Integrated Transcriptomic, Proteomic, and Pharmacologic Profiling of 2D and 3D Patient‐Derived GCTB Cell Lines Reveals Culture‐Dependent Drug Response Determinants

ABSTRACT

Giant cell tumor of bone (GCTB) is an intermediate bone neoplasm defined by recurrent H3F3A mutations and limited systemic treatment options beyond denosumab. Patient‐derived cancer cell lines (PDCs) offer a scalable platform for mechanistic studies and therapeutic discovery, yet the extent to which culture dimensionality alters baseline molecular states and drug response in GCTB remains unclear. Here, we performed integrated transcriptomic, proteomic, and pharmacologic profiling of thirteen patient‐derived GCTB cell lines cultured under two‐dimensional (2D) monolayer and three‐dimensional (3D) spheroid conditions. RNA sequencing and data‐independent acquisition (DIA)–based quantitative proteomics were conducted on paired cultures, and drug sensitivity was assessed using a panel of 221 anticancer agents. 3D culture reproducibly induced compact spheroid formation across all cell lines and was accompanied by broad remodeling of gene expression, protein abundance, and drug‐response profiles. Unsupervised analyses consistently demonstrated that samples clustered primarily by culture condition rather than by cell‐line identity at both the transcriptome and proteome levels. Although global trends were shared, a substantial fraction of molecules showed RNA–protein discordance, indicating that transcriptomic changes alone do not fully explain culture‐dependent functional remodeling. Pathway analyses highlighted enrichment of extracellular matrix–related processes, stress‐response programs, and metabolic regulation in 3D cultures, with several features more prominent at the protein level. Functionally, 3D culture generally reduced sensitivity to many agents while preserving compound‐dependent vulnerabilities. These results establish culture dimensionality as a key determinant of therapeutic susceptibility in GCTB PDCs and support incorporating proteome‐informed 3D models into translational pipelines to prioritize clinically relevant drug candidates and biomarkers.

1. Introduction

Giant cell tumor of bone (GCTB) is a primary osteolytic neoplasm composed of neoplastic stromal cells and numerous osteoclast‐like multinucleated giant cells. The neoplastic stromal component is defined by recurrent H3F3A mutations, which are recognized as the molecular hallmark of the disease [1, 2]. GCTB is classified as an intermediate malignancy, and it can occasionally exhibit clinically aggressive behavior, including local recurrence and, in a subset of patients, pulmonary metastasis. The tumor most frequently arises in the epiphysis of long bones in young adults. Although denosumab has been introduced as a treatment option, effective systemic chemotherapy remains limited, and its long‐term efficacy and safety continue to be debated [3, 4]. These clinical features highlight the need for deeper biological investigation and improved experimental models.

Patient‐derived cancer cell lines (PDCs) represent a practical and reproducible platform for studying tumor biology and for evaluating candidate therapeutics. Over the past several years, our group has established a series of GCTB PDCs directly from surgical specimens, and these models have been characterized and reported in detail [5, 6, 7, 8, 9, 10, 11, 12, 13]. These PDCs make it possible to perform in vitro studies based on materials that closely reflect the biological properties of patient‐derived tumors. While two‐dimensional (2D) monolayer cultures are useful for functional assays and high‐throughput drug screening, they do not fully reproduce essential tumor‐related properties such as cell–cell communication, extracellular matrix organization, and hypoxic gradients. Three‐dimensional (3D) spheroid or organoid models have therefore been proposed as more physiologically relevant systems that may better approximate aspects of the tumor microenvironment [14, 15]. However, the molecular characteristics of 3D cultures and their relationship to drug responses have not been systematically investigated in GCTB. To date, no study has compared 2D and 3D cultures of GCTB PDCs using transcriptomic and proteomic profiling together with large‐scale drug sensitivity testing. Such information is essential for determining how model systems should be selected for mechanistic studies and for model‐based drug discovery.

Transcriptomic analysis provides information on gene expression at the mRNA level, whereas proteomic profiling directly measures protein abundance, which reflects the functional molecules that carry out cellular processes. Because mRNA levels do not necessarily correspond to protein abundance—owing to regulation at multiple steps including mRNA stability, translation efficiency, and protein turnover—integrating transcriptomic and proteomic data enables a more comprehensive understanding of cellular states [16, 17]. Examining both layers allows the identification of concordant and discordant expression patterns that may reveal regulatory mechanisms underlying culture‐dependent phenotypes. This complementary perspective is particularly valuable when assessing how culture conditions influence molecular features that may be relevant to drug responses.

In this study, we performed a comparative analysis of 13 patient‐derived GCTB cell lines cultured under both 2D and 3D conditions. For each cell line, we conducted RNA sequencing and data‐independent acquisition (DIA) proteomic profiling, and evaluated drug sensitivity using a panel of 221 anticancer agents. By integrating these datasets, we aimed to clarify how baseline molecular profiles differ between 2D and 3D cultures and how these differences correspond to culture‐dependent variations in drug response. We expect that our findings will contribute to the effective use of GCTB cell‐line models in future biological and therapeutic research.

2. Materials and Methods

2.1. Cell Lines and Culture Conditions

A total of 13 patient‐derived GCTB cell lines were used in this study (Table 1). All cell lines were originally established from surgically resected GCTB tumors, and have been fully characterized in previous reports [5, 6, 7, 8, 9, 10, 11, 12, 13]. These cell lines have been deposited in public cell banks and are available for academic research. We clarified that the models used in this study are patient‐derived GCTB cell lines that have been stably maintained and characterized in previous publications, rather than primary short‐term cultures [5, 6, 7, 8, 9, 10, 11, 12, 13].

