed a higher intrinsic capacity compared with AX cells for efflux of the fluorescent dye Hoechst 33342 (Supplementary Fig. S2A), a functional feature of stem cells (20), and treatment of AO cells with verapamil, an efflux pump inhibitor, markedly increased their sensitivity to doxorubicin (Supplementary Fig. S2B). These findings suggested that a high intrinsic efflux capacity contributes to the chemoresistance of AO cells. We next examined the characteristics of chemoresistant cells among OSi cells. The proportion of cells expressing PPARγ, a master regulator of adipogenic differentiation (21), was significantly increased after treatment of OSi cells with doxorubicin (Fig. 1B). Expression of fatty acid binding protein 4 (FABP4), which is encoded by a target gene of PPARγ, was also increased in OSi cells treated with doxorubicin (Supplementary Fig. S3A). Furthermore, the cells remaining after doxorubicin treatment could be induced to undergo differentiation into adipocytes that express perilipin, a marker of terminal adipocyte differentiation, by exposure to an adipogenic cocktail (Fig. 1C; Supplementary Fig. S3B). We also established AO cells that express the fluorescent marker protein mCherry (AO-mCherry cells) as well as AX cells that express GFP (AX-GFP cells), and we exposed a mixture of these cells (AO-mCherry:AX-GFP = 3:7) to doxorubicin. Although the number of AX-GFP cells was greater than that of AO-mCherry cells in the absence of doxorubicin, the number of the latter cells was greater than that of the former after doxorubicin treatment (Fig. 1D). These results thus indicated that AO cells that express PPARγ and possess adipogenic potential manifest a chemoresistant phenotype. In addition, we found that doxorubicin treatment resulted in a significant increase in the normalized abundance of PPARG mRNA in human osteosarcoma cell lines (Supplementary Fig. S4). Moreover, AO-like cells expressing PPARγ were also detected in human recurrent osteosarcoma tissue but not in parosteal osteosarcoma tissue, the latter of which is a typically well-differentiated subtype of osteosarcoma (Supplementary Fig. S5), suggesting that adipogenic potential might also contribute to chemoresistance in human osteosarcoma.

Given that cell growth and differentiation are mutually exclusive within the adipogenic lineage (22, 23), the induction of terminal adipocyte differentiation may provide a novel approach to suppression of the proliferation of chemoresistant stem-like osteosarcoma cells. Dynamic changes in the actin cytoskeleton induced by modulation of RhoA-ROCK signaling have been shown to be essential for adipocyte differentiation (Fig. 2A; refs. 24, 25). Furthermore, we have previously shown that treatment with the ROCK inhibitor Y-27632 alone, through induction of remodeling of the actin cytoskeleton, is sufficient to trigger terminal differentiation of several preadipocyte cell lines (15). Knockdown of ROCK1 expression was also found to attenuate cell proliferation and viability and to induce apoptosis in the human osteosarcoma cell lines KHOS and U2OS, and high levels of ROCK1 expression were found to be associated with poor prognosis in patients with osteosarcoma (26). In addition, ROCK2, but not ROCK1, has been implicated in the migration of osteosarcoma cells (27). These findings thus suggest that ROCK contributes to osteosarcoma development. On the basis of these various observations, we examined whether chemical agents that target RhoA–ROCK signaling or dynamics of the actin cytoskeleton might suppress tumorigenesis by AO cells through induction of terminal adipocyte differentiation.

We first assessed whether fasudil, a ROCK inhibitor administered clinically (28), might induce terminal adipocyte differentiation of AO cells. Treatment of AO cells with fasudil alone resulted in a marked increase in the expression of adipogenic differentiation genes at both mRNA and protein levels in a concentration-dependent manner, as revealed by RT and real-time PCR and immunoblot analyses, respectively (Fig. 2B and C). In contrast, fasudil had no such effects in AX cells (Supplementary Fig. S6A and S6B). Fluorescence microscopy also revealed that fasudil induced remodeling of the actin cytoskeleton as well as the expression of FABP4 and perilipin in AO cells (Fig. 2D; Supplementary Fig. S7A and S7B). Furthermore, fasudil suppressed the proliferation of AO cells in a concentration-dependent manner (Fig. 2E). Flow cytometric analysis of cell-cycle profile showed that the proportion of AO cells in G₀–G₁ phase was significantly increased by fasudil treatment, whereas that of cells in S or G₂–M phases was correspondingly reduced (Fig. 2F). These findings thus indicated that fasudil alone is sufficient to elicit terminal adipocyte differentiation as well as growth suppression in chemoresistant AO cells. In addition, the induction of adipocyte differentiation and growth suppression by fasudil was also observed in three different human osteosarcoma cell lines (Fig. 2G; Supplementary Fig. S8).

