Resolving drug effects in patient-derived cancer cells links organoid responses to genome alterations

Julia Neugebauer, Franziska Maria Zickgraf, Jeongbin Park, Steve Wagner, Xiaoqi Jiang, Katharina Jechow, Kortine Kleinheinz, Umut H. Toprak, Marc A. Schneider, Michael Meister, Saskia Spaich, Marc Sütterlin, Matthias Schlesner, Andreas Trumpp, Martin R. Sprick, Roland Eils, Christian ConradbioRxiv 2017 · DOI

TL;DR

This paper introduces "DeathPro," a novel microscopy-based assay to precisely measure drug-induced cell death and proliferation in patient-derived cancer cells. They demonstrate that drug responses, particularly for cytostatic drugs, are highly dependent on whether cells are cultured as 2D monolayers or 3D organoids, with organoids showing more diverse responses and revealing genotype-drug sensitivity links (like DNA repair deficiency) not seen in 2D.

What Problem Does This Paper Address?

Personalized oncology and drug discovery rely heavily on testing drugs against patient-derived cancer cells. However, a major challenge is the lack of standardization, particularly concerning the culture format. Traditional 2D monolayer cultures are widely used but may not accurately reflect the complex 3D tumor microenvironment. Advanced 3D organoid models offer a more physiologically relevant system, but robust, high-throughput methods to assess drug effects in these models have been lacking. This gap makes it difficult to compare drug efficacy across different culture systems and to fully leverage genomic information for predicting drug response. The authors aim to bridge this gap by developing a method to precisely compare drug effects in both 2D and 3D patient-derived cancer models and to link these responses to genomic alterations.

Methods

The core innovation is the DeathPro assay, an automated, microscopy-based workflow. It uses live-cell dyes (Hoechst for all cells, Propidium Iodide for dead cells) to quantify drug-induced cell death and proliferation inhibition over time. Key features include:

  • Confocal Microscopy: Captures detailed images of cells.
  • Maximum Intensity Projection (MIP): Processes 3D image stacks into 2D representations for analysis.
  • Custom Image Analysis Workflow: Developed in KNIME, this workflow uses adaptive thresholding and outlier filtering to accurately measure the area of dead (PI-stained) and total (Hoechst-stained) cells.
  • Quantitative Metrics: Calculates LD50 (lethal dose for 50% of cells), AUCd (Area Under the Curve for cell death), and AUCpi (Area Under the Curve for proliferation inhibition) to provide comprehensive drug response profiles.
  • Whole Genome Sequencing (WGS): Performed on patient-derived cell lines to identify genetic alterations, including copy number alterations (CNAs), indels, and single nucleotide variations (SNVs).
  • Homologous Recombination Deficiency (HRD) Scoring: Calculated based on loss of heterozygosity regions, linking genomic instability to drug response.

The DeathPro assay is novel in its ability to simultaneously resolve cell death and proliferation inhibition over time in both 2D and 3D cultures, overcoming limitations of simpler viability assays.

Datasets

The study utilized patient-derived ovarian cancer cell lines (PDCLs) and normal ovarian epithelial cells (HOSEpiC).

  • Patient-Derived Ovarian Cancer Cell Lines (PDCLs): Established from serous ovarian cancer patients (FIGO stage IIIc-IV). Nine such cell lines were used for drug screening.
  • Normal Ovarian Epithelial Cells (HOSEpiC): Included to assess potential side effects and normalize drug efficacy.
  • Whole Genome Sequencing (WGS) Data: Generated for the PDCLs and HOSEpiC.
  • Publicly Available Genomic Databases: Used for comparison of genetic alterations (e.g., COSMIC, ICGC).

No new benchmarks were introduced, but the study established a comprehensive dataset of drug responses across different culture formats and linked it to genomic data.

