Clonal evolution in patients with chronic lymphocytic leukaemia developing resistance to BTK inhibition

Abstract

Resistance to the Bruton's tyrosine kinase (BTK) inhibitor ibrutinib has been attributed solely to mutations in BTK and related pathway molecules. Using whole-exome and deep-targeted sequencing, we dissect evolution of ibrutinib resistance in serial samples from five chronic lymphocytic leukaemia patients. In two patients, we detect BTK-C481S mutation or multiple PLCG2 mutations. The other three patients exhibit an expansion of clones harbouring del(8p) with additional driver mutations (EP300, MLL2 and EIF2A), with one patient developing trans-differentiation into CD19-negative histiocytic sarcoma. Using droplet-microfluidic technology and growth kinetic analyses, we demonstrate the presence of ibrutinib-resistant subclones and estimate subclone size before treatment initiation. Haploinsufficiency of TRAIL-R, a consequence of del(8p), results in TRAIL insensitivity, which may contribute to ibrutinib resistance. These findings demonstrate that the ibrutinib therapy favours selection and expansion of rare subclones already present before ibrutinib treatment, and provide insight into the heterogeneity of genetic changes associated with ibrutinib resistance.

Figure 1. Evidence of clonal evolution with late disease progression following ibrutinib.

(a) White blood cell counts and treatment course of Patient 1. Peripheral blood specimens were sampled at five time points (indicated by arrows), and CLL cells underwent whole-exome sequencing. Following somatic mutation calling, CCF of somatic variants was inferred by ABSOLUTE analysis of deep-sequencing data of the detected mutations (see Supplementary Figs 1 and 2). Asterisk indicates that this sample had less purity, and hence clone sizes are estimates. (b) A phylogenetic tree was inferred based on PHYLOGIC, a novel algorithmic extension of ABSOLUTE. Driver mutations associated with each clone are indicated (a complete listing of somatic mutations and allelic fractions found for each clone in Supplementary Table 2 and Supplementary Figs 1 and 2). (c) Multiplexed detection of somatic mutations in 134–172 single cells of Patient 1 at TP1, TP2 (pre-ibrutinib) and TP5 (ibrutinib relapse). Between all three time points, shifting cell subpopulations with SF3B1 mutation are observed. At TP5, SF3B1-K666T is detected in all cells, while the various PLCG2 mutations are detected in distinct subpopulations. Single-cell expression levels of wild-type (WT) and mutated (MUT) PLCG2 are shown in Supplementary Fig. 3B. (d) The clonal kinetics during ibrutinib treatment. Filled circles—measurement of the number of cells comprising each subclone at each time point based on the subclone CCF and the corresponding ALCs. Measurements are shown with 95% confidence interval (CI) obtained from posterior distributions of CCFs. Empty circles—upper bound estimates (1% of total CLL cells) for subclones that were below the detection threshold of targeted deep sequencing. Solid lines denote predicted kinetics for clones detected on at least two measurements. Dashed lines represent kinetics with minimal absolute growth rates for clones detected in only one measurement. (e) Extrapolation of clone size with 95% CI at the time of treatment initiation for the PLCG2-MUT subclones.

Figure 2. Clonal evolution with early disease progression following ibrutinib.

White blood cell counts and treatment courses of Patients 2 (a) and 3 (d). Peripheral blood specimens were sampled at serial time points (indicated by arrows), and CLL cells underwent whole-exome sequencing. Following somatic mutation calling, CCFs of somatic variants were inferred by ABSOLUTE analysis (Supplementary Figs 1 and 2). The phylogenetic trees for Patients 2 (b) and 3 (e) were inferred based on Phylogic. Driver mutations associated with each clone are indicated (a complete list of somatic mutations and allelic fractions found for each clone in Supplementary Table 2 and Supplementary Figs 1 and 2). Clonal kinetics during ibrutinib treatment for Patients 2 (c) and 3 (f). Filled circles—measurements combining clonal fractions and ALC counts. Empty circles are upper bound estimates (1% of total CLL cells) for clones that were below detection. Solid lines denote predicted kinetics for clones with at least two measurements. For Patient 2, dashed lines represent kinetics with minimal absolute growth rates for clones with only one measurement, while for Patient 3 the dashed lines represent kinetics obtained from fitting to ALCs. Measurements are shown with 95% CI obtained from posterior distributions of CCFs. For Patient 3, we assumed clones 1 and 2 have the same rates of decline and clones 4 and 5 have the same growth rates during treatment.

