BCL-2 inhibition and AML: can we best Darwin?

In this issue of Blood, DiNardo and colleagues¹  provide fascinating details on the behavior of leukemic clones in less-fit patients treated for acute myeloid leukemia (AML) with the combination of venetoclax and either a hypomethylating agent or a low-dose cytarabine.

These data, derived from patient samples collected during 2 highly influential recent clinical trials,²  not only describe the molecular correlates of disease response and resistance but also provoke important questions that should be addressed in the next wave of clinical studies investigating these backbone regimens. The authors make multiple interesting points. The following 3 take-aways stand out:

1. Patients with mutations in the nucleophosmin gene (NPM1mut) or in isocitrate dehydrogenase 2 (IDH2mut) have the best chance of a durable response with a venetoclax-based regimen. These are patients in whom we need to study time-limited therapy, including intensive or measurable residual disease–directed consolidation.

2. Sequential molecular studies on the remaining patients reveal confounding polyclonal patterns of resistance that emerge sometimes very early in treatment. These individuals (and their physicians) face a challenging game of whack-a-mole to control increasingly unmanageable leukemic subclones. Clinical research might focus on upfront combinations with other molecularly targeted treatments, with (or randomized against) dose-attenuated cytotoxic therapy or even novel chemoimmunotherapeutic approaches.

3. If we are going to depend on repetitive attempts at molecularly targeted therapy for disease control, we will need to develop molecular sequencing strategies that are accessible, validated, standardized, and cost-contained.

After years of therapeutic nihilism, the armamentarium for treating older individuals with AML is expanding. Since 2017, there has been a spate of drug approvals, none more widely heralded than the BCL-2 inhibitor venetoclax. Early-phase clinical studies combined it with low-dose cytarabine or with hypomethylating agents, and very encouraging results prompted Food and Drug Administration approval. Despite the absence of randomized data, there has been rapid clinical uptake of venetoclax-based combinations in older, possibly less-fit individuals.³  It used to be that the challenge of facing a newly diagnosed, older patient was determining if they were robust enough for cytotoxic induction. That question remains, but now one needs to also ask: do I have all the cytogenetic and molecular testing I need to make a decision? Is my patient best served by cytotoxic induction, by CPX-351 (a gentler cytotoxic regimen), by a venetoclax-based regimen, by kinase or IDH inhibition, or with a clinical trial? Also, does this decision truly hinge on fitness? The absence of head-to-head studies makes picking a regimen feel like educated guess work.⁴ 

Enter the study published herein. Investigators studied sequential samples from 81 patients on 1 of the 2 prospective published venetoclax combination studies. Subjects were divided into 3 subgroups: those who achieved complete remission/complete remission with incomplete count recovery (CR/CRi) lasting >12 months (n = 18); those who achieved CR/CRi but who relapsed within a year (n = 25); and, finally, those who were refractory to treatment (n = 20). Interestingly, 18 patients achieved CR/CRi but came off the drug within a year for reasons other than relapse. Single-cell sequencing was performed in some instances to track clonal evolution along the treatment trajectory.

Some of the results are not surprising. Prior exposure to hypomethylating agents is associated with inferior survival. Response to therapy did not always mean that treatment continued. Adverse risk cytogenetics portends adverse risk.

Some of the results are provocative and compelling. Patients with NPM1 mutations (n = 11), IDH2 mutations (n = 7), or both mutations (n = 4) had a 2-year overall survival >70%. These data certainly provide a good argument for considering a venetoclax-based combination for less-fit individuals with NPM1mut or IDH2mut. With the caveat that we are talking about only 22 patients, a validation of this response in the ongoing randomized studies of venetoclax (NCT02993525 or NCT03069352) would be significant for patients in whom cytotoxic induction is truly too hazardous.

Some of the results are sobering. A total of 25 patients had disease that initially responded, but then relapsed at a median of 6.4 months. Serial analysis of samples as well as single-cell sequencing techniques illustrates that kinase activation, including FLT3-ITD, and clones that acquired biallelic silencing of TP53 accounted for many of these clinical outcomes. What is the lesson? Selection pressure means that AML remains a wily disease. In some subjects, the resistant subclones that blossomed at relapse were present at diagnosis. In other cases, they were novel. Importantly, they emerged quickly, sometimes just days to weeks into therapy.

Given these patterns of resistance, perhaps earlier combined targeted therapy is warranted (ie, triplet regimens that use FLT3 inhibition or TP53 stabilization strategies)? However, that approach brings its own new questions: how good are our tests at detecting subclones? Do we need to target mutations that have not yet emerged? How many targets are good enough, and for how long? If a barrage of molecularly directed agents at the start of treatment guts the proliferation potential of the latent clones, well, that’s excellent. However, I would be less enthusiastic about sequencing 1 agent after another, consigning patients to a permanent parade of consecutive treatments. I think it is a mistake to give up on a goal of true disease eradication. Combinations that use cytotoxic agents (possibly in attenuated doses or novel delivery mechanisms) or harness immune therapies alongside targeted therapies should be investigated. Evidence of polyclonal resistance supports the case that, as a research community, we need to focus on what the leukemic clones have in common and target those commonalities early on.⁵ 

Finally, a logistical point is presented. The authors argue that “serial molecular studies can identify patterns of drug sensitivity and resistance at the sub-clonal level.” I fear this recommendation is premature. Are we ready to roll such a practice into prime time, given the costs, the lack of standardization, and the paucity of effective options for refractory disease?

DiNardo, Wei, and colleagues have done much to bring venetoclax to AML patients, a remarkable accomplishment. The data presented here show us that our work has only just begun.