Esophageal epithelial cell-state transitions underlie the severity of pediatric eosinophilic esophagitis

This study reveals that in pediatric eosinophilic esophagitis, disease severity is driven not merely by eosinophil burden but primarily by epithelial cell-state transitions involving impaired differentiation, metabolic stress, and structural remodeling, offering a new framework for understanding and treating the condition.

The Gist

The Big Picture: What is EoE?

Imagine the esophagus (the tube connecting your throat to your stomach) as a highway. In a healthy person, cars (food) drive smoothly down this highway.

In Eosinophilic Esophagitis (EoE), the highway gets clogged with construction crews (immune cells called eosinophils) and debris. This causes traffic jams, potholes, and eventually, the road gets so damaged it starts to narrow or close up completely. This makes it hard for kids to eat, causing pain, vomiting, and poor growth.

Doctors have traditionally measured how bad the traffic jam is by simply counting the number of construction workers (eosinophils) in a tiny sample of the road. They call this the "Peak Eosophil Count" (PEC).

The Problem: Counting the workers doesn't tell the whole story. You might have a few workers, but if they are tearing up the asphalt and destroying the guardrails, the road is still in terrible shape. Conversely, you might have a lot of workers, but if the road is still smooth, the damage isn't as bad.

The New Discovery: It's About the Road, Not Just the Workers

This study looked at 121 children with EoE and used advanced genetic "microscopes" to look at the actual road surface (the esophageal cells), not just the workers. They used a new scoring system called I-SEE, which measures the total damage (symptoms, endoscopic look, and tissue damage), not just the worker count.

The Main Finding:
The severity of the disease isn't just about how many immune cells are there. It's about how the road itself is reacting.

When the disease gets severe, the road cells (epithelial cells) stop acting like a smooth, protective highway. Instead, they start acting like panicked construction workers who are:

1. Trying to rebuild too fast (proliferating).
2. Confused and losing their identity (transitional states).
3. Stressed and breaking apart (metabolic dysfunction).

The "Construction Site" Analogy

Think of the esophagus as a construction site.

• The Old View (PEC): The foreman only counts how many hard hats are on the site. "Oh, there are 50 hard hats, so the site is a disaster!"
• The New View (I-SEE & This Study): The foreman looks at the behavior of the workers.
  • Mild Disease: The workers are there, but the road is mostly fine.
  • Severe Disease: The workers aren't just standing around; they are panic-building. They are frantically trying to lay new pavement (cell division) but doing it so poorly that the new road is bumpy, weak, and full of holes. They are also tearing down the old guardrails (barrier function) and getting so stressed they are running out of fuel (metabolic stress).

The study found that the most severe cases were driven by these "panic-building" cells, not just the number of immune cells.

The "Scissor" Tool: Finding the Culprits

The researchers used a clever computer tool called Scissor. Imagine you have a giant pile of mixed-up puzzle pieces from a construction site (a mix of immune cells, blood vessels, and road cells). You want to find the specific pieces that cause the worst traffic jams.

The "Scissor" tool looked at the genetic blueprints of every single cell and asked: "Which of these cells looks most like the ones found in the kids with the worst symptoms?"

The Result:
The tool pointed its finger almost exclusively at the road cells (epithelial cells), specifically two types:

1. The "New Hires" (Proliferating cells): Cells that are dividing rapidly but haven't learned their job yet. They are trying to fix the road but lack the skills, leading to a messy, unstable surface.
2. The "Confused Interns" (Transitional cells): Cells that are in the middle of changing from one type to another. They have lost their old identity (protection) but haven't gained a new one, leaving the road vulnerable.

Why This Matters

1. We need a new ruler.
Just counting immune cells (PEC) is like judging a house fire only by the number of firefighters present. You need to look at the damage to the house (the epithelial cells) to know how bad the fire really is.

2. New ways to treat it.
Currently, most treatments try to tell the firefighters to go home (suppress the immune system). This study suggests we also need to help the construction crew learn their jobs.

