Trabectedin Overrides Osteosarcoma Differentiative Block and Reprograms the Tumor Immune Environment Enabling Effective Combination with Immune Checkpoint Inhibitors

This article is featured in Highlights of This Issue, p. 4945

Osteosarcoma constitutes a high-grade malignant stromal tumor composed of mesenchymal cells producing osteoid and immature bone (1). Although relatively rare, osteosarcoma is a tumor with high social impact owing to its peak of incidence in the second decade of life. Genetically, osteosarcoma is characterized by an exceptionally high frequency of gene and chromosomal aberrations including copy number changes and complex derivative chromosomes harboring multiple fusion sequences (2), impeding the identification of pathognomonic mutations and potential targeted therapies.

Clinically, osteosarcoma occurs abruptly and progresses rapidly. Most osteosarcomas are diagnosed in stage IIB or III already with occult micrometastases, consistent with the high mortality rate within 6 months of diagnosis following surgery treatment alone. State-of-the-art therapy comprises surgery and neoadjuvant multidrug chemotherapy with conventional cytotoxic drugs (3–5). Although tumor localization at diagnosis is associated with 60% to 70% chance of complete remission, the prognosis for metastatic disease remains dismal with few treatment options available upon relapse after first line therapies. The search for potential new therapeutic avenues could include manipulating the tumor microenvironment and its immune infiltration (6). However, few, contradictory studies on tumor–immune cell interaction are available in osteosarcoma. For example, although tumor-associated macrophages appear protumorigenic with poor prognostic value in several tumor types (7), a single osteosarcoma study shows their positive association, regardless of their M1 or M2 phenotype, with metastasis suppression in high-grade osteosarcoma (8).

To clarify the role of myelomonocytic cells in osteosarcoma progression and metastasis, we exploited immunocompetent models based on our recently developed osteosarcoma cell lines with different histotypes and differentiation grades, which can grow subcutaneously and orthotopically in the bone cavity, spontaneously metastasize to the lungs in the latter situation, phenocopying the clinical setting. We used these models to test the therapeutic efficacy and mechanisms of action of trabectedin on osteosarcoma. Trabectedin is a marine-derived chemotherapeutic agent already approved for clinical practice in Europe as a second-line single-agent treatment for soft-tissue sarcomas (9) or in combination with doxorubicin in ovarian cancer (10). The cytotoxic effects of trabectedin also target macrophages and monocytes (11, 12). Trabectedin treatment of fibrosarcoma, ovarian, and lung tumor mouse models significantly decreased peripheral blood monocyte and tumor site macrophage numbers, a key tumor growth-inhibitory effect (11). We report here that trabectedin significantly inhibited osteosarcoma primary tumor growth and metastasis through impacting both tumor- and immune-infiltrating cells and exhibited increased therapeutic efficacy when combined with PD-1–blocking antibody.

Heterozygous p53 (C129S2(B6)-Trp53^tm1Tyj/J), BALB/c background mice, which develop a wide spectrum of neoplasias (13), were provided by Dr. Lollini (University of Bologna, Bologna, Italy). Female 8-week-old BALB/c and nu/nu mice were purchased from Charles River Laboratories (Calco). Mice were maintained in the Animal Facility of Fondazione IRCCS Istituto Nazionale dei Tumori. Animal experiments were authorized by the Institute Ethical Committee and Italian Ministry of Health and performed in accordance to national law (D.lgs 26/2014).

p53^+/− mice were subjected to lethal irradiation followed by bone marrow transplantation with p53-competent bone marrow cells. Osteosarcoma cell lines (mOS13, mOS69, and mOS14) were established from the resultant bone lesions.

Cells were grown in a humidified incubator at 37°C with 5% CO₂ and maintained in standard medium, DMEM plus 10% FBS. Cell lines are checked for mycoplasma infection every 4 to 6 months by the PCR Mycoplasma Detection Kit (PanReac Applichem).

Following subcutaneous tumor cell (2 × 10⁵) injection, tumor mass was measured with a caliper and tumor volume (mm³) calculated [long diameter × (short diameter)²/2]. Trabectedin (YONDELIS, 0.25 mg powder for infusion; PharmaMar) was administered intravenously (0.15 mg/kg/body weight) once weekly for 3 weeks starting when tumors reached 4 to 5 mm diameter. Anti-mouse PD-1 (RMP1-14 clone, InvivoMab, BioXCell) or control antibody (rat IgG2a Isotype Control, clone 2A3 anti-Trinitrophenol InvivoMab, BioXCell) was administered intraperitoneally (200 μg/mouse) twice weekly, starting 48 hours after the first trabectedin injection.

