Agreement between two pathologists

Abstract Introduction: Osteosarcoma is the most common primary bone cancer in adolescents and young adults. It is characterized by high heterogeneity and a hostile bone microenvironment shaped by osteoclasts, endothelial cells, and immunosuppressive myeloid populations, underscoring the need for innovative models to evaluate novel therapeutic strategies. Objective: This study aims to develop and characterize novel three-dimensional (3D) culture models that recapitulate the complexity and the microenvironment of osteosarcoma. Methods: Tumor samples from patient-derived xenografts (PDXs) established from patients with osteosarcoma at relapse or treatment failure were processed to generate 3D tumor cultures. Tumor tissues were mechanically and enzymatically dissociated into single-cell suspensions and seeded into round-bottom plates with very low adhesion to promote spontaneous formation of 3D tumoroids. Culture conditions were optimized using medium supplemented with specific growth factors to support cell viability and proliferation. Morphology and growth of the tumoroids were monitored over time. The resulting 3D models were characterized by histology (H&E and specific immunostaining like SATB2, SPP1, SOX9⋯), and by RNA sequencing and Whole exome sequencing (WES) to assess transcriptional stability relative to the parental primary/PDX tumors. In addition, these 3D tumoroid models were used for drug testing to evaluate therapeutic responses and identify potential treatment sensitivities. Results: We established five 3D osteosarcoma models from PDX with an establishment rate of 83%. The tumoroids formed spontaneously under low-adhesion conditions within approximately 7 to 30 days, (depending on the derived sample) and remained viable and morphologically stable during this culture period. Histological analysis revealed that the 3D tumoroids recapitulated key features of the parental PDX tumors and the patient tumors, including cellular heterogeneity, hypoxic regions, and osteoid-like matrix deposition. RNA sequencing and drug testing studies to evaluate therapeutic responses are currently ongoing. Conclusions: Patient-derived 3D osteosarcoma models faithfully preserve tumor heterogeneity and microenvironmental features, providing a robust and scalable system for preclinical drug testing. This approach holds a strong potential to accelerate the development of personalized combination therapies targeting both tumor cells and their supportive bone microenvironment in osteosarcoma. Citation Format: Margaux Chantoiseau, Pierre Khneisser, Birgit Geoerger, Nathalie Gaspar, Maria Eugénia Marques da Costa, Antonin Marchais, . Development and characterization of 3D osteosarcoma models recapitulating tumor heterogeneity and bone microenvironment [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 6170.

CosMx protein panel showing stainings in MGC

Immunohistochemistry for representative Visium markers

Cancer metabolism depends on multifaceted mechanisms, including bidirectional inter-organelle communication between mitochondria and the nucleus, facilitating cellular adaptation at the transcriptomic, proteomic, and metabolomic levels. The mitochondrial protein complex composed of apoptosis-inducing factor (AIF) and coiled-coil-helix-coiled-coil-helix domain-containing protein 4 (CHCHD4) is essential for this mitochondrio-nuclear communication. The AIF/CHCHD4 complex mediates the mitochondrial import of cysteine-enriched nuclear gene-encoded proteins, thereby adapting the mitochondrial proteome to cellular energy demands. Here, we report the discovery of M30-E05, a compound that binds to the NADH pocket of AIF, preventing its dimerization and disrupting the AIF/CHCHD4 complex, as demonstrated by molecular docking and gel electrophoresis analysis of mitochondrial AIF/CHCHD4 substrate expression. In cancer cells, M30-E05 reduces the expression of nuclear gene-encoded mitochondrial proteins such as AIF, CHCHD4, cytochrome c oxidase copper chaperone (COX17), and Mitochondrial calcium uptake 1 (MICU1). In addition, M30-E05 fragments the mitochondrial network and impairs mitochondrial respiration, causing profound alterations, particularly in glucose, lipid, and amino acid metabolism, as revealed by kinetic measurements of oxygen consumption and mass spectrometric metabolomics. Importantly, M30-E05 significantly reduces the viability of a human adult and pediatric osteosarcoma cancer cell panel, including those from patient-derived xenografts (PDX) of osteosarcomas, and induces apoptosis. When orally administered for two weeks to immunodeficient NSG mice, M30-E05 inhibited tumor growth in a subcutaneous PDX xenograft model without apparent toxicity. We anticipate that M30-E05, as a first-in-class metabolic inhibitor, could serve as the lead compound for a new class of antineoplastic agents.

Spatial omics technologies map molecular information within intact tissue architecture, revealing how cellular organization and interactions shape tumor biology, therapeutic responses, resistance, and relapse. Advances in artificial intelligence and computational pathology are bridging research discovery and clinical practice, enabling prognostic extraction from routine histology and cost-effective molecular inference. Nevertheless, barriers including high costs, protocol complexity, and lack of standardized workflows remain. We outline how these converging fields can deliver translational value in oncology and identify key enablers for clinical adoption.

