Single-cell metabolic fingerprints discover a cluster of circulating tumor cells with distinct metastatic potential

Colorectal cancer metastasis is a leading cause of death, and early detection is crucial. While circulating tumor cells (CTCs) are recognized as metastasis seeds, simply counting CTCs may not accurately predict metastatic risk due to CTC heterogeneity. Current methods like MRI and CT often miss small lesions. This research aimed to develop a molecular typing system that predicts colorectal cancer metastasis based on the metabolic profiles of individual CTCs, offering a more precise approach for early risk assessment and improved patient outcomes. The study addresses the unmet need for a more accurate and sensitive method for early detection of colorectal cancer metastasis, potentially improving prognosis and treatment strategies by focusing on the metabolic heterogeneity of CTCs.

Existing literature highlights the importance of CTCs in metastasis, but acknowledges the limitations of using CTC count alone as a prognostic indicator. Studies have demonstrated the heterogeneity of CTCs, with variations in their metastatic potential. Previous work has utilized mass spectrometry-based metabolomics to study cancer cell metabolism, but quantitative, single-cell analysis has been challenging. This study builds upon these previous efforts by developing a novel methodology for high-throughput, single-cell metabolic analysis of CTCs, using a home-built platform, allowing for the identification of metabolic subgroups within the CTC population with varying metastatic potential.

The study employed a multi-step methodology. First, mass spectrometry-based untargeted metabolomics was used to profile two pairs of human colorectal cancer cell lines with differing metastatic abilities (SW480 vs. SW620; HT-29 vs. COLO 205). This identified metabolites with differential abundance correlated with metastatic potential. Second, a home-built single-cell quantitative mass spectrometry platform was developed to overcome challenges in single-cell analysis, such as controlled subpicoliter extraction, adjustment of non-biological variations, and improved repeatability and sensitivity. Electro-osmotic extraction using nanocapillaries was employed, along with careful optimization of multiple reaction monitoring (MRM) transitions and internal standards to ensure accurate quantification. Third, CTCs were isolated from colorectal cancer patients, and the levels of 14 selected metabolites were measured in individual CTCs using the developed platform. Finally, a machine learning approach combining non-negative matrix factorization (NMF) and logistic regression was used to cluster the CTCs into subgroups based on their metabolic profiles. The resulting subgroups were then evaluated for their association with metastasis using in vitro and in vivo functional assays, including transwell assays and establishment of CTC-derived explant (CDX) models. The performance of the molecular typing system was validated in a test cohort and an independent prospective cohort.

Untargeted metabolomics of colorectal cancer cell lines identified 14 metabolites significantly associated with metastatic potential, primarily enriched in amino acid and glutathione metabolism pathways. The home-built single-cell quantitative mass spectrometry platform achieved high sensitivity, reproducibility, and an extraction efficiency of 60%. NMF analysis of single CTC metabolic profiles revealed two distinct subgroups (C1 and C2). The number of CTCs in the C2 subgroup was significantly associated with metastasis incidence in both the training and test cohorts (p < 0.0001). In vitro and in vivo experiments confirmed that the C2 subgroup exhibited increased proliferative and migratory capacities. The 4-metabolite fingerprint classifier (glutamic acid, lactic acid, aspartic acid, and malic acid) achieved high sensitivity (78.8%), specificity (96.3%), accuracy (86.7%), and AUC (0.927) in predicting metastasis. The study also found that total CTC count alone was a less effective predictor compared to the C2 CTC count.

The findings demonstrate that a specific metabolic fingerprint at the single-cell level can accurately predict colorectal cancer metastasis potential, surpassing the predictive power of total CTC counts alone. This highlights the importance of considering cellular heterogeneity in assessing metastatic risk. The identified metabolites and pathways are biologically plausible, linking to known processes involved in cancer metastasis, such as amino acid metabolism and the Warburg effect. The success of the home-built single-cell platform offers a valuable tool for future studies exploring single-cell metabolic heterogeneity in various cancers. The high accuracy of the C2 CTC count as a predictor offers potential for improved clinical risk stratification and personalized treatment strategies.

This study successfully established a molecular typing system for predicting colorectal cancer metastasis based on single CTC metabolic fingerprints. The high predictive accuracy of the C2 CTC subgroup, characterized by specific metabolite levels, surpasses the predictive power of traditional methods using only total CTC counts. This innovative approach offers valuable insights into metastasis mechanisms and holds significant potential for improving colorectal cancer diagnosis, prognosis, and therapeutic strategies. Future research could focus on exploring the underlying metabolic mechanisms driving the distinct metastatic potential of the C2 subgroup and validating these findings in larger, multi-center clinical trials.

The study's sample size, while substantial, might be considered limited for definitive conclusions. The generalizability of these findings across diverse colorectal cancer subtypes and patient populations requires further investigation. The study primarily focuses on colorectal cancer; further research is needed to determine whether this approach is applicable to other cancer types.