To evaluate metastatic potential, 10⁵ and 2 × 10⁵ cells were injected intra-vein and intra-bone, respectively. Primary tumors in the tibiae were measured with a caliper evaluating the leg diameter; following intra-bone injection, trabectedin treatment was started when the injected leg diameter reached 7 to 8 mm. After intravenous injection, treatment started at day 10 for mOS69 and day 18 for mOS13. Lung metastases were counted by histologic evaluation on serial lung sections (5 sections for each sample), stained with hematoxylin and eosin (H&E) and metastatic area quantified using Leica Las Core software under a DM4B Leica microscope (details in Supplementary Data).

Tumors were excised, washed in PBS, fixed in 10% neutral-buffered formalin, and paraffin-embedded. All bone samples and subcutaneous mOS13 tumors were decalcified using an EDTA-based decalcifying solution (MicroDec EDTA-based, Diapath). Sections (4-μm-thick) were H&E-stained for tumor histotype definition.

Immunohistochemistry (IHC) was performed using a polymer detection method. Briefly, tissue sections were deparaffinized, rehydrated, antigen-unmasked using Epitope Retrieval Solution (Novocastra) at 98°C for 30 minutes, and then brought to room temperature and washed in PBS. Endogenous peroxidase was neutralized with 3% H₂O₂ and Fc blocked (Novocastra). Samples were incubated for 1 hour with primary antibodies (Supplementary Information) and counterstained with nuclear Fast-red (Sigma) or Harris hematoxylin (Novocastra). Von Kossa and Masson's Trichrome (for bone matrix deposition quantification) staining were performed with Sigma-Aldrich kits (detailed in Supplementary Data). Slides were analyzed under an Axioscope A1, and microphotographs were collected using a Axiocam 503 Color with Zen 2.0 Software (Zeiss). Ki67 proliferation marker quantitative IHC data were obtained by calculating the average Ki67⁺ cell percentage displaying nuclear reactivity among five fields (40× magnification).

Tumor sample leukocyte infiltration was evaluated by flow cytometric analysis as described previously (14), using the following antibodies: CD45.2; Gr-1; CD11b; F4/80, Ly6C, CD19, CD4, CD8, CD44, CD62L, and PD-1 (eBioscience). For T regulatory cell detection, after CD4 surface staining, cells were fixed, permeabilized, and stained with FoxP3 antibody, following manufacturer's instructions (eBioscience). Samples acquired using a BD LSR II Fortessa instrument were analyzed singly with FlowJo software (TreeStar).

Tumor samples collected and stored in RNAlater were mechanically disrupted in TRIzol (Invitrogen). RNA was purified by phenol/chloroform extraction then loaded onto RNeasy MINI or MICRO kits (Qiagen) with on-column DNAse treatment. RNA purity and yield were assessed using NanoDrop. RNA was reverse-transcribed using High Capacity cDNA Reverse Transcription Kit (Thermo Fisher). TaqMan array mouse immune panel (Thermo Fisher) was used to evaluate the tumor immune landscape following manufacturer instruction.

For validating specific array results, quantitative PCR reactions prepared using TaqMan Fast Universal PCR Master Mix were run on a 7900 HT Fast Real-time PCR System (Thermo Fisher) using probes detailed in Supplementary Table S1.

Osteosarcoma cell lines (2 × 10⁵) seeded in 60-mm plates were treated after 24 hours with trabectedin (0.2–0.4 nmol/L) or untreated up to 7 days. Following methanol/acetone (3:7) fixation for 10 minutes, osteoblastic differentiation was assessed at 4 and 7 days by alkaline phosphatase (ALP) staining (Leukocyte Alkaline Phosphatase Kit, Sigma) following manufacturer's instruction with Mayer' hematoxylin counterstaining for 5 minutes at room temperature.

Osteosarcoma cells seeded on coverslips (Sigma) were treated after 24 hours with trabectedin (0.5–1 nmol/L) for 4 hours (1 to 24 hours), washed in PBS, fixed in 4% paraformaldehyde, permeabilized in Triton X-100/0.15% PBS, blocked in 4% BSA, and incubated overnight with anti-Runx2 primary antibody (Santa Cruz Biotechnology, clone M70) and then goat anti-mouse FITC (1:100, Pierce) secondary antibody. Nuclei were counterstained with Hoechst 33258. Images were acquired with a Nikon ECLIPSE 90i with Plan Apo 60x/NA 1.4 DIC N2 and captured using a digital color camera (Nikon DS5MC) with NIS-Elements AR 3.10 software (Nikon).

