Single-cell analysis of patient-derived PDAC organoids reveals cell state heterogeneity and a conserved developmental hierarchy

Teresa G. Krieger, Solange Le Blanc, Julia Jabs, Foo Wei Ten, Naveed Ishaque, Katharina Jechow, Olivia Debnath, Carl-Stephan Leonhardt, Anamika Giri, Roland Eils, Oliver Strobel, Christian Conrad

Abstract

Pancreatic ductal adenocarcinoma (PDAC) is projected to be the second leading cause of cancer mortality by 2030. Bulk transcriptomic analyses have distinguished 'classical' from 'basal-like' tumors with more aggressive clinical behavior. We derive PDAC organoids from 18 primary tumors and two matched liver metastases, and show that 'classical' and 'basal-like' cells coexist in individual organoids. By single-cell transcriptome analysis of PDAC organoids and primary PDAC, we identify distinct tumor cell states shared across patients, including a cycling progenitor cell state and a differentiated secretory state. Cell states are connected by a differentiation hierarchy, with 'classical' cells concentrated at the endpoint. In an imaging-based drug screen, expression of 'classical' subtype genes correlates with better drug response. Our results thus uncover a functional hierarchy of PDAC cell states linked to transcriptional tumor subtypes, and support the use of PDAC organoids as a clinically relevant model for in vitro studies of tumor heterogeneity.

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is characterized by profound intratumoral heterogeneity, which complicates treatment and contributes to its high mortality rate. Previous research using bulk sequencing identified two primary clinical subtypes: Classical, which generally correlates with better patient outcomes, and Basal-like, which is associated with more aggressive disease. However, bulk analysis obscures whether these subtypes represent distinct tumor types or a mixture of cell states within a single tumor. This study utilizes patient-derived organoids and single-cell RNA sequencing (scRNA-seq) to map the developmental hierarchy of PDAC cells and determine if organoids faithfully recapitulate the complexity of human tumors.

Methods and Experimental Design

The researchers established 24 organoid lines from 18 patients, including primary tumor biopsies and matched liver metastases. To characterize these models at high resolution, the team employed several advanced methodologies:

  • scRNA-seq: Over 93,000 individual cells were sequenced to identify gene expression profiles at the single-cell level.
  • RNA Velocity: This computational approach uses the ratio of unspliced to spliced mRNA to predict future cellular states, effectively mapping the direction of differentiation.
  • DeathPro Workflow: An automated imaging platform was used for drug screening, allowing researchers to distinguish between cytotoxic (cell-killing) and cytostatic (growth-stalling) effects.
Experimental workflow for generating PDAC organoids and performing single-cell RNA sequencing, alongside representative brightfield images of organoid morphology.

Key Insights: A Conserved Developmental Hierarchy

The study revealed a conserved "developmental map" shared across patients. Rather than being static identities, the 'classical' and 'basal-like' subtypes appear to be functional states within a shared hierarchy. Cells typically originate in a cycling progenitor state—characterized by rapid division and 'basal-like' features—and progress toward a differentiated secretory state, which represents the 'classical' endpoint.

Crucially, 'classical' and 'basal-like' cells were found to coexist within the same organoids. The degree of this "subtype mixing" correlated with clinical severity; organoids with higher heterogeneity often originated from patients with more aggressive, higher-grade tumors. Furthermore, metastatic tumors in the liver were found to "reset" and re-establish this same developmental hierarchy, suggesting the program is a fundamental biological feature of PDAC.

Results and Clinical Implications

Analysis identified nine distinct cell states shared across the patient cohort, including five cycling phases and two non-cycling secretory states. The molecular heterogeneity of these states is reflected in the varied expression of lineage-specific markers across different patient samples.

Dot plot illustrating the heterogeneous expression of key marker genes, such as Claudins (CLDN), Keratins (KRT), and Mucins (MUC), across different patient-derived organoid lines.

Key findings regarding treatment and prognosis include:

  • Drug Sensitivity: Screening with six clinical drugs (e.g., 5-FU, Gemcitabine) showed that a strong 'classical' gene signature correlates with improved drug response. Cells further along the differentiation path were more susceptible to growth inhibition.
  • Prognostic Value: Patients whose organoids exhibited a "homogeneous classical" profile had significantly longer overall survival compared to those with heterogeneous or predominantly basal-like organoids.

Discussion and Future Directions

While this study demonstrates that organoids are powerful models for studying PDAC heterogeneity, some limitations remain. In vitro culture may inadvertently select for 'classical' cells, and the current model lacks the complex tumor microenvironment (stroma, immune cells) present in vivo.

Future research may focus on the identified "bifurcation point" where cells transition from progenitors to differentiated states. Developing therapies that force aggressive progenitor cells to differentiate into the more treatable 'classical' state could represent a novel therapeutic strategy for this deadly disease.

Summary updated

Abstract

Pancreatic ductal adenocarcinoma (PDAC) is projected to be the second leading cause of cancer mortality by 2030. Bulk transcriptomic analyses have distinguished 'classical' from 'basal-like' tumors with more aggressive clinical behavior. We derive PDAC organoids from 18 primary tumors and two matched liver metastases, and show that 'classical' and 'basal-like' cells coexist in individual organoids. By single-cell transcriptome analysis of PDAC organoids and primary PDAC, we identify distinct tumor cell states shared across patients, including a cycling progenitor cell state and a differentiated secretory state. Cell states are connected by a differentiation hierarchy, with 'classical' cells concentrated at the endpoint. In an imaging-based drug screen, expression of 'classical' subtype genes correlates with better drug response. Our results thus uncover a functional hierarchy of PDAC cell states linked to transcriptional tumor subtypes, and support the use of PDAC organoids as a clinically relevant model for in vitro studies of tumor heterogeneity.
BibTeX
@inproceedings{krieger2021single,
  author = {Teresa G. Krieger and Solange Le Blanc and Julia Jabs and Foo Wei Ten and Naveed Ishaque and Katharina Jechow and Olivia Debnath and Carl-Stephan Leonhardt and Anamika Giri and Roland Eils and Oliver Strobel and Christian Conrad},
  title = {{Single-cell analysis of patient-derived PDAC organoids reveals cell state heterogeneity and a conserved developmental hierarchy}},
  booktitle = {Nature Communications},
  year = {2021},
  doi = {10.1038/s41467-021-26059-4},
}
Figure