# Stromal cell-expressed malignant gene patterns contribute to the progression of squamous cell carcinomas across different sites

## TL;DR

This paper shows that stromal cell gene patterns in squamous cell carcinomas affect cancer progression and could improve classification systems.

## Contribution

The study identifies two novel stromal gene patterns (V1 and V5) linked to SCC progression and prognosis across multiple cancer types.

## Key findings

- Pattern V5 is associated with mixed molecular subtypes in HNSCC and shorter PFI with gender and cancer-type specificity.
- Pattern V1 is linked to poor PFI in HNSCC, LUSC, and CESC patients after surgery.
- Adjuvant radiotherapy may reduce the recurrence and metastasis effects of high V1 expression in HNSCC.

## Abstract

Squamous cell carcinomas (SCCs) across different anatomical locations possess common molecular features. Recent studies showed that stromal cells may contribute to tumor progression and metastasis of SCCs. Limited by current sequencing technology and analysis methods, it has been difficult to combine stroma expression profiles with a large number of clinical information.

With the help of transfer learning on the cell line, single-cell, and bulk tumor sequencing data, we identified and validated 2 malignant gene patterns (V1 and V5) expressed by stromal cells of SCCs from head and neck (HNSCC), lung (LUSC), cervix (CESC), esophagus, and breast.

Pattern V5 reflected a novel malignant feature that explained the mixed signals of HNSCC molecular subtypes. Higher expression of pattern V5 was related to shorter PFI with gender and cancer-type specificity. The other stromal gene pattern V1 was associated with poor PFI in patients after surgery in all the three squamous cancer types (HNSCC p = 0.0055, LUSC p = 0.0292, CESC p = 0.0451). Cancer-associated fibroblasts could induce HNSCC cancer cells to express pattern V1. Adjuvant radiotherapy may weaken the effect of high V1 on recurrence and metastasis, depending on the tumor radiosensitivity.

Considering the prognostic value of stromal gene patterns and its universality, we suggest that the genetic subtype classification of SCCs may be improved to a new system that integrates both malignant and non-malignant components.