Abstract
Background: Chronic myeloid leukemia (CML) patients treated with tyrosine kinase inhibitors (TKIs) have excellent survival outcomes, however TKI side-effects impact patients' quality-of-life. Some patients can withdraw TKIs without molecular recurrence, termed ‘treatment-free remission’ (TFR). The DESTINY trial (NCT01804985) enrolled CML patients in stable deep molecular remission (DMR; BCR::ABL1:ABL1 <0.01%IS) or major molecular response (MMR; BCR::ABL1:ABL1 <0.1%IS) following ≥3 years treatment with imatinib, nilotinib or dasatinib, for TKI discontinuation. Patients were de-escalated to a half-dose of TKI for 12 months prior to TKI cessation, with 24 months of further follow-up. TFR at the 36-month endpoint was reported as 72% and 36% for DMR and MMR patients, respectively.
Presently, there are no established biomarkers to predict the likelihood of molecular recurrence after TKI discontinuation, although recent evidence suggests that variants in genes related to clonal hematopoiesis (CH) could drive BCR::ABL1+ clonal escape in this context. However, the available clinical evidence on this is inconclusive.
Objectives: The aim of this study was to determine whether variants in genes commonly associated with CH were predictive of TFR, using bone marrow (BM) mononuclear cell (MNC)-derived DNA from DESTINY trial participants.
Methods: DNA was isolated from a subset of DESTINY patients (n=94) at the point of TKI de-escalation and synthesized into next generation sequencing (NGS) libraries using a custom QIAseq panel (n=43 genes). Libraries were sequenced on an Illumina NextSeq platform. A custom bioinformatics pipeline aligning to GRCh37p13 was used for data analysis and variant calling, validated to detect single nucleotide variants (SNVs) and small indels (<20bp) at 5% variant allele frequency (VAF) where a minimum read depth of 200x is achieved. Variant interpretation was based on current best evidence.
Results: Contingency analyses using Fisher's Exact Test found that variants in GATA2 [42.5% of cohort], EZH2 c.118-4del [17.0%], SH2B3 c.1236+24_1236+28del [8.5%] orc.557G>T [3.2%], or BCOR c.1692A>G [25.5%] were positively associated with TFR, while variants in EZH2 (excluding those at position 118 [25.5%]) were associated with relapse after TKI discontinuation (p<0.1). Using univariate logistic regression, only those variants in EZH2, GATA2 or SH2B3 were predictive of molecular recurrence and were used to build a multivariate regression model, whereby the strongest model included EZH2 c.118-4del and GATA2 c.1018-19C>T [38.3%], both intronic variants, as positive predictors of sustained TFR, with a negative predictive value for recurrence of 76.74%. In addition, no relationships between VAF and outcome were observed for these variants.
Molecular recurrence-free survival (MRFS) was the highest in patients with EZH2 c.118-4del +/- GATA2 variants (14/16 patients were recurrence-free at ~1092 days post-trial endpoint), while GATA2 variants alone were associated with a lower but still high MRFS (median not reached, 18/27 patients in TFR ~1092 days post-endpoint). In addition, given its known roles in LSC persistence and CH, patients with any DNMT3A variant (but no EZH2/GATA2 variants [18.1%]) were stratified separately, showing the poorest MRFS (median=475 days, 3/17 patients in TFR ~1092 days post-endpoint) and creating a four-group Kaplan–Meier curve (p<0.0001), where patients with none of these variants had an intermediate MRFS (median=1222 days, 16/34 patients in TFR ~1092 days post-endpoint).
Of note, GATA2 was only significantly predictive of TFR in the DMR patient group but not in the MMR group (univariate regression OR for recurrence = 0.170, CIs [0.025–0.693] in DMR vs. OR=1.313, CIs [0.230–10.500] in MMR); meanwhile, all 5/5 MMR patients with EZH2 c.118-4del sustained TFR, but this variant was not predictive of outcome in the DMR group (univariate OR for recurrence = 0.410, CIs [0.060–1.851]).
Conclusion: Using BM MNCs from patients in the DESTINY trial, we found that intronic variants in EZH2 and GATA2 were positively associated with sustained TFR, while DNMT3A variants were associated with recurrence. Together, these findings support targeted investigation into the genomic and molecular mechanisms underpinning BCR::ABL1 clonal escape after TKI withdrawal. Furthermore, this study supports a strong rationale to validate targeted NGS as a means of risk stratifying CML patients who are eligible for TKI discontinuation.