In vitro and in vivo chromatin immunoprecipitation (ChIP) assays were performed as described (15). Precleared chromatin was immunoprecipitated using anti-Runx2 (M-70; Santa Cruz Biotechnology) antibody.

qPCR utilized flanking RUNX2-containing target promoter fragments: osteocalcin (OCN) promoter 5′-CTG GCA GTC TCC GAT TGT G-3′ (forward), 5′-ATG TGC TCA GTG GGT CAA AC-3′ (reverse); p21^WAF/CIP1 promoter 5′-ACA TTT CCC TCA TTT TGG ACC C-3′ (forward), 5′-TTC TCA GAC CAC GGA CTA CC-3′ (reverse). Data are indicated as fold enrichment respective to untreated cells and calculated as: % of recruitment = 2^ΔC_t × input chromatin percentage, where ΔC_t = C_t (input) − C_t (Runx IP; refs. 16, 17).

For OCN or p21 protein analysis, cells were treated with trabectedin (0.2–0.4 nmol/L) up to 7 days or 8 to 24 hours, respectively, and then lysed as described (18). The following primary antibodies were used: anti-OCN (C-7), anti-p21 (Santa Cruz Biotechnology), and anti-β-actin (Merck-Millipore); and the secondary antibodies used were horseradish peroxidase–conjugated anti-rabbit or anti-mouse (GE Healthcare).

For lung metastases evaluation, data are represented singularly with graphs showing the median; statistical significance was evaluated with Mann–Whitney U tests using Prism 5 software. Else, data are represented as the mean ± SD or SEM with the Student unpaired two-tailed t test or Mann–Whitney U test used for statistical analysis. *, P ≤ 0.05; **, P ≤ 0.005; ***, P ≤ 0.001.

To define trabectedin and anti-PD-1 antibody synergistic or addictive effects, their activity patterns were evaluated by a biostatistician using generalized linear models. A two-way ANOVA was implemented by considering the main effects and first-order interaction term.

Osteosarcoma cell lines were developed from p53^+/− mice receiving bone marrow transplantation with p53-competent bone marrow cells to avoid predominant lymphomagenesis. These mice developed bone lesions with higher frequency in comparison to parental p53^+/− animals (not shown).

After in vitro characterization and osteoblastic lineage assessment (Supplementary Fig. S1A and S1B), mOS13 (osteoblastic, well-differentiated) and mOS69 (fibroblastic, less differentiated) were chosen as representative of different histotypes and differentiation grades. Upon subcutaneous and intra-bone injection, mOS13 cells formed smaller tumors with increased bone matrix deposition and lower metastatic potential (upon intra-bone injection) than mOS69 (Supplementary Fig. S1C–S1E). Subcutaneous tumor sample IHC analyses revealed higher leukocyte (CD45⁺) infiltration including macrophages (CD68⁺), granulocytes (Gr-1), and lymphocytes (CD4⁺ and CD8⁺) in well-differentiated mOS13 compared with poorly differentiated mOS69 tumors (Supplementary Fig. S1F and S1G). The data on a higher monocytic/macrophagic infiltration in less aggressive, more differentiated tumors might suggest that, differently from other solid tumor histotypes, these cells could be associated with a better prognosis, as described previously (8) and recently confirmed in clinical tissue samples of patients treated at Rizzoli Institute according to protocol ISG-OS1 (19).

Enhanced mOS13 tumor immune cell infiltration was confirmed by qRT-PCR using the TaqMan array mouse immune panel representing 96 immune-related genes including specific immune cell subset, cytokine, chemokine, and other inflammatory mediator markers. mOS13 tumors exhibited upregulated expression of several immune- and inflammation-related genes, such as immune cell lineage (Cd68, Cd3, Cd8, Cd4, and Cd34) and function/activation state (MHC class II, Cd80, Cd86, Cd40l, Ctla-4, Icos, Fas/FasL, and Tbet) markers, cytokines and growth factors (Il1a, Il1b, Il7, Il10, Il12b, Il15, Il18, Csf1, Csf2, Csf3, Nos2, Tnf, and Tgfb1), and chemokines and chemokine receptors (Ccl2, 3, and 5; Cxcl10 and 11; and Ccr2, Ccr4, and Cxcr3; Supplementary Fig. S1H–S1L; Supplementary Table S2).

