The FMS-like tyrosine kinase 3- internal tandem duplication (FLT3/ITD) aberration is common in acute myeloid leukemia (AML) and associated with poor patient outcome. Inhibitors targeting FLT3/ITD are in development, but clinical responses are transient. This project focussed on elucidating molecular signalling consequences of FLT3/ITD inhibition, to identify rational drug combinations for future development. A Multicolour Phospho Flow Cytometry (MPFC) assay was developed to assess signalling events downstream of FLT3/ITD in primary patient samples, focusing on alterations in ERK, STAT5, Akt, and S6. STAT5 signalling appeared to be important exclusively in FLT3/ITD samples. MPFC accurately predicted the presence of FLT3/ITD, inhibitor sensitivity and the initial positive clinical response of a trial patient receiving a FLT3/ITD inhibitor. PI3K pathway upregulation was observed in a Sorafenib-resistant FLT3/ITD cell line established to study resistance mechanisms of FLT3 inhibition. Further, combination FLT3 and PI3K inhibition demonstrated synergy, suggesting potential clinical relevance to this therapeutic strategy.

The FMS-like tyrosine kinase 3- internal tandem duplication (FLT3/ITD) aberration is common in acute myeloid leukemia (AML) and associated with poor patient outcome. Inhibitors targeting FLT3/ITD are in development, but clinical responses are transient. This project focussed on elucidating molecular signalling consequences of FLT3/ITD inhibition, to identify rational drug combinations for future development. A Multicolour Phospho Flow Cytometry (MPFC) assay was developed to assess signalling events downstream of FLT3/ITD in primary patient samples, focusing on alterations in ERK, STAT5, Akt, and S6. STAT5 signalling appeared to be important exclusively in FLT3/ITD samples. MPFC accurately predicted the presence of FLT3/ITD, inhibitor sensitivity and the initial positive clinical response of a trial patient receiving a FLT3/ITD inhibitor. PI3K pathway upregulation was observed in a Sorafenib-resistant FLT3/ITD cell line established to study resistance mechanisms of FLT3 inhibition. Further, combination FLT3 and PI3K inhibition demonstrated synergy, suggesting potential clinical relevance to this therapeutic strategy.

EGFR TKIs are standard therapy for advanced NSCLC. In order to define their role in early disease, we implemented a phase II trial of neoadjuvant gefitinib in clinical stage I NSCLC. Tumour shrinkage was seen in 43% of patients, with 11% achieving RECIST partial response (PR). Analysis of molecular markers showed EGFR TKD mutations in 17% of cases, being the only associated with PR. For the first time we defined the histopathological response of NSCLC to these agents, characterized by reduction in tumour cellularity and proliferative index as well as presence of non-mucinous BAC histology. Clinical PR tumours also presented large areas of stromal fibrosis with presence of focal residual tumour. In a characterization of intracellular signalling response, EGFR dephosphorylation in the residues Y1068 and Y1173 was not concordant and only the former was significantly reduced. pAkt Ser473/Akt and Thr308/Akt ratios were significantly reduced but observed among both, clinical responders and resistant patients. Interestingly, reduction in pEGFR Y1068 was significantly associated with increase in tumour cellularity (p=0.047), Ki-67 index (p=0.018) and tumour growth (p=0.019) with a residual perinuclear localization been detected, suggesting a novel mechanism of resistance involving receptor internalization. Finally, we determined that the EGFR protein remains stable up to one hour of post resection ischemia but two to three tumour samples are necessary for an adequate tumour representation. Furthermore, EGFR cytoplasmic compartment presented the best association with clinical response in our cohort. Taking all together, we were able to generate the first clinical trial exploring the use of an EGFR TKI in early NSCLC, characterizing for the first time the histopathological and signalling responses to these agents with an evidence of a potential novel mechanism of resistance. Finally, we observed that multiple samples collection for an adequate tumour representation, and assessment of the cytoplasmic compartment, are warrant.

