Novel treatment strategies for chronic myeloid leukemia

Visual Abstract

Abstract

Starting with imatinib, tyrosine kinase inhibitors (TKIs) have turned chronic myeloid leukemia (CML) from a lethal blood cancer into a chronic condition. As patients with access to advanced CML care have an almost normal life expectancy, there is a perception that CML is a problem of the past, and one should direct research resources elsewhere. However, a closer look at the current CML landscape reveals a more nuanced picture. Most patients still require life-long TKI therapy to avoid recurrence of active CML. Chronic TKI toxicity and the high costs of the well-tolerated agents remain challenging. Progression to blast phase still occurs, particularly in socioeconomically disadvantaged parts of the world, where high-risk CML at diagnosis is common. Here, we review the prospects of further improving TKIs to achieve optimal suppression of BCR::ABL1 kinase activity, the potential of combining different classes of TKIs, and the current state of BCR::ABL1 degraders. We cover combination therapy approaches to address TKI resistance in the setting of residual leukemia and in advanced CML. Despite the unprecedented success of TKIs in CML, more work is needed to truly finish the job, and we hope to stimulate innovative research aiming to achieve this goal.

This Review Series highlights the transformative impact of tyrosine kinase inhibitors (TKIs) on therapy for chronic myeloid leukemia (CML) in the last 25 years. Brian J. Druker introduces the series by providing a Perspective that recalls the initial presentation at the 1999 American Society of Hematology annual meeting, revealing a 100% response rate at effective doses of imatinib in the phase 1 study. Guilhot and Hehlmann review long-term outcomes of patients with CML treated with TKIs in major studies and note that survival of patients with CML now approaches that of the general population. Hughes et al address the issue of treatment-free remission and explore the criteria for safely discontinuing TKIs for select patients. Finally, Cruz-Rodriguez and Deininger discuss mechanisms of resistance, emerging therapeutic strategies, and the promise of BCR::ABL1 degraders and TKI combinations.

Introduction

Starting with imatinib, successive generations of increasingly effective tyrosine kinase inhibitors (TKIs) have turned chronic phase chronic myeloid leukemia (CP-CML) from a lethal disease into a chronic condition compatible with almost normal life expectancy.¹ Asciminib, the first approved allosteric TKI and latest addition to the CML armamentarium, has set a new tolerability standard.² Considering this impressive success, one might conclude that little remains to be done. The reality, however, is that most patients with CML continue to face life-long TKI therapy, even those with good risk disease and access to state-of-the-art CML management. For patients with poor risk CML, few in developed countries but numerous in socioeconomically disadvantaged parts of the world, a CML diagnosis still means reduced length and quality of life.³ Thus, despite the unprecedented success of TKIs, the CML story is far from complete. What if there was a cure for most patients with CML similar to an infection that is no more than an episode in a patient’s medical history? What if we had an effective therapy for blast phase CML (BP-CML) that did not include the mortality and morbidity of allogeneic stem cell transplant? As granting agencies must be convinced to fund CML research, we need to focus on innovative strategies, avoid unnecessary duplication, and learn from the past 25 years of CML research to maximize return on investment. Here, we review novel treatment strategies and discuss possible future research directions. We do not cover the most effective strategy to improve outcomes on a global scale: ensuring universal access to affordable CML care.

Improving BCR::ABL1 TKIs

Have we reached the efficacy limit of targeting BCR::ABL1 kinase activity?