Trabectedin treatment significantly inhibited subcutaneous nodules from both cell lines (Fig. 1A), with higher tumor differentiation and bone matrix deposition as determined by histopathologic analysis in treated mice versus untreated controls (Fig. 1B; Supplementary Fig. S2A and S2B). This observation was more striking in the less differentiated mOS69. The tumor cell differentiation effect was confirmed by Masson trichrome staining, OCN IHC, and Von Kossa staining (the latter on mOS69 samples only as mOS13 tumors required decalcification; Fig. 1C–G and Supplementary Fig. S2A and S2B).

To evaluate the effect of trabectedin on tumor cell proliferation, Ki67 immunostaining was performed. We found that proliferation was significantly reduced in trabectedin-treated mice (Fig. 1H and I).

Trabectedin efficacy was also tested on the more recently developed osteosarcoma cell line, mOS14, which gives rise to high-grade, poorly differentiated, osteoblastic osteosarcoma. Confirming our results, trabectedin significantly inhibited subcutaneous mOS14 tumor growth, although neoplastic cell differentiation was less evident than in mOS69 and mOS13 tumors. Ki67 immunostaining confirmed cell growth inhibition (Supplementary Fig. S2C and S2D).

Trabectedin differentiative activity on tumor cells occurs in specific tumor histotypes such as myxoid liposarcoma and Ewing sarcoma, wherein it apparently overcomes lipogenic and neural cell differentiation blocks caused by oncogenic fusion proteins, altering their DNA binding (20, 21), but it has been also documented in human osteosarcoma cell lines in vitro (22). To evaluate the differentiative effect on our mouse osteosarcoma cell lines, we therefore treated them in vitro with trabectedin (0.2–0.4 nmol/L) for 4 or 7 days. ALP staining showed marked osteoblastic differentiation in low-dose–treated mOS69 cells at both time points, whereas the higher dose (0.4 nmol/L) had less effect, likely because of dose-related cytotoxic effects (Fig. 2A). As observed in vivo, trabectedin-induced mOS13 differentiation in vitro was less evident (not shown), likely because of their high basal differentiation state.

As trabectedin causes gene- and promoter-dependent transcription factor modulation (23), we analyzed its effect on the transcription factor Runx2, a master regulator of osteoblastic fate (24). Following mouse osteosarcoma cell lines trabectedin treatment, we performed ChIP to monitor Runx2 binding to its recognized target promoters: OCN, involved in the osteogenic program and p21, encoding a negative regulator of proliferation. Trabectedin increased Runx2 promoter recruitment in mOS69 cells, reaching maximal level 4 hours after treatment (Fig. 2B). mOS13 cells exhibited a similar trend albeit with faster kinetics for OCN, suggesting that the differentiation program operating in well-differentiated cells is driven by a more active transcription factor machinery than in poorly differentiated cells (not shown). Trabectedin-mediated OCN and p21 protein expression upregulation was confirmed by Western blotting (Fig. 2C and D). Accordingly, immunofluorescence on in vitro mOS69 cells showed rapid upmodulation of Runx2 and its translocation to the nucleus upon treatment (Fig. 2E).

Even more interestingly, Runx2 OCN and p21 promoter binding were also detected in in vivo mOS69 tumor samples from trabectedin-treated mice 24 hours after treatment in comparison to the control group, decreasing to the basal condition 7 days later (Fig. 2F). In vivo cell differentiation was confirmed by increased OCN expression, as detected by IHC (Fig. 2G). Furthermore, RUNX2 IHC clearly showed enrichment of RUNX2-positive cells with osteoblastic morphology in trabectedin-treated mOS69 as well as mOS14 tumors (Fig. 2H and Supplementary Fig. S2E), suggesting also an overall increased RUNX2 expression.

As metastatic disease represents the major challenge in the clinical management of osteosarcoma, we tested trabectedin efficacy against lung metastases induced experimentally by direct tumor cell intravenous injection or spontaneously arising upon intra-bone injection. In both cases, trabectedin significantly reduced lung metastases number and size for both mOS69 and mOS13 lines (Fig. 3), whereas primary tumors implanted orthotopically in the bone cavity were not significantly reduced in size (not shown). This could be ascribed to trabectedin-mediated tumor cell differentiation and consequent bone matrix deposition increase, as observed by H&E and Masson trichrome staining and quantification (Supplementary Fig. S3). Accordingly, metastatic lesions of treated mice were also characterized by conspicuous bone matrix deposition (Fig. 3C).

