Intrinsic cellular resistance to BCR::ABL1 inhibitors

Abstract

The clinical implementation of BCR::ABL1 tyrosine kinase inhibitors (TKI) for the treatment of chronic myeloid leukemia (CML) represents one of the big successes of mechanism-based cancer therapy. In 2025, the survival of patients who start TKI therapy while in the chronic phase is approaching that of age-matched controls. Despite this paradigm shift, significant challenges remain. Some patients still develop overt TKI resistance and progress to blast phase, and the majority continue to harbor residual leukemia and require life-long TKI therapy. Growth and survival signals arising from the microenvironment or from within the leukemia cells confer various degrees of resistance to support a spectrum of leukemic activity ranging from overt acute leukemia in blast phase to persistence of minimal residual disease in patients with a deep molecular response. Here we review cell-intrinsic resistance, covering both reactivation of BCR::ABL1 kinase activity and the less well-defined mechanisms underlying BCR::ABL1-independent TKI resistance. We propose that the pathways used by CML to escape TKI effects reflect the potential and the constraints of BCR::ABL1-driven reprogramming of hematopoietic stem and progenitor cells and that the role of BCR::ABL1 functions other than kinase activity may be underappreciated, providing a rationale for the clinical development of BCR::ABL1 degraders.

Introduction

Chronic myeloid leukemia (CML) is caused by BCR::ABL1, a constitutively active tyrosine kinase generated as a result of the t(9;22)(q34;q11.2) reciprocal translocation, cytogenetically detected as the Philadelphia chromosome (Ph).^1,2 In the Western world, most patients are diagnosed in the chronic phase (CP-CML), which is characterized by expansion of myeloid progenitors that maintain terminal differentiation capacity. Without effective treatment, CP-CML invariably progresses to the blast phase (BP-CML), a therapy-resistant acute leukemia of myeloid or lymphoid phenotype. Transformation to BP-CML may occur suddenly or gradually through an intermediary stage historically termed accelerated phase (AP-CML). Although AP criteria were included in earlier classifications, the most recent World Health Organization (WHO) classification no longer recognizes AP-CML as a distinct disease phase, reflecting continuous debate regarding its diagnostic and biological validity.^3,4 Advanced CML at diagnosis is more common in countries with lower socioeconomic status, which is thought to be the result of delayed diagnosis or, less likely, ethnic differences in disease biology.⁵

Starting with imatinib, successive generations of tyrosine kinase inhibitors (TKI) have dramatically improved CML prognosis.⁶ Today, the survival of patients diagnosed in CP-CML with access to modern health care is close to that of age-matched control populations, a quantum leap compared to the 5-year survival in the pre-TKI era.^7,8 However, transformation to BP-CML still occurs in 10-15% of patients, particularly in those with more advanced CML at presentation, inconsistent access to health care, or poor compliance.⁹ Only one quarter of patients can discontinue TKI permanently and achieve a state of functional cure termed treatment-free remission.¹⁰ The remainder require lifelong therapy to avoid recurrence of active leukemia, suggesting that fully leukemogenic CML stem cells (LSC) survive despite prolonged exposure to TKI.

In the three decades since the first clinical use of imatinib, an extensive literature on TKI resistance in CML has accumulated. While both intrinsic cell-autonomous and extrinsic microenvironmental mechanisms contribute to resistance, in the present manuscript we focus on cell-intrinsic TKI resistance, emphasizing overarching principles and clinical context whenever possible. We have structured the paper around the theoretical concept that the patterns of TKI resistance in CML reflect the biological constraints and capabilities of hematopoietic stem and progenitor cells (HSPC) transformed by BCR::ABL1. We hope that this approach will provide orientation and stimulate future research.