TABLE 1.

List of GCTB cell lines in this study.

	Cell line name	Gender	Age	Original site	References (PMID)
1	NCC‐GCTB1‐C1	Male	53	distal radius	[5]
2	NCC‐GCTB3‐C1	Female	20	proximal tibia	[6]
3	NCC‐GCTB4‐C1	Male	52	proximal tibia	[7]
4	NCC‐GCTB5‐C1	Male	52	Knee	[8]
5	NCC‐GCTB6‐C1	Female	38	distal femur	[9]
6	NCC‐GCTB7‐C1	Male	31	Radius	[9]
7	NCC‐GCTB8‐C1	Male	32	proximal tibi	[10]
8	NCC‐GCTB9‐C1	Female	33	distal tibia	[10]
9	NCC‐GCTB10‐C1	Male	27	proximal tibia	[11]
10	NCC‐GCTB11‐C1	Female	24	proximal tibia	[12]
11	NCC‐GCTB12‐C1	Female	23	proximal tibia	[12]
12	NCC‐GCTB14‐C1	Female	24	distal femur	[13]
13	NCC‐GCTB15‐C1	Male	63	distal femur	[13]

2.2. Two‐Dimensional (2D) Culture

GCTB cell lines were cultured in DMEM/F‐12 medium (Thermo Fisher Scientific Inc., Waltham, MA, USA) supplemented with Sarcoma Medium; GlutaMAX (Thermo Fisher Scientific Inc.), 5% heat‐inactivated fetal bovine serum (FBS; Thermo Fisher Scientific Inc.), 10 µM Y‐27632 (ROCK inhibitor; Selleck Chemicals, Houston, TX, USA), 10 ng/mL basic fibroblast growth factor (bFGF; Sigma‐Aldrich Co. LLC, St. Louis, MO, USA), 5 ng/mL epidermal growth factor (EGF; Sigma‐Aldrich Co. LLC), 5 µg/mL insulin (Sigma‐Aldrich Co. LLC), and 0.4 µg/mL hydrocortisone (Sigma‐Aldrich Co. LLC). The medium was additionally supplemented with 100 µg/mL penicillin and 100 µg/mL streptomycin (Nacalai Tesque Inc., Kyoto, Japan).

For 2D culture, cell suspensions were seeded onto 10 cm tissue culture dishes (Nunc Easydish; Thermo Fisher Scientific Inc.) and maintained at 37°C in a humidified incubator with 5% CO₂. Culture medium was refreshed every 2–3 days by replacing approximately half of the medium volume with fresh medium. Cells were allowed to grow until approximately 80% confluence and were then used for passaging, drug sensitivity assays, RNA and protein extraction, or preparation for three‐dimensional (3D) culture.

2.3. Three‐Dimensional (3D) Culture

For 3D culture, GCTB cell lines grown under 2D conditions were washed once with phosphate‐buffered saline without calcium and magnesium (D‐PBS (−), 1×; Nacalai Tesque Inc., Kyoto, Japan) and dissociated using 0.5 g/L trypsin and 0.53 mmol/L EDTA solution with phenol red (Cat. No. 32778‐34; Nacalai Tesque Inc.). Cells were collected primarily as small cell aggregates, resuspended in Sarcoma Medium, and seeded into ultra‐low attachment cultureware.

Specifically, cells were plated onto ultra‐low attachment 6‐cm dishes (PrimeSurface MS‐90600; Sumitomo Bakelite Co., Ltd., Tokyo, Japan), V‐bottom 96‐well ultra‐low attachment plates (PrimeSurface MS‐9096 V; Sumitomo Bakelite Co., Ltd.), or round‐bottom 384‐well ultra‐low attachment plates (Cat. No. 3830; Corning Inc., Glendale, AZ, USA). Cells were maintained under the same incubation conditions as 2D cultures (37°C, 5% CO₂). Spheroid cultures were used for downstream analyses, including FFPE sample preparation, drug sensitivity assays, and RNA and protein extraction, within one week after seeding.

2.4. Drug Sensitivity Assay

For drug sensitivity assays, GCTB cells grown under 2D conditions were harvested using 0.5 g/L trypsin and 0.53 mmol/L EDTA solution with phenol red (Cat. No. 32778‐34; Nacalai Tesque Inc., Kyoto, Japan). Cells were counted and seeded at a density of 5 × 10³ cells per well. For 2D assays, cells were plated into flat‐bottom 384‐well plates (Cat. No. 781186; Greiner Bio‐One Co., Tokyo, Japan). For 3D assays, cells were seeded at the same density into ultra‐low attachment round‐bottom 384‐well plates (Cat. No. 3830; Corning Inc., Glendale, AZ, USA). After seeding, cells were incubated for 24 h at 37°C in a humidified atmosphere containing 5% CO₂ to allow cell attachment (2D) or spheroid formation (3D).

Subsequently, a panel of 221 anticancer agents, including U.S. Food and Drug Administration (FDA)‐approved drugs (Selleck Chemicals, Houston, TX, USA), was dispensed into each well in duplicate using an automated liquid handling system (Bravo; Agilent Technologies Japan, Tokyo, Japan). The fixed concentration of 10 µM was selected for this exploratory screen to ensure the detection of a broad range of potential therapeutic vulnerabilities across diverse chemical classes, although we acknowledge that single‐concentration testing may not capture the full potency profile of each agent. The list of compounds is provided in Table S1. Following drug addition, cells were maintained for an additional 72 h under standard culture conditions (37°C, 5% CO₂).