EGFR TKIs are standard therapy for advanced NSCLC. In order to define their role in early disease, we implemented a phase II trial of neoadjuvant gefitinib in clinical stage I NSCLC. Tumour shrinkage was seen in 43% of patients, with 11% achieving RECIST partial response (PR). Analysis of molecular markers showed EGFR TKD mutations in 17% of cases, being the only associated with PR. For the first time we defined the histopathological response of NSCLC to these agents, characterized by reduction in tumour cellularity and proliferative index as well as presence of non-mucinous BAC histology. Clinical PR tumours also presented large areas of stromal fibrosis with presence of focal residual tumour. In a characterization of intracellular signalling response, EGFR dephosphorylation in the residues Y1068 and Y1173 was not concordant and only the former was significantly reduced. pAkt Ser473/Akt and Thr308/Akt ratios were significantly reduced but observed among both, clinical responders and resistant patients. Interestingly, reduction in pEGFR Y1068 was significantly associated with increase in tumour cellularity (p=0.047), Ki-67 index (p=0.018) and tumour growth (p=0.019) with a residual perinuclear localization been detected, suggesting a novel mechanism of resistance involving receptor internalization. Finally, we determined that the EGFR protein remains stable up to one hour of post resection ischemia but two to three tumour samples are necessary for an adequate tumour representation. Furthermore, EGFR cytoplasmic compartment presented the best association with clinical response in our cohort. Taking all together, we were able to generate the first clinical trial exploring the use of an EGFR TKI in early NSCLC, characterizing for the first time the histopathological and signalling responses to these agents with an evidence of a potential novel mechanism of resistance. Finally, we observed that multiple samples collection for an adequate tumour representation, and assessment of the cytoplasmic compartment, are warrant.

Les mastocytoses systémiques (SM) constituent un groupe hétérogène de maladies rares, caractérisées par l’accumulation anormale de mastocytes malins dans la moelle osseuse et dans d’autres organes extra-cutanés. La majorité des patients avec SM ont une mutation activatrice du gène KIT, le plus souvent la mutation KIT D816V, retrouvée chez plus de 90% de tous les patients. Cette mutation induit l’activation constitutive du récepteur KIT en déclenchant de façon aberrante une cascade de voies de signalisation, dont la voie PI3K/AKT et JAK/STAT5, aboutissant à l’inhibition de l’apoptose et à l’augmentation de la prolifération et de la survie des mastocytes malins. Cependant, l’efficacité des inhibiteurs de tyrosines kinases (ITKs) sur cette mutation est limitée à cause de la résistance et/ou de toxicité liée à un manque de spécificité. Il est donc nécessaire de trouver de nouvelles approches thérapeutiques afin de contourner cette résistance au cours des SM KIT D816V+ avancées. Nous avons utilisé une approche consistant à cibler de façon combinée des molécules activées en aval de KIT D816V, comme AKT et STAT5, par des inhibiteurs pharmacologiques. Ceci nous a permis d’identifier une combinaison synergique entre un inhibiteur d’AKT (GSK690693) et un inhibiteur de STAT5 (BP-1-102). Ces composés sont capables d’inhiber la prolifération des cellules KIT D816V+ seuls ou en combinaison, mais à de très fortes concentrations, malheureusement non utilisables en thérapeutique. Néanmoins, ces premiers résultats ont permis de valider AKT et STAT5 comme cibles potentielles dans le traitement des SM avancées. La seconde approche employée a été de cibler directement le récepteur KIT D816V par des inhibiteurs pharmacologiques. A l’issu d’un criblage, nous avons identifié trois composés - BLU2317, BLU2718 et DCC-2618 - capables d’inhiber sélectivement la phosphorylation de KIT D816V. Ces