Definitions

In 2025, CML resistance to therapy is almost synonymous with resistance to TKI, although one should remember that BP-CML is frequently resistant to multiple modalities, including cytotoxic chemotherapy and hematopoietic stem cell (HSC) transplantation.¹¹ Mechanistically, TKI resistance may be extrinsic, involving growth and survival factors provided by the microenvironment, or intrinsic, mediated by pathways that originate from within the CML cell. While this concept is logical, distinguishing clearly between extrinsic and intrinsic mechanisms is challenging, and clinical resistance may often reflect a combination. This applies both at the single-cell and population level, where multiple cell clones may use different extrinsic or intrinsic strategies to achieve TKI resistance.^12,13 Clinically, resistance can be primary (failure to achieve a desired level of response) or secondary (acquired), occurring after an initial response to TKI. Lastly, the level of residual leukemia must be considered. In CP-CML, failure to achieve a complete hematologic response is very uncommon, while a substantial proportion of patients fail to achieve a deep molecular response, i.e., a reduction of BCR::ABL1 transcripts by 4-log or more. The term persistence is used to describe a situation in which patients harbor low-level residual CML but have a very low relapse risk. However, the rare occurrences of sudden BP in patients even with undetectable BCR::ABL1 is evidence that no level of residual CML offers absolute protection from transformation to BP.^14,15 Considering how difficult it is to draw the line between resistance and persistence, it may be preferable to conceptualize TKI resistance as a continuum.

Reactivation of BCR:ABL1 kinase

BCR::ABL1 mutations

In physiological conditions, ABL1 is maintained in an inactive conformation by complex mechanisms that involve binding of myristate to an allosteric pocket at the base of the kinase domain. In this auto-inhibited conformation, the ABL1 SH2 and SH3 domains are packed on the kinase domain, the DFG motif is in an outward position, and the activation loop assumes a closed conformation.^16,17 TKI can be separated into several classes according to their binding modes. In the case of ABL1, only type I inhibitors (binding an active conformation), type II inhibitors (binding an inactive conformation) and type IV inhibitors (allosteric, binding to the myristoyl pocket) are presently of practical relevance, but TKI belonging to other classes may be developed in the future. Point mutations in BCR::ABL1 that impair drug binding are the best-defined mechanism of TKI resistance.¹⁸ The spectrum of resistance mutations depends on the type of TKI. Within the ATP-competitive TKI, the scope of mutations is largest for the first-generation TKI imatinib. In addition to the commonly observed T315I mutation, multiple other mutations have been reported across different studies (Table 1). Although most of these variants are in the kinase domain of BCR::ABL1, some locate to the SH2 domain or the SH1-SH2 linker.¹⁹ The second-generation TKI nilotinib, bosutinib and dasatinib exhibit both unique and shared resistance mutations within an overall narrower spectrum of BCR::ABL1 genotypes. Collectively, they cover all single mutants except T315I, which is sensitive only to the third-generation TKI ponatinib and olverembatinib. While no single BCR::ABL1 mutant confers absolute resistance to ponatinib, some are less sensitive, most importantly E255V. Clinical resistance to asciminib is associated with mutations not only in the myristoyl (allosteric) pocket, but also in the SH3 domain and in several kinase domain residues already known to confer resistance to ATP-competitive TKI.²⁰ The sensitivity of BCR::ABL1 mutants to the various TKI is typically reported in the form of heatmaps that report half-maximal inhibitory concentration (IC₅₀) values based on growth inhibition of BaF3 or 32Dcl3 cells expressing the variants. While this provides useful orientation, in vitro profiling predicts clinical resistance much more reliably than clinical response. Not all BCR::ABL1 mutations identified in patients have been characterized functionally, and comprehensive validation of predictions based on observed clinical responses is not yet available. BCR::ABL1 mutations may confer resistance through several mechanisms. Most straightforward is steric hindrance, initially reported for the T315I gatekeeper mutation that creates a bulky substation.¹⁸ A more indirect mechanism is conformational rearrangements that stabilize an active kinase conformation, which excludes type II TKI. For instance ABL1, kinase-domain mutations such as M351T (C-lobe αE helix region) and H396P (activation loop) favor an active kinase conformation that decreases affinity for type II but not type I TKI.⁶ In the case of asciminib, mutations that interfere with packing of the SH3 domain on the SH1 domain prevent the assembly of the kinase in the autoinhibited conformation, although asciminib still binds to the myristate pocket. These mutations include M244V, located within the N-lobe ABL1 kinase domain adjacent to the P-loop and F359V in the C-lobe. Rare BCR::ABL1 variants lack ABL1 exon 2 (such as e13a3), which prevents the assembly of an autoinhibited conformation, conferring primary asciminib resistance.^20,21 Several mechanisms may cooperate to achieve high level resistance. In the case of T315I, in addition to creating a steric clash, the mutation disrupts a hydrogen bond with imatinib and stabilizes the hydrophobic spine of the kinase, favoring an active conformation.²² The ability to leverage more than one mechanism may contribute to the ‘success’ of T315I in conferring resistance to multiple different TKI. Another layer of complexity is added by compound mutations, i.e., the presence of two mutations in cis.²³ Compound mutations that encompass T315I render BCR::ABL1 resistant to all ATP-competitive TKI, including ponatinib and, unexpectedly, also asciminib, perhaps by preventing assembly of the myristate pocket.²⁴ However, T315I-inclusive compound mutants are sensitive to combinations of asciminib and ponatinib.²⁴ Computational simulations suggest that the addition of asciminib reduces Gibbs free energy (ΔG⁰) for compound mutant ABL1 to levels equivalent to those of wild-type ABL1.²⁴ Solving the structure of compound mutant ABL1 in complex with both TKI will be required to definitively clarify the mechanism. Data for olverembatinib are unavailable in the public domain, but given the similarities between the two molecules, the suspicion is that its activity profile might resemble that of ponatinib.²⁵ The fascinating variety of mutational resistance mechanisms in BCR::ABL1 may reflect the complex mechanism of physiological ABL1 regulation that involves allosteric and ATP binding pockets, offering multiple opportunities to disrupt TKI binding.

