Unleashing Blocked Apoptosis in Cancer Cells: New MCL1 Inhibitors Find Their Groove

Evasion of programmed cell death is essential for the development and progression of tumors. BCL2 family proteins are the arbitrators of the major form of programmed cell death in mammalian cells known as apoptosis (1). These proteins decide cell fate via a series of competitive direct protein–protein interactions mediated by binding of their common BCL2 homology 3 (BH3) domain to a BH3-binding site found on both pro- and antiapoptotic proteins. Although all BCL2 family proteins have a BH3 domain, only the multidomain executioner proteins BAX and BAK and their inhibitors BCL2, BCL-xL, and MCL1 have functional BH3-binding sites. Interaction of the BH3 sequence of the proapoptotic executioner proteins with the BH3-binding sites on an inhibitor protein prevents the former from oligomerizing and effecting mitochondrial outer membrane permeabilization (MOMP) and committing the cell to apoptosis. Interactions with the third BCL2 family protein type, BH3 proteins, are more complicated. Most bind to and thereby sequester in inactive complexes a subset of the inhibitory proteins. For example, NOXA binds only to MCL1, whereas BIM binds to BCL2, BCL-xL, and MCL1. Moreover, a subset of the BH3 proteins including BIM, BID, and PUMA also activate the pore-forming activity of BAX/BAK (Fig. 1A). Although some of the specificity of these interactions comes from the sequences of the BH3 domains and the structures of the BH3-binding sites, the interactions are also regulated by posttranslational modifications, regulated protein turnover, and competitive binding interactions with non-BCL2 family proteins. The relevance for cancer biology and therapeutics is that the survival of some tumors is exquisitely dependent on specific interactions among subsets of these proteins (Fig. 1B). This is because, unlike most normal tissue, tumors are generally being instructed to undergo apoptosis, so during development they have shut down this response, for example by overexpressing BCL2, which prevents tonic proapoptotic signaling from killing the cells, thereby enabling continued growth. In this case, inhibition of BCL2 is sufficient to unleash widespread apoptosis, and therefore such inhibition represents an important target for therapy (2). A therapeutic index arises because in normal cells apoptosis is balanced by multiple interactions between the different BCL2 family proteins, and they generally contain an excess of antiapoptotic proteins as a buffer against different stressors.

As the binding interfaces in the BH3 pocket to their proapoptotic ligands are extensive and deep, they presented a major challenge to drug development. It was therefore a tour de force in medicinal chemistry when navitoclax was developed by AbbVie (3). This inhibitor has broad specificity against BCL2 and BCL-xL, but notably not MCL1 as the binding pocket of the latter differed significantly from its cousins. Multiple in vitro models with cell lines and patient samples showed that it was effective at hitting its target in cells and in eliciting apoptosis inhibited by BCL2 or BCL-xL; however, in the clinic, this medication has the on-target effect of thrombocytopenia, as BCL-xL is involved in physiologic regulation of platelet turnover. To solve this problem, AbbVie developed venetoclax, which showed specificity for BCL2 with little clinically relevant affinity for BCL-xL, thus preventing thrombocytopenia while retaining proapoptotic activity (4). In cancers with marked addiction to BCL2, the drug was so potent that initial treatment in patients with chronic lymphocytic leukemia caused marked tumor lysis, and subsequent clinical development was accompanied by careful dose titration to avoid this potentially lethal complication. This has been achieved, and the drug has been awarded accelerated approval by the FDA for the treatment of chronic lymphocytic leukemia and acute myeloid leukemia (AML) in the elderly. Nevertheless, it is clear that many tumor types depend on more than BCL2 to inhibit apoptosis, and that resistance to BCL2 or BCL-xL inhibition is conferred by MCL1 that is unaffected by both clinically available inhibitors. Furthermore, MCL1 is one of the most commonly amplified genes across all cancer types (5).

The shallow binding groove of MCL1 compared with BCL2 and BCL-xL and the high affinity of MCL1 for its natural proapoptotic ligands has frustrated and delayed the development of inhibitors. It was therefore a major development when an MCL1 inhibitor was reported last year, and is now in clinical trials (6). In this issue of Cancer Discovery, two more novel, potent, and selective MCL1 inhibitors are described in three articles.

Caenepeel and colleagues initially identified small molecules that bound to the MCL1 binding groove, and expanded the size of the compounds by controlled additions while restricting flexibility to reduce nonbinding conformations that would lead to off-target effects (7). They ultimately produced the AM-8621, a spiromacrocyclic ring with superb sensitivity in the picomolar range of binding to the MCL1 groove, and the ability to displace BIM from MCL1 in the nanomolar range, with substantially lower binding affinity to BCL2 or BCL-xL. The consequence of disrupting the BIM–MCL1 interaction was the subsequent activation of BAK leading to MOMP and caspase activation in many of the 952 cancer cell lines tested; the most sensitive derived from multiple myeloma, AML, and lymphoma. When comparing the relative sensitivity to AM-8621 with venetoclax, multiple myeloma showed predominant addiction to MCL1, whereas AML showed a mixed pattern with addiction to BCL2, MCL1, or both. The profound sensitivity of AML cell lines in vitro was confirmed with the more pharmacokinetically attractive derivative AMG 176 administered orally to mice harboring AML cell lines with schedules as infrequently as once weekly. Significantly, administration of AMG 176 in mice with humanized MCL1 knocked-in demonstrated a dose-dependent decrease in B cells and neutrophils as a pharmacodynamic biomarker. In AML cell lines and patient-derived samples, AMG 176 showed synergistic activity with standard chemotherapies and with venetoclax, indicating a codependency of many AMLs on both BCL2 and MCL1. Importantly, in mice treated with both compounds, no significant organ toxicity was evident.