As shown in four different mouse tumor models, trabectedin efficacy is partly mediated by a cytotoxic effect on monocytes/macrophages either resident at the tumor site or circulating in the blood and spleen (11). To assess this mechanism in osteosarcoma, tumor section IHC and flow cytometric analysis on freshly isolated in vivo samples were performed to characterize distribution and number variations of immune cell subsets infiltrating mOS69 tumors following in vivo trabectedin treatment. Myelo/monocytic subsets identified through flow cytometry by high Ly6C and intermediate Gr-1 marker expression, compatible with monocytes and monocytic myeloid-derived suppressor cells (M-MDSC), were decreased in tumor cell suspensions. However, differentiated macrophages (F4/80 high), granulocytes, and Gr-1 high polymorphonuclear MDSCs (PMN-MDSCs) were not significantly modulated (Fig. 4A). Notably, CD8 and CD4 T-cell percentages were increased in tumors from trabectedin-treated mice (Fig. 4B), as confirmed by IHC (Fig. 4C).

FACS analysis on peripheral blood, bone marrow, and spleen from mOS69-bearing mice indicated systemic drug effects on immune cells with reduced myelo/monocytic cell numbers. Conversely, granulocytes and PMN-MDSCs, evaluated as Gr-1 high cells, although decreased in the spleen, trended toward increase in peripheral blood and bone marrow (Fig. 4D–F), potentially underlying a trabectedin effect on bone marrow precursors. Indeed, although treated mice showed significant overall reduction in cKit⁺ lineage–negative cell numbers, granulocyte macrophage progenitor percentages increased (not shown), likely explaining the enhanced bone marrow granulopoiesis and peripheral circulation granulocyte expansion. CD4 and CD8 T-cell percentages were also increased in the spleen of treated mice (Fig. 4G); however, among CD4 T cells, the ratio between effector T cells and FoxP3⁺ T regulatory cells did not exhibit significant difference (not shown).

To more deeply evaluate the trabectedin-induced tumor immune landscape modifications, in vivo mOS69 tumors from treated and untreated mice were analyzed using the TaqMan array mouse immune panel. Confirming flow cytometric and IHC results, trabectedin increased expression of T-cell markers including Cd3 and Cd8, as well as other T-cell–related genes such as Grzmb, Fas/FasL, Il7, Cd80, Ccl5, Cxcl10, and Cxcl11 (Fig. 4H), suggesting increased T-cell recruitment and adaptive immune response activation (Supplementary Table S3). Differential expression was validated by single-probe qPCR (Supplementary Fig. S4A). As chemokines can be produced by a variety of cells, including both tumor and infiltrating immune cells, we also assessed whether trabectedin was able to impact the secretion of the above chemokines directly by mOS69 cells in vitro. Although trabectedin treatment directly increased Ccl5, Cxcl10, and Cxcl11 expression by mOS69 cells in vitro, in vivo tumor samples exhibited markedly higher expression in the same qPCR analysis (not shown), suggesting that other cells in the tumor microenvironment contribute to cytokine secretion in vivo or induce their upmodulation in neoplastic cells.

Given the enhanced CD8 and CD4 T-cell recruitment, we tested whether adaptive immune cell response was involved in trabectedin efficacy using immunodeficient nude (nu/nu) mice. Although mOS69 tumor growth was accelerated in nude mice, suggesting the lack of an initial antitumor response likely occurring in immunocompetent mice, trabectedin therapeutic efficacy was maintained, indicating that adaptive T-cell response was likely not required (Fig. 5A). Histologic analysis showed a more differentiated phenotype and higher matrix deposition in tumors from trabectedin-treated versus untreated immunodeficient mice, comparable to immunocompetent mice (Fig. 5B).

