Polyclonal and clonal organoid models of Barrett oesophagus and oesophageal adenocarcinoma reveal heterogeneity in progression and therapy response

This study establishes a comprehensive biobank of polyclonal and single-cell-derived clonal organoids from normal tissue, Barrett's esophagus, and esophageal adenocarcinoma that faithfully recapitulate disease heterogeneity and molecular features, enabling the isolation of high-risk subclones and the evaluation of diverse responses to cancer therapies.

The Gist

Imagine your body is a vast, bustling city. The lining of your esophagus (the food pipe) is usually a sturdy, brick-walled neighborhood (squamous epithelium). But for some people, due to chronic acid reflux (like a constant flood of acidic rain), this neighborhood gets damaged. In an attempt to survive, the cells try to rebuild themselves using materials from a different part of the city—the stomach. This new, mismatched neighborhood is called Barrett's Esophagus (BO).

The big problem is that while most of this new neighborhood is fine, a few "rogue construction crews" (cells) sometimes decide to turn into a dangerous, chaotic city of crime and destruction: Esophageal Adenocarcinoma (OAC).

The challenge for doctors has been that this transition is incredibly messy and unpredictable. It's like trying to predict a riot by looking at a single, blurry photo of a crowd. Every patient's "riot" looks different, and the cells are constantly shifting, making it hard to test new medicines or understand exactly how the trouble starts.

This paper is about building a giant, high-tech "mini-city" laboratory to solve this mystery. Here is the breakdown in simple terms:

1. The "Mini-City" Lab (Organoids)

Instead of using animals (which are like using a model of a different city to study traffic jams in New York), the scientists grew Organoids.

• The Analogy: Think of these as tiny, 3D Lego cities grown from a patient's own cells. They are small enough to fit in a test tube but complex enough to act like real human tissue.
• The Achievement: They built a massive "biobank" of 116 of these mini-cities. Some represent healthy stomach tissue, some represent the early warning stage (Barrett's), and some represent the full-blown cancer. This allows them to study the entire journey from a healthy neighborhood to a crime-ridden city.

2. Finding the "Hidden Criminals" (Clonal Organoids)

One of the biggest headaches in cancer research is heterogeneity.

• The Analogy: Imagine a crowd of 1,000 people. If you take a photo of the whole group, you see a blur. You might miss the one person in the back wearing a red hat who is actually the leader of the gang. In a standard "bulk" sample, the dangerous cells get drowned out by the normal ones.
• The Innovation: The scientists developed a new trick called Single-Cell Derived Organoids. They took just one single cell from the crowd and grew a whole new mini-city from it.
• Why it matters: This allowed them to isolate the "red hat" leaders. They found dangerous, high-risk cells that were hiding in the crowd, cells that standard tests would have missed. They could watch these specific "rogue crews" evolve and see exactly how they become dangerous.

3. The "City Planning" Report (Genomics and Morphology)

The scientists didn't just grow the cities; they mapped them.

• The DNA Map: They read the "blueprints" (DNA) of every mini-city. They found that even though the mini-cities looked like the original patient tissue, they had their own unique quirks. Some had lost their "police force" (tumor suppressor genes like TP53), while others had extra "construction equipment" (gene amplifications).
• The Shape Shifters: They used time-lapse cameras to watch the cities grow.
  • Healthy/Early Stage Cities: Looked like neat, organized houses with a central courtyard (a "lumen").
  • Cancer Cities: Looked like solid, chaotic concrete blocks with no courtyards.
  • The Insight: They even built a computer simulation (a video game version) to prove that when the cells lose their "sense of direction" (polarity), the neat houses collapse into solid blocks. This explains how the tissue physically changes shape as it becomes cancerous.

4. The "Stress Test" (Therapy Response)

The most exciting part is testing how these mini-cities react to medicine.

• The Analogy: Doctors usually give the same chemotherapy drug to everyone, hoping it works. It's like giving the same key to every lock in a building, hoping one opens.
• The Experiment: They took their mini-cities and hit them with the standard cancer drugs (chemotherapy, radiation, and targeted pills).
• The Result: The results were all over the place!
  • Some mini-cities were destroyed by the drugs.
  • Some were completely immune.
  • Some were sensitive to radiation but not to pills.
  • The Takeaway: This proves that one size does not fit all. Because every patient's "mini-city" is genetically unique, they will react differently to treatment. This paves the way for Precision Medicine: testing a patient's specific mini-city in the lab first to see which drug works, before ever giving it to the patient.