Ramsey and colleagues (8) used a related potency optimization to improve their previously reported MCL1 inhibitor to produce VU661013 that also has subnanomolar affinity for the MCL1 binding groove, and displaces BAK to initiate MOMP. VU661013 did not significantly inhibit BCL-xL or BCL2. Similar to the results reported with AM-8621, a series of AML cell lines showed sensitivity to both venetoclax and VU661013, with some showing dual sensitivity. Furthermore, cell lines that were sensitive in vitro also showed tumor inhibition and prolongation of survival of xenografted mice after VU661013 treatment. Intriguingly, xenograft-bearing mice treated with venetoclax would ultimately develop resistance that was reversed by exposure to VU661013. The same development of resistance with acquired sensitivity was also noted when VU661013 was used first, followed by venetoclax to treat resistant cases. Synergy was noted with dual therapy. Importantly, this sensitivity of venetoclax-exposed leukemic cells to VU661013 was confirmed in clinical samples from patients who had been treated with the combination of venetoclax and cytarabine but had lost response. Sensitivity was regained in vitro by adding VU661013 to venetoclax and was more evident at relapse than at original diagnosis, indicating that during progression on venetoclax, MCL1 upregulation had occurred conferring sensitivity to VU661013. Taken together, these studies with different potent MCL1 inhibitors indicate that in AML, the ability to inhibit a resistance factor to venetoclax and to treat that subset of patients with predominant MCL1 dependence will increase our ability to elicit apoptosis in these clinical circumstances (Fig 1B-i).

However, optimal initial treatment selection remains an issue. MCL1 levels were not predictive. Extra information was helpful as BAK levels were somewhat predictive and high levels of BCL-xL correlated with resistance but were not sufficiently precise to use as biomarkers. Indeed, when Caenepeel and colleagues attempted to divine a set of predictive gene-expression markers, a 165-gene signature was required across all the cell lines used for training. The signature was not tested against a different test set of cell lines; thus, potential overtraining may have contributed to the signature. The only change universally predictable (but not predictive) was that addition of an MCL1 inhibitor dramatically increased the amount of MCL1 protein in cells. Therefore, alternative methods will be required to identify patients most likely to benefit from inhibition of MCL1.

In many cases, the best combination therapy will include agents like standard or targeted cancer therapies that elicit addiction that did not exist previously by loading empty antiapoptotic proteins with proapoptotic ligands. An elegant example of identifying the relevant players is described in the third article. Here, Nangia and colleagues built on their previous analysis to analyze EGFR-resistant non–small cell lung cancer (NSCLC; ref. 9). They had shown that a substantial proportion of cases are driven by mutations and/or overactivity of the KRAS/MAP kinase pathway that renders them (somewhat) sensitive to MEK inhibitors; siRNA screening identified BCL-xL knockdown as a significant enhancer of MEK inhibition. Trametinib prevented phosphorylation-induced degradation of BIM, thereby increasing the amount of BIM bound to BCL-xL that can be displaced by navitoclax (Fig 1B-ii). In the current article, they used AM-8621 to demonstrate that an equivalent inhibitor-sensitive loading of MCL1 occurred as it also synergized with trametinib—thus indicating joint addiction to BCL-xL and MCL1 in these tumors (10). In NSCLC lines and xenografts, therapy with trametinib and either AM-8621 or navitoclax caused tumor regression that was significant but incomplete. Combined administration of both inhibitors either simultaneously or in a defined sequence with trametinib was markedly more effective, presumably because any BIM that is “mopped up” by MCL1 after having been displaced from BCL-xL by navitoclax is then fair game for displacement from MCL1 by AM-8621. However, for sequential administration the order is important as brief navitoclax preexposure sensitized cells and mice xenografts to treatment with trametinib and AMG 176, but the reverse sequence did not show enhanced killing because of the prolonged intracellular retention of navitoclax compared with AM-8621 (Fig. 1B-iii).

The practical importance of this observation based upon a solid mechanistic understanding of rational exploitation of the apoptotic machinery in cancer cells is that it should limit off-target toxicity while retaining efficacy. Optimal scheduling will be particularly relevant to mitigate the on-target blood cytopenias expected with all these new agents. However, this clever and exciting strategy should be testable in clinical trials, and is further proof of principle for screening for other types of targeted therapy–induced addiction in cancers that do not show sensitivity to monotherapy with any of the now-expanded number of drugs that inhibit BCL2/MCL1/BCL-xL (11).

B. Leber has received honoraria from the speakers bureau of AbbVie and is a consultant/advisory board member for AbbVie and Amgen. No potential conflicts of interest were disclosed by the other authors.

D.W. Andrews is a Tier 1 Canada Research Chair and is funded by grant FRN12517 from the Canadian Institute of Health Research (CIHR).