A more detailed FACS analysis of tumor-infiltrating T cells was next performed to evaluate their activation state via CD44 and CD62L marker expression and that of the checkpoint inhibitor PD-1, known to mediate T-cell exhaustion/anergy. Naïve (CD44⁻CD62L⁺), effector memory (CD44⁺CD62L⁻), and central memory (CD44⁺CD62L⁺) T cells did not significantly differ (not shown); however, trabectedin treatment increased PD-1 expression on tumor-infiltrating CD8 T cells (Fig. 5C). Real-time qPCR analysis of in vivo tumor samples confirmed upregulated expression of Pd-1 and its ligands, Pd1-lg1 and Pd1-lg2 (Fig. 5D), with increased PD-1⁺ cell numbers further confirmed by IHC staining (Fig. 5E). Trabectedin-mediated PD-1/PDL-1 upregulation may explain the irrelevance of T-cell presence for trabectedin efficacy, yet provides a rationale for combining the checkpoint inhibitor anti-PD-1 antibody with trabectedin.

To assess whether blocking the PD-1 immune checkpoint could reactivate T cells recruited at the tumor site by trabectedin treatment, mOS69-bearing mice were treated with trabectedin and anti-PD-1 antibody, either alone or in combination. The combination improved trabectedin efficacy (with addictive effect), whereas anti-PD-1 alone did not significantly impact osteosarcoma growth, likely because of the paucity of tumor-infiltrating T cells in the absence of trabectedin (Fig. 6A–C). Notably, combination treatment further increased recruited CD4 and CD8 T-cell numbers (not shown) and shifted CD4 T-cell phenotype toward effector memory cells at the expense of naïve cells (Fig. 6D). TaqMan array mouse immune panel analysis showed a significant increase in genes associated with adaptive immune response following combination treatment above the levels attained by trabectedin treatment alone (Fig. 6E and F; Supplementary Fig. S5). Upregulated genes included T-cell markers (Cd3, Cd4, Cd8), genes associated with their cytotoxic activity (Gzmb, Fas, Prf1, Ifng), T-cell recruitment (Cxcl11, Cxcl10, Ccl5), and effective antigen presentation (Cd80, Cd86, and Cd40l; Supplementary Table S4). The trabectedin-mediated immune checkpoint molecule Ctla-4 upregulation was also enhanced by combination with anti-PD-1, suggesting that this pathway may represent another suitable partner for combination with trabectedin. Conversely, trabectedin-mediated Pd1 and Pdlg1 upregulation was not modified by anti-PD-1 antibody combination (Supplementary Fig. S5). However, Pd1 mRNA expression may be irrelevant, as the cell surface protein comprises the anti-PD-1 antibody target.

Despite the recent progress in targeted therapies and immunotherapies, treatment options for osteosarcoma remain anchored to surgery and conventional chemotherapy (3). Hence, there is an urgent need for the implementation of new, less toxic therapeutic approaches for likely combination with standard chemotherapy.

The involvement of different immune cell subsets in tumor development and progression (6), and their emerging role in the most effective standard (25) or targeted (26) therapeutic treatments for several tumor histotypes, suggest that immunomodulatory drugs represent a potentially appealing therapeutic option, although the data available for osteosarcoma are still limited.

To address the requirement for suitable immunocompetent preclinical models to assess the role of immune cells, in particular adaptive T cells, in osteosarcoma tumorigenesis, we have developed and extensively characterized, both in vitro and in vivo, syngeneic murine osteosarcoma cell lines that grow both subcutaneously and in the bone cavity, spontaneously metastasizing to the lungs in the latter case and therefore well resembling the human disease. We then used these transplantable osteosarcoma models to evaluate the efficacy of trabectedin against osteosarcoma primary and metastatic tumors. Trabectedin mechanisms of action are still poorly understood and appear not to be unique. In addition to inducing DNA damage, trabectedin efficacy is, in part, mediated by transcriptional interference through its binding to the DNA minor groove, which induces structural changes in DNA to block or alter transcription factor binding, thereby modifying the cell transcription profile. This action is chiefly relevant against soft-tissue sarcomas, particularly those characterized by the oncogenic expression of fusion proteins such as myxoid liposarcoma, where trabectedin blocks FUS-CHOP protein expression, allowing tumor cell differentiation into benign lipoblasts (20). A third mechanism has also been recently described involving the direct targeting of myelo/monocyitc cells, which in many histotypes support tumor growth (11).

From a clinical point of view, very few, heavily pretreated, OS patients have received trabectedin in clinical trials, with limited anti-tumor efficacy (27). Here we show, using multiple mouse osteosarcoma cell lines, that trabectedin exerts potent in vivo antitumor and antimetastatic activities, either following direct intravenous tumor cell injection or against spontaneous development after intra-bone injection. In mice, we observed that trabectedin induced cellular differentiation and osteoid matrix deposition. Accordingly, as neoplastic cells differentiated in vivo, they proliferated less.