We next examined the effects of fasudil on AOT cells, which have lost their adipogenic potential in AO cells (11). In contrast to AO cells, AOT cells were highly sensitive to doxorubicin (Supplementary Fig. S9A) and failed to undergo adipocyte differentiation in response to fasudil treatment (Supplementary Fig. S9B). Moreover, fasudil had no effect on the proliferation or cell-cycle distribution of AOT cells (Supplementary Fig. S9C and S9D), implicating that induction of the adipocyte differentiation program is involved in growth arrest by fasudil treatment. We also noticed that the relative proportion of G₀–G₁ phase in AO cells at steady state was greater than those in AOT cells (74.6% vs. 56%; Fig. 2F; Supplementary Fig. S9D). These findings suggest that AO cells represent a quiescent slow-cycling phenotype, and we thus believe that the inhibitory effect of fasudil on cell-cycle distribution in AO cells is modest for this intrinsic dormancy. We further performed a sphere formation assay with AO and AOT cells. Fasudil impaired sphere formation under nonadherent culture conditions in both cell types (Supplementary Fig. S9E), suggesting that adipogenic potential is not related to this inhibitory effect. In addition, we assessed whether one or both ROCK isoforms (ROCK1 and ROCK2) are required for prevention of terminal adipocyte differentiation in AO cells. RNA interference (RNAi)–mediated depletion of either ROCK1 or ROCK2 alone induced remodeling of the actin cytoskeleton and terminal adipocyte differentiation as well as suppressed the growth of AO cells (Supplementary Fig. S10A–S10D), suggesting that fasudil induces terminal adipocyte differentiation and growth arrest in chemoresistant AO cells by targeting both ROCK1 and ROCK2.

To clarify further whether actin remodeling contributes to terminal adipocyte differentiation in AO cells, we examined the effects of chemical agents that induce depolymerization of the actin cytoskeleton. Treatment with the ROCK inhibitor Y-27632, the Rho inhibitor CT-04, or the actin-depolymerizing agent latrunculin A resulted in a marked increase in the expression of adipogenic differentiation genes and elicited terminal adipocyte differentiation in AO cells (Fig. 3A and B; Supplementary Fig. S11). The effects of these various agents were less pronounced than those of fasudil, however. These results suggested that the Rho–ROCK–actin axis indeed contributes to regulation of terminal adipocyte differentiation in chemoresistant AO cells.

Megakaryoblastic leukemia 1 (MKL1, also known as MAL or MRTF-A) is a transcriptional coactivator of serum response factor (SRF), and the binding of MKL1 to monomeric G-actin prevents its translocation to the nucleus and thereby inhibits its coactivator function (29, 30). We previously showed that depolymerization of the actin cytoskeleton drives adipocyte differentiation by preventing the nuclear translocation of and transcriptional activation by MKL1 (15). We examined the expression of MKL1 in human osteosarcoma tissues. IHC analysis of a tissue array containing 49 human osteosarcoma specimens revealed that 45 (91.8%) samples were positive for MKL1 (Supplementary Fig. S12). The nuclear localization of MKL1 was also apparent in 29 samples (59.2%), suggesting that MKL1 is frequently activated in human osteosarcoma. We next tested the effect of ROCK inhibition on the localization and activity of MKL1 in AO cells. Consistent with previous results (29), MKL1 was localized to the cytoplasm of serum-deprived AO cells (Fig. 3C). Serum stimulation induced the nuclear translocation of MKL1, but this effect was prevented by fasudil treatment (Fig. 3C). We also found that fasudil suppressed the expression of the previously identified (30, 31) MKL1 target genes Acta2, Myl9, Mmp9, and Tnc (Fig. 3D). To examine the role of MKL1 in fasudil-induced terminal adipocyte differentiation in AO cells, we established AO cells that stably express a FLAG epitope–tagged constitutively active mutant of MKL1 (FLAG-MKL1-N100) that lacks the actin binding domain (18) under the control of the TetOn promoter (AO-TetOn-FLAG-MKL1-N100 cells). Exposure of these cells to the inducer doxycycline triggered the nuclear accumulation of and transcriptional activation by MKL1 as well as impaired fasudil-induced terminal adipocyte differentiation (Fig. 3E and F; Supplementary Fig. S13). Furthermore, to test whether depletion of MKL1 alone elicits adipocyte differentiation of AO cells, we established AO cells stably depleted of Mkl1 mRNA by retrovirus-mediated expression of specific short hairpin RNAs (shRNA; Fig. 3G). Although such depletion of MKL1 resulted in marked downregulation of the expression of MKL1 target genes, it induced significant upregulation of the expression of adipocyte differentiation genes (Fig. 3G). Together, these findings suggested that attenuation of MKL1 function contributes to the fasudil-induced terminal adipocyte differentiation of AO cells.

