Clonal Basis of Resistance and Response to Ivosidenib Combination Therapies Is Established Early during Treatment in IDH1-Mutated Myeloid Malignancies

Introduction: There is a major need to improve outcome in acute myeloid leukemia (AML) patients unfit for intensive chemotherapy. In a recent phase Ib/II trial, the combination of ivosidenib (IVO) and venetoclax (VEN) ± azacitidine (AZA) in IDH1-mutated AML led to an impressive overall response rate of 94%. However, relapse was still common. Whilst we previously identified some genetic correlates of acquired resistance, kinetics of clonal trajectories leading to therapy resistance or sustained response, to identify the earliest post-therapy timepoints when treatment outcome could be determined, were not studied.

Methods: We applied TARGET-seq+, a method for combined high fidelity single-cell genotyping, RNA-seq, and immunophenotyping, to interrogate sequential bone marrow (BM) samples from 8 patients (6 relapsed and 2 in sustained remission; 6 treated with IVO/VEN and 2 with IVO/VEN/AZA). We assigned recurrent somatic mutations and copy number abnormalities (the latter identified by karyotyping and inferred in single cells from RNA expression). 14,383 cells were assigned to clones, and clonal identity was linked to cell immunophenotype. For each patient, we studied samples from baseline, early post-therapy response, and either relapse or longer-term remission.

Results: In 3/6 relapsing patients, relapse was associated with emergent clones not detected at trial entry, with recurrent somatic AML driver mutations. In all 3 patients, resistant clones became detectable within 3 cycles of treatment, in specific CD34⁺ stem/progenitor (HSPC) or CD34^-CD117⁺ precursor BM compartments, previously shown to contain leukemia stem cell (LSC) potential, while patients were in morphologic complete remission (CR). In 1/3 of these patients, expansion of the relapse-associated IDH1-mutant/FLT3-ITD-mutant clone ensued quickly, within 1 month after initial detection of the clone. In the other 2 patients, genetically evolved IDH1 wild type (WT) relapse-associated clones also became detectable within HSPC/precursor compartments at CR, within 1-3 cycles of therapy. However, in contrast to the prior patient, relapse occurred years after initial detection of the resistant clones. In one of these patients, a newly detected RUNX1-mutated clone emerged within BM progenitors after 1 treatment cycle but only expanded slowly over 24 months, in a complex competing multi-clonal milieu. Progressive disease occurred only ~2 years later and was associated with a substantial increase in the variant allele frequency of the RUNX1 mutation. In the second patient, at CR, an ancestral del(5q) clone acquired a previously undetected SF3B1 mutation in the HSPC compartment. The patient developed myelodysplastic syndrome almost 4 years later.

In the remaining 3 relapsing patients, relapse-associated clones were already detectable in the pre-treatment sample but were not dominant. Interestingly, in all 3 patients, these clones fully differentiated into mature myeloid cells prior to treatment. However, within only 3 therapy cycles, the differentiation capacity of these clones decreased, without acquisition of known myeloid driver mutations, and they expanded within HSPC/precursor compartments. At relapse, which followed within ~4–12 months after attainment of CR, these clones dominated the abnormally expanded HSPC/precursor compartments. In 2/3 of these patients, resistant clones harbored the IDH1 mutation.

Finally, in the 2 patients who remained in sustained remission (48 and 57 months), therapy eradicated all leukemic clones within 3 cycles. Remission was marked by early selection of WT cells and non-leukemic clones with single mutations in genes associated with clonal hematopoiesis (ASXL1 and GNAS).

Conclusions: Clonal selection occurred rapidly within 3 cycles of therapy and was predictive of long-term outcome. Relapse was associated with either genetic clonal evolution, or impaired differentiation of pre-existing clones, within HSPC/precursor cell compartments known to demonstrate LSC potential. Despite the small cohort size, our study has important implications for clinical practice and future randomized clinical trials. If our observations are reproduced in larger cohorts, comprehensive BM assessment rapidly post-treatment initiation could provide an early clinically actionable endpoint. Scalable, cost-effective methods for this assessment need to be developed.