BH3 profiling discriminates response to cytarabine-based treatment of acute myelogenous leukemia

Abstract

As acute myelogenous leukemia (AML) patient response to cytarabine-based standard-of-care treatment is variable, stratification into subgroups by biomarker-predicted response may lead to improved clinical outcomes. Here, we assess cell mitochondrial depolarization to proapoptotic signaling BH3-only peptides as a surrogate for the function of Bcl-2 family proteins to address clinical response to cytarabine-based therapy in patients with AML (N = 62). Peripheral blood mononuclear cell (PBMC) or bone marrow aspirate specimens were obtained from newly diagnosed patients with AML, viably preserved, and assayed by flow cytometry following BH3 profile assay with individual BH3 peptides. Mann-Whitney analysis indicates biomarker correlation with response to induction therapy: Notably, BIM priming was highly significant (P = 2 × 10(-6)) with a compelling sensitivity/specificity profile [area under curve (AUC) = 0.83; 95% confidence interval (CI), 0.73-0.94; P = 2 × 10(-10)]. Multivariate analysis indicates improved profiles for BIM readout + patient age (AUC = 0.89; 95% CI, 0.81-0.97) and BIM + patient age + cytogenetic status (AUC = 0.91; 95% CI, 0.83-0.98). When patients were stratified by cytogenetic status, BIM readout was significant for both intermediate (P = 0.0017; AUC = 0.88; 95% CI, 0.71-1.04) and unfavorable (P = 0.023; AUC = 0.79; 95% CI, 0.58-1.00) risk groups, demonstrating predictive power independent of cytogenetics. Additional analyses of secondary clinical endpoints displayed correlation between overall survival (P = 0.037) and event-free survival (P = 0.044) when patients were stratified into tertiles by BIM peptide response. Taken together, these results highlight the potential utility of BH3 profiling in personalized diagnostics of AML by offering actionable information for patient management decisions.

Figure 1. Dot-plot and ROC-plot depictions of BIM patient response discrimination

A. Dot-plot for the mean %priming (± S.D.) of BIM(0.1) comparing the 22 non-response (NR) and 39 clinical response (CR) patients. Higher BIM(0.1)%primed patients have statistically better clinical response. Statistical significance was determined using the Mann-Whitney Analysis B. ROC-plot of the sensitivity and specificity of BIM(0.1) as a predictor of clinical outcome as determined by the ROC curve (AUC=0.83) using the 61 patients for which there was available data for this marker. C. Multivariate analysis ROC curve for BIM(0.1) alone (Black), or BIM(0.1) combined with significant adjustment clinicopathological variables (+patient age [Red] or +patient age+cytogenetic status [Green]).

Figure 2. BH3 peptides response prediction stratified by cytogenetic status

Response prediction stratified by intermediate cytogenetic status (A–F) depicted in dot-plots (A, C, and E) and ROC-plots (B, D, and F) for BIM(0.1) (A and B), BAD (C and D), or HRK (E and F) or unfavorable cytogenetic status for BIM(0.1) as a dot-plot (G) and ROC-plot(H). BIM(0.1), BAD, and HRK were all statistically significant for intermediate cytogenetic status (p value <0.01 and AUC >0.85). Statistical significance was by Mann-Whitney Analysis. (Mean ± S.D. for each set is indicated on dot-plots).

Figure 3. Correlation of BIM (0.1) Priming and BIM (BCL2L11) Protein Levels and Response Prediction

BIM protein levels (BCL2L11) determined by RPPA assay were compared to BIM(0.1) BH3 profiling % priming yielding an R²=0.0396 indicating that the two metrics are distinct from one another (A). When statistical analysis is restricted to only patients for which there is BH3 profiling and corresponding BIM RPPA data, BIM(0.1) BH3 profiling remains significant predictor of response (B and C) while BIM protein levels are not (D and E). (Mean ± S.D. for each set is indicated on dot-plots).

Figure 4. OS and EFS vs. AML Patients Subgrouped by BIM % Priming Tertiles

AML patients were stratified by low, medium, and high BIM % priming scores and then analyzed for OS (panel A) and EFS (panel B). Logrank analyses indicates borderline significant associations between OS and BIM(0.1) priming (p=0.037), and EFS and BIM(0.1) priming (p=0.044). For OS, low, medium, and high priming tertiles comprised 19, 18, and 18 patients respectively (total n=55 for which OS data was available; patients who received stem cell transplant were censured). For EFS, low, medium, and high priming tertiles comprised 20, 20, and 21 patients respectively (total n=61 for which EFS data was available).