We and others have shown that primitive CML cells (from here on referred to as leukemia stem cells [LSCs]) can survive independently of BCR::ABL kinase activity, and yet a subset of patients in treatment-free remission (TFR) have no detectable residual leukemia.^{4, 5, 6} How can these observations be reconciled? One explanation is that another, perhaps immune-based mechanism keeps residual leukemia below the detection limit.⁴ Alternatively, minimal residual kinase activity, unmeasurable with current technology, may contribute to LSC survival. The observation that more potent second-generation (2G) TKIs increase the percentage of TFR-eligible patients, and consequently the number of patients achieving TFR, suggests that more potent BCR::ABL1 inhibition may further improve results.^7,8 The aborted phase 3 study of ponatinib in newly diagnosed CP-CML (Evaluation of Ponatinib versus Imatinib in Chronic Myeloid Leukemia [EPIC]) and data from asciminib trials provide insights into the potential of sustained, near-complete BCR::ABL1 inhibition.^2,9,10 The rates of molecular response 4 (MR4; 4-log reduction of BCR::ABL1) and MR4.5 in the EPIC trial are the highest reported in any TKI study of CML (Table 1), with the important caveats of small patient numbers and ponatinib’s activity against additional targets.⁹ Asciminib demonstrated superior major molecular response (MMR) rates over physician’s choice of imatinib, dasatinib, nilotinib, or bosutinib, although MMR was not statistically higher compared with 2G TKIs alone.² Results from a single-armed asciminib study in newly diagnosed CP-CML were similar.¹¹ In both trials, 12-months MR4.5 rates with an 80-mg asciminib total daily dose were 20% to 30% lower than the 60% in the EPIC study. However, asciminib, 80 mg daily, is only one-fifth of the 200 mg twice daily (BID) recommended by the US Food and Drug Administration (but not the European Medicines Agency) for patients with BCR::ABL^T315I. As this high dose of asciminib is generally well tolerated, there may be additional therapeutic space to improve the rates of deep MR (DMR; MR4 or better), and hence rates of TFR. In fact, the cumulative 12-month MR4.5 rate in patients with CP-CML with BCR::ABL1^T315I treated with asciminib, 200 mg BID, was 32%.¹⁰ In view of the greater resistance of BCR::ABL1^T315I, these data suggest asciminib doses >80 mg daily may increase MR4.5 rates in newly diagnosed CP-CML, and that the efficacy limit of BCR::ABL1 inhibition has not been reached with currently approved drugs and dosing schedules.¹² It is important to remember that toxicity could still turn out to be the ultimate limit of ABL1-directed therapies. Although the approved adenosine triphosphate (ATP)–competitive TKIs differ profoundly in target spectrum, potency, and pharmacokinetics, there is a class effect with respect to cardiovascular (CV) toxicity, with the exception of imatinib.¹³ Are CV adverse events (AEs) of BCR::ABL1 inhibitors partially due to ABL1/2 inhibition? And does imatinib’s excellent CV safety reflect the benefit of a relatively weak TKI that gives ABL1/2 kinases in endothelial cells sufficient time to breathe? Asciminib data suggest a low, but not zero, CV risk at 80 mg daily.¹⁴ Long-term follow-up of the various cohorts, particularly those treated with 200 mg BID, will be important.¹⁰

Table 1.

MRs in major randomized trials in newly diagnosed CP-CML

TKI	IC50 (nM)	Daily dose (mg)∗	MR3 (%)				MR4 (%)				MR4.5 (%)
			3 mo	6 mo	9 mo	12 mo	3 mo	6 mo	9 mo	12 mo	3 mo	6 mo	9 mo	12 mo
Imatinib	100–500	400	<1-3	8-18	18-30	22-40	0	5	8	12-15	0	<1	3	3-5
Dasatinib†	0.8–1.8	100	8	27	39	46
Nilotinib	<10-25	600/800	3-9	22-33	33-43	38-44	0	1	4	0	0	0	0	0
Bosutinib	41.6	400/500	4-7	27-35	35-42	41-47	<1	10	14	21	0	2	5	8
Ponatinib‡	0.5	45	31	62	86	80	7	32	64	60	5	16	32	60
Asciminib	0.5	80	48	68		68-77	14	34		39-48	7	18		17-20
Asciminib	2	400§		58		58		37		37		21		32

IC₅₀, 50% inhibitory concentration.

^∗Total daily dose.

^†Responses are cumulative.

^‡N = 109, 69, 22, and 10 for 3, 6, 9, and 12 months, respectively.

^§BCR::ABL1^T315I, no prior ponatinib.

Can we design better TKIs?

The term potency is often used to describe the composite of multiple factors that determine TKI activity (Figure 1). A direct comparison of target affinity between different TKIs is hampered by the lack of a standardized assay. Measuring kinase activity in cell-free systems is the most reproducible approach, but comparisons between studies are complicated by the use of different ATP concentrations, kinase constructs, and substrates as well as phosphorylated vs dephosphorylated enzyme. An often neglected characteristic is residence time, which reflects the ratio of on rates (binding to the target) and off rates (dissociating from the target).¹⁵ For many BCR::ABL1 TKIs, residence time was reported only based on cell-free systems, which do not capture the intracellular environment, where activators and inhibitors influence kinase conformation. Biology downstream of BCR::ABL1 kinase inhibition further complicates the situation, particularly whether depth or duration of BCR::ABL1 inhibition drives efficacy. In the case of dasatinib, the notion that peak plasma concentrations govern efficacy led to a once-daily dosing regimen.^{16, 17, 18} In the meantime, we have learned that the continuity of therapy afforded by lower, less toxic, doses is equally important.¹⁹ Another question is whether the full spectrum of engaging ABL1 has been exploited. The binding modes of TKIs differ in the positions of key structural features, including the aspartate-phenylalanine-glycine (DFG) motif, the activation loop, helix αC, and the assembly of residues in the catalytic core known as the spine.²⁰ Only a subset of TKIs extend into a hydrophobic pocket on the back of the catalytic site, access to which is controlled by the gatekeeper threonine 315. Depending on these features, TKIs can be classified into 6 types (Table 2).²⁰ Type I and type II inhibitors bind within the catalytic site, recognizing an active (type I) or inactive (type II) conformation. Type III inhibitors bind allosteric pockets close to the catalytic site without competing with ATP. Type IV inhibitors recognize allosteric sites distant from the catalytic pocket, type V inhibitors are bifunctional (binding different pockets), and type VI inhibitors are covalent inhibitors. All approved BCR::ABL1 TKIs, except asciminib, are type I or II inhibitors (Table 2). To our knowledge, no type III or type V ABL1 inhibitors and only preclinical type VI inhibitors have been reported, suggesting that additional design options could be explored to increase potency. For instance, nocodazole- and pyridopyrimidine-based compounds with picomolar 50% inhibitory concentration (IC₅₀) have been reported, significantly lower than the nanomolar IC₅₀ of the most potent clinical ABL1 inhibitors.^{12,21, 22, 23, 24} Another interesting development is the recently reported covalent ABL1 inhibitors targeting the catalytic lysine.^{25, 26, 27} The list of preclinical ABL1/2 inhibitors is extensive, and a comprehensive review of preclinical ABL1 inhibitors is beyond the scope of this review.