To unveil the molecular basis for the differentiative effect on osteosarcoma cells, we assessed whether trabectedin altered transcription factor binding related to the cellular differentiation program, as described in myxoid liposarcoma (20) and Ewing sarcoma (21). Focusing on RUNX2, a master osteoblastogenesis-regulatory gene (28), we showed that trabectedin induces, both in in vitro cell lines and in in vivo tumors, Runx-2 transcription factor binding on OCN promoter, likely causing osteogenic differentiation, and on p21 promoter, further supporting trabectedin antiproliferative, prodifferentiative actions. As these are likely not the only promoters occupied by RUNX2 upon trabectedin treatment, further analysis such as by ChIP sequencing may reveal additional transcriptome modifications in neoplastic as well as surrounding accessory cells.

As trabectedin was reported to affect the myelo/monocytic compartment in the tumor microenvironment (11), we also analyzed whether changes were induced in osteosarcoma tumor–infiltrating leukocytes. We detected some reduction in the myelo/monocytic pool, albeit much less significant than in the reported tumor models (11). This may be explained by the marked myelo/monocytic compartment expansion triggered by the growth of such tumors, which was not observed in osteosarcoma, potentially indicating that trabectedin preferentially affects monocytes and macrophages that are pathologically expanded upon disease progression. Alternatively, macrophages infiltrating osteosarcoma tumors may differ from those in fibrosarcomas and Lewis lung carcinoma; for example, in DR5 expression, the TRAIL receptor expressed on monocytes/macrophages infiltrating these cancers where it critically mediates trabectedin-induced cell death. This aspect is under investigation.

Unlike other tumor histotypes, tumor-associated macrophages in osteosarcoma appear not to be related to aggressiveness and poor prognosis, but rather with better prognosis and less aggressive disease (8), as confirmed in clinical tissue samples of patients treated at Rizzoli Institute (protocol ISG-OS1; ref. 19). Consistent with this, our mouse model of well-differentiated and poorly aggressive osteosarcoma mOS13 exhibits greater macrophage and other immune cell infiltration than the more aggressive mOS69 model.

Despite the negligible effect on myelo/monocytic cells, trabectedin was very efficient in recruiting CD8 and CD4 T cells locally in mOS69 tumors; moreover, T-cell–related gene expression, such as Grzmb, Il7, Fas, Cxcl10, and Cxcl11, increased in in vivo mOS69 tumors from trabectedin-treated mice. These markers may be indicative of tumor microenvironment reshaping toward a potentially T lymphocyte–enriched, immune-responsive environment. However, results obtained in T-cell–deficient mice suggested that an adaptive T-cell response was irrelevant for trabectedin therapeutic activity. Analysis of tumor-infiltrating T-cell phenotype and activation state from trabectedin-treated mice showed increased PD-1 checkpoint inhibitor expression on CD8 T cells, compatible with their impaired function. This result provided the rationale for combining trabectedin with anti-PD-1 antibody. Accordingly, anti-PD-1 antibody administration starting after the first trabectedin treatment significantly enhanced tumor inhibition. In addition, the combination therapy further increased expression of many genes already upmodulated by trabectedin alone and associated with a more activated and "effector" infiltrating T-cell phenotype. On the other hand, according to the low T-cell infiltration of osteosarcoma tumors in untreated mice, the immune checkpoint blocker alone did not induce significant tumor growth inhibition.

The mechanism by which trabectedin induces PD-1 upregulation on T cells remains to be elucidated; however, it is currently under investigation, as it may provide useful information for new therapeutic combinations. Trabectedin treatment of naïve T cells in vitro elicited no significant modulation of PD-1 expression, suggesting that the effect is not direct but likely occurs locally in vivo through other mediators. In vivo, trabectedin treatment induces increased overall expression of IFNγ, which is known to induce PD-1 expression on T cells; however, directly treating mOS cell lines with trabectedin in vitro does not induce their IFNγ expression, further supporting the occurrence of a more complicated mechanism in vivo.