Table 1.Mutations identified in ABL1 and their association with clinical and in vitro resistance.

Alternative mechanisms of BCR::ABL1 kinase reactivation

Compared to BCR::ABL1 mutations, other mechanisms of reactivating the kinase are less well characterized. CML cell lines cultured in vitro in the presence of increasing concentrations of TKI frequently overexpress BCR::ABL1 through genomic amplification or transcriptional upregulation.²⁶ The high concentrations of BCR::ABL1 protein appear to be poorly tolerated, and upon TKI withdrawal, cells revert to lower levels.²⁶ Possible explanations for this include sequestration of critical signaling molecules by kinase-inactive BCR:ABL1, protein aggregation eliciting cellular stress responses, and the energy expense associated with maintaining high levels of BCR::ABL expression. Lastly, an overdose of BCR::ABL1 kinase activity may be detrimental, as reported for B-cell acute lymphoblastic leukemia.²⁷ Overexpression of drug efflux pumps can lower intracellular TKI levels, promoting BCR::ABL1 reactivation, and certain polymorphisms in the MDR1 drug transporter are associated with response to imatinib, but there is no evidence that this mechanism can support overt TKI resistance.²⁸ The significance of mechanisms such as increasing expression of MDR1 may be to provide an immediately available bridge until a definitive, genetically encoded mechanism of BCR::ABL1 reactivation is established.²⁹ Understanding how LSC respond to the very first TKI stress event might reveal how this primordial defense can be overcome.

Activation of mitogen activated protein kinase (MAPK) signaling by mutations in NRAS or KRAS is common in patients with acute myeloid leukemia with resistance to FLT3 or IDH1/2 inhibitors, but very rare in CML.^30-32 Why do CML cells strive to reactivate BCR::ABL1 kinase rather than acquiring a different gain-of-function mutation? One explanation may be that CML leukemia stem/progenitor cells (LSPC) typically lack mutations in epigenetic regulators that would allow them to tolerate strong mutational MAPK activation rather than undergoing apoptosis or senescence.^33-35 Another consideration is balanced signaling output. During malignant transformation of B cells, direct mutational activation of MAPK and JAK/STAT signaling is mutually exclusive, suggesting that fine-tuning signal strength is critical.³⁶ It is tempting to speculate that similar principles apply to CML, but experimental verification is unavailable.