5. The "Neighborhood Watch" (Assembloids)

Finally, they realized cancer doesn't happen in a vacuum. It needs help from the surroundings.

• The Analogy: A criminal gang doesn't just exist; they need a corrupt mayor or a bribed police officer to help them. In the body, these helpers are fibroblasts (support cells).
• The Experiment: They mixed the cancer mini-cities with these support cells to create "Assembloids" (a hybrid city).
• The Discovery: They found that the cancer cells could actually brainwash the support cells, turning them into "enablers" that help the cancer grow and spread. This helps explain why the disease gets worse over time.

The Bottom Line

This paper is a massive leap forward. The scientists have built a playground of 116 different "mini-patients" that perfectly mimic the real disease.

• They can see the hidden dangers (single-cell clones).
• They can watch the crime unfold (growth and shape changes).
• They can test the weapons (drugs) before using them on real people.

It's like having a crystal ball that lets doctors see exactly how a specific patient's cancer will behave and which medicine will stop it, moving us away from "guessing" and toward "knowing."

Technical Summary

1. Problem Statement

Oesophageal adenocarcinoma (OAC) is a lethal malignancy with a poor five-year survival rate (<20%). Its precursor lesion, Barrett's oesophagus (BO), exhibits substantial inter- and intra-lesional heterogeneity at both phenotypic and molecular levels. This heterogeneity complicates early detection, risk stratification, and the development of targeted therapies.
Current limitations in modeling this disease include:

• Animal models: Poorly recapitulate human disease heterogeneity and are resource-intensive.
• Existing Organoid Models: While Patient-Derived Organoids (PDOs) exist for OAC, they often fail to capture the full spectrum of heterogeneity in the pre-malignant (BO) stages. Bulk PDO cultures may mask rare, high-risk subclonal populations that drive progression, and they lack the stromal microenvironment (fibroblasts) crucial for disease pathogenesis.
• Therapeutic Prediction: It remains uncertain whether existing PDO models can reliably predict the variable responses of OAC patients to standard systemic therapies (chemotherapy, radiotherapy) and targeted agents.

2. Methodology

The authors established a comprehensive biobank and developed novel modeling techniques to address these gaps:

• PDO Biobank Generation:
  • Generated 116 organoids spanning the disease spectrum: normal gastric tissue, non-dysplastic BO, dysplastic BO, and treatment-naïve/treated OAC.
  • Established high-efficiency derivation protocols for BO (68% for non-dysplastic, 61% for dysplastic) using specific culture media supplements (Gastrin, FGF10).
• Single-Cell Derived Clonal Organoids (sc-organoids):
  • Developed a microdissection technique to isolate individual organoids from bulk cultures and expand them into clonal lines.
  • Generated sc-organoid panels from both non-dysplastic and dysplastic BO tissues to isolate and maintain high-risk subclonal populations often lost in bulk passaging.
• Assembloid Co-cultures:
  • Created "assembloids" by co-culturing BO/OAC PDOs with primary human oesophageal fibroblasts.
  • Applied microenvironmental stressors (reflux acid/bile salts and IL-10) to study epithelial-stromal interactions and fibroblast reprogramming.
• Multi-Omics Characterization:
  • Genomics: Whole Genome Sequencing (WGS) to assess mutational burden, driver gene alterations (e.g., TP53, CDKN2A), copy number variations, and mutational signatures (e.g., SBS17).
  • Transcriptomics: RNA-seq to analyze lineage markers (gastric vs. intestinal), differential gene expression, and pathway enrichment.
  • Phenotyping: High-resolution time-lapse microscopy and in silico Cellular Potts Monte Carlo modeling to analyze morphology, growth kinetics, and polarity.
• Therapeutic Screening:
  • Tested OAC PDOs against standard-of-care chemotherapy (FLOT regimen components: docetaxel, 5-FU, oxaliplatin), external beam radiotherapy (EBRT), and targeted CDK4/6 inhibition (abemaciclib).