Figure 1.

Factors influencing the potency of TKIs. Key factors that contribute to the overall potency of TKIs include the following: (1) cellular uptake: the efficiency of drug entry into target cells; (2) mutant coverage: the ability of the TKI to inhibit different mutant forms of the kinase, particularly variants associated with resistance; (3) pharmacokinetics: the absorption, distribution, metabolism, and excretion of the drug, influencing bioavailability; (4) target affinity: the strength of the binding interaction between the TKI and its target kinase; (5) residence time: the duration for which the drug remains bound to its target; and (6) protein binding: the degree to which the TKI binds to plasma proteins, affecting the free drug concentration.

Table 2.

Types and key features of kinase inhibitors

Type	I	I1/2	II	III	IV	V	VI
Conformation	Active	Inactive	Inactive	Allosteric	Allosteric	Bivalent	Covalent
Extension into back cleft	No	Variable	Variable	Yes	No	Variable	Variable
DFG	In	In	Out	Variable	No	Variable	Variable
A-loop	Out	Variable	Variable	Variable	Variable	Variable	Variable
Helix αC	In	Variable	Variable	Out	Variable	Variable	Variable
Spine	Linear	Distorted	Distorted	Distorted	Variable	Variable	Variable
ATP competitive	Yes	Yes	Yes	No	No	Variable	Variable
Reversible	Yes	Yes	Yes	Yes	Yes	Yes	No
Approved TKIs	DAS	BOS	IM, NIL, RAD, PON	NA	ASC	NA	NA

ASC, asciminib; BOS, bosutinib; DAS, dasatinib; IM, imatinib; NA, not applicable; NIL, nilotinib; PON, ponatinib; RAD, radotinib.

Novel clinical BCR::ABL1 inhibitors

At least 4 new TKIs are currently in clinical development in the United States.

Olverembatinib, structurally remarkably similar to ponatinib (Figure 2), showed significant activity in a phase 1/2 study of patients with CP-CML or accelerated phase (AP)-CML who had failed prior TKI therapy, including >80% exposed to ≥2 TKIs. For CP-CML, 3-year cumulative incidences of complete cytogenetic response (CCyR), MR3, MR4, and MR4.5 were 69%, 56%, 44%, and 39%, respectively. Common AEs included skin hyperpigmentation, hypertriglyceridemia, proteinuria, and thrombocytopenia.²⁸ In multivariate analysis, the BCR::ABL1 genotype showed the strongest correlation with response, with the best results in patients with a single BCR::ABL1^T315I mutation, possibly reflecting the genetic homogeneity of this population. Interestingly, activity extended to patients with prior ponatinib resistance, and to patients with BCR::ABL1^T315I-inclusive compound mutations, which are ponatinib resistant in vitro.²⁹ The latter data require validation, especially because the reported variant allele frequencies for compound mutations (eg, T315I/F359V [7.1%]; T315I [73.0%]; and F359V [10.4%]) are possible only if at least 1 mutation was acquired twice. CV AEs possibly related to olverembatinib were observed in 53 patients (32.1%), but arterial occlusive events were rare. Given the median age of only 42 years and the stringent exclusion criteria for CV risk, validating the toxicity profile in a more real-world population will be important. This being said, the fact that hyperpigmentation is common with olverembatinib but rare with ponatinib is a reminder that even small structural changes can have profound biological effects. Olverembatinib (40 mg every other day) is approved in China for the treatment of TKI-resistant CP-CML or AP-CML. Trials in CML and Philadelphia chromosome (Ph)⁺ acute lymphoblastic leukemia are underway in the United States.

Figure 2.

