Elsevier

Leukemia Research

Volume 110, November 2021, 106663
Leukemia Research

Alternative strategies for optimizing treatment of chronic lymphocytic leukemia with complex clonal architecture

https://doi.org/10.1016/j.leukres.2021.106663Get rights and content

Highlights

  • Clonal architecture and dynamics play key roles in CLL treatment outcomes.

  • In silico simulation can predict outcomes based on NGS assessed clonal dynamics.

  • Our model indicates a clinical trial of time-limited BTK inhibition is warranted.

  • Simulations enhance efficiency/reduce cost of developing complex treatment strategies.

Abstract

In silico simulation of pre-clinical and clinical data may accelerate pre-clinical and clinical trial advances, leading to benefits for therapeutic outcomes, toxicity and cost savings. Combining this with clonal architecture data may permit truly personalized therapy. Chronic lymphocytic leukemia (CLL) exhibits clonal diversity, evolution and selection, spontaneously and under treatment pressure. We apply a dynamic simulation model to published CLL clonal architecture data to explore alternative therapeutic strategies, focusing on BTK inhibition. By deriving parameters of clonal growth and death behavior we model continuous vs time-limited ibrutinib therapy, and find that, despite persistence of disease, time to clinical progression may not differ. This is a testable hypothesis. We model IgVH-mutated CLL vs unmutated CLL by varying proliferation and find, based on the limited available data about clonal dynamics after such therapy, that there are differences predicted in response to anti-CD20 efficacy. These models can suggest potential clinical trials, and also indicate what additional data are needed to improve predictions. Ongoing work will expand modeling to agents such as venetoclax and to T cell therapies.

Introduction

Chronic lymphocytic leukemia (CLL) exhibits clonal diversity, clonal evolution and clonal selection spontaneously and under treatment pressure [1,2]. This is exemplified in FISH reports of a certain percentage of, for example, deletion 17p in a CLL sample, as well as the knowledge that the presence of del17p is higher in relapsed CLL than at initial diagnosis or treatment. Each patient with CLL will have unique biology based on their underlying clonal composition, and this will then be further influenced by therapeutic interventions.

There is a long history of using mathematical models based on experimental data to devise cancer treatment paradigms. Skipper and Schabel [3] proposed, based on rapidly growing tumor models, that each cycle of therapy killed a constant fraction of remaining viable cells, or log cell kill, and that small tumors were more sensitive. This led to application of repeated cycles of maximum tolerated therapy, an approach effective in rapidly growing cancers. Norton and Simon [4] combined concepts of Gompertzian tumor growth kinetics and drugs available at the time that killed rapidly growing cells to develop treatments using late intensification. Goldie and Coldman [5] incorporated the idea of random development of drug resistant cells to propose use of non-cross resistant combination regimens to minimize acquired resistance, leading to many clinically relevant combination and/or alternating treatment regimens. With increased understanding of cancer biology, awareness of individualization not only of each cancer type, but of each patient, and even of sub- clones within that patient, along with the plethora of new agents with novel mechanisms of action, mathematical models are poised to become more individualized as well [6].

Mathematical modeling has long been used to provide testable hypotheses for developing cancer treatments. Our previous simulations of the progression and combination drug treatment of non-Hodgkin lymphomas analyzed malignant cell population data from both single and multiple drug pre-clinical experiments [7], the effects of immune system response [8], and a mechanism for transformation of indolent to aggressive disease [9]. Combination drug scheduling strategies for the treatment of CLL were also investigated [10]. In general, in silico simulation [11] to model multifactorial combinations, dose and schedule of therapies using both pre-clinical and clinical data has the potential to rationally select those most likely to be of benefit, narrowing the number needed to be tested, thereby accelerating clinical trial development.

Simulation of competing CLL clonal populations, including clones with intrinsic chemotherapy resistance (e.g. mutant p53/deletion 17p CLL), intrinsic resistance to targeted therapy (e.g. mutations in BTK or PLCG2), as well as those that evolve under selective pressure, would permit targeted clinical intervention with a different class of therapeutics. As clonal architectures can now be detected and simulated [1,2,12,13], dynamic simulation can then be used to predict the outcomes of early interventions, to prevent or substantially delay clinical recurrence of disease. The work of Wu and colleagues [2] demonstrates the power of next generation sequencing (NGS) to evaluate the evolutionary landscape of CLL. They have shown the ability to quantitate and track subclones, and to identify mutations and transcriptional changes that enhance the understanding of the biology and evolution of CLL, which could permit the rational selection of subsequent therapies. Here we demonstrate that these advances in NGS that permit tracking of individual CLL clones then permit derivation of growth and response parameters for individual agents that can then be used to model in silico an array of combinations and schedules. Specifically, we build on their published patient clonal data [12,13] by adding predictive modeling to demonstrate clinically relevant approaches that may improve outcomes by optimizing use of existing therapeutic agents. Modeling can inform specific testable hypotheses that are proposed regarding CLL therapy, such as intermittent vs continuous BTKi, limiting the number of combinations and schedules that need to be formally investigated. This approach would markedly reduce the time and resources required to develop and test multiple novel strategies.

Section snippets

Simulation methodology

The components of our computational model are illustrated in Fig. 1 and described in the figure legend. The equations derived from this model are presented in more detail in the Supplemental section.

The model’s inputs are clonal population dynamics data obtained from clinical sample - derived next-generation sequencing. The model derives discrete values of K’ and K” associated with individual therapeutics and their associated biological mechanisms. Once the values of these parameters are

Modeling patient data determines parameter values

To determine parameters for our simulation model, we first applied it to published patient data [12,13] acquired with deep sequencing techniques. Clonal dynamic data from a patient who was treated with single agent ibrutinib was reported in Landau et al. [12], reproduced in Fig. 2A, and simulated in Fig. 2B. The patient presented with a common ancestral clone possessing del17p. This clone was already present at the start of ibrutinib therapy. It is not uncommon that such a del17p clone is

Discussion

With advances in deep sequencing able to define clonal architecture and dynamics, detection of drug- resistant clones in CLL is clinically feasible [[2], [3], [4]]. When combined with a data-driven simulation of clonal architectures, the course of the disease can be projected and alternative interventions considered. We demonstrate here that modeling of CLL clinical behavior based on dynamic clonal architecture assessment is achievable by current molecular methodology. CLL biology is

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Declaration of Competing Interest

The authors report no declarations of interest.

Acknowledgements

The authors dedicate this work to Dr. Robert Weiss who passed away just prior to its final resubmission. He has been the driving force behind its development. He will be sorely missed by all of his colleagues, across a range of disciplines, who were fortunate to have had the opportunity to work with him.

The authors acknowledge the valuable assistance of M.G. Miller, PhD and D.M. Coppola.

References (31)

  • R.F. Weiss et al.

    Parametric model of combination therapy for non-Hodgkin lymphoma

    PLoS One

    (2012)
  • R.F. Weiss et al.

    Immune response to combination therapy for non-Hodgkin lymphomas

    PLoS One

    (2013)
  • R.F. Weiss et al.

    Progression and transformation of clonally heterogeneous B-cell lymphoma

    PloS one

    (2015)
  • J.M. Greene et al.

    Mathematical approach to differentiate spontaneous and induced evolution to drug resistance during cancer treatment

    JCO Clin. Cancer Inform.

    (2019)
  • D.A. Landau et al.

    The evolutionary landscape of chronic lymphocytic leukemia treated with ibrutinib targeted therapy

    Nat. Commun.

    (2017)
  • Cited by (0)

    View full text