Abstract
Circulating tumor cells (CTCs) are recognized as direct seeds of metastasis. However, CTC count may not be the "best" indicator of metastatic risk because their heterogeneity is generally neglected. In this study, we develop a molecular typing system to predict colorectal cancer metastasis potential based on the metabolic fingerprints of single CTCs. After identification of the metabolites potentially related to metastasis using mass spectrometry-based untargeted metabolomics, setup of a home-built single-cell quantitative mass spectrometric platform for target metabolite analysis in individual CTCs and use of a machine learning method composed of non-negative matrix factorization and logistic regression, CTCs are divided into two subgroups, C1 and C2, based on a 4-metabolite fingerprint. Both in vitro and in vivo experiments demonstrate that CTC count in C2 subgroup is closely associated with metastasis incidence. This is an interesting report on the presence of a specific population of CTCs with distinct metastatic potential at the single-cell metabolite level.