• Instead of just calming the immune system, maybe we need therapies that help the "panic-building" cells calm down and learn to build a smooth, strong road again.
• We might need to fix the "fuel supply" (metabolism) so the cells don't get stressed and break down.

The Takeaway

This paper tells us that in severe pediatric EoE, the real trouble isn't just the immune system attacking the body. The trouble is that the body's own lining is breaking down, panicking, and trying to rebuild itself in a chaotic way.

By understanding that the "road" itself is the problem, doctors can develop better ways to measure the disease and create treatments that actually fix the road, not just clear out the traffic.

Technical Summary

1. Problem Statement

Eosinophilic Esophagitis (EoE) is a chronic, immune-mediated inflammatory disease. While the Peak Eosinophil Count (PEC) in biopsies is the standard metric for diagnosing and assessing disease activity, it fails to capture the full spectrum of clinical severity, which includes symptoms, endoscopic findings, and complications (e.g., strictures, malnutrition).

• The Gap: The Index of Severity for Eosinophilic Esophagitis (I-SEE) was developed to provide a multidimensional severity score. However, the specific molecular and cellular mechanisms driving I-SEE-defined severity, particularly in pediatric populations, remain poorly understood. Previous studies have relied on pre-defined gene panels or mixed adult/child cohorts, lacking an unbiased, pediatric-specific transcriptomic view linking cellular states to clinical severity.

2. Methodology

The study employed an integrative multi-omics approach combining bulk transcriptomics, clinical phenotyping, and single-cell resolution analysis.

• Cohort: 121 pediatric patients (ages 6–18) from Vanderbilt University Medical Center.
  • Groups: 75 EoE patients (43 active EoE [aEoE] with $\ge$15 eos/hpf; 32 inactive EoE [iEoE] with <15 eos/hpf) and 46 controls.
  • Data: Bulk RNA sequencing (RNA-seq) of distal esophageal biopsies and comprehensive clinical metadata used to calculate I-SEE scores.
• Analytical Pipeline:
  1. Bulk RNA-seq Analysis: Differential expression analysis (DESeq2) comparing aEoE, iEoE, and controls. Pathway enrichment (GO, KEGG, Reactome, Hallmark) and correlation with I-SEE scores.
  2. Cell-Type Deconvolution: Used BayesPrism with a matched human esophageal scRNA-seq reference (Rochman et al., 2022) to estimate cell-type proportions in bulk samples.
  3. Phenotype-Cell Integration (Scissor): Applied the Scissor algorithm to link bulk RNA-seq I-SEE scores with the reference scRNA-seq dataset. This identifies specific single cells whose transcriptional profiles correlate most strongly with high or low disease severity.
  4. Subtype-Specific Analysis: Performed pseudo-bulk differential expression and Gene Set Enrichment Analysis (GSEA) on Scissor-identified cell populations (e.g., proliferating vs. transitional epithelial cells) to isolate severity-specific programs.

3. Key Contributions

• Decoupling Activity from Severity: Demonstrated that while bulk transcriptomic profiles primarily reflect disease activity (PEC status), specific epithelial cell-state transitions are the dominant drivers of clinical severity (I-SEE scores).
• Identification of Severity-Associated Cell States: Identified that proliferating and transitional (Trans1) epithelial cells, rather than immune cells, are the primary cellular correlates of severe pediatric EoE.
• Discovery of Novel Pathways: Uncovered epithelial-intrinsic programs (metabolic stress, oxidative stress, junctional remodeling) associated with severity that are masked by inflammatory signals in bulk tissue analysis.
• Pediatric-Specific Insights: Provided the first unbiased, single-cell-informed molecular map of I-SEE severity specifically in children, distinguishing pediatric inflammatory phenotypes from adult fibrostenotic transitions.

