Computational methods to understand clonal evolutionary dynamics of leukemia under treatment

Leukemia is one of the most common cancers being diagnosed in children and adults. There are different types of leukemia, and each of them exhibits a range of histopathological features/molecular profiles that require different therapeutic approaches. With such a wide spectrum, there is an urgent need to properly profile the heterogeneity of leukemia’s architecture under different treatments, in order to administer therapies tailored to each patient (Chapter 1). With the advent of high-throughput sequencing technologies, we can comprehensively quantify the variants’ abundances using deep targeted DNA sequencing, to detect variants that may have been missed by the traditional sequencing approaches. We have implemented a bioinformatic pipeline, which profiles background sequencing errors, to identify somatic candidate variants, with statistical confidence, at a sensitive threshold (Chapter 2). Using chronic lymphocytic leukemia (CLL) as an illustration, we performed deep targeted DNA sequencing on 33 CLL patients, which are TP53 naïve or relapsed/refractory, treated with ibrutinib monotherapy, and their specimens were collected at homogenous, longitudinal timepoints up to 144 weeks, including before treatment initiation, along with their matched-normal samples. Previous studies have shown that minimal residual disease (MRD) of ibrutinib-treated patients rarely achieve negativity, although their progression-free survival is significantly improved by inactivating Bruton’s tyrosine kinase (BTK) and thus the B cell receptor (BCR) signaling pathway with ibrutinib. By applying our implemented pipeline, we identify a total of 959 variants in all patients, and the genomic architecture of CLL under ibrutinib is very similar across most patients (average Nei’s standard genetic distance is ~0.0006), with the exception of patient 16-014 that manifests genomic heterogeneity starting from 72 weeks after receiving therapy (pairwise genetic distances between week 96 and previous timepoints are ≤0.01). Furthermore, none of the relapsed patients (n = 6) is detected with BTK and/or phospholipase C gamma 2 (PLCG2) mutations, with the exception of patient 04-005, in which we identify clonal expansion of a BTK mutation (at allele frequency of 2-39% from weeks 48-72) and four minute PLCG2 mutations (≤2% at week 72). This suggests that DNA mutations may not play role in shaping the MRD architecture (Chapter 3). In addition, we conducted single-cell RNA-seq on three CLL patients (17-102, 04-002, and 17-100) from the same cohort, up to 24 weeks after ibrutinib treatment, to study the transcriptomic profile of MRD architecture. We identify a union set of 219 biologically variable genes (for all patients) and 164 co-expressed genes (in at least one patient). We detect downregulation of the BCR pathway (with its possibly reactivation in some week 24 cells of 17-102 and 17-100) and FCER2/CD23 gene, as well as up-regulation of the MAPK, NF-κB, and RAS signaling pathways and CXCR4/CD184 gene as the response signatures in these three patients, even just after two weeks of treatment, supporting the notion that non-genetic factors are influencing the evolutionary trajectory of CLL. Further investigations are needed to fully understand molecular mechanisms behind these transcriptomic signatures, with the goal of designing combination therapeutics to target sustaining MRD (Chapter 4).Ph.D.Includes bibliographical reference