Aberrant brain-type neuronal programs in large-cell pancreatic neuroendocrine carcinoma

Olivia Debnath, Hilmar Berger, Katharina Detjen, Philipp Kirchner, Ilaria Marinoni, I Eichhorn, Felix Bolduan, Thomas Conrad, David Horst, Roland Eils, Bertram Wiedenmann, Aurel Perren, Naveed IshaquebioRxiv 2025 · DOI

TL;DR

This study uses single-nucleus transcriptomics to reveal that aggressive pancreatic neuroendocrine carcinomas (panNECs) exhibit aberrant neuronal programs, mimicking brain cell development rather than normal pancreatic cells. This discovery highlights new therapeutic vulnerabilities, suggesting combination therapies targeting these aberrant states could be effective.

What Problem Does This Paper Address?

Pancreatic neuroendocrine carcinomas (panNECs) are rare but aggressive cancers with limited tissue availability, making them challenging to study and treat. Existing research suggests panNECs might originate from exocrine cells, but their precise biological classification and the heterogeneity within these tumors remain unclear. This lack of understanding hinders the development of effective targeted therapies. This paper aims to unravel the complex cellular landscape and developmental origins of large-cell panNEC (LC-panNEC) to identify potential therapeutic strategies.

Methods

The researchers employed single-nucleus RNA sequencing (snRNA-seq) on five LC-panNEC tumor samples to generate a detailed cellular map. This technique allows for the analysis of gene expression at the individual cell nucleus level, providing high resolution. They used computational analysis pipelines, including:

  • CellRanger for initial data processing and alignment.
  • SoupX to remove ambient RNA contamination.
  • Scanpy for data integration, normalization, principal component analysis (PCA), and UMAP visualization to identify cell clusters.
  • Harmony and BBKNN for batch correction to harmonize data across samples and reduce technical variability.
  • Leiden algorithm for clustering and cell state identification.
  • pySCENIC for inferring transcription factor activities and regulatory networks.
  • DESeq2 for differential gene expression analysis between tissues.
  • Celltypist LR-based classifier for semi-supervised classification of panNEC cell states into brain cell types/states, using a first-trimester developing brain atlas as a reference.
  • Immunohistochemistry (IHC) for validation of key marker expressions in tumor tissues.

Datasets

  • Five large-cell pancreatic neuroendocrine carcinoma (LC-panNEC) tumor samples were collected and analyzed using snRNA-seq.
  • GTEx (Genotype-Tissue Expression) bulk RNA-seq data for brain and pancreas tissues was used for differential gene expression analysis.
  • Human brain atlas data (specifically, first-trimester developing brain data) was used as a reference for classifying panNEC cell states.
  • Adult pancreas snRNA-seq dataset from Tosti et al. 2020 was integrated for marker verification and comparative analysis.

Key Insights and Ideas

  • PanNECs exhibit "brain-like" programs: The most striking finding is that panNEC cells, particularly the neuroendocrine (NE) and NE proliferating states, express genes and pathways characteristic of developing neurons, not normal pancreatic cells. This suggests a significant dedifferentiation and aberrant developmental trajectory.
  • Two distinct NE cell states: Within the NE cells, two clinically relevant states were identified: a highly proliferative state and a stress-responsive state enriched for heat stress, hypoxia, and glycolysis.
  • Therapeutic vulnerabilities: The identification of these aberrant states and their associated pathways points to potential therapeutic targets. For instance, the proliferative state is linked to EZH2, and the hypoxic state to HSP90, suggesting combination therapies could be highly effective.
  • PTF1A's dual role: The transcription factor PTF1A, known for its role in both pancreatic and brain development, appears to be a key regulator in driving these aberrant brain-like programs in panNEC.

Key Results

  • Five main cell types were identified: neuroendocrine (NE), amphicrine acinar, amphicrine progenitor-like, normal stroma, and immune cells.
  • The NE cell type was further refined into four distinct states: NE, NE proliferating, NE HSP+ (hypoxic), and NE stromal-like.
  • RBFOX1, a marker strongly associated with neuronal function, was found to be highly expressed in the NE and NE proliferating cell states, supporting their brain-like identity.
  • PTF1A-regulated brain targets were significantly enriched in the NE proliferating cell state, while PTF1A-regulated pancreatic targets were found in the amphicrine acinar state, highlighting a divergence in regulatory programs.
  • Gene set enrichment analysis (GSEA) identified therapeutic vulnerabilities, with Cisplatin emerging as a key inhibitor for the NE proliferating cell state and HSP90 inhibitors (like Geldanamycin and Tanespimycin) for the NE HSP+ (hypoxic) state.

Limitations and Future Work

  • Limited sample size: The study analyzed only five LC-panNEC cases, which is a limitation due to the rarity of the disease. Larger cohorts would be beneficial for validating these findings.
  • Technical artifacts: While efforts were made to correct for batch effects, some variations between samples (e.g., patient P5's distinct profile) might still reflect technical nuances or unique biological characteristics that warrant further investigation.
  • In vitro/in vivo validation: The identified therapeutic vulnerabilities (e.g., combination therapies) require rigorous validation in preclinical models and clinical trials.
  • Further exploration of cell states: The paper mentions that details on remaining cell states are provided in the supplementary information, suggesting there's more to uncover about the full cellular heterogeneity.
  • Investigating resistance mechanisms: The study hints at potential resistance mechanisms (e.g., EZH1/2 co-expression) and suggests dual EZH1/2 targeting, which could be a promising avenue for future research.

Summary updated

Abstract

Abstract High-grade pancreatic neuroendocrine carcinoma (panNEC) is a rare, aggressive cancer with limited tissue availability. Single-nucleus transcriptomics of five large-cell panNECs revealed two clinically relevant neuroendocrine cell states: one highly proliferative with an aberrant brain-specific neuronal differentiation program and another stress-responsive state enriched for heat stress, hypoxia, and glycolysis. Our findings suggest potentially new therapeutic vulnerabilities, highlighting the need to evaluate the efficacy of combination therapies.
BibTeX
@misc{debnath2025aberrant,
  author = {Olivia Debnath and Hilmar Berger and Katharina Detjen and Philipp Kirchner and Ilaria Marinoni and I Eichhorn and Felix Bolduan and Thomas Conrad and David Horst and Roland Eils and Bertram Wiedenmann and Aurel Perren and Naveed Ishaque},
  title = {{Aberrant brain-type neuronal programs in large-cell pancreatic neuroendocrine carcinoma}},
  year = {2025},
  doi = {10.1101/2025.06.18.660360},
  howpublished = {bioRxiv},
}
Figure