Modeling BCR-ABL Transcript Dynamics in CML Patients under Imatinib Treatment.

Treatment of chronic myeloid leukemia (CML) with the tyrosine kinase inhibitor imatinib represents a successful application of a molecularly targeted therapy. A rapid hematologic and cytogenetic response can be induced for the majority of patients even in advanced disease. However, the time course of disappearance of leukemia cells, characterized by the expression of the BCR-ABL fusion protein, varies between patients, and a complete eradication of the malignant cells is a rare event. The reasons for the heterogeneous response and the persistence of the malignant clone in many patients are currently not known.

We propose a mathematical model which consistently explains short and long-term dynamics of BCR-ABL transcript levels in populations of CML patients under imatinib monotherapy. The model is based on the concept that normal and malignant cell clones compete for growth environments in which they behave slightly differently with regard to homing and cell cycle activation/deactivation. This concept has been successfully applied for understanding time-dependent chimerism in mice [Roeder et al.: Blood 105(2):609]. Applying the model to data sets from two independent cohorts of imatinib treated CML patients, we demonstrate the potential of our model to quantitatively describe the typical biphasic decline in BCR-ABL transcript levels during the first year of treatment. Besides the median transcript dynamics in the patient population the model is able to represent the heterogeneity in individual transcript time courses. Qualitative differences in the imatinib response are explained by small quantitative differences in the drug effects regarding proliferation inhibition and/or induction of apoptosis for BCR-ABL positive cells. As demonstrated by comparison with five years follow-up data of 69 unselected newly diagnosed CML patients recruited into the IRIS trial in Germany [Mueller et al.: Leukemia 17(12):2392] the model also correctly describes long-term BCR-ABL dynamics. The observed median BCR-ABL transcript levels, including the vanishing decline after year four of treatment, can quantitatively be explained by a decreasing treatment efficiency in a subset of patients, potentially caused by imatinib-resistant clones. Sensitivity analyses show that moderate functional differences of the resistance mutations can lead to remarkable differences in long-term treatment efficiency. On the other hand, in patients not developing resistance mutations our model predicts the general chance of an eradication of the malignant clone in the long run. This is supported by data in a patient subgroup showing a continued decline of BCR-ABL transcript levels over five years of treatment.

Beyond the consistent description of the clinically observed BCR-ABL dynamics we provide testable predictions for effects of different combination treatments. Based on the explanation of CML as a clonal competition of malignant and normal hematopoietic stem cells, our model particularly predicts that the therapeutic benefit of imatinib can be augmented by a combination with proliferation stimulating treatment strategies. In addition the model permits to describe the heterogeneity of the effect of resistance mutations with respect to specific treatment strategies. In summary, our model describes CML dynamics under imatinib therapy with potential implications for the design of future treatment strategies.