Our result of the poor efficacy of anti PD-1 alone appears to contrast with a recent study showing that blockade of the PD-1/PD-L1 axis is sufficient to reduce growth of the K7M2 osteosarcoma model (29); however, no clear data are shown regarding the overall immune landscape in this model or the number of infiltrating CD8 T cells. In a subsequent study, the authors reported that combining anti-PD-1 with anti-CTLA-4 antibody prevents tumor immune escape and leads to better disease control (30). This efficacy arises consequent to anti-PD-1 treatment upregulating CTLA-4 expression on infiltrating T cells, impairing their ability to mediate tumor rejection. These data are consistent with our observation of significantly increased CTLA-4 expression in in vivo tumor samples from mice treated with trabectedin alone and in combination with anti-PD-1 antibody, again suggesting the possibility of combining trabectedin with both immune checkpoint inhibitors. Furthermore, the expression of CTLA-4 on tumor-infiltrating CD8 T cells was recently identified as the single parameter having a statistically significant association with clinical response to anti PD-1 therapy and progression-free survival in a cohort of patients with metastatic melanoma (31). This CTLA-4^high CD8 T subset also expressed the highest level of PD-1, suggesting that the relative abundance of CTLA-4^high PD-1^high CD8 T cells may be used to predict response to anti PD-1 therapy.

The ability of trabectedin to recruit T cells at the tumor site could be exploited to convert "cold" tumors, such as osteosarcoma, into "hot" ones, richer in T cells (32, 33), making them amenable to checkpoint inhibitor–based immunotherapies. Association with PD-1 and likely CTLA-4 blockers may further offer a new opportunity for trabectedin to be effectively employed in the treatment of osteosarcoma, overcoming the disappointing results previously obtained, although in only a few, heavily pretreated, patients (27).

To the best of our knowledge, the combination of trabectedin with anti-PD-1 antibody has only been reported once previously, in a mouse model of ovarian cancer, where it could induce an antitumor T-cell response, reshaping, similarly to our work, the tumor microenvironment from an immunosuppressive to a stimulatory state (34). In contrast to present results, trabectedin alone did not appear to affect the overall number of adaptive T cells infiltrating the tumor site. This discrepancy may be explained by the different tumor histotype studied; that is, ovarian cancer grown in the peritoneum, from which the tumor-infiltrating leukocytes were obtained by peritoneal lavage.

Overall, our data extend the antitumor efficacy of trabectedin to osteosarcoma, a neoplasia for which an urgent need exists for additional, less toxic therapeutic approaches, and shed new light on its multitarget mechanisms of action. We showed, for the first time, the ability of this drug to induce T-cell recruitment/expansion at the tumor site, providing the foundation for combination therapies with immune checkpoint inhibitors, and indeed we demonstrated here the superior therapeutic activity of its combination with anti-PD-1 antibody over the sole trabectedin treatment We also reported the ability of trabectedin to promote tumor cell differentiation by inducing RUNX2 recruitment to OCN and p21 promoters, suggesting that its ability to interfere with, or modulate, gene transcription is not limited to "transcription-addicted" tumor histotypes, such as myxoid liposarcoma or Ewing sarcoma, but may also be involved in its antitumor activity in other tumors that do not express oncogenic fusion proteins.

No potential conflicts of interest were disclosed.

Conception and design: K. Scotlandi, M.P. Colombo, C. Chiodoni

Development of methodology: C. Ratti, M.C. Manara, C. Tripodo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Ratti, L. Botti, V. Cancila, S. Galvan, I. Torselli, C. Garofalo, L. Bongiovanni, A. Burocchi, M. Parenza, B. Cappetti, S. Sangaletti

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Ratti, V. Cancila, S. Galvan, I. Torselli, C. Garofalo, M.C. Manara, L. Bongiovanni, C.F. Valenti, C. Tripodo, K. Scotlandi, C. Chiodoni

Writing, review, and/or revision of the manuscript: C. Tripodo, K. Scotlandi, M.P. Colombo, C. Chiodoni

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Cancila, L. Bongiovanni, C.F. Valenti

Study supervision: K. Scotlandi, M.P. Colombo, C. Chiodoni

Other (co-last author): M.P. Colombo

The authors thank Ivano Arioli for technical support in the in vivo experiments; Ester Grande for administrative support, the Unit of Medical Statistics, Biometry and Bioinformatics, the Immunohistochemistry Facility for evaluation of additive versus synergistic effect; and the Animal Facility where the in vivo experiments were performed.

This work has been supported by AIRC IG grant (17261) to C. Chiodoni, AIRC IG grant (10137) to M.P. Colombo, AIRC IG grant (14049) to K. Scotlandi and the Italian Ministry of Health. I. Torselli has been recipient of a FIRC fellowship.

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.