Trellis Single-Cell Screening Reveals Stromal Regulation of Patient-Derived Organoid Drug Responses

María Ramos Zapatero, Alexander Tong, Jahangir Sufi, Petra Vlckova, Ferran Cardoso Rodriguez, Callum Baird Nattress, Xiao Qin, Daniel Hochhauser, Smita Krishnaswamy, Christopher J. TapebioRxiv 2022 · DOI

TL;DR

This paper introduces "Trellis," a novel computational framework for analyzing large-scale single-cell drug screening data from patient-derived organoids (PDOs) and their tumor microenvironment (TME). Trellis reveals that cancer-associated fibroblasts (CAFs) can protect cancer cells from chemotherapy by altering their cell state, and this protection can be reversed by targeting YAP signaling.

What Problem Does This Paper Address?

Personalized cancer therapy relies on understanding how individual tumors respond to drugs. Patient-derived organoids (PDOs) are promising models for this, but current methods often fail to capture the complex interactions within the tumor microenvironment (TME), such as those involving cancer-associated fibroblasts (CAFs). Furthermore, existing drug screening technologies typically provide only bulk measurements, lacking the single-cell resolution needed to understand the underlying mechanisms of drug response or resistance. This gap limits our ability to develop effective therapies that account for both intrinsic tumor biology and extrinsic TME influences.

Methods

The authors developed two key innovations:

  • TOBis Mass Cytometry: A highly multiplexed mass cytometry platform to simultaneously measure multiple cellular features (cell-type, cell-state, DNA-damage response, and post-translational modification (PTM) signaling) at the single-cell level. This allows for a deep mechanistic understanding of drug responses in thousands of PDO-CAF co-cultures.
  • Trellis: A novel, highly scalable, hierarchical tree-based computational method for analyzing large single-cell screening datasets. Trellis uses a tree structure to represent biological hierarchies (like cell types or states) and computes generalized optimal transport distances to compare treatment effects across thousands of conditions, effectively normalizing for controls and batch effects. It can also handle "paired" comparisons, where treatment effects are analyzed relative to their specific controls.

Datasets

The study utilized:

  • Colorectal Cancer (CRC) PDOs and CAFs: Over 2,500 patient-derived organoid cultures, co-cultured with cancer-associated fibroblasts, were treated with various clinical therapies.
  • Mass Cytometry Data: Resulting from the TOBis platform, generating over 25 million single-cell PTM signaling datasets.

Key Insights and Ideas

  • Patient-Specific Drug Responses: Drug responses are not solely determined by the drug itself or the cancer's genetic makeup, but are highly patient-specific and influenced by cell-intrinsic signaling states.
  • CAF Protective Role: CAFs are not just passive bystanders; they actively protect chemosensitive PDOs from certain chemotherapies by shifting cancer cells into a slow-cycling, "revSC-like" state. This highlights the critical importance of the TME in drug response.
  • YAP as a Therapeutic Target: The protective effect of CAFs can be reversed by inhibiting YAP signaling, suggesting a novel therapeutic strategy to overcome CAF-mediated resistance.
  • Mechanistic Insights Over Bulk Metrics: Single-cell PTM analysis, combined with Trellis, provides mechanistic insights into drug responses that are missed by traditional bulk viability assays.

Key Results

  • High-Throughput Screening: The study generated over 2,500 single-cell PTM signaling, DNA-damage, cell-cycle, and apoptosis response datasets from drug-treated PDOs and CAFs.
  • Trellis Performance: Trellis demonstrated superior ability to distinguish treatment effects and reduce batch effects compared to other methods, enabling the analysis of thousands of single-cell conditions.
  • CAF-Mediated Resistance: CAFs were found to protect chemosensitive PDOs from apoptosis induced by SN-38, 5-FU, and oxaliplatin in a patient-specific manner.
  • YAP Inhibition Re-sensitization: Inhibiting YAP with Verteporfin completely re-sensitized CAF-protected PDOs to SN-38-induced apoptosis, without affecting on-target DNA damage or cell-cycle arrest.

Limitations and Future Work

  • Generalizability of CAFs: The study suggests that CAF effects are patient-specific and should not be generalized across all patients or cancer types. Future work should explore this further.
  • TME Complexity: While CAFs were investigated, other TME components like immune cells were not included. The authors propose that single-cell technologies could be extended to analyze organoid-leukocyte co-cultures for a more comprehensive understanding.
  • Therapeutic Applications: The authors suggest that Trellis and single-cell PTM analysis could be adapted to study biological therapies, such as CAR-T cells, and to rapidly identify rational drug synergies for overcoming therapy resistance.
  • Broader Application of Trellis: Trellis's tree-based structure is flexible and could be adapted for other biological questions, such as immune profiling or kinase inhibitor screens, by designing different branching structures.

Summary updated

Abstract

Abstract Patient-derived organoids (PDOs) can model personalized therapy responses, however current screening technologies cannot reveal drug response mechanisms or study how tumor microenvironment cells alter therapeutic performance. To address this, we developed a highly-multiplexed mass cytometry platform to measure post translational modification (PTM) signaling in >2,500 colorectal cancer (CRC) PDOs and cancer-associated fibroblasts (CAFs) in response to clinical therapies at single-cell resolution. To compare patient- and microenvironment-specific drug responses in thousands of single-cell datasets, we developed Trellis — a highly-scalable, hierarchical tree-based treatment effect analysis method. Trellis single-cell screening revealed that on-target cell-cycle blockage and DNA-damage drug effects are common, even in chemorefractory PDOs. However, drug-induced apoptosis is patient-specific. We found drug-induced apoptosis does not correlate with genotype or clinical staging but does align with cell-intrinsic PTM signaling in PDOs. CAFs protect chemosensitive PDOs by shifting cancer cells into a slow-cycling cell-state and CAF chemoprotection can be reversed by inhibiting YAP. Highlights >2,500 single-cell PTM signaling, DNA-damage, cell-cycle, and apoptosis responses from drug-treated PDOs and CAFs. Trellis: hierarchical tree-based treatment effect method for single-cell screening analysis. PDOs have patient-specific drug responses that align with cell-intrinsic PTM signaling states. CAFs chemoprotect PDOs by altering PDO cell-state via YAP signaling.
BibTeX
@misc{zapatero2022trellis,
  author = {María Ramos Zapatero and Alexander Tong and Jahangir Sufi and Petra Vlckova and Ferran Cardoso Rodriguez and Callum Baird Nattress and Xiao Qin and Daniel Hochhauser and Smita Krishnaswamy and Christopher J. Tape},
  title = {{Trellis Single-Cell Screening Reveals Stromal Regulation of Patient-Derived Organoid Drug Responses}},
  year = {2022},
  doi = {10.1101/2022.10.19.512668},
  howpublished = {bioRxiv},
}
Figure