Cell viability was assessed using different assays optimized for each culture system. Due to technical limitations in reagent penetration and signal linearity in spheroid cultures, different viability assays optimized for each culture format were employed. For 2D cultures, cell viability was measured by absorbance at 450 nm using the Cell Counting Kit‐8 (Dojindo Molecular Technologies, Kumamoto, Japan). For 3D cultures, viability was evaluated using PrestoBlue Cell Viability Reagent (Thermo Fisher Scientific Inc., Waltham, MA, USA), and fluorescence was measured at 560/590 nm. Cell viability in each well was calculated relative to the dimethyl sulfoxide (DMSO)‐treated control wells.

2.5. Hematoxylin and Eosin (H&E) Staining

2.5.1. H&E Staining of 2D‐Cultured Cells

For histological evaluation of 2D cultures, GCTB cells were harvested using 0.5 g/L trypsin and 0.53 mmol/L EDTA solution with phenol red. Cells were seeded at a density of 5 × 10³ cells per well in duplicate onto 8‐well slide chambers (Cat. No. 192‐008; WATSON Co., Tokyo, Japan) and cultured at 37°C in a humidified atmosphere containing 5% CO₂. After 72 h of culture, wells were washed with D‐PBS and cells were fixed with 4% paraformaldehyde (PFA) for 10 min at room temperature. Following fixation, cells were washed again with D‐PBS and stained with Tissue‐Tek Hematoxylin 3G (Cat. No. 9131‐4P; Sakura Finetek Japan Co., Tokyo, Japan) according to the manufacturer's recommended protocol for hematoxylin and eosin (H&E) staining.

2.5.2. H&E Staining of 3D‐cultured Spheroids

For 3D cultures, GCTB cells harvested from 2D cultures were resuspended and seeded at a density of 1 × 10⁴ cells per well into ultra‐low attachment V‐bottom 96‐well plates (PrimeSurface MS‐9096 V; Sumitomo Bakelite Co., Ltd.) (n = 6). Cells were cultured at 37°C with 5% CO₂ to allow spheroid formation. After 72 h, spheroids were washed with D‐PBS and fixed with 4% PFA overnight at 4°C. Fixed spheroids were dehydrated through a graded ethanol series and subsequently cleared with xylene, followed by paraffin infiltration and embedding.

Paraffin blocks were sectioned at a thickness of 4 µm using a microtome, and sections were mounted onto glass slides. Paraffin sections were deparaffinized with xylene and rehydrated through graded ethanol prior to staining. H&E staining was performed using Tissue‐Tek Hematoxylin 3G following the manufacturer's recommended protocol.

2.6. Transcriptomic Study

2.6.1. RNA Extraction and Sample Preparation

For RNA sequencing, GCTB cells grown under 2D conditions were harvested using 0.5 g/L trypsin and 0.53 mmol/L EDTA solution with phenol red. Cells were counted and seeded at a density of 1.5 × 10⁴ cells per well into flat‐bottom 96‐well plates (Cat. No. 167008; Thermo Fisher Scientific Inc., Waltham, MA, USA) or ultra‐low attachment V‐bottom 96‐well plates (PrimeSurface MS‐9096 V; Sumitomo Bakelite Co., Ltd.), corresponding to 2D and 3D culture conditions, respectively. Cells were cultured overnight at 37°C in a humidified atmosphere containing 5% CO₂.

After incubation, cells were washed twice with PBS directly in the plates. Cells were then resuspended in 57 µL of PBS per well, followed by the addition of 171 µL of QIAzol Lysis Reagent (QIAGEN K.K., Tokyo, Japan) for cell lysis. Cell lysates were collected and shipped to Zeon BioSolutions Inc. (Hyogo, Japan), where total RNA extraction and purification were performed using TRIzol reagent according to the manufacturer's protocol.

2.6.2. RNA Quality Assessment

RNA purity was assessed by spectrophotometric measurement of the A260/280 ratio, and RNA integrity was evaluated using an Agilent Bioanalyzer or TapeStation system. Only RNA samples with an RNA integrity number (RIN) greater than 8.0 were used for downstream analyses. RNA concentration and RNA integrity were applied as the primary quality‐control criteria for RNA sequencing. For library preparation, 20 ng of total RNA per sample was used as starting material.

2.6.3. RNA‐seq Library Preparation and Sequencing

RNA‐seq libraries were prepared using the Quartz‐Seq2 protocol [18]. Libraries were multiplexed and sequenced on an Illumina NextSeq platform to generate paired‐end reads of 101 bp. The mean sequencing depth was 2.16 ± 1.83 million reads per sample (mean ± SD). Quartz‐Seq2 is designed for low‐input transcriptome profiling and has been shown to provide robust relative expression quantification at modest sequencing depth. Quality control of raw sequencing data was performed using FastQC to evaluate read quality, base composition, and potential adapter contamination prior to downstream bioinformatic analyses. Transcript abundance was normalized as transcripts per million (TPM) to allow for comparison across samples.

2.7. Proteomic Study

2.7.1. Protein Extraction and Peptide Preparation

GCTB cells grown under 2D conditions were harvested using 0.5 g/L trypsin and 0.53 mmol/L EDTA solution with phenol red. Cells were counted and seeded at a density of 1 × 10⁶ cells per dish onto 10 cm tissue culture dishes (Nunc Easydish; Thermo Fisher Scientific Inc., Waltham, MA, USA) for 2D culture or ultra‐low attachment 6‐cm dishes (PrimeSurface MS‐90600; Sumitomo Bakelite Co., Ltd.) for 3D culture. Cells were cultured for 48 h at 37°C in a humidified atmosphere containing 5% CO₂.