Figure 3. Droplet-based detection of resistance subclones at the time of treatment initiation.

(a) Schema of the experimental workflow. (b) Specificity of the mutation-detection primers visualized on an agarose gel in bulk cell line populations transfected to express minigenes encoding the WT versus mutated (MUT) allele (for PLCG2 and RPS15), or in bulk patient cDNA at pre-treatment and relapse time points (Patient 3, DGKA; c). Droplet apparatus, and detection of bright droplets following amplification. (d) Detection of MUT RPS15-specific single cells in Patient 2 samples and a PBMC control (left) and of MUT DGKA-specific single cells in Patient 3 samples and a PBMC control (right). (e) Standard curve for the detection of the PLCG2-M1141R template, established based on known input quantities on cell line (murine 30,019 cells, with error bars shown) expressing the MUT template, and detection of PLCG2-M1141R in the pre-treatment sample of Patient 1, but not in controls (Supplementary Fig. 4B).

Figure 4. Histiocytic sarcoma trans-differentiation of CLL during ibrutinib therapy.

(a) Regression of lymph node (LN) disease, visualized with CT scan, following ibrutinib exposure (at time point (TP) 2), compared with TP1. (b) At TP2 (autopsy), histologic sections of liver and LN, stained with haematoxylin and eosin, showed histiocytic sarcoma with sheets of large atypical cells with irregular shaped nuclei, dense nuclear chromatin and abundant cytoplasm (at × 100 and × 500 inserts). Occasional large neoplastic cells demonstrated one or two prominent eosinophilic nuclei. No lymphoid aggregates were seen. The neoplastic cells within the LN were strongly positive for CD163 and are negative for CD19, CD1a and S100 proteins (all at × 500). (c) White blood cell counts and clinical course for Patient 5. Whole-exome sequencing and CCF measurements were made before ibrutinib initiation (TP1) and from post-mortem specimens of the liver and LN (TP2). The fraction of cells that shared the mutations that define the histiocytic sarcoma parent clones are represented with black diagonal lines. Phylogenetic analysis was performed based on PHYLOGIC. A complete list of somatic mutations and allelic fractions for each clone is provided in Supplementary Table 2 and Supplementary Figs 1 and 2.

Figure 5. Impact of del(8p) on apoptosis in response to ibrutinib and/or TRAIL in CLL.

(a) Representative interphase and metaphase FISH results following hybridization for probes specific for chromosome 8p21.3 (red) and chromosome 8 centromere (green), showing a CLL cell with a normal disomic hybridization pattern or with deletion of chromosome 8p (details of the FISH probe in Supplementary Fig. 5a,b,d). (b) FISH hybridization of pre-treatment and relapse samples from Patients 2 and 3 to detect del(8p). For each case, 100 nuclei were scored as summarized in the associated bar graphs. (c) Primary CLL cells were isolated from peripheral blood and treated with ibrutinib (1 μM) and/or TRAIL (200 ng ml⁻¹). Cell death was assessed with Annexin V and propidium iodide (PI) staining and flow cytometry. P-values calculated for absolute change in viability. In agreement with the known pleiotropic effects of TRAIL on CLL cells, we found that TRAIL treatment induced apoptosis in six out of nine of non-del(8p) samples, yet could also enhance survival in three out of nine samples. Red—samples with a decrease in cell viability of at least 15% following exposure to TRAIL or ibrutinib. Circle—independent CLL samples, not previously exposed to ibrutinib; triangles—relapse samples from Patients 2 and 3. n.s., not significant. (d) Cell viability measurements based on flow cytometric analysis following Annexin V and PI of CLL cells from ibrutinib-resistant Patients 2 and 3 (also see Supplementary Fig. 6a). CLL cells collected before ibrutinib (grey bars) or at the time of ibrutinib resistance (orange bars) were incubated with ibrutinib (1 μM) and/or TRAIL (200 ng ml⁻¹) or staurosporine and assessed for viability. At time of ibrutinib resistance, CLL cells were less responsive to TRAIL-induced apoptosis and largely are ibrutinib-resistant, but remained sensitive to staurosporine. Cell viability measurements were equivalent whether based on annexin V/PI or by live cell counts (Supplementary Fig. 6b).