FOS3D Toolkit Uses Fluorescence to Characterize 3D Osteosarcoma Models

Researchers have developed a new tissue-engineering platform that allows osteosarcoma—the most common primary bone cancer in adolescents—to be grown, watched, and drug-tested in the laboratory in real time, without destroying the sample. The system, called FOS3D (Fluorescent Osteosarcoma 3D model), combines a mechanically tunable gelatin-based hydrogel with stably glowing, GFP-labeled cancer cells, giving scientists a continuous, quantitative window into how tumor cells proliferate, organize themselves, and respond to chemotherapy inside a three-dimensional environment that closely mimics the soft, collagen-rich stroma of bone.

The need for such a model is pressing. Despite decades of research, the five-year event-free survival rate for non-metastatic osteosarcoma has remained stuck between 60 and 70 percent for more than four decades, dropping to roughly 25 percent for metastatic cases. Chemoresistance is reported in up to 77 percent of patients under current treatment regimens, and only about 7 percent of oncological drugs entering clinical trials ultimately receive regulatory approval. Animal models, while providing a systemic context, suffer from interspecies differences in bone biology and tumor evolution, high costs, ethical concerns, and limited experimental controllability. Conventional two-dimensional cell cultures, meanwhile, strip away the extracellular matrix architecture, mechanical cues, and spatial constraints that shape real tumor behavior. FOS3D is designed to occupy the sweet spot between these extremes: physiologically relevant, controllable, and scalable.

At the heart of the platform is GelMA—gelatin methacryloyl—a hydrogel derived from denatured collagen that can be crosslinked into a stable polymer network using a photo-initiator (lithium phenyl-2,4,6-trimethylbenzoylphosphinate, or LAP) and precisely timed exposure to 405-nanometer violet light. Because GelMA retains collagen-derived biochemical signals such as arginine-glycine-aspartic acid (RGD) sequences, cells can adhere to it as they would to natural tissue, while its stiffness and viscoelastic behavior can be dialed up or down by adjusting polymer concentration and light exposure. In their experiments, the researchers cast 35-microliter suspensions of MG-63-GFP human osteosarcoma cells into 3.3-cubic-millimeter PDMS molds, varying GelMA concentration from 6 to 10 percent and crosslinking duration from one to three minutes. The resulting compressive moduli ranged from approximately 6.2 to 50.1 kilopascals—a span that overlaps with the reported stiffness of bone marrow stroma (roughly 0.25 to 24.7 kilopascals) and tumor-associated fibrotic regions, making the constructs a plausible mechanical stand-in for the environment a tumor actually encounters in bone.

The team’s second key ingredient is fluorescence. Rather than using short-lived cytoplasmic dyes, whose signal dilutes with every cell division, the researchers virally transduced osteosarcoma cells with a GFP reporter gene under the control of the highly active SV40 promoter, producing constitutive cytoplasmic fluorescence that persists—and grows—as cells proliferate. This means the same hydrogel can be scanned repeatedly over days or weeks. Using a plate reader, each well is interrogated with a 30-by-30-point scan matrix at a fixed focal height, generating both a numerical intensity value (measured at excitation/emission of 470–15/520–20 nanometers) and a topographical fluorescence map of the entire construct after background subtraction at a threshold of 1,000 arbitrary units.

Crucially, the researchers did not simply assume GFP brightness equals cell number—they proved it. In both 2D culture and 3D hydrogels, GFP signal scaled linearly with cell number, achieving a correlation coefficient of r² = 0.9979 in 2D. The team then cross-validated fluorescence against two orthogonal, destructive benchmarks: DNA quantification and metabolic activity assays (CellTiter-Blue). Linear regression and Pearson’s correlation analyses confirmed strong agreement between GFP intensity and both gold-standard measures in 2D and 3D formats. In longitudinal tests, untreated samples showed steadily rising fluorescence from day 1 through day 7, while samples treated with the cytostatic drug paclitaxel showed suppressed fluorescence growth—an early demonstration that the readout captures genuine drug-induced cytostasis rather than optical noise.

The fluorescence-based approach also revealed striking differences in cell behavior between culture formats. By day 7, cells grown in conventional 2D monolayers displayed the uniform, elongated, spindle-like morphology typical of cells spreading on a rigid plastic surface. In contrast, cells embedded within 6 percent GelMA hydrogels adopted a heterogeneous, tumor-like phenotype—a mixed population of rounded and elongated cells—reflecting matrix confinement, three-dimensional cell–ECM interactions, and spatially regulated growth characteristic of solid tumors. Importantly, these comparisons were made under matched theoretical cell–cell spacing, with average center-to-center distances calculated assuming homogeneous distribution across either a 2D plane or a 3D volume, ensuring the comparison was fair.

To confirm that the 3D environment genuinely changes tumor biology—and not just appearance—the researchers coupled fluorescence tracking with RT-qPCR transcriptional profiling, which verified that genes were expressed differently in 3D culture relative to 2D monolayers. Immunohistochemistry further confirmed expression of ABCB1, also known as P-glycoprotein, a well-established prognostic biomarker and drug-efflux pump in osteosarcoma. Together, these molecular analyses demonstrate that FOS3D is not merely a prettier container for cells but a system in which clinically relevant tumor phenotypes emerge.

The platform’s real payoff came in drug testing. The researchers exposed two GFP-labeled osteosarcoma cell lines—MG-63-GFP and Saos-2-GFP—to frontline chemotherapeutics cisplatin (CDDP) and doxorubicin (DOX) in both 2D and 3D formats. The choice of drugs demanded careful optical controls: doxorubicin is intrinsically fluorescent and can quench GFP signals depending on concentration and environment, potentially confounding intensity-based readouts. By calibrating spectra and concentrations, and by benchmarking fluorescence against metabolic and DNA measurements alongside imaging confirmation, the team distinguished true cytotoxicity and cytostasis from optical artifacts. This time-resolved capability matters because fluorescence tracking captures the full kinetics of drug response—onset, lag phase, and rebound—rather than a single endpoint snapshot, and because matrix confinement and altered mechanosensing in 3D can shift drug sensitivity relative to flat cultures.

The researchers ultimately selected 6 percent GelMA crosslinked for 2 minutes as their standard formulation, balancing mechanical robustness for handling with an environment permissive to tumor cell expansion. Notably, across all tested stiffnesses, metabolic activity remained statistically comparable, indicating that cell viability is preserved throughout the mechanical design space, while DNA content showed modest sensitivity to polymer density at shorter crosslinking durations—an effect that vanished at longer exposure times. Because the platform operates in standard multi-well plates, it is inherently scalable: large numbers of conditions can be analyzed in parallel, making it dramatically more cost-effective than animal studies and well suited to high-throughput drug screening and combinatorial testing. It is also compatible with light-sheet microscopy, enabling whole-construct volumetric imaging that resolves cell clustering and morphology in three dimensions over time.

The FOS3D toolkit, published in Advanced Science, offers researchers studying rare and stubborn cancers a way to iterate rapidly: build complexity, test drugs, watch the response unfold live, and then preserve the same sample for downstream molecular analysis. For a disease whose clinical outcomes have barely moved in forty years, a scalable, non-destructive, physiologically relevant model may be exactly the accelerator drug development has been waiting for.