# ESRP1-Associated CD44 Alternative Splicing Stratifies Epithelial–Mesenchymal Identity States in a Non-Transformed Human Cell System

## TL;DR

The study explores how CD44 splicing and ESRP1 relate to epithelial-mesenchymal identity states in non-cancer human cells.

## Contribution

It introduces a non-transformed human cell system to study epithelial–mesenchymal plasticity outside of cancer contexts.

## Key findings

- CD44 splicing patterns correlate with epithelial–mesenchymal identity states in non-transformed cells.
- ESRP1 regulates CD44 splicing, with silencing shifting splicing toward epithelial-associated forms.
- Notch-related transcription varies across identity states, showing context-dependent behavior.

## Abstract

Epithelial–mesenchymal plasticity encompasses a spectrum of epithelial and mesenchymal identity states that enable cells to adapt to changing biological contexts. While CD44 isoform usage and epithelial splicing regulators ESRP1/2 are well-characterized in cancer-associated epithelial–mesenchymal transition (EMT), their regulation across physiological, non-transformed identity states remains less well defined. Here, we employed a non-malignant human cellular system comprising primary dermal fibroblasts, induced pluripotent stem (iPS) cells, and iPS-derived mesenchymal stem cells (iPS-MSCs) to define discrete epithelial, intermediate epithelial/mesenchymal, and mesenchymal identity states positioned along an epithelial–mesenchymal identity axis. Morphological assessment, lineage marker profiling, and RT-qPCR analyses revealed reproducible population-level stratification of these states. CD44 expression and alternative splicing followed this hierarchy, with CD44s predominating in fibroblasts, broad variant exon inclusion in iPS cells, and intermediate patterns in iPS-MSCs. ESRP1 expression mirrored CD44 splicing architecture, and ESRP1 silencing in iPS cells induced a shift toward CD44s, confirming its functional contribution to epithelial-associated CD44 splicing. In contrast, Notch-related transcriptional readouts displayed distinct, context-dependent profiles across the examined identity states. Together, this study establishes a tractable non-transformed human model that captures selected molecular features associated with epithelial–mesenchymal plasticity beyond malignant contexts.