For 2D cultures, cells were washed twice with D‐PBS, followed by the addition of ice‐cold 10% trichloroacetic acid (TCA) and incubation on ice for 30 min to precipitate proteins. Cells were then collected using a cell scraper into 1.5 mL tubes, washed once with ice‐cold D‐PBS, centrifuged, and the supernatant was removed. The resulting pellets were stored at −80°C until further processing.

For 3D cultures, spheroids were collected directly into 1.5 mL tubes, washed twice with D‐PBS, and treated with ice‐cold 10% TCA for 30 min on ice. Subsequent handling, including washing, centrifugation, and storage at −80°C, was performed using the same procedure as for 2D samples.

Frozen cell pellets were solubilized in SDS‐containing lysis buffer and centrifuged at 15000 × g for 5 min. The supernatants containing soluble proteins were collected, and protein concentrations were determined using the BCA Protein Assay Kit (Thermo Fisher Scientific Inc.).

Proteins were reduced with dithiothreitol (DTT) and alkylated with iodoacetamide (IAA), followed by purification using the SP3 (single‐pot solid‐phase–enhanced sample preparation) method. Proteins were subsequently digested overnight at 37°C using Trypsin/Lys‐C Mix, Mass Spec Grade (Promega, Madison, WI, USA). After digestion, peptides were desalted and eluted using GL‐Tip SDB (GL Sciences, Tokyo, Japan), and peptide concentrations were determined using a Qubit fluorometer (Thermo Fisher Scientific Inc.).

2.7.2. Liquid Chromatography–Tandem Mass Spectrometry and Data Analysis

For mass spectrometric analysis, 100 ng of peptides were loaded onto an Ultimate 3000 RSLCnano system (Thermo Fisher Scientific Inc.). Peptides were first trapped on a C18 trap column (Acclaim PepMap 100 C18, 100 µm × 2 cm, 5 µm; Thermo Fisher Scientific Inc.) and subsequently separated on an analytical column (Aurora XS C18, 75 µm × 15 cm, 1.7 µm; IonOpticks, Fitzroy, Australia).

Peptide separation was performed using a segmented linear gradient of solvent B (0.1% formic acid in acetonitrile) in solvent A (0.1% formic acid in water), consisting of 2%–30% solvent B over 43 min, followed by 30%–80% solvent B over 1 min, and 80% solvent B for 2 min, at a flow rate of 300 nL/min.

Data acquisition was carried out on a ZenoTOF 7600+ mass spectrometer (SCIEX, Framingham, MA, USA) operated in data‐independent acquisition (DIA) mode using the Zeno SWATH workflow [19]. DIA data were processed using DIA‐NN (version 2.0) [20] in library‐free mode with the UniProt human reference proteome. Database searches were performed allowing one missed cleavage, with cysteine carbamidomethylation specified as a fixed modification and no variable modifications.

Precursor intensities were normalized using retention time–dependent cross‐run normalization within DIA‐NN. Peptides and proteins were filtered at a false discovery rate (FDR) < 1% for both identification and quantification.

2.8. Statistical Analysis and Data Visualization

Each GCTB cell line cultured under 2D and 3D conditions was treated as an independent biological unit. Uniform manifold approximation and projection (UMAP) was used for nonlinear dimensionality reduction and visualization of high‐dimensional datasets.

Differential expression analysis for both transcriptomic and proteomic datasets was performed using the SciPy (version 1.9.3) library, and significantly altered genes or proteins were visualized using volcano plots. Pathway enrichment analyses were conducted using widely used databases such as the WikiPathway_2023_Human gene set collection from MSigDB, applying rank‐based gene set enrichment analysis (GSEA) or similar methods with gseapy library (version 1.1.2).

Cluster heatmaps, UMAP plots, volcano plots, and pathway enrichment maps were generated using standard visualization packages. The normalized expression values were scaled to z‐scores prior to heatmap visualization. Integrated analyses were performed to evaluate associations between transcriptomic/proteomic features and drug sensitivity patterns. UMAP embedding was computed using the umap‐learn package (version 0.5.3) with the parameters n_neighbors = 15, min_dist = 0.1, metric = “euclidea”, and random_state = 1234.

3. Results

3.1. GCTB Cell Lines Under the 2D and 3D Culture Conditions

Representative morphological features of GCTB cells (NCC‐GCTB6‐C1) cultured under 2D and 3D conditions are shown in Figure 1. Under 2D culture conditions, phase‐contrast microscopy revealed a typical adherent monolayer morphology (Figure 1A), which was further confirmed by H&E staining (Figure 1B). In contrast, under 3D culture conditions, GCTB cells formed compact and well‐defined spheroid structures, as observed by phase‐contrast microscopy (Figure 1C) and histologically confirmed by H&E staining of paraffin‐embedded spheroid sections (Figure 1D). These morphological differences between 2D and 3D cultures were consistently observed across the GCTB cell‐line panel, confirming that stable and reproducible spheroid formation was achieved under the applied 3D culture conditions.

FIGURE 1.