3. Key Contributions

1. Scalable Biobank: Creation of the largest and most diverse PDO resource for BO and OAC to date, covering normal, pre-malignant, and malignant stages.
2. Clonal Resolution: Introduction of sc-organoids as a tool to unmask and maintain rare, high-risk subclones (e.g., TP53 or APC mutants) that are invisible in bulk sequencing but critical for understanding progression.
3. Quantitative Framework: Development of a multi-level toolkit (genomic, transcriptomic, morphological) to accurately classify the disease state of an organoid, acknowledging that a single feature is insufficient due to heterogeneity.
4. Stromal Modeling: Demonstration that BO PDOs can reprogram naïve fibroblasts into a pro-tumorigenic niche (S3 subtype) via assembloids, mimicking the in vivo microenvironment.
5. Therapeutic Heterogeneity Mapping: Systematic mapping of how molecular and phenotypic diversity translates to variable drug and radiation sensitivity across the patient population.

4. Key Results

• Genomic Fidelity and Heterogeneity:
  • PDOs recapitulated the mutational signatures of their parent tissues (high cosine similarity, particularly for SBS17a/b).
  • Clonal Drift: Bulk PDOs showed temporal evolution and clonal drift. sc-organoids revealed that bulk cultures often miss high-risk subclones; for example, one sc-organoid harbored TP53 and SMARCA4 mutations absent in the bulk culture and matching tissue.
  • Concordance: While mutational signatures were highly concordant, specific driver mutations (like TP53) and copy number alterations showed variable concordance between PDOs and matching tissues, highlighting the need for direct organoid profiling.
• Morphology and Growth Kinetics:
  • Non-dysplastic BO: Exhibited four morphologies (single-layer single-lumen, single-layer multi-lumen, multi-layer lumen, solid). Luminal types expanded robustly; solid types showed minimal growth.
  • Dysplastic BO & OAC: Showed a loss of single-layer multi-lumen structures and a dominance of solid morphology (99% in OAC).
  • In Silico Modeling: Confirmed that loss of cellular polarity drives the transition from luminal to solid structures.
• Transcriptomic Landscape:
  • BO PDOs displayed an intermediate gastric-intestinal phenotype.
  • Dysplastic PDOs showed upregulation of HOX genes (HOXB2, B6, B8) and pathways related to development and migration, distinguishing them from non-dysplastic BO.
• Stromal Reprogramming:
  • BO PDOs, particularly under reflux acid stress, induced a shift in fibroblasts from the CD34+ S1 subset to the pro-tumorigenic αSMA+ S3 subset. Normal gastric PDOs required combined stressors to induce this shift, whereas BO PDOs did so more readily.
• Therapeutic Response:
  • Chemotherapy: OAC PDOs showed variable sensitivity to 5-FU and oxaliplatin (IC50 ranges of 8–51 µM and 7–96 µM, respectively), though docetaxel sensitivity was more uniform.
  • Radiotherapy: Significant variation in organoid forming efficiency after EBRT (AUC 0.46–0.98).
  • Targeted Therapy: Response to CDK4/6 inhibitors (abemaciclib) correlated with specific genomic vulnerabilities (e.g., CDK6 amplification or KRAS ecDNA), but sensitivity varied widely even among genetically similar models.
  • Lack of Correlation: Sensitivity to chemotherapy did not predict sensitivity to radiotherapy or targeted agents, supporting the need for multi-modal testing.

5. Significance

• Precision Oncology Platform: This study provides a scalable, functionally validated platform for testing personalized therapeutic strategies in OAC, moving beyond bulk averages to account for subclonal heterogeneity.
• Mechanistic Insight: The use of sc-organoids and assembloids offers a new window into the early drivers of malignant transformation and the role of the microenvironment in BO progression, which was previously difficult to model.
• Clinical Translation: The demonstrated variability in drug response mirrors clinical reality, suggesting that PDO biobanks can be used to stratify patients for clinical trials or to guide treatment selection, particularly for the 50% of OAC cases with potential CDK4/6 vulnerabilities.
• Methodological Advance: The integration of single-cell derivation, in silico morphological modeling, and stromal co-culture sets a new standard for modeling complex, heterogeneous epithelial cancers.

In conclusion, this work establishes a robust, multi-dimensional resource that captures the full biological complexity of the Barrett's to OAC progression, enabling deeper mechanistic studies and more accurate prediction of therapeutic outcomes.