# Mass spectrometry-based multi-omics analysis elucidates immune microenvironmental characteristics and the risk of distant metastasis in N1c colorectal cancer

## TL;DR

This study uses mass spectrometry and machine learning to uncover molecular and immune features of N1c colorectal cancer, identifying genes and proteins linked to metastasis and poor immunotherapy response.

## Contribution

The study introduces a novel machine learning model for predicting prognosis and immunotherapy response in N1c CRC, and identifies PLOD1 as a potential biomarker and therapeutic target.

## Key findings

- N1c CRC samples show increased ECM remodeling and EMT, linked to metastasis and poor prognosis.
- A machine learning model based on 10 TDRGs accurately predicts DFS and immune activity across multiple cohorts.
- PLOD1 expression in fibroblasts correlates with collagen fiber abundance and promotes CRC cell migration and invasion.

## Abstract

Colorectal cancer (CRC) ranks third in global cancer incidence and second in mortality. Tumor deposit (TD), a specific regional spread form, is a crucial independent risk factor for survival and have biological differences from lymph node metastasis (LNM). However, it is underestimated in current staging systems, which results in biased treatment decisions and prognosis evaluation. Moreover, the biological features and distant metastasis patterns of N1c CRC remain largely unknown.

We performed Data-independent Acquisition Mass Spectrometry (DIA-MS) analysis of formalin-fixed, paraffin-embedded (FFPE) samples from 13 T1-T4N1cM1 CRC patients to reveal their molecular characteristics. 9 machine learning algorithms identified 10 TD-related metastasis genes (TDRGs), the multi-cohort validation in 1,582 CRC patients confirmed their role in prognosis. Immune landscape and immunotherapy response were assessed by the CIBERSORT, tumor mutation burden (TMB), Tumor Immune Dysfunction and Exclusion (TIDE) score, Consensus molecular subtype (CMS), immune checkpoint gene expression. scRNA-seq analysis identified Procollagen-Lysine,2-Oxoglutarate 5-Dioxygenase 1 (PLOD1) expression in CRC, Immunohistochemical staining (IHC) and Masson’s trichrome staining were used to assess PLOD1 expression and collagen fiber content in CRC, its role in tumor invasion and migration was elucidated by wound healing and transwell assays.

N1c samples exhibit enhanced extracellular matrix (ECM) remodeling and epithelial mesenchymal transition (EMT). The TDRGs identified by machine learning robustly predicting disease-free survival (DFS) across multiple cohorts (Mean C-index = 0.72) and immune activity. A higher risk score predicted early metastasis and poorer response to immunotherapy, marked by lower level of immune infiltration, higher TIDE scores, lower TMB, and downregulated immune checkpoint genes. scRNA-seq analysis pinpoints highest PLOD1 expression in fibroblasts. Histological analysis of N1c samples demonstrated PLOD1 expression patterns and their significant correlation with stromal collagen fiber abundance (p < 0.01). Wound healing and transwell assays indicating the knockdown of PLOD1 hinders the migration and invasion of CRC DLD-1 cell.

We assessed the protein expression profiles and pathway characteristics of N1c CRC. The model developed based on TDRGs, effectively predicted DFS and immunotherapy response, supporting precision treatment and staging system optimization. PLOD1’s role in ECM remodeling and CRC cell migration and invasion suggests its potential as a prognostic biomarker and therapeutic target.