3D culture reorganizes GCTB cells into compact spheroid structures. Representative morphological features of patient‐derived giant cell tumor of bone (GCTB) cells (NCC‐GCTB6‐C1) cultured under two‐dimensional (2D) and three‐dimensional (3D) conditions. (A) Phase‐contrast microscopy image of GCTB cells cultured under 2D conditions, showing a typical adherent monolayer morphology. (B) Hematoxylin and eosin (H&E) staining of 2D‐cultured GCTB cells, confirming uniform cellular architecture. (C) Phase‐contrast microscopy image of GCTB cells cultured under 3D ultra‐low attachment conditions, demonstrating the formation of compact spheroid structures. (D) H&E staining of paraffin‐embedded sections of 3D spheroids, revealing dense cellular organization within the spheroid architecture. Scale bars indicate 100µm. Representative images are shown; similar morphological features were consistently observed across the GCTB cell‐line panel.

3.2. Transcriptomic Profiling of GCTB Cells Under the 2D and 3D Culture Conditions

To characterize transcriptomic differences between 2D and 3D cultures, we performed differential gene expression analysis using RNA sequencing (RNA‐seq) data. A volcano plot comparing gene expression in 2D cultures relative to 3D cultures is shown in Figure 2A (Tables S2 and S3). Genes located on the right side represent transcripts upregulated in 2D cultures, whereas those on the left side represent transcripts downregulated in 2D cultures, indicating widespread transcriptional changes associated with culture dimensionality.

FIGURE 2.

Culture dimensionality reshapes the transcriptomic landscape of GCTB cells. (A) Transcriptomic profiling of GCTB cell lines cultured under 2D and 3D conditions using RNA sequencing (RNA‐seq). (A) Volcano plot showing DEGs between 2D and 3D cultures. Genes on the right are upregulated in 2D cultures relative to 3D cultures, whereas genes on the left are downregulated in 2D cultures. (B) Heatmap of DEGs sorted by log2 fold differences, demonstrating clear segregation of samples according to culture condition. (C) Uniform manifold approximation and projection (UMAP) of transcriptomic profiles, showing distinct clustering of 2D‐ and 3D‐cultured samples. Together, these analyses indicate that culture dimensionality strongly influences the transcriptomic state of GCTB cells.

Based on these differentially expressed RNAs, we generated a heatmap to visualize global expression patterns across samples (Figure 2B, Table S3). The heatmap demonstrated the distinct transcriptome profiles between 2D‐ and 3D‐cultured GCTB cells.

To further assess global transcriptomic differences in an unbiased manner, uniform manifold approximation, and projection (UMAP) was applied to the same dataset. UMAP analysis revealed clear segregation of samples according to culture condition (Figure 2C, Table S4), consistent with the results obtained from differential expression and clustering analyses. Together, these analyses indicate that culture dimensionality is a major determinant of the transcriptomic state of GCTB cell lines.

3.3. Proteomic Profiling of GCTB Cells Under the 2D and 3D Culture Conditions

To characterize proteomic differences between 2D and 3D cultures, we performed quantitative proteomic analysis using data‐independent acquisition (DIA) mass spectrometry. A volcano plot comparing protein abundance in 2D cultures relative to 3D cultures is shown in Figure 3A (Tables S5 and S6. Proteins located on the right side represent those upregulated in 2D cultures, whereas those on the left side represent proteins downregulated in 2D cultures, indicating widespread culture‐dependent changes at the protein level.

FIGURE 3.

Proteomic remodeling reveals culture‐dependent cellular states in GCTB. (A) Quantitative proteomic profiling of GCTB cell lines cultured under 2D and 3D conditions using data‐independent acquisition (DIA) mass spectrometry. (A) Volcano plot showing differentially abundant proteins between 2D and 3D cultures. Proteins on the right are more abundant in 2D cultures, whereas proteins on the left are reduced in 2D cultures relative to 3D cultures. (B) Heatmap of differentially abundant proteins sorted by log2 fold differences, demonstrating separation of samples by culture condition. (C) UMAP of quantitative proteomic data, revealing distinct proteomic states associated with 2D and 3D culture conditions. These results demonstrate that culture dimensionality exerts a strong and reproducible effect on the proteomic landscape of GCTB cells.

Based on these differentially abundant proteins, we generated a heatmap to visualize global proteomic patterns across samples (Figure 3B, Table S6). The heatmap demonstrated the distinct proteome profiles between 2D‐ and 3D‐cultured GCTB cells.

To further assess global proteomic differences in an unbiased manner, UMAP was applied to the quantitative proteomic dataset. UMAP analysis revealed clear segregation of samples according to culture condition (Figure 3C, Table S7), consistent with the results obtained from differential expression and clustering analyses. Together, these analyses indicate that culture dimensionality is a major determinant of the proteomic state of GCTB cell lines.

3.4. Integrated Pathway‐level Comparison of Transcriptomic and Proteomic Alterations Between 2D and 3D Cultures

To further characterize culture‐dependent molecular remodeling at the pathway level, we performed gene set enrichment analysis (GSEA) using both transcriptomic and proteomic datasets, focusing on the top ten positively and negatively enriched pathways ranked by normalized enrichment score (NES) (Table S8).

In the transcriptomic analysis, GSEA revealed that pathways enriched in 2D cultures were predominantly associated with cell‐cycle regulation, DNA replication, and DNA damage response (Figure 4A, red bars). In contrast, pathways enriched in 3D cultures were mainly related to metabolic processes, and oxidative stress responses (Figure 4A, blue bars).

FIGURE 4.

3D culture differentially activates proliferative and metabolic pathways at the RNA and protein levels. Gene set enrichment analysis (GSEA) comparing 2D and 3D cultures at the transcriptomic and proteomic levels. (A) Top ten positively and negatively enriched pathways based on transcriptomic data, ranked by normalized enrichment score (NES). Red bars indicate pathways enriched in 2D cultures, and blue bars indicate pathways enriched in 3D cultures. (B) Corresponding top enriched pathways based on proteomic data. This figure illustrates both shared and modality‐specific pathway signatures induced by culture dimensionality.