4. Key Results

A. Transcriptomic Distinctions of Disease Activity

• Activity Gradient: There is a clear transcriptomic gradient from Control $\to$ iEoE $\to$ aEoE.
• Active Disease Signature: aEoE showed 1,446 differentially expressed genes (DEGs) vs. controls, characterized by upregulation of Type 2 inflammation (IL-4, IL-13, GATA3), mast cell activation, and tissue remodeling (POSTN, LOX).
• Inactive Disease: iEoE samples clustered closely with controls, showing minimal transcriptional changes (only 1 DEG vs. controls), suggesting a largely quiescent state.
• Bulk vs. Severity: Principal Component Analysis (PCA) showed that I-SEE scores did not create a distinct transcriptional cluster separate from activity groups; high-severity patients grouped with their respective activity (aEoE/iEoE) status.

B. Molecular Correlates of I-SEE Severity

• Severity-Specific Genes: 333 genes correlated with I-SEE scores. While many overlapped with activity genes, 28 were unique to severity, highlighting cytoskeletal reorganization (FLNA, KRT1), ECM remodeling (COL1A1, MMP11), and metabolic/oxidative stress (ALOX15, GFPT2).
• Pathways: Severity-associated genes enriched for epithelial cell death, cytoskeletal destabilization, proteolytic activity, and oxidative stress, distinct from canonical Type 2 inflammation.

C. Single-Cell Resolution: The Epithelial Driver

• Cellular Deconvolution: Higher I-SEE scores correlated with increased proportions of proliferating epithelial cells and mast cells, and a decrease in differentiated epithelial cells.
• Scissor Analysis:
  • Dominance of Epithelium: Nearly 100% of cells positively associated with high I-SEE scores (Scissor-positive) belonged to epithelial lineages.
  • Specific States: Scissor-positive cells were enriched in Proliferating (34.2%) and Trans1 (61.3%) epithelial states. Conversely, Scissor-negative cells were almost exclusively Differentiated epithelial cells.
• Distinct Biological Programs:
  • Proliferating Scissor-Positive Cells: Showed upregulation of biosynthetic machinery (cytoplasmic translation, peptide elongation), metabolic stress adaptation (oxidative phosphorylation, mTORC1), and structural remodeling (desmosome organization).
  • Trans1 Scissor-Positive Cells: Exhibited dedifferentiation (loss of mature markers like CRNN, MAL, KRT4), barrier destabilization, and immune-interacting pathways (chemotaxis, cytokine signaling).

D. Bulk vs. Single-Cell Discrepancy

• Bulk RNA-seq was dominated by inflammatory signals (IL responses) and metabolic suppression.
• Single-cell analysis revealed epithelial-intrinsic injury programs (translational activation, junctional remodeling, loss of differentiation) that were invisible in bulk data but critical for understanding disease severity.

5. Significance and Implications

• Reframing EoE Pathogenesis: The study shifts the paradigm from viewing EoE severity solely as an eosinophil-driven inflammatory burden to a model where epithelial remodeling and loss of differentiation are central drivers of clinical severity.
• Clinical Stratification: The findings suggest that I-SEE scores capture a "fibrostenotic" or "remodeling" trajectory driven by immature epithelial states, even in children who typically present with inflammatory phenotypes.
• Therapeutic Targets:
  • Current therapies target cytokines (e.g., anti-IL-13). This study suggests that restoring epithelial differentiation, stabilizing transitional cell populations, or mitigating epithelial metabolic/oxidative stress could be complementary therapeutic strategies.
  • Severity-associated epithelial signatures could serve as biomarkers for predicting disease progression or treatment response.
• Methodological Framework: The integration of bulk transcriptomics with Scissor-based single-cell mapping provides a robust framework for linking complex clinical phenotypes to specific cellular states in other chronic inflammatory diseases.

Limitations: The study is cross-sectional (cannot infer temporal dynamics), has a small number of severe/fibrostenotic cases (typical for pediatrics), and relies on mucosal biopsies which may miss deeper tissue changes. However, the findings were validated against an independent pediatric cohort (Jacobse et al.).