composés inhibent la prolifération des cellules ROSAKIT D816V et HMC-1.2, et induisent l’apoptose des cellules de façon dose-dépendante. Bien que les effets de ces trois composés soient similaires, DCC-2618 agit à des concentrations plus faibles par rapport aux composés BLU2317 et BLU2718. Afin d’apprécier l’efficacité in vivo de DCC-2618, nous avons d’abord établi un nouveau modèle de SM basé sur l’injection intraveineuse des cellules ROSAKIT D816V-Gluc exprimant la Gaussia luciferase (Gluc) dans des souris NSG. La présence de la Gluc sécrétée par les cellules ROSAKIT D816V-Gluc facilite la mise en évidence de prise de greffe et permet un contrôle précis de la progression de la maladie. Ce modèle reproduit, au bout de 4 semaines, chez toutes les souris greffées, une SM avancée similaire à celle retrouvée chez l’homme, avec atteinte de la moelle osseuse, du sang, de la rate et du foie, tandis que la dégradation de l’état général des souris n’est observée qu’à partir de 12 semaines. Ce nouveau modèle offre suffisamment de temps pour explorer la cinétique de la progression de la maladie et surtout pour effectuer des études pharmacologiques précliniques. L’évaluation de l’effet de DCC-2618 in vivo a été réalisée sur ce modèle. Etonnamment, DCC-2618 n’a pas été capable d’inhiber la progression de la maladie chez les souris traitées, bien qu’atteignant des concentrations élevées dans la moelle osseuse et le plasma des souris traitées. Néanmoins, DCC-2618 s’est montré capable d’inhiber la phosphorylation de KIT dans les cellules issues de la moelle osseuse des souris traitées. En revanche, contrairement aux effets observés in vitro, DCC-2618 a induit une surexpression de phospho-ERK1/2 dans les cellules malignes des souris greffées. Ceci suggère qu’ERK1/2 joue un rôle important dans la résistance au composé DCC-2618 et éventuellement à d’autres ITKs, indépendamment du récepteur KIT. ERK1/2 pourrait donc être une nouvelle cible thérapeutique d’intérêt dans le traitement des SM résistantes aux ITKs / Systemic mastocytosis (SM) is a heterogeneous group of rare diseases characterized by abnormal accumulation of malignant mast cells (MCs) in the bone marrow (BM) and other extra-cutaneous organs. The majority of SM patients have an activating mutation in the KIT gene, usually the D816V point mutation, which is found in more than 90% of all patients. This mutation induces constitutive activation of the KIT receptor by triggering a cascade of signaling pathways, including the PI3K/AKT and the JAK/STAT5 pathways, resulting in the inhibition of apoptosis and increased survival and proliferation of malignant mast cells. However, the efficacy of the tyrosine kinase inhibitors (TKIs) on this mutation is limited due to resistance and/or toxicity associated with a lack of specificity. It is therefore critical to find new therapeutic approaches to overcome this resistance to TKIs, particularly for advanced KIT D816V+ SM. In the present thesis, we have used an approach consisting in targeting molecules activated downstream of KIT D816V, such as AKT and STAT5, using pharmacological inhibitors in combination. This allowed us to identify a synergistic combination of an AKT inhibitor (GSK690693) and an inhibitor of STAT5 (BP-1-102). These compounds are able to inhibit proliferation of KIT D816V+ cells, alone or in combination, but at very high concentrations, unfortunately not useful therapeutically. Nevertheless, these initial results have validated STAT5 and AKT as potential targets for the treatment of advanced SM. The second approach used was to target directly the KIT D816V receptor by pharmacological inhibitors. After a large screening, we identified three compounds - BLU2317, BLU2718 and DCC-2618 - which selectively inhibit the phosphorylation of KIT D816V. These compounds inhibit the proliferation of ROSAKIT D816V and HMC-1.2 cells, and induce apoptosis of these cells in a dose-dependent manner. Although