To investigate the effects of fasudil in vivo, we injected AO cells that stably express GFP subcutaneously into syngeneic C57BL/6 mice. Five days after cell injection, the mice were injected intraperitoneally with fasudil or saline daily for 30 days and tumor size was determined at various time points (Fig. 4A). Although it had no effect on body weight in mice bearing AO tumors (Supplementary Fig. S14A), fasudil treatment significantly suppressed tumor growth compared with the control (Fig. 4B and C). IHC and immunohistofluorescence analyses revealed the presence of many differentiated adipocytes in the tumors from fasudil-treated mice but not in those from saline-treated mice (Fig. 4D; Supplementary Fig. S14B). The fasudil-treated tumors were thus positive for a marker of terminal adipocyte differentiation (perilipin) as well as for GFP. These results thus indicated that fasudil treatment induced terminal adipocyte differentiation and suppressed the growth of tumors derived from AO cells. Moreover, we found that fasudil treatment impaired the nuclear localization of MKL1 in the AO-derived tumor cells (Fig. 4E; Supplementary Fig. S14C), implicating MKL1 as a mediator of the effects of fasudil on chemoresistant AO cells in vivo.

We examined whether fasudil might induce terminal adipocyte differentiation even in heterogeneous osteosarcoma cell populations. Only a small proportion of OSi cells underwent terminal adipocyte differentiation on exposure to fasudil in vitro (Supplementary Fig. S15A and S15B). We therefore tested the effects of sequential treatment with doxorubicin and fasudil in our heterogeneous osteosarcoma model. We mixed AO-mCherry cells and AX-GFP cells at a ratio of 3:7 and then exposed the cells consecutively to doxorubicin and fasudil (Supplementary Fig. S16A). Similar to the results shown in Fig. 1D, AO-mCherry cells survived the doxorubicin treatment (Supplementary Fig. S16B). Subsequent fasudil treatment induced the accumulation of lipid droplets as revealed by immunofluorescence analysis of perilipin in the surviving AO-mCherry cells (Supplementary Fig. S16B). We then performed a similar experiment with unfractionated OSi cells (Fig. 5A). Increased expression of adipogenic differentiation markers at both mRNA and protein levels was apparent after doxorubicin treatment (Fig. 5B; Supplementary Fig. S16C), suggesting that AO-type cells were enriched among the surviving OSi cells. Furthermore, subsequent exposure of these cells to fasudil induced a further marked upregulation of these genes as well as the accumulation of lipid droplets and remodeling of the actin cytoskeleton (Fig. 5B and C; Supplementary Fig. S16C and S16D).