The proportion of patients with clinical TKI resistance who exhibit BCR::ABL1 mutations is lowest in CP-CML and highest in lymphoid BP-CML.³⁷ This may reflect TKI exposure history as well as biological constraints. Advanced-phase patients have typically gone through more lines of TKI therapy than CP-CML patients. Additionally, driving advanced CML may require a level of BCR::ABL1 kinase activity that is not achievable by drug efflux of increased BCR::ABL1 expression. An intriguing observation is that individual CML cases tend to relapse with the same mechanism. Thus, a patient who has developed resistance due to a kinase domain mutation is more likely to fail the next line of TKI treatment through acquisition of a different BCR::ABL1 mutation than through other mechanisms.^38,39 The cause of this intriguing mechanistic fidelity is unknown, but it would be interesting to determine whether the preference for one of these mechanisms can be linked to specific mutations in additional genes.

BCR::ABL1 scaffold functions

Kinase-independent functions of BCR::ABL1, often referred to as scaffold functions, contribute to leukemogenesis, and they may be even more important when kinase activity is suppressed. CD34⁺ CML cells exhibit increased proliferation, reduced apoptosis, and aberrant migration and adhesion to bone marrow stroma, a phenotype recapitulated by cord blood CD34⁺ cells transduced with p210BCR::ABL1.^40,41 TKI treatment of these cells normalizes growth, while abnormal adhesion and migration persist. Consistent with these observations, expression of a kinase inactive BCR::ABL1 mutant (p210^{BCR::ABL1K1172R}) has no effect on proliferation, but in part reproduces the migration and adhesion aberrancies.⁴¹ The tumor suppressor p27 is downregulated in CML LSPC.⁴² Expression is restored by TKI, but p27 remains partially mis-localized to the cytoplasm, from where it promotes leukemogenesis.⁴³ BCR::ABL1 expression is required for activation of a JAK2/β-catenin survival/self-renewal pathway in LSPC that involves inhibition of the tumor suppressor phosphatase PP2A. Reproduction of this finding in mouse Lin^-Sca1⁺Kit⁺ (LSK) cells expressing p210^{BCR::ABL1K1172R} indicates that BCR::ABL1 protein rather than kinase activity is required for promoting LSC survival.⁴⁴ Tyrosine 177 of BCR::ABL1 may be a key mediator of scaffold functions. In untreated CML cells, BCR::ABL1^Y177 phosphorylation activates signaling through the RAS/MAPK and phosphatidyl inositol 3’ kinase (PI3’K) pathways.^45,46 In this situation, BCR::ABL1^Y177 is autophosphorylated. However, LYN or JAK2 were shown to maintain pBCR::ABL1^Y177 in TKI-resistant cells treated with TKI.^47-50 Generation of reactive oxygen species (ROS) in CML LSC is thought to confer genetic instability, promoting progression. ROS remain elevated in LSC treated with imatinib compared to HSC, and part of the ROS production is dependent on pBCR::ABL1^Y177, suggesting persistent pBCR::ABL1^Y177 may maintain ROS-induced genetic instability.^51-53 Although the experimental data supporting the latter phenotype are convincing, recent work on primary CML cells has challenged the notion of a BCR::ABL1-induced mutator phenotype.⁵⁴ Serum-induced phosphorylation of the adaptor protein SHC promotes its binding to the BCR::ABL1 SH2 domain and activation of RAS/MAPK signaling in a BCR::ABL1 kinase-independent manner but is abrogated by BCR:ABL1 degradation.⁵⁵ Scaffold functions of BCR::ABL1 might also underline a puzzling clinical observation. Kinase inactive p210^BCR::ABL1 splice forms are present in some CML patients and may be enriched in patients with TKI resistance.^56,57 CML cells carrying these loss-of-function variants should be eliminated, yet they persist, suggesting the variants provide a net functional gain.^19,58 In aggregate, these data strongly suggest that some signals contributing to TKI resistance originate from the BCR::ABL1 scaffold. Scaffold functions are available immediately upon TKI exposure, perhaps buying LSPC critical time to activate salvage mechanisms following the shock of first BCR::ABL1 inhibition. This would not be possible with an effective BCR::ABL1 degrader that rapidly eliminates both kinase activity and scaffold functionalities.⁵⁹