the effects of these three compounds are similar, the DCC-2618 compound acts at lower concentrations relative to BLU2317 and BLU2718 compounds. In order to assess the in vivo efficacy of DCC-2618, we first established a new model of SM based on intravenous injection of cells expressing Gaussia luciferase (Gluc), ROSAKIT D816V-Gluc cells, in NSG mice. The presence of the secreted Gluc in ROSAKIT D816V-Gluc cells facilitates the detection of engraftment and allows precise monitoring of disease progression. This model reproduced within four weeks, in all grafted mice, an advanced SM similar to the one found in humans, with neoplastic MCs infiltration in BM, blood, spleen and liver, while the terminal deterioration of the clinical condition of the mice was observed after 12 weeks. Thus, this new in vivo model allows modulating the aggressiveness of the disease by varying the number of injected cells. It provides sufficient time to explore the kinetics of disease progression and especially to conduct preclinical pharmacological studies. We then evaluated the effect of DCC-2618 compound in vivo on this model. Surprisingly, DCC-2618 was not able to inhibit disease progression in treated mice, although it reached high concentrations in the BM and the plasma of treated mice. Nevertheless, we showed that the compound was able to inhibit the phosphorylation of the KIT receptor in cells derived from the BM of treated mice. In addition, contrasting to the effects observed in vitro, DCC-2618 induced an over-expression of phospho-ERK1/2 in the malignant cells of transplanted mice. This suggests that ERK1/2 may play a critical role in the resistance to DCC-2618, and possibly to other TKIs, independently of the KIT receptor. ERK1/2 could thus be a new interesting therapeutic target in the treatment of advanced SM resistant to TKIs

Veröffentlichung der Dissertation im Rahmen des Promotionsverfahrens. Kumulative Promotion zum Thema renaler und kardialer Nebenwirkungen nach anti-VEGF-Behandlung. Case reports von Patienten mit refraktären Schilddrüsenkarzinomen und möglicher Nebenwirkungen nach TKI-Behandlung.

Cancer is known for its unregulated and mutagenic characteristics. The topic of targeting cancer by inhibiting the metabolic pathways it uses to thrive has been a focus of modern cancer research. Specifically, in lung cancer, the transformation from non-small cell lung cancer (NSCLC) to small cell lung cancer (SCLC) is a focus. This transformation often comes with a grimmer prognosis and reduced survival rate. This is primarily due to SCLC being resistant to epidermal growth factor receptor (EGFR) inhibitors. This frontline treatment for EGFR mutant NSCLC has shown to be quite effective until transformation to SCLC occurs. To further study the metabolic factors responsible for this transformation, a metabolic screening was conducted on SCLC transformed lung tissues and tumor adjacent normal lung tissues. This analysis revealed that the amino acid L-arginine and intermediates in its biosynthetic pathway were severely dysregulated. While L-arginine supplementation has shown to inhibit the growth of breast and colorectal cancers, there is little literature about its effects on lung cancer. Using cell viability and gene expression screening tools, we have identified arginine metabolizing genes ARG2, GATM, and OAT as being upregulated in NSCLC treated with high concentrations of an EGFR inhibitor. These high treatments also correlate with increased expression of neuronal differentiation factor 1 (NEUROD1), which has been shown to drive tumorigenesis, metastasis, and SCLC transformation. These findings show a role for altering arginine metabolism to accomplish drug resistance through SCLC transformation. These findings will hopefully pave the way for later clinical use of arginine converting enzymes and NEUROD1 expression levels as predictive markers of early drug resistance and SCLC transformation.

