Résumé
Background: In many diseases, including cancer, the number, distribution and shape of mitochondria are affected. Under stress conditions and in tumor cells, changes in cellular, nuclear and mitochondrial morphology are frequently observed. Mitochondria are responsible for energy production and metabolic reprogramming is recognized as a hallmark of cancer, including colorectal cancer (CRC), the third most common and second most deadly cancer worldwide. CRC is a heterogeneous disease, with each subtype exhibiting distinct molecular features that lead to diverse clinical outcomes. The relationship between mitochondrial morphology, metabolic status, and CRC progression has not yet been formally investigated. Here, we sought to determine whether quantitative imaging of mitochondrial shapes, in addition to metabolic measurements, could provide useful information for CRC subtyping. Methods: We recently developed a novel wet-and-dry imaging pipeline (MITOMATICS) that enables the quantitative measurement of a wide range of mitochondrial shapes in their native cellular environment using highcontent confocal microscopy screening. This automated pipeline, which includes statistical tests as well as supervised and unsupervised machine learning tools for analysis and visualization, was applied to monitor mitochondrial morphology in a cellular model of colon cancer progression consisting of various CRC cell lines along with paired non-tumoral cell lines. The metabolic phenotype of the multiple cell subsets was also determined in order to draw inter-assay comparisons. Results: We observed that mitochondria in CRC cells were swollen and formed a fragmented network, whereas in their non-tumor counterparts, mitochondria were found to be elongated and organized into a complex branched network. Statistical analysis confirmed a clear separation between normal and CRC cells, as well as among the various CRC subtypes, based on mitochondrial morphology. In addition, our results showed that both glycolysis and OXPHOS increased as a function of CRC progression, with each tumor cell line displaying a specific metabolic signature. Interestingly, combining both types of mito-signatures improved classification accuracy. Conclusion: Integration of mitochondrial shape phenotyping and metabolic profiling improves CRC cancer cell classification and could provide a novel type of biomarker for CRC screening and therapeutic decision-making.