Chemical structures of novel BCR::ABL1 inhibitors in clinical development. Structures of olverembatinib and vodobatinib, 2 novel BCR::ABL1 TKIs currently in clinical development in the United States, with structures of ponatinib and dasatinib shown for comparison. Olverembatinib shares significant structural similarities with ponatinib, including core motifs that confer activity against the BCR::ABL1 T315I mutant. Vodobatinib exhibits structural features common to both ponatinib and dasatinib, indicating shared mechanisms of BCR::ABL1 inhibition.

Vodobatinib, formerly known as K0706, shares structural features with ponatinib and dasatinib (Figure 2). Vodobatinib has activity against a range of BCR::ABL1 genotypes, but not BCR::ABL1^T315I.³⁰ In a phase 1 study (NCT02629692) of patients with CML with failure to ≥3 TKIs and/or comorbidities that restrict the use of nilotinib, dasatinib, and ponatinib, 35 patients, including 27 with CP-CML, received doses ranging from 12 to 240 mg daily. The drug was generally well tolerated. Of 22 patients with CP-CML evaluable for cytogenetic response, 7 achieved and 4 maintained CCyR, 5 achieved MMR, and 2 achieved MR4.5. Most responses were durable. The 2 patients with BCR::ABL^T315I had early progression, leading to a protocol amendment to exclude patients with BCR::ABL^T315I.³¹ An exploratory analysis compared ponatinib-pretreated and ponatinib-naive patients with CML-CP. Efficacy was comparable, with 50% ponatinib-pretreated and 67% ponatinib-naive patients achieving or maintaining CCyR.³² Vodobatinib’s lack of activity against BCR::ABL1^T315I defines it as a 2G BCR::ABL1 inhibitor. Given that several 2G TKIs are already approved, future development of this agent is uncertain.³¹

TGRX-678 is an allosteric TKI with low nanomolar activity against cell lines expressing native or single kinase domain mutant BCR::ABL1, with a slightly higher IC₅₀ (66 nM) for BCR::ABL1^T315I.³³ Activity was demonstrated in xenograft models of human and mouse cell lines expressing native and T315I-mutated BCR::ABL1. Combinations with ponatinib were strongly synergistic. Mutations in the myristoyl pocket or in the SRC homology 3 (SH3)-kinase domain interface reduced efficacy, consistent with an allosteric mode of action. Preliminary results of a phase 1 study in 58 patients with CP-CML and 37 patients with AP-CML were reported in December 2023.³⁴ At baseline, 82% of patients had received ≥3 and 46% had received ≥4 lines of TKI therapy. Of response-evaluable patients with CP-CML, 89% achieved complete hematologic response, 33% achieved CCyR, and 19% achieved MMR. Of 14 patients with CML-CP with T315I, 100% achieved complete hematologic response, 57% achieved CCyR, and 50% achieved MMR. Phase 1 and phase 2 studies are underway in China and the United States. The structure of TGRX-678 has not been reported.

TERN-701 is an allosteric inhibitor with a hitherto undisclosed structure that inhibits proliferation of CML cell lines expressing native or mutant BCR::ABL1, including BCR::ABL1^T315I, with IC₅₀ ranging from 5 to 26 nM and 70 to 1200 nM against active site and myristoyl site mutants, respectively. TERN-701 is highly selective for BCR::ABL1 in kinase and cell line–based assays, and is synergistic with dasatinib and ponatinib.³⁵ An international phase 1 study in CML (NCT06163430) is ongoing.

ELVN-001 is an orthosteric TKI reported to bind a unique active ABL1 conformation characterized by an inwardly rotated P-loop, providing high selectivity (inactive against KIT and PDGFRB). The structure of this molecule has not been disclosed. A phase 1 study (NCT05304377) is recruiting in the United States and Europe.

