Better therapy selection for childhood leukaemia

Acute myeloid leukaemia in children (paediatric AML, paediatric AML) is one of the most aggressive cancers in children. It develops when immature precursor cells in the bone marrow degenerate due to genetic changes and their normal maturation into functional blood cells is blocked. Instead, the defective cells multiply uncontrollably, displacing healthy blood formation and leading to serious symptoms such as anaemia, increased susceptibility to infection, bleeding tendency and organ failure.

Unlike acute lymphocytic leukaemia, which is more common in children, paediatric AML is biologically more diverse and sometimes more difficult to treat. Although advances in chemotherapy and stem cell transplantation have improved survival rates, the prognosis for many patients remains poor despite decades of optimisation of treatment protocols: Some patients do not respond to standard therapies or suffer relapses. Now, a study published in Cell Reports Medicine shows that functional image analysis and molecular characterisation can be combined to create a tool that can detect resistance to therapy at the time of diagnosis.

The work is the result of particularly close collaboration between the research teams of Kaan Boztug, Michael Dworzak and Giulio Superti-Furga – a collaboration between basic research and clinical practice that received €585,000 in funding from the FWF Clinical Research Programme for the project Linking ex-vivo chemosensitivity, treatment and pathway activations for a deeper understanding of paediatric AML (ExTrAct-AML). First author Ben Haladik, PhD student in Kaan Boztug's research group, worked with the researchers to further develop a platform for investigating active substances based on the ‘pharmacoscopy’ method developed at CeMM for high-resolution imaging, artificial intelligence and comprehensive molecular analysis. Using 45 patient samples, they were able to show that robust predictions about treatment response and relapse risk can be derived from this.

Molecular profile as the key to prognosis
Leukaemia cells from blood or bone marrow samples are treated with various drugs in the laboratory and then observed under a microscope to see whether they die or are resistant to the drugs. This is done on a large scale and is fully automated: with the help of deep learning algorithms, the effect of each active ingredient is analysed in parallel in hundreds of thousands of cells. Combined with genetic and epigenetic data, this results in a detailed "chemosensitivity profile".

This revealed clear differences between the risk groups and even subpopulations of cells that are resistant to standard therapy. Particularly striking was a stem cell-like form of leukaemia that was resistant to conventional chemotherapy but vulnerable to new combinations of known active substances such as BCL2 and MDM2 inhibitors or HDAC inhibitors. The results show that the prognosis for childhood AML can be further improved by such functional analyses. While mutations provide important clues, the actual clinical relevance lies in the question of how leukaemia cells respond to drugs. This is precisely where the new method comes in: it makes the functional level visible and allows a direct link between the molecular profile and the actual response to therapy.

From research to the clinic
This form of functional precision medicine has the potential to fundamentally change the treatment of childhood AML. It complements genetic diagnostics and the detection of minimal residual disease, which are the most important tools for risk assessment, with a level that can directly map the response to drugs. This brings us closer to the possibility of identifying high-risk patients at the time of diagnosis and providing them with targeted new therapies.

Lead author Ben Haladik explains the methodology: "We have created a link between molecular biological analyses, bioinformatic methods and artificial intelligence, which should provide a basis for further research into better treatment methods." Kaan Boztug, senior and corresponding author of the study, also sees the study as a social mission for the future. "Our study is the first to show that such ex vivo drug tests can help us identify patients at an early stage whose leukaemia cells are particularly resistant to standard therapy. We can then use the method specifically for such patients to find targeted therapy options for patients with paediatric AML. With our study, we are also positioning ourselves as a major player in European paediatric cancer research in an area that has received little attention to date – the application of AI to paediatric cancer research."

"At CeMM, we have developed Pharmacoscopy, an image-based approach to functional single-cell precision medicine – a technology that enables true personalised medicine in cancer treatment. This has been further developed in the recently published study and successfully tested for the first time for the diagnosis of paediatric AML in the clinic. This is an important milestone in the use of such methods on a larger scale for the benefit of paediatric patients," said Giulio Superti-Furga, co-senior author of the study. "The results of our study open up a completely new approach to the treatment of paediatric AML. By being able to detect resistance at the time of diagnosis, we are laying the foundation for much more targeted and individualised therapies in the future. This means that we can identify high-risk patients at an early stage and offer them more tailored treatment strategies. This is a decisive step towards sustainably better chances of recovery," adds Michael Dworzak, head of the ’Immunodiagnostics" research group at St. Anna Children's Cancer Research and deputy medical director at St. Anna Children's Hospital.

The results now presented are based on a retrospective cohort. The next step is prospective clinical studies in which the method is applied in real time and compared with the actual course of the disease.

Publication: Cell Reports Medicine
Image-based drug screening combined with molecular profiling identifies signatures and drivers of therapy resistance in paediatric AML.
Haladik B, Maurer-Granofszky M, Zoescher P, Jimenez-Heredia R, Frohne A, Segarra-Roca A, Casey C, Kartnig F, Giuliani S, Rashkova C, Repiscak P, Dworzak MN, Superti-Furga G, Boztug K.
DOI: 10.1016/j.xcrm.2025.102304