Hintergrund Die chronische myeloische Leukämie (CML) ist eine myeloproliferative Er- krankung, die aufgrund ihres Modellcharakters unter der Behandlung mit Tyrosin-Kinase- Inhibitoren (TKI) gut für eine Beschreibung mittels computerbasierter Modelle geeignet ist. Grundlage für die Entstehung einer CML ist die Bildung eines Philadelphia-Chromosoms durch eine Translokation der Chromosomen 9 und 22. Es resultiert das Onkogen BCR- ABL1, welches für eine konstitutiv aktive Tyrosinkinase codiert. Diese führt zu ungeregelter Proliferation der betroffen Zellen und zur Verdrängung der gesunden Blutbildung. Das überaktivierte Protein kann durch TKIs gezielt gehemmt werden. Damit ist es möglich, die Tumorlast erheblich zu senken und das Fortschreiten der Erkrankung aufzuhalten. Aktuell werden in der klinischen Anwendung außerhalb von Studien TKIs für die gesamte Lebensdauer der Patienten eingesetzt. Absetzstudien zeigten, dass circa 50% der Patienten nach einer über zwei Jahren nicht nachweisbaren BCR-ABL1-Last nach Behandlungsstopp kein erneutes Anwachsen der Tumorlast aufwiesen. Die Anwendung von computergestützten Modellsimulationen hilft, Zugriff auf die klinisch nur schwer zu messenden leukämischen Stammzellen zu bekommen und darüber Vorhersagen über den weiteren Therapieverlauf zu treffen. Aufgabenstellung Im Rahmen der vorliegenden Arbeit sollen Möglichkeiten der Übertragung von Patientendaten auf das etablierte Modell nach Roeder und Loeffler (2002) verbessert werden. Die vom Modell vorhergesagten Stammzellkinetiken sollen abschließend auf Praxistauglichkeit geprüft werden. Material und Methoden Aufgrund der Vergleichbarkeit zu früheren Untersuchungen erfolgte die Auswahl von 51 Patienten des deutsches Armes der IRIS-Studie. Deren Therapieverläufe wurden analysiert und können über eine biphasische exponentielle (biexponentielle) bzw. über eine stückweise lineare Funktion beschreiben werden. Als Erweiterung der Arbeiten von Horn et al. (2013) wurden alle Parameter der biexponentiellen Funktion in die Entwicklung neuer Methoden einbezogen. Zusätzlich wurde untersucht, ob die Einbeziehung von zensierten Messpunkte die Form der biexponentiellen Funktion verändert. Basierend auf den Therapiedaten der IRIS-Patienten erfolgte die Ermittlung eines Para meterraumes für Eingangsparameter der Modellsimulation (Modellparameter), welcher in 270.400 individuelle Paramterkombinationen unterteilt wurde. Es erfolgten anschließend die Simulation und Auswertung nach der biexponentiellen Beschreibung. Auf Basis dieser erheblich größeren Datengrundlage konnten zwei neue Verfahren der Modellparameteridentifikation für individuelle Patienten entwickelt werden. Einerseits wurde in Anlehnung an die Arbeit von Horn et al. (2013) ein Verfahren unter Nutzung der Regression vorgestellt. Andererseits konnte über den Vergleich der Abstände zwischen simulierten und realen Therapieverläufen eine Suche (lookup-table) etabliert werden. Die Berechnung des Abstandes zwischen Therapieverläufen ermöglicht gleichzeitig den Vergleich der verschiedenen Verfahren und damit eine Aussage über deren Anpassungsgüte. Zum Schluss wurde beispielhaft für einen Patienten das Verfahren der lookup-table angewendet und die resultierende Stammzellkinetik weiter analysiert. Ergebnisse Einführend erfolgte die Analyse der resultierenden biexponentiellen Funktion mit und ohne Einbeziehung von Messunsicherheiten. Es zeigte sich, dass der Verlauf dieser Funktion besonders in Bereichen, die von einbezogenen Messunsicherheiten betroffen sind, abweichend ist. Die Beschreibung des Langzeitverlaufs erfolgt jedoch annähernd gleich. Anschließend erfolgte die Validierung der Größe des vorsimulierten Datenpool anhand eines Vergleichs der statistischen Parameter von Patienten und Simulationen. Dieser zeigte sich dabei für die weiteren Untersuchungen geeignet. Die Nutzung der lookup-table zur Identifikation der am besten zu einem Patienten passenden Therapiesimulation ist überlegen sowohl gegenüber von der Horn et al. (2013) beschriebenen als auch in dieser Arbeit neu entwickelten Regressionsverfahren. Diese ergeben deutliche Abweichungen zwischen Patientendaten und Simulation. Eine Analyse des vorhergesagten Therapieverlaufes im Stammzellkompartiment ergibt jedoch, dass ähnliche Therapieverläufe im peripheren Blut durch stark unterschiedliche Stammzellkonfigurationen beschrieben werden können. Es resultiert eine starke Streuung der vorhergesagten Zeitpunkte eines möglichen Therapieendes. Schlussfolgerungen Die Nutzung der lookup-table zu