Combining allosteric and orthosteric TKIs

The availability of ATP competitive orthosteric and allosteric ABL1 TKIs begs the question whether their combination may be synergistic (Figure 3). Point mutations described initially in the BCR::ABL1 kinase domain and subsequently in the SH2 and SH3 domains, the SH1-SH2 linker, and most recently the myristate pocket are the most extensively studied mechanism of TKI resistance. In aggregate, the approved TKIs effectively address single mutants, with ponatinib providing the most comprehensive coverage.²⁴ One mechanism of synergy is more comprehensive coverage of BCR::ABL1 variants with nonoverlapping resistance to each single TKI. In the initial report of asciminib, synergy between nilotinib and asciminib was demonstrated in KCL22 CML cells, where the combination prevented outgrowth of resistant BCR::ABL1 mutant subclones in vitro and in vivo.¹² However, a more relevant synergy mechanism materializes when orthosteric and allosteric TKIs mutually reenforce their binding to the same BCR::ABL1 molecule. This may be particularly relevant with respect to BCR::ABL1^T315I-inclusive compound mutants that are resistant to all approved TKIs.³⁶ We and others have shown that combining asciminib with ponatinib is highly synergistic, bringing even BCR::ABL1^T315I-inclusive compound mutants into the range of clinically achievable drug concentrations.^29,37 Concordant with the biochemical data, computational simulations suggest that the simultaneous binding of asciminib to ABL1 and ABL1^Y253H/T315I reduces Gibbs free energy (ΔG) by 1.3 and 2.3 log, respectively.²⁹ As asciminib enforces an inactive ABL1 conformation and ponatinib is a type II inhibitor recognizing an inactive kinase conformation, this result is intuitive. Interestingly, a recent article reporting the structure of ABL1 in complex with asciminib and SRC inhibitor 1 reported cooperative binding. As DFG in this complex is “in,” whereas α-C helix is “out,” cooperative binding does not seem to depend on DFG conformation, suggesting that synergy based on mutually reenforced binding may extend to dasatinib, which is known to bind an active ABL1 conformation.^38,39 Solving the structure of native and compound mutant ABL1 in complex with asciminib and potent type I and type II inhibitors is needed to definitively clarify the situation. Another advantage of TKI combinations may be dose reduction, minimizing AEs from off-target effects. The asciminib phase 1 study explored combinations with orthosteric TKIs. Of the patients treated with asciminib/imatinib (AI), asciminib/nilotinib (AN), and asciminib/dasatinib (AD), 72%, 77%, and 66%, respectively, reported grade ≥3 AEs, with arterial occlusive events in 12%, 8%, and 9%, respectively. The maximum tolerated dose for AI was reached at asciminib, 60 mg daily, plus imatinib, 400 mg daily, whereas the maximum tolerated dose for the other combinations was not reached. By week 96, MR3 rates in response-evaluable patients were 45%, 32%, and 46% in the AI, AN, and AD arms, respectively.^40,41 The combination concept was tested systematically in the Frontline Asciminib Combination in Chronic Phase CML (FASCINATION) study, which tested AI, AD, and AN in newly diagnosed CP-CML.⁴² Of 125 patients, 21 (17%) discontinued therapy within the first 12 months, and the MR4 rate at 12 months was 38%, indicating that these combinations were effective but associated with impaired tolerability, and it seems unlikely that this concept will be pursued further. Case reports suggest combinations of asciminib and ponatinib are tolerated and can have clinical activity.^43,44 Efforts are underway to launch a clinical trial to assess this combination in advanced CML.

Figure 3.

Synergy mechanisms in drug combinations targeting BCR::ABL1. Synergy in combination therapies against BCR::ABL1-driven leukemias may be achieved through several mechanisms: (1) complementary inhibition: 2 different TKIs target separate cell clones carrying different BCR::ABL1 mutations; (2) enhanced target affinity: 2 TKIs simultaneously bind allosteric (myristoyl) and orthosteric (ATP binding) pockets of BCR::ABL1 protein, mutually enhancing their affinity to the target; and (3) synthetic lethality: simultaneous inhibition of 2 independent pathways, typically BCR::ABL1 kinase and a compensatory pathway. In isolation, each pathway is dispensable, but their simultaneous disruption is lethal. KD, kinase domain.

Improving how we use approved TKIs

Toxicity caused by the TKI doses recommended for frontline therapy may compromise efficacy. In clinical practice, many patients are started at lower TKI doses to avoid dose interruptions without compromising efficacy. As an early example, the Tyrosine Kinase Inhibitor Optimization and Selectivity (TOPS) trial of imatinib, 400 mg daily vs 800 mg daily, used a rigid protocol to maintain dose intensity, but failed to show superiority for the higher dose, likely reflecting frequent dose interruptions. In contrast, the higher dose was superior in the CML IV trial of the German study group that used an initial lower-dose run in and favored dose reductions over dose interruptions.^45,46 In the Evaluating Nilotinib Efficacy and Safety in Clinical Trials-Newly Diagnosed Patients (ENESTnd) trial, responses with nilotinib, 300 mg twice daily, were equal or slightly superior to those with 400 mg twice daily.⁴⁷ Additional observations supporting lower doses are the 2 phase 3 trials of bosutinib, 500 mg daily (Bosutinib Efficacy and Safety in Newly Diagnosed Chronic Myeloid Leukemia [BELA] study) and 400 mg daily (Bosutinib Trial in First-line Chronic Myelogenous Leukemia Treatment [BFORE] study),^48,49 and a retrospective propensity score analysis that found dasatinib, 50 mg daily, equal or even superior to 100 mg daily.^50,51 An alternative approach is to start with a high dose until the desired response has been achieved, then dose reduce to optimize tolerability. A retrospective comparison of 2 separate trials of frontline dasatinib in CP-CML (PCR-DEPTH and DAS-CHANGE) suggests that dose reduction for early MR or toxicity may improve MR rates.⁵² The Optimizing Ponatinib Treatment in CML (OPTIC) study of ponatinib in third-line therapy of CP-CML evaluated 3 different stating doses and response-driven dose reduction. Efficacy was comparable to the Ponatinib Ph⁺ ALL and CML Evaluation (PACE) study, with lower CV toxicity.⁵³ A limitation of this important space is the retrospective nature of most studies. Additional prospective trials are needed to provide definitive evidence for the concepts of lower starting doses and response-based early dose deescalation. Whether dose reductions in advanced CML are safe is unknown, and they are currently not recommended.

