Session Summaries from the ESH-iCMLf 28th Annual John Goldman Conference on CML: Biology and Therapy - Session 3

Scientific Session: Resistance and transformation

Chair: Simona Soverini (Bologna)

This session explored the biological and clinical complexity of resistance and disease transformation in CML, from the molecular mechanisms underlying TKI and asciminib resistance to genomic changes associated with progression and new approaches for overcoming resistant disease. The presentations highlighted how structural biology, genomic and transcriptomic profiling, computational drug discovery and dynamic clinical risk assessment can improve understanding of resistance and potentially support more individualised treatment decisions.

A recurring theme was that resistance cannot be understood through BCR::ABL1 mutations alone. The biological context of mutations, additional genomic abnormalities, disease phase and the timing of treatment response all provide important information about the risk of progression and the most appropriate therapeutic strategy.

Our key takeaways from the presentations:

Keynote presentation

Understanding and overcoming resistance to TKIs
Speaker: Oliver Hantschel (Marburg)

“There’s of course now the possibility to use alternative targeting strategies outside of the ATP binding pocket.”  - Oliver Hantschel

Key points:

• Evolution: TKI resistance has been recognised since the earliest years of imatinib therapy, with more than 100 BCR::ABL1 kinase-domain point mutations now identified across more than 70 residues. Modern high-throughput approaches can model resistance and predict variants before they are observed clinically.
• Heterogeneity: Resistance mutations differ substantially in their effects: some confer resistance selectively to individual TKIs, while others affect several inhibitors. Mutations may also alter BCR::ABL1 kinase activity and signalling properties, with T315I representing an important gain-of-function example. 
• Complexity: Resistance extends beyond kinase-domain point mutations and includes compound mutations, deletions, mutations outside the kinase domain, increased BCR::ABL1 expression, altered intracellular drug concentrations and activation of parallel signalling pathways. 
• Asciminib: Resistance to asciminib can be primary or acquired. Atypical BCR::ABL1 transcripts lacking ABL1 exon 2 can cause intrinsic resistance, while acquired resistance mutations can either interfere directly with asciminib binding or alter the conformation of BCR::ABL1. 
• Innovation: New approaches to overcome resistance include next-generation allosteric inhibitors and PROTACs, which could potentially address kinase-independent BCR::ABL1 functions as well as resistant disease. 
• Bitopic targeting: Linking an ATP-competitive and an allosteric inhibitor within a single bitopic molecule represents another promising strategy. Preclinical work presented in the keynote showed that this approach can inhibit wild-type BCR::ABL1 as well as highly resistant compound mutants. 

Presentation 1

Structural analysis of class II asciminib resistance mutations of BCR::ABL1
Speaker: Inga B. Leske (Marburg) 

“We identified two classes of asciminib resistance mutations. For the class II mutations a more differentiated approach is required.” - Inga B. Leske

Key points:

• Classification: Asciminib resistance mutations can be divided into two mechanistically distinct classes: class I mutations located within the myristoyl pocket and class II mutations located outside the asciminib-binding site. 
• Class I: Class I mutations, including A337T/V and G463D, prevent effective asciminib binding, resulting in strong resistance.
• Class II: Class II mutations, including F359I/V/C and M244V, retain asciminib binding but disrupt the autoinhibited conformation of BCR::ABL1, resulting in inefficient kinase inhibition. The experiments showed increased kinase activity and autophosphorylation with these mutations.
• Allostery: Molecular dynamics demonstrated that class II mutations disrupt the long-range communication network between the myristoyl and ATP-binding pockets, helping explain how mutations distant from the asciminib-binding site can nevertheless confer resistance. 
• Switching: For class I mutations, switching TKI is the principal therapeutic option, as increasing the asciminib dose is unlikely to overcome the loss of binding.
• Individualisation: Class II mutations require a more differentiated therapeutic approach. Depending on the mutation, options may include asciminib dose escalation or switching TKI, while taking potential cross-resistance into account. 

Presentation 2

Additional gene fusions and specific deletions at treatment failure are strongly associated with subsequent development of blast phase, whereas ASXL1 variants alone at treatment failure do not carry that risk

Speaker: Susan Branford (Adelaide) 
“We found a high frequency of co-occurrence of ABL1 mutations and AGAs…. demonstrating the clonal complexity underlying treatment resistance.” - Susan Branford

Key points: 

