Model-Based Inference and Classification of Immunologic Control Mechanisms from TKI Cessation and Dose Reduction in Patients with CML

Abstract

Recent clinical findings in patients with chronic myeloid leukemia (CML) suggest that the risk of molecular recurrence after stopping tyrosine kinase inhibitor (TKI) treatment substantially depends on an individual's leukemia-specific immune response. However, it is still not possible to prospectively identify patients that will remain in treatment-free remission (TFR). Here, we used an ordinary differential equation model for CML, which explicitly includes an antileukemic immunologic effect, and applied it to 21 patients with CML for whom BCR-ABL1/ABL1 time courses had been quantified before and after TKI cessation. Immunologic control was conceptually necessary to explain TFR as observed in about half of the patients. Fitting the model simulations to data, we identified patientspecific parameters and classified patients into three different groups according to their predicted immune system configuration ("immunologic landscapes"). While one class of patients required complete CML eradication to achieve TFR, other patients were able to control residual leukemia levels after treatment cessation. Among them were a third class of patients that maintained TFR only if an optimal balance between leukemia abundance and immunologic activation was achieved before treatment cessation. Model simulations further suggested that changes in the BCR-ABL1 dynamics resulting from TKI dose reduction convey information about the patient-specific immune system and allow prediction of outcome after treatment cessation. This inference of individual immunologic configurations based on treatment alterations can also be applied to other cancer types in which the endogenous immune system supports maintenance therapy, long-term disease control, or even cure.Significance: This mathematical modeling approach provides strong evidence that different immunologic configurations in patients with CML determine their response to therapy cessation and that dose reductions can help to prospectively infer different risk groups.See related commentary by Triche Jr, p. 2083

Parameter perturbations in a post-treatment chronic myeloid leukemia model capture the essence of pre-diagnosis A-bomb survivor mysteries

“…Using the parameters in Table 1 , Hahnel’s post-diagnosis treatment cessation model (Hahnel et al 2020 ) was applied here to the pre-diagnosis scenario of starting with one proliferating CML cell (Fig. 2 ).…”

“…The TFR model of (Hahnel et al 2020 ) depicted in Fig. 1 represents quiescent CML stem cells , proliferating CML cells , and anti-CML immune cells , as (Eqs.…”

“…1 – 3 ): where and are transition rate constants into and out of an immune privileged niche of quiescent stem cells, cells grow with an initial rate constant toward a carrying capacity is the rate constant of immune system killing of CML cells, captures the rise (when )-and-fall (when ) of immune response amplification as a function of CML load (i.e., ), is a constant flux of cell creation, is the cell apoptosis rate constant, and the initial conditions are (0) = 0, (0) = 1, and (0) = . This model includes parameters for 21 cases (Table 1 ) obtained via fits to BCR-ABL dynamic responses to tyrosine kinase inhibitor cessation (Hahnel et al 2020 ).

Fig.

…”

“…

Fig. 1 Treatment-free remission (TFR) model (Hahnel et al 2020 ) of CML clone interactions with anti-CML immune cells. Here cells x are quiescent CML stem cells that live in an immune privileged niche, y are proliferating CML cells, and z are anti-CML immune cells

…”

“…Patients attempting this can be classified as either unstable (TFR not reached), strongly stable (TFR reached with no signs of disease), or weakly stable (TFR reached but with BCR-ABL levels still detected by polymerase chain reaction and oscillating as if dynamically regulated by the immune system). A mathematical model of CML dynamics across treatment cessations was recently developed (Hahnel et al 2020 ). This model is based on earlier models of CML malignant clone interactions with the immune system (Moore and Li 2004 ; Kim et al 2008 ; Paquin et al 2011 ; Clapp et al 2015 ; Besse et al 2018 ; Cesar Fassoni et al 2019 ).…”