Activation of alternative growth and survival pathways

Numerous signaling pathways have been implicated in TKI resistance of CML, painting a picture of bewildering complexity. We propose that considering the level of residual leukemia defining resistance may provide useful orientation. One extreme is hematologic resistance with or without transformation to BP, the opposite persistence of residual disease. While residual leukemia in patients responding to TKI is thought to reflect survival of LSC, hematologic resistance involves the progenitor cell compartment. The pathways supporting CML cells despite sustained TKI inhibition of BCR::ABL1 seem to differ depending on the level of resistance and the position of resistant cells in the differentiation hierarchy. The more active the disease, the more the alternative pathways mimic the canonical signaling pathways activated by BCR::ABL1 kinase activity, including MAPK, PI3’K, and JAK/STAT, suggesting progenitor cells attempt to recreate the signaling driven by BCR::ABL1. This mimicry may explain why key features of the CP-CML phenotype are often preserved at relapse, such as myeloid left shift or basophilia. In contrast, many of the pathways implicated in LSC survival are HSC pathways, indicating that the CML disease process mostly respects the hierarchy of normal hematopoiesis (Figure 1).

Figure 1.Hierarchical basis of tyrosine kinase inhibitor resistance in chronic myeloid leukemia. Residual chronic myeloid leukemia cells use different survival strategies depending on their position in the hematopoietic hierarchy. Leukemic stem cells persist through kinase-independent pathways that reflect mechanisms used by hematopoietic stem cells. In contrast, progenitor-mediated resistance relies on kinase-dependent canonical pathways that mimic BCR::ABL1 signaling (e.g., JAK/STAT, RAS/MAPK, PI3K), often reinforced by pathway redundancy. CML: chronic myeloid leukemia; LSC: leukemia stem cells.

Epigenetic reprogramming

Numerous studies have revealed perturbed epigenetic regulation in CML LSPC. These include aberrant regulation of the polycomb repressive complexes PRC1 and PRC2. In the canonical model, PRC2 (composed of EZH1/2, Suz12 and EED) catalyzes mono-, di-, and tri-methylation of H3K27, enforcing a major repressive histone mark. PRC1 binds to chromatin through trimethylated histone 3 lysine 27 (H3K27me3) and ubiquitinates H2AK119 through the action of the RING1A/BRING1B ubiquitin ligases, promoting chromatin compaction and transcriptional repression.^60,61 Several PRC2 components are upregulated in LSC, and two independent studies reported a critical role for EZH2 in LSC maintenance, consistent with the rarity of inactivating EZH2 mutations in CML.^62,63 The PRC1 component BMI1 is responsible for ubiquitination of histone H2A K119. BMI1 expression is normally highest in HSC and declines with differentiation, but its expression is elevated in CP-CML LSC compared to in HSC, and increases further in BP-CML.^34,64 Epigenetic reprogramming in CML also involves histone deacetylates and DNA methylation, both at diagnosis and at progression.⁶⁵ Transformation to BP is associated with reduced PRC2 and increased PRC1 activity, and this seems to be independent of which specific additional mutations are present in addition to BCR::ABL1.³⁴ CML cells exhibit increased DNA CpG methylation compared to their normal counterparts, and hypomethylating agents have activity in TKI-resistant CML.^66-68 A central question is whether these epigenetic aberrancies are induced primarily by BCR::ABL1 itself or require additional mutations. In CP-CML, in which BCR::ABL1 is frequently the only detectable mutation, one must postulate that epigenetic changes are induced by BCR::ABL1 signaling. Whether these changes are reversible following TKI therapy is difficult to study, given the low frequency of LSPC in TKI responders with residual disease. However, a doxycycline-controlled BCR::ABL1 mouse model provides important clues: LSPC from leukemic mice exhibit increased DNA methylation compared to mice without BCR::ABL1 expression. While these changes are mostly reversible following remission induced by the addition of doxycycline, some marks persist.⁶⁹ It would be interesting to determine which epigenetic changes are driven by kinase activity rather than scaffold functions. If human CML LSPC maintained a permanent memory of their BCR::ABL1 exposure that cannot be erased by TKI, perhaps this could be accomplished by BCR::ABL1 degraders. An intriguing question is what happens at the epigenetic level in the immediate aftermath of the first TKI-induced BCR::ABL1 inactivation. If a stress-induced epigenetic program helps LSC to escape apoptosis, preemptively blocking this program may contribute to their elimination. Although hypomethylating agents show activity in TKI-resistant CML, they have never been tested in combination with TKI in the frontline, or even as a primer before starting TKI treatment.^70,71