BCR::ABL1 dimerization inhibitors

The first phosphorylation event during BCR::ABL activation occurs in trans and requires dimer formation afforded by the BCR coiled coil domain (ie, dimerization domain).^54,55 Deletion of the dimerization domain abrogates kinase activity, unless it is replaced with an alternative dimerization mechanism.⁵⁶ The dimerization requirement can be exploited by stapled or mutated peptides that block dimer formation.^{57, 58, 59, 60} As pups from Bcr^–/– mice are viable and have no concerning phenotype, BCR dimerization inhibitors would likely be well tolerated.⁶¹ To date, the challenges common to peptide-based therapeutics aiming at intracellular targets have prevented the clinical development of BCR::ABL1 dimerization inhibitors.

BCR::ABL1 degraders

BCR::ABL1 scaffold functions

BCR::ABL1 scaffold functions contribute to leukemogenesis (reviewed previously by Cruz-Rodriguez et al⁶²). For instance, cord blood CD34⁺ cells transduced with p210^BCR::ABL1 exhibit increased proliferation, reduced adhesion to fibronectin, and reduced chemotaxis to stroma-derived factor-1α.⁶³ Imatinib abrogates abnormal growth but does not completely restore normal adhesion and migration. In line with the latter, expression of a kinase inactive BCR::ABL1 mutant (p210^{BCR::ABL1 K1172R}) has no effect on proliferation, but partly recapitulates the aberrant migration and adhesion phenotype, and this seems to require the BCR::ABL1 C-terminus and F-actin binding.^63,64 BCR::ABL1 expression, but not kinase activity, drives a JAK2/β-catenin survival/self-renewal pathway that includes inhibition of the PP2A tumor suppressor in mouse Lin^–Sca1⁺Kit⁺ cells.⁶⁵ BCR::ABL1^Y177 is an autophosphorylation site.⁶⁶ On TKI inhibition of BCR::ABL1, phosphorylation of BCR::ABL1^Y177 is maintained by SYK or JAK2, ensuring continued mitogen-activated kinase and phosphatidyl inositol 3’ kinase signaling.^{66, 67, 68, 69} Altogether, these and additional observations suggest that BCR::ABL1 degradation would be more effective than even near-complete kinase inhibition. This is not unique, as nonenzymatic oncogenic functions have been described for several other cancer-related kinases, including epidermal growth factor receptor, Bruton tyrosine kinase, and BRAF.^{70, 71, 72, 73, 74, 75, 76}

Proteolysis-targeting chimeras

Previous approaches to degrade BCR::ABL1 protein-targeted chaperones include heat shock protein 90 (HSP-90). Direct HSP-90 inhibitors, such as 17-(allylamino)-17-demethoxygeldanamycin and HDAC6 inhibitors that disrupt HSP-90 function by enhancing acetylation, have preclinical activity, but their clinical development failed because of toxicity.⁷⁷ This liability may be overcome by a new generation of compounds that target HSP-90 dimerization via the C-terminal domain.⁷⁸ However, a more promising strategy is proteolysis-targeting chimeras (PROTACs), heterobifunctional molecules composed of a warhead (binding the protein of interest [POI]), a linker of variable length, and an E3 ligase recruiter. Once ubiquitinated, the POI is recognized by the 26s proteasome and degraded. In contrast to TKIs, where inhibition continues only as long as TKI is bound to the target, PROTACs are released following each round of degradation, theoretically establishing iterative degradation cycles until the POI is eliminated. The design principles for PROTACs are less well defined than those for TKIs: target affinity and activity are not proportional, and linker length and charge impact activity in an unpredictable manner. For obvious reasons, expression of the E3 ligase in the cancer cells is sine qua non for any PROTAC to work. Starting in 2016, several BCR::ABL1 PROTACs have been reported (recently reviewed by Cruz-Rodriguez et al⁶²). All approved TKIs, except imatinib, have been used successfully as warheads, and VHL, CBN, and cIAPs as E3 ligases. Unfortunately, many of the published BCR::ABL1 PROTACs are incompletely characterized with respect to specificity, and most reports do not include a degradation-incompetent derivative, an essential control to distinguish between degradation and kinase inhibitory activity. Moreover, data from relevant CML mouse models and primary CML cells are mostly unavailable, as are toxicity studies. The latter is a significant gap, as Abl1^–/– mice are born runted and combined Abl1/Abl2 knockout is embryonically lethal.⁷⁹ Whether ABL1/2 degradation in normal tissues would be tolerated as well as ABL1/2 inhibition remains to be established. Thus far, no BCR::ABL1 PROTAC has entered clinical trials.