Proteome‐based GSEA demonstrated both overlapping and distinct enrichment patterns compared with transcriptomic analysis (Figure 4B). While enrichment of cell‐cycle–related pathways in 2D cultures was also observed at the protein level, pathways related to mitochondrial function, oxidative phosphorylation, and intracellular organization were more prominently enriched in 3D cultures in the proteomic data.

Collectively, these analyses indicate that pathway‐level differences between 2D and 3D cultures are reflected at both transcriptomic and proteomic levels, while also revealing modality‐specific features.

To complement pathway‐based analyses, word cloud representations were generated using genes and proteins contributing to the top enriched pathways (Figures S1A and S1B). Transcriptomic word clouds emphasized regulators of cell‐cycle progression in 2D cultures and metabolic or stress‐response processes in 3D cultures. In contrast, proteomic word clouds highlighted mitochondrial metabolism, redox regulation, and intracellular organization, features that were less apparent at the transcriptomic level.

These analyses indicate that pathway‐level differences between 2D and 3D cultures are reflected at both transcriptomic and proteomic levels, while also revealing modality‐specific features.

3.5. Integrated Analysis of Transcriptomic and Proteomic Alterations Between 2D and 3D Cultures

To evaluate the concordance and divergence of expression changes between the transcriptome and proteome, we performed an integrated analysis of differentially expressed genes (DEGs) and differentially expressed proteins (DEPs) identified in the comparison between 2D and 3D cultures.

First, for each gene quantified at both molecular layers, log₂ fold difference derived from transcriptomic data were plotted against corresponding log₂ fold difference at the proteomic level (Figure 5A, Table S9). A subset of molecules exhibited concordant differential expression at both the mRNA and protein levels (highlighted in red), indicating coordinated regulation across omics layers. In contrast, a substantial proportion of molecules showed differential expression in only one omics layer (dark grey), reflecting discordant regulation between transcript abundance and protein expression.

FIGURE 5.

Limited RNA–protein concordance highlights the necessity of proteomic profiling in 3D culture models. Integrated comparison of transcriptomic and proteomic alterations between 2D and 3D cultures. (A) Scatter plot comparing log₂ fold differences at the transcriptomic (x‐axis) and proteomic (y‐axis) levels for genes quantified in both datasets. Molecules differentially expressed at both levels are highlighted in red, whereas molecules showing layer‐specific regulation are shown in dark gray. (B) Confusion matrix summarizing concordance of pathway enrichment direction between transcriptomic and proteomic analyses, based on the top 30 enriched terms in each dataset. (C) Heatmap of normalized enrichment scores (NES) for pathways commonly enriched at both molecular layers, illustrating similarities and differences in enrichment magnitude and direction. This integrated analysis highlights partial concordance and substantial divergence between transcriptomic and proteomic responses, underscoring the importance of proteomic profiling for capturing functional regulatory layers that are not predictable from transcriptomic data alone.

To further characterize the concordantly regulated molecules, genes identified as differentially expressed at both the transcriptomic and proteomic levels were classified according to the direction of regulation (2D‐enriched versus 3D‐enriched), and their gene names were visualized (Figure S2A, Table S10). This analysis revealed a limited but distinct group of molecules that consistently followed the same direction of change across both molecular layers, suggesting robust culture‐dependent regulation for these genes.

In parallel, molecules exhibiting omics layer–specific regulation were examined separately. Genes showing transcriptome‐specific changes and those showing proteome‐specific changes were grouped and visualized to highlight representative gene sets unique to each molecular layer (Figure S2B, Table S10). This stratification underscores the presence of regulatory mechanisms that selectively affect mRNA abundance or protein levels under different culture conditions.

To assess concordance at the pathway level, enrichment results derived independently from transcriptomic and proteomic analyses were compared using a confusion matrix–based visualization (Figure 5B). Enrichment direction (positive, negative, or no enrichment) was evaluated for the top 30 enriched terms in each omics layer. While a subset of pathways showed concordant enrichment between transcriptomic and proteomic analyses, many terms were enriched in only one molecular layer, indicating partial overlap but substantial divergence in pathway‐level interpretation.

Finally, pathways that were commonly enriched in both transcriptomic and proteomic analyses (outlined in Figure 5C) were further examined by visualizing their normalized enrichment scores (NES) as a heatmap (Figure 5C). Even among shared pathways, differences in the magnitude and direction of enrichment were observed between molecular layers.

These results demonstrate that culture‐dependent molecular remodeling in GCTB cells is only partially concordant between the transcriptome and proteome. This divergence highlights that proteomic profiling is indispensable for capturing functional regulatory layers—such as translational control and protein‐level pathway dominance—that cannot be reliably inferred from transcriptomic data alone.

3.6. Drug Sensitivity Assay of GCTB Cells Under the 2D and 3D Culture Conditions

We next evaluated the drug responses of GCTB cells cultured under 2D and 3D conditions using a panel of 221 anticancer agents tested at a fixed concentration of 10 µM (Table S1). Drug sensitivity profiles differed markedly between the two culture conditions, as visualized by a heatmap of relative cell viability across all compounds and cell lines (Figure 6A).

FIGURE 6.

Drug sensitivity profiles of GCTB cells under 2D and 3D culture conditions. (A) Heatmap showing relative drug sensitivity of GCTB cell lines cultured under 2D and 3D conditions across a panel of 221 anticancer agents. (B) UMAP of drug‐response profiles, demonstrating segregation of samples according to culture condition.