Quantification of TILs, CD68-positive mononuclear macrophages and tumor grade

CosMx 64-plex human immuno-oncology protein panel

Representative video of an isolated MGC in oral cavity SCC from GR cohort

Abstract Osteosarcoma is a malignant bone tumor with a high risk of metastatic relapse and poor outcomes due to primary and acquired chemoresistance. This highlights the medical need to develop effective targeted approaches to overcome chemoresistance. Recent studies have revealed the roles of metabolic reprogramming and mitochondria-nucleus crosstalk in osteosarcoma progression, indicating the potential of these cellular processes as therapeutic targets. The complex formed by mitochondrial apoptosis-inducing factor (AIF) and coiled-coil-helix-coiled-coil-helix domain-containing protein 4 (CHCHD4) orchestrates the import and oxidative folding of cysteine-rich, nuclear-encoded proteins, thereby regulating key mitochondrial functions and metabolism. Here, we identified mitoxantrone as an inhibitor of the AIF/CHCHD4 mitochondrial import machinery and revealed a new mitoxantrone-induced metabolic vulnerability in some osteosarcoma cell line models, characterized by intracellular glutamine accumulation and an increase in nucleotide synthesis. As a result, synergy was found between mitoxantrone and the glutaminase inhibitor telaglenastat in both in vitro and in vivo osteosarcoma models. Collectively, our findings position the AIF/CHCHD4 complex as a druggable therapeutic target and provide a combination strategy for mitoxantrone/telaglenastat treatment to overcome metabolic adaptations and chemoresistance in osteosarcoma.

Despite well-recognized biological heterogeneity, osteosarcoma has been treated as a single disease for over four decades with minimal improvement in survival. Clinical features are inadequate for risk stratification, and no molecular classifiers guide therapy. An international working group evaluated candidate prognostic biomarkers for clinical translation. Pre-treatment circulating tumor DNA is positioned for clinical implementation, while additional classifiers warrant prospective validation. This work establishes a path to risk-adapted, biologically informed treatment.

This table provides the DEG for the seven main clusters of the Visium experiment

Performances of the deep-learning cell detection model in TCGA cohorts (oral cavity, uterine cervix and larynx SCC) and GR cohorts (oral cavity SCC)

10032 Background: Osteosarcoma exhibits profound genomic complexity, complicating biomarker discovery. Although advances in multi-omics profiling have improved biological understanding, genomic data remain fragmented across heterogeneous cohorts, and actionable alterations are rarely reported. Methods: A systematic review identified 20 studies published between 2005 and 2022, spanning heterogeneous sequencing technologies (whole-genome/exome/RNA sequencing, methylome, and targeted gene panels) and reporting genetic drivers and/or potentially actionable mutations. Mutation data were manually extracted from the published results (e.g., tables, figures, oncoplots) and integrated into a binary Mutation Annotation Format - like matrix (1,041 samples, 471 genes) to provide a comprehensive overview of variants reported to date. Results: 1058 patients were included, with tissue derivation (primary, metastatic, recurrent) described in about 10% of the cases. 3816 germlines or somatic SNVs, CNVs or structural alterations were reported in 471 genes. TP53 (40%), RB1 (18%), and CDKN2A/B (17%) were the most frequently altered genes, alongside with recurrent changes in cell cycle regulators ( CCND family) and angiogenesis-related genes ( VEGFA ). Potentially actionable variants were less common: MYC (13%, all amplifications), CDK4 (10%), PDGFRA (8%); single agent activity against such targets is either lacking or not yet available. Potentially actionable epigenetic targets were rarely identified: TET2 (0.2%), IDH2 (0.3%), DNMT3A (0.7%). TCGA-based pathway analysis revealed consistent involvement of TP53, cell cycle, RTK-RAS, PI3K, and MYC signaling. Co-occurrences detected were TP53 with MAP2K4 , RICTOR , PTPRD and RB1 genes, and structural rearrangements affecting RTK-RAS and PI3K pathways. There was relative exclusivity between mutations of RB1 and CDK4 . Conclusions: The analysis identifies a small number of recurrently mutated genes and a large number of rarely affected, though potentially actionable, mutations, encoding proteins involved in cell surface, nuclear and epigenetic interactions. Lack of harmonization in data processing, mutation calling and definition of ‘over’ and ‘under’ expression, together with absent descriptions of tissue derivation hamper effective large scale analysis and phenotypic subgrouping. TP53 mutation co-occurrences such as RICTOR or RB1 were also identified, with potential therapeutic implications including WEE1 and mTOR inhibitors. Tumors with overactive CDK4/6 typically have functional RB1 and might be sensitive to CDK inhibitors. Harmonization of multi-omic methods and collaborative international analyses are essential to advance precision medicine in osteosarcoma and to inform future trials.

The fourth Paediatric Therapeutic Development Workshop focused on osteosarcoma, the most common primary bone cancer in children and young adults. Current treatment of osteosarcoma comprises surgery and chemotherapy. Outcome has shown very little improvement over the last four decades and there are substantial unmet needs including improving survival, especially in metastatic or relapsed disease, and reducing treatment toxicity. There is a lack of new therapeutics in osteosarcoma, with the evaluation of new treatments challenged by complex biology and variation in chemotherapy standard of care regimens. An antibody-drug conjugate (ADC) targeting LRRC15 with an osteosarcoma-relevant payload is a promising therapeutic approach. Small molecule inhibitors and degraders targeting SMARCAL1 should be considered as a very high priority, as it is a target applicable to several high unmet need malignancies. Preclinical evaluation of KIF18A in osteosarcoma models is required. An early phase study of an eIF4A1 inhibitor is warranted. Further preclinical validation of RUNX2 using PROTACs or ADCs is required. Targeting MYC, either indirectly or directly is a high priority. Osteosarcomas have significant genomic instability and impaired DNA repair mechanisms and efforts are ongoing to exploit this vulnerability therapeutically. The goal is that these therapeutic developments will improve survival for patients with osteosarcoma.

Clinical Tables and Statistical Analyses

Association between keratin, MGC and clinical outcome

This table provides the DEG between MGC spots and non-MGC spots

Representative video of granulomas made of multiple MGC in oral cavity SCC from GR cohort