Combination therapy

The realm of combination therapy is at the extremes of the clinical CML spectrum, residual leukemia in TKI responders, and TKI-resistant BP-CML without explanatory BCR::ABL1 kinase mutations. Here, BCR::ABL1 kinase-independent mechanisms contribute to CML cell growth and survival, using mechanisms that show more similarities than one would expect from fundamentally different clinical scenarios.^80,81 The synergy of combination therapy can be based on at least 3 mechanisms (Figure 3).

Let us first consider patients responding to TKIs who harbor residual CML. Reducing the LSC burden in these patients appears to be a rational strategy to improve DMR rates, and hence, rates of TFR. Research over the past 2 decades has implicated a plethora of signaling pathways as potential targets, typically in combination with BCR::ABL1 inhibition. However, only a handful of potential targets have been evaluated in a clinical trial, and even fewer in randomized studies (Table 3). Thus far, only 1 study—combining BCR::ABL1 TKIs with the JAK1/2 inhibitor ruxolitinib—has read out positive, improving the rate of DMR.^82,83 Despite promising results from single-armed studies, several prospective trials testing combinations of TKIs with interferon alfa failed to demonstrate a consistent and meaningful improvement of TFR rates.^84,85 Maximizing the chances of detecting an efficacy signal requires trials in patients responding to TKIs, but convincing such patients to participate in studies that mean more clinic visits and potentially cause new AEs is challenging. Overcoming this barrier may require a reductionist trial design, with fewer clinic visits, less correlative science, and fewer laboratory draws. A more easily fixable issue is semantics. “Eradicate LSCs” is often used when reporting positive studies and implies complete elimination, although LSC kill is typically incomplete. Thus, even if the preclinical data were fully translatable into the clinic, some LSCs would still survive. Perhaps, we should tone down the language, while raising the preclinical bar. LSCs are heterogeneous, even in those patients whose only detectable genetic abnormality is BCR::ABL1. Are the different combinations active against the same or different LSC subpopulations? At first glimpse into the complexity of LSCs is the recent observation that patients with erythroid-megakaryocyte–biased leukemia stem and progenitor cells tend to have more profound responses to TKIs.^86,87 LSC heterogeneity can explain how many pathways can be simultaneously essential (considering a subset of LSCs) and dispensable (considering the totality of LSCs). In view of this heterogeneity, targeting universal mechanisms (eg, epigenetic reprogramming) may hold more promise than targeting specific pathways. For instance, to our knowledge, the combination of TKIs and hypomethylating agents was never evaluated in frontline TKI therapy of newly diagnosed CP-CML. Maybe all we need to do is to prevent epigenetic reprogramming of LSCs in the first hours of TKI exposure?

Table 3.

Pathways implicated in LSC survival and clinical trials

Pathway	Supporting publications	Clinical trial		Status	Final report
		Advanced CML	Early CML or remission
ADAR1	88				NA
Autophagy	89,90	No	Hydroxychloroquine + IM vs IMRAN	Completed	Published, borderline positive91
BCL2	92,93	VEN + DAS		Completed	Published, negative94
		VEN + AZA + FLU		Completed	NR
		VEN + ASA + 3G TKI		Not yet recruiting	NA
		VEN + DEC + PON		Completed	Published, positive95
			VEN after TKI	Recruiting	NA
BCL6	96
CXCL12	97				NA
EZH2	98,99				NA
HDAC	100	Panobinostat		Terminated	NR
		Panobinostat		Terminated	NR
		Panobinostat + IM		Completed	Abstract, negative
Hedgehog	101,102	LDE225 + NIL		Completed	Abstract, negative
		BMS-833923 + DAS		Terminated	NA
JAK1/2-STAT3/5	65,103, 104, 105, 106	RUX + NIL (Germany)		Completed	NR
			RUX + NIL (Tampa)	Completed	Published, positive82
			RUX + NIL (MDACC)	Terminated	NA
		RUX + NIL (Canada)		Completed	NR
			RUX + TKIRAN∗	Completed	Abstract, positive83
5-Lipoxygenase	107	Zileuton + IM	No	Terminated	No
MDM2/TP53	108,109				NA
miR-126	110				NA
miR-183/EGR1/E2F1	111				NA
Mitochondrial metabolism	112, 113, 114, 115, 116				NA
Musashi2	117				NA
MYC	118,119				NA
PIM2	120				NA
PML	121,122	Multiple ATO + TKI		Terminated or completed	NR
			Realgar/indigo + IMR		Published, borderline positive123
PP2A	65,124, 125, 126				NA
PPRγ	127		Pioglitazone + IM	Completed	Published, positive127
PRMT5	128				NA
Rad52	129				NA
SIRT1	130				NA
TGF-β	131				NA
WNT/β-catenin	9,132, 133, 134, 135, 136, 137, 138, 139, 140				NA

