Zhilin Sha,1,* Qingxiang Gao,1,* Lei Wang,2,* Ni An,3 Yingjun Wu,1 Dong Wei,4 Tong Wang,5 Chen Liu,1 Yang Shen1 

1Department I of Biliary Tract Surgery, Eastern Hepatobiliary Surgery Hospital, Naval Medical University, Shanghai, People’s Republic of China; 2Department of General Surgery, Yancheng Hospital of Traditional Chinese Medicine, Yancheng, Jiang Su, People’s Republic of China; 3Department of Anesthesiology, the Eighth Medical Center of Chinese PLA General Hospital, Beijing, People’s Republic of China; 4Department of General Surgery (Second Ward), the No.1 People’s Hospital of Pinghu, Pinghu, Zhe Jiang, People’s Republic of China; 5Department of Anesthesiology, No.32295 Troop of Chinese PLA, Liaoyang, People’s Republic of China

*These authors contributed equally to this work

Correspondence: Yang Shen, Eastern Hepatobiliary Surgery Hospital, Naval Medical University, No. 700 of North Moyu Road, Jia Ding District, Shanghai, 201823, People’s Republic of China, Email drbruceshen@163.com Chen Liu, Eastern Hepatobiliary Surgery Hospital, Naval Medical University, No. 700 of North Moyu Road, Jia Ding District, Shanghai, 201823, People’s Republic of China, Email dinoliu@126.com

Background: Colorectal cancer (CRC) is one of the deadliest causes of death by cancer worldwide. Liver metastasis (LM) is the main cause of death in patients with CRC. Therefore, identification of patients with the greatest risk of liver metastasis is critical for early treatment and reduces the mortality of patients with colorectal cancer liver metastases.
Methods: Initially, we characterized cell composition through single-cell transcriptome analysis. Subsequently, we employed copy number variation (CNV) and pseudotime analysis to delineate the cellular origins of LM and identify LM-related epithelial cells (LMECs). The LM-index was constructed using machine learning algorithms to forecast the relative abundance of LMECs, reflecting the risk of LM. Furthermore, we analyzed drug sensitivity and drug targeted gene expression in LMECs and patients with a high risk of LM. Finally, functional experiments were conducted to determine the biological roles of metastasis-related gene in vitro.
Results: Single-cell RNA sequencing analysis revealed different immune landscapes between primary CRC and LM tumor. LM originated from chromosomal variants with copy number loss of chr1 and chr6p and copy number gain of chr7 and chr20q. We identified the LMECs cluster and found LM-associated pathways such as Wnt/beta-catenin signaling and KRAS signaling. Subsequently, we identified ten metastasis-associated genes, including SOX4, and established the LM-index, which correlates with poorer prognosis, higher stage, and advanced age. Furthermore, we screened two drugs as potential candidates for treating LM, including Linsitinib_1510, Lapatinib_1558. Immunohistochemistry results demonstrated significantly elevated SOX4 expression in tumor samples compared to normal samples. Finally, in vitro experiments verified that silencing SOX4 significantly inhibited tumor cell migration and invasion.
Conclusion: This study reveals the possible cellular origin and driving factors of LM in CRC at the single cell level, and provides a reference for early detection of CRC patients with a high risk of LM.

Keywords: colorectal cancer, liver metastasis, single-cell sequencing, SOX4, prognostic