Overall, cells cultured under 3D conditions exhibited reduced sensitivity to a broad range of cytotoxic and targeted agents compared with those cultured in 2D. Nevertheless, a subset of compounds showed comparable or, in some cases, greater activity under 3D conditions, indicating that the impact of culture dimensionality on drug response is compound‐dependent.

Unsupervised dimensionality reduction using UMAP further demonstrated a clear separation of samples according to culture condition (Figure 6B, Table S11). Drug‐response profiles clustered primarily by 2D versus 3D culture, rather than by individual cell‐line identity, suggesting that global drug sensitivity patterns are strongly shaped by the structural and molecular context imposed by culture dimensionality.

Together, these results indicate that 3D culture profoundly alters drug response landscapes in GCTB cells. The distinct clustering patterns observed in both hierarchical and UMAP analyses support the notion that therapeutic susceptibility is influenced by culture‐dependent cellular states, providing a functional readout that complements the transcriptomic and proteomic differences described above.

4. Discussion

In this study, we demonstrate that culture dimensionality is a dominant contextual determinant of cellular state in GCTB. Transition from 2D monolayer culture to 3D spheroid growth reproducibly reorganized GCTB cells at multiple levels, including cellular architecture, gene expression, protein abundance, and drug response profiles [21, 22]. Morphologically, 3D culture induced compact spheroid structures distinct from monolayer growth, reflecting a fundamental shift in cell–cell organization. At the molecular level, both transcriptomic and proteomic analyses revealed pronounced culture‐dependent remodeling; however, these changes were only partially concordant across molecular layers. Pathway‐level analyses further showed that proliferative programs predominated in 2D cultures, whereas metabolic and stress‐adaptive pathways were preferentially enriched in 3D cultures, with proteomic profiling capturing functional states that were not readily inferred from transcriptomic data alone. Importantly, these molecular differences translated into distinct drug sensitivity landscapes, with 3D cultures generally exhibiting increased drug tolerance [16, 17]. Together, these findings establish that 3D culture induces a discrete and functionally relevant cellular state in GCTB cells, and highlight the necessity of integrated proteomic analysis for accurately interpreting culture‐dependent phenotypes [23, 24].

4.1. Biological Relevance of 3D Culture Models in GCTB

Three‐dimensional culture systems are increasingly used to model tumor architecture and function more accurately than conventional 2D monolayers [25]. However, in GCTB, the biological relevance of 3D models based on patient‐derived cell lines has not been systematically evaluated, particularly from a multi‐omics perspective. In the present study, 3D culture consistently induced the formation of compact spheroids with well‐defined boundaries across all examined GCTB cell lines (Figure 1). These spheroids were generated without exogenous extracellular matrix components, suggesting that intrinsic cell–cell interactions are sufficient to support stable 3D organization. Similar architectural features have been linked to altered signaling and therapeutic responses in other solid tumors, including sarcomas [26]. It is noteworthy that the morphological transition induced by 3D culture was accompanied by coordinated transcriptomic, proteomic, and functional changes. This concordance between structural reorganization and molecular remodeling supports the notion that 3D culture induces a distinct cellular state that may be more relevant for modeling GCTB biology than conventional 2D systems.

4.2. Transcriptomic and Proteomic Remodeling Induced by 3D Culture

Integrated transcriptomic and proteomic analyses revealed that 3D culture is associated with remodeling of pathways related to extracellular matrix (ECM) organization, cellular stress responses, and metabolic regulation [27]. These changes indicate that 3D culture induces qualitative shifts in cellular state rather than simple quantitative modulation. At both RNA and protein levels, 3D culture was associated with increased representation of ECM‐related and adhesion‐associated processes, consistent with enhanced cell–cell interactions in spheroid structures [28, 29]. In parallel, enrichment of stress‐response pathways, including hypoxia‐ and oxidative stress–related signaling, was observed, likely reflecting diffusion gradients within spheroids [30, 31, 32].

Metabolic reprogramming also emerged as a prominent feature of 3D‐cultured GCTB cells. Proteome‐based analyses revealed metabolic pathway alterations that were less apparent at the transcriptomic level, consistent with the notion that metabolic adaptation is often regulated at the protein level [33, 34, 35]. This RNA–protein discordance further emphasizes the value of proteomic profiling for capturing functional metabolic states [36, 37]. Our analysis indicates that molecules showing such discordance were frequently enriched in metabolic processes and extracellular matrix remodeling (Table S10). This indicates that GCTB cells in 3D environments may rely more heavily on translational control or protein stability—regulatory layers that are essential for adapting to the nutrient and oxygen gradients within spheroids, which cannot be accurately predicted by transcriptomic data alone.

4.3. Impact of 3D Culture on Drug Sensitivity in GCTB

Drug sensitivity profiling demonstrated marked differences between 2D‐ and 3D‐cultured GCTB cells. Overall, 3D culture was associated with reduced sensitivity to a broad range of anticancer agents, although the magnitude and direction of this effect were compound‐dependent [38, 39]. Similar trends have been reported in other tumor models, where 3D systems often exhibit increased drug tolerance compared with monolayer cultures [40, 41, 42].

Multiple mechanisms may contribute to altered drug sensitivity in 3D cultures, including limited drug penetration, altered cell‐cycle dynamics, enhanced cell–cell interactions, and activation of stress‐adaptive pathways. The molecular signatures identified in this study are consistent with these mechanisms and provide a molecular context for interpreting culture‐dependent drug responses [43, 44, 45].