ADAR1, adenosine deaminase acting on RNA 1; ATO, arsenic trioxide; DAS, dasatinib; FLU, fludarabine; HDAC, histone deacetylase; IM, imatinib; MDACC, MD Anderson Cancer Center; NA, not applicable; NIL, nilotinib; NR, not reported; PON, ponatinib; RAN, randomized; RUX, ruxolitinib; SIRT, sirtuin; TGF-β, transforming growth factor-β; VEN, venetoclax.

^∗Dasatinib, nilotinib, bosutinib, or imatinib.

A completely different scenario is advanced CML, defined by morphology, genetics, or TKI resistance, particularly in the absence of explanatory BCR::ABL1 mutations. Whole-exome or targeted DNA sequencing and RNA sequencing reveal a complex genetic landscape.^{81,141, 142, 143} Mutations in the BCR::ABL1 kinase aside, ASXL1, RUNX1, and IKZF1 are the most frequently mutated genes. Mutational activation of alternative tyrosine kinases and RAS signaling are uncommon, and few of the mutants are directly targetable. However, there is evidence for conversion of diverse upstream signaling pathways on a common epigenetic program whose main features are reduced activity of the polycomb repressive complex 2 (PRC2) and increased activity of PRC1.⁸¹ Progress in advanced CML may come from at least 2 directions. Targeting PRC1 with specific inhibitors may soon become feasible, as several compounds are in preclinical development. For patients with targetable genetic lesions, such as MLL rearrangements or IDH1/2 mutations, pathway-directed therapy may be effective, provided rapid turnaround of genetic results is feasible. Additionally, precisely reconstructing clonal architecture to identify truncal mutations may guide optimal target selection to slow selection of resistant subclones.

Combinations involving conventional chemotherapeutics

The current recommendation for the management of BP-CML is combinations of TKIs with acute myeloid leukemia– or acute lymphoblastic leukemia–type multiagent chemotherapy regimens. This is based on retrospective analysis of BP-CML outcomes, where treatment with TKI/chemotherapy combinations proved prognostically favorable.¹⁴⁴ Two recent studies reported promising results with ponatinib-based regimens. The Matchpoint trial tested fludarabine, cytarabine, idarubicin, and G-CSF (FLAG-Ida) combined with ponatinib at a starting dose of 30 mg daily in 17 patients with lymphoid or myeloid blast phase.¹⁴⁵ Twelve patients proceeded to allogeneic stem cell transplant. The 3-year overall survival was 41%. The ponatinib and subcutaneous azacitidine in chronic myelogenous leukemia patients in accelerated phase or in myeloid blast crisis (PONAZA) study reported results on 19 patients with myeloid BP treated with a combination of ponatinib (45 mg daily initial dose) and 5-azacytidine. The complete response rate was 68%, and 2-year overall survival was 65%.¹⁴⁶

Immunotherapy

Numerous studies have identified correlations between the likelihood of achieving TFR and various immune parameters, including subsets of natural killer cells, dendritic cells, and T cells, and a comprehensive review of this topic is beyond the scope of the current article. Altogether, these data seem to suggest that immune effects play a role in the control of residual leukemia after successful TKI-based debulking. Thus far, however, a mechanism has remained elusive, in contrast to the myeloperoxidase-directed cytotoxic T-cell responses described in patients with CCyR to interferon.¹⁴⁷ Given the diversity of observations and their correlative nature, at present it is difficult to see how they could be translated into a therapeutic intervention. Given the low risk of attempting TFR, any test to predict successful TRF requires a high negative predictive value to be clinically useful. In contrast to autoimmunity, the value of alloimmunity has stood the test of time, and current thinking still holds that only allotransplant can cure BP-CML. Achieving a second CP before transplant determines outcome, as both relapse rate and transplant-related mortality are excessive in patients transplanted in overt BP. As all other patients managed with allogeneic stem cell transplants, patients with CML stand to benefit from the progress in managing transplant-related complications, and the reader is referred to a recent review of allotransplant in CML.¹⁴⁸

Conclusion

As of 2024, the CML story is not finished, and therapeutic development continues. Generating preclinical data has become much easier, and the challenge is now to prioritize among the numerous options. Careful consideration of clinical relevance and context and avoidance of unnecessary repetition are critical to steer limited research funding to the most impactful areas.

Conflict-of-interest disclosure: M.W.D. is a paid consultant for Novartis, Pfizer, Incyte, Blueprint, and Cogent and has received research funding from Terns Pharma. N.C.-R. declares no competing financial interests.