Integrating High-Throughput Dynamic BH3 Profiling and Molecular Phenotyping to Identify Therapeutic Vulnerabilities in CLL

Abstract

Despite recent advances in chronic lymphocytic leukemia (CLL) therapy, such as the use of targeted agents including, Bruton's tyrosine kinase (BTK) inhibitor ibrutinib and the potent BCL-2 antagonist venetoclax, this disease remains incurable for most patients, who are refractory or become resistant to the novel agents. Thus, identifying new treatment regimens for CLL and building a precision medicine framework that can match CLL patients to the appropriate drugs are of high priority. To address this challenge, we optimized high-throughput dynamic BH3 profiling (HT-DBP), a functional assay that rapidly measures the initiation of apoptotic signaling after ex vivo exposure to drugs, for interrogation of CLL samples. The main advantages of this assay over other ex vivo assays are: (i) rapidity - under 24 hours, which is especially important in CLL, where cell viability substantially decreases after 24 hours; (ii) miniaturization - a very limited number of primary cells are required; and (iii) scalability - allowing to conduct hundreds of drug response tests in parallel on one 384-well plate. These features collectively maximize the information yield from a given patient sample. We recently identified prognostic genetic alterations and molecular subtypes of CLL, based on multi-omics profiling of >1100 CLL samples (Knisbacher et al., Nat Genet, in press). To determine if these molecular findings were associated with novel therapeutic vulnerabilities in CLL, we performed HT-DBP on 65 primary CLL samples previously characterized by exome, transcriptome and methylome profiling, using 42 FDA approved drugs that were selected for potential relevance to CLL biology. We isolated peripheral blood mononuclear cells (PBMCs) and cultured them in conditioned media derived from stroma cells to reduce spontaneous apoptosis. Target cells were treated with a drug for 20 hours followed by BH3 peptide exposure. Mitochondrial outer membrane permeabilization (MOMP) was then measured on digitonin-permeabilized cells in response to BH3-only synthetic peptides that mimic pro-apoptotic BCL-2 family proteins. Mitochondrial cytochrome c release was quantified as a measure of MOMP by flow cytometry, gating on CD19+ and CD5+ cells. This assay measures if the cell has moved closer to the threshold of apoptosis after drug treatment and thereby identifies drugs that enhance apoptosis priming. The peptides used in each experiment were derived from BIM or PUMA to measure increases in overall apoptotic priming, or BAD and MS1 peptides that identified BCL-2 and MCL-1 dependence, respectively. Our screen revealed differential drug-induced apoptotic priming for various drugs. As expected, venetoclax and ibrutinib were highly effective across CLL. Other drugs that demonstrated high priming included navitoclax (BCL-XL/BCL-2), nutlin-3 (MDM2), abexinostat (HDAC), gandotinib (JAK2), duvelisib (PI3K δ/γ), idelalisib (PI3Kδ) and cerdulatinib (SYK/JAK). The assay was robust, as indicated by an 0.92 median Pearson correlation across replicates. Additionally, the majority of drugs had greater effect on CLL samples than on healthy PBMCs (p<0.001, paired t-test), supporting their specificity. We focused our analysis on the differential drug effects among molecular subtypes of CLL. First, we found that IGHV-mutated CLLs (M-CLLs) became more primed to apoptosis than IGHV-unmutated CLLs (U-CLLs) across the panel of drugs (p<0.001, paired t-test) and significantly in response to fludarabine and umbralisib (FDR<0.1, t-test). Second, we compared drug-induced apoptotic priming among 8 CLL subtypes (i.e. RNA expression clusters (ECs); Knisbacher et al.). Notable among the many drug priming-EC relationships was that within M-CLL ECs, venetoclax was most effective in EC-m3 (high IL-10 expression) and least effective in EC-m2 (low IL-10 and enriched in trisomy 12). Interestingly, EC-m1, associated with high TFEC expression and poor outcome, was most sensitive to nutlin-3. For U-CLLs, EC-u1 was most sensitive to gandotinib and EC-u2 to navitoclax. EC-i, associated with the intermediate methylation subtype of CLL, was the most resistant EC to ibrutinib but was very sensitive to navitoclax, more than to any other drug. Altogether, we present a framework that links ex-vivo drug response with molecular features including expression subtypes to highlight new therapeutic opportunities in CLL.