Identifikation einer passenden Therapiesimulation ist hoch effektiv und anderen Verfahren, die auf Regression basieren, überlegen. Die etablierte Computersimulation nach Roeder und Loeffler (2002) bietet Zugriff auf die Therapie in der Ebene der Stammzellen. Die in weiteren Analysen gezeigten Streuungen der vorhergesagten Therapieverläufe im Stammzellkompartiment lassen den Schluss zu, dass Methoden zur Eingrenzung der Stammzellverläufe entwickelt werden müssen, um die Vorhersagen klinisch nutzbar zu machen. Weiterhin muss anhand von Messungen an Knochenmarkproben von realen Patienten geprüft werden, ob die von der Simulation postulierten Verläufe der Tumorlast im Stammzellkompartiment der realen Behandlung entsprechen. Ausblick Die in aktuellen Arbeiten beschriebene Rolle des Immunsystems im Therapieverlauf der CML (Saussele et al. 2016; Clapp et al. 2016) sollte in eine Verbesserung des Stammzellmodells nach Roeder und Loeffler (2002) einfließen. Weiterhin kann die Validierung der im Rahmen der Individualmedizin zu treffenden Absetzvorhersagen letztendlich nur über klinische Absetzuntersuchungen ermöglicht werden. / Background Chronic myeloic leukaemia (CML) is a myeloproliferative disease, which is well suited for modelling approaches. It is characterized by the oncogenic BCR-ABL1 fusion gene originating from an inverse translocation of the chromosomes 9 and 22 leading to the Philadelphia chromosome. The result is a constitutively activated tyrosine-kinase. This is followed by an extensive proliferation of leukaemic stem cells leading to a displacement of normal haematopoesis. The molecular specificity of CML forms the basis of a highly efficient, targeted therapy by tyrosine kinase inhibitors (TKIs). TKIs can decrease the tumour burden and slow down or eventually stop progressing of the disease. Currently, in clinical applications drugs are administered for the remaining life span. Interestingly, in recent treatment cessation trials patients were stopped after two years of non-detectable tumour burden and about 50% remained without relapse. The application of computer-based modelling helps to gain access to stem cell counts being difficult to measure clinically. This forms the basis for predictions of long-term therapy outcomes. Aim of this work This work aims on identifying a suitable algorithm to efficiently identify model simulations that optimally decribe individual patient kinetics. Furthermore, the clinical usability of the new methods was investigated. Material and methods The analysed group of patients was chosen out of the German cohort of the IRIS trial to ensure comparability to former investigations. It consists of 51 individuals. The course of leukaemic burden , i. e. leukaemic vs. non-leukaemic cells on a single patient level can be described as a biphasic exponential (bi-exponential) or a piecewise linear function. As an extension to former methods described by Horn et al. (2013) all parameters are included into further method development. Additionally, an investigation was conducted whether censored data points change the functional behaviour of a bi-exponential fit based on patients’ data. According to therapy data of all patients an input parameter space for the model simulation was delimited, such that all observed patient kinetics can be mimicked by the model. This parameter space was uniformly divided into 270.400 discrete parameter combinations. The therapy simulation of each combination was conducted and described by a bi-exponential function likewise to the patients’ fit. With the help of these huge variety of in silico therapies two new methods of model parameter identification for individual patients were developed. The first one is an advanced approach based on a regression model proposed by Horn et al. (2013). The second one by comparing distances between the patients’ and the models’ bi-exponential functions (lookup table). The comparison of the distances between different therapy courses (either simulated or patients’ data) was also used to compare the quality of different methods. As an example, for one patient the stem cell kinetics from the model were analysed in more detail and checked for robustness. Such a strategy, which might build the basis for clinical applications. Results A comparison between the different bi-exponential functions with and without censored data points revealed differences especially in the area in which censoring was performed. However, for the long-term tumour burden censored data had no influence. Secondly, an