Given these effects in heterogeneous osteosarcoma cells in vitro, we examined whether sequential treatment with doxorubicin and fasudil might induce terminal adipocyte differentiation as well as suppress tumorigenesis in OSi cells in vivo. We injected OSi cells into syngeneic C57BL/6 mice, which were treated (or not) intravenously with doxorubicin on days 7 and 9 and then intraperitoneally with fasudil or saline once daily beginning on day 11 for 14 days (Fig. 5D). Although treatment with doxorubicin or fasudil alone induced a significant reduction in tumor volume compared with the control, sequential treatment with both agents had a significantly greater effect (Fig. 5E). In addition, terminally differentiated adipocytes derived from OSi cells were observed in the tumors of fasudil-treated mice (Fig. 5F; Supplementary Fig. S17). These results suggested that fasudil induces terminal adipocyte differentiation in vitro and in vivo even in our heterogeneous osteosarcoma model, and that sequential treatment with doxorubicin and fasudil might be an effective therapeutic strategy for heterogeneous osteosarcoma.

Our results revealed that AO cells, unlike AX cells, are highly resistant to conventional anticancer drugs (doxorubicin and cisplatin), suggesting that AO cells possess chemoresistant characteristics. We also found that the clinically administered ROCK inhibitor fasudil induced terminal adipocyte differentiation in AO cells through regulation of the actin cytoskeleton, and suppressed the growth of AO-derived tumors. On the basis of our observations, given that human osteosarcoma tumors are composed of heterogeneous cell types, we propose a novel therapeutic option for osteosarcoma (Fig. 6). Treatment with doxorubicin to eliminate AX-type chemosensitive tumor cells followed by treatment with fasudil to convert the remaining AO-type chemoresistant cells into terminally differentiated adipocytes.

Tumor heterogeneity is characterized by a cell hierarchy in which CSCs differentiate into progenitor cells and more differentiated cells (2, 3). Given that CSCs manifest multipotency, they are potentially susceptible to induction of cell fates distinct from the original lineage. In our osteosarcoma model, AX cells are highly tumorigenic compared with AO cells (11), and we also found that AX cells express the gene for Sp7 (also known as osterix), a master regulator of osteogenesis, at a high level (Supplementary Fig. S18), suggesting that these cells are committed osteoblastic precursors. Fasudil treatment is therefore an appropriate differentiation therapy that directs chemoresistant osteosarcoma cells toward the adipogenic lineage of low tumorigenicity rather than toward the osteogenic lineage of high tumorigenicity. This “transdifferentiation” approach might thus be an effective option for therapeutic targeting of CSCs.

Thiazolidinediones, such as rosiglitazone, pioglitazone, and troglitazone, are high-affinity PPARγ ligands and inducers of adipogenic differentiation. These drugs also exert antitumor effects on various tumor cells (32). Treatment of osteosarcoma cells with thiazolidinediones has been shown to induce differentiation into adipogenic or osteogenic lineages and to suppress tumor growth (14, 33). However, troglitazone, the first drug of this class to be approved for clinical use, was withdrawn from the market as a result of its hepatic toxicity, and rosiglitazone has been associated with a risk of myocardial infarction (34–36). Moreover, thiazolidinediones are associated with a risk for weight gain as well as bone loss and fracture, with the long-term clinical administration of these drugs thus being problematic (37). However, fasudil was approved in Japan in 1995 and has been administered for the treatment of cerebral vasospasm after surgery in patients with subarachnoid hemorrhage, and it has been found to be safe and effective in clinical trials (28). We found that the antiproliferative effect of fasudil on AO cells was not accompanied by the induction of cell death (Fig. 2E), suggesting that its antitumor action is not associated with cytotoxicity. Notably, screening of 20 anticancer agents (Supplementary Table S1) revealed that fasudil enhanced the sensitivity of AO cells only to doxorubicin among the drugs commonly administered for osteosarcoma therapy in the clinical setting (Supplementary Fig. S19). Given that fasudil also potentiated the antitumor activity of daunorubicin, mitomycin C, and vinblastine, these drugs are also potential options for combination therapy with fasudil. Drug repositioning refers to the identification and development of new therapeutic uses for existing drugs and has the advantage that the safety profiles of the drugs are already known and their development is thus faster and cheaper compared with that of new drugs (38). Our present findings suggest that ROCK inhibitors including fasudil have the potential to be repositioned as anticancer drugs that trigger terminal adipocyte differentiation in chemoresistant osteosarcoma cells.

