Integrating Proteomics and Machine Learning Reveals Characteristics and Risks of Lymph Node-independent Distant Metastasis in Colorectal Cancer

Baiwang Zhu¹Chenxiao Zheng²Yanyu Chen³Numan Shahid⁴Yiwang Hu²Hajar Mansoor Ahmed Ali Husain⁵Binbin Ou⁶Qiongying Zhang⁷Haobo Jin⁸Yating Zheng⁹Peng Li^7*Yifei Pan^2*Xiaodong Zhang^9*

• ¹Department of Colorecatl Anal Surgery, First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China
• ²Colorectal Anal Surgery, First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China
• ³Hangzhou Medical University, Hangzhou, China
• ⁴Department of Hepato-Pancreato-Biliary (HPB) Surgery, King’s College Hospital., London, United Kingdom
• ⁵Department of Hepatology, King's College Hospital., London, United Kingdom
• ⁶General Practice Department, Hangzhou Gongshu Hospital of Integrated Traditional and Western Medicine, Hangzhou, China
• ⁷Department of Pathology, First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China
• ⁸Laboratory Animal Centre, Wenzhou Medical University, Wenzhou, China
• ⁹Department of Colorectal Anal Surgery, First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China

Metastatic colorectal cancer (mCRC) poses significant treatment challenges, especially liver metastasis (CRLM). A notable proportion of CRC has synchronous metastasis independent of lymph node metastasis (LNM). The biological traits of lymph node-independent metastasis in CRC are unclear, and early synchronous metastasis is hard to predict with current imaging or clinicopathological methods.We collected samples from 12 CRC patients with synchronous distant metastasis without LNM (T1-3N0M1). Data-Independent Acquisition Mass Spectrometry (DIA-MS), multi-omics data integration, and machine learning were used to develop a Lymph node-Independent Metastasis Genes (LIMGs) signature to predict synchronous distant metastasis risk in stage I-II CRC patients and validate it in multi-cohort. Immune microenvironment across risk subgroups was calculated by Estimating Relative Subsets of RNA Transcripts (CIBERSORT).Tumor Mutation Burden (TMB), Microsatellite Instability (MSI) score, immune functions and immune checkpoint gene expression were analyzed to evaluate immunotherapy response.Single cell RNA sequencing (scRNA-seq) analysis illustrated the expression profile of integrin α11 (ITGA11) in CRC. Immunohistochemistry (IHC) confirmed its expression pattern, while wound healing and transwell assays elucidated the role of ITGA11 in CRC metastasis.The LIMGs signature demonstrated strong predictive performance of lymph nodeindependent synchronous metastasis across cohorts. The high-risk subgroup exhibited enhanced extracellular matrix (ECM) remodeling, epithelial-mesenchymal transition (EMT) and correlated with immunosuppressive tumor microenvironment (TME), lower TMB and MSI score, indicating worse immunotherapy response. Additionally, machine learning reveal ITGA11's pivotal role in lymph node-independent metastasis. IHC scores showing significant discriminatory ability of ITGA11 across different samples. Wound healing and transwell assays reveal that the knockdown of ITGA11 hinders the migration and invasion of CRC SW480 cells.Our findings suggest that EMT-related signature LIMGs significantly affects lymph nodeindependent distant metastasis in CRC and effectively predicts non-LNM synchronous metastasis in stage I-II CRC patients. LIMG ITGA11 may promote early metastasis by enhancing migration and invasion. These offering insights into precise risk stratification and treatment for CRC patients.