Metabolic reprogramming

In contrast to HSC, LSC depend on mitochondrial function, particularly oxidative phosphorylation. It is not well understood which aspects of metabolic reprogramming are dependent on BCR::ABL1 kinase versus scaffold functions, and which are specific to TKI resistance. A recent report showed that BCR::ABL1 increases Myc expression, which suppresses the TXNIP (thioredoxin-interacting protein) transcription factor thereby increasing glucose uptake, glycolysis, and mitochondrial function. In mouse models TXNIP loss accelerates leukemogenesis, imatinib restores TXNIP expression in TKI-sensitive but not resistant CML cells, and TXNIP re-expression resensitizes resistant cells to TKI.⁷² These findings provide a mechanistic link between BCR::ABL1 signaling and metabolic remodeling.

Targeting mitochondrial metabolism is an attractive strategy to overcome TKI resistance. This can be accomplished by inhibition of mitochondrial translation, as critical respiratory chain proteins are encoded by mitochondrial DNA or by small molecule inhibitors, such as the potent and selective electron transport chain complex I inhibitor IACS-010759.^73,74 Unfortunately, although IACS-010759 showed promising preclinical results, clinical trials had to be discontinued due to lactic acidosis and neurotoxicity, and no efficacy was seen at the doses tolerated.⁷⁵ The Glasgow group recently reported a high-throughput screen that identified lomerizine, a Ca²⁺ channel blocker approved for the treatment of migraines, as an inhibitor of isocitrate dehydrogenase and oxidative phosphorylation that selectively sensitizes CML LSC to imatinib.⁷⁶ Pyruvate anaplerosis is increased in LSC, facilitated by increased mitochondrial pyruvate carrier 1/2 (MPC1/2) expression and pyruvate carboxylase activity compared to HSC. Remarkably, this metabolic abnormality is not reversed by TKI, but is sensitive to the MPC1/2 inhibitor MSDC-0160, which is in clinical development for diabetes.⁷⁷ Arginine auxotrophy represents yet another selective vulnerability. Unlike leukemia cell lines, CD34⁺ LSPC do not express significant levels of argininosuccinase 1 (ASS1) and are unable to upregulate ASS1 upon arginine depletion. As a result, they are sensitive to the arginine-depleting enzyme, BCT-100.⁷⁸ In xenograft studies, BCT-100 markedly reduced LSC survival, and this effect was not enhanced by imatinib, suggesting arginine dependence does not depend on BCR::ABL1 kinase activity. Finally, BCAT1, a cytosolic aminotransferase for branched-chain amino acids, is induced by the RNA-binding protein Musashi2 during progression to BP.^79,80 Whether this is dependent on BCR::ABL1 activity is unknown. BCAT1 inhibition induces differentiation and impairs BP-CML both in vitro and in vivo. Collectively, these studies point to extensive metabolic reprogramming in CML that involves both kinase-dependent and kinase-independent mechanisms and contributes to intrinsic resistance to TKI.

Table 2.Gene detection rate in chronic and advanced stages of chronic myeloid leukemia and its clinical relevance.

