Pan-clonal score predicts first-line treatment response in AML

AML has high patient-to-patient treatment response variability. Pharmacoscopy allows ex vivo drug-response profiling of tumour cells, enabling tailored treatments.

Dr Yannik Severin presented the pharmacoscopy technique

AML is a heterogeneous disease due to a fast natural or treatment-induced clonal expansion of premature myeloid cells leading to the presence of multiple clones at the same time, with different drug-sensitivity profiles. On top of this, myeloid maturation of rare leukaemic stem cell populations can give rise to new tumour cell populations leading to relapse after an initial treatment of the patient. Hence, a successful treatment needs to eradicate all sub-populations of tumour cells present. This requires a tailored treatment regimen for every individual AML patient. 

Dr Yannik Severin (ETH Zurich, Switzerland) presented a technique called pharmacoscopy, which allows ex vivo functional drug-response profiling of (tumour) cells1. In pharmacoscopy, the sensitivity of both healthy cells and tumour cells of the patient to 100 distinct drugs and drug combinations is analysed by automated microscopy2.

Neural network characterised cells automatically into distinct populations

In a prospective, non-interventional study, 180 AML patient biopsies were collected from 44 patients with newly diagnosed AML undergoing intensive induction chemotherapy. Patient-matched bone marrow and blood samples were obtained for each donor at 3 different time points across the course of the treatment (at time of diagnosis, after the first and after the second chemotherapy). Tumour cells were recognised by their expression of CD33, CD117, CD34, or combinations of these markers; healthy cells were recognised by CD3 expression. 

A neural network automatically characterised cells into 1 of 5 distinct myeloid populations (CD34+, CD33+, CD117+/CD33+, CD117+/CD33+/CD34+, or negative) or 2 healthy T-cell subtypes (conventional, activated). Dr Severin showed that this technique is able to illustrate the highly heterogeneous blast composition across samples. In addition, blast composition and maturity of tumour cells could predict first-line treatment (cytarabine/daunorubicin) success, with an overall predictive power of 85% accuracy and a diagnostic odds ratio of 47. For example, the presence of CD34 is predictive of non-response leading to an increased fraction of CD34+ cells after the first line of chemotherapy.

Plan-clonal scores were based on ex vivo drug responses

Based on the ex vivo response to different drugs, Dr Severin calculated a pan-clonal score. This score could predict the response of all clones present in the sample to a specific drug or combination of drugs. This pan-clonal score proved to be highly predictive of response to the first-line chemotherapy in the used patient cohort. In addition, using the pan-clonal score, it is possible to select a (combination of) drug(s) that is most likely to kill most tumour cells, irrespective of their actual abundance, Dr Severin concluded.

1. Severin Y, et al. Intra-patient functional heterogeneity of AML determines first-line treatment response. Abstract S134. EHA2022 Hybrid Congress, 09–12 June.
2. Snijder B, et al. Lancet Haematol. 2017;4(12):e595–e606.