Key Insights and Ideas

  • Culture Format Matters Dramatically: Drug responses, especially for cytostatic drugs, are significantly different between 2D monolayers and 3D organoids. This highlights the need to choose the right model system for preclinical drug testing.
  • Organoids Reveal Hidden Drug Sensitivity: Organoid cultures revealed more diverse drug responses and uncovered links between drug sensitivity and genomic alterations (like DNA repair deficiencies) that were not detectable in 2D cultures.
  • Genotype-Drug Sensitivity Links are Stronger in 3D: Correlations between genetic alterations (e.g., HRD scores) and drug responses were more pronounced in 3D organoid cultures compared to 2D.
  • DeathPro is a Versatile Tool: The assay is robust, scalable, and adaptable to different cancer types (ovarian and lung cancer) and cell models.

Key Results

  • Differential Drug Effects: Cytostatic drugs showed culture-type dependent efficacy. While cell death was similar between 2D and 3D, growth arrest varied significantly.
  • Organoid Diversity: Drug response profiles in organoids were more diverse than in 2D cultures, with 3D profiles clustering into 8 subclusters compared to 4 in 2D.
  • Genomic Correlations in 3D: 90% of statistically significant correlations between drug responses and genomic alterations (specifically HRD scores) were observed in 3D organoid data.
  • HRD and Drug Response: High HRD scores correlated with increased cytotoxicity to carboplatin, paclitaxel, azacytidine, and decitabine, and also with growth arrest induced by temsirolismus.
  • Therapeutic Index: Organoid cultures generally showed less favorable therapeutic indices (indicating higher toxicity to normal cells relative to cancer cells) compared to 2D cultures.
  • Carboplatin Dominance: Carboplatin remained a potent drug across both culture types, with limited toxicity in normal cells, suggesting its continued role as a first-line therapy.

Limitations and Future Work

  • Sample Size: While nine PDCLs were used, the authors acknowledge that a larger sample size would further strengthen the findings regarding genotype-drug sensitivity correlations.
  • Drug Removal Kinetics: The study noted that cell death in 3D cultures might be underestimated after a standard 72h treatment interval, suggesting that longer treatment or drug removal phases (like the 144h time point) are crucial for a complete picture.
  • Future Work: The authors suggest that their DeathPro assay, combined with organoid models, can be used for refined drug screening and for deciphering genotype-drug sensitivity associations. They envision this approach guiding the selection of appropriate culture formats for future translational research and personalized treatment strategies. The ability to adapt the DeathPro workflow to other nuclear stains also opens avenues for broader application.

Summary updated

Abstract

Abstract Cancer drug screening in patient-derived cells holds great promise for personalized oncology and drug discovery but lacks standardization. Whether cells are cultured as conventional monolayer or advanced organoid cultures influences drug effects and thereby drug selection and clinical success. To precisely compare drug profiles in differently cultured primary cells, we developed DeathPro , an automated microscopy-based assay to resolve drug-induced cell death and proliferation inhibition. Using DeathPro , we screened cells from ovarian cancer patients in monolayer or organoid culture with clinically relevant drugs. Drug-induced growth arrest and efficacy of cytostatic drugs differed between the two culture systems. Interestingly, drug effects in organoids were more diverse and had lower therapeutic potential. Genomic analysis revealed novel links between drug sensitivity and DNA repair deficiency in organoids that were undetectable in monolayers. Thus, our results highlight the dependency of cytostatic drugs and pharmacogenomic associations on culture systems, and guide culture selection for drug tests.
BibTeX
@misc{neugebauer2017resolving,
  author = {Julia Neugebauer and Franziska Maria Zickgraf and Jeongbin Park and Steve Wagner and Xiaoqi Jiang and Katharina Jechow and Kortine Kleinheinz and Umut H. Toprak and Marc A. Schneider and Michael Meister and Saskia Spaich and Marc Sütterlin and Matthias Schlesner and Andreas Trumpp and Martin R. Sprick and Roland Eils and Christian Conrad},
  title = {{Resolving drug effects in patient-derived cancer cells links organoid responses to genome alterations}},
  year = {2017},
  doi = {10.1101/124446},
  howpublished = {bioRxiv},
}
Figure