Gaining better understanding of the intratumoral heterogeneity of Ewing sarcoma through single-cell technology and artificial intelligence
By combining artificial intelligence with cell-by-cell analysis of tumors, Unit 830 'Cancer, Heterogeneity, Instability and Plasticity', and Unit 900 'Cancer and genome: bioinformatics, biostatistics and epidemiology' at Institut Curie have succeeded in better identifying the mechanisms related to Ewing sarcoma relapses.
Uriel Chantraine / Institut Curie
Ewing sarcoma tumors are rare pediatric tumors, located in the bone, with around 70 cases a year in France. These tumors are among the most genetically stable cancers. They have a major, causal mutation, identified in 1992 by Olivier Delattre's team at Institut Curie: EWSR1-FLI1, from the merging of the EWSR1 and FLI1 genes.
The major challenge of the study of these tumors is to better understand their cellular origins and their heterogeneity. The teams followed the so-called "single-cell" approach. "This is a new approach in our field. It is particularly suited to studying the intratumoral heterogeneity, to identify cellular sub-populations liable to be resistant to treatment. Identifying these populations can help us better understand the causes of relapses," explains Andrei Zinovyev, who co-leads the 'Biology of cancer systems team', Unit U900, jointly with Emmanuel Barillot.
This work has been published in the international journal Cell Reports, and is the result of a close collaboration between bioinformatics specialists and biologists from Institut Curie, Bioinformatics Unit U900 at Institut Curie headed by Emmanuel Barillot on the one hand, and Unit 830 'Cancer, Heterogeneity, Instability and Plasticity', led by Olivier Delattre.
To better characterize intratumoral heterogeneity at the level of individual cells, the two teams used a machine learning approach, which is an artificial intelligence technique. They were first able to observe relations between the activity of EWSR1-FLI1 and cell proliferation. "Bioinformatics changes the landscape for this type of study since it lets us process the enormous quantity of data generated by single-cell approaches and analyze them using innovative statistical tools," explains Andrei Zinovyev.
After characterizing the dynamics of gene expression under the influence of the oncogene EWSR1-FLI1 in a cell line, the team then went on to look at real tumors. "We observed the level of activation of this oncogene in each tumor cell. The result is surprising. Each level of activity of EWSR1-FLI1 corresponds to specific characteristics of cells. When oncogene activity is high, the cells proliferate and present a mitochondrial energy metabolism. And when oncogene activity is low, they change their metabolism and tend to migrate and form metastases", recounts Marie-Ming Aynaud, biologist and the first author of the article, and who is currently doing a post-doctoral fellowship at Mount Sinai hospital in Toronto.
A discovery that's crucial to gain a better understanding of why this type of cancer recurs. "If tomorrow we find a treatment that inhibits the activity of this oncogene and thus halts proliferation, we will probably have to associate cell migration inhibitors to prevent these cells with a residual rate of EWSR1-FLI1 from migrating remotely," explains Olivier Delattre.
This progress could not have been achieved without the contribution of artificial intelligence. "This type of study is quite different from what we were doing ten years ago, mainly due to the quantity of data processed. Each tumor used to have its own molecular profile. Today each tumor is characterized by thousands of cells with different profiles," explains Olivier Mirabeau, bioinformatics specialist and joint first author of the article. A huge mass of data that can now be used and that can tell us a great deal about the nature of cancer.