Heterogeneity of chronic-phase chronic myeloid leukemia

With the current resolution of genetic testing, 10-15% of CP-CML patients have additional mutations at diagnosis, while in the remainder Ph is the sole detectable genetic abnormality.^33,81,82 How can we reconcile this with the clinical heterogeneity of CP-CML? One consideration is loss of ABL1 or BCR sequences (or both) flanking the translocation breakpoint at the time of the initial translocation event.⁸³ These deletions are associated with adverse outcome in patients treated with interferon-α-based regimens, but their negative effect is largely overcome by TKI.⁸⁴ Attempts to identify putative tumor suppressors in the commonly deleted region(s) were inconclusive, but recent fine-mapping of BCR::ABL1 breakpoints suggest that we may not yet fully understand the extent of breakpoint-associated heterogeneity. Intriguing work from Jyoti Nangalia’s team has revealed that the acquisition of Ph increases HSC growth rates by a large margin, with the highest gains in young patients.⁵⁴ These data are consistent with the clinical observation that young CML patients often present with aggressive disease and suggest that age-related host factors modulate CML biology. Single-cell RNA-sequencing studies on CML HSPC have identified two types of differentiation trajectory. Patients with an erythroid-megakaryocyte trajectory tend to have better responses to TKI and better outcomes than patients with a myelomonocytic trajectory.¹³ Thus, not all HSC are created equal, and whether BCR::ABL1 gets to drive a Prius or a Porsche may profoundly impact CML biology and TKI response.

Mutations in addition to the Philadelphia chromosome

Some patients with BP-CML at diagnosis achieve deep and durable responses to TKI, indicating that the cell clone that lost differentiation capacity behaves like CP-CML. On the other hand, deep responses to the next line of TKI are rare if BP developed during TKI exposure and practically absent in patients lacking BCR::ABL1 mutations, suggesting that BCR::ABL1 kinase independence is partly the result of selective pressure. Recurrent somatic mutations in numerous genes have been identified in BP-CML cells compared to the corresponding CP-CML, but there is no consistent pattern (Table 2).^33-35 A compilation of published data reveals several functional groups of genes. First, mutations in epigenetic regulators and chromatin-remodeling genes, such as ASXL1, IDH1/2, DNMT3A and others, can occur in CP-CML but tend to be more frequent in advanced disease. In contrast, mutations affecting the cohesin complex or RNA-splicing factors exhibit no obvious difference between CP-CML and BP-CML. Another set of genes appears mutated at low frequency but exclusively in AP/BP, suggesting a direct role in transformation. Notably, the genes with the clearest association with the BP-CML phenotype are enriched for transcription factors whose disruption is predicted to impair differentiation, including RUNX1 and IKZF1. A smaller subset of genes mutated with increased frequency in BP is associated with RAS/MAPK activation, but this route of resistance appears to be much less common than in other myeloid neoplasms treated with TKI, e.g., FLT3-ITD⁺ acute myeloid leukemia. The gene with the highest mutation frequency is ASXL1. One explanation for the predominance of ASXL1 mutations in CP-CML and their increasing frequency with progression is that ASXL1 loss-of-function provides a permissive epigenetic lesion that is compatible with CP biology but primes cells for subsequent transformation. Since ASXL1 is a positive regulator of PRC2, its disruption is expected to lower PRC2 activity, loosen epigenetic control of self-renewal and differentiation programs, and thereby reduce the number of additional steps required to achieve a BP transcriptional state.^34,85,86 This would explain why ASXL1-mutant clones are already detectable in CP-CML but become enriched as disease advances to BP. Similarly, mutations in ASXL1 are among the most common genetic abnormalities found in clonal hematopoiesis, often arising years before overt myeloid neoplasia and conferring a competitive advantage to mutant stem cells.^87,88 In contrast to ASXL1, TET2, DTMT3A, and EZH2 and the splicing regulators are rarely mutated and not enriched in BP, suggesting that these variants do not significantly impact disease progression. These data suggest fundamental differences between CML and other myeloid neoplasms. As precise data on the clonal architecture of BP-CML are unavailable, we do not know whether clonal evolution is predominantly linear or parallel. However, BCR::ABL1 independence, epigenetic dysregulation and TKI resistance parallel each other, suggesting that the BCR::ABL1 independence of advanced CML reflects progressive reprogramming to a different cellular state that is less dependent on BCR::ABL1 kinase activity. Remarkably, however, although multi-aberrant genotypes are common in BP-CML, loss of Ph has not been reported in patients progressing on TKI (this is not to be confused with the emergence of clonally unrelated Ph^- myelodysplastic syndromes or acute myeloid leukemia⁸⁹). Thus, BCR::ABL1 may continue to provide a net gain, even if kinase activity is suppressed, and even in BP-CML, BCR::ABL1 kinase independence does not equate BCR::ABL independence (Figure 2).