• Progression: Gene fusions and specific deletions were strongly associated with blast-phase progression in patients treated with frontline TKIs, and were found in 70% of patients who progressed to blast phase. 
• ASXL1: In contrast, mutated ASXL1 was rare in blast phase in this frontline TKI-treated cohort. ASXL1 variants present earlier in disease could disappear as other abnormalities emerged during clonal evolution. 
• Co-occurrence: There was a high frequency of co-occurrence between ABL1 mutations and additional genomic abnormalities (AGAs), demonstrating the clonal complexity underlying treatment resistance.
• Evolution: Longitudinal analysis demonstrated dynamic clonal evolution, including acquisition of new variants, clonal competition and convergent evolution. In many patients with both AGAs and ABL1 mutations, AGAs could be detected earlier. 
• Targeting: Understanding this clonal complexity could influence treatment decisions by creating opportunities to target an ABL1 mutation with a BCR::ABL1 inhibitor alongside another therapy directed against a specific cancer-associated variant.
• Testing: For patients experiencing treatment failure, gene-panel testing covering both myeloid and lymphoid genes may be warranted, importantly including assays capable of detecting gene fusions and deletions that may be missed by standard DNA-based panels.

Presentation 3

Integrated genomic and transcriptomic analyses reveal biological heterogeneity of accelerated-phase chronic myeloid leukemia and identify a blast-phase-like subgroup

Speaker: Xiaoshuai Zhang (Beijing) 

“The key question is which AP patients are biologically BP-like with the worst responses and outcomes.” - Xiaoshuai Zhang

Key points:

• Heterogeneity: Accelerated-phase CML is biologically heterogeneous, and conventional clinical classification does not necessarily identify which patients have already developed blast-phase-like disease. 
• De novo AP: De novo accelerated phase showed genomic and transcriptomic features closer to newly diagnosed chronic phase than to blast phase, suggesting that clinical AP features do not necessarily represent a BP-like biological state. 
• Continuum: In contrast, evolved AP spans a biological continuum from CP to BP, with some patients retaining CP-like features and others acquiring high-risk, BP-like genomic and transcriptional characteristics. 
• Subgroups: Transcriptomic profiling separated evolved AP into two biologically distinct groups: CP-like AP (55%) and BP-like AP (45%).
• BP-like AP: The BP-like AP subgroup was characterised by genomic instability and leukemia stemness activation, together with high-risk genomic features and transformation-associated pathways.
• Prognosis: BP-like AP was associated with significantly poorer survival, whereas CP-like AP behaved more similarly to resistant chronic-phase disease, supporting molecular rather than purely clinical stratification of AP-CML.

Presentation 4

Quantum pharmacophore modelling identifies synergy between LY3009120 and asciminib to inhibit BCR::ABL1 compound mutants

Speaker: Michael Deininger (Ann Arbor) 

“Quantum similarity modelling is a powerful tool to discover synergy at the target level.” - Michael Deininger

Key points:

• Challenge: BCR::ABL1 compound mutations can confer high-level resistance to approved TKIs, particularly when they include T315I, creating a need for effective combination strategies. 
• Modelling: Quantum similarity modelling uses the quantum properties and electron-density characteristics of molecules rather than chemical structure alone, allowing identification of interactions and potential drug synergies that conventional structural approaches may miss. 
• Discovery: Screening approximately two million compounds identified LY3009120, a pan-RAF inhibitor, as a candidate with the desired activity in combination with asciminib, including against highly resistant BCR::ABL1 compound mutants. 
• Synergy: LY3009120 and asciminib were highly synergistic, with a pattern of activity similar to the previously observed synergy between ponatinib and asciminib. 
• Tolerability: The combination may be well tolerated, with preliminary toxicity experiments in mice showing no evident problems over two weeks, although further evaluation is required. 
• Mechanism: The mechanism underlying the strong synergy appears to be complex, with changes in binding affinity playing only a minor role.

Presentation 5

The timing of ELN “warning” predicts transition to “unfavourable” in chronic-phase CML

Speaker: Fumisato Takagi (Hamamatsu) 

“These findings support a two-stage risk framework integrating baseline ELTS risk with the timing of warning.” – Fumisato Takagi

Key points:

• ELTS: Baseline ELTS risk stratified the risk of treatment failure, with particularly clear prognostic value among patients receiving second-generation TKIs. 
• Heterogeneity: Clinical outcomes after an ELN “warning” were heterogeneous: most patients did not subsequently transition to an unfavourable response and of those that did, the majority returned to a favourable category. 
• Timing: The timing of the first warning was critical. Patients experiencing warning at three or six months had substantially higher rates of subsequent unfavourable response than those whose first warning occurred later. 
• Prediction: Once a warning occurred, baseline ELTS was no longer independently associated with progression to unfavourable; instead, a first warning within six months was the only independent predictor, with a hazard ratio of 7.3.
• Management: An early warning may therefore identify patients requiring closer assessment and consideration of treatment modification, while a later warning may support close molecular monitoring rather than immediate TKI switching. 
• Framework: The findings support a two-stage risk framework, combining baseline ELTS risk with the timing of warning to guide individualised monitoring and treatment decisions in CP-CML.