investigation was performed showing the sufficiency of the pre-simulated therapy courses for the new methods, i. e. lookup-table and regression models. The lookup- table turns out to be superior to identify a therapy simulation for a unique patient, since the complexity of linear regression models lead to increased deviations between patients’ therapy courses and the simulations. Unfortunately, distinct stem cell configurations lead to similar therapy descriptions in peripheral blood, assuming the correctness of the model. As a result, the prediction of a safe treatment cessation is often widely spread. Conclusions The new developed lookup-table to identify model simulations suitable for an individual patient is highly effective and superior to other methods using regression models. The simulation of the TKI treatment using the agent-based model of Roeder und Loeffler (2002) gives easy access to therapy courses on the level of leukaemic stem cells. Unfortunately, the finding of a well fitting simulation within the peripheral blood is not enough to provide a point of safe treatment cessation, since different stem cell configurations can lead to similar therapy courses. Additionally, it is necessary to check which of the assumed therapy courses on the stem cell level is appropriate. This could be done by gathering more information from bone-marrow punctures during the course of treatment. Outlook Investigations of new data showed the important role of the immune system in CML treatment (Saussele et al. 2016; Clapp et al. 2016). This should be taken into account by improving the model of Roeder und Loeffler (2002). Additionally, data from cessation trials can be used to validate the model assumptions.

Longitudinal monitoring of BCR-ABL transcript levels in peripheral blood of CML patients treated with tyrosine kinase inhibitors (TKI) revealed a typical biphasic response. Although second generation TKIs like dasatinib proved more efficient in achieving molecular remission compared to first generation TKI imatinib, it is unclear how individual responses differ between the drugs and whether mechanisms of drug action can be deduced from the dynamic data. We use time courses from the DASISION trial to address statistical differences in the dynamic response between first line imatinib vs. dasatinib treatment cohorts and we analyze differences between the cohorts by fitting an established mathematical model of functional CML treatment to individual time courses. On average, dasatinib-treated patients show a steeper initial response, while the long-term response only marginally differed between the treatments. Supplementing each patient time course with a corresponding confidence region, we illustrate the consequences of the uncertainty estimate for the underlying mechanisms of CML remission. Our model suggests that the observed BCR-ABL dynamics may result from different, underlying stem cell dynamics. These results illustrate that the perception and description of CML treatment response as a dynamic process on the level of individual patients is a prerequisite for reliable patient-specific response predictions and treatment optimizations.

Introduction: Discontinuation of tyrosine kinase inhibitor (TKI) treatment is emerging as the main therapy goal for Chronic Myeloid Leukemia (CML) patients. The DESTINY trial showed that TKI dose reduction prior to cessation can lead to an increased number of patients achieving sustained treatment free remission (TFR). However, there has been no systematic investigation to evaluate how dose reduction regimens can further improve the success of TKI stop trials. Methods: Here, we apply an established mathematical model of CML therapy to investigate different TKI dose reduction schemes prior to therapy cessation and evaluate them with respect to the total amount of drug used and the expected TFR success. Results: Our systematic analysis confirms clinical findings that the overall time of TKI treatment is a major determinant of TFR success, while highlighting that lower dose TKI treatment for the same duration is equally sufficient for many patients. Our results further suggest that a stepwise dose reduction prior to TKI cessation can increase the success rate of TFR, while substantially reducing the amount of administered TKI. Discussion: Our findings illustrate the potential of dose reduction schemes prior to treatment cessation and suggest corresponding and clinically testable strategies that are applicable to many CML patients.