Identification of genetic heterogeneity in acute myeloid leukemia (AML) has provided a unique opportunity for the greater individualization of therapy. However the implementation of new therapies has lagged far behind the ability to initially recognize operationally important genetic lesions. Until we have further bridged this gap between the identification of genetic lesions and the resultant knowledge of effective therapies, alternative strategies for rapidly identifying candidate therapies can become an important tool for precision medicine. Since most agents, regardless of whether "cytotoxic" or "targeted" ultimately function by activating the mitochondrial apoptotic pathway, we hypothesized that a tool that measures mitochondrial sensitivity may serve as a broadly predictable biomarker. We developed a dynamic BH3 profiling (DBP) technique that measures early death signaling within 8-16 hours after exposure to drugs. Increased cell death signaling is reflected by increased mitochondrial sensitivity (i.e. increased priming) to standardized BH3 peptides mimicking pro-apoptotic proteins.

To develop a personalized therapy for AML using DBP, we utilized 20 independent patient derived xenograft (PDX) models, established from de novo, primary refractory or relapsed (R/R) patients (available at http://www.PRoXe.org). Human myeloblasts from spleen and bone marrow of xenotransplanted NSG mice were exposed to 30 targeted and 3 standard of care drugs to determine mitochondrial responses. Unsupervised hierarchical clustering of ex-vivo DBP measurements across AML PDXs revealed several distinct clusters. Majority of targeted agents with an ability to induce priming in selective PDXs were enriched within a cluster, including kinase inhibitors, epigenetic modifiers, SMAC mimetic and chemotherapy drugs. In contrast, a discrete subcluster of drugs showed sensitivity across majority of PDXs, including BH3 mimetics, CDK9 inhibitors and HDAC inhibitors. Drugs with identical mechanism of action showed similar priming patterns across PDXs. Of note, 3 non-myeloid PDXs clustered distinctly from AML, an indication of differential priming responses owing to their cells of origin. AML PDXs developed from treatment naïve patients clustered adjacently and showed greater priming responses to a large number of drugs as opposed to PDXs from R/R patients that formed a discrete cluster. These data reveal that mitochondrial priming can stratify AML PDXs according to its predicted sensitivity to targeted agents.

Next, we validated ability of DBP to predict in-vivo responses of single agent birinapant (SMAC mimetic), JQ-1 (BRD-4 inhibitor), quizartinib (FLT-3 inhibitor), and venetoclax (BCL-2 inhibitor) across 6 AML PDX models, prioritized based on their greatest range of priming responses. We found that birinapant was most efficacious nonetheless as expected from ex-vivo DBP studies, responses varied between different PDX models. Myeloblasts of those PDXs that showed the greatest drug-induced changes in apoptotic priming were indeed the PDXs with the highest in-vivo responses. When we compared the ability of DBP to identify sensitive PDXs with additional precision medicine tools such as genomics, we found that DBP was able to accurately predict quizartinib activity in PDXs expressing WT FLT-3, which would have been predicted to be unresponsive based on genomic analysis. Collectively, priming responses obtained from ex-vivo DBP was able to rank different PDX models according to their sensitivities to targeted agents (AUC of ROC curve 0.8731, p<0.005). To investigate if DBP can predict combination therapies in relapsed settings, we first developed resistant models of single agents and then repeated DBP. Myeloblasts from relapsing clone showed reduced overall mitochondrial priming and lacked acquisition of a new chemical dependency compared to initial clone. This suggests that targeting of pre-existing dependencies might be more crucial than therapy induced dependency for AML.

In summary, our findings highlight that mitochondria-based measurements could identifying individualized therapy for a heterogeneous population and may serve as a as a powerful biomarker to identify the best responders to patient therapies.

Disclosures

Letai: AbbVie, AstraZeneca, Novartis: Consultancy, Research Funding.

Author notes

*

Asterisk with author names denotes non-ASH members.

Sign in via your Institution