It will be important to clarify the mechanism by which fasudil induces adipocyte differentiation in osteosarcoma cells. We previously showed that G-actin produced as a result of actin depolymerization triggered by ROCK inactivation prevents the nuclear translocation and thereby abrogates the transcriptional activity of MKL1 and allows adipocyte differentiation (15). MKL1 thus functions as a gatekeeper that controls adipocyte differentiation. MKL1 inhibits adipogenesis through a direct inhibitory interaction with PPARγ (39). In this study, we found that treatment of AO cells with fasudil alone blocked the nuclear translocation and activation of MKL1 as well as induced remodeling of the actin cytoskeleton, and that depletion of MKL1 alone resulted in a marked increase in the expression of adipogenic differentiation genes. Together, these findings thus suggest that fasudil prevents the nuclear translocation and activation of MKL1 by inducing F-actin depolymerization and that the resulting inactivation of MKL1 leads to terminal adipocyte differentiation in AO cells.

The transcription factors Yes-associated protein (YAP) and transcriptional coactivator with PDZ-binding motif (TAZ) have been identified as mechanotransducers that mediate the effects of changes in actin cytoskeleton dynamics or extracellular matrix stiffness, and RhoA–ROCK signaling is essential for the nuclear translocation and function of these factors as well as for those of MKL1 (40, 41). Furthermore, YAP/TAZ and MKL1 physically interact with each other and function in a mutually dependent manner (42). Inhibition of YAP/TAZ function triggers the differentiation of mesenchymal stem cells toward the adipogenic lineage (40), and treatment of osteosarcoma cells with thiazolidinediones simultaneously induces adipocyte differentiation and inhibition of the YAP signaling pathway (14). Collectively, these various observations suggest that fasudil-induced terminal adipocyte differentiation of AO cells might be mediated by inhibition not only of MKL1 but also of YAP/TAZ signaling.

Our findings have suggested that the induction of differentiation into adipocytes through regulation of the actin cytoskeleton is a potential novel approach to the therapeutic targeting of CSCs in osteosarcoma. This approach takes advantage of the multipotency of CSCs to induce their transdifferentiation into another lineage. In glioma stem cells, which originate from transformed neural stem cells, impairment of differentiation into the neuronal lineage promotes tumorigenesis whereas, conversely, forced terminal neuronal differentiation attenuates tumor formation (43, 44). Moreover, regulation of RhoA–ROCK signaling by mechanical cues has been shown to contribute to the neuronal fate decision of neural stem cells, with YAP serving as a downstream effector in this process (45, 46). In addition, fasudil suppresses the progression of glioblastoma in vitro and in vivo (47). These various findings suggest that treatment with fasudil may induce terminal neuronal differentiation in glioma stem cells, resulting in suppression of their tumorigenicity. Induction of trans-terminal differentiation in CSCs through regulation of actin dynamics might thus be a potential treatment for various therapy-resistant tumors.

Hideyuki Saya reports receiving a commercial research grant from JSR Corporation, Nihon Noyaku Co., Ltd., Eisai Co., Ltd., and Daiichi-Sankyo Inc. and is a consultant/advisory board member for RBI Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: N. Takahashi, H. Nobusue, T. Shimizu, N. Onishi, T. Kuroda, H. Saya

Development of methodology: N. Takahashi, H. Nobusue, T. Shimizu, S. Yamaguchi-Iwai, N. Onishi, T. Kuroda

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Takahashi, H. Nobusue, T. Shimizu, E. Sugihara, S. Yamaguchi-Iwai, N. Onishi, H. Kunitomi, T. Kuroda

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Takahashi, H. Nobusue, E. Sugihara, T. Kuroda

Writing, review, and/or revision of the manuscript: N. Takahashi, H. Nobusue, T. Shimizu, T. Kuroda, H. Saya

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Shimizu, S. Yamaguchi-Iwai, N. Onishi, H. Kunitomi, T. Kuroda

Study supervision: H. Nobusue, T. Kuroda, H. Saya

We thank I. Ishimatsu for technical assistance; M. Sato and M. Kobori for help with preparation of the manuscript; and Collaborative Research Resources, School of Medicine, Keio University, for technical support and reagents. The compounds for screening were kindly provided by Japan Society for the Promotion of Science KAKENHI project JP16H06276. This work was supported by grants from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (grant nos. 18K07243 to H. Nobusue and 17H01401 to H. Saya).

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.