Abstract
Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal solid malignancies, necessitating the development of improved therapeutic strategies. Patient-derived xenografts (PDX) and patient-derived organoids (PDO) have emerged as vital preclinical tools for drug discovery and personalized medicine. However, the extent to which these models recapitulate the complex genomic landscape of the original human tumor—particularly at the whole-genome level—has not been fully characterized. This study presents a comprehensive assessment of 16 whole-genome pairs of tumors and matched PDX models, including a unique cohort of five "trios" (primary tumor, PDX, and PDO). By developing a specialized pipeline to score concordance across mutations, structural variations, and copy number changes, we demonstrate that PDX and PDO models effectively preserve the genomic hallmarks of PDAC, including driver mutations and complex structural rearrangements.
Introduction
The clinical management of PDAC is hindered by its profound therapeutic resistance and molecular heterogeneity. To bridge the gap between laboratory research and clinical application, researchers utilize PDX models (human tumors grown in mice) and PDO models (3D cell cultures). While these models are widely used, concerns persist regarding how accurately they mirror the "genomic chaos" characteristic of pancreatic cancer, which often involves massive structural rearrangements and chromothripsis.
Previous validation efforts have largely relied on Whole Exome Sequencing (WES), which captures only a small fraction of the genome. This study utilizes Whole Genome Sequencing (WGS) to provide a high-resolution comparison of the genomic fidelity of these models, ensuring that the complex structural and non-coding landscapes of the original disease are preserved.
Methodology
To ensure high-quality data, the researchers employed Laser-Capture Microdissection (LCM) to isolate tumor cells from surrounding stroma, minimizing contamination from healthy human tissue. The study analyzed 16 pairs, including 10 resected primary tumors and 6 liver metastases. A subset of these formed "trios," where the researchers compared the patient tumor to a PDX and a subsequent PDX-derived PDO.
Genomic concordance was quantified using a specialized bioinformatics pipeline focusing on three areas:
- Simple Somatic Mutations (SSM): Single-nucleotide variants.
- Copy Number Variations (CNV): Large-scale gains or losses of genetic material, analyzed using a ploidy-adjusted scoring system.
- Structural Variations (SV): Complex rearrangements such as inversions and translocations.
The Jaccard Index was utilized to statistically measure the overlap of these events between the source tumors and their respective models.
Key Insights and Results
High Fidelity in Driver Mutations and Copy Numbers
The study found remarkable consistency in the primary drivers of PDAC. Mutations in genes such as KRAS, TP53, and CDKN2A were highly conserved across all models. On average, PDX models retained over 50% of the mutations found in primary tumors, while metastatic models showed even higher retention rates (66-86%). Genome-wide copy number states were also highly stable, with concordance scores frequently exceeding 0.8, indicating that the large-scale genomic architecture remains intact during the transition from patient to model.
Preservation of Genomic Catastrophes
Pancreatic cancer is often defined by "catastrophic" genomic events like chromothripsis, where specific chromosomes undergo massive shattering and haphazard reassembly. A significant finding of this study was that these complex structural variations were effectively transmitted to both PDX and PDO models. While SVs generally showed more variability than single-letter mutations, specific "hotspot" chromosomes demonstrated near-perfect preservation of these rearrangements.
Metastatic Stability vs. Primary Divergence
An interesting trend emerged regarding the source of the tumor: models derived from liver metastases exhibited higher genomic concordance than those derived from primary resected tumors. This suggests that metastatic cells may represent a more evolutionarily stable or "fixed" state of the disease. Conversely, in rare cases of primary tumor models, "subclonal takeover" was observed—where a minority population of cells from the original tumor gained a selective advantage in the PDX environment, leading to a model that diverged significantly from the bulk of the original patient sample.
The Trio Relationship
In the five trios studied, the PDOs (which were derived from PDX tissue) showed extreme similarity to their parent PDX models. This confirms that the transition from a mouse model to a 3D culture does not introduce significant new genomic instability, though the authors note that direct patient-to-organoid models remain an important area for future comparison.
Discussion and Limitations
While this study confirms that PDX and PDO models are faithful genomic "avatars" of PDAC, several considerations remain. The sample size of 16 WGS pairs is robust for this type of deep sequencing but remains a relatively small cohort for identifying rare patterns. Additionally, the study focused on genomic concordance; the next critical step is establishing functional concordance—confirming that therapeutic responses in these models accurately predict patient outcomes.
Conclusion
This whole-genome characterization provides strong evidence that PDX and PDO models recapitulate the essential genomic features of pancreatic ductal adenocarcinoma. By successfully capturing both simple mutations and complex structural catastrophes, these models prove to be reliable tools for studying tumor biology and testing novel therapeutic interventions in a patient-specific context.
Summary updated