Importantly, a subset of compounds retained or showed enhanced activity under 3D conditions, underscoring the limitations of extrapolating drug efficacy from 2D assays alone. These findings support the incorporation of 3D culture models into preclinical drug evaluation pipelines, particularly for assessing context‐dependent therapeutic vulnerabilities [43, 46]. Moreover, the observed molecular differences translated into distinct drug sensitivity landscapes. The widespread reduction in sensitivity to many cytotoxic agents, such as topoisomerase inhibitors and anti‐microtubule agents (Figure 6A), directly corresponds with the downregulation of cell‐cycle regulation and DNA replication pathways identified in our GSEA (Figure 4). This suggests that the reduced proliferative state in 3D culture acts as a primary determinant of drug tolerance. Conversely, the preservation of sensitivity to specific targeted agents in 3D suggests that certain signaling vulnerabilities remain constant regardless of cultural dimensionality, providing a basis for prioritizing drug candidates in a more physiologically relevant context.

4.4. Limitations

Several limitations of this study should be acknowledged. First, molecular profiling was limited to baseline conditions, and drug‐induced transcriptomic or proteomic changes were not assessed. Second, the proteomic analysis focused on global protein abundance and did not include post‐translational modification profiling, such as phosphorylation, which is critical for understanding sensitivity to kinase inhibitors and other signaling‐targeted therapies [47, 48].

In addition, regulatory features including protein–protein interactions, subcellular localization of drug targets, and spatial organization of signaling complexes were not examined, despite their known influence on drug response [49, 50, 51]. Furthermore, 3D spheroid cultures do not fully recapitulate the in vivo tumor microenvironment, and fixed‐concentration drug screening limits pharmacodynamic interpretation.

Despite these limitations, the systematic and paired comparison of 2D and 3D cultures across multiple patient‐derived GCTB cell lines using integrated multi‐omics and functional assays provides a robust framework for future studies incorporating phosphoproteomics, spatial proteomics, and more physiologically relevant model systems [14, 52, 53, 54].

5. Conclusions

In conclusion, this study provides a systematic and integrated comparison of 2D and 3D culture systems using a panel of patient‐derived GCTB cell lines. By combining transcriptomic profiling, quantitative proteomics, and large‐scale drug sensitivity screening under paired culture conditions, we demonstrate that culture dimensionality is a dominant contextual determinant of cellular state, shaping molecular organization and therapeutic responses in a coordinated yet non‐identical manner.

We show that 3D culture reproducibly remodels gene expression, protein abundance, and drug sensitivity profiles, establishing a distinct cellular state that cannot be fully captured by transcriptomic analysis alone. While culture dimensionality was the dominant factor in global clustering, the relative spatial distribution of individual cell lines in our UMAP analyses (Figures 2C, 3C, and 6B) was partially preserved between 2D and 3D conditions. This suggests that patient‐specific molecular “fingerprints” remain detectable despite the universal shifts induced by 3D culture, underscoring the potential of these models for personalized drug screening and patient stratification in GCTB research. The partial uncoupling between RNA‐ and protein‐level changes underscores the necessity of proteomic profiling for revealing functional regulatory layers—particularly those related to metabolism, stress adaptation, and pathway dominance—that are not reliably predictable from mRNA abundance. These molecular differences were directly reflected in altered drug response landscapes, indicating that culture dimensionality critically influences therapeutic susceptibility in GCTB cells.

Although this study did not address post‐translational modifications, spatial proteomics, or drug‐induced molecular responses, it establishes a robust experimental framework for dissecting culture‐dependent phenotypes in rare bone tumors. Collectively, our findings support the use of integrated multi‐omics analyses in patient‐derived 3D models as a necessary complement to conventional 2D systems and provide a foundation for future studies aimed at deeper mechanistic insight and improved translational relevance in GCTB research.

Representative morphological features of patient‐derived GCTB cells (NCC‐GCTB6‐C1) cultured under 2D and 3D conditions. (A) Phase‐contrast microscopy image of GCTB cells cultured under 2D conditions, showing a typical adherent monolayer morphology. (B) H&E staining of 2D‐cultured GCTB cells, confirming uniform cellular architecture. (C) Phase‐contrast microscopy image of GCTB cells cultured under 3D ultra‐low attachment conditions, demonstrating the formation of compact spheroid structures. (D) H&E staining of paraffin‐embedded sections of 3D spheroids, revealing dense cellular organization within the spheroid architecture. Scale bars indicate 100 µm. Representative images are shown; similar morphological features were consistently observed across the GCTB cell‐line panel.

Funding

This research was supported by AMED under Grant Number 256f0137009j0001, JP24ama121018, JSPS KAKENHI Grant Number 24K02574, 25K12471, and 25K18360.

Ethics Statement

Ethical approval was not required for this study, as it involved the use of established patient‐derived cell lines that are publicly available and did not include identifiable human subjects.

Conflicts of Interest

The corresponding author (T.K.) conducted this research in the context of a collaborative research project with Zeon Corporation and Knowledge Palette. I.F and S.H belong to Zeon BioSolutions, and Knowledge Palette, respectively. This collaboration did not influence the study design, data acquisition, data analysis, interpretation of the results, or the decision to submit the manuscript for publication. The other authors declare no conflicts of interest.

Supporting information

Supporting File: pmic70136‐sup‐0001‐SuppMat.zip.

Data Availability Statement

The mass spectrometry–based proteomic data have been deposited in the ProteomeXchange Consortium via the jPOST repository under the dataset identifier JPST004280/PXD072451.