Figure 2.Intrinsic versus extrinsic pathways of tyrosine kinase inhibitor resistance. Resistance to tyrosine kinase inhibitors in BCR::ABL1-positive leukemias can be categorized as intrinsic or extrinsic. Intrinsic mechanisms include BCR::ABL1-dependent resistance, which may be kinase-dependent (e.g., kinase domain mutations) or scaffold-dependent (non-catalytic functions sustaining downstream signaling), and BCR::ABL1-independent resistance, involving activation of alternative pathways, metabolic reprogramming, or epigenetic alterations. Extrinsic mechanisms derive from the bone marrow microenvironment and involve cellular, soluble, and solid components. TKI: tyrosine kinase inhibitors; CML: chronic myeloid leukemia.

Perspective

In resource-rich countries with access to state-of-the-art medical care for most patients, CML has morphed into a chronic condition of steadily increasing prevalence. Although we lack precise data, it is certain that CML outcomes are far worse in economically less fortunate parts of the world and much could be improved, without any new drugs, by enabling better access to care. However, even with optimal care, overt TKI resistance and progression to advanced phase would still occur in some patients, and the majority would still have to remain on TKI therapy. A few considerations may inform strategies to improve on this. First, overt clinical TKI resistance often combines BCR::ABL1-dependent and -independent mechanisms that cannot be mapped by mutational screening alone. Functional assays may provide additional guidance for therapy selection. Second, TKI exposure of LSC does not completely revert BCR::ABL1-induced epigenetic and metabolic reprogramming, suggesting certain vulnerabilities persist. Alternatively, preventing LSC from reprogramming may retain them in a more vulnerable state. Third, BCR::ABL1 scaffold functions may play a bigger role than often appreciated, supporting the development of clinical degraders. Fourth, for many patients CML is only one of several chronic conditions. In the USA, 30% of patients aged 60-70 and 50% of those aged 70-80 are prescribed five of more daily medications. As adherence is known to drop sharply once the list of daily medication exceeds three, we can only guess how much non-adherence contributes to TKI resistance. The solution? Making treatment-free remission a reality for most patients. Clearly, long-term TKI therapy cannot be the final answer to CML.

Footnotes

• Received September 26, 2025
• Accepted February 5, 2026

Correspondence

M. Deininger mdeining@med.umich.edu

Disclosures

No conflicts of interest to disclose.

Contributions

NC-R conceived the study, reviewed the literature, drafted and critically revised the manuscript, and designed the figures. YL-T reviewed the literature, drafted specific sections of the manuscript, prepared the tables, and critically revised the manuscript. MWD conceived and supervised the study, wrote and critically revised the manuscript and gave final approval of the manuscript.

Funding

MWD was supported in part by National Institutes of Health, National Cancer Institute grants R01CA268496, R01CA257602, and R01CA254354. NC-R is a Special Fellow supported by Blood Cancer United, formerly The Leukemia & Lymphoma Society.

Acknowledgments

We thank Dr. Oliver Hantschel for generously sharing his expertise in ABL1 structural biology and for assisting us in verifying the structural localization of the ABL1 mutations included in this review. His insights were invaluable in ensuring the accuracy of our analyses.